diff --git "a/checkpoint-54160/trainer_state.json" "b/checkpoint-54160/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-54160/trainer_state.json" @@ -0,0 +1,325159 @@ +{ + "best_metric": 0.34530189633369446, + "best_model_checkpoint": "data-science-article-titles-engagement/checkpoint-54160", + "epoch": 20.0, + "eval_steps": 1, + "global_step": 54160, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-08, + "loss": 0.6949, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 1.2000000000000002e-07, + "loss": 0.7035, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 1.8e-07, + "loss": 0.7055, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 2.4000000000000003e-07, + "loss": 0.7019, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 3.0000000000000004e-07, + "loss": 0.703, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 3.6e-07, + "loss": 0.7046, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 4.2000000000000006e-07, + "loss": 0.7124, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 4.800000000000001e-07, + "loss": 0.7039, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 5.4e-07, + "loss": 0.7169, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-07, + "loss": 0.6806, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 6.6e-07, + "loss": 0.6981, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 7.2e-07, + "loss": 0.7034, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 7.8e-07, + "loss": 0.6931, + "step": 13 + }, + { + "epoch": 0.01, + "learning_rate": 8.400000000000001e-07, + "loss": 0.6969, + "step": 14 + }, + { + "epoch": 0.01, + "learning_rate": 9e-07, + "loss": 0.7136, + "step": 15 + }, + { + "epoch": 0.01, + "learning_rate": 9.600000000000001e-07, + "loss": 0.7199, + "step": 16 + }, + { + "epoch": 0.01, + "learning_rate": 1.0200000000000002e-06, + "loss": 0.6921, + "step": 17 + }, + { + "epoch": 0.01, + "learning_rate": 1.08e-06, + "loss": 0.7073, + "step": 18 + }, + { + "epoch": 0.01, + "learning_rate": 1.14e-06, + "loss": 0.6998, + "step": 19 + }, + { + "epoch": 0.01, + "learning_rate": 1.2000000000000002e-06, + "loss": 0.6933, + "step": 20 + }, + { + "epoch": 0.01, + "learning_rate": 1.26e-06, + "loss": 0.696, + "step": 21 + }, + { + "epoch": 0.01, + "learning_rate": 1.32e-06, + "loss": 0.6921, + "step": 22 + }, + { + "epoch": 0.01, + "learning_rate": 1.3800000000000001e-06, + "loss": 0.6925, + "step": 23 + }, + { + "epoch": 0.01, + "learning_rate": 1.44e-06, + "loss": 0.7119, + "step": 24 + }, + { + "epoch": 0.01, + "learning_rate": 1.5e-06, + "loss": 0.719, + "step": 25 + }, + { + "epoch": 0.01, + "learning_rate": 1.56e-06, + "loss": 0.6959, + "step": 26 + }, + { + "epoch": 0.01, + "learning_rate": 1.6200000000000002e-06, + "loss": 0.6954, + "step": 27 + }, + { + "epoch": 0.01, + "learning_rate": 1.6800000000000002e-06, + "loss": 0.6918, + "step": 28 + }, + { + "epoch": 0.01, + "learning_rate": 1.7399999999999999e-06, + "loss": 0.6982, + "step": 29 + }, + { + "epoch": 0.01, + "learning_rate": 1.8e-06, + "loss": 0.6882, + "step": 30 + }, + { + "epoch": 0.01, + "learning_rate": 1.86e-06, + "loss": 0.6916, + "step": 31 + }, + { + "epoch": 0.01, + "learning_rate": 1.9200000000000003e-06, + "loss": 0.689, + "step": 32 + }, + { + "epoch": 0.01, + "learning_rate": 1.98e-06, + "loss": 0.6946, + "step": 33 + }, + { + "epoch": 0.01, + "learning_rate": 2.0400000000000004e-06, + "loss": 0.6877, + "step": 34 + }, + { + "epoch": 0.01, + "learning_rate": 2.1e-06, + "loss": 0.6862, + "step": 35 + }, + { + "epoch": 0.01, + "learning_rate": 2.16e-06, + "loss": 0.6946, + "step": 36 + }, + { + "epoch": 0.01, + "learning_rate": 2.22e-06, + "loss": 0.6906, + "step": 37 + }, + { + "epoch": 0.01, + "learning_rate": 2.28e-06, + "loss": 0.7017, + "step": 38 + }, + { + "epoch": 0.01, + "learning_rate": 2.34e-06, + "loss": 0.6909, + "step": 39 + }, + { + "epoch": 0.01, + "learning_rate": 2.4000000000000003e-06, + "loss": 0.6826, + "step": 40 + }, + { + "epoch": 0.02, + "learning_rate": 2.4599999999999997e-06, + "loss": 0.6972, + "step": 41 + }, + { + "epoch": 0.02, + "learning_rate": 2.52e-06, + "loss": 0.7017, + "step": 42 + }, + { + "epoch": 0.02, + "learning_rate": 2.58e-06, + "loss": 0.7069, + "step": 43 + }, + { + "epoch": 0.02, + "learning_rate": 2.64e-06, + "loss": 0.6946, + "step": 44 + }, + { + "epoch": 0.02, + "learning_rate": 2.7e-06, + "loss": 0.6864, + "step": 45 + }, + { + "epoch": 0.02, + "learning_rate": 2.7600000000000003e-06, + "loss": 0.6963, + "step": 46 + }, + { + "epoch": 0.02, + "learning_rate": 2.82e-06, + "loss": 0.6915, + "step": 47 + }, + { + "epoch": 0.02, + "learning_rate": 2.88e-06, + "loss": 0.6914, + "step": 48 + }, + { + "epoch": 0.02, + "learning_rate": 2.9400000000000002e-06, + "loss": 0.6856, + "step": 49 + }, + { + "epoch": 0.02, + "learning_rate": 3e-06, + "loss": 0.7003, + "step": 50 + }, + { + "epoch": 0.02, + "learning_rate": 2.9999630496366547e-06, + "loss": 0.6939, + "step": 51 + }, + { + "epoch": 0.02, + "learning_rate": 2.9999260992733098e-06, + "loss": 0.6946, + "step": 52 + }, + { + "epoch": 0.02, + "learning_rate": 2.9998891489099644e-06, + "loss": 0.6944, + "step": 53 + }, + { + "epoch": 0.02, + "learning_rate": 2.999852198546619e-06, + "loss": 0.6989, + "step": 54 + }, + { + "epoch": 0.02, + "learning_rate": 2.9998152481832737e-06, + "loss": 0.6934, + "step": 55 + }, + { + "epoch": 0.02, + "learning_rate": 2.9997782978199287e-06, + "loss": 0.6965, + "step": 56 + }, + { + "epoch": 0.02, + "learning_rate": 2.9997413474565834e-06, + "loss": 0.6913, + "step": 57 + }, + { + "epoch": 0.02, + "learning_rate": 2.9997043970932384e-06, + "loss": 0.6892, + "step": 58 + }, + { + "epoch": 0.02, + "learning_rate": 2.999667446729893e-06, + "loss": 0.694, + "step": 59 + }, + { + "epoch": 0.02, + "learning_rate": 2.9996304963665477e-06, + "loss": 0.6891, + "step": 60 + }, + { + "epoch": 0.02, + "learning_rate": 2.9995935460032023e-06, + "loss": 0.6879, + "step": 61 + }, + { + "epoch": 0.02, + "learning_rate": 2.999556595639857e-06, + "loss": 0.6906, + "step": 62 + }, + { + "epoch": 0.02, + "learning_rate": 2.999519645276512e-06, + "loss": 0.6933, + "step": 63 + }, + { + "epoch": 0.02, + "learning_rate": 2.9994826949131667e-06, + "loss": 0.7009, + "step": 64 + }, + { + "epoch": 0.02, + "learning_rate": 2.9994457445498217e-06, + "loss": 0.6897, + "step": 65 + }, + { + "epoch": 0.02, + "learning_rate": 2.9994087941864764e-06, + "loss": 0.6976, + "step": 66 + }, + { + "epoch": 0.02, + "learning_rate": 2.999371843823131e-06, + "loss": 0.69, + "step": 67 + }, + { + "epoch": 0.03, + "learning_rate": 2.9993348934597856e-06, + "loss": 0.7013, + "step": 68 + }, + { + "epoch": 0.03, + "learning_rate": 2.9992979430964407e-06, + "loss": 0.702, + "step": 69 + }, + { + "epoch": 0.03, + "learning_rate": 2.9992609927330953e-06, + "loss": 0.7041, + "step": 70 + }, + { + "epoch": 0.03, + "learning_rate": 2.99922404236975e-06, + "loss": 0.6942, + "step": 71 + }, + { + "epoch": 0.03, + "learning_rate": 2.999187092006405e-06, + "loss": 0.6935, + "step": 72 + }, + { + "epoch": 0.03, + "learning_rate": 2.9991501416430593e-06, + "loss": 0.6967, + "step": 73 + }, + { + "epoch": 0.03, + "learning_rate": 2.9991131912797143e-06, + "loss": 0.6884, + "step": 74 + }, + { + "epoch": 0.03, + "learning_rate": 2.999076240916369e-06, + "loss": 0.6842, + "step": 75 + }, + { + "epoch": 0.03, + "learning_rate": 2.999039290553024e-06, + "loss": 0.6962, + "step": 76 + }, + { + "epoch": 0.03, + "learning_rate": 2.9990023401896786e-06, + "loss": 0.6913, + "step": 77 + }, + { + "epoch": 0.03, + "learning_rate": 2.9989653898263337e-06, + "loss": 0.6969, + "step": 78 + }, + { + "epoch": 0.03, + "learning_rate": 2.998928439462988e-06, + "loss": 0.6915, + "step": 79 + }, + { + "epoch": 0.03, + "learning_rate": 2.9988914890996426e-06, + "loss": 0.6861, + "step": 80 + }, + { + "epoch": 0.03, + "learning_rate": 2.9988545387362976e-06, + "loss": 0.6858, + "step": 81 + }, + { + "epoch": 0.03, + "learning_rate": 2.9988175883729523e-06, + "loss": 0.6924, + "step": 82 + }, + { + "epoch": 0.03, + "learning_rate": 2.9987806380096073e-06, + "loss": 0.7002, + "step": 83 + }, + { + "epoch": 0.03, + "learning_rate": 2.998743687646262e-06, + "loss": 0.6975, + "step": 84 + }, + { + "epoch": 0.03, + "learning_rate": 2.9987067372829166e-06, + "loss": 0.6947, + "step": 85 + }, + { + "epoch": 0.03, + "learning_rate": 2.9986697869195712e-06, + "loss": 0.6996, + "step": 86 + }, + { + "epoch": 0.03, + "learning_rate": 2.9986328365562263e-06, + "loss": 0.6982, + "step": 87 + }, + { + "epoch": 0.03, + "learning_rate": 2.998595886192881e-06, + "loss": 0.6879, + "step": 88 + }, + { + "epoch": 0.03, + "learning_rate": 2.998558935829536e-06, + "loss": 0.6841, + "step": 89 + }, + { + "epoch": 0.03, + "learning_rate": 2.9985219854661906e-06, + "loss": 0.6955, + "step": 90 + }, + { + "epoch": 0.03, + "learning_rate": 2.9984850351028453e-06, + "loss": 0.6973, + "step": 91 + }, + { + "epoch": 0.03, + "learning_rate": 2.9984480847395e-06, + "loss": 0.6888, + "step": 92 + }, + { + "epoch": 0.03, + "learning_rate": 2.9984111343761545e-06, + "loss": 0.6794, + "step": 93 + }, + { + "epoch": 0.03, + "learning_rate": 2.9983741840128096e-06, + "loss": 0.6835, + "step": 94 + }, + { + "epoch": 0.04, + "learning_rate": 2.9983372336494642e-06, + "loss": 0.6851, + "step": 95 + }, + { + "epoch": 0.04, + "learning_rate": 2.9983002832861193e-06, + "loss": 0.6928, + "step": 96 + }, + { + "epoch": 0.04, + "learning_rate": 2.998263332922774e-06, + "loss": 0.6902, + "step": 97 + }, + { + "epoch": 0.04, + "learning_rate": 2.9982263825594286e-06, + "loss": 0.6979, + "step": 98 + }, + { + "epoch": 0.04, + "learning_rate": 2.998189432196083e-06, + "loss": 0.6865, + "step": 99 + }, + { + "epoch": 0.04, + "learning_rate": 2.998152481832738e-06, + "loss": 0.6902, + "step": 100 + }, + { + "epoch": 0.04, + "learning_rate": 2.998115531469393e-06, + "loss": 0.6849, + "step": 101 + }, + { + "epoch": 0.04, + "learning_rate": 2.9980785811060475e-06, + "loss": 0.6821, + "step": 102 + }, + { + "epoch": 0.04, + "learning_rate": 2.9980416307427026e-06, + "loss": 0.6974, + "step": 103 + }, + { + "epoch": 0.04, + "learning_rate": 2.9980046803793572e-06, + "loss": 0.6793, + "step": 104 + }, + { + "epoch": 0.04, + "learning_rate": 2.997967730016012e-06, + "loss": 0.6983, + "step": 105 + }, + { + "epoch": 0.04, + "learning_rate": 2.9979307796526665e-06, + "loss": 0.6941, + "step": 106 + }, + { + "epoch": 0.04, + "learning_rate": 2.9978938292893216e-06, + "loss": 0.6962, + "step": 107 + }, + { + "epoch": 0.04, + "learning_rate": 2.997856878925976e-06, + "loss": 0.6909, + "step": 108 + }, + { + "epoch": 0.04, + "learning_rate": 2.9978199285626312e-06, + "loss": 0.6904, + "step": 109 + }, + { + "epoch": 0.04, + "learning_rate": 2.997782978199286e-06, + "loss": 0.6962, + "step": 110 + }, + { + "epoch": 0.04, + "learning_rate": 2.9977460278359405e-06, + "loss": 0.6951, + "step": 111 + }, + { + "epoch": 0.04, + "learning_rate": 2.997709077472595e-06, + "loss": 0.6861, + "step": 112 + }, + { + "epoch": 0.04, + "learning_rate": 2.99767212710925e-06, + "loss": 0.6859, + "step": 113 + }, + { + "epoch": 0.04, + "learning_rate": 2.997635176745905e-06, + "loss": 0.6828, + "step": 114 + }, + { + "epoch": 0.04, + "learning_rate": 2.9975982263825595e-06, + "loss": 0.6914, + "step": 115 + }, + { + "epoch": 0.04, + "learning_rate": 2.9975612760192146e-06, + "loss": 0.6975, + "step": 116 + }, + { + "epoch": 0.04, + "learning_rate": 2.997524325655869e-06, + "loss": 0.6879, + "step": 117 + }, + { + "epoch": 0.04, + "learning_rate": 2.997487375292524e-06, + "loss": 0.7023, + "step": 118 + }, + { + "epoch": 0.04, + "learning_rate": 2.9974504249291785e-06, + "loss": 0.6761, + "step": 119 + }, + { + "epoch": 0.04, + "learning_rate": 2.997413474565833e-06, + "loss": 0.6871, + "step": 120 + }, + { + "epoch": 0.04, + "learning_rate": 2.997376524202488e-06, + "loss": 0.6914, + "step": 121 + }, + { + "epoch": 0.05, + "learning_rate": 2.997339573839143e-06, + "loss": 0.682, + "step": 122 + }, + { + "epoch": 0.05, + "learning_rate": 2.997302623475798e-06, + "loss": 0.6869, + "step": 123 + }, + { + "epoch": 0.05, + "learning_rate": 2.997265673112452e-06, + "loss": 0.691, + "step": 124 + }, + { + "epoch": 0.05, + "learning_rate": 2.997228722749107e-06, + "loss": 0.68, + "step": 125 + }, + { + "epoch": 0.05, + "learning_rate": 2.9971917723857618e-06, + "loss": 0.689, + "step": 126 + }, + { + "epoch": 0.05, + "learning_rate": 2.997154822022417e-06, + "loss": 0.695, + "step": 127 + }, + { + "epoch": 0.05, + "learning_rate": 2.9971178716590715e-06, + "loss": 0.6907, + "step": 128 + }, + { + "epoch": 0.05, + "learning_rate": 2.997080921295726e-06, + "loss": 0.6945, + "step": 129 + }, + { + "epoch": 0.05, + "learning_rate": 2.9970439709323807e-06, + "loss": 0.694, + "step": 130 + }, + { + "epoch": 0.05, + "learning_rate": 2.9970070205690354e-06, + "loss": 0.6767, + "step": 131 + }, + { + "epoch": 0.05, + "learning_rate": 2.9969700702056904e-06, + "loss": 0.674, + "step": 132 + }, + { + "epoch": 0.05, + "learning_rate": 2.996933119842345e-06, + "loss": 0.691, + "step": 133 + }, + { + "epoch": 0.05, + "learning_rate": 2.996896169479e-06, + "loss": 0.6799, + "step": 134 + }, + { + "epoch": 0.05, + "learning_rate": 2.9968592191156548e-06, + "loss": 0.679, + "step": 135 + }, + { + "epoch": 0.05, + "learning_rate": 2.9968222687523094e-06, + "loss": 0.6835, + "step": 136 + }, + { + "epoch": 0.05, + "learning_rate": 2.996785318388964e-06, + "loss": 0.6821, + "step": 137 + }, + { + "epoch": 0.05, + "learning_rate": 2.996748368025619e-06, + "loss": 0.6896, + "step": 138 + }, + { + "epoch": 0.05, + "learning_rate": 2.9967114176622737e-06, + "loss": 0.6776, + "step": 139 + }, + { + "epoch": 0.05, + "learning_rate": 2.9966744672989284e-06, + "loss": 0.6865, + "step": 140 + }, + { + "epoch": 0.05, + "learning_rate": 2.9966375169355834e-06, + "loss": 0.6905, + "step": 141 + }, + { + "epoch": 0.05, + "learning_rate": 2.996600566572238e-06, + "loss": 0.683, + "step": 142 + }, + { + "epoch": 0.05, + "learning_rate": 2.9965636162088927e-06, + "loss": 0.6772, + "step": 143 + }, + { + "epoch": 0.05, + "learning_rate": 2.9965266658455473e-06, + "loss": 0.6782, + "step": 144 + }, + { + "epoch": 0.05, + "learning_rate": 2.9964897154822024e-06, + "loss": 0.6916, + "step": 145 + }, + { + "epoch": 0.05, + "learning_rate": 2.996452765118857e-06, + "loss": 0.7026, + "step": 146 + }, + { + "epoch": 0.05, + "learning_rate": 2.996415814755512e-06, + "loss": 0.678, + "step": 147 + }, + { + "epoch": 0.05, + "learning_rate": 2.9963788643921667e-06, + "loss": 0.6844, + "step": 148 + }, + { + "epoch": 0.06, + "learning_rate": 2.9963419140288214e-06, + "loss": 0.7075, + "step": 149 + }, + { + "epoch": 0.06, + "learning_rate": 2.996304963665476e-06, + "loss": 0.6823, + "step": 150 + }, + { + "epoch": 0.06, + "learning_rate": 2.9962680133021306e-06, + "loss": 0.6623, + "step": 151 + }, + { + "epoch": 0.06, + "learning_rate": 2.9962310629387857e-06, + "loss": 0.6798, + "step": 152 + }, + { + "epoch": 0.06, + "learning_rate": 2.9961941125754403e-06, + "loss": 0.6743, + "step": 153 + }, + { + "epoch": 0.06, + "learning_rate": 2.9961571622120954e-06, + "loss": 0.6893, + "step": 154 + }, + { + "epoch": 0.06, + "learning_rate": 2.99612021184875e-06, + "loss": 0.6935, + "step": 155 + }, + { + "epoch": 0.06, + "learning_rate": 2.9960832614854047e-06, + "loss": 0.7051, + "step": 156 + }, + { + "epoch": 0.06, + "learning_rate": 2.9960463111220593e-06, + "loss": 0.6929, + "step": 157 + }, + { + "epoch": 0.06, + "learning_rate": 2.9960093607587144e-06, + "loss": 0.6949, + "step": 158 + }, + { + "epoch": 0.06, + "learning_rate": 2.995972410395369e-06, + "loss": 0.6999, + "step": 159 + }, + { + "epoch": 0.06, + "learning_rate": 2.9959354600320236e-06, + "loss": 0.68, + "step": 160 + }, + { + "epoch": 0.06, + "learning_rate": 2.9958985096686787e-06, + "loss": 0.6801, + "step": 161 + }, + { + "epoch": 0.06, + "learning_rate": 2.995861559305333e-06, + "loss": 0.6858, + "step": 162 + }, + { + "epoch": 0.06, + "learning_rate": 2.995824608941988e-06, + "loss": 0.6955, + "step": 163 + }, + { + "epoch": 0.06, + "learning_rate": 2.9957876585786426e-06, + "loss": 0.6807, + "step": 164 + }, + { + "epoch": 0.06, + "learning_rate": 2.9957507082152977e-06, + "loss": 0.6724, + "step": 165 + }, + { + "epoch": 0.06, + "learning_rate": 2.9957137578519523e-06, + "loss": 0.6864, + "step": 166 + }, + { + "epoch": 0.06, + "learning_rate": 2.9956768074886074e-06, + "loss": 0.6934, + "step": 167 + }, + { + "epoch": 0.06, + "learning_rate": 2.9956398571252616e-06, + "loss": 0.6864, + "step": 168 + }, + { + "epoch": 0.06, + "learning_rate": 2.9956029067619166e-06, + "loss": 0.6925, + "step": 169 + }, + { + "epoch": 0.06, + "learning_rate": 2.9955659563985713e-06, + "loss": 0.6965, + "step": 170 + }, + { + "epoch": 0.06, + "learning_rate": 2.995529006035226e-06, + "loss": 0.6608, + "step": 171 + }, + { + "epoch": 0.06, + "learning_rate": 2.995492055671881e-06, + "loss": 0.6953, + "step": 172 + }, + { + "epoch": 0.06, + "learning_rate": 2.9954551053085356e-06, + "loss": 0.6631, + "step": 173 + }, + { + "epoch": 0.06, + "learning_rate": 2.9954181549451902e-06, + "loss": 0.6693, + "step": 174 + }, + { + "epoch": 0.06, + "learning_rate": 2.995381204581845e-06, + "loss": 0.684, + "step": 175 + }, + { + "epoch": 0.06, + "learning_rate": 2.9953442542185e-06, + "loss": 0.6858, + "step": 176 + }, + { + "epoch": 0.07, + "learning_rate": 2.9953073038551546e-06, + "loss": 0.688, + "step": 177 + }, + { + "epoch": 0.07, + "learning_rate": 2.9952703534918096e-06, + "loss": 0.6811, + "step": 178 + }, + { + "epoch": 0.07, + "learning_rate": 2.9952334031284643e-06, + "loss": 0.686, + "step": 179 + }, + { + "epoch": 0.07, + "learning_rate": 2.995196452765119e-06, + "loss": 0.6908, + "step": 180 + }, + { + "epoch": 0.07, + "learning_rate": 2.9951595024017735e-06, + "loss": 0.6705, + "step": 181 + }, + { + "epoch": 0.07, + "learning_rate": 2.995122552038428e-06, + "loss": 0.6752, + "step": 182 + }, + { + "epoch": 0.07, + "learning_rate": 2.9950856016750832e-06, + "loss": 0.673, + "step": 183 + }, + { + "epoch": 0.07, + "learning_rate": 2.995048651311738e-06, + "loss": 0.6786, + "step": 184 + }, + { + "epoch": 0.07, + "learning_rate": 2.995011700948393e-06, + "loss": 0.6721, + "step": 185 + }, + { + "epoch": 0.07, + "learning_rate": 2.9949747505850476e-06, + "loss": 0.7057, + "step": 186 + }, + { + "epoch": 0.07, + "learning_rate": 2.994937800221702e-06, + "loss": 0.686, + "step": 187 + }, + { + "epoch": 0.07, + "learning_rate": 2.994900849858357e-06, + "loss": 0.6923, + "step": 188 + }, + { + "epoch": 0.07, + "learning_rate": 2.994863899495012e-06, + "loss": 0.6713, + "step": 189 + }, + { + "epoch": 0.07, + "learning_rate": 2.9948269491316665e-06, + "loss": 0.7029, + "step": 190 + }, + { + "epoch": 0.07, + "learning_rate": 2.994789998768321e-06, + "loss": 0.6699, + "step": 191 + }, + { + "epoch": 0.07, + "learning_rate": 2.9947530484049762e-06, + "loss": 0.695, + "step": 192 + }, + { + "epoch": 0.07, + "learning_rate": 2.994716098041631e-06, + "loss": 0.6971, + "step": 193 + }, + { + "epoch": 0.07, + "learning_rate": 2.9946791476782855e-06, + "loss": 0.6712, + "step": 194 + }, + { + "epoch": 0.07, + "learning_rate": 2.99464219731494e-06, + "loss": 0.6784, + "step": 195 + }, + { + "epoch": 0.07, + "learning_rate": 2.994605246951595e-06, + "loss": 0.6765, + "step": 196 + }, + { + "epoch": 0.07, + "learning_rate": 2.99456829658825e-06, + "loss": 0.6836, + "step": 197 + }, + { + "epoch": 0.07, + "learning_rate": 2.994531346224905e-06, + "loss": 0.7201, + "step": 198 + }, + { + "epoch": 0.07, + "learning_rate": 2.9944943958615595e-06, + "loss": 0.6953, + "step": 199 + }, + { + "epoch": 0.07, + "learning_rate": 2.994457445498214e-06, + "loss": 0.6688, + "step": 200 + }, + { + "epoch": 0.07, + "learning_rate": 2.994420495134869e-06, + "loss": 0.6839, + "step": 201 + }, + { + "epoch": 0.07, + "learning_rate": 2.9943835447715234e-06, + "loss": 0.66, + "step": 202 + }, + { + "epoch": 0.07, + "learning_rate": 2.9943465944081785e-06, + "loss": 0.6771, + "step": 203 + }, + { + "epoch": 0.08, + "learning_rate": 2.994309644044833e-06, + "loss": 0.6819, + "step": 204 + }, + { + "epoch": 0.08, + "learning_rate": 2.994272693681488e-06, + "loss": 0.6896, + "step": 205 + }, + { + "epoch": 0.08, + "learning_rate": 2.994235743318143e-06, + "loss": 0.67, + "step": 206 + }, + { + "epoch": 0.08, + "learning_rate": 2.9941987929547975e-06, + "loss": 0.6818, + "step": 207 + }, + { + "epoch": 0.08, + "learning_rate": 2.994161842591452e-06, + "loss": 0.6828, + "step": 208 + }, + { + "epoch": 0.08, + "learning_rate": 2.994124892228107e-06, + "loss": 0.6851, + "step": 209 + }, + { + "epoch": 0.08, + "learning_rate": 2.994087941864762e-06, + "loss": 0.6937, + "step": 210 + }, + { + "epoch": 0.08, + "learning_rate": 2.9940509915014164e-06, + "loss": 0.6698, + "step": 211 + }, + { + "epoch": 0.08, + "learning_rate": 2.9940140411380715e-06, + "loss": 0.6977, + "step": 212 + }, + { + "epoch": 0.08, + "learning_rate": 2.9939770907747257e-06, + "loss": 0.6713, + "step": 213 + }, + { + "epoch": 0.08, + "learning_rate": 2.9939401404113808e-06, + "loss": 0.7059, + "step": 214 + }, + { + "epoch": 0.08, + "learning_rate": 2.9939031900480354e-06, + "loss": 0.6562, + "step": 215 + }, + { + "epoch": 0.08, + "learning_rate": 2.9938662396846905e-06, + "loss": 0.6807, + "step": 216 + }, + { + "epoch": 0.08, + "learning_rate": 2.993829289321345e-06, + "loss": 0.6877, + "step": 217 + }, + { + "epoch": 0.08, + "learning_rate": 2.993792338958e-06, + "loss": 0.65, + "step": 218 + }, + { + "epoch": 0.08, + "learning_rate": 2.9937553885946544e-06, + "loss": 0.6911, + "step": 219 + }, + { + "epoch": 0.08, + "learning_rate": 2.993718438231309e-06, + "loss": 0.665, + "step": 220 + }, + { + "epoch": 0.08, + "learning_rate": 2.993681487867964e-06, + "loss": 0.6697, + "step": 221 + }, + { + "epoch": 0.08, + "learning_rate": 2.9936445375046187e-06, + "loss": 0.6605, + "step": 222 + }, + { + "epoch": 0.08, + "learning_rate": 2.9936075871412738e-06, + "loss": 0.686, + "step": 223 + }, + { + "epoch": 0.08, + "learning_rate": 2.9935706367779284e-06, + "loss": 0.6488, + "step": 224 + }, + { + "epoch": 0.08, + "learning_rate": 2.993533686414583e-06, + "loss": 0.6743, + "step": 225 + }, + { + "epoch": 0.08, + "learning_rate": 2.9934967360512377e-06, + "loss": 0.6684, + "step": 226 + }, + { + "epoch": 0.08, + "learning_rate": 2.9934597856878927e-06, + "loss": 0.6732, + "step": 227 + }, + { + "epoch": 0.08, + "learning_rate": 2.9934228353245474e-06, + "loss": 0.6757, + "step": 228 + }, + { + "epoch": 0.08, + "learning_rate": 2.9933858849612024e-06, + "loss": 0.6726, + "step": 229 + }, + { + "epoch": 0.08, + "learning_rate": 2.993348934597857e-06, + "loss": 0.6975, + "step": 230 + }, + { + "epoch": 0.09, + "learning_rate": 2.9933119842345117e-06, + "loss": 0.6702, + "step": 231 + }, + { + "epoch": 0.09, + "learning_rate": 2.9932750338711664e-06, + "loss": 0.703, + "step": 232 + }, + { + "epoch": 0.09, + "learning_rate": 2.993238083507821e-06, + "loss": 0.6573, + "step": 233 + }, + { + "epoch": 0.09, + "learning_rate": 2.993201133144476e-06, + "loss": 0.6903, + "step": 234 + }, + { + "epoch": 0.09, + "learning_rate": 2.9931641827811307e-06, + "loss": 0.6776, + "step": 235 + }, + { + "epoch": 0.09, + "learning_rate": 2.9931272324177857e-06, + "loss": 0.6648, + "step": 236 + }, + { + "epoch": 0.09, + "learning_rate": 2.9930902820544404e-06, + "loss": 0.6554, + "step": 237 + }, + { + "epoch": 0.09, + "learning_rate": 2.993053331691095e-06, + "loss": 0.6543, + "step": 238 + }, + { + "epoch": 0.09, + "learning_rate": 2.9930163813277497e-06, + "loss": 0.658, + "step": 239 + }, + { + "epoch": 0.09, + "learning_rate": 2.9929794309644043e-06, + "loss": 0.6414, + "step": 240 + }, + { + "epoch": 0.09, + "learning_rate": 2.9929424806010594e-06, + "loss": 0.6864, + "step": 241 + }, + { + "epoch": 0.09, + "learning_rate": 2.992905530237714e-06, + "loss": 0.6766, + "step": 242 + }, + { + "epoch": 0.09, + "learning_rate": 2.992868579874369e-06, + "loss": 0.7082, + "step": 243 + }, + { + "epoch": 0.09, + "learning_rate": 2.9928316295110237e-06, + "loss": 0.6556, + "step": 244 + }, + { + "epoch": 0.09, + "learning_rate": 2.9927946791476783e-06, + "loss": 0.6403, + "step": 245 + }, + { + "epoch": 0.09, + "learning_rate": 2.992757728784333e-06, + "loss": 0.664, + "step": 246 + }, + { + "epoch": 0.09, + "learning_rate": 2.992720778420988e-06, + "loss": 0.6696, + "step": 247 + }, + { + "epoch": 0.09, + "learning_rate": 2.9926838280576427e-06, + "loss": 0.6702, + "step": 248 + }, + { + "epoch": 0.09, + "learning_rate": 2.9926468776942977e-06, + "loss": 0.6528, + "step": 249 + }, + { + "epoch": 0.09, + "learning_rate": 2.9926099273309524e-06, + "loss": 0.6573, + "step": 250 + }, + { + "epoch": 0.09, + "learning_rate": 2.992572976967607e-06, + "loss": 0.6956, + "step": 251 + }, + { + "epoch": 0.09, + "learning_rate": 2.9925360266042616e-06, + "loss": 0.7164, + "step": 252 + }, + { + "epoch": 0.09, + "learning_rate": 2.9924990762409163e-06, + "loss": 0.6567, + "step": 253 + }, + { + "epoch": 0.09, + "learning_rate": 2.9924621258775713e-06, + "loss": 0.7142, + "step": 254 + }, + { + "epoch": 0.09, + "learning_rate": 2.992425175514226e-06, + "loss": 0.7093, + "step": 255 + }, + { + "epoch": 0.09, + "learning_rate": 2.992388225150881e-06, + "loss": 0.7192, + "step": 256 + }, + { + "epoch": 0.09, + "learning_rate": 2.9923512747875357e-06, + "loss": 0.6616, + "step": 257 + }, + { + "epoch": 0.1, + "learning_rate": 2.9923143244241903e-06, + "loss": 0.655, + "step": 258 + }, + { + "epoch": 0.1, + "learning_rate": 2.992277374060845e-06, + "loss": 0.6785, + "step": 259 + }, + { + "epoch": 0.1, + "learning_rate": 2.9922404236974996e-06, + "loss": 0.7033, + "step": 260 + }, + { + "epoch": 0.1, + "learning_rate": 2.9922034733341546e-06, + "loss": 0.6558, + "step": 261 + }, + { + "epoch": 0.1, + "learning_rate": 2.9921665229708093e-06, + "loss": 0.6676, + "step": 262 + }, + { + "epoch": 0.1, + "learning_rate": 2.9921295726074643e-06, + "loss": 0.6374, + "step": 263 + }, + { + "epoch": 0.1, + "learning_rate": 2.9920926222441185e-06, + "loss": 0.6901, + "step": 264 + }, + { + "epoch": 0.1, + "learning_rate": 2.9920556718807736e-06, + "loss": 0.7188, + "step": 265 + }, + { + "epoch": 0.1, + "learning_rate": 2.9920187215174282e-06, + "loss": 0.6617, + "step": 266 + }, + { + "epoch": 0.1, + "learning_rate": 2.9919817711540833e-06, + "loss": 0.6819, + "step": 267 + }, + { + "epoch": 0.1, + "learning_rate": 2.991944820790738e-06, + "loss": 0.6673, + "step": 268 + }, + { + "epoch": 0.1, + "learning_rate": 2.991907870427393e-06, + "loss": 0.6596, + "step": 269 + }, + { + "epoch": 0.1, + "learning_rate": 2.991870920064047e-06, + "loss": 0.6693, + "step": 270 + }, + { + "epoch": 0.1, + "learning_rate": 2.991833969700702e-06, + "loss": 0.6814, + "step": 271 + }, + { + "epoch": 0.1, + "learning_rate": 2.991797019337357e-06, + "loss": 0.7041, + "step": 272 + }, + { + "epoch": 0.1, + "learning_rate": 2.9917600689740115e-06, + "loss": 0.7023, + "step": 273 + }, + { + "epoch": 0.1, + "learning_rate": 2.9917231186106666e-06, + "loss": 0.6641, + "step": 274 + }, + { + "epoch": 0.1, + "learning_rate": 2.9916861682473212e-06, + "loss": 0.6695, + "step": 275 + }, + { + "epoch": 0.1, + "learning_rate": 2.991649217883976e-06, + "loss": 0.6726, + "step": 276 + }, + { + "epoch": 0.1, + "learning_rate": 2.9916122675206305e-06, + "loss": 0.6507, + "step": 277 + }, + { + "epoch": 0.1, + "learning_rate": 2.9915753171572856e-06, + "loss": 0.6911, + "step": 278 + }, + { + "epoch": 0.1, + "learning_rate": 2.99153836679394e-06, + "loss": 0.6843, + "step": 279 + }, + { + "epoch": 0.1, + "learning_rate": 2.991501416430595e-06, + "loss": 0.6754, + "step": 280 + }, + { + "epoch": 0.1, + "learning_rate": 2.99146446606725e-06, + "loss": 0.6414, + "step": 281 + }, + { + "epoch": 0.1, + "learning_rate": 2.9914275157039045e-06, + "loss": 0.6516, + "step": 282 + }, + { + "epoch": 0.1, + "learning_rate": 2.991390565340559e-06, + "loss": 0.6572, + "step": 283 + }, + { + "epoch": 0.1, + "learning_rate": 2.991353614977214e-06, + "loss": 0.6817, + "step": 284 + }, + { + "epoch": 0.11, + "learning_rate": 2.991316664613869e-06, + "loss": 0.688, + "step": 285 + }, + { + "epoch": 0.11, + "learning_rate": 2.9912797142505235e-06, + "loss": 0.6792, + "step": 286 + }, + { + "epoch": 0.11, + "learning_rate": 2.9912427638871786e-06, + "loss": 0.6625, + "step": 287 + }, + { + "epoch": 0.11, + "learning_rate": 2.991205813523833e-06, + "loss": 0.6091, + "step": 288 + }, + { + "epoch": 0.11, + "learning_rate": 2.991168863160488e-06, + "loss": 0.6628, + "step": 289 + }, + { + "epoch": 0.11, + "learning_rate": 2.9911319127971425e-06, + "loss": 0.7099, + "step": 290 + }, + { + "epoch": 0.11, + "learning_rate": 2.991094962433797e-06, + "loss": 0.6606, + "step": 291 + }, + { + "epoch": 0.11, + "learning_rate": 2.991058012070452e-06, + "loss": 0.6826, + "step": 292 + }, + { + "epoch": 0.11, + "learning_rate": 2.991021061707107e-06, + "loss": 0.674, + "step": 293 + }, + { + "epoch": 0.11, + "learning_rate": 2.990984111343762e-06, + "loss": 0.6483, + "step": 294 + }, + { + "epoch": 0.11, + "learning_rate": 2.9909471609804165e-06, + "loss": 0.7094, + "step": 295 + }, + { + "epoch": 0.11, + "learning_rate": 2.990910210617071e-06, + "loss": 0.6507, + "step": 296 + }, + { + "epoch": 0.11, + "learning_rate": 2.9908732602537258e-06, + "loss": 0.6622, + "step": 297 + }, + { + "epoch": 0.11, + "learning_rate": 2.990836309890381e-06, + "loss": 0.6973, + "step": 298 + }, + { + "epoch": 0.11, + "learning_rate": 2.9907993595270355e-06, + "loss": 0.6553, + "step": 299 + }, + { + "epoch": 0.11, + "learning_rate": 2.99076240916369e-06, + "loss": 0.623, + "step": 300 + }, + { + "epoch": 0.11, + "learning_rate": 2.990725458800345e-06, + "loss": 0.6823, + "step": 301 + }, + { + "epoch": 0.11, + "learning_rate": 2.990688508437e-06, + "loss": 0.6902, + "step": 302 + }, + { + "epoch": 0.11, + "learning_rate": 2.9906515580736544e-06, + "loss": 0.6735, + "step": 303 + }, + { + "epoch": 0.11, + "learning_rate": 2.990614607710309e-06, + "loss": 0.6873, + "step": 304 + }, + { + "epoch": 0.11, + "learning_rate": 2.990577657346964e-06, + "loss": 0.6663, + "step": 305 + }, + { + "epoch": 0.11, + "learning_rate": 2.9905407069836188e-06, + "loss": 0.6969, + "step": 306 + }, + { + "epoch": 0.11, + "learning_rate": 2.990503756620274e-06, + "loss": 0.7182, + "step": 307 + }, + { + "epoch": 0.11, + "learning_rate": 2.9904668062569285e-06, + "loss": 0.6749, + "step": 308 + }, + { + "epoch": 0.11, + "learning_rate": 2.9904298558935827e-06, + "loss": 0.7149, + "step": 309 + }, + { + "epoch": 0.11, + "learning_rate": 2.9903929055302377e-06, + "loss": 0.6864, + "step": 310 + }, + { + "epoch": 0.11, + "learning_rate": 2.9903559551668924e-06, + "loss": 0.6553, + "step": 311 + }, + { + "epoch": 0.12, + "learning_rate": 2.9903190048035474e-06, + "loss": 0.7081, + "step": 312 + }, + { + "epoch": 0.12, + "learning_rate": 2.990282054440202e-06, + "loss": 0.6977, + "step": 313 + }, + { + "epoch": 0.12, + "learning_rate": 2.9902451040768567e-06, + "loss": 0.6332, + "step": 314 + }, + { + "epoch": 0.12, + "learning_rate": 2.9902081537135113e-06, + "loss": 0.7013, + "step": 315 + }, + { + "epoch": 0.12, + "learning_rate": 2.9901712033501664e-06, + "loss": 0.6998, + "step": 316 + }, + { + "epoch": 0.12, + "learning_rate": 2.990134252986821e-06, + "loss": 0.6697, + "step": 317 + }, + { + "epoch": 0.12, + "learning_rate": 2.990097302623476e-06, + "loss": 0.6648, + "step": 318 + }, + { + "epoch": 0.12, + "learning_rate": 2.9900603522601307e-06, + "loss": 0.6697, + "step": 319 + }, + { + "epoch": 0.12, + "learning_rate": 2.9900234018967854e-06, + "loss": 0.6715, + "step": 320 + }, + { + "epoch": 0.12, + "learning_rate": 2.98998645153344e-06, + "loss": 0.6659, + "step": 321 + }, + { + "epoch": 0.12, + "learning_rate": 2.9899495011700946e-06, + "loss": 0.7018, + "step": 322 + }, + { + "epoch": 0.12, + "learning_rate": 2.9899125508067497e-06, + "loss": 0.6665, + "step": 323 + }, + { + "epoch": 0.12, + "learning_rate": 2.9898756004434043e-06, + "loss": 0.6808, + "step": 324 + }, + { + "epoch": 0.12, + "learning_rate": 2.9898386500800594e-06, + "loss": 0.6679, + "step": 325 + }, + { + "epoch": 0.12, + "learning_rate": 2.989801699716714e-06, + "loss": 0.6787, + "step": 326 + }, + { + "epoch": 0.12, + "learning_rate": 2.9897647493533687e-06, + "loss": 0.6742, + "step": 327 + }, + { + "epoch": 0.12, + "learning_rate": 2.9897277989900233e-06, + "loss": 0.6606, + "step": 328 + }, + { + "epoch": 0.12, + "learning_rate": 2.989690848626678e-06, + "loss": 0.6644, + "step": 329 + }, + { + "epoch": 0.12, + "learning_rate": 2.989653898263333e-06, + "loss": 0.6822, + "step": 330 + }, + { + "epoch": 0.12, + "learning_rate": 2.9896169478999876e-06, + "loss": 0.6712, + "step": 331 + }, + { + "epoch": 0.12, + "learning_rate": 2.9895799975366427e-06, + "loss": 0.6696, + "step": 332 + }, + { + "epoch": 0.12, + "learning_rate": 2.9895430471732973e-06, + "loss": 0.678, + "step": 333 + }, + { + "epoch": 0.12, + "learning_rate": 2.989506096809952e-06, + "loss": 0.6863, + "step": 334 + }, + { + "epoch": 0.12, + "learning_rate": 2.9894691464466066e-06, + "loss": 0.6654, + "step": 335 + }, + { + "epoch": 0.12, + "learning_rate": 2.9894321960832617e-06, + "loss": 0.6474, + "step": 336 + }, + { + "epoch": 0.12, + "learning_rate": 2.9893952457199163e-06, + "loss": 0.6807, + "step": 337 + }, + { + "epoch": 0.12, + "learning_rate": 2.9893582953565714e-06, + "loss": 0.6462, + "step": 338 + }, + { + "epoch": 0.13, + "learning_rate": 2.989321344993226e-06, + "loss": 0.6861, + "step": 339 + }, + { + "epoch": 0.13, + "learning_rate": 2.9892843946298806e-06, + "loss": 0.691, + "step": 340 + }, + { + "epoch": 0.13, + "learning_rate": 2.9892474442665353e-06, + "loss": 0.7343, + "step": 341 + }, + { + "epoch": 0.13, + "learning_rate": 2.98921049390319e-06, + "loss": 0.6802, + "step": 342 + }, + { + "epoch": 0.13, + "learning_rate": 2.989173543539845e-06, + "loss": 0.6743, + "step": 343 + }, + { + "epoch": 0.13, + "learning_rate": 2.9891365931764996e-06, + "loss": 0.6861, + "step": 344 + }, + { + "epoch": 0.13, + "learning_rate": 2.9890996428131547e-06, + "loss": 0.6527, + "step": 345 + }, + { + "epoch": 0.13, + "learning_rate": 2.9890626924498093e-06, + "loss": 0.6712, + "step": 346 + }, + { + "epoch": 0.13, + "learning_rate": 2.989025742086464e-06, + "loss": 0.698, + "step": 347 + }, + { + "epoch": 0.13, + "learning_rate": 2.9889887917231186e-06, + "loss": 0.7276, + "step": 348 + }, + { + "epoch": 0.13, + "learning_rate": 2.9889518413597732e-06, + "loss": 0.6947, + "step": 349 + }, + { + "epoch": 0.13, + "learning_rate": 2.9889148909964283e-06, + "loss": 0.6767, + "step": 350 + }, + { + "epoch": 0.13, + "learning_rate": 2.988877940633083e-06, + "loss": 0.7207, + "step": 351 + }, + { + "epoch": 0.13, + "learning_rate": 2.988840990269738e-06, + "loss": 0.6809, + "step": 352 + }, + { + "epoch": 0.13, + "learning_rate": 2.988804039906392e-06, + "loss": 0.7004, + "step": 353 + }, + { + "epoch": 0.13, + "learning_rate": 2.9887670895430472e-06, + "loss": 0.7019, + "step": 354 + }, + { + "epoch": 0.13, + "learning_rate": 2.988730139179702e-06, + "loss": 0.6866, + "step": 355 + }, + { + "epoch": 0.13, + "learning_rate": 2.988693188816357e-06, + "loss": 0.6971, + "step": 356 + }, + { + "epoch": 0.13, + "learning_rate": 2.9886562384530116e-06, + "loss": 0.6669, + "step": 357 + }, + { + "epoch": 0.13, + "learning_rate": 2.9886192880896666e-06, + "loss": 0.653, + "step": 358 + }, + { + "epoch": 0.13, + "learning_rate": 2.988582337726321e-06, + "loss": 0.6573, + "step": 359 + }, + { + "epoch": 0.13, + "learning_rate": 2.9885453873629755e-06, + "loss": 0.6824, + "step": 360 + }, + { + "epoch": 0.13, + "learning_rate": 2.9885084369996305e-06, + "loss": 0.653, + "step": 361 + }, + { + "epoch": 0.13, + "learning_rate": 2.988471486636285e-06, + "loss": 0.6934, + "step": 362 + }, + { + "epoch": 0.13, + "learning_rate": 2.9884345362729402e-06, + "loss": 0.7202, + "step": 363 + }, + { + "epoch": 0.13, + "learning_rate": 2.988397585909595e-06, + "loss": 0.6729, + "step": 364 + }, + { + "epoch": 0.13, + "learning_rate": 2.9883606355462495e-06, + "loss": 0.6888, + "step": 365 + }, + { + "epoch": 0.14, + "learning_rate": 2.988323685182904e-06, + "loss": 0.6555, + "step": 366 + }, + { + "epoch": 0.14, + "learning_rate": 2.9882867348195592e-06, + "loss": 0.7093, + "step": 367 + }, + { + "epoch": 0.14, + "learning_rate": 2.988249784456214e-06, + "loss": 0.641, + "step": 368 + }, + { + "epoch": 0.14, + "learning_rate": 2.9882128340928685e-06, + "loss": 0.6659, + "step": 369 + }, + { + "epoch": 0.14, + "learning_rate": 2.9881758837295235e-06, + "loss": 0.673, + "step": 370 + }, + { + "epoch": 0.14, + "learning_rate": 2.988138933366178e-06, + "loss": 0.6862, + "step": 371 + }, + { + "epoch": 0.14, + "learning_rate": 2.988101983002833e-06, + "loss": 0.6712, + "step": 372 + }, + { + "epoch": 0.14, + "learning_rate": 2.9880650326394875e-06, + "loss": 0.6558, + "step": 373 + }, + { + "epoch": 0.14, + "learning_rate": 2.9880280822761425e-06, + "loss": 0.6684, + "step": 374 + }, + { + "epoch": 0.14, + "learning_rate": 2.987991131912797e-06, + "loss": 0.6831, + "step": 375 + }, + { + "epoch": 0.14, + "learning_rate": 2.987954181549452e-06, + "loss": 0.6894, + "step": 376 + }, + { + "epoch": 0.14, + "learning_rate": 2.987917231186107e-06, + "loss": 0.6487, + "step": 377 + }, + { + "epoch": 0.14, + "learning_rate": 2.9878802808227615e-06, + "loss": 0.6962, + "step": 378 + }, + { + "epoch": 0.14, + "learning_rate": 2.987843330459416e-06, + "loss": 0.6653, + "step": 379 + }, + { + "epoch": 0.14, + "learning_rate": 2.9878063800960708e-06, + "loss": 0.6395, + "step": 380 + }, + { + "epoch": 0.14, + "learning_rate": 2.987769429732726e-06, + "loss": 0.6761, + "step": 381 + }, + { + "epoch": 0.14, + "learning_rate": 2.9877324793693805e-06, + "loss": 0.674, + "step": 382 + }, + { + "epoch": 0.14, + "learning_rate": 2.9876955290060355e-06, + "loss": 0.7003, + "step": 383 + }, + { + "epoch": 0.14, + "learning_rate": 2.98765857864269e-06, + "loss": 0.6694, + "step": 384 + }, + { + "epoch": 0.14, + "learning_rate": 2.9876216282793448e-06, + "loss": 0.6502, + "step": 385 + }, + { + "epoch": 0.14, + "learning_rate": 2.9875846779159994e-06, + "loss": 0.6396, + "step": 386 + }, + { + "epoch": 0.14, + "learning_rate": 2.9875477275526545e-06, + "loss": 0.6931, + "step": 387 + }, + { + "epoch": 0.14, + "learning_rate": 2.987510777189309e-06, + "loss": 0.6969, + "step": 388 + }, + { + "epoch": 0.14, + "learning_rate": 2.9874738268259638e-06, + "loss": 0.6803, + "step": 389 + }, + { + "epoch": 0.14, + "learning_rate": 2.987436876462619e-06, + "loss": 0.6586, + "step": 390 + }, + { + "epoch": 0.14, + "learning_rate": 2.9873999260992735e-06, + "loss": 0.7101, + "step": 391 + }, + { + "epoch": 0.14, + "learning_rate": 2.987362975735928e-06, + "loss": 0.6845, + "step": 392 + }, + { + "epoch": 0.15, + "learning_rate": 2.9873260253725827e-06, + "loss": 0.6887, + "step": 393 + }, + { + "epoch": 0.15, + "learning_rate": 2.9872890750092378e-06, + "loss": 0.683, + "step": 394 + }, + { + "epoch": 0.15, + "learning_rate": 2.9872521246458924e-06, + "loss": 0.6331, + "step": 395 + }, + { + "epoch": 0.15, + "learning_rate": 2.9872151742825475e-06, + "loss": 0.6839, + "step": 396 + }, + { + "epoch": 0.15, + "learning_rate": 2.987178223919202e-06, + "loss": 0.6378, + "step": 397 + }, + { + "epoch": 0.15, + "learning_rate": 2.9871412735558568e-06, + "loss": 0.6266, + "step": 398 + }, + { + "epoch": 0.15, + "learning_rate": 2.9871043231925114e-06, + "loss": 0.6474, + "step": 399 + }, + { + "epoch": 0.15, + "learning_rate": 2.987067372829166e-06, + "loss": 0.6344, + "step": 400 + }, + { + "epoch": 0.15, + "learning_rate": 2.987030422465821e-06, + "loss": 0.6449, + "step": 401 + }, + { + "epoch": 0.15, + "learning_rate": 2.9869934721024757e-06, + "loss": 0.6531, + "step": 402 + }, + { + "epoch": 0.15, + "learning_rate": 2.9869565217391308e-06, + "loss": 0.6751, + "step": 403 + }, + { + "epoch": 0.15, + "learning_rate": 2.986919571375785e-06, + "loss": 0.688, + "step": 404 + }, + { + "epoch": 0.15, + "learning_rate": 2.98688262101244e-06, + "loss": 0.6731, + "step": 405 + }, + { + "epoch": 0.15, + "learning_rate": 2.9868456706490947e-06, + "loss": 0.6893, + "step": 406 + }, + { + "epoch": 0.15, + "learning_rate": 2.9868087202857498e-06, + "loss": 0.6884, + "step": 407 + }, + { + "epoch": 0.15, + "learning_rate": 2.9867717699224044e-06, + "loss": 0.6744, + "step": 408 + }, + { + "epoch": 0.15, + "learning_rate": 2.986734819559059e-06, + "loss": 0.6527, + "step": 409 + }, + { + "epoch": 0.15, + "learning_rate": 2.9866978691957137e-06, + "loss": 0.6496, + "step": 410 + }, + { + "epoch": 0.15, + "learning_rate": 2.9866609188323683e-06, + "loss": 0.6685, + "step": 411 + }, + { + "epoch": 0.15, + "learning_rate": 2.9866239684690234e-06, + "loss": 0.6623, + "step": 412 + }, + { + "epoch": 0.15, + "learning_rate": 2.986587018105678e-06, + "loss": 0.6504, + "step": 413 + }, + { + "epoch": 0.15, + "learning_rate": 2.986550067742333e-06, + "loss": 0.6272, + "step": 414 + }, + { + "epoch": 0.15, + "learning_rate": 2.9865131173789877e-06, + "loss": 0.6874, + "step": 415 + }, + { + "epoch": 0.15, + "learning_rate": 2.9864761670156423e-06, + "loss": 0.6428, + "step": 416 + }, + { + "epoch": 0.15, + "learning_rate": 2.986439216652297e-06, + "loss": 0.6476, + "step": 417 + }, + { + "epoch": 0.15, + "learning_rate": 2.986402266288952e-06, + "loss": 0.6647, + "step": 418 + }, + { + "epoch": 0.15, + "learning_rate": 2.9863653159256067e-06, + "loss": 0.7037, + "step": 419 + }, + { + "epoch": 0.16, + "learning_rate": 2.9863283655622613e-06, + "loss": 0.6922, + "step": 420 + }, + { + "epoch": 0.16, + "learning_rate": 2.9862914151989164e-06, + "loss": 0.6616, + "step": 421 + }, + { + "epoch": 0.16, + "learning_rate": 2.986254464835571e-06, + "loss": 0.652, + "step": 422 + }, + { + "epoch": 0.16, + "learning_rate": 2.9862175144722256e-06, + "loss": 0.6553, + "step": 423 + }, + { + "epoch": 0.16, + "learning_rate": 2.9861805641088803e-06, + "loss": 0.6934, + "step": 424 + }, + { + "epoch": 0.16, + "learning_rate": 2.9861436137455353e-06, + "loss": 0.6464, + "step": 425 + }, + { + "epoch": 0.16, + "learning_rate": 2.98610666338219e-06, + "loss": 0.6736, + "step": 426 + }, + { + "epoch": 0.16, + "learning_rate": 2.986069713018845e-06, + "loss": 0.6263, + "step": 427 + }, + { + "epoch": 0.16, + "learning_rate": 2.9860327626554997e-06, + "loss": 0.6445, + "step": 428 + }, + { + "epoch": 0.16, + "learning_rate": 2.9859958122921543e-06, + "loss": 0.6888, + "step": 429 + }, + { + "epoch": 0.16, + "learning_rate": 2.985958861928809e-06, + "loss": 0.6885, + "step": 430 + }, + { + "epoch": 0.16, + "learning_rate": 2.9859219115654636e-06, + "loss": 0.6658, + "step": 431 + }, + { + "epoch": 0.16, + "learning_rate": 2.9858849612021186e-06, + "loss": 0.6716, + "step": 432 + }, + { + "epoch": 0.16, + "learning_rate": 2.9858480108387733e-06, + "loss": 0.667, + "step": 433 + }, + { + "epoch": 0.16, + "learning_rate": 2.9858110604754283e-06, + "loss": 0.6441, + "step": 434 + }, + { + "epoch": 0.16, + "learning_rate": 2.985774110112083e-06, + "loss": 0.6139, + "step": 435 + }, + { + "epoch": 0.16, + "learning_rate": 2.9857371597487376e-06, + "loss": 0.6186, + "step": 436 + }, + { + "epoch": 0.16, + "learning_rate": 2.9857002093853922e-06, + "loss": 0.6475, + "step": 437 + }, + { + "epoch": 0.16, + "learning_rate": 2.9856632590220473e-06, + "loss": 0.6557, + "step": 438 + }, + { + "epoch": 0.16, + "learning_rate": 2.985626308658702e-06, + "loss": 0.6823, + "step": 439 + }, + { + "epoch": 0.16, + "learning_rate": 2.9855893582953566e-06, + "loss": 0.7024, + "step": 440 + }, + { + "epoch": 0.16, + "learning_rate": 2.9855524079320116e-06, + "loss": 0.6656, + "step": 441 + }, + { + "epoch": 0.16, + "learning_rate": 2.9855154575686663e-06, + "loss": 0.6845, + "step": 442 + }, + { + "epoch": 0.16, + "learning_rate": 2.985478507205321e-06, + "loss": 0.6622, + "step": 443 + }, + { + "epoch": 0.16, + "learning_rate": 2.9854415568419755e-06, + "loss": 0.6547, + "step": 444 + }, + { + "epoch": 0.16, + "learning_rate": 2.9854046064786306e-06, + "loss": 0.7092, + "step": 445 + }, + { + "epoch": 0.16, + "learning_rate": 2.9853676561152852e-06, + "loss": 0.6399, + "step": 446 + }, + { + "epoch": 0.17, + "learning_rate": 2.9853307057519403e-06, + "loss": 0.6566, + "step": 447 + }, + { + "epoch": 0.17, + "learning_rate": 2.985293755388595e-06, + "loss": 0.6422, + "step": 448 + }, + { + "epoch": 0.17, + "learning_rate": 2.985256805025249e-06, + "loss": 0.6816, + "step": 449 + }, + { + "epoch": 0.17, + "learning_rate": 2.985219854661904e-06, + "loss": 0.7145, + "step": 450 + }, + { + "epoch": 0.17, + "learning_rate": 2.985182904298559e-06, + "loss": 0.6112, + "step": 451 + }, + { + "epoch": 0.17, + "learning_rate": 2.985145953935214e-06, + "loss": 0.6472, + "step": 452 + }, + { + "epoch": 0.17, + "learning_rate": 2.9851090035718685e-06, + "loss": 0.6663, + "step": 453 + }, + { + "epoch": 0.17, + "learning_rate": 2.9850720532085236e-06, + "loss": 0.6688, + "step": 454 + }, + { + "epoch": 0.17, + "learning_rate": 2.985035102845178e-06, + "loss": 0.6498, + "step": 455 + }, + { + "epoch": 0.17, + "learning_rate": 2.984998152481833e-06, + "loss": 0.6817, + "step": 456 + }, + { + "epoch": 0.17, + "learning_rate": 2.9849612021184875e-06, + "loss": 0.7134, + "step": 457 + }, + { + "epoch": 0.17, + "learning_rate": 2.9849242517551426e-06, + "loss": 0.6319, + "step": 458 + }, + { + "epoch": 0.17, + "learning_rate": 2.984887301391797e-06, + "loss": 0.6781, + "step": 459 + }, + { + "epoch": 0.17, + "learning_rate": 2.984850351028452e-06, + "loss": 0.6907, + "step": 460 + }, + { + "epoch": 0.17, + "learning_rate": 2.9848134006651065e-06, + "loss": 0.7067, + "step": 461 + }, + { + "epoch": 0.17, + "learning_rate": 2.984776450301761e-06, + "loss": 0.6786, + "step": 462 + }, + { + "epoch": 0.17, + "learning_rate": 2.984739499938416e-06, + "loss": 0.645, + "step": 463 + }, + { + "epoch": 0.17, + "learning_rate": 2.984702549575071e-06, + "loss": 0.65, + "step": 464 + }, + { + "epoch": 0.17, + "learning_rate": 2.984665599211726e-06, + "loss": 0.7057, + "step": 465 + }, + { + "epoch": 0.17, + "learning_rate": 2.9846286488483805e-06, + "loss": 0.6791, + "step": 466 + }, + { + "epoch": 0.17, + "learning_rate": 2.984591698485035e-06, + "loss": 0.678, + "step": 467 + }, + { + "epoch": 0.17, + "learning_rate": 2.9845547481216898e-06, + "loss": 0.646, + "step": 468 + }, + { + "epoch": 0.17, + "learning_rate": 2.9845177977583444e-06, + "loss": 0.6781, + "step": 469 + }, + { + "epoch": 0.17, + "learning_rate": 2.9844808473949995e-06, + "loss": 0.6174, + "step": 470 + }, + { + "epoch": 0.17, + "learning_rate": 2.984443897031654e-06, + "loss": 0.6559, + "step": 471 + }, + { + "epoch": 0.17, + "learning_rate": 2.984406946668309e-06, + "loss": 0.6883, + "step": 472 + }, + { + "epoch": 0.17, + "learning_rate": 2.984369996304964e-06, + "loss": 0.6649, + "step": 473 + }, + { + "epoch": 0.18, + "learning_rate": 2.9843330459416184e-06, + "loss": 0.5986, + "step": 474 + }, + { + "epoch": 0.18, + "learning_rate": 2.984296095578273e-06, + "loss": 0.6722, + "step": 475 + }, + { + "epoch": 0.18, + "learning_rate": 2.984259145214928e-06, + "loss": 0.7196, + "step": 476 + }, + { + "epoch": 0.18, + "learning_rate": 2.9842221948515828e-06, + "loss": 0.6995, + "step": 477 + }, + { + "epoch": 0.18, + "learning_rate": 2.984185244488238e-06, + "loss": 0.695, + "step": 478 + }, + { + "epoch": 0.18, + "learning_rate": 2.9841482941248925e-06, + "loss": 0.6819, + "step": 479 + }, + { + "epoch": 0.18, + "learning_rate": 2.984111343761547e-06, + "loss": 0.6389, + "step": 480 + }, + { + "epoch": 0.18, + "learning_rate": 2.9840743933982017e-06, + "loss": 0.6494, + "step": 481 + }, + { + "epoch": 0.18, + "learning_rate": 2.9840374430348564e-06, + "loss": 0.6767, + "step": 482 + }, + { + "epoch": 0.18, + "learning_rate": 2.9840004926715114e-06, + "loss": 0.6669, + "step": 483 + }, + { + "epoch": 0.18, + "learning_rate": 2.983963542308166e-06, + "loss": 0.6567, + "step": 484 + }, + { + "epoch": 0.18, + "learning_rate": 2.983926591944821e-06, + "loss": 0.6624, + "step": 485 + }, + { + "epoch": 0.18, + "learning_rate": 2.9838896415814758e-06, + "loss": 0.7298, + "step": 486 + }, + { + "epoch": 0.18, + "learning_rate": 2.9838526912181304e-06, + "loss": 0.6649, + "step": 487 + }, + { + "epoch": 0.18, + "learning_rate": 2.983815740854785e-06, + "loss": 0.6726, + "step": 488 + }, + { + "epoch": 0.18, + "learning_rate": 2.9837787904914397e-06, + "loss": 0.6385, + "step": 489 + }, + { + "epoch": 0.18, + "learning_rate": 2.9837418401280947e-06, + "loss": 0.7035, + "step": 490 + }, + { + "epoch": 0.18, + "learning_rate": 2.9837048897647494e-06, + "loss": 0.7128, + "step": 491 + }, + { + "epoch": 0.18, + "learning_rate": 2.9836679394014044e-06, + "loss": 0.7244, + "step": 492 + }, + { + "epoch": 0.18, + "learning_rate": 2.983630989038059e-06, + "loss": 0.6812, + "step": 493 + }, + { + "epoch": 0.18, + "learning_rate": 2.9835940386747137e-06, + "loss": 0.7126, + "step": 494 + }, + { + "epoch": 0.18, + "learning_rate": 2.9835570883113683e-06, + "loss": 0.6557, + "step": 495 + }, + { + "epoch": 0.18, + "learning_rate": 2.9835201379480234e-06, + "loss": 0.6349, + "step": 496 + }, + { + "epoch": 0.18, + "learning_rate": 2.983483187584678e-06, + "loss": 0.6563, + "step": 497 + }, + { + "epoch": 0.18, + "learning_rate": 2.983446237221333e-06, + "loss": 0.6438, + "step": 498 + }, + { + "epoch": 0.18, + "learning_rate": 2.9834092868579877e-06, + "loss": 0.6486, + "step": 499 + }, + { + "epoch": 0.18, + "learning_rate": 2.983372336494642e-06, + "loss": 0.6712, + "step": 500 + }, + { + "epoch": 0.19, + "learning_rate": 2.983335386131297e-06, + "loss": 0.6921, + "step": 501 + }, + { + "epoch": 0.19, + "learning_rate": 2.9832984357679516e-06, + "loss": 0.6821, + "step": 502 + }, + { + "epoch": 0.19, + "learning_rate": 2.9832614854046067e-06, + "loss": 0.6419, + "step": 503 + }, + { + "epoch": 0.19, + "learning_rate": 2.9832245350412613e-06, + "loss": 0.6687, + "step": 504 + }, + { + "epoch": 0.19, + "learning_rate": 2.983187584677916e-06, + "loss": 0.6471, + "step": 505 + }, + { + "epoch": 0.19, + "learning_rate": 2.9831506343145706e-06, + "loss": 0.6843, + "step": 506 + }, + { + "epoch": 0.19, + "learning_rate": 2.9831136839512257e-06, + "loss": 0.692, + "step": 507 + }, + { + "epoch": 0.19, + "learning_rate": 2.9830767335878803e-06, + "loss": 0.6514, + "step": 508 + }, + { + "epoch": 0.19, + "learning_rate": 2.983039783224535e-06, + "loss": 0.6234, + "step": 509 + }, + { + "epoch": 0.19, + "learning_rate": 2.98300283286119e-06, + "loss": 0.6451, + "step": 510 + }, + { + "epoch": 0.19, + "learning_rate": 2.9829658824978446e-06, + "loss": 0.6481, + "step": 511 + }, + { + "epoch": 0.19, + "learning_rate": 2.9829289321344993e-06, + "loss": 0.6799, + "step": 512 + }, + { + "epoch": 0.19, + "learning_rate": 2.982891981771154e-06, + "loss": 0.6738, + "step": 513 + }, + { + "epoch": 0.19, + "learning_rate": 2.982855031407809e-06, + "loss": 0.6332, + "step": 514 + }, + { + "epoch": 0.19, + "learning_rate": 2.9828180810444636e-06, + "loss": 0.6987, + "step": 515 + }, + { + "epoch": 0.19, + "learning_rate": 2.9827811306811187e-06, + "loss": 0.6556, + "step": 516 + }, + { + "epoch": 0.19, + "learning_rate": 2.9827441803177733e-06, + "loss": 0.6641, + "step": 517 + }, + { + "epoch": 0.19, + "learning_rate": 2.982707229954428e-06, + "loss": 0.6709, + "step": 518 + }, + { + "epoch": 0.19, + "learning_rate": 2.9826702795910826e-06, + "loss": 0.6753, + "step": 519 + }, + { + "epoch": 0.19, + "learning_rate": 2.9826333292277372e-06, + "loss": 0.6462, + "step": 520 + }, + { + "epoch": 0.19, + "learning_rate": 2.9825963788643923e-06, + "loss": 0.6865, + "step": 521 + }, + { + "epoch": 0.19, + "learning_rate": 2.982559428501047e-06, + "loss": 0.6948, + "step": 522 + }, + { + "epoch": 0.19, + "learning_rate": 2.982522478137702e-06, + "loss": 0.6717, + "step": 523 + }, + { + "epoch": 0.19, + "learning_rate": 2.9824855277743566e-06, + "loss": 0.7206, + "step": 524 + }, + { + "epoch": 0.19, + "learning_rate": 2.9824485774110113e-06, + "loss": 0.6421, + "step": 525 + }, + { + "epoch": 0.19, + "learning_rate": 2.982411627047666e-06, + "loss": 0.664, + "step": 526 + }, + { + "epoch": 0.19, + "learning_rate": 2.982374676684321e-06, + "loss": 0.6024, + "step": 527 + }, + { + "epoch": 0.19, + "learning_rate": 2.9823377263209756e-06, + "loss": 0.6658, + "step": 528 + }, + { + "epoch": 0.2, + "learning_rate": 2.9823007759576302e-06, + "loss": 0.661, + "step": 529 + }, + { + "epoch": 0.2, + "learning_rate": 2.9822638255942853e-06, + "loss": 0.6674, + "step": 530 + }, + { + "epoch": 0.2, + "learning_rate": 2.98222687523094e-06, + "loss": 0.6911, + "step": 531 + }, + { + "epoch": 0.2, + "learning_rate": 2.9821899248675946e-06, + "loss": 0.716, + "step": 532 + }, + { + "epoch": 0.2, + "learning_rate": 2.982152974504249e-06, + "loss": 0.6676, + "step": 533 + }, + { + "epoch": 0.2, + "learning_rate": 2.9821160241409042e-06, + "loss": 0.6536, + "step": 534 + }, + { + "epoch": 0.2, + "learning_rate": 2.982079073777559e-06, + "loss": 0.6284, + "step": 535 + }, + { + "epoch": 0.2, + "learning_rate": 2.982042123414214e-06, + "loss": 0.6833, + "step": 536 + }, + { + "epoch": 0.2, + "learning_rate": 2.9820051730508686e-06, + "loss": 0.668, + "step": 537 + }, + { + "epoch": 0.2, + "learning_rate": 2.981968222687523e-06, + "loss": 0.6601, + "step": 538 + }, + { + "epoch": 0.2, + "learning_rate": 2.981931272324178e-06, + "loss": 0.6711, + "step": 539 + }, + { + "epoch": 0.2, + "learning_rate": 2.9818943219608325e-06, + "loss": 0.6319, + "step": 540 + }, + { + "epoch": 0.2, + "learning_rate": 2.9818573715974876e-06, + "loss": 0.6437, + "step": 541 + }, + { + "epoch": 0.2, + "learning_rate": 2.981820421234142e-06, + "loss": 0.6165, + "step": 542 + }, + { + "epoch": 0.2, + "learning_rate": 2.9817834708707972e-06, + "loss": 0.6117, + "step": 543 + }, + { + "epoch": 0.2, + "learning_rate": 2.9817465205074515e-06, + "loss": 0.6568, + "step": 544 + }, + { + "epoch": 0.2, + "learning_rate": 2.9817095701441065e-06, + "loss": 0.6446, + "step": 545 + }, + { + "epoch": 0.2, + "learning_rate": 2.981672619780761e-06, + "loss": 0.6791, + "step": 546 + }, + { + "epoch": 0.2, + "learning_rate": 2.9816356694174162e-06, + "loss": 0.6363, + "step": 547 + }, + { + "epoch": 0.2, + "learning_rate": 2.981598719054071e-06, + "loss": 0.6301, + "step": 548 + }, + { + "epoch": 0.2, + "learning_rate": 2.9815617686907255e-06, + "loss": 0.6063, + "step": 549 + }, + { + "epoch": 0.2, + "learning_rate": 2.98152481832738e-06, + "loss": 0.6793, + "step": 550 + }, + { + "epoch": 0.2, + "learning_rate": 2.9814878679640348e-06, + "loss": 0.6749, + "step": 551 + }, + { + "epoch": 0.2, + "learning_rate": 2.98145091760069e-06, + "loss": 0.6646, + "step": 552 + }, + { + "epoch": 0.2, + "learning_rate": 2.9814139672373445e-06, + "loss": 0.6493, + "step": 553 + }, + { + "epoch": 0.2, + "learning_rate": 2.9813770168739995e-06, + "loss": 0.6705, + "step": 554 + }, + { + "epoch": 0.2, + "learning_rate": 2.981340066510654e-06, + "loss": 0.6771, + "step": 555 + }, + { + "epoch": 0.21, + "learning_rate": 2.981303116147309e-06, + "loss": 0.6432, + "step": 556 + }, + { + "epoch": 0.21, + "learning_rate": 2.9812661657839634e-06, + "loss": 0.6213, + "step": 557 + }, + { + "epoch": 0.21, + "learning_rate": 2.981229215420618e-06, + "loss": 0.6224, + "step": 558 + }, + { + "epoch": 0.21, + "learning_rate": 2.981192265057273e-06, + "loss": 0.681, + "step": 559 + }, + { + "epoch": 0.21, + "learning_rate": 2.9811553146939278e-06, + "loss": 0.6955, + "step": 560 + }, + { + "epoch": 0.21, + "learning_rate": 2.981118364330583e-06, + "loss": 0.6695, + "step": 561 + }, + { + "epoch": 0.21, + "learning_rate": 2.9810814139672375e-06, + "loss": 0.6434, + "step": 562 + }, + { + "epoch": 0.21, + "learning_rate": 2.981044463603892e-06, + "loss": 0.7034, + "step": 563 + }, + { + "epoch": 0.21, + "learning_rate": 2.9810075132405467e-06, + "loss": 0.6867, + "step": 564 + }, + { + "epoch": 0.21, + "learning_rate": 2.980970562877202e-06, + "loss": 0.7039, + "step": 565 + }, + { + "epoch": 0.21, + "learning_rate": 2.9809336125138564e-06, + "loss": 0.6474, + "step": 566 + }, + { + "epoch": 0.21, + "learning_rate": 2.9808966621505115e-06, + "loss": 0.6704, + "step": 567 + }, + { + "epoch": 0.21, + "learning_rate": 2.980859711787166e-06, + "loss": 0.6862, + "step": 568 + }, + { + "epoch": 0.21, + "learning_rate": 2.9808227614238208e-06, + "loss": 0.632, + "step": 569 + }, + { + "epoch": 0.21, + "learning_rate": 2.9807858110604754e-06, + "loss": 0.6668, + "step": 570 + }, + { + "epoch": 0.21, + "learning_rate": 2.98074886069713e-06, + "loss": 0.6661, + "step": 571 + }, + { + "epoch": 0.21, + "learning_rate": 2.980711910333785e-06, + "loss": 0.6967, + "step": 572 + }, + { + "epoch": 0.21, + "learning_rate": 2.9806749599704397e-06, + "loss": 0.6197, + "step": 573 + }, + { + "epoch": 0.21, + "learning_rate": 2.980638009607095e-06, + "loss": 0.6462, + "step": 574 + }, + { + "epoch": 0.21, + "learning_rate": 2.9806010592437494e-06, + "loss": 0.671, + "step": 575 + }, + { + "epoch": 0.21, + "learning_rate": 2.980564108880404e-06, + "loss": 0.6551, + "step": 576 + }, + { + "epoch": 0.21, + "learning_rate": 2.9805271585170587e-06, + "loss": 0.7453, + "step": 577 + }, + { + "epoch": 0.21, + "learning_rate": 2.9804902081537133e-06, + "loss": 0.6223, + "step": 578 + }, + { + "epoch": 0.21, + "learning_rate": 2.9804532577903684e-06, + "loss": 0.6858, + "step": 579 + }, + { + "epoch": 0.21, + "learning_rate": 2.980416307427023e-06, + "loss": 0.6229, + "step": 580 + }, + { + "epoch": 0.21, + "learning_rate": 2.980379357063678e-06, + "loss": 0.6161, + "step": 581 + }, + { + "epoch": 0.21, + "learning_rate": 2.9803424067003327e-06, + "loss": 0.6491, + "step": 582 + }, + { + "epoch": 0.22, + "learning_rate": 2.9803054563369874e-06, + "loss": 0.6582, + "step": 583 + }, + { + "epoch": 0.22, + "learning_rate": 2.980268505973642e-06, + "loss": 0.6389, + "step": 584 + }, + { + "epoch": 0.22, + "learning_rate": 2.980231555610297e-06, + "loss": 0.6341, + "step": 585 + }, + { + "epoch": 0.22, + "learning_rate": 2.9801946052469517e-06, + "loss": 0.6008, + "step": 586 + }, + { + "epoch": 0.22, + "learning_rate": 2.9801576548836068e-06, + "loss": 0.6888, + "step": 587 + }, + { + "epoch": 0.22, + "learning_rate": 2.9801207045202614e-06, + "loss": 0.6762, + "step": 588 + }, + { + "epoch": 0.22, + "learning_rate": 2.9800837541569156e-06, + "loss": 0.628, + "step": 589 + }, + { + "epoch": 0.22, + "learning_rate": 2.9800468037935707e-06, + "loss": 0.6812, + "step": 590 + }, + { + "epoch": 0.22, + "learning_rate": 2.9800098534302253e-06, + "loss": 0.7191, + "step": 591 + }, + { + "epoch": 0.22, + "learning_rate": 2.9799729030668804e-06, + "loss": 0.668, + "step": 592 + }, + { + "epoch": 0.22, + "learning_rate": 2.979935952703535e-06, + "loss": 0.7042, + "step": 593 + }, + { + "epoch": 0.22, + "learning_rate": 2.97989900234019e-06, + "loss": 0.6774, + "step": 594 + }, + { + "epoch": 0.22, + "learning_rate": 2.9798620519768443e-06, + "loss": 0.6462, + "step": 595 + }, + { + "epoch": 0.22, + "learning_rate": 2.9798251016134993e-06, + "loss": 0.6412, + "step": 596 + }, + { + "epoch": 0.22, + "learning_rate": 2.979788151250154e-06, + "loss": 0.6808, + "step": 597 + }, + { + "epoch": 0.22, + "learning_rate": 2.9797512008868086e-06, + "loss": 0.6413, + "step": 598 + }, + { + "epoch": 0.22, + "learning_rate": 2.9797142505234637e-06, + "loss": 0.6714, + "step": 599 + }, + { + "epoch": 0.22, + "learning_rate": 2.9796773001601183e-06, + "loss": 0.6699, + "step": 600 + }, + { + "epoch": 0.22, + "learning_rate": 2.979640349796773e-06, + "loss": 0.6932, + "step": 601 + }, + { + "epoch": 0.22, + "learning_rate": 2.9796033994334276e-06, + "loss": 0.6728, + "step": 602 + }, + { + "epoch": 0.22, + "learning_rate": 2.9795664490700826e-06, + "loss": 0.6684, + "step": 603 + }, + { + "epoch": 0.22, + "learning_rate": 2.9795294987067373e-06, + "loss": 0.6482, + "step": 604 + }, + { + "epoch": 0.22, + "learning_rate": 2.9794925483433923e-06, + "loss": 0.6673, + "step": 605 + }, + { + "epoch": 0.22, + "learning_rate": 2.979455597980047e-06, + "loss": 0.6871, + "step": 606 + }, + { + "epoch": 0.22, + "learning_rate": 2.9794186476167016e-06, + "loss": 0.6817, + "step": 607 + }, + { + "epoch": 0.22, + "learning_rate": 2.9793816972533562e-06, + "loss": 0.6233, + "step": 608 + }, + { + "epoch": 0.22, + "learning_rate": 2.979344746890011e-06, + "loss": 0.6599, + "step": 609 + }, + { + "epoch": 0.23, + "learning_rate": 2.979307796526666e-06, + "loss": 0.7121, + "step": 610 + }, + { + "epoch": 0.23, + "learning_rate": 2.9792708461633206e-06, + "loss": 0.6793, + "step": 611 + }, + { + "epoch": 0.23, + "learning_rate": 2.9792338957999756e-06, + "loss": 0.6597, + "step": 612 + }, + { + "epoch": 0.23, + "learning_rate": 2.9791969454366303e-06, + "loss": 0.6818, + "step": 613 + }, + { + "epoch": 0.23, + "learning_rate": 2.979159995073285e-06, + "loss": 0.6518, + "step": 614 + }, + { + "epoch": 0.23, + "learning_rate": 2.9791230447099395e-06, + "loss": 0.6231, + "step": 615 + }, + { + "epoch": 0.23, + "learning_rate": 2.9790860943465946e-06, + "loss": 0.6656, + "step": 616 + }, + { + "epoch": 0.23, + "learning_rate": 2.9790491439832492e-06, + "loss": 0.6369, + "step": 617 + }, + { + "epoch": 0.23, + "learning_rate": 2.979012193619904e-06, + "loss": 0.6597, + "step": 618 + }, + { + "epoch": 0.23, + "learning_rate": 2.978975243256559e-06, + "loss": 0.6388, + "step": 619 + }, + { + "epoch": 0.23, + "learning_rate": 2.9789382928932136e-06, + "loss": 0.6248, + "step": 620 + }, + { + "epoch": 0.23, + "learning_rate": 2.978901342529868e-06, + "loss": 0.6334, + "step": 621 + }, + { + "epoch": 0.23, + "learning_rate": 2.978864392166523e-06, + "loss": 0.6446, + "step": 622 + }, + { + "epoch": 0.23, + "learning_rate": 2.978827441803178e-06, + "loss": 0.6516, + "step": 623 + }, + { + "epoch": 0.23, + "learning_rate": 2.9787904914398325e-06, + "loss": 0.6279, + "step": 624 + }, + { + "epoch": 0.23, + "learning_rate": 2.9787535410764876e-06, + "loss": 0.6346, + "step": 625 + }, + { + "epoch": 0.23, + "learning_rate": 2.9787165907131422e-06, + "loss": 0.6126, + "step": 626 + }, + { + "epoch": 0.23, + "learning_rate": 2.978679640349797e-06, + "loss": 0.6763, + "step": 627 + }, + { + "epoch": 0.23, + "learning_rate": 2.9786426899864515e-06, + "loss": 0.6467, + "step": 628 + }, + { + "epoch": 0.23, + "learning_rate": 2.978605739623106e-06, + "loss": 0.7004, + "step": 629 + }, + { + "epoch": 0.23, + "learning_rate": 2.978568789259761e-06, + "loss": 0.6944, + "step": 630 + }, + { + "epoch": 0.23, + "learning_rate": 2.978531838896416e-06, + "loss": 0.6528, + "step": 631 + }, + { + "epoch": 0.23, + "learning_rate": 2.978494888533071e-06, + "loss": 0.6542, + "step": 632 + }, + { + "epoch": 0.23, + "learning_rate": 2.9784579381697255e-06, + "loss": 0.6817, + "step": 633 + }, + { + "epoch": 0.23, + "learning_rate": 2.97842098780638e-06, + "loss": 0.6416, + "step": 634 + }, + { + "epoch": 0.23, + "learning_rate": 2.978384037443035e-06, + "loss": 0.6669, + "step": 635 + }, + { + "epoch": 0.23, + "learning_rate": 2.97834708707969e-06, + "loss": 0.6785, + "step": 636 + }, + { + "epoch": 0.24, + "learning_rate": 2.9783101367163445e-06, + "loss": 0.6719, + "step": 637 + }, + { + "epoch": 0.24, + "learning_rate": 2.978273186352999e-06, + "loss": 0.653, + "step": 638 + }, + { + "epoch": 0.24, + "learning_rate": 2.978236235989654e-06, + "loss": 0.6627, + "step": 639 + }, + { + "epoch": 0.24, + "learning_rate": 2.9781992856263084e-06, + "loss": 0.6167, + "step": 640 + }, + { + "epoch": 0.24, + "learning_rate": 2.9781623352629635e-06, + "loss": 0.6181, + "step": 641 + }, + { + "epoch": 0.24, + "learning_rate": 2.978125384899618e-06, + "loss": 0.6233, + "step": 642 + }, + { + "epoch": 0.24, + "learning_rate": 2.978088434536273e-06, + "loss": 0.6309, + "step": 643 + }, + { + "epoch": 0.24, + "learning_rate": 2.978051484172928e-06, + "loss": 0.7536, + "step": 644 + }, + { + "epoch": 0.24, + "learning_rate": 2.978014533809583e-06, + "loss": 0.704, + "step": 645 + }, + { + "epoch": 0.24, + "learning_rate": 2.977977583446237e-06, + "loss": 0.6388, + "step": 646 + }, + { + "epoch": 0.24, + "learning_rate": 2.977940633082892e-06, + "loss": 0.7227, + "step": 647 + }, + { + "epoch": 0.24, + "learning_rate": 2.9779036827195468e-06, + "loss": 0.6058, + "step": 648 + }, + { + "epoch": 0.24, + "learning_rate": 2.9778667323562014e-06, + "loss": 0.695, + "step": 649 + }, + { + "epoch": 0.24, + "learning_rate": 2.9778297819928565e-06, + "loss": 0.6952, + "step": 650 + }, + { + "epoch": 0.24, + "learning_rate": 2.977792831629511e-06, + "loss": 0.6593, + "step": 651 + }, + { + "epoch": 0.24, + "learning_rate": 2.9777558812661657e-06, + "loss": 0.6491, + "step": 652 + }, + { + "epoch": 0.24, + "learning_rate": 2.9777189309028204e-06, + "loss": 0.6849, + "step": 653 + }, + { + "epoch": 0.24, + "learning_rate": 2.9776819805394754e-06, + "loss": 0.702, + "step": 654 + }, + { + "epoch": 0.24, + "learning_rate": 2.97764503017613e-06, + "loss": 0.6469, + "step": 655 + }, + { + "epoch": 0.24, + "learning_rate": 2.977608079812785e-06, + "loss": 0.6373, + "step": 656 + }, + { + "epoch": 0.24, + "learning_rate": 2.9775711294494398e-06, + "loss": 0.6748, + "step": 657 + }, + { + "epoch": 0.24, + "learning_rate": 2.9775341790860944e-06, + "loss": 0.6537, + "step": 658 + }, + { + "epoch": 0.24, + "learning_rate": 2.977497228722749e-06, + "loss": 0.6592, + "step": 659 + }, + { + "epoch": 0.24, + "learning_rate": 2.9774602783594037e-06, + "loss": 0.6648, + "step": 660 + }, + { + "epoch": 0.24, + "learning_rate": 2.9774233279960587e-06, + "loss": 0.6652, + "step": 661 + }, + { + "epoch": 0.24, + "learning_rate": 2.9773863776327134e-06, + "loss": 0.6349, + "step": 662 + }, + { + "epoch": 0.24, + "learning_rate": 2.9773494272693684e-06, + "loss": 0.6276, + "step": 663 + }, + { + "epoch": 0.25, + "learning_rate": 2.977312476906023e-06, + "loss": 0.6387, + "step": 664 + }, + { + "epoch": 0.25, + "learning_rate": 2.9772755265426777e-06, + "loss": 0.6593, + "step": 665 + }, + { + "epoch": 0.25, + "learning_rate": 2.9772385761793324e-06, + "loss": 0.6601, + "step": 666 + }, + { + "epoch": 0.25, + "learning_rate": 2.9772016258159874e-06, + "loss": 0.5901, + "step": 667 + }, + { + "epoch": 0.25, + "learning_rate": 2.977164675452642e-06, + "loss": 0.6736, + "step": 668 + }, + { + "epoch": 0.25, + "learning_rate": 2.9771277250892967e-06, + "loss": 0.6248, + "step": 669 + }, + { + "epoch": 0.25, + "learning_rate": 2.9770907747259517e-06, + "loss": 0.6673, + "step": 670 + }, + { + "epoch": 0.25, + "learning_rate": 2.9770538243626064e-06, + "loss": 0.6354, + "step": 671 + }, + { + "epoch": 0.25, + "learning_rate": 2.977016873999261e-06, + "loss": 0.6899, + "step": 672 + }, + { + "epoch": 0.25, + "learning_rate": 2.9769799236359157e-06, + "loss": 0.6624, + "step": 673 + }, + { + "epoch": 0.25, + "learning_rate": 2.9769429732725707e-06, + "loss": 0.6639, + "step": 674 + }, + { + "epoch": 0.25, + "learning_rate": 2.9769060229092254e-06, + "loss": 0.6748, + "step": 675 + }, + { + "epoch": 0.25, + "learning_rate": 2.9768690725458804e-06, + "loss": 0.6061, + "step": 676 + }, + { + "epoch": 0.25, + "learning_rate": 2.976832122182535e-06, + "loss": 0.6521, + "step": 677 + }, + { + "epoch": 0.25, + "learning_rate": 2.9767951718191897e-06, + "loss": 0.7058, + "step": 678 + }, + { + "epoch": 0.25, + "learning_rate": 2.9767582214558443e-06, + "loss": 0.6308, + "step": 679 + }, + { + "epoch": 0.25, + "learning_rate": 2.976721271092499e-06, + "loss": 0.6221, + "step": 680 + }, + { + "epoch": 0.25, + "learning_rate": 2.976684320729154e-06, + "loss": 0.6636, + "step": 681 + }, + { + "epoch": 0.25, + "learning_rate": 2.9766473703658087e-06, + "loss": 0.6993, + "step": 682 + }, + { + "epoch": 0.25, + "learning_rate": 2.9766104200024637e-06, + "loss": 0.6782, + "step": 683 + }, + { + "epoch": 0.25, + "learning_rate": 2.9765734696391183e-06, + "loss": 0.6587, + "step": 684 + }, + { + "epoch": 0.25, + "learning_rate": 2.976536519275773e-06, + "loss": 0.6312, + "step": 685 + }, + { + "epoch": 0.25, + "learning_rate": 2.9764995689124276e-06, + "loss": 0.6029, + "step": 686 + }, + { + "epoch": 0.25, + "learning_rate": 2.9764626185490827e-06, + "loss": 0.6882, + "step": 687 + }, + { + "epoch": 0.25, + "learning_rate": 2.9764256681857373e-06, + "loss": 0.6975, + "step": 688 + }, + { + "epoch": 0.25, + "learning_rate": 2.976388717822392e-06, + "loss": 0.7111, + "step": 689 + }, + { + "epoch": 0.25, + "learning_rate": 2.9763517674590466e-06, + "loss": 0.7386, + "step": 690 + }, + { + "epoch": 0.26, + "learning_rate": 2.9763148170957012e-06, + "loss": 0.6298, + "step": 691 + }, + { + "epoch": 0.26, + "learning_rate": 2.9762778667323563e-06, + "loss": 0.6635, + "step": 692 + }, + { + "epoch": 0.26, + "learning_rate": 2.976240916369011e-06, + "loss": 0.6496, + "step": 693 + }, + { + "epoch": 0.26, + "learning_rate": 2.976203966005666e-06, + "loss": 0.63, + "step": 694 + }, + { + "epoch": 0.26, + "learning_rate": 2.9761670156423206e-06, + "loss": 0.6721, + "step": 695 + }, + { + "epoch": 0.26, + "learning_rate": 2.9761300652789753e-06, + "loss": 0.6357, + "step": 696 + }, + { + "epoch": 0.26, + "learning_rate": 2.97609311491563e-06, + "loss": 0.6167, + "step": 697 + }, + { + "epoch": 0.26, + "learning_rate": 2.9760561645522845e-06, + "loss": 0.6586, + "step": 698 + }, + { + "epoch": 0.26, + "learning_rate": 2.9760192141889396e-06, + "loss": 0.6404, + "step": 699 + }, + { + "epoch": 0.26, + "learning_rate": 2.9759822638255942e-06, + "loss": 0.6836, + "step": 700 + }, + { + "epoch": 0.26, + "learning_rate": 2.9759453134622493e-06, + "loss": 0.7118, + "step": 701 + }, + { + "epoch": 0.26, + "learning_rate": 2.975908363098904e-06, + "loss": 0.6458, + "step": 702 + }, + { + "epoch": 0.26, + "learning_rate": 2.9758714127355586e-06, + "loss": 0.6692, + "step": 703 + }, + { + "epoch": 0.26, + "learning_rate": 2.975834462372213e-06, + "loss": 0.728, + "step": 704 + }, + { + "epoch": 0.26, + "learning_rate": 2.9757975120088683e-06, + "loss": 0.7387, + "step": 705 + }, + { + "epoch": 0.26, + "learning_rate": 2.975760561645523e-06, + "loss": 0.6526, + "step": 706 + }, + { + "epoch": 0.26, + "learning_rate": 2.975723611282178e-06, + "loss": 0.6433, + "step": 707 + }, + { + "epoch": 0.26, + "learning_rate": 2.9756866609188326e-06, + "loss": 0.6073, + "step": 708 + }, + { + "epoch": 0.26, + "learning_rate": 2.9756497105554872e-06, + "loss": 0.6705, + "step": 709 + }, + { + "epoch": 0.26, + "learning_rate": 2.975612760192142e-06, + "loss": 0.7018, + "step": 710 + }, + { + "epoch": 0.26, + "learning_rate": 2.9755758098287965e-06, + "loss": 0.6803, + "step": 711 + }, + { + "epoch": 0.26, + "learning_rate": 2.9755388594654516e-06, + "loss": 0.6282, + "step": 712 + }, + { + "epoch": 0.26, + "learning_rate": 2.975501909102106e-06, + "loss": 0.6934, + "step": 713 + }, + { + "epoch": 0.26, + "learning_rate": 2.9754649587387613e-06, + "loss": 0.6508, + "step": 714 + }, + { + "epoch": 0.26, + "learning_rate": 2.975428008375416e-06, + "loss": 0.615, + "step": 715 + }, + { + "epoch": 0.26, + "learning_rate": 2.9753910580120705e-06, + "loss": 0.6731, + "step": 716 + }, + { + "epoch": 0.26, + "learning_rate": 2.975354107648725e-06, + "loss": 0.6293, + "step": 717 + }, + { + "epoch": 0.27, + "learning_rate": 2.97531715728538e-06, + "loss": 0.6409, + "step": 718 + }, + { + "epoch": 0.27, + "learning_rate": 2.975280206922035e-06, + "loss": 0.6439, + "step": 719 + }, + { + "epoch": 0.27, + "learning_rate": 2.9752432565586895e-06, + "loss": 0.6281, + "step": 720 + }, + { + "epoch": 0.27, + "learning_rate": 2.9752063061953446e-06, + "loss": 0.7183, + "step": 721 + }, + { + "epoch": 0.27, + "learning_rate": 2.975169355831999e-06, + "loss": 0.6626, + "step": 722 + }, + { + "epoch": 0.27, + "learning_rate": 2.975132405468654e-06, + "loss": 0.668, + "step": 723 + }, + { + "epoch": 0.27, + "learning_rate": 2.9750954551053085e-06, + "loss": 0.6491, + "step": 724 + }, + { + "epoch": 0.27, + "learning_rate": 2.9750585047419635e-06, + "loss": 0.6641, + "step": 725 + }, + { + "epoch": 0.27, + "learning_rate": 2.975021554378618e-06, + "loss": 0.6541, + "step": 726 + }, + { + "epoch": 0.27, + "learning_rate": 2.9749846040152732e-06, + "loss": 0.6494, + "step": 727 + }, + { + "epoch": 0.27, + "learning_rate": 2.974947653651928e-06, + "loss": 0.6357, + "step": 728 + }, + { + "epoch": 0.27, + "learning_rate": 2.974910703288582e-06, + "loss": 0.6278, + "step": 729 + }, + { + "epoch": 0.27, + "learning_rate": 2.974873752925237e-06, + "loss": 0.6843, + "step": 730 + }, + { + "epoch": 0.27, + "learning_rate": 2.9748368025618918e-06, + "loss": 0.7004, + "step": 731 + }, + { + "epoch": 0.27, + "learning_rate": 2.974799852198547e-06, + "loss": 0.6334, + "step": 732 + }, + { + "epoch": 0.27, + "learning_rate": 2.9747629018352015e-06, + "loss": 0.6424, + "step": 733 + }, + { + "epoch": 0.27, + "learning_rate": 2.9747259514718565e-06, + "loss": 0.6687, + "step": 734 + }, + { + "epoch": 0.27, + "learning_rate": 2.9746890011085107e-06, + "loss": 0.6398, + "step": 735 + }, + { + "epoch": 0.27, + "learning_rate": 2.974652050745166e-06, + "loss": 0.5992, + "step": 736 + }, + { + "epoch": 0.27, + "learning_rate": 2.9746151003818204e-06, + "loss": 0.6304, + "step": 737 + }, + { + "epoch": 0.27, + "learning_rate": 2.974578150018475e-06, + "loss": 0.6353, + "step": 738 + }, + { + "epoch": 0.27, + "learning_rate": 2.97454119965513e-06, + "loss": 0.6405, + "step": 739 + }, + { + "epoch": 0.27, + "learning_rate": 2.9745042492917848e-06, + "loss": 0.6955, + "step": 740 + }, + { + "epoch": 0.27, + "learning_rate": 2.9744672989284394e-06, + "loss": 0.6324, + "step": 741 + }, + { + "epoch": 0.27, + "learning_rate": 2.974430348565094e-06, + "loss": 0.6904, + "step": 742 + }, + { + "epoch": 0.27, + "learning_rate": 2.974393398201749e-06, + "loss": 0.7168, + "step": 743 + }, + { + "epoch": 0.27, + "learning_rate": 2.9743564478384037e-06, + "loss": 0.684, + "step": 744 + }, + { + "epoch": 0.28, + "learning_rate": 2.974319497475059e-06, + "loss": 0.6511, + "step": 745 + }, + { + "epoch": 0.28, + "learning_rate": 2.9742825471117134e-06, + "loss": 0.6471, + "step": 746 + }, + { + "epoch": 0.28, + "learning_rate": 2.974245596748368e-06, + "loss": 0.6612, + "step": 747 + }, + { + "epoch": 0.28, + "learning_rate": 2.9742086463850227e-06, + "loss": 0.7, + "step": 748 + }, + { + "epoch": 0.28, + "learning_rate": 2.9741716960216773e-06, + "loss": 0.6574, + "step": 749 + }, + { + "epoch": 0.28, + "learning_rate": 2.9741347456583324e-06, + "loss": 0.5981, + "step": 750 + }, + { + "epoch": 0.28, + "learning_rate": 2.974097795294987e-06, + "loss": 0.6574, + "step": 751 + }, + { + "epoch": 0.28, + "learning_rate": 2.974060844931642e-06, + "loss": 0.6452, + "step": 752 + }, + { + "epoch": 0.28, + "learning_rate": 2.9740238945682967e-06, + "loss": 0.6525, + "step": 753 + }, + { + "epoch": 0.28, + "learning_rate": 2.9739869442049514e-06, + "loss": 0.627, + "step": 754 + }, + { + "epoch": 0.28, + "learning_rate": 2.973949993841606e-06, + "loss": 0.6704, + "step": 755 + }, + { + "epoch": 0.28, + "learning_rate": 2.973913043478261e-06, + "loss": 0.6284, + "step": 756 + }, + { + "epoch": 0.28, + "learning_rate": 2.9738760931149157e-06, + "loss": 0.6372, + "step": 757 + }, + { + "epoch": 0.28, + "learning_rate": 2.9738391427515703e-06, + "loss": 0.6238, + "step": 758 + }, + { + "epoch": 0.28, + "learning_rate": 2.9738021923882254e-06, + "loss": 0.6252, + "step": 759 + }, + { + "epoch": 0.28, + "learning_rate": 2.97376524202488e-06, + "loss": 0.6105, + "step": 760 + }, + { + "epoch": 0.28, + "learning_rate": 2.9737282916615347e-06, + "loss": 0.6388, + "step": 761 + }, + { + "epoch": 0.28, + "learning_rate": 2.9736913412981893e-06, + "loss": 0.7042, + "step": 762 + }, + { + "epoch": 0.28, + "learning_rate": 2.9736543909348444e-06, + "loss": 0.6354, + "step": 763 + }, + { + "epoch": 0.28, + "learning_rate": 2.973617440571499e-06, + "loss": 0.6684, + "step": 764 + }, + { + "epoch": 0.28, + "learning_rate": 2.973580490208154e-06, + "loss": 0.6763, + "step": 765 + }, + { + "epoch": 0.28, + "learning_rate": 2.9735435398448087e-06, + "loss": 0.6608, + "step": 766 + }, + { + "epoch": 0.28, + "learning_rate": 2.9735065894814633e-06, + "loss": 0.7511, + "step": 767 + }, + { + "epoch": 0.28, + "learning_rate": 2.973469639118118e-06, + "loss": 0.6842, + "step": 768 + }, + { + "epoch": 0.28, + "learning_rate": 2.9734326887547726e-06, + "loss": 0.6615, + "step": 769 + }, + { + "epoch": 0.28, + "learning_rate": 2.9733957383914277e-06, + "loss": 0.6352, + "step": 770 + }, + { + "epoch": 0.28, + "learning_rate": 2.9733587880280823e-06, + "loss": 0.7054, + "step": 771 + }, + { + "epoch": 0.29, + "learning_rate": 2.9733218376647374e-06, + "loss": 0.6235, + "step": 772 + }, + { + "epoch": 0.29, + "learning_rate": 2.973284887301392e-06, + "loss": 0.6744, + "step": 773 + }, + { + "epoch": 0.29, + "learning_rate": 2.9732479369380466e-06, + "loss": 0.6374, + "step": 774 + }, + { + "epoch": 0.29, + "learning_rate": 2.9732109865747013e-06, + "loss": 0.6336, + "step": 775 + }, + { + "epoch": 0.29, + "learning_rate": 2.9731740362113563e-06, + "loss": 0.6656, + "step": 776 + }, + { + "epoch": 0.29, + "learning_rate": 2.973137085848011e-06, + "loss": 0.6452, + "step": 777 + }, + { + "epoch": 0.29, + "learning_rate": 2.9731001354846656e-06, + "loss": 0.6339, + "step": 778 + }, + { + "epoch": 0.29, + "learning_rate": 2.9730631851213207e-06, + "loss": 0.6972, + "step": 779 + }, + { + "epoch": 0.29, + "learning_rate": 2.973026234757975e-06, + "loss": 0.6806, + "step": 780 + }, + { + "epoch": 0.29, + "learning_rate": 2.97298928439463e-06, + "loss": 0.698, + "step": 781 + }, + { + "epoch": 0.29, + "learning_rate": 2.9729523340312846e-06, + "loss": 0.7057, + "step": 782 + }, + { + "epoch": 0.29, + "learning_rate": 2.9729153836679396e-06, + "loss": 0.64, + "step": 783 + }, + { + "epoch": 0.29, + "learning_rate": 2.9728784333045943e-06, + "loss": 0.6577, + "step": 784 + }, + { + "epoch": 0.29, + "learning_rate": 2.9728414829412493e-06, + "loss": 0.6332, + "step": 785 + }, + { + "epoch": 0.29, + "learning_rate": 2.9728045325779035e-06, + "loss": 0.6547, + "step": 786 + }, + { + "epoch": 0.29, + "learning_rate": 2.972767582214558e-06, + "loss": 0.653, + "step": 787 + }, + { + "epoch": 0.29, + "learning_rate": 2.9727306318512132e-06, + "loss": 0.6668, + "step": 788 + }, + { + "epoch": 0.29, + "learning_rate": 2.972693681487868e-06, + "loss": 0.675, + "step": 789 + }, + { + "epoch": 0.29, + "learning_rate": 2.972656731124523e-06, + "loss": 0.6449, + "step": 790 + }, + { + "epoch": 0.29, + "learning_rate": 2.9726197807611776e-06, + "loss": 0.6937, + "step": 791 + }, + { + "epoch": 0.29, + "learning_rate": 2.9725828303978322e-06, + "loss": 0.6427, + "step": 792 + }, + { + "epoch": 0.29, + "learning_rate": 2.972545880034487e-06, + "loss": 0.6721, + "step": 793 + }, + { + "epoch": 0.29, + "learning_rate": 2.972508929671142e-06, + "loss": 0.6623, + "step": 794 + }, + { + "epoch": 0.29, + "learning_rate": 2.9724719793077965e-06, + "loss": 0.6714, + "step": 795 + }, + { + "epoch": 0.29, + "learning_rate": 2.9724350289444516e-06, + "loss": 0.6751, + "step": 796 + }, + { + "epoch": 0.29, + "learning_rate": 2.9723980785811062e-06, + "loss": 0.6465, + "step": 797 + }, + { + "epoch": 0.29, + "learning_rate": 2.972361128217761e-06, + "loss": 0.6907, + "step": 798 + }, + { + "epoch": 0.3, + "learning_rate": 2.9723241778544155e-06, + "loss": 0.625, + "step": 799 + }, + { + "epoch": 0.3, + "learning_rate": 2.97228722749107e-06, + "loss": 0.6588, + "step": 800 + }, + { + "epoch": 0.3, + "learning_rate": 2.972250277127725e-06, + "loss": 0.7141, + "step": 801 + }, + { + "epoch": 0.3, + "learning_rate": 2.97221332676438e-06, + "loss": 0.677, + "step": 802 + }, + { + "epoch": 0.3, + "learning_rate": 2.972176376401035e-06, + "loss": 0.7012, + "step": 803 + }, + { + "epoch": 0.3, + "learning_rate": 2.9721394260376895e-06, + "loss": 0.6795, + "step": 804 + }, + { + "epoch": 0.3, + "learning_rate": 2.972102475674344e-06, + "loss": 0.715, + "step": 805 + }, + { + "epoch": 0.3, + "learning_rate": 2.972065525310999e-06, + "loss": 0.738, + "step": 806 + }, + { + "epoch": 0.3, + "learning_rate": 2.9720285749476535e-06, + "loss": 0.6424, + "step": 807 + }, + { + "epoch": 0.3, + "learning_rate": 2.9719916245843085e-06, + "loss": 0.6247, + "step": 808 + }, + { + "epoch": 0.3, + "learning_rate": 2.971954674220963e-06, + "loss": 0.6059, + "step": 809 + }, + { + "epoch": 0.3, + "learning_rate": 2.971917723857618e-06, + "loss": 0.6799, + "step": 810 + }, + { + "epoch": 0.3, + "learning_rate": 2.971880773494273e-06, + "loss": 0.6384, + "step": 811 + }, + { + "epoch": 0.3, + "learning_rate": 2.9718438231309275e-06, + "loss": 0.676, + "step": 812 + }, + { + "epoch": 0.3, + "learning_rate": 2.971806872767582e-06, + "loss": 0.6987, + "step": 813 + }, + { + "epoch": 0.3, + "learning_rate": 2.971769922404237e-06, + "loss": 0.6824, + "step": 814 + }, + { + "epoch": 0.3, + "learning_rate": 2.971732972040892e-06, + "loss": 0.6828, + "step": 815 + }, + { + "epoch": 0.3, + "learning_rate": 2.971696021677547e-06, + "loss": 0.6374, + "step": 816 + }, + { + "epoch": 0.3, + "learning_rate": 2.9716590713142015e-06, + "loss": 0.6808, + "step": 817 + }, + { + "epoch": 0.3, + "learning_rate": 2.971622120950856e-06, + "loss": 0.6614, + "step": 818 + }, + { + "epoch": 0.3, + "learning_rate": 2.9715851705875108e-06, + "loss": 0.6946, + "step": 819 + }, + { + "epoch": 0.3, + "learning_rate": 2.9715482202241654e-06, + "loss": 0.6524, + "step": 820 + }, + { + "epoch": 0.3, + "learning_rate": 2.9715112698608205e-06, + "loss": 0.677, + "step": 821 + }, + { + "epoch": 0.3, + "learning_rate": 2.971474319497475e-06, + "loss": 0.6313, + "step": 822 + }, + { + "epoch": 0.3, + "learning_rate": 2.97143736913413e-06, + "loss": 0.649, + "step": 823 + }, + { + "epoch": 0.3, + "learning_rate": 2.971400418770785e-06, + "loss": 0.6568, + "step": 824 + }, + { + "epoch": 0.3, + "learning_rate": 2.9713634684074395e-06, + "loss": 0.6563, + "step": 825 + }, + { + "epoch": 0.31, + "learning_rate": 2.971326518044094e-06, + "loss": 0.7086, + "step": 826 + }, + { + "epoch": 0.31, + "learning_rate": 2.9712895676807487e-06, + "loss": 0.6816, + "step": 827 + }, + { + "epoch": 0.31, + "learning_rate": 2.9712526173174038e-06, + "loss": 0.6603, + "step": 828 + }, + { + "epoch": 0.31, + "learning_rate": 2.9712156669540584e-06, + "loss": 0.6734, + "step": 829 + }, + { + "epoch": 0.31, + "learning_rate": 2.9711787165907135e-06, + "loss": 0.6441, + "step": 830 + }, + { + "epoch": 0.31, + "learning_rate": 2.9711417662273677e-06, + "loss": 0.6831, + "step": 831 + }, + { + "epoch": 0.31, + "learning_rate": 2.9711048158640228e-06, + "loss": 0.6876, + "step": 832 + }, + { + "epoch": 0.31, + "learning_rate": 2.9710678655006774e-06, + "loss": 0.6716, + "step": 833 + }, + { + "epoch": 0.31, + "learning_rate": 2.9710309151373324e-06, + "loss": 0.6876, + "step": 834 + }, + { + "epoch": 0.31, + "learning_rate": 2.970993964773987e-06, + "loss": 0.628, + "step": 835 + }, + { + "epoch": 0.31, + "learning_rate": 2.970957014410642e-06, + "loss": 0.6975, + "step": 836 + }, + { + "epoch": 0.31, + "learning_rate": 2.9709200640472964e-06, + "loss": 0.7118, + "step": 837 + }, + { + "epoch": 0.31, + "learning_rate": 2.970883113683951e-06, + "loss": 0.6632, + "step": 838 + }, + { + "epoch": 0.31, + "learning_rate": 2.970846163320606e-06, + "loss": 0.645, + "step": 839 + }, + { + "epoch": 0.31, + "learning_rate": 2.9708092129572607e-06, + "loss": 0.6315, + "step": 840 + }, + { + "epoch": 0.31, + "learning_rate": 2.9707722625939158e-06, + "loss": 0.6561, + "step": 841 + }, + { + "epoch": 0.31, + "learning_rate": 2.9707353122305704e-06, + "loss": 0.6341, + "step": 842 + }, + { + "epoch": 0.31, + "learning_rate": 2.970698361867225e-06, + "loss": 0.6458, + "step": 843 + }, + { + "epoch": 0.31, + "learning_rate": 2.9706614115038797e-06, + "loss": 0.6593, + "step": 844 + }, + { + "epoch": 0.31, + "learning_rate": 2.9706244611405347e-06, + "loss": 0.7175, + "step": 845 + }, + { + "epoch": 0.31, + "learning_rate": 2.9705875107771894e-06, + "loss": 0.6818, + "step": 846 + }, + { + "epoch": 0.31, + "learning_rate": 2.970550560413844e-06, + "loss": 0.6544, + "step": 847 + }, + { + "epoch": 0.31, + "learning_rate": 2.970513610050499e-06, + "loss": 0.6377, + "step": 848 + }, + { + "epoch": 0.31, + "learning_rate": 2.9704766596871537e-06, + "loss": 0.6779, + "step": 849 + }, + { + "epoch": 0.31, + "learning_rate": 2.9704397093238083e-06, + "loss": 0.6823, + "step": 850 + }, + { + "epoch": 0.31, + "learning_rate": 2.970402758960463e-06, + "loss": 0.6964, + "step": 851 + }, + { + "epoch": 0.31, + "learning_rate": 2.970365808597118e-06, + "loss": 0.7071, + "step": 852 + }, + { + "epoch": 0.31, + "learning_rate": 2.9703288582337727e-06, + "loss": 0.6931, + "step": 853 + }, + { + "epoch": 0.32, + "learning_rate": 2.9702919078704277e-06, + "loss": 0.6457, + "step": 854 + }, + { + "epoch": 0.32, + "learning_rate": 2.9702549575070824e-06, + "loss": 0.69, + "step": 855 + }, + { + "epoch": 0.32, + "learning_rate": 2.970218007143737e-06, + "loss": 0.6778, + "step": 856 + }, + { + "epoch": 0.32, + "learning_rate": 2.9701810567803916e-06, + "loss": 0.6448, + "step": 857 + }, + { + "epoch": 0.32, + "learning_rate": 2.9701441064170463e-06, + "loss": 0.6703, + "step": 858 + }, + { + "epoch": 0.32, + "learning_rate": 2.9701071560537013e-06, + "loss": 0.6364, + "step": 859 + }, + { + "epoch": 0.32, + "learning_rate": 2.970070205690356e-06, + "loss": 0.6435, + "step": 860 + }, + { + "epoch": 0.32, + "learning_rate": 2.970033255327011e-06, + "loss": 0.6702, + "step": 861 + }, + { + "epoch": 0.32, + "learning_rate": 2.9699963049636657e-06, + "loss": 0.6199, + "step": 862 + }, + { + "epoch": 0.32, + "learning_rate": 2.9699593546003203e-06, + "loss": 0.6605, + "step": 863 + }, + { + "epoch": 0.32, + "learning_rate": 2.969922404236975e-06, + "loss": 0.6446, + "step": 864 + }, + { + "epoch": 0.32, + "learning_rate": 2.96988545387363e-06, + "loss": 0.6825, + "step": 865 + }, + { + "epoch": 0.32, + "learning_rate": 2.9698485035102846e-06, + "loss": 0.6461, + "step": 866 + }, + { + "epoch": 0.32, + "learning_rate": 2.9698115531469393e-06, + "loss": 0.6417, + "step": 867 + }, + { + "epoch": 0.32, + "learning_rate": 2.9697746027835943e-06, + "loss": 0.6394, + "step": 868 + }, + { + "epoch": 0.32, + "learning_rate": 2.969737652420249e-06, + "loss": 0.6727, + "step": 869 + }, + { + "epoch": 0.32, + "learning_rate": 2.9697007020569036e-06, + "loss": 0.6476, + "step": 870 + }, + { + "epoch": 0.32, + "learning_rate": 2.9696637516935582e-06, + "loss": 0.635, + "step": 871 + }, + { + "epoch": 0.32, + "learning_rate": 2.9696268013302133e-06, + "loss": 0.6364, + "step": 872 + }, + { + "epoch": 0.32, + "learning_rate": 2.969589850966868e-06, + "loss": 0.6432, + "step": 873 + }, + { + "epoch": 0.32, + "learning_rate": 2.969552900603523e-06, + "loss": 0.6541, + "step": 874 + }, + { + "epoch": 0.32, + "learning_rate": 2.969515950240177e-06, + "loss": 0.6759, + "step": 875 + }, + { + "epoch": 0.32, + "learning_rate": 2.9694789998768323e-06, + "loss": 0.6237, + "step": 876 + }, + { + "epoch": 0.32, + "learning_rate": 2.969442049513487e-06, + "loss": 0.5793, + "step": 877 + }, + { + "epoch": 0.32, + "learning_rate": 2.9694050991501415e-06, + "loss": 0.6719, + "step": 878 + }, + { + "epoch": 0.32, + "learning_rate": 2.9693681487867966e-06, + "loss": 0.6791, + "step": 879 + }, + { + "epoch": 0.32, + "learning_rate": 2.9693311984234512e-06, + "loss": 0.6415, + "step": 880 + }, + { + "epoch": 0.33, + "learning_rate": 2.969294248060106e-06, + "loss": 0.6746, + "step": 881 + }, + { + "epoch": 0.33, + "learning_rate": 2.9692572976967605e-06, + "loss": 0.6265, + "step": 882 + }, + { + "epoch": 0.33, + "learning_rate": 2.9692203473334156e-06, + "loss": 0.6742, + "step": 883 + }, + { + "epoch": 0.33, + "learning_rate": 2.96918339697007e-06, + "loss": 0.6701, + "step": 884 + }, + { + "epoch": 0.33, + "learning_rate": 2.9691464466067253e-06, + "loss": 0.6704, + "step": 885 + }, + { + "epoch": 0.33, + "learning_rate": 2.96910949624338e-06, + "loss": 0.6329, + "step": 886 + }, + { + "epoch": 0.33, + "learning_rate": 2.9690725458800345e-06, + "loss": 0.6785, + "step": 887 + }, + { + "epoch": 0.33, + "learning_rate": 2.969035595516689e-06, + "loss": 0.6193, + "step": 888 + }, + { + "epoch": 0.33, + "learning_rate": 2.968998645153344e-06, + "loss": 0.6934, + "step": 889 + }, + { + "epoch": 0.33, + "learning_rate": 2.968961694789999e-06, + "loss": 0.5881, + "step": 890 + }, + { + "epoch": 0.33, + "learning_rate": 2.9689247444266535e-06, + "loss": 0.6278, + "step": 891 + }, + { + "epoch": 0.33, + "learning_rate": 2.9688877940633086e-06, + "loss": 0.651, + "step": 892 + }, + { + "epoch": 0.33, + "learning_rate": 2.968850843699963e-06, + "loss": 0.6601, + "step": 893 + }, + { + "epoch": 0.33, + "learning_rate": 2.968813893336618e-06, + "loss": 0.6767, + "step": 894 + }, + { + "epoch": 0.33, + "learning_rate": 2.9687769429732725e-06, + "loss": 0.6726, + "step": 895 + }, + { + "epoch": 0.33, + "learning_rate": 2.9687399926099275e-06, + "loss": 0.6401, + "step": 896 + }, + { + "epoch": 0.33, + "learning_rate": 2.968703042246582e-06, + "loss": 0.6995, + "step": 897 + }, + { + "epoch": 0.33, + "learning_rate": 2.968666091883237e-06, + "loss": 0.6294, + "step": 898 + }, + { + "epoch": 0.33, + "learning_rate": 2.968629141519892e-06, + "loss": 0.6244, + "step": 899 + }, + { + "epoch": 0.33, + "learning_rate": 2.9685921911565465e-06, + "loss": 0.6496, + "step": 900 + }, + { + "epoch": 0.33, + "learning_rate": 2.968555240793201e-06, + "loss": 0.6983, + "step": 901 + }, + { + "epoch": 0.33, + "learning_rate": 2.9685182904298558e-06, + "loss": 0.5993, + "step": 902 + }, + { + "epoch": 0.33, + "learning_rate": 2.968481340066511e-06, + "loss": 0.6152, + "step": 903 + }, + { + "epoch": 0.33, + "learning_rate": 2.9684443897031655e-06, + "loss": 0.7089, + "step": 904 + }, + { + "epoch": 0.33, + "learning_rate": 2.9684074393398205e-06, + "loss": 0.6435, + "step": 905 + }, + { + "epoch": 0.33, + "learning_rate": 2.968370488976475e-06, + "loss": 0.6498, + "step": 906 + }, + { + "epoch": 0.33, + "learning_rate": 2.96833353861313e-06, + "loss": 0.6319, + "step": 907 + }, + { + "epoch": 0.34, + "learning_rate": 2.9682965882497844e-06, + "loss": 0.6948, + "step": 908 + }, + { + "epoch": 0.34, + "learning_rate": 2.968259637886439e-06, + "loss": 0.7039, + "step": 909 + }, + { + "epoch": 0.34, + "learning_rate": 2.968222687523094e-06, + "loss": 0.6313, + "step": 910 + }, + { + "epoch": 0.34, + "learning_rate": 2.9681857371597488e-06, + "loss": 0.736, + "step": 911 + }, + { + "epoch": 0.34, + "learning_rate": 2.968148786796404e-06, + "loss": 0.6966, + "step": 912 + }, + { + "epoch": 0.34, + "learning_rate": 2.9681118364330585e-06, + "loss": 0.6927, + "step": 913 + }, + { + "epoch": 0.34, + "learning_rate": 2.968074886069713e-06, + "loss": 0.7414, + "step": 914 + }, + { + "epoch": 0.34, + "learning_rate": 2.9680379357063677e-06, + "loss": 0.6303, + "step": 915 + }, + { + "epoch": 0.34, + "learning_rate": 2.968000985343023e-06, + "loss": 0.6944, + "step": 916 + }, + { + "epoch": 0.34, + "learning_rate": 2.9679640349796774e-06, + "loss": 0.6819, + "step": 917 + }, + { + "epoch": 0.34, + "learning_rate": 2.967927084616332e-06, + "loss": 0.6179, + "step": 918 + }, + { + "epoch": 0.34, + "learning_rate": 2.967890134252987e-06, + "loss": 0.648, + "step": 919 + }, + { + "epoch": 0.34, + "learning_rate": 2.9678531838896413e-06, + "loss": 0.6307, + "step": 920 + }, + { + "epoch": 0.34, + "learning_rate": 2.9678162335262964e-06, + "loss": 0.6309, + "step": 921 + }, + { + "epoch": 0.34, + "learning_rate": 2.967779283162951e-06, + "loss": 0.6183, + "step": 922 + }, + { + "epoch": 0.34, + "learning_rate": 2.967742332799606e-06, + "loss": 0.6658, + "step": 923 + }, + { + "epoch": 0.34, + "learning_rate": 2.9677053824362607e-06, + "loss": 0.6096, + "step": 924 + }, + { + "epoch": 0.34, + "learning_rate": 2.967668432072916e-06, + "loss": 0.6689, + "step": 925 + }, + { + "epoch": 0.34, + "learning_rate": 2.96763148170957e-06, + "loss": 0.6166, + "step": 926 + }, + { + "epoch": 0.34, + "learning_rate": 2.9675945313462246e-06, + "loss": 0.6466, + "step": 927 + }, + { + "epoch": 0.34, + "learning_rate": 2.9675575809828797e-06, + "loss": 0.6689, + "step": 928 + }, + { + "epoch": 0.34, + "learning_rate": 2.9675206306195343e-06, + "loss": 0.6399, + "step": 929 + }, + { + "epoch": 0.34, + "learning_rate": 2.9674836802561894e-06, + "loss": 0.6886, + "step": 930 + }, + { + "epoch": 0.34, + "learning_rate": 2.967446729892844e-06, + "loss": 0.6984, + "step": 931 + }, + { + "epoch": 0.34, + "learning_rate": 2.9674097795294987e-06, + "loss": 0.6746, + "step": 932 + }, + { + "epoch": 0.34, + "learning_rate": 2.9673728291661533e-06, + "loss": 0.615, + "step": 933 + }, + { + "epoch": 0.34, + "learning_rate": 2.9673358788028084e-06, + "loss": 0.6375, + "step": 934 + }, + { + "epoch": 0.35, + "learning_rate": 2.967298928439463e-06, + "loss": 0.6632, + "step": 935 + }, + { + "epoch": 0.35, + "learning_rate": 2.967261978076118e-06, + "loss": 0.6339, + "step": 936 + }, + { + "epoch": 0.35, + "learning_rate": 2.9672250277127727e-06, + "loss": 0.649, + "step": 937 + }, + { + "epoch": 0.35, + "learning_rate": 2.9671880773494273e-06, + "loss": 0.6791, + "step": 938 + }, + { + "epoch": 0.35, + "learning_rate": 2.967151126986082e-06, + "loss": 0.6254, + "step": 939 + }, + { + "epoch": 0.35, + "learning_rate": 2.9671141766227366e-06, + "loss": 0.6472, + "step": 940 + }, + { + "epoch": 0.35, + "learning_rate": 2.9670772262593917e-06, + "loss": 0.67, + "step": 941 + }, + { + "epoch": 0.35, + "learning_rate": 2.9670402758960463e-06, + "loss": 0.6413, + "step": 942 + }, + { + "epoch": 0.35, + "learning_rate": 2.9670033255327014e-06, + "loss": 0.6606, + "step": 943 + }, + { + "epoch": 0.35, + "learning_rate": 2.966966375169356e-06, + "loss": 0.687, + "step": 944 + }, + { + "epoch": 0.35, + "learning_rate": 2.9669294248060106e-06, + "loss": 0.6872, + "step": 945 + }, + { + "epoch": 0.35, + "learning_rate": 2.9668924744426653e-06, + "loss": 0.6415, + "step": 946 + }, + { + "epoch": 0.35, + "learning_rate": 2.96685552407932e-06, + "loss": 0.6661, + "step": 947 + }, + { + "epoch": 0.35, + "learning_rate": 2.966818573715975e-06, + "loss": 0.6453, + "step": 948 + }, + { + "epoch": 0.35, + "learning_rate": 2.9667816233526296e-06, + "loss": 0.68, + "step": 949 + }, + { + "epoch": 0.35, + "learning_rate": 2.9667446729892847e-06, + "loss": 0.6429, + "step": 950 + }, + { + "epoch": 0.35, + "learning_rate": 2.9667077226259393e-06, + "loss": 0.5858, + "step": 951 + }, + { + "epoch": 0.35, + "learning_rate": 2.966670772262594e-06, + "loss": 0.6769, + "step": 952 + }, + { + "epoch": 0.35, + "learning_rate": 2.9666338218992486e-06, + "loss": 0.7275, + "step": 953 + }, + { + "epoch": 0.35, + "learning_rate": 2.9665968715359036e-06, + "loss": 0.6521, + "step": 954 + }, + { + "epoch": 0.35, + "learning_rate": 2.9665599211725583e-06, + "loss": 0.6479, + "step": 955 + }, + { + "epoch": 0.35, + "learning_rate": 2.9665229708092133e-06, + "loss": 0.6989, + "step": 956 + }, + { + "epoch": 0.35, + "learning_rate": 2.966486020445868e-06, + "loss": 0.6878, + "step": 957 + }, + { + "epoch": 0.35, + "learning_rate": 2.9664490700825226e-06, + "loss": 0.6266, + "step": 958 + }, + { + "epoch": 0.35, + "learning_rate": 2.9664121197191773e-06, + "loss": 0.6584, + "step": 959 + }, + { + "epoch": 0.35, + "learning_rate": 2.966375169355832e-06, + "loss": 0.588, + "step": 960 + }, + { + "epoch": 0.35, + "learning_rate": 2.966338218992487e-06, + "loss": 0.6727, + "step": 961 + }, + { + "epoch": 0.36, + "learning_rate": 2.9663012686291416e-06, + "loss": 0.6197, + "step": 962 + }, + { + "epoch": 0.36, + "learning_rate": 2.9662643182657966e-06, + "loss": 0.6846, + "step": 963 + }, + { + "epoch": 0.36, + "learning_rate": 2.9662273679024513e-06, + "loss": 0.7241, + "step": 964 + }, + { + "epoch": 0.36, + "learning_rate": 2.966190417539106e-06, + "loss": 0.6638, + "step": 965 + }, + { + "epoch": 0.36, + "learning_rate": 2.9661534671757606e-06, + "loss": 0.6262, + "step": 966 + }, + { + "epoch": 0.36, + "learning_rate": 2.966116516812415e-06, + "loss": 0.6832, + "step": 967 + }, + { + "epoch": 0.36, + "learning_rate": 2.9660795664490702e-06, + "loss": 0.6735, + "step": 968 + }, + { + "epoch": 0.36, + "learning_rate": 2.966042616085725e-06, + "loss": 0.6072, + "step": 969 + }, + { + "epoch": 0.36, + "learning_rate": 2.96600566572238e-06, + "loss": 0.6308, + "step": 970 + }, + { + "epoch": 0.36, + "learning_rate": 2.965968715359034e-06, + "loss": 0.6274, + "step": 971 + }, + { + "epoch": 0.36, + "learning_rate": 2.9659317649956892e-06, + "loss": 0.6727, + "step": 972 + }, + { + "epoch": 0.36, + "learning_rate": 2.965894814632344e-06, + "loss": 0.6641, + "step": 973 + }, + { + "epoch": 0.36, + "learning_rate": 2.965857864268999e-06, + "loss": 0.6688, + "step": 974 + }, + { + "epoch": 0.36, + "learning_rate": 2.9658209139056536e-06, + "loss": 0.6786, + "step": 975 + }, + { + "epoch": 0.36, + "learning_rate": 2.9657839635423086e-06, + "loss": 0.6684, + "step": 976 + }, + { + "epoch": 0.36, + "learning_rate": 2.965747013178963e-06, + "loss": 0.6407, + "step": 977 + }, + { + "epoch": 0.36, + "learning_rate": 2.9657100628156175e-06, + "loss": 0.7011, + "step": 978 + }, + { + "epoch": 0.36, + "learning_rate": 2.9656731124522725e-06, + "loss": 0.6587, + "step": 979 + }, + { + "epoch": 0.36, + "learning_rate": 2.965636162088927e-06, + "loss": 0.666, + "step": 980 + }, + { + "epoch": 0.36, + "learning_rate": 2.9655992117255822e-06, + "loss": 0.69, + "step": 981 + }, + { + "epoch": 0.36, + "learning_rate": 2.965562261362237e-06, + "loss": 0.6869, + "step": 982 + }, + { + "epoch": 0.36, + "learning_rate": 2.9655253109988915e-06, + "loss": 0.652, + "step": 983 + }, + { + "epoch": 0.36, + "learning_rate": 2.965488360635546e-06, + "loss": 0.6596, + "step": 984 + }, + { + "epoch": 0.36, + "learning_rate": 2.965451410272201e-06, + "loss": 0.6567, + "step": 985 + }, + { + "epoch": 0.36, + "learning_rate": 2.965414459908856e-06, + "loss": 0.6999, + "step": 986 + }, + { + "epoch": 0.36, + "learning_rate": 2.9653775095455105e-06, + "loss": 0.6795, + "step": 987 + }, + { + "epoch": 0.36, + "learning_rate": 2.9653405591821655e-06, + "loss": 0.6778, + "step": 988 + }, + { + "epoch": 0.37, + "learning_rate": 2.96530360881882e-06, + "loss": 0.6801, + "step": 989 + }, + { + "epoch": 0.37, + "learning_rate": 2.965266658455475e-06, + "loss": 0.637, + "step": 990 + }, + { + "epoch": 0.37, + "learning_rate": 2.9652297080921294e-06, + "loss": 0.6387, + "step": 991 + }, + { + "epoch": 0.37, + "learning_rate": 2.9651927577287845e-06, + "loss": 0.6927, + "step": 992 + }, + { + "epoch": 0.37, + "learning_rate": 2.965155807365439e-06, + "loss": 0.6695, + "step": 993 + }, + { + "epoch": 0.37, + "learning_rate": 2.965118857002094e-06, + "loss": 0.6455, + "step": 994 + }, + { + "epoch": 0.37, + "learning_rate": 2.965081906638749e-06, + "loss": 0.6422, + "step": 995 + }, + { + "epoch": 0.37, + "learning_rate": 2.9650449562754035e-06, + "loss": 0.6228, + "step": 996 + }, + { + "epoch": 0.37, + "learning_rate": 2.965008005912058e-06, + "loss": 0.6611, + "step": 997 + }, + { + "epoch": 0.37, + "learning_rate": 2.9649710555487127e-06, + "loss": 0.6732, + "step": 998 + }, + { + "epoch": 0.37, + "learning_rate": 2.964934105185368e-06, + "loss": 0.6767, + "step": 999 + }, + { + "epoch": 0.37, + "learning_rate": 2.9648971548220224e-06, + "loss": 0.6794, + "step": 1000 + }, + { + "epoch": 0.37, + "learning_rate": 2.9648602044586775e-06, + "loss": 0.6821, + "step": 1001 + }, + { + "epoch": 0.37, + "learning_rate": 2.964823254095332e-06, + "loss": 0.6745, + "step": 1002 + }, + { + "epoch": 0.37, + "learning_rate": 2.9647863037319868e-06, + "loss": 0.6915, + "step": 1003 + }, + { + "epoch": 0.37, + "learning_rate": 2.9647493533686414e-06, + "loss": 0.665, + "step": 1004 + }, + { + "epoch": 0.37, + "learning_rate": 2.9647124030052965e-06, + "loss": 0.6675, + "step": 1005 + }, + { + "epoch": 0.37, + "learning_rate": 2.964675452641951e-06, + "loss": 0.6826, + "step": 1006 + }, + { + "epoch": 0.37, + "learning_rate": 2.9646385022786057e-06, + "loss": 0.6648, + "step": 1007 + }, + { + "epoch": 0.37, + "learning_rate": 2.9646015519152608e-06, + "loss": 0.6591, + "step": 1008 + }, + { + "epoch": 0.37, + "learning_rate": 2.9645646015519154e-06, + "loss": 0.7042, + "step": 1009 + }, + { + "epoch": 0.37, + "learning_rate": 2.96452765118857e-06, + "loss": 0.6365, + "step": 1010 + }, + { + "epoch": 0.37, + "learning_rate": 2.9644907008252247e-06, + "loss": 0.646, + "step": 1011 + }, + { + "epoch": 0.37, + "learning_rate": 2.9644537504618798e-06, + "loss": 0.6425, + "step": 1012 + }, + { + "epoch": 0.37, + "learning_rate": 2.9644168000985344e-06, + "loss": 0.6316, + "step": 1013 + }, + { + "epoch": 0.37, + "learning_rate": 2.9643798497351895e-06, + "loss": 0.6415, + "step": 1014 + }, + { + "epoch": 0.37, + "learning_rate": 2.964342899371844e-06, + "loss": 0.657, + "step": 1015 + }, + { + "epoch": 0.38, + "learning_rate": 2.9643059490084983e-06, + "loss": 0.6393, + "step": 1016 + }, + { + "epoch": 0.38, + "learning_rate": 2.9642689986451534e-06, + "loss": 0.6534, + "step": 1017 + }, + { + "epoch": 0.38, + "learning_rate": 2.964232048281808e-06, + "loss": 0.6472, + "step": 1018 + }, + { + "epoch": 0.38, + "learning_rate": 2.964195097918463e-06, + "loss": 0.6129, + "step": 1019 + }, + { + "epoch": 0.38, + "learning_rate": 2.9641581475551177e-06, + "loss": 0.6439, + "step": 1020 + }, + { + "epoch": 0.38, + "learning_rate": 2.9641211971917728e-06, + "loss": 0.6851, + "step": 1021 + }, + { + "epoch": 0.38, + "learning_rate": 2.964084246828427e-06, + "loss": 0.7021, + "step": 1022 + }, + { + "epoch": 0.38, + "learning_rate": 2.964047296465082e-06, + "loss": 0.6818, + "step": 1023 + }, + { + "epoch": 0.38, + "learning_rate": 2.9640103461017367e-06, + "loss": 0.6219, + "step": 1024 + }, + { + "epoch": 0.38, + "learning_rate": 2.9639733957383917e-06, + "loss": 0.6674, + "step": 1025 + }, + { + "epoch": 0.38, + "learning_rate": 2.9639364453750464e-06, + "loss": 0.642, + "step": 1026 + }, + { + "epoch": 0.38, + "learning_rate": 2.963899495011701e-06, + "loss": 0.655, + "step": 1027 + }, + { + "epoch": 0.38, + "learning_rate": 2.9638625446483556e-06, + "loss": 0.708, + "step": 1028 + }, + { + "epoch": 0.38, + "learning_rate": 2.9638255942850103e-06, + "loss": 0.6963, + "step": 1029 + }, + { + "epoch": 0.38, + "learning_rate": 2.9637886439216653e-06, + "loss": 0.7225, + "step": 1030 + }, + { + "epoch": 0.38, + "learning_rate": 2.96375169355832e-06, + "loss": 0.5605, + "step": 1031 + }, + { + "epoch": 0.38, + "learning_rate": 2.963714743194975e-06, + "loss": 0.6973, + "step": 1032 + }, + { + "epoch": 0.38, + "learning_rate": 2.9636777928316297e-06, + "loss": 0.6684, + "step": 1033 + }, + { + "epoch": 0.38, + "learning_rate": 2.9636408424682843e-06, + "loss": 0.6026, + "step": 1034 + }, + { + "epoch": 0.38, + "learning_rate": 2.963603892104939e-06, + "loss": 0.6735, + "step": 1035 + }, + { + "epoch": 0.38, + "learning_rate": 2.9635669417415936e-06, + "loss": 0.6648, + "step": 1036 + }, + { + "epoch": 0.38, + "learning_rate": 2.9635299913782486e-06, + "loss": 0.6507, + "step": 1037 + }, + { + "epoch": 0.38, + "learning_rate": 2.9634930410149033e-06, + "loss": 0.6712, + "step": 1038 + }, + { + "epoch": 0.38, + "learning_rate": 2.9634560906515583e-06, + "loss": 0.6828, + "step": 1039 + }, + { + "epoch": 0.38, + "learning_rate": 2.963419140288213e-06, + "loss": 0.6771, + "step": 1040 + }, + { + "epoch": 0.38, + "learning_rate": 2.9633821899248676e-06, + "loss": 0.6667, + "step": 1041 + }, + { + "epoch": 0.38, + "learning_rate": 2.9633452395615222e-06, + "loss": 0.616, + "step": 1042 + }, + { + "epoch": 0.39, + "learning_rate": 2.9633082891981773e-06, + "loss": 0.6114, + "step": 1043 + }, + { + "epoch": 0.39, + "learning_rate": 2.963271338834832e-06, + "loss": 0.6779, + "step": 1044 + }, + { + "epoch": 0.39, + "learning_rate": 2.963234388471487e-06, + "loss": 0.6741, + "step": 1045 + }, + { + "epoch": 0.39, + "learning_rate": 2.9631974381081416e-06, + "loss": 0.6418, + "step": 1046 + }, + { + "epoch": 0.39, + "learning_rate": 2.9631604877447963e-06, + "loss": 0.6456, + "step": 1047 + }, + { + "epoch": 0.39, + "learning_rate": 2.963123537381451e-06, + "loss": 0.666, + "step": 1048 + }, + { + "epoch": 0.39, + "learning_rate": 2.9630865870181055e-06, + "loss": 0.6551, + "step": 1049 + }, + { + "epoch": 0.39, + "learning_rate": 2.9630496366547606e-06, + "loss": 0.6378, + "step": 1050 + }, + { + "epoch": 0.39, + "learning_rate": 2.9630126862914152e-06, + "loss": 0.5971, + "step": 1051 + }, + { + "epoch": 0.39, + "learning_rate": 2.9629757359280703e-06, + "loss": 0.6492, + "step": 1052 + }, + { + "epoch": 0.39, + "learning_rate": 2.962938785564725e-06, + "loss": 0.7074, + "step": 1053 + }, + { + "epoch": 0.39, + "learning_rate": 2.9629018352013796e-06, + "loss": 0.6124, + "step": 1054 + }, + { + "epoch": 0.39, + "learning_rate": 2.962864884838034e-06, + "loss": 0.7169, + "step": 1055 + }, + { + "epoch": 0.39, + "learning_rate": 2.962827934474689e-06, + "loss": 0.6517, + "step": 1056 + }, + { + "epoch": 0.39, + "learning_rate": 2.962790984111344e-06, + "loss": 0.6478, + "step": 1057 + }, + { + "epoch": 0.39, + "learning_rate": 2.9627540337479985e-06, + "loss": 0.6514, + "step": 1058 + }, + { + "epoch": 0.39, + "learning_rate": 2.9627170833846536e-06, + "loss": 0.6447, + "step": 1059 + }, + { + "epoch": 0.39, + "learning_rate": 2.9626801330213082e-06, + "loss": 0.6618, + "step": 1060 + }, + { + "epoch": 0.39, + "learning_rate": 2.962643182657963e-06, + "loss": 0.6152, + "step": 1061 + }, + { + "epoch": 0.39, + "learning_rate": 2.9626062322946175e-06, + "loss": 0.705, + "step": 1062 + }, + { + "epoch": 0.39, + "learning_rate": 2.9625692819312726e-06, + "loss": 0.667, + "step": 1063 + }, + { + "epoch": 0.39, + "learning_rate": 2.962532331567927e-06, + "loss": 0.6638, + "step": 1064 + }, + { + "epoch": 0.39, + "learning_rate": 2.9624953812045823e-06, + "loss": 0.6188, + "step": 1065 + }, + { + "epoch": 0.39, + "learning_rate": 2.9624584308412365e-06, + "loss": 0.6924, + "step": 1066 + }, + { + "epoch": 0.39, + "learning_rate": 2.962421480477891e-06, + "loss": 0.7033, + "step": 1067 + }, + { + "epoch": 0.39, + "learning_rate": 2.962384530114546e-06, + "loss": 0.6179, + "step": 1068 + }, + { + "epoch": 0.39, + "learning_rate": 2.962347579751201e-06, + "loss": 0.6106, + "step": 1069 + }, + { + "epoch": 0.4, + "learning_rate": 2.962310629387856e-06, + "loss": 0.6831, + "step": 1070 + }, + { + "epoch": 0.4, + "learning_rate": 2.9622736790245105e-06, + "loss": 0.6316, + "step": 1071 + }, + { + "epoch": 0.4, + "learning_rate": 2.962236728661165e-06, + "loss": 0.6422, + "step": 1072 + }, + { + "epoch": 0.4, + "learning_rate": 2.9621997782978198e-06, + "loss": 0.6406, + "step": 1073 + }, + { + "epoch": 0.4, + "learning_rate": 2.962162827934475e-06, + "loss": 0.6509, + "step": 1074 + }, + { + "epoch": 0.4, + "learning_rate": 2.9621258775711295e-06, + "loss": 0.6208, + "step": 1075 + }, + { + "epoch": 0.4, + "learning_rate": 2.962088927207784e-06, + "loss": 0.6783, + "step": 1076 + }, + { + "epoch": 0.4, + "learning_rate": 2.962051976844439e-06, + "loss": 0.7157, + "step": 1077 + }, + { + "epoch": 0.4, + "learning_rate": 2.962015026481094e-06, + "loss": 0.6979, + "step": 1078 + }, + { + "epoch": 0.4, + "learning_rate": 2.9619780761177484e-06, + "loss": 0.7237, + "step": 1079 + }, + { + "epoch": 0.4, + "learning_rate": 2.961941125754403e-06, + "loss": 0.6571, + "step": 1080 + }, + { + "epoch": 0.4, + "learning_rate": 2.961904175391058e-06, + "loss": 0.6542, + "step": 1081 + }, + { + "epoch": 0.4, + "learning_rate": 2.9618672250277128e-06, + "loss": 0.6785, + "step": 1082 + }, + { + "epoch": 0.4, + "learning_rate": 2.961830274664368e-06, + "loss": 0.7025, + "step": 1083 + }, + { + "epoch": 0.4, + "learning_rate": 2.9617933243010225e-06, + "loss": 0.6804, + "step": 1084 + }, + { + "epoch": 0.4, + "learning_rate": 2.961756373937677e-06, + "loss": 0.5979, + "step": 1085 + }, + { + "epoch": 0.4, + "learning_rate": 2.9617194235743317e-06, + "loss": 0.6479, + "step": 1086 + }, + { + "epoch": 0.4, + "learning_rate": 2.9616824732109864e-06, + "loss": 0.6725, + "step": 1087 + }, + { + "epoch": 0.4, + "learning_rate": 2.9616455228476414e-06, + "loss": 0.6342, + "step": 1088 + }, + { + "epoch": 0.4, + "learning_rate": 2.961608572484296e-06, + "loss": 0.6361, + "step": 1089 + }, + { + "epoch": 0.4, + "learning_rate": 2.961571622120951e-06, + "loss": 0.6215, + "step": 1090 + }, + { + "epoch": 0.4, + "learning_rate": 2.9615346717576058e-06, + "loss": 0.6688, + "step": 1091 + }, + { + "epoch": 0.4, + "learning_rate": 2.9614977213942604e-06, + "loss": 0.6753, + "step": 1092 + }, + { + "epoch": 0.4, + "learning_rate": 2.961460771030915e-06, + "loss": 0.6544, + "step": 1093 + }, + { + "epoch": 0.4, + "learning_rate": 2.96142382066757e-06, + "loss": 0.6935, + "step": 1094 + }, + { + "epoch": 0.4, + "learning_rate": 2.9613868703042247e-06, + "loss": 0.6554, + "step": 1095 + }, + { + "epoch": 0.4, + "learning_rate": 2.9613499199408794e-06, + "loss": 0.6415, + "step": 1096 + }, + { + "epoch": 0.41, + "learning_rate": 2.9613129695775344e-06, + "loss": 0.6317, + "step": 1097 + }, + { + "epoch": 0.41, + "learning_rate": 2.961276019214189e-06, + "loss": 0.6287, + "step": 1098 + }, + { + "epoch": 0.41, + "learning_rate": 2.9612390688508437e-06, + "loss": 0.6682, + "step": 1099 + }, + { + "epoch": 0.41, + "learning_rate": 2.9612021184874984e-06, + "loss": 0.5868, + "step": 1100 + }, + { + "epoch": 0.41, + "learning_rate": 2.9611651681241534e-06, + "loss": 0.6526, + "step": 1101 + }, + { + "epoch": 0.41, + "learning_rate": 2.961128217760808e-06, + "loss": 0.6018, + "step": 1102 + }, + { + "epoch": 0.41, + "learning_rate": 2.961091267397463e-06, + "loss": 0.6816, + "step": 1103 + }, + { + "epoch": 0.41, + "learning_rate": 2.9610543170341177e-06, + "loss": 0.621, + "step": 1104 + }, + { + "epoch": 0.41, + "learning_rate": 2.9610173666707724e-06, + "loss": 0.6073, + "step": 1105 + }, + { + "epoch": 0.41, + "learning_rate": 2.960980416307427e-06, + "loss": 0.6158, + "step": 1106 + }, + { + "epoch": 0.41, + "learning_rate": 2.9609434659440817e-06, + "loss": 0.7177, + "step": 1107 + }, + { + "epoch": 0.41, + "learning_rate": 2.9609065155807367e-06, + "loss": 0.6195, + "step": 1108 + }, + { + "epoch": 0.41, + "learning_rate": 2.9608695652173913e-06, + "loss": 0.6395, + "step": 1109 + }, + { + "epoch": 0.41, + "learning_rate": 2.9608326148540464e-06, + "loss": 0.6749, + "step": 1110 + }, + { + "epoch": 0.41, + "learning_rate": 2.9607956644907006e-06, + "loss": 0.6637, + "step": 1111 + }, + { + "epoch": 0.41, + "learning_rate": 2.9607587141273557e-06, + "loss": 0.6354, + "step": 1112 + }, + { + "epoch": 0.41, + "learning_rate": 2.9607217637640103e-06, + "loss": 0.6253, + "step": 1113 + }, + { + "epoch": 0.41, + "learning_rate": 2.9606848134006654e-06, + "loss": 0.6493, + "step": 1114 + }, + { + "epoch": 0.41, + "learning_rate": 2.96064786303732e-06, + "loss": 0.7009, + "step": 1115 + }, + { + "epoch": 0.41, + "learning_rate": 2.9606109126739747e-06, + "loss": 0.6244, + "step": 1116 + }, + { + "epoch": 0.41, + "learning_rate": 2.9605739623106293e-06, + "loss": 0.6105, + "step": 1117 + }, + { + "epoch": 0.41, + "learning_rate": 2.960537011947284e-06, + "loss": 0.7126, + "step": 1118 + }, + { + "epoch": 0.41, + "learning_rate": 2.960500061583939e-06, + "loss": 0.6809, + "step": 1119 + }, + { + "epoch": 0.41, + "learning_rate": 2.9604631112205936e-06, + "loss": 0.6602, + "step": 1120 + }, + { + "epoch": 0.41, + "learning_rate": 2.9604261608572487e-06, + "loss": 0.6385, + "step": 1121 + }, + { + "epoch": 0.41, + "learning_rate": 2.9603892104939033e-06, + "loss": 0.6866, + "step": 1122 + }, + { + "epoch": 0.41, + "learning_rate": 2.960352260130558e-06, + "loss": 0.6212, + "step": 1123 + }, + { + "epoch": 0.42, + "learning_rate": 2.9603153097672126e-06, + "loss": 0.6991, + "step": 1124 + }, + { + "epoch": 0.42, + "learning_rate": 2.9602783594038677e-06, + "loss": 0.6728, + "step": 1125 + }, + { + "epoch": 0.42, + "learning_rate": 2.9602414090405223e-06, + "loss": 0.6728, + "step": 1126 + }, + { + "epoch": 0.42, + "learning_rate": 2.960204458677177e-06, + "loss": 0.6742, + "step": 1127 + }, + { + "epoch": 0.42, + "learning_rate": 2.960167508313832e-06, + "loss": 0.6161, + "step": 1128 + }, + { + "epoch": 0.42, + "learning_rate": 2.9601305579504866e-06, + "loss": 0.6456, + "step": 1129 + }, + { + "epoch": 0.42, + "learning_rate": 2.9600936075871413e-06, + "loss": 0.5777, + "step": 1130 + }, + { + "epoch": 0.42, + "learning_rate": 2.960056657223796e-06, + "loss": 0.7315, + "step": 1131 + }, + { + "epoch": 0.42, + "learning_rate": 2.960019706860451e-06, + "loss": 0.6636, + "step": 1132 + }, + { + "epoch": 0.42, + "learning_rate": 2.9599827564971056e-06, + "loss": 0.6065, + "step": 1133 + }, + { + "epoch": 0.42, + "learning_rate": 2.9599458061337606e-06, + "loss": 0.6932, + "step": 1134 + }, + { + "epoch": 0.42, + "learning_rate": 2.9599088557704153e-06, + "loss": 0.7147, + "step": 1135 + }, + { + "epoch": 0.42, + "learning_rate": 2.95987190540707e-06, + "loss": 0.7074, + "step": 1136 + }, + { + "epoch": 0.42, + "learning_rate": 2.9598349550437246e-06, + "loss": 0.6307, + "step": 1137 + }, + { + "epoch": 0.42, + "learning_rate": 2.959798004680379e-06, + "loss": 0.6163, + "step": 1138 + }, + { + "epoch": 0.42, + "learning_rate": 2.9597610543170343e-06, + "loss": 0.6262, + "step": 1139 + }, + { + "epoch": 0.42, + "learning_rate": 2.959724103953689e-06, + "loss": 0.686, + "step": 1140 + }, + { + "epoch": 0.42, + "learning_rate": 2.959687153590344e-06, + "loss": 0.6278, + "step": 1141 + }, + { + "epoch": 0.42, + "learning_rate": 2.9596502032269986e-06, + "loss": 0.6562, + "step": 1142 + }, + { + "epoch": 0.42, + "learning_rate": 2.9596132528636532e-06, + "loss": 0.7056, + "step": 1143 + }, + { + "epoch": 0.42, + "learning_rate": 2.959576302500308e-06, + "loss": 0.6471, + "step": 1144 + }, + { + "epoch": 0.42, + "learning_rate": 2.959539352136963e-06, + "loss": 0.6776, + "step": 1145 + }, + { + "epoch": 0.42, + "learning_rate": 2.9595024017736176e-06, + "loss": 0.6397, + "step": 1146 + }, + { + "epoch": 0.42, + "learning_rate": 2.959465451410272e-06, + "loss": 0.6448, + "step": 1147 + }, + { + "epoch": 0.42, + "learning_rate": 2.9594285010469273e-06, + "loss": 0.6622, + "step": 1148 + }, + { + "epoch": 0.42, + "learning_rate": 2.959391550683582e-06, + "loss": 0.6503, + "step": 1149 + }, + { + "epoch": 0.42, + "learning_rate": 2.9593546003202365e-06, + "loss": 0.6486, + "step": 1150 + }, + { + "epoch": 0.43, + "learning_rate": 2.959317649956891e-06, + "loss": 0.6703, + "step": 1151 + }, + { + "epoch": 0.43, + "learning_rate": 2.9592806995935462e-06, + "loss": 0.5918, + "step": 1152 + }, + { + "epoch": 0.43, + "learning_rate": 2.959243749230201e-06, + "loss": 0.6981, + "step": 1153 + }, + { + "epoch": 0.43, + "learning_rate": 2.959206798866856e-06, + "loss": 0.6305, + "step": 1154 + }, + { + "epoch": 0.43, + "learning_rate": 2.9591698485035106e-06, + "loss": 0.6332, + "step": 1155 + }, + { + "epoch": 0.43, + "learning_rate": 2.9591328981401648e-06, + "loss": 0.6673, + "step": 1156 + }, + { + "epoch": 0.43, + "learning_rate": 2.95909594777682e-06, + "loss": 0.6433, + "step": 1157 + }, + { + "epoch": 0.43, + "learning_rate": 2.9590589974134745e-06, + "loss": 0.6371, + "step": 1158 + }, + { + "epoch": 0.43, + "learning_rate": 2.9590220470501295e-06, + "loss": 0.6534, + "step": 1159 + }, + { + "epoch": 0.43, + "learning_rate": 2.958985096686784e-06, + "loss": 0.6305, + "step": 1160 + }, + { + "epoch": 0.43, + "learning_rate": 2.9589481463234392e-06, + "loss": 0.6651, + "step": 1161 + }, + { + "epoch": 0.43, + "learning_rate": 2.9589111959600934e-06, + "loss": 0.6801, + "step": 1162 + }, + { + "epoch": 0.43, + "learning_rate": 2.9588742455967485e-06, + "loss": 0.6351, + "step": 1163 + }, + { + "epoch": 0.43, + "learning_rate": 2.958837295233403e-06, + "loss": 0.5879, + "step": 1164 + }, + { + "epoch": 0.43, + "learning_rate": 2.958800344870058e-06, + "loss": 0.6642, + "step": 1165 + }, + { + "epoch": 0.43, + "learning_rate": 2.958763394506713e-06, + "loss": 0.6514, + "step": 1166 + }, + { + "epoch": 0.43, + "learning_rate": 2.9587264441433675e-06, + "loss": 0.622, + "step": 1167 + }, + { + "epoch": 0.43, + "learning_rate": 2.958689493780022e-06, + "loss": 0.6526, + "step": 1168 + }, + { + "epoch": 0.43, + "learning_rate": 2.9586525434166767e-06, + "loss": 0.6611, + "step": 1169 + }, + { + "epoch": 0.43, + "learning_rate": 2.958615593053332e-06, + "loss": 0.6535, + "step": 1170 + }, + { + "epoch": 0.43, + "learning_rate": 2.9585786426899864e-06, + "loss": 0.6452, + "step": 1171 + }, + { + "epoch": 0.43, + "learning_rate": 2.9585416923266415e-06, + "loss": 0.6728, + "step": 1172 + }, + { + "epoch": 0.43, + "learning_rate": 2.958504741963296e-06, + "loss": 0.6241, + "step": 1173 + }, + { + "epoch": 0.43, + "learning_rate": 2.9584677915999508e-06, + "loss": 0.6814, + "step": 1174 + }, + { + "epoch": 0.43, + "learning_rate": 2.9584308412366054e-06, + "loss": 0.6868, + "step": 1175 + }, + { + "epoch": 0.43, + "learning_rate": 2.95839389087326e-06, + "loss": 0.7039, + "step": 1176 + }, + { + "epoch": 0.43, + "learning_rate": 2.958356940509915e-06, + "loss": 0.619, + "step": 1177 + }, + { + "epoch": 0.44, + "learning_rate": 2.9583199901465697e-06, + "loss": 0.611, + "step": 1178 + }, + { + "epoch": 0.44, + "learning_rate": 2.958283039783225e-06, + "loss": 0.6812, + "step": 1179 + }, + { + "epoch": 0.44, + "learning_rate": 2.9582460894198794e-06, + "loss": 0.6517, + "step": 1180 + }, + { + "epoch": 0.44, + "learning_rate": 2.958209139056534e-06, + "loss": 0.6781, + "step": 1181 + }, + { + "epoch": 0.44, + "learning_rate": 2.9581721886931887e-06, + "loss": 0.6506, + "step": 1182 + }, + { + "epoch": 0.44, + "learning_rate": 2.9581352383298438e-06, + "loss": 0.6273, + "step": 1183 + }, + { + "epoch": 0.44, + "learning_rate": 2.9580982879664984e-06, + "loss": 0.6995, + "step": 1184 + }, + { + "epoch": 0.44, + "learning_rate": 2.9580613376031535e-06, + "loss": 0.6467, + "step": 1185 + }, + { + "epoch": 0.44, + "learning_rate": 2.958024387239808e-06, + "loss": 0.6301, + "step": 1186 + }, + { + "epoch": 0.44, + "learning_rate": 2.9579874368764627e-06, + "loss": 0.5867, + "step": 1187 + }, + { + "epoch": 0.44, + "learning_rate": 2.9579504865131174e-06, + "loss": 0.6936, + "step": 1188 + }, + { + "epoch": 0.44, + "learning_rate": 2.957913536149772e-06, + "loss": 0.6568, + "step": 1189 + }, + { + "epoch": 0.44, + "learning_rate": 2.957876585786427e-06, + "loss": 0.6977, + "step": 1190 + }, + { + "epoch": 0.44, + "learning_rate": 2.9578396354230817e-06, + "loss": 0.6291, + "step": 1191 + }, + { + "epoch": 0.44, + "learning_rate": 2.9578026850597368e-06, + "loss": 0.672, + "step": 1192 + }, + { + "epoch": 0.44, + "learning_rate": 2.9577657346963914e-06, + "loss": 0.6916, + "step": 1193 + }, + { + "epoch": 0.44, + "learning_rate": 2.957728784333046e-06, + "loss": 0.5993, + "step": 1194 + }, + { + "epoch": 0.44, + "learning_rate": 2.9576918339697007e-06, + "loss": 0.6724, + "step": 1195 + }, + { + "epoch": 0.44, + "learning_rate": 2.9576548836063553e-06, + "loss": 0.6542, + "step": 1196 + }, + { + "epoch": 0.44, + "learning_rate": 2.9576179332430104e-06, + "loss": 0.6949, + "step": 1197 + }, + { + "epoch": 0.44, + "learning_rate": 2.957580982879665e-06, + "loss": 0.6844, + "step": 1198 + }, + { + "epoch": 0.44, + "learning_rate": 2.95754403251632e-06, + "loss": 0.6299, + "step": 1199 + }, + { + "epoch": 0.44, + "learning_rate": 2.9575070821529747e-06, + "loss": 0.6115, + "step": 1200 + }, + { + "epoch": 0.44, + "learning_rate": 2.9574701317896293e-06, + "loss": 0.6524, + "step": 1201 + }, + { + "epoch": 0.44, + "learning_rate": 2.957433181426284e-06, + "loss": 0.6637, + "step": 1202 + }, + { + "epoch": 0.44, + "learning_rate": 2.957396231062939e-06, + "loss": 0.6766, + "step": 1203 + }, + { + "epoch": 0.44, + "learning_rate": 2.9573592806995937e-06, + "loss": 0.6711, + "step": 1204 + }, + { + "epoch": 0.44, + "learning_rate": 2.9573223303362487e-06, + "loss": 0.6478, + "step": 1205 + }, + { + "epoch": 0.45, + "learning_rate": 2.9572853799729034e-06, + "loss": 0.6938, + "step": 1206 + }, + { + "epoch": 0.45, + "learning_rate": 2.9572484296095576e-06, + "loss": 0.6427, + "step": 1207 + }, + { + "epoch": 0.45, + "learning_rate": 2.9572114792462126e-06, + "loss": 0.633, + "step": 1208 + }, + { + "epoch": 0.45, + "learning_rate": 2.9571745288828673e-06, + "loss": 0.624, + "step": 1209 + }, + { + "epoch": 0.45, + "learning_rate": 2.9571375785195223e-06, + "loss": 0.6114, + "step": 1210 + }, + { + "epoch": 0.45, + "learning_rate": 2.957100628156177e-06, + "loss": 0.6644, + "step": 1211 + }, + { + "epoch": 0.45, + "learning_rate": 2.957063677792832e-06, + "loss": 0.6476, + "step": 1212 + }, + { + "epoch": 0.45, + "learning_rate": 2.9570267274294862e-06, + "loss": 0.6068, + "step": 1213 + }, + { + "epoch": 0.45, + "learning_rate": 2.9569897770661413e-06, + "loss": 0.6498, + "step": 1214 + }, + { + "epoch": 0.45, + "learning_rate": 2.956952826702796e-06, + "loss": 0.6987, + "step": 1215 + }, + { + "epoch": 0.45, + "learning_rate": 2.9569158763394506e-06, + "loss": 0.6027, + "step": 1216 + }, + { + "epoch": 0.45, + "learning_rate": 2.9568789259761056e-06, + "loss": 0.6213, + "step": 1217 + }, + { + "epoch": 0.45, + "learning_rate": 2.9568419756127603e-06, + "loss": 0.6999, + "step": 1218 + }, + { + "epoch": 0.45, + "learning_rate": 2.956805025249415e-06, + "loss": 0.632, + "step": 1219 + }, + { + "epoch": 0.45, + "learning_rate": 2.9567680748860695e-06, + "loss": 0.6071, + "step": 1220 + }, + { + "epoch": 0.45, + "learning_rate": 2.9567311245227246e-06, + "loss": 0.6352, + "step": 1221 + }, + { + "epoch": 0.45, + "learning_rate": 2.9566941741593792e-06, + "loss": 0.6511, + "step": 1222 + }, + { + "epoch": 0.45, + "learning_rate": 2.9566572237960343e-06, + "loss": 0.6886, + "step": 1223 + }, + { + "epoch": 0.45, + "learning_rate": 2.956620273432689e-06, + "loss": 0.6408, + "step": 1224 + }, + { + "epoch": 0.45, + "learning_rate": 2.9565833230693436e-06, + "loss": 0.6243, + "step": 1225 + }, + { + "epoch": 0.45, + "learning_rate": 2.9565463727059982e-06, + "loss": 0.6032, + "step": 1226 + }, + { + "epoch": 0.45, + "learning_rate": 2.956509422342653e-06, + "loss": 0.7096, + "step": 1227 + }, + { + "epoch": 0.45, + "learning_rate": 2.956472471979308e-06, + "loss": 0.6249, + "step": 1228 + }, + { + "epoch": 0.45, + "learning_rate": 2.9564355216159625e-06, + "loss": 0.685, + "step": 1229 + }, + { + "epoch": 0.45, + "learning_rate": 2.9563985712526176e-06, + "loss": 0.642, + "step": 1230 + }, + { + "epoch": 0.45, + "learning_rate": 2.9563616208892722e-06, + "loss": 0.7263, + "step": 1231 + }, + { + "epoch": 0.45, + "learning_rate": 2.956324670525927e-06, + "loss": 0.6813, + "step": 1232 + }, + { + "epoch": 0.46, + "learning_rate": 2.9562877201625815e-06, + "loss": 0.6553, + "step": 1233 + }, + { + "epoch": 0.46, + "learning_rate": 2.9562507697992366e-06, + "loss": 0.666, + "step": 1234 + }, + { + "epoch": 0.46, + "learning_rate": 2.956213819435891e-06, + "loss": 0.5808, + "step": 1235 + }, + { + "epoch": 0.46, + "learning_rate": 2.956176869072546e-06, + "loss": 0.6568, + "step": 1236 + }, + { + "epoch": 0.46, + "learning_rate": 2.956139918709201e-06, + "loss": 0.6814, + "step": 1237 + }, + { + "epoch": 0.46, + "learning_rate": 2.9561029683458555e-06, + "loss": 0.6145, + "step": 1238 + }, + { + "epoch": 0.46, + "learning_rate": 2.95606601798251e-06, + "loss": 0.6264, + "step": 1239 + }, + { + "epoch": 0.46, + "learning_rate": 2.956029067619165e-06, + "loss": 0.6197, + "step": 1240 + }, + { + "epoch": 0.46, + "learning_rate": 2.95599211725582e-06, + "loss": 0.6099, + "step": 1241 + }, + { + "epoch": 0.46, + "learning_rate": 2.9559551668924745e-06, + "loss": 0.6419, + "step": 1242 + }, + { + "epoch": 0.46, + "learning_rate": 2.9559182165291296e-06, + "loss": 0.638, + "step": 1243 + }, + { + "epoch": 0.46, + "learning_rate": 2.955881266165784e-06, + "loss": 0.6548, + "step": 1244 + }, + { + "epoch": 0.46, + "learning_rate": 2.955844315802439e-06, + "loss": 0.6295, + "step": 1245 + }, + { + "epoch": 0.46, + "learning_rate": 2.9558073654390935e-06, + "loss": 0.5553, + "step": 1246 + }, + { + "epoch": 0.46, + "learning_rate": 2.955770415075748e-06, + "loss": 0.6214, + "step": 1247 + }, + { + "epoch": 0.46, + "learning_rate": 2.955733464712403e-06, + "loss": 0.6912, + "step": 1248 + }, + { + "epoch": 0.46, + "learning_rate": 2.955696514349058e-06, + "loss": 0.7175, + "step": 1249 + }, + { + "epoch": 0.46, + "learning_rate": 2.955659563985713e-06, + "loss": 0.6306, + "step": 1250 + }, + { + "epoch": 0.46, + "learning_rate": 2.955622613622367e-06, + "loss": 0.6717, + "step": 1251 + }, + { + "epoch": 0.46, + "learning_rate": 2.955585663259022e-06, + "loss": 0.6751, + "step": 1252 + }, + { + "epoch": 0.46, + "learning_rate": 2.9555487128956768e-06, + "loss": 0.6752, + "step": 1253 + }, + { + "epoch": 0.46, + "learning_rate": 2.955511762532332e-06, + "loss": 0.6799, + "step": 1254 + }, + { + "epoch": 0.46, + "learning_rate": 2.9554748121689865e-06, + "loss": 0.606, + "step": 1255 + }, + { + "epoch": 0.46, + "learning_rate": 2.955437861805641e-06, + "loss": 0.6097, + "step": 1256 + }, + { + "epoch": 0.46, + "learning_rate": 2.9554009114422958e-06, + "loss": 0.6511, + "step": 1257 + }, + { + "epoch": 0.46, + "learning_rate": 2.9553639610789504e-06, + "loss": 0.6519, + "step": 1258 + }, + { + "epoch": 0.46, + "learning_rate": 2.9553270107156054e-06, + "loss": 0.6493, + "step": 1259 + }, + { + "epoch": 0.47, + "learning_rate": 2.95529006035226e-06, + "loss": 0.6452, + "step": 1260 + }, + { + "epoch": 0.47, + "learning_rate": 2.955253109988915e-06, + "loss": 0.6494, + "step": 1261 + }, + { + "epoch": 0.47, + "learning_rate": 2.9552161596255698e-06, + "loss": 0.6626, + "step": 1262 + }, + { + "epoch": 0.47, + "learning_rate": 2.9551792092622244e-06, + "loss": 0.6167, + "step": 1263 + }, + { + "epoch": 0.47, + "learning_rate": 2.955142258898879e-06, + "loss": 0.6648, + "step": 1264 + }, + { + "epoch": 0.47, + "learning_rate": 2.9551053085355337e-06, + "loss": 0.6384, + "step": 1265 + }, + { + "epoch": 0.47, + "learning_rate": 2.9550683581721888e-06, + "loss": 0.6762, + "step": 1266 + }, + { + "epoch": 0.47, + "learning_rate": 2.9550314078088434e-06, + "loss": 0.6331, + "step": 1267 + }, + { + "epoch": 0.47, + "learning_rate": 2.9549944574454984e-06, + "loss": 0.689, + "step": 1268 + }, + { + "epoch": 0.47, + "learning_rate": 2.954957507082153e-06, + "loss": 0.6796, + "step": 1269 + }, + { + "epoch": 0.47, + "learning_rate": 2.9549205567188077e-06, + "loss": 0.6021, + "step": 1270 + }, + { + "epoch": 0.47, + "learning_rate": 2.9548836063554624e-06, + "loss": 0.642, + "step": 1271 + }, + { + "epoch": 0.47, + "learning_rate": 2.9548466559921174e-06, + "loss": 0.6493, + "step": 1272 + }, + { + "epoch": 0.47, + "learning_rate": 2.954809705628772e-06, + "loss": 0.6742, + "step": 1273 + }, + { + "epoch": 0.47, + "learning_rate": 2.954772755265427e-06, + "loss": 0.6133, + "step": 1274 + }, + { + "epoch": 0.47, + "learning_rate": 2.9547358049020818e-06, + "loss": 0.6542, + "step": 1275 + }, + { + "epoch": 0.47, + "learning_rate": 2.9546988545387364e-06, + "loss": 0.6422, + "step": 1276 + }, + { + "epoch": 0.47, + "learning_rate": 2.954661904175391e-06, + "loss": 0.7358, + "step": 1277 + }, + { + "epoch": 0.47, + "learning_rate": 2.9546249538120457e-06, + "loss": 0.6907, + "step": 1278 + }, + { + "epoch": 0.47, + "learning_rate": 2.9545880034487007e-06, + "loss": 0.7035, + "step": 1279 + }, + { + "epoch": 0.47, + "learning_rate": 2.9545510530853554e-06, + "loss": 0.676, + "step": 1280 + }, + { + "epoch": 0.47, + "learning_rate": 2.9545141027220104e-06, + "loss": 0.7075, + "step": 1281 + }, + { + "epoch": 0.47, + "learning_rate": 2.954477152358665e-06, + "loss": 0.6357, + "step": 1282 + }, + { + "epoch": 0.47, + "learning_rate": 2.9544402019953197e-06, + "loss": 0.6724, + "step": 1283 + }, + { + "epoch": 0.47, + "learning_rate": 2.9544032516319743e-06, + "loss": 0.6644, + "step": 1284 + }, + { + "epoch": 0.47, + "learning_rate": 2.954366301268629e-06, + "loss": 0.7079, + "step": 1285 + }, + { + "epoch": 0.47, + "learning_rate": 2.954329350905284e-06, + "loss": 0.6462, + "step": 1286 + }, + { + "epoch": 0.48, + "learning_rate": 2.9542924005419387e-06, + "loss": 0.6738, + "step": 1287 + }, + { + "epoch": 0.48, + "learning_rate": 2.9542554501785937e-06, + "loss": 0.685, + "step": 1288 + }, + { + "epoch": 0.48, + "learning_rate": 2.9542184998152484e-06, + "loss": 0.691, + "step": 1289 + }, + { + "epoch": 0.48, + "learning_rate": 2.954181549451903e-06, + "loss": 0.6005, + "step": 1290 + }, + { + "epoch": 0.48, + "learning_rate": 2.9541445990885576e-06, + "loss": 0.6836, + "step": 1291 + }, + { + "epoch": 0.48, + "learning_rate": 2.9541076487252127e-06, + "loss": 0.6441, + "step": 1292 + }, + { + "epoch": 0.48, + "learning_rate": 2.9540706983618673e-06, + "loss": 0.6437, + "step": 1293 + }, + { + "epoch": 0.48, + "learning_rate": 2.9540337479985224e-06, + "loss": 0.6363, + "step": 1294 + }, + { + "epoch": 0.48, + "learning_rate": 2.953996797635177e-06, + "loss": 0.7048, + "step": 1295 + }, + { + "epoch": 0.48, + "learning_rate": 2.9539598472718312e-06, + "loss": 0.6182, + "step": 1296 + }, + { + "epoch": 0.48, + "learning_rate": 2.9539228969084863e-06, + "loss": 0.66, + "step": 1297 + }, + { + "epoch": 0.48, + "learning_rate": 2.953885946545141e-06, + "loss": 0.6663, + "step": 1298 + }, + { + "epoch": 0.48, + "learning_rate": 2.953848996181796e-06, + "loss": 0.7592, + "step": 1299 + }, + { + "epoch": 0.48, + "learning_rate": 2.9538120458184506e-06, + "loss": 0.6699, + "step": 1300 + }, + { + "epoch": 0.48, + "learning_rate": 2.9537750954551057e-06, + "loss": 0.6314, + "step": 1301 + }, + { + "epoch": 0.48, + "learning_rate": 2.95373814509176e-06, + "loss": 0.6668, + "step": 1302 + }, + { + "epoch": 0.48, + "learning_rate": 2.953701194728415e-06, + "loss": 0.6996, + "step": 1303 + }, + { + "epoch": 0.48, + "learning_rate": 2.9536642443650696e-06, + "loss": 0.6752, + "step": 1304 + }, + { + "epoch": 0.48, + "learning_rate": 2.9536272940017242e-06, + "loss": 0.6754, + "step": 1305 + }, + { + "epoch": 0.48, + "learning_rate": 2.9535903436383793e-06, + "loss": 0.5899, + "step": 1306 + }, + { + "epoch": 0.48, + "learning_rate": 2.953553393275034e-06, + "loss": 0.6525, + "step": 1307 + }, + { + "epoch": 0.48, + "learning_rate": 2.9535164429116886e-06, + "loss": 0.6514, + "step": 1308 + }, + { + "epoch": 0.48, + "learning_rate": 2.953479492548343e-06, + "loss": 0.6837, + "step": 1309 + }, + { + "epoch": 0.48, + "learning_rate": 2.9534425421849983e-06, + "loss": 0.6195, + "step": 1310 + }, + { + "epoch": 0.48, + "learning_rate": 2.953405591821653e-06, + "loss": 0.6262, + "step": 1311 + }, + { + "epoch": 0.48, + "learning_rate": 2.953368641458308e-06, + "loss": 0.6639, + "step": 1312 + }, + { + "epoch": 0.48, + "learning_rate": 2.9533316910949626e-06, + "loss": 0.6482, + "step": 1313 + }, + { + "epoch": 0.49, + "learning_rate": 2.9532947407316172e-06, + "loss": 0.6526, + "step": 1314 + }, + { + "epoch": 0.49, + "learning_rate": 2.953257790368272e-06, + "loss": 0.6765, + "step": 1315 + }, + { + "epoch": 0.49, + "learning_rate": 2.9532208400049265e-06, + "loss": 0.6425, + "step": 1316 + }, + { + "epoch": 0.49, + "learning_rate": 2.9531838896415816e-06, + "loss": 0.6774, + "step": 1317 + }, + { + "epoch": 0.49, + "learning_rate": 2.953146939278236e-06, + "loss": 0.6693, + "step": 1318 + }, + { + "epoch": 0.49, + "learning_rate": 2.9531099889148913e-06, + "loss": 0.6986, + "step": 1319 + }, + { + "epoch": 0.49, + "learning_rate": 2.953073038551546e-06, + "loss": 0.6556, + "step": 1320 + }, + { + "epoch": 0.49, + "learning_rate": 2.9530360881882005e-06, + "loss": 0.6698, + "step": 1321 + }, + { + "epoch": 0.49, + "learning_rate": 2.952999137824855e-06, + "loss": 0.6342, + "step": 1322 + }, + { + "epoch": 0.49, + "learning_rate": 2.9529621874615102e-06, + "loss": 0.6255, + "step": 1323 + }, + { + "epoch": 0.49, + "learning_rate": 2.952925237098165e-06, + "loss": 0.6557, + "step": 1324 + }, + { + "epoch": 0.49, + "learning_rate": 2.9528882867348195e-06, + "loss": 0.6561, + "step": 1325 + }, + { + "epoch": 0.49, + "learning_rate": 2.9528513363714746e-06, + "loss": 0.6113, + "step": 1326 + }, + { + "epoch": 0.49, + "learning_rate": 2.952814386008129e-06, + "loss": 0.6924, + "step": 1327 + }, + { + "epoch": 0.49, + "learning_rate": 2.952777435644784e-06, + "loss": 0.6378, + "step": 1328 + }, + { + "epoch": 0.49, + "learning_rate": 2.9527404852814385e-06, + "loss": 0.6439, + "step": 1329 + }, + { + "epoch": 0.49, + "learning_rate": 2.9527035349180935e-06, + "loss": 0.6055, + "step": 1330 + }, + { + "epoch": 0.49, + "learning_rate": 2.952666584554748e-06, + "loss": 0.6272, + "step": 1331 + }, + { + "epoch": 0.49, + "learning_rate": 2.9526296341914032e-06, + "loss": 0.6517, + "step": 1332 + }, + { + "epoch": 0.49, + "learning_rate": 2.952592683828058e-06, + "loss": 0.6951, + "step": 1333 + }, + { + "epoch": 0.49, + "learning_rate": 2.9525557334647125e-06, + "loss": 0.6539, + "step": 1334 + }, + { + "epoch": 0.49, + "learning_rate": 2.952518783101367e-06, + "loss": 0.6973, + "step": 1335 + }, + { + "epoch": 0.49, + "learning_rate": 2.9524818327380218e-06, + "loss": 0.6392, + "step": 1336 + }, + { + "epoch": 0.49, + "learning_rate": 2.952444882374677e-06, + "loss": 0.6404, + "step": 1337 + }, + { + "epoch": 0.49, + "learning_rate": 2.9524079320113315e-06, + "loss": 0.6591, + "step": 1338 + }, + { + "epoch": 0.49, + "learning_rate": 2.9523709816479865e-06, + "loss": 0.6644, + "step": 1339 + }, + { + "epoch": 0.49, + "learning_rate": 2.952334031284641e-06, + "loss": 0.7159, + "step": 1340 + }, + { + "epoch": 0.5, + "learning_rate": 2.952297080921296e-06, + "loss": 0.6996, + "step": 1341 + }, + { + "epoch": 0.5, + "learning_rate": 2.9522601305579504e-06, + "loss": 0.6181, + "step": 1342 + }, + { + "epoch": 0.5, + "learning_rate": 2.9522231801946055e-06, + "loss": 0.6121, + "step": 1343 + }, + { + "epoch": 0.5, + "learning_rate": 2.95218622983126e-06, + "loss": 0.6654, + "step": 1344 + }, + { + "epoch": 0.5, + "learning_rate": 2.9521492794679148e-06, + "loss": 0.6424, + "step": 1345 + }, + { + "epoch": 0.5, + "learning_rate": 2.95211232910457e-06, + "loss": 0.6783, + "step": 1346 + }, + { + "epoch": 0.5, + "learning_rate": 2.952075378741224e-06, + "loss": 0.6818, + "step": 1347 + }, + { + "epoch": 0.5, + "learning_rate": 2.952038428377879e-06, + "loss": 0.6284, + "step": 1348 + }, + { + "epoch": 0.5, + "learning_rate": 2.9520014780145337e-06, + "loss": 0.6934, + "step": 1349 + }, + { + "epoch": 0.5, + "learning_rate": 2.951964527651189e-06, + "loss": 0.6387, + "step": 1350 + }, + { + "epoch": 0.5, + "learning_rate": 2.9519275772878434e-06, + "loss": 0.6435, + "step": 1351 + }, + { + "epoch": 0.5, + "learning_rate": 2.9518906269244985e-06, + "loss": 0.7045, + "step": 1352 + }, + { + "epoch": 0.5, + "learning_rate": 2.9518536765611527e-06, + "loss": 0.6845, + "step": 1353 + }, + { + "epoch": 0.5, + "learning_rate": 2.9518167261978078e-06, + "loss": 0.6286, + "step": 1354 + }, + { + "epoch": 0.5, + "learning_rate": 2.9517797758344624e-06, + "loss": 0.6746, + "step": 1355 + }, + { + "epoch": 0.5, + "learning_rate": 2.951742825471117e-06, + "loss": 0.6303, + "step": 1356 + }, + { + "epoch": 0.5, + "learning_rate": 2.951705875107772e-06, + "loss": 0.6482, + "step": 1357 + }, + { + "epoch": 0.5, + "learning_rate": 2.9516689247444267e-06, + "loss": 0.6625, + "step": 1358 + }, + { + "epoch": 0.5, + "learning_rate": 2.9516319743810814e-06, + "loss": 0.6561, + "step": 1359 + }, + { + "epoch": 0.5, + "learning_rate": 2.951595024017736e-06, + "loss": 0.6631, + "step": 1360 + }, + { + "epoch": 0.5, + "learning_rate": 2.951558073654391e-06, + "loss": 0.6233, + "step": 1361 + }, + { + "epoch": 0.5, + "learning_rate": 2.9515211232910457e-06, + "loss": 0.6677, + "step": 1362 + }, + { + "epoch": 0.5, + "learning_rate": 2.9514841729277008e-06, + "loss": 0.72, + "step": 1363 + }, + { + "epoch": 0.5, + "learning_rate": 2.9514472225643554e-06, + "loss": 0.6385, + "step": 1364 + }, + { + "epoch": 0.5, + "learning_rate": 2.95141027220101e-06, + "loss": 0.6413, + "step": 1365 + }, + { + "epoch": 0.5, + "learning_rate": 2.9513733218376647e-06, + "loss": 0.6046, + "step": 1366 + }, + { + "epoch": 0.5, + "learning_rate": 2.9513363714743193e-06, + "loss": 0.6564, + "step": 1367 + }, + { + "epoch": 0.51, + "learning_rate": 2.9512994211109744e-06, + "loss": 0.62, + "step": 1368 + }, + { + "epoch": 0.51, + "learning_rate": 2.951262470747629e-06, + "loss": 0.6423, + "step": 1369 + }, + { + "epoch": 0.51, + "learning_rate": 2.951225520384284e-06, + "loss": 0.6605, + "step": 1370 + }, + { + "epoch": 0.51, + "learning_rate": 2.9511885700209387e-06, + "loss": 0.6568, + "step": 1371 + }, + { + "epoch": 0.51, + "learning_rate": 2.9511516196575933e-06, + "loss": 0.6966, + "step": 1372 + }, + { + "epoch": 0.51, + "learning_rate": 2.951114669294248e-06, + "loss": 0.6813, + "step": 1373 + }, + { + "epoch": 0.51, + "learning_rate": 2.951077718930903e-06, + "loss": 0.6081, + "step": 1374 + }, + { + "epoch": 0.51, + "learning_rate": 2.9510407685675577e-06, + "loss": 0.5718, + "step": 1375 + }, + { + "epoch": 0.51, + "learning_rate": 2.9510038182042123e-06, + "loss": 0.6676, + "step": 1376 + }, + { + "epoch": 0.51, + "learning_rate": 2.9509668678408674e-06, + "loss": 0.6415, + "step": 1377 + }, + { + "epoch": 0.51, + "learning_rate": 2.950929917477522e-06, + "loss": 0.5946, + "step": 1378 + }, + { + "epoch": 0.51, + "learning_rate": 2.9508929671141766e-06, + "loss": 0.6498, + "step": 1379 + }, + { + "epoch": 0.51, + "learning_rate": 2.9508560167508313e-06, + "loss": 0.6076, + "step": 1380 + }, + { + "epoch": 0.51, + "learning_rate": 2.9508190663874863e-06, + "loss": 0.6661, + "step": 1381 + }, + { + "epoch": 0.51, + "learning_rate": 2.950782116024141e-06, + "loss": 0.6876, + "step": 1382 + }, + { + "epoch": 0.51, + "learning_rate": 2.950745165660796e-06, + "loss": 0.6074, + "step": 1383 + }, + { + "epoch": 0.51, + "learning_rate": 2.9507082152974507e-06, + "loss": 0.7019, + "step": 1384 + }, + { + "epoch": 0.51, + "learning_rate": 2.9506712649341053e-06, + "loss": 0.6745, + "step": 1385 + }, + { + "epoch": 0.51, + "learning_rate": 2.95063431457076e-06, + "loss": 0.6851, + "step": 1386 + }, + { + "epoch": 0.51, + "learning_rate": 2.9505973642074146e-06, + "loss": 0.6668, + "step": 1387 + }, + { + "epoch": 0.51, + "learning_rate": 2.9505604138440696e-06, + "loss": 0.6692, + "step": 1388 + }, + { + "epoch": 0.51, + "learning_rate": 2.9505234634807243e-06, + "loss": 0.6619, + "step": 1389 + }, + { + "epoch": 0.51, + "learning_rate": 2.9504865131173793e-06, + "loss": 0.6486, + "step": 1390 + }, + { + "epoch": 0.51, + "learning_rate": 2.950449562754034e-06, + "loss": 0.6472, + "step": 1391 + }, + { + "epoch": 0.51, + "learning_rate": 2.9504126123906886e-06, + "loss": 0.6799, + "step": 1392 + }, + { + "epoch": 0.51, + "learning_rate": 2.9503756620273432e-06, + "loss": 0.6037, + "step": 1393 + }, + { + "epoch": 0.51, + "learning_rate": 2.9503387116639983e-06, + "loss": 0.6759, + "step": 1394 + }, + { + "epoch": 0.52, + "learning_rate": 2.950301761300653e-06, + "loss": 0.6131, + "step": 1395 + }, + { + "epoch": 0.52, + "learning_rate": 2.9502648109373076e-06, + "loss": 0.65, + "step": 1396 + }, + { + "epoch": 0.52, + "learning_rate": 2.9502278605739626e-06, + "loss": 0.7248, + "step": 1397 + }, + { + "epoch": 0.52, + "learning_rate": 2.950190910210617e-06, + "loss": 0.67, + "step": 1398 + }, + { + "epoch": 0.52, + "learning_rate": 2.950153959847272e-06, + "loss": 0.6614, + "step": 1399 + }, + { + "epoch": 0.52, + "learning_rate": 2.9501170094839266e-06, + "loss": 0.631, + "step": 1400 + }, + { + "epoch": 0.52, + "learning_rate": 2.9500800591205816e-06, + "loss": 0.7149, + "step": 1401 + }, + { + "epoch": 0.52, + "learning_rate": 2.9500431087572362e-06, + "loss": 0.6127, + "step": 1402 + }, + { + "epoch": 0.52, + "learning_rate": 2.950006158393891e-06, + "loss": 0.6302, + "step": 1403 + }, + { + "epoch": 0.52, + "learning_rate": 2.9499692080305455e-06, + "loss": 0.6853, + "step": 1404 + }, + { + "epoch": 0.52, + "learning_rate": 2.9499322576672e-06, + "loss": 0.6231, + "step": 1405 + }, + { + "epoch": 0.52, + "learning_rate": 2.9498953073038552e-06, + "loss": 0.6295, + "step": 1406 + }, + { + "epoch": 0.52, + "learning_rate": 2.94985835694051e-06, + "loss": 0.6471, + "step": 1407 + }, + { + "epoch": 0.52, + "learning_rate": 2.949821406577165e-06, + "loss": 0.6568, + "step": 1408 + }, + { + "epoch": 0.52, + "learning_rate": 2.9497844562138195e-06, + "loss": 0.6445, + "step": 1409 + }, + { + "epoch": 0.52, + "learning_rate": 2.949747505850474e-06, + "loss": 0.6092, + "step": 1410 + }, + { + "epoch": 0.52, + "learning_rate": 2.949710555487129e-06, + "loss": 0.6409, + "step": 1411 + }, + { + "epoch": 0.52, + "learning_rate": 2.949673605123784e-06, + "loss": 0.6251, + "step": 1412 + }, + { + "epoch": 0.52, + "learning_rate": 2.9496366547604385e-06, + "loss": 0.6474, + "step": 1413 + }, + { + "epoch": 0.52, + "learning_rate": 2.9495997043970936e-06, + "loss": 0.666, + "step": 1414 + }, + { + "epoch": 0.52, + "learning_rate": 2.9495627540337482e-06, + "loss": 0.6425, + "step": 1415 + }, + { + "epoch": 0.52, + "learning_rate": 2.949525803670403e-06, + "loss": 0.6325, + "step": 1416 + }, + { + "epoch": 0.52, + "learning_rate": 2.9494888533070575e-06, + "loss": 0.6539, + "step": 1417 + }, + { + "epoch": 0.52, + "learning_rate": 2.949451902943712e-06, + "loss": 0.699, + "step": 1418 + }, + { + "epoch": 0.52, + "learning_rate": 2.949414952580367e-06, + "loss": 0.6603, + "step": 1419 + }, + { + "epoch": 0.52, + "learning_rate": 2.949378002217022e-06, + "loss": 0.6093, + "step": 1420 + }, + { + "epoch": 0.52, + "learning_rate": 2.949341051853677e-06, + "loss": 0.6429, + "step": 1421 + }, + { + "epoch": 0.53, + "learning_rate": 2.9493041014903315e-06, + "loss": 0.7128, + "step": 1422 + }, + { + "epoch": 0.53, + "learning_rate": 2.949267151126986e-06, + "loss": 0.6313, + "step": 1423 + }, + { + "epoch": 0.53, + "learning_rate": 2.949230200763641e-06, + "loss": 0.6726, + "step": 1424 + }, + { + "epoch": 0.53, + "learning_rate": 2.9491932504002954e-06, + "loss": 0.6511, + "step": 1425 + }, + { + "epoch": 0.53, + "learning_rate": 2.9491563000369505e-06, + "loss": 0.6689, + "step": 1426 + }, + { + "epoch": 0.53, + "learning_rate": 2.949119349673605e-06, + "loss": 0.6629, + "step": 1427 + }, + { + "epoch": 0.53, + "learning_rate": 2.94908239931026e-06, + "loss": 0.6918, + "step": 1428 + }, + { + "epoch": 0.53, + "learning_rate": 2.949045448946915e-06, + "loss": 0.6408, + "step": 1429 + }, + { + "epoch": 0.53, + "learning_rate": 2.9490084985835695e-06, + "loss": 0.6297, + "step": 1430 + }, + { + "epoch": 0.53, + "learning_rate": 2.948971548220224e-06, + "loss": 0.6568, + "step": 1431 + }, + { + "epoch": 0.53, + "learning_rate": 2.948934597856879e-06, + "loss": 0.6118, + "step": 1432 + }, + { + "epoch": 0.53, + "learning_rate": 2.948897647493534e-06, + "loss": 0.6478, + "step": 1433 + }, + { + "epoch": 0.53, + "learning_rate": 2.948860697130189e-06, + "loss": 0.6949, + "step": 1434 + }, + { + "epoch": 0.53, + "learning_rate": 2.9488237467668435e-06, + "loss": 0.6259, + "step": 1435 + }, + { + "epoch": 0.53, + "learning_rate": 2.9487867964034977e-06, + "loss": 0.5911, + "step": 1436 + }, + { + "epoch": 0.53, + "learning_rate": 2.9487498460401528e-06, + "loss": 0.6607, + "step": 1437 + }, + { + "epoch": 0.53, + "learning_rate": 2.9487128956768074e-06, + "loss": 0.6715, + "step": 1438 + }, + { + "epoch": 0.53, + "learning_rate": 2.9486759453134625e-06, + "loss": 0.6197, + "step": 1439 + }, + { + "epoch": 0.53, + "learning_rate": 2.948638994950117e-06, + "loss": 0.6005, + "step": 1440 + }, + { + "epoch": 0.53, + "learning_rate": 2.948602044586772e-06, + "loss": 0.681, + "step": 1441 + }, + { + "epoch": 0.53, + "learning_rate": 2.9485650942234264e-06, + "loss": 0.6157, + "step": 1442 + }, + { + "epoch": 0.53, + "learning_rate": 2.9485281438600814e-06, + "loss": 0.606, + "step": 1443 + }, + { + "epoch": 0.53, + "learning_rate": 2.948491193496736e-06, + "loss": 0.6274, + "step": 1444 + }, + { + "epoch": 0.53, + "learning_rate": 2.9484542431333907e-06, + "loss": 0.6878, + "step": 1445 + }, + { + "epoch": 0.53, + "learning_rate": 2.9484172927700458e-06, + "loss": 0.6988, + "step": 1446 + }, + { + "epoch": 0.53, + "learning_rate": 2.9483803424067004e-06, + "loss": 0.6841, + "step": 1447 + }, + { + "epoch": 0.53, + "learning_rate": 2.948343392043355e-06, + "loss": 0.5983, + "step": 1448 + }, + { + "epoch": 0.54, + "learning_rate": 2.9483064416800097e-06, + "loss": 0.6231, + "step": 1449 + }, + { + "epoch": 0.54, + "learning_rate": 2.9482694913166647e-06, + "loss": 0.62, + "step": 1450 + }, + { + "epoch": 0.54, + "learning_rate": 2.9482325409533194e-06, + "loss": 0.6525, + "step": 1451 + }, + { + "epoch": 0.54, + "learning_rate": 2.9481955905899744e-06, + "loss": 0.6653, + "step": 1452 + }, + { + "epoch": 0.54, + "learning_rate": 2.948158640226629e-06, + "loss": 0.646, + "step": 1453 + }, + { + "epoch": 0.54, + "learning_rate": 2.9481216898632837e-06, + "loss": 0.6512, + "step": 1454 + }, + { + "epoch": 0.54, + "learning_rate": 2.9480847394999383e-06, + "loss": 0.6375, + "step": 1455 + }, + { + "epoch": 0.54, + "learning_rate": 2.948047789136593e-06, + "loss": 0.6664, + "step": 1456 + }, + { + "epoch": 0.54, + "learning_rate": 2.948010838773248e-06, + "loss": 0.6468, + "step": 1457 + }, + { + "epoch": 0.54, + "learning_rate": 2.9479738884099027e-06, + "loss": 0.6618, + "step": 1458 + }, + { + "epoch": 0.54, + "learning_rate": 2.9479369380465577e-06, + "loss": 0.6745, + "step": 1459 + }, + { + "epoch": 0.54, + "learning_rate": 2.9478999876832124e-06, + "loss": 0.6139, + "step": 1460 + }, + { + "epoch": 0.54, + "learning_rate": 2.947863037319867e-06, + "loss": 0.6451, + "step": 1461 + }, + { + "epoch": 0.54, + "learning_rate": 2.9478260869565216e-06, + "loss": 0.6464, + "step": 1462 + }, + { + "epoch": 0.54, + "learning_rate": 2.9477891365931767e-06, + "loss": 0.7, + "step": 1463 + }, + { + "epoch": 0.54, + "learning_rate": 2.9477521862298313e-06, + "loss": 0.6671, + "step": 1464 + }, + { + "epoch": 0.54, + "learning_rate": 2.947715235866486e-06, + "loss": 0.6743, + "step": 1465 + }, + { + "epoch": 0.54, + "learning_rate": 2.947678285503141e-06, + "loss": 0.6243, + "step": 1466 + }, + { + "epoch": 0.54, + "learning_rate": 2.9476413351397957e-06, + "loss": 0.6539, + "step": 1467 + }, + { + "epoch": 0.54, + "learning_rate": 2.9476043847764503e-06, + "loss": 0.7235, + "step": 1468 + }, + { + "epoch": 0.54, + "learning_rate": 2.947567434413105e-06, + "loss": 0.6859, + "step": 1469 + }, + { + "epoch": 0.54, + "learning_rate": 2.94753048404976e-06, + "loss": 0.6222, + "step": 1470 + }, + { + "epoch": 0.54, + "learning_rate": 2.9474935336864146e-06, + "loss": 0.6519, + "step": 1471 + }, + { + "epoch": 0.54, + "learning_rate": 2.9474565833230697e-06, + "loss": 0.5848, + "step": 1472 + }, + { + "epoch": 0.54, + "learning_rate": 2.9474196329597243e-06, + "loss": 0.6688, + "step": 1473 + }, + { + "epoch": 0.54, + "learning_rate": 2.947382682596379e-06, + "loss": 0.6342, + "step": 1474 + }, + { + "epoch": 0.54, + "learning_rate": 2.9473457322330336e-06, + "loss": 0.6393, + "step": 1475 + }, + { + "epoch": 0.55, + "learning_rate": 2.9473087818696882e-06, + "loss": 0.6436, + "step": 1476 + }, + { + "epoch": 0.55, + "learning_rate": 2.9472718315063433e-06, + "loss": 0.6726, + "step": 1477 + }, + { + "epoch": 0.55, + "learning_rate": 2.947234881142998e-06, + "loss": 0.6311, + "step": 1478 + }, + { + "epoch": 0.55, + "learning_rate": 2.947197930779653e-06, + "loss": 0.6223, + "step": 1479 + }, + { + "epoch": 0.55, + "learning_rate": 2.9471609804163076e-06, + "loss": 0.6794, + "step": 1480 + }, + { + "epoch": 0.55, + "learning_rate": 2.9471240300529623e-06, + "loss": 0.6338, + "step": 1481 + }, + { + "epoch": 0.55, + "learning_rate": 2.947087079689617e-06, + "loss": 0.633, + "step": 1482 + }, + { + "epoch": 0.55, + "learning_rate": 2.947050129326272e-06, + "loss": 0.6416, + "step": 1483 + }, + { + "epoch": 0.55, + "learning_rate": 2.9470131789629266e-06, + "loss": 0.7015, + "step": 1484 + }, + { + "epoch": 0.55, + "learning_rate": 2.9469762285995812e-06, + "loss": 0.6328, + "step": 1485 + }, + { + "epoch": 0.55, + "learning_rate": 2.9469392782362363e-06, + "loss": 0.6992, + "step": 1486 + }, + { + "epoch": 0.55, + "learning_rate": 2.9469023278728905e-06, + "loss": 0.5929, + "step": 1487 + }, + { + "epoch": 0.55, + "learning_rate": 2.9468653775095456e-06, + "loss": 0.6508, + "step": 1488 + }, + { + "epoch": 0.55, + "learning_rate": 2.9468284271462e-06, + "loss": 0.7061, + "step": 1489 + }, + { + "epoch": 0.55, + "learning_rate": 2.9467914767828553e-06, + "loss": 0.6155, + "step": 1490 + }, + { + "epoch": 0.55, + "learning_rate": 2.94675452641951e-06, + "loss": 0.7134, + "step": 1491 + }, + { + "epoch": 0.55, + "learning_rate": 2.946717576056165e-06, + "loss": 0.6989, + "step": 1492 + }, + { + "epoch": 0.55, + "learning_rate": 2.946680625692819e-06, + "loss": 0.6657, + "step": 1493 + }, + { + "epoch": 0.55, + "learning_rate": 2.946643675329474e-06, + "loss": 0.6548, + "step": 1494 + }, + { + "epoch": 0.55, + "learning_rate": 2.946606724966129e-06, + "loss": 0.599, + "step": 1495 + }, + { + "epoch": 0.55, + "learning_rate": 2.9465697746027835e-06, + "loss": 0.6675, + "step": 1496 + }, + { + "epoch": 0.55, + "learning_rate": 2.9465328242394386e-06, + "loss": 0.6179, + "step": 1497 + }, + { + "epoch": 0.55, + "learning_rate": 2.946495873876093e-06, + "loss": 0.6089, + "step": 1498 + }, + { + "epoch": 0.55, + "learning_rate": 2.946458923512748e-06, + "loss": 0.6645, + "step": 1499 + }, + { + "epoch": 0.55, + "learning_rate": 2.9464219731494025e-06, + "loss": 0.6375, + "step": 1500 + }, + { + "epoch": 0.55, + "learning_rate": 2.9463850227860575e-06, + "loss": 0.6852, + "step": 1501 + }, + { + "epoch": 0.55, + "learning_rate": 2.946348072422712e-06, + "loss": 0.5592, + "step": 1502 + }, + { + "epoch": 0.56, + "learning_rate": 2.9463111220593672e-06, + "loss": 0.6382, + "step": 1503 + }, + { + "epoch": 0.56, + "learning_rate": 2.946274171696022e-06, + "loss": 0.5968, + "step": 1504 + }, + { + "epoch": 0.56, + "learning_rate": 2.9462372213326765e-06, + "loss": 0.6746, + "step": 1505 + }, + { + "epoch": 0.56, + "learning_rate": 2.946200270969331e-06, + "loss": 0.5884, + "step": 1506 + }, + { + "epoch": 0.56, + "learning_rate": 2.9461633206059858e-06, + "loss": 0.6483, + "step": 1507 + }, + { + "epoch": 0.56, + "learning_rate": 2.946126370242641e-06, + "loss": 0.6447, + "step": 1508 + }, + { + "epoch": 0.56, + "learning_rate": 2.9460894198792955e-06, + "loss": 0.657, + "step": 1509 + }, + { + "epoch": 0.56, + "learning_rate": 2.9460524695159505e-06, + "loss": 0.6739, + "step": 1510 + }, + { + "epoch": 0.56, + "learning_rate": 2.946015519152605e-06, + "loss": 0.6626, + "step": 1511 + }, + { + "epoch": 0.56, + "learning_rate": 2.94597856878926e-06, + "loss": 0.6264, + "step": 1512 + }, + { + "epoch": 0.56, + "learning_rate": 2.9459416184259144e-06, + "loss": 0.6106, + "step": 1513 + }, + { + "epoch": 0.56, + "learning_rate": 2.945904668062569e-06, + "loss": 0.5632, + "step": 1514 + }, + { + "epoch": 0.56, + "learning_rate": 2.945867717699224e-06, + "loss": 0.6359, + "step": 1515 + }, + { + "epoch": 0.56, + "learning_rate": 2.9458307673358788e-06, + "loss": 0.6009, + "step": 1516 + }, + { + "epoch": 0.56, + "learning_rate": 2.945793816972534e-06, + "loss": 0.6375, + "step": 1517 + }, + { + "epoch": 0.56, + "learning_rate": 2.9457568666091885e-06, + "loss": 0.6111, + "step": 1518 + }, + { + "epoch": 0.56, + "learning_rate": 2.945719916245843e-06, + "loss": 0.6583, + "step": 1519 + }, + { + "epoch": 0.56, + "learning_rate": 2.9456829658824977e-06, + "loss": 0.6279, + "step": 1520 + }, + { + "epoch": 0.56, + "learning_rate": 2.945646015519153e-06, + "loss": 0.6233, + "step": 1521 + }, + { + "epoch": 0.56, + "learning_rate": 2.9456090651558074e-06, + "loss": 0.6219, + "step": 1522 + }, + { + "epoch": 0.56, + "learning_rate": 2.9455721147924625e-06, + "loss": 0.6154, + "step": 1523 + }, + { + "epoch": 0.56, + "learning_rate": 2.945535164429117e-06, + "loss": 0.6377, + "step": 1524 + }, + { + "epoch": 0.56, + "learning_rate": 2.9454982140657718e-06, + "loss": 0.6672, + "step": 1525 + }, + { + "epoch": 0.56, + "learning_rate": 2.9454612637024264e-06, + "loss": 0.6012, + "step": 1526 + }, + { + "epoch": 0.56, + "learning_rate": 2.945424313339081e-06, + "loss": 0.6395, + "step": 1527 + }, + { + "epoch": 0.56, + "learning_rate": 2.945387362975736e-06, + "loss": 0.6564, + "step": 1528 + }, + { + "epoch": 0.56, + "learning_rate": 2.9453504126123907e-06, + "loss": 0.6385, + "step": 1529 + }, + { + "epoch": 0.56, + "learning_rate": 2.945313462249046e-06, + "loss": 0.6486, + "step": 1530 + }, + { + "epoch": 0.57, + "learning_rate": 2.9452765118857004e-06, + "loss": 0.7127, + "step": 1531 + }, + { + "epoch": 0.57, + "learning_rate": 2.945239561522355e-06, + "loss": 0.6577, + "step": 1532 + }, + { + "epoch": 0.57, + "learning_rate": 2.9452026111590097e-06, + "loss": 0.5877, + "step": 1533 + }, + { + "epoch": 0.57, + "learning_rate": 2.9451656607956644e-06, + "loss": 0.6839, + "step": 1534 + }, + { + "epoch": 0.57, + "learning_rate": 2.9451287104323194e-06, + "loss": 0.6624, + "step": 1535 + }, + { + "epoch": 0.57, + "learning_rate": 2.945091760068974e-06, + "loss": 0.633, + "step": 1536 + }, + { + "epoch": 0.57, + "learning_rate": 2.945054809705629e-06, + "loss": 0.64, + "step": 1537 + }, + { + "epoch": 0.57, + "learning_rate": 2.9450178593422833e-06, + "loss": 0.6667, + "step": 1538 + }, + { + "epoch": 0.57, + "learning_rate": 2.9449809089789384e-06, + "loss": 0.6121, + "step": 1539 + }, + { + "epoch": 0.57, + "learning_rate": 2.944943958615593e-06, + "loss": 0.6003, + "step": 1540 + }, + { + "epoch": 0.57, + "learning_rate": 2.944907008252248e-06, + "loss": 0.6611, + "step": 1541 + }, + { + "epoch": 0.57, + "learning_rate": 2.9448700578889027e-06, + "loss": 0.6038, + "step": 1542 + }, + { + "epoch": 0.57, + "learning_rate": 2.9448331075255578e-06, + "loss": 0.6645, + "step": 1543 + }, + { + "epoch": 0.57, + "learning_rate": 2.944796157162212e-06, + "loss": 0.6585, + "step": 1544 + }, + { + "epoch": 0.57, + "learning_rate": 2.9447592067988666e-06, + "loss": 0.6312, + "step": 1545 + }, + { + "epoch": 0.57, + "learning_rate": 2.9447222564355217e-06, + "loss": 0.6741, + "step": 1546 + }, + { + "epoch": 0.57, + "learning_rate": 2.9446853060721763e-06, + "loss": 0.6618, + "step": 1547 + }, + { + "epoch": 0.57, + "learning_rate": 2.9446483557088314e-06, + "loss": 0.6907, + "step": 1548 + }, + { + "epoch": 0.57, + "learning_rate": 2.944611405345486e-06, + "loss": 0.5723, + "step": 1549 + }, + { + "epoch": 0.57, + "learning_rate": 2.9445744549821407e-06, + "loss": 0.6311, + "step": 1550 + }, + { + "epoch": 0.57, + "learning_rate": 2.9445375046187953e-06, + "loss": 0.6431, + "step": 1551 + }, + { + "epoch": 0.57, + "learning_rate": 2.9445005542554503e-06, + "loss": 0.6326, + "step": 1552 + }, + { + "epoch": 0.57, + "learning_rate": 2.944463603892105e-06, + "loss": 0.698, + "step": 1553 + }, + { + "epoch": 0.57, + "learning_rate": 2.9444266535287596e-06, + "loss": 0.5756, + "step": 1554 + }, + { + "epoch": 0.57, + "learning_rate": 2.9443897031654147e-06, + "loss": 0.6388, + "step": 1555 + }, + { + "epoch": 0.57, + "learning_rate": 2.9443527528020693e-06, + "loss": 0.6823, + "step": 1556 + }, + { + "epoch": 0.57, + "learning_rate": 2.944315802438724e-06, + "loss": 0.6888, + "step": 1557 + }, + { + "epoch": 0.58, + "learning_rate": 2.9442788520753786e-06, + "loss": 0.6562, + "step": 1558 + }, + { + "epoch": 0.58, + "learning_rate": 2.9442419017120336e-06, + "loss": 0.655, + "step": 1559 + }, + { + "epoch": 0.58, + "learning_rate": 2.9442049513486883e-06, + "loss": 0.7094, + "step": 1560 + }, + { + "epoch": 0.58, + "learning_rate": 2.9441680009853433e-06, + "loss": 0.679, + "step": 1561 + }, + { + "epoch": 0.58, + "learning_rate": 2.944131050621998e-06, + "loss": 0.6235, + "step": 1562 + }, + { + "epoch": 0.58, + "learning_rate": 2.9440941002586526e-06, + "loss": 0.6588, + "step": 1563 + }, + { + "epoch": 0.58, + "learning_rate": 2.9440571498953073e-06, + "loss": 0.5996, + "step": 1564 + }, + { + "epoch": 0.58, + "learning_rate": 2.944020199531962e-06, + "loss": 0.62, + "step": 1565 + }, + { + "epoch": 0.58, + "learning_rate": 2.943983249168617e-06, + "loss": 0.5891, + "step": 1566 + }, + { + "epoch": 0.58, + "learning_rate": 2.9439462988052716e-06, + "loss": 0.6653, + "step": 1567 + }, + { + "epoch": 0.58, + "learning_rate": 2.9439093484419266e-06, + "loss": 0.6541, + "step": 1568 + }, + { + "epoch": 0.58, + "learning_rate": 2.9438723980785813e-06, + "loss": 0.6799, + "step": 1569 + }, + { + "epoch": 0.58, + "learning_rate": 2.943835447715236e-06, + "loss": 0.6436, + "step": 1570 + }, + { + "epoch": 0.58, + "learning_rate": 2.9437984973518906e-06, + "loss": 0.6086, + "step": 1571 + }, + { + "epoch": 0.58, + "learning_rate": 2.9437615469885456e-06, + "loss": 0.6296, + "step": 1572 + }, + { + "epoch": 0.58, + "learning_rate": 2.9437245966252003e-06, + "loss": 0.6084, + "step": 1573 + }, + { + "epoch": 0.58, + "learning_rate": 2.943687646261855e-06, + "loss": 0.6128, + "step": 1574 + }, + { + "epoch": 0.58, + "learning_rate": 2.94365069589851e-06, + "loss": 0.6261, + "step": 1575 + }, + { + "epoch": 0.58, + "learning_rate": 2.9436137455351646e-06, + "loss": 0.5943, + "step": 1576 + }, + { + "epoch": 0.58, + "learning_rate": 2.9435767951718192e-06, + "loss": 0.6367, + "step": 1577 + }, + { + "epoch": 0.58, + "learning_rate": 2.943539844808474e-06, + "loss": 0.5793, + "step": 1578 + }, + { + "epoch": 0.58, + "learning_rate": 2.943502894445129e-06, + "loss": 0.5947, + "step": 1579 + }, + { + "epoch": 0.58, + "learning_rate": 2.9434659440817836e-06, + "loss": 0.6273, + "step": 1580 + }, + { + "epoch": 0.58, + "learning_rate": 2.9434289937184386e-06, + "loss": 0.6725, + "step": 1581 + }, + { + "epoch": 0.58, + "learning_rate": 2.9433920433550933e-06, + "loss": 0.5999, + "step": 1582 + }, + { + "epoch": 0.58, + "learning_rate": 2.943355092991748e-06, + "loss": 0.6991, + "step": 1583 + }, + { + "epoch": 0.58, + "learning_rate": 2.9433181426284025e-06, + "loss": 0.6006, + "step": 1584 + }, + { + "epoch": 0.59, + "learning_rate": 2.943281192265057e-06, + "loss": 0.711, + "step": 1585 + }, + { + "epoch": 0.59, + "learning_rate": 2.9432442419017122e-06, + "loss": 0.7152, + "step": 1586 + }, + { + "epoch": 0.59, + "learning_rate": 2.943207291538367e-06, + "loss": 0.6571, + "step": 1587 + }, + { + "epoch": 0.59, + "learning_rate": 2.943170341175022e-06, + "loss": 0.6386, + "step": 1588 + }, + { + "epoch": 0.59, + "learning_rate": 2.943133390811676e-06, + "loss": 0.6545, + "step": 1589 + }, + { + "epoch": 0.59, + "learning_rate": 2.943096440448331e-06, + "loss": 0.7249, + "step": 1590 + }, + { + "epoch": 0.59, + "learning_rate": 2.943059490084986e-06, + "loss": 0.6511, + "step": 1591 + }, + { + "epoch": 0.59, + "learning_rate": 2.943022539721641e-06, + "loss": 0.733, + "step": 1592 + }, + { + "epoch": 0.59, + "learning_rate": 2.9429855893582955e-06, + "loss": 0.6702, + "step": 1593 + }, + { + "epoch": 0.59, + "learning_rate": 2.94294863899495e-06, + "loss": 0.6485, + "step": 1594 + }, + { + "epoch": 0.59, + "learning_rate": 2.942911688631605e-06, + "loss": 0.7169, + "step": 1595 + }, + { + "epoch": 0.59, + "learning_rate": 2.9428747382682594e-06, + "loss": 0.6858, + "step": 1596 + }, + { + "epoch": 0.59, + "learning_rate": 2.9428377879049145e-06, + "loss": 0.633, + "step": 1597 + }, + { + "epoch": 0.59, + "learning_rate": 2.942800837541569e-06, + "loss": 0.6431, + "step": 1598 + }, + { + "epoch": 0.59, + "learning_rate": 2.942763887178224e-06, + "loss": 0.6938, + "step": 1599 + }, + { + "epoch": 0.59, + "learning_rate": 2.942726936814879e-06, + "loss": 0.6499, + "step": 1600 + }, + { + "epoch": 0.59, + "learning_rate": 2.9426899864515335e-06, + "loss": 0.6835, + "step": 1601 + }, + { + "epoch": 0.59, + "learning_rate": 2.942653036088188e-06, + "loss": 0.6216, + "step": 1602 + }, + { + "epoch": 0.59, + "learning_rate": 2.942616085724843e-06, + "loss": 0.6887, + "step": 1603 + }, + { + "epoch": 0.59, + "learning_rate": 2.942579135361498e-06, + "loss": 0.6765, + "step": 1604 + }, + { + "epoch": 0.59, + "learning_rate": 2.9425421849981524e-06, + "loss": 0.5611, + "step": 1605 + }, + { + "epoch": 0.59, + "learning_rate": 2.9425052346348075e-06, + "loss": 0.6631, + "step": 1606 + }, + { + "epoch": 0.59, + "learning_rate": 2.942468284271462e-06, + "loss": 0.6282, + "step": 1607 + }, + { + "epoch": 0.59, + "learning_rate": 2.9424313339081168e-06, + "loss": 0.6518, + "step": 1608 + }, + { + "epoch": 0.59, + "learning_rate": 2.9423943835447714e-06, + "loss": 0.6498, + "step": 1609 + }, + { + "epoch": 0.59, + "learning_rate": 2.9423574331814265e-06, + "loss": 0.6328, + "step": 1610 + }, + { + "epoch": 0.59, + "learning_rate": 2.942320482818081e-06, + "loss": 0.6861, + "step": 1611 + }, + { + "epoch": 0.6, + "learning_rate": 2.942283532454736e-06, + "loss": 0.6616, + "step": 1612 + }, + { + "epoch": 0.6, + "learning_rate": 2.942246582091391e-06, + "loss": 0.6611, + "step": 1613 + }, + { + "epoch": 0.6, + "learning_rate": 2.9422096317280454e-06, + "loss": 0.6907, + "step": 1614 + }, + { + "epoch": 0.6, + "learning_rate": 2.9421726813647e-06, + "loss": 0.6546, + "step": 1615 + }, + { + "epoch": 0.6, + "learning_rate": 2.9421357310013547e-06, + "loss": 0.6095, + "step": 1616 + }, + { + "epoch": 0.6, + "learning_rate": 2.9420987806380098e-06, + "loss": 0.6909, + "step": 1617 + }, + { + "epoch": 0.6, + "learning_rate": 2.9420618302746644e-06, + "loss": 0.6538, + "step": 1618 + }, + { + "epoch": 0.6, + "learning_rate": 2.9420248799113195e-06, + "loss": 0.6666, + "step": 1619 + }, + { + "epoch": 0.6, + "learning_rate": 2.941987929547974e-06, + "loss": 0.6093, + "step": 1620 + }, + { + "epoch": 0.6, + "learning_rate": 2.9419509791846287e-06, + "loss": 0.6567, + "step": 1621 + }, + { + "epoch": 0.6, + "learning_rate": 2.9419140288212834e-06, + "loss": 0.5979, + "step": 1622 + }, + { + "epoch": 0.6, + "learning_rate": 2.9418770784579384e-06, + "loss": 0.6735, + "step": 1623 + }, + { + "epoch": 0.6, + "learning_rate": 2.941840128094593e-06, + "loss": 0.6183, + "step": 1624 + }, + { + "epoch": 0.6, + "learning_rate": 2.9418031777312477e-06, + "loss": 0.619, + "step": 1625 + }, + { + "epoch": 0.6, + "learning_rate": 2.9417662273679028e-06, + "loss": 0.6606, + "step": 1626 + }, + { + "epoch": 0.6, + "learning_rate": 2.941729277004557e-06, + "loss": 0.6951, + "step": 1627 + }, + { + "epoch": 0.6, + "learning_rate": 2.941692326641212e-06, + "loss": 0.6672, + "step": 1628 + }, + { + "epoch": 0.6, + "learning_rate": 2.9416553762778667e-06, + "loss": 0.6175, + "step": 1629 + }, + { + "epoch": 0.6, + "learning_rate": 2.9416184259145217e-06, + "loss": 0.6363, + "step": 1630 + }, + { + "epoch": 0.6, + "learning_rate": 2.9415814755511764e-06, + "loss": 0.657, + "step": 1631 + }, + { + "epoch": 0.6, + "learning_rate": 2.9415445251878314e-06, + "loss": 0.6957, + "step": 1632 + }, + { + "epoch": 0.6, + "learning_rate": 2.9415075748244856e-06, + "loss": 0.6972, + "step": 1633 + }, + { + "epoch": 0.6, + "learning_rate": 2.9414706244611403e-06, + "loss": 0.6468, + "step": 1634 + }, + { + "epoch": 0.6, + "learning_rate": 2.9414336740977953e-06, + "loss": 0.6245, + "step": 1635 + }, + { + "epoch": 0.6, + "learning_rate": 2.94139672373445e-06, + "loss": 0.6369, + "step": 1636 + }, + { + "epoch": 0.6, + "learning_rate": 2.941359773371105e-06, + "loss": 0.7247, + "step": 1637 + }, + { + "epoch": 0.6, + "learning_rate": 2.9413228230077597e-06, + "loss": 0.6474, + "step": 1638 + }, + { + "epoch": 0.61, + "learning_rate": 2.9412858726444143e-06, + "loss": 0.6251, + "step": 1639 + }, + { + "epoch": 0.61, + "learning_rate": 2.941248922281069e-06, + "loss": 0.6317, + "step": 1640 + }, + { + "epoch": 0.61, + "learning_rate": 2.941211971917724e-06, + "loss": 0.659, + "step": 1641 + }, + { + "epoch": 0.61, + "learning_rate": 2.9411750215543786e-06, + "loss": 0.6615, + "step": 1642 + }, + { + "epoch": 0.61, + "learning_rate": 2.9411380711910337e-06, + "loss": 0.6421, + "step": 1643 + }, + { + "epoch": 0.61, + "learning_rate": 2.9411011208276883e-06, + "loss": 0.5983, + "step": 1644 + }, + { + "epoch": 0.61, + "learning_rate": 2.941064170464343e-06, + "loss": 0.6667, + "step": 1645 + }, + { + "epoch": 0.61, + "learning_rate": 2.9410272201009976e-06, + "loss": 0.6388, + "step": 1646 + }, + { + "epoch": 0.61, + "learning_rate": 2.9409902697376522e-06, + "loss": 0.688, + "step": 1647 + }, + { + "epoch": 0.61, + "learning_rate": 2.9409533193743073e-06, + "loss": 0.647, + "step": 1648 + }, + { + "epoch": 0.61, + "learning_rate": 2.940916369010962e-06, + "loss": 0.6396, + "step": 1649 + }, + { + "epoch": 0.61, + "learning_rate": 2.940879418647617e-06, + "loss": 0.6978, + "step": 1650 + }, + { + "epoch": 0.61, + "learning_rate": 2.9408424682842716e-06, + "loss": 0.6498, + "step": 1651 + }, + { + "epoch": 0.61, + "learning_rate": 2.9408055179209263e-06, + "loss": 0.6407, + "step": 1652 + }, + { + "epoch": 0.61, + "learning_rate": 2.940768567557581e-06, + "loss": 0.6726, + "step": 1653 + }, + { + "epoch": 0.61, + "learning_rate": 2.9407316171942355e-06, + "loss": 0.6407, + "step": 1654 + }, + { + "epoch": 0.61, + "learning_rate": 2.9406946668308906e-06, + "loss": 0.6624, + "step": 1655 + }, + { + "epoch": 0.61, + "learning_rate": 2.9406577164675452e-06, + "loss": 0.6314, + "step": 1656 + }, + { + "epoch": 0.61, + "learning_rate": 2.9406207661042003e-06, + "loss": 0.6596, + "step": 1657 + }, + { + "epoch": 0.61, + "learning_rate": 2.940583815740855e-06, + "loss": 0.6006, + "step": 1658 + }, + { + "epoch": 0.61, + "learning_rate": 2.9405468653775096e-06, + "loss": 0.643, + "step": 1659 + }, + { + "epoch": 0.61, + "learning_rate": 2.940509915014164e-06, + "loss": 0.6769, + "step": 1660 + }, + { + "epoch": 0.61, + "learning_rate": 2.9404729646508193e-06, + "loss": 0.6018, + "step": 1661 + }, + { + "epoch": 0.61, + "learning_rate": 2.940436014287474e-06, + "loss": 0.6288, + "step": 1662 + }, + { + "epoch": 0.61, + "learning_rate": 2.940399063924129e-06, + "loss": 0.6194, + "step": 1663 + }, + { + "epoch": 0.61, + "learning_rate": 2.9403621135607836e-06, + "loss": 0.6529, + "step": 1664 + }, + { + "epoch": 0.61, + "learning_rate": 2.9403251631974382e-06, + "loss": 0.6642, + "step": 1665 + }, + { + "epoch": 0.62, + "learning_rate": 2.940288212834093e-06, + "loss": 0.6447, + "step": 1666 + }, + { + "epoch": 0.62, + "learning_rate": 2.9402512624707475e-06, + "loss": 0.6652, + "step": 1667 + }, + { + "epoch": 0.62, + "learning_rate": 2.9402143121074026e-06, + "loss": 0.6683, + "step": 1668 + }, + { + "epoch": 0.62, + "learning_rate": 2.940177361744057e-06, + "loss": 0.643, + "step": 1669 + }, + { + "epoch": 0.62, + "learning_rate": 2.9401404113807123e-06, + "loss": 0.649, + "step": 1670 + }, + { + "epoch": 0.62, + "learning_rate": 2.940103461017367e-06, + "loss": 0.6279, + "step": 1671 + }, + { + "epoch": 0.62, + "learning_rate": 2.9400665106540215e-06, + "loss": 0.6118, + "step": 1672 + }, + { + "epoch": 0.62, + "learning_rate": 2.940029560290676e-06, + "loss": 0.5856, + "step": 1673 + }, + { + "epoch": 0.62, + "learning_rate": 2.939992609927331e-06, + "loss": 0.6249, + "step": 1674 + }, + { + "epoch": 0.62, + "learning_rate": 2.939955659563986e-06, + "loss": 0.6097, + "step": 1675 + }, + { + "epoch": 0.62, + "learning_rate": 2.9399187092006405e-06, + "loss": 0.6797, + "step": 1676 + }, + { + "epoch": 0.62, + "learning_rate": 2.9398817588372956e-06, + "loss": 0.6054, + "step": 1677 + }, + { + "epoch": 0.62, + "learning_rate": 2.9398448084739498e-06, + "loss": 0.6452, + "step": 1678 + }, + { + "epoch": 0.62, + "learning_rate": 2.939807858110605e-06, + "loss": 0.5745, + "step": 1679 + }, + { + "epoch": 0.62, + "learning_rate": 2.9397709077472595e-06, + "loss": 0.6019, + "step": 1680 + }, + { + "epoch": 0.62, + "learning_rate": 2.9397339573839145e-06, + "loss": 0.6436, + "step": 1681 + }, + { + "epoch": 0.62, + "learning_rate": 2.939697007020569e-06, + "loss": 0.6008, + "step": 1682 + }, + { + "epoch": 0.62, + "learning_rate": 2.939660056657224e-06, + "loss": 0.6229, + "step": 1683 + }, + { + "epoch": 0.62, + "learning_rate": 2.9396231062938785e-06, + "loss": 0.7104, + "step": 1684 + }, + { + "epoch": 0.62, + "learning_rate": 2.939586155930533e-06, + "loss": 0.6645, + "step": 1685 + }, + { + "epoch": 0.62, + "learning_rate": 2.939549205567188e-06, + "loss": 0.5606, + "step": 1686 + }, + { + "epoch": 0.62, + "learning_rate": 2.9395122552038428e-06, + "loss": 0.6264, + "step": 1687 + }, + { + "epoch": 0.62, + "learning_rate": 2.939475304840498e-06, + "loss": 0.6653, + "step": 1688 + }, + { + "epoch": 0.62, + "learning_rate": 2.9394383544771525e-06, + "loss": 0.6488, + "step": 1689 + }, + { + "epoch": 0.62, + "learning_rate": 2.939401404113807e-06, + "loss": 0.6184, + "step": 1690 + }, + { + "epoch": 0.62, + "learning_rate": 2.9393644537504618e-06, + "loss": 0.653, + "step": 1691 + }, + { + "epoch": 0.62, + "learning_rate": 2.939327503387117e-06, + "loss": 0.6369, + "step": 1692 + }, + { + "epoch": 0.63, + "learning_rate": 2.9392905530237714e-06, + "loss": 0.6707, + "step": 1693 + }, + { + "epoch": 0.63, + "learning_rate": 2.939253602660426e-06, + "loss": 0.6014, + "step": 1694 + }, + { + "epoch": 0.63, + "learning_rate": 2.939216652297081e-06, + "loss": 0.6211, + "step": 1695 + }, + { + "epoch": 0.63, + "learning_rate": 2.9391797019337358e-06, + "loss": 0.6324, + "step": 1696 + }, + { + "epoch": 0.63, + "learning_rate": 2.9391427515703904e-06, + "loss": 0.6405, + "step": 1697 + }, + { + "epoch": 0.63, + "learning_rate": 2.939105801207045e-06, + "loss": 0.6984, + "step": 1698 + }, + { + "epoch": 0.63, + "learning_rate": 2.9390688508437e-06, + "loss": 0.6165, + "step": 1699 + }, + { + "epoch": 0.63, + "learning_rate": 2.9390319004803548e-06, + "loss": 0.6193, + "step": 1700 + }, + { + "epoch": 0.63, + "learning_rate": 2.93899495011701e-06, + "loss": 0.6759, + "step": 1701 + }, + { + "epoch": 0.63, + "learning_rate": 2.9389579997536644e-06, + "loss": 0.7292, + "step": 1702 + }, + { + "epoch": 0.63, + "learning_rate": 2.938921049390319e-06, + "loss": 0.624, + "step": 1703 + }, + { + "epoch": 0.63, + "learning_rate": 2.9388840990269737e-06, + "loss": 0.6995, + "step": 1704 + }, + { + "epoch": 0.63, + "learning_rate": 2.9388471486636284e-06, + "loss": 0.6782, + "step": 1705 + }, + { + "epoch": 0.63, + "learning_rate": 2.9388101983002834e-06, + "loss": 0.6298, + "step": 1706 + }, + { + "epoch": 0.63, + "learning_rate": 2.938773247936938e-06, + "loss": 0.6598, + "step": 1707 + }, + { + "epoch": 0.63, + "learning_rate": 2.938736297573593e-06, + "loss": 0.5647, + "step": 1708 + }, + { + "epoch": 0.63, + "learning_rate": 2.9386993472102477e-06, + "loss": 0.5811, + "step": 1709 + }, + { + "epoch": 0.63, + "learning_rate": 2.9386623968469024e-06, + "loss": 0.691, + "step": 1710 + }, + { + "epoch": 0.63, + "learning_rate": 2.938625446483557e-06, + "loss": 0.6507, + "step": 1711 + }, + { + "epoch": 0.63, + "learning_rate": 2.938588496120212e-06, + "loss": 0.6633, + "step": 1712 + }, + { + "epoch": 0.63, + "learning_rate": 2.9385515457568667e-06, + "loss": 0.676, + "step": 1713 + }, + { + "epoch": 0.63, + "learning_rate": 2.9385145953935214e-06, + "loss": 0.5755, + "step": 1714 + }, + { + "epoch": 0.63, + "learning_rate": 2.9384776450301764e-06, + "loss": 0.6897, + "step": 1715 + }, + { + "epoch": 0.63, + "learning_rate": 2.938440694666831e-06, + "loss": 0.5796, + "step": 1716 + }, + { + "epoch": 0.63, + "learning_rate": 2.9384037443034857e-06, + "loss": 0.662, + "step": 1717 + }, + { + "epoch": 0.63, + "learning_rate": 2.9383667939401403e-06, + "loss": 0.6149, + "step": 1718 + }, + { + "epoch": 0.63, + "learning_rate": 2.9383298435767954e-06, + "loss": 0.6755, + "step": 1719 + }, + { + "epoch": 0.64, + "learning_rate": 2.93829289321345e-06, + "loss": 0.6246, + "step": 1720 + }, + { + "epoch": 0.64, + "learning_rate": 2.938255942850105e-06, + "loss": 0.6082, + "step": 1721 + }, + { + "epoch": 0.64, + "learning_rate": 2.9382189924867597e-06, + "loss": 0.6553, + "step": 1722 + }, + { + "epoch": 0.64, + "learning_rate": 2.938182042123414e-06, + "loss": 0.6345, + "step": 1723 + }, + { + "epoch": 0.64, + "learning_rate": 2.938145091760069e-06, + "loss": 0.6701, + "step": 1724 + }, + { + "epoch": 0.64, + "learning_rate": 2.9381081413967236e-06, + "loss": 0.6749, + "step": 1725 + }, + { + "epoch": 0.64, + "learning_rate": 2.9380711910333787e-06, + "loss": 0.664, + "step": 1726 + }, + { + "epoch": 0.64, + "learning_rate": 2.9380342406700333e-06, + "loss": 0.6388, + "step": 1727 + }, + { + "epoch": 0.64, + "learning_rate": 2.9379972903066884e-06, + "loss": 0.6044, + "step": 1728 + }, + { + "epoch": 0.64, + "learning_rate": 2.9379603399433426e-06, + "loss": 0.6571, + "step": 1729 + }, + { + "epoch": 0.64, + "learning_rate": 2.9379233895799977e-06, + "loss": 0.7268, + "step": 1730 + }, + { + "epoch": 0.64, + "learning_rate": 2.9378864392166523e-06, + "loss": 0.6325, + "step": 1731 + }, + { + "epoch": 0.64, + "learning_rate": 2.9378494888533074e-06, + "loss": 0.6975, + "step": 1732 + }, + { + "epoch": 0.64, + "learning_rate": 2.937812538489962e-06, + "loss": 0.612, + "step": 1733 + }, + { + "epoch": 0.64, + "learning_rate": 2.9377755881266166e-06, + "loss": 0.7269, + "step": 1734 + }, + { + "epoch": 0.64, + "learning_rate": 2.9377386377632713e-06, + "loss": 0.6364, + "step": 1735 + }, + { + "epoch": 0.64, + "learning_rate": 2.937701687399926e-06, + "loss": 0.5702, + "step": 1736 + }, + { + "epoch": 0.64, + "learning_rate": 2.937664737036581e-06, + "loss": 0.6371, + "step": 1737 + }, + { + "epoch": 0.64, + "learning_rate": 2.9376277866732356e-06, + "loss": 0.6035, + "step": 1738 + }, + { + "epoch": 0.64, + "learning_rate": 2.9375908363098907e-06, + "loss": 0.6274, + "step": 1739 + }, + { + "epoch": 0.64, + "learning_rate": 2.9375538859465453e-06, + "loss": 0.6476, + "step": 1740 + }, + { + "epoch": 0.64, + "learning_rate": 2.9375169355832e-06, + "loss": 0.6278, + "step": 1741 + }, + { + "epoch": 0.64, + "learning_rate": 2.9374799852198546e-06, + "loss": 0.6607, + "step": 1742 + }, + { + "epoch": 0.64, + "learning_rate": 2.937443034856509e-06, + "loss": 0.6593, + "step": 1743 + }, + { + "epoch": 0.64, + "learning_rate": 2.9374060844931643e-06, + "loss": 0.6754, + "step": 1744 + }, + { + "epoch": 0.64, + "learning_rate": 2.937369134129819e-06, + "loss": 0.6458, + "step": 1745 + }, + { + "epoch": 0.64, + "learning_rate": 2.937332183766474e-06, + "loss": 0.6694, + "step": 1746 + }, + { + "epoch": 0.65, + "learning_rate": 2.9372952334031286e-06, + "loss": 0.6729, + "step": 1747 + }, + { + "epoch": 0.65, + "learning_rate": 2.9372582830397832e-06, + "loss": 0.6607, + "step": 1748 + }, + { + "epoch": 0.65, + "learning_rate": 2.937221332676438e-06, + "loss": 0.6057, + "step": 1749 + }, + { + "epoch": 0.65, + "learning_rate": 2.937184382313093e-06, + "loss": 0.6056, + "step": 1750 + }, + { + "epoch": 0.65, + "learning_rate": 2.9371474319497476e-06, + "loss": 0.6804, + "step": 1751 + }, + { + "epoch": 0.65, + "learning_rate": 2.9371104815864026e-06, + "loss": 0.6864, + "step": 1752 + }, + { + "epoch": 0.65, + "learning_rate": 2.9370735312230573e-06, + "loss": 0.6601, + "step": 1753 + }, + { + "epoch": 0.65, + "learning_rate": 2.937036580859712e-06, + "loss": 0.6613, + "step": 1754 + }, + { + "epoch": 0.65, + "learning_rate": 2.9369996304963665e-06, + "loss": 0.6659, + "step": 1755 + }, + { + "epoch": 0.65, + "learning_rate": 2.936962680133021e-06, + "loss": 0.6644, + "step": 1756 + }, + { + "epoch": 0.65, + "learning_rate": 2.9369257297696762e-06, + "loss": 0.7204, + "step": 1757 + }, + { + "epoch": 0.65, + "learning_rate": 2.936888779406331e-06, + "loss": 0.6246, + "step": 1758 + }, + { + "epoch": 0.65, + "learning_rate": 2.936851829042986e-06, + "loss": 0.756, + "step": 1759 + }, + { + "epoch": 0.65, + "learning_rate": 2.9368148786796406e-06, + "loss": 0.6603, + "step": 1760 + }, + { + "epoch": 0.65, + "learning_rate": 2.936777928316295e-06, + "loss": 0.7158, + "step": 1761 + }, + { + "epoch": 0.65, + "learning_rate": 2.93674097795295e-06, + "loss": 0.6586, + "step": 1762 + }, + { + "epoch": 0.65, + "learning_rate": 2.9367040275896045e-06, + "loss": 0.5985, + "step": 1763 + }, + { + "epoch": 0.65, + "learning_rate": 2.9366670772262595e-06, + "loss": 0.6499, + "step": 1764 + }, + { + "epoch": 0.65, + "learning_rate": 2.936630126862914e-06, + "loss": 0.6086, + "step": 1765 + }, + { + "epoch": 0.65, + "learning_rate": 2.9365931764995692e-06, + "loss": 0.6505, + "step": 1766 + }, + { + "epoch": 0.65, + "learning_rate": 2.936556226136224e-06, + "loss": 0.6723, + "step": 1767 + }, + { + "epoch": 0.65, + "learning_rate": 2.9365192757728785e-06, + "loss": 0.6244, + "step": 1768 + }, + { + "epoch": 0.65, + "learning_rate": 2.936482325409533e-06, + "loss": 0.607, + "step": 1769 + }, + { + "epoch": 0.65, + "learning_rate": 2.936445375046188e-06, + "loss": 0.6385, + "step": 1770 + }, + { + "epoch": 0.65, + "learning_rate": 2.936408424682843e-06, + "loss": 0.6482, + "step": 1771 + }, + { + "epoch": 0.65, + "learning_rate": 2.936371474319498e-06, + "loss": 0.5629, + "step": 1772 + }, + { + "epoch": 0.65, + "learning_rate": 2.9363345239561525e-06, + "loss": 0.6624, + "step": 1773 + }, + { + "epoch": 0.66, + "learning_rate": 2.9362975735928067e-06, + "loss": 0.6977, + "step": 1774 + }, + { + "epoch": 0.66, + "learning_rate": 2.936260623229462e-06, + "loss": 0.6566, + "step": 1775 + }, + { + "epoch": 0.66, + "learning_rate": 2.9362236728661164e-06, + "loss": 0.6369, + "step": 1776 + }, + { + "epoch": 0.66, + "learning_rate": 2.9361867225027715e-06, + "loss": 0.7155, + "step": 1777 + }, + { + "epoch": 0.66, + "learning_rate": 2.936149772139426e-06, + "loss": 0.6607, + "step": 1778 + }, + { + "epoch": 0.66, + "learning_rate": 2.9361128217760808e-06, + "loss": 0.6292, + "step": 1779 + }, + { + "epoch": 0.66, + "learning_rate": 2.9360758714127354e-06, + "loss": 0.5858, + "step": 1780 + }, + { + "epoch": 0.66, + "learning_rate": 2.9360389210493905e-06, + "loss": 0.6299, + "step": 1781 + }, + { + "epoch": 0.66, + "learning_rate": 2.936001970686045e-06, + "loss": 0.6363, + "step": 1782 + }, + { + "epoch": 0.66, + "learning_rate": 2.9359650203226997e-06, + "loss": 0.6741, + "step": 1783 + }, + { + "epoch": 0.66, + "learning_rate": 2.935928069959355e-06, + "loss": 0.6581, + "step": 1784 + }, + { + "epoch": 0.66, + "learning_rate": 2.9358911195960094e-06, + "loss": 0.6431, + "step": 1785 + }, + { + "epoch": 0.66, + "learning_rate": 2.935854169232664e-06, + "loss": 0.6398, + "step": 1786 + }, + { + "epoch": 0.66, + "learning_rate": 2.9358172188693187e-06, + "loss": 0.6573, + "step": 1787 + }, + { + "epoch": 0.66, + "learning_rate": 2.9357802685059738e-06, + "loss": 0.6602, + "step": 1788 + }, + { + "epoch": 0.66, + "learning_rate": 2.9357433181426284e-06, + "loss": 0.6933, + "step": 1789 + }, + { + "epoch": 0.66, + "learning_rate": 2.9357063677792835e-06, + "loss": 0.6601, + "step": 1790 + }, + { + "epoch": 0.66, + "learning_rate": 2.935669417415938e-06, + "loss": 0.6381, + "step": 1791 + }, + { + "epoch": 0.66, + "learning_rate": 2.9356324670525927e-06, + "loss": 0.6453, + "step": 1792 + }, + { + "epoch": 0.66, + "learning_rate": 2.9355955166892474e-06, + "loss": 0.5879, + "step": 1793 + }, + { + "epoch": 0.66, + "learning_rate": 2.935558566325902e-06, + "loss": 0.6063, + "step": 1794 + }, + { + "epoch": 0.66, + "learning_rate": 2.935521615962557e-06, + "loss": 0.6686, + "step": 1795 + }, + { + "epoch": 0.66, + "learning_rate": 2.9354846655992117e-06, + "loss": 0.673, + "step": 1796 + }, + { + "epoch": 0.66, + "learning_rate": 2.9354477152358668e-06, + "loss": 0.6374, + "step": 1797 + }, + { + "epoch": 0.66, + "learning_rate": 2.9354107648725214e-06, + "loss": 0.5917, + "step": 1798 + }, + { + "epoch": 0.66, + "learning_rate": 2.935373814509176e-06, + "loss": 0.6254, + "step": 1799 + }, + { + "epoch": 0.66, + "learning_rate": 2.9353368641458307e-06, + "loss": 0.6345, + "step": 1800 + }, + { + "epoch": 0.67, + "learning_rate": 2.9352999137824857e-06, + "loss": 0.6886, + "step": 1801 + }, + { + "epoch": 0.67, + "learning_rate": 2.9352629634191404e-06, + "loss": 0.7181, + "step": 1802 + }, + { + "epoch": 0.67, + "learning_rate": 2.935226013055795e-06, + "loss": 0.6888, + "step": 1803 + }, + { + "epoch": 0.67, + "learning_rate": 2.93518906269245e-06, + "loss": 0.6828, + "step": 1804 + }, + { + "epoch": 0.67, + "learning_rate": 2.9351521123291047e-06, + "loss": 0.6416, + "step": 1805 + }, + { + "epoch": 0.67, + "learning_rate": 2.9351151619657593e-06, + "loss": 0.7222, + "step": 1806 + }, + { + "epoch": 0.67, + "learning_rate": 2.935078211602414e-06, + "loss": 0.6953, + "step": 1807 + }, + { + "epoch": 0.67, + "learning_rate": 2.935041261239069e-06, + "loss": 0.5766, + "step": 1808 + }, + { + "epoch": 0.67, + "learning_rate": 2.9350043108757237e-06, + "loss": 0.6752, + "step": 1809 + }, + { + "epoch": 0.67, + "learning_rate": 2.9349673605123787e-06, + "loss": 0.6241, + "step": 1810 + }, + { + "epoch": 0.67, + "learning_rate": 2.9349304101490334e-06, + "loss": 0.5864, + "step": 1811 + }, + { + "epoch": 0.67, + "learning_rate": 2.934893459785688e-06, + "loss": 0.6368, + "step": 1812 + }, + { + "epoch": 0.67, + "learning_rate": 2.9348565094223426e-06, + "loss": 0.5978, + "step": 1813 + }, + { + "epoch": 0.67, + "learning_rate": 2.9348195590589973e-06, + "loss": 0.644, + "step": 1814 + }, + { + "epoch": 0.67, + "learning_rate": 2.9347826086956523e-06, + "loss": 0.6501, + "step": 1815 + }, + { + "epoch": 0.67, + "learning_rate": 2.934745658332307e-06, + "loss": 0.6889, + "step": 1816 + }, + { + "epoch": 0.67, + "learning_rate": 2.934708707968962e-06, + "loss": 0.5821, + "step": 1817 + }, + { + "epoch": 0.67, + "learning_rate": 2.9346717576056162e-06, + "loss": 0.6276, + "step": 1818 + }, + { + "epoch": 0.67, + "learning_rate": 2.9346348072422713e-06, + "loss": 0.6846, + "step": 1819 + }, + { + "epoch": 0.67, + "learning_rate": 2.934597856878926e-06, + "loss": 0.6973, + "step": 1820 + }, + { + "epoch": 0.67, + "learning_rate": 2.934560906515581e-06, + "loss": 0.5994, + "step": 1821 + }, + { + "epoch": 0.67, + "learning_rate": 2.9345239561522356e-06, + "loss": 0.6607, + "step": 1822 + }, + { + "epoch": 0.67, + "learning_rate": 2.9344870057888903e-06, + "loss": 0.586, + "step": 1823 + }, + { + "epoch": 0.67, + "learning_rate": 2.934450055425545e-06, + "loss": 0.6965, + "step": 1824 + }, + { + "epoch": 0.67, + "learning_rate": 2.9344131050621996e-06, + "loss": 0.6241, + "step": 1825 + }, + { + "epoch": 0.67, + "learning_rate": 2.9343761546988546e-06, + "loss": 0.6341, + "step": 1826 + }, + { + "epoch": 0.67, + "learning_rate": 2.9343392043355092e-06, + "loss": 0.629, + "step": 1827 + }, + { + "epoch": 0.68, + "learning_rate": 2.9343022539721643e-06, + "loss": 0.601, + "step": 1828 + }, + { + "epoch": 0.68, + "learning_rate": 2.934265303608819e-06, + "loss": 0.6028, + "step": 1829 + }, + { + "epoch": 0.68, + "learning_rate": 2.9342283532454736e-06, + "loss": 0.6417, + "step": 1830 + }, + { + "epoch": 0.68, + "learning_rate": 2.9341914028821282e-06, + "loss": 0.6787, + "step": 1831 + }, + { + "epoch": 0.68, + "learning_rate": 2.9341544525187833e-06, + "loss": 0.6323, + "step": 1832 + }, + { + "epoch": 0.68, + "learning_rate": 2.934117502155438e-06, + "loss": 0.6512, + "step": 1833 + }, + { + "epoch": 0.68, + "learning_rate": 2.9340805517920926e-06, + "loss": 0.6481, + "step": 1834 + }, + { + "epoch": 0.68, + "learning_rate": 2.9340436014287476e-06, + "loss": 0.6482, + "step": 1835 + }, + { + "epoch": 0.68, + "learning_rate": 2.9340066510654022e-06, + "loss": 0.5919, + "step": 1836 + }, + { + "epoch": 0.68, + "learning_rate": 2.933969700702057e-06, + "loss": 0.6502, + "step": 1837 + }, + { + "epoch": 0.68, + "learning_rate": 2.9339327503387115e-06, + "loss": 0.6243, + "step": 1838 + }, + { + "epoch": 0.68, + "learning_rate": 2.9338957999753666e-06, + "loss": 0.5604, + "step": 1839 + }, + { + "epoch": 0.68, + "learning_rate": 2.9338588496120212e-06, + "loss": 0.654, + "step": 1840 + }, + { + "epoch": 0.68, + "learning_rate": 2.9338218992486763e-06, + "loss": 0.67, + "step": 1841 + }, + { + "epoch": 0.68, + "learning_rate": 2.933784948885331e-06, + "loss": 0.6773, + "step": 1842 + }, + { + "epoch": 0.68, + "learning_rate": 2.9337479985219855e-06, + "loss": 0.6722, + "step": 1843 + }, + { + "epoch": 0.68, + "learning_rate": 2.93371104815864e-06, + "loss": 0.6033, + "step": 1844 + }, + { + "epoch": 0.68, + "learning_rate": 2.933674097795295e-06, + "loss": 0.598, + "step": 1845 + }, + { + "epoch": 0.68, + "learning_rate": 2.93363714743195e-06, + "loss": 0.651, + "step": 1846 + }, + { + "epoch": 0.68, + "learning_rate": 2.9336001970686045e-06, + "loss": 0.619, + "step": 1847 + }, + { + "epoch": 0.68, + "learning_rate": 2.9335632467052596e-06, + "loss": 0.634, + "step": 1848 + }, + { + "epoch": 0.68, + "learning_rate": 2.9335262963419142e-06, + "loss": 0.5992, + "step": 1849 + }, + { + "epoch": 0.68, + "learning_rate": 2.933489345978569e-06, + "loss": 0.6174, + "step": 1850 + }, + { + "epoch": 0.68, + "learning_rate": 2.9334523956152235e-06, + "loss": 0.641, + "step": 1851 + }, + { + "epoch": 0.68, + "learning_rate": 2.9334154452518785e-06, + "loss": 0.6235, + "step": 1852 + }, + { + "epoch": 0.68, + "learning_rate": 2.933378494888533e-06, + "loss": 0.6499, + "step": 1853 + }, + { + "epoch": 0.68, + "learning_rate": 2.933341544525188e-06, + "loss": 0.6536, + "step": 1854 + }, + { + "epoch": 0.69, + "learning_rate": 2.933304594161843e-06, + "loss": 0.6673, + "step": 1855 + }, + { + "epoch": 0.69, + "learning_rate": 2.9332676437984975e-06, + "loss": 0.6588, + "step": 1856 + }, + { + "epoch": 0.69, + "learning_rate": 2.933230693435152e-06, + "loss": 0.6281, + "step": 1857 + }, + { + "epoch": 0.69, + "learning_rate": 2.933193743071807e-06, + "loss": 0.6202, + "step": 1858 + }, + { + "epoch": 0.69, + "learning_rate": 2.933156792708462e-06, + "loss": 0.5964, + "step": 1859 + }, + { + "epoch": 0.69, + "learning_rate": 2.9331198423451165e-06, + "loss": 0.5873, + "step": 1860 + }, + { + "epoch": 0.69, + "learning_rate": 2.9330828919817715e-06, + "loss": 0.6153, + "step": 1861 + }, + { + "epoch": 0.69, + "learning_rate": 2.933045941618426e-06, + "loss": 0.6834, + "step": 1862 + }, + { + "epoch": 0.69, + "learning_rate": 2.9330089912550804e-06, + "loss": 0.5974, + "step": 1863 + }, + { + "epoch": 0.69, + "learning_rate": 2.9329720408917355e-06, + "loss": 0.6095, + "step": 1864 + }, + { + "epoch": 0.69, + "learning_rate": 2.93293509052839e-06, + "loss": 0.5942, + "step": 1865 + }, + { + "epoch": 0.69, + "learning_rate": 2.932898140165045e-06, + "loss": 0.7187, + "step": 1866 + }, + { + "epoch": 0.69, + "learning_rate": 2.9328611898016998e-06, + "loss": 0.671, + "step": 1867 + }, + { + "epoch": 0.69, + "learning_rate": 2.932824239438355e-06, + "loss": 0.6305, + "step": 1868 + }, + { + "epoch": 0.69, + "learning_rate": 2.932787289075009e-06, + "loss": 0.5916, + "step": 1869 + }, + { + "epoch": 0.69, + "learning_rate": 2.932750338711664e-06, + "loss": 0.5841, + "step": 1870 + }, + { + "epoch": 0.69, + "learning_rate": 2.9327133883483188e-06, + "loss": 0.7227, + "step": 1871 + }, + { + "epoch": 0.69, + "learning_rate": 2.932676437984974e-06, + "loss": 0.6016, + "step": 1872 + }, + { + "epoch": 0.69, + "learning_rate": 2.9326394876216285e-06, + "loss": 0.61, + "step": 1873 + }, + { + "epoch": 0.69, + "learning_rate": 2.932602537258283e-06, + "loss": 0.664, + "step": 1874 + }, + { + "epoch": 0.69, + "learning_rate": 2.9325655868949377e-06, + "loss": 0.6974, + "step": 1875 + }, + { + "epoch": 0.69, + "learning_rate": 2.9325286365315924e-06, + "loss": 0.6296, + "step": 1876 + }, + { + "epoch": 0.69, + "learning_rate": 2.9324916861682474e-06, + "loss": 0.6786, + "step": 1877 + }, + { + "epoch": 0.69, + "learning_rate": 2.932454735804902e-06, + "loss": 0.6893, + "step": 1878 + }, + { + "epoch": 0.69, + "learning_rate": 2.932417785441557e-06, + "loss": 0.638, + "step": 1879 + }, + { + "epoch": 0.69, + "learning_rate": 2.9323808350782118e-06, + "loss": 0.6723, + "step": 1880 + }, + { + "epoch": 0.69, + "learning_rate": 2.9323438847148664e-06, + "loss": 0.5908, + "step": 1881 + }, + { + "epoch": 0.69, + "learning_rate": 2.932306934351521e-06, + "loss": 0.6087, + "step": 1882 + }, + { + "epoch": 0.7, + "learning_rate": 2.9322699839881757e-06, + "loss": 0.6403, + "step": 1883 + }, + { + "epoch": 0.7, + "learning_rate": 2.9322330336248307e-06, + "loss": 0.6787, + "step": 1884 + }, + { + "epoch": 0.7, + "learning_rate": 2.9321960832614854e-06, + "loss": 0.6165, + "step": 1885 + }, + { + "epoch": 0.7, + "learning_rate": 2.9321591328981404e-06, + "loss": 0.6859, + "step": 1886 + }, + { + "epoch": 0.7, + "learning_rate": 2.932122182534795e-06, + "loss": 0.5903, + "step": 1887 + }, + { + "epoch": 0.7, + "learning_rate": 2.9320852321714497e-06, + "loss": 0.6592, + "step": 1888 + }, + { + "epoch": 0.7, + "learning_rate": 2.9320482818081043e-06, + "loss": 0.658, + "step": 1889 + }, + { + "epoch": 0.7, + "learning_rate": 2.9320113314447594e-06, + "loss": 0.6341, + "step": 1890 + }, + { + "epoch": 0.7, + "learning_rate": 2.931974381081414e-06, + "loss": 0.6742, + "step": 1891 + }, + { + "epoch": 0.7, + "learning_rate": 2.931937430718069e-06, + "loss": 0.6147, + "step": 1892 + }, + { + "epoch": 0.7, + "learning_rate": 2.9319004803547237e-06, + "loss": 0.6962, + "step": 1893 + }, + { + "epoch": 0.7, + "learning_rate": 2.9318635299913784e-06, + "loss": 0.5974, + "step": 1894 + }, + { + "epoch": 0.7, + "learning_rate": 2.931826579628033e-06, + "loss": 0.6288, + "step": 1895 + }, + { + "epoch": 0.7, + "learning_rate": 2.9317896292646876e-06, + "loss": 0.6667, + "step": 1896 + }, + { + "epoch": 0.7, + "learning_rate": 2.9317526789013427e-06, + "loss": 0.5754, + "step": 1897 + }, + { + "epoch": 0.7, + "learning_rate": 2.9317157285379973e-06, + "loss": 0.6078, + "step": 1898 + }, + { + "epoch": 0.7, + "learning_rate": 2.9316787781746524e-06, + "loss": 0.6058, + "step": 1899 + }, + { + "epoch": 0.7, + "learning_rate": 2.931641827811307e-06, + "loss": 0.6332, + "step": 1900 + }, + { + "epoch": 0.7, + "learning_rate": 2.9316048774479617e-06, + "loss": 0.6071, + "step": 1901 + }, + { + "epoch": 0.7, + "learning_rate": 2.9315679270846163e-06, + "loss": 0.6075, + "step": 1902 + }, + { + "epoch": 0.7, + "learning_rate": 2.931530976721271e-06, + "loss": 0.6574, + "step": 1903 + }, + { + "epoch": 0.7, + "learning_rate": 2.931494026357926e-06, + "loss": 0.6069, + "step": 1904 + }, + { + "epoch": 0.7, + "learning_rate": 2.9314570759945806e-06, + "loss": 0.7198, + "step": 1905 + }, + { + "epoch": 0.7, + "learning_rate": 2.9314201256312357e-06, + "loss": 0.652, + "step": 1906 + }, + { + "epoch": 0.7, + "learning_rate": 2.9313831752678903e-06, + "loss": 0.641, + "step": 1907 + }, + { + "epoch": 0.7, + "learning_rate": 2.931346224904545e-06, + "loss": 0.6151, + "step": 1908 + }, + { + "epoch": 0.7, + "learning_rate": 2.9313092745411996e-06, + "loss": 0.6954, + "step": 1909 + }, + { + "epoch": 0.71, + "learning_rate": 2.9312723241778547e-06, + "loss": 0.6324, + "step": 1910 + }, + { + "epoch": 0.71, + "learning_rate": 2.9312353738145093e-06, + "loss": 0.654, + "step": 1911 + }, + { + "epoch": 0.71, + "learning_rate": 2.931198423451164e-06, + "loss": 0.646, + "step": 1912 + }, + { + "epoch": 0.71, + "learning_rate": 2.931161473087819e-06, + "loss": 0.7188, + "step": 1913 + }, + { + "epoch": 0.71, + "learning_rate": 2.931124522724473e-06, + "loss": 0.5574, + "step": 1914 + }, + { + "epoch": 0.71, + "learning_rate": 2.9310875723611283e-06, + "loss": 0.6081, + "step": 1915 + }, + { + "epoch": 0.71, + "learning_rate": 2.931050621997783e-06, + "loss": 0.6799, + "step": 1916 + }, + { + "epoch": 0.71, + "learning_rate": 2.931013671634438e-06, + "loss": 0.6571, + "step": 1917 + }, + { + "epoch": 0.71, + "learning_rate": 2.9309767212710926e-06, + "loss": 0.6374, + "step": 1918 + }, + { + "epoch": 0.71, + "learning_rate": 2.9309397709077477e-06, + "loss": 0.6723, + "step": 1919 + }, + { + "epoch": 0.71, + "learning_rate": 2.930902820544402e-06, + "loss": 0.6147, + "step": 1920 + }, + { + "epoch": 0.71, + "learning_rate": 2.930865870181057e-06, + "loss": 0.6089, + "step": 1921 + }, + { + "epoch": 0.71, + "learning_rate": 2.9308289198177116e-06, + "loss": 0.6779, + "step": 1922 + }, + { + "epoch": 0.71, + "learning_rate": 2.930791969454366e-06, + "loss": 0.7265, + "step": 1923 + }, + { + "epoch": 0.71, + "learning_rate": 2.9307550190910213e-06, + "loss": 0.7013, + "step": 1924 + }, + { + "epoch": 0.71, + "learning_rate": 2.930718068727676e-06, + "loss": 0.5869, + "step": 1925 + }, + { + "epoch": 0.71, + "learning_rate": 2.9306811183643305e-06, + "loss": 0.6009, + "step": 1926 + }, + { + "epoch": 0.71, + "learning_rate": 2.930644168000985e-06, + "loss": 0.6783, + "step": 1927 + }, + { + "epoch": 0.71, + "learning_rate": 2.9306072176376402e-06, + "loss": 0.7128, + "step": 1928 + }, + { + "epoch": 0.71, + "learning_rate": 2.930570267274295e-06, + "loss": 0.6217, + "step": 1929 + }, + { + "epoch": 0.71, + "learning_rate": 2.93053331691095e-06, + "loss": 0.6061, + "step": 1930 + }, + { + "epoch": 0.71, + "learning_rate": 2.9304963665476046e-06, + "loss": 0.5871, + "step": 1931 + }, + { + "epoch": 0.71, + "learning_rate": 2.930459416184259e-06, + "loss": 0.6729, + "step": 1932 + }, + { + "epoch": 0.71, + "learning_rate": 2.930422465820914e-06, + "loss": 0.6823, + "step": 1933 + }, + { + "epoch": 0.71, + "learning_rate": 2.9303855154575685e-06, + "loss": 0.6322, + "step": 1934 + }, + { + "epoch": 0.71, + "learning_rate": 2.9303485650942235e-06, + "loss": 0.5788, + "step": 1935 + }, + { + "epoch": 0.71, + "learning_rate": 2.930311614730878e-06, + "loss": 0.6766, + "step": 1936 + }, + { + "epoch": 0.72, + "learning_rate": 2.9302746643675332e-06, + "loss": 0.641, + "step": 1937 + }, + { + "epoch": 0.72, + "learning_rate": 2.930237714004188e-06, + "loss": 0.5849, + "step": 1938 + }, + { + "epoch": 0.72, + "learning_rate": 2.9302007636408425e-06, + "loss": 0.5893, + "step": 1939 + }, + { + "epoch": 0.72, + "learning_rate": 2.930163813277497e-06, + "loss": 0.6209, + "step": 1940 + }, + { + "epoch": 0.72, + "learning_rate": 2.930126862914152e-06, + "loss": 0.6196, + "step": 1941 + }, + { + "epoch": 0.72, + "learning_rate": 2.930089912550807e-06, + "loss": 0.6669, + "step": 1942 + }, + { + "epoch": 0.72, + "learning_rate": 2.9300529621874615e-06, + "loss": 0.6493, + "step": 1943 + }, + { + "epoch": 0.72, + "learning_rate": 2.9300160118241165e-06, + "loss": 0.5686, + "step": 1944 + }, + { + "epoch": 0.72, + "learning_rate": 2.929979061460771e-06, + "loss": 0.6651, + "step": 1945 + }, + { + "epoch": 0.72, + "learning_rate": 2.929942111097426e-06, + "loss": 0.5953, + "step": 1946 + }, + { + "epoch": 0.72, + "learning_rate": 2.9299051607340804e-06, + "loss": 0.6427, + "step": 1947 + }, + { + "epoch": 0.72, + "learning_rate": 2.9298682103707355e-06, + "loss": 0.5149, + "step": 1948 + }, + { + "epoch": 0.72, + "learning_rate": 2.92983126000739e-06, + "loss": 0.7078, + "step": 1949 + }, + { + "epoch": 0.72, + "learning_rate": 2.929794309644045e-06, + "loss": 0.6399, + "step": 1950 + }, + { + "epoch": 0.72, + "learning_rate": 2.9297573592807e-06, + "loss": 0.6905, + "step": 1951 + }, + { + "epoch": 0.72, + "learning_rate": 2.9297204089173545e-06, + "loss": 0.5688, + "step": 1952 + }, + { + "epoch": 0.72, + "learning_rate": 2.929683458554009e-06, + "loss": 0.6453, + "step": 1953 + }, + { + "epoch": 0.72, + "learning_rate": 2.9296465081906637e-06, + "loss": 0.6785, + "step": 1954 + }, + { + "epoch": 0.72, + "learning_rate": 2.929609557827319e-06, + "loss": 0.6112, + "step": 1955 + }, + { + "epoch": 0.72, + "learning_rate": 2.9295726074639734e-06, + "loss": 0.7018, + "step": 1956 + }, + { + "epoch": 0.72, + "learning_rate": 2.9295356571006285e-06, + "loss": 0.6017, + "step": 1957 + }, + { + "epoch": 0.72, + "learning_rate": 2.929498706737283e-06, + "loss": 0.6672, + "step": 1958 + }, + { + "epoch": 0.72, + "learning_rate": 2.9294617563739378e-06, + "loss": 0.6607, + "step": 1959 + }, + { + "epoch": 0.72, + "learning_rate": 2.9294248060105924e-06, + "loss": 0.6859, + "step": 1960 + }, + { + "epoch": 0.72, + "learning_rate": 2.9293878556472475e-06, + "loss": 0.7195, + "step": 1961 + }, + { + "epoch": 0.72, + "learning_rate": 2.929350905283902e-06, + "loss": 0.6618, + "step": 1962 + }, + { + "epoch": 0.72, + "learning_rate": 2.9293139549205567e-06, + "loss": 0.7037, + "step": 1963 + }, + { + "epoch": 0.73, + "learning_rate": 2.9292770045572114e-06, + "loss": 0.6531, + "step": 1964 + }, + { + "epoch": 0.73, + "learning_rate": 2.929240054193866e-06, + "loss": 0.6307, + "step": 1965 + }, + { + "epoch": 0.73, + "learning_rate": 2.929203103830521e-06, + "loss": 0.6846, + "step": 1966 + }, + { + "epoch": 0.73, + "learning_rate": 2.9291661534671757e-06, + "loss": 0.6658, + "step": 1967 + }, + { + "epoch": 0.73, + "learning_rate": 2.9291292031038308e-06, + "loss": 0.654, + "step": 1968 + }, + { + "epoch": 0.73, + "learning_rate": 2.9290922527404854e-06, + "loss": 0.6327, + "step": 1969 + }, + { + "epoch": 0.73, + "learning_rate": 2.92905530237714e-06, + "loss": 0.6292, + "step": 1970 + }, + { + "epoch": 0.73, + "learning_rate": 2.9290183520137947e-06, + "loss": 0.6169, + "step": 1971 + }, + { + "epoch": 0.73, + "learning_rate": 2.9289814016504493e-06, + "loss": 0.6272, + "step": 1972 + }, + { + "epoch": 0.73, + "learning_rate": 2.9289444512871044e-06, + "loss": 0.6783, + "step": 1973 + }, + { + "epoch": 0.73, + "learning_rate": 2.928907500923759e-06, + "loss": 0.6251, + "step": 1974 + }, + { + "epoch": 0.73, + "learning_rate": 2.928870550560414e-06, + "loss": 0.6779, + "step": 1975 + }, + { + "epoch": 0.73, + "learning_rate": 2.9288336001970687e-06, + "loss": 0.6907, + "step": 1976 + }, + { + "epoch": 0.73, + "learning_rate": 2.9287966498337233e-06, + "loss": 0.661, + "step": 1977 + }, + { + "epoch": 0.73, + "learning_rate": 2.928759699470378e-06, + "loss": 0.6694, + "step": 1978 + }, + { + "epoch": 0.73, + "learning_rate": 2.928722749107033e-06, + "loss": 0.6582, + "step": 1979 + }, + { + "epoch": 0.73, + "learning_rate": 2.9286857987436877e-06, + "loss": 0.6995, + "step": 1980 + }, + { + "epoch": 0.73, + "learning_rate": 2.9286488483803427e-06, + "loss": 0.5579, + "step": 1981 + }, + { + "epoch": 0.73, + "learning_rate": 2.9286118980169974e-06, + "loss": 0.6476, + "step": 1982 + }, + { + "epoch": 0.73, + "learning_rate": 2.928574947653652e-06, + "loss": 0.6416, + "step": 1983 + }, + { + "epoch": 0.73, + "learning_rate": 2.9285379972903067e-06, + "loss": 0.6155, + "step": 1984 + }, + { + "epoch": 0.73, + "learning_rate": 2.9285010469269613e-06, + "loss": 0.6421, + "step": 1985 + }, + { + "epoch": 0.73, + "learning_rate": 2.9284640965636163e-06, + "loss": 0.608, + "step": 1986 + }, + { + "epoch": 0.73, + "learning_rate": 2.928427146200271e-06, + "loss": 0.6592, + "step": 1987 + }, + { + "epoch": 0.73, + "learning_rate": 2.928390195836926e-06, + "loss": 0.6625, + "step": 1988 + }, + { + "epoch": 0.73, + "learning_rate": 2.9283532454735807e-06, + "loss": 0.6553, + "step": 1989 + }, + { + "epoch": 0.73, + "learning_rate": 2.9283162951102353e-06, + "loss": 0.629, + "step": 1990 + }, + { + "epoch": 0.74, + "learning_rate": 2.92827934474689e-06, + "loss": 0.6571, + "step": 1991 + }, + { + "epoch": 0.74, + "learning_rate": 2.9282423943835446e-06, + "loss": 0.6561, + "step": 1992 + }, + { + "epoch": 0.74, + "learning_rate": 2.9282054440201996e-06, + "loss": 0.6924, + "step": 1993 + }, + { + "epoch": 0.74, + "learning_rate": 2.9281684936568543e-06, + "loss": 0.6594, + "step": 1994 + }, + { + "epoch": 0.74, + "learning_rate": 2.9281315432935093e-06, + "loss": 0.6225, + "step": 1995 + }, + { + "epoch": 0.74, + "learning_rate": 2.928094592930164e-06, + "loss": 0.6559, + "step": 1996 + }, + { + "epoch": 0.74, + "learning_rate": 2.9280576425668186e-06, + "loss": 0.6242, + "step": 1997 + }, + { + "epoch": 0.74, + "learning_rate": 2.9280206922034733e-06, + "loss": 0.6428, + "step": 1998 + }, + { + "epoch": 0.74, + "learning_rate": 2.9279837418401283e-06, + "loss": 0.6118, + "step": 1999 + }, + { + "epoch": 0.74, + "learning_rate": 2.927946791476783e-06, + "loss": 0.6476, + "step": 2000 + }, + { + "epoch": 0.74, + "learning_rate": 2.927909841113438e-06, + "loss": 0.614, + "step": 2001 + }, + { + "epoch": 0.74, + "learning_rate": 2.9278728907500926e-06, + "loss": 0.6681, + "step": 2002 + }, + { + "epoch": 0.74, + "learning_rate": 2.927835940386747e-06, + "loss": 0.6138, + "step": 2003 + }, + { + "epoch": 0.74, + "learning_rate": 2.927798990023402e-06, + "loss": 0.6332, + "step": 2004 + }, + { + "epoch": 0.74, + "learning_rate": 2.9277620396600566e-06, + "loss": 0.6967, + "step": 2005 + }, + { + "epoch": 0.74, + "learning_rate": 2.9277250892967116e-06, + "loss": 0.6169, + "step": 2006 + }, + { + "epoch": 0.74, + "learning_rate": 2.9276881389333663e-06, + "loss": 0.6265, + "step": 2007 + }, + { + "epoch": 0.74, + "learning_rate": 2.9276511885700213e-06, + "loss": 0.6127, + "step": 2008 + }, + { + "epoch": 0.74, + "learning_rate": 2.9276142382066755e-06, + "loss": 0.6171, + "step": 2009 + }, + { + "epoch": 0.74, + "learning_rate": 2.9275772878433306e-06, + "loss": 0.6713, + "step": 2010 + }, + { + "epoch": 0.74, + "learning_rate": 2.9275403374799852e-06, + "loss": 0.6719, + "step": 2011 + }, + { + "epoch": 0.74, + "learning_rate": 2.92750338711664e-06, + "loss": 0.6716, + "step": 2012 + }, + { + "epoch": 0.74, + "learning_rate": 2.927466436753295e-06, + "loss": 0.6615, + "step": 2013 + }, + { + "epoch": 0.74, + "learning_rate": 2.9274294863899496e-06, + "loss": 0.7017, + "step": 2014 + }, + { + "epoch": 0.74, + "learning_rate": 2.927392536026604e-06, + "loss": 0.6296, + "step": 2015 + }, + { + "epoch": 0.74, + "learning_rate": 2.927355585663259e-06, + "loss": 0.6197, + "step": 2016 + }, + { + "epoch": 0.74, + "learning_rate": 2.927318635299914e-06, + "loss": 0.6324, + "step": 2017 + }, + { + "epoch": 0.75, + "learning_rate": 2.9272816849365685e-06, + "loss": 0.597, + "step": 2018 + }, + { + "epoch": 0.75, + "learning_rate": 2.9272447345732236e-06, + "loss": 0.6003, + "step": 2019 + }, + { + "epoch": 0.75, + "learning_rate": 2.9272077842098782e-06, + "loss": 0.5971, + "step": 2020 + }, + { + "epoch": 0.75, + "learning_rate": 2.927170833846533e-06, + "loss": 0.6883, + "step": 2021 + }, + { + "epoch": 0.75, + "learning_rate": 2.9271338834831875e-06, + "loss": 0.6773, + "step": 2022 + }, + { + "epoch": 0.75, + "learning_rate": 2.927096933119842e-06, + "loss": 0.6693, + "step": 2023 + }, + { + "epoch": 0.75, + "learning_rate": 2.927059982756497e-06, + "loss": 0.6315, + "step": 2024 + }, + { + "epoch": 0.75, + "learning_rate": 2.927023032393152e-06, + "loss": 0.6784, + "step": 2025 + }, + { + "epoch": 0.75, + "learning_rate": 2.926986082029807e-06, + "loss": 0.6452, + "step": 2026 + }, + { + "epoch": 0.75, + "learning_rate": 2.9269491316664615e-06, + "loss": 0.6615, + "step": 2027 + }, + { + "epoch": 0.75, + "learning_rate": 2.926912181303116e-06, + "loss": 0.6854, + "step": 2028 + }, + { + "epoch": 0.75, + "learning_rate": 2.926875230939771e-06, + "loss": 0.6303, + "step": 2029 + }, + { + "epoch": 0.75, + "learning_rate": 2.926838280576426e-06, + "loss": 0.6701, + "step": 2030 + }, + { + "epoch": 0.75, + "learning_rate": 2.9268013302130805e-06, + "loss": 0.6741, + "step": 2031 + }, + { + "epoch": 0.75, + "learning_rate": 2.926764379849735e-06, + "loss": 0.6447, + "step": 2032 + }, + { + "epoch": 0.75, + "learning_rate": 2.92672742948639e-06, + "loss": 0.6843, + "step": 2033 + }, + { + "epoch": 0.75, + "learning_rate": 2.926690479123045e-06, + "loss": 0.7148, + "step": 2034 + }, + { + "epoch": 0.75, + "learning_rate": 2.9266535287596995e-06, + "loss": 0.6216, + "step": 2035 + }, + { + "epoch": 0.75, + "learning_rate": 2.926616578396354e-06, + "loss": 0.6702, + "step": 2036 + }, + { + "epoch": 0.75, + "learning_rate": 2.926579628033009e-06, + "loss": 0.6337, + "step": 2037 + }, + { + "epoch": 0.75, + "learning_rate": 2.926542677669664e-06, + "loss": 0.7097, + "step": 2038 + }, + { + "epoch": 0.75, + "learning_rate": 2.926505727306319e-06, + "loss": 0.6055, + "step": 2039 + }, + { + "epoch": 0.75, + "learning_rate": 2.9264687769429735e-06, + "loss": 0.6684, + "step": 2040 + }, + { + "epoch": 0.75, + "learning_rate": 2.926431826579628e-06, + "loss": 0.6508, + "step": 2041 + }, + { + "epoch": 0.75, + "learning_rate": 2.9263948762162828e-06, + "loss": 0.6508, + "step": 2042 + }, + { + "epoch": 0.75, + "learning_rate": 2.9263579258529374e-06, + "loss": 0.6121, + "step": 2043 + }, + { + "epoch": 0.75, + "learning_rate": 2.9263209754895925e-06, + "loss": 0.6439, + "step": 2044 + }, + { + "epoch": 0.76, + "learning_rate": 2.926284025126247e-06, + "loss": 0.6066, + "step": 2045 + }, + { + "epoch": 0.76, + "learning_rate": 2.926247074762902e-06, + "loss": 0.5991, + "step": 2046 + }, + { + "epoch": 0.76, + "learning_rate": 2.926210124399557e-06, + "loss": 0.65, + "step": 2047 + }, + { + "epoch": 0.76, + "learning_rate": 2.9261731740362114e-06, + "loss": 0.6472, + "step": 2048 + }, + { + "epoch": 0.76, + "learning_rate": 2.926136223672866e-06, + "loss": 0.6347, + "step": 2049 + }, + { + "epoch": 0.76, + "learning_rate": 2.926099273309521e-06, + "loss": 0.6636, + "step": 2050 + }, + { + "epoch": 0.76, + "learning_rate": 2.9260623229461758e-06, + "loss": 0.6749, + "step": 2051 + }, + { + "epoch": 0.76, + "learning_rate": 2.9260253725828304e-06, + "loss": 0.7068, + "step": 2052 + }, + { + "epoch": 0.76, + "learning_rate": 2.9259884222194855e-06, + "loss": 0.6032, + "step": 2053 + }, + { + "epoch": 0.76, + "learning_rate": 2.9259514718561397e-06, + "loss": 0.6763, + "step": 2054 + }, + { + "epoch": 0.76, + "learning_rate": 2.9259145214927947e-06, + "loss": 0.657, + "step": 2055 + }, + { + "epoch": 0.76, + "learning_rate": 2.9258775711294494e-06, + "loss": 0.603, + "step": 2056 + }, + { + "epoch": 0.76, + "learning_rate": 2.9258406207661044e-06, + "loss": 0.5859, + "step": 2057 + }, + { + "epoch": 0.76, + "learning_rate": 2.925803670402759e-06, + "loss": 0.5858, + "step": 2058 + }, + { + "epoch": 0.76, + "learning_rate": 2.925766720039414e-06, + "loss": 0.6538, + "step": 2059 + }, + { + "epoch": 0.76, + "learning_rate": 2.9257297696760683e-06, + "loss": 0.7012, + "step": 2060 + }, + { + "epoch": 0.76, + "learning_rate": 2.9256928193127234e-06, + "loss": 0.6762, + "step": 2061 + }, + { + "epoch": 0.76, + "learning_rate": 2.925655868949378e-06, + "loss": 0.6923, + "step": 2062 + }, + { + "epoch": 0.76, + "learning_rate": 2.9256189185860327e-06, + "loss": 0.6446, + "step": 2063 + }, + { + "epoch": 0.76, + "learning_rate": 2.9255819682226877e-06, + "loss": 0.6834, + "step": 2064 + }, + { + "epoch": 0.76, + "learning_rate": 2.9255450178593424e-06, + "loss": 0.6236, + "step": 2065 + }, + { + "epoch": 0.76, + "learning_rate": 2.925508067495997e-06, + "loss": 0.6444, + "step": 2066 + }, + { + "epoch": 0.76, + "learning_rate": 2.9254711171326516e-06, + "loss": 0.6087, + "step": 2067 + }, + { + "epoch": 0.76, + "learning_rate": 2.9254341667693067e-06, + "loss": 0.6286, + "step": 2068 + }, + { + "epoch": 0.76, + "learning_rate": 2.9253972164059613e-06, + "loss": 0.6678, + "step": 2069 + }, + { + "epoch": 0.76, + "learning_rate": 2.9253602660426164e-06, + "loss": 0.6367, + "step": 2070 + }, + { + "epoch": 0.76, + "learning_rate": 2.925323315679271e-06, + "loss": 0.6493, + "step": 2071 + }, + { + "epoch": 0.77, + "learning_rate": 2.9252863653159257e-06, + "loss": 0.6693, + "step": 2072 + }, + { + "epoch": 0.77, + "learning_rate": 2.9252494149525803e-06, + "loss": 0.6793, + "step": 2073 + }, + { + "epoch": 0.77, + "learning_rate": 2.925212464589235e-06, + "loss": 0.7172, + "step": 2074 + }, + { + "epoch": 0.77, + "learning_rate": 2.92517551422589e-06, + "loss": 0.6316, + "step": 2075 + }, + { + "epoch": 0.77, + "learning_rate": 2.9251385638625446e-06, + "loss": 0.5999, + "step": 2076 + }, + { + "epoch": 0.77, + "learning_rate": 2.9251016134991997e-06, + "loss": 0.6464, + "step": 2077 + }, + { + "epoch": 0.77, + "learning_rate": 2.9250646631358543e-06, + "loss": 0.7272, + "step": 2078 + }, + { + "epoch": 0.77, + "learning_rate": 2.925027712772509e-06, + "loss": 0.6093, + "step": 2079 + }, + { + "epoch": 0.77, + "learning_rate": 2.9249907624091636e-06, + "loss": 0.588, + "step": 2080 + }, + { + "epoch": 0.77, + "learning_rate": 2.9249538120458187e-06, + "loss": 0.6315, + "step": 2081 + }, + { + "epoch": 0.77, + "learning_rate": 2.9249168616824733e-06, + "loss": 0.6762, + "step": 2082 + }, + { + "epoch": 0.77, + "learning_rate": 2.924879911319128e-06, + "loss": 0.6266, + "step": 2083 + }, + { + "epoch": 0.77, + "learning_rate": 2.924842960955783e-06, + "loss": 0.61, + "step": 2084 + }, + { + "epoch": 0.77, + "learning_rate": 2.9248060105924376e-06, + "loss": 0.6604, + "step": 2085 + }, + { + "epoch": 0.77, + "learning_rate": 2.9247690602290923e-06, + "loss": 0.6586, + "step": 2086 + }, + { + "epoch": 0.77, + "learning_rate": 2.924732109865747e-06, + "loss": 0.5823, + "step": 2087 + }, + { + "epoch": 0.77, + "learning_rate": 2.924695159502402e-06, + "loss": 0.6909, + "step": 2088 + }, + { + "epoch": 0.77, + "learning_rate": 2.9246582091390566e-06, + "loss": 0.7032, + "step": 2089 + }, + { + "epoch": 0.77, + "learning_rate": 2.9246212587757117e-06, + "loss": 0.6606, + "step": 2090 + }, + { + "epoch": 0.77, + "learning_rate": 2.9245843084123663e-06, + "loss": 0.6468, + "step": 2091 + }, + { + "epoch": 0.77, + "learning_rate": 2.924547358049021e-06, + "loss": 0.6012, + "step": 2092 + }, + { + "epoch": 0.77, + "learning_rate": 2.9245104076856756e-06, + "loss": 0.6406, + "step": 2093 + }, + { + "epoch": 0.77, + "learning_rate": 2.92447345732233e-06, + "loss": 0.6552, + "step": 2094 + }, + { + "epoch": 0.77, + "learning_rate": 2.9244365069589853e-06, + "loss": 0.6055, + "step": 2095 + }, + { + "epoch": 0.77, + "learning_rate": 2.92439955659564e-06, + "loss": 0.6331, + "step": 2096 + }, + { + "epoch": 0.77, + "learning_rate": 2.924362606232295e-06, + "loss": 0.646, + "step": 2097 + }, + { + "epoch": 0.77, + "learning_rate": 2.9243256558689496e-06, + "loss": 0.6636, + "step": 2098 + }, + { + "epoch": 0.78, + "learning_rate": 2.9242887055056042e-06, + "loss": 0.6312, + "step": 2099 + }, + { + "epoch": 0.78, + "learning_rate": 2.924251755142259e-06, + "loss": 0.603, + "step": 2100 + }, + { + "epoch": 0.78, + "learning_rate": 2.924214804778914e-06, + "loss": 0.6571, + "step": 2101 + }, + { + "epoch": 0.78, + "learning_rate": 2.9241778544155686e-06, + "loss": 0.6426, + "step": 2102 + }, + { + "epoch": 0.78, + "learning_rate": 2.924140904052223e-06, + "loss": 0.7336, + "step": 2103 + }, + { + "epoch": 0.78, + "learning_rate": 2.9241039536888783e-06, + "loss": 0.6432, + "step": 2104 + }, + { + "epoch": 0.78, + "learning_rate": 2.9240670033255325e-06, + "loss": 0.6792, + "step": 2105 + }, + { + "epoch": 0.78, + "learning_rate": 2.9240300529621875e-06, + "loss": 0.6221, + "step": 2106 + }, + { + "epoch": 0.78, + "learning_rate": 2.923993102598842e-06, + "loss": 0.6022, + "step": 2107 + }, + { + "epoch": 0.78, + "learning_rate": 2.9239561522354972e-06, + "loss": 0.7075, + "step": 2108 + }, + { + "epoch": 0.78, + "learning_rate": 2.923919201872152e-06, + "loss": 0.6723, + "step": 2109 + }, + { + "epoch": 0.78, + "learning_rate": 2.923882251508807e-06, + "loss": 0.6491, + "step": 2110 + }, + { + "epoch": 0.78, + "learning_rate": 2.923845301145461e-06, + "loss": 0.6287, + "step": 2111 + }, + { + "epoch": 0.78, + "learning_rate": 2.9238083507821158e-06, + "loss": 0.6688, + "step": 2112 + }, + { + "epoch": 0.78, + "learning_rate": 2.923771400418771e-06, + "loss": 0.6834, + "step": 2113 + }, + { + "epoch": 0.78, + "learning_rate": 2.9237344500554255e-06, + "loss": 0.5951, + "step": 2114 + }, + { + "epoch": 0.78, + "learning_rate": 2.9236974996920805e-06, + "loss": 0.5908, + "step": 2115 + }, + { + "epoch": 0.78, + "learning_rate": 2.923660549328735e-06, + "loss": 0.6448, + "step": 2116 + }, + { + "epoch": 0.78, + "learning_rate": 2.92362359896539e-06, + "loss": 0.6571, + "step": 2117 + }, + { + "epoch": 0.78, + "learning_rate": 2.9235866486020444e-06, + "loss": 0.5912, + "step": 2118 + }, + { + "epoch": 0.78, + "learning_rate": 2.9235496982386995e-06, + "loss": 0.655, + "step": 2119 + }, + { + "epoch": 0.78, + "learning_rate": 2.923512747875354e-06, + "loss": 0.6652, + "step": 2120 + }, + { + "epoch": 0.78, + "learning_rate": 2.923475797512009e-06, + "loss": 0.6095, + "step": 2121 + }, + { + "epoch": 0.78, + "learning_rate": 2.923438847148664e-06, + "loss": 0.6973, + "step": 2122 + }, + { + "epoch": 0.78, + "learning_rate": 2.9234018967853185e-06, + "loss": 0.6328, + "step": 2123 + }, + { + "epoch": 0.78, + "learning_rate": 2.923364946421973e-06, + "loss": 0.6347, + "step": 2124 + }, + { + "epoch": 0.78, + "learning_rate": 2.9233279960586278e-06, + "loss": 0.6193, + "step": 2125 + }, + { + "epoch": 0.79, + "learning_rate": 2.923291045695283e-06, + "loss": 0.5775, + "step": 2126 + }, + { + "epoch": 0.79, + "learning_rate": 2.9232540953319374e-06, + "loss": 0.7012, + "step": 2127 + }, + { + "epoch": 0.79, + "learning_rate": 2.9232171449685925e-06, + "loss": 0.6961, + "step": 2128 + }, + { + "epoch": 0.79, + "learning_rate": 2.923180194605247e-06, + "loss": 0.6353, + "step": 2129 + }, + { + "epoch": 0.79, + "learning_rate": 2.9231432442419018e-06, + "loss": 0.6723, + "step": 2130 + }, + { + "epoch": 0.79, + "learning_rate": 2.9231062938785564e-06, + "loss": 0.6674, + "step": 2131 + }, + { + "epoch": 0.79, + "learning_rate": 2.923069343515211e-06, + "loss": 0.5896, + "step": 2132 + }, + { + "epoch": 0.79, + "learning_rate": 2.923032393151866e-06, + "loss": 0.5724, + "step": 2133 + }, + { + "epoch": 0.79, + "learning_rate": 2.9229954427885207e-06, + "loss": 0.6777, + "step": 2134 + }, + { + "epoch": 0.79, + "learning_rate": 2.922958492425176e-06, + "loss": 0.6539, + "step": 2135 + }, + { + "epoch": 0.79, + "learning_rate": 2.9229215420618304e-06, + "loss": 0.6656, + "step": 2136 + }, + { + "epoch": 0.79, + "learning_rate": 2.922884591698485e-06, + "loss": 0.6279, + "step": 2137 + }, + { + "epoch": 0.79, + "learning_rate": 2.9228476413351397e-06, + "loss": 0.64, + "step": 2138 + }, + { + "epoch": 0.79, + "learning_rate": 2.9228106909717948e-06, + "loss": 0.6956, + "step": 2139 + }, + { + "epoch": 0.79, + "learning_rate": 2.9227737406084494e-06, + "loss": 0.6264, + "step": 2140 + }, + { + "epoch": 0.79, + "learning_rate": 2.922736790245104e-06, + "loss": 0.6377, + "step": 2141 + }, + { + "epoch": 0.79, + "learning_rate": 2.922699839881759e-06, + "loss": 0.6651, + "step": 2142 + }, + { + "epoch": 0.79, + "learning_rate": 2.9226628895184137e-06, + "loss": 0.7061, + "step": 2143 + }, + { + "epoch": 0.79, + "learning_rate": 2.9226259391550684e-06, + "loss": 0.6545, + "step": 2144 + }, + { + "epoch": 0.79, + "learning_rate": 2.922588988791723e-06, + "loss": 0.6753, + "step": 2145 + }, + { + "epoch": 0.79, + "learning_rate": 2.922552038428378e-06, + "loss": 0.7287, + "step": 2146 + }, + { + "epoch": 0.79, + "learning_rate": 2.9225150880650327e-06, + "loss": 0.6296, + "step": 2147 + }, + { + "epoch": 0.79, + "learning_rate": 2.9224781377016878e-06, + "loss": 0.622, + "step": 2148 + }, + { + "epoch": 0.79, + "learning_rate": 2.922441187338342e-06, + "loss": 0.6258, + "step": 2149 + }, + { + "epoch": 0.79, + "learning_rate": 2.922404236974997e-06, + "loss": 0.6658, + "step": 2150 + }, + { + "epoch": 0.79, + "learning_rate": 2.9223672866116517e-06, + "loss": 0.6823, + "step": 2151 + }, + { + "epoch": 0.79, + "learning_rate": 2.9223303362483063e-06, + "loss": 0.6636, + "step": 2152 + }, + { + "epoch": 0.8, + "learning_rate": 2.9222933858849614e-06, + "loss": 0.5902, + "step": 2153 + }, + { + "epoch": 0.8, + "learning_rate": 2.922256435521616e-06, + "loss": 0.5906, + "step": 2154 + }, + { + "epoch": 0.8, + "learning_rate": 2.9222194851582707e-06, + "loss": 0.6608, + "step": 2155 + }, + { + "epoch": 0.8, + "learning_rate": 2.9221825347949253e-06, + "loss": 0.6323, + "step": 2156 + }, + { + "epoch": 0.8, + "learning_rate": 2.9221455844315804e-06, + "loss": 0.6505, + "step": 2157 + }, + { + "epoch": 0.8, + "learning_rate": 2.922108634068235e-06, + "loss": 0.6679, + "step": 2158 + }, + { + "epoch": 0.8, + "learning_rate": 2.92207168370489e-06, + "loss": 0.6456, + "step": 2159 + }, + { + "epoch": 0.8, + "learning_rate": 2.9220347333415447e-06, + "loss": 0.6091, + "step": 2160 + }, + { + "epoch": 0.8, + "learning_rate": 2.9219977829781993e-06, + "loss": 0.6762, + "step": 2161 + }, + { + "epoch": 0.8, + "learning_rate": 2.921960832614854e-06, + "loss": 0.7021, + "step": 2162 + }, + { + "epoch": 0.8, + "learning_rate": 2.9219238822515086e-06, + "loss": 0.6391, + "step": 2163 + }, + { + "epoch": 0.8, + "learning_rate": 2.9218869318881637e-06, + "loss": 0.6444, + "step": 2164 + }, + { + "epoch": 0.8, + "learning_rate": 2.9218499815248183e-06, + "loss": 0.6875, + "step": 2165 + }, + { + "epoch": 0.8, + "learning_rate": 2.9218130311614734e-06, + "loss": 0.6123, + "step": 2166 + }, + { + "epoch": 0.8, + "learning_rate": 2.921776080798128e-06, + "loss": 0.6586, + "step": 2167 + }, + { + "epoch": 0.8, + "learning_rate": 2.9217391304347826e-06, + "loss": 0.6989, + "step": 2168 + }, + { + "epoch": 0.8, + "learning_rate": 2.9217021800714373e-06, + "loss": 0.6493, + "step": 2169 + }, + { + "epoch": 0.8, + "learning_rate": 2.9216652297080923e-06, + "loss": 0.7166, + "step": 2170 + }, + { + "epoch": 0.8, + "learning_rate": 2.921628279344747e-06, + "loss": 0.6303, + "step": 2171 + }, + { + "epoch": 0.8, + "learning_rate": 2.9215913289814016e-06, + "loss": 0.6559, + "step": 2172 + }, + { + "epoch": 0.8, + "learning_rate": 2.9215543786180567e-06, + "loss": 0.611, + "step": 2173 + }, + { + "epoch": 0.8, + "learning_rate": 2.9215174282547113e-06, + "loss": 0.6007, + "step": 2174 + }, + { + "epoch": 0.8, + "learning_rate": 2.921480477891366e-06, + "loss": 0.6151, + "step": 2175 + }, + { + "epoch": 0.8, + "learning_rate": 2.9214435275280206e-06, + "loss": 0.6118, + "step": 2176 + }, + { + "epoch": 0.8, + "learning_rate": 2.9214065771646756e-06, + "loss": 0.6262, + "step": 2177 + }, + { + "epoch": 0.8, + "learning_rate": 2.9213696268013303e-06, + "loss": 0.6012, + "step": 2178 + }, + { + "epoch": 0.8, + "learning_rate": 2.9213326764379853e-06, + "loss": 0.6969, + "step": 2179 + }, + { + "epoch": 0.81, + "learning_rate": 2.92129572607464e-06, + "loss": 0.6364, + "step": 2180 + }, + { + "epoch": 0.81, + "learning_rate": 2.9212587757112946e-06, + "loss": 0.7024, + "step": 2181 + }, + { + "epoch": 0.81, + "learning_rate": 2.9212218253479492e-06, + "loss": 0.6504, + "step": 2182 + }, + { + "epoch": 0.81, + "learning_rate": 2.921184874984604e-06, + "loss": 0.6513, + "step": 2183 + }, + { + "epoch": 0.81, + "learning_rate": 2.921147924621259e-06, + "loss": 0.5964, + "step": 2184 + }, + { + "epoch": 0.81, + "learning_rate": 2.9211109742579136e-06, + "loss": 0.633, + "step": 2185 + }, + { + "epoch": 0.81, + "learning_rate": 2.9210740238945686e-06, + "loss": 0.6873, + "step": 2186 + }, + { + "epoch": 0.81, + "learning_rate": 2.9210370735312233e-06, + "loss": 0.5958, + "step": 2187 + }, + { + "epoch": 0.81, + "learning_rate": 2.921000123167878e-06, + "loss": 0.631, + "step": 2188 + }, + { + "epoch": 0.81, + "learning_rate": 2.9209631728045325e-06, + "loss": 0.6722, + "step": 2189 + }, + { + "epoch": 0.81, + "learning_rate": 2.9209262224411876e-06, + "loss": 0.6781, + "step": 2190 + }, + { + "epoch": 0.81, + "learning_rate": 2.9208892720778422e-06, + "loss": 0.638, + "step": 2191 + }, + { + "epoch": 0.81, + "learning_rate": 2.920852321714497e-06, + "loss": 0.6124, + "step": 2192 + }, + { + "epoch": 0.81, + "learning_rate": 2.920815371351152e-06, + "loss": 0.6948, + "step": 2193 + }, + { + "epoch": 0.81, + "learning_rate": 2.920778420987806e-06, + "loss": 0.6885, + "step": 2194 + }, + { + "epoch": 0.81, + "learning_rate": 2.920741470624461e-06, + "loss": 0.676, + "step": 2195 + }, + { + "epoch": 0.81, + "learning_rate": 2.920704520261116e-06, + "loss": 0.6185, + "step": 2196 + }, + { + "epoch": 0.81, + "learning_rate": 2.920667569897771e-06, + "loss": 0.6234, + "step": 2197 + }, + { + "epoch": 0.81, + "learning_rate": 2.9206306195344255e-06, + "loss": 0.6858, + "step": 2198 + }, + { + "epoch": 0.81, + "learning_rate": 2.9205936691710806e-06, + "loss": 0.6641, + "step": 2199 + }, + { + "epoch": 0.81, + "learning_rate": 2.920556718807735e-06, + "loss": 0.6412, + "step": 2200 + }, + { + "epoch": 0.81, + "learning_rate": 2.9205197684443894e-06, + "loss": 0.6611, + "step": 2201 + }, + { + "epoch": 0.81, + "learning_rate": 2.9204828180810445e-06, + "loss": 0.5969, + "step": 2202 + }, + { + "epoch": 0.81, + "learning_rate": 2.920445867717699e-06, + "loss": 0.6565, + "step": 2203 + }, + { + "epoch": 0.81, + "learning_rate": 2.920408917354354e-06, + "loss": 0.6864, + "step": 2204 + }, + { + "epoch": 0.81, + "learning_rate": 2.920371966991009e-06, + "loss": 0.6985, + "step": 2205 + }, + { + "epoch": 0.81, + "learning_rate": 2.9203350166276635e-06, + "loss": 0.6869, + "step": 2206 + }, + { + "epoch": 0.81, + "learning_rate": 2.920298066264318e-06, + "loss": 0.644, + "step": 2207 + }, + { + "epoch": 0.82, + "learning_rate": 2.920261115900973e-06, + "loss": 0.6732, + "step": 2208 + }, + { + "epoch": 0.82, + "learning_rate": 2.920224165537628e-06, + "loss": 0.6338, + "step": 2209 + }, + { + "epoch": 0.82, + "learning_rate": 2.920187215174283e-06, + "loss": 0.6396, + "step": 2210 + }, + { + "epoch": 0.82, + "learning_rate": 2.9201502648109375e-06, + "loss": 0.6418, + "step": 2211 + }, + { + "epoch": 0.82, + "learning_rate": 2.920113314447592e-06, + "loss": 0.6753, + "step": 2212 + }, + { + "epoch": 0.82, + "learning_rate": 2.9200763640842468e-06, + "loss": 0.6573, + "step": 2213 + }, + { + "epoch": 0.82, + "learning_rate": 2.9200394137209014e-06, + "loss": 0.5747, + "step": 2214 + }, + { + "epoch": 0.82, + "learning_rate": 2.9200024633575565e-06, + "loss": 0.6839, + "step": 2215 + }, + { + "epoch": 0.82, + "learning_rate": 2.919965512994211e-06, + "loss": 0.648, + "step": 2216 + }, + { + "epoch": 0.82, + "learning_rate": 2.919928562630866e-06, + "loss": 0.6894, + "step": 2217 + }, + { + "epoch": 0.82, + "learning_rate": 2.919891612267521e-06, + "loss": 0.6647, + "step": 2218 + }, + { + "epoch": 0.82, + "learning_rate": 2.9198546619041754e-06, + "loss": 0.6229, + "step": 2219 + }, + { + "epoch": 0.82, + "learning_rate": 2.91981771154083e-06, + "loss": 0.6498, + "step": 2220 + }, + { + "epoch": 0.82, + "learning_rate": 2.9197807611774847e-06, + "loss": 0.7073, + "step": 2221 + }, + { + "epoch": 0.82, + "learning_rate": 2.9197438108141398e-06, + "loss": 0.6865, + "step": 2222 + }, + { + "epoch": 0.82, + "learning_rate": 2.9197068604507944e-06, + "loss": 0.6729, + "step": 2223 + }, + { + "epoch": 0.82, + "learning_rate": 2.9196699100874495e-06, + "loss": 0.6182, + "step": 2224 + }, + { + "epoch": 0.82, + "learning_rate": 2.919632959724104e-06, + "loss": 0.737, + "step": 2225 + }, + { + "epoch": 0.82, + "learning_rate": 2.9195960093607587e-06, + "loss": 0.6023, + "step": 2226 + }, + { + "epoch": 0.82, + "learning_rate": 2.9195590589974134e-06, + "loss": 0.6301, + "step": 2227 + }, + { + "epoch": 0.82, + "learning_rate": 2.9195221086340684e-06, + "loss": 0.5864, + "step": 2228 + }, + { + "epoch": 0.82, + "learning_rate": 2.919485158270723e-06, + "loss": 0.6299, + "step": 2229 + }, + { + "epoch": 0.82, + "learning_rate": 2.919448207907378e-06, + "loss": 0.6731, + "step": 2230 + }, + { + "epoch": 0.82, + "learning_rate": 2.9194112575440328e-06, + "loss": 0.6639, + "step": 2231 + }, + { + "epoch": 0.82, + "learning_rate": 2.9193743071806874e-06, + "loss": 0.7063, + "step": 2232 + }, + { + "epoch": 0.82, + "learning_rate": 2.919337356817342e-06, + "loss": 0.646, + "step": 2233 + }, + { + "epoch": 0.82, + "learning_rate": 2.9193004064539967e-06, + "loss": 0.6871, + "step": 2234 + }, + { + "epoch": 0.83, + "learning_rate": 2.9192634560906517e-06, + "loss": 0.643, + "step": 2235 + }, + { + "epoch": 0.83, + "learning_rate": 2.9192265057273064e-06, + "loss": 0.685, + "step": 2236 + }, + { + "epoch": 0.83, + "learning_rate": 2.9191895553639614e-06, + "loss": 0.64, + "step": 2237 + }, + { + "epoch": 0.83, + "learning_rate": 2.919152605000616e-06, + "loss": 0.6221, + "step": 2238 + }, + { + "epoch": 0.83, + "learning_rate": 2.9191156546372707e-06, + "loss": 0.6171, + "step": 2239 + }, + { + "epoch": 0.83, + "learning_rate": 2.9190787042739253e-06, + "loss": 0.6457, + "step": 2240 + }, + { + "epoch": 0.83, + "learning_rate": 2.91904175391058e-06, + "loss": 0.6054, + "step": 2241 + }, + { + "epoch": 0.83, + "learning_rate": 2.919004803547235e-06, + "loss": 0.681, + "step": 2242 + }, + { + "epoch": 0.83, + "learning_rate": 2.9189678531838897e-06, + "loss": 0.6256, + "step": 2243 + }, + { + "epoch": 0.83, + "learning_rate": 2.9189309028205447e-06, + "loss": 0.6096, + "step": 2244 + }, + { + "epoch": 0.83, + "learning_rate": 2.918893952457199e-06, + "loss": 0.7049, + "step": 2245 + }, + { + "epoch": 0.83, + "learning_rate": 2.918857002093854e-06, + "loss": 0.6792, + "step": 2246 + }, + { + "epoch": 0.83, + "learning_rate": 2.9188200517305086e-06, + "loss": 0.5975, + "step": 2247 + }, + { + "epoch": 0.83, + "learning_rate": 2.9187831013671637e-06, + "loss": 0.6577, + "step": 2248 + }, + { + "epoch": 0.83, + "learning_rate": 2.9187461510038183e-06, + "loss": 0.6629, + "step": 2249 + }, + { + "epoch": 0.83, + "learning_rate": 2.9187092006404734e-06, + "loss": 0.6588, + "step": 2250 + }, + { + "epoch": 0.83, + "learning_rate": 2.9186722502771276e-06, + "loss": 0.6715, + "step": 2251 + }, + { + "epoch": 0.83, + "learning_rate": 2.9186352999137822e-06, + "loss": 0.6484, + "step": 2252 + }, + { + "epoch": 0.83, + "learning_rate": 2.9185983495504373e-06, + "loss": 0.6419, + "step": 2253 + }, + { + "epoch": 0.83, + "learning_rate": 2.918561399187092e-06, + "loss": 0.645, + "step": 2254 + }, + { + "epoch": 0.83, + "learning_rate": 2.918524448823747e-06, + "loss": 0.6744, + "step": 2255 + }, + { + "epoch": 0.83, + "learning_rate": 2.9184874984604016e-06, + "loss": 0.6416, + "step": 2256 + }, + { + "epoch": 0.83, + "learning_rate": 2.9184505480970563e-06, + "loss": 0.6318, + "step": 2257 + }, + { + "epoch": 0.83, + "learning_rate": 2.918413597733711e-06, + "loss": 0.6515, + "step": 2258 + }, + { + "epoch": 0.83, + "learning_rate": 2.918376647370366e-06, + "loss": 0.6234, + "step": 2259 + }, + { + "epoch": 0.83, + "learning_rate": 2.9183396970070206e-06, + "loss": 0.5879, + "step": 2260 + }, + { + "epoch": 0.83, + "learning_rate": 2.9183027466436752e-06, + "loss": 0.6473, + "step": 2261 + }, + { + "epoch": 0.84, + "learning_rate": 2.9182657962803303e-06, + "loss": 0.68, + "step": 2262 + }, + { + "epoch": 0.84, + "learning_rate": 2.918228845916985e-06, + "loss": 0.6674, + "step": 2263 + }, + { + "epoch": 0.84, + "learning_rate": 2.9181918955536396e-06, + "loss": 0.6295, + "step": 2264 + }, + { + "epoch": 0.84, + "learning_rate": 2.9181549451902942e-06, + "loss": 0.5921, + "step": 2265 + }, + { + "epoch": 0.84, + "learning_rate": 2.9181179948269493e-06, + "loss": 0.6165, + "step": 2266 + }, + { + "epoch": 0.84, + "learning_rate": 2.918081044463604e-06, + "loss": 0.6363, + "step": 2267 + }, + { + "epoch": 0.84, + "learning_rate": 2.918044094100259e-06, + "loss": 0.5929, + "step": 2268 + }, + { + "epoch": 0.84, + "learning_rate": 2.9180071437369136e-06, + "loss": 0.6479, + "step": 2269 + }, + { + "epoch": 0.84, + "learning_rate": 2.9179701933735682e-06, + "loss": 0.6177, + "step": 2270 + }, + { + "epoch": 0.84, + "learning_rate": 2.917933243010223e-06, + "loss": 0.657, + "step": 2271 + }, + { + "epoch": 0.84, + "learning_rate": 2.9178962926468775e-06, + "loss": 0.6058, + "step": 2272 + }, + { + "epoch": 0.84, + "learning_rate": 2.9178593422835326e-06, + "loss": 0.7102, + "step": 2273 + }, + { + "epoch": 0.84, + "learning_rate": 2.9178223919201872e-06, + "loss": 0.7008, + "step": 2274 + }, + { + "epoch": 0.84, + "learning_rate": 2.9177854415568423e-06, + "loss": 0.6318, + "step": 2275 + }, + { + "epoch": 0.84, + "learning_rate": 2.917748491193497e-06, + "loss": 0.6487, + "step": 2276 + }, + { + "epoch": 0.84, + "learning_rate": 2.9177115408301515e-06, + "loss": 0.6492, + "step": 2277 + }, + { + "epoch": 0.84, + "learning_rate": 2.917674590466806e-06, + "loss": 0.6473, + "step": 2278 + }, + { + "epoch": 0.84, + "learning_rate": 2.9176376401034612e-06, + "loss": 0.639, + "step": 2279 + }, + { + "epoch": 0.84, + "learning_rate": 2.917600689740116e-06, + "loss": 0.5997, + "step": 2280 + }, + { + "epoch": 0.84, + "learning_rate": 2.9175637393767705e-06, + "loss": 0.6185, + "step": 2281 + }, + { + "epoch": 0.84, + "learning_rate": 2.9175267890134256e-06, + "loss": 0.6176, + "step": 2282 + }, + { + "epoch": 0.84, + "learning_rate": 2.9174898386500802e-06, + "loss": 0.5952, + "step": 2283 + }, + { + "epoch": 0.84, + "learning_rate": 2.917452888286735e-06, + "loss": 0.5897, + "step": 2284 + }, + { + "epoch": 0.84, + "learning_rate": 2.9174159379233895e-06, + "loss": 0.6698, + "step": 2285 + }, + { + "epoch": 0.84, + "learning_rate": 2.9173789875600445e-06, + "loss": 0.6918, + "step": 2286 + }, + { + "epoch": 0.84, + "learning_rate": 2.917342037196699e-06, + "loss": 0.6314, + "step": 2287 + }, + { + "epoch": 0.84, + "learning_rate": 2.9173050868333542e-06, + "loss": 0.6386, + "step": 2288 + }, + { + "epoch": 0.85, + "learning_rate": 2.917268136470009e-06, + "loss": 0.6465, + "step": 2289 + }, + { + "epoch": 0.85, + "learning_rate": 2.9172311861066635e-06, + "loss": 0.5798, + "step": 2290 + }, + { + "epoch": 0.85, + "learning_rate": 2.917194235743318e-06, + "loss": 0.6218, + "step": 2291 + }, + { + "epoch": 0.85, + "learning_rate": 2.9171572853799728e-06, + "loss": 0.6798, + "step": 2292 + }, + { + "epoch": 0.85, + "learning_rate": 2.917120335016628e-06, + "loss": 0.6825, + "step": 2293 + }, + { + "epoch": 0.85, + "learning_rate": 2.9170833846532825e-06, + "loss": 0.6761, + "step": 2294 + }, + { + "epoch": 0.85, + "learning_rate": 2.9170464342899375e-06, + "loss": 0.6594, + "step": 2295 + }, + { + "epoch": 0.85, + "learning_rate": 2.9170094839265918e-06, + "loss": 0.6483, + "step": 2296 + }, + { + "epoch": 0.85, + "learning_rate": 2.916972533563247e-06, + "loss": 0.6188, + "step": 2297 + }, + { + "epoch": 0.85, + "learning_rate": 2.9169355831999015e-06, + "loss": 0.6071, + "step": 2298 + }, + { + "epoch": 0.85, + "learning_rate": 2.9168986328365565e-06, + "loss": 0.6791, + "step": 2299 + }, + { + "epoch": 0.85, + "learning_rate": 2.916861682473211e-06, + "loss": 0.6838, + "step": 2300 + }, + { + "epoch": 0.85, + "learning_rate": 2.9168247321098658e-06, + "loss": 0.6472, + "step": 2301 + }, + { + "epoch": 0.85, + "learning_rate": 2.9167877817465204e-06, + "loss": 0.6666, + "step": 2302 + }, + { + "epoch": 0.85, + "learning_rate": 2.916750831383175e-06, + "loss": 0.5576, + "step": 2303 + }, + { + "epoch": 0.85, + "learning_rate": 2.91671388101983e-06, + "loss": 0.5962, + "step": 2304 + }, + { + "epoch": 0.85, + "learning_rate": 2.9166769306564848e-06, + "loss": 0.5953, + "step": 2305 + }, + { + "epoch": 0.85, + "learning_rate": 2.91663998029314e-06, + "loss": 0.6395, + "step": 2306 + }, + { + "epoch": 0.85, + "learning_rate": 2.9166030299297945e-06, + "loss": 0.6495, + "step": 2307 + }, + { + "epoch": 0.85, + "learning_rate": 2.916566079566449e-06, + "loss": 0.6097, + "step": 2308 + }, + { + "epoch": 0.85, + "learning_rate": 2.9165291292031037e-06, + "loss": 0.6235, + "step": 2309 + }, + { + "epoch": 0.85, + "learning_rate": 2.9164921788397588e-06, + "loss": 0.6504, + "step": 2310 + }, + { + "epoch": 0.85, + "learning_rate": 2.9164552284764134e-06, + "loss": 0.6489, + "step": 2311 + }, + { + "epoch": 0.85, + "learning_rate": 2.916418278113068e-06, + "loss": 0.7167, + "step": 2312 + }, + { + "epoch": 0.85, + "learning_rate": 2.916381327749723e-06, + "loss": 0.693, + "step": 2313 + }, + { + "epoch": 0.85, + "learning_rate": 2.9163443773863778e-06, + "loss": 0.7237, + "step": 2314 + }, + { + "epoch": 0.85, + "learning_rate": 2.9163074270230324e-06, + "loss": 0.5928, + "step": 2315 + }, + { + "epoch": 0.86, + "learning_rate": 2.916270476659687e-06, + "loss": 0.6383, + "step": 2316 + }, + { + "epoch": 0.86, + "learning_rate": 2.916233526296342e-06, + "loss": 0.5807, + "step": 2317 + }, + { + "epoch": 0.86, + "learning_rate": 2.9161965759329967e-06, + "loss": 0.6182, + "step": 2318 + }, + { + "epoch": 0.86, + "learning_rate": 2.9161596255696518e-06, + "loss": 0.6005, + "step": 2319 + }, + { + "epoch": 0.86, + "learning_rate": 2.9161226752063064e-06, + "loss": 0.6791, + "step": 2320 + }, + { + "epoch": 0.86, + "learning_rate": 2.916085724842961e-06, + "loss": 0.6896, + "step": 2321 + }, + { + "epoch": 0.86, + "learning_rate": 2.9160487744796157e-06, + "loss": 0.6164, + "step": 2322 + }, + { + "epoch": 0.86, + "learning_rate": 2.9160118241162703e-06, + "loss": 0.6157, + "step": 2323 + }, + { + "epoch": 0.86, + "learning_rate": 2.9159748737529254e-06, + "loss": 0.635, + "step": 2324 + }, + { + "epoch": 0.86, + "learning_rate": 2.91593792338958e-06, + "loss": 0.5906, + "step": 2325 + }, + { + "epoch": 0.86, + "learning_rate": 2.915900973026235e-06, + "loss": 0.6624, + "step": 2326 + }, + { + "epoch": 0.86, + "learning_rate": 2.9158640226628897e-06, + "loss": 0.6326, + "step": 2327 + }, + { + "epoch": 0.86, + "learning_rate": 2.9158270722995444e-06, + "loss": 0.6007, + "step": 2328 + }, + { + "epoch": 0.86, + "learning_rate": 2.915790121936199e-06, + "loss": 0.725, + "step": 2329 + }, + { + "epoch": 0.86, + "learning_rate": 2.915753171572854e-06, + "loss": 0.6376, + "step": 2330 + }, + { + "epoch": 0.86, + "learning_rate": 2.9157162212095087e-06, + "loss": 0.7209, + "step": 2331 + }, + { + "epoch": 0.86, + "learning_rate": 2.9156792708461633e-06, + "loss": 0.6618, + "step": 2332 + }, + { + "epoch": 0.86, + "learning_rate": 2.9156423204828184e-06, + "loss": 0.5767, + "step": 2333 + }, + { + "epoch": 0.86, + "learning_rate": 2.915605370119473e-06, + "loss": 0.6977, + "step": 2334 + }, + { + "epoch": 0.86, + "learning_rate": 2.9155684197561277e-06, + "loss": 0.7247, + "step": 2335 + }, + { + "epoch": 0.86, + "learning_rate": 2.9155314693927823e-06, + "loss": 0.6415, + "step": 2336 + }, + { + "epoch": 0.86, + "learning_rate": 2.9154945190294374e-06, + "loss": 0.6143, + "step": 2337 + }, + { + "epoch": 0.86, + "learning_rate": 2.915457568666092e-06, + "loss": 0.6127, + "step": 2338 + }, + { + "epoch": 0.86, + "learning_rate": 2.915420618302747e-06, + "loss": 0.6345, + "step": 2339 + }, + { + "epoch": 0.86, + "learning_rate": 2.9153836679394013e-06, + "loss": 0.6697, + "step": 2340 + }, + { + "epoch": 0.86, + "learning_rate": 2.915346717576056e-06, + "loss": 0.5946, + "step": 2341 + }, + { + "epoch": 0.86, + "learning_rate": 2.915309767212711e-06, + "loss": 0.672, + "step": 2342 + }, + { + "epoch": 0.87, + "learning_rate": 2.9152728168493656e-06, + "loss": 0.622, + "step": 2343 + }, + { + "epoch": 0.87, + "learning_rate": 2.9152358664860207e-06, + "loss": 0.6425, + "step": 2344 + }, + { + "epoch": 0.87, + "learning_rate": 2.9151989161226753e-06, + "loss": 0.5924, + "step": 2345 + }, + { + "epoch": 0.87, + "learning_rate": 2.91516196575933e-06, + "loss": 0.6177, + "step": 2346 + }, + { + "epoch": 0.87, + "learning_rate": 2.9151250153959846e-06, + "loss": 0.6295, + "step": 2347 + }, + { + "epoch": 0.87, + "learning_rate": 2.9150880650326396e-06, + "loss": 0.6499, + "step": 2348 + }, + { + "epoch": 0.87, + "learning_rate": 2.9150511146692943e-06, + "loss": 0.6465, + "step": 2349 + }, + { + "epoch": 0.87, + "learning_rate": 2.9150141643059493e-06, + "loss": 0.6423, + "step": 2350 + }, + { + "epoch": 0.87, + "learning_rate": 2.914977213942604e-06, + "loss": 0.5915, + "step": 2351 + }, + { + "epoch": 0.87, + "learning_rate": 2.9149402635792586e-06, + "loss": 0.6486, + "step": 2352 + }, + { + "epoch": 0.87, + "learning_rate": 2.9149033132159132e-06, + "loss": 0.6186, + "step": 2353 + }, + { + "epoch": 0.87, + "learning_rate": 2.914866362852568e-06, + "loss": 0.5895, + "step": 2354 + }, + { + "epoch": 0.87, + "learning_rate": 2.914829412489223e-06, + "loss": 0.5916, + "step": 2355 + }, + { + "epoch": 0.87, + "learning_rate": 2.9147924621258776e-06, + "loss": 0.6453, + "step": 2356 + }, + { + "epoch": 0.87, + "learning_rate": 2.9147555117625326e-06, + "loss": 0.5975, + "step": 2357 + }, + { + "epoch": 0.87, + "learning_rate": 2.9147185613991873e-06, + "loss": 0.7374, + "step": 2358 + }, + { + "epoch": 0.87, + "learning_rate": 2.914681611035842e-06, + "loss": 0.5555, + "step": 2359 + }, + { + "epoch": 0.87, + "learning_rate": 2.9146446606724965e-06, + "loss": 0.6171, + "step": 2360 + }, + { + "epoch": 0.87, + "learning_rate": 2.914607710309151e-06, + "loss": 0.672, + "step": 2361 + }, + { + "epoch": 0.87, + "learning_rate": 2.9145707599458062e-06, + "loss": 0.5805, + "step": 2362 + }, + { + "epoch": 0.87, + "learning_rate": 2.914533809582461e-06, + "loss": 0.6248, + "step": 2363 + }, + { + "epoch": 0.87, + "learning_rate": 2.914496859219116e-06, + "loss": 0.6092, + "step": 2364 + }, + { + "epoch": 0.87, + "learning_rate": 2.9144599088557706e-06, + "loss": 0.694, + "step": 2365 + }, + { + "epoch": 0.87, + "learning_rate": 2.914422958492425e-06, + "loss": 0.6673, + "step": 2366 + }, + { + "epoch": 0.87, + "learning_rate": 2.91438600812908e-06, + "loss": 0.6229, + "step": 2367 + }, + { + "epoch": 0.87, + "learning_rate": 2.914349057765735e-06, + "loss": 0.7061, + "step": 2368 + }, + { + "epoch": 0.87, + "learning_rate": 2.9143121074023895e-06, + "loss": 0.6575, + "step": 2369 + }, + { + "epoch": 0.88, + "learning_rate": 2.914275157039044e-06, + "loss": 0.5738, + "step": 2370 + }, + { + "epoch": 0.88, + "learning_rate": 2.9142382066756992e-06, + "loss": 0.668, + "step": 2371 + }, + { + "epoch": 0.88, + "learning_rate": 2.914201256312354e-06, + "loss": 0.6152, + "step": 2372 + }, + { + "epoch": 0.88, + "learning_rate": 2.9141643059490085e-06, + "loss": 0.6741, + "step": 2373 + }, + { + "epoch": 0.88, + "learning_rate": 2.914127355585663e-06, + "loss": 0.6124, + "step": 2374 + }, + { + "epoch": 0.88, + "learning_rate": 2.914090405222318e-06, + "loss": 0.5765, + "step": 2375 + }, + { + "epoch": 0.88, + "learning_rate": 2.914053454858973e-06, + "loss": 0.7308, + "step": 2376 + }, + { + "epoch": 0.88, + "learning_rate": 2.914016504495628e-06, + "loss": 0.6077, + "step": 2377 + }, + { + "epoch": 0.88, + "learning_rate": 2.9139795541322825e-06, + "loss": 0.6704, + "step": 2378 + }, + { + "epoch": 0.88, + "learning_rate": 2.913942603768937e-06, + "loss": 0.7341, + "step": 2379 + }, + { + "epoch": 0.88, + "learning_rate": 2.913905653405592e-06, + "loss": 0.5718, + "step": 2380 + }, + { + "epoch": 0.88, + "learning_rate": 2.9138687030422464e-06, + "loss": 0.6321, + "step": 2381 + }, + { + "epoch": 0.88, + "learning_rate": 2.9138317526789015e-06, + "loss": 0.6128, + "step": 2382 + }, + { + "epoch": 0.88, + "learning_rate": 2.913794802315556e-06, + "loss": 0.6039, + "step": 2383 + }, + { + "epoch": 0.88, + "learning_rate": 2.913757851952211e-06, + "loss": 0.6831, + "step": 2384 + }, + { + "epoch": 0.88, + "learning_rate": 2.9137209015888654e-06, + "loss": 0.71, + "step": 2385 + }, + { + "epoch": 0.88, + "learning_rate": 2.9136839512255205e-06, + "loss": 0.6571, + "step": 2386 + }, + { + "epoch": 0.88, + "learning_rate": 2.913647000862175e-06, + "loss": 0.6095, + "step": 2387 + }, + { + "epoch": 0.88, + "learning_rate": 2.91361005049883e-06, + "loss": 0.632, + "step": 2388 + }, + { + "epoch": 0.88, + "learning_rate": 2.913573100135485e-06, + "loss": 0.6656, + "step": 2389 + }, + { + "epoch": 0.88, + "learning_rate": 2.9135361497721394e-06, + "loss": 0.6464, + "step": 2390 + }, + { + "epoch": 0.88, + "learning_rate": 2.913499199408794e-06, + "loss": 0.5872, + "step": 2391 + }, + { + "epoch": 0.88, + "learning_rate": 2.9134622490454487e-06, + "loss": 0.6239, + "step": 2392 + }, + { + "epoch": 0.88, + "learning_rate": 2.9134252986821038e-06, + "loss": 0.6223, + "step": 2393 + }, + { + "epoch": 0.88, + "learning_rate": 2.9133883483187584e-06, + "loss": 0.7139, + "step": 2394 + }, + { + "epoch": 0.88, + "learning_rate": 2.9133513979554135e-06, + "loss": 0.6325, + "step": 2395 + }, + { + "epoch": 0.88, + "learning_rate": 2.913314447592068e-06, + "loss": 0.6899, + "step": 2396 + }, + { + "epoch": 0.89, + "learning_rate": 2.9132774972287227e-06, + "loss": 0.6571, + "step": 2397 + }, + { + "epoch": 0.89, + "learning_rate": 2.9132405468653774e-06, + "loss": 0.6483, + "step": 2398 + }, + { + "epoch": 0.89, + "learning_rate": 2.9132035965020324e-06, + "loss": 0.6027, + "step": 2399 + }, + { + "epoch": 0.89, + "learning_rate": 2.913166646138687e-06, + "loss": 0.6104, + "step": 2400 + }, + { + "epoch": 0.89, + "learning_rate": 2.9131296957753417e-06, + "loss": 0.5939, + "step": 2401 + }, + { + "epoch": 0.89, + "learning_rate": 2.9130927454119968e-06, + "loss": 0.6723, + "step": 2402 + }, + { + "epoch": 0.89, + "learning_rate": 2.9130557950486514e-06, + "loss": 0.613, + "step": 2403 + }, + { + "epoch": 0.89, + "learning_rate": 2.913018844685306e-06, + "loss": 0.6209, + "step": 2404 + }, + { + "epoch": 0.89, + "learning_rate": 2.9129818943219607e-06, + "loss": 0.6043, + "step": 2405 + }, + { + "epoch": 0.89, + "learning_rate": 2.9129449439586157e-06, + "loss": 0.6838, + "step": 2406 + }, + { + "epoch": 0.89, + "learning_rate": 2.9129079935952704e-06, + "loss": 0.6433, + "step": 2407 + }, + { + "epoch": 0.89, + "learning_rate": 2.9128710432319254e-06, + "loss": 0.7089, + "step": 2408 + }, + { + "epoch": 0.89, + "learning_rate": 2.91283409286858e-06, + "loss": 0.6227, + "step": 2409 + }, + { + "epoch": 0.89, + "learning_rate": 2.9127971425052347e-06, + "loss": 0.6641, + "step": 2410 + }, + { + "epoch": 0.89, + "learning_rate": 2.9127601921418893e-06, + "loss": 0.6161, + "step": 2411 + }, + { + "epoch": 0.89, + "learning_rate": 2.912723241778544e-06, + "loss": 0.6608, + "step": 2412 + }, + { + "epoch": 0.89, + "learning_rate": 2.912686291415199e-06, + "loss": 0.6718, + "step": 2413 + }, + { + "epoch": 0.89, + "learning_rate": 2.9126493410518537e-06, + "loss": 0.6633, + "step": 2414 + }, + { + "epoch": 0.89, + "learning_rate": 2.9126123906885087e-06, + "loss": 0.6372, + "step": 2415 + }, + { + "epoch": 0.89, + "learning_rate": 2.9125754403251634e-06, + "loss": 0.6665, + "step": 2416 + }, + { + "epoch": 0.89, + "learning_rate": 2.912538489961818e-06, + "loss": 0.6035, + "step": 2417 + }, + { + "epoch": 0.89, + "learning_rate": 2.9125015395984726e-06, + "loss": 0.6249, + "step": 2418 + }, + { + "epoch": 0.89, + "learning_rate": 2.9124645892351277e-06, + "loss": 0.6587, + "step": 2419 + }, + { + "epoch": 0.89, + "learning_rate": 2.9124276388717823e-06, + "loss": 0.6267, + "step": 2420 + }, + { + "epoch": 0.89, + "learning_rate": 2.912390688508437e-06, + "loss": 0.6695, + "step": 2421 + }, + { + "epoch": 0.89, + "learning_rate": 2.912353738145092e-06, + "loss": 0.6512, + "step": 2422 + }, + { + "epoch": 0.89, + "learning_rate": 2.9123167877817467e-06, + "loss": 0.6703, + "step": 2423 + }, + { + "epoch": 0.9, + "learning_rate": 2.9122798374184013e-06, + "loss": 0.6707, + "step": 2424 + }, + { + "epoch": 0.9, + "learning_rate": 2.912242887055056e-06, + "loss": 0.6515, + "step": 2425 + }, + { + "epoch": 0.9, + "learning_rate": 2.912205936691711e-06, + "loss": 0.5942, + "step": 2426 + }, + { + "epoch": 0.9, + "learning_rate": 2.9121689863283656e-06, + "loss": 0.6177, + "step": 2427 + }, + { + "epoch": 0.9, + "learning_rate": 2.9121320359650207e-06, + "loss": 0.6853, + "step": 2428 + }, + { + "epoch": 0.9, + "learning_rate": 2.9120950856016753e-06, + "loss": 0.5848, + "step": 2429 + }, + { + "epoch": 0.9, + "learning_rate": 2.9120581352383296e-06, + "loss": 0.6649, + "step": 2430 + }, + { + "epoch": 0.9, + "learning_rate": 2.9120211848749846e-06, + "loss": 0.6483, + "step": 2431 + }, + { + "epoch": 0.9, + "learning_rate": 2.9119842345116393e-06, + "loss": 0.701, + "step": 2432 + }, + { + "epoch": 0.9, + "learning_rate": 2.9119472841482943e-06, + "loss": 0.6455, + "step": 2433 + }, + { + "epoch": 0.9, + "learning_rate": 2.911910333784949e-06, + "loss": 0.6243, + "step": 2434 + }, + { + "epoch": 0.9, + "learning_rate": 2.911873383421604e-06, + "loss": 0.6031, + "step": 2435 + }, + { + "epoch": 0.9, + "learning_rate": 2.9118364330582582e-06, + "loss": 0.5831, + "step": 2436 + }, + { + "epoch": 0.9, + "learning_rate": 2.9117994826949133e-06, + "loss": 0.6603, + "step": 2437 + }, + { + "epoch": 0.9, + "learning_rate": 2.911762532331568e-06, + "loss": 0.5762, + "step": 2438 + }, + { + "epoch": 0.9, + "learning_rate": 2.911725581968223e-06, + "loss": 0.6213, + "step": 2439 + }, + { + "epoch": 0.9, + "learning_rate": 2.9116886316048776e-06, + "loss": 0.6907, + "step": 2440 + }, + { + "epoch": 0.9, + "learning_rate": 2.9116516812415323e-06, + "loss": 0.5955, + "step": 2441 + }, + { + "epoch": 0.9, + "learning_rate": 2.911614730878187e-06, + "loss": 0.6207, + "step": 2442 + }, + { + "epoch": 0.9, + "learning_rate": 2.9115777805148415e-06, + "loss": 0.6373, + "step": 2443 + }, + { + "epoch": 0.9, + "learning_rate": 2.9115408301514966e-06, + "loss": 0.6485, + "step": 2444 + }, + { + "epoch": 0.9, + "learning_rate": 2.9115038797881512e-06, + "loss": 0.6846, + "step": 2445 + }, + { + "epoch": 0.9, + "learning_rate": 2.9114669294248063e-06, + "loss": 0.6384, + "step": 2446 + }, + { + "epoch": 0.9, + "learning_rate": 2.911429979061461e-06, + "loss": 0.6752, + "step": 2447 + }, + { + "epoch": 0.9, + "learning_rate": 2.9113930286981156e-06, + "loss": 0.6218, + "step": 2448 + }, + { + "epoch": 0.9, + "learning_rate": 2.91135607833477e-06, + "loss": 0.5896, + "step": 2449 + }, + { + "epoch": 0.9, + "learning_rate": 2.911319127971425e-06, + "loss": 0.6612, + "step": 2450 + }, + { + "epoch": 0.91, + "learning_rate": 2.91128217760808e-06, + "loss": 0.6547, + "step": 2451 + }, + { + "epoch": 0.91, + "learning_rate": 2.9112452272447345e-06, + "loss": 0.6472, + "step": 2452 + }, + { + "epoch": 0.91, + "learning_rate": 2.9112082768813896e-06, + "loss": 0.7138, + "step": 2453 + }, + { + "epoch": 0.91, + "learning_rate": 2.9111713265180442e-06, + "loss": 0.631, + "step": 2454 + }, + { + "epoch": 0.91, + "learning_rate": 2.911134376154699e-06, + "loss": 0.6411, + "step": 2455 + }, + { + "epoch": 0.91, + "learning_rate": 2.9110974257913535e-06, + "loss": 0.679, + "step": 2456 + }, + { + "epoch": 0.91, + "learning_rate": 2.9110604754280086e-06, + "loss": 0.6192, + "step": 2457 + }, + { + "epoch": 0.91, + "learning_rate": 2.911023525064663e-06, + "loss": 0.6569, + "step": 2458 + }, + { + "epoch": 0.91, + "learning_rate": 2.9109865747013182e-06, + "loss": 0.5973, + "step": 2459 + }, + { + "epoch": 0.91, + "learning_rate": 2.910949624337973e-06, + "loss": 0.6068, + "step": 2460 + }, + { + "epoch": 0.91, + "learning_rate": 2.9109126739746275e-06, + "loss": 0.6253, + "step": 2461 + }, + { + "epoch": 0.91, + "learning_rate": 2.910875723611282e-06, + "loss": 0.6731, + "step": 2462 + }, + { + "epoch": 0.91, + "learning_rate": 2.910838773247937e-06, + "loss": 0.6281, + "step": 2463 + }, + { + "epoch": 0.91, + "learning_rate": 2.910801822884592e-06, + "loss": 0.6516, + "step": 2464 + }, + { + "epoch": 0.91, + "learning_rate": 2.9107648725212465e-06, + "loss": 0.6042, + "step": 2465 + }, + { + "epoch": 0.91, + "learning_rate": 2.9107279221579016e-06, + "loss": 0.5947, + "step": 2466 + }, + { + "epoch": 0.91, + "learning_rate": 2.910690971794556e-06, + "loss": 0.6716, + "step": 2467 + }, + { + "epoch": 0.91, + "learning_rate": 2.910654021431211e-06, + "loss": 0.65, + "step": 2468 + }, + { + "epoch": 0.91, + "learning_rate": 2.9106170710678655e-06, + "loss": 0.6957, + "step": 2469 + }, + { + "epoch": 0.91, + "learning_rate": 2.91058012070452e-06, + "loss": 0.6512, + "step": 2470 + }, + { + "epoch": 0.91, + "learning_rate": 2.910543170341175e-06, + "loss": 0.5909, + "step": 2471 + }, + { + "epoch": 0.91, + "learning_rate": 2.91050621997783e-06, + "loss": 0.6476, + "step": 2472 + }, + { + "epoch": 0.91, + "learning_rate": 2.910469269614485e-06, + "loss": 0.6303, + "step": 2473 + }, + { + "epoch": 0.91, + "learning_rate": 2.9104323192511395e-06, + "loss": 0.6797, + "step": 2474 + }, + { + "epoch": 0.91, + "learning_rate": 2.910395368887794e-06, + "loss": 0.5963, + "step": 2475 + }, + { + "epoch": 0.91, + "learning_rate": 2.9103584185244488e-06, + "loss": 0.6594, + "step": 2476 + }, + { + "epoch": 0.91, + "learning_rate": 2.910321468161104e-06, + "loss": 0.6886, + "step": 2477 + }, + { + "epoch": 0.92, + "learning_rate": 2.9102845177977585e-06, + "loss": 0.6328, + "step": 2478 + }, + { + "epoch": 0.92, + "learning_rate": 2.9102475674344135e-06, + "loss": 0.6765, + "step": 2479 + }, + { + "epoch": 0.92, + "learning_rate": 2.910210617071068e-06, + "loss": 0.6174, + "step": 2480 + }, + { + "epoch": 0.92, + "learning_rate": 2.9101736667077224e-06, + "loss": 0.6144, + "step": 2481 + }, + { + "epoch": 0.92, + "learning_rate": 2.9101367163443774e-06, + "loss": 0.6584, + "step": 2482 + }, + { + "epoch": 0.92, + "learning_rate": 2.910099765981032e-06, + "loss": 0.6605, + "step": 2483 + }, + { + "epoch": 0.92, + "learning_rate": 2.910062815617687e-06, + "loss": 0.6499, + "step": 2484 + }, + { + "epoch": 0.92, + "learning_rate": 2.9100258652543418e-06, + "loss": 0.6247, + "step": 2485 + }, + { + "epoch": 0.92, + "learning_rate": 2.909988914890997e-06, + "loss": 0.6182, + "step": 2486 + }, + { + "epoch": 0.92, + "learning_rate": 2.909951964527651e-06, + "loss": 0.6224, + "step": 2487 + }, + { + "epoch": 0.92, + "learning_rate": 2.909915014164306e-06, + "loss": 0.7, + "step": 2488 + }, + { + "epoch": 0.92, + "learning_rate": 2.9098780638009607e-06, + "loss": 0.5963, + "step": 2489 + }, + { + "epoch": 0.92, + "learning_rate": 2.9098411134376154e-06, + "loss": 0.6389, + "step": 2490 + }, + { + "epoch": 0.92, + "learning_rate": 2.9098041630742704e-06, + "loss": 0.6249, + "step": 2491 + }, + { + "epoch": 0.92, + "learning_rate": 2.909767212710925e-06, + "loss": 0.5971, + "step": 2492 + }, + { + "epoch": 0.92, + "learning_rate": 2.9097302623475797e-06, + "loss": 0.6301, + "step": 2493 + }, + { + "epoch": 0.92, + "learning_rate": 2.9096933119842343e-06, + "loss": 0.6325, + "step": 2494 + }, + { + "epoch": 0.92, + "learning_rate": 2.9096563616208894e-06, + "loss": 0.6135, + "step": 2495 + }, + { + "epoch": 0.92, + "learning_rate": 2.909619411257544e-06, + "loss": 0.6177, + "step": 2496 + }, + { + "epoch": 0.92, + "learning_rate": 2.909582460894199e-06, + "loss": 0.719, + "step": 2497 + }, + { + "epoch": 0.92, + "learning_rate": 2.9095455105308537e-06, + "loss": 0.5745, + "step": 2498 + }, + { + "epoch": 0.92, + "learning_rate": 2.9095085601675084e-06, + "loss": 0.6023, + "step": 2499 + }, + { + "epoch": 0.92, + "learning_rate": 2.909471609804163e-06, + "loss": 0.6258, + "step": 2500 + }, + { + "epoch": 0.92, + "learning_rate": 2.9094346594408176e-06, + "loss": 0.6326, + "step": 2501 + }, + { + "epoch": 0.92, + "learning_rate": 2.9093977090774727e-06, + "loss": 0.6718, + "step": 2502 + }, + { + "epoch": 0.92, + "learning_rate": 2.9093607587141273e-06, + "loss": 0.6763, + "step": 2503 + }, + { + "epoch": 0.92, + "learning_rate": 2.9093238083507824e-06, + "loss": 0.65, + "step": 2504 + }, + { + "epoch": 0.93, + "learning_rate": 2.909286857987437e-06, + "loss": 0.6042, + "step": 2505 + }, + { + "epoch": 0.93, + "learning_rate": 2.9092499076240917e-06, + "loss": 0.6517, + "step": 2506 + }, + { + "epoch": 0.93, + "learning_rate": 2.9092129572607463e-06, + "loss": 0.6088, + "step": 2507 + }, + { + "epoch": 0.93, + "learning_rate": 2.9091760068974014e-06, + "loss": 0.6363, + "step": 2508 + }, + { + "epoch": 0.93, + "learning_rate": 2.909139056534056e-06, + "loss": 0.6417, + "step": 2509 + }, + { + "epoch": 0.93, + "learning_rate": 2.9091021061707106e-06, + "loss": 0.6724, + "step": 2510 + }, + { + "epoch": 0.93, + "learning_rate": 2.9090651558073657e-06, + "loss": 0.6007, + "step": 2511 + }, + { + "epoch": 0.93, + "learning_rate": 2.9090282054440203e-06, + "loss": 0.7065, + "step": 2512 + }, + { + "epoch": 0.93, + "learning_rate": 2.908991255080675e-06, + "loss": 0.5933, + "step": 2513 + }, + { + "epoch": 0.93, + "learning_rate": 2.9089543047173296e-06, + "loss": 0.6981, + "step": 2514 + }, + { + "epoch": 0.93, + "learning_rate": 2.9089173543539847e-06, + "loss": 0.6826, + "step": 2515 + }, + { + "epoch": 0.93, + "learning_rate": 2.9088804039906393e-06, + "loss": 0.6235, + "step": 2516 + }, + { + "epoch": 0.93, + "learning_rate": 2.9088434536272944e-06, + "loss": 0.6504, + "step": 2517 + }, + { + "epoch": 0.93, + "learning_rate": 2.908806503263949e-06, + "loss": 0.6343, + "step": 2518 + }, + { + "epoch": 0.93, + "learning_rate": 2.9087695529006036e-06, + "loss": 0.6394, + "step": 2519 + }, + { + "epoch": 0.93, + "learning_rate": 2.9087326025372583e-06, + "loss": 0.7111, + "step": 2520 + }, + { + "epoch": 0.93, + "learning_rate": 2.908695652173913e-06, + "loss": 0.5775, + "step": 2521 + }, + { + "epoch": 0.93, + "learning_rate": 2.908658701810568e-06, + "loss": 0.6261, + "step": 2522 + }, + { + "epoch": 0.93, + "learning_rate": 2.9086217514472226e-06, + "loss": 0.5766, + "step": 2523 + }, + { + "epoch": 0.93, + "learning_rate": 2.9085848010838777e-06, + "loss": 0.6153, + "step": 2524 + }, + { + "epoch": 0.93, + "learning_rate": 2.908547850720532e-06, + "loss": 0.6232, + "step": 2525 + }, + { + "epoch": 0.93, + "learning_rate": 2.908510900357187e-06, + "loss": 0.6092, + "step": 2526 + }, + { + "epoch": 0.93, + "learning_rate": 2.9084739499938416e-06, + "loss": 0.6337, + "step": 2527 + }, + { + "epoch": 0.93, + "learning_rate": 2.9084369996304966e-06, + "loss": 0.6103, + "step": 2528 + }, + { + "epoch": 0.93, + "learning_rate": 2.9084000492671513e-06, + "loss": 0.5973, + "step": 2529 + }, + { + "epoch": 0.93, + "learning_rate": 2.908363098903806e-06, + "loss": 0.5725, + "step": 2530 + }, + { + "epoch": 0.93, + "learning_rate": 2.9083261485404605e-06, + "loss": 0.6345, + "step": 2531 + }, + { + "epoch": 0.94, + "learning_rate": 2.908289198177115e-06, + "loss": 0.5863, + "step": 2532 + }, + { + "epoch": 0.94, + "learning_rate": 2.9082522478137702e-06, + "loss": 0.6329, + "step": 2533 + }, + { + "epoch": 0.94, + "learning_rate": 2.908215297450425e-06, + "loss": 0.6427, + "step": 2534 + }, + { + "epoch": 0.94, + "learning_rate": 2.90817834708708e-06, + "loss": 0.6446, + "step": 2535 + }, + { + "epoch": 0.94, + "learning_rate": 2.9081413967237346e-06, + "loss": 0.6398, + "step": 2536 + }, + { + "epoch": 0.94, + "learning_rate": 2.908104446360389e-06, + "loss": 0.5834, + "step": 2537 + }, + { + "epoch": 0.94, + "learning_rate": 2.908067495997044e-06, + "loss": 0.6126, + "step": 2538 + }, + { + "epoch": 0.94, + "learning_rate": 2.908030545633699e-06, + "loss": 0.5902, + "step": 2539 + }, + { + "epoch": 0.94, + "learning_rate": 2.9079935952703535e-06, + "loss": 0.6218, + "step": 2540 + }, + { + "epoch": 0.94, + "learning_rate": 2.907956644907008e-06, + "loss": 0.7149, + "step": 2541 + }, + { + "epoch": 0.94, + "learning_rate": 2.9079196945436632e-06, + "loss": 0.6603, + "step": 2542 + }, + { + "epoch": 0.94, + "learning_rate": 2.907882744180318e-06, + "loss": 0.6247, + "step": 2543 + }, + { + "epoch": 0.94, + "learning_rate": 2.9078457938169725e-06, + "loss": 0.6734, + "step": 2544 + }, + { + "epoch": 0.94, + "learning_rate": 2.907808843453627e-06, + "loss": 0.664, + "step": 2545 + }, + { + "epoch": 0.94, + "learning_rate": 2.907771893090282e-06, + "loss": 0.633, + "step": 2546 + }, + { + "epoch": 0.94, + "learning_rate": 2.907734942726937e-06, + "loss": 0.6251, + "step": 2547 + }, + { + "epoch": 0.94, + "learning_rate": 2.907697992363592e-06, + "loss": 0.6551, + "step": 2548 + }, + { + "epoch": 0.94, + "learning_rate": 2.9076610420002465e-06, + "loss": 0.6642, + "step": 2549 + }, + { + "epoch": 0.94, + "learning_rate": 2.907624091636901e-06, + "loss": 0.5967, + "step": 2550 + }, + { + "epoch": 0.94, + "learning_rate": 2.907587141273556e-06, + "loss": 0.5834, + "step": 2551 + }, + { + "epoch": 0.94, + "learning_rate": 2.9075501909102104e-06, + "loss": 0.5637, + "step": 2552 + }, + { + "epoch": 0.94, + "learning_rate": 2.9075132405468655e-06, + "loss": 0.6825, + "step": 2553 + }, + { + "epoch": 0.94, + "learning_rate": 2.90747629018352e-06, + "loss": 0.6089, + "step": 2554 + }, + { + "epoch": 0.94, + "learning_rate": 2.907439339820175e-06, + "loss": 0.653, + "step": 2555 + }, + { + "epoch": 0.94, + "learning_rate": 2.90740238945683e-06, + "loss": 0.6781, + "step": 2556 + }, + { + "epoch": 0.94, + "learning_rate": 2.9073654390934845e-06, + "loss": 0.5882, + "step": 2557 + }, + { + "epoch": 0.94, + "learning_rate": 2.907328488730139e-06, + "loss": 0.6492, + "step": 2558 + }, + { + "epoch": 0.94, + "learning_rate": 2.907291538366794e-06, + "loss": 0.6495, + "step": 2559 + }, + { + "epoch": 0.95, + "learning_rate": 2.907254588003449e-06, + "loss": 0.5876, + "step": 2560 + }, + { + "epoch": 0.95, + "learning_rate": 2.9072176376401034e-06, + "loss": 0.6089, + "step": 2561 + }, + { + "epoch": 0.95, + "learning_rate": 2.9071806872767585e-06, + "loss": 0.5879, + "step": 2562 + }, + { + "epoch": 0.95, + "learning_rate": 2.907143736913413e-06, + "loss": 0.6635, + "step": 2563 + }, + { + "epoch": 0.95, + "learning_rate": 2.9071067865500678e-06, + "loss": 0.6339, + "step": 2564 + }, + { + "epoch": 0.95, + "learning_rate": 2.9070698361867224e-06, + "loss": 0.6402, + "step": 2565 + }, + { + "epoch": 0.95, + "learning_rate": 2.9070328858233775e-06, + "loss": 0.6965, + "step": 2566 + }, + { + "epoch": 0.95, + "learning_rate": 2.906995935460032e-06, + "loss": 0.6542, + "step": 2567 + }, + { + "epoch": 0.95, + "learning_rate": 2.906958985096687e-06, + "loss": 0.6215, + "step": 2568 + }, + { + "epoch": 0.95, + "learning_rate": 2.906922034733342e-06, + "loss": 0.5736, + "step": 2569 + }, + { + "epoch": 0.95, + "learning_rate": 2.906885084369996e-06, + "loss": 0.6567, + "step": 2570 + }, + { + "epoch": 0.95, + "learning_rate": 2.906848134006651e-06, + "loss": 0.6412, + "step": 2571 + }, + { + "epoch": 0.95, + "learning_rate": 2.9068111836433057e-06, + "loss": 0.6058, + "step": 2572 + }, + { + "epoch": 0.95, + "learning_rate": 2.9067742332799608e-06, + "loss": 0.6328, + "step": 2573 + }, + { + "epoch": 0.95, + "learning_rate": 2.9067372829166154e-06, + "loss": 0.6342, + "step": 2574 + }, + { + "epoch": 0.95, + "learning_rate": 2.9067003325532705e-06, + "loss": 0.6159, + "step": 2575 + }, + { + "epoch": 0.95, + "learning_rate": 2.9066633821899247e-06, + "loss": 0.6664, + "step": 2576 + }, + { + "epoch": 0.95, + "learning_rate": 2.9066264318265797e-06, + "loss": 0.6131, + "step": 2577 + }, + { + "epoch": 0.95, + "learning_rate": 2.9065894814632344e-06, + "loss": 0.6702, + "step": 2578 + }, + { + "epoch": 0.95, + "learning_rate": 2.9065525310998894e-06, + "loss": 0.6469, + "step": 2579 + }, + { + "epoch": 0.95, + "learning_rate": 2.906515580736544e-06, + "loss": 0.61, + "step": 2580 + }, + { + "epoch": 0.95, + "learning_rate": 2.9064786303731987e-06, + "loss": 0.6748, + "step": 2581 + }, + { + "epoch": 0.95, + "learning_rate": 2.9064416800098534e-06, + "loss": 0.5967, + "step": 2582 + }, + { + "epoch": 0.95, + "learning_rate": 2.906404729646508e-06, + "loss": 0.6654, + "step": 2583 + }, + { + "epoch": 0.95, + "learning_rate": 2.906367779283163e-06, + "loss": 0.6444, + "step": 2584 + }, + { + "epoch": 0.95, + "learning_rate": 2.9063308289198177e-06, + "loss": 0.5949, + "step": 2585 + }, + { + "epoch": 0.95, + "learning_rate": 2.9062938785564727e-06, + "loss": 0.6553, + "step": 2586 + }, + { + "epoch": 0.96, + "learning_rate": 2.9062569281931274e-06, + "loss": 0.5995, + "step": 2587 + }, + { + "epoch": 0.96, + "learning_rate": 2.906219977829782e-06, + "loss": 0.628, + "step": 2588 + }, + { + "epoch": 0.96, + "learning_rate": 2.9061830274664367e-06, + "loss": 0.5679, + "step": 2589 + }, + { + "epoch": 0.96, + "learning_rate": 2.9061460771030913e-06, + "loss": 0.6453, + "step": 2590 + }, + { + "epoch": 0.96, + "learning_rate": 2.9061091267397464e-06, + "loss": 0.6435, + "step": 2591 + }, + { + "epoch": 0.96, + "learning_rate": 2.906072176376401e-06, + "loss": 0.647, + "step": 2592 + }, + { + "epoch": 0.96, + "learning_rate": 2.906035226013056e-06, + "loss": 0.6316, + "step": 2593 + }, + { + "epoch": 0.96, + "learning_rate": 2.9059982756497107e-06, + "loss": 0.6845, + "step": 2594 + }, + { + "epoch": 0.96, + "learning_rate": 2.9059613252863653e-06, + "loss": 0.6342, + "step": 2595 + }, + { + "epoch": 0.96, + "learning_rate": 2.90592437492302e-06, + "loss": 0.6477, + "step": 2596 + }, + { + "epoch": 0.96, + "learning_rate": 2.905887424559675e-06, + "loss": 0.6111, + "step": 2597 + }, + { + "epoch": 0.96, + "learning_rate": 2.9058504741963297e-06, + "loss": 0.6157, + "step": 2598 + }, + { + "epoch": 0.96, + "learning_rate": 2.9058135238329847e-06, + "loss": 0.6153, + "step": 2599 + }, + { + "epoch": 0.96, + "learning_rate": 2.9057765734696393e-06, + "loss": 0.6003, + "step": 2600 + }, + { + "epoch": 0.96, + "learning_rate": 2.905739623106294e-06, + "loss": 0.6779, + "step": 2601 + }, + { + "epoch": 0.96, + "learning_rate": 2.9057026727429486e-06, + "loss": 0.6132, + "step": 2602 + }, + { + "epoch": 0.96, + "learning_rate": 2.9056657223796033e-06, + "loss": 0.7066, + "step": 2603 + }, + { + "epoch": 0.96, + "learning_rate": 2.9056287720162583e-06, + "loss": 0.6847, + "step": 2604 + }, + { + "epoch": 0.96, + "learning_rate": 2.905591821652913e-06, + "loss": 0.6245, + "step": 2605 + }, + { + "epoch": 0.96, + "learning_rate": 2.905554871289568e-06, + "loss": 0.6946, + "step": 2606 + }, + { + "epoch": 0.96, + "learning_rate": 2.9055179209262227e-06, + "loss": 0.6305, + "step": 2607 + }, + { + "epoch": 0.96, + "learning_rate": 2.9054809705628773e-06, + "loss": 0.5846, + "step": 2608 + }, + { + "epoch": 0.96, + "learning_rate": 2.905444020199532e-06, + "loss": 0.6372, + "step": 2609 + }, + { + "epoch": 0.96, + "learning_rate": 2.9054070698361866e-06, + "loss": 0.6575, + "step": 2610 + }, + { + "epoch": 0.96, + "learning_rate": 2.9053701194728416e-06, + "loss": 0.667, + "step": 2611 + }, + { + "epoch": 0.96, + "learning_rate": 2.9053331691094963e-06, + "loss": 0.6509, + "step": 2612 + }, + { + "epoch": 0.96, + "learning_rate": 2.9052962187461513e-06, + "loss": 0.6235, + "step": 2613 + }, + { + "epoch": 0.97, + "learning_rate": 2.905259268382806e-06, + "loss": 0.5888, + "step": 2614 + }, + { + "epoch": 0.97, + "learning_rate": 2.9052223180194606e-06, + "loss": 0.6678, + "step": 2615 + }, + { + "epoch": 0.97, + "learning_rate": 2.9051853676561152e-06, + "loss": 0.6912, + "step": 2616 + }, + { + "epoch": 0.97, + "learning_rate": 2.9051484172927703e-06, + "loss": 0.6356, + "step": 2617 + }, + { + "epoch": 0.97, + "learning_rate": 2.905111466929425e-06, + "loss": 0.6421, + "step": 2618 + }, + { + "epoch": 0.97, + "learning_rate": 2.9050745165660796e-06, + "loss": 0.6116, + "step": 2619 + }, + { + "epoch": 0.97, + "learning_rate": 2.9050375662027346e-06, + "loss": 0.5981, + "step": 2620 + }, + { + "epoch": 0.97, + "learning_rate": 2.905000615839389e-06, + "loss": 0.6206, + "step": 2621 + }, + { + "epoch": 0.97, + "learning_rate": 2.904963665476044e-06, + "loss": 0.6395, + "step": 2622 + }, + { + "epoch": 0.97, + "learning_rate": 2.9049267151126985e-06, + "loss": 0.5553, + "step": 2623 + }, + { + "epoch": 0.97, + "learning_rate": 2.9048897647493536e-06, + "loss": 0.6408, + "step": 2624 + }, + { + "epoch": 0.97, + "learning_rate": 2.9048528143860082e-06, + "loss": 0.649, + "step": 2625 + }, + { + "epoch": 0.97, + "learning_rate": 2.9048158640226633e-06, + "loss": 0.6786, + "step": 2626 + }, + { + "epoch": 0.97, + "learning_rate": 2.9047789136593175e-06, + "loss": 0.663, + "step": 2627 + }, + { + "epoch": 0.97, + "learning_rate": 2.9047419632959726e-06, + "loss": 0.5925, + "step": 2628 + }, + { + "epoch": 0.97, + "learning_rate": 2.904705012932627e-06, + "loss": 0.6624, + "step": 2629 + }, + { + "epoch": 0.97, + "learning_rate": 2.904668062569282e-06, + "loss": 0.6451, + "step": 2630 + }, + { + "epoch": 0.97, + "learning_rate": 2.904631112205937e-06, + "loss": 0.6472, + "step": 2631 + }, + { + "epoch": 0.97, + "learning_rate": 2.9045941618425915e-06, + "loss": 0.6506, + "step": 2632 + }, + { + "epoch": 0.97, + "learning_rate": 2.904557211479246e-06, + "loss": 0.6393, + "step": 2633 + }, + { + "epoch": 0.97, + "learning_rate": 2.904520261115901e-06, + "loss": 0.6431, + "step": 2634 + }, + { + "epoch": 0.97, + "learning_rate": 2.904483310752556e-06, + "loss": 0.5956, + "step": 2635 + }, + { + "epoch": 0.97, + "learning_rate": 2.9044463603892105e-06, + "loss": 0.6155, + "step": 2636 + }, + { + "epoch": 0.97, + "learning_rate": 2.9044094100258656e-06, + "loss": 0.6318, + "step": 2637 + }, + { + "epoch": 0.97, + "learning_rate": 2.90437245966252e-06, + "loss": 0.6657, + "step": 2638 + }, + { + "epoch": 0.97, + "learning_rate": 2.904335509299175e-06, + "loss": 0.6394, + "step": 2639 + }, + { + "epoch": 0.97, + "learning_rate": 2.9042985589358295e-06, + "loss": 0.6494, + "step": 2640 + }, + { + "epoch": 0.98, + "learning_rate": 2.904261608572484e-06, + "loss": 0.6403, + "step": 2641 + }, + { + "epoch": 0.98, + "learning_rate": 2.904224658209139e-06, + "loss": 0.6056, + "step": 2642 + }, + { + "epoch": 0.98, + "learning_rate": 2.904187707845794e-06, + "loss": 0.644, + "step": 2643 + }, + { + "epoch": 0.98, + "learning_rate": 2.904150757482449e-06, + "loss": 0.6219, + "step": 2644 + }, + { + "epoch": 0.98, + "learning_rate": 2.9041138071191035e-06, + "loss": 0.6352, + "step": 2645 + }, + { + "epoch": 0.98, + "learning_rate": 2.904076856755758e-06, + "loss": 0.5907, + "step": 2646 + }, + { + "epoch": 0.98, + "learning_rate": 2.9040399063924128e-06, + "loss": 0.6433, + "step": 2647 + }, + { + "epoch": 0.98, + "learning_rate": 2.904002956029068e-06, + "loss": 0.6398, + "step": 2648 + }, + { + "epoch": 0.98, + "learning_rate": 2.9039660056657225e-06, + "loss": 0.5882, + "step": 2649 + }, + { + "epoch": 0.98, + "learning_rate": 2.903929055302377e-06, + "loss": 0.6974, + "step": 2650 + }, + { + "epoch": 0.98, + "learning_rate": 2.903892104939032e-06, + "loss": 0.6534, + "step": 2651 + }, + { + "epoch": 0.98, + "learning_rate": 2.903855154575687e-06, + "loss": 0.6398, + "step": 2652 + }, + { + "epoch": 0.98, + "learning_rate": 2.9038182042123414e-06, + "loss": 0.6151, + "step": 2653 + }, + { + "epoch": 0.98, + "learning_rate": 2.903781253848996e-06, + "loss": 0.5724, + "step": 2654 + }, + { + "epoch": 0.98, + "learning_rate": 2.903744303485651e-06, + "loss": 0.6083, + "step": 2655 + }, + { + "epoch": 0.98, + "learning_rate": 2.9037073531223058e-06, + "loss": 0.6309, + "step": 2656 + }, + { + "epoch": 0.98, + "learning_rate": 2.903670402758961e-06, + "loss": 0.5815, + "step": 2657 + }, + { + "epoch": 0.98, + "learning_rate": 2.9036334523956155e-06, + "loss": 0.5847, + "step": 2658 + }, + { + "epoch": 0.98, + "learning_rate": 2.90359650203227e-06, + "loss": 0.6043, + "step": 2659 + }, + { + "epoch": 0.98, + "learning_rate": 2.9035595516689247e-06, + "loss": 0.558, + "step": 2660 + }, + { + "epoch": 0.98, + "learning_rate": 2.9035226013055794e-06, + "loss": 0.6811, + "step": 2661 + }, + { + "epoch": 0.98, + "learning_rate": 2.9034856509422344e-06, + "loss": 0.6164, + "step": 2662 + }, + { + "epoch": 0.98, + "learning_rate": 2.903448700578889e-06, + "loss": 0.6816, + "step": 2663 + }, + { + "epoch": 0.98, + "learning_rate": 2.903411750215544e-06, + "loss": 0.6914, + "step": 2664 + }, + { + "epoch": 0.98, + "learning_rate": 2.9033747998521988e-06, + "loss": 0.6773, + "step": 2665 + }, + { + "epoch": 0.98, + "learning_rate": 2.9033378494888534e-06, + "loss": 0.6516, + "step": 2666 + }, + { + "epoch": 0.98, + "learning_rate": 2.903300899125508e-06, + "loss": 0.658, + "step": 2667 + }, + { + "epoch": 0.99, + "learning_rate": 2.903263948762163e-06, + "loss": 0.6018, + "step": 2668 + }, + { + "epoch": 0.99, + "learning_rate": 2.9032269983988177e-06, + "loss": 0.6863, + "step": 2669 + }, + { + "epoch": 0.99, + "learning_rate": 2.9031900480354724e-06, + "loss": 0.6231, + "step": 2670 + }, + { + "epoch": 0.99, + "learning_rate": 2.9031530976721274e-06, + "loss": 0.5921, + "step": 2671 + }, + { + "epoch": 0.99, + "learning_rate": 2.9031161473087816e-06, + "loss": 0.571, + "step": 2672 + }, + { + "epoch": 0.99, + "learning_rate": 2.9030791969454367e-06, + "loss": 0.6676, + "step": 2673 + }, + { + "epoch": 0.99, + "learning_rate": 2.9030422465820913e-06, + "loss": 0.6478, + "step": 2674 + }, + { + "epoch": 0.99, + "learning_rate": 2.9030052962187464e-06, + "loss": 0.6057, + "step": 2675 + }, + { + "epoch": 0.99, + "learning_rate": 2.902968345855401e-06, + "loss": 0.666, + "step": 2676 + }, + { + "epoch": 0.99, + "learning_rate": 2.9029313954920557e-06, + "loss": 0.6402, + "step": 2677 + }, + { + "epoch": 0.99, + "learning_rate": 2.9028944451287103e-06, + "loss": 0.6835, + "step": 2678 + }, + { + "epoch": 0.99, + "learning_rate": 2.902857494765365e-06, + "loss": 0.6275, + "step": 2679 + }, + { + "epoch": 0.99, + "learning_rate": 2.90282054440202e-06, + "loss": 0.5998, + "step": 2680 + }, + { + "epoch": 0.99, + "learning_rate": 2.9027835940386746e-06, + "loss": 0.6623, + "step": 2681 + }, + { + "epoch": 0.99, + "learning_rate": 2.9027466436753297e-06, + "loss": 0.5697, + "step": 2682 + }, + { + "epoch": 0.99, + "learning_rate": 2.9027096933119843e-06, + "loss": 0.6055, + "step": 2683 + }, + { + "epoch": 0.99, + "learning_rate": 2.902672742948639e-06, + "loss": 0.6238, + "step": 2684 + }, + { + "epoch": 0.99, + "learning_rate": 2.9026357925852936e-06, + "loss": 0.6059, + "step": 2685 + }, + { + "epoch": 0.99, + "learning_rate": 2.9025988422219487e-06, + "loss": 0.6398, + "step": 2686 + }, + { + "epoch": 0.99, + "learning_rate": 2.9025618918586033e-06, + "loss": 0.6369, + "step": 2687 + }, + { + "epoch": 0.99, + "learning_rate": 2.9025249414952584e-06, + "loss": 0.6262, + "step": 2688 + }, + { + "epoch": 0.99, + "learning_rate": 2.902487991131913e-06, + "loss": 0.6149, + "step": 2689 + }, + { + "epoch": 0.99, + "learning_rate": 2.9024510407685676e-06, + "loss": 0.5325, + "step": 2690 + }, + { + "epoch": 0.99, + "learning_rate": 2.9024140904052223e-06, + "loss": 0.6564, + "step": 2691 + }, + { + "epoch": 0.99, + "learning_rate": 2.902377140041877e-06, + "loss": 0.6901, + "step": 2692 + }, + { + "epoch": 0.99, + "learning_rate": 2.902340189678532e-06, + "loss": 0.5815, + "step": 2693 + }, + { + "epoch": 0.99, + "learning_rate": 2.9023032393151866e-06, + "loss": 0.6068, + "step": 2694 + }, + { + "epoch": 1.0, + "learning_rate": 2.9022662889518417e-06, + "loss": 0.6507, + "step": 2695 + }, + { + "epoch": 1.0, + "learning_rate": 2.9022293385884963e-06, + "loss": 0.6568, + "step": 2696 + }, + { + "epoch": 1.0, + "learning_rate": 2.902192388225151e-06, + "loss": 0.6488, + "step": 2697 + }, + { + "epoch": 1.0, + "learning_rate": 2.9021554378618056e-06, + "loss": 0.6871, + "step": 2698 + }, + { + "epoch": 1.0, + "learning_rate": 2.9021184874984602e-06, + "loss": 0.6279, + "step": 2699 + }, + { + "epoch": 1.0, + "learning_rate": 2.9020815371351153e-06, + "loss": 0.6582, + "step": 2700 + }, + { + "epoch": 1.0, + "learning_rate": 2.90204458677177e-06, + "loss": 0.5906, + "step": 2701 + }, + { + "epoch": 1.0, + "learning_rate": 2.902007636408425e-06, + "loss": 0.6808, + "step": 2702 + }, + { + "epoch": 1.0, + "learning_rate": 2.9019706860450796e-06, + "loss": 0.6254, + "step": 2703 + }, + { + "epoch": 1.0, + "learning_rate": 2.9019337356817342e-06, + "loss": 0.6424, + "step": 2704 + }, + { + "epoch": 1.0, + "learning_rate": 2.901896785318389e-06, + "loss": 0.6992, + "step": 2705 + }, + { + "epoch": 1.0, + "learning_rate": 2.901859834955044e-06, + "loss": 0.651, + "step": 2706 + }, + { + "epoch": 1.0, + "learning_rate": 2.9018228845916986e-06, + "loss": 0.6397, + "step": 2707 + }, + { + "epoch": 1.0, + "learning_rate": 2.9017859342283536e-06, + "loss": 0.4828, + "step": 2708 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.6354296532299025, + "eval_loss": 0.6342946887016296, + "eval_runtime": 26.6934, + "eval_samples_per_second": 1622.649, + "eval_steps_per_second": 25.362, + "step": 2708 + }, + { + "epoch": 1.0, + "learning_rate": 2.9017489838650083e-06, + "loss": 0.606, + "step": 2709 + }, + { + "epoch": 1.0, + "learning_rate": 2.9017120335016625e-06, + "loss": 0.6444, + "step": 2710 + }, + { + "epoch": 1.0, + "learning_rate": 2.9016750831383175e-06, + "loss": 0.6178, + "step": 2711 + }, + { + "epoch": 1.0, + "learning_rate": 2.901638132774972e-06, + "loss": 0.6113, + "step": 2712 + }, + { + "epoch": 1.0, + "learning_rate": 2.9016011824116272e-06, + "loss": 0.5956, + "step": 2713 + }, + { + "epoch": 1.0, + "learning_rate": 2.901564232048282e-06, + "loss": 0.6582, + "step": 2714 + }, + { + "epoch": 1.0, + "learning_rate": 2.901527281684937e-06, + "loss": 0.594, + "step": 2715 + }, + { + "epoch": 1.0, + "learning_rate": 2.901490331321591e-06, + "loss": 0.6475, + "step": 2716 + }, + { + "epoch": 1.0, + "learning_rate": 2.9014533809582462e-06, + "loss": 0.5495, + "step": 2717 + }, + { + "epoch": 1.0, + "learning_rate": 2.901416430594901e-06, + "loss": 0.6086, + "step": 2718 + }, + { + "epoch": 1.0, + "learning_rate": 2.9013794802315555e-06, + "loss": 0.5846, + "step": 2719 + }, + { + "epoch": 1.0, + "learning_rate": 2.9013425298682105e-06, + "loss": 0.7024, + "step": 2720 + }, + { + "epoch": 1.0, + "learning_rate": 2.901305579504865e-06, + "loss": 0.6374, + "step": 2721 + }, + { + "epoch": 1.01, + "learning_rate": 2.90126862914152e-06, + "loss": 0.5816, + "step": 2722 + }, + { + "epoch": 1.01, + "learning_rate": 2.9012316787781745e-06, + "loss": 0.6918, + "step": 2723 + }, + { + "epoch": 1.01, + "learning_rate": 2.9011947284148295e-06, + "loss": 0.5992, + "step": 2724 + }, + { + "epoch": 1.01, + "learning_rate": 2.901157778051484e-06, + "loss": 0.6218, + "step": 2725 + }, + { + "epoch": 1.01, + "learning_rate": 2.901120827688139e-06, + "loss": 0.6068, + "step": 2726 + }, + { + "epoch": 1.01, + "learning_rate": 2.901083877324794e-06, + "loss": 0.5932, + "step": 2727 + }, + { + "epoch": 1.01, + "learning_rate": 2.9010469269614485e-06, + "loss": 0.6044, + "step": 2728 + }, + { + "epoch": 1.01, + "learning_rate": 2.901009976598103e-06, + "loss": 0.6614, + "step": 2729 + }, + { + "epoch": 1.01, + "learning_rate": 2.9009730262347578e-06, + "loss": 0.6677, + "step": 2730 + }, + { + "epoch": 1.01, + "learning_rate": 2.900936075871413e-06, + "loss": 0.6, + "step": 2731 + }, + { + "epoch": 1.01, + "learning_rate": 2.9008991255080675e-06, + "loss": 0.673, + "step": 2732 + }, + { + "epoch": 1.01, + "learning_rate": 2.9008621751447225e-06, + "loss": 0.634, + "step": 2733 + }, + { + "epoch": 1.01, + "learning_rate": 2.900825224781377e-06, + "loss": 0.5724, + "step": 2734 + }, + { + "epoch": 1.01, + "learning_rate": 2.9007882744180318e-06, + "loss": 0.606, + "step": 2735 + }, + { + "epoch": 1.01, + "learning_rate": 2.9007513240546864e-06, + "loss": 0.7043, + "step": 2736 + }, + { + "epoch": 1.01, + "learning_rate": 2.9007143736913415e-06, + "loss": 0.6409, + "step": 2737 + }, + { + "epoch": 1.01, + "learning_rate": 2.900677423327996e-06, + "loss": 0.5874, + "step": 2738 + }, + { + "epoch": 1.01, + "learning_rate": 2.9006404729646508e-06, + "loss": 0.6948, + "step": 2739 + }, + { + "epoch": 1.01, + "learning_rate": 2.900603522601306e-06, + "loss": 0.7038, + "step": 2740 + }, + { + "epoch": 1.01, + "learning_rate": 2.9005665722379605e-06, + "loss": 0.6881, + "step": 2741 + }, + { + "epoch": 1.01, + "learning_rate": 2.900529621874615e-06, + "loss": 0.6852, + "step": 2742 + }, + { + "epoch": 1.01, + "learning_rate": 2.9004926715112697e-06, + "loss": 0.6214, + "step": 2743 + }, + { + "epoch": 1.01, + "learning_rate": 2.9004557211479248e-06, + "loss": 0.594, + "step": 2744 + }, + { + "epoch": 1.01, + "learning_rate": 2.9004187707845794e-06, + "loss": 0.669, + "step": 2745 + }, + { + "epoch": 1.01, + "learning_rate": 2.9003818204212345e-06, + "loss": 0.6077, + "step": 2746 + }, + { + "epoch": 1.01, + "learning_rate": 2.900344870057889e-06, + "loss": 0.6328, + "step": 2747 + }, + { + "epoch": 1.01, + "learning_rate": 2.9003079196945438e-06, + "loss": 0.6512, + "step": 2748 + }, + { + "epoch": 1.02, + "learning_rate": 2.9002709693311984e-06, + "loss": 0.6258, + "step": 2749 + }, + { + "epoch": 1.02, + "learning_rate": 2.900234018967853e-06, + "loss": 0.66, + "step": 2750 + }, + { + "epoch": 1.02, + "learning_rate": 2.900197068604508e-06, + "loss": 0.5837, + "step": 2751 + }, + { + "epoch": 1.02, + "learning_rate": 2.9001601182411627e-06, + "loss": 0.6142, + "step": 2752 + }, + { + "epoch": 1.02, + "learning_rate": 2.9001231678778178e-06, + "loss": 0.6206, + "step": 2753 + }, + { + "epoch": 1.02, + "learning_rate": 2.9000862175144724e-06, + "loss": 0.6794, + "step": 2754 + }, + { + "epoch": 1.02, + "learning_rate": 2.900049267151127e-06, + "loss": 0.6483, + "step": 2755 + }, + { + "epoch": 1.02, + "learning_rate": 2.9000123167877817e-06, + "loss": 0.6677, + "step": 2756 + }, + { + "epoch": 1.02, + "learning_rate": 2.8999753664244368e-06, + "loss": 0.6351, + "step": 2757 + }, + { + "epoch": 1.02, + "learning_rate": 2.8999384160610914e-06, + "loss": 0.6227, + "step": 2758 + }, + { + "epoch": 1.02, + "learning_rate": 2.899901465697746e-06, + "loss": 0.6023, + "step": 2759 + }, + { + "epoch": 1.02, + "learning_rate": 2.899864515334401e-06, + "loss": 0.6225, + "step": 2760 + }, + { + "epoch": 1.02, + "learning_rate": 2.8998275649710553e-06, + "loss": 0.6008, + "step": 2761 + }, + { + "epoch": 1.02, + "learning_rate": 2.8997906146077104e-06, + "loss": 0.661, + "step": 2762 + }, + { + "epoch": 1.02, + "learning_rate": 2.899753664244365e-06, + "loss": 0.6219, + "step": 2763 + }, + { + "epoch": 1.02, + "learning_rate": 2.89971671388102e-06, + "loss": 0.676, + "step": 2764 + }, + { + "epoch": 1.02, + "learning_rate": 2.8996797635176747e-06, + "loss": 0.6585, + "step": 2765 + }, + { + "epoch": 1.02, + "learning_rate": 2.8996428131543298e-06, + "loss": 0.6164, + "step": 2766 + }, + { + "epoch": 1.02, + "learning_rate": 2.899605862790984e-06, + "loss": 0.6487, + "step": 2767 + }, + { + "epoch": 1.02, + "learning_rate": 2.899568912427639e-06, + "loss": 0.5129, + "step": 2768 + }, + { + "epoch": 1.02, + "learning_rate": 2.8995319620642937e-06, + "loss": 0.6705, + "step": 2769 + }, + { + "epoch": 1.02, + "learning_rate": 2.8994950117009483e-06, + "loss": 0.6546, + "step": 2770 + }, + { + "epoch": 1.02, + "learning_rate": 2.8994580613376034e-06, + "loss": 0.6858, + "step": 2771 + }, + { + "epoch": 1.02, + "learning_rate": 2.899421110974258e-06, + "loss": 0.6139, + "step": 2772 + }, + { + "epoch": 1.02, + "learning_rate": 2.8993841606109126e-06, + "loss": 0.6068, + "step": 2773 + }, + { + "epoch": 1.02, + "learning_rate": 2.8993472102475673e-06, + "loss": 0.6555, + "step": 2774 + }, + { + "epoch": 1.02, + "learning_rate": 2.8993102598842223e-06, + "loss": 0.6672, + "step": 2775 + }, + { + "epoch": 1.03, + "learning_rate": 2.899273309520877e-06, + "loss": 0.6818, + "step": 2776 + }, + { + "epoch": 1.03, + "learning_rate": 2.899236359157532e-06, + "loss": 0.6086, + "step": 2777 + }, + { + "epoch": 1.03, + "learning_rate": 2.8991994087941867e-06, + "loss": 0.6536, + "step": 2778 + }, + { + "epoch": 1.03, + "learning_rate": 2.8991624584308413e-06, + "loss": 0.5438, + "step": 2779 + }, + { + "epoch": 1.03, + "learning_rate": 2.899125508067496e-06, + "loss": 0.6804, + "step": 2780 + }, + { + "epoch": 1.03, + "learning_rate": 2.8990885577041506e-06, + "loss": 0.5356, + "step": 2781 + }, + { + "epoch": 1.03, + "learning_rate": 2.8990516073408056e-06, + "loss": 0.669, + "step": 2782 + }, + { + "epoch": 1.03, + "learning_rate": 2.8990146569774603e-06, + "loss": 0.6419, + "step": 2783 + }, + { + "epoch": 1.03, + "learning_rate": 2.8989777066141153e-06, + "loss": 0.7351, + "step": 2784 + }, + { + "epoch": 1.03, + "learning_rate": 2.89894075625077e-06, + "loss": 0.6574, + "step": 2785 + }, + { + "epoch": 1.03, + "learning_rate": 2.8989038058874246e-06, + "loss": 0.6098, + "step": 2786 + }, + { + "epoch": 1.03, + "learning_rate": 2.8988668555240792e-06, + "loss": 0.6309, + "step": 2787 + }, + { + "epoch": 1.03, + "learning_rate": 2.8988299051607343e-06, + "loss": 0.649, + "step": 2788 + }, + { + "epoch": 1.03, + "learning_rate": 2.898792954797389e-06, + "loss": 0.6517, + "step": 2789 + }, + { + "epoch": 1.03, + "learning_rate": 2.8987560044340436e-06, + "loss": 0.6043, + "step": 2790 + }, + { + "epoch": 1.03, + "learning_rate": 2.8987190540706986e-06, + "loss": 0.652, + "step": 2791 + }, + { + "epoch": 1.03, + "learning_rate": 2.8986821037073533e-06, + "loss": 0.6266, + "step": 2792 + }, + { + "epoch": 1.03, + "learning_rate": 2.898645153344008e-06, + "loss": 0.5472, + "step": 2793 + }, + { + "epoch": 1.03, + "learning_rate": 2.8986082029806625e-06, + "loss": 0.6745, + "step": 2794 + }, + { + "epoch": 1.03, + "learning_rate": 2.8985712526173176e-06, + "loss": 0.7164, + "step": 2795 + }, + { + "epoch": 1.03, + "learning_rate": 2.8985343022539722e-06, + "loss": 0.5932, + "step": 2796 + }, + { + "epoch": 1.03, + "learning_rate": 2.8984973518906273e-06, + "loss": 0.6136, + "step": 2797 + }, + { + "epoch": 1.03, + "learning_rate": 2.898460401527282e-06, + "loss": 0.5501, + "step": 2798 + }, + { + "epoch": 1.03, + "learning_rate": 2.8984234511639366e-06, + "loss": 0.6486, + "step": 2799 + }, + { + "epoch": 1.03, + "learning_rate": 2.898386500800591e-06, + "loss": 0.6745, + "step": 2800 + }, + { + "epoch": 1.03, + "learning_rate": 2.898349550437246e-06, + "loss": 0.6071, + "step": 2801 + }, + { + "epoch": 1.03, + "learning_rate": 2.898312600073901e-06, + "loss": 0.6155, + "step": 2802 + }, + { + "epoch": 1.04, + "learning_rate": 2.8982756497105555e-06, + "loss": 0.6972, + "step": 2803 + }, + { + "epoch": 1.04, + "learning_rate": 2.8982386993472106e-06, + "loss": 0.6394, + "step": 2804 + }, + { + "epoch": 1.04, + "learning_rate": 2.8982017489838652e-06, + "loss": 0.7062, + "step": 2805 + }, + { + "epoch": 1.04, + "learning_rate": 2.89816479862052e-06, + "loss": 0.6518, + "step": 2806 + }, + { + "epoch": 1.04, + "learning_rate": 2.8981278482571745e-06, + "loss": 0.6377, + "step": 2807 + }, + { + "epoch": 1.04, + "learning_rate": 2.8980908978938296e-06, + "loss": 0.6514, + "step": 2808 + }, + { + "epoch": 1.04, + "learning_rate": 2.898053947530484e-06, + "loss": 0.7417, + "step": 2809 + }, + { + "epoch": 1.04, + "learning_rate": 2.898016997167139e-06, + "loss": 0.5286, + "step": 2810 + }, + { + "epoch": 1.04, + "learning_rate": 2.897980046803794e-06, + "loss": 0.6527, + "step": 2811 + }, + { + "epoch": 1.04, + "learning_rate": 2.897943096440448e-06, + "loss": 0.5646, + "step": 2812 + }, + { + "epoch": 1.04, + "learning_rate": 2.897906146077103e-06, + "loss": 0.6567, + "step": 2813 + }, + { + "epoch": 1.04, + "learning_rate": 2.897869195713758e-06, + "loss": 0.6035, + "step": 2814 + }, + { + "epoch": 1.04, + "learning_rate": 2.897832245350413e-06, + "loss": 0.5786, + "step": 2815 + }, + { + "epoch": 1.04, + "learning_rate": 2.8977952949870675e-06, + "loss": 0.5604, + "step": 2816 + }, + { + "epoch": 1.04, + "learning_rate": 2.8977583446237226e-06, + "loss": 0.6106, + "step": 2817 + }, + { + "epoch": 1.04, + "learning_rate": 2.8977213942603768e-06, + "loss": 0.67, + "step": 2818 + }, + { + "epoch": 1.04, + "learning_rate": 2.8976844438970314e-06, + "loss": 0.6389, + "step": 2819 + }, + { + "epoch": 1.04, + "learning_rate": 2.8976474935336865e-06, + "loss": 0.5909, + "step": 2820 + }, + { + "epoch": 1.04, + "learning_rate": 2.897610543170341e-06, + "loss": 0.6187, + "step": 2821 + }, + { + "epoch": 1.04, + "learning_rate": 2.897573592806996e-06, + "loss": 0.6212, + "step": 2822 + }, + { + "epoch": 1.04, + "learning_rate": 2.897536642443651e-06, + "loss": 0.57, + "step": 2823 + }, + { + "epoch": 1.04, + "learning_rate": 2.8974996920803054e-06, + "loss": 0.5844, + "step": 2824 + }, + { + "epoch": 1.04, + "learning_rate": 2.89746274171696e-06, + "loss": 0.5631, + "step": 2825 + }, + { + "epoch": 1.04, + "learning_rate": 2.897425791353615e-06, + "loss": 0.6698, + "step": 2826 + }, + { + "epoch": 1.04, + "learning_rate": 2.8973888409902698e-06, + "loss": 0.6819, + "step": 2827 + }, + { + "epoch": 1.04, + "learning_rate": 2.897351890626925e-06, + "loss": 0.6316, + "step": 2828 + }, + { + "epoch": 1.04, + "learning_rate": 2.8973149402635795e-06, + "loss": 0.6727, + "step": 2829 + }, + { + "epoch": 1.05, + "learning_rate": 2.897277989900234e-06, + "loss": 0.6014, + "step": 2830 + }, + { + "epoch": 1.05, + "learning_rate": 2.8972410395368887e-06, + "loss": 0.6189, + "step": 2831 + }, + { + "epoch": 1.05, + "learning_rate": 2.8972040891735434e-06, + "loss": 0.6184, + "step": 2832 + }, + { + "epoch": 1.05, + "learning_rate": 2.8971671388101984e-06, + "loss": 0.6533, + "step": 2833 + }, + { + "epoch": 1.05, + "learning_rate": 2.897130188446853e-06, + "loss": 0.6141, + "step": 2834 + }, + { + "epoch": 1.05, + "learning_rate": 2.897093238083508e-06, + "loss": 0.6794, + "step": 2835 + }, + { + "epoch": 1.05, + "learning_rate": 2.8970562877201628e-06, + "loss": 0.698, + "step": 2836 + }, + { + "epoch": 1.05, + "learning_rate": 2.8970193373568174e-06, + "loss": 0.6087, + "step": 2837 + }, + { + "epoch": 1.05, + "learning_rate": 2.896982386993472e-06, + "loss": 0.5872, + "step": 2838 + }, + { + "epoch": 1.05, + "learning_rate": 2.8969454366301267e-06, + "loss": 0.6707, + "step": 2839 + }, + { + "epoch": 1.05, + "learning_rate": 2.8969084862667817e-06, + "loss": 0.5599, + "step": 2840 + }, + { + "epoch": 1.05, + "learning_rate": 2.8968715359034364e-06, + "loss": 0.6989, + "step": 2841 + }, + { + "epoch": 1.05, + "learning_rate": 2.8968345855400914e-06, + "loss": 0.7086, + "step": 2842 + }, + { + "epoch": 1.05, + "learning_rate": 2.896797635176746e-06, + "loss": 0.6384, + "step": 2843 + }, + { + "epoch": 1.05, + "learning_rate": 2.8967606848134007e-06, + "loss": 0.5873, + "step": 2844 + }, + { + "epoch": 1.05, + "learning_rate": 2.8967237344500553e-06, + "loss": 0.6034, + "step": 2845 + }, + { + "epoch": 1.05, + "learning_rate": 2.8966867840867104e-06, + "loss": 0.6363, + "step": 2846 + }, + { + "epoch": 1.05, + "learning_rate": 2.896649833723365e-06, + "loss": 0.6513, + "step": 2847 + }, + { + "epoch": 1.05, + "learning_rate": 2.8966128833600197e-06, + "loss": 0.6533, + "step": 2848 + }, + { + "epoch": 1.05, + "learning_rate": 2.8965759329966747e-06, + "loss": 0.6825, + "step": 2849 + }, + { + "epoch": 1.05, + "learning_rate": 2.8965389826333294e-06, + "loss": 0.647, + "step": 2850 + }, + { + "epoch": 1.05, + "learning_rate": 2.896502032269984e-06, + "loss": 0.585, + "step": 2851 + }, + { + "epoch": 1.05, + "learning_rate": 2.8964650819066386e-06, + "loss": 0.6993, + "step": 2852 + }, + { + "epoch": 1.05, + "learning_rate": 2.8964281315432937e-06, + "loss": 0.6556, + "step": 2853 + }, + { + "epoch": 1.05, + "learning_rate": 2.8963911811799483e-06, + "loss": 0.5918, + "step": 2854 + }, + { + "epoch": 1.05, + "learning_rate": 2.8963542308166034e-06, + "loss": 0.6394, + "step": 2855 + }, + { + "epoch": 1.05, + "learning_rate": 2.896317280453258e-06, + "loss": 0.6096, + "step": 2856 + }, + { + "epoch": 1.06, + "learning_rate": 2.8962803300899127e-06, + "loss": 0.5496, + "step": 2857 + }, + { + "epoch": 1.06, + "learning_rate": 2.8962433797265673e-06, + "loss": 0.6731, + "step": 2858 + }, + { + "epoch": 1.06, + "learning_rate": 2.896206429363222e-06, + "loss": 0.5731, + "step": 2859 + }, + { + "epoch": 1.06, + "learning_rate": 2.896169478999877e-06, + "loss": 0.5838, + "step": 2860 + }, + { + "epoch": 1.06, + "learning_rate": 2.8961325286365316e-06, + "loss": 0.6345, + "step": 2861 + }, + { + "epoch": 1.06, + "learning_rate": 2.8960955782731867e-06, + "loss": 0.6634, + "step": 2862 + }, + { + "epoch": 1.06, + "learning_rate": 2.896058627909841e-06, + "loss": 0.6395, + "step": 2863 + }, + { + "epoch": 1.06, + "learning_rate": 2.896021677546496e-06, + "loss": 0.6418, + "step": 2864 + }, + { + "epoch": 1.06, + "learning_rate": 2.8959847271831506e-06, + "loss": 0.6964, + "step": 2865 + }, + { + "epoch": 1.06, + "learning_rate": 2.8959477768198057e-06, + "loss": 0.6731, + "step": 2866 + }, + { + "epoch": 1.06, + "learning_rate": 2.8959108264564603e-06, + "loss": 0.6133, + "step": 2867 + }, + { + "epoch": 1.06, + "learning_rate": 2.895873876093115e-06, + "loss": 0.5355, + "step": 2868 + }, + { + "epoch": 1.06, + "learning_rate": 2.8958369257297696e-06, + "loss": 0.603, + "step": 2869 + }, + { + "epoch": 1.06, + "learning_rate": 2.8957999753664242e-06, + "loss": 0.6555, + "step": 2870 + }, + { + "epoch": 1.06, + "learning_rate": 2.8957630250030793e-06, + "loss": 0.6478, + "step": 2871 + }, + { + "epoch": 1.06, + "learning_rate": 2.895726074639734e-06, + "loss": 0.5828, + "step": 2872 + }, + { + "epoch": 1.06, + "learning_rate": 2.895689124276389e-06, + "loss": 0.6346, + "step": 2873 + }, + { + "epoch": 1.06, + "learning_rate": 2.8956521739130436e-06, + "loss": 0.5945, + "step": 2874 + }, + { + "epoch": 1.06, + "learning_rate": 2.8956152235496983e-06, + "loss": 0.6694, + "step": 2875 + }, + { + "epoch": 1.06, + "learning_rate": 2.895578273186353e-06, + "loss": 0.6851, + "step": 2876 + }, + { + "epoch": 1.06, + "learning_rate": 2.895541322823008e-06, + "loss": 0.6633, + "step": 2877 + }, + { + "epoch": 1.06, + "learning_rate": 2.8955043724596626e-06, + "loss": 0.5853, + "step": 2878 + }, + { + "epoch": 1.06, + "learning_rate": 2.8954674220963172e-06, + "loss": 0.6075, + "step": 2879 + }, + { + "epoch": 1.06, + "learning_rate": 2.8954304717329723e-06, + "loss": 0.613, + "step": 2880 + }, + { + "epoch": 1.06, + "learning_rate": 2.895393521369627e-06, + "loss": 0.6352, + "step": 2881 + }, + { + "epoch": 1.06, + "learning_rate": 2.8953565710062816e-06, + "loss": 0.6216, + "step": 2882 + }, + { + "epoch": 1.06, + "learning_rate": 2.895319620642936e-06, + "loss": 0.7005, + "step": 2883 + }, + { + "epoch": 1.06, + "learning_rate": 2.8952826702795912e-06, + "loss": 0.6518, + "step": 2884 + }, + { + "epoch": 1.07, + "learning_rate": 2.895245719916246e-06, + "loss": 0.6657, + "step": 2885 + }, + { + "epoch": 1.07, + "learning_rate": 2.895208769552901e-06, + "loss": 0.6311, + "step": 2886 + }, + { + "epoch": 1.07, + "learning_rate": 2.8951718191895556e-06, + "loss": 0.6953, + "step": 2887 + }, + { + "epoch": 1.07, + "learning_rate": 2.8951348688262102e-06, + "loss": 0.6667, + "step": 2888 + }, + { + "epoch": 1.07, + "learning_rate": 2.895097918462865e-06, + "loss": 0.6183, + "step": 2889 + }, + { + "epoch": 1.07, + "learning_rate": 2.8950609680995195e-06, + "loss": 0.6477, + "step": 2890 + }, + { + "epoch": 1.07, + "learning_rate": 2.8950240177361746e-06, + "loss": 0.5517, + "step": 2891 + }, + { + "epoch": 1.07, + "learning_rate": 2.894987067372829e-06, + "loss": 0.6682, + "step": 2892 + }, + { + "epoch": 1.07, + "learning_rate": 2.8949501170094842e-06, + "loss": 0.6832, + "step": 2893 + }, + { + "epoch": 1.07, + "learning_rate": 2.894913166646139e-06, + "loss": 0.6791, + "step": 2894 + }, + { + "epoch": 1.07, + "learning_rate": 2.8948762162827935e-06, + "loss": 0.6981, + "step": 2895 + }, + { + "epoch": 1.07, + "learning_rate": 2.894839265919448e-06, + "loss": 0.6763, + "step": 2896 + }, + { + "epoch": 1.07, + "learning_rate": 2.8948023155561032e-06, + "loss": 0.6315, + "step": 2897 + }, + { + "epoch": 1.07, + "learning_rate": 2.894765365192758e-06, + "loss": 0.6598, + "step": 2898 + }, + { + "epoch": 1.07, + "learning_rate": 2.8947284148294125e-06, + "loss": 0.6281, + "step": 2899 + }, + { + "epoch": 1.07, + "learning_rate": 2.8946914644660675e-06, + "loss": 0.6116, + "step": 2900 + }, + { + "epoch": 1.07, + "learning_rate": 2.8946545141027218e-06, + "loss": 0.6, + "step": 2901 + }, + { + "epoch": 1.07, + "learning_rate": 2.894617563739377e-06, + "loss": 0.5936, + "step": 2902 + }, + { + "epoch": 1.07, + "learning_rate": 2.8945806133760315e-06, + "loss": 0.6111, + "step": 2903 + }, + { + "epoch": 1.07, + "learning_rate": 2.8945436630126865e-06, + "loss": 0.5599, + "step": 2904 + }, + { + "epoch": 1.07, + "learning_rate": 2.894506712649341e-06, + "loss": 0.616, + "step": 2905 + }, + { + "epoch": 1.07, + "learning_rate": 2.8944697622859962e-06, + "loss": 0.6658, + "step": 2906 + }, + { + "epoch": 1.07, + "learning_rate": 2.8944328119226504e-06, + "loss": 0.6408, + "step": 2907 + }, + { + "epoch": 1.07, + "learning_rate": 2.894395861559305e-06, + "loss": 0.6401, + "step": 2908 + }, + { + "epoch": 1.07, + "learning_rate": 2.89435891119596e-06, + "loss": 0.6623, + "step": 2909 + }, + { + "epoch": 1.07, + "learning_rate": 2.8943219608326148e-06, + "loss": 0.6525, + "step": 2910 + }, + { + "epoch": 1.07, + "learning_rate": 2.89428501046927e-06, + "loss": 0.6576, + "step": 2911 + }, + { + "epoch": 1.08, + "learning_rate": 2.8942480601059245e-06, + "loss": 0.718, + "step": 2912 + }, + { + "epoch": 1.08, + "learning_rate": 2.894211109742579e-06, + "loss": 0.592, + "step": 2913 + }, + { + "epoch": 1.08, + "learning_rate": 2.8941741593792337e-06, + "loss": 0.6126, + "step": 2914 + }, + { + "epoch": 1.08, + "learning_rate": 2.894137209015889e-06, + "loss": 0.6411, + "step": 2915 + }, + { + "epoch": 1.08, + "learning_rate": 2.8941002586525434e-06, + "loss": 0.6371, + "step": 2916 + }, + { + "epoch": 1.08, + "learning_rate": 2.8940633082891985e-06, + "loss": 0.6139, + "step": 2917 + }, + { + "epoch": 1.08, + "learning_rate": 2.894026357925853e-06, + "loss": 0.6166, + "step": 2918 + }, + { + "epoch": 1.08, + "learning_rate": 2.8939894075625078e-06, + "loss": 0.698, + "step": 2919 + }, + { + "epoch": 1.08, + "learning_rate": 2.8939524571991624e-06, + "loss": 0.6571, + "step": 2920 + }, + { + "epoch": 1.08, + "learning_rate": 2.893915506835817e-06, + "loss": 0.6525, + "step": 2921 + }, + { + "epoch": 1.08, + "learning_rate": 2.893878556472472e-06, + "loss": 0.6399, + "step": 2922 + }, + { + "epoch": 1.08, + "learning_rate": 2.8938416061091267e-06, + "loss": 0.5809, + "step": 2923 + }, + { + "epoch": 1.08, + "learning_rate": 2.893804655745782e-06, + "loss": 0.6511, + "step": 2924 + }, + { + "epoch": 1.08, + "learning_rate": 2.8937677053824364e-06, + "loss": 0.6466, + "step": 2925 + }, + { + "epoch": 1.08, + "learning_rate": 2.893730755019091e-06, + "loss": 0.6492, + "step": 2926 + }, + { + "epoch": 1.08, + "learning_rate": 2.8936938046557457e-06, + "loss": 0.6587, + "step": 2927 + }, + { + "epoch": 1.08, + "learning_rate": 2.8936568542924003e-06, + "loss": 0.6547, + "step": 2928 + }, + { + "epoch": 1.08, + "learning_rate": 2.8936199039290554e-06, + "loss": 0.6547, + "step": 2929 + }, + { + "epoch": 1.08, + "learning_rate": 2.89358295356571e-06, + "loss": 0.6144, + "step": 2930 + }, + { + "epoch": 1.08, + "learning_rate": 2.893546003202365e-06, + "loss": 0.6319, + "step": 2931 + }, + { + "epoch": 1.08, + "learning_rate": 2.8935090528390197e-06, + "loss": 0.5648, + "step": 2932 + }, + { + "epoch": 1.08, + "learning_rate": 2.8934721024756744e-06, + "loss": 0.6093, + "step": 2933 + }, + { + "epoch": 1.08, + "learning_rate": 2.893435152112329e-06, + "loss": 0.71, + "step": 2934 + }, + { + "epoch": 1.08, + "learning_rate": 2.893398201748984e-06, + "loss": 0.6514, + "step": 2935 + }, + { + "epoch": 1.08, + "learning_rate": 2.8933612513856387e-06, + "loss": 0.6641, + "step": 2936 + }, + { + "epoch": 1.08, + "learning_rate": 2.8933243010222938e-06, + "loss": 0.6271, + "step": 2937 + }, + { + "epoch": 1.08, + "learning_rate": 2.8932873506589484e-06, + "loss": 0.633, + "step": 2938 + }, + { + "epoch": 1.09, + "learning_rate": 2.893250400295603e-06, + "loss": 0.6034, + "step": 2939 + }, + { + "epoch": 1.09, + "learning_rate": 2.8932134499322577e-06, + "loss": 0.619, + "step": 2940 + }, + { + "epoch": 1.09, + "learning_rate": 2.8931764995689123e-06, + "loss": 0.6353, + "step": 2941 + }, + { + "epoch": 1.09, + "learning_rate": 2.8931395492055674e-06, + "loss": 0.6036, + "step": 2942 + }, + { + "epoch": 1.09, + "learning_rate": 2.893102598842222e-06, + "loss": 0.6061, + "step": 2943 + }, + { + "epoch": 1.09, + "learning_rate": 2.893065648478877e-06, + "loss": 0.6081, + "step": 2944 + }, + { + "epoch": 1.09, + "learning_rate": 2.8930286981155317e-06, + "loss": 0.633, + "step": 2945 + }, + { + "epoch": 1.09, + "learning_rate": 2.8929917477521863e-06, + "loss": 0.6295, + "step": 2946 + }, + { + "epoch": 1.09, + "learning_rate": 2.892954797388841e-06, + "loss": 0.6466, + "step": 2947 + }, + { + "epoch": 1.09, + "learning_rate": 2.8929178470254956e-06, + "loss": 0.6297, + "step": 2948 + }, + { + "epoch": 1.09, + "learning_rate": 2.8928808966621507e-06, + "loss": 0.5694, + "step": 2949 + }, + { + "epoch": 1.09, + "learning_rate": 2.8928439462988053e-06, + "loss": 0.6773, + "step": 2950 + }, + { + "epoch": 1.09, + "learning_rate": 2.8928069959354604e-06, + "loss": 0.6936, + "step": 2951 + }, + { + "epoch": 1.09, + "learning_rate": 2.8927700455721146e-06, + "loss": 0.6075, + "step": 2952 + }, + { + "epoch": 1.09, + "learning_rate": 2.8927330952087696e-06, + "loss": 0.6737, + "step": 2953 + }, + { + "epoch": 1.09, + "learning_rate": 2.8926961448454243e-06, + "loss": 0.5973, + "step": 2954 + }, + { + "epoch": 1.09, + "learning_rate": 2.8926591944820793e-06, + "loss": 0.6572, + "step": 2955 + }, + { + "epoch": 1.09, + "learning_rate": 2.892622244118734e-06, + "loss": 0.5345, + "step": 2956 + }, + { + "epoch": 1.09, + "learning_rate": 2.892585293755389e-06, + "loss": 0.5831, + "step": 2957 + }, + { + "epoch": 1.09, + "learning_rate": 2.8925483433920432e-06, + "loss": 0.6459, + "step": 2958 + }, + { + "epoch": 1.09, + "learning_rate": 2.892511393028698e-06, + "loss": 0.5262, + "step": 2959 + }, + { + "epoch": 1.09, + "learning_rate": 2.892474442665353e-06, + "loss": 0.5741, + "step": 2960 + }, + { + "epoch": 1.09, + "learning_rate": 2.8924374923020076e-06, + "loss": 0.5593, + "step": 2961 + }, + { + "epoch": 1.09, + "learning_rate": 2.8924005419386626e-06, + "loss": 0.5239, + "step": 2962 + }, + { + "epoch": 1.09, + "learning_rate": 2.8923635915753173e-06, + "loss": 0.6079, + "step": 2963 + }, + { + "epoch": 1.09, + "learning_rate": 2.892326641211972e-06, + "loss": 0.692, + "step": 2964 + }, + { + "epoch": 1.09, + "learning_rate": 2.8922896908486265e-06, + "loss": 0.6563, + "step": 2965 + }, + { + "epoch": 1.1, + "learning_rate": 2.8922527404852816e-06, + "loss": 0.6655, + "step": 2966 + }, + { + "epoch": 1.1, + "learning_rate": 2.8922157901219362e-06, + "loss": 0.7533, + "step": 2967 + }, + { + "epoch": 1.1, + "learning_rate": 2.892178839758591e-06, + "loss": 0.695, + "step": 2968 + }, + { + "epoch": 1.1, + "learning_rate": 2.892141889395246e-06, + "loss": 0.685, + "step": 2969 + }, + { + "epoch": 1.1, + "learning_rate": 2.8921049390319006e-06, + "loss": 0.585, + "step": 2970 + }, + { + "epoch": 1.1, + "learning_rate": 2.892067988668555e-06, + "loss": 0.6181, + "step": 2971 + }, + { + "epoch": 1.1, + "learning_rate": 2.89203103830521e-06, + "loss": 0.6158, + "step": 2972 + }, + { + "epoch": 1.1, + "learning_rate": 2.891994087941865e-06, + "loss": 0.6995, + "step": 2973 + }, + { + "epoch": 1.1, + "learning_rate": 2.8919571375785195e-06, + "loss": 0.6005, + "step": 2974 + }, + { + "epoch": 1.1, + "learning_rate": 2.8919201872151746e-06, + "loss": 0.6104, + "step": 2975 + }, + { + "epoch": 1.1, + "learning_rate": 2.8918832368518292e-06, + "loss": 0.6589, + "step": 2976 + }, + { + "epoch": 1.1, + "learning_rate": 2.891846286488484e-06, + "loss": 0.6131, + "step": 2977 + }, + { + "epoch": 1.1, + "learning_rate": 2.8918093361251385e-06, + "loss": 0.5597, + "step": 2978 + }, + { + "epoch": 1.1, + "learning_rate": 2.891772385761793e-06, + "loss": 0.584, + "step": 2979 + }, + { + "epoch": 1.1, + "learning_rate": 2.891735435398448e-06, + "loss": 0.5657, + "step": 2980 + }, + { + "epoch": 1.1, + "learning_rate": 2.891698485035103e-06, + "loss": 0.6375, + "step": 2981 + }, + { + "epoch": 1.1, + "learning_rate": 2.891661534671758e-06, + "loss": 0.5823, + "step": 2982 + }, + { + "epoch": 1.1, + "learning_rate": 2.8916245843084125e-06, + "loss": 0.6667, + "step": 2983 + }, + { + "epoch": 1.1, + "learning_rate": 2.891587633945067e-06, + "loss": 0.5795, + "step": 2984 + }, + { + "epoch": 1.1, + "learning_rate": 2.891550683581722e-06, + "loss": 0.606, + "step": 2985 + }, + { + "epoch": 1.1, + "learning_rate": 2.891513733218377e-06, + "loss": 0.6368, + "step": 2986 + }, + { + "epoch": 1.1, + "learning_rate": 2.8914767828550315e-06, + "loss": 0.586, + "step": 2987 + }, + { + "epoch": 1.1, + "learning_rate": 2.891439832491686e-06, + "loss": 0.6035, + "step": 2988 + }, + { + "epoch": 1.1, + "learning_rate": 2.891402882128341e-06, + "loss": 0.5833, + "step": 2989 + }, + { + "epoch": 1.1, + "learning_rate": 2.891365931764996e-06, + "loss": 0.5939, + "step": 2990 + }, + { + "epoch": 1.1, + "learning_rate": 2.8913289814016505e-06, + "loss": 0.6004, + "step": 2991 + }, + { + "epoch": 1.1, + "learning_rate": 2.891292031038305e-06, + "loss": 0.6157, + "step": 2992 + }, + { + "epoch": 1.11, + "learning_rate": 2.89125508067496e-06, + "loss": 0.6361, + "step": 2993 + }, + { + "epoch": 1.11, + "learning_rate": 2.891218130311615e-06, + "loss": 0.6219, + "step": 2994 + }, + { + "epoch": 1.11, + "learning_rate": 2.89118117994827e-06, + "loss": 0.6425, + "step": 2995 + }, + { + "epoch": 1.11, + "learning_rate": 2.8911442295849245e-06, + "loss": 0.6397, + "step": 2996 + }, + { + "epoch": 1.11, + "learning_rate": 2.891107279221579e-06, + "loss": 0.7114, + "step": 2997 + }, + { + "epoch": 1.11, + "learning_rate": 2.8910703288582338e-06, + "loss": 0.5351, + "step": 2998 + }, + { + "epoch": 1.11, + "learning_rate": 2.8910333784948884e-06, + "loss": 0.5902, + "step": 2999 + }, + { + "epoch": 1.11, + "learning_rate": 2.8909964281315435e-06, + "loss": 0.7047, + "step": 3000 + }, + { + "epoch": 1.11, + "learning_rate": 2.890959477768198e-06, + "loss": 0.6546, + "step": 3001 + }, + { + "epoch": 1.11, + "learning_rate": 2.890922527404853e-06, + "loss": 0.6418, + "step": 3002 + }, + { + "epoch": 1.11, + "learning_rate": 2.8908855770415074e-06, + "loss": 0.5967, + "step": 3003 + }, + { + "epoch": 1.11, + "learning_rate": 2.8908486266781624e-06, + "loss": 0.6319, + "step": 3004 + }, + { + "epoch": 1.11, + "learning_rate": 2.890811676314817e-06, + "loss": 0.6792, + "step": 3005 + }, + { + "epoch": 1.11, + "learning_rate": 2.890774725951472e-06, + "loss": 0.6459, + "step": 3006 + }, + { + "epoch": 1.11, + "learning_rate": 2.8907377755881268e-06, + "loss": 0.6106, + "step": 3007 + }, + { + "epoch": 1.11, + "learning_rate": 2.8907008252247814e-06, + "loss": 0.6719, + "step": 3008 + }, + { + "epoch": 1.11, + "learning_rate": 2.890663874861436e-06, + "loss": 0.635, + "step": 3009 + }, + { + "epoch": 1.11, + "learning_rate": 2.8906269244980907e-06, + "loss": 0.6927, + "step": 3010 + }, + { + "epoch": 1.11, + "learning_rate": 2.8905899741347457e-06, + "loss": 0.6971, + "step": 3011 + }, + { + "epoch": 1.11, + "learning_rate": 2.8905530237714004e-06, + "loss": 0.6259, + "step": 3012 + }, + { + "epoch": 1.11, + "learning_rate": 2.8905160734080554e-06, + "loss": 0.6758, + "step": 3013 + }, + { + "epoch": 1.11, + "learning_rate": 2.89047912304471e-06, + "loss": 0.6396, + "step": 3014 + }, + { + "epoch": 1.11, + "learning_rate": 2.8904421726813647e-06, + "loss": 0.5802, + "step": 3015 + }, + { + "epoch": 1.11, + "learning_rate": 2.8904052223180194e-06, + "loss": 0.602, + "step": 3016 + }, + { + "epoch": 1.11, + "learning_rate": 2.8903682719546744e-06, + "loss": 0.6421, + "step": 3017 + }, + { + "epoch": 1.11, + "learning_rate": 2.890331321591329e-06, + "loss": 0.6595, + "step": 3018 + }, + { + "epoch": 1.11, + "learning_rate": 2.8902943712279837e-06, + "loss": 0.5982, + "step": 3019 + }, + { + "epoch": 1.12, + "learning_rate": 2.8902574208646387e-06, + "loss": 0.6248, + "step": 3020 + }, + { + "epoch": 1.12, + "learning_rate": 2.8902204705012934e-06, + "loss": 0.5738, + "step": 3021 + }, + { + "epoch": 1.12, + "learning_rate": 2.890183520137948e-06, + "loss": 0.6364, + "step": 3022 + }, + { + "epoch": 1.12, + "learning_rate": 2.8901465697746027e-06, + "loss": 0.5621, + "step": 3023 + }, + { + "epoch": 1.12, + "learning_rate": 2.8901096194112577e-06, + "loss": 0.6634, + "step": 3024 + }, + { + "epoch": 1.12, + "learning_rate": 2.8900726690479124e-06, + "loss": 0.6328, + "step": 3025 + }, + { + "epoch": 1.12, + "learning_rate": 2.8900357186845674e-06, + "loss": 0.5667, + "step": 3026 + }, + { + "epoch": 1.12, + "learning_rate": 2.889998768321222e-06, + "loss": 0.6509, + "step": 3027 + }, + { + "epoch": 1.12, + "learning_rate": 2.8899618179578767e-06, + "loss": 0.5927, + "step": 3028 + }, + { + "epoch": 1.12, + "learning_rate": 2.8899248675945313e-06, + "loss": 0.6506, + "step": 3029 + }, + { + "epoch": 1.12, + "learning_rate": 2.889887917231186e-06, + "loss": 0.5815, + "step": 3030 + }, + { + "epoch": 1.12, + "learning_rate": 2.889850966867841e-06, + "loss": 0.6414, + "step": 3031 + }, + { + "epoch": 1.12, + "learning_rate": 2.8898140165044957e-06, + "loss": 0.5803, + "step": 3032 + }, + { + "epoch": 1.12, + "learning_rate": 2.8897770661411507e-06, + "loss": 0.6238, + "step": 3033 + }, + { + "epoch": 1.12, + "learning_rate": 2.8897401157778053e-06, + "loss": 0.693, + "step": 3034 + }, + { + "epoch": 1.12, + "learning_rate": 2.88970316541446e-06, + "loss": 0.698, + "step": 3035 + }, + { + "epoch": 1.12, + "learning_rate": 2.8896662150511146e-06, + "loss": 0.6007, + "step": 3036 + }, + { + "epoch": 1.12, + "learning_rate": 2.8896292646877697e-06, + "loss": 0.6966, + "step": 3037 + }, + { + "epoch": 1.12, + "learning_rate": 2.8895923143244243e-06, + "loss": 0.6733, + "step": 3038 + }, + { + "epoch": 1.12, + "learning_rate": 2.889555363961079e-06, + "loss": 0.7034, + "step": 3039 + }, + { + "epoch": 1.12, + "learning_rate": 2.889518413597734e-06, + "loss": 0.5925, + "step": 3040 + }, + { + "epoch": 1.12, + "learning_rate": 2.8894814632343887e-06, + "loss": 0.6639, + "step": 3041 + }, + { + "epoch": 1.12, + "learning_rate": 2.8894445128710433e-06, + "loss": 0.6322, + "step": 3042 + }, + { + "epoch": 1.12, + "learning_rate": 2.889407562507698e-06, + "loss": 0.6548, + "step": 3043 + }, + { + "epoch": 1.12, + "learning_rate": 2.889370612144353e-06, + "loss": 0.6403, + "step": 3044 + }, + { + "epoch": 1.12, + "learning_rate": 2.8893336617810076e-06, + "loss": 0.6857, + "step": 3045 + }, + { + "epoch": 1.12, + "learning_rate": 2.8892967114176627e-06, + "loss": 0.5828, + "step": 3046 + }, + { + "epoch": 1.13, + "learning_rate": 2.8892597610543173e-06, + "loss": 0.6402, + "step": 3047 + }, + { + "epoch": 1.13, + "learning_rate": 2.8892228106909715e-06, + "loss": 0.6316, + "step": 3048 + }, + { + "epoch": 1.13, + "learning_rate": 2.8891858603276266e-06, + "loss": 0.5796, + "step": 3049 + }, + { + "epoch": 1.13, + "learning_rate": 2.8891489099642812e-06, + "loss": 0.5481, + "step": 3050 + }, + { + "epoch": 1.13, + "learning_rate": 2.8891119596009363e-06, + "loss": 0.6102, + "step": 3051 + }, + { + "epoch": 1.13, + "learning_rate": 2.889075009237591e-06, + "loss": 0.6765, + "step": 3052 + }, + { + "epoch": 1.13, + "learning_rate": 2.8890380588742456e-06, + "loss": 0.5621, + "step": 3053 + }, + { + "epoch": 1.13, + "learning_rate": 2.8890011085109e-06, + "loss": 0.6691, + "step": 3054 + }, + { + "epoch": 1.13, + "learning_rate": 2.8889641581475553e-06, + "loss": 0.607, + "step": 3055 + }, + { + "epoch": 1.13, + "learning_rate": 2.88892720778421e-06, + "loss": 0.7186, + "step": 3056 + }, + { + "epoch": 1.13, + "learning_rate": 2.888890257420865e-06, + "loss": 0.5999, + "step": 3057 + }, + { + "epoch": 1.13, + "learning_rate": 2.8888533070575196e-06, + "loss": 0.6336, + "step": 3058 + }, + { + "epoch": 1.13, + "learning_rate": 2.8888163566941742e-06, + "loss": 0.7104, + "step": 3059 + }, + { + "epoch": 1.13, + "learning_rate": 2.888779406330829e-06, + "loss": 0.6176, + "step": 3060 + }, + { + "epoch": 1.13, + "learning_rate": 2.8887424559674835e-06, + "loss": 0.6196, + "step": 3061 + }, + { + "epoch": 1.13, + "learning_rate": 2.8887055056041386e-06, + "loss": 0.6434, + "step": 3062 + }, + { + "epoch": 1.13, + "learning_rate": 2.888668555240793e-06, + "loss": 0.6039, + "step": 3063 + }, + { + "epoch": 1.13, + "learning_rate": 2.8886316048774483e-06, + "loss": 0.6693, + "step": 3064 + }, + { + "epoch": 1.13, + "learning_rate": 2.888594654514103e-06, + "loss": 0.6464, + "step": 3065 + }, + { + "epoch": 1.13, + "learning_rate": 2.8885577041507575e-06, + "loss": 0.6423, + "step": 3066 + }, + { + "epoch": 1.13, + "learning_rate": 2.888520753787412e-06, + "loss": 0.675, + "step": 3067 + }, + { + "epoch": 1.13, + "learning_rate": 2.888483803424067e-06, + "loss": 0.6138, + "step": 3068 + }, + { + "epoch": 1.13, + "learning_rate": 2.888446853060722e-06, + "loss": 0.5631, + "step": 3069 + }, + { + "epoch": 1.13, + "learning_rate": 2.8884099026973765e-06, + "loss": 0.5761, + "step": 3070 + }, + { + "epoch": 1.13, + "learning_rate": 2.8883729523340316e-06, + "loss": 0.5957, + "step": 3071 + }, + { + "epoch": 1.13, + "learning_rate": 2.888336001970686e-06, + "loss": 0.6241, + "step": 3072 + }, + { + "epoch": 1.13, + "learning_rate": 2.888299051607341e-06, + "loss": 0.557, + "step": 3073 + }, + { + "epoch": 1.14, + "learning_rate": 2.8882621012439955e-06, + "loss": 0.5857, + "step": 3074 + }, + { + "epoch": 1.14, + "learning_rate": 2.8882251508806505e-06, + "loss": 0.5987, + "step": 3075 + }, + { + "epoch": 1.14, + "learning_rate": 2.888188200517305e-06, + "loss": 0.6478, + "step": 3076 + }, + { + "epoch": 1.14, + "learning_rate": 2.88815125015396e-06, + "loss": 0.5918, + "step": 3077 + }, + { + "epoch": 1.14, + "learning_rate": 2.888114299790615e-06, + "loss": 0.5469, + "step": 3078 + }, + { + "epoch": 1.14, + "learning_rate": 2.8880773494272695e-06, + "loss": 0.6916, + "step": 3079 + }, + { + "epoch": 1.14, + "learning_rate": 2.888040399063924e-06, + "loss": 0.6005, + "step": 3080 + }, + { + "epoch": 1.14, + "learning_rate": 2.8880034487005788e-06, + "loss": 0.6025, + "step": 3081 + }, + { + "epoch": 1.14, + "learning_rate": 2.887966498337234e-06, + "loss": 0.6503, + "step": 3082 + }, + { + "epoch": 1.14, + "learning_rate": 2.8879295479738885e-06, + "loss": 0.5502, + "step": 3083 + }, + { + "epoch": 1.14, + "learning_rate": 2.8878925976105435e-06, + "loss": 0.6257, + "step": 3084 + }, + { + "epoch": 1.14, + "learning_rate": 2.887855647247198e-06, + "loss": 0.6059, + "step": 3085 + }, + { + "epoch": 1.14, + "learning_rate": 2.887818696883853e-06, + "loss": 0.6214, + "step": 3086 + }, + { + "epoch": 1.14, + "learning_rate": 2.8877817465205074e-06, + "loss": 0.5739, + "step": 3087 + }, + { + "epoch": 1.14, + "learning_rate": 2.887744796157162e-06, + "loss": 0.57, + "step": 3088 + }, + { + "epoch": 1.14, + "learning_rate": 2.887707845793817e-06, + "loss": 0.5491, + "step": 3089 + }, + { + "epoch": 1.14, + "learning_rate": 2.8876708954304718e-06, + "loss": 0.6878, + "step": 3090 + }, + { + "epoch": 1.14, + "learning_rate": 2.887633945067127e-06, + "loss": 0.6018, + "step": 3091 + }, + { + "epoch": 1.14, + "learning_rate": 2.887596994703781e-06, + "loss": 0.7155, + "step": 3092 + }, + { + "epoch": 1.14, + "learning_rate": 2.887560044340436e-06, + "loss": 0.5857, + "step": 3093 + }, + { + "epoch": 1.14, + "learning_rate": 2.8875230939770907e-06, + "loss": 0.6755, + "step": 3094 + }, + { + "epoch": 1.14, + "learning_rate": 2.887486143613746e-06, + "loss": 0.5945, + "step": 3095 + }, + { + "epoch": 1.14, + "learning_rate": 2.8874491932504004e-06, + "loss": 0.5916, + "step": 3096 + }, + { + "epoch": 1.14, + "learning_rate": 2.887412242887055e-06, + "loss": 0.5874, + "step": 3097 + }, + { + "epoch": 1.14, + "learning_rate": 2.8873752925237097e-06, + "loss": 0.6, + "step": 3098 + }, + { + "epoch": 1.14, + "learning_rate": 2.8873383421603643e-06, + "loss": 0.6376, + "step": 3099 + }, + { + "epoch": 1.14, + "learning_rate": 2.8873013917970194e-06, + "loss": 0.7345, + "step": 3100 + }, + { + "epoch": 1.15, + "learning_rate": 2.887264441433674e-06, + "loss": 0.5791, + "step": 3101 + }, + { + "epoch": 1.15, + "learning_rate": 2.887227491070329e-06, + "loss": 0.6081, + "step": 3102 + }, + { + "epoch": 1.15, + "learning_rate": 2.8871905407069837e-06, + "loss": 0.6272, + "step": 3103 + }, + { + "epoch": 1.15, + "learning_rate": 2.8871535903436384e-06, + "loss": 0.6169, + "step": 3104 + }, + { + "epoch": 1.15, + "learning_rate": 2.887116639980293e-06, + "loss": 0.6173, + "step": 3105 + }, + { + "epoch": 1.15, + "learning_rate": 2.887079689616948e-06, + "loss": 0.6124, + "step": 3106 + }, + { + "epoch": 1.15, + "learning_rate": 2.8870427392536027e-06, + "loss": 0.7065, + "step": 3107 + }, + { + "epoch": 1.15, + "learning_rate": 2.8870057888902573e-06, + "loss": 0.5348, + "step": 3108 + }, + { + "epoch": 1.15, + "learning_rate": 2.8869688385269124e-06, + "loss": 0.6295, + "step": 3109 + }, + { + "epoch": 1.15, + "learning_rate": 2.886931888163567e-06, + "loss": 0.6268, + "step": 3110 + }, + { + "epoch": 1.15, + "learning_rate": 2.8868949378002217e-06, + "loss": 0.6675, + "step": 3111 + }, + { + "epoch": 1.15, + "learning_rate": 2.8868579874368763e-06, + "loss": 0.6484, + "step": 3112 + }, + { + "epoch": 1.15, + "learning_rate": 2.8868210370735314e-06, + "loss": 0.6346, + "step": 3113 + }, + { + "epoch": 1.15, + "learning_rate": 2.886784086710186e-06, + "loss": 0.639, + "step": 3114 + }, + { + "epoch": 1.15, + "learning_rate": 2.886747136346841e-06, + "loss": 0.6343, + "step": 3115 + }, + { + "epoch": 1.15, + "learning_rate": 2.8867101859834957e-06, + "loss": 0.5874, + "step": 3116 + }, + { + "epoch": 1.15, + "learning_rate": 2.8866732356201503e-06, + "loss": 0.555, + "step": 3117 + }, + { + "epoch": 1.15, + "learning_rate": 2.886636285256805e-06, + "loss": 0.5765, + "step": 3118 + }, + { + "epoch": 1.15, + "learning_rate": 2.8865993348934596e-06, + "loss": 0.6484, + "step": 3119 + }, + { + "epoch": 1.15, + "learning_rate": 2.8865623845301147e-06, + "loss": 0.6057, + "step": 3120 + }, + { + "epoch": 1.15, + "learning_rate": 2.8865254341667693e-06, + "loss": 0.5647, + "step": 3121 + }, + { + "epoch": 1.15, + "learning_rate": 2.8864884838034244e-06, + "loss": 0.6106, + "step": 3122 + }, + { + "epoch": 1.15, + "learning_rate": 2.886451533440079e-06, + "loss": 0.5504, + "step": 3123 + }, + { + "epoch": 1.15, + "learning_rate": 2.8864145830767336e-06, + "loss": 0.6176, + "step": 3124 + }, + { + "epoch": 1.15, + "learning_rate": 2.8863776327133883e-06, + "loss": 0.6257, + "step": 3125 + }, + { + "epoch": 1.15, + "learning_rate": 2.8863406823500433e-06, + "loss": 0.6972, + "step": 3126 + }, + { + "epoch": 1.15, + "learning_rate": 2.886303731986698e-06, + "loss": 0.6592, + "step": 3127 + }, + { + "epoch": 1.16, + "learning_rate": 2.8862667816233526e-06, + "loss": 0.505, + "step": 3128 + }, + { + "epoch": 1.16, + "learning_rate": 2.8862298312600077e-06, + "loss": 0.6354, + "step": 3129 + }, + { + "epoch": 1.16, + "learning_rate": 2.8861928808966623e-06, + "loss": 0.6024, + "step": 3130 + }, + { + "epoch": 1.16, + "learning_rate": 2.886155930533317e-06, + "loss": 0.689, + "step": 3131 + }, + { + "epoch": 1.16, + "learning_rate": 2.8861189801699716e-06, + "loss": 0.6939, + "step": 3132 + }, + { + "epoch": 1.16, + "learning_rate": 2.8860820298066266e-06, + "loss": 0.6066, + "step": 3133 + }, + { + "epoch": 1.16, + "learning_rate": 2.8860450794432813e-06, + "loss": 0.529, + "step": 3134 + }, + { + "epoch": 1.16, + "learning_rate": 2.8860081290799363e-06, + "loss": 0.7264, + "step": 3135 + }, + { + "epoch": 1.16, + "learning_rate": 2.885971178716591e-06, + "loss": 0.6215, + "step": 3136 + }, + { + "epoch": 1.16, + "learning_rate": 2.885934228353245e-06, + "loss": 0.6212, + "step": 3137 + }, + { + "epoch": 1.16, + "learning_rate": 2.8858972779899002e-06, + "loss": 0.5247, + "step": 3138 + }, + { + "epoch": 1.16, + "learning_rate": 2.885860327626555e-06, + "loss": 0.6209, + "step": 3139 + }, + { + "epoch": 1.16, + "learning_rate": 2.88582337726321e-06, + "loss": 0.7025, + "step": 3140 + }, + { + "epoch": 1.16, + "learning_rate": 2.8857864268998646e-06, + "loss": 0.5796, + "step": 3141 + }, + { + "epoch": 1.16, + "learning_rate": 2.8857494765365196e-06, + "loss": 0.6134, + "step": 3142 + }, + { + "epoch": 1.16, + "learning_rate": 2.885712526173174e-06, + "loss": 0.4862, + "step": 3143 + }, + { + "epoch": 1.16, + "learning_rate": 2.885675575809829e-06, + "loss": 0.6525, + "step": 3144 + }, + { + "epoch": 1.16, + "learning_rate": 2.8856386254464835e-06, + "loss": 0.6111, + "step": 3145 + }, + { + "epoch": 1.16, + "learning_rate": 2.8856016750831386e-06, + "loss": 0.5865, + "step": 3146 + }, + { + "epoch": 1.16, + "learning_rate": 2.8855647247197932e-06, + "loss": 0.5284, + "step": 3147 + }, + { + "epoch": 1.16, + "learning_rate": 2.885527774356448e-06, + "loss": 0.6176, + "step": 3148 + }, + { + "epoch": 1.16, + "learning_rate": 2.8854908239931025e-06, + "loss": 0.6929, + "step": 3149 + }, + { + "epoch": 1.16, + "learning_rate": 2.885453873629757e-06, + "loss": 0.5547, + "step": 3150 + }, + { + "epoch": 1.16, + "learning_rate": 2.885416923266412e-06, + "loss": 0.5945, + "step": 3151 + }, + { + "epoch": 1.16, + "learning_rate": 2.885379972903067e-06, + "loss": 0.6563, + "step": 3152 + }, + { + "epoch": 1.16, + "learning_rate": 2.885343022539722e-06, + "loss": 0.6759, + "step": 3153 + }, + { + "epoch": 1.16, + "learning_rate": 2.8853060721763765e-06, + "loss": 0.6275, + "step": 3154 + }, + { + "epoch": 1.17, + "learning_rate": 2.885269121813031e-06, + "loss": 0.6633, + "step": 3155 + }, + { + "epoch": 1.17, + "learning_rate": 2.885232171449686e-06, + "loss": 0.584, + "step": 3156 + }, + { + "epoch": 1.17, + "learning_rate": 2.8851952210863405e-06, + "loss": 0.7144, + "step": 3157 + }, + { + "epoch": 1.17, + "learning_rate": 2.8851582707229955e-06, + "loss": 0.7127, + "step": 3158 + }, + { + "epoch": 1.17, + "learning_rate": 2.88512132035965e-06, + "loss": 0.5383, + "step": 3159 + }, + { + "epoch": 1.17, + "learning_rate": 2.885084369996305e-06, + "loss": 0.6356, + "step": 3160 + }, + { + "epoch": 1.17, + "learning_rate": 2.88504741963296e-06, + "loss": 0.6351, + "step": 3161 + }, + { + "epoch": 1.17, + "learning_rate": 2.8850104692696145e-06, + "loss": 0.6309, + "step": 3162 + }, + { + "epoch": 1.17, + "learning_rate": 2.884973518906269e-06, + "loss": 0.5995, + "step": 3163 + }, + { + "epoch": 1.17, + "learning_rate": 2.884936568542924e-06, + "loss": 0.5964, + "step": 3164 + }, + { + "epoch": 1.17, + "learning_rate": 2.884899618179579e-06, + "loss": 0.5717, + "step": 3165 + }, + { + "epoch": 1.17, + "learning_rate": 2.884862667816234e-06, + "loss": 0.655, + "step": 3166 + }, + { + "epoch": 1.17, + "learning_rate": 2.8848257174528885e-06, + "loss": 0.6699, + "step": 3167 + }, + { + "epoch": 1.17, + "learning_rate": 2.884788767089543e-06, + "loss": 0.5744, + "step": 3168 + }, + { + "epoch": 1.17, + "learning_rate": 2.8847518167261978e-06, + "loss": 0.5763, + "step": 3169 + }, + { + "epoch": 1.17, + "learning_rate": 2.8847148663628524e-06, + "loss": 0.6319, + "step": 3170 + }, + { + "epoch": 1.17, + "learning_rate": 2.8846779159995075e-06, + "loss": 0.583, + "step": 3171 + }, + { + "epoch": 1.17, + "learning_rate": 2.884640965636162e-06, + "loss": 0.6088, + "step": 3172 + }, + { + "epoch": 1.17, + "learning_rate": 2.884604015272817e-06, + "loss": 0.6785, + "step": 3173 + }, + { + "epoch": 1.17, + "learning_rate": 2.884567064909472e-06, + "loss": 0.612, + "step": 3174 + }, + { + "epoch": 1.17, + "learning_rate": 2.8845301145461265e-06, + "loss": 0.6331, + "step": 3175 + }, + { + "epoch": 1.17, + "learning_rate": 2.884493164182781e-06, + "loss": 0.6552, + "step": 3176 + }, + { + "epoch": 1.17, + "learning_rate": 2.8844562138194357e-06, + "loss": 0.6437, + "step": 3177 + }, + { + "epoch": 1.17, + "learning_rate": 2.8844192634560908e-06, + "loss": 0.6748, + "step": 3178 + }, + { + "epoch": 1.17, + "learning_rate": 2.8843823130927454e-06, + "loss": 0.5556, + "step": 3179 + }, + { + "epoch": 1.17, + "learning_rate": 2.8843453627294005e-06, + "loss": 0.6144, + "step": 3180 + }, + { + "epoch": 1.17, + "learning_rate": 2.884308412366055e-06, + "loss": 0.6051, + "step": 3181 + }, + { + "epoch": 1.18, + "learning_rate": 2.8842714620027098e-06, + "loss": 0.7402, + "step": 3182 + }, + { + "epoch": 1.18, + "learning_rate": 2.8842345116393644e-06, + "loss": 0.5873, + "step": 3183 + }, + { + "epoch": 1.18, + "learning_rate": 2.8841975612760194e-06, + "loss": 0.6001, + "step": 3184 + }, + { + "epoch": 1.18, + "learning_rate": 2.884160610912674e-06, + "loss": 0.5769, + "step": 3185 + }, + { + "epoch": 1.18, + "learning_rate": 2.884123660549329e-06, + "loss": 0.6738, + "step": 3186 + }, + { + "epoch": 1.18, + "learning_rate": 2.8840867101859838e-06, + "loss": 0.6373, + "step": 3187 + }, + { + "epoch": 1.18, + "learning_rate": 2.884049759822638e-06, + "loss": 0.5956, + "step": 3188 + }, + { + "epoch": 1.18, + "learning_rate": 2.884012809459293e-06, + "loss": 0.6333, + "step": 3189 + }, + { + "epoch": 1.18, + "learning_rate": 2.8839758590959477e-06, + "loss": 0.6018, + "step": 3190 + }, + { + "epoch": 1.18, + "learning_rate": 2.8839389087326028e-06, + "loss": 0.6047, + "step": 3191 + }, + { + "epoch": 1.18, + "learning_rate": 2.8839019583692574e-06, + "loss": 0.615, + "step": 3192 + }, + { + "epoch": 1.18, + "learning_rate": 2.8838650080059124e-06, + "loss": 0.6519, + "step": 3193 + }, + { + "epoch": 1.18, + "learning_rate": 2.8838280576425667e-06, + "loss": 0.613, + "step": 3194 + }, + { + "epoch": 1.18, + "learning_rate": 2.8837911072792217e-06, + "loss": 0.6076, + "step": 3195 + }, + { + "epoch": 1.18, + "learning_rate": 2.8837541569158764e-06, + "loss": 0.6368, + "step": 3196 + }, + { + "epoch": 1.18, + "learning_rate": 2.883717206552531e-06, + "loss": 0.5573, + "step": 3197 + }, + { + "epoch": 1.18, + "learning_rate": 2.883680256189186e-06, + "loss": 0.5746, + "step": 3198 + }, + { + "epoch": 1.18, + "learning_rate": 2.8836433058258407e-06, + "loss": 0.6536, + "step": 3199 + }, + { + "epoch": 1.18, + "learning_rate": 2.8836063554624953e-06, + "loss": 0.7124, + "step": 3200 + }, + { + "epoch": 1.18, + "learning_rate": 2.88356940509915e-06, + "loss": 0.6219, + "step": 3201 + }, + { + "epoch": 1.18, + "learning_rate": 2.883532454735805e-06, + "loss": 0.5089, + "step": 3202 + }, + { + "epoch": 1.18, + "learning_rate": 2.8834955043724597e-06, + "loss": 0.6407, + "step": 3203 + }, + { + "epoch": 1.18, + "learning_rate": 2.8834585540091147e-06, + "loss": 0.5848, + "step": 3204 + }, + { + "epoch": 1.18, + "learning_rate": 2.8834216036457694e-06, + "loss": 0.6392, + "step": 3205 + }, + { + "epoch": 1.18, + "learning_rate": 2.883384653282424e-06, + "loss": 0.6545, + "step": 3206 + }, + { + "epoch": 1.18, + "learning_rate": 2.8833477029190786e-06, + "loss": 0.6388, + "step": 3207 + }, + { + "epoch": 1.18, + "learning_rate": 2.8833107525557333e-06, + "loss": 0.6178, + "step": 3208 + }, + { + "epoch": 1.19, + "learning_rate": 2.8832738021923883e-06, + "loss": 0.5897, + "step": 3209 + }, + { + "epoch": 1.19, + "learning_rate": 2.883236851829043e-06, + "loss": 0.631, + "step": 3210 + }, + { + "epoch": 1.19, + "learning_rate": 2.883199901465698e-06, + "loss": 0.6091, + "step": 3211 + }, + { + "epoch": 1.19, + "learning_rate": 2.8831629511023527e-06, + "loss": 0.6643, + "step": 3212 + }, + { + "epoch": 1.19, + "learning_rate": 2.8831260007390073e-06, + "loss": 0.6758, + "step": 3213 + }, + { + "epoch": 1.19, + "learning_rate": 2.883089050375662e-06, + "loss": 0.6286, + "step": 3214 + }, + { + "epoch": 1.19, + "learning_rate": 2.883052100012317e-06, + "loss": 0.6571, + "step": 3215 + }, + { + "epoch": 1.19, + "learning_rate": 2.8830151496489716e-06, + "loss": 0.6603, + "step": 3216 + }, + { + "epoch": 1.19, + "learning_rate": 2.8829781992856263e-06, + "loss": 0.6308, + "step": 3217 + }, + { + "epoch": 1.19, + "learning_rate": 2.8829412489222813e-06, + "loss": 0.6168, + "step": 3218 + }, + { + "epoch": 1.19, + "learning_rate": 2.882904298558936e-06, + "loss": 0.616, + "step": 3219 + }, + { + "epoch": 1.19, + "learning_rate": 2.8828673481955906e-06, + "loss": 0.563, + "step": 3220 + }, + { + "epoch": 1.19, + "learning_rate": 2.8828303978322452e-06, + "loss": 0.6813, + "step": 3221 + }, + { + "epoch": 1.19, + "learning_rate": 2.8827934474689003e-06, + "loss": 0.5913, + "step": 3222 + }, + { + "epoch": 1.19, + "learning_rate": 2.882756497105555e-06, + "loss": 0.6693, + "step": 3223 + }, + { + "epoch": 1.19, + "learning_rate": 2.88271954674221e-06, + "loss": 0.5937, + "step": 3224 + }, + { + "epoch": 1.19, + "learning_rate": 2.8826825963788646e-06, + "loss": 0.6236, + "step": 3225 + }, + { + "epoch": 1.19, + "learning_rate": 2.8826456460155193e-06, + "loss": 0.7334, + "step": 3226 + }, + { + "epoch": 1.19, + "learning_rate": 2.882608695652174e-06, + "loss": 0.6238, + "step": 3227 + }, + { + "epoch": 1.19, + "learning_rate": 2.8825717452888285e-06, + "loss": 0.6985, + "step": 3228 + }, + { + "epoch": 1.19, + "learning_rate": 2.8825347949254836e-06, + "loss": 0.6366, + "step": 3229 + }, + { + "epoch": 1.19, + "learning_rate": 2.8824978445621382e-06, + "loss": 0.5904, + "step": 3230 + }, + { + "epoch": 1.19, + "learning_rate": 2.8824608941987933e-06, + "loss": 0.6293, + "step": 3231 + }, + { + "epoch": 1.19, + "learning_rate": 2.882423943835448e-06, + "loss": 0.6646, + "step": 3232 + }, + { + "epoch": 1.19, + "learning_rate": 2.8823869934721026e-06, + "loss": 0.6712, + "step": 3233 + }, + { + "epoch": 1.19, + "learning_rate": 2.882350043108757e-06, + "loss": 0.6477, + "step": 3234 + }, + { + "epoch": 1.19, + "learning_rate": 2.8823130927454123e-06, + "loss": 0.7115, + "step": 3235 + }, + { + "epoch": 1.19, + "learning_rate": 2.882276142382067e-06, + "loss": 0.5891, + "step": 3236 + }, + { + "epoch": 1.2, + "learning_rate": 2.8822391920187215e-06, + "loss": 0.5818, + "step": 3237 + }, + { + "epoch": 1.2, + "learning_rate": 2.882202241655376e-06, + "loss": 0.5697, + "step": 3238 + }, + { + "epoch": 1.2, + "learning_rate": 2.882165291292031e-06, + "loss": 0.6246, + "step": 3239 + }, + { + "epoch": 1.2, + "learning_rate": 2.882128340928686e-06, + "loss": 0.6483, + "step": 3240 + }, + { + "epoch": 1.2, + "learning_rate": 2.8820913905653405e-06, + "loss": 0.6633, + "step": 3241 + }, + { + "epoch": 1.2, + "learning_rate": 2.8820544402019956e-06, + "loss": 0.6383, + "step": 3242 + }, + { + "epoch": 1.2, + "learning_rate": 2.88201748983865e-06, + "loss": 0.6011, + "step": 3243 + }, + { + "epoch": 1.2, + "learning_rate": 2.881980539475305e-06, + "loss": 0.6375, + "step": 3244 + }, + { + "epoch": 1.2, + "learning_rate": 2.8819435891119595e-06, + "loss": 0.6952, + "step": 3245 + }, + { + "epoch": 1.2, + "learning_rate": 2.8819066387486145e-06, + "loss": 0.7263, + "step": 3246 + }, + { + "epoch": 1.2, + "learning_rate": 2.881869688385269e-06, + "loss": 0.6268, + "step": 3247 + }, + { + "epoch": 1.2, + "learning_rate": 2.881832738021924e-06, + "loss": 0.6505, + "step": 3248 + }, + { + "epoch": 1.2, + "learning_rate": 2.881795787658579e-06, + "loss": 0.6283, + "step": 3249 + }, + { + "epoch": 1.2, + "learning_rate": 2.8817588372952335e-06, + "loss": 0.581, + "step": 3250 + }, + { + "epoch": 1.2, + "learning_rate": 2.881721886931888e-06, + "loss": 0.6284, + "step": 3251 + }, + { + "epoch": 1.2, + "learning_rate": 2.8816849365685428e-06, + "loss": 0.5707, + "step": 3252 + }, + { + "epoch": 1.2, + "learning_rate": 2.881647986205198e-06, + "loss": 0.7195, + "step": 3253 + }, + { + "epoch": 1.2, + "learning_rate": 2.8816110358418525e-06, + "loss": 0.6113, + "step": 3254 + }, + { + "epoch": 1.2, + "learning_rate": 2.8815740854785075e-06, + "loss": 0.5564, + "step": 3255 + }, + { + "epoch": 1.2, + "learning_rate": 2.881537135115162e-06, + "loss": 0.6098, + "step": 3256 + }, + { + "epoch": 1.2, + "learning_rate": 2.881500184751817e-06, + "loss": 0.6024, + "step": 3257 + }, + { + "epoch": 1.2, + "learning_rate": 2.8814632343884714e-06, + "loss": 0.6596, + "step": 3258 + }, + { + "epoch": 1.2, + "learning_rate": 2.881426284025126e-06, + "loss": 0.6159, + "step": 3259 + }, + { + "epoch": 1.2, + "learning_rate": 2.881389333661781e-06, + "loss": 0.6205, + "step": 3260 + }, + { + "epoch": 1.2, + "learning_rate": 2.8813523832984358e-06, + "loss": 0.6425, + "step": 3261 + }, + { + "epoch": 1.2, + "learning_rate": 2.881315432935091e-06, + "loss": 0.6552, + "step": 3262 + }, + { + "epoch": 1.2, + "learning_rate": 2.8812784825717455e-06, + "loss": 0.6174, + "step": 3263 + }, + { + "epoch": 1.21, + "learning_rate": 2.8812415322084e-06, + "loss": 0.6694, + "step": 3264 + }, + { + "epoch": 1.21, + "learning_rate": 2.8812045818450547e-06, + "loss": 0.5784, + "step": 3265 + }, + { + "epoch": 1.21, + "learning_rate": 2.88116763148171e-06, + "loss": 0.6299, + "step": 3266 + }, + { + "epoch": 1.21, + "learning_rate": 2.8811306811183644e-06, + "loss": 0.5915, + "step": 3267 + }, + { + "epoch": 1.21, + "learning_rate": 2.881093730755019e-06, + "loss": 0.6635, + "step": 3268 + }, + { + "epoch": 1.21, + "learning_rate": 2.881056780391674e-06, + "loss": 0.6902, + "step": 3269 + }, + { + "epoch": 1.21, + "learning_rate": 2.8810198300283288e-06, + "loss": 0.6225, + "step": 3270 + }, + { + "epoch": 1.21, + "learning_rate": 2.8809828796649834e-06, + "loss": 0.6132, + "step": 3271 + }, + { + "epoch": 1.21, + "learning_rate": 2.880945929301638e-06, + "loss": 0.6637, + "step": 3272 + }, + { + "epoch": 1.21, + "learning_rate": 2.880908978938293e-06, + "loss": 0.5917, + "step": 3273 + }, + { + "epoch": 1.21, + "learning_rate": 2.8808720285749477e-06, + "loss": 0.5875, + "step": 3274 + }, + { + "epoch": 1.21, + "learning_rate": 2.880835078211603e-06, + "loss": 0.5852, + "step": 3275 + }, + { + "epoch": 1.21, + "learning_rate": 2.8807981278482574e-06, + "loss": 0.5717, + "step": 3276 + }, + { + "epoch": 1.21, + "learning_rate": 2.8807611774849116e-06, + "loss": 0.6039, + "step": 3277 + }, + { + "epoch": 1.21, + "learning_rate": 2.8807242271215667e-06, + "loss": 0.6195, + "step": 3278 + }, + { + "epoch": 1.21, + "learning_rate": 2.8806872767582213e-06, + "loss": 0.5549, + "step": 3279 + }, + { + "epoch": 1.21, + "learning_rate": 2.8806503263948764e-06, + "loss": 0.6132, + "step": 3280 + }, + { + "epoch": 1.21, + "learning_rate": 2.880613376031531e-06, + "loss": 0.615, + "step": 3281 + }, + { + "epoch": 1.21, + "learning_rate": 2.880576425668186e-06, + "loss": 0.602, + "step": 3282 + }, + { + "epoch": 1.21, + "learning_rate": 2.8805394753048403e-06, + "loss": 0.568, + "step": 3283 + }, + { + "epoch": 1.21, + "learning_rate": 2.8805025249414954e-06, + "loss": 0.6023, + "step": 3284 + }, + { + "epoch": 1.21, + "learning_rate": 2.88046557457815e-06, + "loss": 0.6606, + "step": 3285 + }, + { + "epoch": 1.21, + "learning_rate": 2.880428624214805e-06, + "loss": 0.6992, + "step": 3286 + }, + { + "epoch": 1.21, + "learning_rate": 2.8803916738514597e-06, + "loss": 0.6854, + "step": 3287 + }, + { + "epoch": 1.21, + "learning_rate": 2.8803547234881143e-06, + "loss": 0.6771, + "step": 3288 + }, + { + "epoch": 1.21, + "learning_rate": 2.880317773124769e-06, + "loss": 0.575, + "step": 3289 + }, + { + "epoch": 1.21, + "learning_rate": 2.8802808227614236e-06, + "loss": 0.625, + "step": 3290 + }, + { + "epoch": 1.22, + "learning_rate": 2.8802438723980787e-06, + "loss": 0.6643, + "step": 3291 + }, + { + "epoch": 1.22, + "learning_rate": 2.8802069220347333e-06, + "loss": 0.6693, + "step": 3292 + }, + { + "epoch": 1.22, + "learning_rate": 2.8801699716713884e-06, + "loss": 0.6289, + "step": 3293 + }, + { + "epoch": 1.22, + "learning_rate": 2.880133021308043e-06, + "loss": 0.6229, + "step": 3294 + }, + { + "epoch": 1.22, + "learning_rate": 2.8800960709446976e-06, + "loss": 0.6698, + "step": 3295 + }, + { + "epoch": 1.22, + "learning_rate": 2.8800591205813523e-06, + "loss": 0.6655, + "step": 3296 + }, + { + "epoch": 1.22, + "learning_rate": 2.880022170218007e-06, + "loss": 0.5949, + "step": 3297 + }, + { + "epoch": 1.22, + "learning_rate": 2.879985219854662e-06, + "loss": 0.6638, + "step": 3298 + }, + { + "epoch": 1.22, + "learning_rate": 2.8799482694913166e-06, + "loss": 0.6328, + "step": 3299 + }, + { + "epoch": 1.22, + "learning_rate": 2.8799113191279717e-06, + "loss": 0.6648, + "step": 3300 + }, + { + "epoch": 1.22, + "learning_rate": 2.8798743687646263e-06, + "loss": 0.6549, + "step": 3301 + }, + { + "epoch": 1.22, + "learning_rate": 2.879837418401281e-06, + "loss": 0.5975, + "step": 3302 + }, + { + "epoch": 1.22, + "learning_rate": 2.8798004680379356e-06, + "loss": 0.7055, + "step": 3303 + }, + { + "epoch": 1.22, + "learning_rate": 2.8797635176745906e-06, + "loss": 0.623, + "step": 3304 + }, + { + "epoch": 1.22, + "learning_rate": 2.8797265673112453e-06, + "loss": 0.6476, + "step": 3305 + }, + { + "epoch": 1.22, + "learning_rate": 2.8796896169479e-06, + "loss": 0.5958, + "step": 3306 + }, + { + "epoch": 1.22, + "learning_rate": 2.879652666584555e-06, + "loss": 0.6978, + "step": 3307 + }, + { + "epoch": 1.22, + "learning_rate": 2.8796157162212096e-06, + "loss": 0.5987, + "step": 3308 + }, + { + "epoch": 1.22, + "learning_rate": 2.8795787658578642e-06, + "loss": 0.6696, + "step": 3309 + }, + { + "epoch": 1.22, + "learning_rate": 2.879541815494519e-06, + "loss": 0.5954, + "step": 3310 + }, + { + "epoch": 1.22, + "learning_rate": 2.879504865131174e-06, + "loss": 0.5564, + "step": 3311 + }, + { + "epoch": 1.22, + "learning_rate": 2.8794679147678286e-06, + "loss": 0.6332, + "step": 3312 + }, + { + "epoch": 1.22, + "learning_rate": 2.8794309644044836e-06, + "loss": 0.6694, + "step": 3313 + }, + { + "epoch": 1.22, + "learning_rate": 2.8793940140411383e-06, + "loss": 0.6281, + "step": 3314 + }, + { + "epoch": 1.22, + "learning_rate": 2.879357063677793e-06, + "loss": 0.6133, + "step": 3315 + }, + { + "epoch": 1.22, + "learning_rate": 2.8793201133144476e-06, + "loss": 0.6389, + "step": 3316 + }, + { + "epoch": 1.22, + "learning_rate": 2.879283162951102e-06, + "loss": 0.6252, + "step": 3317 + }, + { + "epoch": 1.23, + "learning_rate": 2.8792462125877572e-06, + "loss": 0.6941, + "step": 3318 + }, + { + "epoch": 1.23, + "learning_rate": 2.879209262224412e-06, + "loss": 0.5979, + "step": 3319 + }, + { + "epoch": 1.23, + "learning_rate": 2.879172311861067e-06, + "loss": 0.6032, + "step": 3320 + }, + { + "epoch": 1.23, + "learning_rate": 2.8791353614977216e-06, + "loss": 0.6795, + "step": 3321 + }, + { + "epoch": 1.23, + "learning_rate": 2.8790984111343762e-06, + "loss": 0.595, + "step": 3322 + }, + { + "epoch": 1.23, + "learning_rate": 2.879061460771031e-06, + "loss": 0.6616, + "step": 3323 + }, + { + "epoch": 1.23, + "learning_rate": 2.879024510407686e-06, + "loss": 0.6589, + "step": 3324 + }, + { + "epoch": 1.23, + "learning_rate": 2.8789875600443406e-06, + "loss": 0.6937, + "step": 3325 + }, + { + "epoch": 1.23, + "learning_rate": 2.878950609680995e-06, + "loss": 0.6601, + "step": 3326 + }, + { + "epoch": 1.23, + "learning_rate": 2.8789136593176502e-06, + "loss": 0.5705, + "step": 3327 + }, + { + "epoch": 1.23, + "learning_rate": 2.8788767089543045e-06, + "loss": 0.6226, + "step": 3328 + }, + { + "epoch": 1.23, + "learning_rate": 2.8788397585909595e-06, + "loss": 0.5655, + "step": 3329 + }, + { + "epoch": 1.23, + "learning_rate": 2.878802808227614e-06, + "loss": 0.6144, + "step": 3330 + }, + { + "epoch": 1.23, + "learning_rate": 2.8787658578642692e-06, + "loss": 0.6808, + "step": 3331 + }, + { + "epoch": 1.23, + "learning_rate": 2.878728907500924e-06, + "loss": 0.6266, + "step": 3332 + }, + { + "epoch": 1.23, + "learning_rate": 2.878691957137579e-06, + "loss": 0.6497, + "step": 3333 + }, + { + "epoch": 1.23, + "learning_rate": 2.878655006774233e-06, + "loss": 0.6521, + "step": 3334 + }, + { + "epoch": 1.23, + "learning_rate": 2.878618056410888e-06, + "loss": 0.6525, + "step": 3335 + }, + { + "epoch": 1.23, + "learning_rate": 2.878581106047543e-06, + "loss": 0.6001, + "step": 3336 + }, + { + "epoch": 1.23, + "learning_rate": 2.8785441556841975e-06, + "loss": 0.6503, + "step": 3337 + }, + { + "epoch": 1.23, + "learning_rate": 2.8785072053208525e-06, + "loss": 0.5598, + "step": 3338 + }, + { + "epoch": 1.23, + "learning_rate": 2.878470254957507e-06, + "loss": 0.6106, + "step": 3339 + }, + { + "epoch": 1.23, + "learning_rate": 2.878433304594162e-06, + "loss": 0.6138, + "step": 3340 + }, + { + "epoch": 1.23, + "learning_rate": 2.8783963542308164e-06, + "loss": 0.6345, + "step": 3341 + }, + { + "epoch": 1.23, + "learning_rate": 2.8783594038674715e-06, + "loss": 0.6485, + "step": 3342 + }, + { + "epoch": 1.23, + "learning_rate": 2.878322453504126e-06, + "loss": 0.6134, + "step": 3343 + }, + { + "epoch": 1.23, + "learning_rate": 2.878285503140781e-06, + "loss": 0.6242, + "step": 3344 + }, + { + "epoch": 1.24, + "learning_rate": 2.878248552777436e-06, + "loss": 0.6247, + "step": 3345 + }, + { + "epoch": 1.24, + "learning_rate": 2.8782116024140905e-06, + "loss": 0.5743, + "step": 3346 + }, + { + "epoch": 1.24, + "learning_rate": 2.878174652050745e-06, + "loss": 0.6306, + "step": 3347 + }, + { + "epoch": 1.24, + "learning_rate": 2.8781377016873997e-06, + "loss": 0.5929, + "step": 3348 + }, + { + "epoch": 1.24, + "learning_rate": 2.878100751324055e-06, + "loss": 0.5905, + "step": 3349 + }, + { + "epoch": 1.24, + "learning_rate": 2.8780638009607094e-06, + "loss": 0.5982, + "step": 3350 + }, + { + "epoch": 1.24, + "learning_rate": 2.8780268505973645e-06, + "loss": 0.6332, + "step": 3351 + }, + { + "epoch": 1.24, + "learning_rate": 2.877989900234019e-06, + "loss": 0.6135, + "step": 3352 + }, + { + "epoch": 1.24, + "learning_rate": 2.8779529498706738e-06, + "loss": 0.6515, + "step": 3353 + }, + { + "epoch": 1.24, + "learning_rate": 2.8779159995073284e-06, + "loss": 0.639, + "step": 3354 + }, + { + "epoch": 1.24, + "learning_rate": 2.8778790491439835e-06, + "loss": 0.6001, + "step": 3355 + }, + { + "epoch": 1.24, + "learning_rate": 2.877842098780638e-06, + "loss": 0.6107, + "step": 3356 + }, + { + "epoch": 1.24, + "learning_rate": 2.8778051484172927e-06, + "loss": 0.5756, + "step": 3357 + }, + { + "epoch": 1.24, + "learning_rate": 2.8777681980539478e-06, + "loss": 0.6604, + "step": 3358 + }, + { + "epoch": 1.24, + "learning_rate": 2.8777312476906024e-06, + "loss": 0.6054, + "step": 3359 + }, + { + "epoch": 1.24, + "learning_rate": 2.877694297327257e-06, + "loss": 0.7048, + "step": 3360 + }, + { + "epoch": 1.24, + "learning_rate": 2.8776573469639117e-06, + "loss": 0.5853, + "step": 3361 + }, + { + "epoch": 1.24, + "learning_rate": 2.8776203966005668e-06, + "loss": 0.568, + "step": 3362 + }, + { + "epoch": 1.24, + "learning_rate": 2.8775834462372214e-06, + "loss": 0.5418, + "step": 3363 + }, + { + "epoch": 1.24, + "learning_rate": 2.8775464958738765e-06, + "loss": 0.554, + "step": 3364 + }, + { + "epoch": 1.24, + "learning_rate": 2.877509545510531e-06, + "loss": 0.6458, + "step": 3365 + }, + { + "epoch": 1.24, + "learning_rate": 2.8774725951471857e-06, + "loss": 0.647, + "step": 3366 + }, + { + "epoch": 1.24, + "learning_rate": 2.8774356447838404e-06, + "loss": 0.5638, + "step": 3367 + }, + { + "epoch": 1.24, + "learning_rate": 2.877398694420495e-06, + "loss": 0.6441, + "step": 3368 + }, + { + "epoch": 1.24, + "learning_rate": 2.87736174405715e-06, + "loss": 0.6109, + "step": 3369 + }, + { + "epoch": 1.24, + "learning_rate": 2.8773247936938047e-06, + "loss": 0.6611, + "step": 3370 + }, + { + "epoch": 1.24, + "learning_rate": 2.8772878433304598e-06, + "loss": 0.7233, + "step": 3371 + }, + { + "epoch": 1.25, + "learning_rate": 2.8772508929671144e-06, + "loss": 0.6042, + "step": 3372 + }, + { + "epoch": 1.25, + "learning_rate": 2.877213942603769e-06, + "loss": 0.62, + "step": 3373 + }, + { + "epoch": 1.25, + "learning_rate": 2.8771769922404237e-06, + "loss": 0.6771, + "step": 3374 + }, + { + "epoch": 1.25, + "learning_rate": 2.8771400418770787e-06, + "loss": 0.6768, + "step": 3375 + }, + { + "epoch": 1.25, + "learning_rate": 2.8771030915137334e-06, + "loss": 0.6806, + "step": 3376 + }, + { + "epoch": 1.25, + "learning_rate": 2.877066141150388e-06, + "loss": 0.5888, + "step": 3377 + }, + { + "epoch": 1.25, + "learning_rate": 2.877029190787043e-06, + "loss": 0.7121, + "step": 3378 + }, + { + "epoch": 1.25, + "learning_rate": 2.8769922404236973e-06, + "loss": 0.5824, + "step": 3379 + }, + { + "epoch": 1.25, + "learning_rate": 2.8769552900603523e-06, + "loss": 0.5846, + "step": 3380 + }, + { + "epoch": 1.25, + "learning_rate": 2.876918339697007e-06, + "loss": 0.6074, + "step": 3381 + }, + { + "epoch": 1.25, + "learning_rate": 2.876881389333662e-06, + "loss": 0.5828, + "step": 3382 + }, + { + "epoch": 1.25, + "learning_rate": 2.8768444389703167e-06, + "loss": 0.6857, + "step": 3383 + }, + { + "epoch": 1.25, + "learning_rate": 2.8768074886069717e-06, + "loss": 0.6424, + "step": 3384 + }, + { + "epoch": 1.25, + "learning_rate": 2.876770538243626e-06, + "loss": 0.5595, + "step": 3385 + }, + { + "epoch": 1.25, + "learning_rate": 2.8767335878802806e-06, + "loss": 0.6467, + "step": 3386 + }, + { + "epoch": 1.25, + "learning_rate": 2.8766966375169356e-06, + "loss": 0.6864, + "step": 3387 + }, + { + "epoch": 1.25, + "learning_rate": 2.8766596871535903e-06, + "loss": 0.7213, + "step": 3388 + }, + { + "epoch": 1.25, + "learning_rate": 2.8766227367902453e-06, + "loss": 0.5876, + "step": 3389 + }, + { + "epoch": 1.25, + "learning_rate": 2.8765857864269e-06, + "loss": 0.6465, + "step": 3390 + }, + { + "epoch": 1.25, + "learning_rate": 2.8765488360635546e-06, + "loss": 0.6923, + "step": 3391 + }, + { + "epoch": 1.25, + "learning_rate": 2.8765118857002092e-06, + "loss": 0.612, + "step": 3392 + }, + { + "epoch": 1.25, + "learning_rate": 2.8764749353368643e-06, + "loss": 0.563, + "step": 3393 + }, + { + "epoch": 1.25, + "learning_rate": 2.876437984973519e-06, + "loss": 0.6296, + "step": 3394 + }, + { + "epoch": 1.25, + "learning_rate": 2.876401034610174e-06, + "loss": 0.6414, + "step": 3395 + }, + { + "epoch": 1.25, + "learning_rate": 2.8763640842468286e-06, + "loss": 0.6239, + "step": 3396 + }, + { + "epoch": 1.25, + "learning_rate": 2.8763271338834833e-06, + "loss": 0.6443, + "step": 3397 + }, + { + "epoch": 1.25, + "learning_rate": 2.876290183520138e-06, + "loss": 0.6149, + "step": 3398 + }, + { + "epoch": 1.26, + "learning_rate": 2.8762532331567925e-06, + "loss": 0.6242, + "step": 3399 + }, + { + "epoch": 1.26, + "learning_rate": 2.8762162827934476e-06, + "loss": 0.6187, + "step": 3400 + }, + { + "epoch": 1.26, + "learning_rate": 2.8761793324301022e-06, + "loss": 0.6131, + "step": 3401 + }, + { + "epoch": 1.26, + "learning_rate": 2.8761423820667573e-06, + "loss": 0.6205, + "step": 3402 + }, + { + "epoch": 1.26, + "learning_rate": 2.876105431703412e-06, + "loss": 0.6247, + "step": 3403 + }, + { + "epoch": 1.26, + "learning_rate": 2.8760684813400666e-06, + "loss": 0.6071, + "step": 3404 + }, + { + "epoch": 1.26, + "learning_rate": 2.876031530976721e-06, + "loss": 0.7212, + "step": 3405 + }, + { + "epoch": 1.26, + "learning_rate": 2.875994580613376e-06, + "loss": 0.6383, + "step": 3406 + }, + { + "epoch": 1.26, + "learning_rate": 2.875957630250031e-06, + "loss": 0.6888, + "step": 3407 + }, + { + "epoch": 1.26, + "learning_rate": 2.8759206798866855e-06, + "loss": 0.5703, + "step": 3408 + }, + { + "epoch": 1.26, + "learning_rate": 2.8758837295233406e-06, + "loss": 0.6171, + "step": 3409 + }, + { + "epoch": 1.26, + "learning_rate": 2.8758467791599952e-06, + "loss": 0.579, + "step": 3410 + }, + { + "epoch": 1.26, + "learning_rate": 2.87580982879665e-06, + "loss": 0.6565, + "step": 3411 + }, + { + "epoch": 1.26, + "learning_rate": 2.8757728784333045e-06, + "loss": 0.6418, + "step": 3412 + }, + { + "epoch": 1.26, + "learning_rate": 2.8757359280699596e-06, + "loss": 0.6496, + "step": 3413 + }, + { + "epoch": 1.26, + "learning_rate": 2.875698977706614e-06, + "loss": 0.6169, + "step": 3414 + }, + { + "epoch": 1.26, + "learning_rate": 2.8756620273432693e-06, + "loss": 0.6278, + "step": 3415 + }, + { + "epoch": 1.26, + "learning_rate": 2.875625076979924e-06, + "loss": 0.6296, + "step": 3416 + }, + { + "epoch": 1.26, + "learning_rate": 2.8755881266165785e-06, + "loss": 0.6542, + "step": 3417 + }, + { + "epoch": 1.26, + "learning_rate": 2.875551176253233e-06, + "loss": 0.6211, + "step": 3418 + }, + { + "epoch": 1.26, + "learning_rate": 2.875514225889888e-06, + "loss": 0.6357, + "step": 3419 + }, + { + "epoch": 1.26, + "learning_rate": 2.875477275526543e-06, + "loss": 0.5841, + "step": 3420 + }, + { + "epoch": 1.26, + "learning_rate": 2.8754403251631975e-06, + "loss": 0.595, + "step": 3421 + }, + { + "epoch": 1.26, + "learning_rate": 2.8754033747998526e-06, + "loss": 0.7045, + "step": 3422 + }, + { + "epoch": 1.26, + "learning_rate": 2.8753664244365068e-06, + "loss": 0.5824, + "step": 3423 + }, + { + "epoch": 1.26, + "learning_rate": 2.875329474073162e-06, + "loss": 0.6024, + "step": 3424 + }, + { + "epoch": 1.26, + "learning_rate": 2.8752925237098165e-06, + "loss": 0.6567, + "step": 3425 + }, + { + "epoch": 1.27, + "learning_rate": 2.875255573346471e-06, + "loss": 0.6673, + "step": 3426 + }, + { + "epoch": 1.27, + "learning_rate": 2.875218622983126e-06, + "loss": 0.5457, + "step": 3427 + }, + { + "epoch": 1.27, + "learning_rate": 2.875181672619781e-06, + "loss": 0.6639, + "step": 3428 + }, + { + "epoch": 1.27, + "learning_rate": 2.8751447222564354e-06, + "loss": 0.6701, + "step": 3429 + }, + { + "epoch": 1.27, + "learning_rate": 2.87510777189309e-06, + "loss": 0.6681, + "step": 3430 + }, + { + "epoch": 1.27, + "learning_rate": 2.875070821529745e-06, + "loss": 0.6258, + "step": 3431 + }, + { + "epoch": 1.27, + "learning_rate": 2.8750338711663998e-06, + "loss": 0.5715, + "step": 3432 + }, + { + "epoch": 1.27, + "learning_rate": 2.874996920803055e-06, + "loss": 0.593, + "step": 3433 + }, + { + "epoch": 1.27, + "learning_rate": 2.8749599704397095e-06, + "loss": 0.6631, + "step": 3434 + }, + { + "epoch": 1.27, + "learning_rate": 2.874923020076364e-06, + "loss": 0.6396, + "step": 3435 + }, + { + "epoch": 1.27, + "learning_rate": 2.8748860697130187e-06, + "loss": 0.6096, + "step": 3436 + }, + { + "epoch": 1.27, + "learning_rate": 2.8748491193496734e-06, + "loss": 0.5831, + "step": 3437 + }, + { + "epoch": 1.27, + "learning_rate": 2.8748121689863284e-06, + "loss": 0.5947, + "step": 3438 + }, + { + "epoch": 1.27, + "learning_rate": 2.874775218622983e-06, + "loss": 0.622, + "step": 3439 + }, + { + "epoch": 1.27, + "learning_rate": 2.874738268259638e-06, + "loss": 0.5869, + "step": 3440 + }, + { + "epoch": 1.27, + "learning_rate": 2.8747013178962928e-06, + "loss": 0.6477, + "step": 3441 + }, + { + "epoch": 1.27, + "learning_rate": 2.8746643675329474e-06, + "loss": 0.6296, + "step": 3442 + }, + { + "epoch": 1.27, + "learning_rate": 2.874627417169602e-06, + "loss": 0.5803, + "step": 3443 + }, + { + "epoch": 1.27, + "learning_rate": 2.874590466806257e-06, + "loss": 0.6135, + "step": 3444 + }, + { + "epoch": 1.27, + "learning_rate": 2.8745535164429117e-06, + "loss": 0.6097, + "step": 3445 + }, + { + "epoch": 1.27, + "learning_rate": 2.8745165660795664e-06, + "loss": 0.609, + "step": 3446 + }, + { + "epoch": 1.27, + "learning_rate": 2.8744796157162214e-06, + "loss": 0.5726, + "step": 3447 + }, + { + "epoch": 1.27, + "learning_rate": 2.874442665352876e-06, + "loss": 0.5765, + "step": 3448 + }, + { + "epoch": 1.27, + "learning_rate": 2.8744057149895307e-06, + "loss": 0.6126, + "step": 3449 + }, + { + "epoch": 1.27, + "learning_rate": 2.8743687646261854e-06, + "loss": 0.5982, + "step": 3450 + }, + { + "epoch": 1.27, + "learning_rate": 2.8743318142628404e-06, + "loss": 0.6984, + "step": 3451 + }, + { + "epoch": 1.27, + "learning_rate": 2.874294863899495e-06, + "loss": 0.6155, + "step": 3452 + }, + { + "epoch": 1.28, + "learning_rate": 2.87425791353615e-06, + "loss": 0.6456, + "step": 3453 + }, + { + "epoch": 1.28, + "learning_rate": 2.8742209631728047e-06, + "loss": 0.6593, + "step": 3454 + }, + { + "epoch": 1.28, + "learning_rate": 2.8741840128094594e-06, + "loss": 0.5536, + "step": 3455 + }, + { + "epoch": 1.28, + "learning_rate": 2.874147062446114e-06, + "loss": 0.7036, + "step": 3456 + }, + { + "epoch": 1.28, + "learning_rate": 2.8741101120827687e-06, + "loss": 0.6469, + "step": 3457 + }, + { + "epoch": 1.28, + "learning_rate": 2.8740731617194237e-06, + "loss": 0.6321, + "step": 3458 + }, + { + "epoch": 1.28, + "learning_rate": 2.8740362113560783e-06, + "loss": 0.6003, + "step": 3459 + }, + { + "epoch": 1.28, + "learning_rate": 2.8739992609927334e-06, + "loss": 0.661, + "step": 3460 + }, + { + "epoch": 1.28, + "learning_rate": 2.873962310629388e-06, + "loss": 0.6248, + "step": 3461 + }, + { + "epoch": 1.28, + "learning_rate": 2.8739253602660427e-06, + "loss": 0.6851, + "step": 3462 + }, + { + "epoch": 1.28, + "learning_rate": 2.8738884099026973e-06, + "loss": 0.5818, + "step": 3463 + }, + { + "epoch": 1.28, + "learning_rate": 2.8738514595393524e-06, + "loss": 0.5713, + "step": 3464 + }, + { + "epoch": 1.28, + "learning_rate": 2.873814509176007e-06, + "loss": 0.5321, + "step": 3465 + }, + { + "epoch": 1.28, + "learning_rate": 2.8737775588126617e-06, + "loss": 0.6358, + "step": 3466 + }, + { + "epoch": 1.28, + "learning_rate": 2.8737406084493167e-06, + "loss": 0.6604, + "step": 3467 + }, + { + "epoch": 1.28, + "learning_rate": 2.873703658085971e-06, + "loss": 0.5813, + "step": 3468 + }, + { + "epoch": 1.28, + "learning_rate": 2.873666707722626e-06, + "loss": 0.6604, + "step": 3469 + }, + { + "epoch": 1.28, + "learning_rate": 2.8736297573592806e-06, + "loss": 0.5888, + "step": 3470 + }, + { + "epoch": 1.28, + "learning_rate": 2.8735928069959357e-06, + "loss": 0.6596, + "step": 3471 + }, + { + "epoch": 1.28, + "learning_rate": 2.8735558566325903e-06, + "loss": 0.6336, + "step": 3472 + }, + { + "epoch": 1.28, + "learning_rate": 2.8735189062692454e-06, + "loss": 0.6271, + "step": 3473 + }, + { + "epoch": 1.28, + "learning_rate": 2.8734819559058996e-06, + "loss": 0.6483, + "step": 3474 + }, + { + "epoch": 1.28, + "learning_rate": 2.8734450055425547e-06, + "loss": 0.6825, + "step": 3475 + }, + { + "epoch": 1.28, + "learning_rate": 2.8734080551792093e-06, + "loss": 0.6575, + "step": 3476 + }, + { + "epoch": 1.28, + "learning_rate": 2.873371104815864e-06, + "loss": 0.6088, + "step": 3477 + }, + { + "epoch": 1.28, + "learning_rate": 2.873334154452519e-06, + "loss": 0.5651, + "step": 3478 + }, + { + "epoch": 1.28, + "learning_rate": 2.8732972040891736e-06, + "loss": 0.5967, + "step": 3479 + }, + { + "epoch": 1.29, + "learning_rate": 2.8732602537258283e-06, + "loss": 0.5991, + "step": 3480 + }, + { + "epoch": 1.29, + "learning_rate": 2.873223303362483e-06, + "loss": 0.6485, + "step": 3481 + }, + { + "epoch": 1.29, + "learning_rate": 2.873186352999138e-06, + "loss": 0.6568, + "step": 3482 + }, + { + "epoch": 1.29, + "learning_rate": 2.8731494026357926e-06, + "loss": 0.602, + "step": 3483 + }, + { + "epoch": 1.29, + "learning_rate": 2.8731124522724476e-06, + "loss": 0.631, + "step": 3484 + }, + { + "epoch": 1.29, + "learning_rate": 2.8730755019091023e-06, + "loss": 0.6322, + "step": 3485 + }, + { + "epoch": 1.29, + "learning_rate": 2.873038551545757e-06, + "loss": 0.6427, + "step": 3486 + }, + { + "epoch": 1.29, + "learning_rate": 2.8730016011824116e-06, + "loss": 0.6655, + "step": 3487 + }, + { + "epoch": 1.29, + "learning_rate": 2.872964650819066e-06, + "loss": 0.635, + "step": 3488 + }, + { + "epoch": 1.29, + "learning_rate": 2.8729277004557213e-06, + "loss": 0.5685, + "step": 3489 + }, + { + "epoch": 1.29, + "learning_rate": 2.872890750092376e-06, + "loss": 0.584, + "step": 3490 + }, + { + "epoch": 1.29, + "learning_rate": 2.872853799729031e-06, + "loss": 0.5468, + "step": 3491 + }, + { + "epoch": 1.29, + "learning_rate": 2.8728168493656856e-06, + "loss": 0.6746, + "step": 3492 + }, + { + "epoch": 1.29, + "learning_rate": 2.8727798990023402e-06, + "loss": 0.6684, + "step": 3493 + }, + { + "epoch": 1.29, + "learning_rate": 2.872742948638995e-06, + "loss": 0.5227, + "step": 3494 + }, + { + "epoch": 1.29, + "learning_rate": 2.87270599827565e-06, + "loss": 0.6332, + "step": 3495 + }, + { + "epoch": 1.29, + "learning_rate": 2.8726690479123046e-06, + "loss": 0.5714, + "step": 3496 + }, + { + "epoch": 1.29, + "learning_rate": 2.872632097548959e-06, + "loss": 0.6282, + "step": 3497 + }, + { + "epoch": 1.29, + "learning_rate": 2.8725951471856143e-06, + "loss": 0.6014, + "step": 3498 + }, + { + "epoch": 1.29, + "learning_rate": 2.872558196822269e-06, + "loss": 0.6317, + "step": 3499 + }, + { + "epoch": 1.29, + "learning_rate": 2.8725212464589235e-06, + "loss": 0.5299, + "step": 3500 + }, + { + "epoch": 1.29, + "learning_rate": 2.872484296095578e-06, + "loss": 0.5785, + "step": 3501 + }, + { + "epoch": 1.29, + "learning_rate": 2.8724473457322332e-06, + "loss": 0.5699, + "step": 3502 + }, + { + "epoch": 1.29, + "learning_rate": 2.872410395368888e-06, + "loss": 0.6333, + "step": 3503 + }, + { + "epoch": 1.29, + "learning_rate": 2.872373445005543e-06, + "loss": 0.6571, + "step": 3504 + }, + { + "epoch": 1.29, + "learning_rate": 2.8723364946421976e-06, + "loss": 0.5725, + "step": 3505 + }, + { + "epoch": 1.29, + "learning_rate": 2.872299544278852e-06, + "loss": 0.6296, + "step": 3506 + }, + { + "epoch": 1.3, + "learning_rate": 2.872262593915507e-06, + "loss": 0.6103, + "step": 3507 + }, + { + "epoch": 1.3, + "learning_rate": 2.8722256435521615e-06, + "loss": 0.5468, + "step": 3508 + }, + { + "epoch": 1.3, + "learning_rate": 2.8721886931888165e-06, + "loss": 0.6097, + "step": 3509 + }, + { + "epoch": 1.3, + "learning_rate": 2.872151742825471e-06, + "loss": 0.6229, + "step": 3510 + }, + { + "epoch": 1.3, + "learning_rate": 2.8721147924621262e-06, + "loss": 0.5769, + "step": 3511 + }, + { + "epoch": 1.3, + "learning_rate": 2.872077842098781e-06, + "loss": 0.6577, + "step": 3512 + }, + { + "epoch": 1.3, + "learning_rate": 2.8720408917354355e-06, + "loss": 0.5252, + "step": 3513 + }, + { + "epoch": 1.3, + "learning_rate": 2.87200394137209e-06, + "loss": 0.6103, + "step": 3514 + }, + { + "epoch": 1.3, + "learning_rate": 2.871966991008745e-06, + "loss": 0.6583, + "step": 3515 + }, + { + "epoch": 1.3, + "learning_rate": 2.8719300406454e-06, + "loss": 0.6382, + "step": 3516 + }, + { + "epoch": 1.3, + "learning_rate": 2.8718930902820545e-06, + "loss": 0.614, + "step": 3517 + }, + { + "epoch": 1.3, + "learning_rate": 2.8718561399187095e-06, + "loss": 0.6682, + "step": 3518 + }, + { + "epoch": 1.3, + "learning_rate": 2.8718191895553637e-06, + "loss": 0.5905, + "step": 3519 + }, + { + "epoch": 1.3, + "learning_rate": 2.871782239192019e-06, + "loss": 0.5861, + "step": 3520 + }, + { + "epoch": 1.3, + "learning_rate": 2.8717452888286734e-06, + "loss": 0.6483, + "step": 3521 + }, + { + "epoch": 1.3, + "learning_rate": 2.8717083384653285e-06, + "loss": 0.5962, + "step": 3522 + }, + { + "epoch": 1.3, + "learning_rate": 2.871671388101983e-06, + "loss": 0.6606, + "step": 3523 + }, + { + "epoch": 1.3, + "learning_rate": 2.871634437738638e-06, + "loss": 0.576, + "step": 3524 + }, + { + "epoch": 1.3, + "learning_rate": 2.8715974873752924e-06, + "loss": 0.6988, + "step": 3525 + }, + { + "epoch": 1.3, + "learning_rate": 2.871560537011947e-06, + "loss": 0.6267, + "step": 3526 + }, + { + "epoch": 1.3, + "learning_rate": 2.871523586648602e-06, + "loss": 0.6083, + "step": 3527 + }, + { + "epoch": 1.3, + "learning_rate": 2.8714866362852567e-06, + "loss": 0.635, + "step": 3528 + }, + { + "epoch": 1.3, + "learning_rate": 2.871449685921912e-06, + "loss": 0.6303, + "step": 3529 + }, + { + "epoch": 1.3, + "learning_rate": 2.8714127355585664e-06, + "loss": 0.7399, + "step": 3530 + }, + { + "epoch": 1.3, + "learning_rate": 2.871375785195221e-06, + "loss": 0.6345, + "step": 3531 + }, + { + "epoch": 1.3, + "learning_rate": 2.8713388348318757e-06, + "loss": 0.6317, + "step": 3532 + }, + { + "epoch": 1.3, + "learning_rate": 2.8713018844685308e-06, + "loss": 0.6579, + "step": 3533 + }, + { + "epoch": 1.31, + "learning_rate": 2.8712649341051854e-06, + "loss": 0.5437, + "step": 3534 + }, + { + "epoch": 1.31, + "learning_rate": 2.87122798374184e-06, + "loss": 0.6279, + "step": 3535 + }, + { + "epoch": 1.31, + "learning_rate": 2.871191033378495e-06, + "loss": 0.553, + "step": 3536 + }, + { + "epoch": 1.31, + "learning_rate": 2.8711540830151497e-06, + "loss": 0.5637, + "step": 3537 + }, + { + "epoch": 1.31, + "learning_rate": 2.8711171326518044e-06, + "loss": 0.6331, + "step": 3538 + }, + { + "epoch": 1.31, + "learning_rate": 2.871080182288459e-06, + "loss": 0.5732, + "step": 3539 + }, + { + "epoch": 1.31, + "learning_rate": 2.871043231925114e-06, + "loss": 0.7318, + "step": 3540 + }, + { + "epoch": 1.31, + "learning_rate": 2.8710062815617687e-06, + "loss": 0.6994, + "step": 3541 + }, + { + "epoch": 1.31, + "learning_rate": 2.8709693311984238e-06, + "loss": 0.6212, + "step": 3542 + }, + { + "epoch": 1.31, + "learning_rate": 2.8709323808350784e-06, + "loss": 0.5357, + "step": 3543 + }, + { + "epoch": 1.31, + "learning_rate": 2.870895430471733e-06, + "loss": 0.6429, + "step": 3544 + }, + { + "epoch": 1.31, + "learning_rate": 2.8708584801083877e-06, + "loss": 0.5698, + "step": 3545 + }, + { + "epoch": 1.31, + "learning_rate": 2.8708215297450423e-06, + "loss": 0.6581, + "step": 3546 + }, + { + "epoch": 1.31, + "learning_rate": 2.8707845793816974e-06, + "loss": 0.6767, + "step": 3547 + }, + { + "epoch": 1.31, + "learning_rate": 2.870747629018352e-06, + "loss": 0.6981, + "step": 3548 + }, + { + "epoch": 1.31, + "learning_rate": 2.870710678655007e-06, + "loss": 0.6416, + "step": 3549 + }, + { + "epoch": 1.31, + "learning_rate": 2.8706737282916617e-06, + "loss": 0.6897, + "step": 3550 + }, + { + "epoch": 1.31, + "learning_rate": 2.8706367779283163e-06, + "loss": 0.6691, + "step": 3551 + }, + { + "epoch": 1.31, + "learning_rate": 2.870599827564971e-06, + "loss": 0.5839, + "step": 3552 + }, + { + "epoch": 1.31, + "learning_rate": 2.870562877201626e-06, + "loss": 0.5741, + "step": 3553 + }, + { + "epoch": 1.31, + "learning_rate": 2.8705259268382807e-06, + "loss": 0.7064, + "step": 3554 + }, + { + "epoch": 1.31, + "learning_rate": 2.8704889764749353e-06, + "loss": 0.6234, + "step": 3555 + }, + { + "epoch": 1.31, + "learning_rate": 2.8704520261115904e-06, + "loss": 0.6783, + "step": 3556 + }, + { + "epoch": 1.31, + "learning_rate": 2.870415075748245e-06, + "loss": 0.6226, + "step": 3557 + }, + { + "epoch": 1.31, + "learning_rate": 2.8703781253848996e-06, + "loss": 0.5874, + "step": 3558 + }, + { + "epoch": 1.31, + "learning_rate": 2.8703411750215543e-06, + "loss": 0.6478, + "step": 3559 + }, + { + "epoch": 1.31, + "learning_rate": 2.8703042246582093e-06, + "loss": 0.5923, + "step": 3560 + }, + { + "epoch": 1.31, + "learning_rate": 2.870267274294864e-06, + "loss": 0.5861, + "step": 3561 + }, + { + "epoch": 1.32, + "learning_rate": 2.870230323931519e-06, + "loss": 0.603, + "step": 3562 + }, + { + "epoch": 1.32, + "learning_rate": 2.8701933735681737e-06, + "loss": 0.5845, + "step": 3563 + }, + { + "epoch": 1.32, + "learning_rate": 2.8701564232048283e-06, + "loss": 0.6666, + "step": 3564 + }, + { + "epoch": 1.32, + "learning_rate": 2.870119472841483e-06, + "loss": 0.6447, + "step": 3565 + }, + { + "epoch": 1.32, + "learning_rate": 2.8700825224781376e-06, + "loss": 0.535, + "step": 3566 + }, + { + "epoch": 1.32, + "learning_rate": 2.8700455721147926e-06, + "loss": 0.5753, + "step": 3567 + }, + { + "epoch": 1.32, + "learning_rate": 2.8700086217514473e-06, + "loss": 0.6377, + "step": 3568 + }, + { + "epoch": 1.32, + "learning_rate": 2.8699716713881023e-06, + "loss": 0.6362, + "step": 3569 + }, + { + "epoch": 1.32, + "learning_rate": 2.8699347210247565e-06, + "loss": 0.6031, + "step": 3570 + }, + { + "epoch": 1.32, + "learning_rate": 2.8698977706614116e-06, + "loss": 0.7061, + "step": 3571 + }, + { + "epoch": 1.32, + "learning_rate": 2.8698608202980662e-06, + "loss": 0.6795, + "step": 3572 + }, + { + "epoch": 1.32, + "learning_rate": 2.8698238699347213e-06, + "loss": 0.6936, + "step": 3573 + }, + { + "epoch": 1.32, + "learning_rate": 2.869786919571376e-06, + "loss": 0.6202, + "step": 3574 + }, + { + "epoch": 1.32, + "learning_rate": 2.8697499692080306e-06, + "loss": 0.7065, + "step": 3575 + }, + { + "epoch": 1.32, + "learning_rate": 2.8697130188446852e-06, + "loss": 0.6013, + "step": 3576 + }, + { + "epoch": 1.32, + "learning_rate": 2.86967606848134e-06, + "loss": 0.575, + "step": 3577 + }, + { + "epoch": 1.32, + "learning_rate": 2.869639118117995e-06, + "loss": 0.6351, + "step": 3578 + }, + { + "epoch": 1.32, + "learning_rate": 2.8696021677546495e-06, + "loss": 0.6686, + "step": 3579 + }, + { + "epoch": 1.32, + "learning_rate": 2.8695652173913046e-06, + "loss": 0.6051, + "step": 3580 + }, + { + "epoch": 1.32, + "learning_rate": 2.8695282670279592e-06, + "loss": 0.7179, + "step": 3581 + }, + { + "epoch": 1.32, + "learning_rate": 2.869491316664614e-06, + "loss": 0.5838, + "step": 3582 + }, + { + "epoch": 1.32, + "learning_rate": 2.8694543663012685e-06, + "loss": 0.6472, + "step": 3583 + }, + { + "epoch": 1.32, + "learning_rate": 2.8694174159379236e-06, + "loss": 0.5699, + "step": 3584 + }, + { + "epoch": 1.32, + "learning_rate": 2.869380465574578e-06, + "loss": 0.616, + "step": 3585 + }, + { + "epoch": 1.32, + "learning_rate": 2.869343515211233e-06, + "loss": 0.635, + "step": 3586 + }, + { + "epoch": 1.32, + "learning_rate": 2.869306564847888e-06, + "loss": 0.6475, + "step": 3587 + }, + { + "epoch": 1.32, + "learning_rate": 2.8692696144845425e-06, + "loss": 0.6467, + "step": 3588 + }, + { + "epoch": 1.33, + "learning_rate": 2.869232664121197e-06, + "loss": 0.6713, + "step": 3589 + }, + { + "epoch": 1.33, + "learning_rate": 2.869195713757852e-06, + "loss": 0.5558, + "step": 3590 + }, + { + "epoch": 1.33, + "learning_rate": 2.869158763394507e-06, + "loss": 0.5473, + "step": 3591 + }, + { + "epoch": 1.33, + "learning_rate": 2.8691218130311615e-06, + "loss": 0.6589, + "step": 3592 + }, + { + "epoch": 1.33, + "learning_rate": 2.8690848626678166e-06, + "loss": 0.6595, + "step": 3593 + }, + { + "epoch": 1.33, + "learning_rate": 2.869047912304471e-06, + "loss": 0.6671, + "step": 3594 + }, + { + "epoch": 1.33, + "learning_rate": 2.869010961941126e-06, + "loss": 0.5791, + "step": 3595 + }, + { + "epoch": 1.33, + "learning_rate": 2.8689740115777805e-06, + "loss": 0.6865, + "step": 3596 + }, + { + "epoch": 1.33, + "learning_rate": 2.868937061214435e-06, + "loss": 0.5897, + "step": 3597 + }, + { + "epoch": 1.33, + "learning_rate": 2.86890011085109e-06, + "loss": 0.6539, + "step": 3598 + }, + { + "epoch": 1.33, + "learning_rate": 2.868863160487745e-06, + "loss": 0.6615, + "step": 3599 + }, + { + "epoch": 1.33, + "learning_rate": 2.8688262101244e-06, + "loss": 0.6418, + "step": 3600 + }, + { + "epoch": 1.33, + "learning_rate": 2.8687892597610545e-06, + "loss": 0.6906, + "step": 3601 + }, + { + "epoch": 1.33, + "learning_rate": 2.868752309397709e-06, + "loss": 0.6234, + "step": 3602 + }, + { + "epoch": 1.33, + "learning_rate": 2.8687153590343638e-06, + "loss": 0.5894, + "step": 3603 + }, + { + "epoch": 1.33, + "learning_rate": 2.868678408671019e-06, + "loss": 0.662, + "step": 3604 + }, + { + "epoch": 1.33, + "learning_rate": 2.8686414583076735e-06, + "loss": 0.6134, + "step": 3605 + }, + { + "epoch": 1.33, + "learning_rate": 2.868604507944328e-06, + "loss": 0.6359, + "step": 3606 + }, + { + "epoch": 1.33, + "learning_rate": 2.868567557580983e-06, + "loss": 0.678, + "step": 3607 + }, + { + "epoch": 1.33, + "learning_rate": 2.868530607217638e-06, + "loss": 0.6202, + "step": 3608 + }, + { + "epoch": 1.33, + "learning_rate": 2.8684936568542924e-06, + "loss": 0.6439, + "step": 3609 + }, + { + "epoch": 1.33, + "learning_rate": 2.868456706490947e-06, + "loss": 0.6334, + "step": 3610 + }, + { + "epoch": 1.33, + "learning_rate": 2.868419756127602e-06, + "loss": 0.6306, + "step": 3611 + }, + { + "epoch": 1.33, + "learning_rate": 2.8683828057642568e-06, + "loss": 0.6804, + "step": 3612 + }, + { + "epoch": 1.33, + "learning_rate": 2.868345855400912e-06, + "loss": 0.6534, + "step": 3613 + }, + { + "epoch": 1.33, + "learning_rate": 2.868308905037566e-06, + "loss": 0.6575, + "step": 3614 + }, + { + "epoch": 1.33, + "learning_rate": 2.8682719546742207e-06, + "loss": 0.6047, + "step": 3615 + }, + { + "epoch": 1.34, + "learning_rate": 2.8682350043108758e-06, + "loss": 0.5703, + "step": 3616 + }, + { + "epoch": 1.34, + "learning_rate": 2.8681980539475304e-06, + "loss": 0.6025, + "step": 3617 + }, + { + "epoch": 1.34, + "learning_rate": 2.8681611035841854e-06, + "loss": 0.6196, + "step": 3618 + }, + { + "epoch": 1.34, + "learning_rate": 2.86812415322084e-06, + "loss": 0.6202, + "step": 3619 + }, + { + "epoch": 1.34, + "learning_rate": 2.8680872028574947e-06, + "loss": 0.604, + "step": 3620 + }, + { + "epoch": 1.34, + "learning_rate": 2.8680502524941494e-06, + "loss": 0.541, + "step": 3621 + }, + { + "epoch": 1.34, + "learning_rate": 2.8680133021308044e-06, + "loss": 0.6368, + "step": 3622 + }, + { + "epoch": 1.34, + "learning_rate": 2.867976351767459e-06, + "loss": 0.5986, + "step": 3623 + }, + { + "epoch": 1.34, + "learning_rate": 2.867939401404114e-06, + "loss": 0.6494, + "step": 3624 + }, + { + "epoch": 1.34, + "learning_rate": 2.8679024510407687e-06, + "loss": 0.5964, + "step": 3625 + }, + { + "epoch": 1.34, + "learning_rate": 2.8678655006774234e-06, + "loss": 0.6022, + "step": 3626 + }, + { + "epoch": 1.34, + "learning_rate": 2.867828550314078e-06, + "loss": 0.5923, + "step": 3627 + }, + { + "epoch": 1.34, + "learning_rate": 2.8677915999507327e-06, + "loss": 0.6556, + "step": 3628 + }, + { + "epoch": 1.34, + "learning_rate": 2.8677546495873877e-06, + "loss": 0.619, + "step": 3629 + }, + { + "epoch": 1.34, + "learning_rate": 2.8677176992240424e-06, + "loss": 0.6408, + "step": 3630 + }, + { + "epoch": 1.34, + "learning_rate": 2.8676807488606974e-06, + "loss": 0.6559, + "step": 3631 + }, + { + "epoch": 1.34, + "learning_rate": 2.867643798497352e-06, + "loss": 0.6643, + "step": 3632 + }, + { + "epoch": 1.34, + "learning_rate": 2.8676068481340067e-06, + "loss": 0.6142, + "step": 3633 + }, + { + "epoch": 1.34, + "learning_rate": 2.8675698977706613e-06, + "loss": 0.6474, + "step": 3634 + }, + { + "epoch": 1.34, + "learning_rate": 2.867532947407316e-06, + "loss": 0.5825, + "step": 3635 + }, + { + "epoch": 1.34, + "learning_rate": 2.867495997043971e-06, + "loss": 0.6403, + "step": 3636 + }, + { + "epoch": 1.34, + "learning_rate": 2.8674590466806257e-06, + "loss": 0.5791, + "step": 3637 + }, + { + "epoch": 1.34, + "learning_rate": 2.8674220963172807e-06, + "loss": 0.6127, + "step": 3638 + }, + { + "epoch": 1.34, + "learning_rate": 2.8673851459539354e-06, + "loss": 0.5704, + "step": 3639 + }, + { + "epoch": 1.34, + "learning_rate": 2.86734819559059e-06, + "loss": 0.5731, + "step": 3640 + }, + { + "epoch": 1.34, + "learning_rate": 2.8673112452272446e-06, + "loss": 0.6305, + "step": 3641 + }, + { + "epoch": 1.34, + "learning_rate": 2.8672742948638997e-06, + "loss": 0.6329, + "step": 3642 + }, + { + "epoch": 1.35, + "learning_rate": 2.8672373445005543e-06, + "loss": 0.5881, + "step": 3643 + }, + { + "epoch": 1.35, + "learning_rate": 2.8672003941372094e-06, + "loss": 0.5798, + "step": 3644 + }, + { + "epoch": 1.35, + "learning_rate": 2.867163443773864e-06, + "loss": 0.5737, + "step": 3645 + }, + { + "epoch": 1.35, + "learning_rate": 2.8671264934105187e-06, + "loss": 0.6411, + "step": 3646 + }, + { + "epoch": 1.35, + "learning_rate": 2.8670895430471733e-06, + "loss": 0.6532, + "step": 3647 + }, + { + "epoch": 1.35, + "learning_rate": 2.867052592683828e-06, + "loss": 0.5776, + "step": 3648 + }, + { + "epoch": 1.35, + "learning_rate": 2.867015642320483e-06, + "loss": 0.5742, + "step": 3649 + }, + { + "epoch": 1.35, + "learning_rate": 2.8669786919571376e-06, + "loss": 0.557, + "step": 3650 + }, + { + "epoch": 1.35, + "learning_rate": 2.8669417415937927e-06, + "loss": 0.6627, + "step": 3651 + }, + { + "epoch": 1.35, + "learning_rate": 2.8669047912304473e-06, + "loss": 0.6377, + "step": 3652 + }, + { + "epoch": 1.35, + "learning_rate": 2.866867840867102e-06, + "loss": 0.6507, + "step": 3653 + }, + { + "epoch": 1.35, + "learning_rate": 2.8668308905037566e-06, + "loss": 0.6463, + "step": 3654 + }, + { + "epoch": 1.35, + "learning_rate": 2.8667939401404112e-06, + "loss": 0.6338, + "step": 3655 + }, + { + "epoch": 1.35, + "learning_rate": 2.8667569897770663e-06, + "loss": 0.6716, + "step": 3656 + }, + { + "epoch": 1.35, + "learning_rate": 2.866720039413721e-06, + "loss": 0.6606, + "step": 3657 + }, + { + "epoch": 1.35, + "learning_rate": 2.866683089050376e-06, + "loss": 0.6147, + "step": 3658 + }, + { + "epoch": 1.35, + "learning_rate": 2.86664613868703e-06, + "loss": 0.6508, + "step": 3659 + }, + { + "epoch": 1.35, + "learning_rate": 2.8666091883236853e-06, + "loss": 0.5789, + "step": 3660 + }, + { + "epoch": 1.35, + "learning_rate": 2.86657223796034e-06, + "loss": 0.5882, + "step": 3661 + }, + { + "epoch": 1.35, + "learning_rate": 2.866535287596995e-06, + "loss": 0.6699, + "step": 3662 + }, + { + "epoch": 1.35, + "learning_rate": 2.8664983372336496e-06, + "loss": 0.5794, + "step": 3663 + }, + { + "epoch": 1.35, + "learning_rate": 2.8664613868703047e-06, + "loss": 0.6808, + "step": 3664 + }, + { + "epoch": 1.35, + "learning_rate": 2.866424436506959e-06, + "loss": 0.7513, + "step": 3665 + }, + { + "epoch": 1.35, + "learning_rate": 2.8663874861436135e-06, + "loss": 0.6154, + "step": 3666 + }, + { + "epoch": 1.35, + "learning_rate": 2.8663505357802686e-06, + "loss": 0.691, + "step": 3667 + }, + { + "epoch": 1.35, + "learning_rate": 2.866313585416923e-06, + "loss": 0.636, + "step": 3668 + }, + { + "epoch": 1.35, + "learning_rate": 2.8662766350535783e-06, + "loss": 0.6015, + "step": 3669 + }, + { + "epoch": 1.36, + "learning_rate": 2.866239684690233e-06, + "loss": 0.5615, + "step": 3670 + }, + { + "epoch": 1.36, + "learning_rate": 2.8662027343268875e-06, + "loss": 0.6174, + "step": 3671 + }, + { + "epoch": 1.36, + "learning_rate": 2.866165783963542e-06, + "loss": 0.6101, + "step": 3672 + }, + { + "epoch": 1.36, + "learning_rate": 2.8661288336001972e-06, + "loss": 0.645, + "step": 3673 + }, + { + "epoch": 1.36, + "learning_rate": 2.866091883236852e-06, + "loss": 0.6425, + "step": 3674 + }, + { + "epoch": 1.36, + "learning_rate": 2.8660549328735065e-06, + "loss": 0.6105, + "step": 3675 + }, + { + "epoch": 1.36, + "learning_rate": 2.8660179825101616e-06, + "loss": 0.5244, + "step": 3676 + }, + { + "epoch": 1.36, + "learning_rate": 2.865981032146816e-06, + "loss": 0.5996, + "step": 3677 + }, + { + "epoch": 1.36, + "learning_rate": 2.865944081783471e-06, + "loss": 0.641, + "step": 3678 + }, + { + "epoch": 1.36, + "learning_rate": 2.8659071314201255e-06, + "loss": 0.5792, + "step": 3679 + }, + { + "epoch": 1.36, + "learning_rate": 2.8658701810567805e-06, + "loss": 0.6894, + "step": 3680 + }, + { + "epoch": 1.36, + "learning_rate": 2.865833230693435e-06, + "loss": 0.6223, + "step": 3681 + }, + { + "epoch": 1.36, + "learning_rate": 2.8657962803300902e-06, + "loss": 0.5966, + "step": 3682 + }, + { + "epoch": 1.36, + "learning_rate": 2.865759329966745e-06, + "loss": 0.6466, + "step": 3683 + }, + { + "epoch": 1.36, + "learning_rate": 2.8657223796033995e-06, + "loss": 0.6273, + "step": 3684 + }, + { + "epoch": 1.36, + "learning_rate": 2.865685429240054e-06, + "loss": 0.663, + "step": 3685 + }, + { + "epoch": 1.36, + "learning_rate": 2.8656484788767088e-06, + "loss": 0.6311, + "step": 3686 + }, + { + "epoch": 1.36, + "learning_rate": 2.865611528513364e-06, + "loss": 0.5892, + "step": 3687 + }, + { + "epoch": 1.36, + "learning_rate": 2.8655745781500185e-06, + "loss": 0.5676, + "step": 3688 + }, + { + "epoch": 1.36, + "learning_rate": 2.8655376277866735e-06, + "loss": 0.6374, + "step": 3689 + }, + { + "epoch": 1.36, + "learning_rate": 2.865500677423328e-06, + "loss": 0.5572, + "step": 3690 + }, + { + "epoch": 1.36, + "learning_rate": 2.865463727059983e-06, + "loss": 0.6045, + "step": 3691 + }, + { + "epoch": 1.36, + "learning_rate": 2.8654267766966374e-06, + "loss": 0.6234, + "step": 3692 + }, + { + "epoch": 1.36, + "learning_rate": 2.8653898263332925e-06, + "loss": 0.6085, + "step": 3693 + }, + { + "epoch": 1.36, + "learning_rate": 2.865352875969947e-06, + "loss": 0.6119, + "step": 3694 + }, + { + "epoch": 1.36, + "learning_rate": 2.8653159256066018e-06, + "loss": 0.5859, + "step": 3695 + }, + { + "epoch": 1.36, + "learning_rate": 2.865278975243257e-06, + "loss": 0.6121, + "step": 3696 + }, + { + "epoch": 1.37, + "learning_rate": 2.8652420248799115e-06, + "loss": 0.5945, + "step": 3697 + }, + { + "epoch": 1.37, + "learning_rate": 2.865205074516566e-06, + "loss": 0.6196, + "step": 3698 + }, + { + "epoch": 1.37, + "learning_rate": 2.8651681241532207e-06, + "loss": 0.5294, + "step": 3699 + }, + { + "epoch": 1.37, + "learning_rate": 2.865131173789876e-06, + "loss": 0.6545, + "step": 3700 + }, + { + "epoch": 1.37, + "learning_rate": 2.8650942234265304e-06, + "loss": 0.6681, + "step": 3701 + }, + { + "epoch": 1.37, + "learning_rate": 2.8650572730631855e-06, + "loss": 0.6096, + "step": 3702 + }, + { + "epoch": 1.37, + "learning_rate": 2.86502032269984e-06, + "loss": 0.6379, + "step": 3703 + }, + { + "epoch": 1.37, + "learning_rate": 2.8649833723364948e-06, + "loss": 0.6022, + "step": 3704 + }, + { + "epoch": 1.37, + "learning_rate": 2.8649464219731494e-06, + "loss": 0.6318, + "step": 3705 + }, + { + "epoch": 1.37, + "learning_rate": 2.864909471609804e-06, + "loss": 0.5942, + "step": 3706 + }, + { + "epoch": 1.37, + "learning_rate": 2.864872521246459e-06, + "loss": 0.5988, + "step": 3707 + }, + { + "epoch": 1.37, + "learning_rate": 2.8648355708831137e-06, + "loss": 0.6733, + "step": 3708 + }, + { + "epoch": 1.37, + "learning_rate": 2.864798620519769e-06, + "loss": 0.6738, + "step": 3709 + }, + { + "epoch": 1.37, + "learning_rate": 2.864761670156423e-06, + "loss": 0.6356, + "step": 3710 + }, + { + "epoch": 1.37, + "learning_rate": 2.864724719793078e-06, + "loss": 0.5969, + "step": 3711 + }, + { + "epoch": 1.37, + "learning_rate": 2.8646877694297327e-06, + "loss": 0.6729, + "step": 3712 + }, + { + "epoch": 1.37, + "learning_rate": 2.8646508190663878e-06, + "loss": 0.6417, + "step": 3713 + }, + { + "epoch": 1.37, + "learning_rate": 2.8646138687030424e-06, + "loss": 0.6708, + "step": 3714 + }, + { + "epoch": 1.37, + "learning_rate": 2.864576918339697e-06, + "loss": 0.6072, + "step": 3715 + }, + { + "epoch": 1.37, + "learning_rate": 2.8645399679763517e-06, + "loss": 0.6176, + "step": 3716 + }, + { + "epoch": 1.37, + "learning_rate": 2.8645030176130063e-06, + "loss": 0.6261, + "step": 3717 + }, + { + "epoch": 1.37, + "learning_rate": 2.8644660672496614e-06, + "loss": 0.6494, + "step": 3718 + }, + { + "epoch": 1.37, + "learning_rate": 2.864429116886316e-06, + "loss": 0.6476, + "step": 3719 + }, + { + "epoch": 1.37, + "learning_rate": 2.864392166522971e-06, + "loss": 0.5914, + "step": 3720 + }, + { + "epoch": 1.37, + "learning_rate": 2.8643552161596257e-06, + "loss": 0.6507, + "step": 3721 + }, + { + "epoch": 1.37, + "learning_rate": 2.8643182657962803e-06, + "loss": 0.7605, + "step": 3722 + }, + { + "epoch": 1.37, + "learning_rate": 2.864281315432935e-06, + "loss": 0.6258, + "step": 3723 + }, + { + "epoch": 1.38, + "learning_rate": 2.86424436506959e-06, + "loss": 0.6196, + "step": 3724 + }, + { + "epoch": 1.38, + "learning_rate": 2.8642074147062447e-06, + "loss": 0.5822, + "step": 3725 + }, + { + "epoch": 1.38, + "learning_rate": 2.8641704643428993e-06, + "loss": 0.6912, + "step": 3726 + }, + { + "epoch": 1.38, + "learning_rate": 2.8641335139795544e-06, + "loss": 0.6124, + "step": 3727 + }, + { + "epoch": 1.38, + "learning_rate": 2.864096563616209e-06, + "loss": 0.6419, + "step": 3728 + }, + { + "epoch": 1.38, + "learning_rate": 2.8640596132528636e-06, + "loss": 0.6293, + "step": 3729 + }, + { + "epoch": 1.38, + "learning_rate": 2.8640226628895183e-06, + "loss": 0.5541, + "step": 3730 + }, + { + "epoch": 1.38, + "learning_rate": 2.8639857125261733e-06, + "loss": 0.5368, + "step": 3731 + }, + { + "epoch": 1.38, + "learning_rate": 2.863948762162828e-06, + "loss": 0.6625, + "step": 3732 + }, + { + "epoch": 1.38, + "learning_rate": 2.863911811799483e-06, + "loss": 0.6697, + "step": 3733 + }, + { + "epoch": 1.38, + "learning_rate": 2.8638748614361377e-06, + "loss": 0.6176, + "step": 3734 + }, + { + "epoch": 1.38, + "learning_rate": 2.8638379110727923e-06, + "loss": 0.5983, + "step": 3735 + }, + { + "epoch": 1.38, + "learning_rate": 2.863800960709447e-06, + "loss": 0.6314, + "step": 3736 + }, + { + "epoch": 1.38, + "learning_rate": 2.8637640103461016e-06, + "loss": 0.6265, + "step": 3737 + }, + { + "epoch": 1.38, + "learning_rate": 2.8637270599827566e-06, + "loss": 0.6251, + "step": 3738 + }, + { + "epoch": 1.38, + "learning_rate": 2.8636901096194113e-06, + "loss": 0.6873, + "step": 3739 + }, + { + "epoch": 1.38, + "learning_rate": 2.8636531592560663e-06, + "loss": 0.5747, + "step": 3740 + }, + { + "epoch": 1.38, + "learning_rate": 2.863616208892721e-06, + "loss": 0.6313, + "step": 3741 + }, + { + "epoch": 1.38, + "learning_rate": 2.8635792585293756e-06, + "loss": 0.5392, + "step": 3742 + }, + { + "epoch": 1.38, + "learning_rate": 2.8635423081660302e-06, + "loss": 0.5875, + "step": 3743 + }, + { + "epoch": 1.38, + "learning_rate": 2.8635053578026853e-06, + "loss": 0.642, + "step": 3744 + }, + { + "epoch": 1.38, + "learning_rate": 2.86346840743934e-06, + "loss": 0.6419, + "step": 3745 + }, + { + "epoch": 1.38, + "learning_rate": 2.8634314570759946e-06, + "loss": 0.6542, + "step": 3746 + }, + { + "epoch": 1.38, + "learning_rate": 2.8633945067126496e-06, + "loss": 0.6792, + "step": 3747 + }, + { + "epoch": 1.38, + "learning_rate": 2.8633575563493043e-06, + "loss": 0.6368, + "step": 3748 + }, + { + "epoch": 1.38, + "learning_rate": 2.863320605985959e-06, + "loss": 0.6492, + "step": 3749 + }, + { + "epoch": 1.38, + "learning_rate": 2.8632836556226136e-06, + "loss": 0.6643, + "step": 3750 + }, + { + "epoch": 1.39, + "learning_rate": 2.8632467052592686e-06, + "loss": 0.6926, + "step": 3751 + }, + { + "epoch": 1.39, + "learning_rate": 2.8632097548959232e-06, + "loss": 0.5811, + "step": 3752 + }, + { + "epoch": 1.39, + "learning_rate": 2.8631728045325783e-06, + "loss": 0.6406, + "step": 3753 + }, + { + "epoch": 1.39, + "learning_rate": 2.863135854169233e-06, + "loss": 0.6144, + "step": 3754 + }, + { + "epoch": 1.39, + "learning_rate": 2.863098903805887e-06, + "loss": 0.6281, + "step": 3755 + }, + { + "epoch": 1.39, + "learning_rate": 2.8630619534425422e-06, + "loss": 0.5666, + "step": 3756 + }, + { + "epoch": 1.39, + "learning_rate": 2.863025003079197e-06, + "loss": 0.634, + "step": 3757 + }, + { + "epoch": 1.39, + "learning_rate": 2.862988052715852e-06, + "loss": 0.6252, + "step": 3758 + }, + { + "epoch": 1.39, + "learning_rate": 2.8629511023525065e-06, + "loss": 0.6601, + "step": 3759 + }, + { + "epoch": 1.39, + "learning_rate": 2.8629141519891616e-06, + "loss": 0.5615, + "step": 3760 + }, + { + "epoch": 1.39, + "learning_rate": 2.862877201625816e-06, + "loss": 0.6154, + "step": 3761 + }, + { + "epoch": 1.39, + "learning_rate": 2.862840251262471e-06, + "loss": 0.6077, + "step": 3762 + }, + { + "epoch": 1.39, + "learning_rate": 2.8628033008991255e-06, + "loss": 0.6168, + "step": 3763 + }, + { + "epoch": 1.39, + "learning_rate": 2.86276635053578e-06, + "loss": 0.6062, + "step": 3764 + }, + { + "epoch": 1.39, + "learning_rate": 2.8627294001724352e-06, + "loss": 0.5392, + "step": 3765 + }, + { + "epoch": 1.39, + "learning_rate": 2.86269244980909e-06, + "loss": 0.5484, + "step": 3766 + }, + { + "epoch": 1.39, + "learning_rate": 2.8626554994457445e-06, + "loss": 0.5826, + "step": 3767 + }, + { + "epoch": 1.39, + "learning_rate": 2.862618549082399e-06, + "loss": 0.6796, + "step": 3768 + }, + { + "epoch": 1.39, + "learning_rate": 2.862581598719054e-06, + "loss": 0.6033, + "step": 3769 + }, + { + "epoch": 1.39, + "learning_rate": 2.862544648355709e-06, + "loss": 0.6033, + "step": 3770 + }, + { + "epoch": 1.39, + "learning_rate": 2.862507697992364e-06, + "loss": 0.5647, + "step": 3771 + }, + { + "epoch": 1.39, + "learning_rate": 2.8624707476290185e-06, + "loss": 0.6179, + "step": 3772 + }, + { + "epoch": 1.39, + "learning_rate": 2.862433797265673e-06, + "loss": 0.5503, + "step": 3773 + }, + { + "epoch": 1.39, + "learning_rate": 2.862396846902328e-06, + "loss": 0.5811, + "step": 3774 + }, + { + "epoch": 1.39, + "learning_rate": 2.8623598965389824e-06, + "loss": 0.5707, + "step": 3775 + }, + { + "epoch": 1.39, + "learning_rate": 2.8623229461756375e-06, + "loss": 0.6003, + "step": 3776 + }, + { + "epoch": 1.39, + "learning_rate": 2.862285995812292e-06, + "loss": 0.6327, + "step": 3777 + }, + { + "epoch": 1.4, + "learning_rate": 2.862249045448947e-06, + "loss": 0.6405, + "step": 3778 + }, + { + "epoch": 1.4, + "learning_rate": 2.862212095085602e-06, + "loss": 0.5763, + "step": 3779 + }, + { + "epoch": 1.4, + "learning_rate": 2.8621751447222565e-06, + "loss": 0.6278, + "step": 3780 + }, + { + "epoch": 1.4, + "learning_rate": 2.862138194358911e-06, + "loss": 0.609, + "step": 3781 + }, + { + "epoch": 1.4, + "learning_rate": 2.862101243995566e-06, + "loss": 0.6177, + "step": 3782 + }, + { + "epoch": 1.4, + "learning_rate": 2.862064293632221e-06, + "loss": 0.6135, + "step": 3783 + }, + { + "epoch": 1.4, + "learning_rate": 2.8620273432688754e-06, + "loss": 0.714, + "step": 3784 + }, + { + "epoch": 1.4, + "learning_rate": 2.8619903929055305e-06, + "loss": 0.652, + "step": 3785 + }, + { + "epoch": 1.4, + "learning_rate": 2.861953442542185e-06, + "loss": 0.5344, + "step": 3786 + }, + { + "epoch": 1.4, + "learning_rate": 2.8619164921788398e-06, + "loss": 0.5623, + "step": 3787 + }, + { + "epoch": 1.4, + "learning_rate": 2.8618795418154944e-06, + "loss": 0.5827, + "step": 3788 + }, + { + "epoch": 1.4, + "learning_rate": 2.8618425914521495e-06, + "loss": 0.6388, + "step": 3789 + }, + { + "epoch": 1.4, + "learning_rate": 2.861805641088804e-06, + "loss": 0.6551, + "step": 3790 + }, + { + "epoch": 1.4, + "learning_rate": 2.861768690725459e-06, + "loss": 0.6982, + "step": 3791 + }, + { + "epoch": 1.4, + "learning_rate": 2.8617317403621138e-06, + "loss": 0.5888, + "step": 3792 + }, + { + "epoch": 1.4, + "learning_rate": 2.8616947899987684e-06, + "loss": 0.6677, + "step": 3793 + }, + { + "epoch": 1.4, + "learning_rate": 2.861657839635423e-06, + "loss": 0.6082, + "step": 3794 + }, + { + "epoch": 1.4, + "learning_rate": 2.8616208892720777e-06, + "loss": 0.6343, + "step": 3795 + }, + { + "epoch": 1.4, + "learning_rate": 2.8615839389087328e-06, + "loss": 0.5193, + "step": 3796 + }, + { + "epoch": 1.4, + "learning_rate": 2.8615469885453874e-06, + "loss": 0.5951, + "step": 3797 + }, + { + "epoch": 1.4, + "learning_rate": 2.8615100381820425e-06, + "loss": 0.6624, + "step": 3798 + }, + { + "epoch": 1.4, + "learning_rate": 2.8614730878186967e-06, + "loss": 0.7311, + "step": 3799 + }, + { + "epoch": 1.4, + "learning_rate": 2.8614361374553517e-06, + "loss": 0.6372, + "step": 3800 + }, + { + "epoch": 1.4, + "learning_rate": 2.8613991870920064e-06, + "loss": 0.5355, + "step": 3801 + }, + { + "epoch": 1.4, + "learning_rate": 2.8613622367286614e-06, + "loss": 0.6626, + "step": 3802 + }, + { + "epoch": 1.4, + "learning_rate": 2.861325286365316e-06, + "loss": 0.6121, + "step": 3803 + }, + { + "epoch": 1.4, + "learning_rate": 2.8612883360019707e-06, + "loss": 0.631, + "step": 3804 + }, + { + "epoch": 1.41, + "learning_rate": 2.8612513856386253e-06, + "loss": 0.6215, + "step": 3805 + }, + { + "epoch": 1.41, + "learning_rate": 2.86121443527528e-06, + "loss": 0.6091, + "step": 3806 + }, + { + "epoch": 1.41, + "learning_rate": 2.861177484911935e-06, + "loss": 0.5705, + "step": 3807 + }, + { + "epoch": 1.41, + "learning_rate": 2.8611405345485897e-06, + "loss": 0.6177, + "step": 3808 + }, + { + "epoch": 1.41, + "learning_rate": 2.8611035841852447e-06, + "loss": 0.6829, + "step": 3809 + }, + { + "epoch": 1.41, + "learning_rate": 2.8610666338218994e-06, + "loss": 0.6219, + "step": 3810 + }, + { + "epoch": 1.41, + "learning_rate": 2.861029683458554e-06, + "loss": 0.5898, + "step": 3811 + }, + { + "epoch": 1.41, + "learning_rate": 2.8609927330952086e-06, + "loss": 0.6099, + "step": 3812 + }, + { + "epoch": 1.41, + "learning_rate": 2.8609557827318637e-06, + "loss": 0.6035, + "step": 3813 + }, + { + "epoch": 1.41, + "learning_rate": 2.8609188323685183e-06, + "loss": 0.5672, + "step": 3814 + }, + { + "epoch": 1.41, + "learning_rate": 2.860881882005173e-06, + "loss": 0.6537, + "step": 3815 + }, + { + "epoch": 1.41, + "learning_rate": 2.860844931641828e-06, + "loss": 0.6023, + "step": 3816 + }, + { + "epoch": 1.41, + "learning_rate": 2.8608079812784827e-06, + "loss": 0.6294, + "step": 3817 + }, + { + "epoch": 1.41, + "learning_rate": 2.8607710309151373e-06, + "loss": 0.5664, + "step": 3818 + }, + { + "epoch": 1.41, + "learning_rate": 2.860734080551792e-06, + "loss": 0.5453, + "step": 3819 + }, + { + "epoch": 1.41, + "learning_rate": 2.860697130188447e-06, + "loss": 0.6447, + "step": 3820 + }, + { + "epoch": 1.41, + "learning_rate": 2.8606601798251016e-06, + "loss": 0.6186, + "step": 3821 + }, + { + "epoch": 1.41, + "learning_rate": 2.8606232294617567e-06, + "loss": 0.6233, + "step": 3822 + }, + { + "epoch": 1.41, + "learning_rate": 2.8605862790984113e-06, + "loss": 0.587, + "step": 3823 + }, + { + "epoch": 1.41, + "learning_rate": 2.860549328735066e-06, + "loss": 0.6023, + "step": 3824 + }, + { + "epoch": 1.41, + "learning_rate": 2.8605123783717206e-06, + "loss": 0.5758, + "step": 3825 + }, + { + "epoch": 1.41, + "learning_rate": 2.8604754280083752e-06, + "loss": 0.669, + "step": 3826 + }, + { + "epoch": 1.41, + "learning_rate": 2.8604384776450303e-06, + "loss": 0.725, + "step": 3827 + }, + { + "epoch": 1.41, + "learning_rate": 2.860401527281685e-06, + "loss": 0.6885, + "step": 3828 + }, + { + "epoch": 1.41, + "learning_rate": 2.86036457691834e-06, + "loss": 0.6056, + "step": 3829 + }, + { + "epoch": 1.41, + "learning_rate": 2.8603276265549946e-06, + "loss": 0.5587, + "step": 3830 + }, + { + "epoch": 1.41, + "learning_rate": 2.8602906761916493e-06, + "loss": 0.6112, + "step": 3831 + }, + { + "epoch": 1.42, + "learning_rate": 2.860253725828304e-06, + "loss": 0.6104, + "step": 3832 + }, + { + "epoch": 1.42, + "learning_rate": 2.860216775464959e-06, + "loss": 0.5507, + "step": 3833 + }, + { + "epoch": 1.42, + "learning_rate": 2.8601798251016136e-06, + "loss": 0.5734, + "step": 3834 + }, + { + "epoch": 1.42, + "learning_rate": 2.8601428747382682e-06, + "loss": 0.6398, + "step": 3835 + }, + { + "epoch": 1.42, + "learning_rate": 2.8601059243749233e-06, + "loss": 0.6573, + "step": 3836 + }, + { + "epoch": 1.42, + "learning_rate": 2.860068974011578e-06, + "loss": 0.8146, + "step": 3837 + }, + { + "epoch": 1.42, + "learning_rate": 2.8600320236482326e-06, + "loss": 0.6003, + "step": 3838 + }, + { + "epoch": 1.42, + "learning_rate": 2.859995073284887e-06, + "loss": 0.6376, + "step": 3839 + }, + { + "epoch": 1.42, + "learning_rate": 2.8599581229215423e-06, + "loss": 0.678, + "step": 3840 + }, + { + "epoch": 1.42, + "learning_rate": 2.859921172558197e-06, + "loss": 0.5996, + "step": 3841 + }, + { + "epoch": 1.42, + "learning_rate": 2.859884222194852e-06, + "loss": 0.6724, + "step": 3842 + }, + { + "epoch": 1.42, + "learning_rate": 2.8598472718315066e-06, + "loss": 0.6507, + "step": 3843 + }, + { + "epoch": 1.42, + "learning_rate": 2.859810321468161e-06, + "loss": 0.6926, + "step": 3844 + }, + { + "epoch": 1.42, + "learning_rate": 2.859773371104816e-06, + "loss": 0.603, + "step": 3845 + }, + { + "epoch": 1.42, + "learning_rate": 2.8597364207414705e-06, + "loss": 0.633, + "step": 3846 + }, + { + "epoch": 1.42, + "learning_rate": 2.8596994703781256e-06, + "loss": 0.6086, + "step": 3847 + }, + { + "epoch": 1.42, + "learning_rate": 2.85966252001478e-06, + "loss": 0.6142, + "step": 3848 + }, + { + "epoch": 1.42, + "learning_rate": 2.8596255696514353e-06, + "loss": 0.646, + "step": 3849 + }, + { + "epoch": 1.42, + "learning_rate": 2.8595886192880895e-06, + "loss": 0.5736, + "step": 3850 + }, + { + "epoch": 1.42, + "learning_rate": 2.8595516689247445e-06, + "loss": 0.6382, + "step": 3851 + }, + { + "epoch": 1.42, + "learning_rate": 2.859514718561399e-06, + "loss": 0.6166, + "step": 3852 + }, + { + "epoch": 1.42, + "learning_rate": 2.8594777681980542e-06, + "loss": 0.6761, + "step": 3853 + }, + { + "epoch": 1.42, + "learning_rate": 2.859440817834709e-06, + "loss": 0.556, + "step": 3854 + }, + { + "epoch": 1.42, + "learning_rate": 2.8594038674713635e-06, + "loss": 0.6142, + "step": 3855 + }, + { + "epoch": 1.42, + "learning_rate": 2.859366917108018e-06, + "loss": 0.5762, + "step": 3856 + }, + { + "epoch": 1.42, + "learning_rate": 2.8593299667446728e-06, + "loss": 0.5986, + "step": 3857 + }, + { + "epoch": 1.42, + "learning_rate": 2.859293016381328e-06, + "loss": 0.631, + "step": 3858 + }, + { + "epoch": 1.43, + "learning_rate": 2.8592560660179825e-06, + "loss": 0.6362, + "step": 3859 + }, + { + "epoch": 1.43, + "learning_rate": 2.8592191156546375e-06, + "loss": 0.6177, + "step": 3860 + }, + { + "epoch": 1.43, + "learning_rate": 2.859182165291292e-06, + "loss": 0.6403, + "step": 3861 + }, + { + "epoch": 1.43, + "learning_rate": 2.859145214927947e-06, + "loss": 0.5793, + "step": 3862 + }, + { + "epoch": 1.43, + "learning_rate": 2.8591082645646014e-06, + "loss": 0.6778, + "step": 3863 + }, + { + "epoch": 1.43, + "learning_rate": 2.859071314201256e-06, + "loss": 0.6173, + "step": 3864 + }, + { + "epoch": 1.43, + "learning_rate": 2.859034363837911e-06, + "loss": 0.5822, + "step": 3865 + }, + { + "epoch": 1.43, + "learning_rate": 2.8589974134745658e-06, + "loss": 0.5848, + "step": 3866 + }, + { + "epoch": 1.43, + "learning_rate": 2.858960463111221e-06, + "loss": 0.6742, + "step": 3867 + }, + { + "epoch": 1.43, + "learning_rate": 2.8589235127478755e-06, + "loss": 0.672, + "step": 3868 + }, + { + "epoch": 1.43, + "learning_rate": 2.85888656238453e-06, + "loss": 0.5786, + "step": 3869 + }, + { + "epoch": 1.43, + "learning_rate": 2.8588496120211847e-06, + "loss": 0.6223, + "step": 3870 + }, + { + "epoch": 1.43, + "learning_rate": 2.85881266165784e-06, + "loss": 0.6097, + "step": 3871 + }, + { + "epoch": 1.43, + "learning_rate": 2.8587757112944944e-06, + "loss": 0.5925, + "step": 3872 + }, + { + "epoch": 1.43, + "learning_rate": 2.8587387609311495e-06, + "loss": 0.6005, + "step": 3873 + }, + { + "epoch": 1.43, + "learning_rate": 2.858701810567804e-06, + "loss": 0.6559, + "step": 3874 + }, + { + "epoch": 1.43, + "learning_rate": 2.8586648602044588e-06, + "loss": 0.6544, + "step": 3875 + }, + { + "epoch": 1.43, + "learning_rate": 2.8586279098411134e-06, + "loss": 0.5788, + "step": 3876 + }, + { + "epoch": 1.43, + "learning_rate": 2.858590959477768e-06, + "loss": 0.5739, + "step": 3877 + }, + { + "epoch": 1.43, + "learning_rate": 2.858554009114423e-06, + "loss": 0.5549, + "step": 3878 + }, + { + "epoch": 1.43, + "learning_rate": 2.8585170587510777e-06, + "loss": 0.6456, + "step": 3879 + }, + { + "epoch": 1.43, + "learning_rate": 2.858480108387733e-06, + "loss": 0.6663, + "step": 3880 + }, + { + "epoch": 1.43, + "learning_rate": 2.8584431580243874e-06, + "loss": 0.641, + "step": 3881 + }, + { + "epoch": 1.43, + "learning_rate": 2.858406207661042e-06, + "loss": 0.6046, + "step": 3882 + }, + { + "epoch": 1.43, + "learning_rate": 2.8583692572976967e-06, + "loss": 0.6088, + "step": 3883 + }, + { + "epoch": 1.43, + "learning_rate": 2.8583323069343514e-06, + "loss": 0.6468, + "step": 3884 + }, + { + "epoch": 1.43, + "learning_rate": 2.8582953565710064e-06, + "loss": 0.6377, + "step": 3885 + }, + { + "epoch": 1.44, + "learning_rate": 2.858258406207661e-06, + "loss": 0.5596, + "step": 3886 + }, + { + "epoch": 1.44, + "learning_rate": 2.858221455844316e-06, + "loss": 0.5799, + "step": 3887 + }, + { + "epoch": 1.44, + "learning_rate": 2.8581845054809707e-06, + "loss": 0.6338, + "step": 3888 + }, + { + "epoch": 1.44, + "learning_rate": 2.8581475551176254e-06, + "loss": 0.5772, + "step": 3889 + }, + { + "epoch": 1.44, + "learning_rate": 2.85811060475428e-06, + "loss": 0.6399, + "step": 3890 + }, + { + "epoch": 1.44, + "learning_rate": 2.858073654390935e-06, + "loss": 0.6191, + "step": 3891 + }, + { + "epoch": 1.44, + "learning_rate": 2.8580367040275897e-06, + "loss": 0.6445, + "step": 3892 + }, + { + "epoch": 1.44, + "learning_rate": 2.8579997536642448e-06, + "loss": 0.6648, + "step": 3893 + }, + { + "epoch": 1.44, + "learning_rate": 2.8579628033008994e-06, + "loss": 0.5634, + "step": 3894 + }, + { + "epoch": 1.44, + "learning_rate": 2.8579258529375536e-06, + "loss": 0.6334, + "step": 3895 + }, + { + "epoch": 1.44, + "learning_rate": 2.8578889025742087e-06, + "loss": 0.6049, + "step": 3896 + }, + { + "epoch": 1.44, + "learning_rate": 2.8578519522108633e-06, + "loss": 0.6264, + "step": 3897 + }, + { + "epoch": 1.44, + "learning_rate": 2.8578150018475184e-06, + "loss": 0.5904, + "step": 3898 + }, + { + "epoch": 1.44, + "learning_rate": 2.857778051484173e-06, + "loss": 0.5453, + "step": 3899 + }, + { + "epoch": 1.44, + "learning_rate": 2.857741101120828e-06, + "loss": 0.7174, + "step": 3900 + }, + { + "epoch": 1.44, + "learning_rate": 2.8577041507574823e-06, + "loss": 0.5828, + "step": 3901 + }, + { + "epoch": 1.44, + "learning_rate": 2.8576672003941373e-06, + "loss": 0.5951, + "step": 3902 + }, + { + "epoch": 1.44, + "learning_rate": 2.857630250030792e-06, + "loss": 0.6364, + "step": 3903 + }, + { + "epoch": 1.44, + "learning_rate": 2.8575932996674466e-06, + "loss": 0.5729, + "step": 3904 + }, + { + "epoch": 1.44, + "learning_rate": 2.8575563493041017e-06, + "loss": 0.6281, + "step": 3905 + }, + { + "epoch": 1.44, + "learning_rate": 2.8575193989407563e-06, + "loss": 0.6136, + "step": 3906 + }, + { + "epoch": 1.44, + "learning_rate": 2.857482448577411e-06, + "loss": 0.5378, + "step": 3907 + }, + { + "epoch": 1.44, + "learning_rate": 2.8574454982140656e-06, + "loss": 0.559, + "step": 3908 + }, + { + "epoch": 1.44, + "learning_rate": 2.8574085478507206e-06, + "loss": 0.6194, + "step": 3909 + }, + { + "epoch": 1.44, + "learning_rate": 2.8573715974873753e-06, + "loss": 0.601, + "step": 3910 + }, + { + "epoch": 1.44, + "learning_rate": 2.8573346471240303e-06, + "loss": 0.6207, + "step": 3911 + }, + { + "epoch": 1.44, + "learning_rate": 2.857297696760685e-06, + "loss": 0.6128, + "step": 3912 + }, + { + "epoch": 1.44, + "learning_rate": 2.8572607463973396e-06, + "loss": 0.569, + "step": 3913 + }, + { + "epoch": 1.45, + "learning_rate": 2.8572237960339943e-06, + "loss": 0.5925, + "step": 3914 + }, + { + "epoch": 1.45, + "learning_rate": 2.857186845670649e-06, + "loss": 0.5553, + "step": 3915 + }, + { + "epoch": 1.45, + "learning_rate": 2.857149895307304e-06, + "loss": 0.6303, + "step": 3916 + }, + { + "epoch": 1.45, + "learning_rate": 2.8571129449439586e-06, + "loss": 0.6451, + "step": 3917 + }, + { + "epoch": 1.45, + "learning_rate": 2.8570759945806136e-06, + "loss": 0.5955, + "step": 3918 + }, + { + "epoch": 1.45, + "learning_rate": 2.8570390442172683e-06, + "loss": 0.5328, + "step": 3919 + }, + { + "epoch": 1.45, + "learning_rate": 2.857002093853923e-06, + "loss": 0.5851, + "step": 3920 + }, + { + "epoch": 1.45, + "learning_rate": 2.8569651434905776e-06, + "loss": 0.6468, + "step": 3921 + }, + { + "epoch": 1.45, + "learning_rate": 2.8569281931272326e-06, + "loss": 0.6437, + "step": 3922 + }, + { + "epoch": 1.45, + "learning_rate": 2.8568912427638873e-06, + "loss": 0.7167, + "step": 3923 + }, + { + "epoch": 1.45, + "learning_rate": 2.856854292400542e-06, + "loss": 0.6255, + "step": 3924 + }, + { + "epoch": 1.45, + "learning_rate": 2.856817342037197e-06, + "loss": 0.629, + "step": 3925 + }, + { + "epoch": 1.45, + "learning_rate": 2.8567803916738516e-06, + "loss": 0.5417, + "step": 3926 + }, + { + "epoch": 1.45, + "learning_rate": 2.8567434413105062e-06, + "loss": 0.6143, + "step": 3927 + }, + { + "epoch": 1.45, + "learning_rate": 2.856706490947161e-06, + "loss": 0.594, + "step": 3928 + }, + { + "epoch": 1.45, + "learning_rate": 2.856669540583816e-06, + "loss": 0.6218, + "step": 3929 + }, + { + "epoch": 1.45, + "learning_rate": 2.8566325902204706e-06, + "loss": 0.5675, + "step": 3930 + }, + { + "epoch": 1.45, + "learning_rate": 2.8565956398571256e-06, + "loss": 0.6284, + "step": 3931 + }, + { + "epoch": 1.45, + "learning_rate": 2.8565586894937803e-06, + "loss": 0.645, + "step": 3932 + }, + { + "epoch": 1.45, + "learning_rate": 2.856521739130435e-06, + "loss": 0.6229, + "step": 3933 + }, + { + "epoch": 1.45, + "learning_rate": 2.8564847887670895e-06, + "loss": 0.561, + "step": 3934 + }, + { + "epoch": 1.45, + "learning_rate": 2.856447838403744e-06, + "loss": 0.5714, + "step": 3935 + }, + { + "epoch": 1.45, + "learning_rate": 2.8564108880403992e-06, + "loss": 0.5892, + "step": 3936 + }, + { + "epoch": 1.45, + "learning_rate": 2.856373937677054e-06, + "loss": 0.547, + "step": 3937 + }, + { + "epoch": 1.45, + "learning_rate": 2.856336987313709e-06, + "loss": 0.543, + "step": 3938 + }, + { + "epoch": 1.45, + "learning_rate": 2.8563000369503636e-06, + "loss": 0.6437, + "step": 3939 + }, + { + "epoch": 1.45, + "learning_rate": 2.856263086587018e-06, + "loss": 0.6677, + "step": 3940 + }, + { + "epoch": 1.46, + "learning_rate": 2.856226136223673e-06, + "loss": 0.6827, + "step": 3941 + }, + { + "epoch": 1.46, + "learning_rate": 2.856189185860328e-06, + "loss": 0.5706, + "step": 3942 + }, + { + "epoch": 1.46, + "learning_rate": 2.8561522354969825e-06, + "loss": 0.5991, + "step": 3943 + }, + { + "epoch": 1.46, + "learning_rate": 2.856115285133637e-06, + "loss": 0.6557, + "step": 3944 + }, + { + "epoch": 1.46, + "learning_rate": 2.8560783347702922e-06, + "loss": 0.5456, + "step": 3945 + }, + { + "epoch": 1.46, + "learning_rate": 2.8560413844069464e-06, + "loss": 0.5743, + "step": 3946 + }, + { + "epoch": 1.46, + "learning_rate": 2.8560044340436015e-06, + "loss": 0.62, + "step": 3947 + }, + { + "epoch": 1.46, + "learning_rate": 2.855967483680256e-06, + "loss": 0.6274, + "step": 3948 + }, + { + "epoch": 1.46, + "learning_rate": 2.855930533316911e-06, + "loss": 0.6148, + "step": 3949 + }, + { + "epoch": 1.46, + "learning_rate": 2.855893582953566e-06, + "loss": 0.574, + "step": 3950 + }, + { + "epoch": 1.46, + "learning_rate": 2.8558566325902205e-06, + "loss": 0.6812, + "step": 3951 + }, + { + "epoch": 1.46, + "learning_rate": 2.855819682226875e-06, + "loss": 0.5522, + "step": 3952 + }, + { + "epoch": 1.46, + "learning_rate": 2.85578273186353e-06, + "loss": 0.6688, + "step": 3953 + }, + { + "epoch": 1.46, + "learning_rate": 2.855745781500185e-06, + "loss": 0.6417, + "step": 3954 + }, + { + "epoch": 1.46, + "learning_rate": 2.8557088311368394e-06, + "loss": 0.5859, + "step": 3955 + }, + { + "epoch": 1.46, + "learning_rate": 2.8556718807734945e-06, + "loss": 0.5959, + "step": 3956 + }, + { + "epoch": 1.46, + "learning_rate": 2.855634930410149e-06, + "loss": 0.5962, + "step": 3957 + }, + { + "epoch": 1.46, + "learning_rate": 2.8555979800468038e-06, + "loss": 0.6166, + "step": 3958 + }, + { + "epoch": 1.46, + "learning_rate": 2.8555610296834584e-06, + "loss": 0.6094, + "step": 3959 + }, + { + "epoch": 1.46, + "learning_rate": 2.8555240793201135e-06, + "loss": 0.6791, + "step": 3960 + }, + { + "epoch": 1.46, + "learning_rate": 2.855487128956768e-06, + "loss": 0.5859, + "step": 3961 + }, + { + "epoch": 1.46, + "learning_rate": 2.855450178593423e-06, + "loss": 0.6242, + "step": 3962 + }, + { + "epoch": 1.46, + "learning_rate": 2.855413228230078e-06, + "loss": 0.6574, + "step": 3963 + }, + { + "epoch": 1.46, + "learning_rate": 2.8553762778667324e-06, + "loss": 0.6112, + "step": 3964 + }, + { + "epoch": 1.46, + "learning_rate": 2.855339327503387e-06, + "loss": 0.6356, + "step": 3965 + }, + { + "epoch": 1.46, + "learning_rate": 2.8553023771400417e-06, + "loss": 0.5943, + "step": 3966 + }, + { + "epoch": 1.46, + "learning_rate": 2.8552654267766968e-06, + "loss": 0.5881, + "step": 3967 + }, + { + "epoch": 1.47, + "learning_rate": 2.8552284764133514e-06, + "loss": 0.6539, + "step": 3968 + }, + { + "epoch": 1.47, + "learning_rate": 2.8551915260500065e-06, + "loss": 0.5269, + "step": 3969 + }, + { + "epoch": 1.47, + "learning_rate": 2.855154575686661e-06, + "loss": 0.6497, + "step": 3970 + }, + { + "epoch": 1.47, + "learning_rate": 2.8551176253233157e-06, + "loss": 0.6083, + "step": 3971 + }, + { + "epoch": 1.47, + "learning_rate": 2.8550806749599704e-06, + "loss": 0.5823, + "step": 3972 + }, + { + "epoch": 1.47, + "learning_rate": 2.8550437245966254e-06, + "loss": 0.6255, + "step": 3973 + }, + { + "epoch": 1.47, + "learning_rate": 2.85500677423328e-06, + "loss": 0.6327, + "step": 3974 + }, + { + "epoch": 1.47, + "learning_rate": 2.8549698238699347e-06, + "loss": 0.6521, + "step": 3975 + }, + { + "epoch": 1.47, + "learning_rate": 2.8549328735065898e-06, + "loss": 0.6006, + "step": 3976 + }, + { + "epoch": 1.47, + "learning_rate": 2.8548959231432444e-06, + "loss": 0.5495, + "step": 3977 + }, + { + "epoch": 1.47, + "learning_rate": 2.854858972779899e-06, + "loss": 0.6759, + "step": 3978 + }, + { + "epoch": 1.47, + "learning_rate": 2.8548220224165537e-06, + "loss": 0.703, + "step": 3979 + }, + { + "epoch": 1.47, + "learning_rate": 2.8547850720532087e-06, + "loss": 0.5521, + "step": 3980 + }, + { + "epoch": 1.47, + "learning_rate": 2.8547481216898634e-06, + "loss": 0.7045, + "step": 3981 + }, + { + "epoch": 1.47, + "learning_rate": 2.8547111713265184e-06, + "loss": 0.6527, + "step": 3982 + }, + { + "epoch": 1.47, + "learning_rate": 2.854674220963173e-06, + "loss": 0.6029, + "step": 3983 + }, + { + "epoch": 1.47, + "learning_rate": 2.8546372705998273e-06, + "loss": 0.6955, + "step": 3984 + }, + { + "epoch": 1.47, + "learning_rate": 2.8546003202364823e-06, + "loss": 0.6516, + "step": 3985 + }, + { + "epoch": 1.47, + "learning_rate": 2.854563369873137e-06, + "loss": 0.6444, + "step": 3986 + }, + { + "epoch": 1.47, + "learning_rate": 2.854526419509792e-06, + "loss": 0.616, + "step": 3987 + }, + { + "epoch": 1.47, + "learning_rate": 2.8544894691464467e-06, + "loss": 0.6349, + "step": 3988 + }, + { + "epoch": 1.47, + "learning_rate": 2.8544525187831017e-06, + "loss": 0.5902, + "step": 3989 + }, + { + "epoch": 1.47, + "learning_rate": 2.854415568419756e-06, + "loss": 0.5766, + "step": 3990 + }, + { + "epoch": 1.47, + "learning_rate": 2.854378618056411e-06, + "loss": 0.5686, + "step": 3991 + }, + { + "epoch": 1.47, + "learning_rate": 2.8543416676930656e-06, + "loss": 0.57, + "step": 3992 + }, + { + "epoch": 1.47, + "learning_rate": 2.8543047173297207e-06, + "loss": 0.6217, + "step": 3993 + }, + { + "epoch": 1.47, + "learning_rate": 2.8542677669663753e-06, + "loss": 0.5889, + "step": 3994 + }, + { + "epoch": 1.48, + "learning_rate": 2.85423081660303e-06, + "loss": 0.6, + "step": 3995 + }, + { + "epoch": 1.48, + "learning_rate": 2.8541938662396846e-06, + "loss": 0.5765, + "step": 3996 + }, + { + "epoch": 1.48, + "learning_rate": 2.8541569158763392e-06, + "loss": 0.5551, + "step": 3997 + }, + { + "epoch": 1.48, + "learning_rate": 2.8541199655129943e-06, + "loss": 0.643, + "step": 3998 + }, + { + "epoch": 1.48, + "learning_rate": 2.854083015149649e-06, + "loss": 0.644, + "step": 3999 + }, + { + "epoch": 1.48, + "learning_rate": 2.854046064786304e-06, + "loss": 0.7074, + "step": 4000 + }, + { + "epoch": 1.48, + "learning_rate": 2.8540091144229586e-06, + "loss": 0.6239, + "step": 4001 + }, + { + "epoch": 1.48, + "learning_rate": 2.8539721640596133e-06, + "loss": 0.6394, + "step": 4002 + }, + { + "epoch": 1.48, + "learning_rate": 2.853935213696268e-06, + "loss": 0.6073, + "step": 4003 + }, + { + "epoch": 1.48, + "learning_rate": 2.8538982633329225e-06, + "loss": 0.6482, + "step": 4004 + }, + { + "epoch": 1.48, + "learning_rate": 2.8538613129695776e-06, + "loss": 0.7428, + "step": 4005 + }, + { + "epoch": 1.48, + "learning_rate": 2.8538243626062322e-06, + "loss": 0.5691, + "step": 4006 + }, + { + "epoch": 1.48, + "learning_rate": 2.8537874122428873e-06, + "loss": 0.6221, + "step": 4007 + }, + { + "epoch": 1.48, + "learning_rate": 2.853750461879542e-06, + "loss": 0.5077, + "step": 4008 + }, + { + "epoch": 1.48, + "learning_rate": 2.8537135115161966e-06, + "loss": 0.6606, + "step": 4009 + }, + { + "epoch": 1.48, + "learning_rate": 2.853676561152851e-06, + "loss": 0.6693, + "step": 4010 + }, + { + "epoch": 1.48, + "learning_rate": 2.8536396107895063e-06, + "loss": 0.5209, + "step": 4011 + }, + { + "epoch": 1.48, + "learning_rate": 2.853602660426161e-06, + "loss": 0.6473, + "step": 4012 + }, + { + "epoch": 1.48, + "learning_rate": 2.8535657100628155e-06, + "loss": 0.5906, + "step": 4013 + }, + { + "epoch": 1.48, + "learning_rate": 2.8535287596994706e-06, + "loss": 0.6115, + "step": 4014 + }, + { + "epoch": 1.48, + "learning_rate": 2.8534918093361252e-06, + "loss": 0.5866, + "step": 4015 + }, + { + "epoch": 1.48, + "learning_rate": 2.85345485897278e-06, + "loss": 0.6902, + "step": 4016 + }, + { + "epoch": 1.48, + "learning_rate": 2.8534179086094345e-06, + "loss": 0.5829, + "step": 4017 + }, + { + "epoch": 1.48, + "learning_rate": 2.8533809582460896e-06, + "loss": 0.6329, + "step": 4018 + }, + { + "epoch": 1.48, + "learning_rate": 2.853344007882744e-06, + "loss": 0.6448, + "step": 4019 + }, + { + "epoch": 1.48, + "learning_rate": 2.8533070575193993e-06, + "loss": 0.6797, + "step": 4020 + }, + { + "epoch": 1.48, + "learning_rate": 2.853270107156054e-06, + "loss": 0.5425, + "step": 4021 + }, + { + "epoch": 1.49, + "learning_rate": 2.8532331567927085e-06, + "loss": 0.6131, + "step": 4022 + }, + { + "epoch": 1.49, + "learning_rate": 2.853196206429363e-06, + "loss": 0.5899, + "step": 4023 + }, + { + "epoch": 1.49, + "learning_rate": 2.853159256066018e-06, + "loss": 0.6011, + "step": 4024 + }, + { + "epoch": 1.49, + "learning_rate": 2.853122305702673e-06, + "loss": 0.644, + "step": 4025 + }, + { + "epoch": 1.49, + "learning_rate": 2.8530853553393275e-06, + "loss": 0.5703, + "step": 4026 + }, + { + "epoch": 1.49, + "learning_rate": 2.8530484049759826e-06, + "loss": 0.553, + "step": 4027 + }, + { + "epoch": 1.49, + "learning_rate": 2.853011454612637e-06, + "loss": 0.6428, + "step": 4028 + }, + { + "epoch": 1.49, + "learning_rate": 2.852974504249292e-06, + "loss": 0.5581, + "step": 4029 + }, + { + "epoch": 1.49, + "learning_rate": 2.8529375538859465e-06, + "loss": 0.6221, + "step": 4030 + }, + { + "epoch": 1.49, + "learning_rate": 2.8529006035226015e-06, + "loss": 0.6292, + "step": 4031 + }, + { + "epoch": 1.49, + "learning_rate": 2.852863653159256e-06, + "loss": 0.5623, + "step": 4032 + }, + { + "epoch": 1.49, + "learning_rate": 2.852826702795911e-06, + "loss": 0.6147, + "step": 4033 + }, + { + "epoch": 1.49, + "learning_rate": 2.852789752432566e-06, + "loss": 0.6451, + "step": 4034 + }, + { + "epoch": 1.49, + "learning_rate": 2.85275280206922e-06, + "loss": 0.6118, + "step": 4035 + }, + { + "epoch": 1.49, + "learning_rate": 2.852715851705875e-06, + "loss": 0.507, + "step": 4036 + }, + { + "epoch": 1.49, + "learning_rate": 2.8526789013425298e-06, + "loss": 0.6051, + "step": 4037 + }, + { + "epoch": 1.49, + "learning_rate": 2.852641950979185e-06, + "loss": 0.7059, + "step": 4038 + }, + { + "epoch": 1.49, + "learning_rate": 2.8526050006158395e-06, + "loss": 0.5567, + "step": 4039 + }, + { + "epoch": 1.49, + "learning_rate": 2.8525680502524945e-06, + "loss": 0.6016, + "step": 4040 + }, + { + "epoch": 1.49, + "learning_rate": 2.8525310998891488e-06, + "loss": 0.647, + "step": 4041 + }, + { + "epoch": 1.49, + "learning_rate": 2.852494149525804e-06, + "loss": 0.6632, + "step": 4042 + }, + { + "epoch": 1.49, + "learning_rate": 2.8524571991624584e-06, + "loss": 0.6048, + "step": 4043 + }, + { + "epoch": 1.49, + "learning_rate": 2.852420248799113e-06, + "loss": 0.6071, + "step": 4044 + }, + { + "epoch": 1.49, + "learning_rate": 2.852383298435768e-06, + "loss": 0.7389, + "step": 4045 + }, + { + "epoch": 1.49, + "learning_rate": 2.8523463480724228e-06, + "loss": 0.6656, + "step": 4046 + }, + { + "epoch": 1.49, + "learning_rate": 2.8523093977090774e-06, + "loss": 0.6675, + "step": 4047 + }, + { + "epoch": 1.49, + "learning_rate": 2.852272447345732e-06, + "loss": 0.5657, + "step": 4048 + }, + { + "epoch": 1.5, + "learning_rate": 2.852235496982387e-06, + "loss": 0.6184, + "step": 4049 + }, + { + "epoch": 1.5, + "learning_rate": 2.8521985466190418e-06, + "loss": 0.6261, + "step": 4050 + }, + { + "epoch": 1.5, + "learning_rate": 2.852161596255697e-06, + "loss": 0.5893, + "step": 4051 + }, + { + "epoch": 1.5, + "learning_rate": 2.8521246458923514e-06, + "loss": 0.6565, + "step": 4052 + }, + { + "epoch": 1.5, + "learning_rate": 2.852087695529006e-06, + "loss": 0.5802, + "step": 4053 + }, + { + "epoch": 1.5, + "learning_rate": 2.8520507451656607e-06, + "loss": 0.6168, + "step": 4054 + }, + { + "epoch": 1.5, + "learning_rate": 2.8520137948023154e-06, + "loss": 0.5163, + "step": 4055 + }, + { + "epoch": 1.5, + "learning_rate": 2.8519768444389704e-06, + "loss": 0.5761, + "step": 4056 + }, + { + "epoch": 1.5, + "learning_rate": 2.851939894075625e-06, + "loss": 0.5896, + "step": 4057 + }, + { + "epoch": 1.5, + "learning_rate": 2.85190294371228e-06, + "loss": 0.6874, + "step": 4058 + }, + { + "epoch": 1.5, + "learning_rate": 2.8518659933489347e-06, + "loss": 0.6215, + "step": 4059 + }, + { + "epoch": 1.5, + "learning_rate": 2.8518290429855894e-06, + "loss": 0.7064, + "step": 4060 + }, + { + "epoch": 1.5, + "learning_rate": 2.851792092622244e-06, + "loss": 0.6443, + "step": 4061 + }, + { + "epoch": 1.5, + "learning_rate": 2.851755142258899e-06, + "loss": 0.6211, + "step": 4062 + }, + { + "epoch": 1.5, + "learning_rate": 2.8517181918955537e-06, + "loss": 0.6576, + "step": 4063 + }, + { + "epoch": 1.5, + "learning_rate": 2.8516812415322084e-06, + "loss": 0.5842, + "step": 4064 + }, + { + "epoch": 1.5, + "learning_rate": 2.8516442911688634e-06, + "loss": 0.6033, + "step": 4065 + }, + { + "epoch": 1.5, + "learning_rate": 2.851607340805518e-06, + "loss": 0.599, + "step": 4066 + }, + { + "epoch": 1.5, + "learning_rate": 2.8515703904421727e-06, + "loss": 0.6213, + "step": 4067 + }, + { + "epoch": 1.5, + "learning_rate": 2.8515334400788273e-06, + "loss": 0.6698, + "step": 4068 + }, + { + "epoch": 1.5, + "learning_rate": 2.8514964897154824e-06, + "loss": 0.6498, + "step": 4069 + }, + { + "epoch": 1.5, + "learning_rate": 2.851459539352137e-06, + "loss": 0.6553, + "step": 4070 + }, + { + "epoch": 1.5, + "learning_rate": 2.851422588988792e-06, + "loss": 0.6052, + "step": 4071 + }, + { + "epoch": 1.5, + "learning_rate": 2.8513856386254467e-06, + "loss": 0.6615, + "step": 4072 + }, + { + "epoch": 1.5, + "learning_rate": 2.8513486882621014e-06, + "loss": 0.5786, + "step": 4073 + }, + { + "epoch": 1.5, + "learning_rate": 2.851311737898756e-06, + "loss": 0.5848, + "step": 4074 + }, + { + "epoch": 1.5, + "learning_rate": 2.8512747875354106e-06, + "loss": 0.6304, + "step": 4075 + }, + { + "epoch": 1.51, + "learning_rate": 2.8512378371720657e-06, + "loss": 0.6134, + "step": 4076 + }, + { + "epoch": 1.51, + "learning_rate": 2.8512008868087203e-06, + "loss": 0.5446, + "step": 4077 + }, + { + "epoch": 1.51, + "learning_rate": 2.8511639364453754e-06, + "loss": 0.5628, + "step": 4078 + }, + { + "epoch": 1.51, + "learning_rate": 2.85112698608203e-06, + "loss": 0.5955, + "step": 4079 + }, + { + "epoch": 1.51, + "learning_rate": 2.8510900357186847e-06, + "loss": 0.5369, + "step": 4080 + }, + { + "epoch": 1.51, + "learning_rate": 2.8510530853553393e-06, + "loss": 0.5894, + "step": 4081 + }, + { + "epoch": 1.51, + "learning_rate": 2.8510161349919944e-06, + "loss": 0.607, + "step": 4082 + }, + { + "epoch": 1.51, + "learning_rate": 2.850979184628649e-06, + "loss": 0.5863, + "step": 4083 + }, + { + "epoch": 1.51, + "learning_rate": 2.8509422342653036e-06, + "loss": 0.6631, + "step": 4084 + }, + { + "epoch": 1.51, + "learning_rate": 2.8509052839019587e-06, + "loss": 0.6748, + "step": 4085 + }, + { + "epoch": 1.51, + "learning_rate": 2.850868333538613e-06, + "loss": 0.5803, + "step": 4086 + }, + { + "epoch": 1.51, + "learning_rate": 2.850831383175268e-06, + "loss": 0.556, + "step": 4087 + }, + { + "epoch": 1.51, + "learning_rate": 2.8507944328119226e-06, + "loss": 0.6574, + "step": 4088 + }, + { + "epoch": 1.51, + "learning_rate": 2.8507574824485777e-06, + "loss": 0.6188, + "step": 4089 + }, + { + "epoch": 1.51, + "learning_rate": 2.8507205320852323e-06, + "loss": 0.6314, + "step": 4090 + }, + { + "epoch": 1.51, + "learning_rate": 2.8506835817218873e-06, + "loss": 0.5954, + "step": 4091 + }, + { + "epoch": 1.51, + "learning_rate": 2.8506466313585416e-06, + "loss": 0.6694, + "step": 4092 + }, + { + "epoch": 1.51, + "learning_rate": 2.850609680995196e-06, + "loss": 0.5611, + "step": 4093 + }, + { + "epoch": 1.51, + "learning_rate": 2.8505727306318513e-06, + "loss": 0.6283, + "step": 4094 + }, + { + "epoch": 1.51, + "learning_rate": 2.850535780268506e-06, + "loss": 0.5426, + "step": 4095 + }, + { + "epoch": 1.51, + "learning_rate": 2.850498829905161e-06, + "loss": 0.7052, + "step": 4096 + }, + { + "epoch": 1.51, + "learning_rate": 2.8504618795418156e-06, + "loss": 0.6489, + "step": 4097 + }, + { + "epoch": 1.51, + "learning_rate": 2.8504249291784702e-06, + "loss": 0.686, + "step": 4098 + }, + { + "epoch": 1.51, + "learning_rate": 2.850387978815125e-06, + "loss": 0.554, + "step": 4099 + }, + { + "epoch": 1.51, + "learning_rate": 2.85035102845178e-06, + "loss": 0.618, + "step": 4100 + }, + { + "epoch": 1.51, + "learning_rate": 2.8503140780884346e-06, + "loss": 0.5814, + "step": 4101 + }, + { + "epoch": 1.51, + "learning_rate": 2.8502771277250896e-06, + "loss": 0.6408, + "step": 4102 + }, + { + "epoch": 1.52, + "learning_rate": 2.8502401773617443e-06, + "loss": 0.6347, + "step": 4103 + }, + { + "epoch": 1.52, + "learning_rate": 2.850203226998399e-06, + "loss": 0.7202, + "step": 4104 + }, + { + "epoch": 1.52, + "learning_rate": 2.8501662766350535e-06, + "loss": 0.6571, + "step": 4105 + }, + { + "epoch": 1.52, + "learning_rate": 2.850129326271708e-06, + "loss": 0.635, + "step": 4106 + }, + { + "epoch": 1.52, + "learning_rate": 2.8500923759083632e-06, + "loss": 0.6777, + "step": 4107 + }, + { + "epoch": 1.52, + "learning_rate": 2.850055425545018e-06, + "loss": 0.5874, + "step": 4108 + }, + { + "epoch": 1.52, + "learning_rate": 2.850018475181673e-06, + "loss": 0.5959, + "step": 4109 + }, + { + "epoch": 1.52, + "learning_rate": 2.8499815248183276e-06, + "loss": 0.6846, + "step": 4110 + }, + { + "epoch": 1.52, + "learning_rate": 2.849944574454982e-06, + "loss": 0.5675, + "step": 4111 + }, + { + "epoch": 1.52, + "learning_rate": 2.849907624091637e-06, + "loss": 0.6841, + "step": 4112 + }, + { + "epoch": 1.52, + "learning_rate": 2.8498706737282915e-06, + "loss": 0.5888, + "step": 4113 + }, + { + "epoch": 1.52, + "learning_rate": 2.8498337233649465e-06, + "loss": 0.5772, + "step": 4114 + }, + { + "epoch": 1.52, + "learning_rate": 2.849796773001601e-06, + "loss": 0.5267, + "step": 4115 + }, + { + "epoch": 1.52, + "learning_rate": 2.8497598226382562e-06, + "loss": 0.5754, + "step": 4116 + }, + { + "epoch": 1.52, + "learning_rate": 2.849722872274911e-06, + "loss": 0.5602, + "step": 4117 + }, + { + "epoch": 1.52, + "learning_rate": 2.8496859219115655e-06, + "loss": 0.6082, + "step": 4118 + }, + { + "epoch": 1.52, + "learning_rate": 2.84964897154822e-06, + "loss": 0.5759, + "step": 4119 + }, + { + "epoch": 1.52, + "learning_rate": 2.849612021184875e-06, + "loss": 0.6218, + "step": 4120 + }, + { + "epoch": 1.52, + "learning_rate": 2.84957507082153e-06, + "loss": 0.6359, + "step": 4121 + }, + { + "epoch": 1.52, + "learning_rate": 2.849538120458185e-06, + "loss": 0.5585, + "step": 4122 + }, + { + "epoch": 1.52, + "learning_rate": 2.8495011700948395e-06, + "loss": 0.6082, + "step": 4123 + }, + { + "epoch": 1.52, + "learning_rate": 2.849464219731494e-06, + "loss": 0.6773, + "step": 4124 + }, + { + "epoch": 1.52, + "learning_rate": 2.849427269368149e-06, + "loss": 0.5921, + "step": 4125 + }, + { + "epoch": 1.52, + "learning_rate": 2.8493903190048034e-06, + "loss": 0.5918, + "step": 4126 + }, + { + "epoch": 1.52, + "learning_rate": 2.8493533686414585e-06, + "loss": 0.6961, + "step": 4127 + }, + { + "epoch": 1.52, + "learning_rate": 2.849316418278113e-06, + "loss": 0.6989, + "step": 4128 + }, + { + "epoch": 1.52, + "learning_rate": 2.849279467914768e-06, + "loss": 0.5901, + "step": 4129 + }, + { + "epoch": 1.53, + "learning_rate": 2.849242517551423e-06, + "loss": 0.6293, + "step": 4130 + }, + { + "epoch": 1.53, + "learning_rate": 2.8492055671880775e-06, + "loss": 0.7415, + "step": 4131 + }, + { + "epoch": 1.53, + "learning_rate": 2.849168616824732e-06, + "loss": 0.6074, + "step": 4132 + }, + { + "epoch": 1.53, + "learning_rate": 2.8491316664613867e-06, + "loss": 0.576, + "step": 4133 + }, + { + "epoch": 1.53, + "learning_rate": 2.849094716098042e-06, + "loss": 0.6104, + "step": 4134 + }, + { + "epoch": 1.53, + "learning_rate": 2.8490577657346964e-06, + "loss": 0.5402, + "step": 4135 + }, + { + "epoch": 1.53, + "learning_rate": 2.8490208153713515e-06, + "loss": 0.5367, + "step": 4136 + }, + { + "epoch": 1.53, + "learning_rate": 2.8489838650080057e-06, + "loss": 0.5963, + "step": 4137 + }, + { + "epoch": 1.53, + "learning_rate": 2.8489469146446608e-06, + "loss": 0.673, + "step": 4138 + }, + { + "epoch": 1.53, + "learning_rate": 2.8489099642813154e-06, + "loss": 0.6451, + "step": 4139 + }, + { + "epoch": 1.53, + "learning_rate": 2.8488730139179705e-06, + "loss": 0.6452, + "step": 4140 + }, + { + "epoch": 1.53, + "learning_rate": 2.848836063554625e-06, + "loss": 0.5906, + "step": 4141 + }, + { + "epoch": 1.53, + "learning_rate": 2.8487991131912797e-06, + "loss": 0.6592, + "step": 4142 + }, + { + "epoch": 1.53, + "learning_rate": 2.8487621628279344e-06, + "loss": 0.5621, + "step": 4143 + }, + { + "epoch": 1.53, + "learning_rate": 2.848725212464589e-06, + "loss": 0.663, + "step": 4144 + }, + { + "epoch": 1.53, + "learning_rate": 2.848688262101244e-06, + "loss": 0.5841, + "step": 4145 + }, + { + "epoch": 1.53, + "learning_rate": 2.8486513117378987e-06, + "loss": 0.6453, + "step": 4146 + }, + { + "epoch": 1.53, + "learning_rate": 2.8486143613745538e-06, + "loss": 0.6331, + "step": 4147 + }, + { + "epoch": 1.53, + "learning_rate": 2.8485774110112084e-06, + "loss": 0.5673, + "step": 4148 + }, + { + "epoch": 1.53, + "learning_rate": 2.848540460647863e-06, + "loss": 0.5992, + "step": 4149 + }, + { + "epoch": 1.53, + "learning_rate": 2.8485035102845177e-06, + "loss": 0.5741, + "step": 4150 + }, + { + "epoch": 1.53, + "learning_rate": 2.8484665599211727e-06, + "loss": 0.5312, + "step": 4151 + }, + { + "epoch": 1.53, + "learning_rate": 2.8484296095578274e-06, + "loss": 0.5897, + "step": 4152 + }, + { + "epoch": 1.53, + "learning_rate": 2.848392659194482e-06, + "loss": 0.6181, + "step": 4153 + }, + { + "epoch": 1.53, + "learning_rate": 2.848355708831137e-06, + "loss": 0.5189, + "step": 4154 + }, + { + "epoch": 1.53, + "learning_rate": 2.8483187584677917e-06, + "loss": 0.652, + "step": 4155 + }, + { + "epoch": 1.53, + "learning_rate": 2.8482818081044463e-06, + "loss": 0.5758, + "step": 4156 + }, + { + "epoch": 1.54, + "learning_rate": 2.848244857741101e-06, + "loss": 0.6446, + "step": 4157 + }, + { + "epoch": 1.54, + "learning_rate": 2.848207907377756e-06, + "loss": 0.5855, + "step": 4158 + }, + { + "epoch": 1.54, + "learning_rate": 2.8481709570144107e-06, + "loss": 0.7059, + "step": 4159 + }, + { + "epoch": 1.54, + "learning_rate": 2.8481340066510657e-06, + "loss": 0.6162, + "step": 4160 + }, + { + "epoch": 1.54, + "learning_rate": 2.8480970562877204e-06, + "loss": 0.6381, + "step": 4161 + }, + { + "epoch": 1.54, + "learning_rate": 2.848060105924375e-06, + "loss": 0.5981, + "step": 4162 + }, + { + "epoch": 1.54, + "learning_rate": 2.8480231555610296e-06, + "loss": 0.6835, + "step": 4163 + }, + { + "epoch": 1.54, + "learning_rate": 2.8479862051976843e-06, + "loss": 0.6578, + "step": 4164 + }, + { + "epoch": 1.54, + "learning_rate": 2.8479492548343393e-06, + "loss": 0.5497, + "step": 4165 + }, + { + "epoch": 1.54, + "learning_rate": 2.847912304470994e-06, + "loss": 0.5831, + "step": 4166 + }, + { + "epoch": 1.54, + "learning_rate": 2.847875354107649e-06, + "loss": 0.6538, + "step": 4167 + }, + { + "epoch": 1.54, + "learning_rate": 2.8478384037443037e-06, + "loss": 0.5002, + "step": 4168 + }, + { + "epoch": 1.54, + "learning_rate": 2.8478014533809583e-06, + "loss": 0.5826, + "step": 4169 + }, + { + "epoch": 1.54, + "learning_rate": 2.847764503017613e-06, + "loss": 0.6046, + "step": 4170 + }, + { + "epoch": 1.54, + "learning_rate": 2.847727552654268e-06, + "loss": 0.6292, + "step": 4171 + }, + { + "epoch": 1.54, + "learning_rate": 2.8476906022909226e-06, + "loss": 0.6513, + "step": 4172 + }, + { + "epoch": 1.54, + "learning_rate": 2.8476536519275773e-06, + "loss": 0.6812, + "step": 4173 + }, + { + "epoch": 1.54, + "learning_rate": 2.8476167015642323e-06, + "loss": 0.5825, + "step": 4174 + }, + { + "epoch": 1.54, + "learning_rate": 2.8475797512008866e-06, + "loss": 0.638, + "step": 4175 + }, + { + "epoch": 1.54, + "learning_rate": 2.8475428008375416e-06, + "loss": 0.6407, + "step": 4176 + }, + { + "epoch": 1.54, + "learning_rate": 2.8475058504741962e-06, + "loss": 0.6315, + "step": 4177 + }, + { + "epoch": 1.54, + "learning_rate": 2.8474689001108513e-06, + "loss": 0.6555, + "step": 4178 + }, + { + "epoch": 1.54, + "learning_rate": 2.847431949747506e-06, + "loss": 0.7252, + "step": 4179 + }, + { + "epoch": 1.54, + "learning_rate": 2.847394999384161e-06, + "loss": 0.6318, + "step": 4180 + }, + { + "epoch": 1.54, + "learning_rate": 2.8473580490208152e-06, + "loss": 0.676, + "step": 4181 + }, + { + "epoch": 1.54, + "learning_rate": 2.8473210986574703e-06, + "loss": 0.6538, + "step": 4182 + }, + { + "epoch": 1.54, + "learning_rate": 2.847284148294125e-06, + "loss": 0.6646, + "step": 4183 + }, + { + "epoch": 1.55, + "learning_rate": 2.8472471979307795e-06, + "loss": 0.6243, + "step": 4184 + }, + { + "epoch": 1.55, + "learning_rate": 2.8472102475674346e-06, + "loss": 0.475, + "step": 4185 + }, + { + "epoch": 1.55, + "learning_rate": 2.8471732972040892e-06, + "loss": 0.5643, + "step": 4186 + }, + { + "epoch": 1.55, + "learning_rate": 2.847136346840744e-06, + "loss": 0.6444, + "step": 4187 + }, + { + "epoch": 1.55, + "learning_rate": 2.8470993964773985e-06, + "loss": 0.5928, + "step": 4188 + }, + { + "epoch": 1.55, + "learning_rate": 2.8470624461140536e-06, + "loss": 0.5651, + "step": 4189 + }, + { + "epoch": 1.55, + "learning_rate": 2.8470254957507082e-06, + "loss": 0.6614, + "step": 4190 + }, + { + "epoch": 1.55, + "learning_rate": 2.8469885453873633e-06, + "loss": 0.5428, + "step": 4191 + }, + { + "epoch": 1.55, + "learning_rate": 2.846951595024018e-06, + "loss": 0.581, + "step": 4192 + }, + { + "epoch": 1.55, + "learning_rate": 2.8469146446606725e-06, + "loss": 0.6524, + "step": 4193 + }, + { + "epoch": 1.55, + "learning_rate": 2.846877694297327e-06, + "loss": 0.5292, + "step": 4194 + }, + { + "epoch": 1.55, + "learning_rate": 2.846840743933982e-06, + "loss": 0.6533, + "step": 4195 + }, + { + "epoch": 1.55, + "learning_rate": 2.846803793570637e-06, + "loss": 0.6561, + "step": 4196 + }, + { + "epoch": 1.55, + "learning_rate": 2.8467668432072915e-06, + "loss": 0.6156, + "step": 4197 + }, + { + "epoch": 1.55, + "learning_rate": 2.8467298928439466e-06, + "loss": 0.5667, + "step": 4198 + }, + { + "epoch": 1.55, + "learning_rate": 2.8466929424806012e-06, + "loss": 0.5799, + "step": 4199 + }, + { + "epoch": 1.55, + "learning_rate": 2.846655992117256e-06, + "loss": 0.6171, + "step": 4200 + }, + { + "epoch": 1.55, + "learning_rate": 2.8466190417539105e-06, + "loss": 0.6923, + "step": 4201 + }, + { + "epoch": 1.55, + "learning_rate": 2.8465820913905655e-06, + "loss": 0.5824, + "step": 4202 + }, + { + "epoch": 1.55, + "learning_rate": 2.84654514102722e-06, + "loss": 0.6369, + "step": 4203 + }, + { + "epoch": 1.55, + "learning_rate": 2.846508190663875e-06, + "loss": 0.5792, + "step": 4204 + }, + { + "epoch": 1.55, + "learning_rate": 2.84647124030053e-06, + "loss": 0.6744, + "step": 4205 + }, + { + "epoch": 1.55, + "learning_rate": 2.8464342899371845e-06, + "loss": 0.5681, + "step": 4206 + }, + { + "epoch": 1.55, + "learning_rate": 2.846397339573839e-06, + "loss": 0.6885, + "step": 4207 + }, + { + "epoch": 1.55, + "learning_rate": 2.846360389210494e-06, + "loss": 0.5581, + "step": 4208 + }, + { + "epoch": 1.55, + "learning_rate": 2.846323438847149e-06, + "loss": 0.6449, + "step": 4209 + }, + { + "epoch": 1.55, + "learning_rate": 2.8462864884838035e-06, + "loss": 0.6037, + "step": 4210 + }, + { + "epoch": 1.56, + "learning_rate": 2.8462495381204585e-06, + "loss": 0.6062, + "step": 4211 + }, + { + "epoch": 1.56, + "learning_rate": 2.846212587757113e-06, + "loss": 0.6145, + "step": 4212 + }, + { + "epoch": 1.56, + "learning_rate": 2.846175637393768e-06, + "loss": 0.6116, + "step": 4213 + }, + { + "epoch": 1.56, + "learning_rate": 2.8461386870304225e-06, + "loss": 0.5947, + "step": 4214 + }, + { + "epoch": 1.56, + "learning_rate": 2.846101736667077e-06, + "loss": 0.6399, + "step": 4215 + }, + { + "epoch": 1.56, + "learning_rate": 2.846064786303732e-06, + "loss": 0.6807, + "step": 4216 + }, + { + "epoch": 1.56, + "learning_rate": 2.8460278359403868e-06, + "loss": 0.5784, + "step": 4217 + }, + { + "epoch": 1.56, + "learning_rate": 2.845990885577042e-06, + "loss": 0.5613, + "step": 4218 + }, + { + "epoch": 1.56, + "learning_rate": 2.8459539352136965e-06, + "loss": 0.5981, + "step": 4219 + }, + { + "epoch": 1.56, + "learning_rate": 2.845916984850351e-06, + "loss": 0.6121, + "step": 4220 + }, + { + "epoch": 1.56, + "learning_rate": 2.8458800344870058e-06, + "loss": 0.6828, + "step": 4221 + }, + { + "epoch": 1.56, + "learning_rate": 2.845843084123661e-06, + "loss": 0.583, + "step": 4222 + }, + { + "epoch": 1.56, + "learning_rate": 2.8458061337603155e-06, + "loss": 0.6225, + "step": 4223 + }, + { + "epoch": 1.56, + "learning_rate": 2.84576918339697e-06, + "loss": 0.6164, + "step": 4224 + }, + { + "epoch": 1.56, + "learning_rate": 2.845732233033625e-06, + "loss": 0.6374, + "step": 4225 + }, + { + "epoch": 1.56, + "learning_rate": 2.8456952826702794e-06, + "loss": 0.6065, + "step": 4226 + }, + { + "epoch": 1.56, + "learning_rate": 2.8456583323069344e-06, + "loss": 0.5042, + "step": 4227 + }, + { + "epoch": 1.56, + "learning_rate": 2.845621381943589e-06, + "loss": 0.5908, + "step": 4228 + }, + { + "epoch": 1.56, + "learning_rate": 2.845584431580244e-06, + "loss": 0.5766, + "step": 4229 + }, + { + "epoch": 1.56, + "learning_rate": 2.8455474812168988e-06, + "loss": 0.6159, + "step": 4230 + }, + { + "epoch": 1.56, + "learning_rate": 2.845510530853554e-06, + "loss": 0.6756, + "step": 4231 + }, + { + "epoch": 1.56, + "learning_rate": 2.845473580490208e-06, + "loss": 0.5727, + "step": 4232 + }, + { + "epoch": 1.56, + "learning_rate": 2.8454366301268627e-06, + "loss": 0.6126, + "step": 4233 + }, + { + "epoch": 1.56, + "learning_rate": 2.8453996797635177e-06, + "loss": 0.6491, + "step": 4234 + }, + { + "epoch": 1.56, + "learning_rate": 2.8453627294001724e-06, + "loss": 0.6566, + "step": 4235 + }, + { + "epoch": 1.56, + "learning_rate": 2.8453257790368274e-06, + "loss": 0.5738, + "step": 4236 + }, + { + "epoch": 1.56, + "learning_rate": 2.845288828673482e-06, + "loss": 0.7405, + "step": 4237 + }, + { + "epoch": 1.56, + "learning_rate": 2.8452518783101367e-06, + "loss": 0.657, + "step": 4238 + }, + { + "epoch": 1.57, + "learning_rate": 2.8452149279467913e-06, + "loss": 0.6614, + "step": 4239 + }, + { + "epoch": 1.57, + "learning_rate": 2.8451779775834464e-06, + "loss": 0.5795, + "step": 4240 + }, + { + "epoch": 1.57, + "learning_rate": 2.845141027220101e-06, + "loss": 0.6085, + "step": 4241 + }, + { + "epoch": 1.57, + "learning_rate": 2.8451040768567557e-06, + "loss": 0.603, + "step": 4242 + }, + { + "epoch": 1.57, + "learning_rate": 2.8450671264934107e-06, + "loss": 0.5642, + "step": 4243 + }, + { + "epoch": 1.57, + "learning_rate": 2.8450301761300654e-06, + "loss": 0.6115, + "step": 4244 + }, + { + "epoch": 1.57, + "learning_rate": 2.84499322576672e-06, + "loss": 0.6109, + "step": 4245 + }, + { + "epoch": 1.57, + "learning_rate": 2.8449562754033746e-06, + "loss": 0.6815, + "step": 4246 + }, + { + "epoch": 1.57, + "learning_rate": 2.8449193250400297e-06, + "loss": 0.6114, + "step": 4247 + }, + { + "epoch": 1.57, + "learning_rate": 2.8448823746766843e-06, + "loss": 0.529, + "step": 4248 + }, + { + "epoch": 1.57, + "learning_rate": 2.8448454243133394e-06, + "loss": 0.5978, + "step": 4249 + }, + { + "epoch": 1.57, + "learning_rate": 2.844808473949994e-06, + "loss": 0.6162, + "step": 4250 + }, + { + "epoch": 1.57, + "learning_rate": 2.8447715235866487e-06, + "loss": 0.6432, + "step": 4251 + }, + { + "epoch": 1.57, + "learning_rate": 2.8447345732233033e-06, + "loss": 0.4943, + "step": 4252 + }, + { + "epoch": 1.57, + "learning_rate": 2.844697622859958e-06, + "loss": 0.5589, + "step": 4253 + }, + { + "epoch": 1.57, + "learning_rate": 2.844660672496613e-06, + "loss": 0.5776, + "step": 4254 + }, + { + "epoch": 1.57, + "learning_rate": 2.8446237221332676e-06, + "loss": 0.6166, + "step": 4255 + }, + { + "epoch": 1.57, + "learning_rate": 2.8445867717699227e-06, + "loss": 0.6733, + "step": 4256 + }, + { + "epoch": 1.57, + "learning_rate": 2.8445498214065773e-06, + "loss": 0.5921, + "step": 4257 + }, + { + "epoch": 1.57, + "learning_rate": 2.844512871043232e-06, + "loss": 0.6082, + "step": 4258 + }, + { + "epoch": 1.57, + "learning_rate": 2.8444759206798866e-06, + "loss": 0.5839, + "step": 4259 + }, + { + "epoch": 1.57, + "learning_rate": 2.8444389703165417e-06, + "loss": 0.5801, + "step": 4260 + }, + { + "epoch": 1.57, + "learning_rate": 2.8444020199531963e-06, + "loss": 0.5297, + "step": 4261 + }, + { + "epoch": 1.57, + "learning_rate": 2.844365069589851e-06, + "loss": 0.6069, + "step": 4262 + }, + { + "epoch": 1.57, + "learning_rate": 2.844328119226506e-06, + "loss": 0.5881, + "step": 4263 + }, + { + "epoch": 1.57, + "learning_rate": 2.8442911688631606e-06, + "loss": 0.5584, + "step": 4264 + }, + { + "epoch": 1.57, + "learning_rate": 2.8442542184998153e-06, + "loss": 0.6818, + "step": 4265 + }, + { + "epoch": 1.58, + "learning_rate": 2.84421726813647e-06, + "loss": 0.5699, + "step": 4266 + }, + { + "epoch": 1.58, + "learning_rate": 2.844180317773125e-06, + "loss": 0.6562, + "step": 4267 + }, + { + "epoch": 1.58, + "learning_rate": 2.8441433674097796e-06, + "loss": 0.6698, + "step": 4268 + }, + { + "epoch": 1.58, + "learning_rate": 2.8441064170464347e-06, + "loss": 0.6617, + "step": 4269 + }, + { + "epoch": 1.58, + "learning_rate": 2.8440694666830893e-06, + "loss": 0.6664, + "step": 4270 + }, + { + "epoch": 1.58, + "learning_rate": 2.844032516319744e-06, + "loss": 0.5712, + "step": 4271 + }, + { + "epoch": 1.58, + "learning_rate": 2.8439955659563986e-06, + "loss": 0.588, + "step": 4272 + }, + { + "epoch": 1.58, + "learning_rate": 2.843958615593053e-06, + "loss": 0.648, + "step": 4273 + }, + { + "epoch": 1.58, + "learning_rate": 2.8439216652297083e-06, + "loss": 0.5604, + "step": 4274 + }, + { + "epoch": 1.58, + "learning_rate": 2.843884714866363e-06, + "loss": 0.6338, + "step": 4275 + }, + { + "epoch": 1.58, + "learning_rate": 2.843847764503018e-06, + "loss": 0.591, + "step": 4276 + }, + { + "epoch": 1.58, + "learning_rate": 2.843810814139672e-06, + "loss": 0.6357, + "step": 4277 + }, + { + "epoch": 1.58, + "learning_rate": 2.8437738637763272e-06, + "loss": 0.6517, + "step": 4278 + }, + { + "epoch": 1.58, + "learning_rate": 2.843736913412982e-06, + "loss": 0.6212, + "step": 4279 + }, + { + "epoch": 1.58, + "learning_rate": 2.843699963049637e-06, + "loss": 0.6719, + "step": 4280 + }, + { + "epoch": 1.58, + "learning_rate": 2.8436630126862916e-06, + "loss": 0.6434, + "step": 4281 + }, + { + "epoch": 1.58, + "learning_rate": 2.843626062322946e-06, + "loss": 0.6844, + "step": 4282 + }, + { + "epoch": 1.58, + "learning_rate": 2.843589111959601e-06, + "loss": 0.6135, + "step": 4283 + }, + { + "epoch": 1.58, + "learning_rate": 2.8435521615962555e-06, + "loss": 0.5946, + "step": 4284 + }, + { + "epoch": 1.58, + "learning_rate": 2.8435152112329105e-06, + "loss": 0.6348, + "step": 4285 + }, + { + "epoch": 1.58, + "learning_rate": 2.843478260869565e-06, + "loss": 0.5877, + "step": 4286 + }, + { + "epoch": 1.58, + "learning_rate": 2.8434413105062202e-06, + "loss": 0.6247, + "step": 4287 + }, + { + "epoch": 1.58, + "learning_rate": 2.843404360142875e-06, + "loss": 0.5768, + "step": 4288 + }, + { + "epoch": 1.58, + "learning_rate": 2.8433674097795295e-06, + "loss": 0.6184, + "step": 4289 + }, + { + "epoch": 1.58, + "learning_rate": 2.843330459416184e-06, + "loss": 0.6424, + "step": 4290 + }, + { + "epoch": 1.58, + "learning_rate": 2.843293509052839e-06, + "loss": 0.6428, + "step": 4291 + }, + { + "epoch": 1.58, + "learning_rate": 2.843256558689494e-06, + "loss": 0.5527, + "step": 4292 + }, + { + "epoch": 1.59, + "learning_rate": 2.8432196083261485e-06, + "loss": 0.6693, + "step": 4293 + }, + { + "epoch": 1.59, + "learning_rate": 2.8431826579628035e-06, + "loss": 0.5746, + "step": 4294 + }, + { + "epoch": 1.59, + "learning_rate": 2.843145707599458e-06, + "loss": 0.6052, + "step": 4295 + }, + { + "epoch": 1.59, + "learning_rate": 2.843108757236113e-06, + "loss": 0.6983, + "step": 4296 + }, + { + "epoch": 1.59, + "learning_rate": 2.8430718068727674e-06, + "loss": 0.6435, + "step": 4297 + }, + { + "epoch": 1.59, + "learning_rate": 2.8430348565094225e-06, + "loss": 0.6161, + "step": 4298 + }, + { + "epoch": 1.59, + "learning_rate": 2.842997906146077e-06, + "loss": 0.6139, + "step": 4299 + }, + { + "epoch": 1.59, + "learning_rate": 2.842960955782732e-06, + "loss": 0.5758, + "step": 4300 + }, + { + "epoch": 1.59, + "learning_rate": 2.842924005419387e-06, + "loss": 0.6588, + "step": 4301 + }, + { + "epoch": 1.59, + "learning_rate": 2.8428870550560415e-06, + "loss": 0.6247, + "step": 4302 + }, + { + "epoch": 1.59, + "learning_rate": 2.842850104692696e-06, + "loss": 0.5742, + "step": 4303 + }, + { + "epoch": 1.59, + "learning_rate": 2.8428131543293507e-06, + "loss": 0.5636, + "step": 4304 + }, + { + "epoch": 1.59, + "learning_rate": 2.842776203966006e-06, + "loss": 0.6372, + "step": 4305 + }, + { + "epoch": 1.59, + "learning_rate": 2.8427392536026604e-06, + "loss": 0.5975, + "step": 4306 + }, + { + "epoch": 1.59, + "learning_rate": 2.8427023032393155e-06, + "loss": 0.5917, + "step": 4307 + }, + { + "epoch": 1.59, + "learning_rate": 2.84266535287597e-06, + "loss": 0.6135, + "step": 4308 + }, + { + "epoch": 1.59, + "learning_rate": 2.8426284025126248e-06, + "loss": 0.6255, + "step": 4309 + }, + { + "epoch": 1.59, + "learning_rate": 2.8425914521492794e-06, + "loss": 0.6408, + "step": 4310 + }, + { + "epoch": 1.59, + "learning_rate": 2.8425545017859345e-06, + "loss": 0.6438, + "step": 4311 + }, + { + "epoch": 1.59, + "learning_rate": 2.842517551422589e-06, + "loss": 0.6425, + "step": 4312 + }, + { + "epoch": 1.59, + "learning_rate": 2.8424806010592437e-06, + "loss": 0.5933, + "step": 4313 + }, + { + "epoch": 1.59, + "learning_rate": 2.842443650695899e-06, + "loss": 0.6728, + "step": 4314 + }, + { + "epoch": 1.59, + "learning_rate": 2.8424067003325534e-06, + "loss": 0.4889, + "step": 4315 + }, + { + "epoch": 1.59, + "learning_rate": 2.842369749969208e-06, + "loss": 0.574, + "step": 4316 + }, + { + "epoch": 1.59, + "learning_rate": 2.8423327996058627e-06, + "loss": 0.6236, + "step": 4317 + }, + { + "epoch": 1.59, + "learning_rate": 2.8422958492425178e-06, + "loss": 0.6072, + "step": 4318 + }, + { + "epoch": 1.59, + "learning_rate": 2.8422588988791724e-06, + "loss": 0.6137, + "step": 4319 + }, + { + "epoch": 1.6, + "learning_rate": 2.8422219485158275e-06, + "loss": 0.5921, + "step": 4320 + }, + { + "epoch": 1.6, + "learning_rate": 2.842184998152482e-06, + "loss": 0.5975, + "step": 4321 + }, + { + "epoch": 1.6, + "learning_rate": 2.8421480477891363e-06, + "loss": 0.6568, + "step": 4322 + }, + { + "epoch": 1.6, + "learning_rate": 2.8421110974257914e-06, + "loss": 0.509, + "step": 4323 + }, + { + "epoch": 1.6, + "learning_rate": 2.842074147062446e-06, + "loss": 0.5842, + "step": 4324 + }, + { + "epoch": 1.6, + "learning_rate": 2.842037196699101e-06, + "loss": 0.5959, + "step": 4325 + }, + { + "epoch": 1.6, + "learning_rate": 2.8420002463357557e-06, + "loss": 0.6023, + "step": 4326 + }, + { + "epoch": 1.6, + "learning_rate": 2.8419632959724103e-06, + "loss": 0.6057, + "step": 4327 + }, + { + "epoch": 1.6, + "learning_rate": 2.841926345609065e-06, + "loss": 0.6696, + "step": 4328 + }, + { + "epoch": 1.6, + "learning_rate": 2.84188939524572e-06, + "loss": 0.6381, + "step": 4329 + }, + { + "epoch": 1.6, + "learning_rate": 2.8418524448823747e-06, + "loss": 0.5697, + "step": 4330 + }, + { + "epoch": 1.6, + "learning_rate": 2.8418154945190297e-06, + "loss": 0.6052, + "step": 4331 + }, + { + "epoch": 1.6, + "learning_rate": 2.8417785441556844e-06, + "loss": 0.6333, + "step": 4332 + }, + { + "epoch": 1.6, + "learning_rate": 2.841741593792339e-06, + "loss": 0.608, + "step": 4333 + }, + { + "epoch": 1.6, + "learning_rate": 2.8417046434289936e-06, + "loss": 0.6899, + "step": 4334 + }, + { + "epoch": 1.6, + "learning_rate": 2.8416676930656483e-06, + "loss": 0.6206, + "step": 4335 + }, + { + "epoch": 1.6, + "learning_rate": 2.8416307427023033e-06, + "loss": 0.5749, + "step": 4336 + }, + { + "epoch": 1.6, + "learning_rate": 2.841593792338958e-06, + "loss": 0.6702, + "step": 4337 + }, + { + "epoch": 1.6, + "learning_rate": 2.841556841975613e-06, + "loss": 0.7148, + "step": 4338 + }, + { + "epoch": 1.6, + "learning_rate": 2.8415198916122677e-06, + "loss": 0.5464, + "step": 4339 + }, + { + "epoch": 1.6, + "learning_rate": 2.8414829412489223e-06, + "loss": 0.7168, + "step": 4340 + }, + { + "epoch": 1.6, + "learning_rate": 2.841445990885577e-06, + "loss": 0.5937, + "step": 4341 + }, + { + "epoch": 1.6, + "learning_rate": 2.8414090405222316e-06, + "loss": 0.6142, + "step": 4342 + }, + { + "epoch": 1.6, + "learning_rate": 2.8413720901588866e-06, + "loss": 0.5812, + "step": 4343 + }, + { + "epoch": 1.6, + "learning_rate": 2.8413351397955413e-06, + "loss": 0.6671, + "step": 4344 + }, + { + "epoch": 1.6, + "learning_rate": 2.8412981894321963e-06, + "loss": 0.5662, + "step": 4345 + }, + { + "epoch": 1.6, + "learning_rate": 2.841261239068851e-06, + "loss": 0.6422, + "step": 4346 + }, + { + "epoch": 1.61, + "learning_rate": 2.8412242887055056e-06, + "loss": 0.6328, + "step": 4347 + }, + { + "epoch": 1.61, + "learning_rate": 2.8411873383421603e-06, + "loss": 0.6091, + "step": 4348 + }, + { + "epoch": 1.61, + "learning_rate": 2.8411503879788153e-06, + "loss": 0.6264, + "step": 4349 + }, + { + "epoch": 1.61, + "learning_rate": 2.84111343761547e-06, + "loss": 0.6122, + "step": 4350 + }, + { + "epoch": 1.61, + "learning_rate": 2.841076487252125e-06, + "loss": 0.6447, + "step": 4351 + }, + { + "epoch": 1.61, + "learning_rate": 2.8410395368887796e-06, + "loss": 0.6363, + "step": 4352 + }, + { + "epoch": 1.61, + "learning_rate": 2.8410025865254343e-06, + "loss": 0.5797, + "step": 4353 + }, + { + "epoch": 1.61, + "learning_rate": 2.840965636162089e-06, + "loss": 0.747, + "step": 4354 + }, + { + "epoch": 1.61, + "learning_rate": 2.8409286857987436e-06, + "loss": 0.6728, + "step": 4355 + }, + { + "epoch": 1.61, + "learning_rate": 2.8408917354353986e-06, + "loss": 0.6508, + "step": 4356 + }, + { + "epoch": 1.61, + "learning_rate": 2.8408547850720533e-06, + "loss": 0.6542, + "step": 4357 + }, + { + "epoch": 1.61, + "learning_rate": 2.8408178347087083e-06, + "loss": 0.671, + "step": 4358 + }, + { + "epoch": 1.61, + "learning_rate": 2.840780884345363e-06, + "loss": 0.5851, + "step": 4359 + }, + { + "epoch": 1.61, + "learning_rate": 2.8407439339820176e-06, + "loss": 0.569, + "step": 4360 + }, + { + "epoch": 1.61, + "learning_rate": 2.8407069836186722e-06, + "loss": 0.6682, + "step": 4361 + }, + { + "epoch": 1.61, + "learning_rate": 2.840670033255327e-06, + "loss": 0.6231, + "step": 4362 + }, + { + "epoch": 1.61, + "learning_rate": 2.840633082891982e-06, + "loss": 0.6642, + "step": 4363 + }, + { + "epoch": 1.61, + "learning_rate": 2.8405961325286366e-06, + "loss": 0.5942, + "step": 4364 + }, + { + "epoch": 1.61, + "learning_rate": 2.8405591821652916e-06, + "loss": 0.5956, + "step": 4365 + }, + { + "epoch": 1.61, + "learning_rate": 2.840522231801946e-06, + "loss": 0.5361, + "step": 4366 + }, + { + "epoch": 1.61, + "learning_rate": 2.840485281438601e-06, + "loss": 0.5809, + "step": 4367 + }, + { + "epoch": 1.61, + "learning_rate": 2.8404483310752555e-06, + "loss": 0.5972, + "step": 4368 + }, + { + "epoch": 1.61, + "learning_rate": 2.8404113807119106e-06, + "loss": 0.612, + "step": 4369 + }, + { + "epoch": 1.61, + "learning_rate": 2.8403744303485652e-06, + "loss": 0.6355, + "step": 4370 + }, + { + "epoch": 1.61, + "learning_rate": 2.8403374799852203e-06, + "loss": 0.6047, + "step": 4371 + }, + { + "epoch": 1.61, + "learning_rate": 2.8403005296218745e-06, + "loss": 0.5539, + "step": 4372 + }, + { + "epoch": 1.61, + "learning_rate": 2.840263579258529e-06, + "loss": 0.5295, + "step": 4373 + }, + { + "epoch": 1.62, + "learning_rate": 2.840226628895184e-06, + "loss": 0.5846, + "step": 4374 + }, + { + "epoch": 1.62, + "learning_rate": 2.840189678531839e-06, + "loss": 0.6926, + "step": 4375 + }, + { + "epoch": 1.62, + "learning_rate": 2.840152728168494e-06, + "loss": 0.6359, + "step": 4376 + }, + { + "epoch": 1.62, + "learning_rate": 2.8401157778051485e-06, + "loss": 0.6289, + "step": 4377 + }, + { + "epoch": 1.62, + "learning_rate": 2.840078827441803e-06, + "loss": 0.6043, + "step": 4378 + }, + { + "epoch": 1.62, + "learning_rate": 2.840041877078458e-06, + "loss": 0.6359, + "step": 4379 + }, + { + "epoch": 1.62, + "learning_rate": 2.840004926715113e-06, + "loss": 0.6405, + "step": 4380 + }, + { + "epoch": 1.62, + "learning_rate": 2.8399679763517675e-06, + "loss": 0.6626, + "step": 4381 + }, + { + "epoch": 1.62, + "learning_rate": 2.839931025988422e-06, + "loss": 0.628, + "step": 4382 + }, + { + "epoch": 1.62, + "learning_rate": 2.839894075625077e-06, + "loss": 0.6081, + "step": 4383 + }, + { + "epoch": 1.62, + "learning_rate": 2.839857125261732e-06, + "loss": 0.5393, + "step": 4384 + }, + { + "epoch": 1.62, + "learning_rate": 2.8398201748983865e-06, + "loss": 0.6369, + "step": 4385 + }, + { + "epoch": 1.62, + "learning_rate": 2.839783224535041e-06, + "loss": 0.6045, + "step": 4386 + }, + { + "epoch": 1.62, + "learning_rate": 2.839746274171696e-06, + "loss": 0.6527, + "step": 4387 + }, + { + "epoch": 1.62, + "learning_rate": 2.839709323808351e-06, + "loss": 0.6283, + "step": 4388 + }, + { + "epoch": 1.62, + "learning_rate": 2.839672373445006e-06, + "loss": 0.6918, + "step": 4389 + }, + { + "epoch": 1.62, + "learning_rate": 2.8396354230816605e-06, + "loss": 0.6028, + "step": 4390 + }, + { + "epoch": 1.62, + "learning_rate": 2.839598472718315e-06, + "loss": 0.6932, + "step": 4391 + }, + { + "epoch": 1.62, + "learning_rate": 2.8395615223549698e-06, + "loss": 0.6609, + "step": 4392 + }, + { + "epoch": 1.62, + "learning_rate": 2.8395245719916244e-06, + "loss": 0.5924, + "step": 4393 + }, + { + "epoch": 1.62, + "learning_rate": 2.8394876216282795e-06, + "loss": 0.5835, + "step": 4394 + }, + { + "epoch": 1.62, + "learning_rate": 2.839450671264934e-06, + "loss": 0.606, + "step": 4395 + }, + { + "epoch": 1.62, + "learning_rate": 2.839413720901589e-06, + "loss": 0.6142, + "step": 4396 + }, + { + "epoch": 1.62, + "learning_rate": 2.839376770538244e-06, + "loss": 0.6027, + "step": 4397 + }, + { + "epoch": 1.62, + "learning_rate": 2.8393398201748984e-06, + "loss": 0.6288, + "step": 4398 + }, + { + "epoch": 1.62, + "learning_rate": 2.839302869811553e-06, + "loss": 0.5656, + "step": 4399 + }, + { + "epoch": 1.62, + "learning_rate": 2.839265919448208e-06, + "loss": 0.6092, + "step": 4400 + }, + { + "epoch": 1.63, + "learning_rate": 2.8392289690848628e-06, + "loss": 0.6986, + "step": 4401 + }, + { + "epoch": 1.63, + "learning_rate": 2.8391920187215174e-06, + "loss": 0.7583, + "step": 4402 + }, + { + "epoch": 1.63, + "learning_rate": 2.8391550683581725e-06, + "loss": 0.7096, + "step": 4403 + }, + { + "epoch": 1.63, + "learning_rate": 2.839118117994827e-06, + "loss": 0.6013, + "step": 4404 + }, + { + "epoch": 1.63, + "learning_rate": 2.8390811676314817e-06, + "loss": 0.602, + "step": 4405 + }, + { + "epoch": 1.63, + "learning_rate": 2.8390442172681364e-06, + "loss": 0.5782, + "step": 4406 + }, + { + "epoch": 1.63, + "learning_rate": 2.8390072669047914e-06, + "loss": 0.6599, + "step": 4407 + }, + { + "epoch": 1.63, + "learning_rate": 2.838970316541446e-06, + "loss": 0.6481, + "step": 4408 + }, + { + "epoch": 1.63, + "learning_rate": 2.838933366178101e-06, + "loss": 0.6588, + "step": 4409 + }, + { + "epoch": 1.63, + "learning_rate": 2.8388964158147558e-06, + "loss": 0.6304, + "step": 4410 + }, + { + "epoch": 1.63, + "learning_rate": 2.8388594654514104e-06, + "loss": 0.6148, + "step": 4411 + }, + { + "epoch": 1.63, + "learning_rate": 2.838822515088065e-06, + "loss": 0.5991, + "step": 4412 + }, + { + "epoch": 1.63, + "learning_rate": 2.8387855647247197e-06, + "loss": 0.6195, + "step": 4413 + }, + { + "epoch": 1.63, + "learning_rate": 2.8387486143613747e-06, + "loss": 0.5696, + "step": 4414 + }, + { + "epoch": 1.63, + "learning_rate": 2.8387116639980294e-06, + "loss": 0.5901, + "step": 4415 + }, + { + "epoch": 1.63, + "learning_rate": 2.8386747136346844e-06, + "loss": 0.6771, + "step": 4416 + }, + { + "epoch": 1.63, + "learning_rate": 2.8386377632713386e-06, + "loss": 0.5881, + "step": 4417 + }, + { + "epoch": 1.63, + "learning_rate": 2.8386008129079937e-06, + "loss": 0.7036, + "step": 4418 + }, + { + "epoch": 1.63, + "learning_rate": 2.8385638625446483e-06, + "loss": 0.5867, + "step": 4419 + }, + { + "epoch": 1.63, + "learning_rate": 2.8385269121813034e-06, + "loss": 0.6539, + "step": 4420 + }, + { + "epoch": 1.63, + "learning_rate": 2.838489961817958e-06, + "loss": 0.6314, + "step": 4421 + }, + { + "epoch": 1.63, + "learning_rate": 2.8384530114546127e-06, + "loss": 0.6507, + "step": 4422 + }, + { + "epoch": 1.63, + "learning_rate": 2.8384160610912673e-06, + "loss": 0.5394, + "step": 4423 + }, + { + "epoch": 1.63, + "learning_rate": 2.838379110727922e-06, + "loss": 0.5514, + "step": 4424 + }, + { + "epoch": 1.63, + "learning_rate": 2.838342160364577e-06, + "loss": 0.5732, + "step": 4425 + }, + { + "epoch": 1.63, + "learning_rate": 2.8383052100012316e-06, + "loss": 0.4812, + "step": 4426 + }, + { + "epoch": 1.63, + "learning_rate": 2.8382682596378867e-06, + "loss": 0.6428, + "step": 4427 + }, + { + "epoch": 1.64, + "learning_rate": 2.8382313092745413e-06, + "loss": 0.5997, + "step": 4428 + }, + { + "epoch": 1.64, + "learning_rate": 2.838194358911196e-06, + "loss": 0.5726, + "step": 4429 + }, + { + "epoch": 1.64, + "learning_rate": 2.8381574085478506e-06, + "loss": 0.6081, + "step": 4430 + }, + { + "epoch": 1.64, + "learning_rate": 2.8381204581845057e-06, + "loss": 0.557, + "step": 4431 + }, + { + "epoch": 1.64, + "learning_rate": 2.8380835078211603e-06, + "loss": 0.6186, + "step": 4432 + }, + { + "epoch": 1.64, + "learning_rate": 2.838046557457815e-06, + "loss": 0.6473, + "step": 4433 + }, + { + "epoch": 1.64, + "learning_rate": 2.83800960709447e-06, + "loss": 0.6272, + "step": 4434 + }, + { + "epoch": 1.64, + "learning_rate": 2.8379726567311246e-06, + "loss": 0.5923, + "step": 4435 + }, + { + "epoch": 1.64, + "learning_rate": 2.8379357063677793e-06, + "loss": 0.6901, + "step": 4436 + }, + { + "epoch": 1.64, + "learning_rate": 2.837898756004434e-06, + "loss": 0.6774, + "step": 4437 + }, + { + "epoch": 1.64, + "learning_rate": 2.837861805641089e-06, + "loss": 0.5917, + "step": 4438 + }, + { + "epoch": 1.64, + "learning_rate": 2.8378248552777436e-06, + "loss": 0.6211, + "step": 4439 + }, + { + "epoch": 1.64, + "learning_rate": 2.8377879049143987e-06, + "loss": 0.5894, + "step": 4440 + }, + { + "epoch": 1.64, + "learning_rate": 2.8377509545510533e-06, + "loss": 0.5254, + "step": 4441 + }, + { + "epoch": 1.64, + "learning_rate": 2.837714004187708e-06, + "loss": 0.5824, + "step": 4442 + }, + { + "epoch": 1.64, + "learning_rate": 2.8376770538243626e-06, + "loss": 0.6849, + "step": 4443 + }, + { + "epoch": 1.64, + "learning_rate": 2.837640103461017e-06, + "loss": 0.629, + "step": 4444 + }, + { + "epoch": 1.64, + "learning_rate": 2.8376031530976723e-06, + "loss": 0.6096, + "step": 4445 + }, + { + "epoch": 1.64, + "learning_rate": 2.837566202734327e-06, + "loss": 0.6431, + "step": 4446 + }, + { + "epoch": 1.64, + "learning_rate": 2.837529252370982e-06, + "loss": 0.5868, + "step": 4447 + }, + { + "epoch": 1.64, + "learning_rate": 2.8374923020076366e-06, + "loss": 0.5792, + "step": 4448 + }, + { + "epoch": 1.64, + "learning_rate": 2.8374553516442912e-06, + "loss": 0.6683, + "step": 4449 + }, + { + "epoch": 1.64, + "learning_rate": 2.837418401280946e-06, + "loss": 0.628, + "step": 4450 + }, + { + "epoch": 1.64, + "learning_rate": 2.837381450917601e-06, + "loss": 0.6094, + "step": 4451 + }, + { + "epoch": 1.64, + "learning_rate": 2.8373445005542556e-06, + "loss": 0.5893, + "step": 4452 + }, + { + "epoch": 1.64, + "learning_rate": 2.83730755019091e-06, + "loss": 0.5893, + "step": 4453 + }, + { + "epoch": 1.64, + "learning_rate": 2.8372705998275653e-06, + "loss": 0.6071, + "step": 4454 + }, + { + "epoch": 1.65, + "learning_rate": 2.83723364946422e-06, + "loss": 0.5554, + "step": 4455 + }, + { + "epoch": 1.65, + "learning_rate": 2.8371966991008745e-06, + "loss": 0.6385, + "step": 4456 + }, + { + "epoch": 1.65, + "learning_rate": 2.837159748737529e-06, + "loss": 0.6268, + "step": 4457 + }, + { + "epoch": 1.65, + "learning_rate": 2.8371227983741842e-06, + "loss": 0.5215, + "step": 4458 + }, + { + "epoch": 1.65, + "learning_rate": 2.837085848010839e-06, + "loss": 0.5094, + "step": 4459 + }, + { + "epoch": 1.65, + "learning_rate": 2.837048897647494e-06, + "loss": 0.5899, + "step": 4460 + }, + { + "epoch": 1.65, + "learning_rate": 2.8370119472841486e-06, + "loss": 0.514, + "step": 4461 + }, + { + "epoch": 1.65, + "learning_rate": 2.8369749969208028e-06, + "loss": 0.6216, + "step": 4462 + }, + { + "epoch": 1.65, + "learning_rate": 2.836938046557458e-06, + "loss": 0.5979, + "step": 4463 + }, + { + "epoch": 1.65, + "learning_rate": 2.8369010961941125e-06, + "loss": 0.7243, + "step": 4464 + }, + { + "epoch": 1.65, + "learning_rate": 2.8368641458307675e-06, + "loss": 0.5316, + "step": 4465 + }, + { + "epoch": 1.65, + "learning_rate": 2.836827195467422e-06, + "loss": 0.6373, + "step": 4466 + }, + { + "epoch": 1.65, + "learning_rate": 2.8367902451040772e-06, + "loss": 0.6353, + "step": 4467 + }, + { + "epoch": 1.65, + "learning_rate": 2.8367532947407314e-06, + "loss": 0.5728, + "step": 4468 + }, + { + "epoch": 1.65, + "learning_rate": 2.8367163443773865e-06, + "loss": 0.574, + "step": 4469 + }, + { + "epoch": 1.65, + "learning_rate": 2.836679394014041e-06, + "loss": 0.5546, + "step": 4470 + }, + { + "epoch": 1.65, + "learning_rate": 2.8366424436506958e-06, + "loss": 0.5895, + "step": 4471 + }, + { + "epoch": 1.65, + "learning_rate": 2.836605493287351e-06, + "loss": 0.5537, + "step": 4472 + }, + { + "epoch": 1.65, + "learning_rate": 2.8365685429240055e-06, + "loss": 0.6244, + "step": 4473 + }, + { + "epoch": 1.65, + "learning_rate": 2.83653159256066e-06, + "loss": 0.6178, + "step": 4474 + }, + { + "epoch": 1.65, + "learning_rate": 2.8364946421973148e-06, + "loss": 0.5408, + "step": 4475 + }, + { + "epoch": 1.65, + "learning_rate": 2.83645769183397e-06, + "loss": 0.6437, + "step": 4476 + }, + { + "epoch": 1.65, + "learning_rate": 2.8364207414706244e-06, + "loss": 0.6936, + "step": 4477 + }, + { + "epoch": 1.65, + "learning_rate": 2.8363837911072795e-06, + "loss": 0.6752, + "step": 4478 + }, + { + "epoch": 1.65, + "learning_rate": 2.836346840743934e-06, + "loss": 0.5932, + "step": 4479 + }, + { + "epoch": 1.65, + "learning_rate": 2.8363098903805888e-06, + "loss": 0.5621, + "step": 4480 + }, + { + "epoch": 1.65, + "learning_rate": 2.8362729400172434e-06, + "loss": 0.6413, + "step": 4481 + }, + { + "epoch": 1.66, + "learning_rate": 2.836235989653898e-06, + "loss": 0.5905, + "step": 4482 + }, + { + "epoch": 1.66, + "learning_rate": 2.836199039290553e-06, + "loss": 0.6008, + "step": 4483 + }, + { + "epoch": 1.66, + "learning_rate": 2.8361620889272077e-06, + "loss": 0.5896, + "step": 4484 + }, + { + "epoch": 1.66, + "learning_rate": 2.836125138563863e-06, + "loss": 0.5776, + "step": 4485 + }, + { + "epoch": 1.66, + "learning_rate": 2.8360881882005174e-06, + "loss": 0.5956, + "step": 4486 + }, + { + "epoch": 1.66, + "learning_rate": 2.836051237837172e-06, + "loss": 0.5681, + "step": 4487 + }, + { + "epoch": 1.66, + "learning_rate": 2.8360142874738267e-06, + "loss": 0.6123, + "step": 4488 + }, + { + "epoch": 1.66, + "learning_rate": 2.8359773371104818e-06, + "loss": 0.5886, + "step": 4489 + }, + { + "epoch": 1.66, + "learning_rate": 2.8359403867471364e-06, + "loss": 0.582, + "step": 4490 + }, + { + "epoch": 1.66, + "learning_rate": 2.835903436383791e-06, + "loss": 0.7072, + "step": 4491 + }, + { + "epoch": 1.66, + "learning_rate": 2.835866486020446e-06, + "loss": 0.683, + "step": 4492 + }, + { + "epoch": 1.66, + "learning_rate": 2.8358295356571007e-06, + "loss": 0.636, + "step": 4493 + }, + { + "epoch": 1.66, + "learning_rate": 2.8357925852937554e-06, + "loss": 0.5712, + "step": 4494 + }, + { + "epoch": 1.66, + "learning_rate": 2.83575563493041e-06, + "loss": 0.678, + "step": 4495 + }, + { + "epoch": 1.66, + "learning_rate": 2.835718684567065e-06, + "loss": 0.6215, + "step": 4496 + }, + { + "epoch": 1.66, + "learning_rate": 2.8356817342037197e-06, + "loss": 0.6208, + "step": 4497 + }, + { + "epoch": 1.66, + "learning_rate": 2.8356447838403748e-06, + "loss": 0.6087, + "step": 4498 + }, + { + "epoch": 1.66, + "learning_rate": 2.8356078334770294e-06, + "loss": 0.5212, + "step": 4499 + }, + { + "epoch": 1.66, + "learning_rate": 2.835570883113684e-06, + "loss": 0.5589, + "step": 4500 + }, + { + "epoch": 1.66, + "learning_rate": 2.8355339327503387e-06, + "loss": 0.5712, + "step": 4501 + }, + { + "epoch": 1.66, + "learning_rate": 2.8354969823869933e-06, + "loss": 0.6224, + "step": 4502 + }, + { + "epoch": 1.66, + "learning_rate": 2.8354600320236484e-06, + "loss": 0.6307, + "step": 4503 + }, + { + "epoch": 1.66, + "learning_rate": 2.835423081660303e-06, + "loss": 0.6113, + "step": 4504 + }, + { + "epoch": 1.66, + "learning_rate": 2.835386131296958e-06, + "loss": 0.5899, + "step": 4505 + }, + { + "epoch": 1.66, + "learning_rate": 2.8353491809336127e-06, + "loss": 0.5809, + "step": 4506 + }, + { + "epoch": 1.66, + "learning_rate": 2.8353122305702674e-06, + "loss": 0.6244, + "step": 4507 + }, + { + "epoch": 1.66, + "learning_rate": 2.835275280206922e-06, + "loss": 0.6434, + "step": 4508 + }, + { + "epoch": 1.67, + "learning_rate": 2.835238329843577e-06, + "loss": 0.6248, + "step": 4509 + }, + { + "epoch": 1.67, + "learning_rate": 2.8352013794802317e-06, + "loss": 0.5713, + "step": 4510 + }, + { + "epoch": 1.67, + "learning_rate": 2.8351644291168863e-06, + "loss": 0.5737, + "step": 4511 + }, + { + "epoch": 1.67, + "learning_rate": 2.835127478753541e-06, + "loss": 0.5911, + "step": 4512 + }, + { + "epoch": 1.67, + "learning_rate": 2.8350905283901956e-06, + "loss": 0.5855, + "step": 4513 + }, + { + "epoch": 1.67, + "learning_rate": 2.8350535780268507e-06, + "loss": 0.5994, + "step": 4514 + }, + { + "epoch": 1.67, + "learning_rate": 2.8350166276635053e-06, + "loss": 0.6206, + "step": 4515 + }, + { + "epoch": 1.67, + "learning_rate": 2.8349796773001604e-06, + "loss": 0.7298, + "step": 4516 + }, + { + "epoch": 1.67, + "learning_rate": 2.834942726936815e-06, + "loss": 0.6309, + "step": 4517 + }, + { + "epoch": 1.67, + "learning_rate": 2.8349057765734696e-06, + "loss": 0.6804, + "step": 4518 + }, + { + "epoch": 1.67, + "learning_rate": 2.8348688262101243e-06, + "loss": 0.6515, + "step": 4519 + }, + { + "epoch": 1.67, + "learning_rate": 2.8348318758467793e-06, + "loss": 0.5671, + "step": 4520 + }, + { + "epoch": 1.67, + "learning_rate": 2.834794925483434e-06, + "loss": 0.6273, + "step": 4521 + }, + { + "epoch": 1.67, + "learning_rate": 2.8347579751200886e-06, + "loss": 0.5985, + "step": 4522 + }, + { + "epoch": 1.67, + "learning_rate": 2.8347210247567437e-06, + "loss": 0.5754, + "step": 4523 + }, + { + "epoch": 1.67, + "learning_rate": 2.8346840743933983e-06, + "loss": 0.6271, + "step": 4524 + }, + { + "epoch": 1.67, + "learning_rate": 2.834647124030053e-06, + "loss": 0.5548, + "step": 4525 + }, + { + "epoch": 1.67, + "learning_rate": 2.8346101736667076e-06, + "loss": 0.6251, + "step": 4526 + }, + { + "epoch": 1.67, + "learning_rate": 2.8345732233033626e-06, + "loss": 0.652, + "step": 4527 + }, + { + "epoch": 1.67, + "learning_rate": 2.8345362729400173e-06, + "loss": 0.587, + "step": 4528 + }, + { + "epoch": 1.67, + "learning_rate": 2.8344993225766723e-06, + "loss": 0.6307, + "step": 4529 + }, + { + "epoch": 1.67, + "learning_rate": 2.834462372213327e-06, + "loss": 0.6924, + "step": 4530 + }, + { + "epoch": 1.67, + "learning_rate": 2.8344254218499816e-06, + "loss": 0.5855, + "step": 4531 + }, + { + "epoch": 1.67, + "learning_rate": 2.8343884714866362e-06, + "loss": 0.5465, + "step": 4532 + }, + { + "epoch": 1.67, + "learning_rate": 2.834351521123291e-06, + "loss": 0.6269, + "step": 4533 + }, + { + "epoch": 1.67, + "learning_rate": 2.834314570759946e-06, + "loss": 0.6045, + "step": 4534 + }, + { + "epoch": 1.67, + "learning_rate": 2.8342776203966006e-06, + "loss": 0.6195, + "step": 4535 + }, + { + "epoch": 1.68, + "learning_rate": 2.8342406700332556e-06, + "loss": 0.6119, + "step": 4536 + }, + { + "epoch": 1.68, + "learning_rate": 2.8342037196699103e-06, + "loss": 0.5929, + "step": 4537 + }, + { + "epoch": 1.68, + "learning_rate": 2.834166769306565e-06, + "loss": 0.7326, + "step": 4538 + }, + { + "epoch": 1.68, + "learning_rate": 2.8341298189432195e-06, + "loss": 0.5574, + "step": 4539 + }, + { + "epoch": 1.68, + "learning_rate": 2.8340928685798746e-06, + "loss": 0.6093, + "step": 4540 + }, + { + "epoch": 1.68, + "learning_rate": 2.8340559182165292e-06, + "loss": 0.5644, + "step": 4541 + }, + { + "epoch": 1.68, + "learning_rate": 2.834018967853184e-06, + "loss": 0.6679, + "step": 4542 + }, + { + "epoch": 1.68, + "learning_rate": 2.833982017489839e-06, + "loss": 0.647, + "step": 4543 + }, + { + "epoch": 1.68, + "learning_rate": 2.8339450671264936e-06, + "loss": 0.5767, + "step": 4544 + }, + { + "epoch": 1.68, + "learning_rate": 2.833908116763148e-06, + "loss": 0.6822, + "step": 4545 + }, + { + "epoch": 1.68, + "learning_rate": 2.833871166399803e-06, + "loss": 0.5828, + "step": 4546 + }, + { + "epoch": 1.68, + "learning_rate": 2.833834216036458e-06, + "loss": 0.5808, + "step": 4547 + }, + { + "epoch": 1.68, + "learning_rate": 2.8337972656731125e-06, + "loss": 0.6762, + "step": 4548 + }, + { + "epoch": 1.68, + "learning_rate": 2.8337603153097676e-06, + "loss": 0.5853, + "step": 4549 + }, + { + "epoch": 1.68, + "learning_rate": 2.8337233649464222e-06, + "loss": 0.6596, + "step": 4550 + }, + { + "epoch": 1.68, + "learning_rate": 2.8336864145830764e-06, + "loss": 0.7053, + "step": 4551 + }, + { + "epoch": 1.68, + "learning_rate": 2.8336494642197315e-06, + "loss": 0.6289, + "step": 4552 + }, + { + "epoch": 1.68, + "learning_rate": 2.833612513856386e-06, + "loss": 0.5929, + "step": 4553 + }, + { + "epoch": 1.68, + "learning_rate": 2.833575563493041e-06, + "loss": 0.5925, + "step": 4554 + }, + { + "epoch": 1.68, + "learning_rate": 2.833538613129696e-06, + "loss": 0.6337, + "step": 4555 + }, + { + "epoch": 1.68, + "learning_rate": 2.833501662766351e-06, + "loss": 0.6472, + "step": 4556 + }, + { + "epoch": 1.68, + "learning_rate": 2.833464712403005e-06, + "loss": 0.6006, + "step": 4557 + }, + { + "epoch": 1.68, + "learning_rate": 2.83342776203966e-06, + "loss": 0.6093, + "step": 4558 + }, + { + "epoch": 1.68, + "learning_rate": 2.833390811676315e-06, + "loss": 0.5615, + "step": 4559 + }, + { + "epoch": 1.68, + "learning_rate": 2.83335386131297e-06, + "loss": 0.5978, + "step": 4560 + }, + { + "epoch": 1.68, + "learning_rate": 2.8333169109496245e-06, + "loss": 0.6145, + "step": 4561 + }, + { + "epoch": 1.68, + "learning_rate": 2.833279960586279e-06, + "loss": 0.5968, + "step": 4562 + }, + { + "epoch": 1.69, + "learning_rate": 2.8332430102229338e-06, + "loss": 0.641, + "step": 4563 + }, + { + "epoch": 1.69, + "learning_rate": 2.8332060598595884e-06, + "loss": 0.6188, + "step": 4564 + }, + { + "epoch": 1.69, + "learning_rate": 2.8331691094962435e-06, + "loss": 0.6425, + "step": 4565 + }, + { + "epoch": 1.69, + "learning_rate": 2.833132159132898e-06, + "loss": 0.667, + "step": 4566 + }, + { + "epoch": 1.69, + "learning_rate": 2.833095208769553e-06, + "loss": 0.6076, + "step": 4567 + }, + { + "epoch": 1.69, + "learning_rate": 2.833058258406208e-06, + "loss": 0.6626, + "step": 4568 + }, + { + "epoch": 1.69, + "learning_rate": 2.8330213080428624e-06, + "loss": 0.6552, + "step": 4569 + }, + { + "epoch": 1.69, + "learning_rate": 2.832984357679517e-06, + "loss": 0.6812, + "step": 4570 + }, + { + "epoch": 1.69, + "learning_rate": 2.8329474073161717e-06, + "loss": 0.6654, + "step": 4571 + }, + { + "epoch": 1.69, + "learning_rate": 2.8329104569528268e-06, + "loss": 0.564, + "step": 4572 + }, + { + "epoch": 1.69, + "learning_rate": 2.8328735065894814e-06, + "loss": 0.5421, + "step": 4573 + }, + { + "epoch": 1.69, + "learning_rate": 2.8328365562261365e-06, + "loss": 0.6245, + "step": 4574 + }, + { + "epoch": 1.69, + "learning_rate": 2.832799605862791e-06, + "loss": 0.723, + "step": 4575 + }, + { + "epoch": 1.69, + "learning_rate": 2.8327626554994457e-06, + "loss": 0.6121, + "step": 4576 + }, + { + "epoch": 1.69, + "learning_rate": 2.8327257051361004e-06, + "loss": 0.6515, + "step": 4577 + }, + { + "epoch": 1.69, + "learning_rate": 2.8326887547727554e-06, + "loss": 0.5475, + "step": 4578 + }, + { + "epoch": 1.69, + "learning_rate": 2.83265180440941e-06, + "loss": 0.6225, + "step": 4579 + }, + { + "epoch": 1.69, + "learning_rate": 2.832614854046065e-06, + "loss": 0.5676, + "step": 4580 + }, + { + "epoch": 1.69, + "learning_rate": 2.8325779036827198e-06, + "loss": 0.6485, + "step": 4581 + }, + { + "epoch": 1.69, + "learning_rate": 2.8325409533193744e-06, + "loss": 0.6319, + "step": 4582 + }, + { + "epoch": 1.69, + "learning_rate": 2.832504002956029e-06, + "loss": 0.6977, + "step": 4583 + }, + { + "epoch": 1.69, + "learning_rate": 2.8324670525926837e-06, + "loss": 0.6065, + "step": 4584 + }, + { + "epoch": 1.69, + "learning_rate": 2.8324301022293387e-06, + "loss": 0.5522, + "step": 4585 + }, + { + "epoch": 1.69, + "learning_rate": 2.8323931518659934e-06, + "loss": 0.6126, + "step": 4586 + }, + { + "epoch": 1.69, + "learning_rate": 2.8323562015026484e-06, + "loss": 0.5221, + "step": 4587 + }, + { + "epoch": 1.69, + "learning_rate": 2.832319251139303e-06, + "loss": 0.5805, + "step": 4588 + }, + { + "epoch": 1.69, + "learning_rate": 2.8322823007759577e-06, + "loss": 0.5832, + "step": 4589 + }, + { + "epoch": 1.69, + "learning_rate": 2.8322453504126123e-06, + "loss": 0.6188, + "step": 4590 + }, + { + "epoch": 1.7, + "learning_rate": 2.832208400049267e-06, + "loss": 0.5967, + "step": 4591 + }, + { + "epoch": 1.7, + "learning_rate": 2.832171449685922e-06, + "loss": 0.6138, + "step": 4592 + }, + { + "epoch": 1.7, + "learning_rate": 2.8321344993225767e-06, + "loss": 0.5691, + "step": 4593 + }, + { + "epoch": 1.7, + "learning_rate": 2.8320975489592317e-06, + "loss": 0.5906, + "step": 4594 + }, + { + "epoch": 1.7, + "learning_rate": 2.8320605985958864e-06, + "loss": 0.6293, + "step": 4595 + }, + { + "epoch": 1.7, + "learning_rate": 2.832023648232541e-06, + "loss": 0.661, + "step": 4596 + }, + { + "epoch": 1.7, + "learning_rate": 2.8319866978691956e-06, + "loss": 0.5963, + "step": 4597 + }, + { + "epoch": 1.7, + "learning_rate": 2.8319497475058507e-06, + "loss": 0.5437, + "step": 4598 + }, + { + "epoch": 1.7, + "learning_rate": 2.8319127971425053e-06, + "loss": 0.5946, + "step": 4599 + }, + { + "epoch": 1.7, + "learning_rate": 2.8318758467791604e-06, + "loss": 0.6208, + "step": 4600 + }, + { + "epoch": 1.7, + "learning_rate": 2.831838896415815e-06, + "loss": 0.656, + "step": 4601 + }, + { + "epoch": 1.7, + "learning_rate": 2.8318019460524692e-06, + "loss": 0.536, + "step": 4602 + }, + { + "epoch": 1.7, + "learning_rate": 2.8317649956891243e-06, + "loss": 0.6756, + "step": 4603 + }, + { + "epoch": 1.7, + "learning_rate": 2.831728045325779e-06, + "loss": 0.6446, + "step": 4604 + }, + { + "epoch": 1.7, + "learning_rate": 2.831691094962434e-06, + "loss": 0.5843, + "step": 4605 + }, + { + "epoch": 1.7, + "learning_rate": 2.8316541445990886e-06, + "loss": 0.7073, + "step": 4606 + }, + { + "epoch": 1.7, + "learning_rate": 2.8316171942357437e-06, + "loss": 0.586, + "step": 4607 + }, + { + "epoch": 1.7, + "learning_rate": 2.831580243872398e-06, + "loss": 0.6551, + "step": 4608 + }, + { + "epoch": 1.7, + "learning_rate": 2.831543293509053e-06, + "loss": 0.6606, + "step": 4609 + }, + { + "epoch": 1.7, + "learning_rate": 2.8315063431457076e-06, + "loss": 0.5673, + "step": 4610 + }, + { + "epoch": 1.7, + "learning_rate": 2.8314693927823622e-06, + "loss": 0.6042, + "step": 4611 + }, + { + "epoch": 1.7, + "learning_rate": 2.8314324424190173e-06, + "loss": 0.637, + "step": 4612 + }, + { + "epoch": 1.7, + "learning_rate": 2.831395492055672e-06, + "loss": 0.5977, + "step": 4613 + }, + { + "epoch": 1.7, + "learning_rate": 2.8313585416923266e-06, + "loss": 0.6647, + "step": 4614 + }, + { + "epoch": 1.7, + "learning_rate": 2.8313215913289812e-06, + "loss": 0.6799, + "step": 4615 + }, + { + "epoch": 1.7, + "learning_rate": 2.8312846409656363e-06, + "loss": 0.661, + "step": 4616 + }, + { + "epoch": 1.7, + "learning_rate": 2.831247690602291e-06, + "loss": 0.6908, + "step": 4617 + }, + { + "epoch": 1.71, + "learning_rate": 2.831210740238946e-06, + "loss": 0.6145, + "step": 4618 + }, + { + "epoch": 1.71, + "learning_rate": 2.8311737898756006e-06, + "loss": 0.6761, + "step": 4619 + }, + { + "epoch": 1.71, + "learning_rate": 2.8311368395122552e-06, + "loss": 0.6937, + "step": 4620 + }, + { + "epoch": 1.71, + "learning_rate": 2.83109988914891e-06, + "loss": 0.5906, + "step": 4621 + }, + { + "epoch": 1.71, + "learning_rate": 2.8310629387855645e-06, + "loss": 0.52, + "step": 4622 + }, + { + "epoch": 1.71, + "learning_rate": 2.8310259884222196e-06, + "loss": 0.6355, + "step": 4623 + }, + { + "epoch": 1.71, + "learning_rate": 2.8309890380588742e-06, + "loss": 0.6308, + "step": 4624 + }, + { + "epoch": 1.71, + "learning_rate": 2.8309520876955293e-06, + "loss": 0.5371, + "step": 4625 + }, + { + "epoch": 1.71, + "learning_rate": 2.830915137332184e-06, + "loss": 0.596, + "step": 4626 + }, + { + "epoch": 1.71, + "learning_rate": 2.8308781869688385e-06, + "loss": 0.6194, + "step": 4627 + }, + { + "epoch": 1.71, + "learning_rate": 2.830841236605493e-06, + "loss": 0.5903, + "step": 4628 + }, + { + "epoch": 1.71, + "learning_rate": 2.8308042862421482e-06, + "loss": 0.5978, + "step": 4629 + }, + { + "epoch": 1.71, + "learning_rate": 2.830767335878803e-06, + "loss": 0.6383, + "step": 4630 + }, + { + "epoch": 1.71, + "learning_rate": 2.8307303855154575e-06, + "loss": 0.5931, + "step": 4631 + }, + { + "epoch": 1.71, + "learning_rate": 2.8306934351521126e-06, + "loss": 0.5971, + "step": 4632 + }, + { + "epoch": 1.71, + "learning_rate": 2.8306564847887672e-06, + "loss": 0.5904, + "step": 4633 + }, + { + "epoch": 1.71, + "learning_rate": 2.830619534425422e-06, + "loss": 0.5651, + "step": 4634 + }, + { + "epoch": 1.71, + "learning_rate": 2.8305825840620765e-06, + "loss": 0.557, + "step": 4635 + }, + { + "epoch": 1.71, + "learning_rate": 2.8305456336987315e-06, + "loss": 0.7208, + "step": 4636 + }, + { + "epoch": 1.71, + "learning_rate": 2.830508683335386e-06, + "loss": 0.6509, + "step": 4637 + }, + { + "epoch": 1.71, + "learning_rate": 2.8304717329720412e-06, + "loss": 0.5971, + "step": 4638 + }, + { + "epoch": 1.71, + "learning_rate": 2.830434782608696e-06, + "loss": 0.5806, + "step": 4639 + }, + { + "epoch": 1.71, + "learning_rate": 2.8303978322453505e-06, + "loss": 0.6173, + "step": 4640 + }, + { + "epoch": 1.71, + "learning_rate": 2.830360881882005e-06, + "loss": 0.7246, + "step": 4641 + }, + { + "epoch": 1.71, + "learning_rate": 2.8303239315186598e-06, + "loss": 0.6624, + "step": 4642 + }, + { + "epoch": 1.71, + "learning_rate": 2.830286981155315e-06, + "loss": 0.5634, + "step": 4643 + }, + { + "epoch": 1.71, + "learning_rate": 2.8302500307919695e-06, + "loss": 0.6156, + "step": 4644 + }, + { + "epoch": 1.72, + "learning_rate": 2.8302130804286245e-06, + "loss": 0.7008, + "step": 4645 + }, + { + "epoch": 1.72, + "learning_rate": 2.830176130065279e-06, + "loss": 0.5625, + "step": 4646 + }, + { + "epoch": 1.72, + "learning_rate": 2.830139179701934e-06, + "loss": 0.5991, + "step": 4647 + }, + { + "epoch": 1.72, + "learning_rate": 2.8301022293385885e-06, + "loss": 0.6319, + "step": 4648 + }, + { + "epoch": 1.72, + "learning_rate": 2.8300652789752435e-06, + "loss": 0.6145, + "step": 4649 + }, + { + "epoch": 1.72, + "learning_rate": 2.830028328611898e-06, + "loss": 0.6855, + "step": 4650 + }, + { + "epoch": 1.72, + "learning_rate": 2.8299913782485528e-06, + "loss": 0.55, + "step": 4651 + }, + { + "epoch": 1.72, + "learning_rate": 2.829954427885208e-06, + "loss": 0.6331, + "step": 4652 + }, + { + "epoch": 1.72, + "learning_rate": 2.829917477521862e-06, + "loss": 0.6308, + "step": 4653 + }, + { + "epoch": 1.72, + "learning_rate": 2.829880527158517e-06, + "loss": 0.6013, + "step": 4654 + }, + { + "epoch": 1.72, + "learning_rate": 2.8298435767951718e-06, + "loss": 0.5998, + "step": 4655 + }, + { + "epoch": 1.72, + "learning_rate": 2.829806626431827e-06, + "loss": 0.632, + "step": 4656 + }, + { + "epoch": 1.72, + "learning_rate": 2.8297696760684815e-06, + "loss": 0.5499, + "step": 4657 + }, + { + "epoch": 1.72, + "learning_rate": 2.8297327257051365e-06, + "loss": 0.571, + "step": 4658 + }, + { + "epoch": 1.72, + "learning_rate": 2.8296957753417907e-06, + "loss": 0.5966, + "step": 4659 + }, + { + "epoch": 1.72, + "learning_rate": 2.8296588249784458e-06, + "loss": 0.5663, + "step": 4660 + }, + { + "epoch": 1.72, + "learning_rate": 2.8296218746151004e-06, + "loss": 0.7238, + "step": 4661 + }, + { + "epoch": 1.72, + "learning_rate": 2.829584924251755e-06, + "loss": 0.5773, + "step": 4662 + }, + { + "epoch": 1.72, + "learning_rate": 2.82954797388841e-06, + "loss": 0.4866, + "step": 4663 + }, + { + "epoch": 1.72, + "learning_rate": 2.8295110235250648e-06, + "loss": 0.5945, + "step": 4664 + }, + { + "epoch": 1.72, + "learning_rate": 2.8294740731617194e-06, + "loss": 0.6513, + "step": 4665 + }, + { + "epoch": 1.72, + "learning_rate": 2.829437122798374e-06, + "loss": 0.5896, + "step": 4666 + }, + { + "epoch": 1.72, + "learning_rate": 2.829400172435029e-06, + "loss": 0.5781, + "step": 4667 + }, + { + "epoch": 1.72, + "learning_rate": 2.8293632220716837e-06, + "loss": 0.7019, + "step": 4668 + }, + { + "epoch": 1.72, + "learning_rate": 2.8293262717083388e-06, + "loss": 0.6042, + "step": 4669 + }, + { + "epoch": 1.72, + "learning_rate": 2.8292893213449934e-06, + "loss": 0.6648, + "step": 4670 + }, + { + "epoch": 1.72, + "learning_rate": 2.829252370981648e-06, + "loss": 0.6121, + "step": 4671 + }, + { + "epoch": 1.73, + "learning_rate": 2.8292154206183027e-06, + "loss": 0.6727, + "step": 4672 + }, + { + "epoch": 1.73, + "learning_rate": 2.8291784702549573e-06, + "loss": 0.6431, + "step": 4673 + }, + { + "epoch": 1.73, + "learning_rate": 2.8291415198916124e-06, + "loss": 0.6539, + "step": 4674 + }, + { + "epoch": 1.73, + "learning_rate": 2.829104569528267e-06, + "loss": 0.6045, + "step": 4675 + }, + { + "epoch": 1.73, + "learning_rate": 2.829067619164922e-06, + "loss": 0.6174, + "step": 4676 + }, + { + "epoch": 1.73, + "learning_rate": 2.8290306688015767e-06, + "loss": 0.6795, + "step": 4677 + }, + { + "epoch": 1.73, + "learning_rate": 2.8289937184382314e-06, + "loss": 0.5323, + "step": 4678 + }, + { + "epoch": 1.73, + "learning_rate": 2.828956768074886e-06, + "loss": 0.5857, + "step": 4679 + }, + { + "epoch": 1.73, + "learning_rate": 2.828919817711541e-06, + "loss": 0.592, + "step": 4680 + }, + { + "epoch": 1.73, + "learning_rate": 2.8288828673481957e-06, + "loss": 0.6682, + "step": 4681 + }, + { + "epoch": 1.73, + "learning_rate": 2.8288459169848503e-06, + "loss": 0.6581, + "step": 4682 + }, + { + "epoch": 1.73, + "learning_rate": 2.8288089666215054e-06, + "loss": 0.6253, + "step": 4683 + }, + { + "epoch": 1.73, + "learning_rate": 2.82877201625816e-06, + "loss": 0.5897, + "step": 4684 + }, + { + "epoch": 1.73, + "learning_rate": 2.8287350658948147e-06, + "loss": 0.6074, + "step": 4685 + }, + { + "epoch": 1.73, + "learning_rate": 2.8286981155314693e-06, + "loss": 0.6006, + "step": 4686 + }, + { + "epoch": 1.73, + "learning_rate": 2.8286611651681244e-06, + "loss": 0.5512, + "step": 4687 + }, + { + "epoch": 1.73, + "learning_rate": 2.828624214804779e-06, + "loss": 0.6559, + "step": 4688 + }, + { + "epoch": 1.73, + "learning_rate": 2.828587264441434e-06, + "loss": 0.6656, + "step": 4689 + }, + { + "epoch": 1.73, + "learning_rate": 2.8285503140780887e-06, + "loss": 0.6149, + "step": 4690 + }, + { + "epoch": 1.73, + "learning_rate": 2.8285133637147433e-06, + "loss": 0.6243, + "step": 4691 + }, + { + "epoch": 1.73, + "learning_rate": 2.828476413351398e-06, + "loss": 0.6677, + "step": 4692 + }, + { + "epoch": 1.73, + "learning_rate": 2.8284394629880526e-06, + "loss": 0.5966, + "step": 4693 + }, + { + "epoch": 1.73, + "learning_rate": 2.8284025126247077e-06, + "loss": 0.6148, + "step": 4694 + }, + { + "epoch": 1.73, + "learning_rate": 2.8283655622613623e-06, + "loss": 0.7229, + "step": 4695 + }, + { + "epoch": 1.73, + "learning_rate": 2.8283286118980174e-06, + "loss": 0.6371, + "step": 4696 + }, + { + "epoch": 1.73, + "learning_rate": 2.828291661534672e-06, + "loss": 0.5994, + "step": 4697 + }, + { + "epoch": 1.73, + "learning_rate": 2.8282547111713266e-06, + "loss": 0.6404, + "step": 4698 + }, + { + "epoch": 1.74, + "learning_rate": 2.8282177608079813e-06, + "loss": 0.691, + "step": 4699 + }, + { + "epoch": 1.74, + "learning_rate": 2.828180810444636e-06, + "loss": 0.6015, + "step": 4700 + }, + { + "epoch": 1.74, + "learning_rate": 2.828143860081291e-06, + "loss": 0.6019, + "step": 4701 + }, + { + "epoch": 1.74, + "learning_rate": 2.8281069097179456e-06, + "loss": 0.5828, + "step": 4702 + }, + { + "epoch": 1.74, + "learning_rate": 2.8280699593546002e-06, + "loss": 0.6201, + "step": 4703 + }, + { + "epoch": 1.74, + "learning_rate": 2.828033008991255e-06, + "loss": 0.6047, + "step": 4704 + }, + { + "epoch": 1.74, + "learning_rate": 2.82799605862791e-06, + "loss": 0.6317, + "step": 4705 + }, + { + "epoch": 1.74, + "learning_rate": 2.8279591082645646e-06, + "loss": 0.5433, + "step": 4706 + }, + { + "epoch": 1.74, + "learning_rate": 2.8279221579012196e-06, + "loss": 0.6676, + "step": 4707 + }, + { + "epoch": 1.74, + "learning_rate": 2.8278852075378743e-06, + "loss": 0.6353, + "step": 4708 + }, + { + "epoch": 1.74, + "learning_rate": 2.827848257174529e-06, + "loss": 0.6202, + "step": 4709 + }, + { + "epoch": 1.74, + "learning_rate": 2.8278113068111835e-06, + "loss": 0.6722, + "step": 4710 + }, + { + "epoch": 1.74, + "learning_rate": 2.827774356447838e-06, + "loss": 0.6131, + "step": 4711 + }, + { + "epoch": 1.74, + "learning_rate": 2.8277374060844932e-06, + "loss": 0.5813, + "step": 4712 + }, + { + "epoch": 1.74, + "learning_rate": 2.827700455721148e-06, + "loss": 0.581, + "step": 4713 + }, + { + "epoch": 1.74, + "learning_rate": 2.827663505357803e-06, + "loss": 0.7, + "step": 4714 + }, + { + "epoch": 1.74, + "learning_rate": 2.8276265549944576e-06, + "loss": 0.649, + "step": 4715 + }, + { + "epoch": 1.74, + "learning_rate": 2.827589604631112e-06, + "loss": 0.5531, + "step": 4716 + }, + { + "epoch": 1.74, + "learning_rate": 2.827552654267767e-06, + "loss": 0.6717, + "step": 4717 + }, + { + "epoch": 1.74, + "learning_rate": 2.827515703904422e-06, + "loss": 0.5638, + "step": 4718 + }, + { + "epoch": 1.74, + "learning_rate": 2.8274787535410765e-06, + "loss": 0.5814, + "step": 4719 + }, + { + "epoch": 1.74, + "learning_rate": 2.827441803177731e-06, + "loss": 0.5642, + "step": 4720 + }, + { + "epoch": 1.74, + "learning_rate": 2.8274048528143862e-06, + "loss": 0.6798, + "step": 4721 + }, + { + "epoch": 1.74, + "learning_rate": 2.827367902451041e-06, + "loss": 0.7265, + "step": 4722 + }, + { + "epoch": 1.74, + "learning_rate": 2.8273309520876955e-06, + "loss": 0.6619, + "step": 4723 + }, + { + "epoch": 1.74, + "learning_rate": 2.82729400172435e-06, + "loss": 0.5445, + "step": 4724 + }, + { + "epoch": 1.74, + "learning_rate": 2.827257051361005e-06, + "loss": 0.582, + "step": 4725 + }, + { + "epoch": 1.75, + "learning_rate": 2.82722010099766e-06, + "loss": 0.6873, + "step": 4726 + }, + { + "epoch": 1.75, + "learning_rate": 2.827183150634315e-06, + "loss": 0.5942, + "step": 4727 + }, + { + "epoch": 1.75, + "learning_rate": 2.8271462002709695e-06, + "loss": 0.5669, + "step": 4728 + }, + { + "epoch": 1.75, + "learning_rate": 2.827109249907624e-06, + "loss": 0.6367, + "step": 4729 + }, + { + "epoch": 1.75, + "learning_rate": 2.827072299544279e-06, + "loss": 0.5776, + "step": 4730 + }, + { + "epoch": 1.75, + "learning_rate": 2.8270353491809334e-06, + "loss": 0.6288, + "step": 4731 + }, + { + "epoch": 1.75, + "learning_rate": 2.8269983988175885e-06, + "loss": 0.6235, + "step": 4732 + }, + { + "epoch": 1.75, + "learning_rate": 2.826961448454243e-06, + "loss": 0.6429, + "step": 4733 + }, + { + "epoch": 1.75, + "learning_rate": 2.826924498090898e-06, + "loss": 0.6377, + "step": 4734 + }, + { + "epoch": 1.75, + "learning_rate": 2.826887547727553e-06, + "loss": 0.6456, + "step": 4735 + }, + { + "epoch": 1.75, + "learning_rate": 2.8268505973642075e-06, + "loss": 0.5821, + "step": 4736 + }, + { + "epoch": 1.75, + "learning_rate": 2.826813647000862e-06, + "loss": 0.6462, + "step": 4737 + }, + { + "epoch": 1.75, + "learning_rate": 2.826776696637517e-06, + "loss": 0.6499, + "step": 4738 + }, + { + "epoch": 1.75, + "learning_rate": 2.826739746274172e-06, + "loss": 0.6015, + "step": 4739 + }, + { + "epoch": 1.75, + "learning_rate": 2.8267027959108264e-06, + "loss": 0.6423, + "step": 4740 + }, + { + "epoch": 1.75, + "learning_rate": 2.8266658455474815e-06, + "loss": 0.6322, + "step": 4741 + }, + { + "epoch": 1.75, + "learning_rate": 2.8266288951841357e-06, + "loss": 0.6638, + "step": 4742 + }, + { + "epoch": 1.75, + "learning_rate": 2.8265919448207908e-06, + "loss": 0.565, + "step": 4743 + }, + { + "epoch": 1.75, + "learning_rate": 2.8265549944574454e-06, + "loss": 0.6522, + "step": 4744 + }, + { + "epoch": 1.75, + "learning_rate": 2.8265180440941005e-06, + "loss": 0.6015, + "step": 4745 + }, + { + "epoch": 1.75, + "learning_rate": 2.826481093730755e-06, + "loss": 0.6889, + "step": 4746 + }, + { + "epoch": 1.75, + "learning_rate": 2.82644414336741e-06, + "loss": 0.5853, + "step": 4747 + }, + { + "epoch": 1.75, + "learning_rate": 2.8264071930040644e-06, + "loss": 0.593, + "step": 4748 + }, + { + "epoch": 1.75, + "learning_rate": 2.8263702426407194e-06, + "loss": 0.688, + "step": 4749 + }, + { + "epoch": 1.75, + "learning_rate": 2.826333292277374e-06, + "loss": 0.611, + "step": 4750 + }, + { + "epoch": 1.75, + "learning_rate": 2.8262963419140287e-06, + "loss": 0.6239, + "step": 4751 + }, + { + "epoch": 1.75, + "learning_rate": 2.8262593915506838e-06, + "loss": 0.6074, + "step": 4752 + }, + { + "epoch": 1.76, + "learning_rate": 2.8262224411873384e-06, + "loss": 0.6049, + "step": 4753 + }, + { + "epoch": 1.76, + "learning_rate": 2.826185490823993e-06, + "loss": 0.5879, + "step": 4754 + }, + { + "epoch": 1.76, + "learning_rate": 2.8261485404606477e-06, + "loss": 0.6528, + "step": 4755 + }, + { + "epoch": 1.76, + "learning_rate": 2.8261115900973027e-06, + "loss": 0.6898, + "step": 4756 + }, + { + "epoch": 1.76, + "learning_rate": 2.8260746397339574e-06, + "loss": 0.6209, + "step": 4757 + }, + { + "epoch": 1.76, + "learning_rate": 2.8260376893706124e-06, + "loss": 0.601, + "step": 4758 + }, + { + "epoch": 1.76, + "learning_rate": 2.826000739007267e-06, + "loss": 0.6596, + "step": 4759 + }, + { + "epoch": 1.76, + "learning_rate": 2.8259637886439217e-06, + "loss": 0.5737, + "step": 4760 + }, + { + "epoch": 1.76, + "learning_rate": 2.8259268382805763e-06, + "loss": 0.6072, + "step": 4761 + }, + { + "epoch": 1.76, + "learning_rate": 2.825889887917231e-06, + "loss": 0.6123, + "step": 4762 + }, + { + "epoch": 1.76, + "learning_rate": 2.825852937553886e-06, + "loss": 0.6018, + "step": 4763 + }, + { + "epoch": 1.76, + "learning_rate": 2.8258159871905407e-06, + "loss": 0.6205, + "step": 4764 + }, + { + "epoch": 1.76, + "learning_rate": 2.8257790368271957e-06, + "loss": 0.5665, + "step": 4765 + }, + { + "epoch": 1.76, + "learning_rate": 2.8257420864638504e-06, + "loss": 0.5952, + "step": 4766 + }, + { + "epoch": 1.76, + "learning_rate": 2.825705136100505e-06, + "loss": 0.7064, + "step": 4767 + }, + { + "epoch": 1.76, + "learning_rate": 2.8256681857371596e-06, + "loss": 0.6935, + "step": 4768 + }, + { + "epoch": 1.76, + "learning_rate": 2.8256312353738147e-06, + "loss": 0.6772, + "step": 4769 + }, + { + "epoch": 1.76, + "learning_rate": 2.8255942850104693e-06, + "loss": 0.6118, + "step": 4770 + }, + { + "epoch": 1.76, + "learning_rate": 2.825557334647124e-06, + "loss": 0.6453, + "step": 4771 + }, + { + "epoch": 1.76, + "learning_rate": 2.825520384283779e-06, + "loss": 0.635, + "step": 4772 + }, + { + "epoch": 1.76, + "learning_rate": 2.8254834339204337e-06, + "loss": 0.5987, + "step": 4773 + }, + { + "epoch": 1.76, + "learning_rate": 2.8254464835570883e-06, + "loss": 0.5494, + "step": 4774 + }, + { + "epoch": 1.76, + "learning_rate": 2.825409533193743e-06, + "loss": 0.6614, + "step": 4775 + }, + { + "epoch": 1.76, + "learning_rate": 2.825372582830398e-06, + "loss": 0.6211, + "step": 4776 + }, + { + "epoch": 1.76, + "learning_rate": 2.8253356324670526e-06, + "loss": 0.6717, + "step": 4777 + }, + { + "epoch": 1.76, + "learning_rate": 2.8252986821037077e-06, + "loss": 0.5791, + "step": 4778 + }, + { + "epoch": 1.76, + "learning_rate": 2.8252617317403623e-06, + "loss": 0.6426, + "step": 4779 + }, + { + "epoch": 1.77, + "learning_rate": 2.825224781377017e-06, + "loss": 0.5525, + "step": 4780 + }, + { + "epoch": 1.77, + "learning_rate": 2.8251878310136716e-06, + "loss": 0.5634, + "step": 4781 + }, + { + "epoch": 1.77, + "learning_rate": 2.8251508806503263e-06, + "loss": 0.6924, + "step": 4782 + }, + { + "epoch": 1.77, + "learning_rate": 2.8251139302869813e-06, + "loss": 0.547, + "step": 4783 + }, + { + "epoch": 1.77, + "learning_rate": 2.825076979923636e-06, + "loss": 0.6859, + "step": 4784 + }, + { + "epoch": 1.77, + "learning_rate": 2.825040029560291e-06, + "loss": 0.6238, + "step": 4785 + }, + { + "epoch": 1.77, + "learning_rate": 2.8250030791969456e-06, + "loss": 0.6285, + "step": 4786 + }, + { + "epoch": 1.77, + "learning_rate": 2.8249661288336003e-06, + "loss": 0.5826, + "step": 4787 + }, + { + "epoch": 1.77, + "learning_rate": 2.824929178470255e-06, + "loss": 0.5599, + "step": 4788 + }, + { + "epoch": 1.77, + "learning_rate": 2.82489222810691e-06, + "loss": 0.6587, + "step": 4789 + }, + { + "epoch": 1.77, + "learning_rate": 2.8248552777435646e-06, + "loss": 0.5459, + "step": 4790 + }, + { + "epoch": 1.77, + "learning_rate": 2.8248183273802193e-06, + "loss": 0.5902, + "step": 4791 + }, + { + "epoch": 1.77, + "learning_rate": 2.8247813770168743e-06, + "loss": 0.6726, + "step": 4792 + }, + { + "epoch": 1.77, + "learning_rate": 2.8247444266535285e-06, + "loss": 0.6612, + "step": 4793 + }, + { + "epoch": 1.77, + "learning_rate": 2.8247074762901836e-06, + "loss": 0.6245, + "step": 4794 + }, + { + "epoch": 1.77, + "learning_rate": 2.8246705259268382e-06, + "loss": 0.5829, + "step": 4795 + }, + { + "epoch": 1.77, + "learning_rate": 2.8246335755634933e-06, + "loss": 0.633, + "step": 4796 + }, + { + "epoch": 1.77, + "learning_rate": 2.824596625200148e-06, + "loss": 0.5904, + "step": 4797 + }, + { + "epoch": 1.77, + "learning_rate": 2.824559674836803e-06, + "loss": 0.6464, + "step": 4798 + }, + { + "epoch": 1.77, + "learning_rate": 2.824522724473457e-06, + "loss": 0.5776, + "step": 4799 + }, + { + "epoch": 1.77, + "learning_rate": 2.824485774110112e-06, + "loss": 0.7112, + "step": 4800 + }, + { + "epoch": 1.77, + "learning_rate": 2.824448823746767e-06, + "loss": 0.6197, + "step": 4801 + }, + { + "epoch": 1.77, + "learning_rate": 2.8244118733834215e-06, + "loss": 0.5575, + "step": 4802 + }, + { + "epoch": 1.77, + "learning_rate": 2.8243749230200766e-06, + "loss": 0.6361, + "step": 4803 + }, + { + "epoch": 1.77, + "learning_rate": 2.8243379726567312e-06, + "loss": 0.6642, + "step": 4804 + }, + { + "epoch": 1.77, + "learning_rate": 2.824301022293386e-06, + "loss": 0.6871, + "step": 4805 + }, + { + "epoch": 1.77, + "learning_rate": 2.8242640719300405e-06, + "loss": 0.6236, + "step": 4806 + }, + { + "epoch": 1.78, + "learning_rate": 2.8242271215666956e-06, + "loss": 0.6632, + "step": 4807 + }, + { + "epoch": 1.78, + "learning_rate": 2.82419017120335e-06, + "loss": 0.6368, + "step": 4808 + }, + { + "epoch": 1.78, + "learning_rate": 2.8241532208400052e-06, + "loss": 0.7171, + "step": 4809 + }, + { + "epoch": 1.78, + "learning_rate": 2.82411627047666e-06, + "loss": 0.6283, + "step": 4810 + }, + { + "epoch": 1.78, + "learning_rate": 2.8240793201133145e-06, + "loss": 0.6598, + "step": 4811 + }, + { + "epoch": 1.78, + "learning_rate": 2.824042369749969e-06, + "loss": 0.6414, + "step": 4812 + }, + { + "epoch": 1.78, + "learning_rate": 2.824005419386624e-06, + "loss": 0.5992, + "step": 4813 + }, + { + "epoch": 1.78, + "learning_rate": 2.823968469023279e-06, + "loss": 0.5823, + "step": 4814 + }, + { + "epoch": 1.78, + "learning_rate": 2.8239315186599335e-06, + "loss": 0.6975, + "step": 4815 + }, + { + "epoch": 1.78, + "learning_rate": 2.8238945682965886e-06, + "loss": 0.5721, + "step": 4816 + }, + { + "epoch": 1.78, + "learning_rate": 2.823857617933243e-06, + "loss": 0.619, + "step": 4817 + }, + { + "epoch": 1.78, + "learning_rate": 2.823820667569898e-06, + "loss": 0.6072, + "step": 4818 + }, + { + "epoch": 1.78, + "learning_rate": 2.8237837172065525e-06, + "loss": 0.6836, + "step": 4819 + }, + { + "epoch": 1.78, + "learning_rate": 2.823746766843207e-06, + "loss": 0.5793, + "step": 4820 + }, + { + "epoch": 1.78, + "learning_rate": 2.823709816479862e-06, + "loss": 0.6206, + "step": 4821 + }, + { + "epoch": 1.78, + "learning_rate": 2.823672866116517e-06, + "loss": 0.5605, + "step": 4822 + }, + { + "epoch": 1.78, + "learning_rate": 2.823635915753172e-06, + "loss": 0.6452, + "step": 4823 + }, + { + "epoch": 1.78, + "learning_rate": 2.8235989653898265e-06, + "loss": 0.6491, + "step": 4824 + }, + { + "epoch": 1.78, + "learning_rate": 2.823562015026481e-06, + "loss": 0.6145, + "step": 4825 + }, + { + "epoch": 1.78, + "learning_rate": 2.8235250646631358e-06, + "loss": 0.6616, + "step": 4826 + }, + { + "epoch": 1.78, + "learning_rate": 2.823488114299791e-06, + "loss": 0.5009, + "step": 4827 + }, + { + "epoch": 1.78, + "learning_rate": 2.8234511639364455e-06, + "loss": 0.6433, + "step": 4828 + }, + { + "epoch": 1.78, + "learning_rate": 2.8234142135731005e-06, + "loss": 0.5899, + "step": 4829 + }, + { + "epoch": 1.78, + "learning_rate": 2.823377263209755e-06, + "loss": 0.7139, + "step": 4830 + }, + { + "epoch": 1.78, + "learning_rate": 2.82334031284641e-06, + "loss": 0.5398, + "step": 4831 + }, + { + "epoch": 1.78, + "learning_rate": 2.8233033624830644e-06, + "loss": 0.5776, + "step": 4832 + }, + { + "epoch": 1.78, + "learning_rate": 2.823266412119719e-06, + "loss": 0.6084, + "step": 4833 + }, + { + "epoch": 1.79, + "learning_rate": 2.823229461756374e-06, + "loss": 0.6289, + "step": 4834 + }, + { + "epoch": 1.79, + "learning_rate": 2.8231925113930288e-06, + "loss": 0.6185, + "step": 4835 + }, + { + "epoch": 1.79, + "learning_rate": 2.823155561029684e-06, + "loss": 0.6066, + "step": 4836 + }, + { + "epoch": 1.79, + "learning_rate": 2.8231186106663385e-06, + "loss": 0.5919, + "step": 4837 + }, + { + "epoch": 1.79, + "learning_rate": 2.823081660302993e-06, + "loss": 0.6171, + "step": 4838 + }, + { + "epoch": 1.79, + "learning_rate": 2.8230447099396477e-06, + "loss": 0.5757, + "step": 4839 + }, + { + "epoch": 1.79, + "learning_rate": 2.8230077595763024e-06, + "loss": 0.5908, + "step": 4840 + }, + { + "epoch": 1.79, + "learning_rate": 2.8229708092129574e-06, + "loss": 0.587, + "step": 4841 + }, + { + "epoch": 1.79, + "learning_rate": 2.822933858849612e-06, + "loss": 0.7037, + "step": 4842 + }, + { + "epoch": 1.79, + "learning_rate": 2.822896908486267e-06, + "loss": 0.5943, + "step": 4843 + }, + { + "epoch": 1.79, + "learning_rate": 2.8228599581229213e-06, + "loss": 0.6249, + "step": 4844 + }, + { + "epoch": 1.79, + "learning_rate": 2.8228230077595764e-06, + "loss": 0.6154, + "step": 4845 + }, + { + "epoch": 1.79, + "learning_rate": 2.822786057396231e-06, + "loss": 0.5569, + "step": 4846 + }, + { + "epoch": 1.79, + "learning_rate": 2.822749107032886e-06, + "loss": 0.6233, + "step": 4847 + }, + { + "epoch": 1.79, + "learning_rate": 2.8227121566695407e-06, + "loss": 0.6025, + "step": 4848 + }, + { + "epoch": 1.79, + "learning_rate": 2.8226752063061958e-06, + "loss": 0.5896, + "step": 4849 + }, + { + "epoch": 1.79, + "learning_rate": 2.82263825594285e-06, + "loss": 0.6836, + "step": 4850 + }, + { + "epoch": 1.79, + "learning_rate": 2.8226013055795046e-06, + "loss": 0.6068, + "step": 4851 + }, + { + "epoch": 1.79, + "learning_rate": 2.8225643552161597e-06, + "loss": 0.633, + "step": 4852 + }, + { + "epoch": 1.79, + "learning_rate": 2.8225274048528143e-06, + "loss": 0.621, + "step": 4853 + }, + { + "epoch": 1.79, + "learning_rate": 2.8224904544894694e-06, + "loss": 0.631, + "step": 4854 + }, + { + "epoch": 1.79, + "learning_rate": 2.822453504126124e-06, + "loss": 0.5588, + "step": 4855 + }, + { + "epoch": 1.79, + "learning_rate": 2.8224165537627787e-06, + "loss": 0.6105, + "step": 4856 + }, + { + "epoch": 1.79, + "learning_rate": 2.8223796033994333e-06, + "loss": 0.5633, + "step": 4857 + }, + { + "epoch": 1.79, + "learning_rate": 2.8223426530360884e-06, + "loss": 0.6835, + "step": 4858 + }, + { + "epoch": 1.79, + "learning_rate": 2.822305702672743e-06, + "loss": 0.6182, + "step": 4859 + }, + { + "epoch": 1.79, + "learning_rate": 2.8222687523093976e-06, + "loss": 0.5023, + "step": 4860 + }, + { + "epoch": 1.8, + "learning_rate": 2.8222318019460527e-06, + "loss": 0.6051, + "step": 4861 + }, + { + "epoch": 1.8, + "learning_rate": 2.8221948515827073e-06, + "loss": 0.6873, + "step": 4862 + }, + { + "epoch": 1.8, + "learning_rate": 2.822157901219362e-06, + "loss": 0.6053, + "step": 4863 + }, + { + "epoch": 1.8, + "learning_rate": 2.8221209508560166e-06, + "loss": 0.6118, + "step": 4864 + }, + { + "epoch": 1.8, + "learning_rate": 2.8220840004926717e-06, + "loss": 0.5804, + "step": 4865 + }, + { + "epoch": 1.8, + "learning_rate": 2.8220470501293263e-06, + "loss": 0.6246, + "step": 4866 + }, + { + "epoch": 1.8, + "learning_rate": 2.8220100997659814e-06, + "loss": 0.5137, + "step": 4867 + }, + { + "epoch": 1.8, + "learning_rate": 2.821973149402636e-06, + "loss": 0.642, + "step": 4868 + }, + { + "epoch": 1.8, + "learning_rate": 2.8219361990392906e-06, + "loss": 0.6524, + "step": 4869 + }, + { + "epoch": 1.8, + "learning_rate": 2.8218992486759453e-06, + "loss": 0.6363, + "step": 4870 + }, + { + "epoch": 1.8, + "learning_rate": 2.8218622983126e-06, + "loss": 0.6051, + "step": 4871 + }, + { + "epoch": 1.8, + "learning_rate": 2.821825347949255e-06, + "loss": 0.5693, + "step": 4872 + }, + { + "epoch": 1.8, + "learning_rate": 2.8217883975859096e-06, + "loss": 0.5645, + "step": 4873 + }, + { + "epoch": 1.8, + "learning_rate": 2.8217514472225647e-06, + "loss": 0.6356, + "step": 4874 + }, + { + "epoch": 1.8, + "learning_rate": 2.8217144968592193e-06, + "loss": 0.6628, + "step": 4875 + }, + { + "epoch": 1.8, + "learning_rate": 2.821677546495874e-06, + "loss": 0.6249, + "step": 4876 + }, + { + "epoch": 1.8, + "learning_rate": 2.8216405961325286e-06, + "loss": 0.6571, + "step": 4877 + }, + { + "epoch": 1.8, + "learning_rate": 2.8216036457691836e-06, + "loss": 0.6492, + "step": 4878 + }, + { + "epoch": 1.8, + "learning_rate": 2.8215666954058383e-06, + "loss": 0.5648, + "step": 4879 + }, + { + "epoch": 1.8, + "learning_rate": 2.821529745042493e-06, + "loss": 0.7436, + "step": 4880 + }, + { + "epoch": 1.8, + "learning_rate": 2.821492794679148e-06, + "loss": 0.6213, + "step": 4881 + }, + { + "epoch": 1.8, + "learning_rate": 2.8214558443158026e-06, + "loss": 0.6084, + "step": 4882 + }, + { + "epoch": 1.8, + "learning_rate": 2.8214188939524572e-06, + "loss": 0.5845, + "step": 4883 + }, + { + "epoch": 1.8, + "learning_rate": 2.821381943589112e-06, + "loss": 0.691, + "step": 4884 + }, + { + "epoch": 1.8, + "learning_rate": 2.821344993225767e-06, + "loss": 0.6052, + "step": 4885 + }, + { + "epoch": 1.8, + "learning_rate": 2.8213080428624216e-06, + "loss": 0.6217, + "step": 4886 + }, + { + "epoch": 1.8, + "learning_rate": 2.8212710924990766e-06, + "loss": 0.6364, + "step": 4887 + }, + { + "epoch": 1.81, + "learning_rate": 2.821234142135731e-06, + "loss": 0.5913, + "step": 4888 + }, + { + "epoch": 1.81, + "learning_rate": 2.821197191772386e-06, + "loss": 0.5672, + "step": 4889 + }, + { + "epoch": 1.81, + "learning_rate": 2.8211602414090405e-06, + "loss": 0.6703, + "step": 4890 + }, + { + "epoch": 1.81, + "learning_rate": 2.821123291045695e-06, + "loss": 0.6171, + "step": 4891 + }, + { + "epoch": 1.81, + "learning_rate": 2.8210863406823502e-06, + "loss": 0.5281, + "step": 4892 + }, + { + "epoch": 1.81, + "learning_rate": 2.821049390319005e-06, + "loss": 0.5913, + "step": 4893 + }, + { + "epoch": 1.81, + "learning_rate": 2.8210124399556595e-06, + "loss": 0.5997, + "step": 4894 + }, + { + "epoch": 1.81, + "learning_rate": 2.820975489592314e-06, + "loss": 0.6191, + "step": 4895 + }, + { + "epoch": 1.81, + "learning_rate": 2.820938539228969e-06, + "loss": 0.7026, + "step": 4896 + }, + { + "epoch": 1.81, + "learning_rate": 2.820901588865624e-06, + "loss": 0.5957, + "step": 4897 + }, + { + "epoch": 1.81, + "learning_rate": 2.820864638502279e-06, + "loss": 0.6575, + "step": 4898 + }, + { + "epoch": 1.81, + "learning_rate": 2.8208276881389335e-06, + "loss": 0.592, + "step": 4899 + }, + { + "epoch": 1.81, + "learning_rate": 2.820790737775588e-06, + "loss": 0.6811, + "step": 4900 + }, + { + "epoch": 1.81, + "learning_rate": 2.820753787412243e-06, + "loss": 0.592, + "step": 4901 + }, + { + "epoch": 1.81, + "learning_rate": 2.8207168370488974e-06, + "loss": 0.6404, + "step": 4902 + }, + { + "epoch": 1.81, + "learning_rate": 2.8206798866855525e-06, + "loss": 0.5639, + "step": 4903 + }, + { + "epoch": 1.81, + "learning_rate": 2.820642936322207e-06, + "loss": 0.6734, + "step": 4904 + }, + { + "epoch": 1.81, + "learning_rate": 2.820605985958862e-06, + "loss": 0.6265, + "step": 4905 + }, + { + "epoch": 1.81, + "learning_rate": 2.820569035595517e-06, + "loss": 0.7114, + "step": 4906 + }, + { + "epoch": 1.81, + "learning_rate": 2.8205320852321715e-06, + "loss": 0.5622, + "step": 4907 + }, + { + "epoch": 1.81, + "learning_rate": 2.820495134868826e-06, + "loss": 0.6321, + "step": 4908 + }, + { + "epoch": 1.81, + "learning_rate": 2.820458184505481e-06, + "loss": 0.6567, + "step": 4909 + }, + { + "epoch": 1.81, + "learning_rate": 2.820421234142136e-06, + "loss": 0.5972, + "step": 4910 + }, + { + "epoch": 1.81, + "learning_rate": 2.8203842837787904e-06, + "loss": 0.6371, + "step": 4911 + }, + { + "epoch": 1.81, + "learning_rate": 2.8203473334154455e-06, + "loss": 0.6366, + "step": 4912 + }, + { + "epoch": 1.81, + "learning_rate": 2.8203103830521e-06, + "loss": 0.612, + "step": 4913 + }, + { + "epoch": 1.81, + "learning_rate": 2.8202734326887548e-06, + "loss": 0.5316, + "step": 4914 + }, + { + "epoch": 1.81, + "learning_rate": 2.8202364823254094e-06, + "loss": 0.6484, + "step": 4915 + }, + { + "epoch": 1.82, + "learning_rate": 2.8201995319620645e-06, + "loss": 0.6089, + "step": 4916 + }, + { + "epoch": 1.82, + "learning_rate": 2.820162581598719e-06, + "loss": 0.5687, + "step": 4917 + }, + { + "epoch": 1.82, + "learning_rate": 2.820125631235374e-06, + "loss": 0.5923, + "step": 4918 + }, + { + "epoch": 1.82, + "learning_rate": 2.820088680872029e-06, + "loss": 0.5464, + "step": 4919 + }, + { + "epoch": 1.82, + "learning_rate": 2.8200517305086834e-06, + "loss": 0.5866, + "step": 4920 + }, + { + "epoch": 1.82, + "learning_rate": 2.820014780145338e-06, + "loss": 0.7691, + "step": 4921 + }, + { + "epoch": 1.82, + "learning_rate": 2.8199778297819927e-06, + "loss": 0.5779, + "step": 4922 + }, + { + "epoch": 1.82, + "learning_rate": 2.8199408794186478e-06, + "loss": 0.6107, + "step": 4923 + }, + { + "epoch": 1.82, + "learning_rate": 2.8199039290553024e-06, + "loss": 0.6354, + "step": 4924 + }, + { + "epoch": 1.82, + "learning_rate": 2.8198669786919575e-06, + "loss": 0.5789, + "step": 4925 + }, + { + "epoch": 1.82, + "learning_rate": 2.819830028328612e-06, + "loss": 0.5578, + "step": 4926 + }, + { + "epoch": 1.82, + "learning_rate": 2.8197930779652667e-06, + "loss": 0.6602, + "step": 4927 + }, + { + "epoch": 1.82, + "learning_rate": 2.8197561276019214e-06, + "loss": 0.5296, + "step": 4928 + }, + { + "epoch": 1.82, + "learning_rate": 2.819719177238576e-06, + "loss": 0.6268, + "step": 4929 + }, + { + "epoch": 1.82, + "learning_rate": 2.819682226875231e-06, + "loss": 0.6291, + "step": 4930 + }, + { + "epoch": 1.82, + "learning_rate": 2.8196452765118857e-06, + "loss": 0.6121, + "step": 4931 + }, + { + "epoch": 1.82, + "learning_rate": 2.8196083261485408e-06, + "loss": 0.6728, + "step": 4932 + }, + { + "epoch": 1.82, + "learning_rate": 2.819571375785195e-06, + "loss": 0.6118, + "step": 4933 + }, + { + "epoch": 1.82, + "learning_rate": 2.81953442542185e-06, + "loss": 0.6313, + "step": 4934 + }, + { + "epoch": 1.82, + "learning_rate": 2.8194974750585047e-06, + "loss": 0.6262, + "step": 4935 + }, + { + "epoch": 1.82, + "learning_rate": 2.8194605246951597e-06, + "loss": 0.5741, + "step": 4936 + }, + { + "epoch": 1.82, + "learning_rate": 2.8194235743318144e-06, + "loss": 0.5642, + "step": 4937 + }, + { + "epoch": 1.82, + "learning_rate": 2.8193866239684694e-06, + "loss": 0.5792, + "step": 4938 + }, + { + "epoch": 1.82, + "learning_rate": 2.8193496736051237e-06, + "loss": 0.5866, + "step": 4939 + }, + { + "epoch": 1.82, + "learning_rate": 2.8193127232417783e-06, + "loss": 0.6346, + "step": 4940 + }, + { + "epoch": 1.82, + "learning_rate": 2.8192757728784334e-06, + "loss": 0.6276, + "step": 4941 + }, + { + "epoch": 1.82, + "learning_rate": 2.819238822515088e-06, + "loss": 0.6482, + "step": 4942 + }, + { + "epoch": 1.83, + "learning_rate": 2.819201872151743e-06, + "loss": 0.5693, + "step": 4943 + }, + { + "epoch": 1.83, + "learning_rate": 2.8191649217883977e-06, + "loss": 0.6026, + "step": 4944 + }, + { + "epoch": 1.83, + "learning_rate": 2.8191279714250523e-06, + "loss": 0.614, + "step": 4945 + }, + { + "epoch": 1.83, + "learning_rate": 2.819091021061707e-06, + "loss": 0.6279, + "step": 4946 + }, + { + "epoch": 1.83, + "learning_rate": 2.819054070698362e-06, + "loss": 0.5587, + "step": 4947 + }, + { + "epoch": 1.83, + "learning_rate": 2.8190171203350167e-06, + "loss": 0.5534, + "step": 4948 + }, + { + "epoch": 1.83, + "learning_rate": 2.8189801699716713e-06, + "loss": 0.5773, + "step": 4949 + }, + { + "epoch": 1.83, + "learning_rate": 2.8189432196083263e-06, + "loss": 0.6441, + "step": 4950 + }, + { + "epoch": 1.83, + "learning_rate": 2.818906269244981e-06, + "loss": 0.632, + "step": 4951 + }, + { + "epoch": 1.83, + "learning_rate": 2.8188693188816356e-06, + "loss": 0.4987, + "step": 4952 + }, + { + "epoch": 1.83, + "learning_rate": 2.8188323685182903e-06, + "loss": 0.5185, + "step": 4953 + }, + { + "epoch": 1.83, + "learning_rate": 2.8187954181549453e-06, + "loss": 0.6284, + "step": 4954 + }, + { + "epoch": 1.83, + "learning_rate": 2.8187584677916e-06, + "loss": 0.6021, + "step": 4955 + }, + { + "epoch": 1.83, + "learning_rate": 2.818721517428255e-06, + "loss": 0.6325, + "step": 4956 + }, + { + "epoch": 1.83, + "learning_rate": 2.8186845670649097e-06, + "loss": 0.6015, + "step": 4957 + }, + { + "epoch": 1.83, + "learning_rate": 2.8186476167015643e-06, + "loss": 0.5595, + "step": 4958 + }, + { + "epoch": 1.83, + "learning_rate": 2.818610666338219e-06, + "loss": 0.5834, + "step": 4959 + }, + { + "epoch": 1.83, + "learning_rate": 2.8185737159748736e-06, + "loss": 0.5384, + "step": 4960 + }, + { + "epoch": 1.83, + "learning_rate": 2.8185367656115286e-06, + "loss": 0.5813, + "step": 4961 + }, + { + "epoch": 1.83, + "learning_rate": 2.8184998152481833e-06, + "loss": 0.547, + "step": 4962 + }, + { + "epoch": 1.83, + "learning_rate": 2.8184628648848383e-06, + "loss": 0.5613, + "step": 4963 + }, + { + "epoch": 1.83, + "learning_rate": 2.818425914521493e-06, + "loss": 0.6661, + "step": 4964 + }, + { + "epoch": 1.83, + "learning_rate": 2.8183889641581476e-06, + "loss": 0.6129, + "step": 4965 + }, + { + "epoch": 1.83, + "learning_rate": 2.8183520137948022e-06, + "loss": 0.634, + "step": 4966 + }, + { + "epoch": 1.83, + "learning_rate": 2.8183150634314573e-06, + "loss": 0.6518, + "step": 4967 + }, + { + "epoch": 1.83, + "learning_rate": 2.818278113068112e-06, + "loss": 0.576, + "step": 4968 + }, + { + "epoch": 1.83, + "learning_rate": 2.8182411627047666e-06, + "loss": 0.6487, + "step": 4969 + }, + { + "epoch": 1.84, + "learning_rate": 2.8182042123414216e-06, + "loss": 0.5932, + "step": 4970 + }, + { + "epoch": 1.84, + "learning_rate": 2.8181672619780763e-06, + "loss": 0.686, + "step": 4971 + }, + { + "epoch": 1.84, + "learning_rate": 2.818130311614731e-06, + "loss": 0.6162, + "step": 4972 + }, + { + "epoch": 1.84, + "learning_rate": 2.8180933612513855e-06, + "loss": 0.5829, + "step": 4973 + }, + { + "epoch": 1.84, + "learning_rate": 2.8180564108880406e-06, + "loss": 0.571, + "step": 4974 + }, + { + "epoch": 1.84, + "learning_rate": 2.8180194605246952e-06, + "loss": 0.6201, + "step": 4975 + }, + { + "epoch": 1.84, + "learning_rate": 2.8179825101613503e-06, + "loss": 0.597, + "step": 4976 + }, + { + "epoch": 1.84, + "learning_rate": 2.817945559798005e-06, + "loss": 0.7011, + "step": 4977 + }, + { + "epoch": 1.84, + "learning_rate": 2.8179086094346596e-06, + "loss": 0.6022, + "step": 4978 + }, + { + "epoch": 1.84, + "learning_rate": 2.817871659071314e-06, + "loss": 0.5215, + "step": 4979 + }, + { + "epoch": 1.84, + "learning_rate": 2.817834708707969e-06, + "loss": 0.6785, + "step": 4980 + }, + { + "epoch": 1.84, + "learning_rate": 2.817797758344624e-06, + "loss": 0.5612, + "step": 4981 + }, + { + "epoch": 1.84, + "learning_rate": 2.8177608079812785e-06, + "loss": 0.6606, + "step": 4982 + }, + { + "epoch": 1.84, + "learning_rate": 2.8177238576179336e-06, + "loss": 0.6027, + "step": 4983 + }, + { + "epoch": 1.84, + "learning_rate": 2.817686907254588e-06, + "loss": 0.6412, + "step": 4984 + }, + { + "epoch": 1.84, + "learning_rate": 2.817649956891243e-06, + "loss": 0.6605, + "step": 4985 + }, + { + "epoch": 1.84, + "learning_rate": 2.8176130065278975e-06, + "loss": 0.6038, + "step": 4986 + }, + { + "epoch": 1.84, + "learning_rate": 2.8175760561645526e-06, + "loss": 0.6299, + "step": 4987 + }, + { + "epoch": 1.84, + "learning_rate": 2.817539105801207e-06, + "loss": 0.6456, + "step": 4988 + }, + { + "epoch": 1.84, + "learning_rate": 2.817502155437862e-06, + "loss": 0.6403, + "step": 4989 + }, + { + "epoch": 1.84, + "learning_rate": 2.8174652050745165e-06, + "loss": 0.575, + "step": 4990 + }, + { + "epoch": 1.84, + "learning_rate": 2.817428254711171e-06, + "loss": 0.6094, + "step": 4991 + }, + { + "epoch": 1.84, + "learning_rate": 2.817391304347826e-06, + "loss": 0.5647, + "step": 4992 + }, + { + "epoch": 1.84, + "learning_rate": 2.817354353984481e-06, + "loss": 0.6276, + "step": 4993 + }, + { + "epoch": 1.84, + "learning_rate": 2.817317403621136e-06, + "loss": 0.6008, + "step": 4994 + }, + { + "epoch": 1.84, + "learning_rate": 2.8172804532577905e-06, + "loss": 0.4907, + "step": 4995 + }, + { + "epoch": 1.84, + "learning_rate": 2.817243502894445e-06, + "loss": 0.586, + "step": 4996 + }, + { + "epoch": 1.85, + "learning_rate": 2.8172065525310998e-06, + "loss": 0.6153, + "step": 4997 + }, + { + "epoch": 1.85, + "learning_rate": 2.817169602167755e-06, + "loss": 0.6068, + "step": 4998 + }, + { + "epoch": 1.85, + "learning_rate": 2.8171326518044095e-06, + "loss": 0.6137, + "step": 4999 + }, + { + "epoch": 1.85, + "learning_rate": 2.817095701441064e-06, + "loss": 0.5209, + "step": 5000 + }, + { + "epoch": 1.85, + "learning_rate": 2.817058751077719e-06, + "loss": 0.6653, + "step": 5001 + }, + { + "epoch": 1.85, + "learning_rate": 2.817021800714374e-06, + "loss": 0.6102, + "step": 5002 + }, + { + "epoch": 1.85, + "learning_rate": 2.8169848503510284e-06, + "loss": 0.5612, + "step": 5003 + }, + { + "epoch": 1.85, + "learning_rate": 2.816947899987683e-06, + "loss": 0.6872, + "step": 5004 + }, + { + "epoch": 1.85, + "learning_rate": 2.816910949624338e-06, + "loss": 0.5434, + "step": 5005 + }, + { + "epoch": 1.85, + "learning_rate": 2.8168739992609928e-06, + "loss": 0.564, + "step": 5006 + }, + { + "epoch": 1.85, + "learning_rate": 2.816837048897648e-06, + "loss": 0.5721, + "step": 5007 + }, + { + "epoch": 1.85, + "learning_rate": 2.8168000985343025e-06, + "loss": 0.5559, + "step": 5008 + }, + { + "epoch": 1.85, + "learning_rate": 2.816763148170957e-06, + "loss": 0.574, + "step": 5009 + }, + { + "epoch": 1.85, + "learning_rate": 2.8167261978076117e-06, + "loss": 0.6116, + "step": 5010 + }, + { + "epoch": 1.85, + "learning_rate": 2.8166892474442664e-06, + "loss": 0.6347, + "step": 5011 + }, + { + "epoch": 1.85, + "learning_rate": 2.8166522970809214e-06, + "loss": 0.587, + "step": 5012 + }, + { + "epoch": 1.85, + "learning_rate": 2.816615346717576e-06, + "loss": 0.6148, + "step": 5013 + }, + { + "epoch": 1.85, + "learning_rate": 2.816578396354231e-06, + "loss": 0.6179, + "step": 5014 + }, + { + "epoch": 1.85, + "learning_rate": 2.8165414459908858e-06, + "loss": 0.6718, + "step": 5015 + }, + { + "epoch": 1.85, + "learning_rate": 2.8165044956275404e-06, + "loss": 0.633, + "step": 5016 + }, + { + "epoch": 1.85, + "learning_rate": 2.816467545264195e-06, + "loss": 0.6596, + "step": 5017 + }, + { + "epoch": 1.85, + "learning_rate": 2.81643059490085e-06, + "loss": 0.5813, + "step": 5018 + }, + { + "epoch": 1.85, + "learning_rate": 2.8163936445375047e-06, + "loss": 0.5383, + "step": 5019 + }, + { + "epoch": 1.85, + "learning_rate": 2.8163566941741594e-06, + "loss": 0.5712, + "step": 5020 + }, + { + "epoch": 1.85, + "learning_rate": 2.8163197438108144e-06, + "loss": 0.6187, + "step": 5021 + }, + { + "epoch": 1.85, + "learning_rate": 2.816282793447469e-06, + "loss": 0.6505, + "step": 5022 + }, + { + "epoch": 1.85, + "learning_rate": 2.8162458430841237e-06, + "loss": 0.581, + "step": 5023 + }, + { + "epoch": 1.86, + "learning_rate": 2.8162088927207783e-06, + "loss": 0.6491, + "step": 5024 + }, + { + "epoch": 1.86, + "learning_rate": 2.8161719423574334e-06, + "loss": 0.6632, + "step": 5025 + }, + { + "epoch": 1.86, + "learning_rate": 2.816134991994088e-06, + "loss": 0.6406, + "step": 5026 + }, + { + "epoch": 1.86, + "learning_rate": 2.816098041630743e-06, + "loss": 0.6808, + "step": 5027 + }, + { + "epoch": 1.86, + "learning_rate": 2.8160610912673977e-06, + "loss": 0.6103, + "step": 5028 + }, + { + "epoch": 1.86, + "learning_rate": 2.816024140904052e-06, + "loss": 0.6532, + "step": 5029 + }, + { + "epoch": 1.86, + "learning_rate": 2.815987190540707e-06, + "loss": 0.6364, + "step": 5030 + }, + { + "epoch": 1.86, + "learning_rate": 2.8159502401773616e-06, + "loss": 0.7133, + "step": 5031 + }, + { + "epoch": 1.86, + "learning_rate": 2.8159132898140167e-06, + "loss": 0.6771, + "step": 5032 + }, + { + "epoch": 1.86, + "learning_rate": 2.8158763394506713e-06, + "loss": 0.5879, + "step": 5033 + }, + { + "epoch": 1.86, + "learning_rate": 2.8158393890873264e-06, + "loss": 0.6414, + "step": 5034 + }, + { + "epoch": 1.86, + "learning_rate": 2.8158024387239806e-06, + "loss": 0.6014, + "step": 5035 + }, + { + "epoch": 1.86, + "learning_rate": 2.8157654883606357e-06, + "loss": 0.5732, + "step": 5036 + }, + { + "epoch": 1.86, + "learning_rate": 2.8157285379972903e-06, + "loss": 0.671, + "step": 5037 + }, + { + "epoch": 1.86, + "learning_rate": 2.8156915876339454e-06, + "loss": 0.6277, + "step": 5038 + }, + { + "epoch": 1.86, + "learning_rate": 2.8156546372706e-06, + "loss": 0.6396, + "step": 5039 + }, + { + "epoch": 1.86, + "learning_rate": 2.8156176869072546e-06, + "loss": 0.5949, + "step": 5040 + }, + { + "epoch": 1.86, + "learning_rate": 2.8155807365439093e-06, + "loss": 0.6147, + "step": 5041 + }, + { + "epoch": 1.86, + "learning_rate": 2.815543786180564e-06, + "loss": 0.544, + "step": 5042 + }, + { + "epoch": 1.86, + "learning_rate": 2.815506835817219e-06, + "loss": 0.5736, + "step": 5043 + }, + { + "epoch": 1.86, + "learning_rate": 2.8154698854538736e-06, + "loss": 0.615, + "step": 5044 + }, + { + "epoch": 1.86, + "learning_rate": 2.8154329350905287e-06, + "loss": 0.6306, + "step": 5045 + }, + { + "epoch": 1.86, + "learning_rate": 2.8153959847271833e-06, + "loss": 0.6096, + "step": 5046 + }, + { + "epoch": 1.86, + "learning_rate": 2.815359034363838e-06, + "loss": 0.5759, + "step": 5047 + }, + { + "epoch": 1.86, + "learning_rate": 2.8153220840004926e-06, + "loss": 0.6518, + "step": 5048 + }, + { + "epoch": 1.86, + "learning_rate": 2.8152851336371472e-06, + "loss": 0.5517, + "step": 5049 + }, + { + "epoch": 1.86, + "learning_rate": 2.8152481832738023e-06, + "loss": 0.6612, + "step": 5050 + }, + { + "epoch": 1.87, + "learning_rate": 2.815211232910457e-06, + "loss": 0.5931, + "step": 5051 + }, + { + "epoch": 1.87, + "learning_rate": 2.815174282547112e-06, + "loss": 0.6453, + "step": 5052 + }, + { + "epoch": 1.87, + "learning_rate": 2.8151373321837666e-06, + "loss": 0.6219, + "step": 5053 + }, + { + "epoch": 1.87, + "learning_rate": 2.8151003818204212e-06, + "loss": 0.5613, + "step": 5054 + }, + { + "epoch": 1.87, + "learning_rate": 2.815063431457076e-06, + "loss": 0.5891, + "step": 5055 + }, + { + "epoch": 1.87, + "learning_rate": 2.815026481093731e-06, + "loss": 0.6709, + "step": 5056 + }, + { + "epoch": 1.87, + "learning_rate": 2.8149895307303856e-06, + "loss": 0.5834, + "step": 5057 + }, + { + "epoch": 1.87, + "learning_rate": 2.8149525803670406e-06, + "loss": 0.6584, + "step": 5058 + }, + { + "epoch": 1.87, + "learning_rate": 2.8149156300036953e-06, + "loss": 0.5728, + "step": 5059 + }, + { + "epoch": 1.87, + "learning_rate": 2.81487867964035e-06, + "loss": 0.5318, + "step": 5060 + }, + { + "epoch": 1.87, + "learning_rate": 2.8148417292770045e-06, + "loss": 0.6144, + "step": 5061 + }, + { + "epoch": 1.87, + "learning_rate": 2.814804778913659e-06, + "loss": 0.6408, + "step": 5062 + }, + { + "epoch": 1.87, + "learning_rate": 2.8147678285503142e-06, + "loss": 0.6098, + "step": 5063 + }, + { + "epoch": 1.87, + "learning_rate": 2.814730878186969e-06, + "loss": 0.6761, + "step": 5064 + }, + { + "epoch": 1.87, + "learning_rate": 2.814693927823624e-06, + "loss": 0.6445, + "step": 5065 + }, + { + "epoch": 1.87, + "learning_rate": 2.8146569774602786e-06, + "loss": 0.5574, + "step": 5066 + }, + { + "epoch": 1.87, + "learning_rate": 2.8146200270969332e-06, + "loss": 0.607, + "step": 5067 + }, + { + "epoch": 1.87, + "learning_rate": 2.814583076733588e-06, + "loss": 0.5442, + "step": 5068 + }, + { + "epoch": 1.87, + "learning_rate": 2.8145461263702425e-06, + "loss": 0.6911, + "step": 5069 + }, + { + "epoch": 1.87, + "learning_rate": 2.8145091760068975e-06, + "loss": 0.5667, + "step": 5070 + }, + { + "epoch": 1.87, + "learning_rate": 2.814472225643552e-06, + "loss": 0.5235, + "step": 5071 + }, + { + "epoch": 1.87, + "learning_rate": 2.8144352752802072e-06, + "loss": 0.6998, + "step": 5072 + }, + { + "epoch": 1.87, + "learning_rate": 2.8143983249168615e-06, + "loss": 0.5771, + "step": 5073 + }, + { + "epoch": 1.87, + "learning_rate": 2.8143613745535165e-06, + "loss": 0.6027, + "step": 5074 + }, + { + "epoch": 1.87, + "learning_rate": 2.814324424190171e-06, + "loss": 0.5495, + "step": 5075 + }, + { + "epoch": 1.87, + "learning_rate": 2.814287473826826e-06, + "loss": 0.6304, + "step": 5076 + }, + { + "epoch": 1.87, + "learning_rate": 2.814250523463481e-06, + "loss": 0.6218, + "step": 5077 + }, + { + "epoch": 1.88, + "learning_rate": 2.814213573100136e-06, + "loss": 0.6755, + "step": 5078 + }, + { + "epoch": 1.88, + "learning_rate": 2.81417662273679e-06, + "loss": 0.6573, + "step": 5079 + }, + { + "epoch": 1.88, + "learning_rate": 2.8141396723734448e-06, + "loss": 0.6688, + "step": 5080 + }, + { + "epoch": 1.88, + "learning_rate": 2.8141027220101e-06, + "loss": 0.6138, + "step": 5081 + }, + { + "epoch": 1.88, + "learning_rate": 2.8140657716467545e-06, + "loss": 0.6568, + "step": 5082 + }, + { + "epoch": 1.88, + "learning_rate": 2.8140288212834095e-06, + "loss": 0.599, + "step": 5083 + }, + { + "epoch": 1.88, + "learning_rate": 2.813991870920064e-06, + "loss": 0.6233, + "step": 5084 + }, + { + "epoch": 1.88, + "learning_rate": 2.8139549205567188e-06, + "loss": 0.5957, + "step": 5085 + }, + { + "epoch": 1.88, + "learning_rate": 2.8139179701933734e-06, + "loss": 0.6606, + "step": 5086 + }, + { + "epoch": 1.88, + "learning_rate": 2.8138810198300285e-06, + "loss": 0.5409, + "step": 5087 + }, + { + "epoch": 1.88, + "learning_rate": 2.813844069466683e-06, + "loss": 0.6347, + "step": 5088 + }, + { + "epoch": 1.88, + "learning_rate": 2.8138071191033378e-06, + "loss": 0.5536, + "step": 5089 + }, + { + "epoch": 1.88, + "learning_rate": 2.813770168739993e-06, + "loss": 0.6594, + "step": 5090 + }, + { + "epoch": 1.88, + "learning_rate": 2.8137332183766475e-06, + "loss": 0.5888, + "step": 5091 + }, + { + "epoch": 1.88, + "learning_rate": 2.813696268013302e-06, + "loss": 0.6521, + "step": 5092 + }, + { + "epoch": 1.88, + "learning_rate": 2.8136593176499567e-06, + "loss": 0.5463, + "step": 5093 + }, + { + "epoch": 1.88, + "learning_rate": 2.8136223672866118e-06, + "loss": 0.5061, + "step": 5094 + }, + { + "epoch": 1.88, + "learning_rate": 2.8135854169232664e-06, + "loss": 0.6506, + "step": 5095 + }, + { + "epoch": 1.88, + "learning_rate": 2.8135484665599215e-06, + "loss": 0.5466, + "step": 5096 + }, + { + "epoch": 1.88, + "learning_rate": 2.813511516196576e-06, + "loss": 0.5967, + "step": 5097 + }, + { + "epoch": 1.88, + "learning_rate": 2.8134745658332308e-06, + "loss": 0.5827, + "step": 5098 + }, + { + "epoch": 1.88, + "learning_rate": 2.8134376154698854e-06, + "loss": 0.5682, + "step": 5099 + }, + { + "epoch": 1.88, + "learning_rate": 2.81340066510654e-06, + "loss": 0.6392, + "step": 5100 + }, + { + "epoch": 1.88, + "learning_rate": 2.813363714743195e-06, + "loss": 0.5855, + "step": 5101 + }, + { + "epoch": 1.88, + "learning_rate": 2.8133267643798497e-06, + "loss": 0.6666, + "step": 5102 + }, + { + "epoch": 1.88, + "learning_rate": 2.8132898140165048e-06, + "loss": 0.5739, + "step": 5103 + }, + { + "epoch": 1.88, + "learning_rate": 2.8132528636531594e-06, + "loss": 0.5137, + "step": 5104 + }, + { + "epoch": 1.89, + "learning_rate": 2.813215913289814e-06, + "loss": 0.6219, + "step": 5105 + }, + { + "epoch": 1.89, + "learning_rate": 2.8131789629264687e-06, + "loss": 0.5564, + "step": 5106 + }, + { + "epoch": 1.89, + "learning_rate": 2.8131420125631238e-06, + "loss": 0.6682, + "step": 5107 + }, + { + "epoch": 1.89, + "learning_rate": 2.8131050621997784e-06, + "loss": 0.617, + "step": 5108 + }, + { + "epoch": 1.89, + "learning_rate": 2.813068111836433e-06, + "loss": 0.6022, + "step": 5109 + }, + { + "epoch": 1.89, + "learning_rate": 2.813031161473088e-06, + "loss": 0.6735, + "step": 5110 + }, + { + "epoch": 1.89, + "learning_rate": 2.8129942111097427e-06, + "loss": 0.5941, + "step": 5111 + }, + { + "epoch": 1.89, + "learning_rate": 2.8129572607463974e-06, + "loss": 0.6727, + "step": 5112 + }, + { + "epoch": 1.89, + "learning_rate": 2.812920310383052e-06, + "loss": 0.5911, + "step": 5113 + }, + { + "epoch": 1.89, + "learning_rate": 2.812883360019707e-06, + "loss": 0.6676, + "step": 5114 + }, + { + "epoch": 1.89, + "learning_rate": 2.8128464096563617e-06, + "loss": 0.5943, + "step": 5115 + }, + { + "epoch": 1.89, + "learning_rate": 2.8128094592930167e-06, + "loss": 0.6258, + "step": 5116 + }, + { + "epoch": 1.89, + "learning_rate": 2.8127725089296714e-06, + "loss": 0.6467, + "step": 5117 + }, + { + "epoch": 1.89, + "learning_rate": 2.812735558566326e-06, + "loss": 0.6483, + "step": 5118 + }, + { + "epoch": 1.89, + "learning_rate": 2.8126986082029807e-06, + "loss": 0.6488, + "step": 5119 + }, + { + "epoch": 1.89, + "learning_rate": 2.8126616578396353e-06, + "loss": 0.6113, + "step": 5120 + }, + { + "epoch": 1.89, + "learning_rate": 2.8126247074762904e-06, + "loss": 0.5941, + "step": 5121 + }, + { + "epoch": 1.89, + "learning_rate": 2.812587757112945e-06, + "loss": 0.6176, + "step": 5122 + }, + { + "epoch": 1.89, + "learning_rate": 2.8125508067496e-06, + "loss": 0.6089, + "step": 5123 + }, + { + "epoch": 1.89, + "learning_rate": 2.8125138563862543e-06, + "loss": 0.5176, + "step": 5124 + }, + { + "epoch": 1.89, + "learning_rate": 2.8124769060229093e-06, + "loss": 0.5831, + "step": 5125 + }, + { + "epoch": 1.89, + "learning_rate": 2.812439955659564e-06, + "loss": 0.6959, + "step": 5126 + }, + { + "epoch": 1.89, + "learning_rate": 2.812403005296219e-06, + "loss": 0.569, + "step": 5127 + }, + { + "epoch": 1.89, + "learning_rate": 2.8123660549328737e-06, + "loss": 0.7412, + "step": 5128 + }, + { + "epoch": 1.89, + "learning_rate": 2.8123291045695283e-06, + "loss": 0.5864, + "step": 5129 + }, + { + "epoch": 1.89, + "learning_rate": 2.812292154206183e-06, + "loss": 0.6114, + "step": 5130 + }, + { + "epoch": 1.89, + "learning_rate": 2.8122552038428376e-06, + "loss": 0.5548, + "step": 5131 + }, + { + "epoch": 1.9, + "learning_rate": 2.8122182534794926e-06, + "loss": 0.5394, + "step": 5132 + }, + { + "epoch": 1.9, + "learning_rate": 2.8121813031161473e-06, + "loss": 0.5519, + "step": 5133 + }, + { + "epoch": 1.9, + "learning_rate": 2.8121443527528023e-06, + "loss": 0.6047, + "step": 5134 + }, + { + "epoch": 1.9, + "learning_rate": 2.812107402389457e-06, + "loss": 0.6037, + "step": 5135 + }, + { + "epoch": 1.9, + "learning_rate": 2.8120704520261116e-06, + "loss": 0.6334, + "step": 5136 + }, + { + "epoch": 1.9, + "learning_rate": 2.8120335016627662e-06, + "loss": 0.6203, + "step": 5137 + }, + { + "epoch": 1.9, + "learning_rate": 2.8119965512994213e-06, + "loss": 0.5292, + "step": 5138 + }, + { + "epoch": 1.9, + "learning_rate": 2.811959600936076e-06, + "loss": 0.6368, + "step": 5139 + }, + { + "epoch": 1.9, + "learning_rate": 2.8119226505727306e-06, + "loss": 0.6292, + "step": 5140 + }, + { + "epoch": 1.9, + "learning_rate": 2.8118857002093856e-06, + "loss": 0.6093, + "step": 5141 + }, + { + "epoch": 1.9, + "learning_rate": 2.8118487498460403e-06, + "loss": 0.5897, + "step": 5142 + }, + { + "epoch": 1.9, + "learning_rate": 2.811811799482695e-06, + "loss": 0.5855, + "step": 5143 + }, + { + "epoch": 1.9, + "learning_rate": 2.8117748491193495e-06, + "loss": 0.6614, + "step": 5144 + }, + { + "epoch": 1.9, + "learning_rate": 2.8117378987560046e-06, + "loss": 0.6368, + "step": 5145 + }, + { + "epoch": 1.9, + "learning_rate": 2.8117009483926592e-06, + "loss": 0.6104, + "step": 5146 + }, + { + "epoch": 1.9, + "learning_rate": 2.8116639980293143e-06, + "loss": 0.6177, + "step": 5147 + }, + { + "epoch": 1.9, + "learning_rate": 2.811627047665969e-06, + "loss": 0.6136, + "step": 5148 + }, + { + "epoch": 1.9, + "learning_rate": 2.8115900973026236e-06, + "loss": 0.552, + "step": 5149 + }, + { + "epoch": 1.9, + "learning_rate": 2.811553146939278e-06, + "loss": 0.7161, + "step": 5150 + }, + { + "epoch": 1.9, + "learning_rate": 2.811516196575933e-06, + "loss": 0.5367, + "step": 5151 + }, + { + "epoch": 1.9, + "learning_rate": 2.811479246212588e-06, + "loss": 0.5507, + "step": 5152 + }, + { + "epoch": 1.9, + "learning_rate": 2.8114422958492425e-06, + "loss": 0.7207, + "step": 5153 + }, + { + "epoch": 1.9, + "learning_rate": 2.8114053454858976e-06, + "loss": 0.6069, + "step": 5154 + }, + { + "epoch": 1.9, + "learning_rate": 2.8113683951225522e-06, + "loss": 0.5991, + "step": 5155 + }, + { + "epoch": 1.9, + "learning_rate": 2.811331444759207e-06, + "loss": 0.6223, + "step": 5156 + }, + { + "epoch": 1.9, + "learning_rate": 2.8112944943958615e-06, + "loss": 0.5975, + "step": 5157 + }, + { + "epoch": 1.9, + "learning_rate": 2.811257544032516e-06, + "loss": 0.573, + "step": 5158 + }, + { + "epoch": 1.91, + "learning_rate": 2.811220593669171e-06, + "loss": 0.5947, + "step": 5159 + }, + { + "epoch": 1.91, + "learning_rate": 2.811183643305826e-06, + "loss": 0.572, + "step": 5160 + }, + { + "epoch": 1.91, + "learning_rate": 2.811146692942481e-06, + "loss": 0.5795, + "step": 5161 + }, + { + "epoch": 1.91, + "learning_rate": 2.8111097425791355e-06, + "loss": 0.6237, + "step": 5162 + }, + { + "epoch": 1.91, + "learning_rate": 2.81107279221579e-06, + "loss": 0.6392, + "step": 5163 + }, + { + "epoch": 1.91, + "learning_rate": 2.811035841852445e-06, + "loss": 0.5923, + "step": 5164 + }, + { + "epoch": 1.91, + "learning_rate": 2.8109988914891e-06, + "loss": 0.5818, + "step": 5165 + }, + { + "epoch": 1.91, + "learning_rate": 2.8109619411257545e-06, + "loss": 0.582, + "step": 5166 + }, + { + "epoch": 1.91, + "learning_rate": 2.8109249907624096e-06, + "loss": 0.5589, + "step": 5167 + }, + { + "epoch": 1.91, + "learning_rate": 2.810888040399064e-06, + "loss": 0.6511, + "step": 5168 + }, + { + "epoch": 1.91, + "learning_rate": 2.8108510900357184e-06, + "loss": 0.6682, + "step": 5169 + }, + { + "epoch": 1.91, + "learning_rate": 2.8108141396723735e-06, + "loss": 0.5876, + "step": 5170 + }, + { + "epoch": 1.91, + "learning_rate": 2.810777189309028e-06, + "loss": 0.6653, + "step": 5171 + }, + { + "epoch": 1.91, + "learning_rate": 2.810740238945683e-06, + "loss": 0.614, + "step": 5172 + }, + { + "epoch": 1.91, + "learning_rate": 2.810703288582338e-06, + "loss": 0.6425, + "step": 5173 + }, + { + "epoch": 1.91, + "learning_rate": 2.810666338218993e-06, + "loss": 0.6693, + "step": 5174 + }, + { + "epoch": 1.91, + "learning_rate": 2.810629387855647e-06, + "loss": 0.5442, + "step": 5175 + }, + { + "epoch": 1.91, + "learning_rate": 2.810592437492302e-06, + "loss": 0.616, + "step": 5176 + }, + { + "epoch": 1.91, + "learning_rate": 2.8105554871289568e-06, + "loss": 0.6337, + "step": 5177 + }, + { + "epoch": 1.91, + "learning_rate": 2.8105185367656114e-06, + "loss": 0.5008, + "step": 5178 + }, + { + "epoch": 1.91, + "learning_rate": 2.8104815864022665e-06, + "loss": 0.592, + "step": 5179 + }, + { + "epoch": 1.91, + "learning_rate": 2.810444636038921e-06, + "loss": 0.5672, + "step": 5180 + }, + { + "epoch": 1.91, + "learning_rate": 2.8104076856755757e-06, + "loss": 0.596, + "step": 5181 + }, + { + "epoch": 1.91, + "learning_rate": 2.8103707353122304e-06, + "loss": 0.6241, + "step": 5182 + }, + { + "epoch": 1.91, + "learning_rate": 2.8103337849488854e-06, + "loss": 0.5645, + "step": 5183 + }, + { + "epoch": 1.91, + "learning_rate": 2.81029683458554e-06, + "loss": 0.5461, + "step": 5184 + }, + { + "epoch": 1.91, + "learning_rate": 2.810259884222195e-06, + "loss": 0.5683, + "step": 5185 + }, + { + "epoch": 1.92, + "learning_rate": 2.8102229338588498e-06, + "loss": 0.5881, + "step": 5186 + }, + { + "epoch": 1.92, + "learning_rate": 2.8101859834955044e-06, + "loss": 0.5709, + "step": 5187 + }, + { + "epoch": 1.92, + "learning_rate": 2.810149033132159e-06, + "loss": 0.6464, + "step": 5188 + }, + { + "epoch": 1.92, + "learning_rate": 2.8101120827688137e-06, + "loss": 0.5788, + "step": 5189 + }, + { + "epoch": 1.92, + "learning_rate": 2.8100751324054687e-06, + "loss": 0.6314, + "step": 5190 + }, + { + "epoch": 1.92, + "learning_rate": 2.8100381820421234e-06, + "loss": 0.6282, + "step": 5191 + }, + { + "epoch": 1.92, + "learning_rate": 2.8100012316787784e-06, + "loss": 0.7355, + "step": 5192 + }, + { + "epoch": 1.92, + "learning_rate": 2.809964281315433e-06, + "loss": 0.5781, + "step": 5193 + }, + { + "epoch": 1.92, + "learning_rate": 2.8099273309520877e-06, + "loss": 0.6042, + "step": 5194 + }, + { + "epoch": 1.92, + "learning_rate": 2.8098903805887423e-06, + "loss": 0.6379, + "step": 5195 + }, + { + "epoch": 1.92, + "learning_rate": 2.8098534302253974e-06, + "loss": 0.5287, + "step": 5196 + }, + { + "epoch": 1.92, + "learning_rate": 2.809816479862052e-06, + "loss": 0.5354, + "step": 5197 + }, + { + "epoch": 1.92, + "learning_rate": 2.8097795294987067e-06, + "loss": 0.5608, + "step": 5198 + }, + { + "epoch": 1.92, + "learning_rate": 2.8097425791353617e-06, + "loss": 0.6577, + "step": 5199 + }, + { + "epoch": 1.92, + "learning_rate": 2.8097056287720164e-06, + "loss": 0.6454, + "step": 5200 + }, + { + "epoch": 1.92, + "learning_rate": 2.809668678408671e-06, + "loss": 0.5422, + "step": 5201 + }, + { + "epoch": 1.92, + "learning_rate": 2.8096317280453256e-06, + "loss": 0.5405, + "step": 5202 + }, + { + "epoch": 1.92, + "learning_rate": 2.8095947776819807e-06, + "loss": 0.6137, + "step": 5203 + }, + { + "epoch": 1.92, + "learning_rate": 2.8095578273186353e-06, + "loss": 0.5897, + "step": 5204 + }, + { + "epoch": 1.92, + "learning_rate": 2.8095208769552904e-06, + "loss": 0.6008, + "step": 5205 + }, + { + "epoch": 1.92, + "learning_rate": 2.809483926591945e-06, + "loss": 0.6759, + "step": 5206 + }, + { + "epoch": 1.92, + "learning_rate": 2.8094469762285997e-06, + "loss": 0.6132, + "step": 5207 + }, + { + "epoch": 1.92, + "learning_rate": 2.8094100258652543e-06, + "loss": 0.6306, + "step": 5208 + }, + { + "epoch": 1.92, + "learning_rate": 2.809373075501909e-06, + "loss": 0.5472, + "step": 5209 + }, + { + "epoch": 1.92, + "learning_rate": 2.809336125138564e-06, + "loss": 0.6872, + "step": 5210 + }, + { + "epoch": 1.92, + "learning_rate": 2.8092991747752186e-06, + "loss": 0.6329, + "step": 5211 + }, + { + "epoch": 1.92, + "learning_rate": 2.8092622244118737e-06, + "loss": 0.6025, + "step": 5212 + }, + { + "epoch": 1.93, + "learning_rate": 2.8092252740485283e-06, + "loss": 0.6418, + "step": 5213 + }, + { + "epoch": 1.93, + "learning_rate": 2.809188323685183e-06, + "loss": 0.5938, + "step": 5214 + }, + { + "epoch": 1.93, + "learning_rate": 2.8091513733218376e-06, + "loss": 0.638, + "step": 5215 + }, + { + "epoch": 1.93, + "learning_rate": 2.8091144229584927e-06, + "loss": 0.7062, + "step": 5216 + }, + { + "epoch": 1.93, + "learning_rate": 2.8090774725951473e-06, + "loss": 0.5994, + "step": 5217 + }, + { + "epoch": 1.93, + "learning_rate": 2.809040522231802e-06, + "loss": 0.6014, + "step": 5218 + }, + { + "epoch": 1.93, + "learning_rate": 2.809003571868457e-06, + "loss": 0.6311, + "step": 5219 + }, + { + "epoch": 1.93, + "learning_rate": 2.8089666215051112e-06, + "loss": 0.6212, + "step": 5220 + }, + { + "epoch": 1.93, + "learning_rate": 2.8089296711417663e-06, + "loss": 0.6427, + "step": 5221 + }, + { + "epoch": 1.93, + "learning_rate": 2.808892720778421e-06, + "loss": 0.6301, + "step": 5222 + }, + { + "epoch": 1.93, + "learning_rate": 2.808855770415076e-06, + "loss": 0.6202, + "step": 5223 + }, + { + "epoch": 1.93, + "learning_rate": 2.8088188200517306e-06, + "loss": 0.5464, + "step": 5224 + }, + { + "epoch": 1.93, + "learning_rate": 2.8087818696883853e-06, + "loss": 0.5736, + "step": 5225 + }, + { + "epoch": 1.93, + "learning_rate": 2.80874491932504e-06, + "loss": 0.6669, + "step": 5226 + }, + { + "epoch": 1.93, + "learning_rate": 2.808707968961695e-06, + "loss": 0.5687, + "step": 5227 + }, + { + "epoch": 1.93, + "learning_rate": 2.8086710185983496e-06, + "loss": 0.5856, + "step": 5228 + }, + { + "epoch": 1.93, + "learning_rate": 2.8086340682350042e-06, + "loss": 0.5934, + "step": 5229 + }, + { + "epoch": 1.93, + "learning_rate": 2.8085971178716593e-06, + "loss": 0.6272, + "step": 5230 + }, + { + "epoch": 1.93, + "learning_rate": 2.808560167508314e-06, + "loss": 0.6362, + "step": 5231 + }, + { + "epoch": 1.93, + "learning_rate": 2.8085232171449686e-06, + "loss": 0.4905, + "step": 5232 + }, + { + "epoch": 1.93, + "learning_rate": 2.808486266781623e-06, + "loss": 0.6661, + "step": 5233 + }, + { + "epoch": 1.93, + "learning_rate": 2.8084493164182782e-06, + "loss": 0.6262, + "step": 5234 + }, + { + "epoch": 1.93, + "learning_rate": 2.808412366054933e-06, + "loss": 0.6182, + "step": 5235 + }, + { + "epoch": 1.93, + "learning_rate": 2.808375415691588e-06, + "loss": 0.6448, + "step": 5236 + }, + { + "epoch": 1.93, + "learning_rate": 2.8083384653282426e-06, + "loss": 0.5953, + "step": 5237 + }, + { + "epoch": 1.93, + "learning_rate": 2.8083015149648972e-06, + "loss": 0.6522, + "step": 5238 + }, + { + "epoch": 1.93, + "learning_rate": 2.808264564601552e-06, + "loss": 0.5971, + "step": 5239 + }, + { + "epoch": 1.94, + "learning_rate": 2.8082276142382065e-06, + "loss": 0.5859, + "step": 5240 + }, + { + "epoch": 1.94, + "learning_rate": 2.8081906638748616e-06, + "loss": 0.6347, + "step": 5241 + }, + { + "epoch": 1.94, + "learning_rate": 2.808153713511516e-06, + "loss": 0.5238, + "step": 5242 + }, + { + "epoch": 1.94, + "learning_rate": 2.8081167631481712e-06, + "loss": 0.5966, + "step": 5243 + }, + { + "epoch": 1.94, + "learning_rate": 2.808079812784826e-06, + "loss": 0.599, + "step": 5244 + }, + { + "epoch": 1.94, + "learning_rate": 2.8080428624214805e-06, + "loss": 0.5709, + "step": 5245 + }, + { + "epoch": 1.94, + "learning_rate": 2.808005912058135e-06, + "loss": 0.5737, + "step": 5246 + }, + { + "epoch": 1.94, + "learning_rate": 2.8079689616947902e-06, + "loss": 0.5751, + "step": 5247 + }, + { + "epoch": 1.94, + "learning_rate": 2.807932011331445e-06, + "loss": 0.6744, + "step": 5248 + }, + { + "epoch": 1.94, + "learning_rate": 2.8078950609680995e-06, + "loss": 0.6787, + "step": 5249 + }, + { + "epoch": 1.94, + "learning_rate": 2.8078581106047545e-06, + "loss": 0.6492, + "step": 5250 + }, + { + "epoch": 1.94, + "learning_rate": 2.807821160241409e-06, + "loss": 0.609, + "step": 5251 + }, + { + "epoch": 1.94, + "learning_rate": 2.807784209878064e-06, + "loss": 0.6085, + "step": 5252 + }, + { + "epoch": 1.94, + "learning_rate": 2.8077472595147185e-06, + "loss": 0.6598, + "step": 5253 + }, + { + "epoch": 1.94, + "learning_rate": 2.8077103091513735e-06, + "loss": 0.6741, + "step": 5254 + }, + { + "epoch": 1.94, + "learning_rate": 2.807673358788028e-06, + "loss": 0.5443, + "step": 5255 + }, + { + "epoch": 1.94, + "learning_rate": 2.8076364084246832e-06, + "loss": 0.6225, + "step": 5256 + }, + { + "epoch": 1.94, + "learning_rate": 2.807599458061338e-06, + "loss": 0.6108, + "step": 5257 + }, + { + "epoch": 1.94, + "learning_rate": 2.807562507697992e-06, + "loss": 0.6448, + "step": 5258 + }, + { + "epoch": 1.94, + "learning_rate": 2.807525557334647e-06, + "loss": 0.6221, + "step": 5259 + }, + { + "epoch": 1.94, + "learning_rate": 2.8074886069713018e-06, + "loss": 0.498, + "step": 5260 + }, + { + "epoch": 1.94, + "learning_rate": 2.807451656607957e-06, + "loss": 0.6736, + "step": 5261 + }, + { + "epoch": 1.94, + "learning_rate": 2.8074147062446115e-06, + "loss": 0.519, + "step": 5262 + }, + { + "epoch": 1.94, + "learning_rate": 2.8073777558812665e-06, + "loss": 0.6114, + "step": 5263 + }, + { + "epoch": 1.94, + "learning_rate": 2.8073408055179207e-06, + "loss": 0.6519, + "step": 5264 + }, + { + "epoch": 1.94, + "learning_rate": 2.807303855154576e-06, + "loss": 0.6419, + "step": 5265 + }, + { + "epoch": 1.94, + "learning_rate": 2.8072669047912304e-06, + "loss": 0.6143, + "step": 5266 + }, + { + "epoch": 1.94, + "learning_rate": 2.8072299544278855e-06, + "loss": 0.6457, + "step": 5267 + }, + { + "epoch": 1.95, + "learning_rate": 2.80719300406454e-06, + "loss": 0.664, + "step": 5268 + }, + { + "epoch": 1.95, + "learning_rate": 2.8071560537011948e-06, + "loss": 0.6117, + "step": 5269 + }, + { + "epoch": 1.95, + "learning_rate": 2.8071191033378494e-06, + "loss": 0.6603, + "step": 5270 + }, + { + "epoch": 1.95, + "learning_rate": 2.807082152974504e-06, + "loss": 0.6368, + "step": 5271 + }, + { + "epoch": 1.95, + "learning_rate": 2.807045202611159e-06, + "loss": 0.612, + "step": 5272 + }, + { + "epoch": 1.95, + "learning_rate": 2.8070082522478137e-06, + "loss": 0.58, + "step": 5273 + }, + { + "epoch": 1.95, + "learning_rate": 2.806971301884469e-06, + "loss": 0.6039, + "step": 5274 + }, + { + "epoch": 1.95, + "learning_rate": 2.8069343515211234e-06, + "loss": 0.5959, + "step": 5275 + }, + { + "epoch": 1.95, + "learning_rate": 2.806897401157778e-06, + "loss": 0.5608, + "step": 5276 + }, + { + "epoch": 1.95, + "learning_rate": 2.8068604507944327e-06, + "loss": 0.6021, + "step": 5277 + }, + { + "epoch": 1.95, + "learning_rate": 2.8068235004310873e-06, + "loss": 0.6327, + "step": 5278 + }, + { + "epoch": 1.95, + "learning_rate": 2.8067865500677424e-06, + "loss": 0.6348, + "step": 5279 + }, + { + "epoch": 1.95, + "learning_rate": 2.806749599704397e-06, + "loss": 0.5879, + "step": 5280 + }, + { + "epoch": 1.95, + "learning_rate": 2.806712649341052e-06, + "loss": 0.6683, + "step": 5281 + }, + { + "epoch": 1.95, + "learning_rate": 2.8066756989777067e-06, + "loss": 0.55, + "step": 5282 + }, + { + "epoch": 1.95, + "learning_rate": 2.8066387486143614e-06, + "loss": 0.5099, + "step": 5283 + }, + { + "epoch": 1.95, + "learning_rate": 2.806601798251016e-06, + "loss": 0.5847, + "step": 5284 + }, + { + "epoch": 1.95, + "learning_rate": 2.806564847887671e-06, + "loss": 0.5451, + "step": 5285 + }, + { + "epoch": 1.95, + "learning_rate": 2.8065278975243257e-06, + "loss": 0.6013, + "step": 5286 + }, + { + "epoch": 1.95, + "learning_rate": 2.8064909471609808e-06, + "loss": 0.5861, + "step": 5287 + }, + { + "epoch": 1.95, + "learning_rate": 2.8064539967976354e-06, + "loss": 0.5306, + "step": 5288 + }, + { + "epoch": 1.95, + "learning_rate": 2.80641704643429e-06, + "loss": 0.593, + "step": 5289 + }, + { + "epoch": 1.95, + "learning_rate": 2.8063800960709447e-06, + "loss": 0.6271, + "step": 5290 + }, + { + "epoch": 1.95, + "learning_rate": 2.8063431457075993e-06, + "loss": 0.6694, + "step": 5291 + }, + { + "epoch": 1.95, + "learning_rate": 2.8063061953442544e-06, + "loss": 0.6608, + "step": 5292 + }, + { + "epoch": 1.95, + "learning_rate": 2.806269244980909e-06, + "loss": 0.6713, + "step": 5293 + }, + { + "epoch": 1.95, + "learning_rate": 2.806232294617564e-06, + "loss": 0.6787, + "step": 5294 + }, + { + "epoch": 1.96, + "learning_rate": 2.8061953442542187e-06, + "loss": 0.6237, + "step": 5295 + }, + { + "epoch": 1.96, + "learning_rate": 2.8061583938908733e-06, + "loss": 0.6207, + "step": 5296 + }, + { + "epoch": 1.96, + "learning_rate": 2.806121443527528e-06, + "loss": 0.5749, + "step": 5297 + }, + { + "epoch": 1.96, + "learning_rate": 2.8060844931641826e-06, + "loss": 0.6103, + "step": 5298 + }, + { + "epoch": 1.96, + "learning_rate": 2.8060475428008377e-06, + "loss": 0.5687, + "step": 5299 + }, + { + "epoch": 1.96, + "learning_rate": 2.8060105924374923e-06, + "loss": 0.6339, + "step": 5300 + }, + { + "epoch": 1.96, + "learning_rate": 2.8059736420741474e-06, + "loss": 0.7155, + "step": 5301 + }, + { + "epoch": 1.96, + "learning_rate": 2.805936691710802e-06, + "loss": 0.6072, + "step": 5302 + }, + { + "epoch": 1.96, + "learning_rate": 2.8058997413474566e-06, + "loss": 0.6611, + "step": 5303 + }, + { + "epoch": 1.96, + "learning_rate": 2.8058627909841113e-06, + "loss": 0.572, + "step": 5304 + }, + { + "epoch": 1.96, + "learning_rate": 2.8058258406207663e-06, + "loss": 0.55, + "step": 5305 + }, + { + "epoch": 1.96, + "learning_rate": 2.805788890257421e-06, + "loss": 0.6197, + "step": 5306 + }, + { + "epoch": 1.96, + "learning_rate": 2.805751939894076e-06, + "loss": 0.6087, + "step": 5307 + }, + { + "epoch": 1.96, + "learning_rate": 2.8057149895307307e-06, + "loss": 0.6705, + "step": 5308 + }, + { + "epoch": 1.96, + "learning_rate": 2.805678039167385e-06, + "loss": 0.6072, + "step": 5309 + }, + { + "epoch": 1.96, + "learning_rate": 2.80564108880404e-06, + "loss": 0.583, + "step": 5310 + }, + { + "epoch": 1.96, + "learning_rate": 2.8056041384406946e-06, + "loss": 0.5937, + "step": 5311 + }, + { + "epoch": 1.96, + "learning_rate": 2.8055671880773496e-06, + "loss": 0.6082, + "step": 5312 + }, + { + "epoch": 1.96, + "learning_rate": 2.8055302377140043e-06, + "loss": 0.7022, + "step": 5313 + }, + { + "epoch": 1.96, + "learning_rate": 2.8054932873506593e-06, + "loss": 0.6522, + "step": 5314 + }, + { + "epoch": 1.96, + "learning_rate": 2.8054563369873135e-06, + "loss": 0.6386, + "step": 5315 + }, + { + "epoch": 1.96, + "learning_rate": 2.8054193866239686e-06, + "loss": 0.5893, + "step": 5316 + }, + { + "epoch": 1.96, + "learning_rate": 2.8053824362606232e-06, + "loss": 0.6326, + "step": 5317 + }, + { + "epoch": 1.96, + "learning_rate": 2.805345485897278e-06, + "loss": 0.5623, + "step": 5318 + }, + { + "epoch": 1.96, + "learning_rate": 2.805308535533933e-06, + "loss": 0.6161, + "step": 5319 + }, + { + "epoch": 1.96, + "learning_rate": 2.8052715851705876e-06, + "loss": 0.6049, + "step": 5320 + }, + { + "epoch": 1.96, + "learning_rate": 2.805234634807242e-06, + "loss": 0.673, + "step": 5321 + }, + { + "epoch": 1.97, + "learning_rate": 2.805197684443897e-06, + "loss": 0.6708, + "step": 5322 + }, + { + "epoch": 1.97, + "learning_rate": 2.805160734080552e-06, + "loss": 0.5612, + "step": 5323 + }, + { + "epoch": 1.97, + "learning_rate": 2.8051237837172065e-06, + "loss": 0.6137, + "step": 5324 + }, + { + "epoch": 1.97, + "learning_rate": 2.8050868333538616e-06, + "loss": 0.6473, + "step": 5325 + }, + { + "epoch": 1.97, + "learning_rate": 2.8050498829905162e-06, + "loss": 0.5954, + "step": 5326 + }, + { + "epoch": 1.97, + "learning_rate": 2.805012932627171e-06, + "loss": 0.5439, + "step": 5327 + }, + { + "epoch": 1.97, + "learning_rate": 2.8049759822638255e-06, + "loss": 0.6023, + "step": 5328 + }, + { + "epoch": 1.97, + "learning_rate": 2.80493903190048e-06, + "loss": 0.5697, + "step": 5329 + }, + { + "epoch": 1.97, + "learning_rate": 2.804902081537135e-06, + "loss": 0.6135, + "step": 5330 + }, + { + "epoch": 1.97, + "learning_rate": 2.80486513117379e-06, + "loss": 0.6091, + "step": 5331 + }, + { + "epoch": 1.97, + "learning_rate": 2.804828180810445e-06, + "loss": 0.5986, + "step": 5332 + }, + { + "epoch": 1.97, + "learning_rate": 2.8047912304470995e-06, + "loss": 0.5667, + "step": 5333 + }, + { + "epoch": 1.97, + "learning_rate": 2.804754280083754e-06, + "loss": 0.6077, + "step": 5334 + }, + { + "epoch": 1.97, + "learning_rate": 2.804717329720409e-06, + "loss": 0.6363, + "step": 5335 + }, + { + "epoch": 1.97, + "learning_rate": 2.804680379357064e-06, + "loss": 0.5871, + "step": 5336 + }, + { + "epoch": 1.97, + "learning_rate": 2.8046434289937185e-06, + "loss": 0.6588, + "step": 5337 + }, + { + "epoch": 1.97, + "learning_rate": 2.804606478630373e-06, + "loss": 0.7014, + "step": 5338 + }, + { + "epoch": 1.97, + "learning_rate": 2.804569528267028e-06, + "loss": 0.5956, + "step": 5339 + }, + { + "epoch": 1.97, + "learning_rate": 2.804532577903683e-06, + "loss": 0.5866, + "step": 5340 + }, + { + "epoch": 1.97, + "learning_rate": 2.8044956275403375e-06, + "loss": 0.5501, + "step": 5341 + }, + { + "epoch": 1.97, + "learning_rate": 2.804458677176992e-06, + "loss": 0.5857, + "step": 5342 + }, + { + "epoch": 1.97, + "learning_rate": 2.804421726813647e-06, + "loss": 0.5945, + "step": 5343 + }, + { + "epoch": 1.97, + "learning_rate": 2.804384776450302e-06, + "loss": 0.5038, + "step": 5344 + }, + { + "epoch": 1.97, + "learning_rate": 2.804347826086957e-06, + "loss": 0.6167, + "step": 5345 + }, + { + "epoch": 1.97, + "learning_rate": 2.8043108757236115e-06, + "loss": 0.5946, + "step": 5346 + }, + { + "epoch": 1.97, + "learning_rate": 2.804273925360266e-06, + "loss": 0.7085, + "step": 5347 + }, + { + "epoch": 1.97, + "learning_rate": 2.8042369749969208e-06, + "loss": 0.689, + "step": 5348 + }, + { + "epoch": 1.98, + "learning_rate": 2.8042000246335754e-06, + "loss": 0.6378, + "step": 5349 + }, + { + "epoch": 1.98, + "learning_rate": 2.8041630742702305e-06, + "loss": 0.5694, + "step": 5350 + }, + { + "epoch": 1.98, + "learning_rate": 2.804126123906885e-06, + "loss": 0.6668, + "step": 5351 + }, + { + "epoch": 1.98, + "learning_rate": 2.80408917354354e-06, + "loss": 0.578, + "step": 5352 + }, + { + "epoch": 1.98, + "learning_rate": 2.804052223180195e-06, + "loss": 0.5541, + "step": 5353 + }, + { + "epoch": 1.98, + "learning_rate": 2.8040152728168494e-06, + "loss": 0.5641, + "step": 5354 + }, + { + "epoch": 1.98, + "learning_rate": 2.803978322453504e-06, + "loss": 0.6399, + "step": 5355 + }, + { + "epoch": 1.98, + "learning_rate": 2.803941372090159e-06, + "loss": 0.5757, + "step": 5356 + }, + { + "epoch": 1.98, + "learning_rate": 2.8039044217268138e-06, + "loss": 0.6215, + "step": 5357 + }, + { + "epoch": 1.98, + "learning_rate": 2.8038674713634684e-06, + "loss": 0.6793, + "step": 5358 + }, + { + "epoch": 1.98, + "learning_rate": 2.8038305210001235e-06, + "loss": 0.5831, + "step": 5359 + }, + { + "epoch": 1.98, + "learning_rate": 2.8037935706367777e-06, + "loss": 0.6815, + "step": 5360 + }, + { + "epoch": 1.98, + "learning_rate": 2.8037566202734327e-06, + "loss": 0.5944, + "step": 5361 + }, + { + "epoch": 1.98, + "learning_rate": 2.8037196699100874e-06, + "loss": 0.6354, + "step": 5362 + }, + { + "epoch": 1.98, + "learning_rate": 2.8036827195467424e-06, + "loss": 0.5519, + "step": 5363 + }, + { + "epoch": 1.98, + "learning_rate": 2.803645769183397e-06, + "loss": 0.7257, + "step": 5364 + }, + { + "epoch": 1.98, + "learning_rate": 2.803608818820052e-06, + "loss": 0.616, + "step": 5365 + }, + { + "epoch": 1.98, + "learning_rate": 2.8035718684567064e-06, + "loss": 0.6248, + "step": 5366 + }, + { + "epoch": 1.98, + "learning_rate": 2.8035349180933614e-06, + "loss": 0.6651, + "step": 5367 + }, + { + "epoch": 1.98, + "learning_rate": 2.803497967730016e-06, + "loss": 0.6514, + "step": 5368 + }, + { + "epoch": 1.98, + "learning_rate": 2.8034610173666707e-06, + "loss": 0.6106, + "step": 5369 + }, + { + "epoch": 1.98, + "learning_rate": 2.8034240670033257e-06, + "loss": 0.6879, + "step": 5370 + }, + { + "epoch": 1.98, + "learning_rate": 2.8033871166399804e-06, + "loss": 0.5377, + "step": 5371 + }, + { + "epoch": 1.98, + "learning_rate": 2.803350166276635e-06, + "loss": 0.5608, + "step": 5372 + }, + { + "epoch": 1.98, + "learning_rate": 2.8033132159132897e-06, + "loss": 0.6248, + "step": 5373 + }, + { + "epoch": 1.98, + "learning_rate": 2.8032762655499447e-06, + "loss": 0.601, + "step": 5374 + }, + { + "epoch": 1.98, + "learning_rate": 2.8032393151865994e-06, + "loss": 0.5707, + "step": 5375 + }, + { + "epoch": 1.99, + "learning_rate": 2.8032023648232544e-06, + "loss": 0.6197, + "step": 5376 + }, + { + "epoch": 1.99, + "learning_rate": 2.803165414459909e-06, + "loss": 0.597, + "step": 5377 + }, + { + "epoch": 1.99, + "learning_rate": 2.8031284640965637e-06, + "loss": 0.6169, + "step": 5378 + }, + { + "epoch": 1.99, + "learning_rate": 2.8030915137332183e-06, + "loss": 0.6114, + "step": 5379 + }, + { + "epoch": 1.99, + "learning_rate": 2.803054563369873e-06, + "loss": 0.6435, + "step": 5380 + }, + { + "epoch": 1.99, + "learning_rate": 2.803017613006528e-06, + "loss": 0.6205, + "step": 5381 + }, + { + "epoch": 1.99, + "learning_rate": 2.8029806626431827e-06, + "loss": 0.6292, + "step": 5382 + }, + { + "epoch": 1.99, + "learning_rate": 2.8029437122798377e-06, + "loss": 0.6343, + "step": 5383 + }, + { + "epoch": 1.99, + "learning_rate": 2.8029067619164923e-06, + "loss": 0.6228, + "step": 5384 + }, + { + "epoch": 1.99, + "learning_rate": 2.802869811553147e-06, + "loss": 0.5836, + "step": 5385 + }, + { + "epoch": 1.99, + "learning_rate": 2.8028328611898016e-06, + "loss": 0.631, + "step": 5386 + }, + { + "epoch": 1.99, + "learning_rate": 2.8027959108264563e-06, + "loss": 0.6426, + "step": 5387 + }, + { + "epoch": 1.99, + "learning_rate": 2.8027589604631113e-06, + "loss": 0.6358, + "step": 5388 + }, + { + "epoch": 1.99, + "learning_rate": 2.802722010099766e-06, + "loss": 0.5657, + "step": 5389 + }, + { + "epoch": 1.99, + "learning_rate": 2.802685059736421e-06, + "loss": 0.5855, + "step": 5390 + }, + { + "epoch": 1.99, + "learning_rate": 2.8026481093730757e-06, + "loss": 0.6674, + "step": 5391 + }, + { + "epoch": 1.99, + "learning_rate": 2.8026111590097303e-06, + "loss": 0.5954, + "step": 5392 + }, + { + "epoch": 1.99, + "learning_rate": 2.802574208646385e-06, + "loss": 0.5877, + "step": 5393 + }, + { + "epoch": 1.99, + "learning_rate": 2.80253725828304e-06, + "loss": 0.6284, + "step": 5394 + }, + { + "epoch": 1.99, + "learning_rate": 2.8025003079196946e-06, + "loss": 0.6532, + "step": 5395 + }, + { + "epoch": 1.99, + "learning_rate": 2.8024633575563497e-06, + "loss": 0.5536, + "step": 5396 + }, + { + "epoch": 1.99, + "learning_rate": 2.8024264071930043e-06, + "loss": 0.5326, + "step": 5397 + }, + { + "epoch": 1.99, + "learning_rate": 2.802389456829659e-06, + "loss": 0.6428, + "step": 5398 + }, + { + "epoch": 1.99, + "learning_rate": 2.8023525064663136e-06, + "loss": 0.6579, + "step": 5399 + }, + { + "epoch": 1.99, + "learning_rate": 2.8023155561029682e-06, + "loss": 0.5547, + "step": 5400 + }, + { + "epoch": 1.99, + "learning_rate": 2.8022786057396233e-06, + "loss": 0.5418, + "step": 5401 + }, + { + "epoch": 1.99, + "learning_rate": 2.802241655376278e-06, + "loss": 0.5092, + "step": 5402 + }, + { + "epoch": 2.0, + "learning_rate": 2.802204705012933e-06, + "loss": 0.6465, + "step": 5403 + }, + { + "epoch": 2.0, + "learning_rate": 2.8021677546495876e-06, + "loss": 0.6142, + "step": 5404 + }, + { + "epoch": 2.0, + "learning_rate": 2.8021308042862423e-06, + "loss": 0.6129, + "step": 5405 + }, + { + "epoch": 2.0, + "learning_rate": 2.802093853922897e-06, + "loss": 0.6309, + "step": 5406 + }, + { + "epoch": 2.0, + "learning_rate": 2.8020569035595515e-06, + "loss": 0.6123, + "step": 5407 + }, + { + "epoch": 2.0, + "learning_rate": 2.8020199531962066e-06, + "loss": 0.6657, + "step": 5408 + }, + { + "epoch": 2.0, + "learning_rate": 2.8019830028328612e-06, + "loss": 0.5732, + "step": 5409 + }, + { + "epoch": 2.0, + "learning_rate": 2.8019460524695163e-06, + "loss": 0.6167, + "step": 5410 + }, + { + "epoch": 2.0, + "learning_rate": 2.8019091021061705e-06, + "loss": 0.7006, + "step": 5411 + }, + { + "epoch": 2.0, + "learning_rate": 2.8018721517428256e-06, + "loss": 0.6158, + "step": 5412 + }, + { + "epoch": 2.0, + "learning_rate": 2.80183520137948e-06, + "loss": 0.5683, + "step": 5413 + }, + { + "epoch": 2.0, + "learning_rate": 2.8017982510161353e-06, + "loss": 0.6437, + "step": 5414 + }, + { + "epoch": 2.0, + "learning_rate": 2.80176130065279e-06, + "loss": 0.6448, + "step": 5415 + }, + { + "epoch": 2.0, + "learning_rate": 2.8017243502894445e-06, + "loss": 0.5267, + "step": 5416 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.6681211617490881, + "eval_loss": 0.6051530838012695, + "eval_runtime": 25.3258, + "eval_samples_per_second": 1710.273, + "eval_steps_per_second": 26.732, + "step": 5416 + }, + { + "epoch": 2.0, + "learning_rate": 2.801687399926099e-06, + "loss": 0.6801, + "step": 5417 + }, + { + "epoch": 2.0, + "learning_rate": 2.801650449562754e-06, + "loss": 0.6001, + "step": 5418 + }, + { + "epoch": 2.0, + "learning_rate": 2.801613499199409e-06, + "loss": 0.6276, + "step": 5419 + }, + { + "epoch": 2.0, + "learning_rate": 2.8015765488360635e-06, + "loss": 0.6301, + "step": 5420 + }, + { + "epoch": 2.0, + "learning_rate": 2.8015395984727186e-06, + "loss": 0.6028, + "step": 5421 + }, + { + "epoch": 2.0, + "learning_rate": 2.801502648109373e-06, + "loss": 0.5368, + "step": 5422 + }, + { + "epoch": 2.0, + "learning_rate": 2.801465697746028e-06, + "loss": 0.53, + "step": 5423 + }, + { + "epoch": 2.0, + "learning_rate": 2.8014287473826825e-06, + "loss": 0.5899, + "step": 5424 + }, + { + "epoch": 2.0, + "learning_rate": 2.8013917970193375e-06, + "loss": 0.5953, + "step": 5425 + }, + { + "epoch": 2.0, + "learning_rate": 2.801354846655992e-06, + "loss": 0.6092, + "step": 5426 + }, + { + "epoch": 2.0, + "learning_rate": 2.801317896292647e-06, + "loss": 0.5146, + "step": 5427 + }, + { + "epoch": 2.0, + "learning_rate": 2.801280945929302e-06, + "loss": 0.605, + "step": 5428 + }, + { + "epoch": 2.0, + "learning_rate": 2.8012439955659565e-06, + "loss": 0.6687, + "step": 5429 + }, + { + "epoch": 2.01, + "learning_rate": 2.801207045202611e-06, + "loss": 0.6178, + "step": 5430 + }, + { + "epoch": 2.01, + "learning_rate": 2.8011700948392658e-06, + "loss": 0.6121, + "step": 5431 + }, + { + "epoch": 2.01, + "learning_rate": 2.801133144475921e-06, + "loss": 0.6235, + "step": 5432 + }, + { + "epoch": 2.01, + "learning_rate": 2.8010961941125755e-06, + "loss": 0.5544, + "step": 5433 + }, + { + "epoch": 2.01, + "learning_rate": 2.8010592437492305e-06, + "loss": 0.6102, + "step": 5434 + }, + { + "epoch": 2.01, + "learning_rate": 2.801022293385885e-06, + "loss": 0.6221, + "step": 5435 + }, + { + "epoch": 2.01, + "learning_rate": 2.80098534302254e-06, + "loss": 0.5995, + "step": 5436 + }, + { + "epoch": 2.01, + "learning_rate": 2.8009483926591944e-06, + "loss": 0.6045, + "step": 5437 + }, + { + "epoch": 2.01, + "learning_rate": 2.800911442295849e-06, + "loss": 0.6283, + "step": 5438 + }, + { + "epoch": 2.01, + "learning_rate": 2.800874491932504e-06, + "loss": 0.5985, + "step": 5439 + }, + { + "epoch": 2.01, + "learning_rate": 2.8008375415691588e-06, + "loss": 0.5795, + "step": 5440 + }, + { + "epoch": 2.01, + "learning_rate": 2.800800591205814e-06, + "loss": 0.6495, + "step": 5441 + }, + { + "epoch": 2.01, + "learning_rate": 2.8007636408424685e-06, + "loss": 0.6602, + "step": 5442 + }, + { + "epoch": 2.01, + "learning_rate": 2.800726690479123e-06, + "loss": 0.6746, + "step": 5443 + }, + { + "epoch": 2.01, + "learning_rate": 2.8006897401157777e-06, + "loss": 0.582, + "step": 5444 + }, + { + "epoch": 2.01, + "learning_rate": 2.800652789752433e-06, + "loss": 0.6316, + "step": 5445 + }, + { + "epoch": 2.01, + "learning_rate": 2.8006158393890874e-06, + "loss": 0.6253, + "step": 5446 + }, + { + "epoch": 2.01, + "learning_rate": 2.800578889025742e-06, + "loss": 0.6007, + "step": 5447 + }, + { + "epoch": 2.01, + "learning_rate": 2.800541938662397e-06, + "loss": 0.5797, + "step": 5448 + }, + { + "epoch": 2.01, + "learning_rate": 2.8005049882990513e-06, + "loss": 0.611, + "step": 5449 + }, + { + "epoch": 2.01, + "learning_rate": 2.8004680379357064e-06, + "loss": 0.5466, + "step": 5450 + }, + { + "epoch": 2.01, + "learning_rate": 2.800431087572361e-06, + "loss": 0.6167, + "step": 5451 + }, + { + "epoch": 2.01, + "learning_rate": 2.800394137209016e-06, + "loss": 0.5514, + "step": 5452 + }, + { + "epoch": 2.01, + "learning_rate": 2.8003571868456707e-06, + "loss": 0.6254, + "step": 5453 + }, + { + "epoch": 2.01, + "learning_rate": 2.800320236482326e-06, + "loss": 0.5953, + "step": 5454 + }, + { + "epoch": 2.01, + "learning_rate": 2.80028328611898e-06, + "loss": 0.5739, + "step": 5455 + }, + { + "epoch": 2.01, + "learning_rate": 2.800246335755635e-06, + "loss": 0.6459, + "step": 5456 + }, + { + "epoch": 2.02, + "learning_rate": 2.8002093853922897e-06, + "loss": 0.664, + "step": 5457 + }, + { + "epoch": 2.02, + "learning_rate": 2.8001724350289443e-06, + "loss": 0.5583, + "step": 5458 + }, + { + "epoch": 2.02, + "learning_rate": 2.8001354846655994e-06, + "loss": 0.526, + "step": 5459 + }, + { + "epoch": 2.02, + "learning_rate": 2.800098534302254e-06, + "loss": 0.5473, + "step": 5460 + }, + { + "epoch": 2.02, + "learning_rate": 2.8000615839389087e-06, + "loss": 0.6117, + "step": 5461 + }, + { + "epoch": 2.02, + "learning_rate": 2.8000246335755633e-06, + "loss": 0.6234, + "step": 5462 + }, + { + "epoch": 2.02, + "learning_rate": 2.7999876832122184e-06, + "loss": 0.626, + "step": 5463 + }, + { + "epoch": 2.02, + "learning_rate": 2.799950732848873e-06, + "loss": 0.5777, + "step": 5464 + }, + { + "epoch": 2.02, + "learning_rate": 2.799913782485528e-06, + "loss": 0.549, + "step": 5465 + }, + { + "epoch": 2.02, + "learning_rate": 2.7998768321221827e-06, + "loss": 0.6076, + "step": 5466 + }, + { + "epoch": 2.02, + "learning_rate": 2.7998398817588373e-06, + "loss": 0.5715, + "step": 5467 + }, + { + "epoch": 2.02, + "learning_rate": 2.799802931395492e-06, + "loss": 0.6187, + "step": 5468 + }, + { + "epoch": 2.02, + "learning_rate": 2.7997659810321466e-06, + "loss": 0.5385, + "step": 5469 + }, + { + "epoch": 2.02, + "learning_rate": 2.7997290306688017e-06, + "loss": 0.5838, + "step": 5470 + }, + { + "epoch": 2.02, + "learning_rate": 2.7996920803054563e-06, + "loss": 0.5532, + "step": 5471 + }, + { + "epoch": 2.02, + "learning_rate": 2.7996551299421114e-06, + "loss": 0.5947, + "step": 5472 + }, + { + "epoch": 2.02, + "learning_rate": 2.799618179578766e-06, + "loss": 0.5963, + "step": 5473 + }, + { + "epoch": 2.02, + "learning_rate": 2.7995812292154206e-06, + "loss": 0.5059, + "step": 5474 + }, + { + "epoch": 2.02, + "learning_rate": 2.7995442788520753e-06, + "loss": 0.7516, + "step": 5475 + }, + { + "epoch": 2.02, + "learning_rate": 2.7995073284887303e-06, + "loss": 0.6222, + "step": 5476 + }, + { + "epoch": 2.02, + "learning_rate": 2.799470378125385e-06, + "loss": 0.6269, + "step": 5477 + }, + { + "epoch": 2.02, + "learning_rate": 2.7994334277620396e-06, + "loss": 0.5667, + "step": 5478 + }, + { + "epoch": 2.02, + "learning_rate": 2.7993964773986947e-06, + "loss": 0.5083, + "step": 5479 + }, + { + "epoch": 2.02, + "learning_rate": 2.7993595270353493e-06, + "loss": 0.5569, + "step": 5480 + }, + { + "epoch": 2.02, + "learning_rate": 2.799322576672004e-06, + "loss": 0.6916, + "step": 5481 + }, + { + "epoch": 2.02, + "learning_rate": 2.7992856263086586e-06, + "loss": 0.6677, + "step": 5482 + }, + { + "epoch": 2.02, + "learning_rate": 2.7992486759453136e-06, + "loss": 0.6108, + "step": 5483 + }, + { + "epoch": 2.03, + "learning_rate": 2.7992117255819683e-06, + "loss": 0.6561, + "step": 5484 + }, + { + "epoch": 2.03, + "learning_rate": 2.7991747752186233e-06, + "loss": 0.6096, + "step": 5485 + }, + { + "epoch": 2.03, + "learning_rate": 2.799137824855278e-06, + "loss": 0.5448, + "step": 5486 + }, + { + "epoch": 2.03, + "learning_rate": 2.7991008744919326e-06, + "loss": 0.5079, + "step": 5487 + }, + { + "epoch": 2.03, + "learning_rate": 2.7990639241285872e-06, + "loss": 0.691, + "step": 5488 + }, + { + "epoch": 2.03, + "learning_rate": 2.799026973765242e-06, + "loss": 0.5925, + "step": 5489 + }, + { + "epoch": 2.03, + "learning_rate": 2.798990023401897e-06, + "loss": 0.5506, + "step": 5490 + }, + { + "epoch": 2.03, + "learning_rate": 2.7989530730385516e-06, + "loss": 0.4803, + "step": 5491 + }, + { + "epoch": 2.03, + "learning_rate": 2.7989161226752066e-06, + "loss": 0.5749, + "step": 5492 + }, + { + "epoch": 2.03, + "learning_rate": 2.7988791723118613e-06, + "loss": 0.6511, + "step": 5493 + }, + { + "epoch": 2.03, + "learning_rate": 2.798842221948516e-06, + "loss": 0.6142, + "step": 5494 + }, + { + "epoch": 2.03, + "learning_rate": 2.7988052715851705e-06, + "loss": 0.5263, + "step": 5495 + }, + { + "epoch": 2.03, + "learning_rate": 2.7987683212218256e-06, + "loss": 0.5341, + "step": 5496 + }, + { + "epoch": 2.03, + "learning_rate": 2.7987313708584802e-06, + "loss": 0.5419, + "step": 5497 + }, + { + "epoch": 2.03, + "learning_rate": 2.798694420495135e-06, + "loss": 0.6175, + "step": 5498 + }, + { + "epoch": 2.03, + "learning_rate": 2.79865747013179e-06, + "loss": 0.6156, + "step": 5499 + }, + { + "epoch": 2.03, + "learning_rate": 2.798620519768444e-06, + "loss": 0.5883, + "step": 5500 + }, + { + "epoch": 2.03, + "learning_rate": 2.798583569405099e-06, + "loss": 0.5676, + "step": 5501 + }, + { + "epoch": 2.03, + "learning_rate": 2.798546619041754e-06, + "loss": 0.533, + "step": 5502 + }, + { + "epoch": 2.03, + "learning_rate": 2.798509668678409e-06, + "loss": 0.5353, + "step": 5503 + }, + { + "epoch": 2.03, + "learning_rate": 2.7984727183150635e-06, + "loss": 0.6559, + "step": 5504 + }, + { + "epoch": 2.03, + "learning_rate": 2.7984357679517186e-06, + "loss": 0.6971, + "step": 5505 + }, + { + "epoch": 2.03, + "learning_rate": 2.798398817588373e-06, + "loss": 0.6693, + "step": 5506 + }, + { + "epoch": 2.03, + "learning_rate": 2.7983618672250275e-06, + "loss": 0.5263, + "step": 5507 + }, + { + "epoch": 2.03, + "learning_rate": 2.7983249168616825e-06, + "loss": 0.6463, + "step": 5508 + }, + { + "epoch": 2.03, + "learning_rate": 2.798287966498337e-06, + "loss": 0.6066, + "step": 5509 + }, + { + "epoch": 2.03, + "learning_rate": 2.798251016134992e-06, + "loss": 0.5163, + "step": 5510 + }, + { + "epoch": 2.04, + "learning_rate": 2.798214065771647e-06, + "loss": 0.5725, + "step": 5511 + }, + { + "epoch": 2.04, + "learning_rate": 2.7981771154083015e-06, + "loss": 0.5788, + "step": 5512 + }, + { + "epoch": 2.04, + "learning_rate": 2.798140165044956e-06, + "loss": 0.532, + "step": 5513 + }, + { + "epoch": 2.04, + "learning_rate": 2.798103214681611e-06, + "loss": 0.4918, + "step": 5514 + }, + { + "epoch": 2.04, + "learning_rate": 2.798066264318266e-06, + "loss": 0.5149, + "step": 5515 + }, + { + "epoch": 2.04, + "learning_rate": 2.798029313954921e-06, + "loss": 0.6519, + "step": 5516 + }, + { + "epoch": 2.04, + "learning_rate": 2.7979923635915755e-06, + "loss": 0.6378, + "step": 5517 + }, + { + "epoch": 2.04, + "learning_rate": 2.79795541322823e-06, + "loss": 0.6328, + "step": 5518 + }, + { + "epoch": 2.04, + "learning_rate": 2.7979184628648848e-06, + "loss": 0.5523, + "step": 5519 + }, + { + "epoch": 2.04, + "learning_rate": 2.7978815125015394e-06, + "loss": 0.6481, + "step": 5520 + }, + { + "epoch": 2.04, + "learning_rate": 2.7978445621381945e-06, + "loss": 0.6286, + "step": 5521 + }, + { + "epoch": 2.04, + "learning_rate": 2.797807611774849e-06, + "loss": 0.6493, + "step": 5522 + }, + { + "epoch": 2.04, + "learning_rate": 2.797770661411504e-06, + "loss": 0.5615, + "step": 5523 + }, + { + "epoch": 2.04, + "learning_rate": 2.797733711048159e-06, + "loss": 0.5817, + "step": 5524 + }, + { + "epoch": 2.04, + "learning_rate": 2.7976967606848134e-06, + "loss": 0.528, + "step": 5525 + }, + { + "epoch": 2.04, + "learning_rate": 2.797659810321468e-06, + "loss": 0.6099, + "step": 5526 + }, + { + "epoch": 2.04, + "learning_rate": 2.7976228599581227e-06, + "loss": 0.6445, + "step": 5527 + }, + { + "epoch": 2.04, + "learning_rate": 2.7975859095947778e-06, + "loss": 0.5881, + "step": 5528 + }, + { + "epoch": 2.04, + "learning_rate": 2.7975489592314324e-06, + "loss": 0.5704, + "step": 5529 + }, + { + "epoch": 2.04, + "learning_rate": 2.7975120088680875e-06, + "loss": 0.5697, + "step": 5530 + }, + { + "epoch": 2.04, + "learning_rate": 2.797475058504742e-06, + "loss": 0.605, + "step": 5531 + }, + { + "epoch": 2.04, + "learning_rate": 2.7974381081413968e-06, + "loss": 0.5807, + "step": 5532 + }, + { + "epoch": 2.04, + "learning_rate": 2.7974011577780514e-06, + "loss": 0.5294, + "step": 5533 + }, + { + "epoch": 2.04, + "learning_rate": 2.7973642074147064e-06, + "loss": 0.5971, + "step": 5534 + }, + { + "epoch": 2.04, + "learning_rate": 2.797327257051361e-06, + "loss": 0.5739, + "step": 5535 + }, + { + "epoch": 2.04, + "learning_rate": 2.797290306688016e-06, + "loss": 0.649, + "step": 5536 + }, + { + "epoch": 2.04, + "learning_rate": 2.7972533563246708e-06, + "loss": 0.5803, + "step": 5537 + }, + { + "epoch": 2.05, + "learning_rate": 2.7972164059613254e-06, + "loss": 0.5208, + "step": 5538 + }, + { + "epoch": 2.05, + "learning_rate": 2.79717945559798e-06, + "loss": 0.5947, + "step": 5539 + }, + { + "epoch": 2.05, + "learning_rate": 2.7971425052346347e-06, + "loss": 0.6089, + "step": 5540 + }, + { + "epoch": 2.05, + "learning_rate": 2.7971055548712898e-06, + "loss": 0.5784, + "step": 5541 + }, + { + "epoch": 2.05, + "learning_rate": 2.7970686045079444e-06, + "loss": 0.5764, + "step": 5542 + }, + { + "epoch": 2.05, + "learning_rate": 2.7970316541445994e-06, + "loss": 0.4907, + "step": 5543 + }, + { + "epoch": 2.05, + "learning_rate": 2.796994703781254e-06, + "loss": 0.6319, + "step": 5544 + }, + { + "epoch": 2.05, + "learning_rate": 2.7969577534179087e-06, + "loss": 0.576, + "step": 5545 + }, + { + "epoch": 2.05, + "learning_rate": 2.7969208030545634e-06, + "loss": 0.6336, + "step": 5546 + }, + { + "epoch": 2.05, + "learning_rate": 2.796883852691218e-06, + "loss": 0.5683, + "step": 5547 + }, + { + "epoch": 2.05, + "learning_rate": 2.796846902327873e-06, + "loss": 0.556, + "step": 5548 + }, + { + "epoch": 2.05, + "learning_rate": 2.7968099519645277e-06, + "loss": 0.6208, + "step": 5549 + }, + { + "epoch": 2.05, + "learning_rate": 2.7967730016011827e-06, + "loss": 0.6337, + "step": 5550 + }, + { + "epoch": 2.05, + "learning_rate": 2.796736051237837e-06, + "loss": 0.572, + "step": 5551 + }, + { + "epoch": 2.05, + "learning_rate": 2.796699100874492e-06, + "loss": 0.5497, + "step": 5552 + }, + { + "epoch": 2.05, + "learning_rate": 2.7966621505111467e-06, + "loss": 0.614, + "step": 5553 + }, + { + "epoch": 2.05, + "learning_rate": 2.7966252001478017e-06, + "loss": 0.5126, + "step": 5554 + }, + { + "epoch": 2.05, + "learning_rate": 2.7965882497844564e-06, + "loss": 0.6278, + "step": 5555 + }, + { + "epoch": 2.05, + "learning_rate": 2.7965512994211114e-06, + "loss": 0.5636, + "step": 5556 + }, + { + "epoch": 2.05, + "learning_rate": 2.7965143490577656e-06, + "loss": 0.6991, + "step": 5557 + }, + { + "epoch": 2.05, + "learning_rate": 2.7964773986944203e-06, + "loss": 0.5547, + "step": 5558 + }, + { + "epoch": 2.05, + "learning_rate": 2.7964404483310753e-06, + "loss": 0.6055, + "step": 5559 + }, + { + "epoch": 2.05, + "learning_rate": 2.79640349796773e-06, + "loss": 0.509, + "step": 5560 + }, + { + "epoch": 2.05, + "learning_rate": 2.796366547604385e-06, + "loss": 0.591, + "step": 5561 + }, + { + "epoch": 2.05, + "learning_rate": 2.7963295972410397e-06, + "loss": 0.6799, + "step": 5562 + }, + { + "epoch": 2.05, + "learning_rate": 2.7962926468776943e-06, + "loss": 0.6578, + "step": 5563 + }, + { + "epoch": 2.05, + "learning_rate": 2.796255696514349e-06, + "loss": 0.6878, + "step": 5564 + }, + { + "epoch": 2.06, + "learning_rate": 2.796218746151004e-06, + "loss": 0.5863, + "step": 5565 + }, + { + "epoch": 2.06, + "learning_rate": 2.7961817957876586e-06, + "loss": 0.5825, + "step": 5566 + }, + { + "epoch": 2.06, + "learning_rate": 2.7961448454243133e-06, + "loss": 0.5708, + "step": 5567 + }, + { + "epoch": 2.06, + "learning_rate": 2.7961078950609683e-06, + "loss": 0.6532, + "step": 5568 + }, + { + "epoch": 2.06, + "learning_rate": 2.796070944697623e-06, + "loss": 0.5695, + "step": 5569 + }, + { + "epoch": 2.06, + "learning_rate": 2.7960339943342776e-06, + "loss": 0.5922, + "step": 5570 + }, + { + "epoch": 2.06, + "learning_rate": 2.7959970439709322e-06, + "loss": 0.5778, + "step": 5571 + }, + { + "epoch": 2.06, + "learning_rate": 2.7959600936075873e-06, + "loss": 0.7098, + "step": 5572 + }, + { + "epoch": 2.06, + "learning_rate": 2.795923143244242e-06, + "loss": 0.5777, + "step": 5573 + }, + { + "epoch": 2.06, + "learning_rate": 2.795886192880897e-06, + "loss": 0.6148, + "step": 5574 + }, + { + "epoch": 2.06, + "learning_rate": 2.7958492425175516e-06, + "loss": 0.651, + "step": 5575 + }, + { + "epoch": 2.06, + "learning_rate": 2.7958122921542063e-06, + "loss": 0.5251, + "step": 5576 + }, + { + "epoch": 2.06, + "learning_rate": 2.795775341790861e-06, + "loss": 0.5509, + "step": 5577 + }, + { + "epoch": 2.06, + "learning_rate": 2.7957383914275155e-06, + "loss": 0.4887, + "step": 5578 + }, + { + "epoch": 2.06, + "learning_rate": 2.7957014410641706e-06, + "loss": 0.606, + "step": 5579 + }, + { + "epoch": 2.06, + "learning_rate": 2.7956644907008252e-06, + "loss": 0.6277, + "step": 5580 + }, + { + "epoch": 2.06, + "learning_rate": 2.7956275403374803e-06, + "loss": 0.5437, + "step": 5581 + }, + { + "epoch": 2.06, + "learning_rate": 2.795590589974135e-06, + "loss": 0.6482, + "step": 5582 + }, + { + "epoch": 2.06, + "learning_rate": 2.7955536396107896e-06, + "loss": 0.582, + "step": 5583 + }, + { + "epoch": 2.06, + "learning_rate": 2.795516689247444e-06, + "loss": 0.6466, + "step": 5584 + }, + { + "epoch": 2.06, + "learning_rate": 2.7954797388840993e-06, + "loss": 0.5917, + "step": 5585 + }, + { + "epoch": 2.06, + "learning_rate": 2.795442788520754e-06, + "loss": 0.6331, + "step": 5586 + }, + { + "epoch": 2.06, + "learning_rate": 2.7954058381574085e-06, + "loss": 0.6095, + "step": 5587 + }, + { + "epoch": 2.06, + "learning_rate": 2.7953688877940636e-06, + "loss": 0.6277, + "step": 5588 + }, + { + "epoch": 2.06, + "learning_rate": 2.7953319374307182e-06, + "loss": 0.5646, + "step": 5589 + }, + { + "epoch": 2.06, + "learning_rate": 2.795294987067373e-06, + "loss": 0.6031, + "step": 5590 + }, + { + "epoch": 2.06, + "learning_rate": 2.7952580367040275e-06, + "loss": 0.4958, + "step": 5591 + }, + { + "epoch": 2.06, + "learning_rate": 2.7952210863406826e-06, + "loss": 0.611, + "step": 5592 + }, + { + "epoch": 2.07, + "learning_rate": 2.795184135977337e-06, + "loss": 0.5098, + "step": 5593 + }, + { + "epoch": 2.07, + "learning_rate": 2.7951471856139923e-06, + "loss": 0.5533, + "step": 5594 + }, + { + "epoch": 2.07, + "learning_rate": 2.795110235250647e-06, + "loss": 0.6067, + "step": 5595 + }, + { + "epoch": 2.07, + "learning_rate": 2.7950732848873015e-06, + "loss": 0.6289, + "step": 5596 + }, + { + "epoch": 2.07, + "learning_rate": 2.795036334523956e-06, + "loss": 0.6142, + "step": 5597 + }, + { + "epoch": 2.07, + "learning_rate": 2.794999384160611e-06, + "loss": 0.6841, + "step": 5598 + }, + { + "epoch": 2.07, + "learning_rate": 2.794962433797266e-06, + "loss": 0.6546, + "step": 5599 + }, + { + "epoch": 2.07, + "learning_rate": 2.7949254834339205e-06, + "loss": 0.5551, + "step": 5600 + }, + { + "epoch": 2.07, + "learning_rate": 2.794888533070575e-06, + "loss": 0.602, + "step": 5601 + }, + { + "epoch": 2.07, + "learning_rate": 2.7948515827072298e-06, + "loss": 0.6345, + "step": 5602 + }, + { + "epoch": 2.07, + "learning_rate": 2.794814632343885e-06, + "loss": 0.6168, + "step": 5603 + }, + { + "epoch": 2.07, + "learning_rate": 2.7947776819805395e-06, + "loss": 0.5477, + "step": 5604 + }, + { + "epoch": 2.07, + "learning_rate": 2.7947407316171945e-06, + "loss": 0.6678, + "step": 5605 + }, + { + "epoch": 2.07, + "learning_rate": 2.794703781253849e-06, + "loss": 0.6733, + "step": 5606 + }, + { + "epoch": 2.07, + "learning_rate": 2.794666830890504e-06, + "loss": 0.5678, + "step": 5607 + }, + { + "epoch": 2.07, + "learning_rate": 2.7946298805271584e-06, + "loss": 0.5968, + "step": 5608 + }, + { + "epoch": 2.07, + "learning_rate": 2.794592930163813e-06, + "loss": 0.7559, + "step": 5609 + }, + { + "epoch": 2.07, + "learning_rate": 2.794555979800468e-06, + "loss": 0.5145, + "step": 5610 + }, + { + "epoch": 2.07, + "learning_rate": 2.7945190294371228e-06, + "loss": 0.5524, + "step": 5611 + }, + { + "epoch": 2.07, + "learning_rate": 2.794482079073778e-06, + "loss": 0.5393, + "step": 5612 + }, + { + "epoch": 2.07, + "learning_rate": 2.7944451287104325e-06, + "loss": 0.5737, + "step": 5613 + }, + { + "epoch": 2.07, + "learning_rate": 2.794408178347087e-06, + "loss": 0.5653, + "step": 5614 + }, + { + "epoch": 2.07, + "learning_rate": 2.7943712279837417e-06, + "loss": 0.7482, + "step": 5615 + }, + { + "epoch": 2.07, + "learning_rate": 2.794334277620397e-06, + "loss": 0.6512, + "step": 5616 + }, + { + "epoch": 2.07, + "learning_rate": 2.7942973272570514e-06, + "loss": 0.6159, + "step": 5617 + }, + { + "epoch": 2.07, + "learning_rate": 2.794260376893706e-06, + "loss": 0.6191, + "step": 5618 + }, + { + "epoch": 2.07, + "learning_rate": 2.794223426530361e-06, + "loss": 0.5781, + "step": 5619 + }, + { + "epoch": 2.08, + "learning_rate": 2.7941864761670158e-06, + "loss": 0.5492, + "step": 5620 + }, + { + "epoch": 2.08, + "learning_rate": 2.7941495258036704e-06, + "loss": 0.5694, + "step": 5621 + }, + { + "epoch": 2.08, + "learning_rate": 2.794112575440325e-06, + "loss": 0.5898, + "step": 5622 + }, + { + "epoch": 2.08, + "learning_rate": 2.79407562507698e-06, + "loss": 0.5882, + "step": 5623 + }, + { + "epoch": 2.08, + "learning_rate": 2.7940386747136347e-06, + "loss": 0.542, + "step": 5624 + }, + { + "epoch": 2.08, + "learning_rate": 2.79400172435029e-06, + "loss": 0.5628, + "step": 5625 + }, + { + "epoch": 2.08, + "learning_rate": 2.7939647739869444e-06, + "loss": 0.5018, + "step": 5626 + }, + { + "epoch": 2.08, + "learning_rate": 2.793927823623599e-06, + "loss": 0.6168, + "step": 5627 + }, + { + "epoch": 2.08, + "learning_rate": 2.7938908732602537e-06, + "loss": 0.5731, + "step": 5628 + }, + { + "epoch": 2.08, + "learning_rate": 2.7938539228969083e-06, + "loss": 0.6232, + "step": 5629 + }, + { + "epoch": 2.08, + "learning_rate": 2.7938169725335634e-06, + "loss": 0.5909, + "step": 5630 + }, + { + "epoch": 2.08, + "learning_rate": 2.793780022170218e-06, + "loss": 0.5827, + "step": 5631 + }, + { + "epoch": 2.08, + "learning_rate": 2.793743071806873e-06, + "loss": 0.5902, + "step": 5632 + }, + { + "epoch": 2.08, + "learning_rate": 2.7937061214435277e-06, + "loss": 0.5628, + "step": 5633 + }, + { + "epoch": 2.08, + "learning_rate": 2.7936691710801824e-06, + "loss": 0.5584, + "step": 5634 + }, + { + "epoch": 2.08, + "learning_rate": 2.793632220716837e-06, + "loss": 0.6058, + "step": 5635 + }, + { + "epoch": 2.08, + "learning_rate": 2.7935952703534916e-06, + "loss": 0.5631, + "step": 5636 + }, + { + "epoch": 2.08, + "learning_rate": 2.7935583199901467e-06, + "loss": 0.5854, + "step": 5637 + }, + { + "epoch": 2.08, + "learning_rate": 2.7935213696268013e-06, + "loss": 0.6194, + "step": 5638 + }, + { + "epoch": 2.08, + "learning_rate": 2.7934844192634564e-06, + "loss": 0.6956, + "step": 5639 + }, + { + "epoch": 2.08, + "learning_rate": 2.7934474689001106e-06, + "loss": 0.6412, + "step": 5640 + }, + { + "epoch": 2.08, + "learning_rate": 2.7934105185367657e-06, + "loss": 0.5675, + "step": 5641 + }, + { + "epoch": 2.08, + "learning_rate": 2.7933735681734203e-06, + "loss": 0.6407, + "step": 5642 + }, + { + "epoch": 2.08, + "learning_rate": 2.7933366178100754e-06, + "loss": 0.5387, + "step": 5643 + }, + { + "epoch": 2.08, + "learning_rate": 2.79329966744673e-06, + "loss": 0.6203, + "step": 5644 + }, + { + "epoch": 2.08, + "learning_rate": 2.793262717083385e-06, + "loss": 0.5083, + "step": 5645 + }, + { + "epoch": 2.08, + "learning_rate": 2.7932257667200393e-06, + "loss": 0.5774, + "step": 5646 + }, + { + "epoch": 2.09, + "learning_rate": 2.793188816356694e-06, + "loss": 0.5804, + "step": 5647 + }, + { + "epoch": 2.09, + "learning_rate": 2.793151865993349e-06, + "loss": 0.6037, + "step": 5648 + }, + { + "epoch": 2.09, + "learning_rate": 2.7931149156300036e-06, + "loss": 0.538, + "step": 5649 + }, + { + "epoch": 2.09, + "learning_rate": 2.7930779652666587e-06, + "loss": 0.6504, + "step": 5650 + }, + { + "epoch": 2.09, + "learning_rate": 2.7930410149033133e-06, + "loss": 0.6097, + "step": 5651 + }, + { + "epoch": 2.09, + "learning_rate": 2.793004064539968e-06, + "loss": 0.5976, + "step": 5652 + }, + { + "epoch": 2.09, + "learning_rate": 2.7929671141766226e-06, + "loss": 0.542, + "step": 5653 + }, + { + "epoch": 2.09, + "learning_rate": 2.7929301638132776e-06, + "loss": 0.6007, + "step": 5654 + }, + { + "epoch": 2.09, + "learning_rate": 2.7928932134499323e-06, + "loss": 0.5441, + "step": 5655 + }, + { + "epoch": 2.09, + "learning_rate": 2.792856263086587e-06, + "loss": 0.6031, + "step": 5656 + }, + { + "epoch": 2.09, + "learning_rate": 2.792819312723242e-06, + "loss": 0.5637, + "step": 5657 + }, + { + "epoch": 2.09, + "learning_rate": 2.7927823623598966e-06, + "loss": 0.6134, + "step": 5658 + }, + { + "epoch": 2.09, + "learning_rate": 2.7927454119965512e-06, + "loss": 0.5771, + "step": 5659 + }, + { + "epoch": 2.09, + "learning_rate": 2.792708461633206e-06, + "loss": 0.583, + "step": 5660 + }, + { + "epoch": 2.09, + "learning_rate": 2.792671511269861e-06, + "loss": 0.622, + "step": 5661 + }, + { + "epoch": 2.09, + "learning_rate": 2.7926345609065156e-06, + "loss": 0.6785, + "step": 5662 + }, + { + "epoch": 2.09, + "learning_rate": 2.7925976105431706e-06, + "loss": 0.5652, + "step": 5663 + }, + { + "epoch": 2.09, + "learning_rate": 2.7925606601798253e-06, + "loss": 0.574, + "step": 5664 + }, + { + "epoch": 2.09, + "learning_rate": 2.79252370981648e-06, + "loss": 0.632, + "step": 5665 + }, + { + "epoch": 2.09, + "learning_rate": 2.7924867594531346e-06, + "loss": 0.7062, + "step": 5666 + }, + { + "epoch": 2.09, + "learning_rate": 2.792449809089789e-06, + "loss": 0.5037, + "step": 5667 + }, + { + "epoch": 2.09, + "learning_rate": 2.7924128587264442e-06, + "loss": 0.5119, + "step": 5668 + }, + { + "epoch": 2.09, + "learning_rate": 2.792375908363099e-06, + "loss": 0.611, + "step": 5669 + }, + { + "epoch": 2.09, + "learning_rate": 2.792338957999754e-06, + "loss": 0.6275, + "step": 5670 + }, + { + "epoch": 2.09, + "learning_rate": 2.7923020076364086e-06, + "loss": 0.6562, + "step": 5671 + }, + { + "epoch": 2.09, + "learning_rate": 2.7922650572730632e-06, + "loss": 0.5277, + "step": 5672 + }, + { + "epoch": 2.09, + "learning_rate": 2.792228106909718e-06, + "loss": 0.5774, + "step": 5673 + }, + { + "epoch": 2.1, + "learning_rate": 2.792191156546373e-06, + "loss": 0.6997, + "step": 5674 + }, + { + "epoch": 2.1, + "learning_rate": 2.7921542061830275e-06, + "loss": 0.5983, + "step": 5675 + }, + { + "epoch": 2.1, + "learning_rate": 2.792117255819682e-06, + "loss": 0.6433, + "step": 5676 + }, + { + "epoch": 2.1, + "learning_rate": 2.7920803054563372e-06, + "loss": 0.5568, + "step": 5677 + }, + { + "epoch": 2.1, + "learning_rate": 2.792043355092992e-06, + "loss": 0.5747, + "step": 5678 + }, + { + "epoch": 2.1, + "learning_rate": 2.7920064047296465e-06, + "loss": 0.658, + "step": 5679 + }, + { + "epoch": 2.1, + "learning_rate": 2.791969454366301e-06, + "loss": 0.5365, + "step": 5680 + }, + { + "epoch": 2.1, + "learning_rate": 2.7919325040029562e-06, + "loss": 0.5626, + "step": 5681 + }, + { + "epoch": 2.1, + "learning_rate": 2.791895553639611e-06, + "loss": 0.4568, + "step": 5682 + }, + { + "epoch": 2.1, + "learning_rate": 2.791858603276266e-06, + "loss": 0.5563, + "step": 5683 + }, + { + "epoch": 2.1, + "learning_rate": 2.7918216529129205e-06, + "loss": 0.5949, + "step": 5684 + }, + { + "epoch": 2.1, + "learning_rate": 2.791784702549575e-06, + "loss": 0.5919, + "step": 5685 + }, + { + "epoch": 2.1, + "learning_rate": 2.79174775218623e-06, + "loss": 0.7028, + "step": 5686 + }, + { + "epoch": 2.1, + "learning_rate": 2.7917108018228845e-06, + "loss": 0.5662, + "step": 5687 + }, + { + "epoch": 2.1, + "learning_rate": 2.7916738514595395e-06, + "loss": 0.6532, + "step": 5688 + }, + { + "epoch": 2.1, + "learning_rate": 2.791636901096194e-06, + "loss": 0.6158, + "step": 5689 + }, + { + "epoch": 2.1, + "learning_rate": 2.7915999507328492e-06, + "loss": 0.5464, + "step": 5690 + }, + { + "epoch": 2.1, + "learning_rate": 2.7915630003695034e-06, + "loss": 0.5565, + "step": 5691 + }, + { + "epoch": 2.1, + "learning_rate": 2.7915260500061585e-06, + "loss": 0.5936, + "step": 5692 + }, + { + "epoch": 2.1, + "learning_rate": 2.791489099642813e-06, + "loss": 0.5791, + "step": 5693 + }, + { + "epoch": 2.1, + "learning_rate": 2.791452149279468e-06, + "loss": 0.5355, + "step": 5694 + }, + { + "epoch": 2.1, + "learning_rate": 2.791415198916123e-06, + "loss": 0.5148, + "step": 5695 + }, + { + "epoch": 2.1, + "learning_rate": 2.7913782485527775e-06, + "loss": 0.5116, + "step": 5696 + }, + { + "epoch": 2.1, + "learning_rate": 2.791341298189432e-06, + "loss": 0.6086, + "step": 5697 + }, + { + "epoch": 2.1, + "learning_rate": 2.7913043478260867e-06, + "loss": 0.5457, + "step": 5698 + }, + { + "epoch": 2.1, + "learning_rate": 2.791267397462742e-06, + "loss": 0.4782, + "step": 5699 + }, + { + "epoch": 2.1, + "learning_rate": 2.7912304470993964e-06, + "loss": 0.6476, + "step": 5700 + }, + { + "epoch": 2.11, + "learning_rate": 2.7911934967360515e-06, + "loss": 0.5779, + "step": 5701 + }, + { + "epoch": 2.11, + "learning_rate": 2.791156546372706e-06, + "loss": 0.5975, + "step": 5702 + }, + { + "epoch": 2.11, + "learning_rate": 2.7911195960093608e-06, + "loss": 0.4795, + "step": 5703 + }, + { + "epoch": 2.11, + "learning_rate": 2.7910826456460154e-06, + "loss": 0.6362, + "step": 5704 + }, + { + "epoch": 2.11, + "learning_rate": 2.7910456952826705e-06, + "loss": 0.6513, + "step": 5705 + }, + { + "epoch": 2.11, + "learning_rate": 2.791008744919325e-06, + "loss": 0.5584, + "step": 5706 + }, + { + "epoch": 2.11, + "learning_rate": 2.7909717945559797e-06, + "loss": 0.6369, + "step": 5707 + }, + { + "epoch": 2.11, + "learning_rate": 2.7909348441926348e-06, + "loss": 0.4862, + "step": 5708 + }, + { + "epoch": 2.11, + "learning_rate": 2.7908978938292894e-06, + "loss": 0.5721, + "step": 5709 + }, + { + "epoch": 2.11, + "learning_rate": 2.790860943465944e-06, + "loss": 0.6143, + "step": 5710 + }, + { + "epoch": 2.11, + "learning_rate": 2.7908239931025987e-06, + "loss": 0.5925, + "step": 5711 + }, + { + "epoch": 2.11, + "learning_rate": 2.7907870427392538e-06, + "loss": 0.5483, + "step": 5712 + }, + { + "epoch": 2.11, + "learning_rate": 2.7907500923759084e-06, + "loss": 0.5655, + "step": 5713 + }, + { + "epoch": 2.11, + "learning_rate": 2.7907131420125635e-06, + "loss": 0.5763, + "step": 5714 + }, + { + "epoch": 2.11, + "learning_rate": 2.790676191649218e-06, + "loss": 0.5608, + "step": 5715 + }, + { + "epoch": 2.11, + "learning_rate": 2.7906392412858727e-06, + "loss": 0.5615, + "step": 5716 + }, + { + "epoch": 2.11, + "learning_rate": 2.7906022909225274e-06, + "loss": 0.6183, + "step": 5717 + }, + { + "epoch": 2.11, + "learning_rate": 2.790565340559182e-06, + "loss": 0.5747, + "step": 5718 + }, + { + "epoch": 2.11, + "learning_rate": 2.790528390195837e-06, + "loss": 0.4682, + "step": 5719 + }, + { + "epoch": 2.11, + "learning_rate": 2.7904914398324917e-06, + "loss": 0.6754, + "step": 5720 + }, + { + "epoch": 2.11, + "learning_rate": 2.7904544894691468e-06, + "loss": 0.6339, + "step": 5721 + }, + { + "epoch": 2.11, + "learning_rate": 2.7904175391058014e-06, + "loss": 0.5505, + "step": 5722 + }, + { + "epoch": 2.11, + "learning_rate": 2.790380588742456e-06, + "loss": 0.6626, + "step": 5723 + }, + { + "epoch": 2.11, + "learning_rate": 2.7903436383791107e-06, + "loss": 0.609, + "step": 5724 + }, + { + "epoch": 2.11, + "learning_rate": 2.7903066880157657e-06, + "loss": 0.5698, + "step": 5725 + }, + { + "epoch": 2.11, + "learning_rate": 2.7902697376524204e-06, + "loss": 0.6433, + "step": 5726 + }, + { + "epoch": 2.11, + "learning_rate": 2.790232787289075e-06, + "loss": 0.5548, + "step": 5727 + }, + { + "epoch": 2.12, + "learning_rate": 2.79019583692573e-06, + "loss": 0.6329, + "step": 5728 + }, + { + "epoch": 2.12, + "learning_rate": 2.7901588865623847e-06, + "loss": 0.6321, + "step": 5729 + }, + { + "epoch": 2.12, + "learning_rate": 2.7901219361990393e-06, + "loss": 0.4903, + "step": 5730 + }, + { + "epoch": 2.12, + "learning_rate": 2.790084985835694e-06, + "loss": 0.5686, + "step": 5731 + }, + { + "epoch": 2.12, + "learning_rate": 2.790048035472349e-06, + "loss": 0.5959, + "step": 5732 + }, + { + "epoch": 2.12, + "learning_rate": 2.7900110851090037e-06, + "loss": 0.6065, + "step": 5733 + }, + { + "epoch": 2.12, + "learning_rate": 2.7899741347456587e-06, + "loss": 0.7152, + "step": 5734 + }, + { + "epoch": 2.12, + "learning_rate": 2.7899371843823134e-06, + "loss": 0.5818, + "step": 5735 + }, + { + "epoch": 2.12, + "learning_rate": 2.7899002340189676e-06, + "loss": 0.5896, + "step": 5736 + }, + { + "epoch": 2.12, + "learning_rate": 2.7898632836556226e-06, + "loss": 0.5514, + "step": 5737 + }, + { + "epoch": 2.12, + "learning_rate": 2.7898263332922773e-06, + "loss": 0.6432, + "step": 5738 + }, + { + "epoch": 2.12, + "learning_rate": 2.7897893829289323e-06, + "loss": 0.5986, + "step": 5739 + }, + { + "epoch": 2.12, + "learning_rate": 2.789752432565587e-06, + "loss": 0.6176, + "step": 5740 + }, + { + "epoch": 2.12, + "learning_rate": 2.789715482202242e-06, + "loss": 0.6298, + "step": 5741 + }, + { + "epoch": 2.12, + "learning_rate": 2.7896785318388962e-06, + "loss": 0.5896, + "step": 5742 + }, + { + "epoch": 2.12, + "learning_rate": 2.7896415814755513e-06, + "loss": 0.6169, + "step": 5743 + }, + { + "epoch": 2.12, + "learning_rate": 2.789604631112206e-06, + "loss": 0.51, + "step": 5744 + }, + { + "epoch": 2.12, + "learning_rate": 2.789567680748861e-06, + "loss": 0.6005, + "step": 5745 + }, + { + "epoch": 2.12, + "learning_rate": 2.7895307303855156e-06, + "loss": 0.663, + "step": 5746 + }, + { + "epoch": 2.12, + "learning_rate": 2.7894937800221703e-06, + "loss": 0.694, + "step": 5747 + }, + { + "epoch": 2.12, + "learning_rate": 2.789456829658825e-06, + "loss": 0.5829, + "step": 5748 + }, + { + "epoch": 2.12, + "learning_rate": 2.7894198792954795e-06, + "loss": 0.6033, + "step": 5749 + }, + { + "epoch": 2.12, + "learning_rate": 2.7893829289321346e-06, + "loss": 0.5526, + "step": 5750 + }, + { + "epoch": 2.12, + "learning_rate": 2.7893459785687892e-06, + "loss": 0.536, + "step": 5751 + }, + { + "epoch": 2.12, + "learning_rate": 2.7893090282054443e-06, + "loss": 0.6123, + "step": 5752 + }, + { + "epoch": 2.12, + "learning_rate": 2.789272077842099e-06, + "loss": 0.6895, + "step": 5753 + }, + { + "epoch": 2.12, + "learning_rate": 2.7892351274787536e-06, + "loss": 0.4608, + "step": 5754 + }, + { + "epoch": 2.13, + "learning_rate": 2.789198177115408e-06, + "loss": 0.6202, + "step": 5755 + }, + { + "epoch": 2.13, + "learning_rate": 2.789161226752063e-06, + "loss": 0.5698, + "step": 5756 + }, + { + "epoch": 2.13, + "learning_rate": 2.789124276388718e-06, + "loss": 0.5922, + "step": 5757 + }, + { + "epoch": 2.13, + "learning_rate": 2.7890873260253725e-06, + "loss": 0.6044, + "step": 5758 + }, + { + "epoch": 2.13, + "learning_rate": 2.7890503756620276e-06, + "loss": 0.5817, + "step": 5759 + }, + { + "epoch": 2.13, + "learning_rate": 2.7890134252986822e-06, + "loss": 0.6493, + "step": 5760 + }, + { + "epoch": 2.13, + "learning_rate": 2.788976474935337e-06, + "loss": 0.63, + "step": 5761 + }, + { + "epoch": 2.13, + "learning_rate": 2.7889395245719915e-06, + "loss": 0.6132, + "step": 5762 + }, + { + "epoch": 2.13, + "learning_rate": 2.7889025742086466e-06, + "loss": 0.5721, + "step": 5763 + }, + { + "epoch": 2.13, + "learning_rate": 2.788865623845301e-06, + "loss": 0.586, + "step": 5764 + }, + { + "epoch": 2.13, + "learning_rate": 2.7888286734819563e-06, + "loss": 0.59, + "step": 5765 + }, + { + "epoch": 2.13, + "learning_rate": 2.788791723118611e-06, + "loss": 0.6322, + "step": 5766 + }, + { + "epoch": 2.13, + "learning_rate": 2.7887547727552655e-06, + "loss": 0.6278, + "step": 5767 + }, + { + "epoch": 2.13, + "learning_rate": 2.78871782239192e-06, + "loss": 0.6313, + "step": 5768 + }, + { + "epoch": 2.13, + "learning_rate": 2.788680872028575e-06, + "loss": 0.6127, + "step": 5769 + }, + { + "epoch": 2.13, + "learning_rate": 2.78864392166523e-06, + "loss": 0.5643, + "step": 5770 + }, + { + "epoch": 2.13, + "learning_rate": 2.7886069713018845e-06, + "loss": 0.5635, + "step": 5771 + }, + { + "epoch": 2.13, + "learning_rate": 2.7885700209385396e-06, + "loss": 0.478, + "step": 5772 + }, + { + "epoch": 2.13, + "learning_rate": 2.788533070575194e-06, + "loss": 0.538, + "step": 5773 + }, + { + "epoch": 2.13, + "learning_rate": 2.788496120211849e-06, + "loss": 0.5673, + "step": 5774 + }, + { + "epoch": 2.13, + "learning_rate": 2.7884591698485035e-06, + "loss": 0.5398, + "step": 5775 + }, + { + "epoch": 2.13, + "learning_rate": 2.788422219485158e-06, + "loss": 0.5796, + "step": 5776 + }, + { + "epoch": 2.13, + "learning_rate": 2.788385269121813e-06, + "loss": 0.5522, + "step": 5777 + }, + { + "epoch": 2.13, + "learning_rate": 2.788348318758468e-06, + "loss": 0.6317, + "step": 5778 + }, + { + "epoch": 2.13, + "learning_rate": 2.788311368395123e-06, + "loss": 0.602, + "step": 5779 + }, + { + "epoch": 2.13, + "learning_rate": 2.7882744180317775e-06, + "loss": 0.6355, + "step": 5780 + }, + { + "epoch": 2.13, + "learning_rate": 2.788237467668432e-06, + "loss": 0.5414, + "step": 5781 + }, + { + "epoch": 2.14, + "learning_rate": 2.7882005173050868e-06, + "loss": 0.5695, + "step": 5782 + }, + { + "epoch": 2.14, + "learning_rate": 2.788163566941742e-06, + "loss": 0.6465, + "step": 5783 + }, + { + "epoch": 2.14, + "learning_rate": 2.7881266165783965e-06, + "loss": 0.5587, + "step": 5784 + }, + { + "epoch": 2.14, + "learning_rate": 2.7880896662150515e-06, + "loss": 0.5951, + "step": 5785 + }, + { + "epoch": 2.14, + "learning_rate": 2.7880527158517057e-06, + "loss": 0.6325, + "step": 5786 + }, + { + "epoch": 2.14, + "learning_rate": 2.7880157654883604e-06, + "loss": 0.5738, + "step": 5787 + }, + { + "epoch": 2.14, + "learning_rate": 2.7879788151250154e-06, + "loss": 0.6289, + "step": 5788 + }, + { + "epoch": 2.14, + "learning_rate": 2.78794186476167e-06, + "loss": 0.6377, + "step": 5789 + }, + { + "epoch": 2.14, + "learning_rate": 2.787904914398325e-06, + "loss": 0.5632, + "step": 5790 + }, + { + "epoch": 2.14, + "learning_rate": 2.7878679640349798e-06, + "loss": 0.5621, + "step": 5791 + }, + { + "epoch": 2.14, + "learning_rate": 2.7878310136716344e-06, + "loss": 0.6629, + "step": 5792 + }, + { + "epoch": 2.14, + "learning_rate": 2.787794063308289e-06, + "loss": 0.5452, + "step": 5793 + }, + { + "epoch": 2.14, + "learning_rate": 2.787757112944944e-06, + "loss": 0.6732, + "step": 5794 + }, + { + "epoch": 2.14, + "learning_rate": 2.7877201625815987e-06, + "loss": 0.6284, + "step": 5795 + }, + { + "epoch": 2.14, + "learning_rate": 2.7876832122182534e-06, + "loss": 0.6289, + "step": 5796 + }, + { + "epoch": 2.14, + "learning_rate": 2.7876462618549084e-06, + "loss": 0.5349, + "step": 5797 + }, + { + "epoch": 2.14, + "learning_rate": 2.787609311491563e-06, + "loss": 0.5373, + "step": 5798 + }, + { + "epoch": 2.14, + "learning_rate": 2.7875723611282177e-06, + "loss": 0.5523, + "step": 5799 + }, + { + "epoch": 2.14, + "learning_rate": 2.7875354107648724e-06, + "loss": 0.5966, + "step": 5800 + }, + { + "epoch": 2.14, + "learning_rate": 2.7874984604015274e-06, + "loss": 0.5904, + "step": 5801 + }, + { + "epoch": 2.14, + "learning_rate": 2.787461510038182e-06, + "loss": 0.5419, + "step": 5802 + }, + { + "epoch": 2.14, + "learning_rate": 2.787424559674837e-06, + "loss": 0.6494, + "step": 5803 + }, + { + "epoch": 2.14, + "learning_rate": 2.7873876093114917e-06, + "loss": 0.6136, + "step": 5804 + }, + { + "epoch": 2.14, + "learning_rate": 2.7873506589481464e-06, + "loss": 0.6194, + "step": 5805 + }, + { + "epoch": 2.14, + "learning_rate": 2.787313708584801e-06, + "loss": 0.5713, + "step": 5806 + }, + { + "epoch": 2.14, + "learning_rate": 2.7872767582214557e-06, + "loss": 0.5799, + "step": 5807 + }, + { + "epoch": 2.14, + "learning_rate": 2.7872398078581107e-06, + "loss": 0.6607, + "step": 5808 + }, + { + "epoch": 2.15, + "learning_rate": 2.7872028574947653e-06, + "loss": 0.5871, + "step": 5809 + }, + { + "epoch": 2.15, + "learning_rate": 2.7871659071314204e-06, + "loss": 0.5306, + "step": 5810 + }, + { + "epoch": 2.15, + "learning_rate": 2.787128956768075e-06, + "loss": 0.624, + "step": 5811 + }, + { + "epoch": 2.15, + "learning_rate": 2.7870920064047297e-06, + "loss": 0.6069, + "step": 5812 + }, + { + "epoch": 2.15, + "learning_rate": 2.7870550560413843e-06, + "loss": 0.5171, + "step": 5813 + }, + { + "epoch": 2.15, + "learning_rate": 2.7870181056780394e-06, + "loss": 0.566, + "step": 5814 + }, + { + "epoch": 2.15, + "learning_rate": 2.786981155314694e-06, + "loss": 0.642, + "step": 5815 + }, + { + "epoch": 2.15, + "learning_rate": 2.7869442049513487e-06, + "loss": 0.585, + "step": 5816 + }, + { + "epoch": 2.15, + "learning_rate": 2.7869072545880037e-06, + "loss": 0.5608, + "step": 5817 + }, + { + "epoch": 2.15, + "learning_rate": 2.7868703042246583e-06, + "loss": 0.5588, + "step": 5818 + }, + { + "epoch": 2.15, + "learning_rate": 2.786833353861313e-06, + "loss": 0.5801, + "step": 5819 + }, + { + "epoch": 2.15, + "learning_rate": 2.7867964034979676e-06, + "loss": 0.678, + "step": 5820 + }, + { + "epoch": 2.15, + "learning_rate": 2.7867594531346227e-06, + "loss": 0.5743, + "step": 5821 + }, + { + "epoch": 2.15, + "learning_rate": 2.7867225027712773e-06, + "loss": 0.5609, + "step": 5822 + }, + { + "epoch": 2.15, + "learning_rate": 2.7866855524079324e-06, + "loss": 0.5846, + "step": 5823 + }, + { + "epoch": 2.15, + "learning_rate": 2.786648602044587e-06, + "loss": 0.7, + "step": 5824 + }, + { + "epoch": 2.15, + "learning_rate": 2.7866116516812416e-06, + "loss": 0.6432, + "step": 5825 + }, + { + "epoch": 2.15, + "learning_rate": 2.7865747013178963e-06, + "loss": 0.68, + "step": 5826 + }, + { + "epoch": 2.15, + "learning_rate": 2.786537750954551e-06, + "loss": 0.5992, + "step": 5827 + }, + { + "epoch": 2.15, + "learning_rate": 2.786500800591206e-06, + "loss": 0.7236, + "step": 5828 + }, + { + "epoch": 2.15, + "learning_rate": 2.7864638502278606e-06, + "loss": 0.623, + "step": 5829 + }, + { + "epoch": 2.15, + "learning_rate": 2.7864268998645157e-06, + "loss": 0.6449, + "step": 5830 + }, + { + "epoch": 2.15, + "learning_rate": 2.78638994950117e-06, + "loss": 0.6422, + "step": 5831 + }, + { + "epoch": 2.15, + "learning_rate": 2.786352999137825e-06, + "loss": 0.6059, + "step": 5832 + }, + { + "epoch": 2.15, + "learning_rate": 2.7863160487744796e-06, + "loss": 0.6921, + "step": 5833 + }, + { + "epoch": 2.15, + "learning_rate": 2.7862790984111346e-06, + "loss": 0.5878, + "step": 5834 + }, + { + "epoch": 2.15, + "learning_rate": 2.7862421480477893e-06, + "loss": 0.6487, + "step": 5835 + }, + { + "epoch": 2.16, + "learning_rate": 2.786205197684444e-06, + "loss": 0.6719, + "step": 5836 + }, + { + "epoch": 2.16, + "learning_rate": 2.7861682473210986e-06, + "loss": 0.5413, + "step": 5837 + }, + { + "epoch": 2.16, + "learning_rate": 2.786131296957753e-06, + "loss": 0.5996, + "step": 5838 + }, + { + "epoch": 2.16, + "learning_rate": 2.7860943465944083e-06, + "loss": 0.6603, + "step": 5839 + }, + { + "epoch": 2.16, + "learning_rate": 2.786057396231063e-06, + "loss": 0.5893, + "step": 5840 + }, + { + "epoch": 2.16, + "learning_rate": 2.786020445867718e-06, + "loss": 0.6005, + "step": 5841 + }, + { + "epoch": 2.16, + "learning_rate": 2.7859834955043726e-06, + "loss": 0.5278, + "step": 5842 + }, + { + "epoch": 2.16, + "learning_rate": 2.7859465451410272e-06, + "loss": 0.6851, + "step": 5843 + }, + { + "epoch": 2.16, + "learning_rate": 2.785909594777682e-06, + "loss": 0.5441, + "step": 5844 + }, + { + "epoch": 2.16, + "learning_rate": 2.785872644414337e-06, + "loss": 0.6576, + "step": 5845 + }, + { + "epoch": 2.16, + "learning_rate": 2.7858356940509916e-06, + "loss": 0.6506, + "step": 5846 + }, + { + "epoch": 2.16, + "learning_rate": 2.785798743687646e-06, + "loss": 0.703, + "step": 5847 + }, + { + "epoch": 2.16, + "learning_rate": 2.7857617933243013e-06, + "loss": 0.5309, + "step": 5848 + }, + { + "epoch": 2.16, + "learning_rate": 2.785724842960956e-06, + "loss": 0.5134, + "step": 5849 + }, + { + "epoch": 2.16, + "learning_rate": 2.7856878925976105e-06, + "loss": 0.6403, + "step": 5850 + }, + { + "epoch": 2.16, + "learning_rate": 2.785650942234265e-06, + "loss": 0.5257, + "step": 5851 + }, + { + "epoch": 2.16, + "learning_rate": 2.7856139918709202e-06, + "loss": 0.5963, + "step": 5852 + }, + { + "epoch": 2.16, + "learning_rate": 2.785577041507575e-06, + "loss": 0.521, + "step": 5853 + }, + { + "epoch": 2.16, + "learning_rate": 2.78554009114423e-06, + "loss": 0.6215, + "step": 5854 + }, + { + "epoch": 2.16, + "learning_rate": 2.7855031407808846e-06, + "loss": 0.6474, + "step": 5855 + }, + { + "epoch": 2.16, + "learning_rate": 2.785466190417539e-06, + "loss": 0.6281, + "step": 5856 + }, + { + "epoch": 2.16, + "learning_rate": 2.785429240054194e-06, + "loss": 0.5985, + "step": 5857 + }, + { + "epoch": 2.16, + "learning_rate": 2.7853922896908485e-06, + "loss": 0.7267, + "step": 5858 + }, + { + "epoch": 2.16, + "learning_rate": 2.7853553393275035e-06, + "loss": 0.5393, + "step": 5859 + }, + { + "epoch": 2.16, + "learning_rate": 2.785318388964158e-06, + "loss": 0.5643, + "step": 5860 + }, + { + "epoch": 2.16, + "learning_rate": 2.7852814386008132e-06, + "loss": 0.6895, + "step": 5861 + }, + { + "epoch": 2.16, + "learning_rate": 2.785244488237468e-06, + "loss": 0.5366, + "step": 5862 + }, + { + "epoch": 2.17, + "learning_rate": 2.7852075378741225e-06, + "loss": 0.5703, + "step": 5863 + }, + { + "epoch": 2.17, + "learning_rate": 2.785170587510777e-06, + "loss": 0.5481, + "step": 5864 + }, + { + "epoch": 2.17, + "learning_rate": 2.7851336371474318e-06, + "loss": 0.5627, + "step": 5865 + }, + { + "epoch": 2.17, + "learning_rate": 2.785096686784087e-06, + "loss": 0.5303, + "step": 5866 + }, + { + "epoch": 2.17, + "learning_rate": 2.7850597364207415e-06, + "loss": 0.6514, + "step": 5867 + }, + { + "epoch": 2.17, + "learning_rate": 2.7850227860573965e-06, + "loss": 0.6479, + "step": 5868 + }, + { + "epoch": 2.17, + "learning_rate": 2.784985835694051e-06, + "loss": 0.6806, + "step": 5869 + }, + { + "epoch": 2.17, + "learning_rate": 2.784948885330706e-06, + "loss": 0.5881, + "step": 5870 + }, + { + "epoch": 2.17, + "learning_rate": 2.7849119349673604e-06, + "loss": 0.5374, + "step": 5871 + }, + { + "epoch": 2.17, + "learning_rate": 2.7848749846040155e-06, + "loss": 0.5561, + "step": 5872 + }, + { + "epoch": 2.17, + "learning_rate": 2.78483803424067e-06, + "loss": 0.601, + "step": 5873 + }, + { + "epoch": 2.17, + "learning_rate": 2.784801083877325e-06, + "loss": 0.5546, + "step": 5874 + }, + { + "epoch": 2.17, + "learning_rate": 2.78476413351398e-06, + "loss": 0.6089, + "step": 5875 + }, + { + "epoch": 2.17, + "learning_rate": 2.784727183150634e-06, + "loss": 0.569, + "step": 5876 + }, + { + "epoch": 2.17, + "learning_rate": 2.784690232787289e-06, + "loss": 0.6219, + "step": 5877 + }, + { + "epoch": 2.17, + "learning_rate": 2.7846532824239437e-06, + "loss": 0.5334, + "step": 5878 + }, + { + "epoch": 2.17, + "learning_rate": 2.784616332060599e-06, + "loss": 0.5904, + "step": 5879 + }, + { + "epoch": 2.17, + "learning_rate": 2.7845793816972534e-06, + "loss": 0.5837, + "step": 5880 + }, + { + "epoch": 2.17, + "learning_rate": 2.7845424313339085e-06, + "loss": 0.5651, + "step": 5881 + }, + { + "epoch": 2.17, + "learning_rate": 2.7845054809705627e-06, + "loss": 0.6038, + "step": 5882 + }, + { + "epoch": 2.17, + "learning_rate": 2.7844685306072178e-06, + "loss": 0.5903, + "step": 5883 + }, + { + "epoch": 2.17, + "learning_rate": 2.7844315802438724e-06, + "loss": 0.55, + "step": 5884 + }, + { + "epoch": 2.17, + "learning_rate": 2.784394629880527e-06, + "loss": 0.4924, + "step": 5885 + }, + { + "epoch": 2.17, + "learning_rate": 2.784357679517182e-06, + "loss": 0.6151, + "step": 5886 + }, + { + "epoch": 2.17, + "learning_rate": 2.7843207291538367e-06, + "loss": 0.5644, + "step": 5887 + }, + { + "epoch": 2.17, + "learning_rate": 2.7842837787904914e-06, + "loss": 0.539, + "step": 5888 + }, + { + "epoch": 2.17, + "learning_rate": 2.784246828427146e-06, + "loss": 0.5461, + "step": 5889 + }, + { + "epoch": 2.18, + "learning_rate": 2.784209878063801e-06, + "loss": 0.5628, + "step": 5890 + }, + { + "epoch": 2.18, + "learning_rate": 2.7841729277004557e-06, + "loss": 0.5854, + "step": 5891 + }, + { + "epoch": 2.18, + "learning_rate": 2.7841359773371108e-06, + "loss": 0.4532, + "step": 5892 + }, + { + "epoch": 2.18, + "learning_rate": 2.7840990269737654e-06, + "loss": 0.5875, + "step": 5893 + }, + { + "epoch": 2.18, + "learning_rate": 2.78406207661042e-06, + "loss": 0.7275, + "step": 5894 + }, + { + "epoch": 2.18, + "learning_rate": 2.7840251262470747e-06, + "loss": 0.6621, + "step": 5895 + }, + { + "epoch": 2.18, + "learning_rate": 2.7839881758837293e-06, + "loss": 0.5293, + "step": 5896 + }, + { + "epoch": 2.18, + "learning_rate": 2.7839512255203844e-06, + "loss": 0.5971, + "step": 5897 + }, + { + "epoch": 2.18, + "learning_rate": 2.783914275157039e-06, + "loss": 0.6606, + "step": 5898 + }, + { + "epoch": 2.18, + "learning_rate": 2.783877324793694e-06, + "loss": 0.546, + "step": 5899 + }, + { + "epoch": 2.18, + "learning_rate": 2.7838403744303487e-06, + "loss": 0.6567, + "step": 5900 + }, + { + "epoch": 2.18, + "learning_rate": 2.7838034240670033e-06, + "loss": 0.5436, + "step": 5901 + }, + { + "epoch": 2.18, + "learning_rate": 2.783766473703658e-06, + "loss": 0.5636, + "step": 5902 + }, + { + "epoch": 2.18, + "learning_rate": 2.783729523340313e-06, + "loss": 0.5938, + "step": 5903 + }, + { + "epoch": 2.18, + "learning_rate": 2.7836925729769677e-06, + "loss": 0.5175, + "step": 5904 + }, + { + "epoch": 2.18, + "learning_rate": 2.7836556226136223e-06, + "loss": 0.5885, + "step": 5905 + }, + { + "epoch": 2.18, + "learning_rate": 2.7836186722502774e-06, + "loss": 0.6216, + "step": 5906 + }, + { + "epoch": 2.18, + "learning_rate": 2.783581721886932e-06, + "loss": 0.5766, + "step": 5907 + }, + { + "epoch": 2.18, + "learning_rate": 2.7835447715235866e-06, + "loss": 0.5436, + "step": 5908 + }, + { + "epoch": 2.18, + "learning_rate": 2.7835078211602413e-06, + "loss": 0.6115, + "step": 5909 + }, + { + "epoch": 2.18, + "learning_rate": 2.7834708707968963e-06, + "loss": 0.6466, + "step": 5910 + }, + { + "epoch": 2.18, + "learning_rate": 2.783433920433551e-06, + "loss": 0.5656, + "step": 5911 + }, + { + "epoch": 2.18, + "learning_rate": 2.783396970070206e-06, + "loss": 0.5008, + "step": 5912 + }, + { + "epoch": 2.18, + "learning_rate": 2.7833600197068607e-06, + "loss": 0.6073, + "step": 5913 + }, + { + "epoch": 2.18, + "learning_rate": 2.7833230693435153e-06, + "loss": 0.5706, + "step": 5914 + }, + { + "epoch": 2.18, + "learning_rate": 2.78328611898017e-06, + "loss": 0.5676, + "step": 5915 + }, + { + "epoch": 2.18, + "learning_rate": 2.7832491686168246e-06, + "loss": 0.6461, + "step": 5916 + }, + { + "epoch": 2.19, + "learning_rate": 2.7832122182534796e-06, + "loss": 0.5307, + "step": 5917 + }, + { + "epoch": 2.19, + "learning_rate": 2.7831752678901343e-06, + "loss": 0.5743, + "step": 5918 + }, + { + "epoch": 2.19, + "learning_rate": 2.7831383175267893e-06, + "loss": 0.5202, + "step": 5919 + }, + { + "epoch": 2.19, + "learning_rate": 2.783101367163444e-06, + "loss": 0.528, + "step": 5920 + }, + { + "epoch": 2.19, + "learning_rate": 2.7830644168000986e-06, + "loss": 0.6005, + "step": 5921 + }, + { + "epoch": 2.19, + "learning_rate": 2.7830274664367532e-06, + "loss": 0.6394, + "step": 5922 + }, + { + "epoch": 2.19, + "learning_rate": 2.7829905160734083e-06, + "loss": 0.5563, + "step": 5923 + }, + { + "epoch": 2.19, + "learning_rate": 2.782953565710063e-06, + "loss": 0.5354, + "step": 5924 + }, + { + "epoch": 2.19, + "learning_rate": 2.7829166153467176e-06, + "loss": 0.631, + "step": 5925 + }, + { + "epoch": 2.19, + "learning_rate": 2.7828796649833726e-06, + "loss": 0.5409, + "step": 5926 + }, + { + "epoch": 2.19, + "learning_rate": 2.782842714620027e-06, + "loss": 0.5697, + "step": 5927 + }, + { + "epoch": 2.19, + "learning_rate": 2.782805764256682e-06, + "loss": 0.6345, + "step": 5928 + }, + { + "epoch": 2.19, + "learning_rate": 2.7827688138933365e-06, + "loss": 0.5969, + "step": 5929 + }, + { + "epoch": 2.19, + "learning_rate": 2.7827318635299916e-06, + "loss": 0.6084, + "step": 5930 + }, + { + "epoch": 2.19, + "learning_rate": 2.7826949131666462e-06, + "loss": 0.6114, + "step": 5931 + }, + { + "epoch": 2.19, + "learning_rate": 2.7826579628033013e-06, + "loss": 0.6429, + "step": 5932 + }, + { + "epoch": 2.19, + "learning_rate": 2.7826210124399555e-06, + "loss": 0.5893, + "step": 5933 + }, + { + "epoch": 2.19, + "learning_rate": 2.7825840620766106e-06, + "loss": 0.5672, + "step": 5934 + }, + { + "epoch": 2.19, + "learning_rate": 2.782547111713265e-06, + "loss": 0.4641, + "step": 5935 + }, + { + "epoch": 2.19, + "learning_rate": 2.78251016134992e-06, + "loss": 0.49, + "step": 5936 + }, + { + "epoch": 2.19, + "learning_rate": 2.782473210986575e-06, + "loss": 0.6096, + "step": 5937 + }, + { + "epoch": 2.19, + "learning_rate": 2.7824362606232295e-06, + "loss": 0.6348, + "step": 5938 + }, + { + "epoch": 2.19, + "learning_rate": 2.782399310259884e-06, + "loss": 0.6189, + "step": 5939 + }, + { + "epoch": 2.19, + "learning_rate": 2.782362359896539e-06, + "loss": 0.5802, + "step": 5940 + }, + { + "epoch": 2.19, + "learning_rate": 2.782325409533194e-06, + "loss": 0.648, + "step": 5941 + }, + { + "epoch": 2.19, + "learning_rate": 2.7822884591698485e-06, + "loss": 0.6198, + "step": 5942 + }, + { + "epoch": 2.19, + "learning_rate": 2.7822515088065036e-06, + "loss": 0.5436, + "step": 5943 + }, + { + "epoch": 2.19, + "learning_rate": 2.782214558443158e-06, + "loss": 0.596, + "step": 5944 + }, + { + "epoch": 2.2, + "learning_rate": 2.782177608079813e-06, + "loss": 0.6184, + "step": 5945 + }, + { + "epoch": 2.2, + "learning_rate": 2.7821406577164675e-06, + "loss": 0.684, + "step": 5946 + }, + { + "epoch": 2.2, + "learning_rate": 2.782103707353122e-06, + "loss": 0.5385, + "step": 5947 + }, + { + "epoch": 2.2, + "learning_rate": 2.782066756989777e-06, + "loss": 0.5204, + "step": 5948 + }, + { + "epoch": 2.2, + "learning_rate": 2.782029806626432e-06, + "loss": 0.6027, + "step": 5949 + }, + { + "epoch": 2.2, + "learning_rate": 2.781992856263087e-06, + "loss": 0.539, + "step": 5950 + }, + { + "epoch": 2.2, + "learning_rate": 2.7819559058997415e-06, + "loss": 0.6001, + "step": 5951 + }, + { + "epoch": 2.2, + "learning_rate": 2.781918955536396e-06, + "loss": 0.6248, + "step": 5952 + }, + { + "epoch": 2.2, + "learning_rate": 2.7818820051730508e-06, + "loss": 0.5672, + "step": 5953 + }, + { + "epoch": 2.2, + "learning_rate": 2.781845054809706e-06, + "loss": 0.5205, + "step": 5954 + }, + { + "epoch": 2.2, + "learning_rate": 2.7818081044463605e-06, + "loss": 0.5987, + "step": 5955 + }, + { + "epoch": 2.2, + "learning_rate": 2.781771154083015e-06, + "loss": 0.6307, + "step": 5956 + }, + { + "epoch": 2.2, + "learning_rate": 2.78173420371967e-06, + "loss": 0.5926, + "step": 5957 + }, + { + "epoch": 2.2, + "learning_rate": 2.781697253356325e-06, + "loss": 0.5814, + "step": 5958 + }, + { + "epoch": 2.2, + "learning_rate": 2.7816603029929794e-06, + "loss": 0.5466, + "step": 5959 + }, + { + "epoch": 2.2, + "learning_rate": 2.781623352629634e-06, + "loss": 0.6939, + "step": 5960 + }, + { + "epoch": 2.2, + "learning_rate": 2.781586402266289e-06, + "loss": 0.5637, + "step": 5961 + }, + { + "epoch": 2.2, + "learning_rate": 2.7815494519029438e-06, + "loss": 0.6743, + "step": 5962 + }, + { + "epoch": 2.2, + "learning_rate": 2.781512501539599e-06, + "loss": 0.5652, + "step": 5963 + }, + { + "epoch": 2.2, + "learning_rate": 2.7814755511762535e-06, + "loss": 0.5341, + "step": 5964 + }, + { + "epoch": 2.2, + "learning_rate": 2.781438600812908e-06, + "loss": 0.515, + "step": 5965 + }, + { + "epoch": 2.2, + "learning_rate": 2.7814016504495628e-06, + "loss": 0.5823, + "step": 5966 + }, + { + "epoch": 2.2, + "learning_rate": 2.7813647000862174e-06, + "loss": 0.6267, + "step": 5967 + }, + { + "epoch": 2.2, + "learning_rate": 2.7813277497228724e-06, + "loss": 0.529, + "step": 5968 + }, + { + "epoch": 2.2, + "learning_rate": 2.781290799359527e-06, + "loss": 0.6847, + "step": 5969 + }, + { + "epoch": 2.2, + "learning_rate": 2.781253848996182e-06, + "loss": 0.6564, + "step": 5970 + }, + { + "epoch": 2.2, + "learning_rate": 2.7812168986328368e-06, + "loss": 0.6277, + "step": 5971 + }, + { + "epoch": 2.21, + "learning_rate": 2.7811799482694914e-06, + "loss": 0.5141, + "step": 5972 + }, + { + "epoch": 2.21, + "learning_rate": 2.781142997906146e-06, + "loss": 0.5438, + "step": 5973 + }, + { + "epoch": 2.21, + "learning_rate": 2.781106047542801e-06, + "loss": 0.5219, + "step": 5974 + }, + { + "epoch": 2.21, + "learning_rate": 2.7810690971794557e-06, + "loss": 0.5577, + "step": 5975 + }, + { + "epoch": 2.21, + "learning_rate": 2.7810321468161104e-06, + "loss": 0.5875, + "step": 5976 + }, + { + "epoch": 2.21, + "learning_rate": 2.780995196452765e-06, + "loss": 0.6428, + "step": 5977 + }, + { + "epoch": 2.21, + "learning_rate": 2.7809582460894197e-06, + "loss": 0.52, + "step": 5978 + }, + { + "epoch": 2.21, + "learning_rate": 2.7809212957260747e-06, + "loss": 0.6364, + "step": 5979 + }, + { + "epoch": 2.21, + "learning_rate": 2.7808843453627294e-06, + "loss": 0.6011, + "step": 5980 + }, + { + "epoch": 2.21, + "learning_rate": 2.7808473949993844e-06, + "loss": 0.5171, + "step": 5981 + }, + { + "epoch": 2.21, + "learning_rate": 2.780810444636039e-06, + "loss": 0.6798, + "step": 5982 + }, + { + "epoch": 2.21, + "learning_rate": 2.7807734942726937e-06, + "loss": 0.6083, + "step": 5983 + }, + { + "epoch": 2.21, + "learning_rate": 2.7807365439093483e-06, + "loss": 0.4959, + "step": 5984 + }, + { + "epoch": 2.21, + "learning_rate": 2.780699593546003e-06, + "loss": 0.5451, + "step": 5985 + }, + { + "epoch": 2.21, + "learning_rate": 2.780662643182658e-06, + "loss": 0.6058, + "step": 5986 + }, + { + "epoch": 2.21, + "learning_rate": 2.7806256928193127e-06, + "loss": 0.6581, + "step": 5987 + }, + { + "epoch": 2.21, + "learning_rate": 2.7805887424559677e-06, + "loss": 0.5788, + "step": 5988 + }, + { + "epoch": 2.21, + "learning_rate": 2.7805517920926224e-06, + "loss": 0.6448, + "step": 5989 + }, + { + "epoch": 2.21, + "learning_rate": 2.780514841729277e-06, + "loss": 0.6146, + "step": 5990 + }, + { + "epoch": 2.21, + "learning_rate": 2.7804778913659316e-06, + "loss": 0.5394, + "step": 5991 + }, + { + "epoch": 2.21, + "learning_rate": 2.7804409410025867e-06, + "loss": 0.551, + "step": 5992 + }, + { + "epoch": 2.21, + "learning_rate": 2.7804039906392413e-06, + "loss": 0.5747, + "step": 5993 + }, + { + "epoch": 2.21, + "learning_rate": 2.7803670402758964e-06, + "loss": 0.6167, + "step": 5994 + }, + { + "epoch": 2.21, + "learning_rate": 2.780330089912551e-06, + "loss": 0.7107, + "step": 5995 + }, + { + "epoch": 2.21, + "learning_rate": 2.7802931395492057e-06, + "loss": 0.7001, + "step": 5996 + }, + { + "epoch": 2.21, + "learning_rate": 2.7802561891858603e-06, + "loss": 0.5712, + "step": 5997 + }, + { + "epoch": 2.21, + "learning_rate": 2.780219238822515e-06, + "loss": 0.5173, + "step": 5998 + }, + { + "epoch": 2.22, + "learning_rate": 2.78018228845917e-06, + "loss": 0.474, + "step": 5999 + }, + { + "epoch": 2.22, + "learning_rate": 2.7801453380958246e-06, + "loss": 0.5993, + "step": 6000 + }, + { + "epoch": 2.22, + "learning_rate": 2.7801083877324797e-06, + "loss": 0.6437, + "step": 6001 + }, + { + "epoch": 2.22, + "learning_rate": 2.7800714373691343e-06, + "loss": 0.7018, + "step": 6002 + }, + { + "epoch": 2.22, + "learning_rate": 2.780034487005789e-06, + "loss": 0.5888, + "step": 6003 + }, + { + "epoch": 2.22, + "learning_rate": 2.7799975366424436e-06, + "loss": 0.5906, + "step": 6004 + }, + { + "epoch": 2.22, + "learning_rate": 2.7799605862790982e-06, + "loss": 0.568, + "step": 6005 + }, + { + "epoch": 2.22, + "learning_rate": 2.7799236359157533e-06, + "loss": 0.5897, + "step": 6006 + }, + { + "epoch": 2.22, + "learning_rate": 2.779886685552408e-06, + "loss": 0.6087, + "step": 6007 + }, + { + "epoch": 2.22, + "learning_rate": 2.779849735189063e-06, + "loss": 0.6628, + "step": 6008 + }, + { + "epoch": 2.22, + "learning_rate": 2.7798127848257176e-06, + "loss": 0.556, + "step": 6009 + }, + { + "epoch": 2.22, + "learning_rate": 2.7797758344623723e-06, + "loss": 0.5826, + "step": 6010 + }, + { + "epoch": 2.22, + "learning_rate": 2.779738884099027e-06, + "loss": 0.6516, + "step": 6011 + }, + { + "epoch": 2.22, + "learning_rate": 2.779701933735682e-06, + "loss": 0.6373, + "step": 6012 + }, + { + "epoch": 2.22, + "learning_rate": 2.7796649833723366e-06, + "loss": 0.5387, + "step": 6013 + }, + { + "epoch": 2.22, + "learning_rate": 2.7796280330089917e-06, + "loss": 0.7071, + "step": 6014 + }, + { + "epoch": 2.22, + "learning_rate": 2.7795910826456463e-06, + "loss": 0.6063, + "step": 6015 + }, + { + "epoch": 2.22, + "learning_rate": 2.7795541322823005e-06, + "loss": 0.5565, + "step": 6016 + }, + { + "epoch": 2.22, + "learning_rate": 2.7795171819189556e-06, + "loss": 0.5703, + "step": 6017 + }, + { + "epoch": 2.22, + "learning_rate": 2.77948023155561e-06, + "loss": 0.6246, + "step": 6018 + }, + { + "epoch": 2.22, + "learning_rate": 2.7794432811922653e-06, + "loss": 0.601, + "step": 6019 + }, + { + "epoch": 2.22, + "learning_rate": 2.77940633082892e-06, + "loss": 0.5015, + "step": 6020 + }, + { + "epoch": 2.22, + "learning_rate": 2.779369380465575e-06, + "loss": 0.6811, + "step": 6021 + }, + { + "epoch": 2.22, + "learning_rate": 2.779332430102229e-06, + "loss": 0.6049, + "step": 6022 + }, + { + "epoch": 2.22, + "learning_rate": 2.7792954797388842e-06, + "loss": 0.5279, + "step": 6023 + }, + { + "epoch": 2.22, + "learning_rate": 2.779258529375539e-06, + "loss": 0.646, + "step": 6024 + }, + { + "epoch": 2.22, + "learning_rate": 2.7792215790121935e-06, + "loss": 0.5567, + "step": 6025 + }, + { + "epoch": 2.23, + "learning_rate": 2.7791846286488486e-06, + "loss": 0.5295, + "step": 6026 + }, + { + "epoch": 2.23, + "learning_rate": 2.779147678285503e-06, + "loss": 0.5872, + "step": 6027 + }, + { + "epoch": 2.23, + "learning_rate": 2.779110727922158e-06, + "loss": 0.5526, + "step": 6028 + }, + { + "epoch": 2.23, + "learning_rate": 2.7790737775588125e-06, + "loss": 0.4825, + "step": 6029 + }, + { + "epoch": 2.23, + "learning_rate": 2.7790368271954675e-06, + "loss": 0.6046, + "step": 6030 + }, + { + "epoch": 2.23, + "learning_rate": 2.778999876832122e-06, + "loss": 0.6408, + "step": 6031 + }, + { + "epoch": 2.23, + "learning_rate": 2.7789629264687772e-06, + "loss": 0.5688, + "step": 6032 + }, + { + "epoch": 2.23, + "learning_rate": 2.778925976105432e-06, + "loss": 0.5814, + "step": 6033 + }, + { + "epoch": 2.23, + "learning_rate": 2.7788890257420865e-06, + "loss": 0.5614, + "step": 6034 + }, + { + "epoch": 2.23, + "learning_rate": 2.778852075378741e-06, + "loss": 0.6214, + "step": 6035 + }, + { + "epoch": 2.23, + "learning_rate": 2.7788151250153958e-06, + "loss": 0.6108, + "step": 6036 + }, + { + "epoch": 2.23, + "learning_rate": 2.778778174652051e-06, + "loss": 0.5442, + "step": 6037 + }, + { + "epoch": 2.23, + "learning_rate": 2.7787412242887055e-06, + "loss": 0.5675, + "step": 6038 + }, + { + "epoch": 2.23, + "learning_rate": 2.7787042739253605e-06, + "loss": 0.5415, + "step": 6039 + }, + { + "epoch": 2.23, + "learning_rate": 2.778667323562015e-06, + "loss": 0.6544, + "step": 6040 + }, + { + "epoch": 2.23, + "learning_rate": 2.77863037319867e-06, + "loss": 0.5194, + "step": 6041 + }, + { + "epoch": 2.23, + "learning_rate": 2.7785934228353244e-06, + "loss": 0.5896, + "step": 6042 + }, + { + "epoch": 2.23, + "learning_rate": 2.7785564724719795e-06, + "loss": 0.5495, + "step": 6043 + }, + { + "epoch": 2.23, + "learning_rate": 2.778519522108634e-06, + "loss": 0.6637, + "step": 6044 + }, + { + "epoch": 2.23, + "learning_rate": 2.7784825717452888e-06, + "loss": 0.6611, + "step": 6045 + }, + { + "epoch": 2.23, + "learning_rate": 2.778445621381944e-06, + "loss": 0.5804, + "step": 6046 + }, + { + "epoch": 2.23, + "learning_rate": 2.7784086710185985e-06, + "loss": 0.6014, + "step": 6047 + }, + { + "epoch": 2.23, + "learning_rate": 2.778371720655253e-06, + "loss": 0.609, + "step": 6048 + }, + { + "epoch": 2.23, + "learning_rate": 2.7783347702919077e-06, + "loss": 0.5761, + "step": 6049 + }, + { + "epoch": 2.23, + "learning_rate": 2.778297819928563e-06, + "loss": 0.6287, + "step": 6050 + }, + { + "epoch": 2.23, + "learning_rate": 2.7782608695652174e-06, + "loss": 0.5826, + "step": 6051 + }, + { + "epoch": 2.23, + "learning_rate": 2.7782239192018725e-06, + "loss": 0.5501, + "step": 6052 + }, + { + "epoch": 2.24, + "learning_rate": 2.778186968838527e-06, + "loss": 0.5288, + "step": 6053 + }, + { + "epoch": 2.24, + "learning_rate": 2.7781500184751818e-06, + "loss": 0.6485, + "step": 6054 + }, + { + "epoch": 2.24, + "learning_rate": 2.7781130681118364e-06, + "loss": 0.5552, + "step": 6055 + }, + { + "epoch": 2.24, + "learning_rate": 2.778076117748491e-06, + "loss": 0.6279, + "step": 6056 + }, + { + "epoch": 2.24, + "learning_rate": 2.778039167385146e-06, + "loss": 0.5953, + "step": 6057 + }, + { + "epoch": 2.24, + "learning_rate": 2.7780022170218007e-06, + "loss": 0.6308, + "step": 6058 + }, + { + "epoch": 2.24, + "learning_rate": 2.777965266658456e-06, + "loss": 0.5456, + "step": 6059 + }, + { + "epoch": 2.24, + "learning_rate": 2.7779283162951104e-06, + "loss": 0.5671, + "step": 6060 + }, + { + "epoch": 2.24, + "learning_rate": 2.777891365931765e-06, + "loss": 0.617, + "step": 6061 + }, + { + "epoch": 2.24, + "learning_rate": 2.7778544155684197e-06, + "loss": 0.5732, + "step": 6062 + }, + { + "epoch": 2.24, + "learning_rate": 2.7778174652050748e-06, + "loss": 0.6147, + "step": 6063 + }, + { + "epoch": 2.24, + "learning_rate": 2.7777805148417294e-06, + "loss": 0.5898, + "step": 6064 + }, + { + "epoch": 2.24, + "learning_rate": 2.777743564478384e-06, + "loss": 0.5232, + "step": 6065 + }, + { + "epoch": 2.24, + "learning_rate": 2.777706614115039e-06, + "loss": 0.6076, + "step": 6066 + }, + { + "epoch": 2.24, + "learning_rate": 2.7776696637516933e-06, + "loss": 0.5062, + "step": 6067 + }, + { + "epoch": 2.24, + "learning_rate": 2.7776327133883484e-06, + "loss": 0.5243, + "step": 6068 + }, + { + "epoch": 2.24, + "learning_rate": 2.777595763025003e-06, + "loss": 0.6073, + "step": 6069 + }, + { + "epoch": 2.24, + "learning_rate": 2.777558812661658e-06, + "loss": 0.6637, + "step": 6070 + }, + { + "epoch": 2.24, + "learning_rate": 2.7775218622983127e-06, + "loss": 0.5499, + "step": 6071 + }, + { + "epoch": 2.24, + "learning_rate": 2.7774849119349678e-06, + "loss": 0.6657, + "step": 6072 + }, + { + "epoch": 2.24, + "learning_rate": 2.777447961571622e-06, + "loss": 0.5528, + "step": 6073 + }, + { + "epoch": 2.24, + "learning_rate": 2.777411011208277e-06, + "loss": 0.6556, + "step": 6074 + }, + { + "epoch": 2.24, + "learning_rate": 2.7773740608449317e-06, + "loss": 0.5553, + "step": 6075 + }, + { + "epoch": 2.24, + "learning_rate": 2.7773371104815863e-06, + "loss": 0.6257, + "step": 6076 + }, + { + "epoch": 2.24, + "learning_rate": 2.7773001601182414e-06, + "loss": 0.5534, + "step": 6077 + }, + { + "epoch": 2.24, + "learning_rate": 2.777263209754896e-06, + "loss": 0.5836, + "step": 6078 + }, + { + "epoch": 2.24, + "learning_rate": 2.7772262593915506e-06, + "loss": 0.59, + "step": 6079 + }, + { + "epoch": 2.25, + "learning_rate": 2.7771893090282053e-06, + "loss": 0.6444, + "step": 6080 + }, + { + "epoch": 2.25, + "learning_rate": 2.7771523586648603e-06, + "loss": 0.5759, + "step": 6081 + }, + { + "epoch": 2.25, + "learning_rate": 2.777115408301515e-06, + "loss": 0.7231, + "step": 6082 + }, + { + "epoch": 2.25, + "learning_rate": 2.77707845793817e-06, + "loss": 0.5448, + "step": 6083 + }, + { + "epoch": 2.25, + "learning_rate": 2.7770415075748247e-06, + "loss": 0.5224, + "step": 6084 + }, + { + "epoch": 2.25, + "learning_rate": 2.7770045572114793e-06, + "loss": 0.613, + "step": 6085 + }, + { + "epoch": 2.25, + "learning_rate": 2.776967606848134e-06, + "loss": 0.5295, + "step": 6086 + }, + { + "epoch": 2.25, + "learning_rate": 2.7769306564847886e-06, + "loss": 0.6438, + "step": 6087 + }, + { + "epoch": 2.25, + "learning_rate": 2.7768937061214436e-06, + "loss": 0.4883, + "step": 6088 + }, + { + "epoch": 2.25, + "learning_rate": 2.7768567557580983e-06, + "loss": 0.6437, + "step": 6089 + }, + { + "epoch": 2.25, + "learning_rate": 2.7768198053947533e-06, + "loss": 0.5746, + "step": 6090 + }, + { + "epoch": 2.25, + "learning_rate": 2.776782855031408e-06, + "loss": 0.5984, + "step": 6091 + }, + { + "epoch": 2.25, + "learning_rate": 2.7767459046680626e-06, + "loss": 0.5743, + "step": 6092 + }, + { + "epoch": 2.25, + "learning_rate": 2.7767089543047172e-06, + "loss": 0.5855, + "step": 6093 + }, + { + "epoch": 2.25, + "learning_rate": 2.776672003941372e-06, + "loss": 0.5977, + "step": 6094 + }, + { + "epoch": 2.25, + "learning_rate": 2.776635053578027e-06, + "loss": 0.6743, + "step": 6095 + }, + { + "epoch": 2.25, + "learning_rate": 2.7765981032146816e-06, + "loss": 0.6102, + "step": 6096 + }, + { + "epoch": 2.25, + "learning_rate": 2.7765611528513366e-06, + "loss": 0.5495, + "step": 6097 + }, + { + "epoch": 2.25, + "learning_rate": 2.7765242024879913e-06, + "loss": 0.5705, + "step": 6098 + }, + { + "epoch": 2.25, + "learning_rate": 2.776487252124646e-06, + "loss": 0.6015, + "step": 6099 + }, + { + "epoch": 2.25, + "learning_rate": 2.7764503017613006e-06, + "loss": 0.6337, + "step": 6100 + }, + { + "epoch": 2.25, + "learning_rate": 2.7764133513979556e-06, + "loss": 0.6686, + "step": 6101 + }, + { + "epoch": 2.25, + "learning_rate": 2.7763764010346102e-06, + "loss": 0.5611, + "step": 6102 + }, + { + "epoch": 2.25, + "learning_rate": 2.7763394506712653e-06, + "loss": 0.6006, + "step": 6103 + }, + { + "epoch": 2.25, + "learning_rate": 2.77630250030792e-06, + "loss": 0.5801, + "step": 6104 + }, + { + "epoch": 2.25, + "learning_rate": 2.7762655499445746e-06, + "loss": 0.6269, + "step": 6105 + }, + { + "epoch": 2.25, + "learning_rate": 2.7762285995812292e-06, + "loss": 0.6151, + "step": 6106 + }, + { + "epoch": 2.26, + "learning_rate": 2.776191649217884e-06, + "loss": 0.7392, + "step": 6107 + }, + { + "epoch": 2.26, + "learning_rate": 2.776154698854539e-06, + "loss": 0.6025, + "step": 6108 + }, + { + "epoch": 2.26, + "learning_rate": 2.7761177484911935e-06, + "loss": 0.6264, + "step": 6109 + }, + { + "epoch": 2.26, + "learning_rate": 2.7760807981278486e-06, + "loss": 0.6456, + "step": 6110 + }, + { + "epoch": 2.26, + "learning_rate": 2.7760438477645032e-06, + "loss": 0.6402, + "step": 6111 + }, + { + "epoch": 2.26, + "learning_rate": 2.776006897401158e-06, + "loss": 0.6425, + "step": 6112 + }, + { + "epoch": 2.26, + "learning_rate": 2.7759699470378125e-06, + "loss": 0.5549, + "step": 6113 + }, + { + "epoch": 2.26, + "learning_rate": 2.775932996674467e-06, + "loss": 0.6248, + "step": 6114 + }, + { + "epoch": 2.26, + "learning_rate": 2.7758960463111222e-06, + "loss": 0.5356, + "step": 6115 + }, + { + "epoch": 2.26, + "learning_rate": 2.775859095947777e-06, + "loss": 0.5203, + "step": 6116 + }, + { + "epoch": 2.26, + "learning_rate": 2.775822145584432e-06, + "loss": 0.5765, + "step": 6117 + }, + { + "epoch": 2.26, + "learning_rate": 2.775785195221086e-06, + "loss": 0.5413, + "step": 6118 + }, + { + "epoch": 2.26, + "learning_rate": 2.775748244857741e-06, + "loss": 0.5373, + "step": 6119 + }, + { + "epoch": 2.26, + "learning_rate": 2.775711294494396e-06, + "loss": 0.5719, + "step": 6120 + }, + { + "epoch": 2.26, + "learning_rate": 2.775674344131051e-06, + "loss": 0.5125, + "step": 6121 + }, + { + "epoch": 2.26, + "learning_rate": 2.7756373937677055e-06, + "loss": 0.5187, + "step": 6122 + }, + { + "epoch": 2.26, + "learning_rate": 2.7756004434043606e-06, + "loss": 0.5865, + "step": 6123 + }, + { + "epoch": 2.26, + "learning_rate": 2.775563493041015e-06, + "loss": 0.5524, + "step": 6124 + }, + { + "epoch": 2.26, + "learning_rate": 2.7755265426776694e-06, + "loss": 0.5401, + "step": 6125 + }, + { + "epoch": 2.26, + "learning_rate": 2.7754895923143245e-06, + "loss": 0.5694, + "step": 6126 + }, + { + "epoch": 2.26, + "learning_rate": 2.775452641950979e-06, + "loss": 0.5524, + "step": 6127 + }, + { + "epoch": 2.26, + "learning_rate": 2.775415691587634e-06, + "loss": 0.6051, + "step": 6128 + }, + { + "epoch": 2.26, + "learning_rate": 2.775378741224289e-06, + "loss": 0.5433, + "step": 6129 + }, + { + "epoch": 2.26, + "learning_rate": 2.7753417908609435e-06, + "loss": 0.5895, + "step": 6130 + }, + { + "epoch": 2.26, + "learning_rate": 2.775304840497598e-06, + "loss": 0.605, + "step": 6131 + }, + { + "epoch": 2.26, + "learning_rate": 2.775267890134253e-06, + "loss": 0.5901, + "step": 6132 + }, + { + "epoch": 2.26, + "learning_rate": 2.7752309397709078e-06, + "loss": 0.5854, + "step": 6133 + }, + { + "epoch": 2.27, + "learning_rate": 2.7751939894075624e-06, + "loss": 0.7142, + "step": 6134 + }, + { + "epoch": 2.27, + "learning_rate": 2.7751570390442175e-06, + "loss": 0.6666, + "step": 6135 + }, + { + "epoch": 2.27, + "learning_rate": 2.775120088680872e-06, + "loss": 0.5255, + "step": 6136 + }, + { + "epoch": 2.27, + "learning_rate": 2.7750831383175268e-06, + "loss": 0.626, + "step": 6137 + }, + { + "epoch": 2.27, + "learning_rate": 2.7750461879541814e-06, + "loss": 0.6088, + "step": 6138 + }, + { + "epoch": 2.27, + "learning_rate": 2.7750092375908365e-06, + "loss": 0.5229, + "step": 6139 + }, + { + "epoch": 2.27, + "learning_rate": 2.774972287227491e-06, + "loss": 0.6238, + "step": 6140 + }, + { + "epoch": 2.27, + "learning_rate": 2.774935336864146e-06, + "loss": 0.5579, + "step": 6141 + }, + { + "epoch": 2.27, + "learning_rate": 2.7748983865008008e-06, + "loss": 0.5189, + "step": 6142 + }, + { + "epoch": 2.27, + "learning_rate": 2.7748614361374554e-06, + "loss": 0.5906, + "step": 6143 + }, + { + "epoch": 2.27, + "learning_rate": 2.77482448577411e-06, + "loss": 0.6643, + "step": 6144 + }, + { + "epoch": 2.27, + "learning_rate": 2.7747875354107647e-06, + "loss": 0.4463, + "step": 6145 + }, + { + "epoch": 2.27, + "learning_rate": 2.7747505850474198e-06, + "loss": 0.6896, + "step": 6146 + }, + { + "epoch": 2.27, + "learning_rate": 2.7747136346840744e-06, + "loss": 0.6299, + "step": 6147 + }, + { + "epoch": 2.27, + "learning_rate": 2.7746766843207295e-06, + "loss": 0.7012, + "step": 6148 + }, + { + "epoch": 2.27, + "learning_rate": 2.774639733957384e-06, + "loss": 0.4997, + "step": 6149 + }, + { + "epoch": 2.27, + "learning_rate": 2.7746027835940387e-06, + "loss": 0.6138, + "step": 6150 + }, + { + "epoch": 2.27, + "learning_rate": 2.7745658332306934e-06, + "loss": 0.5437, + "step": 6151 + }, + { + "epoch": 2.27, + "learning_rate": 2.7745288828673484e-06, + "loss": 0.6493, + "step": 6152 + }, + { + "epoch": 2.27, + "learning_rate": 2.774491932504003e-06, + "loss": 0.5679, + "step": 6153 + }, + { + "epoch": 2.27, + "learning_rate": 2.7744549821406577e-06, + "loss": 0.6475, + "step": 6154 + }, + { + "epoch": 2.27, + "learning_rate": 2.7744180317773128e-06, + "loss": 0.5523, + "step": 6155 + }, + { + "epoch": 2.27, + "learning_rate": 2.7743810814139674e-06, + "loss": 0.5666, + "step": 6156 + }, + { + "epoch": 2.27, + "learning_rate": 2.774344131050622e-06, + "loss": 0.4738, + "step": 6157 + }, + { + "epoch": 2.27, + "learning_rate": 2.7743071806872767e-06, + "loss": 0.5216, + "step": 6158 + }, + { + "epoch": 2.27, + "learning_rate": 2.7742702303239317e-06, + "loss": 0.5461, + "step": 6159 + }, + { + "epoch": 2.27, + "learning_rate": 2.7742332799605864e-06, + "loss": 0.6539, + "step": 6160 + }, + { + "epoch": 2.28, + "learning_rate": 2.7741963295972414e-06, + "loss": 0.5482, + "step": 6161 + }, + { + "epoch": 2.28, + "learning_rate": 2.7741593792338956e-06, + "loss": 0.5625, + "step": 6162 + }, + { + "epoch": 2.28, + "learning_rate": 2.7741224288705507e-06, + "loss": 0.577, + "step": 6163 + }, + { + "epoch": 2.28, + "learning_rate": 2.7740854785072053e-06, + "loss": 0.6101, + "step": 6164 + }, + { + "epoch": 2.28, + "learning_rate": 2.77404852814386e-06, + "loss": 0.581, + "step": 6165 + }, + { + "epoch": 2.28, + "learning_rate": 2.774011577780515e-06, + "loss": 0.5291, + "step": 6166 + }, + { + "epoch": 2.28, + "learning_rate": 2.7739746274171697e-06, + "loss": 0.5046, + "step": 6167 + }, + { + "epoch": 2.28, + "learning_rate": 2.7739376770538243e-06, + "loss": 0.6231, + "step": 6168 + }, + { + "epoch": 2.28, + "learning_rate": 2.773900726690479e-06, + "loss": 0.6094, + "step": 6169 + }, + { + "epoch": 2.28, + "learning_rate": 2.773863776327134e-06, + "loss": 0.5636, + "step": 6170 + }, + { + "epoch": 2.28, + "learning_rate": 2.7738268259637886e-06, + "loss": 0.5683, + "step": 6171 + }, + { + "epoch": 2.28, + "learning_rate": 2.7737898756004437e-06, + "loss": 0.6241, + "step": 6172 + }, + { + "epoch": 2.28, + "learning_rate": 2.7737529252370983e-06, + "loss": 0.5596, + "step": 6173 + }, + { + "epoch": 2.28, + "learning_rate": 2.773715974873753e-06, + "loss": 0.6016, + "step": 6174 + }, + { + "epoch": 2.28, + "learning_rate": 2.7736790245104076e-06, + "loss": 0.597, + "step": 6175 + }, + { + "epoch": 2.28, + "learning_rate": 2.7736420741470622e-06, + "loss": 0.5673, + "step": 6176 + }, + { + "epoch": 2.28, + "learning_rate": 2.7736051237837173e-06, + "loss": 0.5895, + "step": 6177 + }, + { + "epoch": 2.28, + "learning_rate": 2.773568173420372e-06, + "loss": 0.5711, + "step": 6178 + }, + { + "epoch": 2.28, + "learning_rate": 2.773531223057027e-06, + "loss": 0.6817, + "step": 6179 + }, + { + "epoch": 2.28, + "learning_rate": 2.7734942726936816e-06, + "loss": 0.589, + "step": 6180 + }, + { + "epoch": 2.28, + "learning_rate": 2.7734573223303363e-06, + "loss": 0.6195, + "step": 6181 + }, + { + "epoch": 2.28, + "learning_rate": 2.773420371966991e-06, + "loss": 0.4973, + "step": 6182 + }, + { + "epoch": 2.28, + "learning_rate": 2.773383421603646e-06, + "loss": 0.6058, + "step": 6183 + }, + { + "epoch": 2.28, + "learning_rate": 2.7733464712403006e-06, + "loss": 0.6329, + "step": 6184 + }, + { + "epoch": 2.28, + "learning_rate": 2.7733095208769552e-06, + "loss": 0.6106, + "step": 6185 + }, + { + "epoch": 2.28, + "learning_rate": 2.7732725705136103e-06, + "loss": 0.6229, + "step": 6186 + }, + { + "epoch": 2.28, + "learning_rate": 2.773235620150265e-06, + "loss": 0.5982, + "step": 6187 + }, + { + "epoch": 2.29, + "learning_rate": 2.7731986697869196e-06, + "loss": 0.6068, + "step": 6188 + }, + { + "epoch": 2.29, + "learning_rate": 2.773161719423574e-06, + "loss": 0.5832, + "step": 6189 + }, + { + "epoch": 2.29, + "learning_rate": 2.7731247690602293e-06, + "loss": 0.6217, + "step": 6190 + }, + { + "epoch": 2.29, + "learning_rate": 2.773087818696884e-06, + "loss": 0.6013, + "step": 6191 + }, + { + "epoch": 2.29, + "learning_rate": 2.773050868333539e-06, + "loss": 0.5861, + "step": 6192 + }, + { + "epoch": 2.29, + "learning_rate": 2.7730139179701936e-06, + "loss": 0.6099, + "step": 6193 + }, + { + "epoch": 2.29, + "learning_rate": 2.7729769676068482e-06, + "loss": 0.6111, + "step": 6194 + }, + { + "epoch": 2.29, + "learning_rate": 2.772940017243503e-06, + "loss": 0.6571, + "step": 6195 + }, + { + "epoch": 2.29, + "learning_rate": 2.7729030668801575e-06, + "loss": 0.5411, + "step": 6196 + }, + { + "epoch": 2.29, + "learning_rate": 2.7728661165168126e-06, + "loss": 0.5492, + "step": 6197 + }, + { + "epoch": 2.29, + "learning_rate": 2.772829166153467e-06, + "loss": 0.5849, + "step": 6198 + }, + { + "epoch": 2.29, + "learning_rate": 2.7727922157901223e-06, + "loss": 0.611, + "step": 6199 + }, + { + "epoch": 2.29, + "learning_rate": 2.772755265426777e-06, + "loss": 0.5865, + "step": 6200 + }, + { + "epoch": 2.29, + "learning_rate": 2.7727183150634315e-06, + "loss": 0.6756, + "step": 6201 + }, + { + "epoch": 2.29, + "learning_rate": 2.772681364700086e-06, + "loss": 0.5182, + "step": 6202 + }, + { + "epoch": 2.29, + "learning_rate": 2.7726444143367412e-06, + "loss": 0.5953, + "step": 6203 + }, + { + "epoch": 2.29, + "learning_rate": 2.772607463973396e-06, + "loss": 0.5458, + "step": 6204 + }, + { + "epoch": 2.29, + "learning_rate": 2.7725705136100505e-06, + "loss": 0.566, + "step": 6205 + }, + { + "epoch": 2.29, + "learning_rate": 2.7725335632467056e-06, + "loss": 0.6254, + "step": 6206 + }, + { + "epoch": 2.29, + "learning_rate": 2.7724966128833598e-06, + "loss": 0.6126, + "step": 6207 + }, + { + "epoch": 2.29, + "learning_rate": 2.772459662520015e-06, + "loss": 0.6291, + "step": 6208 + }, + { + "epoch": 2.29, + "learning_rate": 2.7724227121566695e-06, + "loss": 0.6081, + "step": 6209 + }, + { + "epoch": 2.29, + "learning_rate": 2.7723857617933245e-06, + "loss": 0.5362, + "step": 6210 + }, + { + "epoch": 2.29, + "learning_rate": 2.772348811429979e-06, + "loss": 0.6278, + "step": 6211 + }, + { + "epoch": 2.29, + "learning_rate": 2.7723118610666342e-06, + "loss": 0.5556, + "step": 6212 + }, + { + "epoch": 2.29, + "learning_rate": 2.7722749107032884e-06, + "loss": 0.5702, + "step": 6213 + }, + { + "epoch": 2.29, + "learning_rate": 2.772237960339943e-06, + "loss": 0.4896, + "step": 6214 + }, + { + "epoch": 2.3, + "learning_rate": 2.772201009976598e-06, + "loss": 0.6031, + "step": 6215 + }, + { + "epoch": 2.3, + "learning_rate": 2.7721640596132528e-06, + "loss": 0.6087, + "step": 6216 + }, + { + "epoch": 2.3, + "learning_rate": 2.772127109249908e-06, + "loss": 0.4995, + "step": 6217 + }, + { + "epoch": 2.3, + "learning_rate": 2.7720901588865625e-06, + "loss": 0.526, + "step": 6218 + }, + { + "epoch": 2.3, + "learning_rate": 2.772053208523217e-06, + "loss": 0.6629, + "step": 6219 + }, + { + "epoch": 2.3, + "learning_rate": 2.7720162581598717e-06, + "loss": 0.7049, + "step": 6220 + }, + { + "epoch": 2.3, + "learning_rate": 2.771979307796527e-06, + "loss": 0.6033, + "step": 6221 + }, + { + "epoch": 2.3, + "learning_rate": 2.7719423574331814e-06, + "loss": 0.6643, + "step": 6222 + }, + { + "epoch": 2.3, + "learning_rate": 2.7719054070698365e-06, + "loss": 0.4836, + "step": 6223 + }, + { + "epoch": 2.3, + "learning_rate": 2.771868456706491e-06, + "loss": 0.6157, + "step": 6224 + }, + { + "epoch": 2.3, + "learning_rate": 2.7718315063431458e-06, + "loss": 0.5685, + "step": 6225 + }, + { + "epoch": 2.3, + "learning_rate": 2.7717945559798004e-06, + "loss": 0.6766, + "step": 6226 + }, + { + "epoch": 2.3, + "learning_rate": 2.771757605616455e-06, + "loss": 0.7093, + "step": 6227 + }, + { + "epoch": 2.3, + "learning_rate": 2.77172065525311e-06, + "loss": 0.5974, + "step": 6228 + }, + { + "epoch": 2.3, + "learning_rate": 2.7716837048897647e-06, + "loss": 0.6449, + "step": 6229 + }, + { + "epoch": 2.3, + "learning_rate": 2.77164675452642e-06, + "loss": 0.6803, + "step": 6230 + }, + { + "epoch": 2.3, + "learning_rate": 2.7716098041630744e-06, + "loss": 0.6125, + "step": 6231 + }, + { + "epoch": 2.3, + "learning_rate": 2.771572853799729e-06, + "loss": 0.5934, + "step": 6232 + }, + { + "epoch": 2.3, + "learning_rate": 2.7715359034363837e-06, + "loss": 0.5679, + "step": 6233 + }, + { + "epoch": 2.3, + "learning_rate": 2.7714989530730383e-06, + "loss": 0.6894, + "step": 6234 + }, + { + "epoch": 2.3, + "learning_rate": 2.7714620027096934e-06, + "loss": 0.5653, + "step": 6235 + }, + { + "epoch": 2.3, + "learning_rate": 2.771425052346348e-06, + "loss": 0.543, + "step": 6236 + }, + { + "epoch": 2.3, + "learning_rate": 2.771388101983003e-06, + "loss": 0.6524, + "step": 6237 + }, + { + "epoch": 2.3, + "learning_rate": 2.7713511516196577e-06, + "loss": 0.6221, + "step": 6238 + }, + { + "epoch": 2.3, + "learning_rate": 2.7713142012563124e-06, + "loss": 0.561, + "step": 6239 + }, + { + "epoch": 2.3, + "learning_rate": 2.771277250892967e-06, + "loss": 0.5863, + "step": 6240 + }, + { + "epoch": 2.3, + "learning_rate": 2.771240300529622e-06, + "loss": 0.6032, + "step": 6241 + }, + { + "epoch": 2.31, + "learning_rate": 2.7712033501662767e-06, + "loss": 0.5308, + "step": 6242 + }, + { + "epoch": 2.31, + "learning_rate": 2.7711663998029318e-06, + "loss": 0.6259, + "step": 6243 + }, + { + "epoch": 2.31, + "learning_rate": 2.7711294494395864e-06, + "loss": 0.6551, + "step": 6244 + }, + { + "epoch": 2.31, + "learning_rate": 2.771092499076241e-06, + "loss": 0.5676, + "step": 6245 + }, + { + "epoch": 2.31, + "learning_rate": 2.7710555487128957e-06, + "loss": 0.6347, + "step": 6246 + }, + { + "epoch": 2.31, + "learning_rate": 2.7710185983495503e-06, + "loss": 0.553, + "step": 6247 + }, + { + "epoch": 2.31, + "learning_rate": 2.7709816479862054e-06, + "loss": 0.5912, + "step": 6248 + }, + { + "epoch": 2.31, + "learning_rate": 2.77094469762286e-06, + "loss": 0.5119, + "step": 6249 + }, + { + "epoch": 2.31, + "learning_rate": 2.770907747259515e-06, + "loss": 0.5448, + "step": 6250 + }, + { + "epoch": 2.31, + "learning_rate": 2.7708707968961697e-06, + "loss": 0.5935, + "step": 6251 + }, + { + "epoch": 2.31, + "learning_rate": 2.7708338465328243e-06, + "loss": 0.613, + "step": 6252 + }, + { + "epoch": 2.31, + "learning_rate": 2.770796896169479e-06, + "loss": 0.6266, + "step": 6253 + }, + { + "epoch": 2.31, + "learning_rate": 2.7707599458061336e-06, + "loss": 0.6867, + "step": 6254 + }, + { + "epoch": 2.31, + "learning_rate": 2.7707229954427887e-06, + "loss": 0.6546, + "step": 6255 + }, + { + "epoch": 2.31, + "learning_rate": 2.7706860450794433e-06, + "loss": 0.5702, + "step": 6256 + }, + { + "epoch": 2.31, + "learning_rate": 2.7706490947160984e-06, + "loss": 0.5867, + "step": 6257 + }, + { + "epoch": 2.31, + "learning_rate": 2.7706121443527526e-06, + "loss": 0.6339, + "step": 6258 + }, + { + "epoch": 2.31, + "learning_rate": 2.7705751939894076e-06, + "loss": 0.5168, + "step": 6259 + }, + { + "epoch": 2.31, + "learning_rate": 2.7705382436260623e-06, + "loss": 0.6193, + "step": 6260 + }, + { + "epoch": 2.31, + "learning_rate": 2.7705012932627173e-06, + "loss": 0.5849, + "step": 6261 + }, + { + "epoch": 2.31, + "learning_rate": 2.770464342899372e-06, + "loss": 0.6107, + "step": 6262 + }, + { + "epoch": 2.31, + "learning_rate": 2.770427392536027e-06, + "loss": 0.609, + "step": 6263 + }, + { + "epoch": 2.31, + "learning_rate": 2.7703904421726813e-06, + "loss": 0.6492, + "step": 6264 + }, + { + "epoch": 2.31, + "learning_rate": 2.770353491809336e-06, + "loss": 0.5393, + "step": 6265 + }, + { + "epoch": 2.31, + "learning_rate": 2.770316541445991e-06, + "loss": 0.6664, + "step": 6266 + }, + { + "epoch": 2.31, + "learning_rate": 2.7702795910826456e-06, + "loss": 0.5568, + "step": 6267 + }, + { + "epoch": 2.31, + "learning_rate": 2.7702426407193006e-06, + "loss": 0.6573, + "step": 6268 + }, + { + "epoch": 2.31, + "learning_rate": 2.7702056903559553e-06, + "loss": 0.5917, + "step": 6269 + }, + { + "epoch": 2.32, + "learning_rate": 2.77016873999261e-06, + "loss": 0.6066, + "step": 6270 + }, + { + "epoch": 2.32, + "learning_rate": 2.7701317896292646e-06, + "loss": 0.6247, + "step": 6271 + }, + { + "epoch": 2.32, + "learning_rate": 2.7700948392659196e-06, + "loss": 0.6349, + "step": 6272 + }, + { + "epoch": 2.32, + "learning_rate": 2.7700578889025743e-06, + "loss": 0.5772, + "step": 6273 + }, + { + "epoch": 2.32, + "learning_rate": 2.770020938539229e-06, + "loss": 0.5926, + "step": 6274 + }, + { + "epoch": 2.32, + "learning_rate": 2.769983988175884e-06, + "loss": 0.5031, + "step": 6275 + }, + { + "epoch": 2.32, + "learning_rate": 2.7699470378125386e-06, + "loss": 0.662, + "step": 6276 + }, + { + "epoch": 2.32, + "learning_rate": 2.7699100874491932e-06, + "loss": 0.5919, + "step": 6277 + }, + { + "epoch": 2.32, + "learning_rate": 2.769873137085848e-06, + "loss": 0.6492, + "step": 6278 + }, + { + "epoch": 2.32, + "learning_rate": 2.769836186722503e-06, + "loss": 0.5964, + "step": 6279 + }, + { + "epoch": 2.32, + "learning_rate": 2.7697992363591576e-06, + "loss": 0.54, + "step": 6280 + }, + { + "epoch": 2.32, + "learning_rate": 2.7697622859958126e-06, + "loss": 0.6459, + "step": 6281 + }, + { + "epoch": 2.32, + "learning_rate": 2.7697253356324673e-06, + "loss": 0.6239, + "step": 6282 + }, + { + "epoch": 2.32, + "learning_rate": 2.769688385269122e-06, + "loss": 0.5999, + "step": 6283 + }, + { + "epoch": 2.32, + "learning_rate": 2.7696514349057765e-06, + "loss": 0.6531, + "step": 6284 + }, + { + "epoch": 2.32, + "learning_rate": 2.769614484542431e-06, + "loss": 0.5163, + "step": 6285 + }, + { + "epoch": 2.32, + "learning_rate": 2.7695775341790862e-06, + "loss": 0.6413, + "step": 6286 + }, + { + "epoch": 2.32, + "learning_rate": 2.769540583815741e-06, + "loss": 0.5518, + "step": 6287 + }, + { + "epoch": 2.32, + "learning_rate": 2.769503633452396e-06, + "loss": 0.5511, + "step": 6288 + }, + { + "epoch": 2.32, + "learning_rate": 2.7694666830890506e-06, + "loss": 0.6025, + "step": 6289 + }, + { + "epoch": 2.32, + "learning_rate": 2.769429732725705e-06, + "loss": 0.5878, + "step": 6290 + }, + { + "epoch": 2.32, + "learning_rate": 2.76939278236236e-06, + "loss": 0.5584, + "step": 6291 + }, + { + "epoch": 2.32, + "learning_rate": 2.769355831999015e-06, + "loss": 0.5967, + "step": 6292 + }, + { + "epoch": 2.32, + "learning_rate": 2.7693188816356695e-06, + "loss": 0.5886, + "step": 6293 + }, + { + "epoch": 2.32, + "learning_rate": 2.769281931272324e-06, + "loss": 0.6358, + "step": 6294 + }, + { + "epoch": 2.32, + "learning_rate": 2.7692449809089792e-06, + "loss": 0.5298, + "step": 6295 + }, + { + "epoch": 2.32, + "learning_rate": 2.769208030545634e-06, + "loss": 0.5585, + "step": 6296 + }, + { + "epoch": 2.33, + "learning_rate": 2.7691710801822885e-06, + "loss": 0.5054, + "step": 6297 + }, + { + "epoch": 2.33, + "learning_rate": 2.769134129818943e-06, + "loss": 0.6127, + "step": 6298 + }, + { + "epoch": 2.33, + "learning_rate": 2.769097179455598e-06, + "loss": 0.5779, + "step": 6299 + }, + { + "epoch": 2.33, + "learning_rate": 2.769060229092253e-06, + "loss": 0.5256, + "step": 6300 + }, + { + "epoch": 2.33, + "learning_rate": 2.769023278728908e-06, + "loss": 0.5145, + "step": 6301 + }, + { + "epoch": 2.33, + "learning_rate": 2.7689863283655625e-06, + "loss": 0.669, + "step": 6302 + }, + { + "epoch": 2.33, + "learning_rate": 2.768949378002217e-06, + "loss": 0.5331, + "step": 6303 + }, + { + "epoch": 2.33, + "learning_rate": 2.768912427638872e-06, + "loss": 0.6451, + "step": 6304 + }, + { + "epoch": 2.33, + "learning_rate": 2.7688754772755264e-06, + "loss": 0.5412, + "step": 6305 + }, + { + "epoch": 2.33, + "learning_rate": 2.7688385269121815e-06, + "loss": 0.5952, + "step": 6306 + }, + { + "epoch": 2.33, + "learning_rate": 2.768801576548836e-06, + "loss": 0.5617, + "step": 6307 + }, + { + "epoch": 2.33, + "learning_rate": 2.768764626185491e-06, + "loss": 0.6045, + "step": 6308 + }, + { + "epoch": 2.33, + "learning_rate": 2.7687276758221454e-06, + "loss": 0.5646, + "step": 6309 + }, + { + "epoch": 2.33, + "learning_rate": 2.7686907254588005e-06, + "loss": 0.5784, + "step": 6310 + }, + { + "epoch": 2.33, + "learning_rate": 2.768653775095455e-06, + "loss": 0.6495, + "step": 6311 + }, + { + "epoch": 2.33, + "learning_rate": 2.76861682473211e-06, + "loss": 0.5389, + "step": 6312 + }, + { + "epoch": 2.33, + "learning_rate": 2.768579874368765e-06, + "loss": 0.6783, + "step": 6313 + }, + { + "epoch": 2.33, + "learning_rate": 2.7685429240054194e-06, + "loss": 0.5818, + "step": 6314 + }, + { + "epoch": 2.33, + "learning_rate": 2.768505973642074e-06, + "loss": 0.5839, + "step": 6315 + }, + { + "epoch": 2.33, + "learning_rate": 2.7684690232787287e-06, + "loss": 0.5349, + "step": 6316 + }, + { + "epoch": 2.33, + "learning_rate": 2.7684320729153838e-06, + "loss": 0.5951, + "step": 6317 + }, + { + "epoch": 2.33, + "learning_rate": 2.7683951225520384e-06, + "loss": 0.5943, + "step": 6318 + }, + { + "epoch": 2.33, + "learning_rate": 2.7683581721886935e-06, + "loss": 0.5854, + "step": 6319 + }, + { + "epoch": 2.33, + "learning_rate": 2.768321221825348e-06, + "loss": 0.569, + "step": 6320 + }, + { + "epoch": 2.33, + "learning_rate": 2.7682842714620027e-06, + "loss": 0.5366, + "step": 6321 + }, + { + "epoch": 2.33, + "learning_rate": 2.7682473210986574e-06, + "loss": 0.6561, + "step": 6322 + }, + { + "epoch": 2.33, + "learning_rate": 2.768210370735312e-06, + "loss": 0.4879, + "step": 6323 + }, + { + "epoch": 2.34, + "learning_rate": 2.768173420371967e-06, + "loss": 0.5741, + "step": 6324 + }, + { + "epoch": 2.34, + "learning_rate": 2.7681364700086217e-06, + "loss": 0.7088, + "step": 6325 + }, + { + "epoch": 2.34, + "learning_rate": 2.7680995196452768e-06, + "loss": 0.6164, + "step": 6326 + }, + { + "epoch": 2.34, + "learning_rate": 2.7680625692819314e-06, + "loss": 0.5811, + "step": 6327 + }, + { + "epoch": 2.34, + "learning_rate": 2.768025618918586e-06, + "loss": 0.6887, + "step": 6328 + }, + { + "epoch": 2.34, + "learning_rate": 2.7679886685552407e-06, + "loss": 0.6055, + "step": 6329 + }, + { + "epoch": 2.34, + "learning_rate": 2.7679517181918957e-06, + "loss": 0.6048, + "step": 6330 + }, + { + "epoch": 2.34, + "learning_rate": 2.7679147678285504e-06, + "loss": 0.613, + "step": 6331 + }, + { + "epoch": 2.34, + "learning_rate": 2.7678778174652054e-06, + "loss": 0.5595, + "step": 6332 + }, + { + "epoch": 2.34, + "learning_rate": 2.76784086710186e-06, + "loss": 0.5943, + "step": 6333 + }, + { + "epoch": 2.34, + "learning_rate": 2.7678039167385147e-06, + "loss": 0.5972, + "step": 6334 + }, + { + "epoch": 2.34, + "learning_rate": 2.7677669663751693e-06, + "loss": 0.5926, + "step": 6335 + }, + { + "epoch": 2.34, + "learning_rate": 2.767730016011824e-06, + "loss": 0.6289, + "step": 6336 + }, + { + "epoch": 2.34, + "learning_rate": 2.767693065648479e-06, + "loss": 0.5732, + "step": 6337 + }, + { + "epoch": 2.34, + "learning_rate": 2.7676561152851337e-06, + "loss": 0.6355, + "step": 6338 + }, + { + "epoch": 2.34, + "learning_rate": 2.7676191649217887e-06, + "loss": 0.5963, + "step": 6339 + }, + { + "epoch": 2.34, + "learning_rate": 2.7675822145584434e-06, + "loss": 0.5036, + "step": 6340 + }, + { + "epoch": 2.34, + "learning_rate": 2.767545264195098e-06, + "loss": 0.5205, + "step": 6341 + }, + { + "epoch": 2.34, + "learning_rate": 2.7675083138317526e-06, + "loss": 0.542, + "step": 6342 + }, + { + "epoch": 2.34, + "learning_rate": 2.7674713634684073e-06, + "loss": 0.604, + "step": 6343 + }, + { + "epoch": 2.34, + "learning_rate": 2.7674344131050623e-06, + "loss": 0.5266, + "step": 6344 + }, + { + "epoch": 2.34, + "learning_rate": 2.767397462741717e-06, + "loss": 0.6107, + "step": 6345 + }, + { + "epoch": 2.34, + "learning_rate": 2.767360512378372e-06, + "loss": 0.568, + "step": 6346 + }, + { + "epoch": 2.34, + "learning_rate": 2.7673235620150262e-06, + "loss": 0.5432, + "step": 6347 + }, + { + "epoch": 2.34, + "learning_rate": 2.7672866116516813e-06, + "loss": 0.547, + "step": 6348 + }, + { + "epoch": 2.34, + "learning_rate": 2.767249661288336e-06, + "loss": 0.5731, + "step": 6349 + }, + { + "epoch": 2.34, + "learning_rate": 2.767212710924991e-06, + "loss": 0.6222, + "step": 6350 + }, + { + "epoch": 2.35, + "learning_rate": 2.7671757605616456e-06, + "loss": 0.6953, + "step": 6351 + }, + { + "epoch": 2.35, + "learning_rate": 2.7671388101983007e-06, + "loss": 0.5329, + "step": 6352 + }, + { + "epoch": 2.35, + "learning_rate": 2.767101859834955e-06, + "loss": 0.6091, + "step": 6353 + }, + { + "epoch": 2.35, + "learning_rate": 2.7670649094716095e-06, + "loss": 0.5508, + "step": 6354 + }, + { + "epoch": 2.35, + "learning_rate": 2.7670279591082646e-06, + "loss": 0.5628, + "step": 6355 + }, + { + "epoch": 2.35, + "learning_rate": 2.7669910087449192e-06, + "loss": 0.5919, + "step": 6356 + }, + { + "epoch": 2.35, + "learning_rate": 2.7669540583815743e-06, + "loss": 0.5995, + "step": 6357 + }, + { + "epoch": 2.35, + "learning_rate": 2.766917108018229e-06, + "loss": 0.6627, + "step": 6358 + }, + { + "epoch": 2.35, + "learning_rate": 2.7668801576548836e-06, + "loss": 0.5088, + "step": 6359 + }, + { + "epoch": 2.35, + "learning_rate": 2.766843207291538e-06, + "loss": 0.5372, + "step": 6360 + }, + { + "epoch": 2.35, + "learning_rate": 2.7668062569281933e-06, + "loss": 0.6013, + "step": 6361 + }, + { + "epoch": 2.35, + "learning_rate": 2.766769306564848e-06, + "loss": 0.6067, + "step": 6362 + }, + { + "epoch": 2.35, + "learning_rate": 2.7667323562015025e-06, + "loss": 0.6647, + "step": 6363 + }, + { + "epoch": 2.35, + "learning_rate": 2.7666954058381576e-06, + "loss": 0.6374, + "step": 6364 + }, + { + "epoch": 2.35, + "learning_rate": 2.7666584554748122e-06, + "loss": 0.5502, + "step": 6365 + }, + { + "epoch": 2.35, + "learning_rate": 2.766621505111467e-06, + "loss": 0.6815, + "step": 6366 + }, + { + "epoch": 2.35, + "learning_rate": 2.7665845547481215e-06, + "loss": 0.6648, + "step": 6367 + }, + { + "epoch": 2.35, + "learning_rate": 2.7665476043847766e-06, + "loss": 0.6769, + "step": 6368 + }, + { + "epoch": 2.35, + "learning_rate": 2.766510654021431e-06, + "loss": 0.5977, + "step": 6369 + }, + { + "epoch": 2.35, + "learning_rate": 2.7664737036580863e-06, + "loss": 0.584, + "step": 6370 + }, + { + "epoch": 2.35, + "learning_rate": 2.766436753294741e-06, + "loss": 0.5561, + "step": 6371 + }, + { + "epoch": 2.35, + "learning_rate": 2.7663998029313955e-06, + "loss": 0.591, + "step": 6372 + }, + { + "epoch": 2.35, + "learning_rate": 2.76636285256805e-06, + "loss": 0.6098, + "step": 6373 + }, + { + "epoch": 2.35, + "learning_rate": 2.766325902204705e-06, + "loss": 0.5587, + "step": 6374 + }, + { + "epoch": 2.35, + "learning_rate": 2.76628895184136e-06, + "loss": 0.5819, + "step": 6375 + }, + { + "epoch": 2.35, + "learning_rate": 2.7662520014780145e-06, + "loss": 0.6006, + "step": 6376 + }, + { + "epoch": 2.35, + "learning_rate": 2.7662150511146696e-06, + "loss": 0.6084, + "step": 6377 + }, + { + "epoch": 2.36, + "learning_rate": 2.766178100751324e-06, + "loss": 0.5322, + "step": 6378 + }, + { + "epoch": 2.36, + "learning_rate": 2.766141150387979e-06, + "loss": 0.5481, + "step": 6379 + }, + { + "epoch": 2.36, + "learning_rate": 2.7661042000246335e-06, + "loss": 0.5694, + "step": 6380 + }, + { + "epoch": 2.36, + "learning_rate": 2.7660672496612885e-06, + "loss": 0.5441, + "step": 6381 + }, + { + "epoch": 2.36, + "learning_rate": 2.766030299297943e-06, + "loss": 0.6376, + "step": 6382 + }, + { + "epoch": 2.36, + "learning_rate": 2.765993348934598e-06, + "loss": 0.6537, + "step": 6383 + }, + { + "epoch": 2.36, + "learning_rate": 2.765956398571253e-06, + "loss": 0.6936, + "step": 6384 + }, + { + "epoch": 2.36, + "learning_rate": 2.7659194482079075e-06, + "loss": 0.5466, + "step": 6385 + }, + { + "epoch": 2.36, + "learning_rate": 2.765882497844562e-06, + "loss": 0.5752, + "step": 6386 + }, + { + "epoch": 2.36, + "learning_rate": 2.7658455474812168e-06, + "loss": 0.6079, + "step": 6387 + }, + { + "epoch": 2.36, + "learning_rate": 2.765808597117872e-06, + "loss": 0.6421, + "step": 6388 + }, + { + "epoch": 2.36, + "learning_rate": 2.7657716467545265e-06, + "loss": 0.5948, + "step": 6389 + }, + { + "epoch": 2.36, + "learning_rate": 2.7657346963911815e-06, + "loss": 0.6116, + "step": 6390 + }, + { + "epoch": 2.36, + "learning_rate": 2.765697746027836e-06, + "loss": 0.5819, + "step": 6391 + }, + { + "epoch": 2.36, + "learning_rate": 2.765660795664491e-06, + "loss": 0.5419, + "step": 6392 + }, + { + "epoch": 2.36, + "learning_rate": 2.7656238453011454e-06, + "loss": 0.4473, + "step": 6393 + }, + { + "epoch": 2.36, + "learning_rate": 2.7655868949378e-06, + "loss": 0.5948, + "step": 6394 + }, + { + "epoch": 2.36, + "learning_rate": 2.765549944574455e-06, + "loss": 0.5697, + "step": 6395 + }, + { + "epoch": 2.36, + "learning_rate": 2.7655129942111098e-06, + "loss": 0.6605, + "step": 6396 + }, + { + "epoch": 2.36, + "learning_rate": 2.765476043847765e-06, + "loss": 0.5683, + "step": 6397 + }, + { + "epoch": 2.36, + "learning_rate": 2.765439093484419e-06, + "loss": 0.5956, + "step": 6398 + }, + { + "epoch": 2.36, + "learning_rate": 2.765402143121074e-06, + "loss": 0.5763, + "step": 6399 + }, + { + "epoch": 2.36, + "learning_rate": 2.7653651927577288e-06, + "loss": 0.5784, + "step": 6400 + }, + { + "epoch": 2.36, + "learning_rate": 2.765328242394384e-06, + "loss": 0.6001, + "step": 6401 + }, + { + "epoch": 2.36, + "learning_rate": 2.7652912920310384e-06, + "loss": 0.6438, + "step": 6402 + }, + { + "epoch": 2.36, + "learning_rate": 2.765254341667693e-06, + "loss": 0.5302, + "step": 6403 + }, + { + "epoch": 2.36, + "learning_rate": 2.7652173913043477e-06, + "loss": 0.7242, + "step": 6404 + }, + { + "epoch": 2.37, + "learning_rate": 2.7651804409410024e-06, + "loss": 0.6403, + "step": 6405 + }, + { + "epoch": 2.37, + "learning_rate": 2.7651434905776574e-06, + "loss": 0.5575, + "step": 6406 + }, + { + "epoch": 2.37, + "learning_rate": 2.765106540214312e-06, + "loss": 0.617, + "step": 6407 + }, + { + "epoch": 2.37, + "learning_rate": 2.765069589850967e-06, + "loss": 0.6788, + "step": 6408 + }, + { + "epoch": 2.37, + "learning_rate": 2.7650326394876217e-06, + "loss": 0.5849, + "step": 6409 + }, + { + "epoch": 2.37, + "learning_rate": 2.7649956891242764e-06, + "loss": 0.6052, + "step": 6410 + }, + { + "epoch": 2.37, + "learning_rate": 2.764958738760931e-06, + "loss": 0.6302, + "step": 6411 + }, + { + "epoch": 2.37, + "learning_rate": 2.764921788397586e-06, + "loss": 0.6352, + "step": 6412 + }, + { + "epoch": 2.37, + "learning_rate": 2.7648848380342407e-06, + "loss": 0.5163, + "step": 6413 + }, + { + "epoch": 2.37, + "learning_rate": 2.7648478876708954e-06, + "loss": 0.5859, + "step": 6414 + }, + { + "epoch": 2.37, + "learning_rate": 2.7648109373075504e-06, + "loss": 0.6094, + "step": 6415 + }, + { + "epoch": 2.37, + "learning_rate": 2.764773986944205e-06, + "loss": 0.5235, + "step": 6416 + }, + { + "epoch": 2.37, + "learning_rate": 2.7647370365808597e-06, + "loss": 0.6492, + "step": 6417 + }, + { + "epoch": 2.37, + "learning_rate": 2.7647000862175143e-06, + "loss": 0.6151, + "step": 6418 + }, + { + "epoch": 2.37, + "learning_rate": 2.7646631358541694e-06, + "loss": 0.5694, + "step": 6419 + }, + { + "epoch": 2.37, + "learning_rate": 2.764626185490824e-06, + "loss": 0.5894, + "step": 6420 + }, + { + "epoch": 2.37, + "learning_rate": 2.764589235127479e-06, + "loss": 0.5877, + "step": 6421 + }, + { + "epoch": 2.37, + "learning_rate": 2.7645522847641337e-06, + "loss": 0.5626, + "step": 6422 + }, + { + "epoch": 2.37, + "learning_rate": 2.7645153344007884e-06, + "loss": 0.6235, + "step": 6423 + }, + { + "epoch": 2.37, + "learning_rate": 2.764478384037443e-06, + "loss": 0.6737, + "step": 6424 + }, + { + "epoch": 2.37, + "learning_rate": 2.7644414336740976e-06, + "loss": 0.5449, + "step": 6425 + }, + { + "epoch": 2.37, + "learning_rate": 2.7644044833107527e-06, + "loss": 0.5746, + "step": 6426 + }, + { + "epoch": 2.37, + "learning_rate": 2.7643675329474073e-06, + "loss": 0.5821, + "step": 6427 + }, + { + "epoch": 2.37, + "learning_rate": 2.7643305825840624e-06, + "loss": 0.5855, + "step": 6428 + }, + { + "epoch": 2.37, + "learning_rate": 2.764293632220717e-06, + "loss": 0.5969, + "step": 6429 + }, + { + "epoch": 2.37, + "learning_rate": 2.7642566818573717e-06, + "loss": 0.6026, + "step": 6430 + }, + { + "epoch": 2.37, + "learning_rate": 2.7642197314940263e-06, + "loss": 0.6419, + "step": 6431 + }, + { + "epoch": 2.38, + "learning_rate": 2.7641827811306814e-06, + "loss": 0.5704, + "step": 6432 + }, + { + "epoch": 2.38, + "learning_rate": 2.764145830767336e-06, + "loss": 0.5428, + "step": 6433 + }, + { + "epoch": 2.38, + "learning_rate": 2.7641088804039906e-06, + "loss": 0.6111, + "step": 6434 + }, + { + "epoch": 2.38, + "learning_rate": 2.7640719300406457e-06, + "loss": 0.5103, + "step": 6435 + }, + { + "epoch": 2.38, + "learning_rate": 2.7640349796773003e-06, + "loss": 0.6013, + "step": 6436 + }, + { + "epoch": 2.38, + "learning_rate": 2.763998029313955e-06, + "loss": 0.6283, + "step": 6437 + }, + { + "epoch": 2.38, + "learning_rate": 2.7639610789506096e-06, + "loss": 0.6836, + "step": 6438 + }, + { + "epoch": 2.38, + "learning_rate": 2.7639241285872647e-06, + "loss": 0.4813, + "step": 6439 + }, + { + "epoch": 2.38, + "learning_rate": 2.7638871782239193e-06, + "loss": 0.5823, + "step": 6440 + }, + { + "epoch": 2.38, + "learning_rate": 2.7638502278605743e-06, + "loss": 0.599, + "step": 6441 + }, + { + "epoch": 2.38, + "learning_rate": 2.763813277497229e-06, + "loss": 0.5135, + "step": 6442 + }, + { + "epoch": 2.38, + "learning_rate": 2.763776327133883e-06, + "loss": 0.6654, + "step": 6443 + }, + { + "epoch": 2.38, + "learning_rate": 2.7637393767705383e-06, + "loss": 0.5309, + "step": 6444 + }, + { + "epoch": 2.38, + "learning_rate": 2.763702426407193e-06, + "loss": 0.6505, + "step": 6445 + }, + { + "epoch": 2.38, + "learning_rate": 2.763665476043848e-06, + "loss": 0.5486, + "step": 6446 + }, + { + "epoch": 2.38, + "learning_rate": 2.7636285256805026e-06, + "loss": 0.5068, + "step": 6447 + }, + { + "epoch": 2.38, + "learning_rate": 2.7635915753171577e-06, + "loss": 0.6766, + "step": 6448 + }, + { + "epoch": 2.38, + "learning_rate": 2.763554624953812e-06, + "loss": 0.5908, + "step": 6449 + }, + { + "epoch": 2.38, + "learning_rate": 2.763517674590467e-06, + "loss": 0.5727, + "step": 6450 + }, + { + "epoch": 2.38, + "learning_rate": 2.7634807242271216e-06, + "loss": 0.5775, + "step": 6451 + }, + { + "epoch": 2.38, + "learning_rate": 2.7634437738637766e-06, + "loss": 0.5338, + "step": 6452 + }, + { + "epoch": 2.38, + "learning_rate": 2.7634068235004313e-06, + "loss": 0.5725, + "step": 6453 + }, + { + "epoch": 2.38, + "learning_rate": 2.763369873137086e-06, + "loss": 0.5801, + "step": 6454 + }, + { + "epoch": 2.38, + "learning_rate": 2.7633329227737405e-06, + "loss": 0.5086, + "step": 6455 + }, + { + "epoch": 2.38, + "learning_rate": 2.763295972410395e-06, + "loss": 0.5923, + "step": 6456 + }, + { + "epoch": 2.38, + "learning_rate": 2.7632590220470502e-06, + "loss": 0.6104, + "step": 6457 + }, + { + "epoch": 2.38, + "learning_rate": 2.763222071683705e-06, + "loss": 0.5552, + "step": 6458 + }, + { + "epoch": 2.39, + "learning_rate": 2.76318512132036e-06, + "loss": 0.5966, + "step": 6459 + }, + { + "epoch": 2.39, + "learning_rate": 2.7631481709570146e-06, + "loss": 0.6503, + "step": 6460 + }, + { + "epoch": 2.39, + "learning_rate": 2.763111220593669e-06, + "loss": 0.6749, + "step": 6461 + }, + { + "epoch": 2.39, + "learning_rate": 2.763074270230324e-06, + "loss": 0.5423, + "step": 6462 + }, + { + "epoch": 2.39, + "learning_rate": 2.7630373198669785e-06, + "loss": 0.6462, + "step": 6463 + }, + { + "epoch": 2.39, + "learning_rate": 2.7630003695036335e-06, + "loss": 0.6302, + "step": 6464 + }, + { + "epoch": 2.39, + "learning_rate": 2.762963419140288e-06, + "loss": 0.5956, + "step": 6465 + }, + { + "epoch": 2.39, + "learning_rate": 2.7629264687769432e-06, + "loss": 0.6201, + "step": 6466 + }, + { + "epoch": 2.39, + "learning_rate": 2.762889518413598e-06, + "loss": 0.5528, + "step": 6467 + }, + { + "epoch": 2.39, + "learning_rate": 2.7628525680502525e-06, + "loss": 0.5737, + "step": 6468 + }, + { + "epoch": 2.39, + "learning_rate": 2.762815617686907e-06, + "loss": 0.6401, + "step": 6469 + }, + { + "epoch": 2.39, + "learning_rate": 2.762778667323562e-06, + "loss": 0.6427, + "step": 6470 + }, + { + "epoch": 2.39, + "learning_rate": 2.762741716960217e-06, + "loss": 0.5686, + "step": 6471 + }, + { + "epoch": 2.39, + "learning_rate": 2.762704766596872e-06, + "loss": 0.6563, + "step": 6472 + }, + { + "epoch": 2.39, + "learning_rate": 2.7626678162335265e-06, + "loss": 0.5426, + "step": 6473 + }, + { + "epoch": 2.39, + "learning_rate": 2.762630865870181e-06, + "loss": 0.5737, + "step": 6474 + }, + { + "epoch": 2.39, + "learning_rate": 2.762593915506836e-06, + "loss": 0.6259, + "step": 6475 + }, + { + "epoch": 2.39, + "learning_rate": 2.7625569651434904e-06, + "loss": 0.6518, + "step": 6476 + }, + { + "epoch": 2.39, + "learning_rate": 2.7625200147801455e-06, + "loss": 0.5661, + "step": 6477 + }, + { + "epoch": 2.39, + "learning_rate": 2.7624830644168e-06, + "loss": 0.4968, + "step": 6478 + }, + { + "epoch": 2.39, + "learning_rate": 2.762446114053455e-06, + "loss": 0.6039, + "step": 6479 + }, + { + "epoch": 2.39, + "learning_rate": 2.76240916369011e-06, + "loss": 0.5612, + "step": 6480 + }, + { + "epoch": 2.39, + "learning_rate": 2.7623722133267645e-06, + "loss": 0.5711, + "step": 6481 + }, + { + "epoch": 2.39, + "learning_rate": 2.762335262963419e-06, + "loss": 0.7565, + "step": 6482 + }, + { + "epoch": 2.39, + "learning_rate": 2.7622983126000737e-06, + "loss": 0.5517, + "step": 6483 + }, + { + "epoch": 2.39, + "learning_rate": 2.762261362236729e-06, + "loss": 0.5029, + "step": 6484 + }, + { + "epoch": 2.39, + "learning_rate": 2.7622244118733834e-06, + "loss": 0.465, + "step": 6485 + }, + { + "epoch": 2.4, + "learning_rate": 2.7621874615100385e-06, + "loss": 0.5503, + "step": 6486 + }, + { + "epoch": 2.4, + "learning_rate": 2.762150511146693e-06, + "loss": 0.5478, + "step": 6487 + }, + { + "epoch": 2.4, + "learning_rate": 2.7621135607833478e-06, + "loss": 0.5625, + "step": 6488 + }, + { + "epoch": 2.4, + "learning_rate": 2.7620766104200024e-06, + "loss": 0.5646, + "step": 6489 + }, + { + "epoch": 2.4, + "learning_rate": 2.7620396600566575e-06, + "loss": 0.5079, + "step": 6490 + }, + { + "epoch": 2.4, + "learning_rate": 2.762002709693312e-06, + "loss": 0.5256, + "step": 6491 + }, + { + "epoch": 2.4, + "learning_rate": 2.761965759329967e-06, + "loss": 0.5888, + "step": 6492 + }, + { + "epoch": 2.4, + "learning_rate": 2.761928808966622e-06, + "loss": 0.5327, + "step": 6493 + }, + { + "epoch": 2.4, + "learning_rate": 2.761891858603276e-06, + "loss": 0.5799, + "step": 6494 + }, + { + "epoch": 2.4, + "learning_rate": 2.761854908239931e-06, + "loss": 0.6068, + "step": 6495 + }, + { + "epoch": 2.4, + "learning_rate": 2.7618179578765857e-06, + "loss": 0.5889, + "step": 6496 + }, + { + "epoch": 2.4, + "learning_rate": 2.7617810075132408e-06, + "loss": 0.6057, + "step": 6497 + }, + { + "epoch": 2.4, + "learning_rate": 2.7617440571498954e-06, + "loss": 0.5223, + "step": 6498 + }, + { + "epoch": 2.4, + "learning_rate": 2.7617071067865505e-06, + "loss": 0.606, + "step": 6499 + }, + { + "epoch": 2.4, + "learning_rate": 2.7616701564232047e-06, + "loss": 0.5697, + "step": 6500 + }, + { + "epoch": 2.4, + "learning_rate": 2.7616332060598597e-06, + "loss": 0.545, + "step": 6501 + }, + { + "epoch": 2.4, + "learning_rate": 2.7615962556965144e-06, + "loss": 0.5973, + "step": 6502 + }, + { + "epoch": 2.4, + "learning_rate": 2.761559305333169e-06, + "loss": 0.6822, + "step": 6503 + }, + { + "epoch": 2.4, + "learning_rate": 2.761522354969824e-06, + "loss": 0.5886, + "step": 6504 + }, + { + "epoch": 2.4, + "learning_rate": 2.7614854046064787e-06, + "loss": 0.5747, + "step": 6505 + }, + { + "epoch": 2.4, + "learning_rate": 2.7614484542431333e-06, + "loss": 0.6354, + "step": 6506 + }, + { + "epoch": 2.4, + "learning_rate": 2.761411503879788e-06, + "loss": 0.6095, + "step": 6507 + }, + { + "epoch": 2.4, + "learning_rate": 2.761374553516443e-06, + "loss": 0.5647, + "step": 6508 + }, + { + "epoch": 2.4, + "learning_rate": 2.7613376031530977e-06, + "loss": 0.528, + "step": 6509 + }, + { + "epoch": 2.4, + "learning_rate": 2.7613006527897527e-06, + "loss": 0.5884, + "step": 6510 + }, + { + "epoch": 2.4, + "learning_rate": 2.7612637024264074e-06, + "loss": 0.6715, + "step": 6511 + }, + { + "epoch": 2.4, + "learning_rate": 2.761226752063062e-06, + "loss": 0.6267, + "step": 6512 + }, + { + "epoch": 2.41, + "learning_rate": 2.7611898016997166e-06, + "loss": 0.4847, + "step": 6513 + }, + { + "epoch": 2.41, + "learning_rate": 2.7611528513363713e-06, + "loss": 0.6439, + "step": 6514 + }, + { + "epoch": 2.41, + "learning_rate": 2.7611159009730263e-06, + "loss": 0.5816, + "step": 6515 + }, + { + "epoch": 2.41, + "learning_rate": 2.761078950609681e-06, + "loss": 0.5626, + "step": 6516 + }, + { + "epoch": 2.41, + "learning_rate": 2.761042000246336e-06, + "loss": 0.6438, + "step": 6517 + }, + { + "epoch": 2.41, + "learning_rate": 2.7610050498829907e-06, + "loss": 0.5847, + "step": 6518 + }, + { + "epoch": 2.41, + "learning_rate": 2.7609680995196453e-06, + "loss": 0.6001, + "step": 6519 + }, + { + "epoch": 2.41, + "learning_rate": 2.7609311491563e-06, + "loss": 0.6138, + "step": 6520 + }, + { + "epoch": 2.41, + "learning_rate": 2.760894198792955e-06, + "loss": 0.5943, + "step": 6521 + }, + { + "epoch": 2.41, + "learning_rate": 2.7608572484296096e-06, + "loss": 0.6407, + "step": 6522 + }, + { + "epoch": 2.41, + "learning_rate": 2.7608202980662643e-06, + "loss": 0.6472, + "step": 6523 + }, + { + "epoch": 2.41, + "learning_rate": 2.7607833477029193e-06, + "loss": 0.5173, + "step": 6524 + }, + { + "epoch": 2.41, + "learning_rate": 2.760746397339574e-06, + "loss": 0.6077, + "step": 6525 + }, + { + "epoch": 2.41, + "learning_rate": 2.7607094469762286e-06, + "loss": 0.6205, + "step": 6526 + }, + { + "epoch": 2.41, + "learning_rate": 2.7606724966128832e-06, + "loss": 0.5608, + "step": 6527 + }, + { + "epoch": 2.41, + "learning_rate": 2.7606355462495383e-06, + "loss": 0.5244, + "step": 6528 + }, + { + "epoch": 2.41, + "learning_rate": 2.760598595886193e-06, + "loss": 0.5518, + "step": 6529 + }, + { + "epoch": 2.41, + "learning_rate": 2.760561645522848e-06, + "loss": 0.6701, + "step": 6530 + }, + { + "epoch": 2.41, + "learning_rate": 2.7605246951595026e-06, + "loss": 0.6301, + "step": 6531 + }, + { + "epoch": 2.41, + "learning_rate": 2.7604877447961573e-06, + "loss": 0.5964, + "step": 6532 + }, + { + "epoch": 2.41, + "learning_rate": 2.760450794432812e-06, + "loss": 0.6201, + "step": 6533 + }, + { + "epoch": 2.41, + "learning_rate": 2.7604138440694665e-06, + "loss": 0.5103, + "step": 6534 + }, + { + "epoch": 2.41, + "learning_rate": 2.7603768937061216e-06, + "loss": 0.5099, + "step": 6535 + }, + { + "epoch": 2.41, + "learning_rate": 2.7603399433427762e-06, + "loss": 0.5643, + "step": 6536 + }, + { + "epoch": 2.41, + "learning_rate": 2.7603029929794313e-06, + "loss": 0.5604, + "step": 6537 + }, + { + "epoch": 2.41, + "learning_rate": 2.7602660426160855e-06, + "loss": 0.7363, + "step": 6538 + }, + { + "epoch": 2.41, + "learning_rate": 2.7602290922527406e-06, + "loss": 0.6312, + "step": 6539 + }, + { + "epoch": 2.42, + "learning_rate": 2.7601921418893952e-06, + "loss": 0.5298, + "step": 6540 + }, + { + "epoch": 2.42, + "learning_rate": 2.7601551915260503e-06, + "loss": 0.5821, + "step": 6541 + }, + { + "epoch": 2.42, + "learning_rate": 2.760118241162705e-06, + "loss": 0.6096, + "step": 6542 + }, + { + "epoch": 2.42, + "learning_rate": 2.7600812907993595e-06, + "loss": 0.5545, + "step": 6543 + }, + { + "epoch": 2.42, + "learning_rate": 2.760044340436014e-06, + "loss": 0.548, + "step": 6544 + }, + { + "epoch": 2.42, + "learning_rate": 2.760007390072669e-06, + "loss": 0.6927, + "step": 6545 + }, + { + "epoch": 2.42, + "learning_rate": 2.759970439709324e-06, + "loss": 0.5897, + "step": 6546 + }, + { + "epoch": 2.42, + "learning_rate": 2.7599334893459785e-06, + "loss": 0.6142, + "step": 6547 + }, + { + "epoch": 2.42, + "learning_rate": 2.7598965389826336e-06, + "loss": 0.511, + "step": 6548 + }, + { + "epoch": 2.42, + "learning_rate": 2.7598595886192882e-06, + "loss": 0.5769, + "step": 6549 + }, + { + "epoch": 2.42, + "learning_rate": 2.759822638255943e-06, + "loss": 0.6697, + "step": 6550 + }, + { + "epoch": 2.42, + "learning_rate": 2.7597856878925975e-06, + "loss": 0.523, + "step": 6551 + }, + { + "epoch": 2.42, + "learning_rate": 2.759748737529252e-06, + "loss": 0.5873, + "step": 6552 + }, + { + "epoch": 2.42, + "learning_rate": 2.759711787165907e-06, + "loss": 0.5864, + "step": 6553 + }, + { + "epoch": 2.42, + "learning_rate": 2.759674836802562e-06, + "loss": 0.5575, + "step": 6554 + }, + { + "epoch": 2.42, + "learning_rate": 2.759637886439217e-06, + "loss": 0.6519, + "step": 6555 + }, + { + "epoch": 2.42, + "learning_rate": 2.7596009360758715e-06, + "loss": 0.5466, + "step": 6556 + }, + { + "epoch": 2.42, + "learning_rate": 2.759563985712526e-06, + "loss": 0.5697, + "step": 6557 + }, + { + "epoch": 2.42, + "learning_rate": 2.759527035349181e-06, + "loss": 0.5083, + "step": 6558 + }, + { + "epoch": 2.42, + "learning_rate": 2.759490084985836e-06, + "loss": 0.6097, + "step": 6559 + }, + { + "epoch": 2.42, + "learning_rate": 2.7594531346224905e-06, + "loss": 0.6216, + "step": 6560 + }, + { + "epoch": 2.42, + "learning_rate": 2.7594161842591455e-06, + "loss": 0.5352, + "step": 6561 + }, + { + "epoch": 2.42, + "learning_rate": 2.7593792338958e-06, + "loss": 0.5843, + "step": 6562 + }, + { + "epoch": 2.42, + "learning_rate": 2.759342283532455e-06, + "loss": 0.5724, + "step": 6563 + }, + { + "epoch": 2.42, + "learning_rate": 2.7593053331691095e-06, + "loss": 0.5544, + "step": 6564 + }, + { + "epoch": 2.42, + "learning_rate": 2.759268382805764e-06, + "loss": 0.6567, + "step": 6565 + }, + { + "epoch": 2.42, + "learning_rate": 2.759231432442419e-06, + "loss": 0.696, + "step": 6566 + }, + { + "epoch": 2.43, + "learning_rate": 2.7591944820790738e-06, + "loss": 0.5684, + "step": 6567 + }, + { + "epoch": 2.43, + "learning_rate": 2.759157531715729e-06, + "loss": 0.5783, + "step": 6568 + }, + { + "epoch": 2.43, + "learning_rate": 2.7591205813523835e-06, + "loss": 0.6573, + "step": 6569 + }, + { + "epoch": 2.43, + "learning_rate": 2.759083630989038e-06, + "loss": 0.5624, + "step": 6570 + }, + { + "epoch": 2.43, + "learning_rate": 2.7590466806256928e-06, + "loss": 0.6003, + "step": 6571 + }, + { + "epoch": 2.43, + "learning_rate": 2.7590097302623474e-06, + "loss": 0.6254, + "step": 6572 + }, + { + "epoch": 2.43, + "learning_rate": 2.7589727798990025e-06, + "loss": 0.6205, + "step": 6573 + }, + { + "epoch": 2.43, + "learning_rate": 2.758935829535657e-06, + "loss": 0.6256, + "step": 6574 + }, + { + "epoch": 2.43, + "learning_rate": 2.758898879172312e-06, + "loss": 0.6042, + "step": 6575 + }, + { + "epoch": 2.43, + "learning_rate": 2.7588619288089668e-06, + "loss": 0.5532, + "step": 6576 + }, + { + "epoch": 2.43, + "learning_rate": 2.7588249784456214e-06, + "loss": 0.5729, + "step": 6577 + }, + { + "epoch": 2.43, + "learning_rate": 2.758788028082276e-06, + "loss": 0.6548, + "step": 6578 + }, + { + "epoch": 2.43, + "learning_rate": 2.758751077718931e-06, + "loss": 0.5828, + "step": 6579 + }, + { + "epoch": 2.43, + "learning_rate": 2.7587141273555858e-06, + "loss": 0.658, + "step": 6580 + }, + { + "epoch": 2.43, + "learning_rate": 2.758677176992241e-06, + "loss": 0.595, + "step": 6581 + }, + { + "epoch": 2.43, + "learning_rate": 2.7586402266288955e-06, + "loss": 0.5551, + "step": 6582 + }, + { + "epoch": 2.43, + "learning_rate": 2.7586032762655497e-06, + "loss": 0.6476, + "step": 6583 + }, + { + "epoch": 2.43, + "learning_rate": 2.7585663259022047e-06, + "loss": 0.6618, + "step": 6584 + }, + { + "epoch": 2.43, + "learning_rate": 2.7585293755388594e-06, + "loss": 0.5299, + "step": 6585 + }, + { + "epoch": 2.43, + "learning_rate": 2.7584924251755144e-06, + "loss": 0.6368, + "step": 6586 + }, + { + "epoch": 2.43, + "learning_rate": 2.758455474812169e-06, + "loss": 0.6597, + "step": 6587 + }, + { + "epoch": 2.43, + "learning_rate": 2.758418524448824e-06, + "loss": 0.6154, + "step": 6588 + }, + { + "epoch": 2.43, + "learning_rate": 2.7583815740854783e-06, + "loss": 0.543, + "step": 6589 + }, + { + "epoch": 2.43, + "learning_rate": 2.7583446237221334e-06, + "loss": 0.5424, + "step": 6590 + }, + { + "epoch": 2.43, + "learning_rate": 2.758307673358788e-06, + "loss": 0.6304, + "step": 6591 + }, + { + "epoch": 2.43, + "learning_rate": 2.7582707229954427e-06, + "loss": 0.5761, + "step": 6592 + }, + { + "epoch": 2.43, + "learning_rate": 2.7582337726320977e-06, + "loss": 0.556, + "step": 6593 + }, + { + "epoch": 2.44, + "learning_rate": 2.7581968222687524e-06, + "loss": 0.6264, + "step": 6594 + }, + { + "epoch": 2.44, + "learning_rate": 2.758159871905407e-06, + "loss": 0.5655, + "step": 6595 + }, + { + "epoch": 2.44, + "learning_rate": 2.7581229215420616e-06, + "loss": 0.562, + "step": 6596 + }, + { + "epoch": 2.44, + "learning_rate": 2.7580859711787167e-06, + "loss": 0.4636, + "step": 6597 + }, + { + "epoch": 2.44, + "learning_rate": 2.7580490208153713e-06, + "loss": 0.5893, + "step": 6598 + }, + { + "epoch": 2.44, + "learning_rate": 2.7580120704520264e-06, + "loss": 0.5772, + "step": 6599 + }, + { + "epoch": 2.44, + "learning_rate": 2.757975120088681e-06, + "loss": 0.6261, + "step": 6600 + }, + { + "epoch": 2.44, + "learning_rate": 2.7579381697253357e-06, + "loss": 0.5892, + "step": 6601 + }, + { + "epoch": 2.44, + "learning_rate": 2.7579012193619903e-06, + "loss": 0.6079, + "step": 6602 + }, + { + "epoch": 2.44, + "learning_rate": 2.757864268998645e-06, + "loss": 0.565, + "step": 6603 + }, + { + "epoch": 2.44, + "learning_rate": 2.7578273186353e-06, + "loss": 0.5978, + "step": 6604 + }, + { + "epoch": 2.44, + "learning_rate": 2.7577903682719546e-06, + "loss": 0.6427, + "step": 6605 + }, + { + "epoch": 2.44, + "learning_rate": 2.7577534179086097e-06, + "loss": 0.5644, + "step": 6606 + }, + { + "epoch": 2.44, + "learning_rate": 2.7577164675452643e-06, + "loss": 0.6447, + "step": 6607 + }, + { + "epoch": 2.44, + "learning_rate": 2.757679517181919e-06, + "loss": 0.654, + "step": 6608 + }, + { + "epoch": 2.44, + "learning_rate": 2.7576425668185736e-06, + "loss": 0.5652, + "step": 6609 + }, + { + "epoch": 2.44, + "learning_rate": 2.7576056164552287e-06, + "loss": 0.6134, + "step": 6610 + }, + { + "epoch": 2.44, + "learning_rate": 2.7575686660918833e-06, + "loss": 0.5882, + "step": 6611 + }, + { + "epoch": 2.44, + "learning_rate": 2.757531715728538e-06, + "loss": 0.5809, + "step": 6612 + }, + { + "epoch": 2.44, + "learning_rate": 2.757494765365193e-06, + "loss": 0.6161, + "step": 6613 + }, + { + "epoch": 2.44, + "learning_rate": 2.7574578150018476e-06, + "loss": 0.6086, + "step": 6614 + }, + { + "epoch": 2.44, + "learning_rate": 2.7574208646385023e-06, + "loss": 0.5442, + "step": 6615 + }, + { + "epoch": 2.44, + "learning_rate": 2.757383914275157e-06, + "loss": 0.651, + "step": 6616 + }, + { + "epoch": 2.44, + "learning_rate": 2.757346963911812e-06, + "loss": 0.5563, + "step": 6617 + }, + { + "epoch": 2.44, + "learning_rate": 2.7573100135484666e-06, + "loss": 0.615, + "step": 6618 + }, + { + "epoch": 2.44, + "learning_rate": 2.7572730631851217e-06, + "loss": 0.5324, + "step": 6619 + }, + { + "epoch": 2.44, + "learning_rate": 2.7572361128217763e-06, + "loss": 0.5632, + "step": 6620 + }, + { + "epoch": 2.44, + "learning_rate": 2.757199162458431e-06, + "loss": 0.6213, + "step": 6621 + }, + { + "epoch": 2.45, + "learning_rate": 2.7571622120950856e-06, + "loss": 0.6375, + "step": 6622 + }, + { + "epoch": 2.45, + "learning_rate": 2.75712526173174e-06, + "loss": 0.6126, + "step": 6623 + }, + { + "epoch": 2.45, + "learning_rate": 2.7570883113683953e-06, + "loss": 0.5385, + "step": 6624 + }, + { + "epoch": 2.45, + "learning_rate": 2.75705136100505e-06, + "loss": 0.5601, + "step": 6625 + }, + { + "epoch": 2.45, + "learning_rate": 2.757014410641705e-06, + "loss": 0.6899, + "step": 6626 + }, + { + "epoch": 2.45, + "learning_rate": 2.7569774602783596e-06, + "loss": 0.5829, + "step": 6627 + }, + { + "epoch": 2.45, + "learning_rate": 2.7569405099150142e-06, + "loss": 0.596, + "step": 6628 + }, + { + "epoch": 2.45, + "learning_rate": 2.756903559551669e-06, + "loss": 0.6277, + "step": 6629 + }, + { + "epoch": 2.45, + "learning_rate": 2.756866609188324e-06, + "loss": 0.5972, + "step": 6630 + }, + { + "epoch": 2.45, + "learning_rate": 2.7568296588249786e-06, + "loss": 0.6019, + "step": 6631 + }, + { + "epoch": 2.45, + "learning_rate": 2.756792708461633e-06, + "loss": 0.5757, + "step": 6632 + }, + { + "epoch": 2.45, + "learning_rate": 2.7567557580982883e-06, + "loss": 0.6594, + "step": 6633 + }, + { + "epoch": 2.45, + "learning_rate": 2.7567188077349425e-06, + "loss": 0.5531, + "step": 6634 + }, + { + "epoch": 2.45, + "learning_rate": 2.7566818573715975e-06, + "loss": 0.6456, + "step": 6635 + }, + { + "epoch": 2.45, + "learning_rate": 2.756644907008252e-06, + "loss": 0.5983, + "step": 6636 + }, + { + "epoch": 2.45, + "learning_rate": 2.7566079566449072e-06, + "loss": 0.5592, + "step": 6637 + }, + { + "epoch": 2.45, + "learning_rate": 2.756571006281562e-06, + "loss": 0.5949, + "step": 6638 + }, + { + "epoch": 2.45, + "learning_rate": 2.756534055918217e-06, + "loss": 0.5782, + "step": 6639 + }, + { + "epoch": 2.45, + "learning_rate": 2.756497105554871e-06, + "loss": 0.6002, + "step": 6640 + }, + { + "epoch": 2.45, + "learning_rate": 2.756460155191526e-06, + "loss": 0.6118, + "step": 6641 + }, + { + "epoch": 2.45, + "learning_rate": 2.756423204828181e-06, + "loss": 0.6293, + "step": 6642 + }, + { + "epoch": 2.45, + "learning_rate": 2.7563862544648355e-06, + "loss": 0.5195, + "step": 6643 + }, + { + "epoch": 2.45, + "learning_rate": 2.7563493041014905e-06, + "loss": 0.6037, + "step": 6644 + }, + { + "epoch": 2.45, + "learning_rate": 2.756312353738145e-06, + "loss": 0.4794, + "step": 6645 + }, + { + "epoch": 2.45, + "learning_rate": 2.7562754033748e-06, + "loss": 0.6772, + "step": 6646 + }, + { + "epoch": 2.45, + "learning_rate": 2.7562384530114544e-06, + "loss": 0.5777, + "step": 6647 + }, + { + "epoch": 2.45, + "learning_rate": 2.7562015026481095e-06, + "loss": 0.5819, + "step": 6648 + }, + { + "epoch": 2.46, + "learning_rate": 2.756164552284764e-06, + "loss": 0.6203, + "step": 6649 + }, + { + "epoch": 2.46, + "learning_rate": 2.756127601921419e-06, + "loss": 0.5134, + "step": 6650 + }, + { + "epoch": 2.46, + "learning_rate": 2.756090651558074e-06, + "loss": 0.5036, + "step": 6651 + }, + { + "epoch": 2.46, + "learning_rate": 2.7560537011947285e-06, + "loss": 0.597, + "step": 6652 + }, + { + "epoch": 2.46, + "learning_rate": 2.756016750831383e-06, + "loss": 0.5165, + "step": 6653 + }, + { + "epoch": 2.46, + "learning_rate": 2.7559798004680377e-06, + "loss": 0.5826, + "step": 6654 + }, + { + "epoch": 2.46, + "learning_rate": 2.755942850104693e-06, + "loss": 0.6093, + "step": 6655 + }, + { + "epoch": 2.46, + "learning_rate": 2.7559058997413474e-06, + "loss": 0.6081, + "step": 6656 + }, + { + "epoch": 2.46, + "learning_rate": 2.7558689493780025e-06, + "loss": 0.5323, + "step": 6657 + }, + { + "epoch": 2.46, + "learning_rate": 2.755831999014657e-06, + "loss": 0.5525, + "step": 6658 + }, + { + "epoch": 2.46, + "learning_rate": 2.7557950486513118e-06, + "loss": 0.4722, + "step": 6659 + }, + { + "epoch": 2.46, + "learning_rate": 2.7557580982879664e-06, + "loss": 0.5658, + "step": 6660 + }, + { + "epoch": 2.46, + "learning_rate": 2.7557211479246215e-06, + "loss": 0.6592, + "step": 6661 + }, + { + "epoch": 2.46, + "learning_rate": 2.755684197561276e-06, + "loss": 0.5612, + "step": 6662 + }, + { + "epoch": 2.46, + "learning_rate": 2.7556472471979307e-06, + "loss": 0.6516, + "step": 6663 + }, + { + "epoch": 2.46, + "learning_rate": 2.755610296834586e-06, + "loss": 0.5669, + "step": 6664 + }, + { + "epoch": 2.46, + "learning_rate": 2.7555733464712404e-06, + "loss": 0.5269, + "step": 6665 + }, + { + "epoch": 2.46, + "learning_rate": 2.755536396107895e-06, + "loss": 0.5996, + "step": 6666 + }, + { + "epoch": 2.46, + "learning_rate": 2.7554994457445497e-06, + "loss": 0.6474, + "step": 6667 + }, + { + "epoch": 2.46, + "learning_rate": 2.7554624953812048e-06, + "loss": 0.6394, + "step": 6668 + }, + { + "epoch": 2.46, + "learning_rate": 2.7554255450178594e-06, + "loss": 0.5232, + "step": 6669 + }, + { + "epoch": 2.46, + "learning_rate": 2.7553885946545145e-06, + "loss": 0.5821, + "step": 6670 + }, + { + "epoch": 2.46, + "learning_rate": 2.755351644291169e-06, + "loss": 0.5404, + "step": 6671 + }, + { + "epoch": 2.46, + "learning_rate": 2.7553146939278237e-06, + "loss": 0.5599, + "step": 6672 + }, + { + "epoch": 2.46, + "learning_rate": 2.7552777435644784e-06, + "loss": 0.5117, + "step": 6673 + }, + { + "epoch": 2.46, + "learning_rate": 2.755240793201133e-06, + "loss": 0.547, + "step": 6674 + }, + { + "epoch": 2.46, + "learning_rate": 2.755203842837788e-06, + "loss": 0.5854, + "step": 6675 + }, + { + "epoch": 2.47, + "learning_rate": 2.7551668924744427e-06, + "loss": 0.5972, + "step": 6676 + }, + { + "epoch": 2.47, + "learning_rate": 2.7551299421110978e-06, + "loss": 0.5587, + "step": 6677 + }, + { + "epoch": 2.47, + "learning_rate": 2.7550929917477524e-06, + "loss": 0.5419, + "step": 6678 + }, + { + "epoch": 2.47, + "learning_rate": 2.755056041384407e-06, + "loss": 0.5523, + "step": 6679 + }, + { + "epoch": 2.47, + "learning_rate": 2.7550190910210617e-06, + "loss": 0.6387, + "step": 6680 + }, + { + "epoch": 2.47, + "learning_rate": 2.7549821406577167e-06, + "loss": 0.5257, + "step": 6681 + }, + { + "epoch": 2.47, + "learning_rate": 2.7549451902943714e-06, + "loss": 0.6166, + "step": 6682 + }, + { + "epoch": 2.47, + "learning_rate": 2.754908239931026e-06, + "loss": 0.6274, + "step": 6683 + }, + { + "epoch": 2.47, + "learning_rate": 2.754871289567681e-06, + "loss": 0.6426, + "step": 6684 + }, + { + "epoch": 2.47, + "learning_rate": 2.7548343392043353e-06, + "loss": 0.6679, + "step": 6685 + }, + { + "epoch": 2.47, + "learning_rate": 2.7547973888409903e-06, + "loss": 0.6553, + "step": 6686 + }, + { + "epoch": 2.47, + "learning_rate": 2.754760438477645e-06, + "loss": 0.6088, + "step": 6687 + }, + { + "epoch": 2.47, + "learning_rate": 2.7547234881143e-06, + "loss": 0.6133, + "step": 6688 + }, + { + "epoch": 2.47, + "learning_rate": 2.7546865377509547e-06, + "loss": 0.5691, + "step": 6689 + }, + { + "epoch": 2.47, + "learning_rate": 2.7546495873876093e-06, + "loss": 0.6444, + "step": 6690 + }, + { + "epoch": 2.47, + "learning_rate": 2.754612637024264e-06, + "loss": 0.5492, + "step": 6691 + }, + { + "epoch": 2.47, + "learning_rate": 2.7545756866609186e-06, + "loss": 0.6257, + "step": 6692 + }, + { + "epoch": 2.47, + "learning_rate": 2.7545387362975736e-06, + "loss": 0.5724, + "step": 6693 + }, + { + "epoch": 2.47, + "learning_rate": 2.7545017859342283e-06, + "loss": 0.5625, + "step": 6694 + }, + { + "epoch": 2.47, + "learning_rate": 2.7544648355708833e-06, + "loss": 0.5383, + "step": 6695 + }, + { + "epoch": 2.47, + "learning_rate": 2.754427885207538e-06, + "loss": 0.6076, + "step": 6696 + }, + { + "epoch": 2.47, + "learning_rate": 2.7543909348441926e-06, + "loss": 0.6804, + "step": 6697 + }, + { + "epoch": 2.47, + "learning_rate": 2.7543539844808473e-06, + "loss": 0.7185, + "step": 6698 + }, + { + "epoch": 2.47, + "learning_rate": 2.7543170341175023e-06, + "loss": 0.6428, + "step": 6699 + }, + { + "epoch": 2.47, + "learning_rate": 2.754280083754157e-06, + "loss": 0.5046, + "step": 6700 + }, + { + "epoch": 2.47, + "learning_rate": 2.754243133390812e-06, + "loss": 0.6194, + "step": 6701 + }, + { + "epoch": 2.47, + "learning_rate": 2.7542061830274666e-06, + "loss": 0.5653, + "step": 6702 + }, + { + "epoch": 2.48, + "learning_rate": 2.7541692326641213e-06, + "loss": 0.6297, + "step": 6703 + }, + { + "epoch": 2.48, + "learning_rate": 2.754132282300776e-06, + "loss": 0.6407, + "step": 6704 + }, + { + "epoch": 2.48, + "learning_rate": 2.7540953319374306e-06, + "loss": 0.5676, + "step": 6705 + }, + { + "epoch": 2.48, + "learning_rate": 2.7540583815740856e-06, + "loss": 0.6042, + "step": 6706 + }, + { + "epoch": 2.48, + "learning_rate": 2.7540214312107403e-06, + "loss": 0.6404, + "step": 6707 + }, + { + "epoch": 2.48, + "learning_rate": 2.7539844808473953e-06, + "loss": 0.5771, + "step": 6708 + }, + { + "epoch": 2.48, + "learning_rate": 2.75394753048405e-06, + "loss": 0.4919, + "step": 6709 + }, + { + "epoch": 2.48, + "learning_rate": 2.7539105801207046e-06, + "loss": 0.5057, + "step": 6710 + }, + { + "epoch": 2.48, + "learning_rate": 2.7538736297573592e-06, + "loss": 0.492, + "step": 6711 + }, + { + "epoch": 2.48, + "learning_rate": 2.753836679394014e-06, + "loss": 0.5402, + "step": 6712 + }, + { + "epoch": 2.48, + "learning_rate": 2.753799729030669e-06, + "loss": 0.5088, + "step": 6713 + }, + { + "epoch": 2.48, + "learning_rate": 2.7537627786673236e-06, + "loss": 0.5706, + "step": 6714 + }, + { + "epoch": 2.48, + "learning_rate": 2.7537258283039786e-06, + "loss": 0.6526, + "step": 6715 + }, + { + "epoch": 2.48, + "learning_rate": 2.7536888779406333e-06, + "loss": 0.6171, + "step": 6716 + }, + { + "epoch": 2.48, + "learning_rate": 2.753651927577288e-06, + "loss": 0.6267, + "step": 6717 + }, + { + "epoch": 2.48, + "learning_rate": 2.7536149772139425e-06, + "loss": 0.639, + "step": 6718 + }, + { + "epoch": 2.48, + "learning_rate": 2.7535780268505976e-06, + "loss": 0.7012, + "step": 6719 + }, + { + "epoch": 2.48, + "learning_rate": 2.7535410764872522e-06, + "loss": 0.6038, + "step": 6720 + }, + { + "epoch": 2.48, + "learning_rate": 2.7535041261239073e-06, + "loss": 0.5849, + "step": 6721 + }, + { + "epoch": 2.48, + "learning_rate": 2.753467175760562e-06, + "loss": 0.524, + "step": 6722 + }, + { + "epoch": 2.48, + "learning_rate": 2.753430225397216e-06, + "loss": 0.6464, + "step": 6723 + }, + { + "epoch": 2.48, + "learning_rate": 2.753393275033871e-06, + "loss": 0.5592, + "step": 6724 + }, + { + "epoch": 2.48, + "learning_rate": 2.753356324670526e-06, + "loss": 0.4929, + "step": 6725 + }, + { + "epoch": 2.48, + "learning_rate": 2.753319374307181e-06, + "loss": 0.6188, + "step": 6726 + }, + { + "epoch": 2.48, + "learning_rate": 2.7532824239438355e-06, + "loss": 0.648, + "step": 6727 + }, + { + "epoch": 2.48, + "learning_rate": 2.7532454735804906e-06, + "loss": 0.6311, + "step": 6728 + }, + { + "epoch": 2.48, + "learning_rate": 2.753208523217145e-06, + "loss": 0.6037, + "step": 6729 + }, + { + "epoch": 2.49, + "learning_rate": 2.7531715728538e-06, + "loss": 0.5576, + "step": 6730 + }, + { + "epoch": 2.49, + "learning_rate": 2.7531346224904545e-06, + "loss": 0.6027, + "step": 6731 + }, + { + "epoch": 2.49, + "learning_rate": 2.753097672127109e-06, + "loss": 0.6634, + "step": 6732 + }, + { + "epoch": 2.49, + "learning_rate": 2.753060721763764e-06, + "loss": 0.6061, + "step": 6733 + }, + { + "epoch": 2.49, + "learning_rate": 2.753023771400419e-06, + "loss": 0.5861, + "step": 6734 + }, + { + "epoch": 2.49, + "learning_rate": 2.7529868210370735e-06, + "loss": 0.527, + "step": 6735 + }, + { + "epoch": 2.49, + "learning_rate": 2.752949870673728e-06, + "loss": 0.581, + "step": 6736 + }, + { + "epoch": 2.49, + "learning_rate": 2.752912920310383e-06, + "loss": 0.6509, + "step": 6737 + }, + { + "epoch": 2.49, + "learning_rate": 2.752875969947038e-06, + "loss": 0.5682, + "step": 6738 + }, + { + "epoch": 2.49, + "learning_rate": 2.752839019583693e-06, + "loss": 0.5827, + "step": 6739 + }, + { + "epoch": 2.49, + "learning_rate": 2.7528020692203475e-06, + "loss": 0.5701, + "step": 6740 + }, + { + "epoch": 2.49, + "learning_rate": 2.752765118857002e-06, + "loss": 0.5172, + "step": 6741 + }, + { + "epoch": 2.49, + "learning_rate": 2.7527281684936568e-06, + "loss": 0.5378, + "step": 6742 + }, + { + "epoch": 2.49, + "learning_rate": 2.7526912181303114e-06, + "loss": 0.6386, + "step": 6743 + }, + { + "epoch": 2.49, + "learning_rate": 2.7526542677669665e-06, + "loss": 0.5763, + "step": 6744 + }, + { + "epoch": 2.49, + "learning_rate": 2.752617317403621e-06, + "loss": 0.5962, + "step": 6745 + }, + { + "epoch": 2.49, + "learning_rate": 2.752580367040276e-06, + "loss": 0.5888, + "step": 6746 + }, + { + "epoch": 2.49, + "learning_rate": 2.752543416676931e-06, + "loss": 0.5452, + "step": 6747 + }, + { + "epoch": 2.49, + "learning_rate": 2.7525064663135854e-06, + "loss": 0.5415, + "step": 6748 + }, + { + "epoch": 2.49, + "learning_rate": 2.75246951595024e-06, + "loss": 0.6121, + "step": 6749 + }, + { + "epoch": 2.49, + "learning_rate": 2.752432565586895e-06, + "loss": 0.6173, + "step": 6750 + }, + { + "epoch": 2.49, + "learning_rate": 2.7523956152235498e-06, + "loss": 0.523, + "step": 6751 + }, + { + "epoch": 2.49, + "learning_rate": 2.7523586648602044e-06, + "loss": 0.5139, + "step": 6752 + }, + { + "epoch": 2.49, + "learning_rate": 2.7523217144968595e-06, + "loss": 0.7478, + "step": 6753 + }, + { + "epoch": 2.49, + "learning_rate": 2.752284764133514e-06, + "loss": 0.6182, + "step": 6754 + }, + { + "epoch": 2.49, + "learning_rate": 2.7522478137701687e-06, + "loss": 0.7248, + "step": 6755 + }, + { + "epoch": 2.49, + "learning_rate": 2.7522108634068234e-06, + "loss": 0.6009, + "step": 6756 + }, + { + "epoch": 2.5, + "learning_rate": 2.7521739130434784e-06, + "loss": 0.5483, + "step": 6757 + }, + { + "epoch": 2.5, + "learning_rate": 2.752136962680133e-06, + "loss": 0.5282, + "step": 6758 + }, + { + "epoch": 2.5, + "learning_rate": 2.752100012316788e-06, + "loss": 0.6202, + "step": 6759 + }, + { + "epoch": 2.5, + "learning_rate": 2.7520630619534428e-06, + "loss": 0.6276, + "step": 6760 + }, + { + "epoch": 2.5, + "learning_rate": 2.7520261115900974e-06, + "loss": 0.6525, + "step": 6761 + }, + { + "epoch": 2.5, + "learning_rate": 2.751989161226752e-06, + "loss": 0.5932, + "step": 6762 + }, + { + "epoch": 2.5, + "learning_rate": 2.7519522108634067e-06, + "loss": 0.6995, + "step": 6763 + }, + { + "epoch": 2.5, + "learning_rate": 2.7519152605000617e-06, + "loss": 0.4987, + "step": 6764 + }, + { + "epoch": 2.5, + "learning_rate": 2.7518783101367164e-06, + "loss": 0.5992, + "step": 6765 + }, + { + "epoch": 2.5, + "learning_rate": 2.7518413597733714e-06, + "loss": 0.5622, + "step": 6766 + }, + { + "epoch": 2.5, + "learning_rate": 2.751804409410026e-06, + "loss": 0.6438, + "step": 6767 + }, + { + "epoch": 2.5, + "learning_rate": 2.7517674590466807e-06, + "loss": 0.6018, + "step": 6768 + }, + { + "epoch": 2.5, + "learning_rate": 2.7517305086833353e-06, + "loss": 0.5924, + "step": 6769 + }, + { + "epoch": 2.5, + "learning_rate": 2.7516935583199904e-06, + "loss": 0.6016, + "step": 6770 + }, + { + "epoch": 2.5, + "learning_rate": 2.751656607956645e-06, + "loss": 0.74, + "step": 6771 + }, + { + "epoch": 2.5, + "learning_rate": 2.7516196575932997e-06, + "loss": 0.6208, + "step": 6772 + }, + { + "epoch": 2.5, + "learning_rate": 2.7515827072299547e-06, + "loss": 0.6438, + "step": 6773 + }, + { + "epoch": 2.5, + "learning_rate": 2.751545756866609e-06, + "loss": 0.5871, + "step": 6774 + }, + { + "epoch": 2.5, + "learning_rate": 2.751508806503264e-06, + "loss": 0.5163, + "step": 6775 + }, + { + "epoch": 2.5, + "learning_rate": 2.7514718561399186e-06, + "loss": 0.6158, + "step": 6776 + }, + { + "epoch": 2.5, + "learning_rate": 2.7514349057765737e-06, + "loss": 0.6401, + "step": 6777 + }, + { + "epoch": 2.5, + "learning_rate": 2.7513979554132283e-06, + "loss": 0.5671, + "step": 6778 + }, + { + "epoch": 2.5, + "learning_rate": 2.7513610050498834e-06, + "loss": 0.6108, + "step": 6779 + }, + { + "epoch": 2.5, + "learning_rate": 2.7513240546865376e-06, + "loss": 0.5163, + "step": 6780 + }, + { + "epoch": 2.5, + "learning_rate": 2.7512871043231922e-06, + "loss": 0.6119, + "step": 6781 + }, + { + "epoch": 2.5, + "learning_rate": 2.7512501539598473e-06, + "loss": 0.5809, + "step": 6782 + }, + { + "epoch": 2.5, + "learning_rate": 2.751213203596502e-06, + "loss": 0.5655, + "step": 6783 + }, + { + "epoch": 2.51, + "learning_rate": 2.751176253233157e-06, + "loss": 0.5433, + "step": 6784 + }, + { + "epoch": 2.51, + "learning_rate": 2.7511393028698116e-06, + "loss": 0.5747, + "step": 6785 + }, + { + "epoch": 2.51, + "learning_rate": 2.7511023525064663e-06, + "loss": 0.5593, + "step": 6786 + }, + { + "epoch": 2.51, + "learning_rate": 2.751065402143121e-06, + "loss": 0.6229, + "step": 6787 + }, + { + "epoch": 2.51, + "learning_rate": 2.751028451779776e-06, + "loss": 0.6087, + "step": 6788 + }, + { + "epoch": 2.51, + "learning_rate": 2.7509915014164306e-06, + "loss": 0.5966, + "step": 6789 + }, + { + "epoch": 2.51, + "learning_rate": 2.7509545510530857e-06, + "loss": 0.6195, + "step": 6790 + }, + { + "epoch": 2.51, + "learning_rate": 2.7509176006897403e-06, + "loss": 0.5819, + "step": 6791 + }, + { + "epoch": 2.51, + "learning_rate": 2.750880650326395e-06, + "loss": 0.6367, + "step": 6792 + }, + { + "epoch": 2.51, + "learning_rate": 2.7508436999630496e-06, + "loss": 0.5465, + "step": 6793 + }, + { + "epoch": 2.51, + "learning_rate": 2.750806749599704e-06, + "loss": 0.5964, + "step": 6794 + }, + { + "epoch": 2.51, + "learning_rate": 2.7507697992363593e-06, + "loss": 0.583, + "step": 6795 + }, + { + "epoch": 2.51, + "learning_rate": 2.750732848873014e-06, + "loss": 0.6954, + "step": 6796 + }, + { + "epoch": 2.51, + "learning_rate": 2.750695898509669e-06, + "loss": 0.5962, + "step": 6797 + }, + { + "epoch": 2.51, + "learning_rate": 2.7506589481463236e-06, + "loss": 0.5676, + "step": 6798 + }, + { + "epoch": 2.51, + "learning_rate": 2.7506219977829782e-06, + "loss": 0.6195, + "step": 6799 + }, + { + "epoch": 2.51, + "learning_rate": 2.750585047419633e-06, + "loss": 0.5872, + "step": 6800 + }, + { + "epoch": 2.51, + "learning_rate": 2.7505480970562875e-06, + "loss": 0.6381, + "step": 6801 + }, + { + "epoch": 2.51, + "learning_rate": 2.7505111466929426e-06, + "loss": 0.5679, + "step": 6802 + }, + { + "epoch": 2.51, + "learning_rate": 2.750474196329597e-06, + "loss": 0.5178, + "step": 6803 + }, + { + "epoch": 2.51, + "learning_rate": 2.7504372459662523e-06, + "loss": 0.5547, + "step": 6804 + }, + { + "epoch": 2.51, + "learning_rate": 2.750400295602907e-06, + "loss": 0.5856, + "step": 6805 + }, + { + "epoch": 2.51, + "learning_rate": 2.7503633452395615e-06, + "loss": 0.5432, + "step": 6806 + }, + { + "epoch": 2.51, + "learning_rate": 2.750326394876216e-06, + "loss": 0.6455, + "step": 6807 + }, + { + "epoch": 2.51, + "learning_rate": 2.7502894445128712e-06, + "loss": 0.5668, + "step": 6808 + }, + { + "epoch": 2.51, + "learning_rate": 2.750252494149526e-06, + "loss": 0.6511, + "step": 6809 + }, + { + "epoch": 2.51, + "learning_rate": 2.750215543786181e-06, + "loss": 0.5772, + "step": 6810 + }, + { + "epoch": 2.52, + "learning_rate": 2.7501785934228356e-06, + "loss": 0.5771, + "step": 6811 + }, + { + "epoch": 2.52, + "learning_rate": 2.75014164305949e-06, + "loss": 0.5744, + "step": 6812 + }, + { + "epoch": 2.52, + "learning_rate": 2.750104692696145e-06, + "loss": 0.603, + "step": 6813 + }, + { + "epoch": 2.52, + "learning_rate": 2.7500677423327995e-06, + "loss": 0.5279, + "step": 6814 + }, + { + "epoch": 2.52, + "learning_rate": 2.7500307919694545e-06, + "loss": 0.6325, + "step": 6815 + }, + { + "epoch": 2.52, + "learning_rate": 2.749993841606109e-06, + "loss": 0.5695, + "step": 6816 + }, + { + "epoch": 2.52, + "learning_rate": 2.7499568912427642e-06, + "loss": 0.6026, + "step": 6817 + }, + { + "epoch": 2.52, + "learning_rate": 2.749919940879419e-06, + "loss": 0.5638, + "step": 6818 + }, + { + "epoch": 2.52, + "learning_rate": 2.7498829905160735e-06, + "loss": 0.537, + "step": 6819 + }, + { + "epoch": 2.52, + "learning_rate": 2.749846040152728e-06, + "loss": 0.5711, + "step": 6820 + }, + { + "epoch": 2.52, + "learning_rate": 2.7498090897893828e-06, + "loss": 0.4908, + "step": 6821 + }, + { + "epoch": 2.52, + "learning_rate": 2.749772139426038e-06, + "loss": 0.5791, + "step": 6822 + }, + { + "epoch": 2.52, + "learning_rate": 2.7497351890626925e-06, + "loss": 0.6028, + "step": 6823 + }, + { + "epoch": 2.52, + "learning_rate": 2.7496982386993475e-06, + "loss": 0.6194, + "step": 6824 + }, + { + "epoch": 2.52, + "learning_rate": 2.7496612883360018e-06, + "loss": 0.6708, + "step": 6825 + }, + { + "epoch": 2.52, + "learning_rate": 2.749624337972657e-06, + "loss": 0.5627, + "step": 6826 + }, + { + "epoch": 2.52, + "learning_rate": 2.7495873876093114e-06, + "loss": 0.4977, + "step": 6827 + }, + { + "epoch": 2.52, + "learning_rate": 2.7495504372459665e-06, + "loss": 0.5605, + "step": 6828 + }, + { + "epoch": 2.52, + "learning_rate": 2.749513486882621e-06, + "loss": 0.5822, + "step": 6829 + }, + { + "epoch": 2.52, + "learning_rate": 2.749476536519276e-06, + "loss": 0.5778, + "step": 6830 + }, + { + "epoch": 2.52, + "learning_rate": 2.7494395861559304e-06, + "loss": 0.5977, + "step": 6831 + }, + { + "epoch": 2.52, + "learning_rate": 2.749402635792585e-06, + "loss": 0.6443, + "step": 6832 + }, + { + "epoch": 2.52, + "learning_rate": 2.74936568542924e-06, + "loss": 0.609, + "step": 6833 + }, + { + "epoch": 2.52, + "learning_rate": 2.7493287350658947e-06, + "loss": 0.5658, + "step": 6834 + }, + { + "epoch": 2.52, + "learning_rate": 2.74929178470255e-06, + "loss": 0.6265, + "step": 6835 + }, + { + "epoch": 2.52, + "learning_rate": 2.7492548343392044e-06, + "loss": 0.5088, + "step": 6836 + }, + { + "epoch": 2.52, + "learning_rate": 2.749217883975859e-06, + "loss": 0.5617, + "step": 6837 + }, + { + "epoch": 2.53, + "learning_rate": 2.7491809336125137e-06, + "loss": 0.6732, + "step": 6838 + }, + { + "epoch": 2.53, + "learning_rate": 2.7491439832491688e-06, + "loss": 0.5809, + "step": 6839 + }, + { + "epoch": 2.53, + "learning_rate": 2.7491070328858234e-06, + "loss": 0.6094, + "step": 6840 + }, + { + "epoch": 2.53, + "learning_rate": 2.749070082522478e-06, + "loss": 0.6743, + "step": 6841 + }, + { + "epoch": 2.53, + "learning_rate": 2.749033132159133e-06, + "loss": 0.6428, + "step": 6842 + }, + { + "epoch": 2.53, + "learning_rate": 2.7489961817957877e-06, + "loss": 0.4807, + "step": 6843 + }, + { + "epoch": 2.53, + "learning_rate": 2.7489592314324424e-06, + "loss": 0.5062, + "step": 6844 + }, + { + "epoch": 2.53, + "learning_rate": 2.748922281069097e-06, + "loss": 0.4932, + "step": 6845 + }, + { + "epoch": 2.53, + "learning_rate": 2.748885330705752e-06, + "loss": 0.5996, + "step": 6846 + }, + { + "epoch": 2.53, + "learning_rate": 2.7488483803424067e-06, + "loss": 0.5731, + "step": 6847 + }, + { + "epoch": 2.53, + "learning_rate": 2.7488114299790618e-06, + "loss": 0.489, + "step": 6848 + }, + { + "epoch": 2.53, + "learning_rate": 2.7487744796157164e-06, + "loss": 0.6046, + "step": 6849 + }, + { + "epoch": 2.53, + "learning_rate": 2.748737529252371e-06, + "loss": 0.5752, + "step": 6850 + }, + { + "epoch": 2.53, + "learning_rate": 2.7487005788890257e-06, + "loss": 0.5946, + "step": 6851 + }, + { + "epoch": 2.53, + "learning_rate": 2.7486636285256803e-06, + "loss": 0.5444, + "step": 6852 + }, + { + "epoch": 2.53, + "learning_rate": 2.7486266781623354e-06, + "loss": 0.5384, + "step": 6853 + }, + { + "epoch": 2.53, + "learning_rate": 2.74858972779899e-06, + "loss": 0.6101, + "step": 6854 + }, + { + "epoch": 2.53, + "learning_rate": 2.748552777435645e-06, + "loss": 0.5854, + "step": 6855 + }, + { + "epoch": 2.53, + "learning_rate": 2.7485158270722997e-06, + "loss": 0.5742, + "step": 6856 + }, + { + "epoch": 2.53, + "learning_rate": 2.7484788767089544e-06, + "loss": 0.6175, + "step": 6857 + }, + { + "epoch": 2.53, + "learning_rate": 2.748441926345609e-06, + "loss": 0.5746, + "step": 6858 + }, + { + "epoch": 2.53, + "learning_rate": 2.748404975982264e-06, + "loss": 0.6434, + "step": 6859 + }, + { + "epoch": 2.53, + "learning_rate": 2.7483680256189187e-06, + "loss": 0.5892, + "step": 6860 + }, + { + "epoch": 2.53, + "learning_rate": 2.7483310752555733e-06, + "loss": 0.6152, + "step": 6861 + }, + { + "epoch": 2.53, + "learning_rate": 2.7482941248922284e-06, + "loss": 0.584, + "step": 6862 + }, + { + "epoch": 2.53, + "learning_rate": 2.748257174528883e-06, + "loss": 0.5293, + "step": 6863 + }, + { + "epoch": 2.53, + "learning_rate": 2.7482202241655377e-06, + "loss": 0.5439, + "step": 6864 + }, + { + "epoch": 2.54, + "learning_rate": 2.7481832738021923e-06, + "loss": 0.6256, + "step": 6865 + }, + { + "epoch": 2.54, + "learning_rate": 2.7481463234388474e-06, + "loss": 0.6241, + "step": 6866 + }, + { + "epoch": 2.54, + "learning_rate": 2.748109373075502e-06, + "loss": 0.5862, + "step": 6867 + }, + { + "epoch": 2.54, + "learning_rate": 2.748072422712157e-06, + "loss": 0.6391, + "step": 6868 + }, + { + "epoch": 2.54, + "learning_rate": 2.7480354723488117e-06, + "loss": 0.6063, + "step": 6869 + }, + { + "epoch": 2.54, + "learning_rate": 2.7479985219854663e-06, + "loss": 0.5747, + "step": 6870 + }, + { + "epoch": 2.54, + "learning_rate": 2.747961571622121e-06, + "loss": 0.588, + "step": 6871 + }, + { + "epoch": 2.54, + "learning_rate": 2.7479246212587756e-06, + "loss": 0.5924, + "step": 6872 + }, + { + "epoch": 2.54, + "learning_rate": 2.7478876708954307e-06, + "loss": 0.6076, + "step": 6873 + }, + { + "epoch": 2.54, + "learning_rate": 2.7478507205320853e-06, + "loss": 0.5511, + "step": 6874 + }, + { + "epoch": 2.54, + "learning_rate": 2.74781377016874e-06, + "loss": 0.5103, + "step": 6875 + }, + { + "epoch": 2.54, + "learning_rate": 2.7477768198053946e-06, + "loss": 0.6861, + "step": 6876 + }, + { + "epoch": 2.54, + "learning_rate": 2.7477398694420496e-06, + "loss": 0.5755, + "step": 6877 + }, + { + "epoch": 2.54, + "learning_rate": 2.7477029190787043e-06, + "loss": 0.6557, + "step": 6878 + }, + { + "epoch": 2.54, + "learning_rate": 2.7476659687153593e-06, + "loss": 0.5782, + "step": 6879 + }, + { + "epoch": 2.54, + "learning_rate": 2.747629018352014e-06, + "loss": 0.5577, + "step": 6880 + }, + { + "epoch": 2.54, + "learning_rate": 2.7475920679886686e-06, + "loss": 0.6351, + "step": 6881 + }, + { + "epoch": 2.54, + "learning_rate": 2.7475551176253232e-06, + "loss": 0.5509, + "step": 6882 + }, + { + "epoch": 2.54, + "learning_rate": 2.747518167261978e-06, + "loss": 0.6487, + "step": 6883 + }, + { + "epoch": 2.54, + "learning_rate": 2.747481216898633e-06, + "loss": 0.6014, + "step": 6884 + }, + { + "epoch": 2.54, + "learning_rate": 2.7474442665352876e-06, + "loss": 0.5473, + "step": 6885 + }, + { + "epoch": 2.54, + "learning_rate": 2.7474073161719426e-06, + "loss": 0.4771, + "step": 6886 + }, + { + "epoch": 2.54, + "learning_rate": 2.7473703658085973e-06, + "loss": 0.5632, + "step": 6887 + }, + { + "epoch": 2.54, + "learning_rate": 2.747333415445252e-06, + "loss": 0.5955, + "step": 6888 + }, + { + "epoch": 2.54, + "learning_rate": 2.7472964650819065e-06, + "loss": 0.6858, + "step": 6889 + }, + { + "epoch": 2.54, + "learning_rate": 2.7472595147185616e-06, + "loss": 0.514, + "step": 6890 + }, + { + "epoch": 2.54, + "learning_rate": 2.7472225643552162e-06, + "loss": 0.5902, + "step": 6891 + }, + { + "epoch": 2.55, + "learning_rate": 2.747185613991871e-06, + "loss": 0.5191, + "step": 6892 + }, + { + "epoch": 2.55, + "learning_rate": 2.747148663628526e-06, + "loss": 0.6736, + "step": 6893 + }, + { + "epoch": 2.55, + "learning_rate": 2.7471117132651806e-06, + "loss": 0.5561, + "step": 6894 + }, + { + "epoch": 2.55, + "learning_rate": 2.747074762901835e-06, + "loss": 0.6148, + "step": 6895 + }, + { + "epoch": 2.55, + "learning_rate": 2.74703781253849e-06, + "loss": 0.6772, + "step": 6896 + }, + { + "epoch": 2.55, + "learning_rate": 2.747000862175145e-06, + "loss": 0.5669, + "step": 6897 + }, + { + "epoch": 2.55, + "learning_rate": 2.7469639118117995e-06, + "loss": 0.5132, + "step": 6898 + }, + { + "epoch": 2.55, + "learning_rate": 2.7469269614484546e-06, + "loss": 0.6487, + "step": 6899 + }, + { + "epoch": 2.55, + "learning_rate": 2.7468900110851092e-06, + "loss": 0.6772, + "step": 6900 + }, + { + "epoch": 2.55, + "learning_rate": 2.746853060721764e-06, + "loss": 0.5473, + "step": 6901 + }, + { + "epoch": 2.55, + "learning_rate": 2.7468161103584185e-06, + "loss": 0.5142, + "step": 6902 + }, + { + "epoch": 2.55, + "learning_rate": 2.746779159995073e-06, + "loss": 0.6417, + "step": 6903 + }, + { + "epoch": 2.55, + "learning_rate": 2.746742209631728e-06, + "loss": 0.6025, + "step": 6904 + }, + { + "epoch": 2.55, + "learning_rate": 2.746705259268383e-06, + "loss": 0.6839, + "step": 6905 + }, + { + "epoch": 2.55, + "learning_rate": 2.746668308905038e-06, + "loss": 0.5691, + "step": 6906 + }, + { + "epoch": 2.55, + "learning_rate": 2.7466313585416925e-06, + "loss": 0.5251, + "step": 6907 + }, + { + "epoch": 2.55, + "learning_rate": 2.746594408178347e-06, + "loss": 0.5351, + "step": 6908 + }, + { + "epoch": 2.55, + "learning_rate": 2.746557457815002e-06, + "loss": 0.5755, + "step": 6909 + }, + { + "epoch": 2.55, + "learning_rate": 2.746520507451657e-06, + "loss": 0.5232, + "step": 6910 + }, + { + "epoch": 2.55, + "learning_rate": 2.7464835570883115e-06, + "loss": 0.4681, + "step": 6911 + }, + { + "epoch": 2.55, + "learning_rate": 2.746446606724966e-06, + "loss": 0.596, + "step": 6912 + }, + { + "epoch": 2.55, + "learning_rate": 2.746409656361621e-06, + "loss": 0.6457, + "step": 6913 + }, + { + "epoch": 2.55, + "learning_rate": 2.7463727059982754e-06, + "loss": 0.5832, + "step": 6914 + }, + { + "epoch": 2.55, + "learning_rate": 2.7463357556349305e-06, + "loss": 0.6215, + "step": 6915 + }, + { + "epoch": 2.55, + "learning_rate": 2.746298805271585e-06, + "loss": 0.6222, + "step": 6916 + }, + { + "epoch": 2.55, + "learning_rate": 2.74626185490824e-06, + "loss": 0.6385, + "step": 6917 + }, + { + "epoch": 2.55, + "learning_rate": 2.746224904544895e-06, + "loss": 0.6058, + "step": 6918 + }, + { + "epoch": 2.56, + "learning_rate": 2.74618795418155e-06, + "loss": 0.6466, + "step": 6919 + }, + { + "epoch": 2.56, + "learning_rate": 2.746151003818204e-06, + "loss": 0.5749, + "step": 6920 + }, + { + "epoch": 2.56, + "learning_rate": 2.7461140534548587e-06, + "loss": 0.6198, + "step": 6921 + }, + { + "epoch": 2.56, + "learning_rate": 2.7460771030915138e-06, + "loss": 0.5834, + "step": 6922 + }, + { + "epoch": 2.56, + "learning_rate": 2.7460401527281684e-06, + "loss": 0.588, + "step": 6923 + }, + { + "epoch": 2.56, + "learning_rate": 2.7460032023648235e-06, + "loss": 0.6206, + "step": 6924 + }, + { + "epoch": 2.56, + "learning_rate": 2.745966252001478e-06, + "loss": 0.6039, + "step": 6925 + }, + { + "epoch": 2.56, + "learning_rate": 2.7459293016381327e-06, + "loss": 0.5708, + "step": 6926 + }, + { + "epoch": 2.56, + "learning_rate": 2.7458923512747874e-06, + "loss": 0.6373, + "step": 6927 + }, + { + "epoch": 2.56, + "learning_rate": 2.7458554009114424e-06, + "loss": 0.5341, + "step": 6928 + }, + { + "epoch": 2.56, + "learning_rate": 2.745818450548097e-06, + "loss": 0.5665, + "step": 6929 + }, + { + "epoch": 2.56, + "learning_rate": 2.745781500184752e-06, + "loss": 0.6017, + "step": 6930 + }, + { + "epoch": 2.56, + "learning_rate": 2.7457445498214068e-06, + "loss": 0.5818, + "step": 6931 + }, + { + "epoch": 2.56, + "learning_rate": 2.7457075994580614e-06, + "loss": 0.5712, + "step": 6932 + }, + { + "epoch": 2.56, + "learning_rate": 2.745670649094716e-06, + "loss": 0.5421, + "step": 6933 + }, + { + "epoch": 2.56, + "learning_rate": 2.7456336987313707e-06, + "loss": 0.5618, + "step": 6934 + }, + { + "epoch": 2.56, + "learning_rate": 2.7455967483680257e-06, + "loss": 0.6046, + "step": 6935 + }, + { + "epoch": 2.56, + "learning_rate": 2.7455597980046804e-06, + "loss": 0.5855, + "step": 6936 + }, + { + "epoch": 2.56, + "learning_rate": 2.7455228476413354e-06, + "loss": 0.5532, + "step": 6937 + }, + { + "epoch": 2.56, + "learning_rate": 2.74548589727799e-06, + "loss": 0.6921, + "step": 6938 + }, + { + "epoch": 2.56, + "learning_rate": 2.7454489469146447e-06, + "loss": 0.557, + "step": 6939 + }, + { + "epoch": 2.56, + "learning_rate": 2.7454119965512993e-06, + "loss": 0.619, + "step": 6940 + }, + { + "epoch": 2.56, + "learning_rate": 2.745375046187954e-06, + "loss": 0.5967, + "step": 6941 + }, + { + "epoch": 2.56, + "learning_rate": 2.745338095824609e-06, + "loss": 0.5781, + "step": 6942 + }, + { + "epoch": 2.56, + "learning_rate": 2.7453011454612637e-06, + "loss": 0.6138, + "step": 6943 + }, + { + "epoch": 2.56, + "learning_rate": 2.7452641950979187e-06, + "loss": 0.567, + "step": 6944 + }, + { + "epoch": 2.56, + "learning_rate": 2.7452272447345734e-06, + "loss": 0.5602, + "step": 6945 + }, + { + "epoch": 2.56, + "learning_rate": 2.745190294371228e-06, + "loss": 0.6323, + "step": 6946 + }, + { + "epoch": 2.57, + "learning_rate": 2.7451533440078826e-06, + "loss": 0.6286, + "step": 6947 + }, + { + "epoch": 2.57, + "learning_rate": 2.7451163936445377e-06, + "loss": 0.5712, + "step": 6948 + }, + { + "epoch": 2.57, + "learning_rate": 2.7450794432811923e-06, + "loss": 0.4862, + "step": 6949 + }, + { + "epoch": 2.57, + "learning_rate": 2.7450424929178474e-06, + "loss": 0.5804, + "step": 6950 + }, + { + "epoch": 2.57, + "learning_rate": 2.745005542554502e-06, + "loss": 0.6266, + "step": 6951 + }, + { + "epoch": 2.57, + "learning_rate": 2.7449685921911567e-06, + "loss": 0.6668, + "step": 6952 + }, + { + "epoch": 2.57, + "learning_rate": 2.7449316418278113e-06, + "loss": 0.5057, + "step": 6953 + }, + { + "epoch": 2.57, + "learning_rate": 2.744894691464466e-06, + "loss": 0.6038, + "step": 6954 + }, + { + "epoch": 2.57, + "learning_rate": 2.744857741101121e-06, + "loss": 0.5661, + "step": 6955 + }, + { + "epoch": 2.57, + "learning_rate": 2.7448207907377756e-06, + "loss": 0.5181, + "step": 6956 + }, + { + "epoch": 2.57, + "learning_rate": 2.7447838403744307e-06, + "loss": 0.5041, + "step": 6957 + }, + { + "epoch": 2.57, + "learning_rate": 2.7447468900110853e-06, + "loss": 0.5767, + "step": 6958 + }, + { + "epoch": 2.57, + "learning_rate": 2.74470993964774e-06, + "loss": 0.6344, + "step": 6959 + }, + { + "epoch": 2.57, + "learning_rate": 2.7446729892843946e-06, + "loss": 0.5688, + "step": 6960 + }, + { + "epoch": 2.57, + "learning_rate": 2.7446360389210492e-06, + "loss": 0.555, + "step": 6961 + }, + { + "epoch": 2.57, + "learning_rate": 2.7445990885577043e-06, + "loss": 0.5587, + "step": 6962 + }, + { + "epoch": 2.57, + "learning_rate": 2.744562138194359e-06, + "loss": 0.6248, + "step": 6963 + }, + { + "epoch": 2.57, + "learning_rate": 2.744525187831014e-06, + "loss": 0.6271, + "step": 6964 + }, + { + "epoch": 2.57, + "learning_rate": 2.7444882374676682e-06, + "loss": 0.5739, + "step": 6965 + }, + { + "epoch": 2.57, + "learning_rate": 2.7444512871043233e-06, + "loss": 0.5886, + "step": 6966 + }, + { + "epoch": 2.57, + "learning_rate": 2.744414336740978e-06, + "loss": 0.5598, + "step": 6967 + }, + { + "epoch": 2.57, + "learning_rate": 2.744377386377633e-06, + "loss": 0.5917, + "step": 6968 + }, + { + "epoch": 2.57, + "learning_rate": 2.7443404360142876e-06, + "loss": 0.4972, + "step": 6969 + }, + { + "epoch": 2.57, + "learning_rate": 2.7443034856509427e-06, + "loss": 0.5309, + "step": 6970 + }, + { + "epoch": 2.57, + "learning_rate": 2.744266535287597e-06, + "loss": 0.5594, + "step": 6971 + }, + { + "epoch": 2.57, + "learning_rate": 2.7442295849242515e-06, + "loss": 0.6554, + "step": 6972 + }, + { + "epoch": 2.57, + "learning_rate": 2.7441926345609066e-06, + "loss": 0.5747, + "step": 6973 + }, + { + "epoch": 2.58, + "learning_rate": 2.7441556841975612e-06, + "loss": 0.5703, + "step": 6974 + }, + { + "epoch": 2.58, + "learning_rate": 2.7441187338342163e-06, + "loss": 0.6276, + "step": 6975 + }, + { + "epoch": 2.58, + "learning_rate": 2.744081783470871e-06, + "loss": 0.5785, + "step": 6976 + }, + { + "epoch": 2.58, + "learning_rate": 2.7440448331075255e-06, + "loss": 0.6577, + "step": 6977 + }, + { + "epoch": 2.58, + "learning_rate": 2.74400788274418e-06, + "loss": 0.5308, + "step": 6978 + }, + { + "epoch": 2.58, + "learning_rate": 2.7439709323808352e-06, + "loss": 0.6033, + "step": 6979 + }, + { + "epoch": 2.58, + "learning_rate": 2.74393398201749e-06, + "loss": 0.5408, + "step": 6980 + }, + { + "epoch": 2.58, + "learning_rate": 2.7438970316541445e-06, + "loss": 0.5165, + "step": 6981 + }, + { + "epoch": 2.58, + "learning_rate": 2.7438600812907996e-06, + "loss": 0.5709, + "step": 6982 + }, + { + "epoch": 2.58, + "learning_rate": 2.7438231309274542e-06, + "loss": 0.6528, + "step": 6983 + }, + { + "epoch": 2.58, + "learning_rate": 2.743786180564109e-06, + "loss": 0.5263, + "step": 6984 + }, + { + "epoch": 2.58, + "learning_rate": 2.7437492302007635e-06, + "loss": 0.5701, + "step": 6985 + }, + { + "epoch": 2.58, + "learning_rate": 2.7437122798374185e-06, + "loss": 0.5995, + "step": 6986 + }, + { + "epoch": 2.58, + "learning_rate": 2.743675329474073e-06, + "loss": 0.6841, + "step": 6987 + }, + { + "epoch": 2.58, + "learning_rate": 2.7436383791107282e-06, + "loss": 0.6046, + "step": 6988 + }, + { + "epoch": 2.58, + "learning_rate": 2.743601428747383e-06, + "loss": 0.5711, + "step": 6989 + }, + { + "epoch": 2.58, + "learning_rate": 2.7435644783840375e-06, + "loss": 0.5281, + "step": 6990 + }, + { + "epoch": 2.58, + "learning_rate": 2.743527528020692e-06, + "loss": 0.6044, + "step": 6991 + }, + { + "epoch": 2.58, + "learning_rate": 2.7434905776573468e-06, + "loss": 0.6045, + "step": 6992 + }, + { + "epoch": 2.58, + "learning_rate": 2.743453627294002e-06, + "loss": 0.7619, + "step": 6993 + }, + { + "epoch": 2.58, + "learning_rate": 2.7434166769306565e-06, + "loss": 0.6868, + "step": 6994 + }, + { + "epoch": 2.58, + "learning_rate": 2.7433797265673115e-06, + "loss": 0.617, + "step": 6995 + }, + { + "epoch": 2.58, + "learning_rate": 2.743342776203966e-06, + "loss": 0.5116, + "step": 6996 + }, + { + "epoch": 2.58, + "learning_rate": 2.743305825840621e-06, + "loss": 0.5434, + "step": 6997 + }, + { + "epoch": 2.58, + "learning_rate": 2.7432688754772755e-06, + "loss": 0.5975, + "step": 6998 + }, + { + "epoch": 2.58, + "learning_rate": 2.7432319251139305e-06, + "loss": 0.5549, + "step": 6999 + }, + { + "epoch": 2.58, + "learning_rate": 2.743194974750585e-06, + "loss": 0.5276, + "step": 7000 + }, + { + "epoch": 2.59, + "learning_rate": 2.7431580243872398e-06, + "loss": 0.5946, + "step": 7001 + }, + { + "epoch": 2.59, + "learning_rate": 2.743121074023895e-06, + "loss": 0.523, + "step": 7002 + }, + { + "epoch": 2.59, + "learning_rate": 2.7430841236605495e-06, + "loss": 0.5863, + "step": 7003 + }, + { + "epoch": 2.59, + "learning_rate": 2.743047173297204e-06, + "loss": 0.6578, + "step": 7004 + }, + { + "epoch": 2.59, + "learning_rate": 2.7430102229338588e-06, + "loss": 0.5232, + "step": 7005 + }, + { + "epoch": 2.59, + "learning_rate": 2.742973272570514e-06, + "loss": 0.5342, + "step": 7006 + }, + { + "epoch": 2.59, + "learning_rate": 2.7429363222071685e-06, + "loss": 0.5948, + "step": 7007 + }, + { + "epoch": 2.59, + "learning_rate": 2.7428993718438235e-06, + "loss": 0.602, + "step": 7008 + }, + { + "epoch": 2.59, + "learning_rate": 2.742862421480478e-06, + "loss": 0.5349, + "step": 7009 + }, + { + "epoch": 2.59, + "learning_rate": 2.7428254711171324e-06, + "loss": 0.6016, + "step": 7010 + }, + { + "epoch": 2.59, + "learning_rate": 2.7427885207537874e-06, + "loss": 0.5604, + "step": 7011 + }, + { + "epoch": 2.59, + "learning_rate": 2.742751570390442e-06, + "loss": 0.5375, + "step": 7012 + }, + { + "epoch": 2.59, + "learning_rate": 2.742714620027097e-06, + "loss": 0.5933, + "step": 7013 + }, + { + "epoch": 2.59, + "learning_rate": 2.7426776696637518e-06, + "loss": 0.5842, + "step": 7014 + }, + { + "epoch": 2.59, + "learning_rate": 2.742640719300407e-06, + "loss": 0.6531, + "step": 7015 + }, + { + "epoch": 2.59, + "learning_rate": 2.742603768937061e-06, + "loss": 0.5891, + "step": 7016 + }, + { + "epoch": 2.59, + "learning_rate": 2.742566818573716e-06, + "loss": 0.6619, + "step": 7017 + }, + { + "epoch": 2.59, + "learning_rate": 2.7425298682103707e-06, + "loss": 0.5659, + "step": 7018 + }, + { + "epoch": 2.59, + "learning_rate": 2.7424929178470258e-06, + "loss": 0.6503, + "step": 7019 + }, + { + "epoch": 2.59, + "learning_rate": 2.7424559674836804e-06, + "loss": 0.654, + "step": 7020 + }, + { + "epoch": 2.59, + "learning_rate": 2.742419017120335e-06, + "loss": 0.5337, + "step": 7021 + }, + { + "epoch": 2.59, + "learning_rate": 2.7423820667569897e-06, + "loss": 0.534, + "step": 7022 + }, + { + "epoch": 2.59, + "learning_rate": 2.7423451163936443e-06, + "loss": 0.5461, + "step": 7023 + }, + { + "epoch": 2.59, + "learning_rate": 2.7423081660302994e-06, + "loss": 0.5989, + "step": 7024 + }, + { + "epoch": 2.59, + "learning_rate": 2.742271215666954e-06, + "loss": 0.6097, + "step": 7025 + }, + { + "epoch": 2.59, + "learning_rate": 2.742234265303609e-06, + "loss": 0.626, + "step": 7026 + }, + { + "epoch": 2.59, + "learning_rate": 2.7421973149402637e-06, + "loss": 0.5294, + "step": 7027 + }, + { + "epoch": 2.6, + "learning_rate": 2.7421603645769184e-06, + "loss": 0.5384, + "step": 7028 + }, + { + "epoch": 2.6, + "learning_rate": 2.742123414213573e-06, + "loss": 0.5924, + "step": 7029 + }, + { + "epoch": 2.6, + "learning_rate": 2.7420864638502276e-06, + "loss": 0.5559, + "step": 7030 + }, + { + "epoch": 2.6, + "learning_rate": 2.7420495134868827e-06, + "loss": 0.5454, + "step": 7031 + }, + { + "epoch": 2.6, + "learning_rate": 2.7420125631235373e-06, + "loss": 0.6129, + "step": 7032 + }, + { + "epoch": 2.6, + "learning_rate": 2.7419756127601924e-06, + "loss": 0.5586, + "step": 7033 + }, + { + "epoch": 2.6, + "learning_rate": 2.741938662396847e-06, + "loss": 0.7082, + "step": 7034 + }, + { + "epoch": 2.6, + "learning_rate": 2.7419017120335017e-06, + "loss": 0.6518, + "step": 7035 + }, + { + "epoch": 2.6, + "learning_rate": 2.7418647616701563e-06, + "loss": 0.5667, + "step": 7036 + }, + { + "epoch": 2.6, + "learning_rate": 2.7418278113068114e-06, + "loss": 0.633, + "step": 7037 + }, + { + "epoch": 2.6, + "learning_rate": 2.741790860943466e-06, + "loss": 0.5724, + "step": 7038 + }, + { + "epoch": 2.6, + "learning_rate": 2.741753910580121e-06, + "loss": 0.6014, + "step": 7039 + }, + { + "epoch": 2.6, + "learning_rate": 2.7417169602167757e-06, + "loss": 0.6683, + "step": 7040 + }, + { + "epoch": 2.6, + "learning_rate": 2.7416800098534303e-06, + "loss": 0.5912, + "step": 7041 + }, + { + "epoch": 2.6, + "learning_rate": 2.741643059490085e-06, + "loss": 0.5045, + "step": 7042 + }, + { + "epoch": 2.6, + "learning_rate": 2.7416061091267396e-06, + "loss": 0.5188, + "step": 7043 + }, + { + "epoch": 2.6, + "learning_rate": 2.7415691587633947e-06, + "loss": 0.688, + "step": 7044 + }, + { + "epoch": 2.6, + "learning_rate": 2.7415322084000493e-06, + "loss": 0.5404, + "step": 7045 + }, + { + "epoch": 2.6, + "learning_rate": 2.7414952580367044e-06, + "loss": 0.5535, + "step": 7046 + }, + { + "epoch": 2.6, + "learning_rate": 2.741458307673359e-06, + "loss": 0.5713, + "step": 7047 + }, + { + "epoch": 2.6, + "learning_rate": 2.7414213573100136e-06, + "loss": 0.5923, + "step": 7048 + }, + { + "epoch": 2.6, + "learning_rate": 2.7413844069466683e-06, + "loss": 0.5972, + "step": 7049 + }, + { + "epoch": 2.6, + "learning_rate": 2.741347456583323e-06, + "loss": 0.6272, + "step": 7050 + }, + { + "epoch": 2.6, + "learning_rate": 2.741310506219978e-06, + "loss": 0.5402, + "step": 7051 + }, + { + "epoch": 2.6, + "learning_rate": 2.7412735558566326e-06, + "loss": 0.5037, + "step": 7052 + }, + { + "epoch": 2.6, + "learning_rate": 2.7412366054932877e-06, + "loss": 0.5509, + "step": 7053 + }, + { + "epoch": 2.6, + "learning_rate": 2.7411996551299423e-06, + "loss": 0.6314, + "step": 7054 + }, + { + "epoch": 2.61, + "learning_rate": 2.741162704766597e-06, + "loss": 0.5851, + "step": 7055 + }, + { + "epoch": 2.61, + "learning_rate": 2.7411257544032516e-06, + "loss": 0.597, + "step": 7056 + }, + { + "epoch": 2.61, + "learning_rate": 2.7410888040399066e-06, + "loss": 0.5556, + "step": 7057 + }, + { + "epoch": 2.61, + "learning_rate": 2.7410518536765613e-06, + "loss": 0.5179, + "step": 7058 + }, + { + "epoch": 2.61, + "learning_rate": 2.7410149033132163e-06, + "loss": 0.5547, + "step": 7059 + }, + { + "epoch": 2.61, + "learning_rate": 2.7409779529498705e-06, + "loss": 0.5522, + "step": 7060 + }, + { + "epoch": 2.61, + "learning_rate": 2.740941002586525e-06, + "loss": 0.5524, + "step": 7061 + }, + { + "epoch": 2.61, + "learning_rate": 2.7409040522231802e-06, + "loss": 0.5429, + "step": 7062 + }, + { + "epoch": 2.61, + "learning_rate": 2.740867101859835e-06, + "loss": 0.7112, + "step": 7063 + }, + { + "epoch": 2.61, + "learning_rate": 2.74083015149649e-06, + "loss": 0.5635, + "step": 7064 + }, + { + "epoch": 2.61, + "learning_rate": 2.7407932011331446e-06, + "loss": 0.6246, + "step": 7065 + }, + { + "epoch": 2.61, + "learning_rate": 2.740756250769799e-06, + "loss": 0.5457, + "step": 7066 + }, + { + "epoch": 2.61, + "learning_rate": 2.740719300406454e-06, + "loss": 0.6171, + "step": 7067 + }, + { + "epoch": 2.61, + "learning_rate": 2.740682350043109e-06, + "loss": 0.6451, + "step": 7068 + }, + { + "epoch": 2.61, + "learning_rate": 2.7406453996797635e-06, + "loss": 0.5855, + "step": 7069 + }, + { + "epoch": 2.61, + "learning_rate": 2.740608449316418e-06, + "loss": 0.5735, + "step": 7070 + }, + { + "epoch": 2.61, + "learning_rate": 2.7405714989530732e-06, + "loss": 0.4875, + "step": 7071 + }, + { + "epoch": 2.61, + "learning_rate": 2.740534548589728e-06, + "loss": 0.582, + "step": 7072 + }, + { + "epoch": 2.61, + "learning_rate": 2.7404975982263825e-06, + "loss": 0.4943, + "step": 7073 + }, + { + "epoch": 2.61, + "learning_rate": 2.740460647863037e-06, + "loss": 0.7199, + "step": 7074 + }, + { + "epoch": 2.61, + "learning_rate": 2.740423697499692e-06, + "loss": 0.5164, + "step": 7075 + }, + { + "epoch": 2.61, + "learning_rate": 2.740386747136347e-06, + "loss": 0.4991, + "step": 7076 + }, + { + "epoch": 2.61, + "learning_rate": 2.740349796773002e-06, + "loss": 0.5274, + "step": 7077 + }, + { + "epoch": 2.61, + "learning_rate": 2.7403128464096565e-06, + "loss": 0.4618, + "step": 7078 + }, + { + "epoch": 2.61, + "learning_rate": 2.740275896046311e-06, + "loss": 0.5586, + "step": 7079 + }, + { + "epoch": 2.61, + "learning_rate": 2.740238945682966e-06, + "loss": 0.5726, + "step": 7080 + }, + { + "epoch": 2.61, + "learning_rate": 2.7402019953196204e-06, + "loss": 0.6019, + "step": 7081 + }, + { + "epoch": 2.62, + "learning_rate": 2.7401650449562755e-06, + "loss": 0.5854, + "step": 7082 + }, + { + "epoch": 2.62, + "learning_rate": 2.74012809459293e-06, + "loss": 0.7006, + "step": 7083 + }, + { + "epoch": 2.62, + "learning_rate": 2.740091144229585e-06, + "loss": 0.6072, + "step": 7084 + }, + { + "epoch": 2.62, + "learning_rate": 2.74005419386624e-06, + "loss": 0.5639, + "step": 7085 + }, + { + "epoch": 2.62, + "learning_rate": 2.7400172435028945e-06, + "loss": 0.596, + "step": 7086 + }, + { + "epoch": 2.62, + "learning_rate": 2.739980293139549e-06, + "loss": 0.5805, + "step": 7087 + }, + { + "epoch": 2.62, + "learning_rate": 2.739943342776204e-06, + "loss": 0.5832, + "step": 7088 + }, + { + "epoch": 2.62, + "learning_rate": 2.739906392412859e-06, + "loss": 0.6013, + "step": 7089 + }, + { + "epoch": 2.62, + "learning_rate": 2.7398694420495134e-06, + "loss": 0.5453, + "step": 7090 + }, + { + "epoch": 2.62, + "learning_rate": 2.7398324916861685e-06, + "loss": 0.5769, + "step": 7091 + }, + { + "epoch": 2.62, + "learning_rate": 2.739795541322823e-06, + "loss": 0.5642, + "step": 7092 + }, + { + "epoch": 2.62, + "learning_rate": 2.7397585909594778e-06, + "loss": 0.6786, + "step": 7093 + }, + { + "epoch": 2.62, + "learning_rate": 2.7397216405961324e-06, + "loss": 0.5499, + "step": 7094 + }, + { + "epoch": 2.62, + "learning_rate": 2.7396846902327875e-06, + "loss": 0.6296, + "step": 7095 + }, + { + "epoch": 2.62, + "learning_rate": 2.739647739869442e-06, + "loss": 0.588, + "step": 7096 + }, + { + "epoch": 2.62, + "learning_rate": 2.739610789506097e-06, + "loss": 0.6601, + "step": 7097 + }, + { + "epoch": 2.62, + "learning_rate": 2.739573839142752e-06, + "loss": 0.6682, + "step": 7098 + }, + { + "epoch": 2.62, + "learning_rate": 2.7395368887794064e-06, + "loss": 0.5962, + "step": 7099 + }, + { + "epoch": 2.62, + "learning_rate": 2.739499938416061e-06, + "loss": 0.6303, + "step": 7100 + }, + { + "epoch": 2.62, + "learning_rate": 2.7394629880527157e-06, + "loss": 0.5416, + "step": 7101 + }, + { + "epoch": 2.62, + "learning_rate": 2.7394260376893708e-06, + "loss": 0.5185, + "step": 7102 + }, + { + "epoch": 2.62, + "learning_rate": 2.7393890873260254e-06, + "loss": 0.6349, + "step": 7103 + }, + { + "epoch": 2.62, + "learning_rate": 2.7393521369626805e-06, + "loss": 0.6269, + "step": 7104 + }, + { + "epoch": 2.62, + "learning_rate": 2.7393151865993347e-06, + "loss": 0.5343, + "step": 7105 + }, + { + "epoch": 2.62, + "learning_rate": 2.7392782362359897e-06, + "loss": 0.5511, + "step": 7106 + }, + { + "epoch": 2.62, + "learning_rate": 2.7392412858726444e-06, + "loss": 0.6322, + "step": 7107 + }, + { + "epoch": 2.62, + "learning_rate": 2.7392043355092994e-06, + "loss": 0.4825, + "step": 7108 + }, + { + "epoch": 2.63, + "learning_rate": 2.739167385145954e-06, + "loss": 0.5177, + "step": 7109 + }, + { + "epoch": 2.63, + "learning_rate": 2.7391304347826087e-06, + "loss": 0.5846, + "step": 7110 + }, + { + "epoch": 2.63, + "learning_rate": 2.7390934844192633e-06, + "loss": 0.4686, + "step": 7111 + }, + { + "epoch": 2.63, + "learning_rate": 2.739056534055918e-06, + "loss": 0.5194, + "step": 7112 + }, + { + "epoch": 2.63, + "learning_rate": 2.739019583692573e-06, + "loss": 0.6272, + "step": 7113 + }, + { + "epoch": 2.63, + "learning_rate": 2.7389826333292277e-06, + "loss": 0.654, + "step": 7114 + }, + { + "epoch": 2.63, + "learning_rate": 2.7389456829658827e-06, + "loss": 0.6191, + "step": 7115 + }, + { + "epoch": 2.63, + "learning_rate": 2.7389087326025374e-06, + "loss": 0.639, + "step": 7116 + }, + { + "epoch": 2.63, + "learning_rate": 2.738871782239192e-06, + "loss": 0.5182, + "step": 7117 + }, + { + "epoch": 2.63, + "learning_rate": 2.7388348318758466e-06, + "loss": 0.4999, + "step": 7118 + }, + { + "epoch": 2.63, + "learning_rate": 2.7387978815125017e-06, + "loss": 0.5543, + "step": 7119 + }, + { + "epoch": 2.63, + "learning_rate": 2.7387609311491563e-06, + "loss": 0.584, + "step": 7120 + }, + { + "epoch": 2.63, + "learning_rate": 2.738723980785811e-06, + "loss": 0.6178, + "step": 7121 + }, + { + "epoch": 2.63, + "learning_rate": 2.738687030422466e-06, + "loss": 0.5848, + "step": 7122 + }, + { + "epoch": 2.63, + "learning_rate": 2.7386500800591207e-06, + "loss": 0.539, + "step": 7123 + }, + { + "epoch": 2.63, + "learning_rate": 2.7386131296957753e-06, + "loss": 0.5522, + "step": 7124 + }, + { + "epoch": 2.63, + "learning_rate": 2.73857617933243e-06, + "loss": 0.5172, + "step": 7125 + }, + { + "epoch": 2.63, + "learning_rate": 2.738539228969085e-06, + "loss": 0.5379, + "step": 7126 + }, + { + "epoch": 2.63, + "learning_rate": 2.7385022786057396e-06, + "loss": 0.5907, + "step": 7127 + }, + { + "epoch": 2.63, + "learning_rate": 2.7384653282423947e-06, + "loss": 0.5995, + "step": 7128 + }, + { + "epoch": 2.63, + "learning_rate": 2.7384283778790493e-06, + "loss": 0.5174, + "step": 7129 + }, + { + "epoch": 2.63, + "learning_rate": 2.738391427515704e-06, + "loss": 0.752, + "step": 7130 + }, + { + "epoch": 2.63, + "learning_rate": 2.7383544771523586e-06, + "loss": 0.6606, + "step": 7131 + }, + { + "epoch": 2.63, + "learning_rate": 2.7383175267890133e-06, + "loss": 0.5171, + "step": 7132 + }, + { + "epoch": 2.63, + "learning_rate": 2.7382805764256683e-06, + "loss": 0.6093, + "step": 7133 + }, + { + "epoch": 2.63, + "learning_rate": 2.738243626062323e-06, + "loss": 0.5149, + "step": 7134 + }, + { + "epoch": 2.63, + "learning_rate": 2.738206675698978e-06, + "loss": 0.6454, + "step": 7135 + }, + { + "epoch": 2.64, + "learning_rate": 2.7381697253356326e-06, + "loss": 0.6162, + "step": 7136 + }, + { + "epoch": 2.64, + "learning_rate": 2.7381327749722873e-06, + "loss": 0.5386, + "step": 7137 + }, + { + "epoch": 2.64, + "learning_rate": 2.738095824608942e-06, + "loss": 0.5747, + "step": 7138 + }, + { + "epoch": 2.64, + "learning_rate": 2.738058874245597e-06, + "loss": 0.5529, + "step": 7139 + }, + { + "epoch": 2.64, + "learning_rate": 2.7380219238822516e-06, + "loss": 0.6119, + "step": 7140 + }, + { + "epoch": 2.64, + "learning_rate": 2.7379849735189063e-06, + "loss": 0.5541, + "step": 7141 + }, + { + "epoch": 2.64, + "learning_rate": 2.7379480231555613e-06, + "loss": 0.6215, + "step": 7142 + }, + { + "epoch": 2.64, + "learning_rate": 2.737911072792216e-06, + "loss": 0.5712, + "step": 7143 + }, + { + "epoch": 2.64, + "learning_rate": 2.7378741224288706e-06, + "loss": 0.6385, + "step": 7144 + }, + { + "epoch": 2.64, + "learning_rate": 2.7378371720655252e-06, + "loss": 0.5665, + "step": 7145 + }, + { + "epoch": 2.64, + "learning_rate": 2.7378002217021803e-06, + "loss": 0.5954, + "step": 7146 + }, + { + "epoch": 2.64, + "learning_rate": 2.737763271338835e-06, + "loss": 0.5794, + "step": 7147 + }, + { + "epoch": 2.64, + "learning_rate": 2.73772632097549e-06, + "loss": 0.6066, + "step": 7148 + }, + { + "epoch": 2.64, + "learning_rate": 2.7376893706121446e-06, + "loss": 0.6379, + "step": 7149 + }, + { + "epoch": 2.64, + "learning_rate": 2.737652420248799e-06, + "loss": 0.6134, + "step": 7150 + }, + { + "epoch": 2.64, + "learning_rate": 2.737615469885454e-06, + "loss": 0.4969, + "step": 7151 + }, + { + "epoch": 2.64, + "learning_rate": 2.7375785195221085e-06, + "loss": 0.5263, + "step": 7152 + }, + { + "epoch": 2.64, + "learning_rate": 2.7375415691587636e-06, + "loss": 0.4863, + "step": 7153 + }, + { + "epoch": 2.64, + "learning_rate": 2.7375046187954182e-06, + "loss": 0.5578, + "step": 7154 + }, + { + "epoch": 2.64, + "learning_rate": 2.7374676684320733e-06, + "loss": 0.5617, + "step": 7155 + }, + { + "epoch": 2.64, + "learning_rate": 2.7374307180687275e-06, + "loss": 0.6141, + "step": 7156 + }, + { + "epoch": 2.64, + "learning_rate": 2.7373937677053826e-06, + "loss": 0.4765, + "step": 7157 + }, + { + "epoch": 2.64, + "learning_rate": 2.737356817342037e-06, + "loss": 0.6306, + "step": 7158 + }, + { + "epoch": 2.64, + "learning_rate": 2.7373198669786922e-06, + "loss": 0.6031, + "step": 7159 + }, + { + "epoch": 2.64, + "learning_rate": 2.737282916615347e-06, + "loss": 0.5853, + "step": 7160 + }, + { + "epoch": 2.64, + "learning_rate": 2.7372459662520015e-06, + "loss": 0.4721, + "step": 7161 + }, + { + "epoch": 2.64, + "learning_rate": 2.737209015888656e-06, + "loss": 0.5593, + "step": 7162 + }, + { + "epoch": 2.65, + "learning_rate": 2.737172065525311e-06, + "loss": 0.6536, + "step": 7163 + }, + { + "epoch": 2.65, + "learning_rate": 2.737135115161966e-06, + "loss": 0.5898, + "step": 7164 + }, + { + "epoch": 2.65, + "learning_rate": 2.7370981647986205e-06, + "loss": 0.4609, + "step": 7165 + }, + { + "epoch": 2.65, + "learning_rate": 2.7370612144352755e-06, + "loss": 0.6327, + "step": 7166 + }, + { + "epoch": 2.65, + "learning_rate": 2.73702426407193e-06, + "loss": 0.5772, + "step": 7167 + }, + { + "epoch": 2.65, + "learning_rate": 2.736987313708585e-06, + "loss": 0.6487, + "step": 7168 + }, + { + "epoch": 2.65, + "learning_rate": 2.7369503633452395e-06, + "loss": 0.5731, + "step": 7169 + }, + { + "epoch": 2.65, + "learning_rate": 2.736913412981894e-06, + "loss": 0.6145, + "step": 7170 + }, + { + "epoch": 2.65, + "learning_rate": 2.736876462618549e-06, + "loss": 0.5707, + "step": 7171 + }, + { + "epoch": 2.65, + "learning_rate": 2.736839512255204e-06, + "loss": 0.5307, + "step": 7172 + }, + { + "epoch": 2.65, + "learning_rate": 2.736802561891859e-06, + "loss": 0.6378, + "step": 7173 + }, + { + "epoch": 2.65, + "learning_rate": 2.7367656115285135e-06, + "loss": 0.5329, + "step": 7174 + }, + { + "epoch": 2.65, + "learning_rate": 2.736728661165168e-06, + "loss": 0.5962, + "step": 7175 + }, + { + "epoch": 2.65, + "learning_rate": 2.7366917108018228e-06, + "loss": 0.6072, + "step": 7176 + }, + { + "epoch": 2.65, + "learning_rate": 2.736654760438478e-06, + "loss": 0.5354, + "step": 7177 + }, + { + "epoch": 2.65, + "learning_rate": 2.7366178100751325e-06, + "loss": 0.587, + "step": 7178 + }, + { + "epoch": 2.65, + "learning_rate": 2.7365808597117875e-06, + "loss": 0.6044, + "step": 7179 + }, + { + "epoch": 2.65, + "learning_rate": 2.736543909348442e-06, + "loss": 0.5775, + "step": 7180 + }, + { + "epoch": 2.65, + "learning_rate": 2.736506958985097e-06, + "loss": 0.5469, + "step": 7181 + }, + { + "epoch": 2.65, + "learning_rate": 2.7364700086217514e-06, + "loss": 0.5746, + "step": 7182 + }, + { + "epoch": 2.65, + "learning_rate": 2.736433058258406e-06, + "loss": 0.652, + "step": 7183 + }, + { + "epoch": 2.65, + "learning_rate": 2.736396107895061e-06, + "loss": 0.643, + "step": 7184 + }, + { + "epoch": 2.65, + "learning_rate": 2.7363591575317158e-06, + "loss": 0.6908, + "step": 7185 + }, + { + "epoch": 2.65, + "learning_rate": 2.736322207168371e-06, + "loss": 0.6707, + "step": 7186 + }, + { + "epoch": 2.65, + "learning_rate": 2.7362852568050255e-06, + "loss": 0.6547, + "step": 7187 + }, + { + "epoch": 2.65, + "learning_rate": 2.73624830644168e-06, + "loss": 0.6024, + "step": 7188 + }, + { + "epoch": 2.65, + "learning_rate": 2.7362113560783347e-06, + "loss": 0.6632, + "step": 7189 + }, + { + "epoch": 2.66, + "learning_rate": 2.7361744057149894e-06, + "loss": 0.6172, + "step": 7190 + }, + { + "epoch": 2.66, + "learning_rate": 2.7361374553516444e-06, + "loss": 0.5101, + "step": 7191 + }, + { + "epoch": 2.66, + "learning_rate": 2.736100504988299e-06, + "loss": 0.7212, + "step": 7192 + }, + { + "epoch": 2.66, + "learning_rate": 2.736063554624954e-06, + "loss": 0.5879, + "step": 7193 + }, + { + "epoch": 2.66, + "learning_rate": 2.7360266042616088e-06, + "loss": 0.5483, + "step": 7194 + }, + { + "epoch": 2.66, + "learning_rate": 2.7359896538982634e-06, + "loss": 0.5366, + "step": 7195 + }, + { + "epoch": 2.66, + "learning_rate": 2.735952703534918e-06, + "loss": 0.5771, + "step": 7196 + }, + { + "epoch": 2.66, + "learning_rate": 2.735915753171573e-06, + "loss": 0.6889, + "step": 7197 + }, + { + "epoch": 2.66, + "learning_rate": 2.7358788028082277e-06, + "loss": 0.5634, + "step": 7198 + }, + { + "epoch": 2.66, + "learning_rate": 2.7358418524448828e-06, + "loss": 0.5241, + "step": 7199 + }, + { + "epoch": 2.66, + "learning_rate": 2.7358049020815374e-06, + "loss": 0.6417, + "step": 7200 + }, + { + "epoch": 2.66, + "learning_rate": 2.7357679517181916e-06, + "loss": 0.5737, + "step": 7201 + }, + { + "epoch": 2.66, + "learning_rate": 2.7357310013548467e-06, + "loss": 0.6365, + "step": 7202 + }, + { + "epoch": 2.66, + "learning_rate": 2.7356940509915013e-06, + "loss": 0.5948, + "step": 7203 + }, + { + "epoch": 2.66, + "learning_rate": 2.7356571006281564e-06, + "loss": 0.5587, + "step": 7204 + }, + { + "epoch": 2.66, + "learning_rate": 2.735620150264811e-06, + "loss": 0.544, + "step": 7205 + }, + { + "epoch": 2.66, + "learning_rate": 2.735583199901466e-06, + "loss": 0.5479, + "step": 7206 + }, + { + "epoch": 2.66, + "learning_rate": 2.7355462495381203e-06, + "loss": 0.6376, + "step": 7207 + }, + { + "epoch": 2.66, + "learning_rate": 2.7355092991747754e-06, + "loss": 0.6549, + "step": 7208 + }, + { + "epoch": 2.66, + "learning_rate": 2.73547234881143e-06, + "loss": 0.5609, + "step": 7209 + }, + { + "epoch": 2.66, + "learning_rate": 2.7354353984480846e-06, + "loss": 0.565, + "step": 7210 + }, + { + "epoch": 2.66, + "learning_rate": 2.7353984480847397e-06, + "loss": 0.6103, + "step": 7211 + }, + { + "epoch": 2.66, + "learning_rate": 2.7353614977213943e-06, + "loss": 0.5266, + "step": 7212 + }, + { + "epoch": 2.66, + "learning_rate": 2.735324547358049e-06, + "loss": 0.5774, + "step": 7213 + }, + { + "epoch": 2.66, + "learning_rate": 2.7352875969947036e-06, + "loss": 0.6296, + "step": 7214 + }, + { + "epoch": 2.66, + "learning_rate": 2.7352506466313587e-06, + "loss": 0.5538, + "step": 7215 + }, + { + "epoch": 2.66, + "learning_rate": 2.7352136962680133e-06, + "loss": 0.6071, + "step": 7216 + }, + { + "epoch": 2.67, + "learning_rate": 2.7351767459046684e-06, + "loss": 0.6316, + "step": 7217 + }, + { + "epoch": 2.67, + "learning_rate": 2.735139795541323e-06, + "loss": 0.5777, + "step": 7218 + }, + { + "epoch": 2.67, + "learning_rate": 2.7351028451779776e-06, + "loss": 0.6541, + "step": 7219 + }, + { + "epoch": 2.67, + "learning_rate": 2.7350658948146323e-06, + "loss": 0.6098, + "step": 7220 + }, + { + "epoch": 2.67, + "learning_rate": 2.735028944451287e-06, + "loss": 0.5988, + "step": 7221 + }, + { + "epoch": 2.67, + "learning_rate": 2.734991994087942e-06, + "loss": 0.565, + "step": 7222 + }, + { + "epoch": 2.67, + "learning_rate": 2.7349550437245966e-06, + "loss": 0.4779, + "step": 7223 + }, + { + "epoch": 2.67, + "learning_rate": 2.7349180933612517e-06, + "loss": 0.5822, + "step": 7224 + }, + { + "epoch": 2.67, + "learning_rate": 2.7348811429979063e-06, + "loss": 0.5386, + "step": 7225 + }, + { + "epoch": 2.67, + "learning_rate": 2.734844192634561e-06, + "loss": 0.6422, + "step": 7226 + }, + { + "epoch": 2.67, + "learning_rate": 2.7348072422712156e-06, + "loss": 0.5487, + "step": 7227 + }, + { + "epoch": 2.67, + "learning_rate": 2.7347702919078706e-06, + "loss": 0.5912, + "step": 7228 + }, + { + "epoch": 2.67, + "learning_rate": 2.7347333415445253e-06, + "loss": 0.5524, + "step": 7229 + }, + { + "epoch": 2.67, + "learning_rate": 2.73469639118118e-06, + "loss": 0.5953, + "step": 7230 + }, + { + "epoch": 2.67, + "learning_rate": 2.734659440817835e-06, + "loss": 0.5899, + "step": 7231 + }, + { + "epoch": 2.67, + "learning_rate": 2.7346224904544896e-06, + "loss": 0.5356, + "step": 7232 + }, + { + "epoch": 2.67, + "learning_rate": 2.7345855400911442e-06, + "loss": 0.5103, + "step": 7233 + }, + { + "epoch": 2.67, + "learning_rate": 2.734548589727799e-06, + "loss": 0.5492, + "step": 7234 + }, + { + "epoch": 2.67, + "learning_rate": 2.734511639364454e-06, + "loss": 0.5933, + "step": 7235 + }, + { + "epoch": 2.67, + "learning_rate": 2.7344746890011086e-06, + "loss": 0.528, + "step": 7236 + }, + { + "epoch": 2.67, + "learning_rate": 2.7344377386377636e-06, + "loss": 0.569, + "step": 7237 + }, + { + "epoch": 2.67, + "learning_rate": 2.7344007882744183e-06, + "loss": 0.524, + "step": 7238 + }, + { + "epoch": 2.67, + "learning_rate": 2.734363837911073e-06, + "loss": 0.5195, + "step": 7239 + }, + { + "epoch": 2.67, + "learning_rate": 2.7343268875477275e-06, + "loss": 0.46, + "step": 7240 + }, + { + "epoch": 2.67, + "learning_rate": 2.734289937184382e-06, + "loss": 0.6224, + "step": 7241 + }, + { + "epoch": 2.67, + "learning_rate": 2.7342529868210372e-06, + "loss": 0.5498, + "step": 7242 + }, + { + "epoch": 2.67, + "learning_rate": 2.734216036457692e-06, + "loss": 0.661, + "step": 7243 + }, + { + "epoch": 2.68, + "learning_rate": 2.734179086094347e-06, + "loss": 0.5905, + "step": 7244 + }, + { + "epoch": 2.68, + "learning_rate": 2.7341421357310016e-06, + "loss": 0.6658, + "step": 7245 + }, + { + "epoch": 2.68, + "learning_rate": 2.734105185367656e-06, + "loss": 0.6617, + "step": 7246 + }, + { + "epoch": 2.68, + "learning_rate": 2.734068235004311e-06, + "loss": 0.5101, + "step": 7247 + }, + { + "epoch": 2.68, + "learning_rate": 2.734031284640966e-06, + "loss": 0.6474, + "step": 7248 + }, + { + "epoch": 2.68, + "learning_rate": 2.7339943342776205e-06, + "loss": 0.6208, + "step": 7249 + }, + { + "epoch": 2.68, + "learning_rate": 2.733957383914275e-06, + "loss": 0.4752, + "step": 7250 + }, + { + "epoch": 2.68, + "learning_rate": 2.73392043355093e-06, + "loss": 0.651, + "step": 7251 + }, + { + "epoch": 2.68, + "learning_rate": 2.7338834831875844e-06, + "loss": 0.6331, + "step": 7252 + }, + { + "epoch": 2.68, + "learning_rate": 2.7338465328242395e-06, + "loss": 0.5655, + "step": 7253 + }, + { + "epoch": 2.68, + "learning_rate": 2.733809582460894e-06, + "loss": 0.7154, + "step": 7254 + }, + { + "epoch": 2.68, + "learning_rate": 2.733772632097549e-06, + "loss": 0.7233, + "step": 7255 + }, + { + "epoch": 2.68, + "learning_rate": 2.733735681734204e-06, + "loss": 0.6514, + "step": 7256 + }, + { + "epoch": 2.68, + "learning_rate": 2.7336987313708585e-06, + "loss": 0.6742, + "step": 7257 + }, + { + "epoch": 2.68, + "learning_rate": 2.733661781007513e-06, + "loss": 0.5984, + "step": 7258 + }, + { + "epoch": 2.68, + "learning_rate": 2.7336248306441677e-06, + "loss": 0.6545, + "step": 7259 + }, + { + "epoch": 2.68, + "learning_rate": 2.733587880280823e-06, + "loss": 0.6044, + "step": 7260 + }, + { + "epoch": 2.68, + "learning_rate": 2.7335509299174774e-06, + "loss": 0.5353, + "step": 7261 + }, + { + "epoch": 2.68, + "learning_rate": 2.7335139795541325e-06, + "loss": 0.5768, + "step": 7262 + }, + { + "epoch": 2.68, + "learning_rate": 2.733477029190787e-06, + "loss": 0.5814, + "step": 7263 + }, + { + "epoch": 2.68, + "learning_rate": 2.7334400788274418e-06, + "loss": 0.5316, + "step": 7264 + }, + { + "epoch": 2.68, + "learning_rate": 2.7334031284640964e-06, + "loss": 0.515, + "step": 7265 + }, + { + "epoch": 2.68, + "learning_rate": 2.7333661781007515e-06, + "loss": 0.6368, + "step": 7266 + }, + { + "epoch": 2.68, + "learning_rate": 2.733329227737406e-06, + "loss": 0.5345, + "step": 7267 + }, + { + "epoch": 2.68, + "learning_rate": 2.733292277374061e-06, + "loss": 0.6193, + "step": 7268 + }, + { + "epoch": 2.68, + "learning_rate": 2.733255327010716e-06, + "loss": 0.5297, + "step": 7269 + }, + { + "epoch": 2.68, + "learning_rate": 2.7332183766473704e-06, + "loss": 0.5931, + "step": 7270 + }, + { + "epoch": 2.69, + "learning_rate": 2.733181426284025e-06, + "loss": 0.5301, + "step": 7271 + }, + { + "epoch": 2.69, + "learning_rate": 2.7331444759206797e-06, + "loss": 0.5871, + "step": 7272 + }, + { + "epoch": 2.69, + "learning_rate": 2.7331075255573348e-06, + "loss": 0.5574, + "step": 7273 + }, + { + "epoch": 2.69, + "learning_rate": 2.7330705751939894e-06, + "loss": 0.5543, + "step": 7274 + }, + { + "epoch": 2.69, + "learning_rate": 2.7330336248306445e-06, + "loss": 0.5767, + "step": 7275 + }, + { + "epoch": 2.69, + "learning_rate": 2.732996674467299e-06, + "loss": 0.5927, + "step": 7276 + }, + { + "epoch": 2.69, + "learning_rate": 2.7329597241039537e-06, + "loss": 0.6337, + "step": 7277 + }, + { + "epoch": 2.69, + "learning_rate": 2.7329227737406084e-06, + "loss": 0.5724, + "step": 7278 + }, + { + "epoch": 2.69, + "learning_rate": 2.732885823377263e-06, + "loss": 0.5559, + "step": 7279 + }, + { + "epoch": 2.69, + "learning_rate": 2.732848873013918e-06, + "loss": 0.6162, + "step": 7280 + }, + { + "epoch": 2.69, + "learning_rate": 2.7328119226505727e-06, + "loss": 0.562, + "step": 7281 + }, + { + "epoch": 2.69, + "learning_rate": 2.7327749722872278e-06, + "loss": 0.6096, + "step": 7282 + }, + { + "epoch": 2.69, + "learning_rate": 2.7327380219238824e-06, + "loss": 0.5773, + "step": 7283 + }, + { + "epoch": 2.69, + "learning_rate": 2.732701071560537e-06, + "loss": 0.5619, + "step": 7284 + }, + { + "epoch": 2.69, + "learning_rate": 2.7326641211971917e-06, + "loss": 0.6098, + "step": 7285 + }, + { + "epoch": 2.69, + "learning_rate": 2.7326271708338467e-06, + "loss": 0.6102, + "step": 7286 + }, + { + "epoch": 2.69, + "learning_rate": 2.7325902204705014e-06, + "loss": 0.5264, + "step": 7287 + }, + { + "epoch": 2.69, + "learning_rate": 2.7325532701071564e-06, + "loss": 0.5195, + "step": 7288 + }, + { + "epoch": 2.69, + "learning_rate": 2.732516319743811e-06, + "loss": 0.5395, + "step": 7289 + }, + { + "epoch": 2.69, + "learning_rate": 2.7324793693804653e-06, + "loss": 0.6032, + "step": 7290 + }, + { + "epoch": 2.69, + "learning_rate": 2.7324424190171204e-06, + "loss": 0.5842, + "step": 7291 + }, + { + "epoch": 2.69, + "learning_rate": 2.732405468653775e-06, + "loss": 0.5866, + "step": 7292 + }, + { + "epoch": 2.69, + "learning_rate": 2.73236851829043e-06, + "loss": 0.5857, + "step": 7293 + }, + { + "epoch": 2.69, + "learning_rate": 2.7323315679270847e-06, + "loss": 0.6281, + "step": 7294 + }, + { + "epoch": 2.69, + "learning_rate": 2.7322946175637397e-06, + "loss": 0.528, + "step": 7295 + }, + { + "epoch": 2.69, + "learning_rate": 2.732257667200394e-06, + "loss": 0.5242, + "step": 7296 + }, + { + "epoch": 2.69, + "learning_rate": 2.732220716837049e-06, + "loss": 0.5827, + "step": 7297 + }, + { + "epoch": 2.69, + "learning_rate": 2.7321837664737037e-06, + "loss": 0.6573, + "step": 7298 + }, + { + "epoch": 2.7, + "learning_rate": 2.7321468161103583e-06, + "loss": 0.6344, + "step": 7299 + }, + { + "epoch": 2.7, + "learning_rate": 2.7321098657470133e-06, + "loss": 0.5144, + "step": 7300 + }, + { + "epoch": 2.7, + "learning_rate": 2.732072915383668e-06, + "loss": 0.5555, + "step": 7301 + }, + { + "epoch": 2.7, + "learning_rate": 2.7320359650203226e-06, + "loss": 0.573, + "step": 7302 + }, + { + "epoch": 2.7, + "learning_rate": 2.7319990146569773e-06, + "loss": 0.554, + "step": 7303 + }, + { + "epoch": 2.7, + "learning_rate": 2.7319620642936323e-06, + "loss": 0.523, + "step": 7304 + }, + { + "epoch": 2.7, + "learning_rate": 2.731925113930287e-06, + "loss": 0.5573, + "step": 7305 + }, + { + "epoch": 2.7, + "learning_rate": 2.731888163566942e-06, + "loss": 0.5369, + "step": 7306 + }, + { + "epoch": 2.7, + "learning_rate": 2.7318512132035967e-06, + "loss": 0.648, + "step": 7307 + }, + { + "epoch": 2.7, + "learning_rate": 2.7318142628402513e-06, + "loss": 0.5918, + "step": 7308 + }, + { + "epoch": 2.7, + "learning_rate": 2.731777312476906e-06, + "loss": 0.6142, + "step": 7309 + }, + { + "epoch": 2.7, + "learning_rate": 2.7317403621135606e-06, + "loss": 0.5276, + "step": 7310 + }, + { + "epoch": 2.7, + "learning_rate": 2.7317034117502156e-06, + "loss": 0.665, + "step": 7311 + }, + { + "epoch": 2.7, + "learning_rate": 2.7316664613868703e-06, + "loss": 0.5272, + "step": 7312 + }, + { + "epoch": 2.7, + "learning_rate": 2.7316295110235253e-06, + "loss": 0.5297, + "step": 7313 + }, + { + "epoch": 2.7, + "learning_rate": 2.73159256066018e-06, + "loss": 0.6818, + "step": 7314 + }, + { + "epoch": 2.7, + "learning_rate": 2.7315556102968346e-06, + "loss": 0.5645, + "step": 7315 + }, + { + "epoch": 2.7, + "learning_rate": 2.7315186599334892e-06, + "loss": 0.4372, + "step": 7316 + }, + { + "epoch": 2.7, + "learning_rate": 2.7314817095701443e-06, + "loss": 0.6264, + "step": 7317 + }, + { + "epoch": 2.7, + "learning_rate": 2.731444759206799e-06, + "loss": 0.5968, + "step": 7318 + }, + { + "epoch": 2.7, + "learning_rate": 2.7314078088434536e-06, + "loss": 0.5688, + "step": 7319 + }, + { + "epoch": 2.7, + "learning_rate": 2.7313708584801086e-06, + "loss": 0.6291, + "step": 7320 + }, + { + "epoch": 2.7, + "learning_rate": 2.7313339081167633e-06, + "loss": 0.6093, + "step": 7321 + }, + { + "epoch": 2.7, + "learning_rate": 2.731296957753418e-06, + "loss": 0.4621, + "step": 7322 + }, + { + "epoch": 2.7, + "learning_rate": 2.7312600073900725e-06, + "loss": 0.561, + "step": 7323 + }, + { + "epoch": 2.7, + "learning_rate": 2.7312230570267276e-06, + "loss": 0.6287, + "step": 7324 + }, + { + "epoch": 2.7, + "learning_rate": 2.7311861066633822e-06, + "loss": 0.6465, + "step": 7325 + }, + { + "epoch": 2.71, + "learning_rate": 2.7311491563000373e-06, + "loss": 0.5471, + "step": 7326 + }, + { + "epoch": 2.71, + "learning_rate": 2.731112205936692e-06, + "loss": 0.5284, + "step": 7327 + }, + { + "epoch": 2.71, + "learning_rate": 2.7310752555733466e-06, + "loss": 0.6465, + "step": 7328 + }, + { + "epoch": 2.71, + "learning_rate": 2.731038305210001e-06, + "loss": 0.5685, + "step": 7329 + }, + { + "epoch": 2.71, + "learning_rate": 2.731001354846656e-06, + "loss": 0.5868, + "step": 7330 + }, + { + "epoch": 2.71, + "learning_rate": 2.730964404483311e-06, + "loss": 0.5324, + "step": 7331 + }, + { + "epoch": 2.71, + "learning_rate": 2.7309274541199655e-06, + "loss": 0.569, + "step": 7332 + }, + { + "epoch": 2.71, + "learning_rate": 2.7308905037566206e-06, + "loss": 0.7863, + "step": 7333 + }, + { + "epoch": 2.71, + "learning_rate": 2.7308535533932752e-06, + "loss": 0.6042, + "step": 7334 + }, + { + "epoch": 2.71, + "learning_rate": 2.73081660302993e-06, + "loss": 0.5954, + "step": 7335 + }, + { + "epoch": 2.71, + "learning_rate": 2.7307796526665845e-06, + "loss": 0.5461, + "step": 7336 + }, + { + "epoch": 2.71, + "learning_rate": 2.7307427023032396e-06, + "loss": 0.6807, + "step": 7337 + }, + { + "epoch": 2.71, + "learning_rate": 2.730705751939894e-06, + "loss": 0.5531, + "step": 7338 + }, + { + "epoch": 2.71, + "learning_rate": 2.730668801576549e-06, + "loss": 0.6075, + "step": 7339 + }, + { + "epoch": 2.71, + "learning_rate": 2.730631851213204e-06, + "loss": 0.5336, + "step": 7340 + }, + { + "epoch": 2.71, + "learning_rate": 2.730594900849858e-06, + "loss": 0.5416, + "step": 7341 + }, + { + "epoch": 2.71, + "learning_rate": 2.730557950486513e-06, + "loss": 0.5472, + "step": 7342 + }, + { + "epoch": 2.71, + "learning_rate": 2.730521000123168e-06, + "loss": 0.6321, + "step": 7343 + }, + { + "epoch": 2.71, + "learning_rate": 2.730484049759823e-06, + "loss": 0.5982, + "step": 7344 + }, + { + "epoch": 2.71, + "learning_rate": 2.7304470993964775e-06, + "loss": 0.5763, + "step": 7345 + }, + { + "epoch": 2.71, + "learning_rate": 2.7304101490331326e-06, + "loss": 0.6473, + "step": 7346 + }, + { + "epoch": 2.71, + "learning_rate": 2.7303731986697868e-06, + "loss": 0.6751, + "step": 7347 + }, + { + "epoch": 2.71, + "learning_rate": 2.730336248306442e-06, + "loss": 0.6144, + "step": 7348 + }, + { + "epoch": 2.71, + "learning_rate": 2.7302992979430965e-06, + "loss": 0.5698, + "step": 7349 + }, + { + "epoch": 2.71, + "learning_rate": 2.730262347579751e-06, + "loss": 0.6338, + "step": 7350 + }, + { + "epoch": 2.71, + "learning_rate": 2.730225397216406e-06, + "loss": 0.6558, + "step": 7351 + }, + { + "epoch": 2.71, + "learning_rate": 2.730188446853061e-06, + "loss": 0.6257, + "step": 7352 + }, + { + "epoch": 2.72, + "learning_rate": 2.7301514964897154e-06, + "loss": 0.55, + "step": 7353 + }, + { + "epoch": 2.72, + "learning_rate": 2.73011454612637e-06, + "loss": 0.6431, + "step": 7354 + }, + { + "epoch": 2.72, + "learning_rate": 2.730077595763025e-06, + "loss": 0.605, + "step": 7355 + }, + { + "epoch": 2.72, + "learning_rate": 2.7300406453996798e-06, + "loss": 0.6333, + "step": 7356 + }, + { + "epoch": 2.72, + "learning_rate": 2.730003695036335e-06, + "loss": 0.6662, + "step": 7357 + }, + { + "epoch": 2.72, + "learning_rate": 2.7299667446729895e-06, + "loss": 0.6, + "step": 7358 + }, + { + "epoch": 2.72, + "learning_rate": 2.729929794309644e-06, + "loss": 0.6777, + "step": 7359 + }, + { + "epoch": 2.72, + "learning_rate": 2.7298928439462987e-06, + "loss": 0.5278, + "step": 7360 + }, + { + "epoch": 2.72, + "learning_rate": 2.7298558935829534e-06, + "loss": 0.6004, + "step": 7361 + }, + { + "epoch": 2.72, + "learning_rate": 2.7298189432196084e-06, + "loss": 0.5951, + "step": 7362 + }, + { + "epoch": 2.72, + "learning_rate": 2.729781992856263e-06, + "loss": 0.5238, + "step": 7363 + }, + { + "epoch": 2.72, + "learning_rate": 2.729745042492918e-06, + "loss": 0.6166, + "step": 7364 + }, + { + "epoch": 2.72, + "learning_rate": 2.7297080921295728e-06, + "loss": 0.6043, + "step": 7365 + }, + { + "epoch": 2.72, + "learning_rate": 2.7296711417662274e-06, + "loss": 0.5923, + "step": 7366 + }, + { + "epoch": 2.72, + "learning_rate": 2.729634191402882e-06, + "loss": 0.5563, + "step": 7367 + }, + { + "epoch": 2.72, + "learning_rate": 2.729597241039537e-06, + "loss": 0.5617, + "step": 7368 + }, + { + "epoch": 2.72, + "learning_rate": 2.7295602906761917e-06, + "loss": 0.6371, + "step": 7369 + }, + { + "epoch": 2.72, + "learning_rate": 2.7295233403128464e-06, + "loss": 0.5674, + "step": 7370 + }, + { + "epoch": 2.72, + "learning_rate": 2.7294863899495014e-06, + "loss": 0.5808, + "step": 7371 + }, + { + "epoch": 2.72, + "learning_rate": 2.729449439586156e-06, + "loss": 0.499, + "step": 7372 + }, + { + "epoch": 2.72, + "learning_rate": 2.7294124892228107e-06, + "loss": 0.5696, + "step": 7373 + }, + { + "epoch": 2.72, + "learning_rate": 2.7293755388594653e-06, + "loss": 0.5894, + "step": 7374 + }, + { + "epoch": 2.72, + "learning_rate": 2.7293385884961204e-06, + "loss": 0.5827, + "step": 7375 + }, + { + "epoch": 2.72, + "learning_rate": 2.729301638132775e-06, + "loss": 0.6162, + "step": 7376 + }, + { + "epoch": 2.72, + "learning_rate": 2.72926468776943e-06, + "loss": 0.4997, + "step": 7377 + }, + { + "epoch": 2.72, + "learning_rate": 2.7292277374060847e-06, + "loss": 0.5062, + "step": 7378 + }, + { + "epoch": 2.72, + "learning_rate": 2.7291907870427394e-06, + "loss": 0.5892, + "step": 7379 + }, + { + "epoch": 2.73, + "learning_rate": 2.729153836679394e-06, + "loss": 0.6033, + "step": 7380 + }, + { + "epoch": 2.73, + "learning_rate": 2.7291168863160486e-06, + "loss": 0.5596, + "step": 7381 + }, + { + "epoch": 2.73, + "learning_rate": 2.7290799359527037e-06, + "loss": 0.5944, + "step": 7382 + }, + { + "epoch": 2.73, + "learning_rate": 2.7290429855893583e-06, + "loss": 0.5587, + "step": 7383 + }, + { + "epoch": 2.73, + "learning_rate": 2.7290060352260134e-06, + "loss": 0.5971, + "step": 7384 + }, + { + "epoch": 2.73, + "learning_rate": 2.728969084862668e-06, + "loss": 0.4832, + "step": 7385 + }, + { + "epoch": 2.73, + "learning_rate": 2.7289321344993227e-06, + "loss": 0.4986, + "step": 7386 + }, + { + "epoch": 2.73, + "learning_rate": 2.7288951841359773e-06, + "loss": 0.5535, + "step": 7387 + }, + { + "epoch": 2.73, + "learning_rate": 2.7288582337726324e-06, + "loss": 0.537, + "step": 7388 + }, + { + "epoch": 2.73, + "learning_rate": 2.728821283409287e-06, + "loss": 0.5038, + "step": 7389 + }, + { + "epoch": 2.73, + "learning_rate": 2.7287843330459416e-06, + "loss": 0.5357, + "step": 7390 + }, + { + "epoch": 2.73, + "learning_rate": 2.7287473826825967e-06, + "loss": 0.5996, + "step": 7391 + }, + { + "epoch": 2.73, + "learning_rate": 2.728710432319251e-06, + "loss": 0.5606, + "step": 7392 + }, + { + "epoch": 2.73, + "learning_rate": 2.728673481955906e-06, + "loss": 0.5768, + "step": 7393 + }, + { + "epoch": 2.73, + "learning_rate": 2.7286365315925606e-06, + "loss": 0.6767, + "step": 7394 + }, + { + "epoch": 2.73, + "learning_rate": 2.7285995812292157e-06, + "loss": 0.5425, + "step": 7395 + }, + { + "epoch": 2.73, + "learning_rate": 2.7285626308658703e-06, + "loss": 0.5575, + "step": 7396 + }, + { + "epoch": 2.73, + "learning_rate": 2.7285256805025254e-06, + "loss": 0.5871, + "step": 7397 + }, + { + "epoch": 2.73, + "learning_rate": 2.7284887301391796e-06, + "loss": 0.618, + "step": 7398 + }, + { + "epoch": 2.73, + "learning_rate": 2.7284517797758342e-06, + "loss": 0.5663, + "step": 7399 + }, + { + "epoch": 2.73, + "learning_rate": 2.7284148294124893e-06, + "loss": 0.6346, + "step": 7400 + }, + { + "epoch": 2.73, + "learning_rate": 2.728377879049144e-06, + "loss": 0.7113, + "step": 7401 + }, + { + "epoch": 2.73, + "learning_rate": 2.728340928685799e-06, + "loss": 0.6585, + "step": 7402 + }, + { + "epoch": 2.73, + "learning_rate": 2.7283039783224536e-06, + "loss": 0.6634, + "step": 7403 + }, + { + "epoch": 2.73, + "learning_rate": 2.7282670279591082e-06, + "loss": 0.6567, + "step": 7404 + }, + { + "epoch": 2.73, + "learning_rate": 2.728230077595763e-06, + "loss": 0.6621, + "step": 7405 + }, + { + "epoch": 2.73, + "learning_rate": 2.728193127232418e-06, + "loss": 0.5219, + "step": 7406 + }, + { + "epoch": 2.74, + "learning_rate": 2.7281561768690726e-06, + "loss": 0.5775, + "step": 7407 + }, + { + "epoch": 2.74, + "learning_rate": 2.7281192265057276e-06, + "loss": 0.6212, + "step": 7408 + }, + { + "epoch": 2.74, + "learning_rate": 2.7280822761423823e-06, + "loss": 0.5399, + "step": 7409 + }, + { + "epoch": 2.74, + "learning_rate": 2.728045325779037e-06, + "loss": 0.4813, + "step": 7410 + }, + { + "epoch": 2.74, + "learning_rate": 2.7280083754156915e-06, + "loss": 0.6356, + "step": 7411 + }, + { + "epoch": 2.74, + "learning_rate": 2.727971425052346e-06, + "loss": 0.6348, + "step": 7412 + }, + { + "epoch": 2.74, + "learning_rate": 2.7279344746890012e-06, + "loss": 0.5715, + "step": 7413 + }, + { + "epoch": 2.74, + "learning_rate": 2.727897524325656e-06, + "loss": 0.5134, + "step": 7414 + }, + { + "epoch": 2.74, + "learning_rate": 2.727860573962311e-06, + "loss": 0.5819, + "step": 7415 + }, + { + "epoch": 2.74, + "learning_rate": 2.7278236235989656e-06, + "loss": 0.5726, + "step": 7416 + }, + { + "epoch": 2.74, + "learning_rate": 2.72778667323562e-06, + "loss": 0.5954, + "step": 7417 + }, + { + "epoch": 2.74, + "learning_rate": 2.727749722872275e-06, + "loss": 0.6554, + "step": 7418 + }, + { + "epoch": 2.74, + "learning_rate": 2.7277127725089295e-06, + "loss": 0.7002, + "step": 7419 + }, + { + "epoch": 2.74, + "learning_rate": 2.7276758221455845e-06, + "loss": 0.5559, + "step": 7420 + }, + { + "epoch": 2.74, + "learning_rate": 2.727638871782239e-06, + "loss": 0.5836, + "step": 7421 + }, + { + "epoch": 2.74, + "learning_rate": 2.7276019214188942e-06, + "loss": 0.6569, + "step": 7422 + }, + { + "epoch": 2.74, + "learning_rate": 2.727564971055549e-06, + "loss": 0.5885, + "step": 7423 + }, + { + "epoch": 2.74, + "learning_rate": 2.7275280206922035e-06, + "loss": 0.62, + "step": 7424 + }, + { + "epoch": 2.74, + "learning_rate": 2.727491070328858e-06, + "loss": 0.6023, + "step": 7425 + }, + { + "epoch": 2.74, + "learning_rate": 2.727454119965513e-06, + "loss": 0.5959, + "step": 7426 + }, + { + "epoch": 2.74, + "learning_rate": 2.727417169602168e-06, + "loss": 0.6641, + "step": 7427 + }, + { + "epoch": 2.74, + "learning_rate": 2.727380219238823e-06, + "loss": 0.5158, + "step": 7428 + }, + { + "epoch": 2.74, + "learning_rate": 2.7273432688754775e-06, + "loss": 0.6516, + "step": 7429 + }, + { + "epoch": 2.74, + "learning_rate": 2.727306318512132e-06, + "loss": 0.6257, + "step": 7430 + }, + { + "epoch": 2.74, + "learning_rate": 2.727269368148787e-06, + "loss": 0.6873, + "step": 7431 + }, + { + "epoch": 2.74, + "learning_rate": 2.7272324177854415e-06, + "loss": 0.5582, + "step": 7432 + }, + { + "epoch": 2.74, + "learning_rate": 2.7271954674220965e-06, + "loss": 0.4577, + "step": 7433 + }, + { + "epoch": 2.75, + "learning_rate": 2.727158517058751e-06, + "loss": 0.6037, + "step": 7434 + }, + { + "epoch": 2.75, + "learning_rate": 2.727121566695406e-06, + "loss": 0.6228, + "step": 7435 + }, + { + "epoch": 2.75, + "learning_rate": 2.7270846163320604e-06, + "loss": 0.7018, + "step": 7436 + }, + { + "epoch": 2.75, + "learning_rate": 2.7270476659687155e-06, + "loss": 0.4978, + "step": 7437 + }, + { + "epoch": 2.75, + "learning_rate": 2.72701071560537e-06, + "loss": 0.552, + "step": 7438 + }, + { + "epoch": 2.75, + "learning_rate": 2.7269737652420248e-06, + "loss": 0.5968, + "step": 7439 + }, + { + "epoch": 2.75, + "learning_rate": 2.72693681487868e-06, + "loss": 0.6073, + "step": 7440 + }, + { + "epoch": 2.75, + "learning_rate": 2.7268998645153345e-06, + "loss": 0.548, + "step": 7441 + }, + { + "epoch": 2.75, + "learning_rate": 2.726862914151989e-06, + "loss": 0.5328, + "step": 7442 + }, + { + "epoch": 2.75, + "learning_rate": 2.7268259637886437e-06, + "loss": 0.5912, + "step": 7443 + }, + { + "epoch": 2.75, + "learning_rate": 2.7267890134252988e-06, + "loss": 0.6651, + "step": 7444 + }, + { + "epoch": 2.75, + "learning_rate": 2.7267520630619534e-06, + "loss": 0.6525, + "step": 7445 + }, + { + "epoch": 2.75, + "learning_rate": 2.7267151126986085e-06, + "loss": 0.6138, + "step": 7446 + }, + { + "epoch": 2.75, + "learning_rate": 2.726678162335263e-06, + "loss": 0.6174, + "step": 7447 + }, + { + "epoch": 2.75, + "learning_rate": 2.7266412119719178e-06, + "loss": 0.5781, + "step": 7448 + }, + { + "epoch": 2.75, + "learning_rate": 2.7266042616085724e-06, + "loss": 0.585, + "step": 7449 + }, + { + "epoch": 2.75, + "learning_rate": 2.726567311245227e-06, + "loss": 0.5612, + "step": 7450 + }, + { + "epoch": 2.75, + "learning_rate": 2.726530360881882e-06, + "loss": 0.5714, + "step": 7451 + }, + { + "epoch": 2.75, + "learning_rate": 2.7264934105185367e-06, + "loss": 0.5889, + "step": 7452 + }, + { + "epoch": 2.75, + "learning_rate": 2.7264564601551918e-06, + "loss": 0.578, + "step": 7453 + }, + { + "epoch": 2.75, + "learning_rate": 2.7264195097918464e-06, + "loss": 0.5995, + "step": 7454 + }, + { + "epoch": 2.75, + "learning_rate": 2.726382559428501e-06, + "loss": 0.6235, + "step": 7455 + }, + { + "epoch": 2.75, + "learning_rate": 2.7263456090651557e-06, + "loss": 0.5395, + "step": 7456 + }, + { + "epoch": 2.75, + "learning_rate": 2.7263086587018108e-06, + "loss": 0.5887, + "step": 7457 + }, + { + "epoch": 2.75, + "learning_rate": 2.7262717083384654e-06, + "loss": 0.6027, + "step": 7458 + }, + { + "epoch": 2.75, + "learning_rate": 2.72623475797512e-06, + "loss": 0.6005, + "step": 7459 + }, + { + "epoch": 2.75, + "learning_rate": 2.726197807611775e-06, + "loss": 0.573, + "step": 7460 + }, + { + "epoch": 2.76, + "learning_rate": 2.7261608572484297e-06, + "loss": 0.5712, + "step": 7461 + }, + { + "epoch": 2.76, + "learning_rate": 2.7261239068850844e-06, + "loss": 0.541, + "step": 7462 + }, + { + "epoch": 2.76, + "learning_rate": 2.726086956521739e-06, + "loss": 0.4951, + "step": 7463 + }, + { + "epoch": 2.76, + "learning_rate": 2.726050006158394e-06, + "loss": 0.6321, + "step": 7464 + }, + { + "epoch": 2.76, + "learning_rate": 2.7260130557950487e-06, + "loss": 0.5462, + "step": 7465 + }, + { + "epoch": 2.76, + "learning_rate": 2.7259761054317037e-06, + "loss": 0.5498, + "step": 7466 + }, + { + "epoch": 2.76, + "learning_rate": 2.7259391550683584e-06, + "loss": 0.6308, + "step": 7467 + }, + { + "epoch": 2.76, + "learning_rate": 2.725902204705013e-06, + "loss": 0.511, + "step": 7468 + }, + { + "epoch": 2.76, + "learning_rate": 2.7258652543416677e-06, + "loss": 0.5383, + "step": 7469 + }, + { + "epoch": 2.76, + "learning_rate": 2.7258283039783223e-06, + "loss": 0.656, + "step": 7470 + }, + { + "epoch": 2.76, + "learning_rate": 2.7257913536149774e-06, + "loss": 0.5705, + "step": 7471 + }, + { + "epoch": 2.76, + "learning_rate": 2.725754403251632e-06, + "loss": 0.6638, + "step": 7472 + }, + { + "epoch": 2.76, + "learning_rate": 2.725717452888287e-06, + "loss": 0.5566, + "step": 7473 + }, + { + "epoch": 2.76, + "learning_rate": 2.7256805025249417e-06, + "loss": 0.594, + "step": 7474 + }, + { + "epoch": 2.76, + "learning_rate": 2.7256435521615963e-06, + "loss": 0.5925, + "step": 7475 + }, + { + "epoch": 2.76, + "learning_rate": 2.725606601798251e-06, + "loss": 0.6266, + "step": 7476 + }, + { + "epoch": 2.76, + "learning_rate": 2.725569651434906e-06, + "loss": 0.5658, + "step": 7477 + }, + { + "epoch": 2.76, + "learning_rate": 2.7255327010715607e-06, + "loss": 0.5435, + "step": 7478 + }, + { + "epoch": 2.76, + "learning_rate": 2.7254957507082153e-06, + "loss": 0.6231, + "step": 7479 + }, + { + "epoch": 2.76, + "learning_rate": 2.7254588003448704e-06, + "loss": 0.5898, + "step": 7480 + }, + { + "epoch": 2.76, + "learning_rate": 2.7254218499815246e-06, + "loss": 0.6273, + "step": 7481 + }, + { + "epoch": 2.76, + "learning_rate": 2.7253848996181796e-06, + "loss": 0.5849, + "step": 7482 + }, + { + "epoch": 2.76, + "learning_rate": 2.7253479492548343e-06, + "loss": 0.6377, + "step": 7483 + }, + { + "epoch": 2.76, + "learning_rate": 2.7253109988914893e-06, + "loss": 0.5819, + "step": 7484 + }, + { + "epoch": 2.76, + "learning_rate": 2.725274048528144e-06, + "loss": 0.5784, + "step": 7485 + }, + { + "epoch": 2.76, + "learning_rate": 2.725237098164799e-06, + "loss": 0.6412, + "step": 7486 + }, + { + "epoch": 2.76, + "learning_rate": 2.7252001478014532e-06, + "loss": 0.5309, + "step": 7487 + }, + { + "epoch": 2.77, + "learning_rate": 2.725163197438108e-06, + "loss": 0.5789, + "step": 7488 + }, + { + "epoch": 2.77, + "learning_rate": 2.725126247074763e-06, + "loss": 0.5874, + "step": 7489 + }, + { + "epoch": 2.77, + "learning_rate": 2.7250892967114176e-06, + "loss": 0.6165, + "step": 7490 + }, + { + "epoch": 2.77, + "learning_rate": 2.7250523463480726e-06, + "loss": 0.5214, + "step": 7491 + }, + { + "epoch": 2.77, + "learning_rate": 2.7250153959847273e-06, + "loss": 0.5081, + "step": 7492 + }, + { + "epoch": 2.77, + "learning_rate": 2.724978445621382e-06, + "loss": 0.5168, + "step": 7493 + }, + { + "epoch": 2.77, + "learning_rate": 2.7249414952580365e-06, + "loss": 0.5446, + "step": 7494 + }, + { + "epoch": 2.77, + "learning_rate": 2.7249045448946916e-06, + "loss": 0.5829, + "step": 7495 + }, + { + "epoch": 2.77, + "learning_rate": 2.7248675945313462e-06, + "loss": 0.6262, + "step": 7496 + }, + { + "epoch": 2.77, + "learning_rate": 2.7248306441680013e-06, + "loss": 0.5607, + "step": 7497 + }, + { + "epoch": 2.77, + "learning_rate": 2.724793693804656e-06, + "loss": 0.5275, + "step": 7498 + }, + { + "epoch": 2.77, + "learning_rate": 2.7247567434413106e-06, + "loss": 0.5411, + "step": 7499 + }, + { + "epoch": 2.77, + "learning_rate": 2.724719793077965e-06, + "loss": 0.6187, + "step": 7500 + }, + { + "epoch": 2.77, + "learning_rate": 2.72468284271462e-06, + "loss": 0.6332, + "step": 7501 + }, + { + "epoch": 2.77, + "learning_rate": 2.724645892351275e-06, + "loss": 0.5883, + "step": 7502 + }, + { + "epoch": 2.77, + "learning_rate": 2.7246089419879295e-06, + "loss": 0.4787, + "step": 7503 + }, + { + "epoch": 2.77, + "learning_rate": 2.7245719916245846e-06, + "loss": 0.5447, + "step": 7504 + }, + { + "epoch": 2.77, + "learning_rate": 2.7245350412612392e-06, + "loss": 0.6748, + "step": 7505 + }, + { + "epoch": 2.77, + "learning_rate": 2.724498090897894e-06, + "loss": 0.5289, + "step": 7506 + }, + { + "epoch": 2.77, + "learning_rate": 2.7244611405345485e-06, + "loss": 0.6716, + "step": 7507 + }, + { + "epoch": 2.77, + "learning_rate": 2.724424190171203e-06, + "loss": 0.6237, + "step": 7508 + }, + { + "epoch": 2.77, + "learning_rate": 2.724387239807858e-06, + "loss": 0.6189, + "step": 7509 + }, + { + "epoch": 2.77, + "learning_rate": 2.724350289444513e-06, + "loss": 0.605, + "step": 7510 + }, + { + "epoch": 2.77, + "learning_rate": 2.724313339081168e-06, + "loss": 0.5984, + "step": 7511 + }, + { + "epoch": 2.77, + "learning_rate": 2.7242763887178225e-06, + "loss": 0.5982, + "step": 7512 + }, + { + "epoch": 2.77, + "learning_rate": 2.724239438354477e-06, + "loss": 0.5367, + "step": 7513 + }, + { + "epoch": 2.77, + "learning_rate": 2.724202487991132e-06, + "loss": 0.4898, + "step": 7514 + }, + { + "epoch": 2.78, + "learning_rate": 2.724165537627787e-06, + "loss": 0.5944, + "step": 7515 + }, + { + "epoch": 2.78, + "learning_rate": 2.7241285872644415e-06, + "loss": 0.5908, + "step": 7516 + }, + { + "epoch": 2.78, + "learning_rate": 2.7240916369010966e-06, + "loss": 0.5742, + "step": 7517 + }, + { + "epoch": 2.78, + "learning_rate": 2.724054686537751e-06, + "loss": 0.5738, + "step": 7518 + }, + { + "epoch": 2.78, + "learning_rate": 2.724017736174406e-06, + "loss": 0.5855, + "step": 7519 + }, + { + "epoch": 2.78, + "learning_rate": 2.7239807858110605e-06, + "loss": 0.5824, + "step": 7520 + }, + { + "epoch": 2.78, + "learning_rate": 2.723943835447715e-06, + "loss": 0.6931, + "step": 7521 + }, + { + "epoch": 2.78, + "learning_rate": 2.72390688508437e-06, + "loss": 0.5615, + "step": 7522 + }, + { + "epoch": 2.78, + "learning_rate": 2.723869934721025e-06, + "loss": 0.5269, + "step": 7523 + }, + { + "epoch": 2.78, + "learning_rate": 2.72383298435768e-06, + "loss": 0.5949, + "step": 7524 + }, + { + "epoch": 2.78, + "learning_rate": 2.7237960339943345e-06, + "loss": 0.6491, + "step": 7525 + }, + { + "epoch": 2.78, + "learning_rate": 2.723759083630989e-06, + "loss": 0.5372, + "step": 7526 + }, + { + "epoch": 2.78, + "learning_rate": 2.7237221332676438e-06, + "loss": 0.6071, + "step": 7527 + }, + { + "epoch": 2.78, + "learning_rate": 2.7236851829042984e-06, + "loss": 0.4929, + "step": 7528 + }, + { + "epoch": 2.78, + "learning_rate": 2.7236482325409535e-06, + "loss": 0.5951, + "step": 7529 + }, + { + "epoch": 2.78, + "learning_rate": 2.723611282177608e-06, + "loss": 0.5745, + "step": 7530 + }, + { + "epoch": 2.78, + "learning_rate": 2.723574331814263e-06, + "loss": 0.6257, + "step": 7531 + }, + { + "epoch": 2.78, + "learning_rate": 2.7235373814509174e-06, + "loss": 0.6263, + "step": 7532 + }, + { + "epoch": 2.78, + "learning_rate": 2.7235004310875724e-06, + "loss": 0.4941, + "step": 7533 + }, + { + "epoch": 2.78, + "learning_rate": 2.723463480724227e-06, + "loss": 0.6387, + "step": 7534 + }, + { + "epoch": 2.78, + "learning_rate": 2.723426530360882e-06, + "loss": 0.6142, + "step": 7535 + }, + { + "epoch": 2.78, + "learning_rate": 2.7233895799975368e-06, + "loss": 0.5285, + "step": 7536 + }, + { + "epoch": 2.78, + "learning_rate": 2.723352629634192e-06, + "loss": 0.5001, + "step": 7537 + }, + { + "epoch": 2.78, + "learning_rate": 2.723315679270846e-06, + "loss": 0.6168, + "step": 7538 + }, + { + "epoch": 2.78, + "learning_rate": 2.7232787289075007e-06, + "loss": 0.559, + "step": 7539 + }, + { + "epoch": 2.78, + "learning_rate": 2.7232417785441557e-06, + "loss": 0.6693, + "step": 7540 + }, + { + "epoch": 2.78, + "learning_rate": 2.7232048281808104e-06, + "loss": 0.6316, + "step": 7541 + }, + { + "epoch": 2.79, + "learning_rate": 2.7231678778174654e-06, + "loss": 0.6186, + "step": 7542 + }, + { + "epoch": 2.79, + "learning_rate": 2.72313092745412e-06, + "loss": 0.5946, + "step": 7543 + }, + { + "epoch": 2.79, + "learning_rate": 2.7230939770907747e-06, + "loss": 0.5282, + "step": 7544 + }, + { + "epoch": 2.79, + "learning_rate": 2.7230570267274293e-06, + "loss": 0.5663, + "step": 7545 + }, + { + "epoch": 2.79, + "learning_rate": 2.7230200763640844e-06, + "loss": 0.513, + "step": 7546 + }, + { + "epoch": 2.79, + "learning_rate": 2.722983126000739e-06, + "loss": 0.6089, + "step": 7547 + }, + { + "epoch": 2.79, + "learning_rate": 2.7229461756373937e-06, + "loss": 0.6464, + "step": 7548 + }, + { + "epoch": 2.79, + "learning_rate": 2.7229092252740487e-06, + "loss": 0.6521, + "step": 7549 + }, + { + "epoch": 2.79, + "learning_rate": 2.7228722749107034e-06, + "loss": 0.5739, + "step": 7550 + }, + { + "epoch": 2.79, + "learning_rate": 2.722835324547358e-06, + "loss": 0.6165, + "step": 7551 + }, + { + "epoch": 2.79, + "learning_rate": 2.7227983741840126e-06, + "loss": 0.5418, + "step": 7552 + }, + { + "epoch": 2.79, + "learning_rate": 2.7227614238206677e-06, + "loss": 0.5441, + "step": 7553 + }, + { + "epoch": 2.79, + "learning_rate": 2.7227244734573223e-06, + "loss": 0.5427, + "step": 7554 + }, + { + "epoch": 2.79, + "learning_rate": 2.7226875230939774e-06, + "loss": 0.6153, + "step": 7555 + }, + { + "epoch": 2.79, + "learning_rate": 2.722650572730632e-06, + "loss": 0.5484, + "step": 7556 + }, + { + "epoch": 2.79, + "learning_rate": 2.7226136223672867e-06, + "loss": 0.5658, + "step": 7557 + }, + { + "epoch": 2.79, + "learning_rate": 2.7225766720039413e-06, + "loss": 0.5991, + "step": 7558 + }, + { + "epoch": 2.79, + "learning_rate": 2.722539721640596e-06, + "loss": 0.5622, + "step": 7559 + }, + { + "epoch": 2.79, + "learning_rate": 2.722502771277251e-06, + "loss": 0.5855, + "step": 7560 + }, + { + "epoch": 2.79, + "learning_rate": 2.7224658209139056e-06, + "loss": 0.5766, + "step": 7561 + }, + { + "epoch": 2.79, + "learning_rate": 2.7224288705505607e-06, + "loss": 0.549, + "step": 7562 + }, + { + "epoch": 2.79, + "learning_rate": 2.7223919201872153e-06, + "loss": 0.5842, + "step": 7563 + }, + { + "epoch": 2.79, + "learning_rate": 2.72235496982387e-06, + "loss": 0.5592, + "step": 7564 + }, + { + "epoch": 2.79, + "learning_rate": 2.7223180194605246e-06, + "loss": 0.5661, + "step": 7565 + }, + { + "epoch": 2.79, + "learning_rate": 2.7222810690971797e-06, + "loss": 0.5679, + "step": 7566 + }, + { + "epoch": 2.79, + "learning_rate": 2.7222441187338343e-06, + "loss": 0.6087, + "step": 7567 + }, + { + "epoch": 2.79, + "learning_rate": 2.722207168370489e-06, + "loss": 0.5774, + "step": 7568 + }, + { + "epoch": 2.8, + "learning_rate": 2.722170218007144e-06, + "loss": 0.5682, + "step": 7569 + }, + { + "epoch": 2.8, + "learning_rate": 2.7221332676437986e-06, + "loss": 0.5684, + "step": 7570 + }, + { + "epoch": 2.8, + "learning_rate": 2.7220963172804533e-06, + "loss": 0.6613, + "step": 7571 + }, + { + "epoch": 2.8, + "learning_rate": 2.722059366917108e-06, + "loss": 0.6211, + "step": 7572 + }, + { + "epoch": 2.8, + "learning_rate": 2.722022416553763e-06, + "loss": 0.5862, + "step": 7573 + }, + { + "epoch": 2.8, + "learning_rate": 2.7219854661904176e-06, + "loss": 0.5622, + "step": 7574 + }, + { + "epoch": 2.8, + "learning_rate": 2.7219485158270727e-06, + "loss": 0.5547, + "step": 7575 + }, + { + "epoch": 2.8, + "learning_rate": 2.7219115654637273e-06, + "loss": 0.5109, + "step": 7576 + }, + { + "epoch": 2.8, + "learning_rate": 2.721874615100382e-06, + "loss": 0.6494, + "step": 7577 + }, + { + "epoch": 2.8, + "learning_rate": 2.7218376647370366e-06, + "loss": 0.5821, + "step": 7578 + }, + { + "epoch": 2.8, + "learning_rate": 2.7218007143736912e-06, + "loss": 0.6095, + "step": 7579 + }, + { + "epoch": 2.8, + "learning_rate": 2.7217637640103463e-06, + "loss": 0.6382, + "step": 7580 + }, + { + "epoch": 2.8, + "learning_rate": 2.721726813647001e-06, + "loss": 0.5054, + "step": 7581 + }, + { + "epoch": 2.8, + "learning_rate": 2.721689863283656e-06, + "loss": 0.6388, + "step": 7582 + }, + { + "epoch": 2.8, + "learning_rate": 2.72165291292031e-06, + "loss": 0.5411, + "step": 7583 + }, + { + "epoch": 2.8, + "learning_rate": 2.7216159625569652e-06, + "loss": 0.6117, + "step": 7584 + }, + { + "epoch": 2.8, + "learning_rate": 2.72157901219362e-06, + "loss": 0.6042, + "step": 7585 + }, + { + "epoch": 2.8, + "learning_rate": 2.721542061830275e-06, + "loss": 0.5744, + "step": 7586 + }, + { + "epoch": 2.8, + "learning_rate": 2.7215051114669296e-06, + "loss": 0.5841, + "step": 7587 + }, + { + "epoch": 2.8, + "learning_rate": 2.7214681611035842e-06, + "loss": 0.5518, + "step": 7588 + }, + { + "epoch": 2.8, + "learning_rate": 2.721431210740239e-06, + "loss": 0.5551, + "step": 7589 + }, + { + "epoch": 2.8, + "learning_rate": 2.7213942603768935e-06, + "loss": 0.5718, + "step": 7590 + }, + { + "epoch": 2.8, + "learning_rate": 2.7213573100135486e-06, + "loss": 0.6384, + "step": 7591 + }, + { + "epoch": 2.8, + "learning_rate": 2.721320359650203e-06, + "loss": 0.4752, + "step": 7592 + }, + { + "epoch": 2.8, + "learning_rate": 2.7212834092868582e-06, + "loss": 0.5058, + "step": 7593 + }, + { + "epoch": 2.8, + "learning_rate": 2.721246458923513e-06, + "loss": 0.6208, + "step": 7594 + }, + { + "epoch": 2.8, + "learning_rate": 2.7212095085601675e-06, + "loss": 0.6231, + "step": 7595 + }, + { + "epoch": 2.81, + "learning_rate": 2.721172558196822e-06, + "loss": 0.5787, + "step": 7596 + }, + { + "epoch": 2.81, + "learning_rate": 2.7211356078334772e-06, + "loss": 0.5777, + "step": 7597 + }, + { + "epoch": 2.81, + "learning_rate": 2.721098657470132e-06, + "loss": 0.5129, + "step": 7598 + }, + { + "epoch": 2.81, + "learning_rate": 2.7210617071067865e-06, + "loss": 0.5664, + "step": 7599 + }, + { + "epoch": 2.81, + "learning_rate": 2.7210247567434415e-06, + "loss": 0.6126, + "step": 7600 + }, + { + "epoch": 2.81, + "learning_rate": 2.720987806380096e-06, + "loss": 0.6908, + "step": 7601 + }, + { + "epoch": 2.81, + "learning_rate": 2.720950856016751e-06, + "loss": 0.5517, + "step": 7602 + }, + { + "epoch": 2.81, + "learning_rate": 2.7209139056534055e-06, + "loss": 0.5533, + "step": 7603 + }, + { + "epoch": 2.81, + "learning_rate": 2.7208769552900605e-06, + "loss": 0.5769, + "step": 7604 + }, + { + "epoch": 2.81, + "learning_rate": 2.720840004926715e-06, + "loss": 0.6323, + "step": 7605 + }, + { + "epoch": 2.81, + "learning_rate": 2.7208030545633702e-06, + "loss": 0.6558, + "step": 7606 + }, + { + "epoch": 2.81, + "learning_rate": 2.720766104200025e-06, + "loss": 0.5943, + "step": 7607 + }, + { + "epoch": 2.81, + "learning_rate": 2.7207291538366795e-06, + "loss": 0.5907, + "step": 7608 + }, + { + "epoch": 2.81, + "learning_rate": 2.720692203473334e-06, + "loss": 0.6758, + "step": 7609 + }, + { + "epoch": 2.81, + "learning_rate": 2.7206552531099888e-06, + "loss": 0.5998, + "step": 7610 + }, + { + "epoch": 2.81, + "learning_rate": 2.720618302746644e-06, + "loss": 0.5984, + "step": 7611 + }, + { + "epoch": 2.81, + "learning_rate": 2.7205813523832985e-06, + "loss": 0.6256, + "step": 7612 + }, + { + "epoch": 2.81, + "learning_rate": 2.7205444020199535e-06, + "loss": 0.614, + "step": 7613 + }, + { + "epoch": 2.81, + "learning_rate": 2.720507451656608e-06, + "loss": 0.5002, + "step": 7614 + }, + { + "epoch": 2.81, + "learning_rate": 2.720470501293263e-06, + "loss": 0.6095, + "step": 7615 + }, + { + "epoch": 2.81, + "learning_rate": 2.7204335509299174e-06, + "loss": 0.5835, + "step": 7616 + }, + { + "epoch": 2.81, + "learning_rate": 2.7203966005665725e-06, + "loss": 0.5732, + "step": 7617 + }, + { + "epoch": 2.81, + "learning_rate": 2.720359650203227e-06, + "loss": 0.6159, + "step": 7618 + }, + { + "epoch": 2.81, + "learning_rate": 2.7203226998398818e-06, + "loss": 0.5455, + "step": 7619 + }, + { + "epoch": 2.81, + "learning_rate": 2.720285749476537e-06, + "loss": 0.6318, + "step": 7620 + }, + { + "epoch": 2.81, + "learning_rate": 2.720248799113191e-06, + "loss": 0.566, + "step": 7621 + }, + { + "epoch": 2.81, + "learning_rate": 2.720211848749846e-06, + "loss": 0.6016, + "step": 7622 + }, + { + "epoch": 2.81, + "learning_rate": 2.7201748983865007e-06, + "loss": 0.6657, + "step": 7623 + }, + { + "epoch": 2.82, + "learning_rate": 2.7201379480231558e-06, + "loss": 0.5417, + "step": 7624 + }, + { + "epoch": 2.82, + "learning_rate": 2.7201009976598104e-06, + "loss": 0.5752, + "step": 7625 + }, + { + "epoch": 2.82, + "learning_rate": 2.7200640472964655e-06, + "loss": 0.5693, + "step": 7626 + }, + { + "epoch": 2.82, + "learning_rate": 2.7200270969331197e-06, + "loss": 0.6167, + "step": 7627 + }, + { + "epoch": 2.82, + "learning_rate": 2.7199901465697743e-06, + "loss": 0.6463, + "step": 7628 + }, + { + "epoch": 2.82, + "learning_rate": 2.7199531962064294e-06, + "loss": 0.5552, + "step": 7629 + }, + { + "epoch": 2.82, + "learning_rate": 2.719916245843084e-06, + "loss": 0.584, + "step": 7630 + }, + { + "epoch": 2.82, + "learning_rate": 2.719879295479739e-06, + "loss": 0.5771, + "step": 7631 + }, + { + "epoch": 2.82, + "learning_rate": 2.7198423451163937e-06, + "loss": 0.5666, + "step": 7632 + }, + { + "epoch": 2.82, + "learning_rate": 2.7198053947530484e-06, + "loss": 0.5552, + "step": 7633 + }, + { + "epoch": 2.82, + "learning_rate": 2.719768444389703e-06, + "loss": 0.5597, + "step": 7634 + }, + { + "epoch": 2.82, + "learning_rate": 2.719731494026358e-06, + "loss": 0.5548, + "step": 7635 + }, + { + "epoch": 2.82, + "learning_rate": 2.7196945436630127e-06, + "loss": 0.6028, + "step": 7636 + }, + { + "epoch": 2.82, + "learning_rate": 2.7196575932996678e-06, + "loss": 0.5112, + "step": 7637 + }, + { + "epoch": 2.82, + "learning_rate": 2.7196206429363224e-06, + "loss": 0.624, + "step": 7638 + }, + { + "epoch": 2.82, + "learning_rate": 2.719583692572977e-06, + "loss": 0.6711, + "step": 7639 + }, + { + "epoch": 2.82, + "learning_rate": 2.7195467422096317e-06, + "loss": 0.6384, + "step": 7640 + }, + { + "epoch": 2.82, + "learning_rate": 2.7195097918462863e-06, + "loss": 0.5757, + "step": 7641 + }, + { + "epoch": 2.82, + "learning_rate": 2.7194728414829414e-06, + "loss": 0.7251, + "step": 7642 + }, + { + "epoch": 2.82, + "learning_rate": 2.719435891119596e-06, + "loss": 0.5264, + "step": 7643 + }, + { + "epoch": 2.82, + "learning_rate": 2.719398940756251e-06, + "loss": 0.6372, + "step": 7644 + }, + { + "epoch": 2.82, + "learning_rate": 2.7193619903929057e-06, + "loss": 0.5176, + "step": 7645 + }, + { + "epoch": 2.82, + "learning_rate": 2.7193250400295603e-06, + "loss": 0.6062, + "step": 7646 + }, + { + "epoch": 2.82, + "learning_rate": 2.719288089666215e-06, + "loss": 0.5441, + "step": 7647 + }, + { + "epoch": 2.82, + "learning_rate": 2.7192511393028696e-06, + "loss": 0.5766, + "step": 7648 + }, + { + "epoch": 2.82, + "learning_rate": 2.7192141889395247e-06, + "loss": 0.5964, + "step": 7649 + }, + { + "epoch": 2.82, + "learning_rate": 2.7191772385761793e-06, + "loss": 0.5679, + "step": 7650 + }, + { + "epoch": 2.83, + "learning_rate": 2.7191402882128344e-06, + "loss": 0.6704, + "step": 7651 + }, + { + "epoch": 2.83, + "learning_rate": 2.719103337849489e-06, + "loss": 0.6023, + "step": 7652 + }, + { + "epoch": 2.83, + "learning_rate": 2.7190663874861436e-06, + "loss": 0.465, + "step": 7653 + }, + { + "epoch": 2.83, + "learning_rate": 2.7190294371227983e-06, + "loss": 0.5345, + "step": 7654 + }, + { + "epoch": 2.83, + "learning_rate": 2.7189924867594533e-06, + "loss": 0.5672, + "step": 7655 + }, + { + "epoch": 2.83, + "learning_rate": 2.718955536396108e-06, + "loss": 0.592, + "step": 7656 + }, + { + "epoch": 2.83, + "learning_rate": 2.718918586032763e-06, + "loss": 0.5675, + "step": 7657 + }, + { + "epoch": 2.83, + "learning_rate": 2.7188816356694177e-06, + "loss": 0.5561, + "step": 7658 + }, + { + "epoch": 2.83, + "learning_rate": 2.7188446853060723e-06, + "loss": 0.591, + "step": 7659 + }, + { + "epoch": 2.83, + "learning_rate": 2.718807734942727e-06, + "loss": 0.5974, + "step": 7660 + }, + { + "epoch": 2.83, + "learning_rate": 2.7187707845793816e-06, + "loss": 0.5827, + "step": 7661 + }, + { + "epoch": 2.83, + "learning_rate": 2.7187338342160366e-06, + "loss": 0.5851, + "step": 7662 + }, + { + "epoch": 2.83, + "learning_rate": 2.7186968838526913e-06, + "loss": 0.6034, + "step": 7663 + }, + { + "epoch": 2.83, + "learning_rate": 2.7186599334893463e-06, + "loss": 0.5461, + "step": 7664 + }, + { + "epoch": 2.83, + "learning_rate": 2.718622983126001e-06, + "loss": 0.5009, + "step": 7665 + }, + { + "epoch": 2.83, + "learning_rate": 2.7185860327626556e-06, + "loss": 0.5872, + "step": 7666 + }, + { + "epoch": 2.83, + "learning_rate": 2.7185490823993102e-06, + "loss": 0.5134, + "step": 7667 + }, + { + "epoch": 2.83, + "learning_rate": 2.718512132035965e-06, + "loss": 0.5903, + "step": 7668 + }, + { + "epoch": 2.83, + "learning_rate": 2.71847518167262e-06, + "loss": 0.5484, + "step": 7669 + }, + { + "epoch": 2.83, + "learning_rate": 2.7184382313092746e-06, + "loss": 0.5064, + "step": 7670 + }, + { + "epoch": 2.83, + "learning_rate": 2.7184012809459296e-06, + "loss": 0.5397, + "step": 7671 + }, + { + "epoch": 2.83, + "learning_rate": 2.718364330582584e-06, + "loss": 0.7255, + "step": 7672 + }, + { + "epoch": 2.83, + "learning_rate": 2.718327380219239e-06, + "loss": 0.5056, + "step": 7673 + }, + { + "epoch": 2.83, + "learning_rate": 2.7182904298558935e-06, + "loss": 0.6251, + "step": 7674 + }, + { + "epoch": 2.83, + "learning_rate": 2.7182534794925486e-06, + "loss": 0.5867, + "step": 7675 + }, + { + "epoch": 2.83, + "learning_rate": 2.7182165291292032e-06, + "loss": 0.6264, + "step": 7676 + }, + { + "epoch": 2.83, + "learning_rate": 2.7181795787658583e-06, + "loss": 0.6061, + "step": 7677 + }, + { + "epoch": 2.84, + "learning_rate": 2.7181426284025125e-06, + "loss": 0.4574, + "step": 7678 + }, + { + "epoch": 2.84, + "learning_rate": 2.718105678039167e-06, + "loss": 0.5308, + "step": 7679 + }, + { + "epoch": 2.84, + "learning_rate": 2.718068727675822e-06, + "loss": 0.6425, + "step": 7680 + }, + { + "epoch": 2.84, + "learning_rate": 2.718031777312477e-06, + "loss": 0.589, + "step": 7681 + }, + { + "epoch": 2.84, + "learning_rate": 2.717994826949132e-06, + "loss": 0.4963, + "step": 7682 + }, + { + "epoch": 2.84, + "learning_rate": 2.7179578765857865e-06, + "loss": 0.5114, + "step": 7683 + }, + { + "epoch": 2.84, + "learning_rate": 2.717920926222441e-06, + "loss": 0.4967, + "step": 7684 + }, + { + "epoch": 2.84, + "learning_rate": 2.717883975859096e-06, + "loss": 0.5435, + "step": 7685 + }, + { + "epoch": 2.84, + "learning_rate": 2.717847025495751e-06, + "loss": 0.5624, + "step": 7686 + }, + { + "epoch": 2.84, + "learning_rate": 2.7178100751324055e-06, + "loss": 0.6425, + "step": 7687 + }, + { + "epoch": 2.84, + "learning_rate": 2.71777312476906e-06, + "loss": 0.6628, + "step": 7688 + }, + { + "epoch": 2.84, + "learning_rate": 2.717736174405715e-06, + "loss": 0.5664, + "step": 7689 + }, + { + "epoch": 2.84, + "learning_rate": 2.71769922404237e-06, + "loss": 0.5792, + "step": 7690 + }, + { + "epoch": 2.84, + "learning_rate": 2.7176622736790245e-06, + "loss": 0.6237, + "step": 7691 + }, + { + "epoch": 2.84, + "learning_rate": 2.717625323315679e-06, + "loss": 0.5139, + "step": 7692 + }, + { + "epoch": 2.84, + "learning_rate": 2.717588372952334e-06, + "loss": 0.5467, + "step": 7693 + }, + { + "epoch": 2.84, + "learning_rate": 2.717551422588989e-06, + "loss": 0.6195, + "step": 7694 + }, + { + "epoch": 2.84, + "learning_rate": 2.717514472225644e-06, + "loss": 0.5345, + "step": 7695 + }, + { + "epoch": 2.84, + "learning_rate": 2.7174775218622985e-06, + "loss": 0.5709, + "step": 7696 + }, + { + "epoch": 2.84, + "learning_rate": 2.717440571498953e-06, + "loss": 0.5534, + "step": 7697 + }, + { + "epoch": 2.84, + "learning_rate": 2.7174036211356078e-06, + "loss": 0.5824, + "step": 7698 + }, + { + "epoch": 2.84, + "learning_rate": 2.7173666707722624e-06, + "loss": 0.5952, + "step": 7699 + }, + { + "epoch": 2.84, + "learning_rate": 2.7173297204089175e-06, + "loss": 0.5487, + "step": 7700 + }, + { + "epoch": 2.84, + "learning_rate": 2.717292770045572e-06, + "loss": 0.6829, + "step": 7701 + }, + { + "epoch": 2.84, + "learning_rate": 2.717255819682227e-06, + "loss": 0.602, + "step": 7702 + }, + { + "epoch": 2.84, + "learning_rate": 2.717218869318882e-06, + "loss": 0.6247, + "step": 7703 + }, + { + "epoch": 2.84, + "learning_rate": 2.7171819189555364e-06, + "loss": 0.5986, + "step": 7704 + }, + { + "epoch": 2.85, + "learning_rate": 2.717144968592191e-06, + "loss": 0.6214, + "step": 7705 + }, + { + "epoch": 2.85, + "learning_rate": 2.717108018228846e-06, + "loss": 0.6476, + "step": 7706 + }, + { + "epoch": 2.85, + "learning_rate": 2.7170710678655008e-06, + "loss": 0.5274, + "step": 7707 + }, + { + "epoch": 2.85, + "learning_rate": 2.7170341175021554e-06, + "loss": 0.6266, + "step": 7708 + }, + { + "epoch": 2.85, + "learning_rate": 2.7169971671388105e-06, + "loss": 0.4971, + "step": 7709 + }, + { + "epoch": 2.85, + "learning_rate": 2.716960216775465e-06, + "loss": 0.4978, + "step": 7710 + }, + { + "epoch": 2.85, + "learning_rate": 2.7169232664121197e-06, + "loss": 0.6372, + "step": 7711 + }, + { + "epoch": 2.85, + "learning_rate": 2.7168863160487744e-06, + "loss": 0.5746, + "step": 7712 + }, + { + "epoch": 2.85, + "learning_rate": 2.7168493656854294e-06, + "loss": 0.5789, + "step": 7713 + }, + { + "epoch": 2.85, + "learning_rate": 2.716812415322084e-06, + "loss": 0.5423, + "step": 7714 + }, + { + "epoch": 2.85, + "learning_rate": 2.716775464958739e-06, + "loss": 0.6405, + "step": 7715 + }, + { + "epoch": 2.85, + "learning_rate": 2.7167385145953938e-06, + "loss": 0.5223, + "step": 7716 + }, + { + "epoch": 2.85, + "learning_rate": 2.716701564232048e-06, + "loss": 0.5948, + "step": 7717 + }, + { + "epoch": 2.85, + "learning_rate": 2.716664613868703e-06, + "loss": 0.5475, + "step": 7718 + }, + { + "epoch": 2.85, + "learning_rate": 2.7166276635053577e-06, + "loss": 0.6239, + "step": 7719 + }, + { + "epoch": 2.85, + "learning_rate": 2.7165907131420127e-06, + "loss": 0.484, + "step": 7720 + }, + { + "epoch": 2.85, + "learning_rate": 2.7165537627786674e-06, + "loss": 0.5602, + "step": 7721 + }, + { + "epoch": 2.85, + "learning_rate": 2.7165168124153224e-06, + "loss": 0.541, + "step": 7722 + }, + { + "epoch": 2.85, + "learning_rate": 2.7164798620519767e-06, + "loss": 0.4976, + "step": 7723 + }, + { + "epoch": 2.85, + "learning_rate": 2.7164429116886317e-06, + "loss": 0.5175, + "step": 7724 + }, + { + "epoch": 2.85, + "learning_rate": 2.7164059613252863e-06, + "loss": 0.5856, + "step": 7725 + }, + { + "epoch": 2.85, + "learning_rate": 2.7163690109619414e-06, + "loss": 0.5555, + "step": 7726 + }, + { + "epoch": 2.85, + "learning_rate": 2.716332060598596e-06, + "loss": 0.5186, + "step": 7727 + }, + { + "epoch": 2.85, + "learning_rate": 2.7162951102352507e-06, + "loss": 0.5854, + "step": 7728 + }, + { + "epoch": 2.85, + "learning_rate": 2.7162581598719053e-06, + "loss": 0.5788, + "step": 7729 + }, + { + "epoch": 2.85, + "learning_rate": 2.71622120950856e-06, + "loss": 0.5784, + "step": 7730 + }, + { + "epoch": 2.85, + "learning_rate": 2.716184259145215e-06, + "loss": 0.4696, + "step": 7731 + }, + { + "epoch": 2.86, + "learning_rate": 2.7161473087818697e-06, + "loss": 0.7346, + "step": 7732 + }, + { + "epoch": 2.86, + "learning_rate": 2.7161103584185247e-06, + "loss": 0.5499, + "step": 7733 + }, + { + "epoch": 2.86, + "learning_rate": 2.7160734080551793e-06, + "loss": 0.5569, + "step": 7734 + }, + { + "epoch": 2.86, + "learning_rate": 2.716036457691834e-06, + "loss": 0.6429, + "step": 7735 + }, + { + "epoch": 2.86, + "learning_rate": 2.7159995073284886e-06, + "loss": 0.4796, + "step": 7736 + }, + { + "epoch": 2.86, + "learning_rate": 2.7159625569651433e-06, + "loss": 0.6002, + "step": 7737 + }, + { + "epoch": 2.86, + "learning_rate": 2.7159256066017983e-06, + "loss": 0.6788, + "step": 7738 + }, + { + "epoch": 2.86, + "learning_rate": 2.715888656238453e-06, + "loss": 0.592, + "step": 7739 + }, + { + "epoch": 2.86, + "learning_rate": 2.715851705875108e-06, + "loss": 0.6136, + "step": 7740 + }, + { + "epoch": 2.86, + "learning_rate": 2.7158147555117627e-06, + "loss": 0.6709, + "step": 7741 + }, + { + "epoch": 2.86, + "learning_rate": 2.7157778051484173e-06, + "loss": 0.5904, + "step": 7742 + }, + { + "epoch": 2.86, + "learning_rate": 2.715740854785072e-06, + "loss": 0.5499, + "step": 7743 + }, + { + "epoch": 2.86, + "learning_rate": 2.715703904421727e-06, + "loss": 0.5471, + "step": 7744 + }, + { + "epoch": 2.86, + "learning_rate": 2.7156669540583816e-06, + "loss": 0.6417, + "step": 7745 + }, + { + "epoch": 2.86, + "learning_rate": 2.7156300036950367e-06, + "loss": 0.6333, + "step": 7746 + }, + { + "epoch": 2.86, + "learning_rate": 2.7155930533316913e-06, + "loss": 0.5237, + "step": 7747 + }, + { + "epoch": 2.86, + "learning_rate": 2.715556102968346e-06, + "loss": 0.5773, + "step": 7748 + }, + { + "epoch": 2.86, + "learning_rate": 2.7155191526050006e-06, + "loss": 0.6346, + "step": 7749 + }, + { + "epoch": 2.86, + "learning_rate": 2.7154822022416552e-06, + "loss": 0.6433, + "step": 7750 + }, + { + "epoch": 2.86, + "learning_rate": 2.7154452518783103e-06, + "loss": 0.6207, + "step": 7751 + }, + { + "epoch": 2.86, + "learning_rate": 2.715408301514965e-06, + "loss": 0.578, + "step": 7752 + }, + { + "epoch": 2.86, + "learning_rate": 2.71537135115162e-06, + "loss": 0.5748, + "step": 7753 + }, + { + "epoch": 2.86, + "learning_rate": 2.7153344007882746e-06, + "loss": 0.5126, + "step": 7754 + }, + { + "epoch": 2.86, + "learning_rate": 2.7152974504249293e-06, + "loss": 0.5313, + "step": 7755 + }, + { + "epoch": 2.86, + "learning_rate": 2.715260500061584e-06, + "loss": 0.6144, + "step": 7756 + }, + { + "epoch": 2.86, + "learning_rate": 2.7152235496982385e-06, + "loss": 0.6214, + "step": 7757 + }, + { + "epoch": 2.86, + "learning_rate": 2.7151865993348936e-06, + "loss": 0.5989, + "step": 7758 + }, + { + "epoch": 2.87, + "learning_rate": 2.7151496489715482e-06, + "loss": 0.5802, + "step": 7759 + }, + { + "epoch": 2.87, + "learning_rate": 2.7151126986082033e-06, + "loss": 0.634, + "step": 7760 + }, + { + "epoch": 2.87, + "learning_rate": 2.715075748244858e-06, + "loss": 0.6212, + "step": 7761 + }, + { + "epoch": 2.87, + "learning_rate": 2.7150387978815126e-06, + "loss": 0.5883, + "step": 7762 + }, + { + "epoch": 2.87, + "learning_rate": 2.715001847518167e-06, + "loss": 0.4866, + "step": 7763 + }, + { + "epoch": 2.87, + "learning_rate": 2.7149648971548223e-06, + "loss": 0.6616, + "step": 7764 + }, + { + "epoch": 2.87, + "learning_rate": 2.714927946791477e-06, + "loss": 0.5254, + "step": 7765 + }, + { + "epoch": 2.87, + "learning_rate": 2.714890996428132e-06, + "loss": 0.5659, + "step": 7766 + }, + { + "epoch": 2.87, + "learning_rate": 2.7148540460647866e-06, + "loss": 0.5475, + "step": 7767 + }, + { + "epoch": 2.87, + "learning_rate": 2.714817095701441e-06, + "loss": 0.5813, + "step": 7768 + }, + { + "epoch": 2.87, + "learning_rate": 2.714780145338096e-06, + "loss": 0.5502, + "step": 7769 + }, + { + "epoch": 2.87, + "learning_rate": 2.7147431949747505e-06, + "loss": 0.5984, + "step": 7770 + }, + { + "epoch": 2.87, + "learning_rate": 2.7147062446114056e-06, + "loss": 0.53, + "step": 7771 + }, + { + "epoch": 2.87, + "learning_rate": 2.71466929424806e-06, + "loss": 0.5718, + "step": 7772 + }, + { + "epoch": 2.87, + "learning_rate": 2.7146323438847153e-06, + "loss": 0.5973, + "step": 7773 + }, + { + "epoch": 2.87, + "learning_rate": 2.7145953935213695e-06, + "loss": 0.654, + "step": 7774 + }, + { + "epoch": 2.87, + "learning_rate": 2.7145584431580245e-06, + "loss": 0.5062, + "step": 7775 + }, + { + "epoch": 2.87, + "learning_rate": 2.714521492794679e-06, + "loss": 0.5617, + "step": 7776 + }, + { + "epoch": 2.87, + "learning_rate": 2.714484542431334e-06, + "loss": 0.5182, + "step": 7777 + }, + { + "epoch": 2.87, + "learning_rate": 2.714447592067989e-06, + "loss": 0.5551, + "step": 7778 + }, + { + "epoch": 2.87, + "learning_rate": 2.7144106417046435e-06, + "loss": 0.5616, + "step": 7779 + }, + { + "epoch": 2.87, + "learning_rate": 2.714373691341298e-06, + "loss": 0.6624, + "step": 7780 + }, + { + "epoch": 2.87, + "learning_rate": 2.7143367409779528e-06, + "loss": 0.5711, + "step": 7781 + }, + { + "epoch": 2.87, + "learning_rate": 2.714299790614608e-06, + "loss": 0.5469, + "step": 7782 + }, + { + "epoch": 2.87, + "learning_rate": 2.7142628402512625e-06, + "loss": 0.603, + "step": 7783 + }, + { + "epoch": 2.87, + "learning_rate": 2.7142258898879175e-06, + "loss": 0.5691, + "step": 7784 + }, + { + "epoch": 2.87, + "learning_rate": 2.714188939524572e-06, + "loss": 0.5734, + "step": 7785 + }, + { + "epoch": 2.88, + "learning_rate": 2.714151989161227e-06, + "loss": 0.6488, + "step": 7786 + }, + { + "epoch": 2.88, + "learning_rate": 2.7141150387978814e-06, + "loss": 0.5779, + "step": 7787 + }, + { + "epoch": 2.88, + "learning_rate": 2.714078088434536e-06, + "loss": 0.5181, + "step": 7788 + }, + { + "epoch": 2.88, + "learning_rate": 2.714041138071191e-06, + "loss": 0.5864, + "step": 7789 + }, + { + "epoch": 2.88, + "learning_rate": 2.7140041877078458e-06, + "loss": 0.5474, + "step": 7790 + }, + { + "epoch": 2.88, + "learning_rate": 2.713967237344501e-06, + "loss": 0.5647, + "step": 7791 + }, + { + "epoch": 2.88, + "learning_rate": 2.7139302869811555e-06, + "loss": 0.5917, + "step": 7792 + }, + { + "epoch": 2.88, + "learning_rate": 2.71389333661781e-06, + "loss": 0.5825, + "step": 7793 + }, + { + "epoch": 2.88, + "learning_rate": 2.7138563862544647e-06, + "loss": 0.5557, + "step": 7794 + }, + { + "epoch": 2.88, + "learning_rate": 2.71381943589112e-06, + "loss": 0.5846, + "step": 7795 + }, + { + "epoch": 2.88, + "learning_rate": 2.7137824855277744e-06, + "loss": 0.6214, + "step": 7796 + }, + { + "epoch": 2.88, + "learning_rate": 2.713745535164429e-06, + "loss": 0.4948, + "step": 7797 + }, + { + "epoch": 2.88, + "learning_rate": 2.713708584801084e-06, + "loss": 0.5947, + "step": 7798 + }, + { + "epoch": 2.88, + "learning_rate": 2.7136716344377388e-06, + "loss": 0.6303, + "step": 7799 + }, + { + "epoch": 2.88, + "learning_rate": 2.7136346840743934e-06, + "loss": 0.5791, + "step": 7800 + }, + { + "epoch": 2.88, + "learning_rate": 2.713597733711048e-06, + "loss": 0.5551, + "step": 7801 + }, + { + "epoch": 2.88, + "learning_rate": 2.713560783347703e-06, + "loss": 0.5436, + "step": 7802 + }, + { + "epoch": 2.88, + "learning_rate": 2.7135238329843577e-06, + "loss": 0.5918, + "step": 7803 + }, + { + "epoch": 2.88, + "learning_rate": 2.713486882621013e-06, + "loss": 0.6285, + "step": 7804 + }, + { + "epoch": 2.88, + "learning_rate": 2.7134499322576674e-06, + "loss": 0.5459, + "step": 7805 + }, + { + "epoch": 2.88, + "learning_rate": 2.713412981894322e-06, + "loss": 0.5363, + "step": 7806 + }, + { + "epoch": 2.88, + "learning_rate": 2.7133760315309767e-06, + "loss": 0.5822, + "step": 7807 + }, + { + "epoch": 2.88, + "learning_rate": 2.7133390811676313e-06, + "loss": 0.5022, + "step": 7808 + }, + { + "epoch": 2.88, + "learning_rate": 2.7133021308042864e-06, + "loss": 0.6503, + "step": 7809 + }, + { + "epoch": 2.88, + "learning_rate": 2.713265180440941e-06, + "loss": 0.6239, + "step": 7810 + }, + { + "epoch": 2.88, + "learning_rate": 2.713228230077596e-06, + "loss": 0.5895, + "step": 7811 + }, + { + "epoch": 2.88, + "learning_rate": 2.7131912797142503e-06, + "loss": 0.5679, + "step": 7812 + }, + { + "epoch": 2.89, + "learning_rate": 2.7131543293509054e-06, + "loss": 0.6244, + "step": 7813 + }, + { + "epoch": 2.89, + "learning_rate": 2.71311737898756e-06, + "loss": 0.6507, + "step": 7814 + }, + { + "epoch": 2.89, + "learning_rate": 2.713080428624215e-06, + "loss": 0.5102, + "step": 7815 + }, + { + "epoch": 2.89, + "learning_rate": 2.7130434782608697e-06, + "loss": 0.5393, + "step": 7816 + }, + { + "epoch": 2.89, + "learning_rate": 2.7130065278975243e-06, + "loss": 0.5751, + "step": 7817 + }, + { + "epoch": 2.89, + "learning_rate": 2.712969577534179e-06, + "loss": 0.5921, + "step": 7818 + }, + { + "epoch": 2.89, + "learning_rate": 2.7129326271708336e-06, + "loss": 0.5471, + "step": 7819 + }, + { + "epoch": 2.89, + "learning_rate": 2.7128956768074887e-06, + "loss": 0.5767, + "step": 7820 + }, + { + "epoch": 2.89, + "learning_rate": 2.7128587264441433e-06, + "loss": 0.5203, + "step": 7821 + }, + { + "epoch": 2.89, + "learning_rate": 2.7128217760807984e-06, + "loss": 0.5655, + "step": 7822 + }, + { + "epoch": 2.89, + "learning_rate": 2.712784825717453e-06, + "loss": 0.5808, + "step": 7823 + }, + { + "epoch": 2.89, + "learning_rate": 2.7127478753541076e-06, + "loss": 0.5018, + "step": 7824 + }, + { + "epoch": 2.89, + "learning_rate": 2.7127109249907623e-06, + "loss": 0.6338, + "step": 7825 + }, + { + "epoch": 2.89, + "learning_rate": 2.7126739746274173e-06, + "loss": 0.6002, + "step": 7826 + }, + { + "epoch": 2.89, + "learning_rate": 2.712637024264072e-06, + "loss": 0.7114, + "step": 7827 + }, + { + "epoch": 2.89, + "learning_rate": 2.7126000739007266e-06, + "loss": 0.5893, + "step": 7828 + }, + { + "epoch": 2.89, + "learning_rate": 2.7125631235373817e-06, + "loss": 0.5104, + "step": 7829 + }, + { + "epoch": 2.89, + "learning_rate": 2.7125261731740363e-06, + "loss": 0.5883, + "step": 7830 + }, + { + "epoch": 2.89, + "learning_rate": 2.712489222810691e-06, + "loss": 0.5794, + "step": 7831 + }, + { + "epoch": 2.89, + "learning_rate": 2.7124522724473456e-06, + "loss": 0.5446, + "step": 7832 + }, + { + "epoch": 2.89, + "learning_rate": 2.7124153220840006e-06, + "loss": 0.5397, + "step": 7833 + }, + { + "epoch": 2.89, + "learning_rate": 2.7123783717206553e-06, + "loss": 0.5876, + "step": 7834 + }, + { + "epoch": 2.89, + "learning_rate": 2.7123414213573103e-06, + "loss": 0.5454, + "step": 7835 + }, + { + "epoch": 2.89, + "learning_rate": 2.712304470993965e-06, + "loss": 0.6025, + "step": 7836 + }, + { + "epoch": 2.89, + "learning_rate": 2.7122675206306196e-06, + "loss": 0.6078, + "step": 7837 + }, + { + "epoch": 2.89, + "learning_rate": 2.7122305702672742e-06, + "loss": 0.5236, + "step": 7838 + }, + { + "epoch": 2.89, + "learning_rate": 2.712193619903929e-06, + "loss": 0.5484, + "step": 7839 + }, + { + "epoch": 2.9, + "learning_rate": 2.712156669540584e-06, + "loss": 0.5204, + "step": 7840 + }, + { + "epoch": 2.9, + "learning_rate": 2.7121197191772386e-06, + "loss": 0.6271, + "step": 7841 + }, + { + "epoch": 2.9, + "learning_rate": 2.7120827688138936e-06, + "loss": 0.6165, + "step": 7842 + }, + { + "epoch": 2.9, + "learning_rate": 2.7120458184505483e-06, + "loss": 0.61, + "step": 7843 + }, + { + "epoch": 2.9, + "learning_rate": 2.712008868087203e-06, + "loss": 0.64, + "step": 7844 + }, + { + "epoch": 2.9, + "learning_rate": 2.7119719177238575e-06, + "loss": 0.5712, + "step": 7845 + }, + { + "epoch": 2.9, + "learning_rate": 2.7119349673605126e-06, + "loss": 0.5532, + "step": 7846 + }, + { + "epoch": 2.9, + "learning_rate": 2.7118980169971672e-06, + "loss": 0.5299, + "step": 7847 + }, + { + "epoch": 2.9, + "learning_rate": 2.711861066633822e-06, + "loss": 0.518, + "step": 7848 + }, + { + "epoch": 2.9, + "learning_rate": 2.711824116270477e-06, + "loss": 0.6281, + "step": 7849 + }, + { + "epoch": 2.9, + "learning_rate": 2.7117871659071316e-06, + "loss": 0.5831, + "step": 7850 + }, + { + "epoch": 2.9, + "learning_rate": 2.711750215543786e-06, + "loss": 0.5925, + "step": 7851 + }, + { + "epoch": 2.9, + "learning_rate": 2.711713265180441e-06, + "loss": 0.5644, + "step": 7852 + }, + { + "epoch": 2.9, + "learning_rate": 2.711676314817096e-06, + "loss": 0.6939, + "step": 7853 + }, + { + "epoch": 2.9, + "learning_rate": 2.7116393644537505e-06, + "loss": 0.7147, + "step": 7854 + }, + { + "epoch": 2.9, + "learning_rate": 2.7116024140904056e-06, + "loss": 0.6375, + "step": 7855 + }, + { + "epoch": 2.9, + "learning_rate": 2.7115654637270602e-06, + "loss": 0.5585, + "step": 7856 + }, + { + "epoch": 2.9, + "learning_rate": 2.7115285133637145e-06, + "loss": 0.6138, + "step": 7857 + }, + { + "epoch": 2.9, + "learning_rate": 2.7114915630003695e-06, + "loss": 0.5939, + "step": 7858 + }, + { + "epoch": 2.9, + "learning_rate": 2.711454612637024e-06, + "loss": 0.5599, + "step": 7859 + }, + { + "epoch": 2.9, + "learning_rate": 2.711417662273679e-06, + "loss": 0.5571, + "step": 7860 + }, + { + "epoch": 2.9, + "learning_rate": 2.711380711910334e-06, + "loss": 0.5686, + "step": 7861 + }, + { + "epoch": 2.9, + "learning_rate": 2.711343761546989e-06, + "loss": 0.5827, + "step": 7862 + }, + { + "epoch": 2.9, + "learning_rate": 2.711306811183643e-06, + "loss": 0.6452, + "step": 7863 + }, + { + "epoch": 2.9, + "learning_rate": 2.711269860820298e-06, + "loss": 0.5201, + "step": 7864 + }, + { + "epoch": 2.9, + "learning_rate": 2.711232910456953e-06, + "loss": 0.5812, + "step": 7865 + }, + { + "epoch": 2.9, + "learning_rate": 2.711195960093608e-06, + "loss": 0.5647, + "step": 7866 + }, + { + "epoch": 2.91, + "learning_rate": 2.7111590097302625e-06, + "loss": 0.5421, + "step": 7867 + }, + { + "epoch": 2.91, + "learning_rate": 2.711122059366917e-06, + "loss": 0.6232, + "step": 7868 + }, + { + "epoch": 2.91, + "learning_rate": 2.7110851090035718e-06, + "loss": 0.5277, + "step": 7869 + }, + { + "epoch": 2.91, + "learning_rate": 2.7110481586402264e-06, + "loss": 0.4696, + "step": 7870 + }, + { + "epoch": 2.91, + "learning_rate": 2.7110112082768815e-06, + "loss": 0.5968, + "step": 7871 + }, + { + "epoch": 2.91, + "learning_rate": 2.710974257913536e-06, + "loss": 0.6079, + "step": 7872 + }, + { + "epoch": 2.91, + "learning_rate": 2.710937307550191e-06, + "loss": 0.4916, + "step": 7873 + }, + { + "epoch": 2.91, + "learning_rate": 2.710900357186846e-06, + "loss": 0.5328, + "step": 7874 + }, + { + "epoch": 2.91, + "learning_rate": 2.7108634068235004e-06, + "loss": 0.5208, + "step": 7875 + }, + { + "epoch": 2.91, + "learning_rate": 2.710826456460155e-06, + "loss": 0.5476, + "step": 7876 + }, + { + "epoch": 2.91, + "learning_rate": 2.7107895060968097e-06, + "loss": 0.6214, + "step": 7877 + }, + { + "epoch": 2.91, + "learning_rate": 2.7107525557334648e-06, + "loss": 0.5181, + "step": 7878 + }, + { + "epoch": 2.91, + "learning_rate": 2.7107156053701194e-06, + "loss": 0.5352, + "step": 7879 + }, + { + "epoch": 2.91, + "learning_rate": 2.7106786550067745e-06, + "loss": 0.5775, + "step": 7880 + }, + { + "epoch": 2.91, + "learning_rate": 2.710641704643429e-06, + "loss": 0.5856, + "step": 7881 + }, + { + "epoch": 2.91, + "learning_rate": 2.7106047542800838e-06, + "loss": 0.4832, + "step": 7882 + }, + { + "epoch": 2.91, + "learning_rate": 2.7105678039167384e-06, + "loss": 0.586, + "step": 7883 + }, + { + "epoch": 2.91, + "learning_rate": 2.7105308535533934e-06, + "loss": 0.5472, + "step": 7884 + }, + { + "epoch": 2.91, + "learning_rate": 2.710493903190048e-06, + "loss": 0.5135, + "step": 7885 + }, + { + "epoch": 2.91, + "learning_rate": 2.710456952826703e-06, + "loss": 0.512, + "step": 7886 + }, + { + "epoch": 2.91, + "learning_rate": 2.7104200024633578e-06, + "loss": 0.5919, + "step": 7887 + }, + { + "epoch": 2.91, + "learning_rate": 2.7103830521000124e-06, + "loss": 0.4807, + "step": 7888 + }, + { + "epoch": 2.91, + "learning_rate": 2.710346101736667e-06, + "loss": 0.5452, + "step": 7889 + }, + { + "epoch": 2.91, + "learning_rate": 2.7103091513733217e-06, + "loss": 0.5722, + "step": 7890 + }, + { + "epoch": 2.91, + "learning_rate": 2.7102722010099768e-06, + "loss": 0.5736, + "step": 7891 + }, + { + "epoch": 2.91, + "learning_rate": 2.7102352506466314e-06, + "loss": 0.6063, + "step": 7892 + }, + { + "epoch": 2.91, + "learning_rate": 2.7101983002832864e-06, + "loss": 0.5535, + "step": 7893 + }, + { + "epoch": 2.92, + "learning_rate": 2.710161349919941e-06, + "loss": 0.5153, + "step": 7894 + }, + { + "epoch": 2.92, + "learning_rate": 2.7101243995565957e-06, + "loss": 0.5888, + "step": 7895 + }, + { + "epoch": 2.92, + "learning_rate": 2.7100874491932504e-06, + "loss": 0.6264, + "step": 7896 + }, + { + "epoch": 2.92, + "learning_rate": 2.710050498829905e-06, + "loss": 0.5828, + "step": 7897 + }, + { + "epoch": 2.92, + "learning_rate": 2.71001354846656e-06, + "loss": 0.6003, + "step": 7898 + }, + { + "epoch": 2.92, + "learning_rate": 2.7099765981032147e-06, + "loss": 0.4906, + "step": 7899 + }, + { + "epoch": 2.92, + "learning_rate": 2.7099396477398697e-06, + "loss": 0.5354, + "step": 7900 + }, + { + "epoch": 2.92, + "learning_rate": 2.7099026973765244e-06, + "loss": 0.7299, + "step": 7901 + }, + { + "epoch": 2.92, + "learning_rate": 2.709865747013179e-06, + "loss": 0.6488, + "step": 7902 + }, + { + "epoch": 2.92, + "learning_rate": 2.7098287966498337e-06, + "loss": 0.6196, + "step": 7903 + }, + { + "epoch": 2.92, + "learning_rate": 2.7097918462864887e-06, + "loss": 0.598, + "step": 7904 + }, + { + "epoch": 2.92, + "learning_rate": 2.7097548959231434e-06, + "loss": 0.5626, + "step": 7905 + }, + { + "epoch": 2.92, + "learning_rate": 2.7097179455597984e-06, + "loss": 0.6721, + "step": 7906 + }, + { + "epoch": 2.92, + "learning_rate": 2.709680995196453e-06, + "loss": 0.5993, + "step": 7907 + }, + { + "epoch": 2.92, + "learning_rate": 2.7096440448331073e-06, + "loss": 0.5333, + "step": 7908 + }, + { + "epoch": 2.92, + "learning_rate": 2.7096070944697623e-06, + "loss": 0.5547, + "step": 7909 + }, + { + "epoch": 2.92, + "learning_rate": 2.709570144106417e-06, + "loss": 0.6105, + "step": 7910 + }, + { + "epoch": 2.92, + "learning_rate": 2.709533193743072e-06, + "loss": 0.6728, + "step": 7911 + }, + { + "epoch": 2.92, + "learning_rate": 2.7094962433797267e-06, + "loss": 0.5493, + "step": 7912 + }, + { + "epoch": 2.92, + "learning_rate": 2.7094592930163817e-06, + "loss": 0.5277, + "step": 7913 + }, + { + "epoch": 2.92, + "learning_rate": 2.709422342653036e-06, + "loss": 0.5032, + "step": 7914 + }, + { + "epoch": 2.92, + "learning_rate": 2.709385392289691e-06, + "loss": 0.6829, + "step": 7915 + }, + { + "epoch": 2.92, + "learning_rate": 2.7093484419263456e-06, + "loss": 0.5028, + "step": 7916 + }, + { + "epoch": 2.92, + "learning_rate": 2.7093114915630003e-06, + "loss": 0.61, + "step": 7917 + }, + { + "epoch": 2.92, + "learning_rate": 2.7092745411996553e-06, + "loss": 0.5633, + "step": 7918 + }, + { + "epoch": 2.92, + "learning_rate": 2.70923759083631e-06, + "loss": 0.5393, + "step": 7919 + }, + { + "epoch": 2.92, + "learning_rate": 2.7092006404729646e-06, + "loss": 0.6815, + "step": 7920 + }, + { + "epoch": 2.93, + "learning_rate": 2.7091636901096192e-06, + "loss": 0.5755, + "step": 7921 + }, + { + "epoch": 2.93, + "learning_rate": 2.7091267397462743e-06, + "loss": 0.5974, + "step": 7922 + }, + { + "epoch": 2.93, + "learning_rate": 2.709089789382929e-06, + "loss": 0.7051, + "step": 7923 + }, + { + "epoch": 2.93, + "learning_rate": 2.709052839019584e-06, + "loss": 0.548, + "step": 7924 + }, + { + "epoch": 2.93, + "learning_rate": 2.7090158886562386e-06, + "loss": 0.6005, + "step": 7925 + }, + { + "epoch": 2.93, + "learning_rate": 2.7089789382928933e-06, + "loss": 0.5421, + "step": 7926 + }, + { + "epoch": 2.93, + "learning_rate": 2.708941987929548e-06, + "loss": 0.5673, + "step": 7927 + }, + { + "epoch": 2.93, + "learning_rate": 2.7089050375662025e-06, + "loss": 0.5591, + "step": 7928 + }, + { + "epoch": 2.93, + "learning_rate": 2.7088680872028576e-06, + "loss": 0.5697, + "step": 7929 + }, + { + "epoch": 2.93, + "learning_rate": 2.7088311368395122e-06, + "loss": 0.5807, + "step": 7930 + }, + { + "epoch": 2.93, + "learning_rate": 2.7087941864761673e-06, + "loss": 0.584, + "step": 7931 + }, + { + "epoch": 2.93, + "learning_rate": 2.708757236112822e-06, + "loss": 0.5059, + "step": 7932 + }, + { + "epoch": 2.93, + "learning_rate": 2.7087202857494766e-06, + "loss": 0.5241, + "step": 7933 + }, + { + "epoch": 2.93, + "learning_rate": 2.708683335386131e-06, + "loss": 0.5579, + "step": 7934 + }, + { + "epoch": 2.93, + "learning_rate": 2.7086463850227863e-06, + "loss": 0.5698, + "step": 7935 + }, + { + "epoch": 2.93, + "learning_rate": 2.708609434659441e-06, + "loss": 0.5089, + "step": 7936 + }, + { + "epoch": 2.93, + "learning_rate": 2.7085724842960955e-06, + "loss": 0.5721, + "step": 7937 + }, + { + "epoch": 2.93, + "learning_rate": 2.7085355339327506e-06, + "loss": 0.5719, + "step": 7938 + }, + { + "epoch": 2.93, + "learning_rate": 2.7084985835694052e-06, + "loss": 0.5876, + "step": 7939 + }, + { + "epoch": 2.93, + "learning_rate": 2.70846163320606e-06, + "loss": 0.5636, + "step": 7940 + }, + { + "epoch": 2.93, + "learning_rate": 2.7084246828427145e-06, + "loss": 0.581, + "step": 7941 + }, + { + "epoch": 2.93, + "learning_rate": 2.7083877324793696e-06, + "loss": 0.5062, + "step": 7942 + }, + { + "epoch": 2.93, + "learning_rate": 2.708350782116024e-06, + "loss": 0.5495, + "step": 7943 + }, + { + "epoch": 2.93, + "learning_rate": 2.7083138317526793e-06, + "loss": 0.6218, + "step": 7944 + }, + { + "epoch": 2.93, + "learning_rate": 2.708276881389334e-06, + "loss": 0.4971, + "step": 7945 + }, + { + "epoch": 2.93, + "learning_rate": 2.7082399310259885e-06, + "loss": 0.4987, + "step": 7946 + }, + { + "epoch": 2.93, + "learning_rate": 2.708202980662643e-06, + "loss": 0.6033, + "step": 7947 + }, + { + "epoch": 2.94, + "learning_rate": 2.708166030299298e-06, + "loss": 0.6197, + "step": 7948 + }, + { + "epoch": 2.94, + "learning_rate": 2.708129079935953e-06, + "loss": 0.6597, + "step": 7949 + }, + { + "epoch": 2.94, + "learning_rate": 2.7080921295726075e-06, + "loss": 0.553, + "step": 7950 + }, + { + "epoch": 2.94, + "learning_rate": 2.7080551792092626e-06, + "loss": 0.6029, + "step": 7951 + }, + { + "epoch": 2.94, + "learning_rate": 2.708018228845917e-06, + "loss": 0.5591, + "step": 7952 + }, + { + "epoch": 2.94, + "learning_rate": 2.707981278482572e-06, + "loss": 0.6495, + "step": 7953 + }, + { + "epoch": 2.94, + "learning_rate": 2.7079443281192265e-06, + "loss": 0.6026, + "step": 7954 + }, + { + "epoch": 2.94, + "learning_rate": 2.7079073777558815e-06, + "loss": 0.6123, + "step": 7955 + }, + { + "epoch": 2.94, + "learning_rate": 2.707870427392536e-06, + "loss": 0.5666, + "step": 7956 + }, + { + "epoch": 2.94, + "learning_rate": 2.707833477029191e-06, + "loss": 0.5399, + "step": 7957 + }, + { + "epoch": 2.94, + "learning_rate": 2.707796526665846e-06, + "loss": 0.6044, + "step": 7958 + }, + { + "epoch": 2.94, + "learning_rate": 2.7077595763025e-06, + "loss": 0.5042, + "step": 7959 + }, + { + "epoch": 2.94, + "learning_rate": 2.707722625939155e-06, + "loss": 0.5048, + "step": 7960 + }, + { + "epoch": 2.94, + "learning_rate": 2.7076856755758098e-06, + "loss": 0.63, + "step": 7961 + }, + { + "epoch": 2.94, + "learning_rate": 2.707648725212465e-06, + "loss": 0.4926, + "step": 7962 + }, + { + "epoch": 2.94, + "learning_rate": 2.7076117748491195e-06, + "loss": 0.5589, + "step": 7963 + }, + { + "epoch": 2.94, + "learning_rate": 2.707574824485774e-06, + "loss": 0.4678, + "step": 7964 + }, + { + "epoch": 2.94, + "learning_rate": 2.7075378741224287e-06, + "loss": 0.5712, + "step": 7965 + }, + { + "epoch": 2.94, + "learning_rate": 2.7075009237590834e-06, + "loss": 0.5763, + "step": 7966 + }, + { + "epoch": 2.94, + "learning_rate": 2.7074639733957384e-06, + "loss": 0.6103, + "step": 7967 + }, + { + "epoch": 2.94, + "learning_rate": 2.707427023032393e-06, + "loss": 0.5646, + "step": 7968 + }, + { + "epoch": 2.94, + "learning_rate": 2.707390072669048e-06, + "loss": 0.5253, + "step": 7969 + }, + { + "epoch": 2.94, + "learning_rate": 2.7073531223057028e-06, + "loss": 0.6661, + "step": 7970 + }, + { + "epoch": 2.94, + "learning_rate": 2.7073161719423574e-06, + "loss": 0.5537, + "step": 7971 + }, + { + "epoch": 2.94, + "learning_rate": 2.707279221579012e-06, + "loss": 0.578, + "step": 7972 + }, + { + "epoch": 2.94, + "learning_rate": 2.707242271215667e-06, + "loss": 0.5873, + "step": 7973 + }, + { + "epoch": 2.94, + "learning_rate": 2.7072053208523217e-06, + "loss": 0.5235, + "step": 7974 + }, + { + "epoch": 2.94, + "learning_rate": 2.707168370488977e-06, + "loss": 0.6484, + "step": 7975 + }, + { + "epoch": 2.95, + "learning_rate": 2.7071314201256314e-06, + "loss": 0.6572, + "step": 7976 + }, + { + "epoch": 2.95, + "learning_rate": 2.707094469762286e-06, + "loss": 0.6511, + "step": 7977 + }, + { + "epoch": 2.95, + "learning_rate": 2.7070575193989407e-06, + "loss": 0.5272, + "step": 7978 + }, + { + "epoch": 2.95, + "learning_rate": 2.7070205690355953e-06, + "loss": 0.5082, + "step": 7979 + }, + { + "epoch": 2.95, + "learning_rate": 2.7069836186722504e-06, + "loss": 0.565, + "step": 7980 + }, + { + "epoch": 2.95, + "learning_rate": 2.706946668308905e-06, + "loss": 0.4338, + "step": 7981 + }, + { + "epoch": 2.95, + "learning_rate": 2.70690971794556e-06, + "loss": 0.563, + "step": 7982 + }, + { + "epoch": 2.95, + "learning_rate": 2.7068727675822147e-06, + "loss": 0.5197, + "step": 7983 + }, + { + "epoch": 2.95, + "learning_rate": 2.7068358172188694e-06, + "loss": 0.5745, + "step": 7984 + }, + { + "epoch": 2.95, + "learning_rate": 2.706798866855524e-06, + "loss": 0.5684, + "step": 7985 + }, + { + "epoch": 2.95, + "learning_rate": 2.7067619164921786e-06, + "loss": 0.6517, + "step": 7986 + }, + { + "epoch": 2.95, + "learning_rate": 2.7067249661288337e-06, + "loss": 0.593, + "step": 7987 + }, + { + "epoch": 2.95, + "learning_rate": 2.7066880157654883e-06, + "loss": 0.4944, + "step": 7988 + }, + { + "epoch": 2.95, + "learning_rate": 2.7066510654021434e-06, + "loss": 0.6571, + "step": 7989 + }, + { + "epoch": 2.95, + "learning_rate": 2.706614115038798e-06, + "loss": 0.5691, + "step": 7990 + }, + { + "epoch": 2.95, + "learning_rate": 2.7065771646754527e-06, + "loss": 0.5275, + "step": 7991 + }, + { + "epoch": 2.95, + "learning_rate": 2.7065402143121073e-06, + "loss": 0.6407, + "step": 7992 + }, + { + "epoch": 2.95, + "learning_rate": 2.7065032639487624e-06, + "loss": 0.6164, + "step": 7993 + }, + { + "epoch": 2.95, + "learning_rate": 2.706466313585417e-06, + "loss": 0.564, + "step": 7994 + }, + { + "epoch": 2.95, + "learning_rate": 2.706429363222072e-06, + "loss": 0.5785, + "step": 7995 + }, + { + "epoch": 2.95, + "learning_rate": 2.7063924128587267e-06, + "loss": 0.5315, + "step": 7996 + }, + { + "epoch": 2.95, + "learning_rate": 2.706355462495381e-06, + "loss": 0.6872, + "step": 7997 + }, + { + "epoch": 2.95, + "learning_rate": 2.706318512132036e-06, + "loss": 0.6058, + "step": 7998 + }, + { + "epoch": 2.95, + "learning_rate": 2.7062815617686906e-06, + "loss": 0.5547, + "step": 7999 + }, + { + "epoch": 2.95, + "learning_rate": 2.7062446114053457e-06, + "loss": 0.6574, + "step": 8000 + }, + { + "epoch": 2.95, + "learning_rate": 2.7062076610420003e-06, + "loss": 0.6462, + "step": 8001 + }, + { + "epoch": 2.95, + "learning_rate": 2.7061707106786554e-06, + "loss": 0.5808, + "step": 8002 + }, + { + "epoch": 2.96, + "learning_rate": 2.7061337603153096e-06, + "loss": 0.6072, + "step": 8003 + }, + { + "epoch": 2.96, + "learning_rate": 2.7060968099519646e-06, + "loss": 0.5119, + "step": 8004 + }, + { + "epoch": 2.96, + "learning_rate": 2.7060598595886193e-06, + "loss": 0.5685, + "step": 8005 + }, + { + "epoch": 2.96, + "learning_rate": 2.706022909225274e-06, + "loss": 0.673, + "step": 8006 + }, + { + "epoch": 2.96, + "learning_rate": 2.705985958861929e-06, + "loss": 0.6209, + "step": 8007 + }, + { + "epoch": 2.96, + "learning_rate": 2.7059490084985836e-06, + "loss": 0.6781, + "step": 8008 + }, + { + "epoch": 2.96, + "learning_rate": 2.7059120581352382e-06, + "loss": 0.6805, + "step": 8009 + }, + { + "epoch": 2.96, + "learning_rate": 2.705875107771893e-06, + "loss": 0.5946, + "step": 8010 + }, + { + "epoch": 2.96, + "learning_rate": 2.705838157408548e-06, + "loss": 0.6146, + "step": 8011 + }, + { + "epoch": 2.96, + "learning_rate": 2.7058012070452026e-06, + "loss": 0.5579, + "step": 8012 + }, + { + "epoch": 2.96, + "learning_rate": 2.7057642566818576e-06, + "loss": 0.6735, + "step": 8013 + }, + { + "epoch": 2.96, + "learning_rate": 2.7057273063185123e-06, + "loss": 0.6592, + "step": 8014 + }, + { + "epoch": 2.96, + "learning_rate": 2.705690355955167e-06, + "loss": 0.546, + "step": 8015 + }, + { + "epoch": 2.96, + "learning_rate": 2.7056534055918216e-06, + "loss": 0.6232, + "step": 8016 + }, + { + "epoch": 2.96, + "learning_rate": 2.705616455228476e-06, + "loss": 0.5701, + "step": 8017 + }, + { + "epoch": 2.96, + "learning_rate": 2.7055795048651312e-06, + "loss": 0.5679, + "step": 8018 + }, + { + "epoch": 2.96, + "learning_rate": 2.705542554501786e-06, + "loss": 0.578, + "step": 8019 + }, + { + "epoch": 2.96, + "learning_rate": 2.705505604138441e-06, + "loss": 0.5268, + "step": 8020 + }, + { + "epoch": 2.96, + "learning_rate": 2.7054686537750956e-06, + "loss": 0.5136, + "step": 8021 + }, + { + "epoch": 2.96, + "learning_rate": 2.7054317034117502e-06, + "loss": 0.6692, + "step": 8022 + }, + { + "epoch": 2.96, + "learning_rate": 2.705394753048405e-06, + "loss": 0.5883, + "step": 8023 + }, + { + "epoch": 2.96, + "learning_rate": 2.70535780268506e-06, + "loss": 0.5155, + "step": 8024 + }, + { + "epoch": 2.96, + "learning_rate": 2.7053208523217145e-06, + "loss": 0.5745, + "step": 8025 + }, + { + "epoch": 2.96, + "learning_rate": 2.705283901958369e-06, + "loss": 0.5413, + "step": 8026 + }, + { + "epoch": 2.96, + "learning_rate": 2.7052469515950242e-06, + "loss": 0.5814, + "step": 8027 + }, + { + "epoch": 2.96, + "learning_rate": 2.705210001231679e-06, + "loss": 0.6048, + "step": 8028 + }, + { + "epoch": 2.96, + "learning_rate": 2.7051730508683335e-06, + "loss": 0.515, + "step": 8029 + }, + { + "epoch": 2.97, + "learning_rate": 2.705136100504988e-06, + "loss": 0.6272, + "step": 8030 + }, + { + "epoch": 2.97, + "learning_rate": 2.7050991501416432e-06, + "loss": 0.5228, + "step": 8031 + }, + { + "epoch": 2.97, + "learning_rate": 2.705062199778298e-06, + "loss": 0.6861, + "step": 8032 + }, + { + "epoch": 2.97, + "learning_rate": 2.705025249414953e-06, + "loss": 0.6745, + "step": 8033 + }, + { + "epoch": 2.97, + "learning_rate": 2.7049882990516075e-06, + "loss": 0.6334, + "step": 8034 + }, + { + "epoch": 2.97, + "learning_rate": 2.704951348688262e-06, + "loss": 0.6105, + "step": 8035 + }, + { + "epoch": 2.97, + "learning_rate": 2.704914398324917e-06, + "loss": 0.5858, + "step": 8036 + }, + { + "epoch": 2.97, + "learning_rate": 2.7048774479615715e-06, + "loss": 0.5911, + "step": 8037 + }, + { + "epoch": 2.97, + "learning_rate": 2.7048404975982265e-06, + "loss": 0.6032, + "step": 8038 + }, + { + "epoch": 2.97, + "learning_rate": 2.704803547234881e-06, + "loss": 0.5474, + "step": 8039 + }, + { + "epoch": 2.97, + "learning_rate": 2.7047665968715362e-06, + "loss": 0.5806, + "step": 8040 + }, + { + "epoch": 2.97, + "learning_rate": 2.704729646508191e-06, + "loss": 0.606, + "step": 8041 + }, + { + "epoch": 2.97, + "learning_rate": 2.7046926961448455e-06, + "loss": 0.5318, + "step": 8042 + }, + { + "epoch": 2.97, + "learning_rate": 2.7046557457815e-06, + "loss": 0.6063, + "step": 8043 + }, + { + "epoch": 2.97, + "learning_rate": 2.704618795418155e-06, + "loss": 0.6317, + "step": 8044 + }, + { + "epoch": 2.97, + "learning_rate": 2.70458184505481e-06, + "loss": 0.5517, + "step": 8045 + }, + { + "epoch": 2.97, + "learning_rate": 2.7045448946914645e-06, + "loss": 0.641, + "step": 8046 + }, + { + "epoch": 2.97, + "learning_rate": 2.7045079443281195e-06, + "loss": 0.5622, + "step": 8047 + }, + { + "epoch": 2.97, + "learning_rate": 2.7044709939647737e-06, + "loss": 0.5922, + "step": 8048 + }, + { + "epoch": 2.97, + "learning_rate": 2.704434043601429e-06, + "loss": 0.5277, + "step": 8049 + }, + { + "epoch": 2.97, + "learning_rate": 2.7043970932380834e-06, + "loss": 0.6828, + "step": 8050 + }, + { + "epoch": 2.97, + "learning_rate": 2.7043601428747385e-06, + "loss": 0.6462, + "step": 8051 + }, + { + "epoch": 2.97, + "learning_rate": 2.704323192511393e-06, + "loss": 0.6091, + "step": 8052 + }, + { + "epoch": 2.97, + "learning_rate": 2.704286242148048e-06, + "loss": 0.5043, + "step": 8053 + }, + { + "epoch": 2.97, + "learning_rate": 2.7042492917847024e-06, + "loss": 0.5515, + "step": 8054 + }, + { + "epoch": 2.97, + "learning_rate": 2.7042123414213575e-06, + "loss": 0.559, + "step": 8055 + }, + { + "epoch": 2.97, + "learning_rate": 2.704175391058012e-06, + "loss": 0.5477, + "step": 8056 + }, + { + "epoch": 2.98, + "learning_rate": 2.7041384406946667e-06, + "loss": 0.6331, + "step": 8057 + }, + { + "epoch": 2.98, + "learning_rate": 2.7041014903313218e-06, + "loss": 0.6353, + "step": 8058 + }, + { + "epoch": 2.98, + "learning_rate": 2.7040645399679764e-06, + "loss": 0.4931, + "step": 8059 + }, + { + "epoch": 2.98, + "learning_rate": 2.704027589604631e-06, + "loss": 0.6005, + "step": 8060 + }, + { + "epoch": 2.98, + "learning_rate": 2.7039906392412857e-06, + "loss": 0.5966, + "step": 8061 + }, + { + "epoch": 2.98, + "learning_rate": 2.7039536888779408e-06, + "loss": 0.5911, + "step": 8062 + }, + { + "epoch": 2.98, + "learning_rate": 2.7039167385145954e-06, + "loss": 0.5379, + "step": 8063 + }, + { + "epoch": 2.98, + "learning_rate": 2.7038797881512505e-06, + "loss": 0.6121, + "step": 8064 + }, + { + "epoch": 2.98, + "learning_rate": 2.703842837787905e-06, + "loss": 0.5543, + "step": 8065 + }, + { + "epoch": 2.98, + "learning_rate": 2.7038058874245597e-06, + "loss": 0.5774, + "step": 8066 + }, + { + "epoch": 2.98, + "learning_rate": 2.7037689370612144e-06, + "loss": 0.5779, + "step": 8067 + }, + { + "epoch": 2.98, + "learning_rate": 2.703731986697869e-06, + "loss": 0.5705, + "step": 8068 + }, + { + "epoch": 2.98, + "learning_rate": 2.703695036334524e-06, + "loss": 0.634, + "step": 8069 + }, + { + "epoch": 2.98, + "learning_rate": 2.7036580859711787e-06, + "loss": 0.6038, + "step": 8070 + }, + { + "epoch": 2.98, + "learning_rate": 2.7036211356078338e-06, + "loss": 0.5298, + "step": 8071 + }, + { + "epoch": 2.98, + "learning_rate": 2.7035841852444884e-06, + "loss": 0.5863, + "step": 8072 + }, + { + "epoch": 2.98, + "learning_rate": 2.703547234881143e-06, + "loss": 0.6278, + "step": 8073 + }, + { + "epoch": 2.98, + "learning_rate": 2.7035102845177977e-06, + "loss": 0.7061, + "step": 8074 + }, + { + "epoch": 2.98, + "learning_rate": 2.7034733341544527e-06, + "loss": 0.5572, + "step": 8075 + }, + { + "epoch": 2.98, + "learning_rate": 2.7034363837911074e-06, + "loss": 0.5805, + "step": 8076 + }, + { + "epoch": 2.98, + "learning_rate": 2.703399433427762e-06, + "loss": 0.5248, + "step": 8077 + }, + { + "epoch": 2.98, + "learning_rate": 2.703362483064417e-06, + "loss": 0.5684, + "step": 8078 + }, + { + "epoch": 2.98, + "learning_rate": 2.7033255327010717e-06, + "loss": 0.5394, + "step": 8079 + }, + { + "epoch": 2.98, + "learning_rate": 2.7032885823377263e-06, + "loss": 0.605, + "step": 8080 + }, + { + "epoch": 2.98, + "learning_rate": 2.703251631974381e-06, + "loss": 0.5487, + "step": 8081 + }, + { + "epoch": 2.98, + "learning_rate": 2.703214681611036e-06, + "loss": 0.5512, + "step": 8082 + }, + { + "epoch": 2.98, + "learning_rate": 2.7031777312476907e-06, + "loss": 0.5491, + "step": 8083 + }, + { + "epoch": 2.99, + "learning_rate": 2.7031407808843457e-06, + "loss": 0.5816, + "step": 8084 + }, + { + "epoch": 2.99, + "learning_rate": 2.7031038305210004e-06, + "loss": 0.5858, + "step": 8085 + }, + { + "epoch": 2.99, + "learning_rate": 2.703066880157655e-06, + "loss": 0.5347, + "step": 8086 + }, + { + "epoch": 2.99, + "learning_rate": 2.7030299297943096e-06, + "loss": 0.6166, + "step": 8087 + }, + { + "epoch": 2.99, + "learning_rate": 2.7029929794309643e-06, + "loss": 0.6214, + "step": 8088 + }, + { + "epoch": 2.99, + "learning_rate": 2.7029560290676193e-06, + "loss": 0.5563, + "step": 8089 + }, + { + "epoch": 2.99, + "learning_rate": 2.702919078704274e-06, + "loss": 0.5994, + "step": 8090 + }, + { + "epoch": 2.99, + "learning_rate": 2.702882128340929e-06, + "loss": 0.5988, + "step": 8091 + }, + { + "epoch": 2.99, + "learning_rate": 2.7028451779775837e-06, + "loss": 0.5684, + "step": 8092 + }, + { + "epoch": 2.99, + "learning_rate": 2.7028082276142383e-06, + "loss": 0.554, + "step": 8093 + }, + { + "epoch": 2.99, + "learning_rate": 2.702771277250893e-06, + "loss": 0.6096, + "step": 8094 + }, + { + "epoch": 2.99, + "learning_rate": 2.702734326887548e-06, + "loss": 0.5689, + "step": 8095 + }, + { + "epoch": 2.99, + "learning_rate": 2.7026973765242026e-06, + "loss": 0.5603, + "step": 8096 + }, + { + "epoch": 2.99, + "learning_rate": 2.7026604261608573e-06, + "loss": 0.5402, + "step": 8097 + }, + { + "epoch": 2.99, + "learning_rate": 2.7026234757975123e-06, + "loss": 0.5669, + "step": 8098 + }, + { + "epoch": 2.99, + "learning_rate": 2.7025865254341665e-06, + "loss": 0.6016, + "step": 8099 + }, + { + "epoch": 2.99, + "learning_rate": 2.7025495750708216e-06, + "loss": 0.6279, + "step": 8100 + }, + { + "epoch": 2.99, + "learning_rate": 2.7025126247074762e-06, + "loss": 0.551, + "step": 8101 + }, + { + "epoch": 2.99, + "learning_rate": 2.7024756743441313e-06, + "loss": 0.6063, + "step": 8102 + }, + { + "epoch": 2.99, + "learning_rate": 2.702438723980786e-06, + "loss": 0.5313, + "step": 8103 + }, + { + "epoch": 2.99, + "learning_rate": 2.702401773617441e-06, + "loss": 0.6554, + "step": 8104 + }, + { + "epoch": 2.99, + "learning_rate": 2.702364823254095e-06, + "loss": 0.5996, + "step": 8105 + }, + { + "epoch": 2.99, + "learning_rate": 2.70232787289075e-06, + "loss": 0.5484, + "step": 8106 + }, + { + "epoch": 2.99, + "learning_rate": 2.702290922527405e-06, + "loss": 0.5571, + "step": 8107 + }, + { + "epoch": 2.99, + "learning_rate": 2.7022539721640595e-06, + "loss": 0.5728, + "step": 8108 + }, + { + "epoch": 2.99, + "learning_rate": 2.7022170218007146e-06, + "loss": 0.5805, + "step": 8109 + }, + { + "epoch": 2.99, + "learning_rate": 2.7021800714373692e-06, + "loss": 0.5774, + "step": 8110 + }, + { + "epoch": 3.0, + "learning_rate": 2.702143121074024e-06, + "loss": 0.5275, + "step": 8111 + }, + { + "epoch": 3.0, + "learning_rate": 2.7021061707106785e-06, + "loss": 0.5815, + "step": 8112 + }, + { + "epoch": 3.0, + "learning_rate": 2.7020692203473336e-06, + "loss": 0.6059, + "step": 8113 + }, + { + "epoch": 3.0, + "learning_rate": 2.702032269983988e-06, + "loss": 0.6077, + "step": 8114 + }, + { + "epoch": 3.0, + "learning_rate": 2.7019953196206433e-06, + "loss": 0.61, + "step": 8115 + }, + { + "epoch": 3.0, + "learning_rate": 2.701958369257298e-06, + "loss": 0.5737, + "step": 8116 + }, + { + "epoch": 3.0, + "learning_rate": 2.7019214188939525e-06, + "loss": 0.5411, + "step": 8117 + }, + { + "epoch": 3.0, + "learning_rate": 2.701884468530607e-06, + "loss": 0.4931, + "step": 8118 + }, + { + "epoch": 3.0, + "learning_rate": 2.701847518167262e-06, + "loss": 0.5329, + "step": 8119 + }, + { + "epoch": 3.0, + "learning_rate": 2.701810567803917e-06, + "loss": 0.6033, + "step": 8120 + }, + { + "epoch": 3.0, + "learning_rate": 2.7017736174405715e-06, + "loss": 0.6679, + "step": 8121 + }, + { + "epoch": 3.0, + "learning_rate": 2.7017366670772266e-06, + "loss": 0.6119, + "step": 8122 + }, + { + "epoch": 3.0, + "learning_rate": 2.701699716713881e-06, + "loss": 0.593, + "step": 8123 + }, + { + "epoch": 3.0, + "learning_rate": 2.701662766350536e-06, + "loss": 0.3855, + "step": 8124 + }, + { + "epoch": 3.0, + "eval_accuracy": 0.6937941543150021, + "eval_loss": 0.5787221193313599, + "eval_runtime": 25.2063, + "eval_samples_per_second": 1718.382, + "eval_steps_per_second": 26.858, + "step": 8124 + }, + { + "epoch": 3.0, + "learning_rate": 2.7016258159871905e-06, + "loss": 0.6075, + "step": 8125 + }, + { + "epoch": 3.0, + "learning_rate": 2.701588865623845e-06, + "loss": 0.5218, + "step": 8126 + }, + { + "epoch": 3.0, + "learning_rate": 2.7015519152605e-06, + "loss": 0.624, + "step": 8127 + }, + { + "epoch": 3.0, + "learning_rate": 2.701514964897155e-06, + "loss": 0.4764, + "step": 8128 + }, + { + "epoch": 3.0, + "learning_rate": 2.70147801453381e-06, + "loss": 0.6013, + "step": 8129 + }, + { + "epoch": 3.0, + "learning_rate": 2.7014410641704645e-06, + "loss": 0.5767, + "step": 8130 + }, + { + "epoch": 3.0, + "learning_rate": 2.701404113807119e-06, + "loss": 0.578, + "step": 8131 + }, + { + "epoch": 3.0, + "learning_rate": 2.7013671634437738e-06, + "loss": 0.6124, + "step": 8132 + }, + { + "epoch": 3.0, + "learning_rate": 2.701330213080429e-06, + "loss": 0.6175, + "step": 8133 + }, + { + "epoch": 3.0, + "learning_rate": 2.7012932627170835e-06, + "loss": 0.6011, + "step": 8134 + }, + { + "epoch": 3.0, + "learning_rate": 2.7012563123537385e-06, + "loss": 0.5298, + "step": 8135 + }, + { + "epoch": 3.0, + "learning_rate": 2.701219361990393e-06, + "loss": 0.5598, + "step": 8136 + }, + { + "epoch": 3.0, + "learning_rate": 2.701182411627048e-06, + "loss": 0.5853, + "step": 8137 + }, + { + "epoch": 3.01, + "learning_rate": 2.7011454612637024e-06, + "loss": 0.5708, + "step": 8138 + }, + { + "epoch": 3.01, + "learning_rate": 2.701108510900357e-06, + "loss": 0.5577, + "step": 8139 + }, + { + "epoch": 3.01, + "learning_rate": 2.701071560537012e-06, + "loss": 0.5737, + "step": 8140 + }, + { + "epoch": 3.01, + "learning_rate": 2.7010346101736668e-06, + "loss": 0.6146, + "step": 8141 + }, + { + "epoch": 3.01, + "learning_rate": 2.700997659810322e-06, + "loss": 0.5629, + "step": 8142 + }, + { + "epoch": 3.01, + "learning_rate": 2.7009607094469765e-06, + "loss": 0.7467, + "step": 8143 + }, + { + "epoch": 3.01, + "learning_rate": 2.700923759083631e-06, + "loss": 0.462, + "step": 8144 + }, + { + "epoch": 3.01, + "learning_rate": 2.7008868087202857e-06, + "loss": 0.5525, + "step": 8145 + }, + { + "epoch": 3.01, + "learning_rate": 2.7008498583569404e-06, + "loss": 0.5871, + "step": 8146 + }, + { + "epoch": 3.01, + "learning_rate": 2.7008129079935954e-06, + "loss": 0.6453, + "step": 8147 + }, + { + "epoch": 3.01, + "learning_rate": 2.70077595763025e-06, + "loss": 0.5402, + "step": 8148 + }, + { + "epoch": 3.01, + "learning_rate": 2.7007390072669047e-06, + "loss": 0.5205, + "step": 8149 + }, + { + "epoch": 3.01, + "learning_rate": 2.7007020569035594e-06, + "loss": 0.544, + "step": 8150 + }, + { + "epoch": 3.01, + "learning_rate": 2.7006651065402144e-06, + "loss": 0.6196, + "step": 8151 + }, + { + "epoch": 3.01, + "learning_rate": 2.700628156176869e-06, + "loss": 0.5575, + "step": 8152 + }, + { + "epoch": 3.01, + "learning_rate": 2.700591205813524e-06, + "loss": 0.5895, + "step": 8153 + }, + { + "epoch": 3.01, + "learning_rate": 2.7005542554501787e-06, + "loss": 0.5927, + "step": 8154 + }, + { + "epoch": 3.01, + "learning_rate": 2.7005173050868334e-06, + "loss": 0.658, + "step": 8155 + }, + { + "epoch": 3.01, + "learning_rate": 2.700480354723488e-06, + "loss": 0.6492, + "step": 8156 + }, + { + "epoch": 3.01, + "learning_rate": 2.7004434043601427e-06, + "loss": 0.5744, + "step": 8157 + }, + { + "epoch": 3.01, + "learning_rate": 2.7004064539967977e-06, + "loss": 0.5273, + "step": 8158 + }, + { + "epoch": 3.01, + "learning_rate": 2.7003695036334523e-06, + "loss": 0.4995, + "step": 8159 + }, + { + "epoch": 3.01, + "learning_rate": 2.7003325532701074e-06, + "loss": 0.4811, + "step": 8160 + }, + { + "epoch": 3.01, + "learning_rate": 2.700295602906762e-06, + "loss": 0.5582, + "step": 8161 + }, + { + "epoch": 3.01, + "learning_rate": 2.7002586525434167e-06, + "loss": 0.4532, + "step": 8162 + }, + { + "epoch": 3.01, + "learning_rate": 2.7002217021800713e-06, + "loss": 0.5765, + "step": 8163 + }, + { + "epoch": 3.01, + "learning_rate": 2.7001847518167264e-06, + "loss": 0.5713, + "step": 8164 + }, + { + "epoch": 3.02, + "learning_rate": 2.700147801453381e-06, + "loss": 0.6028, + "step": 8165 + }, + { + "epoch": 3.02, + "learning_rate": 2.7001108510900357e-06, + "loss": 0.6146, + "step": 8166 + }, + { + "epoch": 3.02, + "learning_rate": 2.7000739007266907e-06, + "loss": 0.5977, + "step": 8167 + }, + { + "epoch": 3.02, + "learning_rate": 2.7000369503633453e-06, + "loss": 0.5055, + "step": 8168 + }, + { + "epoch": 3.02, + "learning_rate": 2.7e-06, + "loss": 0.5516, + "step": 8169 + }, + { + "epoch": 3.02, + "learning_rate": 2.6999630496366546e-06, + "loss": 0.5501, + "step": 8170 + }, + { + "epoch": 3.02, + "learning_rate": 2.6999260992733097e-06, + "loss": 0.56, + "step": 8171 + }, + { + "epoch": 3.02, + "learning_rate": 2.6998891489099643e-06, + "loss": 0.5349, + "step": 8172 + }, + { + "epoch": 3.02, + "learning_rate": 2.6998521985466194e-06, + "loss": 0.5252, + "step": 8173 + }, + { + "epoch": 3.02, + "learning_rate": 2.699815248183274e-06, + "loss": 0.5092, + "step": 8174 + }, + { + "epoch": 3.02, + "learning_rate": 2.6997782978199286e-06, + "loss": 0.6935, + "step": 8175 + }, + { + "epoch": 3.02, + "learning_rate": 2.6997413474565833e-06, + "loss": 0.6572, + "step": 8176 + }, + { + "epoch": 3.02, + "learning_rate": 2.699704397093238e-06, + "loss": 0.5823, + "step": 8177 + }, + { + "epoch": 3.02, + "learning_rate": 2.699667446729893e-06, + "loss": 0.5364, + "step": 8178 + }, + { + "epoch": 3.02, + "learning_rate": 2.6996304963665476e-06, + "loss": 0.6652, + "step": 8179 + }, + { + "epoch": 3.02, + "learning_rate": 2.6995935460032027e-06, + "loss": 0.5839, + "step": 8180 + }, + { + "epoch": 3.02, + "learning_rate": 2.6995565956398573e-06, + "loss": 0.5481, + "step": 8181 + }, + { + "epoch": 3.02, + "learning_rate": 2.699519645276512e-06, + "loss": 0.5572, + "step": 8182 + }, + { + "epoch": 3.02, + "learning_rate": 2.6994826949131666e-06, + "loss": 0.5837, + "step": 8183 + }, + { + "epoch": 3.02, + "learning_rate": 2.6994457445498216e-06, + "loss": 0.5316, + "step": 8184 + }, + { + "epoch": 3.02, + "learning_rate": 2.6994087941864763e-06, + "loss": 0.5846, + "step": 8185 + }, + { + "epoch": 3.02, + "learning_rate": 2.699371843823131e-06, + "loss": 0.5872, + "step": 8186 + }, + { + "epoch": 3.02, + "learning_rate": 2.699334893459786e-06, + "loss": 0.5724, + "step": 8187 + }, + { + "epoch": 3.02, + "learning_rate": 2.69929794309644e-06, + "loss": 0.522, + "step": 8188 + }, + { + "epoch": 3.02, + "learning_rate": 2.6992609927330953e-06, + "loss": 0.6407, + "step": 8189 + }, + { + "epoch": 3.02, + "learning_rate": 2.69922404236975e-06, + "loss": 0.5112, + "step": 8190 + }, + { + "epoch": 3.02, + "learning_rate": 2.699187092006405e-06, + "loss": 0.5686, + "step": 8191 + }, + { + "epoch": 3.03, + "learning_rate": 2.6991501416430596e-06, + "loss": 0.5138, + "step": 8192 + }, + { + "epoch": 3.03, + "learning_rate": 2.6991131912797146e-06, + "loss": 0.5576, + "step": 8193 + }, + { + "epoch": 3.03, + "learning_rate": 2.699076240916369e-06, + "loss": 0.4984, + "step": 8194 + }, + { + "epoch": 3.03, + "learning_rate": 2.6990392905530235e-06, + "loss": 0.7009, + "step": 8195 + }, + { + "epoch": 3.03, + "learning_rate": 2.6990023401896786e-06, + "loss": 0.6083, + "step": 8196 + }, + { + "epoch": 3.03, + "learning_rate": 2.698965389826333e-06, + "loss": 0.5012, + "step": 8197 + }, + { + "epoch": 3.03, + "learning_rate": 2.6989284394629883e-06, + "loss": 0.5736, + "step": 8198 + }, + { + "epoch": 3.03, + "learning_rate": 2.698891489099643e-06, + "loss": 0.5662, + "step": 8199 + }, + { + "epoch": 3.03, + "learning_rate": 2.6988545387362975e-06, + "loss": 0.4761, + "step": 8200 + }, + { + "epoch": 3.03, + "learning_rate": 2.698817588372952e-06, + "loss": 0.5827, + "step": 8201 + }, + { + "epoch": 3.03, + "learning_rate": 2.6987806380096072e-06, + "loss": 0.4666, + "step": 8202 + }, + { + "epoch": 3.03, + "learning_rate": 2.698743687646262e-06, + "loss": 0.5513, + "step": 8203 + }, + { + "epoch": 3.03, + "learning_rate": 2.698706737282917e-06, + "loss": 0.5527, + "step": 8204 + }, + { + "epoch": 3.03, + "learning_rate": 2.6986697869195716e-06, + "loss": 0.5296, + "step": 8205 + }, + { + "epoch": 3.03, + "learning_rate": 2.698632836556226e-06, + "loss": 0.6427, + "step": 8206 + }, + { + "epoch": 3.03, + "learning_rate": 2.698595886192881e-06, + "loss": 0.6231, + "step": 8207 + }, + { + "epoch": 3.03, + "learning_rate": 2.6985589358295355e-06, + "loss": 0.5462, + "step": 8208 + }, + { + "epoch": 3.03, + "learning_rate": 2.6985219854661905e-06, + "loss": 0.5937, + "step": 8209 + }, + { + "epoch": 3.03, + "learning_rate": 2.698485035102845e-06, + "loss": 0.594, + "step": 8210 + }, + { + "epoch": 3.03, + "learning_rate": 2.6984480847395002e-06, + "loss": 0.5788, + "step": 8211 + }, + { + "epoch": 3.03, + "learning_rate": 2.698411134376155e-06, + "loss": 0.5383, + "step": 8212 + }, + { + "epoch": 3.03, + "learning_rate": 2.6983741840128095e-06, + "loss": 0.5018, + "step": 8213 + }, + { + "epoch": 3.03, + "learning_rate": 2.698337233649464e-06, + "loss": 0.6255, + "step": 8214 + }, + { + "epoch": 3.03, + "learning_rate": 2.6983002832861188e-06, + "loss": 0.5789, + "step": 8215 + }, + { + "epoch": 3.03, + "learning_rate": 2.698263332922774e-06, + "loss": 0.5014, + "step": 8216 + }, + { + "epoch": 3.03, + "learning_rate": 2.6982263825594285e-06, + "loss": 0.5886, + "step": 8217 + }, + { + "epoch": 3.03, + "learning_rate": 2.6981894321960835e-06, + "loss": 0.5646, + "step": 8218 + }, + { + "epoch": 3.04, + "learning_rate": 2.698152481832738e-06, + "loss": 0.5211, + "step": 8219 + }, + { + "epoch": 3.04, + "learning_rate": 2.698115531469393e-06, + "loss": 0.5529, + "step": 8220 + }, + { + "epoch": 3.04, + "learning_rate": 2.6980785811060474e-06, + "loss": 0.7047, + "step": 8221 + }, + { + "epoch": 3.04, + "learning_rate": 2.6980416307427025e-06, + "loss": 0.6306, + "step": 8222 + }, + { + "epoch": 3.04, + "learning_rate": 2.698004680379357e-06, + "loss": 0.6016, + "step": 8223 + }, + { + "epoch": 3.04, + "learning_rate": 2.697967730016012e-06, + "loss": 0.508, + "step": 8224 + }, + { + "epoch": 3.04, + "learning_rate": 2.697930779652667e-06, + "loss": 0.6164, + "step": 8225 + }, + { + "epoch": 3.04, + "learning_rate": 2.6978938292893215e-06, + "loss": 0.5497, + "step": 8226 + }, + { + "epoch": 3.04, + "learning_rate": 2.697856878925976e-06, + "loss": 0.5008, + "step": 8227 + }, + { + "epoch": 3.04, + "learning_rate": 2.6978199285626307e-06, + "loss": 0.5269, + "step": 8228 + }, + { + "epoch": 3.04, + "learning_rate": 2.697782978199286e-06, + "loss": 0.5349, + "step": 8229 + }, + { + "epoch": 3.04, + "learning_rate": 2.6977460278359404e-06, + "loss": 0.5647, + "step": 8230 + }, + { + "epoch": 3.04, + "learning_rate": 2.6977090774725955e-06, + "loss": 0.548, + "step": 8231 + }, + { + "epoch": 3.04, + "learning_rate": 2.69767212710925e-06, + "loss": 0.3913, + "step": 8232 + }, + { + "epoch": 3.04, + "learning_rate": 2.6976351767459048e-06, + "loss": 0.5028, + "step": 8233 + }, + { + "epoch": 3.04, + "learning_rate": 2.6975982263825594e-06, + "loss": 0.5078, + "step": 8234 + }, + { + "epoch": 3.04, + "learning_rate": 2.697561276019214e-06, + "loss": 0.6426, + "step": 8235 + }, + { + "epoch": 3.04, + "learning_rate": 2.697524325655869e-06, + "loss": 0.5902, + "step": 8236 + }, + { + "epoch": 3.04, + "learning_rate": 2.6974873752925237e-06, + "loss": 0.5246, + "step": 8237 + }, + { + "epoch": 3.04, + "learning_rate": 2.697450424929179e-06, + "loss": 0.566, + "step": 8238 + }, + { + "epoch": 3.04, + "learning_rate": 2.697413474565833e-06, + "loss": 0.6079, + "step": 8239 + }, + { + "epoch": 3.04, + "learning_rate": 2.697376524202488e-06, + "loss": 0.5437, + "step": 8240 + }, + { + "epoch": 3.04, + "learning_rate": 2.6973395738391427e-06, + "loss": 0.5143, + "step": 8241 + }, + { + "epoch": 3.04, + "learning_rate": 2.6973026234757978e-06, + "loss": 0.5355, + "step": 8242 + }, + { + "epoch": 3.04, + "learning_rate": 2.6972656731124524e-06, + "loss": 0.5014, + "step": 8243 + }, + { + "epoch": 3.04, + "learning_rate": 2.6972287227491075e-06, + "loss": 0.6648, + "step": 8244 + }, + { + "epoch": 3.04, + "learning_rate": 2.6971917723857617e-06, + "loss": 0.5548, + "step": 8245 + }, + { + "epoch": 3.05, + "learning_rate": 2.6971548220224163e-06, + "loss": 0.6127, + "step": 8246 + }, + { + "epoch": 3.05, + "learning_rate": 2.6971178716590714e-06, + "loss": 0.5779, + "step": 8247 + }, + { + "epoch": 3.05, + "learning_rate": 2.697080921295726e-06, + "loss": 0.4973, + "step": 8248 + }, + { + "epoch": 3.05, + "learning_rate": 2.697043970932381e-06, + "loss": 0.6247, + "step": 8249 + }, + { + "epoch": 3.05, + "learning_rate": 2.6970070205690357e-06, + "loss": 0.5826, + "step": 8250 + }, + { + "epoch": 3.05, + "learning_rate": 2.6969700702056903e-06, + "loss": 0.6617, + "step": 8251 + }, + { + "epoch": 3.05, + "learning_rate": 2.696933119842345e-06, + "loss": 0.5662, + "step": 8252 + }, + { + "epoch": 3.05, + "learning_rate": 2.696896169479e-06, + "loss": 0.5123, + "step": 8253 + }, + { + "epoch": 3.05, + "learning_rate": 2.6968592191156547e-06, + "loss": 0.5382, + "step": 8254 + }, + { + "epoch": 3.05, + "learning_rate": 2.6968222687523093e-06, + "loss": 0.6104, + "step": 8255 + }, + { + "epoch": 3.05, + "learning_rate": 2.6967853183889644e-06, + "loss": 0.4932, + "step": 8256 + }, + { + "epoch": 3.05, + "learning_rate": 2.696748368025619e-06, + "loss": 0.482, + "step": 8257 + }, + { + "epoch": 3.05, + "learning_rate": 2.6967114176622736e-06, + "loss": 0.5397, + "step": 8258 + }, + { + "epoch": 3.05, + "learning_rate": 2.6966744672989283e-06, + "loss": 0.5927, + "step": 8259 + }, + { + "epoch": 3.05, + "learning_rate": 2.6966375169355833e-06, + "loss": 0.6794, + "step": 8260 + }, + { + "epoch": 3.05, + "learning_rate": 2.696600566572238e-06, + "loss": 0.5216, + "step": 8261 + }, + { + "epoch": 3.05, + "learning_rate": 2.696563616208893e-06, + "loss": 0.5199, + "step": 8262 + }, + { + "epoch": 3.05, + "learning_rate": 2.6965266658455477e-06, + "loss": 0.5745, + "step": 8263 + }, + { + "epoch": 3.05, + "learning_rate": 2.6964897154822023e-06, + "loss": 0.5884, + "step": 8264 + }, + { + "epoch": 3.05, + "learning_rate": 2.696452765118857e-06, + "loss": 0.4891, + "step": 8265 + }, + { + "epoch": 3.05, + "learning_rate": 2.6964158147555116e-06, + "loss": 0.5586, + "step": 8266 + }, + { + "epoch": 3.05, + "learning_rate": 2.6963788643921666e-06, + "loss": 0.5078, + "step": 8267 + }, + { + "epoch": 3.05, + "learning_rate": 2.6963419140288213e-06, + "loss": 0.4981, + "step": 8268 + }, + { + "epoch": 3.05, + "learning_rate": 2.6963049636654763e-06, + "loss": 0.6299, + "step": 8269 + }, + { + "epoch": 3.05, + "learning_rate": 2.696268013302131e-06, + "loss": 0.5487, + "step": 8270 + }, + { + "epoch": 3.05, + "learning_rate": 2.6962310629387856e-06, + "loss": 0.4901, + "step": 8271 + }, + { + "epoch": 3.05, + "learning_rate": 2.6961941125754402e-06, + "loss": 0.5208, + "step": 8272 + }, + { + "epoch": 3.06, + "learning_rate": 2.6961571622120953e-06, + "loss": 0.5826, + "step": 8273 + }, + { + "epoch": 3.06, + "learning_rate": 2.69612021184875e-06, + "loss": 0.5032, + "step": 8274 + }, + { + "epoch": 3.06, + "learning_rate": 2.6960832614854046e-06, + "loss": 0.5138, + "step": 8275 + }, + { + "epoch": 3.06, + "learning_rate": 2.6960463111220596e-06, + "loss": 0.5626, + "step": 8276 + }, + { + "epoch": 3.06, + "learning_rate": 2.6960093607587143e-06, + "loss": 0.5969, + "step": 8277 + }, + { + "epoch": 3.06, + "learning_rate": 2.695972410395369e-06, + "loss": 0.5913, + "step": 8278 + }, + { + "epoch": 3.06, + "learning_rate": 2.6959354600320235e-06, + "loss": 0.5598, + "step": 8279 + }, + { + "epoch": 3.06, + "learning_rate": 2.6958985096686786e-06, + "loss": 0.5032, + "step": 8280 + }, + { + "epoch": 3.06, + "learning_rate": 2.6958615593053332e-06, + "loss": 0.5678, + "step": 8281 + }, + { + "epoch": 3.06, + "learning_rate": 2.6958246089419883e-06, + "loss": 0.6385, + "step": 8282 + }, + { + "epoch": 3.06, + "learning_rate": 2.695787658578643e-06, + "loss": 0.6282, + "step": 8283 + }, + { + "epoch": 3.06, + "learning_rate": 2.6957507082152976e-06, + "loss": 0.5673, + "step": 8284 + }, + { + "epoch": 3.06, + "learning_rate": 2.695713757851952e-06, + "loss": 0.5916, + "step": 8285 + }, + { + "epoch": 3.06, + "learning_rate": 2.695676807488607e-06, + "loss": 0.5342, + "step": 8286 + }, + { + "epoch": 3.06, + "learning_rate": 2.695639857125262e-06, + "loss": 0.5039, + "step": 8287 + }, + { + "epoch": 3.06, + "learning_rate": 2.6956029067619165e-06, + "loss": 0.4898, + "step": 8288 + }, + { + "epoch": 3.06, + "learning_rate": 2.6955659563985716e-06, + "loss": 0.5444, + "step": 8289 + }, + { + "epoch": 3.06, + "learning_rate": 2.695529006035226e-06, + "loss": 0.5319, + "step": 8290 + }, + { + "epoch": 3.06, + "learning_rate": 2.695492055671881e-06, + "loss": 0.5193, + "step": 8291 + }, + { + "epoch": 3.06, + "learning_rate": 2.6954551053085355e-06, + "loss": 0.5084, + "step": 8292 + }, + { + "epoch": 3.06, + "learning_rate": 2.6954181549451906e-06, + "loss": 0.5526, + "step": 8293 + }, + { + "epoch": 3.06, + "learning_rate": 2.695381204581845e-06, + "loss": 0.4986, + "step": 8294 + }, + { + "epoch": 3.06, + "learning_rate": 2.6953442542185e-06, + "loss": 0.5231, + "step": 8295 + }, + { + "epoch": 3.06, + "learning_rate": 2.6953073038551545e-06, + "loss": 0.5808, + "step": 8296 + }, + { + "epoch": 3.06, + "learning_rate": 2.695270353491809e-06, + "loss": 0.5121, + "step": 8297 + }, + { + "epoch": 3.06, + "learning_rate": 2.695233403128464e-06, + "loss": 0.6143, + "step": 8298 + }, + { + "epoch": 3.06, + "learning_rate": 2.695196452765119e-06, + "loss": 0.5084, + "step": 8299 + }, + { + "epoch": 3.06, + "learning_rate": 2.695159502401774e-06, + "loss": 0.5236, + "step": 8300 + }, + { + "epoch": 3.07, + "learning_rate": 2.6951225520384285e-06, + "loss": 0.5636, + "step": 8301 + }, + { + "epoch": 3.07, + "learning_rate": 2.695085601675083e-06, + "loss": 0.5175, + "step": 8302 + }, + { + "epoch": 3.07, + "learning_rate": 2.6950486513117378e-06, + "loss": 0.4742, + "step": 8303 + }, + { + "epoch": 3.07, + "learning_rate": 2.695011700948393e-06, + "loss": 0.6147, + "step": 8304 + }, + { + "epoch": 3.07, + "learning_rate": 2.6949747505850475e-06, + "loss": 0.6025, + "step": 8305 + }, + { + "epoch": 3.07, + "learning_rate": 2.694937800221702e-06, + "loss": 0.5652, + "step": 8306 + }, + { + "epoch": 3.07, + "learning_rate": 2.694900849858357e-06, + "loss": 0.6361, + "step": 8307 + }, + { + "epoch": 3.07, + "learning_rate": 2.694863899495012e-06, + "loss": 0.4805, + "step": 8308 + }, + { + "epoch": 3.07, + "learning_rate": 2.6948269491316664e-06, + "loss": 0.5852, + "step": 8309 + }, + { + "epoch": 3.07, + "learning_rate": 2.694789998768321e-06, + "loss": 0.5909, + "step": 8310 + }, + { + "epoch": 3.07, + "learning_rate": 2.694753048404976e-06, + "loss": 0.5211, + "step": 8311 + }, + { + "epoch": 3.07, + "learning_rate": 2.6947160980416308e-06, + "loss": 0.4432, + "step": 8312 + }, + { + "epoch": 3.07, + "learning_rate": 2.694679147678286e-06, + "loss": 0.4874, + "step": 8313 + }, + { + "epoch": 3.07, + "learning_rate": 2.6946421973149405e-06, + "loss": 0.6035, + "step": 8314 + }, + { + "epoch": 3.07, + "learning_rate": 2.694605246951595e-06, + "loss": 0.5247, + "step": 8315 + }, + { + "epoch": 3.07, + "learning_rate": 2.6945682965882498e-06, + "loss": 0.5791, + "step": 8316 + }, + { + "epoch": 3.07, + "learning_rate": 2.6945313462249044e-06, + "loss": 0.6283, + "step": 8317 + }, + { + "epoch": 3.07, + "learning_rate": 2.6944943958615594e-06, + "loss": 0.6144, + "step": 8318 + }, + { + "epoch": 3.07, + "learning_rate": 2.694457445498214e-06, + "loss": 0.5907, + "step": 8319 + }, + { + "epoch": 3.07, + "learning_rate": 2.694420495134869e-06, + "loss": 0.4598, + "step": 8320 + }, + { + "epoch": 3.07, + "learning_rate": 2.6943835447715238e-06, + "loss": 0.5238, + "step": 8321 + }, + { + "epoch": 3.07, + "learning_rate": 2.6943465944081784e-06, + "loss": 0.6302, + "step": 8322 + }, + { + "epoch": 3.07, + "learning_rate": 2.694309644044833e-06, + "loss": 0.5288, + "step": 8323 + }, + { + "epoch": 3.07, + "learning_rate": 2.694272693681488e-06, + "loss": 0.5107, + "step": 8324 + }, + { + "epoch": 3.07, + "learning_rate": 2.6942357433181427e-06, + "loss": 0.5781, + "step": 8325 + }, + { + "epoch": 3.07, + "learning_rate": 2.6941987929547974e-06, + "loss": 0.4907, + "step": 8326 + }, + { + "epoch": 3.07, + "learning_rate": 2.6941618425914524e-06, + "loss": 0.6763, + "step": 8327 + }, + { + "epoch": 3.08, + "learning_rate": 2.694124892228107e-06, + "loss": 0.5464, + "step": 8328 + }, + { + "epoch": 3.08, + "learning_rate": 2.6940879418647617e-06, + "loss": 0.4814, + "step": 8329 + }, + { + "epoch": 3.08, + "learning_rate": 2.6940509915014164e-06, + "loss": 0.6754, + "step": 8330 + }, + { + "epoch": 3.08, + "learning_rate": 2.6940140411380714e-06, + "loss": 0.5236, + "step": 8331 + }, + { + "epoch": 3.08, + "learning_rate": 2.693977090774726e-06, + "loss": 0.5295, + "step": 8332 + }, + { + "epoch": 3.08, + "learning_rate": 2.693940140411381e-06, + "loss": 0.5884, + "step": 8333 + }, + { + "epoch": 3.08, + "learning_rate": 2.6939031900480353e-06, + "loss": 0.457, + "step": 8334 + }, + { + "epoch": 3.08, + "learning_rate": 2.69386623968469e-06, + "loss": 0.5767, + "step": 8335 + }, + { + "epoch": 3.08, + "learning_rate": 2.693829289321345e-06, + "loss": 0.6048, + "step": 8336 + }, + { + "epoch": 3.08, + "learning_rate": 2.6937923389579997e-06, + "loss": 0.5911, + "step": 8337 + }, + { + "epoch": 3.08, + "learning_rate": 2.6937553885946547e-06, + "loss": 0.5531, + "step": 8338 + }, + { + "epoch": 3.08, + "learning_rate": 2.6937184382313094e-06, + "loss": 0.6412, + "step": 8339 + }, + { + "epoch": 3.08, + "learning_rate": 2.693681487867964e-06, + "loss": 0.5577, + "step": 8340 + }, + { + "epoch": 3.08, + "learning_rate": 2.6936445375046186e-06, + "loss": 0.5216, + "step": 8341 + }, + { + "epoch": 3.08, + "learning_rate": 2.6936075871412737e-06, + "loss": 0.5142, + "step": 8342 + }, + { + "epoch": 3.08, + "learning_rate": 2.6935706367779283e-06, + "loss": 0.6643, + "step": 8343 + }, + { + "epoch": 3.08, + "learning_rate": 2.6935336864145834e-06, + "loss": 0.465, + "step": 8344 + }, + { + "epoch": 3.08, + "learning_rate": 2.693496736051238e-06, + "loss": 0.6575, + "step": 8345 + }, + { + "epoch": 3.08, + "learning_rate": 2.6934597856878927e-06, + "loss": 0.5751, + "step": 8346 + }, + { + "epoch": 3.08, + "learning_rate": 2.6934228353245473e-06, + "loss": 0.5016, + "step": 8347 + }, + { + "epoch": 3.08, + "learning_rate": 2.693385884961202e-06, + "loss": 0.5652, + "step": 8348 + }, + { + "epoch": 3.08, + "learning_rate": 2.693348934597857e-06, + "loss": 0.4821, + "step": 8349 + }, + { + "epoch": 3.08, + "learning_rate": 2.6933119842345116e-06, + "loss": 0.5293, + "step": 8350 + }, + { + "epoch": 3.08, + "learning_rate": 2.6932750338711667e-06, + "loss": 0.5414, + "step": 8351 + }, + { + "epoch": 3.08, + "learning_rate": 2.6932380835078213e-06, + "loss": 0.5655, + "step": 8352 + }, + { + "epoch": 3.08, + "learning_rate": 2.693201133144476e-06, + "loss": 0.5984, + "step": 8353 + }, + { + "epoch": 3.08, + "learning_rate": 2.6931641827811306e-06, + "loss": 0.5714, + "step": 8354 + }, + { + "epoch": 3.09, + "learning_rate": 2.6931272324177852e-06, + "loss": 0.5685, + "step": 8355 + }, + { + "epoch": 3.09, + "learning_rate": 2.6930902820544403e-06, + "loss": 0.5201, + "step": 8356 + }, + { + "epoch": 3.09, + "learning_rate": 2.693053331691095e-06, + "loss": 0.6359, + "step": 8357 + }, + { + "epoch": 3.09, + "learning_rate": 2.69301638132775e-06, + "loss": 0.5631, + "step": 8358 + }, + { + "epoch": 3.09, + "learning_rate": 2.6929794309644046e-06, + "loss": 0.4717, + "step": 8359 + }, + { + "epoch": 3.09, + "learning_rate": 2.6929424806010593e-06, + "loss": 0.5352, + "step": 8360 + }, + { + "epoch": 3.09, + "learning_rate": 2.692905530237714e-06, + "loss": 0.6454, + "step": 8361 + }, + { + "epoch": 3.09, + "learning_rate": 2.692868579874369e-06, + "loss": 0.5098, + "step": 8362 + }, + { + "epoch": 3.09, + "learning_rate": 2.6928316295110236e-06, + "loss": 0.6374, + "step": 8363 + }, + { + "epoch": 3.09, + "learning_rate": 2.6927946791476787e-06, + "loss": 0.5039, + "step": 8364 + }, + { + "epoch": 3.09, + "learning_rate": 2.6927577287843333e-06, + "loss": 0.5519, + "step": 8365 + }, + { + "epoch": 3.09, + "learning_rate": 2.692720778420988e-06, + "loss": 0.6859, + "step": 8366 + }, + { + "epoch": 3.09, + "learning_rate": 2.6926838280576426e-06, + "loss": 0.5028, + "step": 8367 + }, + { + "epoch": 3.09, + "learning_rate": 2.692646877694297e-06, + "loss": 0.6723, + "step": 8368 + }, + { + "epoch": 3.09, + "learning_rate": 2.6926099273309523e-06, + "loss": 0.5991, + "step": 8369 + }, + { + "epoch": 3.09, + "learning_rate": 2.692572976967607e-06, + "loss": 0.5736, + "step": 8370 + }, + { + "epoch": 3.09, + "learning_rate": 2.692536026604262e-06, + "loss": 0.4724, + "step": 8371 + }, + { + "epoch": 3.09, + "learning_rate": 2.6924990762409166e-06, + "loss": 0.5455, + "step": 8372 + }, + { + "epoch": 3.09, + "learning_rate": 2.6924621258775712e-06, + "loss": 0.4541, + "step": 8373 + }, + { + "epoch": 3.09, + "learning_rate": 2.692425175514226e-06, + "loss": 0.5827, + "step": 8374 + }, + { + "epoch": 3.09, + "learning_rate": 2.6923882251508805e-06, + "loss": 0.6085, + "step": 8375 + }, + { + "epoch": 3.09, + "learning_rate": 2.6923512747875356e-06, + "loss": 0.5269, + "step": 8376 + }, + { + "epoch": 3.09, + "learning_rate": 2.69231432442419e-06, + "loss": 0.5307, + "step": 8377 + }, + { + "epoch": 3.09, + "learning_rate": 2.6922773740608453e-06, + "loss": 0.5911, + "step": 8378 + }, + { + "epoch": 3.09, + "learning_rate": 2.6922404236974995e-06, + "loss": 0.4953, + "step": 8379 + }, + { + "epoch": 3.09, + "learning_rate": 2.6922034733341545e-06, + "loss": 0.4382, + "step": 8380 + }, + { + "epoch": 3.09, + "learning_rate": 2.692166522970809e-06, + "loss": 0.5952, + "step": 8381 + }, + { + "epoch": 3.1, + "learning_rate": 2.6921295726074642e-06, + "loss": 0.5233, + "step": 8382 + }, + { + "epoch": 3.1, + "learning_rate": 2.692092622244119e-06, + "loss": 0.5889, + "step": 8383 + }, + { + "epoch": 3.1, + "learning_rate": 2.692055671880774e-06, + "loss": 0.6239, + "step": 8384 + }, + { + "epoch": 3.1, + "learning_rate": 2.692018721517428e-06, + "loss": 0.6172, + "step": 8385 + }, + { + "epoch": 3.1, + "learning_rate": 2.6919817711540828e-06, + "loss": 0.6215, + "step": 8386 + }, + { + "epoch": 3.1, + "learning_rate": 2.691944820790738e-06, + "loss": 0.5275, + "step": 8387 + }, + { + "epoch": 3.1, + "learning_rate": 2.6919078704273925e-06, + "loss": 0.6323, + "step": 8388 + }, + { + "epoch": 3.1, + "learning_rate": 2.6918709200640475e-06, + "loss": 0.5413, + "step": 8389 + }, + { + "epoch": 3.1, + "learning_rate": 2.691833969700702e-06, + "loss": 0.5717, + "step": 8390 + }, + { + "epoch": 3.1, + "learning_rate": 2.691797019337357e-06, + "loss": 0.5846, + "step": 8391 + }, + { + "epoch": 3.1, + "learning_rate": 2.6917600689740114e-06, + "loss": 0.5969, + "step": 8392 + }, + { + "epoch": 3.1, + "learning_rate": 2.6917231186106665e-06, + "loss": 0.5179, + "step": 8393 + }, + { + "epoch": 3.1, + "learning_rate": 2.691686168247321e-06, + "loss": 0.5838, + "step": 8394 + }, + { + "epoch": 3.1, + "learning_rate": 2.6916492178839758e-06, + "loss": 0.5536, + "step": 8395 + }, + { + "epoch": 3.1, + "learning_rate": 2.691612267520631e-06, + "loss": 0.6028, + "step": 8396 + }, + { + "epoch": 3.1, + "learning_rate": 2.6915753171572855e-06, + "loss": 0.5853, + "step": 8397 + }, + { + "epoch": 3.1, + "learning_rate": 2.69153836679394e-06, + "loss": 0.6088, + "step": 8398 + }, + { + "epoch": 3.1, + "learning_rate": 2.6915014164305947e-06, + "loss": 0.5072, + "step": 8399 + }, + { + "epoch": 3.1, + "learning_rate": 2.69146446606725e-06, + "loss": 0.619, + "step": 8400 + }, + { + "epoch": 3.1, + "learning_rate": 2.6914275157039044e-06, + "loss": 0.4879, + "step": 8401 + }, + { + "epoch": 3.1, + "learning_rate": 2.6913905653405595e-06, + "loss": 0.51, + "step": 8402 + }, + { + "epoch": 3.1, + "learning_rate": 2.691353614977214e-06, + "loss": 0.6015, + "step": 8403 + }, + { + "epoch": 3.1, + "learning_rate": 2.6913166646138688e-06, + "loss": 0.5415, + "step": 8404 + }, + { + "epoch": 3.1, + "learning_rate": 2.6912797142505234e-06, + "loss": 0.6289, + "step": 8405 + }, + { + "epoch": 3.1, + "learning_rate": 2.691242763887178e-06, + "loss": 0.5339, + "step": 8406 + }, + { + "epoch": 3.1, + "learning_rate": 2.691205813523833e-06, + "loss": 0.6033, + "step": 8407 + }, + { + "epoch": 3.1, + "learning_rate": 2.6911688631604877e-06, + "loss": 0.6558, + "step": 8408 + }, + { + "epoch": 3.11, + "learning_rate": 2.691131912797143e-06, + "loss": 0.5741, + "step": 8409 + }, + { + "epoch": 3.11, + "learning_rate": 2.6910949624337974e-06, + "loss": 0.5631, + "step": 8410 + }, + { + "epoch": 3.11, + "learning_rate": 2.691058012070452e-06, + "loss": 0.5921, + "step": 8411 + }, + { + "epoch": 3.11, + "learning_rate": 2.6910210617071067e-06, + "loss": 0.5442, + "step": 8412 + }, + { + "epoch": 3.11, + "learning_rate": 2.6909841113437618e-06, + "loss": 0.5442, + "step": 8413 + }, + { + "epoch": 3.11, + "learning_rate": 2.6909471609804164e-06, + "loss": 0.4619, + "step": 8414 + }, + { + "epoch": 3.11, + "learning_rate": 2.690910210617071e-06, + "loss": 0.5284, + "step": 8415 + }, + { + "epoch": 3.11, + "learning_rate": 2.690873260253726e-06, + "loss": 0.5589, + "step": 8416 + }, + { + "epoch": 3.11, + "learning_rate": 2.6908363098903807e-06, + "loss": 0.5888, + "step": 8417 + }, + { + "epoch": 3.11, + "learning_rate": 2.6907993595270354e-06, + "loss": 0.6431, + "step": 8418 + }, + { + "epoch": 3.11, + "learning_rate": 2.69076240916369e-06, + "loss": 0.4924, + "step": 8419 + }, + { + "epoch": 3.11, + "learning_rate": 2.690725458800345e-06, + "loss": 0.6281, + "step": 8420 + }, + { + "epoch": 3.11, + "learning_rate": 2.6906885084369997e-06, + "loss": 0.5421, + "step": 8421 + }, + { + "epoch": 3.11, + "learning_rate": 2.6906515580736548e-06, + "loss": 0.4903, + "step": 8422 + }, + { + "epoch": 3.11, + "learning_rate": 2.6906146077103094e-06, + "loss": 0.5527, + "step": 8423 + }, + { + "epoch": 3.11, + "learning_rate": 2.6905776573469636e-06, + "loss": 0.5797, + "step": 8424 + }, + { + "epoch": 3.11, + "learning_rate": 2.6905407069836187e-06, + "loss": 0.5775, + "step": 8425 + }, + { + "epoch": 3.11, + "learning_rate": 2.6905037566202733e-06, + "loss": 0.5044, + "step": 8426 + }, + { + "epoch": 3.11, + "learning_rate": 2.6904668062569284e-06, + "loss": 0.5996, + "step": 8427 + }, + { + "epoch": 3.11, + "learning_rate": 2.690429855893583e-06, + "loss": 0.628, + "step": 8428 + }, + { + "epoch": 3.11, + "learning_rate": 2.690392905530238e-06, + "loss": 0.5541, + "step": 8429 + }, + { + "epoch": 3.11, + "learning_rate": 2.6903559551668923e-06, + "loss": 0.5136, + "step": 8430 + }, + { + "epoch": 3.11, + "learning_rate": 2.6903190048035473e-06, + "loss": 0.6466, + "step": 8431 + }, + { + "epoch": 3.11, + "learning_rate": 2.690282054440202e-06, + "loss": 0.494, + "step": 8432 + }, + { + "epoch": 3.11, + "learning_rate": 2.690245104076857e-06, + "loss": 0.5884, + "step": 8433 + }, + { + "epoch": 3.11, + "learning_rate": 2.6902081537135117e-06, + "loss": 0.569, + "step": 8434 + }, + { + "epoch": 3.11, + "learning_rate": 2.6901712033501663e-06, + "loss": 0.5382, + "step": 8435 + }, + { + "epoch": 3.12, + "learning_rate": 2.690134252986821e-06, + "loss": 0.5269, + "step": 8436 + }, + { + "epoch": 3.12, + "learning_rate": 2.6900973026234756e-06, + "loss": 0.6423, + "step": 8437 + }, + { + "epoch": 3.12, + "learning_rate": 2.6900603522601306e-06, + "loss": 0.6081, + "step": 8438 + }, + { + "epoch": 3.12, + "learning_rate": 2.6900234018967853e-06, + "loss": 0.5455, + "step": 8439 + }, + { + "epoch": 3.12, + "learning_rate": 2.6899864515334403e-06, + "loss": 0.555, + "step": 8440 + }, + { + "epoch": 3.12, + "learning_rate": 2.689949501170095e-06, + "loss": 0.5672, + "step": 8441 + }, + { + "epoch": 3.12, + "learning_rate": 2.6899125508067496e-06, + "loss": 0.5496, + "step": 8442 + }, + { + "epoch": 3.12, + "learning_rate": 2.6898756004434042e-06, + "loss": 0.6527, + "step": 8443 + }, + { + "epoch": 3.12, + "learning_rate": 2.689838650080059e-06, + "loss": 0.5138, + "step": 8444 + }, + { + "epoch": 3.12, + "learning_rate": 2.689801699716714e-06, + "loss": 0.5279, + "step": 8445 + }, + { + "epoch": 3.12, + "learning_rate": 2.6897647493533686e-06, + "loss": 0.6536, + "step": 8446 + }, + { + "epoch": 3.12, + "learning_rate": 2.6897277989900236e-06, + "loss": 0.4839, + "step": 8447 + }, + { + "epoch": 3.12, + "learning_rate": 2.6896908486266783e-06, + "loss": 0.4523, + "step": 8448 + }, + { + "epoch": 3.12, + "learning_rate": 2.689653898263333e-06, + "loss": 0.6428, + "step": 8449 + }, + { + "epoch": 3.12, + "learning_rate": 2.6896169478999876e-06, + "loss": 0.4912, + "step": 8450 + }, + { + "epoch": 3.12, + "learning_rate": 2.6895799975366426e-06, + "loss": 0.5904, + "step": 8451 + }, + { + "epoch": 3.12, + "learning_rate": 2.6895430471732972e-06, + "loss": 0.5608, + "step": 8452 + }, + { + "epoch": 3.12, + "learning_rate": 2.6895060968099523e-06, + "loss": 0.5325, + "step": 8453 + }, + { + "epoch": 3.12, + "learning_rate": 2.689469146446607e-06, + "loss": 0.5927, + "step": 8454 + }, + { + "epoch": 3.12, + "learning_rate": 2.6894321960832616e-06, + "loss": 0.6493, + "step": 8455 + }, + { + "epoch": 3.12, + "learning_rate": 2.6893952457199162e-06, + "loss": 0.6131, + "step": 8456 + }, + { + "epoch": 3.12, + "learning_rate": 2.689358295356571e-06, + "loss": 0.6408, + "step": 8457 + }, + { + "epoch": 3.12, + "learning_rate": 2.689321344993226e-06, + "loss": 0.5551, + "step": 8458 + }, + { + "epoch": 3.12, + "learning_rate": 2.6892843946298805e-06, + "loss": 0.4342, + "step": 8459 + }, + { + "epoch": 3.12, + "learning_rate": 2.6892474442665356e-06, + "loss": 0.5115, + "step": 8460 + }, + { + "epoch": 3.12, + "learning_rate": 2.6892104939031902e-06, + "loss": 0.4691, + "step": 8461 + }, + { + "epoch": 3.12, + "learning_rate": 2.689173543539845e-06, + "loss": 0.585, + "step": 8462 + }, + { + "epoch": 3.13, + "learning_rate": 2.6891365931764995e-06, + "loss": 0.5154, + "step": 8463 + }, + { + "epoch": 3.13, + "learning_rate": 2.689099642813154e-06, + "loss": 0.6603, + "step": 8464 + }, + { + "epoch": 3.13, + "learning_rate": 2.6890626924498092e-06, + "loss": 0.6448, + "step": 8465 + }, + { + "epoch": 3.13, + "learning_rate": 2.689025742086464e-06, + "loss": 0.6072, + "step": 8466 + }, + { + "epoch": 3.13, + "learning_rate": 2.688988791723119e-06, + "loss": 0.5151, + "step": 8467 + }, + { + "epoch": 3.13, + "learning_rate": 2.6889518413597735e-06, + "loss": 0.5598, + "step": 8468 + }, + { + "epoch": 3.13, + "learning_rate": 2.688914890996428e-06, + "loss": 0.562, + "step": 8469 + }, + { + "epoch": 3.13, + "learning_rate": 2.688877940633083e-06, + "loss": 0.5959, + "step": 8470 + }, + { + "epoch": 3.13, + "learning_rate": 2.688840990269738e-06, + "loss": 0.5475, + "step": 8471 + }, + { + "epoch": 3.13, + "learning_rate": 2.6888040399063925e-06, + "loss": 0.5153, + "step": 8472 + }, + { + "epoch": 3.13, + "learning_rate": 2.6887670895430476e-06, + "loss": 0.5187, + "step": 8473 + }, + { + "epoch": 3.13, + "learning_rate": 2.6887301391797022e-06, + "loss": 0.6142, + "step": 8474 + }, + { + "epoch": 3.13, + "learning_rate": 2.6886931888163564e-06, + "loss": 0.4853, + "step": 8475 + }, + { + "epoch": 3.13, + "learning_rate": 2.6886562384530115e-06, + "loss": 0.493, + "step": 8476 + }, + { + "epoch": 3.13, + "learning_rate": 2.688619288089666e-06, + "loss": 0.5982, + "step": 8477 + }, + { + "epoch": 3.13, + "learning_rate": 2.688582337726321e-06, + "loss": 0.5698, + "step": 8478 + }, + { + "epoch": 3.13, + "learning_rate": 2.688545387362976e-06, + "loss": 0.5854, + "step": 8479 + }, + { + "epoch": 3.13, + "learning_rate": 2.688508436999631e-06, + "loss": 0.5311, + "step": 8480 + }, + { + "epoch": 3.13, + "learning_rate": 2.688471486636285e-06, + "loss": 0.5633, + "step": 8481 + }, + { + "epoch": 3.13, + "learning_rate": 2.68843453627294e-06, + "loss": 0.6372, + "step": 8482 + }, + { + "epoch": 3.13, + "learning_rate": 2.6883975859095948e-06, + "loss": 0.5183, + "step": 8483 + }, + { + "epoch": 3.13, + "learning_rate": 2.6883606355462494e-06, + "loss": 0.5883, + "step": 8484 + }, + { + "epoch": 3.13, + "learning_rate": 2.6883236851829045e-06, + "loss": 0.5032, + "step": 8485 + }, + { + "epoch": 3.13, + "learning_rate": 2.688286734819559e-06, + "loss": 0.5496, + "step": 8486 + }, + { + "epoch": 3.13, + "learning_rate": 2.6882497844562138e-06, + "loss": 0.5539, + "step": 8487 + }, + { + "epoch": 3.13, + "learning_rate": 2.6882128340928684e-06, + "loss": 0.5705, + "step": 8488 + }, + { + "epoch": 3.13, + "learning_rate": 2.6881758837295235e-06, + "loss": 0.5743, + "step": 8489 + }, + { + "epoch": 3.14, + "learning_rate": 2.688138933366178e-06, + "loss": 0.57, + "step": 8490 + }, + { + "epoch": 3.14, + "learning_rate": 2.688101983002833e-06, + "loss": 0.5991, + "step": 8491 + }, + { + "epoch": 3.14, + "learning_rate": 2.6880650326394878e-06, + "loss": 0.5866, + "step": 8492 + }, + { + "epoch": 3.14, + "learning_rate": 2.6880280822761424e-06, + "loss": 0.5551, + "step": 8493 + }, + { + "epoch": 3.14, + "learning_rate": 2.687991131912797e-06, + "loss": 0.5104, + "step": 8494 + }, + { + "epoch": 3.14, + "learning_rate": 2.6879541815494517e-06, + "loss": 0.5764, + "step": 8495 + }, + { + "epoch": 3.14, + "learning_rate": 2.6879172311861068e-06, + "loss": 0.5444, + "step": 8496 + }, + { + "epoch": 3.14, + "learning_rate": 2.6878802808227614e-06, + "loss": 0.5011, + "step": 8497 + }, + { + "epoch": 3.14, + "learning_rate": 2.6878433304594165e-06, + "loss": 0.5044, + "step": 8498 + }, + { + "epoch": 3.14, + "learning_rate": 2.687806380096071e-06, + "loss": 0.5291, + "step": 8499 + }, + { + "epoch": 3.14, + "learning_rate": 2.6877694297327257e-06, + "loss": 0.5808, + "step": 8500 + }, + { + "epoch": 3.14, + "learning_rate": 2.6877324793693804e-06, + "loss": 0.5039, + "step": 8501 + }, + { + "epoch": 3.14, + "learning_rate": 2.6876955290060354e-06, + "loss": 0.5247, + "step": 8502 + }, + { + "epoch": 3.14, + "learning_rate": 2.68765857864269e-06, + "loss": 0.6191, + "step": 8503 + }, + { + "epoch": 3.14, + "learning_rate": 2.6876216282793447e-06, + "loss": 0.571, + "step": 8504 + }, + { + "epoch": 3.14, + "learning_rate": 2.6875846779159998e-06, + "loss": 0.5149, + "step": 8505 + }, + { + "epoch": 3.14, + "learning_rate": 2.6875477275526544e-06, + "loss": 0.5305, + "step": 8506 + }, + { + "epoch": 3.14, + "learning_rate": 2.687510777189309e-06, + "loss": 0.4829, + "step": 8507 + }, + { + "epoch": 3.14, + "learning_rate": 2.6874738268259637e-06, + "loss": 0.6054, + "step": 8508 + }, + { + "epoch": 3.14, + "learning_rate": 2.6874368764626187e-06, + "loss": 0.7147, + "step": 8509 + }, + { + "epoch": 3.14, + "learning_rate": 2.6873999260992734e-06, + "loss": 0.5789, + "step": 8510 + }, + { + "epoch": 3.14, + "learning_rate": 2.6873629757359284e-06, + "loss": 0.5821, + "step": 8511 + }, + { + "epoch": 3.14, + "learning_rate": 2.687326025372583e-06, + "loss": 0.5297, + "step": 8512 + }, + { + "epoch": 3.14, + "learning_rate": 2.6872890750092377e-06, + "loss": 0.6215, + "step": 8513 + }, + { + "epoch": 3.14, + "learning_rate": 2.6872521246458923e-06, + "loss": 0.5373, + "step": 8514 + }, + { + "epoch": 3.14, + "learning_rate": 2.687215174282547e-06, + "loss": 0.5884, + "step": 8515 + }, + { + "epoch": 3.14, + "learning_rate": 2.687178223919202e-06, + "loss": 0.5256, + "step": 8516 + }, + { + "epoch": 3.15, + "learning_rate": 2.6871412735558567e-06, + "loss": 0.6425, + "step": 8517 + }, + { + "epoch": 3.15, + "learning_rate": 2.6871043231925117e-06, + "loss": 0.5123, + "step": 8518 + }, + { + "epoch": 3.15, + "learning_rate": 2.6870673728291664e-06, + "loss": 0.5002, + "step": 8519 + }, + { + "epoch": 3.15, + "learning_rate": 2.687030422465821e-06, + "loss": 0.6072, + "step": 8520 + }, + { + "epoch": 3.15, + "learning_rate": 2.6869934721024756e-06, + "loss": 0.636, + "step": 8521 + }, + { + "epoch": 3.15, + "learning_rate": 2.6869565217391307e-06, + "loss": 0.649, + "step": 8522 + }, + { + "epoch": 3.15, + "learning_rate": 2.6869195713757853e-06, + "loss": 0.579, + "step": 8523 + }, + { + "epoch": 3.15, + "learning_rate": 2.68688262101244e-06, + "loss": 0.5057, + "step": 8524 + }, + { + "epoch": 3.15, + "learning_rate": 2.6868456706490946e-06, + "loss": 0.5426, + "step": 8525 + }, + { + "epoch": 3.15, + "learning_rate": 2.6868087202857492e-06, + "loss": 0.507, + "step": 8526 + }, + { + "epoch": 3.15, + "learning_rate": 2.6867717699224043e-06, + "loss": 0.6115, + "step": 8527 + }, + { + "epoch": 3.15, + "learning_rate": 2.686734819559059e-06, + "loss": 0.5788, + "step": 8528 + }, + { + "epoch": 3.15, + "learning_rate": 2.686697869195714e-06, + "loss": 0.5611, + "step": 8529 + }, + { + "epoch": 3.15, + "learning_rate": 2.6866609188323686e-06, + "loss": 0.5005, + "step": 8530 + }, + { + "epoch": 3.15, + "learning_rate": 2.6866239684690233e-06, + "loss": 0.6396, + "step": 8531 + }, + { + "epoch": 3.15, + "learning_rate": 2.686587018105678e-06, + "loss": 0.5522, + "step": 8532 + }, + { + "epoch": 3.15, + "learning_rate": 2.686550067742333e-06, + "loss": 0.5583, + "step": 8533 + }, + { + "epoch": 3.15, + "learning_rate": 2.6865131173789876e-06, + "loss": 0.5522, + "step": 8534 + }, + { + "epoch": 3.15, + "learning_rate": 2.6864761670156422e-06, + "loss": 0.5193, + "step": 8535 + }, + { + "epoch": 3.15, + "learning_rate": 2.6864392166522973e-06, + "loss": 0.5048, + "step": 8536 + }, + { + "epoch": 3.15, + "learning_rate": 2.686402266288952e-06, + "loss": 0.5974, + "step": 8537 + }, + { + "epoch": 3.15, + "learning_rate": 2.6863653159256066e-06, + "loss": 0.6047, + "step": 8538 + }, + { + "epoch": 3.15, + "learning_rate": 2.686328365562261e-06, + "loss": 0.5769, + "step": 8539 + }, + { + "epoch": 3.15, + "learning_rate": 2.6862914151989163e-06, + "loss": 0.5707, + "step": 8540 + }, + { + "epoch": 3.15, + "learning_rate": 2.686254464835571e-06, + "loss": 0.497, + "step": 8541 + }, + { + "epoch": 3.15, + "learning_rate": 2.686217514472226e-06, + "loss": 0.5044, + "step": 8542 + }, + { + "epoch": 3.15, + "learning_rate": 2.6861805641088806e-06, + "loss": 0.5242, + "step": 8543 + }, + { + "epoch": 3.16, + "learning_rate": 2.6861436137455352e-06, + "loss": 0.5559, + "step": 8544 + }, + { + "epoch": 3.16, + "learning_rate": 2.68610666338219e-06, + "loss": 0.5144, + "step": 8545 + }, + { + "epoch": 3.16, + "learning_rate": 2.6860697130188445e-06, + "loss": 0.5848, + "step": 8546 + }, + { + "epoch": 3.16, + "learning_rate": 2.6860327626554996e-06, + "loss": 0.7599, + "step": 8547 + }, + { + "epoch": 3.16, + "learning_rate": 2.685995812292154e-06, + "loss": 0.5447, + "step": 8548 + }, + { + "epoch": 3.16, + "learning_rate": 2.6859588619288093e-06, + "loss": 0.5659, + "step": 8549 + }, + { + "epoch": 3.16, + "learning_rate": 2.685921911565464e-06, + "loss": 0.5908, + "step": 8550 + }, + { + "epoch": 3.16, + "learning_rate": 2.6858849612021185e-06, + "loss": 0.5493, + "step": 8551 + }, + { + "epoch": 3.16, + "learning_rate": 2.685848010838773e-06, + "loss": 0.6076, + "step": 8552 + }, + { + "epoch": 3.16, + "learning_rate": 2.6858110604754282e-06, + "loss": 0.4878, + "step": 8553 + }, + { + "epoch": 3.16, + "learning_rate": 2.685774110112083e-06, + "loss": 0.5583, + "step": 8554 + }, + { + "epoch": 3.16, + "learning_rate": 2.6857371597487375e-06, + "loss": 0.6189, + "step": 8555 + }, + { + "epoch": 3.16, + "learning_rate": 2.6857002093853926e-06, + "loss": 0.6233, + "step": 8556 + }, + { + "epoch": 3.16, + "learning_rate": 2.685663259022047e-06, + "loss": 0.5759, + "step": 8557 + }, + { + "epoch": 3.16, + "learning_rate": 2.685626308658702e-06, + "loss": 0.5848, + "step": 8558 + }, + { + "epoch": 3.16, + "learning_rate": 2.6855893582953565e-06, + "loss": 0.5592, + "step": 8559 + }, + { + "epoch": 3.16, + "learning_rate": 2.6855524079320115e-06, + "loss": 0.5179, + "step": 8560 + }, + { + "epoch": 3.16, + "learning_rate": 2.685515457568666e-06, + "loss": 0.4963, + "step": 8561 + }, + { + "epoch": 3.16, + "learning_rate": 2.6854785072053212e-06, + "loss": 0.5027, + "step": 8562 + }, + { + "epoch": 3.16, + "learning_rate": 2.685441556841976e-06, + "loss": 0.527, + "step": 8563 + }, + { + "epoch": 3.16, + "learning_rate": 2.68540460647863e-06, + "loss": 0.5098, + "step": 8564 + }, + { + "epoch": 3.16, + "learning_rate": 2.685367656115285e-06, + "loss": 0.5226, + "step": 8565 + }, + { + "epoch": 3.16, + "learning_rate": 2.6853307057519398e-06, + "loss": 0.6199, + "step": 8566 + }, + { + "epoch": 3.16, + "learning_rate": 2.685293755388595e-06, + "loss": 0.5388, + "step": 8567 + }, + { + "epoch": 3.16, + "learning_rate": 2.6852568050252495e-06, + "loss": 0.5754, + "step": 8568 + }, + { + "epoch": 3.16, + "learning_rate": 2.6852198546619045e-06, + "loss": 0.5357, + "step": 8569 + }, + { + "epoch": 3.16, + "learning_rate": 2.6851829042985587e-06, + "loss": 0.5712, + "step": 8570 + }, + { + "epoch": 3.17, + "learning_rate": 2.685145953935214e-06, + "loss": 0.5056, + "step": 8571 + }, + { + "epoch": 3.17, + "learning_rate": 2.6851090035718684e-06, + "loss": 0.7225, + "step": 8572 + }, + { + "epoch": 3.17, + "learning_rate": 2.6850720532085235e-06, + "loss": 0.4717, + "step": 8573 + }, + { + "epoch": 3.17, + "learning_rate": 2.685035102845178e-06, + "loss": 0.6779, + "step": 8574 + }, + { + "epoch": 3.17, + "learning_rate": 2.6849981524818328e-06, + "loss": 0.5827, + "step": 8575 + }, + { + "epoch": 3.17, + "learning_rate": 2.6849612021184874e-06, + "loss": 0.5926, + "step": 8576 + }, + { + "epoch": 3.17, + "learning_rate": 2.684924251755142e-06, + "loss": 0.5047, + "step": 8577 + }, + { + "epoch": 3.17, + "learning_rate": 2.684887301391797e-06, + "loss": 0.6276, + "step": 8578 + }, + { + "epoch": 3.17, + "learning_rate": 2.6848503510284517e-06, + "loss": 0.507, + "step": 8579 + }, + { + "epoch": 3.17, + "learning_rate": 2.684813400665107e-06, + "loss": 0.4824, + "step": 8580 + }, + { + "epoch": 3.17, + "learning_rate": 2.6847764503017614e-06, + "loss": 0.4972, + "step": 8581 + }, + { + "epoch": 3.17, + "learning_rate": 2.684739499938416e-06, + "loss": 0.5075, + "step": 8582 + }, + { + "epoch": 3.17, + "learning_rate": 2.6847025495750707e-06, + "loss": 0.6243, + "step": 8583 + }, + { + "epoch": 3.17, + "learning_rate": 2.6846655992117253e-06, + "loss": 0.5099, + "step": 8584 + }, + { + "epoch": 3.17, + "learning_rate": 2.6846286488483804e-06, + "loss": 0.6494, + "step": 8585 + }, + { + "epoch": 3.17, + "learning_rate": 2.684591698485035e-06, + "loss": 0.6456, + "step": 8586 + }, + { + "epoch": 3.17, + "learning_rate": 2.68455474812169e-06, + "loss": 0.497, + "step": 8587 + }, + { + "epoch": 3.17, + "learning_rate": 2.6845177977583447e-06, + "loss": 0.5164, + "step": 8588 + }, + { + "epoch": 3.17, + "learning_rate": 2.6844808473949994e-06, + "loss": 0.5804, + "step": 8589 + }, + { + "epoch": 3.17, + "learning_rate": 2.684443897031654e-06, + "loss": 0.5913, + "step": 8590 + }, + { + "epoch": 3.17, + "learning_rate": 2.684406946668309e-06, + "loss": 0.5937, + "step": 8591 + }, + { + "epoch": 3.17, + "learning_rate": 2.6843699963049637e-06, + "loss": 0.495, + "step": 8592 + }, + { + "epoch": 3.17, + "learning_rate": 2.6843330459416188e-06, + "loss": 0.5768, + "step": 8593 + }, + { + "epoch": 3.17, + "learning_rate": 2.6842960955782734e-06, + "loss": 0.5689, + "step": 8594 + }, + { + "epoch": 3.17, + "learning_rate": 2.684259145214928e-06, + "loss": 0.5974, + "step": 8595 + }, + { + "epoch": 3.17, + "learning_rate": 2.6842221948515827e-06, + "loss": 0.5648, + "step": 8596 + }, + { + "epoch": 3.17, + "learning_rate": 2.6841852444882373e-06, + "loss": 0.5219, + "step": 8597 + }, + { + "epoch": 3.18, + "learning_rate": 2.6841482941248924e-06, + "loss": 0.5139, + "step": 8598 + }, + { + "epoch": 3.18, + "learning_rate": 2.684111343761547e-06, + "loss": 0.5759, + "step": 8599 + }, + { + "epoch": 3.18, + "learning_rate": 2.684074393398202e-06, + "loss": 0.5599, + "step": 8600 + }, + { + "epoch": 3.18, + "learning_rate": 2.6840374430348567e-06, + "loss": 0.548, + "step": 8601 + }, + { + "epoch": 3.18, + "learning_rate": 2.6840004926715113e-06, + "loss": 0.5905, + "step": 8602 + }, + { + "epoch": 3.18, + "learning_rate": 2.683963542308166e-06, + "loss": 0.6649, + "step": 8603 + }, + { + "epoch": 3.18, + "learning_rate": 2.6839265919448206e-06, + "loss": 0.7093, + "step": 8604 + }, + { + "epoch": 3.18, + "learning_rate": 2.6838896415814757e-06, + "loss": 0.4812, + "step": 8605 + }, + { + "epoch": 3.18, + "learning_rate": 2.6838526912181303e-06, + "loss": 0.618, + "step": 8606 + }, + { + "epoch": 3.18, + "learning_rate": 2.6838157408547854e-06, + "loss": 0.6034, + "step": 8607 + }, + { + "epoch": 3.18, + "learning_rate": 2.68377879049144e-06, + "loss": 0.7002, + "step": 8608 + }, + { + "epoch": 3.18, + "learning_rate": 2.6837418401280946e-06, + "loss": 0.6164, + "step": 8609 + }, + { + "epoch": 3.18, + "learning_rate": 2.6837048897647493e-06, + "loss": 0.5233, + "step": 8610 + }, + { + "epoch": 3.18, + "learning_rate": 2.6836679394014043e-06, + "loss": 0.5955, + "step": 8611 + }, + { + "epoch": 3.18, + "learning_rate": 2.683630989038059e-06, + "loss": 0.5662, + "step": 8612 + }, + { + "epoch": 3.18, + "learning_rate": 2.683594038674714e-06, + "loss": 0.6224, + "step": 8613 + }, + { + "epoch": 3.18, + "learning_rate": 2.6835570883113687e-06, + "loss": 0.4895, + "step": 8614 + }, + { + "epoch": 3.18, + "learning_rate": 2.683520137948023e-06, + "loss": 0.4803, + "step": 8615 + }, + { + "epoch": 3.18, + "learning_rate": 2.683483187584678e-06, + "loss": 0.5966, + "step": 8616 + }, + { + "epoch": 3.18, + "learning_rate": 2.6834462372213326e-06, + "loss": 0.5713, + "step": 8617 + }, + { + "epoch": 3.18, + "learning_rate": 2.6834092868579876e-06, + "loss": 0.5474, + "step": 8618 + }, + { + "epoch": 3.18, + "learning_rate": 2.6833723364946423e-06, + "loss": 0.6595, + "step": 8619 + }, + { + "epoch": 3.18, + "learning_rate": 2.6833353861312973e-06, + "loss": 0.6896, + "step": 8620 + }, + { + "epoch": 3.18, + "learning_rate": 2.6832984357679516e-06, + "loss": 0.6494, + "step": 8621 + }, + { + "epoch": 3.18, + "learning_rate": 2.6832614854046066e-06, + "loss": 0.6167, + "step": 8622 + }, + { + "epoch": 3.18, + "learning_rate": 2.6832245350412613e-06, + "loss": 0.7099, + "step": 8623 + }, + { + "epoch": 3.18, + "learning_rate": 2.683187584677916e-06, + "loss": 0.5787, + "step": 8624 + }, + { + "epoch": 3.19, + "learning_rate": 2.683150634314571e-06, + "loss": 0.6323, + "step": 8625 + }, + { + "epoch": 3.19, + "learning_rate": 2.6831136839512256e-06, + "loss": 0.4858, + "step": 8626 + }, + { + "epoch": 3.19, + "learning_rate": 2.6830767335878802e-06, + "loss": 0.6192, + "step": 8627 + }, + { + "epoch": 3.19, + "learning_rate": 2.683039783224535e-06, + "loss": 0.5144, + "step": 8628 + }, + { + "epoch": 3.19, + "learning_rate": 2.68300283286119e-06, + "loss": 0.5459, + "step": 8629 + }, + { + "epoch": 3.19, + "learning_rate": 2.6829658824978446e-06, + "loss": 0.6165, + "step": 8630 + }, + { + "epoch": 3.19, + "learning_rate": 2.6829289321344996e-06, + "loss": 0.5943, + "step": 8631 + }, + { + "epoch": 3.19, + "learning_rate": 2.6828919817711543e-06, + "loss": 0.6398, + "step": 8632 + }, + { + "epoch": 3.19, + "learning_rate": 2.682855031407809e-06, + "loss": 0.6722, + "step": 8633 + }, + { + "epoch": 3.19, + "learning_rate": 2.6828180810444635e-06, + "loss": 0.6821, + "step": 8634 + }, + { + "epoch": 3.19, + "learning_rate": 2.682781130681118e-06, + "loss": 0.5571, + "step": 8635 + }, + { + "epoch": 3.19, + "learning_rate": 2.6827441803177732e-06, + "loss": 0.5735, + "step": 8636 + }, + { + "epoch": 3.19, + "learning_rate": 2.682707229954428e-06, + "loss": 0.5465, + "step": 8637 + }, + { + "epoch": 3.19, + "learning_rate": 2.682670279591083e-06, + "loss": 0.6143, + "step": 8638 + }, + { + "epoch": 3.19, + "learning_rate": 2.6826333292277376e-06, + "loss": 0.5424, + "step": 8639 + }, + { + "epoch": 3.19, + "learning_rate": 2.682596378864392e-06, + "loss": 0.5387, + "step": 8640 + }, + { + "epoch": 3.19, + "learning_rate": 2.682559428501047e-06, + "loss": 0.6023, + "step": 8641 + }, + { + "epoch": 3.19, + "learning_rate": 2.682522478137702e-06, + "loss": 0.5171, + "step": 8642 + }, + { + "epoch": 3.19, + "learning_rate": 2.6824855277743565e-06, + "loss": 0.4745, + "step": 8643 + }, + { + "epoch": 3.19, + "learning_rate": 2.682448577411011e-06, + "loss": 0.5711, + "step": 8644 + }, + { + "epoch": 3.19, + "learning_rate": 2.6824116270476662e-06, + "loss": 0.6159, + "step": 8645 + }, + { + "epoch": 3.19, + "learning_rate": 2.682374676684321e-06, + "loss": 0.4714, + "step": 8646 + }, + { + "epoch": 3.19, + "learning_rate": 2.6823377263209755e-06, + "loss": 0.5566, + "step": 8647 + }, + { + "epoch": 3.19, + "learning_rate": 2.68230077595763e-06, + "loss": 0.5136, + "step": 8648 + }, + { + "epoch": 3.19, + "learning_rate": 2.682263825594285e-06, + "loss": 0.5048, + "step": 8649 + }, + { + "epoch": 3.19, + "learning_rate": 2.68222687523094e-06, + "loss": 0.5951, + "step": 8650 + }, + { + "epoch": 3.19, + "learning_rate": 2.682189924867595e-06, + "loss": 0.6142, + "step": 8651 + }, + { + "epoch": 3.19, + "learning_rate": 2.6821529745042495e-06, + "loss": 0.6035, + "step": 8652 + }, + { + "epoch": 3.2, + "learning_rate": 2.682116024140904e-06, + "loss": 0.6433, + "step": 8653 + }, + { + "epoch": 3.2, + "learning_rate": 2.682079073777559e-06, + "loss": 0.5608, + "step": 8654 + }, + { + "epoch": 3.2, + "learning_rate": 2.6820421234142134e-06, + "loss": 0.6065, + "step": 8655 + }, + { + "epoch": 3.2, + "learning_rate": 2.6820051730508685e-06, + "loss": 0.6147, + "step": 8656 + }, + { + "epoch": 3.2, + "learning_rate": 2.681968222687523e-06, + "loss": 0.5539, + "step": 8657 + }, + { + "epoch": 3.2, + "learning_rate": 2.681931272324178e-06, + "loss": 0.5546, + "step": 8658 + }, + { + "epoch": 3.2, + "learning_rate": 2.681894321960833e-06, + "loss": 0.5441, + "step": 8659 + }, + { + "epoch": 3.2, + "learning_rate": 2.6818573715974875e-06, + "loss": 0.6052, + "step": 8660 + }, + { + "epoch": 3.2, + "learning_rate": 2.681820421234142e-06, + "loss": 0.5671, + "step": 8661 + }, + { + "epoch": 3.2, + "learning_rate": 2.681783470870797e-06, + "loss": 0.4861, + "step": 8662 + }, + { + "epoch": 3.2, + "learning_rate": 2.681746520507452e-06, + "loss": 0.5983, + "step": 8663 + }, + { + "epoch": 3.2, + "learning_rate": 2.6817095701441064e-06, + "loss": 0.5777, + "step": 8664 + }, + { + "epoch": 3.2, + "learning_rate": 2.6816726197807615e-06, + "loss": 0.5496, + "step": 8665 + }, + { + "epoch": 3.2, + "learning_rate": 2.6816356694174157e-06, + "loss": 0.5693, + "step": 8666 + }, + { + "epoch": 3.2, + "learning_rate": 2.6815987190540708e-06, + "loss": 0.571, + "step": 8667 + }, + { + "epoch": 3.2, + "learning_rate": 2.6815617686907254e-06, + "loss": 0.6708, + "step": 8668 + }, + { + "epoch": 3.2, + "learning_rate": 2.6815248183273805e-06, + "loss": 0.5745, + "step": 8669 + }, + { + "epoch": 3.2, + "learning_rate": 2.681487867964035e-06, + "loss": 0.5247, + "step": 8670 + }, + { + "epoch": 3.2, + "learning_rate": 2.68145091760069e-06, + "loss": 0.6413, + "step": 8671 + }, + { + "epoch": 3.2, + "learning_rate": 2.6814139672373444e-06, + "loss": 0.4794, + "step": 8672 + }, + { + "epoch": 3.2, + "learning_rate": 2.681377016873999e-06, + "loss": 0.5819, + "step": 8673 + }, + { + "epoch": 3.2, + "learning_rate": 2.681340066510654e-06, + "loss": 0.5174, + "step": 8674 + }, + { + "epoch": 3.2, + "learning_rate": 2.6813031161473087e-06, + "loss": 0.6449, + "step": 8675 + }, + { + "epoch": 3.2, + "learning_rate": 2.6812661657839638e-06, + "loss": 0.5852, + "step": 8676 + }, + { + "epoch": 3.2, + "learning_rate": 2.6812292154206184e-06, + "loss": 0.633, + "step": 8677 + }, + { + "epoch": 3.2, + "learning_rate": 2.681192265057273e-06, + "loss": 0.5002, + "step": 8678 + }, + { + "epoch": 3.2, + "learning_rate": 2.6811553146939277e-06, + "loss": 0.5384, + "step": 8679 + }, + { + "epoch": 3.21, + "learning_rate": 2.6811183643305827e-06, + "loss": 0.432, + "step": 8680 + }, + { + "epoch": 3.21, + "learning_rate": 2.6810814139672374e-06, + "loss": 0.5099, + "step": 8681 + }, + { + "epoch": 3.21, + "learning_rate": 2.6810444636038924e-06, + "loss": 0.5198, + "step": 8682 + }, + { + "epoch": 3.21, + "learning_rate": 2.681007513240547e-06, + "loss": 0.5389, + "step": 8683 + }, + { + "epoch": 3.21, + "learning_rate": 2.6809705628772017e-06, + "loss": 0.4915, + "step": 8684 + }, + { + "epoch": 3.21, + "learning_rate": 2.6809336125138563e-06, + "loss": 0.5958, + "step": 8685 + }, + { + "epoch": 3.21, + "learning_rate": 2.680896662150511e-06, + "loss": 0.598, + "step": 8686 + }, + { + "epoch": 3.21, + "learning_rate": 2.680859711787166e-06, + "loss": 0.6442, + "step": 8687 + }, + { + "epoch": 3.21, + "learning_rate": 2.6808227614238207e-06, + "loss": 0.4999, + "step": 8688 + }, + { + "epoch": 3.21, + "learning_rate": 2.6807858110604757e-06, + "loss": 0.5611, + "step": 8689 + }, + { + "epoch": 3.21, + "learning_rate": 2.6807488606971304e-06, + "loss": 0.5672, + "step": 8690 + }, + { + "epoch": 3.21, + "learning_rate": 2.680711910333785e-06, + "loss": 0.5039, + "step": 8691 + }, + { + "epoch": 3.21, + "learning_rate": 2.6806749599704396e-06, + "loss": 0.4435, + "step": 8692 + }, + { + "epoch": 3.21, + "learning_rate": 2.6806380096070943e-06, + "loss": 0.6329, + "step": 8693 + }, + { + "epoch": 3.21, + "learning_rate": 2.6806010592437493e-06, + "loss": 0.5414, + "step": 8694 + }, + { + "epoch": 3.21, + "learning_rate": 2.680564108880404e-06, + "loss": 0.5124, + "step": 8695 + }, + { + "epoch": 3.21, + "learning_rate": 2.680527158517059e-06, + "loss": 0.4848, + "step": 8696 + }, + { + "epoch": 3.21, + "learning_rate": 2.6804902081537137e-06, + "loss": 0.6093, + "step": 8697 + }, + { + "epoch": 3.21, + "learning_rate": 2.6804532577903683e-06, + "loss": 0.5146, + "step": 8698 + }, + { + "epoch": 3.21, + "learning_rate": 2.680416307427023e-06, + "loss": 0.487, + "step": 8699 + }, + { + "epoch": 3.21, + "learning_rate": 2.680379357063678e-06, + "loss": 0.5199, + "step": 8700 + }, + { + "epoch": 3.21, + "learning_rate": 2.6803424067003326e-06, + "loss": 0.6197, + "step": 8701 + }, + { + "epoch": 3.21, + "learning_rate": 2.6803054563369877e-06, + "loss": 0.6038, + "step": 8702 + }, + { + "epoch": 3.21, + "learning_rate": 2.6802685059736423e-06, + "loss": 0.5387, + "step": 8703 + }, + { + "epoch": 3.21, + "learning_rate": 2.680231555610297e-06, + "loss": 0.4655, + "step": 8704 + }, + { + "epoch": 3.21, + "learning_rate": 2.6801946052469516e-06, + "loss": 0.5902, + "step": 8705 + }, + { + "epoch": 3.21, + "learning_rate": 2.6801576548836062e-06, + "loss": 0.5659, + "step": 8706 + }, + { + "epoch": 3.22, + "learning_rate": 2.6801207045202613e-06, + "loss": 0.5637, + "step": 8707 + }, + { + "epoch": 3.22, + "learning_rate": 2.680083754156916e-06, + "loss": 0.5482, + "step": 8708 + }, + { + "epoch": 3.22, + "learning_rate": 2.680046803793571e-06, + "loss": 0.6111, + "step": 8709 + }, + { + "epoch": 3.22, + "learning_rate": 2.680009853430225e-06, + "loss": 0.7074, + "step": 8710 + }, + { + "epoch": 3.22, + "learning_rate": 2.6799729030668803e-06, + "loss": 0.5516, + "step": 8711 + }, + { + "epoch": 3.22, + "learning_rate": 2.679935952703535e-06, + "loss": 0.4692, + "step": 8712 + }, + { + "epoch": 3.22, + "learning_rate": 2.6798990023401895e-06, + "loss": 0.6332, + "step": 8713 + }, + { + "epoch": 3.22, + "learning_rate": 2.6798620519768446e-06, + "loss": 0.6207, + "step": 8714 + }, + { + "epoch": 3.22, + "learning_rate": 2.6798251016134992e-06, + "loss": 0.5263, + "step": 8715 + }, + { + "epoch": 3.22, + "learning_rate": 2.679788151250154e-06, + "loss": 0.4439, + "step": 8716 + }, + { + "epoch": 3.22, + "learning_rate": 2.6797512008868085e-06, + "loss": 0.5466, + "step": 8717 + }, + { + "epoch": 3.22, + "learning_rate": 2.6797142505234636e-06, + "loss": 0.6487, + "step": 8718 + }, + { + "epoch": 3.22, + "learning_rate": 2.679677300160118e-06, + "loss": 0.5936, + "step": 8719 + }, + { + "epoch": 3.22, + "learning_rate": 2.6796403497967733e-06, + "loss": 0.51, + "step": 8720 + }, + { + "epoch": 3.22, + "learning_rate": 2.679603399433428e-06, + "loss": 0.4879, + "step": 8721 + }, + { + "epoch": 3.22, + "learning_rate": 2.6795664490700825e-06, + "loss": 0.5847, + "step": 8722 + }, + { + "epoch": 3.22, + "learning_rate": 2.679529498706737e-06, + "loss": 0.568, + "step": 8723 + }, + { + "epoch": 3.22, + "learning_rate": 2.679492548343392e-06, + "loss": 0.5942, + "step": 8724 + }, + { + "epoch": 3.22, + "learning_rate": 2.679455597980047e-06, + "loss": 0.5462, + "step": 8725 + }, + { + "epoch": 3.22, + "learning_rate": 2.6794186476167015e-06, + "loss": 0.5939, + "step": 8726 + }, + { + "epoch": 3.22, + "learning_rate": 2.6793816972533566e-06, + "loss": 0.5193, + "step": 8727 + }, + { + "epoch": 3.22, + "learning_rate": 2.679344746890011e-06, + "loss": 0.6035, + "step": 8728 + }, + { + "epoch": 3.22, + "learning_rate": 2.679307796526666e-06, + "loss": 0.5664, + "step": 8729 + }, + { + "epoch": 3.22, + "learning_rate": 2.6792708461633205e-06, + "loss": 0.6085, + "step": 8730 + }, + { + "epoch": 3.22, + "learning_rate": 2.6792338957999755e-06, + "loss": 0.5214, + "step": 8731 + }, + { + "epoch": 3.22, + "learning_rate": 2.67919694543663e-06, + "loss": 0.4942, + "step": 8732 + }, + { + "epoch": 3.22, + "learning_rate": 2.679159995073285e-06, + "loss": 0.5114, + "step": 8733 + }, + { + "epoch": 3.23, + "learning_rate": 2.67912304470994e-06, + "loss": 0.5993, + "step": 8734 + }, + { + "epoch": 3.23, + "learning_rate": 2.6790860943465945e-06, + "loss": 0.5566, + "step": 8735 + }, + { + "epoch": 3.23, + "learning_rate": 2.679049143983249e-06, + "loss": 0.5567, + "step": 8736 + }, + { + "epoch": 3.23, + "learning_rate": 2.6790121936199038e-06, + "loss": 0.5191, + "step": 8737 + }, + { + "epoch": 3.23, + "learning_rate": 2.678975243256559e-06, + "loss": 0.5312, + "step": 8738 + }, + { + "epoch": 3.23, + "learning_rate": 2.6789382928932135e-06, + "loss": 0.5029, + "step": 8739 + }, + { + "epoch": 3.23, + "learning_rate": 2.6789013425298685e-06, + "loss": 0.6371, + "step": 8740 + }, + { + "epoch": 3.23, + "learning_rate": 2.678864392166523e-06, + "loss": 0.5115, + "step": 8741 + }, + { + "epoch": 3.23, + "learning_rate": 2.678827441803178e-06, + "loss": 0.6623, + "step": 8742 + }, + { + "epoch": 3.23, + "learning_rate": 2.6787904914398324e-06, + "loss": 0.5444, + "step": 8743 + }, + { + "epoch": 3.23, + "learning_rate": 2.678753541076487e-06, + "loss": 0.4872, + "step": 8744 + }, + { + "epoch": 3.23, + "learning_rate": 2.678716590713142e-06, + "loss": 0.5084, + "step": 8745 + }, + { + "epoch": 3.23, + "learning_rate": 2.6786796403497968e-06, + "loss": 0.6627, + "step": 8746 + }, + { + "epoch": 3.23, + "learning_rate": 2.678642689986452e-06, + "loss": 0.5672, + "step": 8747 + }, + { + "epoch": 3.23, + "learning_rate": 2.6786057396231065e-06, + "loss": 0.5512, + "step": 8748 + }, + { + "epoch": 3.23, + "learning_rate": 2.678568789259761e-06, + "loss": 0.581, + "step": 8749 + }, + { + "epoch": 3.23, + "learning_rate": 2.6785318388964157e-06, + "loss": 0.5617, + "step": 8750 + }, + { + "epoch": 3.23, + "learning_rate": 2.678494888533071e-06, + "loss": 0.6028, + "step": 8751 + }, + { + "epoch": 3.23, + "learning_rate": 2.6784579381697254e-06, + "loss": 0.5264, + "step": 8752 + }, + { + "epoch": 3.23, + "learning_rate": 2.67842098780638e-06, + "loss": 0.5565, + "step": 8753 + }, + { + "epoch": 3.23, + "learning_rate": 2.678384037443035e-06, + "loss": 0.6523, + "step": 8754 + }, + { + "epoch": 3.23, + "learning_rate": 2.6783470870796894e-06, + "loss": 0.5315, + "step": 8755 + }, + { + "epoch": 3.23, + "learning_rate": 2.6783101367163444e-06, + "loss": 0.5677, + "step": 8756 + }, + { + "epoch": 3.23, + "learning_rate": 2.678273186352999e-06, + "loss": 0.5857, + "step": 8757 + }, + { + "epoch": 3.23, + "learning_rate": 2.678236235989654e-06, + "loss": 0.5486, + "step": 8758 + }, + { + "epoch": 3.23, + "learning_rate": 2.6781992856263087e-06, + "loss": 0.5623, + "step": 8759 + }, + { + "epoch": 3.23, + "learning_rate": 2.678162335262964e-06, + "loss": 0.5437, + "step": 8760 + }, + { + "epoch": 3.24, + "learning_rate": 2.678125384899618e-06, + "loss": 0.6628, + "step": 8761 + }, + { + "epoch": 3.24, + "learning_rate": 2.678088434536273e-06, + "loss": 0.5449, + "step": 8762 + }, + { + "epoch": 3.24, + "learning_rate": 2.6780514841729277e-06, + "loss": 0.5569, + "step": 8763 + }, + { + "epoch": 3.24, + "learning_rate": 2.6780145338095824e-06, + "loss": 0.5269, + "step": 8764 + }, + { + "epoch": 3.24, + "learning_rate": 2.6779775834462374e-06, + "loss": 0.5094, + "step": 8765 + }, + { + "epoch": 3.24, + "learning_rate": 2.677940633082892e-06, + "loss": 0.5099, + "step": 8766 + }, + { + "epoch": 3.24, + "learning_rate": 2.6779036827195467e-06, + "loss": 0.6331, + "step": 8767 + }, + { + "epoch": 3.24, + "learning_rate": 2.6778667323562013e-06, + "loss": 0.6323, + "step": 8768 + }, + { + "epoch": 3.24, + "learning_rate": 2.6778297819928564e-06, + "loss": 0.6491, + "step": 8769 + }, + { + "epoch": 3.24, + "learning_rate": 2.677792831629511e-06, + "loss": 0.5772, + "step": 8770 + }, + { + "epoch": 3.24, + "learning_rate": 2.677755881266166e-06, + "loss": 0.5227, + "step": 8771 + }, + { + "epoch": 3.24, + "learning_rate": 2.6777189309028207e-06, + "loss": 0.6368, + "step": 8772 + }, + { + "epoch": 3.24, + "learning_rate": 2.6776819805394754e-06, + "loss": 0.569, + "step": 8773 + }, + { + "epoch": 3.24, + "learning_rate": 2.67764503017613e-06, + "loss": 0.6955, + "step": 8774 + }, + { + "epoch": 3.24, + "learning_rate": 2.6776080798127846e-06, + "loss": 0.5457, + "step": 8775 + }, + { + "epoch": 3.24, + "learning_rate": 2.6775711294494397e-06, + "loss": 0.5373, + "step": 8776 + }, + { + "epoch": 3.24, + "learning_rate": 2.6775341790860943e-06, + "loss": 0.5479, + "step": 8777 + }, + { + "epoch": 3.24, + "learning_rate": 2.6774972287227494e-06, + "loss": 0.5029, + "step": 8778 + }, + { + "epoch": 3.24, + "learning_rate": 2.677460278359404e-06, + "loss": 0.6292, + "step": 8779 + }, + { + "epoch": 3.24, + "learning_rate": 2.6774233279960587e-06, + "loss": 0.5685, + "step": 8780 + }, + { + "epoch": 3.24, + "learning_rate": 2.6773863776327133e-06, + "loss": 0.5092, + "step": 8781 + }, + { + "epoch": 3.24, + "learning_rate": 2.6773494272693684e-06, + "loss": 0.54, + "step": 8782 + }, + { + "epoch": 3.24, + "learning_rate": 2.677312476906023e-06, + "loss": 0.4858, + "step": 8783 + }, + { + "epoch": 3.24, + "learning_rate": 2.6772755265426776e-06, + "loss": 0.5078, + "step": 8784 + }, + { + "epoch": 3.24, + "learning_rate": 2.6772385761793327e-06, + "loss": 0.5743, + "step": 8785 + }, + { + "epoch": 3.24, + "learning_rate": 2.6772016258159873e-06, + "loss": 0.5221, + "step": 8786 + }, + { + "epoch": 3.24, + "learning_rate": 2.677164675452642e-06, + "loss": 0.4871, + "step": 8787 + }, + { + "epoch": 3.25, + "learning_rate": 2.6771277250892966e-06, + "loss": 0.6055, + "step": 8788 + }, + { + "epoch": 3.25, + "learning_rate": 2.6770907747259517e-06, + "loss": 0.5919, + "step": 8789 + }, + { + "epoch": 3.25, + "learning_rate": 2.6770538243626063e-06, + "loss": 0.5345, + "step": 8790 + }, + { + "epoch": 3.25, + "learning_rate": 2.6770168739992613e-06, + "loss": 0.6692, + "step": 8791 + }, + { + "epoch": 3.25, + "learning_rate": 2.676979923635916e-06, + "loss": 0.5694, + "step": 8792 + }, + { + "epoch": 3.25, + "learning_rate": 2.6769429732725706e-06, + "loss": 0.4921, + "step": 8793 + }, + { + "epoch": 3.25, + "learning_rate": 2.6769060229092253e-06, + "loss": 0.6034, + "step": 8794 + }, + { + "epoch": 3.25, + "learning_rate": 2.67686907254588e-06, + "loss": 0.5293, + "step": 8795 + }, + { + "epoch": 3.25, + "learning_rate": 2.676832122182535e-06, + "loss": 0.5862, + "step": 8796 + }, + { + "epoch": 3.25, + "learning_rate": 2.6767951718191896e-06, + "loss": 0.7126, + "step": 8797 + }, + { + "epoch": 3.25, + "learning_rate": 2.6767582214558447e-06, + "loss": 0.513, + "step": 8798 + }, + { + "epoch": 3.25, + "learning_rate": 2.6767212710924993e-06, + "loss": 0.605, + "step": 8799 + }, + { + "epoch": 3.25, + "learning_rate": 2.676684320729154e-06, + "loss": 0.5548, + "step": 8800 + }, + { + "epoch": 3.25, + "learning_rate": 2.6766473703658086e-06, + "loss": 0.577, + "step": 8801 + }, + { + "epoch": 3.25, + "learning_rate": 2.6766104200024636e-06, + "loss": 0.6138, + "step": 8802 + }, + { + "epoch": 3.25, + "learning_rate": 2.6765734696391183e-06, + "loss": 0.5921, + "step": 8803 + }, + { + "epoch": 3.25, + "learning_rate": 2.676536519275773e-06, + "loss": 0.4392, + "step": 8804 + }, + { + "epoch": 3.25, + "learning_rate": 2.676499568912428e-06, + "loss": 0.5096, + "step": 8805 + }, + { + "epoch": 3.25, + "learning_rate": 2.676462618549082e-06, + "loss": 0.5668, + "step": 8806 + }, + { + "epoch": 3.25, + "learning_rate": 2.6764256681857372e-06, + "loss": 0.5727, + "step": 8807 + }, + { + "epoch": 3.25, + "learning_rate": 2.676388717822392e-06, + "loss": 0.5462, + "step": 8808 + }, + { + "epoch": 3.25, + "learning_rate": 2.676351767459047e-06, + "loss": 0.4251, + "step": 8809 + }, + { + "epoch": 3.25, + "learning_rate": 2.6763148170957016e-06, + "loss": 0.5144, + "step": 8810 + }, + { + "epoch": 3.25, + "learning_rate": 2.6762778667323566e-06, + "loss": 0.4664, + "step": 8811 + }, + { + "epoch": 3.25, + "learning_rate": 2.676240916369011e-06, + "loss": 0.5242, + "step": 8812 + }, + { + "epoch": 3.25, + "learning_rate": 2.6762039660056655e-06, + "loss": 0.5298, + "step": 8813 + }, + { + "epoch": 3.25, + "learning_rate": 2.6761670156423205e-06, + "loss": 0.5588, + "step": 8814 + }, + { + "epoch": 3.26, + "learning_rate": 2.676130065278975e-06, + "loss": 0.4579, + "step": 8815 + }, + { + "epoch": 3.26, + "learning_rate": 2.6760931149156302e-06, + "loss": 0.6042, + "step": 8816 + }, + { + "epoch": 3.26, + "learning_rate": 2.676056164552285e-06, + "loss": 0.4944, + "step": 8817 + }, + { + "epoch": 3.26, + "learning_rate": 2.6760192141889395e-06, + "loss": 0.4415, + "step": 8818 + }, + { + "epoch": 3.26, + "learning_rate": 2.675982263825594e-06, + "loss": 0.5541, + "step": 8819 + }, + { + "epoch": 3.26, + "learning_rate": 2.675945313462249e-06, + "loss": 0.5339, + "step": 8820 + }, + { + "epoch": 3.26, + "learning_rate": 2.675908363098904e-06, + "loss": 0.5481, + "step": 8821 + }, + { + "epoch": 3.26, + "learning_rate": 2.675871412735559e-06, + "loss": 0.5741, + "step": 8822 + }, + { + "epoch": 3.26, + "learning_rate": 2.6758344623722135e-06, + "loss": 0.6363, + "step": 8823 + }, + { + "epoch": 3.26, + "learning_rate": 2.675797512008868e-06, + "loss": 0.6197, + "step": 8824 + }, + { + "epoch": 3.26, + "learning_rate": 2.675760561645523e-06, + "loss": 0.6264, + "step": 8825 + }, + { + "epoch": 3.26, + "learning_rate": 2.6757236112821774e-06, + "loss": 0.5668, + "step": 8826 + }, + { + "epoch": 3.26, + "learning_rate": 2.6756866609188325e-06, + "loss": 0.5296, + "step": 8827 + }, + { + "epoch": 3.26, + "learning_rate": 2.675649710555487e-06, + "loss": 0.5354, + "step": 8828 + }, + { + "epoch": 3.26, + "learning_rate": 2.675612760192142e-06, + "loss": 0.5237, + "step": 8829 + }, + { + "epoch": 3.26, + "learning_rate": 2.675575809828797e-06, + "loss": 0.5123, + "step": 8830 + }, + { + "epoch": 3.26, + "learning_rate": 2.6755388594654515e-06, + "loss": 0.5973, + "step": 8831 + }, + { + "epoch": 3.26, + "learning_rate": 2.675501909102106e-06, + "loss": 0.4797, + "step": 8832 + }, + { + "epoch": 3.26, + "learning_rate": 2.6754649587387607e-06, + "loss": 0.5641, + "step": 8833 + }, + { + "epoch": 3.26, + "learning_rate": 2.675428008375416e-06, + "loss": 0.5003, + "step": 8834 + }, + { + "epoch": 3.26, + "learning_rate": 2.6753910580120704e-06, + "loss": 0.6346, + "step": 8835 + }, + { + "epoch": 3.26, + "learning_rate": 2.6753541076487255e-06, + "loss": 0.621, + "step": 8836 + }, + { + "epoch": 3.26, + "learning_rate": 2.67531715728538e-06, + "loss": 0.6586, + "step": 8837 + }, + { + "epoch": 3.26, + "learning_rate": 2.6752802069220348e-06, + "loss": 0.5068, + "step": 8838 + }, + { + "epoch": 3.26, + "learning_rate": 2.6752432565586894e-06, + "loss": 0.4706, + "step": 8839 + }, + { + "epoch": 3.26, + "learning_rate": 2.6752063061953445e-06, + "loss": 0.5801, + "step": 8840 + }, + { + "epoch": 3.26, + "learning_rate": 2.675169355831999e-06, + "loss": 0.5308, + "step": 8841 + }, + { + "epoch": 3.27, + "learning_rate": 2.675132405468654e-06, + "loss": 0.6211, + "step": 8842 + }, + { + "epoch": 3.27, + "learning_rate": 2.675095455105309e-06, + "loss": 0.5681, + "step": 8843 + }, + { + "epoch": 3.27, + "learning_rate": 2.6750585047419634e-06, + "loss": 0.5421, + "step": 8844 + }, + { + "epoch": 3.27, + "learning_rate": 2.675021554378618e-06, + "loss": 0.4911, + "step": 8845 + }, + { + "epoch": 3.27, + "learning_rate": 2.6749846040152727e-06, + "loss": 0.5146, + "step": 8846 + }, + { + "epoch": 3.27, + "learning_rate": 2.6749476536519278e-06, + "loss": 0.6228, + "step": 8847 + }, + { + "epoch": 3.27, + "learning_rate": 2.6749107032885824e-06, + "loss": 0.5975, + "step": 8848 + }, + { + "epoch": 3.27, + "learning_rate": 2.6748737529252375e-06, + "loss": 0.5014, + "step": 8849 + }, + { + "epoch": 3.27, + "learning_rate": 2.674836802561892e-06, + "loss": 0.5052, + "step": 8850 + }, + { + "epoch": 3.27, + "learning_rate": 2.6747998521985467e-06, + "loss": 0.6181, + "step": 8851 + }, + { + "epoch": 3.27, + "learning_rate": 2.6747629018352014e-06, + "loss": 0.6044, + "step": 8852 + }, + { + "epoch": 3.27, + "learning_rate": 2.674725951471856e-06, + "loss": 0.5736, + "step": 8853 + }, + { + "epoch": 3.27, + "learning_rate": 2.674689001108511e-06, + "loss": 0.6622, + "step": 8854 + }, + { + "epoch": 3.27, + "learning_rate": 2.6746520507451657e-06, + "loss": 0.5248, + "step": 8855 + }, + { + "epoch": 3.27, + "learning_rate": 2.6746151003818208e-06, + "loss": 0.5087, + "step": 8856 + }, + { + "epoch": 3.27, + "learning_rate": 2.674578150018475e-06, + "loss": 0.5781, + "step": 8857 + }, + { + "epoch": 3.27, + "learning_rate": 2.67454119965513e-06, + "loss": 0.5727, + "step": 8858 + }, + { + "epoch": 3.27, + "learning_rate": 2.6745042492917847e-06, + "loss": 0.56, + "step": 8859 + }, + { + "epoch": 3.27, + "learning_rate": 2.6744672989284397e-06, + "loss": 0.5648, + "step": 8860 + }, + { + "epoch": 3.27, + "learning_rate": 2.6744303485650944e-06, + "loss": 0.4614, + "step": 8861 + }, + { + "epoch": 3.27, + "learning_rate": 2.674393398201749e-06, + "loss": 0.5269, + "step": 8862 + }, + { + "epoch": 3.27, + "learning_rate": 2.6743564478384036e-06, + "loss": 0.6112, + "step": 8863 + }, + { + "epoch": 3.27, + "learning_rate": 2.6743194974750583e-06, + "loss": 0.5434, + "step": 8864 + }, + { + "epoch": 3.27, + "learning_rate": 2.6742825471117133e-06, + "loss": 0.5918, + "step": 8865 + }, + { + "epoch": 3.27, + "learning_rate": 2.674245596748368e-06, + "loss": 0.5604, + "step": 8866 + }, + { + "epoch": 3.27, + "learning_rate": 2.674208646385023e-06, + "loss": 0.5561, + "step": 8867 + }, + { + "epoch": 3.27, + "learning_rate": 2.6741716960216777e-06, + "loss": 0.5562, + "step": 8868 + }, + { + "epoch": 3.28, + "learning_rate": 2.6741347456583323e-06, + "loss": 0.5435, + "step": 8869 + }, + { + "epoch": 3.28, + "learning_rate": 2.674097795294987e-06, + "loss": 0.5325, + "step": 8870 + }, + { + "epoch": 3.28, + "learning_rate": 2.674060844931642e-06, + "loss": 0.4823, + "step": 8871 + }, + { + "epoch": 3.28, + "learning_rate": 2.6740238945682966e-06, + "loss": 0.5433, + "step": 8872 + }, + { + "epoch": 3.28, + "learning_rate": 2.6739869442049513e-06, + "loss": 0.6704, + "step": 8873 + }, + { + "epoch": 3.28, + "learning_rate": 2.6739499938416063e-06, + "loss": 0.5856, + "step": 8874 + }, + { + "epoch": 3.28, + "learning_rate": 2.673913043478261e-06, + "loss": 0.5924, + "step": 8875 + }, + { + "epoch": 3.28, + "learning_rate": 2.6738760931149156e-06, + "loss": 0.4889, + "step": 8876 + }, + { + "epoch": 3.28, + "learning_rate": 2.6738391427515702e-06, + "loss": 0.5034, + "step": 8877 + }, + { + "epoch": 3.28, + "learning_rate": 2.6738021923882253e-06, + "loss": 0.5955, + "step": 8878 + }, + { + "epoch": 3.28, + "learning_rate": 2.67376524202488e-06, + "loss": 0.6007, + "step": 8879 + }, + { + "epoch": 3.28, + "learning_rate": 2.673728291661535e-06, + "loss": 0.5666, + "step": 8880 + }, + { + "epoch": 3.28, + "learning_rate": 2.6736913412981896e-06, + "loss": 0.6237, + "step": 8881 + }, + { + "epoch": 3.28, + "learning_rate": 2.6736543909348443e-06, + "loss": 0.5331, + "step": 8882 + }, + { + "epoch": 3.28, + "learning_rate": 2.673617440571499e-06, + "loss": 0.5392, + "step": 8883 + }, + { + "epoch": 3.28, + "learning_rate": 2.6735804902081535e-06, + "loss": 0.5772, + "step": 8884 + }, + { + "epoch": 3.28, + "learning_rate": 2.6735435398448086e-06, + "loss": 0.4883, + "step": 8885 + }, + { + "epoch": 3.28, + "learning_rate": 2.6735065894814632e-06, + "loss": 0.6699, + "step": 8886 + }, + { + "epoch": 3.28, + "learning_rate": 2.6734696391181183e-06, + "loss": 0.6638, + "step": 8887 + }, + { + "epoch": 3.28, + "learning_rate": 2.673432688754773e-06, + "loss": 0.5526, + "step": 8888 + }, + { + "epoch": 3.28, + "learning_rate": 2.6733957383914276e-06, + "loss": 0.6069, + "step": 8889 + }, + { + "epoch": 3.28, + "learning_rate": 2.6733587880280822e-06, + "loss": 0.5187, + "step": 8890 + }, + { + "epoch": 3.28, + "learning_rate": 2.6733218376647373e-06, + "loss": 0.6169, + "step": 8891 + }, + { + "epoch": 3.28, + "learning_rate": 2.673284887301392e-06, + "loss": 0.6468, + "step": 8892 + }, + { + "epoch": 3.28, + "learning_rate": 2.6732479369380465e-06, + "loss": 0.6252, + "step": 8893 + }, + { + "epoch": 3.28, + "learning_rate": 2.6732109865747016e-06, + "loss": 0.5781, + "step": 8894 + }, + { + "epoch": 3.28, + "learning_rate": 2.673174036211356e-06, + "loss": 0.5391, + "step": 8895 + }, + { + "epoch": 3.29, + "learning_rate": 2.673137085848011e-06, + "loss": 0.6612, + "step": 8896 + }, + { + "epoch": 3.29, + "learning_rate": 2.6731001354846655e-06, + "loss": 0.5408, + "step": 8897 + }, + { + "epoch": 3.29, + "learning_rate": 2.6730631851213206e-06, + "loss": 0.5989, + "step": 8898 + }, + { + "epoch": 3.29, + "learning_rate": 2.6730262347579752e-06, + "loss": 0.5017, + "step": 8899 + }, + { + "epoch": 3.29, + "learning_rate": 2.6729892843946303e-06, + "loss": 0.5911, + "step": 8900 + }, + { + "epoch": 3.29, + "learning_rate": 2.6729523340312845e-06, + "loss": 0.5677, + "step": 8901 + }, + { + "epoch": 3.29, + "learning_rate": 2.672915383667939e-06, + "loss": 0.5836, + "step": 8902 + }, + { + "epoch": 3.29, + "learning_rate": 2.672878433304594e-06, + "loss": 0.6299, + "step": 8903 + }, + { + "epoch": 3.29, + "learning_rate": 2.672841482941249e-06, + "loss": 0.6255, + "step": 8904 + }, + { + "epoch": 3.29, + "learning_rate": 2.672804532577904e-06, + "loss": 0.5119, + "step": 8905 + }, + { + "epoch": 3.29, + "learning_rate": 2.6727675822145585e-06, + "loss": 0.5347, + "step": 8906 + }, + { + "epoch": 3.29, + "learning_rate": 2.672730631851213e-06, + "loss": 0.5748, + "step": 8907 + }, + { + "epoch": 3.29, + "learning_rate": 2.6726936814878678e-06, + "loss": 0.4909, + "step": 8908 + }, + { + "epoch": 3.29, + "learning_rate": 2.672656731124523e-06, + "loss": 0.5862, + "step": 8909 + }, + { + "epoch": 3.29, + "learning_rate": 2.6726197807611775e-06, + "loss": 0.474, + "step": 8910 + }, + { + "epoch": 3.29, + "learning_rate": 2.6725828303978325e-06, + "loss": 0.4875, + "step": 8911 + }, + { + "epoch": 3.29, + "learning_rate": 2.672545880034487e-06, + "loss": 0.6427, + "step": 8912 + }, + { + "epoch": 3.29, + "learning_rate": 2.672508929671142e-06, + "loss": 0.5845, + "step": 8913 + }, + { + "epoch": 3.29, + "learning_rate": 2.6724719793077965e-06, + "loss": 0.5934, + "step": 8914 + }, + { + "epoch": 3.29, + "learning_rate": 2.672435028944451e-06, + "loss": 0.5846, + "step": 8915 + }, + { + "epoch": 3.29, + "learning_rate": 2.672398078581106e-06, + "loss": 0.743, + "step": 8916 + }, + { + "epoch": 3.29, + "learning_rate": 2.6723611282177608e-06, + "loss": 0.5657, + "step": 8917 + }, + { + "epoch": 3.29, + "learning_rate": 2.672324177854416e-06, + "loss": 0.5281, + "step": 8918 + }, + { + "epoch": 3.29, + "learning_rate": 2.6722872274910705e-06, + "loss": 0.5337, + "step": 8919 + }, + { + "epoch": 3.29, + "learning_rate": 2.672250277127725e-06, + "loss": 0.66, + "step": 8920 + }, + { + "epoch": 3.29, + "learning_rate": 2.6722133267643798e-06, + "loss": 0.5704, + "step": 8921 + }, + { + "epoch": 3.29, + "learning_rate": 2.6721763764010344e-06, + "loss": 0.6135, + "step": 8922 + }, + { + "epoch": 3.3, + "learning_rate": 2.6721394260376895e-06, + "loss": 0.6048, + "step": 8923 + }, + { + "epoch": 3.3, + "learning_rate": 2.672102475674344e-06, + "loss": 0.5567, + "step": 8924 + }, + { + "epoch": 3.3, + "learning_rate": 2.672065525310999e-06, + "loss": 0.5716, + "step": 8925 + }, + { + "epoch": 3.3, + "learning_rate": 2.6720285749476538e-06, + "loss": 0.528, + "step": 8926 + }, + { + "epoch": 3.3, + "learning_rate": 2.6719916245843084e-06, + "loss": 0.5646, + "step": 8927 + }, + { + "epoch": 3.3, + "learning_rate": 2.671954674220963e-06, + "loss": 0.568, + "step": 8928 + }, + { + "epoch": 3.3, + "learning_rate": 2.671917723857618e-06, + "loss": 0.5682, + "step": 8929 + }, + { + "epoch": 3.3, + "learning_rate": 2.6718807734942728e-06, + "loss": 0.607, + "step": 8930 + }, + { + "epoch": 3.3, + "learning_rate": 2.671843823130928e-06, + "loss": 0.5364, + "step": 8931 + }, + { + "epoch": 3.3, + "learning_rate": 2.6718068727675825e-06, + "loss": 0.5622, + "step": 8932 + }, + { + "epoch": 3.3, + "learning_rate": 2.671769922404237e-06, + "loss": 0.5236, + "step": 8933 + }, + { + "epoch": 3.3, + "learning_rate": 2.6717329720408917e-06, + "loss": 0.56, + "step": 8934 + }, + { + "epoch": 3.3, + "learning_rate": 2.6716960216775464e-06, + "loss": 0.6068, + "step": 8935 + }, + { + "epoch": 3.3, + "learning_rate": 2.6716590713142014e-06, + "loss": 0.5496, + "step": 8936 + }, + { + "epoch": 3.3, + "learning_rate": 2.671622120950856e-06, + "loss": 0.609, + "step": 8937 + }, + { + "epoch": 3.3, + "learning_rate": 2.671585170587511e-06, + "loss": 0.5383, + "step": 8938 + }, + { + "epoch": 3.3, + "learning_rate": 2.6715482202241658e-06, + "loss": 0.5657, + "step": 8939 + }, + { + "epoch": 3.3, + "learning_rate": 2.6715112698608204e-06, + "loss": 0.511, + "step": 8940 + }, + { + "epoch": 3.3, + "learning_rate": 2.671474319497475e-06, + "loss": 0.5852, + "step": 8941 + }, + { + "epoch": 3.3, + "learning_rate": 2.6714373691341297e-06, + "loss": 0.677, + "step": 8942 + }, + { + "epoch": 3.3, + "learning_rate": 2.6714004187707847e-06, + "loss": 0.4946, + "step": 8943 + }, + { + "epoch": 3.3, + "learning_rate": 2.6713634684074394e-06, + "loss": 0.5694, + "step": 8944 + }, + { + "epoch": 3.3, + "learning_rate": 2.6713265180440944e-06, + "loss": 0.6583, + "step": 8945 + }, + { + "epoch": 3.3, + "learning_rate": 2.6712895676807486e-06, + "loss": 0.5748, + "step": 8946 + }, + { + "epoch": 3.3, + "learning_rate": 2.6712526173174037e-06, + "loss": 0.6178, + "step": 8947 + }, + { + "epoch": 3.3, + "learning_rate": 2.6712156669540583e-06, + "loss": 0.4755, + "step": 8948 + }, + { + "epoch": 3.3, + "learning_rate": 2.6711787165907134e-06, + "loss": 0.6211, + "step": 8949 + }, + { + "epoch": 3.31, + "learning_rate": 2.671141766227368e-06, + "loss": 0.5322, + "step": 8950 + }, + { + "epoch": 3.31, + "learning_rate": 2.671104815864023e-06, + "loss": 0.5579, + "step": 8951 + }, + { + "epoch": 3.31, + "learning_rate": 2.6710678655006773e-06, + "loss": 0.6286, + "step": 8952 + }, + { + "epoch": 3.31, + "learning_rate": 2.671030915137332e-06, + "loss": 0.6073, + "step": 8953 + }, + { + "epoch": 3.31, + "learning_rate": 2.670993964773987e-06, + "loss": 0.515, + "step": 8954 + }, + { + "epoch": 3.31, + "learning_rate": 2.6709570144106416e-06, + "loss": 0.5887, + "step": 8955 + }, + { + "epoch": 3.31, + "learning_rate": 2.6709200640472967e-06, + "loss": 0.5971, + "step": 8956 + }, + { + "epoch": 3.31, + "learning_rate": 2.6708831136839513e-06, + "loss": 0.5462, + "step": 8957 + }, + { + "epoch": 3.31, + "learning_rate": 2.670846163320606e-06, + "loss": 0.5019, + "step": 8958 + }, + { + "epoch": 3.31, + "learning_rate": 2.6708092129572606e-06, + "loss": 0.5502, + "step": 8959 + }, + { + "epoch": 3.31, + "learning_rate": 2.6707722625939157e-06, + "loss": 0.6133, + "step": 8960 + }, + { + "epoch": 3.31, + "learning_rate": 2.6707353122305703e-06, + "loss": 0.6095, + "step": 8961 + }, + { + "epoch": 3.31, + "learning_rate": 2.670698361867225e-06, + "loss": 0.5164, + "step": 8962 + }, + { + "epoch": 3.31, + "learning_rate": 2.67066141150388e-06, + "loss": 0.5105, + "step": 8963 + }, + { + "epoch": 3.31, + "learning_rate": 2.6706244611405346e-06, + "loss": 0.5193, + "step": 8964 + }, + { + "epoch": 3.31, + "learning_rate": 2.6705875107771893e-06, + "loss": 0.532, + "step": 8965 + }, + { + "epoch": 3.31, + "learning_rate": 2.670550560413844e-06, + "loss": 0.5377, + "step": 8966 + }, + { + "epoch": 3.31, + "learning_rate": 2.670513610050499e-06, + "loss": 0.4886, + "step": 8967 + }, + { + "epoch": 3.31, + "learning_rate": 2.6704766596871536e-06, + "loss": 0.5178, + "step": 8968 + }, + { + "epoch": 3.31, + "learning_rate": 2.6704397093238087e-06, + "loss": 0.4747, + "step": 8969 + }, + { + "epoch": 3.31, + "learning_rate": 2.6704027589604633e-06, + "loss": 0.5365, + "step": 8970 + }, + { + "epoch": 3.31, + "learning_rate": 2.670365808597118e-06, + "loss": 0.544, + "step": 8971 + }, + { + "epoch": 3.31, + "learning_rate": 2.6703288582337726e-06, + "loss": 0.6037, + "step": 8972 + }, + { + "epoch": 3.31, + "learning_rate": 2.670291907870427e-06, + "loss": 0.5676, + "step": 8973 + }, + { + "epoch": 3.31, + "learning_rate": 2.6702549575070823e-06, + "loss": 0.619, + "step": 8974 + }, + { + "epoch": 3.31, + "learning_rate": 2.670218007143737e-06, + "loss": 0.6212, + "step": 8975 + }, + { + "epoch": 3.31, + "learning_rate": 2.670181056780392e-06, + "loss": 0.7133, + "step": 8976 + }, + { + "epoch": 3.31, + "learning_rate": 2.6701441064170466e-06, + "loss": 0.5653, + "step": 8977 + }, + { + "epoch": 3.32, + "learning_rate": 2.6701071560537012e-06, + "loss": 0.5827, + "step": 8978 + }, + { + "epoch": 3.32, + "learning_rate": 2.670070205690356e-06, + "loss": 0.5978, + "step": 8979 + }, + { + "epoch": 3.32, + "learning_rate": 2.670033255327011e-06, + "loss": 0.616, + "step": 8980 + }, + { + "epoch": 3.32, + "learning_rate": 2.6699963049636656e-06, + "loss": 0.5278, + "step": 8981 + }, + { + "epoch": 3.32, + "learning_rate": 2.66995935460032e-06, + "loss": 0.5836, + "step": 8982 + }, + { + "epoch": 3.32, + "learning_rate": 2.6699224042369753e-06, + "loss": 0.5236, + "step": 8983 + }, + { + "epoch": 3.32, + "learning_rate": 2.66988545387363e-06, + "loss": 0.6016, + "step": 8984 + }, + { + "epoch": 3.32, + "learning_rate": 2.6698485035102845e-06, + "loss": 0.5777, + "step": 8985 + }, + { + "epoch": 3.32, + "learning_rate": 2.669811553146939e-06, + "loss": 0.5157, + "step": 8986 + }, + { + "epoch": 3.32, + "learning_rate": 2.6697746027835942e-06, + "loss": 0.5072, + "step": 8987 + }, + { + "epoch": 3.32, + "learning_rate": 2.669737652420249e-06, + "loss": 0.6338, + "step": 8988 + }, + { + "epoch": 3.32, + "learning_rate": 2.669700702056904e-06, + "loss": 0.548, + "step": 8989 + }, + { + "epoch": 3.32, + "learning_rate": 2.6696637516935586e-06, + "loss": 0.5365, + "step": 8990 + }, + { + "epoch": 3.32, + "learning_rate": 2.669626801330213e-06, + "loss": 0.5473, + "step": 8991 + }, + { + "epoch": 3.32, + "learning_rate": 2.669589850966868e-06, + "loss": 0.583, + "step": 8992 + }, + { + "epoch": 3.32, + "learning_rate": 2.6695529006035225e-06, + "loss": 0.5218, + "step": 8993 + }, + { + "epoch": 3.32, + "learning_rate": 2.6695159502401775e-06, + "loss": 0.4928, + "step": 8994 + }, + { + "epoch": 3.32, + "learning_rate": 2.669478999876832e-06, + "loss": 0.5243, + "step": 8995 + }, + { + "epoch": 3.32, + "learning_rate": 2.6694420495134872e-06, + "loss": 0.6012, + "step": 8996 + }, + { + "epoch": 3.32, + "learning_rate": 2.6694050991501414e-06, + "loss": 0.6001, + "step": 8997 + }, + { + "epoch": 3.32, + "learning_rate": 2.6693681487867965e-06, + "loss": 0.6588, + "step": 8998 + }, + { + "epoch": 3.32, + "learning_rate": 2.669331198423451e-06, + "loss": 0.4861, + "step": 8999 + }, + { + "epoch": 3.32, + "learning_rate": 2.669294248060106e-06, + "loss": 0.5072, + "step": 9000 + }, + { + "epoch": 3.32, + "learning_rate": 2.669257297696761e-06, + "loss": 0.6342, + "step": 9001 + }, + { + "epoch": 3.32, + "learning_rate": 2.6692203473334155e-06, + "loss": 0.6148, + "step": 9002 + }, + { + "epoch": 3.32, + "learning_rate": 2.66918339697007e-06, + "loss": 0.5467, + "step": 9003 + }, + { + "epoch": 3.32, + "learning_rate": 2.6691464466067247e-06, + "loss": 0.5913, + "step": 9004 + }, + { + "epoch": 3.33, + "learning_rate": 2.66910949624338e-06, + "loss": 0.6468, + "step": 9005 + }, + { + "epoch": 3.33, + "learning_rate": 2.6690725458800344e-06, + "loss": 0.6461, + "step": 9006 + }, + { + "epoch": 3.33, + "learning_rate": 2.6690355955166895e-06, + "loss": 0.6468, + "step": 9007 + }, + { + "epoch": 3.33, + "learning_rate": 2.668998645153344e-06, + "loss": 0.5869, + "step": 9008 + }, + { + "epoch": 3.33, + "learning_rate": 2.6689616947899988e-06, + "loss": 0.6446, + "step": 9009 + }, + { + "epoch": 3.33, + "learning_rate": 2.6689247444266534e-06, + "loss": 0.5225, + "step": 9010 + }, + { + "epoch": 3.33, + "learning_rate": 2.6688877940633085e-06, + "loss": 0.5431, + "step": 9011 + }, + { + "epoch": 3.33, + "learning_rate": 2.668850843699963e-06, + "loss": 0.5792, + "step": 9012 + }, + { + "epoch": 3.33, + "learning_rate": 2.6688138933366177e-06, + "loss": 0.5906, + "step": 9013 + }, + { + "epoch": 3.33, + "learning_rate": 2.668776942973273e-06, + "loss": 0.4641, + "step": 9014 + }, + { + "epoch": 3.33, + "learning_rate": 2.6687399926099274e-06, + "loss": 0.5871, + "step": 9015 + }, + { + "epoch": 3.33, + "learning_rate": 2.668703042246582e-06, + "loss": 0.5325, + "step": 9016 + }, + { + "epoch": 3.33, + "learning_rate": 2.6686660918832367e-06, + "loss": 0.6741, + "step": 9017 + }, + { + "epoch": 3.33, + "learning_rate": 2.6686291415198918e-06, + "loss": 0.6285, + "step": 9018 + }, + { + "epoch": 3.33, + "learning_rate": 2.6685921911565464e-06, + "loss": 0.5432, + "step": 9019 + }, + { + "epoch": 3.33, + "learning_rate": 2.6685552407932015e-06, + "loss": 0.5385, + "step": 9020 + }, + { + "epoch": 3.33, + "learning_rate": 2.668518290429856e-06, + "loss": 0.5354, + "step": 9021 + }, + { + "epoch": 3.33, + "learning_rate": 2.6684813400665107e-06, + "loss": 0.5391, + "step": 9022 + }, + { + "epoch": 3.33, + "learning_rate": 2.6684443897031654e-06, + "loss": 0.6106, + "step": 9023 + }, + { + "epoch": 3.33, + "learning_rate": 2.66840743933982e-06, + "loss": 0.5002, + "step": 9024 + }, + { + "epoch": 3.33, + "learning_rate": 2.668370488976475e-06, + "loss": 0.5446, + "step": 9025 + }, + { + "epoch": 3.33, + "learning_rate": 2.6683335386131297e-06, + "loss": 0.4958, + "step": 9026 + }, + { + "epoch": 3.33, + "learning_rate": 2.6682965882497848e-06, + "loss": 0.5413, + "step": 9027 + }, + { + "epoch": 3.33, + "learning_rate": 2.6682596378864394e-06, + "loss": 0.5209, + "step": 9028 + }, + { + "epoch": 3.33, + "learning_rate": 2.668222687523094e-06, + "loss": 0.5679, + "step": 9029 + }, + { + "epoch": 3.33, + "learning_rate": 2.6681857371597487e-06, + "loss": 0.5378, + "step": 9030 + }, + { + "epoch": 3.33, + "learning_rate": 2.6681487867964037e-06, + "loss": 0.5734, + "step": 9031 + }, + { + "epoch": 3.34, + "learning_rate": 2.6681118364330584e-06, + "loss": 0.5417, + "step": 9032 + }, + { + "epoch": 3.34, + "learning_rate": 2.668074886069713e-06, + "loss": 0.5531, + "step": 9033 + }, + { + "epoch": 3.34, + "learning_rate": 2.668037935706368e-06, + "loss": 0.5771, + "step": 9034 + }, + { + "epoch": 3.34, + "learning_rate": 2.6680009853430227e-06, + "loss": 0.434, + "step": 9035 + }, + { + "epoch": 3.34, + "learning_rate": 2.6679640349796773e-06, + "loss": 0.512, + "step": 9036 + }, + { + "epoch": 3.34, + "learning_rate": 2.667927084616332e-06, + "loss": 0.5075, + "step": 9037 + }, + { + "epoch": 3.34, + "learning_rate": 2.667890134252987e-06, + "loss": 0.6095, + "step": 9038 + }, + { + "epoch": 3.34, + "learning_rate": 2.6678531838896417e-06, + "loss": 0.5195, + "step": 9039 + }, + { + "epoch": 3.34, + "learning_rate": 2.6678162335262967e-06, + "loss": 0.4949, + "step": 9040 + }, + { + "epoch": 3.34, + "learning_rate": 2.6677792831629514e-06, + "loss": 0.4987, + "step": 9041 + }, + { + "epoch": 3.34, + "learning_rate": 2.6677423327996056e-06, + "loss": 0.4532, + "step": 9042 + }, + { + "epoch": 3.34, + "learning_rate": 2.6677053824362606e-06, + "loss": 0.5459, + "step": 9043 + }, + { + "epoch": 3.34, + "learning_rate": 2.6676684320729153e-06, + "loss": 0.6692, + "step": 9044 + }, + { + "epoch": 3.34, + "learning_rate": 2.6676314817095703e-06, + "loss": 0.6229, + "step": 9045 + }, + { + "epoch": 3.34, + "learning_rate": 2.667594531346225e-06, + "loss": 0.5521, + "step": 9046 + }, + { + "epoch": 3.34, + "learning_rate": 2.66755758098288e-06, + "loss": 0.4575, + "step": 9047 + }, + { + "epoch": 3.34, + "learning_rate": 2.6675206306195343e-06, + "loss": 0.4498, + "step": 9048 + }, + { + "epoch": 3.34, + "learning_rate": 2.6674836802561893e-06, + "loss": 0.432, + "step": 9049 + }, + { + "epoch": 3.34, + "learning_rate": 2.667446729892844e-06, + "loss": 0.5625, + "step": 9050 + }, + { + "epoch": 3.34, + "learning_rate": 2.667409779529499e-06, + "loss": 0.4922, + "step": 9051 + }, + { + "epoch": 3.34, + "learning_rate": 2.6673728291661536e-06, + "loss": 0.567, + "step": 9052 + }, + { + "epoch": 3.34, + "learning_rate": 2.6673358788028083e-06, + "loss": 0.4728, + "step": 9053 + }, + { + "epoch": 3.34, + "learning_rate": 2.667298928439463e-06, + "loss": 0.6192, + "step": 9054 + }, + { + "epoch": 3.34, + "learning_rate": 2.6672619780761176e-06, + "loss": 0.5005, + "step": 9055 + }, + { + "epoch": 3.34, + "learning_rate": 2.6672250277127726e-06, + "loss": 0.6654, + "step": 9056 + }, + { + "epoch": 3.34, + "learning_rate": 2.6671880773494273e-06, + "loss": 0.5077, + "step": 9057 + }, + { + "epoch": 3.34, + "learning_rate": 2.6671511269860823e-06, + "loss": 0.528, + "step": 9058 + }, + { + "epoch": 3.35, + "learning_rate": 2.667114176622737e-06, + "loss": 0.5034, + "step": 9059 + }, + { + "epoch": 3.35, + "learning_rate": 2.6670772262593916e-06, + "loss": 0.5189, + "step": 9060 + }, + { + "epoch": 3.35, + "learning_rate": 2.6670402758960462e-06, + "loss": 0.6586, + "step": 9061 + }, + { + "epoch": 3.35, + "learning_rate": 2.667003325532701e-06, + "loss": 0.5119, + "step": 9062 + }, + { + "epoch": 3.35, + "learning_rate": 2.666966375169356e-06, + "loss": 0.5415, + "step": 9063 + }, + { + "epoch": 3.35, + "learning_rate": 2.6669294248060106e-06, + "loss": 0.4771, + "step": 9064 + }, + { + "epoch": 3.35, + "learning_rate": 2.6668924744426656e-06, + "loss": 0.6281, + "step": 9065 + }, + { + "epoch": 3.35, + "learning_rate": 2.6668555240793202e-06, + "loss": 0.6029, + "step": 9066 + }, + { + "epoch": 3.35, + "learning_rate": 2.666818573715975e-06, + "loss": 0.5393, + "step": 9067 + }, + { + "epoch": 3.35, + "learning_rate": 2.6667816233526295e-06, + "loss": 0.5552, + "step": 9068 + }, + { + "epoch": 3.35, + "learning_rate": 2.6667446729892846e-06, + "loss": 0.4611, + "step": 9069 + }, + { + "epoch": 3.35, + "learning_rate": 2.6667077226259392e-06, + "loss": 0.6084, + "step": 9070 + }, + { + "epoch": 3.35, + "learning_rate": 2.6666707722625943e-06, + "loss": 0.5915, + "step": 9071 + }, + { + "epoch": 3.35, + "learning_rate": 2.666633821899249e-06, + "loss": 0.5596, + "step": 9072 + }, + { + "epoch": 3.35, + "learning_rate": 2.6665968715359036e-06, + "loss": 0.5565, + "step": 9073 + }, + { + "epoch": 3.35, + "learning_rate": 2.666559921172558e-06, + "loss": 0.4363, + "step": 9074 + }, + { + "epoch": 3.35, + "learning_rate": 2.666522970809213e-06, + "loss": 0.6136, + "step": 9075 + }, + { + "epoch": 3.35, + "learning_rate": 2.666486020445868e-06, + "loss": 0.5459, + "step": 9076 + }, + { + "epoch": 3.35, + "learning_rate": 2.6664490700825225e-06, + "loss": 0.5793, + "step": 9077 + }, + { + "epoch": 3.35, + "learning_rate": 2.6664121197191776e-06, + "loss": 0.6228, + "step": 9078 + }, + { + "epoch": 3.35, + "learning_rate": 2.6663751693558322e-06, + "loss": 0.595, + "step": 9079 + }, + { + "epoch": 3.35, + "learning_rate": 2.666338218992487e-06, + "loss": 0.5608, + "step": 9080 + }, + { + "epoch": 3.35, + "learning_rate": 2.6663012686291415e-06, + "loss": 0.5326, + "step": 9081 + }, + { + "epoch": 3.35, + "learning_rate": 2.666264318265796e-06, + "loss": 0.5945, + "step": 9082 + }, + { + "epoch": 3.35, + "learning_rate": 2.666227367902451e-06, + "loss": 0.54, + "step": 9083 + }, + { + "epoch": 3.35, + "learning_rate": 2.666190417539106e-06, + "loss": 0.5598, + "step": 9084 + }, + { + "epoch": 3.35, + "learning_rate": 2.666153467175761e-06, + "loss": 0.5831, + "step": 9085 + }, + { + "epoch": 3.36, + "learning_rate": 2.666116516812415e-06, + "loss": 0.5911, + "step": 9086 + }, + { + "epoch": 3.36, + "learning_rate": 2.66607956644907e-06, + "loss": 0.6036, + "step": 9087 + }, + { + "epoch": 3.36, + "learning_rate": 2.666042616085725e-06, + "loss": 0.5159, + "step": 9088 + }, + { + "epoch": 3.36, + "learning_rate": 2.66600566572238e-06, + "loss": 0.5169, + "step": 9089 + }, + { + "epoch": 3.36, + "learning_rate": 2.6659687153590345e-06, + "loss": 0.4927, + "step": 9090 + }, + { + "epoch": 3.36, + "learning_rate": 2.6659317649956895e-06, + "loss": 0.5405, + "step": 9091 + }, + { + "epoch": 3.36, + "learning_rate": 2.6658948146323438e-06, + "loss": 0.5083, + "step": 9092 + }, + { + "epoch": 3.36, + "learning_rate": 2.6658578642689984e-06, + "loss": 0.5778, + "step": 9093 + }, + { + "epoch": 3.36, + "learning_rate": 2.6658209139056535e-06, + "loss": 0.5907, + "step": 9094 + }, + { + "epoch": 3.36, + "learning_rate": 2.665783963542308e-06, + "loss": 0.5033, + "step": 9095 + }, + { + "epoch": 3.36, + "learning_rate": 2.665747013178963e-06, + "loss": 0.5472, + "step": 9096 + }, + { + "epoch": 3.36, + "learning_rate": 2.665710062815618e-06, + "loss": 0.5169, + "step": 9097 + }, + { + "epoch": 3.36, + "learning_rate": 2.6656731124522724e-06, + "loss": 0.5646, + "step": 9098 + }, + { + "epoch": 3.36, + "learning_rate": 2.665636162088927e-06, + "loss": 0.6398, + "step": 9099 + }, + { + "epoch": 3.36, + "learning_rate": 2.665599211725582e-06, + "loss": 0.4688, + "step": 9100 + }, + { + "epoch": 3.36, + "learning_rate": 2.6655622613622368e-06, + "loss": 0.5249, + "step": 9101 + }, + { + "epoch": 3.36, + "learning_rate": 2.6655253109988914e-06, + "loss": 0.5474, + "step": 9102 + }, + { + "epoch": 3.36, + "learning_rate": 2.6654883606355465e-06, + "loss": 0.639, + "step": 9103 + }, + { + "epoch": 3.36, + "learning_rate": 2.665451410272201e-06, + "loss": 0.4894, + "step": 9104 + }, + { + "epoch": 3.36, + "learning_rate": 2.6654144599088557e-06, + "loss": 0.5435, + "step": 9105 + }, + { + "epoch": 3.36, + "learning_rate": 2.6653775095455104e-06, + "loss": 0.5233, + "step": 9106 + }, + { + "epoch": 3.36, + "learning_rate": 2.6653405591821654e-06, + "loss": 0.5039, + "step": 9107 + }, + { + "epoch": 3.36, + "learning_rate": 2.66530360881882e-06, + "loss": 0.5818, + "step": 9108 + }, + { + "epoch": 3.36, + "learning_rate": 2.665266658455475e-06, + "loss": 0.5139, + "step": 9109 + }, + { + "epoch": 3.36, + "learning_rate": 2.6652297080921298e-06, + "loss": 0.5283, + "step": 9110 + }, + { + "epoch": 3.36, + "learning_rate": 2.6651927577287844e-06, + "loss": 0.5834, + "step": 9111 + }, + { + "epoch": 3.36, + "learning_rate": 2.665155807365439e-06, + "loss": 0.3791, + "step": 9112 + }, + { + "epoch": 3.37, + "learning_rate": 2.6651188570020937e-06, + "loss": 0.4306, + "step": 9113 + }, + { + "epoch": 3.37, + "learning_rate": 2.6650819066387487e-06, + "loss": 0.6105, + "step": 9114 + }, + { + "epoch": 3.37, + "learning_rate": 2.6650449562754034e-06, + "loss": 0.5231, + "step": 9115 + }, + { + "epoch": 3.37, + "learning_rate": 2.6650080059120584e-06, + "loss": 0.5849, + "step": 9116 + }, + { + "epoch": 3.37, + "learning_rate": 2.664971055548713e-06, + "loss": 0.4948, + "step": 9117 + }, + { + "epoch": 3.37, + "learning_rate": 2.6649341051853677e-06, + "loss": 0.6055, + "step": 9118 + }, + { + "epoch": 3.37, + "learning_rate": 2.6648971548220223e-06, + "loss": 0.5628, + "step": 9119 + }, + { + "epoch": 3.37, + "learning_rate": 2.6648602044586774e-06, + "loss": 0.533, + "step": 9120 + }, + { + "epoch": 3.37, + "learning_rate": 2.664823254095332e-06, + "loss": 0.4637, + "step": 9121 + }, + { + "epoch": 3.37, + "learning_rate": 2.6647863037319867e-06, + "loss": 0.5455, + "step": 9122 + }, + { + "epoch": 3.37, + "learning_rate": 2.6647493533686417e-06, + "loss": 0.5524, + "step": 9123 + }, + { + "epoch": 3.37, + "learning_rate": 2.6647124030052964e-06, + "loss": 0.5668, + "step": 9124 + }, + { + "epoch": 3.37, + "learning_rate": 2.664675452641951e-06, + "loss": 0.5811, + "step": 9125 + }, + { + "epoch": 3.37, + "learning_rate": 2.6646385022786056e-06, + "loss": 0.5358, + "step": 9126 + }, + { + "epoch": 3.37, + "learning_rate": 2.6646015519152607e-06, + "loss": 0.5114, + "step": 9127 + }, + { + "epoch": 3.37, + "learning_rate": 2.6645646015519153e-06, + "loss": 0.5887, + "step": 9128 + }, + { + "epoch": 3.37, + "learning_rate": 2.6645276511885704e-06, + "loss": 0.4803, + "step": 9129 + }, + { + "epoch": 3.37, + "learning_rate": 2.664490700825225e-06, + "loss": 0.5741, + "step": 9130 + }, + { + "epoch": 3.37, + "learning_rate": 2.6644537504618792e-06, + "loss": 0.5247, + "step": 9131 + }, + { + "epoch": 3.37, + "learning_rate": 2.6644168000985343e-06, + "loss": 0.4713, + "step": 9132 + }, + { + "epoch": 3.37, + "learning_rate": 2.664379849735189e-06, + "loss": 0.5931, + "step": 9133 + }, + { + "epoch": 3.37, + "learning_rate": 2.664342899371844e-06, + "loss": 0.4948, + "step": 9134 + }, + { + "epoch": 3.37, + "learning_rate": 2.6643059490084986e-06, + "loss": 0.5524, + "step": 9135 + }, + { + "epoch": 3.37, + "learning_rate": 2.6642689986451537e-06, + "loss": 0.6175, + "step": 9136 + }, + { + "epoch": 3.37, + "learning_rate": 2.664232048281808e-06, + "loss": 0.5778, + "step": 9137 + }, + { + "epoch": 3.37, + "learning_rate": 2.664195097918463e-06, + "loss": 0.5306, + "step": 9138 + }, + { + "epoch": 3.37, + "learning_rate": 2.6641581475551176e-06, + "loss": 0.5404, + "step": 9139 + }, + { + "epoch": 3.38, + "learning_rate": 2.6641211971917727e-06, + "loss": 0.4948, + "step": 9140 + }, + { + "epoch": 3.38, + "learning_rate": 2.6640842468284273e-06, + "loss": 0.4488, + "step": 9141 + }, + { + "epoch": 3.38, + "learning_rate": 2.664047296465082e-06, + "loss": 0.5777, + "step": 9142 + }, + { + "epoch": 3.38, + "learning_rate": 2.6640103461017366e-06, + "loss": 0.5244, + "step": 9143 + }, + { + "epoch": 3.38, + "learning_rate": 2.663973395738391e-06, + "loss": 0.4995, + "step": 9144 + }, + { + "epoch": 3.38, + "learning_rate": 2.6639364453750463e-06, + "loss": 0.6309, + "step": 9145 + }, + { + "epoch": 3.38, + "learning_rate": 2.663899495011701e-06, + "loss": 0.537, + "step": 9146 + }, + { + "epoch": 3.38, + "learning_rate": 2.663862544648356e-06, + "loss": 0.5024, + "step": 9147 + }, + { + "epoch": 3.38, + "learning_rate": 2.6638255942850106e-06, + "loss": 0.3998, + "step": 9148 + }, + { + "epoch": 3.38, + "learning_rate": 2.6637886439216652e-06, + "loss": 0.5541, + "step": 9149 + }, + { + "epoch": 3.38, + "learning_rate": 2.66375169355832e-06, + "loss": 0.5817, + "step": 9150 + }, + { + "epoch": 3.38, + "learning_rate": 2.6637147431949745e-06, + "loss": 0.4733, + "step": 9151 + }, + { + "epoch": 3.38, + "learning_rate": 2.6636777928316296e-06, + "loss": 0.6517, + "step": 9152 + }, + { + "epoch": 3.38, + "learning_rate": 2.663640842468284e-06, + "loss": 0.629, + "step": 9153 + }, + { + "epoch": 3.38, + "learning_rate": 2.6636038921049393e-06, + "loss": 0.4767, + "step": 9154 + }, + { + "epoch": 3.38, + "learning_rate": 2.663566941741594e-06, + "loss": 0.615, + "step": 9155 + }, + { + "epoch": 3.38, + "learning_rate": 2.6635299913782485e-06, + "loss": 0.5786, + "step": 9156 + }, + { + "epoch": 3.38, + "learning_rate": 2.663493041014903e-06, + "loss": 0.6286, + "step": 9157 + }, + { + "epoch": 3.38, + "learning_rate": 2.6634560906515582e-06, + "loss": 0.5973, + "step": 9158 + }, + { + "epoch": 3.38, + "learning_rate": 2.663419140288213e-06, + "loss": 0.5662, + "step": 9159 + }, + { + "epoch": 3.38, + "learning_rate": 2.663382189924868e-06, + "loss": 0.5177, + "step": 9160 + }, + { + "epoch": 3.38, + "learning_rate": 2.6633452395615226e-06, + "loss": 0.5575, + "step": 9161 + }, + { + "epoch": 3.38, + "learning_rate": 2.663308289198177e-06, + "loss": 0.5327, + "step": 9162 + }, + { + "epoch": 3.38, + "learning_rate": 2.663271338834832e-06, + "loss": 0.5724, + "step": 9163 + }, + { + "epoch": 3.38, + "learning_rate": 2.6632343884714865e-06, + "loss": 0.5683, + "step": 9164 + }, + { + "epoch": 3.38, + "learning_rate": 2.6631974381081415e-06, + "loss": 0.5298, + "step": 9165 + }, + { + "epoch": 3.38, + "learning_rate": 2.663160487744796e-06, + "loss": 0.4888, + "step": 9166 + }, + { + "epoch": 3.39, + "learning_rate": 2.6631235373814512e-06, + "loss": 0.5475, + "step": 9167 + }, + { + "epoch": 3.39, + "learning_rate": 2.663086587018106e-06, + "loss": 0.6403, + "step": 9168 + }, + { + "epoch": 3.39, + "learning_rate": 2.6630496366547605e-06, + "loss": 0.4744, + "step": 9169 + }, + { + "epoch": 3.39, + "learning_rate": 2.663012686291415e-06, + "loss": 0.4943, + "step": 9170 + }, + { + "epoch": 3.39, + "learning_rate": 2.6629757359280698e-06, + "loss": 0.5958, + "step": 9171 + }, + { + "epoch": 3.39, + "learning_rate": 2.662938785564725e-06, + "loss": 0.5784, + "step": 9172 + }, + { + "epoch": 3.39, + "learning_rate": 2.6629018352013795e-06, + "loss": 0.5092, + "step": 9173 + }, + { + "epoch": 3.39, + "learning_rate": 2.6628648848380345e-06, + "loss": 0.4935, + "step": 9174 + }, + { + "epoch": 3.39, + "learning_rate": 2.662827934474689e-06, + "loss": 0.6921, + "step": 9175 + }, + { + "epoch": 3.39, + "learning_rate": 2.662790984111344e-06, + "loss": 0.5487, + "step": 9176 + }, + { + "epoch": 3.39, + "learning_rate": 2.6627540337479984e-06, + "loss": 0.5288, + "step": 9177 + }, + { + "epoch": 3.39, + "learning_rate": 2.6627170833846535e-06, + "loss": 0.575, + "step": 9178 + }, + { + "epoch": 3.39, + "learning_rate": 2.662680133021308e-06, + "loss": 0.5657, + "step": 9179 + }, + { + "epoch": 3.39, + "learning_rate": 2.662643182657963e-06, + "loss": 0.5929, + "step": 9180 + }, + { + "epoch": 3.39, + "learning_rate": 2.662606232294618e-06, + "loss": 0.7036, + "step": 9181 + }, + { + "epoch": 3.39, + "learning_rate": 2.662569281931272e-06, + "loss": 0.6376, + "step": 9182 + }, + { + "epoch": 3.39, + "learning_rate": 2.662532331567927e-06, + "loss": 0.5779, + "step": 9183 + }, + { + "epoch": 3.39, + "learning_rate": 2.6624953812045817e-06, + "loss": 0.6108, + "step": 9184 + }, + { + "epoch": 3.39, + "learning_rate": 2.662458430841237e-06, + "loss": 0.5873, + "step": 9185 + }, + { + "epoch": 3.39, + "learning_rate": 2.6624214804778914e-06, + "loss": 0.5656, + "step": 9186 + }, + { + "epoch": 3.39, + "learning_rate": 2.6623845301145465e-06, + "loss": 0.576, + "step": 9187 + }, + { + "epoch": 3.39, + "learning_rate": 2.6623475797512007e-06, + "loss": 0.6003, + "step": 9188 + }, + { + "epoch": 3.39, + "learning_rate": 2.6623106293878558e-06, + "loss": 0.5825, + "step": 9189 + }, + { + "epoch": 3.39, + "learning_rate": 2.6622736790245104e-06, + "loss": 0.5656, + "step": 9190 + }, + { + "epoch": 3.39, + "learning_rate": 2.662236728661165e-06, + "loss": 0.6001, + "step": 9191 + }, + { + "epoch": 3.39, + "learning_rate": 2.66219977829782e-06, + "loss": 0.5557, + "step": 9192 + }, + { + "epoch": 3.39, + "learning_rate": 2.6621628279344747e-06, + "loss": 0.6117, + "step": 9193 + }, + { + "epoch": 3.4, + "learning_rate": 2.6621258775711294e-06, + "loss": 0.4577, + "step": 9194 + }, + { + "epoch": 3.4, + "learning_rate": 2.662088927207784e-06, + "loss": 0.5683, + "step": 9195 + }, + { + "epoch": 3.4, + "learning_rate": 2.662051976844439e-06, + "loss": 0.5502, + "step": 9196 + }, + { + "epoch": 3.4, + "learning_rate": 2.6620150264810937e-06, + "loss": 0.529, + "step": 9197 + }, + { + "epoch": 3.4, + "learning_rate": 2.6619780761177488e-06, + "loss": 0.6023, + "step": 9198 + }, + { + "epoch": 3.4, + "learning_rate": 2.6619411257544034e-06, + "loss": 0.6089, + "step": 9199 + }, + { + "epoch": 3.4, + "learning_rate": 2.661904175391058e-06, + "loss": 0.5767, + "step": 9200 + }, + { + "epoch": 3.4, + "learning_rate": 2.6618672250277127e-06, + "loss": 0.5241, + "step": 9201 + }, + { + "epoch": 3.4, + "learning_rate": 2.6618302746643673e-06, + "loss": 0.6233, + "step": 9202 + }, + { + "epoch": 3.4, + "learning_rate": 2.6617933243010224e-06, + "loss": 0.5492, + "step": 9203 + }, + { + "epoch": 3.4, + "learning_rate": 2.661756373937677e-06, + "loss": 0.6143, + "step": 9204 + }, + { + "epoch": 3.4, + "learning_rate": 2.661719423574332e-06, + "loss": 0.5487, + "step": 9205 + }, + { + "epoch": 3.4, + "learning_rate": 2.6616824732109867e-06, + "loss": 0.5562, + "step": 9206 + }, + { + "epoch": 3.4, + "learning_rate": 2.6616455228476414e-06, + "loss": 0.5438, + "step": 9207 + }, + { + "epoch": 3.4, + "learning_rate": 2.661608572484296e-06, + "loss": 0.5381, + "step": 9208 + }, + { + "epoch": 3.4, + "learning_rate": 2.661571622120951e-06, + "loss": 0.6181, + "step": 9209 + }, + { + "epoch": 3.4, + "learning_rate": 2.6615346717576057e-06, + "loss": 0.542, + "step": 9210 + }, + { + "epoch": 3.4, + "learning_rate": 2.6614977213942603e-06, + "loss": 0.4704, + "step": 9211 + }, + { + "epoch": 3.4, + "learning_rate": 2.6614607710309154e-06, + "loss": 0.5714, + "step": 9212 + }, + { + "epoch": 3.4, + "learning_rate": 2.66142382066757e-06, + "loss": 0.5519, + "step": 9213 + }, + { + "epoch": 3.4, + "learning_rate": 2.6613868703042247e-06, + "loss": 0.5605, + "step": 9214 + }, + { + "epoch": 3.4, + "learning_rate": 2.6613499199408793e-06, + "loss": 0.4521, + "step": 9215 + }, + { + "epoch": 3.4, + "learning_rate": 2.6613129695775343e-06, + "loss": 0.4564, + "step": 9216 + }, + { + "epoch": 3.4, + "learning_rate": 2.661276019214189e-06, + "loss": 0.5908, + "step": 9217 + }, + { + "epoch": 3.4, + "learning_rate": 2.661239068850844e-06, + "loss": 0.6256, + "step": 9218 + }, + { + "epoch": 3.4, + "learning_rate": 2.6612021184874987e-06, + "loss": 0.4946, + "step": 9219 + }, + { + "epoch": 3.4, + "learning_rate": 2.6611651681241533e-06, + "loss": 0.5636, + "step": 9220 + }, + { + "epoch": 3.41, + "learning_rate": 2.661128217760808e-06, + "loss": 0.5845, + "step": 9221 + }, + { + "epoch": 3.41, + "learning_rate": 2.6610912673974626e-06, + "loss": 0.4899, + "step": 9222 + }, + { + "epoch": 3.41, + "learning_rate": 2.6610543170341177e-06, + "loss": 0.6072, + "step": 9223 + }, + { + "epoch": 3.41, + "learning_rate": 2.6610173666707723e-06, + "loss": 0.5712, + "step": 9224 + }, + { + "epoch": 3.41, + "learning_rate": 2.6609804163074273e-06, + "loss": 0.5176, + "step": 9225 + }, + { + "epoch": 3.41, + "learning_rate": 2.660943465944082e-06, + "loss": 0.5773, + "step": 9226 + }, + { + "epoch": 3.41, + "learning_rate": 2.6609065155807366e-06, + "loss": 0.5673, + "step": 9227 + }, + { + "epoch": 3.41, + "learning_rate": 2.6608695652173913e-06, + "loss": 0.5163, + "step": 9228 + }, + { + "epoch": 3.41, + "learning_rate": 2.6608326148540463e-06, + "loss": 0.5542, + "step": 9229 + }, + { + "epoch": 3.41, + "learning_rate": 2.660795664490701e-06, + "loss": 0.6322, + "step": 9230 + }, + { + "epoch": 3.41, + "learning_rate": 2.6607587141273556e-06, + "loss": 0.5535, + "step": 9231 + }, + { + "epoch": 3.41, + "learning_rate": 2.6607217637640107e-06, + "loss": 0.6297, + "step": 9232 + }, + { + "epoch": 3.41, + "learning_rate": 2.660684813400665e-06, + "loss": 0.4845, + "step": 9233 + }, + { + "epoch": 3.41, + "learning_rate": 2.66064786303732e-06, + "loss": 0.4804, + "step": 9234 + }, + { + "epoch": 3.41, + "learning_rate": 2.6606109126739746e-06, + "loss": 0.5359, + "step": 9235 + }, + { + "epoch": 3.41, + "learning_rate": 2.6605739623106296e-06, + "loss": 0.5056, + "step": 9236 + }, + { + "epoch": 3.41, + "learning_rate": 2.6605370119472843e-06, + "loss": 0.6609, + "step": 9237 + }, + { + "epoch": 3.41, + "learning_rate": 2.660500061583939e-06, + "loss": 0.5603, + "step": 9238 + }, + { + "epoch": 3.41, + "learning_rate": 2.6604631112205935e-06, + "loss": 0.5246, + "step": 9239 + }, + { + "epoch": 3.41, + "learning_rate": 2.6604261608572486e-06, + "loss": 0.5671, + "step": 9240 + }, + { + "epoch": 3.41, + "learning_rate": 2.6603892104939032e-06, + "loss": 0.5701, + "step": 9241 + }, + { + "epoch": 3.41, + "learning_rate": 2.660352260130558e-06, + "loss": 0.5873, + "step": 9242 + }, + { + "epoch": 3.41, + "learning_rate": 2.660315309767213e-06, + "loss": 0.5257, + "step": 9243 + }, + { + "epoch": 3.41, + "learning_rate": 2.6602783594038676e-06, + "loss": 0.5006, + "step": 9244 + }, + { + "epoch": 3.41, + "learning_rate": 2.660241409040522e-06, + "loss": 0.5328, + "step": 9245 + }, + { + "epoch": 3.41, + "learning_rate": 2.660204458677177e-06, + "loss": 0.525, + "step": 9246 + }, + { + "epoch": 3.41, + "learning_rate": 2.660167508313832e-06, + "loss": 0.4884, + "step": 9247 + }, + { + "epoch": 3.42, + "learning_rate": 2.6601305579504865e-06, + "loss": 0.4662, + "step": 9248 + }, + { + "epoch": 3.42, + "learning_rate": 2.6600936075871416e-06, + "loss": 0.4206, + "step": 9249 + }, + { + "epoch": 3.42, + "learning_rate": 2.6600566572237962e-06, + "loss": 0.5239, + "step": 9250 + }, + { + "epoch": 3.42, + "learning_rate": 2.660019706860451e-06, + "loss": 0.5016, + "step": 9251 + }, + { + "epoch": 3.42, + "learning_rate": 2.6599827564971055e-06, + "loss": 0.4692, + "step": 9252 + }, + { + "epoch": 3.42, + "learning_rate": 2.65994580613376e-06, + "loss": 0.532, + "step": 9253 + }, + { + "epoch": 3.42, + "learning_rate": 2.659908855770415e-06, + "loss": 0.5565, + "step": 9254 + }, + { + "epoch": 3.42, + "learning_rate": 2.65987190540707e-06, + "loss": 0.5247, + "step": 9255 + }, + { + "epoch": 3.42, + "learning_rate": 2.659834955043725e-06, + "loss": 0.5701, + "step": 9256 + }, + { + "epoch": 3.42, + "learning_rate": 2.6597980046803795e-06, + "loss": 0.7086, + "step": 9257 + }, + { + "epoch": 3.42, + "learning_rate": 2.659761054317034e-06, + "loss": 0.6235, + "step": 9258 + }, + { + "epoch": 3.42, + "learning_rate": 2.659724103953689e-06, + "loss": 0.5419, + "step": 9259 + }, + { + "epoch": 3.42, + "learning_rate": 2.659687153590344e-06, + "loss": 0.57, + "step": 9260 + }, + { + "epoch": 3.42, + "learning_rate": 2.6596502032269985e-06, + "loss": 0.6268, + "step": 9261 + }, + { + "epoch": 3.42, + "learning_rate": 2.659613252863653e-06, + "loss": 0.619, + "step": 9262 + }, + { + "epoch": 3.42, + "learning_rate": 2.659576302500308e-06, + "loss": 0.5144, + "step": 9263 + }, + { + "epoch": 3.42, + "learning_rate": 2.659539352136963e-06, + "loss": 0.5617, + "step": 9264 + }, + { + "epoch": 3.42, + "learning_rate": 2.6595024017736175e-06, + "loss": 0.4989, + "step": 9265 + }, + { + "epoch": 3.42, + "learning_rate": 2.659465451410272e-06, + "loss": 0.4818, + "step": 9266 + }, + { + "epoch": 3.42, + "learning_rate": 2.659428501046927e-06, + "loss": 0.4882, + "step": 9267 + }, + { + "epoch": 3.42, + "learning_rate": 2.659391550683582e-06, + "loss": 0.5546, + "step": 9268 + }, + { + "epoch": 3.42, + "learning_rate": 2.659354600320237e-06, + "loss": 0.5629, + "step": 9269 + }, + { + "epoch": 3.42, + "learning_rate": 2.6593176499568915e-06, + "loss": 0.6082, + "step": 9270 + }, + { + "epoch": 3.42, + "learning_rate": 2.6592806995935457e-06, + "loss": 0.4342, + "step": 9271 + }, + { + "epoch": 3.42, + "learning_rate": 2.6592437492302008e-06, + "loss": 0.6296, + "step": 9272 + }, + { + "epoch": 3.42, + "learning_rate": 2.6592067988668554e-06, + "loss": 0.5153, + "step": 9273 + }, + { + "epoch": 3.42, + "learning_rate": 2.6591698485035105e-06, + "loss": 0.5182, + "step": 9274 + }, + { + "epoch": 3.43, + "learning_rate": 2.659132898140165e-06, + "loss": 0.5489, + "step": 9275 + }, + { + "epoch": 3.43, + "learning_rate": 2.65909594777682e-06, + "loss": 0.611, + "step": 9276 + }, + { + "epoch": 3.43, + "learning_rate": 2.6590589974134744e-06, + "loss": 0.5258, + "step": 9277 + }, + { + "epoch": 3.43, + "learning_rate": 2.6590220470501294e-06, + "loss": 0.5222, + "step": 9278 + }, + { + "epoch": 3.43, + "learning_rate": 2.658985096686784e-06, + "loss": 0.5219, + "step": 9279 + }, + { + "epoch": 3.43, + "learning_rate": 2.658948146323439e-06, + "loss": 0.6179, + "step": 9280 + }, + { + "epoch": 3.43, + "learning_rate": 2.6589111959600938e-06, + "loss": 0.5136, + "step": 9281 + }, + { + "epoch": 3.43, + "learning_rate": 2.6588742455967484e-06, + "loss": 0.5251, + "step": 9282 + }, + { + "epoch": 3.43, + "learning_rate": 2.658837295233403e-06, + "loss": 0.6187, + "step": 9283 + }, + { + "epoch": 3.43, + "learning_rate": 2.6588003448700577e-06, + "loss": 0.5995, + "step": 9284 + }, + { + "epoch": 3.43, + "learning_rate": 2.6587633945067127e-06, + "loss": 0.5204, + "step": 9285 + }, + { + "epoch": 3.43, + "learning_rate": 2.6587264441433674e-06, + "loss": 0.6251, + "step": 9286 + }, + { + "epoch": 3.43, + "learning_rate": 2.6586894937800224e-06, + "loss": 0.6057, + "step": 9287 + }, + { + "epoch": 3.43, + "learning_rate": 2.658652543416677e-06, + "loss": 0.5305, + "step": 9288 + }, + { + "epoch": 3.43, + "learning_rate": 2.6586155930533317e-06, + "loss": 0.6105, + "step": 9289 + }, + { + "epoch": 3.43, + "learning_rate": 2.6585786426899863e-06, + "loss": 0.5283, + "step": 9290 + }, + { + "epoch": 3.43, + "learning_rate": 2.658541692326641e-06, + "loss": 0.5624, + "step": 9291 + }, + { + "epoch": 3.43, + "learning_rate": 2.658504741963296e-06, + "loss": 0.5261, + "step": 9292 + }, + { + "epoch": 3.43, + "learning_rate": 2.6584677915999507e-06, + "loss": 0.5333, + "step": 9293 + }, + { + "epoch": 3.43, + "learning_rate": 2.6584308412366057e-06, + "loss": 0.4614, + "step": 9294 + }, + { + "epoch": 3.43, + "learning_rate": 2.6583938908732604e-06, + "loss": 0.5915, + "step": 9295 + }, + { + "epoch": 3.43, + "learning_rate": 2.658356940509915e-06, + "loss": 0.508, + "step": 9296 + }, + { + "epoch": 3.43, + "learning_rate": 2.6583199901465696e-06, + "loss": 0.5926, + "step": 9297 + }, + { + "epoch": 3.43, + "learning_rate": 2.6582830397832247e-06, + "loss": 0.7274, + "step": 9298 + }, + { + "epoch": 3.43, + "learning_rate": 2.6582460894198793e-06, + "loss": 0.5368, + "step": 9299 + }, + { + "epoch": 3.43, + "learning_rate": 2.6582091390565344e-06, + "loss": 0.6461, + "step": 9300 + }, + { + "epoch": 3.43, + "learning_rate": 2.658172188693189e-06, + "loss": 0.5355, + "step": 9301 + }, + { + "epoch": 3.44, + "learning_rate": 2.6581352383298437e-06, + "loss": 0.558, + "step": 9302 + }, + { + "epoch": 3.44, + "learning_rate": 2.6580982879664983e-06, + "loss": 0.5029, + "step": 9303 + }, + { + "epoch": 3.44, + "learning_rate": 2.658061337603153e-06, + "loss": 0.5268, + "step": 9304 + }, + { + "epoch": 3.44, + "learning_rate": 2.658024387239808e-06, + "loss": 0.5037, + "step": 9305 + }, + { + "epoch": 3.44, + "learning_rate": 2.6579874368764626e-06, + "loss": 0.5211, + "step": 9306 + }, + { + "epoch": 3.44, + "learning_rate": 2.6579504865131177e-06, + "loss": 0.5495, + "step": 9307 + }, + { + "epoch": 3.44, + "learning_rate": 2.6579135361497723e-06, + "loss": 0.5073, + "step": 9308 + }, + { + "epoch": 3.44, + "learning_rate": 2.657876585786427e-06, + "loss": 0.5275, + "step": 9309 + }, + { + "epoch": 3.44, + "learning_rate": 2.6578396354230816e-06, + "loss": 0.5687, + "step": 9310 + }, + { + "epoch": 3.44, + "learning_rate": 2.6578026850597362e-06, + "loss": 0.5001, + "step": 9311 + }, + { + "epoch": 3.44, + "learning_rate": 2.6577657346963913e-06, + "loss": 0.5704, + "step": 9312 + }, + { + "epoch": 3.44, + "learning_rate": 2.657728784333046e-06, + "loss": 0.5823, + "step": 9313 + }, + { + "epoch": 3.44, + "learning_rate": 2.657691833969701e-06, + "loss": 0.5176, + "step": 9314 + }, + { + "epoch": 3.44, + "learning_rate": 2.6576548836063556e-06, + "loss": 0.7158, + "step": 9315 + }, + { + "epoch": 3.44, + "learning_rate": 2.6576179332430103e-06, + "loss": 0.4833, + "step": 9316 + }, + { + "epoch": 3.44, + "learning_rate": 2.657580982879665e-06, + "loss": 0.516, + "step": 9317 + }, + { + "epoch": 3.44, + "learning_rate": 2.65754403251632e-06, + "loss": 0.5747, + "step": 9318 + }, + { + "epoch": 3.44, + "learning_rate": 2.6575070821529746e-06, + "loss": 0.7186, + "step": 9319 + }, + { + "epoch": 3.44, + "learning_rate": 2.6574701317896297e-06, + "loss": 0.5811, + "step": 9320 + }, + { + "epoch": 3.44, + "learning_rate": 2.6574331814262843e-06, + "loss": 0.4815, + "step": 9321 + }, + { + "epoch": 3.44, + "learning_rate": 2.6573962310629385e-06, + "loss": 0.6787, + "step": 9322 + }, + { + "epoch": 3.44, + "learning_rate": 2.6573592806995936e-06, + "loss": 0.5637, + "step": 9323 + }, + { + "epoch": 3.44, + "learning_rate": 2.6573223303362482e-06, + "loss": 0.6196, + "step": 9324 + }, + { + "epoch": 3.44, + "learning_rate": 2.6572853799729033e-06, + "loss": 0.5443, + "step": 9325 + }, + { + "epoch": 3.44, + "learning_rate": 2.657248429609558e-06, + "loss": 0.5629, + "step": 9326 + }, + { + "epoch": 3.44, + "learning_rate": 2.657211479246213e-06, + "loss": 0.619, + "step": 9327 + }, + { + "epoch": 3.44, + "learning_rate": 2.657174528882867e-06, + "loss": 0.4438, + "step": 9328 + }, + { + "epoch": 3.44, + "learning_rate": 2.6571375785195222e-06, + "loss": 0.6278, + "step": 9329 + }, + { + "epoch": 3.45, + "learning_rate": 2.657100628156177e-06, + "loss": 0.6237, + "step": 9330 + }, + { + "epoch": 3.45, + "learning_rate": 2.6570636777928315e-06, + "loss": 0.5298, + "step": 9331 + }, + { + "epoch": 3.45, + "learning_rate": 2.6570267274294866e-06, + "loss": 0.6159, + "step": 9332 + }, + { + "epoch": 3.45, + "learning_rate": 2.6569897770661412e-06, + "loss": 0.5096, + "step": 9333 + }, + { + "epoch": 3.45, + "learning_rate": 2.656952826702796e-06, + "loss": 0.5197, + "step": 9334 + }, + { + "epoch": 3.45, + "learning_rate": 2.6569158763394505e-06, + "loss": 0.5662, + "step": 9335 + }, + { + "epoch": 3.45, + "learning_rate": 2.6568789259761055e-06, + "loss": 0.5624, + "step": 9336 + }, + { + "epoch": 3.45, + "learning_rate": 2.65684197561276e-06, + "loss": 0.5943, + "step": 9337 + }, + { + "epoch": 3.45, + "learning_rate": 2.6568050252494152e-06, + "loss": 0.5326, + "step": 9338 + }, + { + "epoch": 3.45, + "learning_rate": 2.65676807488607e-06, + "loss": 0.4971, + "step": 9339 + }, + { + "epoch": 3.45, + "learning_rate": 2.6567311245227245e-06, + "loss": 0.5427, + "step": 9340 + }, + { + "epoch": 3.45, + "learning_rate": 2.656694174159379e-06, + "loss": 0.5356, + "step": 9341 + }, + { + "epoch": 3.45, + "learning_rate": 2.6566572237960338e-06, + "loss": 0.5397, + "step": 9342 + }, + { + "epoch": 3.45, + "learning_rate": 2.656620273432689e-06, + "loss": 0.5822, + "step": 9343 + }, + { + "epoch": 3.45, + "learning_rate": 2.6565833230693435e-06, + "loss": 0.5835, + "step": 9344 + }, + { + "epoch": 3.45, + "learning_rate": 2.6565463727059985e-06, + "loss": 0.4609, + "step": 9345 + }, + { + "epoch": 3.45, + "learning_rate": 2.656509422342653e-06, + "loss": 0.6806, + "step": 9346 + }, + { + "epoch": 3.45, + "learning_rate": 2.656472471979308e-06, + "loss": 0.5519, + "step": 9347 + }, + { + "epoch": 3.45, + "learning_rate": 2.6564355216159625e-06, + "loss": 0.4325, + "step": 9348 + }, + { + "epoch": 3.45, + "learning_rate": 2.6563985712526175e-06, + "loss": 0.5482, + "step": 9349 + }, + { + "epoch": 3.45, + "learning_rate": 2.656361620889272e-06, + "loss": 0.5355, + "step": 9350 + }, + { + "epoch": 3.45, + "learning_rate": 2.6563246705259268e-06, + "loss": 0.4558, + "step": 9351 + }, + { + "epoch": 3.45, + "learning_rate": 2.656287720162582e-06, + "loss": 0.548, + "step": 9352 + }, + { + "epoch": 3.45, + "learning_rate": 2.6562507697992365e-06, + "loss": 0.5428, + "step": 9353 + }, + { + "epoch": 3.45, + "learning_rate": 2.656213819435891e-06, + "loss": 0.6303, + "step": 9354 + }, + { + "epoch": 3.45, + "learning_rate": 2.6561768690725458e-06, + "loss": 0.6161, + "step": 9355 + }, + { + "epoch": 3.45, + "learning_rate": 2.656139918709201e-06, + "loss": 0.4879, + "step": 9356 + }, + { + "epoch": 3.46, + "learning_rate": 2.6561029683458555e-06, + "loss": 0.6173, + "step": 9357 + }, + { + "epoch": 3.46, + "learning_rate": 2.6560660179825105e-06, + "loss": 0.5737, + "step": 9358 + }, + { + "epoch": 3.46, + "learning_rate": 2.656029067619165e-06, + "loss": 0.5413, + "step": 9359 + }, + { + "epoch": 3.46, + "learning_rate": 2.6559921172558198e-06, + "loss": 0.5667, + "step": 9360 + }, + { + "epoch": 3.46, + "learning_rate": 2.6559551668924744e-06, + "loss": 0.5726, + "step": 9361 + }, + { + "epoch": 3.46, + "learning_rate": 2.655918216529129e-06, + "loss": 0.5224, + "step": 9362 + }, + { + "epoch": 3.46, + "learning_rate": 2.655881266165784e-06, + "loss": 0.5578, + "step": 9363 + }, + { + "epoch": 3.46, + "learning_rate": 2.6558443158024388e-06, + "loss": 0.5144, + "step": 9364 + }, + { + "epoch": 3.46, + "learning_rate": 2.655807365439094e-06, + "loss": 0.5422, + "step": 9365 + }, + { + "epoch": 3.46, + "learning_rate": 2.6557704150757484e-06, + "loss": 0.5687, + "step": 9366 + }, + { + "epoch": 3.46, + "learning_rate": 2.655733464712403e-06, + "loss": 0.4508, + "step": 9367 + }, + { + "epoch": 3.46, + "learning_rate": 2.6556965143490577e-06, + "loss": 0.6025, + "step": 9368 + }, + { + "epoch": 3.46, + "learning_rate": 2.6556595639857128e-06, + "loss": 0.4916, + "step": 9369 + }, + { + "epoch": 3.46, + "learning_rate": 2.6556226136223674e-06, + "loss": 0.5514, + "step": 9370 + }, + { + "epoch": 3.46, + "learning_rate": 2.655585663259022e-06, + "loss": 0.5704, + "step": 9371 + }, + { + "epoch": 3.46, + "learning_rate": 2.655548712895677e-06, + "loss": 0.5302, + "step": 9372 + }, + { + "epoch": 3.46, + "learning_rate": 2.6555117625323313e-06, + "loss": 0.507, + "step": 9373 + }, + { + "epoch": 3.46, + "learning_rate": 2.6554748121689864e-06, + "loss": 0.4931, + "step": 9374 + }, + { + "epoch": 3.46, + "learning_rate": 2.655437861805641e-06, + "loss": 0.6612, + "step": 9375 + }, + { + "epoch": 3.46, + "learning_rate": 2.655400911442296e-06, + "loss": 0.5876, + "step": 9376 + }, + { + "epoch": 3.46, + "learning_rate": 2.6553639610789507e-06, + "loss": 0.5934, + "step": 9377 + }, + { + "epoch": 3.46, + "learning_rate": 2.6553270107156058e-06, + "loss": 0.5351, + "step": 9378 + }, + { + "epoch": 3.46, + "learning_rate": 2.65529006035226e-06, + "loss": 0.643, + "step": 9379 + }, + { + "epoch": 3.46, + "learning_rate": 2.6552531099889146e-06, + "loss": 0.4359, + "step": 9380 + }, + { + "epoch": 3.46, + "learning_rate": 2.6552161596255697e-06, + "loss": 0.4513, + "step": 9381 + }, + { + "epoch": 3.46, + "learning_rate": 2.6551792092622243e-06, + "loss": 0.5447, + "step": 9382 + }, + { + "epoch": 3.46, + "learning_rate": 2.6551422588988794e-06, + "loss": 0.5579, + "step": 9383 + }, + { + "epoch": 3.47, + "learning_rate": 2.655105308535534e-06, + "loss": 0.6522, + "step": 9384 + }, + { + "epoch": 3.47, + "learning_rate": 2.6550683581721887e-06, + "loss": 0.4843, + "step": 9385 + }, + { + "epoch": 3.47, + "learning_rate": 2.6550314078088433e-06, + "loss": 0.5335, + "step": 9386 + }, + { + "epoch": 3.47, + "learning_rate": 2.6549944574454984e-06, + "loss": 0.46, + "step": 9387 + }, + { + "epoch": 3.47, + "learning_rate": 2.654957507082153e-06, + "loss": 0.6219, + "step": 9388 + }, + { + "epoch": 3.47, + "learning_rate": 2.654920556718808e-06, + "loss": 0.5967, + "step": 9389 + }, + { + "epoch": 3.47, + "learning_rate": 2.6548836063554627e-06, + "loss": 0.5167, + "step": 9390 + }, + { + "epoch": 3.47, + "learning_rate": 2.6548466559921173e-06, + "loss": 0.5063, + "step": 9391 + }, + { + "epoch": 3.47, + "learning_rate": 2.654809705628772e-06, + "loss": 0.5669, + "step": 9392 + }, + { + "epoch": 3.47, + "learning_rate": 2.6547727552654266e-06, + "loss": 0.4855, + "step": 9393 + }, + { + "epoch": 3.47, + "learning_rate": 2.6547358049020817e-06, + "loss": 0.5081, + "step": 9394 + }, + { + "epoch": 3.47, + "learning_rate": 2.6546988545387363e-06, + "loss": 0.5283, + "step": 9395 + }, + { + "epoch": 3.47, + "learning_rate": 2.6546619041753914e-06, + "loss": 0.5044, + "step": 9396 + }, + { + "epoch": 3.47, + "learning_rate": 2.654624953812046e-06, + "loss": 0.4107, + "step": 9397 + }, + { + "epoch": 3.47, + "learning_rate": 2.6545880034487006e-06, + "loss": 0.6156, + "step": 9398 + }, + { + "epoch": 3.47, + "learning_rate": 2.6545510530853553e-06, + "loss": 0.4334, + "step": 9399 + }, + { + "epoch": 3.47, + "learning_rate": 2.65451410272201e-06, + "loss": 0.597, + "step": 9400 + }, + { + "epoch": 3.47, + "learning_rate": 2.654477152358665e-06, + "loss": 0.4887, + "step": 9401 + }, + { + "epoch": 3.47, + "learning_rate": 2.6544402019953196e-06, + "loss": 0.4704, + "step": 9402 + }, + { + "epoch": 3.47, + "learning_rate": 2.6544032516319747e-06, + "loss": 0.5577, + "step": 9403 + }, + { + "epoch": 3.47, + "learning_rate": 2.6543663012686293e-06, + "loss": 0.6342, + "step": 9404 + }, + { + "epoch": 3.47, + "learning_rate": 2.654329350905284e-06, + "loss": 0.4694, + "step": 9405 + }, + { + "epoch": 3.47, + "learning_rate": 2.6542924005419386e-06, + "loss": 0.5284, + "step": 9406 + }, + { + "epoch": 3.47, + "learning_rate": 2.6542554501785936e-06, + "loss": 0.5655, + "step": 9407 + }, + { + "epoch": 3.47, + "learning_rate": 2.6542184998152483e-06, + "loss": 0.679, + "step": 9408 + }, + { + "epoch": 3.47, + "learning_rate": 2.6541815494519033e-06, + "loss": 0.5754, + "step": 9409 + }, + { + "epoch": 3.47, + "learning_rate": 2.654144599088558e-06, + "loss": 0.5488, + "step": 9410 + }, + { + "epoch": 3.48, + "learning_rate": 2.6541076487252126e-06, + "loss": 0.6194, + "step": 9411 + }, + { + "epoch": 3.48, + "learning_rate": 2.6540706983618672e-06, + "loss": 0.5829, + "step": 9412 + }, + { + "epoch": 3.48, + "learning_rate": 2.654033747998522e-06, + "loss": 0.4686, + "step": 9413 + }, + { + "epoch": 3.48, + "learning_rate": 2.653996797635177e-06, + "loss": 0.567, + "step": 9414 + }, + { + "epoch": 3.48, + "learning_rate": 2.6539598472718316e-06, + "loss": 0.5519, + "step": 9415 + }, + { + "epoch": 3.48, + "learning_rate": 2.6539228969084866e-06, + "loss": 0.6458, + "step": 9416 + }, + { + "epoch": 3.48, + "learning_rate": 2.6538859465451413e-06, + "loss": 0.5434, + "step": 9417 + }, + { + "epoch": 3.48, + "learning_rate": 2.653848996181796e-06, + "loss": 0.4612, + "step": 9418 + }, + { + "epoch": 3.48, + "learning_rate": 2.6538120458184505e-06, + "loss": 0.553, + "step": 9419 + }, + { + "epoch": 3.48, + "learning_rate": 2.653775095455105e-06, + "loss": 0.6215, + "step": 9420 + }, + { + "epoch": 3.48, + "learning_rate": 2.6537381450917602e-06, + "loss": 0.5041, + "step": 9421 + }, + { + "epoch": 3.48, + "learning_rate": 2.653701194728415e-06, + "loss": 0.5518, + "step": 9422 + }, + { + "epoch": 3.48, + "learning_rate": 2.6536642443650695e-06, + "loss": 0.5418, + "step": 9423 + }, + { + "epoch": 3.48, + "learning_rate": 2.653627294001724e-06, + "loss": 0.5564, + "step": 9424 + }, + { + "epoch": 3.48, + "learning_rate": 2.653590343638379e-06, + "loss": 0.4768, + "step": 9425 + }, + { + "epoch": 3.48, + "learning_rate": 2.653553393275034e-06, + "loss": 0.5158, + "step": 9426 + }, + { + "epoch": 3.48, + "learning_rate": 2.653516442911689e-06, + "loss": 0.4874, + "step": 9427 + }, + { + "epoch": 3.48, + "learning_rate": 2.6534794925483435e-06, + "loss": 0.5465, + "step": 9428 + }, + { + "epoch": 3.48, + "learning_rate": 2.653442542184998e-06, + "loss": 0.4395, + "step": 9429 + }, + { + "epoch": 3.48, + "learning_rate": 2.653405591821653e-06, + "loss": 0.5803, + "step": 9430 + }, + { + "epoch": 3.48, + "learning_rate": 2.6533686414583074e-06, + "loss": 0.6084, + "step": 9431 + }, + { + "epoch": 3.48, + "learning_rate": 2.6533316910949625e-06, + "loss": 0.4795, + "step": 9432 + }, + { + "epoch": 3.48, + "learning_rate": 2.653294740731617e-06, + "loss": 0.5318, + "step": 9433 + }, + { + "epoch": 3.48, + "learning_rate": 2.653257790368272e-06, + "loss": 0.6316, + "step": 9434 + }, + { + "epoch": 3.48, + "learning_rate": 2.653220840004927e-06, + "loss": 0.5497, + "step": 9435 + }, + { + "epoch": 3.48, + "learning_rate": 2.6531838896415815e-06, + "loss": 0.5892, + "step": 9436 + }, + { + "epoch": 3.48, + "learning_rate": 2.653146939278236e-06, + "loss": 0.6049, + "step": 9437 + }, + { + "epoch": 3.49, + "learning_rate": 2.653109988914891e-06, + "loss": 0.6553, + "step": 9438 + }, + { + "epoch": 3.49, + "learning_rate": 2.653073038551546e-06, + "loss": 0.5215, + "step": 9439 + }, + { + "epoch": 3.49, + "learning_rate": 2.6530360881882004e-06, + "loss": 0.5242, + "step": 9440 + }, + { + "epoch": 3.49, + "learning_rate": 2.6529991378248555e-06, + "loss": 0.4665, + "step": 9441 + }, + { + "epoch": 3.49, + "learning_rate": 2.65296218746151e-06, + "loss": 0.5449, + "step": 9442 + }, + { + "epoch": 3.49, + "learning_rate": 2.6529252370981648e-06, + "loss": 0.5431, + "step": 9443 + }, + { + "epoch": 3.49, + "learning_rate": 2.6528882867348194e-06, + "loss": 0.5325, + "step": 9444 + }, + { + "epoch": 3.49, + "learning_rate": 2.6528513363714745e-06, + "loss": 0.5392, + "step": 9445 + }, + { + "epoch": 3.49, + "learning_rate": 2.652814386008129e-06, + "loss": 0.5316, + "step": 9446 + }, + { + "epoch": 3.49, + "learning_rate": 2.652777435644784e-06, + "loss": 0.6055, + "step": 9447 + }, + { + "epoch": 3.49, + "learning_rate": 2.652740485281439e-06, + "loss": 0.6207, + "step": 9448 + }, + { + "epoch": 3.49, + "learning_rate": 2.6527035349180934e-06, + "loss": 0.4585, + "step": 9449 + }, + { + "epoch": 3.49, + "learning_rate": 2.652666584554748e-06, + "loss": 0.5445, + "step": 9450 + }, + { + "epoch": 3.49, + "learning_rate": 2.6526296341914027e-06, + "loss": 0.6538, + "step": 9451 + }, + { + "epoch": 3.49, + "learning_rate": 2.6525926838280578e-06, + "loss": 0.4398, + "step": 9452 + }, + { + "epoch": 3.49, + "learning_rate": 2.6525557334647124e-06, + "loss": 0.614, + "step": 9453 + }, + { + "epoch": 3.49, + "learning_rate": 2.6525187831013675e-06, + "loss": 0.5904, + "step": 9454 + }, + { + "epoch": 3.49, + "learning_rate": 2.652481832738022e-06, + "loss": 0.5736, + "step": 9455 + }, + { + "epoch": 3.49, + "learning_rate": 2.6524448823746767e-06, + "loss": 0.5798, + "step": 9456 + }, + { + "epoch": 3.49, + "learning_rate": 2.6524079320113314e-06, + "loss": 0.5496, + "step": 9457 + }, + { + "epoch": 3.49, + "learning_rate": 2.6523709816479864e-06, + "loss": 0.6591, + "step": 9458 + }, + { + "epoch": 3.49, + "learning_rate": 2.652334031284641e-06, + "loss": 0.4859, + "step": 9459 + }, + { + "epoch": 3.49, + "learning_rate": 2.6522970809212957e-06, + "loss": 0.5179, + "step": 9460 + }, + { + "epoch": 3.49, + "learning_rate": 2.6522601305579508e-06, + "loss": 0.6024, + "step": 9461 + }, + { + "epoch": 3.49, + "learning_rate": 2.652223180194605e-06, + "loss": 0.5028, + "step": 9462 + }, + { + "epoch": 3.49, + "learning_rate": 2.65218622983126e-06, + "loss": 0.4925, + "step": 9463 + }, + { + "epoch": 3.49, + "learning_rate": 2.6521492794679147e-06, + "loss": 0.5073, + "step": 9464 + }, + { + "epoch": 3.5, + "learning_rate": 2.6521123291045697e-06, + "loss": 0.4973, + "step": 9465 + }, + { + "epoch": 3.5, + "learning_rate": 2.6520753787412244e-06, + "loss": 0.3973, + "step": 9466 + }, + { + "epoch": 3.5, + "learning_rate": 2.6520384283778794e-06, + "loss": 0.5875, + "step": 9467 + }, + { + "epoch": 3.5, + "learning_rate": 2.6520014780145336e-06, + "loss": 0.5095, + "step": 9468 + }, + { + "epoch": 3.5, + "learning_rate": 2.6519645276511887e-06, + "loss": 0.6784, + "step": 9469 + }, + { + "epoch": 3.5, + "learning_rate": 2.6519275772878433e-06, + "loss": 0.4751, + "step": 9470 + }, + { + "epoch": 3.5, + "learning_rate": 2.651890626924498e-06, + "loss": 0.6044, + "step": 9471 + }, + { + "epoch": 3.5, + "learning_rate": 2.651853676561153e-06, + "loss": 0.6432, + "step": 9472 + }, + { + "epoch": 3.5, + "learning_rate": 2.6518167261978077e-06, + "loss": 0.5153, + "step": 9473 + }, + { + "epoch": 3.5, + "learning_rate": 2.6517797758344623e-06, + "loss": 0.6742, + "step": 9474 + }, + { + "epoch": 3.5, + "learning_rate": 2.651742825471117e-06, + "loss": 0.5837, + "step": 9475 + }, + { + "epoch": 3.5, + "learning_rate": 2.651705875107772e-06, + "loss": 0.5805, + "step": 9476 + }, + { + "epoch": 3.5, + "learning_rate": 2.6516689247444266e-06, + "loss": 0.505, + "step": 9477 + }, + { + "epoch": 3.5, + "learning_rate": 2.6516319743810817e-06, + "loss": 0.4626, + "step": 9478 + }, + { + "epoch": 3.5, + "learning_rate": 2.6515950240177363e-06, + "loss": 0.5058, + "step": 9479 + }, + { + "epoch": 3.5, + "learning_rate": 2.651558073654391e-06, + "loss": 0.5384, + "step": 9480 + }, + { + "epoch": 3.5, + "learning_rate": 2.6515211232910456e-06, + "loss": 0.5233, + "step": 9481 + }, + { + "epoch": 3.5, + "learning_rate": 2.6514841729277003e-06, + "loss": 0.5664, + "step": 9482 + }, + { + "epoch": 3.5, + "learning_rate": 2.6514472225643553e-06, + "loss": 0.4971, + "step": 9483 + }, + { + "epoch": 3.5, + "learning_rate": 2.65141027220101e-06, + "loss": 0.6409, + "step": 9484 + }, + { + "epoch": 3.5, + "learning_rate": 2.651373321837665e-06, + "loss": 0.5914, + "step": 9485 + }, + { + "epoch": 3.5, + "learning_rate": 2.6513363714743196e-06, + "loss": 0.686, + "step": 9486 + }, + { + "epoch": 3.5, + "learning_rate": 2.6512994211109743e-06, + "loss": 0.6246, + "step": 9487 + }, + { + "epoch": 3.5, + "learning_rate": 2.651262470747629e-06, + "loss": 0.6778, + "step": 9488 + }, + { + "epoch": 3.5, + "learning_rate": 2.651225520384284e-06, + "loss": 0.576, + "step": 9489 + }, + { + "epoch": 3.5, + "learning_rate": 2.6511885700209386e-06, + "loss": 0.5878, + "step": 9490 + }, + { + "epoch": 3.5, + "learning_rate": 2.6511516196575933e-06, + "loss": 0.556, + "step": 9491 + }, + { + "epoch": 3.51, + "learning_rate": 2.6511146692942483e-06, + "loss": 0.4935, + "step": 9492 + }, + { + "epoch": 3.51, + "learning_rate": 2.651077718930903e-06, + "loss": 0.5653, + "step": 9493 + }, + { + "epoch": 3.51, + "learning_rate": 2.6510407685675576e-06, + "loss": 0.5471, + "step": 9494 + }, + { + "epoch": 3.51, + "learning_rate": 2.6510038182042122e-06, + "loss": 0.6087, + "step": 9495 + }, + { + "epoch": 3.51, + "learning_rate": 2.6509668678408673e-06, + "loss": 0.51, + "step": 9496 + }, + { + "epoch": 3.51, + "learning_rate": 2.650929917477522e-06, + "loss": 0.5728, + "step": 9497 + }, + { + "epoch": 3.51, + "learning_rate": 2.650892967114177e-06, + "loss": 0.5196, + "step": 9498 + }, + { + "epoch": 3.51, + "learning_rate": 2.6508560167508316e-06, + "loss": 0.5227, + "step": 9499 + }, + { + "epoch": 3.51, + "learning_rate": 2.6508190663874862e-06, + "loss": 0.4595, + "step": 9500 + }, + { + "epoch": 3.51, + "learning_rate": 2.650782116024141e-06, + "loss": 0.5988, + "step": 9501 + }, + { + "epoch": 3.51, + "learning_rate": 2.6507451656607955e-06, + "loss": 0.5602, + "step": 9502 + }, + { + "epoch": 3.51, + "learning_rate": 2.6507082152974506e-06, + "loss": 0.5687, + "step": 9503 + }, + { + "epoch": 3.51, + "learning_rate": 2.6506712649341052e-06, + "loss": 0.5561, + "step": 9504 + }, + { + "epoch": 3.51, + "learning_rate": 2.6506343145707603e-06, + "loss": 0.5237, + "step": 9505 + }, + { + "epoch": 3.51, + "learning_rate": 2.650597364207415e-06, + "loss": 0.5076, + "step": 9506 + }, + { + "epoch": 3.51, + "learning_rate": 2.6505604138440696e-06, + "loss": 0.5219, + "step": 9507 + }, + { + "epoch": 3.51, + "learning_rate": 2.650523463480724e-06, + "loss": 0.5573, + "step": 9508 + }, + { + "epoch": 3.51, + "learning_rate": 2.6504865131173792e-06, + "loss": 0.6434, + "step": 9509 + }, + { + "epoch": 3.51, + "learning_rate": 2.650449562754034e-06, + "loss": 0.5445, + "step": 9510 + }, + { + "epoch": 3.51, + "learning_rate": 2.6504126123906885e-06, + "loss": 0.5594, + "step": 9511 + }, + { + "epoch": 3.51, + "learning_rate": 2.6503756620273436e-06, + "loss": 0.4872, + "step": 9512 + }, + { + "epoch": 3.51, + "learning_rate": 2.650338711663998e-06, + "loss": 0.5923, + "step": 9513 + }, + { + "epoch": 3.51, + "learning_rate": 2.650301761300653e-06, + "loss": 0.5157, + "step": 9514 + }, + { + "epoch": 3.51, + "learning_rate": 2.6502648109373075e-06, + "loss": 0.6163, + "step": 9515 + }, + { + "epoch": 3.51, + "learning_rate": 2.6502278605739625e-06, + "loss": 0.6718, + "step": 9516 + }, + { + "epoch": 3.51, + "learning_rate": 2.650190910210617e-06, + "loss": 0.588, + "step": 9517 + }, + { + "epoch": 3.51, + "learning_rate": 2.6501539598472722e-06, + "loss": 0.5764, + "step": 9518 + }, + { + "epoch": 3.52, + "learning_rate": 2.6501170094839265e-06, + "loss": 0.5722, + "step": 9519 + }, + { + "epoch": 3.52, + "learning_rate": 2.650080059120581e-06, + "loss": 0.6031, + "step": 9520 + }, + { + "epoch": 3.52, + "learning_rate": 2.650043108757236e-06, + "loss": 0.6016, + "step": 9521 + }, + { + "epoch": 3.52, + "learning_rate": 2.650006158393891e-06, + "loss": 0.6147, + "step": 9522 + }, + { + "epoch": 3.52, + "learning_rate": 2.649969208030546e-06, + "loss": 0.677, + "step": 9523 + }, + { + "epoch": 3.52, + "learning_rate": 2.6499322576672005e-06, + "loss": 0.6264, + "step": 9524 + }, + { + "epoch": 3.52, + "learning_rate": 2.649895307303855e-06, + "loss": 0.4798, + "step": 9525 + }, + { + "epoch": 3.52, + "learning_rate": 2.6498583569405098e-06, + "loss": 0.5171, + "step": 9526 + }, + { + "epoch": 3.52, + "learning_rate": 2.649821406577165e-06, + "loss": 0.5747, + "step": 9527 + }, + { + "epoch": 3.52, + "learning_rate": 2.6497844562138195e-06, + "loss": 0.5629, + "step": 9528 + }, + { + "epoch": 3.52, + "learning_rate": 2.6497475058504745e-06, + "loss": 0.5185, + "step": 9529 + }, + { + "epoch": 3.52, + "learning_rate": 2.649710555487129e-06, + "loss": 0.5013, + "step": 9530 + }, + { + "epoch": 3.52, + "learning_rate": 2.649673605123784e-06, + "loss": 0.5165, + "step": 9531 + }, + { + "epoch": 3.52, + "learning_rate": 2.6496366547604384e-06, + "loss": 0.5134, + "step": 9532 + }, + { + "epoch": 3.52, + "learning_rate": 2.649599704397093e-06, + "loss": 0.5317, + "step": 9533 + }, + { + "epoch": 3.52, + "learning_rate": 2.649562754033748e-06, + "loss": 0.5109, + "step": 9534 + }, + { + "epoch": 3.52, + "learning_rate": 2.6495258036704028e-06, + "loss": 0.6122, + "step": 9535 + }, + { + "epoch": 3.52, + "learning_rate": 2.649488853307058e-06, + "loss": 0.6106, + "step": 9536 + }, + { + "epoch": 3.52, + "learning_rate": 2.6494519029437125e-06, + "loss": 0.4663, + "step": 9537 + }, + { + "epoch": 3.52, + "learning_rate": 2.649414952580367e-06, + "loss": 0.4996, + "step": 9538 + }, + { + "epoch": 3.52, + "learning_rate": 2.6493780022170217e-06, + "loss": 0.6032, + "step": 9539 + }, + { + "epoch": 3.52, + "learning_rate": 2.6493410518536764e-06, + "loss": 0.6133, + "step": 9540 + }, + { + "epoch": 3.52, + "learning_rate": 2.6493041014903314e-06, + "loss": 0.5395, + "step": 9541 + }, + { + "epoch": 3.52, + "learning_rate": 2.649267151126986e-06, + "loss": 0.5786, + "step": 9542 + }, + { + "epoch": 3.52, + "learning_rate": 2.649230200763641e-06, + "loss": 0.5435, + "step": 9543 + }, + { + "epoch": 3.52, + "learning_rate": 2.6491932504002958e-06, + "loss": 0.5296, + "step": 9544 + }, + { + "epoch": 3.52, + "learning_rate": 2.6491563000369504e-06, + "loss": 0.4891, + "step": 9545 + }, + { + "epoch": 3.53, + "learning_rate": 2.649119349673605e-06, + "loss": 0.5521, + "step": 9546 + }, + { + "epoch": 3.53, + "learning_rate": 2.64908239931026e-06, + "loss": 0.6211, + "step": 9547 + }, + { + "epoch": 3.53, + "learning_rate": 2.6490454489469147e-06, + "loss": 0.5312, + "step": 9548 + }, + { + "epoch": 3.53, + "learning_rate": 2.6490084985835698e-06, + "loss": 0.704, + "step": 9549 + }, + { + "epoch": 3.53, + "learning_rate": 2.6489715482202244e-06, + "loss": 0.5195, + "step": 9550 + }, + { + "epoch": 3.53, + "learning_rate": 2.648934597856879e-06, + "loss": 0.6188, + "step": 9551 + }, + { + "epoch": 3.53, + "learning_rate": 2.6488976474935337e-06, + "loss": 0.4872, + "step": 9552 + }, + { + "epoch": 3.53, + "learning_rate": 2.6488606971301883e-06, + "loss": 0.5329, + "step": 9553 + }, + { + "epoch": 3.53, + "learning_rate": 2.6488237467668434e-06, + "loss": 0.5504, + "step": 9554 + }, + { + "epoch": 3.53, + "learning_rate": 2.648786796403498e-06, + "loss": 0.553, + "step": 9555 + }, + { + "epoch": 3.53, + "learning_rate": 2.648749846040153e-06, + "loss": 0.5319, + "step": 9556 + }, + { + "epoch": 3.53, + "learning_rate": 2.6487128956768077e-06, + "loss": 0.5428, + "step": 9557 + }, + { + "epoch": 3.53, + "learning_rate": 2.6486759453134624e-06, + "loss": 0.506, + "step": 9558 + }, + { + "epoch": 3.53, + "learning_rate": 2.648638994950117e-06, + "loss": 0.3758, + "step": 9559 + }, + { + "epoch": 3.53, + "learning_rate": 2.6486020445867716e-06, + "loss": 0.5376, + "step": 9560 + }, + { + "epoch": 3.53, + "learning_rate": 2.6485650942234267e-06, + "loss": 0.4931, + "step": 9561 + }, + { + "epoch": 3.53, + "learning_rate": 2.6485281438600813e-06, + "loss": 0.4307, + "step": 9562 + }, + { + "epoch": 3.53, + "learning_rate": 2.6484911934967364e-06, + "loss": 0.5628, + "step": 9563 + }, + { + "epoch": 3.53, + "learning_rate": 2.6484542431333906e-06, + "loss": 0.4843, + "step": 9564 + }, + { + "epoch": 3.53, + "learning_rate": 2.6484172927700457e-06, + "loss": 0.4806, + "step": 9565 + }, + { + "epoch": 3.53, + "learning_rate": 2.6483803424067003e-06, + "loss": 0.5206, + "step": 9566 + }, + { + "epoch": 3.53, + "learning_rate": 2.6483433920433554e-06, + "loss": 0.5167, + "step": 9567 + }, + { + "epoch": 3.53, + "learning_rate": 2.64830644168001e-06, + "loss": 0.6857, + "step": 9568 + }, + { + "epoch": 3.53, + "learning_rate": 2.6482694913166646e-06, + "loss": 0.588, + "step": 9569 + }, + { + "epoch": 3.53, + "learning_rate": 2.6482325409533193e-06, + "loss": 0.6101, + "step": 9570 + }, + { + "epoch": 3.53, + "learning_rate": 2.648195590589974e-06, + "loss": 0.5782, + "step": 9571 + }, + { + "epoch": 3.53, + "learning_rate": 2.648158640226629e-06, + "loss": 0.4554, + "step": 9572 + }, + { + "epoch": 3.54, + "learning_rate": 2.6481216898632836e-06, + "loss": 0.6088, + "step": 9573 + }, + { + "epoch": 3.54, + "learning_rate": 2.6480847394999387e-06, + "loss": 0.6195, + "step": 9574 + }, + { + "epoch": 3.54, + "learning_rate": 2.6480477891365933e-06, + "loss": 0.5648, + "step": 9575 + }, + { + "epoch": 3.54, + "learning_rate": 2.648010838773248e-06, + "loss": 0.4989, + "step": 9576 + }, + { + "epoch": 3.54, + "learning_rate": 2.6479738884099026e-06, + "loss": 0.5786, + "step": 9577 + }, + { + "epoch": 3.54, + "learning_rate": 2.6479369380465576e-06, + "loss": 0.491, + "step": 9578 + }, + { + "epoch": 3.54, + "learning_rate": 2.6478999876832123e-06, + "loss": 0.5354, + "step": 9579 + }, + { + "epoch": 3.54, + "learning_rate": 2.647863037319867e-06, + "loss": 0.6286, + "step": 9580 + }, + { + "epoch": 3.54, + "learning_rate": 2.647826086956522e-06, + "loss": 0.5348, + "step": 9581 + }, + { + "epoch": 3.54, + "learning_rate": 2.6477891365931766e-06, + "loss": 0.6317, + "step": 9582 + }, + { + "epoch": 3.54, + "learning_rate": 2.6477521862298312e-06, + "loss": 0.6796, + "step": 9583 + }, + { + "epoch": 3.54, + "learning_rate": 2.647715235866486e-06, + "loss": 0.5958, + "step": 9584 + }, + { + "epoch": 3.54, + "learning_rate": 2.647678285503141e-06, + "loss": 0.5105, + "step": 9585 + }, + { + "epoch": 3.54, + "learning_rate": 2.6476413351397956e-06, + "loss": 0.5247, + "step": 9586 + }, + { + "epoch": 3.54, + "learning_rate": 2.6476043847764506e-06, + "loss": 0.535, + "step": 9587 + }, + { + "epoch": 3.54, + "learning_rate": 2.6475674344131053e-06, + "loss": 0.5966, + "step": 9588 + }, + { + "epoch": 3.54, + "learning_rate": 2.64753048404976e-06, + "loss": 0.5914, + "step": 9589 + }, + { + "epoch": 3.54, + "learning_rate": 2.6474935336864145e-06, + "loss": 0.555, + "step": 9590 + }, + { + "epoch": 3.54, + "learning_rate": 2.647456583323069e-06, + "loss": 0.4991, + "step": 9591 + }, + { + "epoch": 3.54, + "learning_rate": 2.6474196329597242e-06, + "loss": 0.5786, + "step": 9592 + }, + { + "epoch": 3.54, + "learning_rate": 2.647382682596379e-06, + "loss": 0.6465, + "step": 9593 + }, + { + "epoch": 3.54, + "learning_rate": 2.647345732233034e-06, + "loss": 0.6737, + "step": 9594 + }, + { + "epoch": 3.54, + "learning_rate": 2.6473087818696886e-06, + "loss": 0.6887, + "step": 9595 + }, + { + "epoch": 3.54, + "learning_rate": 2.647271831506343e-06, + "loss": 0.4782, + "step": 9596 + }, + { + "epoch": 3.54, + "learning_rate": 2.647234881142998e-06, + "loss": 0.6247, + "step": 9597 + }, + { + "epoch": 3.54, + "learning_rate": 2.647197930779653e-06, + "loss": 0.5478, + "step": 9598 + }, + { + "epoch": 3.54, + "learning_rate": 2.6471609804163075e-06, + "loss": 0.4707, + "step": 9599 + }, + { + "epoch": 3.55, + "learning_rate": 2.647124030052962e-06, + "loss": 0.605, + "step": 9600 + }, + { + "epoch": 3.55, + "learning_rate": 2.6470870796896172e-06, + "loss": 0.642, + "step": 9601 + }, + { + "epoch": 3.55, + "learning_rate": 2.647050129326272e-06, + "loss": 0.4633, + "step": 9602 + }, + { + "epoch": 3.55, + "learning_rate": 2.6470131789629265e-06, + "loss": 0.6145, + "step": 9603 + }, + { + "epoch": 3.55, + "learning_rate": 2.646976228599581e-06, + "loss": 0.7256, + "step": 9604 + }, + { + "epoch": 3.55, + "learning_rate": 2.646939278236236e-06, + "loss": 0.4991, + "step": 9605 + }, + { + "epoch": 3.55, + "learning_rate": 2.646902327872891e-06, + "loss": 0.547, + "step": 9606 + }, + { + "epoch": 3.55, + "learning_rate": 2.646865377509546e-06, + "loss": 0.5658, + "step": 9607 + }, + { + "epoch": 3.55, + "learning_rate": 2.6468284271462005e-06, + "loss": 0.5445, + "step": 9608 + }, + { + "epoch": 3.55, + "learning_rate": 2.6467914767828547e-06, + "loss": 0.5572, + "step": 9609 + }, + { + "epoch": 3.55, + "learning_rate": 2.64675452641951e-06, + "loss": 0.5146, + "step": 9610 + }, + { + "epoch": 3.55, + "learning_rate": 2.6467175760561644e-06, + "loss": 0.5165, + "step": 9611 + }, + { + "epoch": 3.55, + "learning_rate": 2.6466806256928195e-06, + "loss": 0.4821, + "step": 9612 + }, + { + "epoch": 3.55, + "learning_rate": 2.646643675329474e-06, + "loss": 0.5045, + "step": 9613 + }, + { + "epoch": 3.55, + "learning_rate": 2.6466067249661288e-06, + "loss": 0.579, + "step": 9614 + }, + { + "epoch": 3.55, + "learning_rate": 2.6465697746027834e-06, + "loss": 0.5054, + "step": 9615 + }, + { + "epoch": 3.55, + "learning_rate": 2.6465328242394385e-06, + "loss": 0.5351, + "step": 9616 + }, + { + "epoch": 3.55, + "learning_rate": 2.646495873876093e-06, + "loss": 0.5786, + "step": 9617 + }, + { + "epoch": 3.55, + "learning_rate": 2.646458923512748e-06, + "loss": 0.5721, + "step": 9618 + }, + { + "epoch": 3.55, + "learning_rate": 2.646421973149403e-06, + "loss": 0.5532, + "step": 9619 + }, + { + "epoch": 3.55, + "learning_rate": 2.6463850227860574e-06, + "loss": 0.5829, + "step": 9620 + }, + { + "epoch": 3.55, + "learning_rate": 2.646348072422712e-06, + "loss": 0.605, + "step": 9621 + }, + { + "epoch": 3.55, + "learning_rate": 2.6463111220593667e-06, + "loss": 0.5555, + "step": 9622 + }, + { + "epoch": 3.55, + "learning_rate": 2.6462741716960218e-06, + "loss": 0.5566, + "step": 9623 + }, + { + "epoch": 3.55, + "learning_rate": 2.6462372213326764e-06, + "loss": 0.541, + "step": 9624 + }, + { + "epoch": 3.55, + "learning_rate": 2.6462002709693315e-06, + "loss": 0.5729, + "step": 9625 + }, + { + "epoch": 3.55, + "learning_rate": 2.646163320605986e-06, + "loss": 0.5141, + "step": 9626 + }, + { + "epoch": 3.56, + "learning_rate": 2.6461263702426407e-06, + "loss": 0.5804, + "step": 9627 + }, + { + "epoch": 3.56, + "learning_rate": 2.6460894198792954e-06, + "loss": 0.4553, + "step": 9628 + }, + { + "epoch": 3.56, + "learning_rate": 2.64605246951595e-06, + "loss": 0.5312, + "step": 9629 + }, + { + "epoch": 3.56, + "learning_rate": 2.646015519152605e-06, + "loss": 0.656, + "step": 9630 + }, + { + "epoch": 3.56, + "learning_rate": 2.6459785687892597e-06, + "loss": 0.6194, + "step": 9631 + }, + { + "epoch": 3.56, + "learning_rate": 2.6459416184259148e-06, + "loss": 0.5452, + "step": 9632 + }, + { + "epoch": 3.56, + "learning_rate": 2.6459046680625694e-06, + "loss": 0.5219, + "step": 9633 + }, + { + "epoch": 3.56, + "learning_rate": 2.645867717699224e-06, + "loss": 0.5771, + "step": 9634 + }, + { + "epoch": 3.56, + "learning_rate": 2.6458307673358787e-06, + "loss": 0.5406, + "step": 9635 + }, + { + "epoch": 3.56, + "learning_rate": 2.6457938169725337e-06, + "loss": 0.4593, + "step": 9636 + }, + { + "epoch": 3.56, + "learning_rate": 2.6457568666091884e-06, + "loss": 0.5038, + "step": 9637 + }, + { + "epoch": 3.56, + "learning_rate": 2.6457199162458434e-06, + "loss": 0.5974, + "step": 9638 + }, + { + "epoch": 3.56, + "learning_rate": 2.645682965882498e-06, + "loss": 0.627, + "step": 9639 + }, + { + "epoch": 3.56, + "learning_rate": 2.6456460155191527e-06, + "loss": 0.5743, + "step": 9640 + }, + { + "epoch": 3.56, + "learning_rate": 2.6456090651558074e-06, + "loss": 0.4596, + "step": 9641 + }, + { + "epoch": 3.56, + "learning_rate": 2.645572114792462e-06, + "loss": 0.5425, + "step": 9642 + }, + { + "epoch": 3.56, + "learning_rate": 2.645535164429117e-06, + "loss": 0.4629, + "step": 9643 + }, + { + "epoch": 3.56, + "learning_rate": 2.6454982140657717e-06, + "loss": 0.5793, + "step": 9644 + }, + { + "epoch": 3.56, + "learning_rate": 2.6454612637024267e-06, + "loss": 0.5482, + "step": 9645 + }, + { + "epoch": 3.56, + "learning_rate": 2.6454243133390814e-06, + "loss": 0.5645, + "step": 9646 + }, + { + "epoch": 3.56, + "learning_rate": 2.645387362975736e-06, + "loss": 0.5024, + "step": 9647 + }, + { + "epoch": 3.56, + "learning_rate": 2.6453504126123907e-06, + "loss": 0.6297, + "step": 9648 + }, + { + "epoch": 3.56, + "learning_rate": 2.6453134622490453e-06, + "loss": 0.5669, + "step": 9649 + }, + { + "epoch": 3.56, + "learning_rate": 2.6452765118857003e-06, + "loss": 0.4988, + "step": 9650 + }, + { + "epoch": 3.56, + "learning_rate": 2.645239561522355e-06, + "loss": 0.5203, + "step": 9651 + }, + { + "epoch": 3.56, + "learning_rate": 2.64520261115901e-06, + "loss": 0.5019, + "step": 9652 + }, + { + "epoch": 3.56, + "learning_rate": 2.6451656607956643e-06, + "loss": 0.4605, + "step": 9653 + }, + { + "epoch": 3.56, + "learning_rate": 2.6451287104323193e-06, + "loss": 0.5364, + "step": 9654 + }, + { + "epoch": 3.57, + "learning_rate": 2.645091760068974e-06, + "loss": 0.5717, + "step": 9655 + }, + { + "epoch": 3.57, + "learning_rate": 2.645054809705629e-06, + "loss": 0.5147, + "step": 9656 + }, + { + "epoch": 3.57, + "learning_rate": 2.6450178593422837e-06, + "loss": 0.6287, + "step": 9657 + }, + { + "epoch": 3.57, + "learning_rate": 2.6449809089789387e-06, + "loss": 0.6027, + "step": 9658 + }, + { + "epoch": 3.57, + "learning_rate": 2.644943958615593e-06, + "loss": 0.5286, + "step": 9659 + }, + { + "epoch": 3.57, + "learning_rate": 2.6449070082522476e-06, + "loss": 0.7801, + "step": 9660 + }, + { + "epoch": 3.57, + "learning_rate": 2.6448700578889026e-06, + "loss": 0.6133, + "step": 9661 + }, + { + "epoch": 3.57, + "learning_rate": 2.6448331075255573e-06, + "loss": 0.6015, + "step": 9662 + }, + { + "epoch": 3.57, + "learning_rate": 2.6447961571622123e-06, + "loss": 0.4809, + "step": 9663 + }, + { + "epoch": 3.57, + "learning_rate": 2.644759206798867e-06, + "loss": 0.5698, + "step": 9664 + }, + { + "epoch": 3.57, + "learning_rate": 2.6447222564355216e-06, + "loss": 0.5571, + "step": 9665 + }, + { + "epoch": 3.57, + "learning_rate": 2.6446853060721762e-06, + "loss": 0.5614, + "step": 9666 + }, + { + "epoch": 3.57, + "learning_rate": 2.6446483557088313e-06, + "loss": 0.4928, + "step": 9667 + }, + { + "epoch": 3.57, + "learning_rate": 2.644611405345486e-06, + "loss": 0.5202, + "step": 9668 + }, + { + "epoch": 3.57, + "learning_rate": 2.6445744549821406e-06, + "loss": 0.5834, + "step": 9669 + }, + { + "epoch": 3.57, + "learning_rate": 2.6445375046187956e-06, + "loss": 0.6878, + "step": 9670 + }, + { + "epoch": 3.57, + "learning_rate": 2.6445005542554503e-06, + "loss": 0.5349, + "step": 9671 + }, + { + "epoch": 3.57, + "learning_rate": 2.644463603892105e-06, + "loss": 0.4989, + "step": 9672 + }, + { + "epoch": 3.57, + "learning_rate": 2.6444266535287595e-06, + "loss": 0.6067, + "step": 9673 + }, + { + "epoch": 3.57, + "learning_rate": 2.6443897031654146e-06, + "loss": 0.5312, + "step": 9674 + }, + { + "epoch": 3.57, + "learning_rate": 2.6443527528020692e-06, + "loss": 0.575, + "step": 9675 + }, + { + "epoch": 3.57, + "learning_rate": 2.6443158024387243e-06, + "loss": 0.4995, + "step": 9676 + }, + { + "epoch": 3.57, + "learning_rate": 2.644278852075379e-06, + "loss": 0.5241, + "step": 9677 + }, + { + "epoch": 3.57, + "learning_rate": 2.6442419017120336e-06, + "loss": 0.5801, + "step": 9678 + }, + { + "epoch": 3.57, + "learning_rate": 2.644204951348688e-06, + "loss": 0.5778, + "step": 9679 + }, + { + "epoch": 3.57, + "learning_rate": 2.644168000985343e-06, + "loss": 0.5459, + "step": 9680 + }, + { + "epoch": 3.57, + "learning_rate": 2.644131050621998e-06, + "loss": 0.5829, + "step": 9681 + }, + { + "epoch": 3.58, + "learning_rate": 2.6440941002586525e-06, + "loss": 0.516, + "step": 9682 + }, + { + "epoch": 3.58, + "learning_rate": 2.6440571498953076e-06, + "loss": 0.6807, + "step": 9683 + }, + { + "epoch": 3.58, + "learning_rate": 2.6440201995319622e-06, + "loss": 0.5541, + "step": 9684 + }, + { + "epoch": 3.58, + "learning_rate": 2.643983249168617e-06, + "loss": 0.6296, + "step": 9685 + }, + { + "epoch": 3.58, + "learning_rate": 2.6439462988052715e-06, + "loss": 0.5928, + "step": 9686 + }, + { + "epoch": 3.58, + "learning_rate": 2.6439093484419266e-06, + "loss": 0.5476, + "step": 9687 + }, + { + "epoch": 3.58, + "learning_rate": 2.643872398078581e-06, + "loss": 0.6139, + "step": 9688 + }, + { + "epoch": 3.58, + "learning_rate": 2.643835447715236e-06, + "loss": 0.5815, + "step": 9689 + }, + { + "epoch": 3.58, + "learning_rate": 2.643798497351891e-06, + "loss": 0.4339, + "step": 9690 + }, + { + "epoch": 3.58, + "learning_rate": 2.6437615469885455e-06, + "loss": 0.5351, + "step": 9691 + }, + { + "epoch": 3.58, + "learning_rate": 2.6437245966252e-06, + "loss": 0.4931, + "step": 9692 + }, + { + "epoch": 3.58, + "learning_rate": 2.643687646261855e-06, + "loss": 0.5051, + "step": 9693 + }, + { + "epoch": 3.58, + "learning_rate": 2.64365069589851e-06, + "loss": 0.5492, + "step": 9694 + }, + { + "epoch": 3.58, + "learning_rate": 2.6436137455351645e-06, + "loss": 0.4921, + "step": 9695 + }, + { + "epoch": 3.58, + "learning_rate": 2.6435767951718196e-06, + "loss": 0.5049, + "step": 9696 + }, + { + "epoch": 3.58, + "learning_rate": 2.643539844808474e-06, + "loss": 0.5724, + "step": 9697 + }, + { + "epoch": 3.58, + "learning_rate": 2.643502894445129e-06, + "loss": 0.6439, + "step": 9698 + }, + { + "epoch": 3.58, + "learning_rate": 2.6434659440817835e-06, + "loss": 0.5709, + "step": 9699 + }, + { + "epoch": 3.58, + "learning_rate": 2.643428993718438e-06, + "loss": 0.5564, + "step": 9700 + }, + { + "epoch": 3.58, + "learning_rate": 2.643392043355093e-06, + "loss": 0.4407, + "step": 9701 + }, + { + "epoch": 3.58, + "learning_rate": 2.643355092991748e-06, + "loss": 0.4997, + "step": 9702 + }, + { + "epoch": 3.58, + "learning_rate": 2.643318142628403e-06, + "loss": 0.5524, + "step": 9703 + }, + { + "epoch": 3.58, + "learning_rate": 2.643281192265057e-06, + "loss": 0.4856, + "step": 9704 + }, + { + "epoch": 3.58, + "learning_rate": 2.643244241901712e-06, + "loss": 0.5348, + "step": 9705 + }, + { + "epoch": 3.58, + "learning_rate": 2.6432072915383668e-06, + "loss": 0.5659, + "step": 9706 + }, + { + "epoch": 3.58, + "learning_rate": 2.643170341175022e-06, + "loss": 0.5471, + "step": 9707 + }, + { + "epoch": 3.58, + "learning_rate": 2.6431333908116765e-06, + "loss": 0.576, + "step": 9708 + }, + { + "epoch": 3.59, + "learning_rate": 2.643096440448331e-06, + "loss": 0.5677, + "step": 9709 + }, + { + "epoch": 3.59, + "learning_rate": 2.6430594900849857e-06, + "loss": 0.4827, + "step": 9710 + }, + { + "epoch": 3.59, + "learning_rate": 2.6430225397216404e-06, + "loss": 0.5257, + "step": 9711 + }, + { + "epoch": 3.59, + "learning_rate": 2.6429855893582954e-06, + "loss": 0.6029, + "step": 9712 + }, + { + "epoch": 3.59, + "learning_rate": 2.64294863899495e-06, + "loss": 0.5312, + "step": 9713 + }, + { + "epoch": 3.59, + "learning_rate": 2.642911688631605e-06, + "loss": 0.4879, + "step": 9714 + }, + { + "epoch": 3.59, + "learning_rate": 2.6428747382682598e-06, + "loss": 0.5409, + "step": 9715 + }, + { + "epoch": 3.59, + "learning_rate": 2.6428377879049144e-06, + "loss": 0.5515, + "step": 9716 + }, + { + "epoch": 3.59, + "learning_rate": 2.642800837541569e-06, + "loss": 0.5474, + "step": 9717 + }, + { + "epoch": 3.59, + "learning_rate": 2.642763887178224e-06, + "loss": 0.622, + "step": 9718 + }, + { + "epoch": 3.59, + "learning_rate": 2.6427269368148787e-06, + "loss": 0.4797, + "step": 9719 + }, + { + "epoch": 3.59, + "learning_rate": 2.6426899864515334e-06, + "loss": 0.498, + "step": 9720 + }, + { + "epoch": 3.59, + "learning_rate": 2.6426530360881884e-06, + "loss": 0.6942, + "step": 9721 + }, + { + "epoch": 3.59, + "learning_rate": 2.642616085724843e-06, + "loss": 0.6705, + "step": 9722 + }, + { + "epoch": 3.59, + "learning_rate": 2.6425791353614977e-06, + "loss": 0.5272, + "step": 9723 + }, + { + "epoch": 3.59, + "learning_rate": 2.6425421849981523e-06, + "loss": 0.5239, + "step": 9724 + }, + { + "epoch": 3.59, + "learning_rate": 2.6425052346348074e-06, + "loss": 0.5934, + "step": 9725 + }, + { + "epoch": 3.59, + "learning_rate": 2.642468284271462e-06, + "loss": 0.4921, + "step": 9726 + }, + { + "epoch": 3.59, + "learning_rate": 2.642431333908117e-06, + "loss": 0.5877, + "step": 9727 + }, + { + "epoch": 3.59, + "learning_rate": 2.6423943835447717e-06, + "loss": 0.6084, + "step": 9728 + }, + { + "epoch": 3.59, + "learning_rate": 2.6423574331814264e-06, + "loss": 0.517, + "step": 9729 + }, + { + "epoch": 3.59, + "learning_rate": 2.642320482818081e-06, + "loss": 0.4938, + "step": 9730 + }, + { + "epoch": 3.59, + "learning_rate": 2.6422835324547356e-06, + "loss": 0.5381, + "step": 9731 + }, + { + "epoch": 3.59, + "learning_rate": 2.6422465820913907e-06, + "loss": 0.5634, + "step": 9732 + }, + { + "epoch": 3.59, + "learning_rate": 2.6422096317280453e-06, + "loss": 0.6455, + "step": 9733 + }, + { + "epoch": 3.59, + "learning_rate": 2.6421726813647004e-06, + "loss": 0.4791, + "step": 9734 + }, + { + "epoch": 3.59, + "learning_rate": 2.642135731001355e-06, + "loss": 0.5906, + "step": 9735 + }, + { + "epoch": 3.6, + "learning_rate": 2.6420987806380097e-06, + "loss": 0.5226, + "step": 9736 + }, + { + "epoch": 3.6, + "learning_rate": 2.6420618302746643e-06, + "loss": 0.6055, + "step": 9737 + }, + { + "epoch": 3.6, + "learning_rate": 2.6420248799113194e-06, + "loss": 0.6367, + "step": 9738 + }, + { + "epoch": 3.6, + "learning_rate": 2.641987929547974e-06, + "loss": 0.4505, + "step": 9739 + }, + { + "epoch": 3.6, + "learning_rate": 2.6419509791846286e-06, + "loss": 0.6025, + "step": 9740 + }, + { + "epoch": 3.6, + "learning_rate": 2.6419140288212837e-06, + "loss": 0.5069, + "step": 9741 + }, + { + "epoch": 3.6, + "learning_rate": 2.6418770784579383e-06, + "loss": 0.4973, + "step": 9742 + }, + { + "epoch": 3.6, + "learning_rate": 2.641840128094593e-06, + "loss": 0.4693, + "step": 9743 + }, + { + "epoch": 3.6, + "learning_rate": 2.6418031777312476e-06, + "loss": 0.4859, + "step": 9744 + }, + { + "epoch": 3.6, + "learning_rate": 2.6417662273679027e-06, + "loss": 0.5359, + "step": 9745 + }, + { + "epoch": 3.6, + "learning_rate": 2.6417292770045573e-06, + "loss": 0.4484, + "step": 9746 + }, + { + "epoch": 3.6, + "learning_rate": 2.6416923266412124e-06, + "loss": 0.5692, + "step": 9747 + }, + { + "epoch": 3.6, + "learning_rate": 2.641655376277867e-06, + "loss": 0.5366, + "step": 9748 + }, + { + "epoch": 3.6, + "learning_rate": 2.6416184259145212e-06, + "loss": 0.5043, + "step": 9749 + }, + { + "epoch": 3.6, + "learning_rate": 2.6415814755511763e-06, + "loss": 0.5918, + "step": 9750 + }, + { + "epoch": 3.6, + "learning_rate": 2.641544525187831e-06, + "loss": 0.5726, + "step": 9751 + }, + { + "epoch": 3.6, + "learning_rate": 2.641507574824486e-06, + "loss": 0.4493, + "step": 9752 + }, + { + "epoch": 3.6, + "learning_rate": 2.6414706244611406e-06, + "loss": 0.59, + "step": 9753 + }, + { + "epoch": 3.6, + "learning_rate": 2.6414336740977957e-06, + "loss": 0.5146, + "step": 9754 + }, + { + "epoch": 3.6, + "learning_rate": 2.64139672373445e-06, + "loss": 0.547, + "step": 9755 + }, + { + "epoch": 3.6, + "learning_rate": 2.641359773371105e-06, + "loss": 0.4905, + "step": 9756 + }, + { + "epoch": 3.6, + "learning_rate": 2.6413228230077596e-06, + "loss": 0.5673, + "step": 9757 + }, + { + "epoch": 3.6, + "learning_rate": 2.6412858726444146e-06, + "loss": 0.6743, + "step": 9758 + }, + { + "epoch": 3.6, + "learning_rate": 2.6412489222810693e-06, + "loss": 0.5716, + "step": 9759 + }, + { + "epoch": 3.6, + "learning_rate": 2.641211971917724e-06, + "loss": 0.6035, + "step": 9760 + }, + { + "epoch": 3.6, + "learning_rate": 2.6411750215543785e-06, + "loss": 0.5924, + "step": 9761 + }, + { + "epoch": 3.6, + "learning_rate": 2.641138071191033e-06, + "loss": 0.5799, + "step": 9762 + }, + { + "epoch": 3.61, + "learning_rate": 2.6411011208276882e-06, + "loss": 0.5781, + "step": 9763 + }, + { + "epoch": 3.61, + "learning_rate": 2.641064170464343e-06, + "loss": 0.4613, + "step": 9764 + }, + { + "epoch": 3.61, + "learning_rate": 2.641027220100998e-06, + "loss": 0.5645, + "step": 9765 + }, + { + "epoch": 3.61, + "learning_rate": 2.6409902697376526e-06, + "loss": 0.4921, + "step": 9766 + }, + { + "epoch": 3.61, + "learning_rate": 2.640953319374307e-06, + "loss": 0.4722, + "step": 9767 + }, + { + "epoch": 3.61, + "learning_rate": 2.640916369010962e-06, + "loss": 0.5251, + "step": 9768 + }, + { + "epoch": 3.61, + "learning_rate": 2.6408794186476165e-06, + "loss": 0.5588, + "step": 9769 + }, + { + "epoch": 3.61, + "learning_rate": 2.6408424682842715e-06, + "loss": 0.6124, + "step": 9770 + }, + { + "epoch": 3.61, + "learning_rate": 2.640805517920926e-06, + "loss": 0.5745, + "step": 9771 + }, + { + "epoch": 3.61, + "learning_rate": 2.6407685675575812e-06, + "loss": 0.5977, + "step": 9772 + }, + { + "epoch": 3.61, + "learning_rate": 2.640731617194236e-06, + "loss": 0.5235, + "step": 9773 + }, + { + "epoch": 3.61, + "learning_rate": 2.6406946668308905e-06, + "loss": 0.5999, + "step": 9774 + }, + { + "epoch": 3.61, + "learning_rate": 2.640657716467545e-06, + "loss": 0.5336, + "step": 9775 + }, + { + "epoch": 3.61, + "learning_rate": 2.6406207661042e-06, + "loss": 0.4725, + "step": 9776 + }, + { + "epoch": 3.61, + "learning_rate": 2.640583815740855e-06, + "loss": 0.6232, + "step": 9777 + }, + { + "epoch": 3.61, + "learning_rate": 2.64054686537751e-06, + "loss": 0.6039, + "step": 9778 + }, + { + "epoch": 3.61, + "learning_rate": 2.6405099150141645e-06, + "loss": 0.5366, + "step": 9779 + }, + { + "epoch": 3.61, + "learning_rate": 2.640472964650819e-06, + "loss": 0.5453, + "step": 9780 + }, + { + "epoch": 3.61, + "learning_rate": 2.640436014287474e-06, + "loss": 0.5902, + "step": 9781 + }, + { + "epoch": 3.61, + "learning_rate": 2.6403990639241285e-06, + "loss": 0.5243, + "step": 9782 + }, + { + "epoch": 3.61, + "learning_rate": 2.6403621135607835e-06, + "loss": 0.6409, + "step": 9783 + }, + { + "epoch": 3.61, + "learning_rate": 2.640325163197438e-06, + "loss": 0.5948, + "step": 9784 + }, + { + "epoch": 3.61, + "learning_rate": 2.640288212834093e-06, + "loss": 0.5538, + "step": 9785 + }, + { + "epoch": 3.61, + "learning_rate": 2.640251262470748e-06, + "loss": 0.5918, + "step": 9786 + }, + { + "epoch": 3.61, + "learning_rate": 2.6402143121074025e-06, + "loss": 0.6025, + "step": 9787 + }, + { + "epoch": 3.61, + "learning_rate": 2.640177361744057e-06, + "loss": 0.4649, + "step": 9788 + }, + { + "epoch": 3.61, + "learning_rate": 2.6401404113807118e-06, + "loss": 0.5215, + "step": 9789 + }, + { + "epoch": 3.62, + "learning_rate": 2.640103461017367e-06, + "loss": 0.496, + "step": 9790 + }, + { + "epoch": 3.62, + "learning_rate": 2.6400665106540215e-06, + "loss": 0.5131, + "step": 9791 + }, + { + "epoch": 3.62, + "learning_rate": 2.6400295602906765e-06, + "loss": 0.498, + "step": 9792 + }, + { + "epoch": 3.62, + "learning_rate": 2.639992609927331e-06, + "loss": 0.5492, + "step": 9793 + }, + { + "epoch": 3.62, + "learning_rate": 2.6399556595639858e-06, + "loss": 0.5535, + "step": 9794 + }, + { + "epoch": 3.62, + "learning_rate": 2.6399187092006404e-06, + "loss": 0.4009, + "step": 9795 + }, + { + "epoch": 3.62, + "learning_rate": 2.6398817588372955e-06, + "loss": 0.6022, + "step": 9796 + }, + { + "epoch": 3.62, + "learning_rate": 2.63984480847395e-06, + "loss": 0.4657, + "step": 9797 + }, + { + "epoch": 3.62, + "learning_rate": 2.6398078581106048e-06, + "loss": 0.5598, + "step": 9798 + }, + { + "epoch": 3.62, + "learning_rate": 2.6397709077472594e-06, + "loss": 0.5423, + "step": 9799 + }, + { + "epoch": 3.62, + "learning_rate": 2.639733957383914e-06, + "loss": 0.4513, + "step": 9800 + }, + { + "epoch": 3.62, + "learning_rate": 2.639697007020569e-06, + "loss": 0.5513, + "step": 9801 + }, + { + "epoch": 3.62, + "learning_rate": 2.6396600566572237e-06, + "loss": 0.6347, + "step": 9802 + }, + { + "epoch": 3.62, + "learning_rate": 2.6396231062938788e-06, + "loss": 0.5346, + "step": 9803 + }, + { + "epoch": 3.62, + "learning_rate": 2.6395861559305334e-06, + "loss": 0.5834, + "step": 9804 + }, + { + "epoch": 3.62, + "learning_rate": 2.639549205567188e-06, + "loss": 0.5306, + "step": 9805 + }, + { + "epoch": 3.62, + "learning_rate": 2.6395122552038427e-06, + "loss": 0.5202, + "step": 9806 + }, + { + "epoch": 3.62, + "learning_rate": 2.6394753048404978e-06, + "loss": 0.5023, + "step": 9807 + }, + { + "epoch": 3.62, + "learning_rate": 2.6394383544771524e-06, + "loss": 0.6514, + "step": 9808 + }, + { + "epoch": 3.62, + "learning_rate": 2.639401404113807e-06, + "loss": 0.4751, + "step": 9809 + }, + { + "epoch": 3.62, + "learning_rate": 2.639364453750462e-06, + "loss": 0.5913, + "step": 9810 + }, + { + "epoch": 3.62, + "learning_rate": 2.6393275033871167e-06, + "loss": 0.5507, + "step": 9811 + }, + { + "epoch": 3.62, + "learning_rate": 2.6392905530237714e-06, + "loss": 0.5521, + "step": 9812 + }, + { + "epoch": 3.62, + "learning_rate": 2.639253602660426e-06, + "loss": 0.6124, + "step": 9813 + }, + { + "epoch": 3.62, + "learning_rate": 2.639216652297081e-06, + "loss": 0.4706, + "step": 9814 + }, + { + "epoch": 3.62, + "learning_rate": 2.6391797019337357e-06, + "loss": 0.5797, + "step": 9815 + }, + { + "epoch": 3.62, + "learning_rate": 2.6391427515703907e-06, + "loss": 0.5032, + "step": 9816 + }, + { + "epoch": 3.63, + "learning_rate": 2.6391058012070454e-06, + "loss": 0.5214, + "step": 9817 + }, + { + "epoch": 3.63, + "learning_rate": 2.6390688508437e-06, + "loss": 0.4656, + "step": 9818 + }, + { + "epoch": 3.63, + "learning_rate": 2.6390319004803547e-06, + "loss": 0.5936, + "step": 9819 + }, + { + "epoch": 3.63, + "learning_rate": 2.6389949501170093e-06, + "loss": 0.5948, + "step": 9820 + }, + { + "epoch": 3.63, + "learning_rate": 2.6389579997536644e-06, + "loss": 0.5806, + "step": 9821 + }, + { + "epoch": 3.63, + "learning_rate": 2.638921049390319e-06, + "loss": 0.5355, + "step": 9822 + }, + { + "epoch": 3.63, + "learning_rate": 2.638884099026974e-06, + "loss": 0.6845, + "step": 9823 + }, + { + "epoch": 3.63, + "learning_rate": 2.6388471486636287e-06, + "loss": 0.6157, + "step": 9824 + }, + { + "epoch": 3.63, + "learning_rate": 2.6388101983002833e-06, + "loss": 0.6715, + "step": 9825 + }, + { + "epoch": 3.63, + "learning_rate": 2.638773247936938e-06, + "loss": 0.5325, + "step": 9826 + }, + { + "epoch": 3.63, + "learning_rate": 2.638736297573593e-06, + "loss": 0.5678, + "step": 9827 + }, + { + "epoch": 3.63, + "learning_rate": 2.6386993472102477e-06, + "loss": 0.5487, + "step": 9828 + }, + { + "epoch": 3.63, + "learning_rate": 2.6386623968469023e-06, + "loss": 0.4972, + "step": 9829 + }, + { + "epoch": 3.63, + "learning_rate": 2.6386254464835574e-06, + "loss": 0.576, + "step": 9830 + }, + { + "epoch": 3.63, + "learning_rate": 2.638588496120212e-06, + "loss": 0.658, + "step": 9831 + }, + { + "epoch": 3.63, + "learning_rate": 2.6385515457568666e-06, + "loss": 0.5128, + "step": 9832 + }, + { + "epoch": 3.63, + "learning_rate": 2.6385145953935213e-06, + "loss": 0.6809, + "step": 9833 + }, + { + "epoch": 3.63, + "learning_rate": 2.6384776450301763e-06, + "loss": 0.6162, + "step": 9834 + }, + { + "epoch": 3.63, + "learning_rate": 2.638440694666831e-06, + "loss": 0.6099, + "step": 9835 + }, + { + "epoch": 3.63, + "learning_rate": 2.638403744303486e-06, + "loss": 0.5661, + "step": 9836 + }, + { + "epoch": 3.63, + "learning_rate": 2.6383667939401407e-06, + "loss": 0.5001, + "step": 9837 + }, + { + "epoch": 3.63, + "learning_rate": 2.638329843576795e-06, + "loss": 0.6011, + "step": 9838 + }, + { + "epoch": 3.63, + "learning_rate": 2.63829289321345e-06, + "loss": 0.5485, + "step": 9839 + }, + { + "epoch": 3.63, + "learning_rate": 2.6382559428501046e-06, + "loss": 0.5209, + "step": 9840 + }, + { + "epoch": 3.63, + "learning_rate": 2.6382189924867596e-06, + "loss": 0.5822, + "step": 9841 + }, + { + "epoch": 3.63, + "learning_rate": 2.6381820421234143e-06, + "loss": 0.5295, + "step": 9842 + }, + { + "epoch": 3.63, + "learning_rate": 2.6381450917600693e-06, + "loss": 0.6369, + "step": 9843 + }, + { + "epoch": 3.64, + "learning_rate": 2.6381081413967235e-06, + "loss": 0.4964, + "step": 9844 + }, + { + "epoch": 3.64, + "learning_rate": 2.6380711910333786e-06, + "loss": 0.5147, + "step": 9845 + }, + { + "epoch": 3.64, + "learning_rate": 2.6380342406700332e-06, + "loss": 0.4881, + "step": 9846 + }, + { + "epoch": 3.64, + "learning_rate": 2.6379972903066883e-06, + "loss": 0.5153, + "step": 9847 + }, + { + "epoch": 3.64, + "learning_rate": 2.637960339943343e-06, + "loss": 0.5022, + "step": 9848 + }, + { + "epoch": 3.64, + "learning_rate": 2.6379233895799976e-06, + "loss": 0.6148, + "step": 9849 + }, + { + "epoch": 3.64, + "learning_rate": 2.637886439216652e-06, + "loss": 0.4885, + "step": 9850 + }, + { + "epoch": 3.64, + "learning_rate": 2.637849488853307e-06, + "loss": 0.5362, + "step": 9851 + }, + { + "epoch": 3.64, + "learning_rate": 2.637812538489962e-06, + "loss": 0.4963, + "step": 9852 + }, + { + "epoch": 3.64, + "learning_rate": 2.6377755881266165e-06, + "loss": 0.459, + "step": 9853 + }, + { + "epoch": 3.64, + "learning_rate": 2.6377386377632716e-06, + "loss": 0.5162, + "step": 9854 + }, + { + "epoch": 3.64, + "learning_rate": 2.6377016873999262e-06, + "loss": 0.5513, + "step": 9855 + }, + { + "epoch": 3.64, + "learning_rate": 2.637664737036581e-06, + "loss": 0.5947, + "step": 9856 + }, + { + "epoch": 3.64, + "learning_rate": 2.6376277866732355e-06, + "loss": 0.608, + "step": 9857 + }, + { + "epoch": 3.64, + "learning_rate": 2.63759083630989e-06, + "loss": 0.5119, + "step": 9858 + }, + { + "epoch": 3.64, + "learning_rate": 2.637553885946545e-06, + "loss": 0.5372, + "step": 9859 + }, + { + "epoch": 3.64, + "learning_rate": 2.6375169355832e-06, + "loss": 0.5661, + "step": 9860 + }, + { + "epoch": 3.64, + "learning_rate": 2.637479985219855e-06, + "loss": 0.609, + "step": 9861 + }, + { + "epoch": 3.64, + "learning_rate": 2.6374430348565095e-06, + "loss": 0.4323, + "step": 9862 + }, + { + "epoch": 3.64, + "learning_rate": 2.637406084493164e-06, + "loss": 0.5377, + "step": 9863 + }, + { + "epoch": 3.64, + "learning_rate": 2.637369134129819e-06, + "loss": 0.5689, + "step": 9864 + }, + { + "epoch": 3.64, + "learning_rate": 2.637332183766474e-06, + "loss": 0.5886, + "step": 9865 + }, + { + "epoch": 3.64, + "learning_rate": 2.6372952334031285e-06, + "loss": 0.5146, + "step": 9866 + }, + { + "epoch": 3.64, + "learning_rate": 2.6372582830397836e-06, + "loss": 0.518, + "step": 9867 + }, + { + "epoch": 3.64, + "learning_rate": 2.637221332676438e-06, + "loss": 0.5799, + "step": 9868 + }, + { + "epoch": 3.64, + "learning_rate": 2.637184382313093e-06, + "loss": 0.5711, + "step": 9869 + }, + { + "epoch": 3.64, + "learning_rate": 2.6371474319497475e-06, + "loss": 0.4704, + "step": 9870 + }, + { + "epoch": 3.65, + "learning_rate": 2.637110481586402e-06, + "loss": 0.5213, + "step": 9871 + }, + { + "epoch": 3.65, + "learning_rate": 2.637073531223057e-06, + "loss": 0.564, + "step": 9872 + }, + { + "epoch": 3.65, + "learning_rate": 2.637036580859712e-06, + "loss": 0.4864, + "step": 9873 + }, + { + "epoch": 3.65, + "learning_rate": 2.636999630496367e-06, + "loss": 0.524, + "step": 9874 + }, + { + "epoch": 3.65, + "learning_rate": 2.6369626801330215e-06, + "loss": 0.4988, + "step": 9875 + }, + { + "epoch": 3.65, + "learning_rate": 2.636925729769676e-06, + "loss": 0.6121, + "step": 9876 + }, + { + "epoch": 3.65, + "learning_rate": 2.6368887794063308e-06, + "loss": 0.5509, + "step": 9877 + }, + { + "epoch": 3.65, + "learning_rate": 2.6368518290429854e-06, + "loss": 0.6289, + "step": 9878 + }, + { + "epoch": 3.65, + "learning_rate": 2.6368148786796405e-06, + "loss": 0.5345, + "step": 9879 + }, + { + "epoch": 3.65, + "learning_rate": 2.636777928316295e-06, + "loss": 0.5325, + "step": 9880 + }, + { + "epoch": 3.65, + "learning_rate": 2.63674097795295e-06, + "loss": 0.5517, + "step": 9881 + }, + { + "epoch": 3.65, + "learning_rate": 2.636704027589605e-06, + "loss": 0.5642, + "step": 9882 + }, + { + "epoch": 3.65, + "learning_rate": 2.6366670772262594e-06, + "loss": 0.5492, + "step": 9883 + }, + { + "epoch": 3.65, + "learning_rate": 2.636630126862914e-06, + "loss": 0.4802, + "step": 9884 + }, + { + "epoch": 3.65, + "learning_rate": 2.636593176499569e-06, + "loss": 0.5598, + "step": 9885 + }, + { + "epoch": 3.65, + "learning_rate": 2.6365562261362238e-06, + "loss": 0.694, + "step": 9886 + }, + { + "epoch": 3.65, + "learning_rate": 2.636519275772879e-06, + "loss": 0.5622, + "step": 9887 + }, + { + "epoch": 3.65, + "learning_rate": 2.6364823254095335e-06, + "loss": 0.5317, + "step": 9888 + }, + { + "epoch": 3.65, + "learning_rate": 2.6364453750461877e-06, + "loss": 0.6103, + "step": 9889 + }, + { + "epoch": 3.65, + "learning_rate": 2.6364084246828427e-06, + "loss": 0.5976, + "step": 9890 + }, + { + "epoch": 3.65, + "learning_rate": 2.6363714743194974e-06, + "loss": 0.5021, + "step": 9891 + }, + { + "epoch": 3.65, + "learning_rate": 2.6363345239561524e-06, + "loss": 0.4976, + "step": 9892 + }, + { + "epoch": 3.65, + "learning_rate": 2.636297573592807e-06, + "loss": 0.465, + "step": 9893 + }, + { + "epoch": 3.65, + "learning_rate": 2.636260623229462e-06, + "loss": 0.5925, + "step": 9894 + }, + { + "epoch": 3.65, + "learning_rate": 2.6362236728661163e-06, + "loss": 0.5552, + "step": 9895 + }, + { + "epoch": 3.65, + "learning_rate": 2.6361867225027714e-06, + "loss": 0.5817, + "step": 9896 + }, + { + "epoch": 3.65, + "learning_rate": 2.636149772139426e-06, + "loss": 0.6276, + "step": 9897 + }, + { + "epoch": 3.66, + "learning_rate": 2.6361128217760807e-06, + "loss": 0.5422, + "step": 9898 + }, + { + "epoch": 3.66, + "learning_rate": 2.6360758714127357e-06, + "loss": 0.5634, + "step": 9899 + }, + { + "epoch": 3.66, + "learning_rate": 2.6360389210493904e-06, + "loss": 0.4725, + "step": 9900 + }, + { + "epoch": 3.66, + "learning_rate": 2.636001970686045e-06, + "loss": 0.5489, + "step": 9901 + }, + { + "epoch": 3.66, + "learning_rate": 2.6359650203226996e-06, + "loss": 0.5924, + "step": 9902 + }, + { + "epoch": 3.66, + "learning_rate": 2.6359280699593547e-06, + "loss": 0.4735, + "step": 9903 + }, + { + "epoch": 3.66, + "learning_rate": 2.6358911195960093e-06, + "loss": 0.5125, + "step": 9904 + }, + { + "epoch": 3.66, + "learning_rate": 2.6358541692326644e-06, + "loss": 0.564, + "step": 9905 + }, + { + "epoch": 3.66, + "learning_rate": 2.635817218869319e-06, + "loss": 0.6359, + "step": 9906 + }, + { + "epoch": 3.66, + "learning_rate": 2.6357802685059737e-06, + "loss": 0.6494, + "step": 9907 + }, + { + "epoch": 3.66, + "learning_rate": 2.6357433181426283e-06, + "loss": 0.5046, + "step": 9908 + }, + { + "epoch": 3.66, + "learning_rate": 2.635706367779283e-06, + "loss": 0.4187, + "step": 9909 + }, + { + "epoch": 3.66, + "learning_rate": 2.635669417415938e-06, + "loss": 0.5085, + "step": 9910 + }, + { + "epoch": 3.66, + "learning_rate": 2.6356324670525926e-06, + "loss": 0.6309, + "step": 9911 + }, + { + "epoch": 3.66, + "learning_rate": 2.6355955166892477e-06, + "loss": 0.594, + "step": 9912 + }, + { + "epoch": 3.66, + "learning_rate": 2.6355585663259023e-06, + "loss": 0.482, + "step": 9913 + }, + { + "epoch": 3.66, + "learning_rate": 2.635521615962557e-06, + "loss": 0.5438, + "step": 9914 + }, + { + "epoch": 3.66, + "learning_rate": 2.6354846655992116e-06, + "loss": 0.5154, + "step": 9915 + }, + { + "epoch": 3.66, + "learning_rate": 2.6354477152358667e-06, + "loss": 0.5246, + "step": 9916 + }, + { + "epoch": 3.66, + "learning_rate": 2.6354107648725213e-06, + "loss": 0.5485, + "step": 9917 + }, + { + "epoch": 3.66, + "learning_rate": 2.635373814509176e-06, + "loss": 0.5594, + "step": 9918 + }, + { + "epoch": 3.66, + "learning_rate": 2.635336864145831e-06, + "loss": 0.5438, + "step": 9919 + }, + { + "epoch": 3.66, + "learning_rate": 2.6352999137824856e-06, + "loss": 0.4503, + "step": 9920 + }, + { + "epoch": 3.66, + "learning_rate": 2.6352629634191403e-06, + "loss": 0.6228, + "step": 9921 + }, + { + "epoch": 3.66, + "learning_rate": 2.635226013055795e-06, + "loss": 0.5832, + "step": 9922 + }, + { + "epoch": 3.66, + "learning_rate": 2.63518906269245e-06, + "loss": 0.4568, + "step": 9923 + }, + { + "epoch": 3.66, + "learning_rate": 2.6351521123291046e-06, + "loss": 0.5049, + "step": 9924 + }, + { + "epoch": 3.67, + "learning_rate": 2.6351151619657597e-06, + "loss": 0.5848, + "step": 9925 + }, + { + "epoch": 3.67, + "learning_rate": 2.6350782116024143e-06, + "loss": 0.5596, + "step": 9926 + }, + { + "epoch": 3.67, + "learning_rate": 2.635041261239069e-06, + "loss": 0.5445, + "step": 9927 + }, + { + "epoch": 3.67, + "learning_rate": 2.6350043108757236e-06, + "loss": 0.6206, + "step": 9928 + }, + { + "epoch": 3.67, + "learning_rate": 2.6349673605123782e-06, + "loss": 0.4063, + "step": 9929 + }, + { + "epoch": 3.67, + "learning_rate": 2.6349304101490333e-06, + "loss": 0.5469, + "step": 9930 + }, + { + "epoch": 3.67, + "learning_rate": 2.634893459785688e-06, + "loss": 0.5506, + "step": 9931 + }, + { + "epoch": 3.67, + "learning_rate": 2.634856509422343e-06, + "loss": 0.5475, + "step": 9932 + }, + { + "epoch": 3.67, + "learning_rate": 2.6348195590589976e-06, + "loss": 0.5545, + "step": 9933 + }, + { + "epoch": 3.67, + "learning_rate": 2.6347826086956522e-06, + "loss": 0.6342, + "step": 9934 + }, + { + "epoch": 3.67, + "learning_rate": 2.634745658332307e-06, + "loss": 0.496, + "step": 9935 + }, + { + "epoch": 3.67, + "learning_rate": 2.634708707968962e-06, + "loss": 0.5337, + "step": 9936 + }, + { + "epoch": 3.67, + "learning_rate": 2.6346717576056166e-06, + "loss": 0.615, + "step": 9937 + }, + { + "epoch": 3.67, + "learning_rate": 2.6346348072422712e-06, + "loss": 0.6669, + "step": 9938 + }, + { + "epoch": 3.67, + "learning_rate": 2.6345978568789263e-06, + "loss": 0.5181, + "step": 9939 + }, + { + "epoch": 3.67, + "learning_rate": 2.6345609065155805e-06, + "loss": 0.6144, + "step": 9940 + }, + { + "epoch": 3.67, + "learning_rate": 2.6345239561522356e-06, + "loss": 0.5692, + "step": 9941 + }, + { + "epoch": 3.67, + "learning_rate": 2.63448700578889e-06, + "loss": 0.6035, + "step": 9942 + }, + { + "epoch": 3.67, + "learning_rate": 2.6344500554255452e-06, + "loss": 0.5057, + "step": 9943 + }, + { + "epoch": 3.67, + "learning_rate": 2.6344131050622e-06, + "loss": 0.6124, + "step": 9944 + }, + { + "epoch": 3.67, + "learning_rate": 2.634376154698855e-06, + "loss": 0.6262, + "step": 9945 + }, + { + "epoch": 3.67, + "learning_rate": 2.634339204335509e-06, + "loss": 0.585, + "step": 9946 + }, + { + "epoch": 3.67, + "learning_rate": 2.6343022539721642e-06, + "loss": 0.4805, + "step": 9947 + }, + { + "epoch": 3.67, + "learning_rate": 2.634265303608819e-06, + "loss": 0.4866, + "step": 9948 + }, + { + "epoch": 3.67, + "learning_rate": 2.6342283532454735e-06, + "loss": 0.503, + "step": 9949 + }, + { + "epoch": 3.67, + "learning_rate": 2.6341914028821285e-06, + "loss": 0.5728, + "step": 9950 + }, + { + "epoch": 3.67, + "learning_rate": 2.634154452518783e-06, + "loss": 0.5013, + "step": 9951 + }, + { + "epoch": 3.68, + "learning_rate": 2.634117502155438e-06, + "loss": 0.5111, + "step": 9952 + }, + { + "epoch": 3.68, + "learning_rate": 2.6340805517920925e-06, + "loss": 0.5465, + "step": 9953 + }, + { + "epoch": 3.68, + "learning_rate": 2.6340436014287475e-06, + "loss": 0.602, + "step": 9954 + }, + { + "epoch": 3.68, + "learning_rate": 2.634006651065402e-06, + "loss": 0.5424, + "step": 9955 + }, + { + "epoch": 3.68, + "learning_rate": 2.6339697007020572e-06, + "loss": 0.5508, + "step": 9956 + }, + { + "epoch": 3.68, + "learning_rate": 2.633932750338712e-06, + "loss": 0.5452, + "step": 9957 + }, + { + "epoch": 3.68, + "learning_rate": 2.6338957999753665e-06, + "loss": 0.5234, + "step": 9958 + }, + { + "epoch": 3.68, + "learning_rate": 2.633858849612021e-06, + "loss": 0.4441, + "step": 9959 + }, + { + "epoch": 3.68, + "learning_rate": 2.6338218992486758e-06, + "loss": 0.5057, + "step": 9960 + }, + { + "epoch": 3.68, + "learning_rate": 2.633784948885331e-06, + "loss": 0.5878, + "step": 9961 + }, + { + "epoch": 3.68, + "learning_rate": 2.6337479985219855e-06, + "loss": 0.5293, + "step": 9962 + }, + { + "epoch": 3.68, + "learning_rate": 2.6337110481586405e-06, + "loss": 0.6574, + "step": 9963 + }, + { + "epoch": 3.68, + "learning_rate": 2.633674097795295e-06, + "loss": 0.6046, + "step": 9964 + }, + { + "epoch": 3.68, + "learning_rate": 2.63363714743195e-06, + "loss": 0.422, + "step": 9965 + }, + { + "epoch": 3.68, + "learning_rate": 2.6336001970686044e-06, + "loss": 0.6031, + "step": 9966 + }, + { + "epoch": 3.68, + "learning_rate": 2.6335632467052595e-06, + "loss": 0.5384, + "step": 9967 + }, + { + "epoch": 3.68, + "learning_rate": 2.633526296341914e-06, + "loss": 0.5852, + "step": 9968 + }, + { + "epoch": 3.68, + "learning_rate": 2.6334893459785688e-06, + "loss": 0.4489, + "step": 9969 + }, + { + "epoch": 3.68, + "learning_rate": 2.633452395615224e-06, + "loss": 0.54, + "step": 9970 + }, + { + "epoch": 3.68, + "learning_rate": 2.6334154452518785e-06, + "loss": 0.5175, + "step": 9971 + }, + { + "epoch": 3.68, + "learning_rate": 2.633378494888533e-06, + "loss": 0.5887, + "step": 9972 + }, + { + "epoch": 3.68, + "learning_rate": 2.6333415445251877e-06, + "loss": 0.5026, + "step": 9973 + }, + { + "epoch": 3.68, + "learning_rate": 2.6333045941618428e-06, + "loss": 0.5194, + "step": 9974 + }, + { + "epoch": 3.68, + "learning_rate": 2.6332676437984974e-06, + "loss": 0.6131, + "step": 9975 + }, + { + "epoch": 3.68, + "learning_rate": 2.6332306934351525e-06, + "loss": 0.5677, + "step": 9976 + }, + { + "epoch": 3.68, + "learning_rate": 2.633193743071807e-06, + "loss": 0.517, + "step": 9977 + }, + { + "epoch": 3.68, + "learning_rate": 2.6331567927084618e-06, + "loss": 0.472, + "step": 9978 + }, + { + "epoch": 3.69, + "learning_rate": 2.6331198423451164e-06, + "loss": 0.5354, + "step": 9979 + }, + { + "epoch": 3.69, + "learning_rate": 2.633082891981771e-06, + "loss": 0.5387, + "step": 9980 + }, + { + "epoch": 3.69, + "learning_rate": 2.633045941618426e-06, + "loss": 0.6941, + "step": 9981 + }, + { + "epoch": 3.69, + "learning_rate": 2.6330089912550807e-06, + "loss": 0.6143, + "step": 9982 + }, + { + "epoch": 3.69, + "learning_rate": 2.6329720408917358e-06, + "loss": 0.4835, + "step": 9983 + }, + { + "epoch": 3.69, + "learning_rate": 2.63293509052839e-06, + "loss": 0.5735, + "step": 9984 + }, + { + "epoch": 3.69, + "learning_rate": 2.632898140165045e-06, + "loss": 0.5739, + "step": 9985 + }, + { + "epoch": 3.69, + "learning_rate": 2.6328611898016997e-06, + "loss": 0.5964, + "step": 9986 + }, + { + "epoch": 3.69, + "learning_rate": 2.6328242394383548e-06, + "loss": 0.5651, + "step": 9987 + }, + { + "epoch": 3.69, + "learning_rate": 2.6327872890750094e-06, + "loss": 0.5504, + "step": 9988 + }, + { + "epoch": 3.69, + "learning_rate": 2.632750338711664e-06, + "loss": 0.487, + "step": 9989 + }, + { + "epoch": 3.69, + "learning_rate": 2.6327133883483187e-06, + "loss": 0.5479, + "step": 9990 + }, + { + "epoch": 3.69, + "learning_rate": 2.6326764379849733e-06, + "loss": 0.5934, + "step": 9991 + }, + { + "epoch": 3.69, + "learning_rate": 2.6326394876216284e-06, + "loss": 0.5001, + "step": 9992 + }, + { + "epoch": 3.69, + "learning_rate": 2.632602537258283e-06, + "loss": 0.5242, + "step": 9993 + }, + { + "epoch": 3.69, + "learning_rate": 2.632565586894938e-06, + "loss": 0.4674, + "step": 9994 + }, + { + "epoch": 3.69, + "learning_rate": 2.6325286365315927e-06, + "loss": 0.5093, + "step": 9995 + }, + { + "epoch": 3.69, + "learning_rate": 2.6324916861682473e-06, + "loss": 0.446, + "step": 9996 + }, + { + "epoch": 3.69, + "learning_rate": 2.632454735804902e-06, + "loss": 0.5, + "step": 9997 + }, + { + "epoch": 3.69, + "learning_rate": 2.6324177854415566e-06, + "loss": 0.5648, + "step": 9998 + }, + { + "epoch": 3.69, + "learning_rate": 2.6323808350782117e-06, + "loss": 0.5784, + "step": 9999 + }, + { + "epoch": 3.69, + "learning_rate": 2.6323438847148663e-06, + "loss": 0.4556, + "step": 10000 + }, + { + "epoch": 3.69, + "learning_rate": 2.6323069343515214e-06, + "loss": 0.5434, + "step": 10001 + }, + { + "epoch": 3.69, + "learning_rate": 2.632269983988176e-06, + "loss": 0.5551, + "step": 10002 + }, + { + "epoch": 3.69, + "learning_rate": 2.6322330336248306e-06, + "loss": 0.5405, + "step": 10003 + }, + { + "epoch": 3.69, + "learning_rate": 2.6321960832614853e-06, + "loss": 0.5097, + "step": 10004 + }, + { + "epoch": 3.69, + "learning_rate": 2.6321591328981403e-06, + "loss": 0.5604, + "step": 10005 + }, + { + "epoch": 3.69, + "learning_rate": 2.632122182534795e-06, + "loss": 0.6801, + "step": 10006 + }, + { + "epoch": 3.7, + "learning_rate": 2.63208523217145e-06, + "loss": 0.5728, + "step": 10007 + }, + { + "epoch": 3.7, + "learning_rate": 2.6320482818081047e-06, + "loss": 0.6318, + "step": 10008 + }, + { + "epoch": 3.7, + "learning_rate": 2.6320113314447593e-06, + "loss": 0.4447, + "step": 10009 + }, + { + "epoch": 3.7, + "learning_rate": 2.631974381081414e-06, + "loss": 0.6077, + "step": 10010 + }, + { + "epoch": 3.7, + "learning_rate": 2.6319374307180686e-06, + "loss": 0.5136, + "step": 10011 + }, + { + "epoch": 3.7, + "learning_rate": 2.6319004803547236e-06, + "loss": 0.57, + "step": 10012 + }, + { + "epoch": 3.7, + "learning_rate": 2.6318635299913783e-06, + "loss": 0.6071, + "step": 10013 + }, + { + "epoch": 3.7, + "learning_rate": 2.6318265796280333e-06, + "loss": 0.5182, + "step": 10014 + }, + { + "epoch": 3.7, + "learning_rate": 2.631789629264688e-06, + "loss": 0.5385, + "step": 10015 + }, + { + "epoch": 3.7, + "learning_rate": 2.6317526789013426e-06, + "loss": 0.4487, + "step": 10016 + }, + { + "epoch": 3.7, + "learning_rate": 2.6317157285379972e-06, + "loss": 0.5609, + "step": 10017 + }, + { + "epoch": 3.7, + "learning_rate": 2.631678778174652e-06, + "loss": 0.6935, + "step": 10018 + }, + { + "epoch": 3.7, + "learning_rate": 2.631641827811307e-06, + "loss": 0.4726, + "step": 10019 + }, + { + "epoch": 3.7, + "learning_rate": 2.6316048774479616e-06, + "loss": 0.5031, + "step": 10020 + }, + { + "epoch": 3.7, + "learning_rate": 2.6315679270846166e-06, + "loss": 0.5396, + "step": 10021 + }, + { + "epoch": 3.7, + "learning_rate": 2.6315309767212713e-06, + "loss": 0.5753, + "step": 10022 + }, + { + "epoch": 3.7, + "learning_rate": 2.631494026357926e-06, + "loss": 0.5373, + "step": 10023 + }, + { + "epoch": 3.7, + "learning_rate": 2.6314570759945805e-06, + "loss": 0.5983, + "step": 10024 + }, + { + "epoch": 3.7, + "learning_rate": 2.6314201256312356e-06, + "loss": 0.4817, + "step": 10025 + }, + { + "epoch": 3.7, + "learning_rate": 2.6313831752678902e-06, + "loss": 0.5363, + "step": 10026 + }, + { + "epoch": 3.7, + "learning_rate": 2.631346224904545e-06, + "loss": 0.639, + "step": 10027 + }, + { + "epoch": 3.7, + "learning_rate": 2.6313092745412e-06, + "loss": 0.5649, + "step": 10028 + }, + { + "epoch": 3.7, + "learning_rate": 2.631272324177854e-06, + "loss": 0.6968, + "step": 10029 + }, + { + "epoch": 3.7, + "learning_rate": 2.631235373814509e-06, + "loss": 0.5988, + "step": 10030 + }, + { + "epoch": 3.7, + "learning_rate": 2.631198423451164e-06, + "loss": 0.576, + "step": 10031 + }, + { + "epoch": 3.7, + "learning_rate": 2.631161473087819e-06, + "loss": 0.5224, + "step": 10032 + }, + { + "epoch": 3.7, + "learning_rate": 2.6311245227244735e-06, + "loss": 0.4726, + "step": 10033 + }, + { + "epoch": 3.71, + "learning_rate": 2.6310875723611286e-06, + "loss": 0.5518, + "step": 10034 + }, + { + "epoch": 3.71, + "learning_rate": 2.631050621997783e-06, + "loss": 0.4312, + "step": 10035 + }, + { + "epoch": 3.71, + "learning_rate": 2.631013671634438e-06, + "loss": 0.5769, + "step": 10036 + }, + { + "epoch": 3.71, + "learning_rate": 2.6309767212710925e-06, + "loss": 0.5274, + "step": 10037 + }, + { + "epoch": 3.71, + "learning_rate": 2.630939770907747e-06, + "loss": 0.5222, + "step": 10038 + }, + { + "epoch": 3.71, + "learning_rate": 2.630902820544402e-06, + "loss": 0.5287, + "step": 10039 + }, + { + "epoch": 3.71, + "learning_rate": 2.630865870181057e-06, + "loss": 0.4684, + "step": 10040 + }, + { + "epoch": 3.71, + "learning_rate": 2.6308289198177115e-06, + "loss": 0.5623, + "step": 10041 + }, + { + "epoch": 3.71, + "learning_rate": 2.630791969454366e-06, + "loss": 0.5219, + "step": 10042 + }, + { + "epoch": 3.71, + "learning_rate": 2.630755019091021e-06, + "loss": 0.6032, + "step": 10043 + }, + { + "epoch": 3.71, + "learning_rate": 2.630718068727676e-06, + "loss": 0.5216, + "step": 10044 + }, + { + "epoch": 3.71, + "learning_rate": 2.630681118364331e-06, + "loss": 0.5715, + "step": 10045 + }, + { + "epoch": 3.71, + "learning_rate": 2.6306441680009855e-06, + "loss": 0.56, + "step": 10046 + }, + { + "epoch": 3.71, + "learning_rate": 2.63060721763764e-06, + "loss": 0.4697, + "step": 10047 + }, + { + "epoch": 3.71, + "learning_rate": 2.6305702672742948e-06, + "loss": 0.6375, + "step": 10048 + }, + { + "epoch": 3.71, + "learning_rate": 2.6305333169109494e-06, + "loss": 0.5319, + "step": 10049 + }, + { + "epoch": 3.71, + "learning_rate": 2.6304963665476045e-06, + "loss": 0.5197, + "step": 10050 + }, + { + "epoch": 3.71, + "learning_rate": 2.630459416184259e-06, + "loss": 0.4964, + "step": 10051 + }, + { + "epoch": 3.71, + "learning_rate": 2.630422465820914e-06, + "loss": 0.6668, + "step": 10052 + }, + { + "epoch": 3.71, + "learning_rate": 2.630385515457569e-06, + "loss": 0.5607, + "step": 10053 + }, + { + "epoch": 3.71, + "learning_rate": 2.6303485650942234e-06, + "loss": 0.5458, + "step": 10054 + }, + { + "epoch": 3.71, + "learning_rate": 2.630311614730878e-06, + "loss": 0.4782, + "step": 10055 + }, + { + "epoch": 3.71, + "learning_rate": 2.630274664367533e-06, + "loss": 0.5323, + "step": 10056 + }, + { + "epoch": 3.71, + "learning_rate": 2.6302377140041878e-06, + "loss": 0.4936, + "step": 10057 + }, + { + "epoch": 3.71, + "learning_rate": 2.6302007636408424e-06, + "loss": 0.6628, + "step": 10058 + }, + { + "epoch": 3.71, + "learning_rate": 2.6301638132774975e-06, + "loss": 0.5337, + "step": 10059 + }, + { + "epoch": 3.71, + "learning_rate": 2.630126862914152e-06, + "loss": 0.5948, + "step": 10060 + }, + { + "epoch": 3.72, + "learning_rate": 2.6300899125508067e-06, + "loss": 0.5745, + "step": 10061 + }, + { + "epoch": 3.72, + "learning_rate": 2.6300529621874614e-06, + "loss": 0.5277, + "step": 10062 + }, + { + "epoch": 3.72, + "learning_rate": 2.6300160118241164e-06, + "loss": 0.4829, + "step": 10063 + }, + { + "epoch": 3.72, + "learning_rate": 2.629979061460771e-06, + "loss": 0.7144, + "step": 10064 + }, + { + "epoch": 3.72, + "learning_rate": 2.629942111097426e-06, + "loss": 0.5419, + "step": 10065 + }, + { + "epoch": 3.72, + "learning_rate": 2.6299051607340808e-06, + "loss": 0.5124, + "step": 10066 + }, + { + "epoch": 3.72, + "learning_rate": 2.6298682103707354e-06, + "loss": 0.5758, + "step": 10067 + }, + { + "epoch": 3.72, + "learning_rate": 2.62983126000739e-06, + "loss": 0.5177, + "step": 10068 + }, + { + "epoch": 3.72, + "learning_rate": 2.6297943096440447e-06, + "loss": 0.4419, + "step": 10069 + }, + { + "epoch": 3.72, + "learning_rate": 2.6297573592806997e-06, + "loss": 0.5029, + "step": 10070 + }, + { + "epoch": 3.72, + "learning_rate": 2.6297204089173544e-06, + "loss": 0.5726, + "step": 10071 + }, + { + "epoch": 3.72, + "learning_rate": 2.6296834585540094e-06, + "loss": 0.6629, + "step": 10072 + }, + { + "epoch": 3.72, + "learning_rate": 2.629646508190664e-06, + "loss": 0.5246, + "step": 10073 + }, + { + "epoch": 3.72, + "learning_rate": 2.6296095578273187e-06, + "loss": 0.5948, + "step": 10074 + }, + { + "epoch": 3.72, + "learning_rate": 2.6295726074639733e-06, + "loss": 0.5282, + "step": 10075 + }, + { + "epoch": 3.72, + "learning_rate": 2.6295356571006284e-06, + "loss": 0.5658, + "step": 10076 + }, + { + "epoch": 3.72, + "learning_rate": 2.629498706737283e-06, + "loss": 0.5073, + "step": 10077 + }, + { + "epoch": 3.72, + "learning_rate": 2.6294617563739377e-06, + "loss": 0.5116, + "step": 10078 + }, + { + "epoch": 3.72, + "learning_rate": 2.6294248060105927e-06, + "loss": 0.6135, + "step": 10079 + }, + { + "epoch": 3.72, + "learning_rate": 2.629387855647247e-06, + "loss": 0.5806, + "step": 10080 + }, + { + "epoch": 3.72, + "learning_rate": 2.629350905283902e-06, + "loss": 0.5392, + "step": 10081 + }, + { + "epoch": 3.72, + "learning_rate": 2.6293139549205567e-06, + "loss": 0.5576, + "step": 10082 + }, + { + "epoch": 3.72, + "learning_rate": 2.6292770045572117e-06, + "loss": 0.5434, + "step": 10083 + }, + { + "epoch": 3.72, + "learning_rate": 2.6292400541938663e-06, + "loss": 0.5032, + "step": 10084 + }, + { + "epoch": 3.72, + "learning_rate": 2.6292031038305214e-06, + "loss": 0.4791, + "step": 10085 + }, + { + "epoch": 3.72, + "learning_rate": 2.6291661534671756e-06, + "loss": 0.5405, + "step": 10086 + }, + { + "epoch": 3.72, + "learning_rate": 2.6291292031038303e-06, + "loss": 0.578, + "step": 10087 + }, + { + "epoch": 3.73, + "learning_rate": 2.6290922527404853e-06, + "loss": 0.577, + "step": 10088 + }, + { + "epoch": 3.73, + "learning_rate": 2.62905530237714e-06, + "loss": 0.4666, + "step": 10089 + }, + { + "epoch": 3.73, + "learning_rate": 2.629018352013795e-06, + "loss": 0.5271, + "step": 10090 + }, + { + "epoch": 3.73, + "learning_rate": 2.6289814016504496e-06, + "loss": 0.6295, + "step": 10091 + }, + { + "epoch": 3.73, + "learning_rate": 2.6289444512871043e-06, + "loss": 0.5712, + "step": 10092 + }, + { + "epoch": 3.73, + "learning_rate": 2.628907500923759e-06, + "loss": 0.5026, + "step": 10093 + }, + { + "epoch": 3.73, + "learning_rate": 2.628870550560414e-06, + "loss": 0.5634, + "step": 10094 + }, + { + "epoch": 3.73, + "learning_rate": 2.6288336001970686e-06, + "loss": 0.5652, + "step": 10095 + }, + { + "epoch": 3.73, + "learning_rate": 2.6287966498337237e-06, + "loss": 0.5545, + "step": 10096 + }, + { + "epoch": 3.73, + "learning_rate": 2.6287596994703783e-06, + "loss": 0.4662, + "step": 10097 + }, + { + "epoch": 3.73, + "learning_rate": 2.628722749107033e-06, + "loss": 0.4752, + "step": 10098 + }, + { + "epoch": 3.73, + "learning_rate": 2.6286857987436876e-06, + "loss": 0.5451, + "step": 10099 + }, + { + "epoch": 3.73, + "learning_rate": 2.6286488483803422e-06, + "loss": 0.562, + "step": 10100 + }, + { + "epoch": 3.73, + "learning_rate": 2.6286118980169973e-06, + "loss": 0.478, + "step": 10101 + }, + { + "epoch": 3.73, + "learning_rate": 2.628574947653652e-06, + "loss": 0.4714, + "step": 10102 + }, + { + "epoch": 3.73, + "learning_rate": 2.628537997290307e-06, + "loss": 0.5392, + "step": 10103 + }, + { + "epoch": 3.73, + "learning_rate": 2.6285010469269616e-06, + "loss": 0.5735, + "step": 10104 + }, + { + "epoch": 3.73, + "learning_rate": 2.6284640965636163e-06, + "loss": 0.5184, + "step": 10105 + }, + { + "epoch": 3.73, + "learning_rate": 2.628427146200271e-06, + "loss": 0.4789, + "step": 10106 + }, + { + "epoch": 3.73, + "learning_rate": 2.6283901958369255e-06, + "loss": 0.5767, + "step": 10107 + }, + { + "epoch": 3.73, + "learning_rate": 2.6283532454735806e-06, + "loss": 0.697, + "step": 10108 + }, + { + "epoch": 3.73, + "learning_rate": 2.6283162951102352e-06, + "loss": 0.4583, + "step": 10109 + }, + { + "epoch": 3.73, + "learning_rate": 2.6282793447468903e-06, + "loss": 0.4988, + "step": 10110 + }, + { + "epoch": 3.73, + "learning_rate": 2.628242394383545e-06, + "loss": 0.4938, + "step": 10111 + }, + { + "epoch": 3.73, + "learning_rate": 2.6282054440201996e-06, + "loss": 0.5093, + "step": 10112 + }, + { + "epoch": 3.73, + "learning_rate": 2.628168493656854e-06, + "loss": 0.6427, + "step": 10113 + }, + { + "epoch": 3.73, + "learning_rate": 2.6281315432935093e-06, + "loss": 0.5215, + "step": 10114 + }, + { + "epoch": 3.74, + "learning_rate": 2.628094592930164e-06, + "loss": 0.4328, + "step": 10115 + }, + { + "epoch": 3.74, + "learning_rate": 2.628057642566819e-06, + "loss": 0.5815, + "step": 10116 + }, + { + "epoch": 3.74, + "learning_rate": 2.6280206922034736e-06, + "loss": 0.4826, + "step": 10117 + }, + { + "epoch": 3.74, + "learning_rate": 2.6279837418401282e-06, + "loss": 0.5069, + "step": 10118 + }, + { + "epoch": 3.74, + "learning_rate": 2.627946791476783e-06, + "loss": 0.4823, + "step": 10119 + }, + { + "epoch": 3.74, + "learning_rate": 2.6279098411134375e-06, + "loss": 0.5743, + "step": 10120 + }, + { + "epoch": 3.74, + "learning_rate": 2.6278728907500926e-06, + "loss": 0.5951, + "step": 10121 + }, + { + "epoch": 3.74, + "learning_rate": 2.627835940386747e-06, + "loss": 0.4864, + "step": 10122 + }, + { + "epoch": 3.74, + "learning_rate": 2.6277989900234023e-06, + "loss": 0.4984, + "step": 10123 + }, + { + "epoch": 3.74, + "learning_rate": 2.627762039660057e-06, + "loss": 0.6377, + "step": 10124 + }, + { + "epoch": 3.74, + "learning_rate": 2.6277250892967115e-06, + "loss": 0.706, + "step": 10125 + }, + { + "epoch": 3.74, + "learning_rate": 2.627688138933366e-06, + "loss": 0.5456, + "step": 10126 + }, + { + "epoch": 3.74, + "learning_rate": 2.627651188570021e-06, + "loss": 0.5416, + "step": 10127 + }, + { + "epoch": 3.74, + "learning_rate": 2.627614238206676e-06, + "loss": 0.615, + "step": 10128 + }, + { + "epoch": 3.74, + "learning_rate": 2.6275772878433305e-06, + "loss": 0.5954, + "step": 10129 + }, + { + "epoch": 3.74, + "learning_rate": 2.6275403374799856e-06, + "loss": 0.6881, + "step": 10130 + }, + { + "epoch": 3.74, + "learning_rate": 2.6275033871166398e-06, + "loss": 0.6257, + "step": 10131 + }, + { + "epoch": 3.74, + "learning_rate": 2.627466436753295e-06, + "loss": 0.5443, + "step": 10132 + }, + { + "epoch": 3.74, + "learning_rate": 2.6274294863899495e-06, + "loss": 0.5056, + "step": 10133 + }, + { + "epoch": 3.74, + "learning_rate": 2.6273925360266045e-06, + "loss": 0.4621, + "step": 10134 + }, + { + "epoch": 3.74, + "learning_rate": 2.627355585663259e-06, + "loss": 0.6424, + "step": 10135 + }, + { + "epoch": 3.74, + "learning_rate": 2.627318635299914e-06, + "loss": 0.5774, + "step": 10136 + }, + { + "epoch": 3.74, + "learning_rate": 2.6272816849365684e-06, + "loss": 0.5576, + "step": 10137 + }, + { + "epoch": 3.74, + "learning_rate": 2.627244734573223e-06, + "loss": 0.4801, + "step": 10138 + }, + { + "epoch": 3.74, + "learning_rate": 2.627207784209878e-06, + "loss": 0.5343, + "step": 10139 + }, + { + "epoch": 3.74, + "learning_rate": 2.6271708338465328e-06, + "loss": 0.5949, + "step": 10140 + }, + { + "epoch": 3.74, + "learning_rate": 2.627133883483188e-06, + "loss": 0.5883, + "step": 10141 + }, + { + "epoch": 3.75, + "learning_rate": 2.6270969331198425e-06, + "loss": 0.493, + "step": 10142 + }, + { + "epoch": 3.75, + "learning_rate": 2.627059982756497e-06, + "loss": 0.4798, + "step": 10143 + }, + { + "epoch": 3.75, + "learning_rate": 2.6270230323931517e-06, + "loss": 0.5016, + "step": 10144 + }, + { + "epoch": 3.75, + "learning_rate": 2.626986082029807e-06, + "loss": 0.6119, + "step": 10145 + }, + { + "epoch": 3.75, + "learning_rate": 2.6269491316664614e-06, + "loss": 0.5702, + "step": 10146 + }, + { + "epoch": 3.75, + "learning_rate": 2.626912181303116e-06, + "loss": 0.5923, + "step": 10147 + }, + { + "epoch": 3.75, + "learning_rate": 2.626875230939771e-06, + "loss": 0.5286, + "step": 10148 + }, + { + "epoch": 3.75, + "learning_rate": 2.6268382805764258e-06, + "loss": 0.5425, + "step": 10149 + }, + { + "epoch": 3.75, + "learning_rate": 2.6268013302130804e-06, + "loss": 0.579, + "step": 10150 + }, + { + "epoch": 3.75, + "learning_rate": 2.626764379849735e-06, + "loss": 0.4564, + "step": 10151 + }, + { + "epoch": 3.75, + "learning_rate": 2.62672742948639e-06, + "loss": 0.4793, + "step": 10152 + }, + { + "epoch": 3.75, + "learning_rate": 2.6266904791230447e-06, + "loss": 0.5488, + "step": 10153 + }, + { + "epoch": 3.75, + "learning_rate": 2.6266535287597e-06, + "loss": 0.521, + "step": 10154 + }, + { + "epoch": 3.75, + "learning_rate": 2.6266165783963544e-06, + "loss": 0.5767, + "step": 10155 + }, + { + "epoch": 3.75, + "learning_rate": 2.626579628033009e-06, + "loss": 0.6723, + "step": 10156 + }, + { + "epoch": 3.75, + "learning_rate": 2.6265426776696637e-06, + "loss": 0.4679, + "step": 10157 + }, + { + "epoch": 3.75, + "learning_rate": 2.6265057273063183e-06, + "loss": 0.5617, + "step": 10158 + }, + { + "epoch": 3.75, + "learning_rate": 2.6264687769429734e-06, + "loss": 0.5456, + "step": 10159 + }, + { + "epoch": 3.75, + "learning_rate": 2.626431826579628e-06, + "loss": 0.5177, + "step": 10160 + }, + { + "epoch": 3.75, + "learning_rate": 2.626394876216283e-06, + "loss": 0.5162, + "step": 10161 + }, + { + "epoch": 3.75, + "learning_rate": 2.6263579258529377e-06, + "loss": 0.4546, + "step": 10162 + }, + { + "epoch": 3.75, + "learning_rate": 2.6263209754895924e-06, + "loss": 0.5557, + "step": 10163 + }, + { + "epoch": 3.75, + "learning_rate": 2.626284025126247e-06, + "loss": 0.5259, + "step": 10164 + }, + { + "epoch": 3.75, + "learning_rate": 2.626247074762902e-06, + "loss": 0.4808, + "step": 10165 + }, + { + "epoch": 3.75, + "learning_rate": 2.6262101243995567e-06, + "loss": 0.561, + "step": 10166 + }, + { + "epoch": 3.75, + "learning_rate": 2.6261731740362113e-06, + "loss": 0.6163, + "step": 10167 + }, + { + "epoch": 3.75, + "learning_rate": 2.6261362236728664e-06, + "loss": 0.5263, + "step": 10168 + }, + { + "epoch": 3.76, + "learning_rate": 2.6260992733095206e-06, + "loss": 0.5528, + "step": 10169 + }, + { + "epoch": 3.76, + "learning_rate": 2.6260623229461757e-06, + "loss": 0.5551, + "step": 10170 + }, + { + "epoch": 3.76, + "learning_rate": 2.6260253725828303e-06, + "loss": 0.5977, + "step": 10171 + }, + { + "epoch": 3.76, + "learning_rate": 2.6259884222194854e-06, + "loss": 0.5246, + "step": 10172 + }, + { + "epoch": 3.76, + "learning_rate": 2.62595147185614e-06, + "loss": 0.583, + "step": 10173 + }, + { + "epoch": 3.76, + "learning_rate": 2.625914521492795e-06, + "loss": 0.5725, + "step": 10174 + }, + { + "epoch": 3.76, + "learning_rate": 2.6258775711294493e-06, + "loss": 0.4844, + "step": 10175 + }, + { + "epoch": 3.76, + "learning_rate": 2.6258406207661043e-06, + "loss": 0.5323, + "step": 10176 + }, + { + "epoch": 3.76, + "learning_rate": 2.625803670402759e-06, + "loss": 0.5422, + "step": 10177 + }, + { + "epoch": 3.76, + "learning_rate": 2.6257667200394136e-06, + "loss": 0.4216, + "step": 10178 + }, + { + "epoch": 3.76, + "learning_rate": 2.6257297696760687e-06, + "loss": 0.5542, + "step": 10179 + }, + { + "epoch": 3.76, + "learning_rate": 2.6256928193127233e-06, + "loss": 0.4725, + "step": 10180 + }, + { + "epoch": 3.76, + "learning_rate": 2.625655868949378e-06, + "loss": 0.4864, + "step": 10181 + }, + { + "epoch": 3.76, + "learning_rate": 2.6256189185860326e-06, + "loss": 0.5483, + "step": 10182 + }, + { + "epoch": 3.76, + "learning_rate": 2.6255819682226876e-06, + "loss": 0.4837, + "step": 10183 + }, + { + "epoch": 3.76, + "learning_rate": 2.6255450178593423e-06, + "loss": 0.4733, + "step": 10184 + }, + { + "epoch": 3.76, + "learning_rate": 2.6255080674959973e-06, + "loss": 0.5877, + "step": 10185 + }, + { + "epoch": 3.76, + "learning_rate": 2.625471117132652e-06, + "loss": 0.5367, + "step": 10186 + }, + { + "epoch": 3.76, + "learning_rate": 2.6254341667693066e-06, + "loss": 0.5652, + "step": 10187 + }, + { + "epoch": 3.76, + "learning_rate": 2.6253972164059612e-06, + "loss": 0.5604, + "step": 10188 + }, + { + "epoch": 3.76, + "learning_rate": 2.625360266042616e-06, + "loss": 0.5104, + "step": 10189 + }, + { + "epoch": 3.76, + "learning_rate": 2.625323315679271e-06, + "loss": 0.5949, + "step": 10190 + }, + { + "epoch": 3.76, + "learning_rate": 2.6252863653159256e-06, + "loss": 0.6051, + "step": 10191 + }, + { + "epoch": 3.76, + "learning_rate": 2.6252494149525806e-06, + "loss": 0.5056, + "step": 10192 + }, + { + "epoch": 3.76, + "learning_rate": 2.6252124645892353e-06, + "loss": 0.5839, + "step": 10193 + }, + { + "epoch": 3.76, + "learning_rate": 2.62517551422589e-06, + "loss": 0.5139, + "step": 10194 + }, + { + "epoch": 3.76, + "learning_rate": 2.6251385638625445e-06, + "loss": 0.7087, + "step": 10195 + }, + { + "epoch": 3.77, + "learning_rate": 2.6251016134991996e-06, + "loss": 0.5734, + "step": 10196 + }, + { + "epoch": 3.77, + "learning_rate": 2.6250646631358542e-06, + "loss": 0.6256, + "step": 10197 + }, + { + "epoch": 3.77, + "learning_rate": 2.625027712772509e-06, + "loss": 0.6108, + "step": 10198 + }, + { + "epoch": 3.77, + "learning_rate": 2.624990762409164e-06, + "loss": 0.5429, + "step": 10199 + }, + { + "epoch": 3.77, + "learning_rate": 2.6249538120458186e-06, + "loss": 0.5503, + "step": 10200 + }, + { + "epoch": 3.77, + "learning_rate": 2.624916861682473e-06, + "loss": 0.3922, + "step": 10201 + }, + { + "epoch": 3.77, + "learning_rate": 2.624879911319128e-06, + "loss": 0.552, + "step": 10202 + }, + { + "epoch": 3.77, + "learning_rate": 2.624842960955783e-06, + "loss": 0.5988, + "step": 10203 + }, + { + "epoch": 3.77, + "learning_rate": 2.6248060105924375e-06, + "loss": 0.5758, + "step": 10204 + }, + { + "epoch": 3.77, + "learning_rate": 2.6247690602290926e-06, + "loss": 0.5855, + "step": 10205 + }, + { + "epoch": 3.77, + "learning_rate": 2.6247321098657472e-06, + "loss": 0.5136, + "step": 10206 + }, + { + "epoch": 3.77, + "learning_rate": 2.624695159502402e-06, + "loss": 0.6076, + "step": 10207 + }, + { + "epoch": 3.77, + "learning_rate": 2.6246582091390565e-06, + "loss": 0.6197, + "step": 10208 + }, + { + "epoch": 3.77, + "learning_rate": 2.624621258775711e-06, + "loss": 0.5952, + "step": 10209 + }, + { + "epoch": 3.77, + "learning_rate": 2.624584308412366e-06, + "loss": 0.498, + "step": 10210 + }, + { + "epoch": 3.77, + "learning_rate": 2.624547358049021e-06, + "loss": 0.5086, + "step": 10211 + }, + { + "epoch": 3.77, + "learning_rate": 2.624510407685676e-06, + "loss": 0.5572, + "step": 10212 + }, + { + "epoch": 3.77, + "learning_rate": 2.6244734573223305e-06, + "loss": 0.4448, + "step": 10213 + }, + { + "epoch": 3.77, + "learning_rate": 2.624436506958985e-06, + "loss": 0.5451, + "step": 10214 + }, + { + "epoch": 3.77, + "learning_rate": 2.62439955659564e-06, + "loss": 0.5588, + "step": 10215 + }, + { + "epoch": 3.77, + "learning_rate": 2.624362606232295e-06, + "loss": 0.5808, + "step": 10216 + }, + { + "epoch": 3.77, + "learning_rate": 2.6243256558689495e-06, + "loss": 0.5063, + "step": 10217 + }, + { + "epoch": 3.77, + "learning_rate": 2.624288705505604e-06, + "loss": 0.4694, + "step": 10218 + }, + { + "epoch": 3.77, + "learning_rate": 2.624251755142259e-06, + "loss": 0.4997, + "step": 10219 + }, + { + "epoch": 3.77, + "learning_rate": 2.6242148047789134e-06, + "loss": 0.578, + "step": 10220 + }, + { + "epoch": 3.77, + "learning_rate": 2.6241778544155685e-06, + "loss": 0.4651, + "step": 10221 + }, + { + "epoch": 3.77, + "learning_rate": 2.624140904052223e-06, + "loss": 0.5549, + "step": 10222 + }, + { + "epoch": 3.78, + "learning_rate": 2.624103953688878e-06, + "loss": 0.4584, + "step": 10223 + }, + { + "epoch": 3.78, + "learning_rate": 2.624067003325533e-06, + "loss": 0.6199, + "step": 10224 + }, + { + "epoch": 3.78, + "learning_rate": 2.624030052962188e-06, + "loss": 0.5349, + "step": 10225 + }, + { + "epoch": 3.78, + "learning_rate": 2.623993102598842e-06, + "loss": 0.5115, + "step": 10226 + }, + { + "epoch": 3.78, + "learning_rate": 2.6239561522354967e-06, + "loss": 0.6035, + "step": 10227 + }, + { + "epoch": 3.78, + "learning_rate": 2.6239192018721518e-06, + "loss": 0.5592, + "step": 10228 + }, + { + "epoch": 3.78, + "learning_rate": 2.6238822515088064e-06, + "loss": 0.3911, + "step": 10229 + }, + { + "epoch": 3.78, + "learning_rate": 2.6238453011454615e-06, + "loss": 0.562, + "step": 10230 + }, + { + "epoch": 3.78, + "learning_rate": 2.623808350782116e-06, + "loss": 0.6523, + "step": 10231 + }, + { + "epoch": 3.78, + "learning_rate": 2.6237714004187708e-06, + "loss": 0.5414, + "step": 10232 + }, + { + "epoch": 3.78, + "learning_rate": 2.6237344500554254e-06, + "loss": 0.4969, + "step": 10233 + }, + { + "epoch": 3.78, + "learning_rate": 2.6236974996920804e-06, + "loss": 0.64, + "step": 10234 + }, + { + "epoch": 3.78, + "learning_rate": 2.623660549328735e-06, + "loss": 0.4195, + "step": 10235 + }, + { + "epoch": 3.78, + "learning_rate": 2.62362359896539e-06, + "loss": 0.4507, + "step": 10236 + }, + { + "epoch": 3.78, + "learning_rate": 2.6235866486020448e-06, + "loss": 0.4919, + "step": 10237 + }, + { + "epoch": 3.78, + "learning_rate": 2.6235496982386994e-06, + "loss": 0.5014, + "step": 10238 + }, + { + "epoch": 3.78, + "learning_rate": 2.623512747875354e-06, + "loss": 0.5814, + "step": 10239 + }, + { + "epoch": 3.78, + "learning_rate": 2.6234757975120087e-06, + "loss": 0.5153, + "step": 10240 + }, + { + "epoch": 3.78, + "learning_rate": 2.6234388471486637e-06, + "loss": 0.5444, + "step": 10241 + }, + { + "epoch": 3.78, + "learning_rate": 2.6234018967853184e-06, + "loss": 0.5545, + "step": 10242 + }, + { + "epoch": 3.78, + "learning_rate": 2.6233649464219734e-06, + "loss": 0.5918, + "step": 10243 + }, + { + "epoch": 3.78, + "learning_rate": 2.623327996058628e-06, + "loss": 0.5298, + "step": 10244 + }, + { + "epoch": 3.78, + "learning_rate": 2.6232910456952827e-06, + "loss": 0.4919, + "step": 10245 + }, + { + "epoch": 3.78, + "learning_rate": 2.6232540953319374e-06, + "loss": 0.5187, + "step": 10246 + }, + { + "epoch": 3.78, + "learning_rate": 2.623217144968592e-06, + "loss": 0.4946, + "step": 10247 + }, + { + "epoch": 3.78, + "learning_rate": 2.623180194605247e-06, + "loss": 0.5933, + "step": 10248 + }, + { + "epoch": 3.78, + "learning_rate": 2.6231432442419017e-06, + "loss": 0.5806, + "step": 10249 + }, + { + "epoch": 3.79, + "learning_rate": 2.6231062938785567e-06, + "loss": 0.6019, + "step": 10250 + }, + { + "epoch": 3.79, + "learning_rate": 2.6230693435152114e-06, + "loss": 0.5239, + "step": 10251 + }, + { + "epoch": 3.79, + "learning_rate": 2.623032393151866e-06, + "loss": 0.5092, + "step": 10252 + }, + { + "epoch": 3.79, + "learning_rate": 2.6229954427885207e-06, + "loss": 0.5693, + "step": 10253 + }, + { + "epoch": 3.79, + "learning_rate": 2.6229584924251757e-06, + "loss": 0.5505, + "step": 10254 + }, + { + "epoch": 3.79, + "learning_rate": 2.6229215420618304e-06, + "loss": 0.4245, + "step": 10255 + }, + { + "epoch": 3.79, + "learning_rate": 2.6228845916984854e-06, + "loss": 0.5879, + "step": 10256 + }, + { + "epoch": 3.79, + "learning_rate": 2.62284764133514e-06, + "loss": 0.5064, + "step": 10257 + }, + { + "epoch": 3.79, + "learning_rate": 2.6228106909717947e-06, + "loss": 0.5822, + "step": 10258 + }, + { + "epoch": 3.79, + "learning_rate": 2.6227737406084493e-06, + "loss": 0.5109, + "step": 10259 + }, + { + "epoch": 3.79, + "learning_rate": 2.622736790245104e-06, + "loss": 0.632, + "step": 10260 + }, + { + "epoch": 3.79, + "learning_rate": 2.622699839881759e-06, + "loss": 0.6227, + "step": 10261 + }, + { + "epoch": 3.79, + "learning_rate": 2.6226628895184137e-06, + "loss": 0.5533, + "step": 10262 + }, + { + "epoch": 3.79, + "learning_rate": 2.6226259391550687e-06, + "loss": 0.4521, + "step": 10263 + }, + { + "epoch": 3.79, + "learning_rate": 2.6225889887917234e-06, + "loss": 0.6601, + "step": 10264 + }, + { + "epoch": 3.79, + "learning_rate": 2.622552038428378e-06, + "loss": 0.5471, + "step": 10265 + }, + { + "epoch": 3.79, + "learning_rate": 2.6225150880650326e-06, + "loss": 0.5735, + "step": 10266 + }, + { + "epoch": 3.79, + "learning_rate": 2.6224781377016873e-06, + "loss": 0.5253, + "step": 10267 + }, + { + "epoch": 3.79, + "learning_rate": 2.6224411873383423e-06, + "loss": 0.4655, + "step": 10268 + }, + { + "epoch": 3.79, + "learning_rate": 2.622404236974997e-06, + "loss": 0.5741, + "step": 10269 + }, + { + "epoch": 3.79, + "learning_rate": 2.622367286611652e-06, + "loss": 0.5896, + "step": 10270 + }, + { + "epoch": 3.79, + "learning_rate": 2.6223303362483062e-06, + "loss": 0.4719, + "step": 10271 + }, + { + "epoch": 3.79, + "learning_rate": 2.6222933858849613e-06, + "loss": 0.5592, + "step": 10272 + }, + { + "epoch": 3.79, + "learning_rate": 2.622256435521616e-06, + "loss": 0.5977, + "step": 10273 + }, + { + "epoch": 3.79, + "learning_rate": 2.622219485158271e-06, + "loss": 0.4609, + "step": 10274 + }, + { + "epoch": 3.79, + "learning_rate": 2.6221825347949256e-06, + "loss": 0.5698, + "step": 10275 + }, + { + "epoch": 3.79, + "learning_rate": 2.6221455844315803e-06, + "loss": 0.6446, + "step": 10276 + }, + { + "epoch": 3.8, + "learning_rate": 2.622108634068235e-06, + "loss": 0.5248, + "step": 10277 + }, + { + "epoch": 3.8, + "learning_rate": 2.6220716837048895e-06, + "loss": 0.5365, + "step": 10278 + }, + { + "epoch": 3.8, + "learning_rate": 2.6220347333415446e-06, + "loss": 0.6436, + "step": 10279 + }, + { + "epoch": 3.8, + "learning_rate": 2.6219977829781992e-06, + "loss": 0.4593, + "step": 10280 + }, + { + "epoch": 3.8, + "learning_rate": 2.6219608326148543e-06, + "loss": 0.5883, + "step": 10281 + }, + { + "epoch": 3.8, + "learning_rate": 2.621923882251509e-06, + "loss": 0.4923, + "step": 10282 + }, + { + "epoch": 3.8, + "learning_rate": 2.6218869318881636e-06, + "loss": 0.5184, + "step": 10283 + }, + { + "epoch": 3.8, + "learning_rate": 2.621849981524818e-06, + "loss": 0.5647, + "step": 10284 + }, + { + "epoch": 3.8, + "learning_rate": 2.6218130311614733e-06, + "loss": 0.4859, + "step": 10285 + }, + { + "epoch": 3.8, + "learning_rate": 2.621776080798128e-06, + "loss": 0.4819, + "step": 10286 + }, + { + "epoch": 3.8, + "learning_rate": 2.6217391304347825e-06, + "loss": 0.501, + "step": 10287 + }, + { + "epoch": 3.8, + "learning_rate": 2.6217021800714376e-06, + "loss": 0.5734, + "step": 10288 + }, + { + "epoch": 3.8, + "learning_rate": 2.6216652297080922e-06, + "loss": 0.536, + "step": 10289 + }, + { + "epoch": 3.8, + "learning_rate": 2.621628279344747e-06, + "loss": 0.5276, + "step": 10290 + }, + { + "epoch": 3.8, + "learning_rate": 2.6215913289814015e-06, + "loss": 0.4968, + "step": 10291 + }, + { + "epoch": 3.8, + "learning_rate": 2.6215543786180566e-06, + "loss": 0.4548, + "step": 10292 + }, + { + "epoch": 3.8, + "learning_rate": 2.621517428254711e-06, + "loss": 0.5707, + "step": 10293 + }, + { + "epoch": 3.8, + "learning_rate": 2.6214804778913663e-06, + "loss": 0.5072, + "step": 10294 + }, + { + "epoch": 3.8, + "learning_rate": 2.621443527528021e-06, + "loss": 0.4268, + "step": 10295 + }, + { + "epoch": 3.8, + "learning_rate": 2.6214065771646755e-06, + "loss": 0.6332, + "step": 10296 + }, + { + "epoch": 3.8, + "learning_rate": 2.62136962680133e-06, + "loss": 0.4853, + "step": 10297 + }, + { + "epoch": 3.8, + "learning_rate": 2.621332676437985e-06, + "loss": 0.4761, + "step": 10298 + }, + { + "epoch": 3.8, + "learning_rate": 2.62129572607464e-06, + "loss": 0.4277, + "step": 10299 + }, + { + "epoch": 3.8, + "learning_rate": 2.6212587757112945e-06, + "loss": 0.5011, + "step": 10300 + }, + { + "epoch": 3.8, + "learning_rate": 2.6212218253479496e-06, + "loss": 0.4994, + "step": 10301 + }, + { + "epoch": 3.8, + "learning_rate": 2.621184874984604e-06, + "loss": 0.4767, + "step": 10302 + }, + { + "epoch": 3.8, + "learning_rate": 2.621147924621259e-06, + "loss": 0.4782, + "step": 10303 + }, + { + "epoch": 3.81, + "learning_rate": 2.6211109742579135e-06, + "loss": 0.6876, + "step": 10304 + }, + { + "epoch": 3.81, + "learning_rate": 2.6210740238945685e-06, + "loss": 0.5724, + "step": 10305 + }, + { + "epoch": 3.81, + "learning_rate": 2.621037073531223e-06, + "loss": 0.5514, + "step": 10306 + }, + { + "epoch": 3.81, + "learning_rate": 2.621000123167878e-06, + "loss": 0.5798, + "step": 10307 + }, + { + "epoch": 3.81, + "learning_rate": 2.620963172804533e-06, + "loss": 0.5888, + "step": 10308 + }, + { + "epoch": 3.81, + "learning_rate": 2.6209262224411875e-06, + "loss": 0.554, + "step": 10309 + }, + { + "epoch": 3.81, + "learning_rate": 2.620889272077842e-06, + "loss": 0.4297, + "step": 10310 + }, + { + "epoch": 3.81, + "learning_rate": 2.6208523217144968e-06, + "loss": 0.4896, + "step": 10311 + }, + { + "epoch": 3.81, + "learning_rate": 2.620815371351152e-06, + "loss": 0.5492, + "step": 10312 + }, + { + "epoch": 3.81, + "learning_rate": 2.6207784209878065e-06, + "loss": 0.4766, + "step": 10313 + }, + { + "epoch": 3.81, + "learning_rate": 2.6207414706244615e-06, + "loss": 0.5351, + "step": 10314 + }, + { + "epoch": 3.81, + "learning_rate": 2.620704520261116e-06, + "loss": 0.5232, + "step": 10315 + }, + { + "epoch": 3.81, + "learning_rate": 2.6206675698977704e-06, + "loss": 0.6448, + "step": 10316 + }, + { + "epoch": 3.81, + "learning_rate": 2.6206306195344254e-06, + "loss": 0.6335, + "step": 10317 + }, + { + "epoch": 3.81, + "learning_rate": 2.62059366917108e-06, + "loss": 0.512, + "step": 10318 + }, + { + "epoch": 3.81, + "learning_rate": 2.620556718807735e-06, + "loss": 0.6443, + "step": 10319 + }, + { + "epoch": 3.81, + "learning_rate": 2.6205197684443898e-06, + "loss": 0.562, + "step": 10320 + }, + { + "epoch": 3.81, + "learning_rate": 2.620482818081045e-06, + "loss": 0.5673, + "step": 10321 + }, + { + "epoch": 3.81, + "learning_rate": 2.620445867717699e-06, + "loss": 0.5399, + "step": 10322 + }, + { + "epoch": 3.81, + "learning_rate": 2.620408917354354e-06, + "loss": 0.5111, + "step": 10323 + }, + { + "epoch": 3.81, + "learning_rate": 2.6203719669910087e-06, + "loss": 0.519, + "step": 10324 + }, + { + "epoch": 3.81, + "learning_rate": 2.620335016627664e-06, + "loss": 0.5686, + "step": 10325 + }, + { + "epoch": 3.81, + "learning_rate": 2.6202980662643184e-06, + "loss": 0.4875, + "step": 10326 + }, + { + "epoch": 3.81, + "learning_rate": 2.620261115900973e-06, + "loss": 0.5985, + "step": 10327 + }, + { + "epoch": 3.81, + "learning_rate": 2.6202241655376277e-06, + "loss": 0.5576, + "step": 10328 + }, + { + "epoch": 3.81, + "learning_rate": 2.6201872151742823e-06, + "loss": 0.454, + "step": 10329 + }, + { + "epoch": 3.81, + "learning_rate": 2.6201502648109374e-06, + "loss": 0.6176, + "step": 10330 + }, + { + "epoch": 3.81, + "learning_rate": 2.620113314447592e-06, + "loss": 0.4815, + "step": 10331 + }, + { + "epoch": 3.82, + "learning_rate": 2.620076364084247e-06, + "loss": 0.5409, + "step": 10332 + }, + { + "epoch": 3.82, + "learning_rate": 2.6200394137209017e-06, + "loss": 0.5444, + "step": 10333 + }, + { + "epoch": 3.82, + "learning_rate": 2.6200024633575564e-06, + "loss": 0.5775, + "step": 10334 + }, + { + "epoch": 3.82, + "learning_rate": 2.619965512994211e-06, + "loss": 0.4974, + "step": 10335 + }, + { + "epoch": 3.82, + "learning_rate": 2.6199285626308656e-06, + "loss": 0.4596, + "step": 10336 + }, + { + "epoch": 3.82, + "learning_rate": 2.6198916122675207e-06, + "loss": 0.5357, + "step": 10337 + }, + { + "epoch": 3.82, + "learning_rate": 2.6198546619041753e-06, + "loss": 0.5074, + "step": 10338 + }, + { + "epoch": 3.82, + "learning_rate": 2.6198177115408304e-06, + "loss": 0.5016, + "step": 10339 + }, + { + "epoch": 3.82, + "learning_rate": 2.619780761177485e-06, + "loss": 0.5221, + "step": 10340 + }, + { + "epoch": 3.82, + "learning_rate": 2.6197438108141397e-06, + "loss": 0.4352, + "step": 10341 + }, + { + "epoch": 3.82, + "learning_rate": 2.6197068604507943e-06, + "loss": 0.5091, + "step": 10342 + }, + { + "epoch": 3.82, + "learning_rate": 2.6196699100874494e-06, + "loss": 0.5669, + "step": 10343 + }, + { + "epoch": 3.82, + "learning_rate": 2.619632959724104e-06, + "loss": 0.4835, + "step": 10344 + }, + { + "epoch": 3.82, + "learning_rate": 2.619596009360759e-06, + "loss": 0.4503, + "step": 10345 + }, + { + "epoch": 3.82, + "learning_rate": 2.6195590589974137e-06, + "loss": 0.5503, + "step": 10346 + }, + { + "epoch": 3.82, + "learning_rate": 2.6195221086340683e-06, + "loss": 0.574, + "step": 10347 + }, + { + "epoch": 3.82, + "learning_rate": 2.619485158270723e-06, + "loss": 0.6807, + "step": 10348 + }, + { + "epoch": 3.82, + "learning_rate": 2.6194482079073776e-06, + "loss": 0.5647, + "step": 10349 + }, + { + "epoch": 3.82, + "learning_rate": 2.6194112575440327e-06, + "loss": 0.6664, + "step": 10350 + }, + { + "epoch": 3.82, + "learning_rate": 2.6193743071806873e-06, + "loss": 0.5542, + "step": 10351 + }, + { + "epoch": 3.82, + "learning_rate": 2.6193373568173424e-06, + "loss": 0.4722, + "step": 10352 + }, + { + "epoch": 3.82, + "learning_rate": 2.619300406453997e-06, + "loss": 0.6013, + "step": 10353 + }, + { + "epoch": 3.82, + "learning_rate": 2.6192634560906516e-06, + "loss": 0.5287, + "step": 10354 + }, + { + "epoch": 3.82, + "learning_rate": 2.6192265057273063e-06, + "loss": 0.4863, + "step": 10355 + }, + { + "epoch": 3.82, + "learning_rate": 2.619189555363961e-06, + "loss": 0.4705, + "step": 10356 + }, + { + "epoch": 3.82, + "learning_rate": 2.619152605000616e-06, + "loss": 0.5285, + "step": 10357 + }, + { + "epoch": 3.82, + "learning_rate": 2.6191156546372706e-06, + "loss": 0.4821, + "step": 10358 + }, + { + "epoch": 3.83, + "learning_rate": 2.6190787042739257e-06, + "loss": 0.5284, + "step": 10359 + }, + { + "epoch": 3.83, + "learning_rate": 2.61904175391058e-06, + "loss": 0.479, + "step": 10360 + }, + { + "epoch": 3.83, + "learning_rate": 2.619004803547235e-06, + "loss": 0.6735, + "step": 10361 + }, + { + "epoch": 3.83, + "learning_rate": 2.6189678531838896e-06, + "loss": 0.5364, + "step": 10362 + }, + { + "epoch": 3.83, + "learning_rate": 2.6189309028205446e-06, + "loss": 0.5386, + "step": 10363 + }, + { + "epoch": 3.83, + "learning_rate": 2.6188939524571993e-06, + "loss": 0.6143, + "step": 10364 + }, + { + "epoch": 3.83, + "learning_rate": 2.6188570020938543e-06, + "loss": 0.5435, + "step": 10365 + }, + { + "epoch": 3.83, + "learning_rate": 2.6188200517305086e-06, + "loss": 0.5198, + "step": 10366 + }, + { + "epoch": 3.83, + "learning_rate": 2.618783101367163e-06, + "loss": 0.4626, + "step": 10367 + }, + { + "epoch": 3.83, + "learning_rate": 2.6187461510038182e-06, + "loss": 0.6519, + "step": 10368 + }, + { + "epoch": 3.83, + "learning_rate": 2.618709200640473e-06, + "loss": 0.6262, + "step": 10369 + }, + { + "epoch": 3.83, + "learning_rate": 2.618672250277128e-06, + "loss": 0.6844, + "step": 10370 + }, + { + "epoch": 3.83, + "learning_rate": 2.6186352999137826e-06, + "loss": 0.5668, + "step": 10371 + }, + { + "epoch": 3.83, + "learning_rate": 2.6185983495504372e-06, + "loss": 0.5467, + "step": 10372 + }, + { + "epoch": 3.83, + "learning_rate": 2.618561399187092e-06, + "loss": 0.6501, + "step": 10373 + }, + { + "epoch": 3.83, + "learning_rate": 2.618524448823747e-06, + "loss": 0.5063, + "step": 10374 + }, + { + "epoch": 3.83, + "learning_rate": 2.6184874984604015e-06, + "loss": 0.5014, + "step": 10375 + }, + { + "epoch": 3.83, + "learning_rate": 2.618450548097056e-06, + "loss": 0.5074, + "step": 10376 + }, + { + "epoch": 3.83, + "learning_rate": 2.6184135977337112e-06, + "loss": 0.4373, + "step": 10377 + }, + { + "epoch": 3.83, + "learning_rate": 2.618376647370366e-06, + "loss": 0.5068, + "step": 10378 + }, + { + "epoch": 3.83, + "learning_rate": 2.6183396970070205e-06, + "loss": 0.6086, + "step": 10379 + }, + { + "epoch": 3.83, + "learning_rate": 2.618302746643675e-06, + "loss": 0.4804, + "step": 10380 + }, + { + "epoch": 3.83, + "learning_rate": 2.6182657962803302e-06, + "loss": 0.613, + "step": 10381 + }, + { + "epoch": 3.83, + "learning_rate": 2.618228845916985e-06, + "loss": 0.6118, + "step": 10382 + }, + { + "epoch": 3.83, + "learning_rate": 2.61819189555364e-06, + "loss": 0.5126, + "step": 10383 + }, + { + "epoch": 3.83, + "learning_rate": 2.6181549451902945e-06, + "loss": 0.5289, + "step": 10384 + }, + { + "epoch": 3.83, + "learning_rate": 2.618117994826949e-06, + "loss": 0.5524, + "step": 10385 + }, + { + "epoch": 3.84, + "learning_rate": 2.618081044463604e-06, + "loss": 0.5204, + "step": 10386 + }, + { + "epoch": 3.84, + "learning_rate": 2.6180440941002585e-06, + "loss": 0.5347, + "step": 10387 + }, + { + "epoch": 3.84, + "learning_rate": 2.6180071437369135e-06, + "loss": 0.6386, + "step": 10388 + }, + { + "epoch": 3.84, + "learning_rate": 2.617970193373568e-06, + "loss": 0.5482, + "step": 10389 + }, + { + "epoch": 3.84, + "learning_rate": 2.6179332430102232e-06, + "loss": 0.619, + "step": 10390 + }, + { + "epoch": 3.84, + "learning_rate": 2.617896292646878e-06, + "loss": 0.608, + "step": 10391 + }, + { + "epoch": 3.84, + "learning_rate": 2.6178593422835325e-06, + "loss": 0.6216, + "step": 10392 + }, + { + "epoch": 3.84, + "learning_rate": 2.617822391920187e-06, + "loss": 0.6297, + "step": 10393 + }, + { + "epoch": 3.84, + "learning_rate": 2.617785441556842e-06, + "loss": 0.4704, + "step": 10394 + }, + { + "epoch": 3.84, + "learning_rate": 2.617748491193497e-06, + "loss": 0.5315, + "step": 10395 + }, + { + "epoch": 3.84, + "learning_rate": 2.6177115408301515e-06, + "loss": 0.4749, + "step": 10396 + }, + { + "epoch": 3.84, + "learning_rate": 2.6176745904668065e-06, + "loss": 0.6698, + "step": 10397 + }, + { + "epoch": 3.84, + "learning_rate": 2.617637640103461e-06, + "loss": 0.5977, + "step": 10398 + }, + { + "epoch": 3.84, + "learning_rate": 2.6176006897401158e-06, + "loss": 0.4819, + "step": 10399 + }, + { + "epoch": 3.84, + "learning_rate": 2.6175637393767704e-06, + "loss": 0.5522, + "step": 10400 + }, + { + "epoch": 3.84, + "learning_rate": 2.6175267890134255e-06, + "loss": 0.5159, + "step": 10401 + }, + { + "epoch": 3.84, + "learning_rate": 2.61748983865008e-06, + "loss": 0.459, + "step": 10402 + }, + { + "epoch": 3.84, + "learning_rate": 2.617452888286735e-06, + "loss": 0.5665, + "step": 10403 + }, + { + "epoch": 3.84, + "learning_rate": 2.61741593792339e-06, + "loss": 0.5836, + "step": 10404 + }, + { + "epoch": 3.84, + "learning_rate": 2.6173789875600445e-06, + "loss": 0.5071, + "step": 10405 + }, + { + "epoch": 3.84, + "learning_rate": 2.617342037196699e-06, + "loss": 0.5725, + "step": 10406 + }, + { + "epoch": 3.84, + "learning_rate": 2.6173050868333537e-06, + "loss": 0.5065, + "step": 10407 + }, + { + "epoch": 3.84, + "learning_rate": 2.6172681364700088e-06, + "loss": 0.5459, + "step": 10408 + }, + { + "epoch": 3.84, + "learning_rate": 2.6172311861066634e-06, + "loss": 0.5772, + "step": 10409 + }, + { + "epoch": 3.84, + "learning_rate": 2.6171942357433185e-06, + "loss": 0.502, + "step": 10410 + }, + { + "epoch": 3.84, + "learning_rate": 2.6171572853799727e-06, + "loss": 0.5876, + "step": 10411 + }, + { + "epoch": 3.84, + "learning_rate": 2.6171203350166278e-06, + "loss": 0.5449, + "step": 10412 + }, + { + "epoch": 3.85, + "learning_rate": 2.6170833846532824e-06, + "loss": 0.6633, + "step": 10413 + }, + { + "epoch": 3.85, + "learning_rate": 2.6170464342899375e-06, + "loss": 0.6104, + "step": 10414 + }, + { + "epoch": 3.85, + "learning_rate": 2.617009483926592e-06, + "loss": 0.568, + "step": 10415 + }, + { + "epoch": 3.85, + "learning_rate": 2.6169725335632467e-06, + "loss": 0.5445, + "step": 10416 + }, + { + "epoch": 3.85, + "learning_rate": 2.6169355831999014e-06, + "loss": 0.6323, + "step": 10417 + }, + { + "epoch": 3.85, + "learning_rate": 2.616898632836556e-06, + "loss": 0.4874, + "step": 10418 + }, + { + "epoch": 3.85, + "learning_rate": 2.616861682473211e-06, + "loss": 0.6517, + "step": 10419 + }, + { + "epoch": 3.85, + "learning_rate": 2.6168247321098657e-06, + "loss": 0.596, + "step": 10420 + }, + { + "epoch": 3.85, + "learning_rate": 2.6167877817465208e-06, + "loss": 0.6827, + "step": 10421 + }, + { + "epoch": 3.85, + "learning_rate": 2.6167508313831754e-06, + "loss": 0.511, + "step": 10422 + }, + { + "epoch": 3.85, + "learning_rate": 2.61671388101983e-06, + "loss": 0.564, + "step": 10423 + }, + { + "epoch": 3.85, + "learning_rate": 2.6166769306564847e-06, + "loss": 0.6556, + "step": 10424 + }, + { + "epoch": 3.85, + "learning_rate": 2.6166399802931397e-06, + "loss": 0.5129, + "step": 10425 + }, + { + "epoch": 3.85, + "learning_rate": 2.6166030299297944e-06, + "loss": 0.5774, + "step": 10426 + }, + { + "epoch": 3.85, + "learning_rate": 2.616566079566449e-06, + "loss": 0.6294, + "step": 10427 + }, + { + "epoch": 3.85, + "learning_rate": 2.616529129203104e-06, + "loss": 0.607, + "step": 10428 + }, + { + "epoch": 3.85, + "learning_rate": 2.6164921788397587e-06, + "loss": 0.5616, + "step": 10429 + }, + { + "epoch": 3.85, + "learning_rate": 2.6164552284764133e-06, + "loss": 0.4546, + "step": 10430 + }, + { + "epoch": 3.85, + "learning_rate": 2.616418278113068e-06, + "loss": 0.5643, + "step": 10431 + }, + { + "epoch": 3.85, + "learning_rate": 2.616381327749723e-06, + "loss": 0.4529, + "step": 10432 + }, + { + "epoch": 3.85, + "learning_rate": 2.6163443773863777e-06, + "loss": 0.5919, + "step": 10433 + }, + { + "epoch": 3.85, + "learning_rate": 2.6163074270230327e-06, + "loss": 0.6888, + "step": 10434 + }, + { + "epoch": 3.85, + "learning_rate": 2.6162704766596874e-06, + "loss": 0.5341, + "step": 10435 + }, + { + "epoch": 3.85, + "learning_rate": 2.616233526296342e-06, + "loss": 0.5218, + "step": 10436 + }, + { + "epoch": 3.85, + "learning_rate": 2.6161965759329966e-06, + "loss": 0.5255, + "step": 10437 + }, + { + "epoch": 3.85, + "learning_rate": 2.6161596255696513e-06, + "loss": 0.472, + "step": 10438 + }, + { + "epoch": 3.85, + "learning_rate": 2.6161226752063063e-06, + "loss": 0.4436, + "step": 10439 + }, + { + "epoch": 3.86, + "learning_rate": 2.616085724842961e-06, + "loss": 0.5985, + "step": 10440 + }, + { + "epoch": 3.86, + "learning_rate": 2.616048774479616e-06, + "loss": 0.5443, + "step": 10441 + }, + { + "epoch": 3.86, + "learning_rate": 2.6160118241162707e-06, + "loss": 0.5063, + "step": 10442 + }, + { + "epoch": 3.86, + "learning_rate": 2.6159748737529253e-06, + "loss": 0.5509, + "step": 10443 + }, + { + "epoch": 3.86, + "learning_rate": 2.61593792338958e-06, + "loss": 0.5048, + "step": 10444 + }, + { + "epoch": 3.86, + "learning_rate": 2.615900973026235e-06, + "loss": 0.5929, + "step": 10445 + }, + { + "epoch": 3.86, + "learning_rate": 2.6158640226628896e-06, + "loss": 0.4958, + "step": 10446 + }, + { + "epoch": 3.86, + "learning_rate": 2.6158270722995443e-06, + "loss": 0.5719, + "step": 10447 + }, + { + "epoch": 3.86, + "learning_rate": 2.6157901219361993e-06, + "loss": 0.5183, + "step": 10448 + }, + { + "epoch": 3.86, + "learning_rate": 2.615753171572854e-06, + "loss": 0.5497, + "step": 10449 + }, + { + "epoch": 3.86, + "learning_rate": 2.6157162212095086e-06, + "loss": 0.5547, + "step": 10450 + }, + { + "epoch": 3.86, + "learning_rate": 2.6156792708461632e-06, + "loss": 0.5514, + "step": 10451 + }, + { + "epoch": 3.86, + "learning_rate": 2.6156423204828183e-06, + "loss": 0.6276, + "step": 10452 + }, + { + "epoch": 3.86, + "learning_rate": 2.615605370119473e-06, + "loss": 0.6192, + "step": 10453 + }, + { + "epoch": 3.86, + "learning_rate": 2.615568419756128e-06, + "loss": 0.5342, + "step": 10454 + }, + { + "epoch": 3.86, + "learning_rate": 2.6155314693927826e-06, + "loss": 0.4796, + "step": 10455 + }, + { + "epoch": 3.86, + "learning_rate": 2.615494519029437e-06, + "loss": 0.4657, + "step": 10456 + }, + { + "epoch": 3.86, + "learning_rate": 2.615457568666092e-06, + "loss": 0.4738, + "step": 10457 + }, + { + "epoch": 3.86, + "learning_rate": 2.6154206183027465e-06, + "loss": 0.5912, + "step": 10458 + }, + { + "epoch": 3.86, + "learning_rate": 2.6153836679394016e-06, + "loss": 0.5412, + "step": 10459 + }, + { + "epoch": 3.86, + "learning_rate": 2.6153467175760562e-06, + "loss": 0.5584, + "step": 10460 + }, + { + "epoch": 3.86, + "learning_rate": 2.6153097672127113e-06, + "loss": 0.5559, + "step": 10461 + }, + { + "epoch": 3.86, + "learning_rate": 2.6152728168493655e-06, + "loss": 0.6136, + "step": 10462 + }, + { + "epoch": 3.86, + "learning_rate": 2.6152358664860206e-06, + "loss": 0.555, + "step": 10463 + }, + { + "epoch": 3.86, + "learning_rate": 2.615198916122675e-06, + "loss": 0.4853, + "step": 10464 + }, + { + "epoch": 3.86, + "learning_rate": 2.6151619657593303e-06, + "loss": 0.4595, + "step": 10465 + }, + { + "epoch": 3.86, + "learning_rate": 2.615125015395985e-06, + "loss": 0.6188, + "step": 10466 + }, + { + "epoch": 3.87, + "learning_rate": 2.6150880650326395e-06, + "loss": 0.5593, + "step": 10467 + }, + { + "epoch": 3.87, + "learning_rate": 2.615051114669294e-06, + "loss": 0.4722, + "step": 10468 + }, + { + "epoch": 3.87, + "learning_rate": 2.615014164305949e-06, + "loss": 0.4842, + "step": 10469 + }, + { + "epoch": 3.87, + "learning_rate": 2.614977213942604e-06, + "loss": 0.5251, + "step": 10470 + }, + { + "epoch": 3.87, + "learning_rate": 2.6149402635792585e-06, + "loss": 0.5338, + "step": 10471 + }, + { + "epoch": 3.87, + "learning_rate": 2.6149033132159136e-06, + "loss": 0.4839, + "step": 10472 + }, + { + "epoch": 3.87, + "learning_rate": 2.614866362852568e-06, + "loss": 0.4628, + "step": 10473 + }, + { + "epoch": 3.87, + "learning_rate": 2.614829412489223e-06, + "loss": 0.5388, + "step": 10474 + }, + { + "epoch": 3.87, + "learning_rate": 2.6147924621258775e-06, + "loss": 0.5509, + "step": 10475 + }, + { + "epoch": 3.87, + "learning_rate": 2.614755511762532e-06, + "loss": 0.5027, + "step": 10476 + }, + { + "epoch": 3.87, + "learning_rate": 2.614718561399187e-06, + "loss": 0.6132, + "step": 10477 + }, + { + "epoch": 3.87, + "learning_rate": 2.614681611035842e-06, + "loss": 0.607, + "step": 10478 + }, + { + "epoch": 3.87, + "learning_rate": 2.614644660672497e-06, + "loss": 0.6178, + "step": 10479 + }, + { + "epoch": 3.87, + "learning_rate": 2.6146077103091515e-06, + "loss": 0.4983, + "step": 10480 + }, + { + "epoch": 3.87, + "learning_rate": 2.614570759945806e-06, + "loss": 0.6065, + "step": 10481 + }, + { + "epoch": 3.87, + "learning_rate": 2.6145338095824608e-06, + "loss": 0.5189, + "step": 10482 + }, + { + "epoch": 3.87, + "learning_rate": 2.614496859219116e-06, + "loss": 0.6221, + "step": 10483 + }, + { + "epoch": 3.87, + "learning_rate": 2.6144599088557705e-06, + "loss": 0.435, + "step": 10484 + }, + { + "epoch": 3.87, + "learning_rate": 2.6144229584924255e-06, + "loss": 0.5264, + "step": 10485 + }, + { + "epoch": 3.87, + "learning_rate": 2.61438600812908e-06, + "loss": 0.5839, + "step": 10486 + }, + { + "epoch": 3.87, + "learning_rate": 2.614349057765735e-06, + "loss": 0.4951, + "step": 10487 + }, + { + "epoch": 3.87, + "learning_rate": 2.6143121074023894e-06, + "loss": 0.58, + "step": 10488 + }, + { + "epoch": 3.87, + "learning_rate": 2.614275157039044e-06, + "loss": 0.5778, + "step": 10489 + }, + { + "epoch": 3.87, + "learning_rate": 2.614238206675699e-06, + "loss": 0.547, + "step": 10490 + }, + { + "epoch": 3.87, + "learning_rate": 2.6142012563123538e-06, + "loss": 0.6497, + "step": 10491 + }, + { + "epoch": 3.87, + "learning_rate": 2.614164305949009e-06, + "loss": 0.5372, + "step": 10492 + }, + { + "epoch": 3.87, + "learning_rate": 2.6141273555856635e-06, + "loss": 0.663, + "step": 10493 + }, + { + "epoch": 3.88, + "learning_rate": 2.614090405222318e-06, + "loss": 0.554, + "step": 10494 + }, + { + "epoch": 3.88, + "learning_rate": 2.6140534548589727e-06, + "loss": 0.5083, + "step": 10495 + }, + { + "epoch": 3.88, + "learning_rate": 2.6140165044956274e-06, + "loss": 0.5465, + "step": 10496 + }, + { + "epoch": 3.88, + "learning_rate": 2.6139795541322824e-06, + "loss": 0.4904, + "step": 10497 + }, + { + "epoch": 3.88, + "learning_rate": 2.613942603768937e-06, + "loss": 0.6269, + "step": 10498 + }, + { + "epoch": 3.88, + "learning_rate": 2.613905653405592e-06, + "loss": 0.6063, + "step": 10499 + }, + { + "epoch": 3.88, + "learning_rate": 2.6138687030422468e-06, + "loss": 0.5255, + "step": 10500 + }, + { + "epoch": 3.88, + "learning_rate": 2.6138317526789014e-06, + "loss": 0.5834, + "step": 10501 + }, + { + "epoch": 3.88, + "learning_rate": 2.613794802315556e-06, + "loss": 0.5281, + "step": 10502 + }, + { + "epoch": 3.88, + "learning_rate": 2.613757851952211e-06, + "loss": 0.4686, + "step": 10503 + }, + { + "epoch": 3.88, + "learning_rate": 2.6137209015888657e-06, + "loss": 0.5462, + "step": 10504 + }, + { + "epoch": 3.88, + "learning_rate": 2.6136839512255204e-06, + "loss": 0.5126, + "step": 10505 + }, + { + "epoch": 3.88, + "learning_rate": 2.6136470008621754e-06, + "loss": 0.5896, + "step": 10506 + }, + { + "epoch": 3.88, + "learning_rate": 2.6136100504988297e-06, + "loss": 0.5791, + "step": 10507 + }, + { + "epoch": 3.88, + "learning_rate": 2.6135731001354847e-06, + "loss": 0.6253, + "step": 10508 + }, + { + "epoch": 3.88, + "learning_rate": 2.6135361497721393e-06, + "loss": 0.5557, + "step": 10509 + }, + { + "epoch": 3.88, + "learning_rate": 2.6134991994087944e-06, + "loss": 0.6867, + "step": 10510 + }, + { + "epoch": 3.88, + "learning_rate": 2.613462249045449e-06, + "loss": 0.5186, + "step": 10511 + }, + { + "epoch": 3.88, + "learning_rate": 2.6134252986821037e-06, + "loss": 0.4666, + "step": 10512 + }, + { + "epoch": 3.88, + "learning_rate": 2.6133883483187583e-06, + "loss": 0.5126, + "step": 10513 + }, + { + "epoch": 3.88, + "learning_rate": 2.6133513979554134e-06, + "loss": 0.5489, + "step": 10514 + }, + { + "epoch": 3.88, + "learning_rate": 2.613314447592068e-06, + "loss": 0.5335, + "step": 10515 + }, + { + "epoch": 3.88, + "learning_rate": 2.6132774972287227e-06, + "loss": 0.5924, + "step": 10516 + }, + { + "epoch": 3.88, + "learning_rate": 2.6132405468653777e-06, + "loss": 0.5641, + "step": 10517 + }, + { + "epoch": 3.88, + "learning_rate": 2.6132035965020323e-06, + "loss": 0.5779, + "step": 10518 + }, + { + "epoch": 3.88, + "learning_rate": 2.613166646138687e-06, + "loss": 0.5401, + "step": 10519 + }, + { + "epoch": 3.88, + "learning_rate": 2.6131296957753416e-06, + "loss": 0.6276, + "step": 10520 + }, + { + "epoch": 3.89, + "learning_rate": 2.6130927454119967e-06, + "loss": 0.5509, + "step": 10521 + }, + { + "epoch": 3.89, + "learning_rate": 2.6130557950486513e-06, + "loss": 0.5603, + "step": 10522 + }, + { + "epoch": 3.89, + "learning_rate": 2.6130188446853064e-06, + "loss": 0.501, + "step": 10523 + }, + { + "epoch": 3.89, + "learning_rate": 2.612981894321961e-06, + "loss": 0.4458, + "step": 10524 + }, + { + "epoch": 3.89, + "learning_rate": 2.6129449439586156e-06, + "loss": 0.5624, + "step": 10525 + }, + { + "epoch": 3.89, + "learning_rate": 2.6129079935952703e-06, + "loss": 0.5487, + "step": 10526 + }, + { + "epoch": 3.89, + "learning_rate": 2.612871043231925e-06, + "loss": 0.4914, + "step": 10527 + }, + { + "epoch": 3.89, + "learning_rate": 2.61283409286858e-06, + "loss": 0.6589, + "step": 10528 + }, + { + "epoch": 3.89, + "learning_rate": 2.6127971425052346e-06, + "loss": 0.5134, + "step": 10529 + }, + { + "epoch": 3.89, + "learning_rate": 2.6127601921418897e-06, + "loss": 0.5141, + "step": 10530 + }, + { + "epoch": 3.89, + "learning_rate": 2.6127232417785443e-06, + "loss": 0.5173, + "step": 10531 + }, + { + "epoch": 3.89, + "learning_rate": 2.612686291415199e-06, + "loss": 0.5767, + "step": 10532 + }, + { + "epoch": 3.89, + "learning_rate": 2.6126493410518536e-06, + "loss": 0.5662, + "step": 10533 + }, + { + "epoch": 3.89, + "learning_rate": 2.6126123906885086e-06, + "loss": 0.5603, + "step": 10534 + }, + { + "epoch": 3.89, + "learning_rate": 2.6125754403251633e-06, + "loss": 0.493, + "step": 10535 + }, + { + "epoch": 3.89, + "learning_rate": 2.612538489961818e-06, + "loss": 0.5092, + "step": 10536 + }, + { + "epoch": 3.89, + "learning_rate": 2.612501539598473e-06, + "loss": 0.5124, + "step": 10537 + }, + { + "epoch": 3.89, + "learning_rate": 2.6124645892351276e-06, + "loss": 0.5422, + "step": 10538 + }, + { + "epoch": 3.89, + "learning_rate": 2.6124276388717823e-06, + "loss": 0.602, + "step": 10539 + }, + { + "epoch": 3.89, + "learning_rate": 2.612390688508437e-06, + "loss": 0.579, + "step": 10540 + }, + { + "epoch": 3.89, + "learning_rate": 2.612353738145092e-06, + "loss": 0.4442, + "step": 10541 + }, + { + "epoch": 3.89, + "learning_rate": 2.6123167877817466e-06, + "loss": 0.5057, + "step": 10542 + }, + { + "epoch": 3.89, + "learning_rate": 2.6122798374184016e-06, + "loss": 0.4313, + "step": 10543 + }, + { + "epoch": 3.89, + "learning_rate": 2.6122428870550563e-06, + "loss": 0.5794, + "step": 10544 + }, + { + "epoch": 3.89, + "learning_rate": 2.6122059366917105e-06, + "loss": 0.4975, + "step": 10545 + }, + { + "epoch": 3.89, + "learning_rate": 2.6121689863283656e-06, + "loss": 0.5672, + "step": 10546 + }, + { + "epoch": 3.89, + "learning_rate": 2.61213203596502e-06, + "loss": 0.5151, + "step": 10547 + }, + { + "epoch": 3.9, + "learning_rate": 2.6120950856016753e-06, + "loss": 0.5642, + "step": 10548 + }, + { + "epoch": 3.9, + "learning_rate": 2.61205813523833e-06, + "loss": 0.4196, + "step": 10549 + }, + { + "epoch": 3.9, + "learning_rate": 2.612021184874985e-06, + "loss": 0.5414, + "step": 10550 + }, + { + "epoch": 3.9, + "learning_rate": 2.611984234511639e-06, + "loss": 0.5396, + "step": 10551 + }, + { + "epoch": 3.9, + "learning_rate": 2.6119472841482942e-06, + "loss": 0.6236, + "step": 10552 + }, + { + "epoch": 3.9, + "learning_rate": 2.611910333784949e-06, + "loss": 0.5778, + "step": 10553 + }, + { + "epoch": 3.9, + "learning_rate": 2.611873383421604e-06, + "loss": 0.4123, + "step": 10554 + }, + { + "epoch": 3.9, + "learning_rate": 2.6118364330582586e-06, + "loss": 0.4658, + "step": 10555 + }, + { + "epoch": 3.9, + "learning_rate": 2.611799482694913e-06, + "loss": 0.6907, + "step": 10556 + }, + { + "epoch": 3.9, + "learning_rate": 2.611762532331568e-06, + "loss": 0.5565, + "step": 10557 + }, + { + "epoch": 3.9, + "learning_rate": 2.6117255819682225e-06, + "loss": 0.4634, + "step": 10558 + }, + { + "epoch": 3.9, + "learning_rate": 2.6116886316048775e-06, + "loss": 0.5633, + "step": 10559 + }, + { + "epoch": 3.9, + "learning_rate": 2.611651681241532e-06, + "loss": 0.5311, + "step": 10560 + }, + { + "epoch": 3.9, + "learning_rate": 2.6116147308781872e-06, + "loss": 0.6093, + "step": 10561 + }, + { + "epoch": 3.9, + "learning_rate": 2.611577780514842e-06, + "loss": 0.6017, + "step": 10562 + }, + { + "epoch": 3.9, + "learning_rate": 2.6115408301514965e-06, + "loss": 0.5153, + "step": 10563 + }, + { + "epoch": 3.9, + "learning_rate": 2.611503879788151e-06, + "loss": 0.5308, + "step": 10564 + }, + { + "epoch": 3.9, + "learning_rate": 2.6114669294248058e-06, + "loss": 0.6102, + "step": 10565 + }, + { + "epoch": 3.9, + "learning_rate": 2.611429979061461e-06, + "loss": 0.5361, + "step": 10566 + }, + { + "epoch": 3.9, + "learning_rate": 2.6113930286981155e-06, + "loss": 0.655, + "step": 10567 + }, + { + "epoch": 3.9, + "learning_rate": 2.6113560783347705e-06, + "loss": 0.5024, + "step": 10568 + }, + { + "epoch": 3.9, + "learning_rate": 2.611319127971425e-06, + "loss": 0.4887, + "step": 10569 + }, + { + "epoch": 3.9, + "learning_rate": 2.61128217760808e-06, + "loss": 0.5142, + "step": 10570 + }, + { + "epoch": 3.9, + "learning_rate": 2.6112452272447344e-06, + "loss": 0.642, + "step": 10571 + }, + { + "epoch": 3.9, + "learning_rate": 2.6112082768813895e-06, + "loss": 0.4878, + "step": 10572 + }, + { + "epoch": 3.9, + "learning_rate": 2.611171326518044e-06, + "loss": 0.5229, + "step": 10573 + }, + { + "epoch": 3.9, + "learning_rate": 2.611134376154699e-06, + "loss": 0.5702, + "step": 10574 + }, + { + "epoch": 3.91, + "learning_rate": 2.611097425791354e-06, + "loss": 0.5535, + "step": 10575 + }, + { + "epoch": 3.91, + "learning_rate": 2.6110604754280085e-06, + "loss": 0.4817, + "step": 10576 + }, + { + "epoch": 3.91, + "learning_rate": 2.611023525064663e-06, + "loss": 0.6219, + "step": 10577 + }, + { + "epoch": 3.91, + "learning_rate": 2.6109865747013177e-06, + "loss": 0.587, + "step": 10578 + }, + { + "epoch": 3.91, + "learning_rate": 2.610949624337973e-06, + "loss": 0.5019, + "step": 10579 + }, + { + "epoch": 3.91, + "learning_rate": 2.6109126739746274e-06, + "loss": 0.4785, + "step": 10580 + }, + { + "epoch": 3.91, + "learning_rate": 2.6108757236112825e-06, + "loss": 0.5583, + "step": 10581 + }, + { + "epoch": 3.91, + "learning_rate": 2.610838773247937e-06, + "loss": 0.6496, + "step": 10582 + }, + { + "epoch": 3.91, + "learning_rate": 2.6108018228845918e-06, + "loss": 0.5187, + "step": 10583 + }, + { + "epoch": 3.91, + "learning_rate": 2.6107648725212464e-06, + "loss": 0.6107, + "step": 10584 + }, + { + "epoch": 3.91, + "learning_rate": 2.610727922157901e-06, + "loss": 0.5093, + "step": 10585 + }, + { + "epoch": 3.91, + "learning_rate": 2.610690971794556e-06, + "loss": 0.5001, + "step": 10586 + }, + { + "epoch": 3.91, + "learning_rate": 2.6106540214312107e-06, + "loss": 0.5741, + "step": 10587 + }, + { + "epoch": 3.91, + "learning_rate": 2.610617071067866e-06, + "loss": 0.6343, + "step": 10588 + }, + { + "epoch": 3.91, + "learning_rate": 2.6105801207045204e-06, + "loss": 0.7148, + "step": 10589 + }, + { + "epoch": 3.91, + "learning_rate": 2.610543170341175e-06, + "loss": 0.5763, + "step": 10590 + }, + { + "epoch": 3.91, + "learning_rate": 2.6105062199778297e-06, + "loss": 0.604, + "step": 10591 + }, + { + "epoch": 3.91, + "learning_rate": 2.6104692696144848e-06, + "loss": 0.5514, + "step": 10592 + }, + { + "epoch": 3.91, + "learning_rate": 2.6104323192511394e-06, + "loss": 0.5092, + "step": 10593 + }, + { + "epoch": 3.91, + "learning_rate": 2.6103953688877945e-06, + "loss": 0.5758, + "step": 10594 + }, + { + "epoch": 3.91, + "learning_rate": 2.610358418524449e-06, + "loss": 0.5009, + "step": 10595 + }, + { + "epoch": 3.91, + "learning_rate": 2.6103214681611033e-06, + "loss": 0.5908, + "step": 10596 + }, + { + "epoch": 3.91, + "learning_rate": 2.6102845177977584e-06, + "loss": 0.5749, + "step": 10597 + }, + { + "epoch": 3.91, + "learning_rate": 2.610247567434413e-06, + "loss": 0.5817, + "step": 10598 + }, + { + "epoch": 3.91, + "learning_rate": 2.610210617071068e-06, + "loss": 0.5681, + "step": 10599 + }, + { + "epoch": 3.91, + "learning_rate": 2.6101736667077227e-06, + "loss": 0.5974, + "step": 10600 + }, + { + "epoch": 3.91, + "learning_rate": 2.6101367163443778e-06, + "loss": 0.5256, + "step": 10601 + }, + { + "epoch": 3.92, + "learning_rate": 2.610099765981032e-06, + "loss": 0.4935, + "step": 10602 + }, + { + "epoch": 3.92, + "learning_rate": 2.610062815617687e-06, + "loss": 0.5394, + "step": 10603 + }, + { + "epoch": 3.92, + "learning_rate": 2.6100258652543417e-06, + "loss": 0.4926, + "step": 10604 + }, + { + "epoch": 3.92, + "learning_rate": 2.6099889148909963e-06, + "loss": 0.5295, + "step": 10605 + }, + { + "epoch": 3.92, + "learning_rate": 2.6099519645276514e-06, + "loss": 0.6148, + "step": 10606 + }, + { + "epoch": 3.92, + "learning_rate": 2.609915014164306e-06, + "loss": 0.4078, + "step": 10607 + }, + { + "epoch": 3.92, + "learning_rate": 2.6098780638009606e-06, + "loss": 0.4302, + "step": 10608 + }, + { + "epoch": 3.92, + "learning_rate": 2.6098411134376153e-06, + "loss": 0.6249, + "step": 10609 + }, + { + "epoch": 3.92, + "learning_rate": 2.6098041630742703e-06, + "loss": 0.5852, + "step": 10610 + }, + { + "epoch": 3.92, + "learning_rate": 2.609767212710925e-06, + "loss": 0.5877, + "step": 10611 + }, + { + "epoch": 3.92, + "learning_rate": 2.60973026234758e-06, + "loss": 0.5461, + "step": 10612 + }, + { + "epoch": 3.92, + "learning_rate": 2.6096933119842347e-06, + "loss": 0.697, + "step": 10613 + }, + { + "epoch": 3.92, + "learning_rate": 2.6096563616208893e-06, + "loss": 0.637, + "step": 10614 + }, + { + "epoch": 3.92, + "learning_rate": 2.609619411257544e-06, + "loss": 0.4974, + "step": 10615 + }, + { + "epoch": 3.92, + "learning_rate": 2.6095824608941986e-06, + "loss": 0.5464, + "step": 10616 + }, + { + "epoch": 3.92, + "learning_rate": 2.6095455105308536e-06, + "loss": 0.546, + "step": 10617 + }, + { + "epoch": 3.92, + "learning_rate": 2.6095085601675083e-06, + "loss": 0.4981, + "step": 10618 + }, + { + "epoch": 3.92, + "learning_rate": 2.6094716098041633e-06, + "loss": 0.5735, + "step": 10619 + }, + { + "epoch": 3.92, + "learning_rate": 2.609434659440818e-06, + "loss": 0.4888, + "step": 10620 + }, + { + "epoch": 3.92, + "learning_rate": 2.6093977090774726e-06, + "loss": 0.4757, + "step": 10621 + }, + { + "epoch": 3.92, + "learning_rate": 2.6093607587141272e-06, + "loss": 0.5987, + "step": 10622 + }, + { + "epoch": 3.92, + "learning_rate": 2.6093238083507823e-06, + "loss": 0.5601, + "step": 10623 + }, + { + "epoch": 3.92, + "learning_rate": 2.609286857987437e-06, + "loss": 0.6398, + "step": 10624 + }, + { + "epoch": 3.92, + "learning_rate": 2.6092499076240916e-06, + "loss": 0.4971, + "step": 10625 + }, + { + "epoch": 3.92, + "learning_rate": 2.6092129572607466e-06, + "loss": 0.5869, + "step": 10626 + }, + { + "epoch": 3.92, + "learning_rate": 2.6091760068974013e-06, + "loss": 0.5284, + "step": 10627 + }, + { + "epoch": 3.92, + "learning_rate": 2.609139056534056e-06, + "loss": 0.5644, + "step": 10628 + }, + { + "epoch": 3.93, + "learning_rate": 2.6091021061707105e-06, + "loss": 0.6064, + "step": 10629 + }, + { + "epoch": 3.93, + "learning_rate": 2.6090651558073656e-06, + "loss": 0.5839, + "step": 10630 + }, + { + "epoch": 3.93, + "learning_rate": 2.6090282054440202e-06, + "loss": 0.6792, + "step": 10631 + }, + { + "epoch": 3.93, + "learning_rate": 2.6089912550806753e-06, + "loss": 0.5689, + "step": 10632 + }, + { + "epoch": 3.93, + "learning_rate": 2.60895430471733e-06, + "loss": 0.5144, + "step": 10633 + }, + { + "epoch": 3.93, + "learning_rate": 2.6089173543539846e-06, + "loss": 0.5666, + "step": 10634 + }, + { + "epoch": 3.93, + "learning_rate": 2.608880403990639e-06, + "loss": 0.5923, + "step": 10635 + }, + { + "epoch": 3.93, + "learning_rate": 2.608843453627294e-06, + "loss": 0.5184, + "step": 10636 + }, + { + "epoch": 3.93, + "learning_rate": 2.608806503263949e-06, + "loss": 0.6103, + "step": 10637 + }, + { + "epoch": 3.93, + "learning_rate": 2.6087695529006035e-06, + "loss": 0.4917, + "step": 10638 + }, + { + "epoch": 3.93, + "learning_rate": 2.6087326025372586e-06, + "loss": 0.4995, + "step": 10639 + }, + { + "epoch": 3.93, + "learning_rate": 2.6086956521739132e-06, + "loss": 0.5016, + "step": 10640 + }, + { + "epoch": 3.93, + "learning_rate": 2.608658701810568e-06, + "loss": 0.5672, + "step": 10641 + }, + { + "epoch": 3.93, + "learning_rate": 2.6086217514472225e-06, + "loss": 0.5576, + "step": 10642 + }, + { + "epoch": 3.93, + "learning_rate": 2.6085848010838776e-06, + "loss": 0.5123, + "step": 10643 + }, + { + "epoch": 3.93, + "learning_rate": 2.608547850720532e-06, + "loss": 0.541, + "step": 10644 + }, + { + "epoch": 3.93, + "learning_rate": 2.608510900357187e-06, + "loss": 0.5013, + "step": 10645 + }, + { + "epoch": 3.93, + "learning_rate": 2.608473949993842e-06, + "loss": 0.4861, + "step": 10646 + }, + { + "epoch": 3.93, + "learning_rate": 2.608436999630496e-06, + "loss": 0.5381, + "step": 10647 + }, + { + "epoch": 3.93, + "learning_rate": 2.608400049267151e-06, + "loss": 0.5504, + "step": 10648 + }, + { + "epoch": 3.93, + "learning_rate": 2.608363098903806e-06, + "loss": 0.5868, + "step": 10649 + }, + { + "epoch": 3.93, + "learning_rate": 2.608326148540461e-06, + "loss": 0.505, + "step": 10650 + }, + { + "epoch": 3.93, + "learning_rate": 2.6082891981771155e-06, + "loss": 0.55, + "step": 10651 + }, + { + "epoch": 3.93, + "learning_rate": 2.6082522478137706e-06, + "loss": 0.4449, + "step": 10652 + }, + { + "epoch": 3.93, + "learning_rate": 2.6082152974504248e-06, + "loss": 0.5197, + "step": 10653 + }, + { + "epoch": 3.93, + "learning_rate": 2.60817834708708e-06, + "loss": 0.4901, + "step": 10654 + }, + { + "epoch": 3.93, + "learning_rate": 2.6081413967237345e-06, + "loss": 0.4937, + "step": 10655 + }, + { + "epoch": 3.94, + "learning_rate": 2.608104446360389e-06, + "loss": 0.6087, + "step": 10656 + }, + { + "epoch": 3.94, + "learning_rate": 2.608067495997044e-06, + "loss": 0.4936, + "step": 10657 + }, + { + "epoch": 3.94, + "learning_rate": 2.608030545633699e-06, + "loss": 0.5951, + "step": 10658 + }, + { + "epoch": 3.94, + "learning_rate": 2.6079935952703534e-06, + "loss": 0.4404, + "step": 10659 + }, + { + "epoch": 3.94, + "learning_rate": 2.607956644907008e-06, + "loss": 0.6012, + "step": 10660 + }, + { + "epoch": 3.94, + "learning_rate": 2.607919694543663e-06, + "loss": 0.4732, + "step": 10661 + }, + { + "epoch": 3.94, + "learning_rate": 2.6078827441803178e-06, + "loss": 0.6009, + "step": 10662 + }, + { + "epoch": 3.94, + "learning_rate": 2.607845793816973e-06, + "loss": 0.5858, + "step": 10663 + }, + { + "epoch": 3.94, + "learning_rate": 2.6078088434536275e-06, + "loss": 0.4929, + "step": 10664 + }, + { + "epoch": 3.94, + "learning_rate": 2.607771893090282e-06, + "loss": 0.5158, + "step": 10665 + }, + { + "epoch": 3.94, + "learning_rate": 2.6077349427269368e-06, + "loss": 0.5066, + "step": 10666 + }, + { + "epoch": 3.94, + "learning_rate": 2.6076979923635914e-06, + "loss": 0.6079, + "step": 10667 + }, + { + "epoch": 3.94, + "learning_rate": 2.6076610420002464e-06, + "loss": 0.6272, + "step": 10668 + }, + { + "epoch": 3.94, + "learning_rate": 2.607624091636901e-06, + "loss": 0.6389, + "step": 10669 + }, + { + "epoch": 3.94, + "learning_rate": 2.607587141273556e-06, + "loss": 0.4806, + "step": 10670 + }, + { + "epoch": 3.94, + "learning_rate": 2.6075501909102108e-06, + "loss": 0.4673, + "step": 10671 + }, + { + "epoch": 3.94, + "learning_rate": 2.6075132405468654e-06, + "loss": 0.4878, + "step": 10672 + }, + { + "epoch": 3.94, + "learning_rate": 2.60747629018352e-06, + "loss": 0.5591, + "step": 10673 + }, + { + "epoch": 3.94, + "learning_rate": 2.607439339820175e-06, + "loss": 0.5098, + "step": 10674 + }, + { + "epoch": 3.94, + "learning_rate": 2.6074023894568297e-06, + "loss": 0.4271, + "step": 10675 + }, + { + "epoch": 3.94, + "learning_rate": 2.6073654390934844e-06, + "loss": 0.5159, + "step": 10676 + }, + { + "epoch": 3.94, + "learning_rate": 2.6073284887301394e-06, + "loss": 0.5136, + "step": 10677 + }, + { + "epoch": 3.94, + "learning_rate": 2.607291538366794e-06, + "loss": 0.4492, + "step": 10678 + }, + { + "epoch": 3.94, + "learning_rate": 2.6072545880034487e-06, + "loss": 0.6286, + "step": 10679 + }, + { + "epoch": 3.94, + "learning_rate": 2.6072176376401034e-06, + "loss": 0.5149, + "step": 10680 + }, + { + "epoch": 3.94, + "learning_rate": 2.6071806872767584e-06, + "loss": 0.5283, + "step": 10681 + }, + { + "epoch": 3.94, + "learning_rate": 2.607143736913413e-06, + "loss": 0.4808, + "step": 10682 + }, + { + "epoch": 3.94, + "learning_rate": 2.607106786550068e-06, + "loss": 0.4212, + "step": 10683 + }, + { + "epoch": 3.95, + "learning_rate": 2.6070698361867227e-06, + "loss": 0.6719, + "step": 10684 + }, + { + "epoch": 3.95, + "learning_rate": 2.6070328858233774e-06, + "loss": 0.6311, + "step": 10685 + }, + { + "epoch": 3.95, + "learning_rate": 2.606995935460032e-06, + "loss": 0.595, + "step": 10686 + }, + { + "epoch": 3.95, + "learning_rate": 2.6069589850966867e-06, + "loss": 0.5775, + "step": 10687 + }, + { + "epoch": 3.95, + "learning_rate": 2.6069220347333417e-06, + "loss": 0.5501, + "step": 10688 + }, + { + "epoch": 3.95, + "learning_rate": 2.6068850843699964e-06, + "loss": 0.4771, + "step": 10689 + }, + { + "epoch": 3.95, + "learning_rate": 2.6068481340066514e-06, + "loss": 0.583, + "step": 10690 + }, + { + "epoch": 3.95, + "learning_rate": 2.606811183643306e-06, + "loss": 0.5973, + "step": 10691 + }, + { + "epoch": 3.95, + "learning_rate": 2.6067742332799607e-06, + "loss": 0.5428, + "step": 10692 + }, + { + "epoch": 3.95, + "learning_rate": 2.6067372829166153e-06, + "loss": 0.575, + "step": 10693 + }, + { + "epoch": 3.95, + "learning_rate": 2.6067003325532704e-06, + "loss": 0.4754, + "step": 10694 + }, + { + "epoch": 3.95, + "learning_rate": 2.606663382189925e-06, + "loss": 0.6769, + "step": 10695 + }, + { + "epoch": 3.95, + "learning_rate": 2.6066264318265797e-06, + "loss": 0.5887, + "step": 10696 + }, + { + "epoch": 3.95, + "learning_rate": 2.6065894814632343e-06, + "loss": 0.4856, + "step": 10697 + }, + { + "epoch": 3.95, + "learning_rate": 2.606552531099889e-06, + "loss": 0.5101, + "step": 10698 + }, + { + "epoch": 3.95, + "learning_rate": 2.606515580736544e-06, + "loss": 0.4137, + "step": 10699 + }, + { + "epoch": 3.95, + "learning_rate": 2.6064786303731986e-06, + "loss": 0.548, + "step": 10700 + }, + { + "epoch": 3.95, + "learning_rate": 2.6064416800098537e-06, + "loss": 0.608, + "step": 10701 + }, + { + "epoch": 3.95, + "learning_rate": 2.6064047296465083e-06, + "loss": 0.5905, + "step": 10702 + }, + { + "epoch": 3.95, + "learning_rate": 2.606367779283163e-06, + "loss": 0.4803, + "step": 10703 + }, + { + "epoch": 3.95, + "learning_rate": 2.6063308289198176e-06, + "loss": 0.5197, + "step": 10704 + }, + { + "epoch": 3.95, + "learning_rate": 2.6062938785564722e-06, + "loss": 0.5897, + "step": 10705 + }, + { + "epoch": 3.95, + "learning_rate": 2.6062569281931273e-06, + "loss": 0.6376, + "step": 10706 + }, + { + "epoch": 3.95, + "learning_rate": 2.606219977829782e-06, + "loss": 0.6161, + "step": 10707 + }, + { + "epoch": 3.95, + "learning_rate": 2.606183027466437e-06, + "loss": 0.6076, + "step": 10708 + }, + { + "epoch": 3.95, + "learning_rate": 2.6061460771030916e-06, + "loss": 0.533, + "step": 10709 + }, + { + "epoch": 3.95, + "learning_rate": 2.6061091267397463e-06, + "loss": 0.6007, + "step": 10710 + }, + { + "epoch": 3.96, + "learning_rate": 2.606072176376401e-06, + "loss": 0.5088, + "step": 10711 + }, + { + "epoch": 3.96, + "learning_rate": 2.606035226013056e-06, + "loss": 0.5638, + "step": 10712 + }, + { + "epoch": 3.96, + "learning_rate": 2.6059982756497106e-06, + "loss": 0.4697, + "step": 10713 + }, + { + "epoch": 3.96, + "learning_rate": 2.6059613252863657e-06, + "loss": 0.462, + "step": 10714 + }, + { + "epoch": 3.96, + "learning_rate": 2.6059243749230203e-06, + "loss": 0.6228, + "step": 10715 + }, + { + "epoch": 3.96, + "learning_rate": 2.605887424559675e-06, + "loss": 0.4949, + "step": 10716 + }, + { + "epoch": 3.96, + "learning_rate": 2.6058504741963296e-06, + "loss": 0.5313, + "step": 10717 + }, + { + "epoch": 3.96, + "learning_rate": 2.605813523832984e-06, + "loss": 0.5651, + "step": 10718 + }, + { + "epoch": 3.96, + "learning_rate": 2.6057765734696393e-06, + "loss": 0.4468, + "step": 10719 + }, + { + "epoch": 3.96, + "learning_rate": 2.605739623106294e-06, + "loss": 0.5935, + "step": 10720 + }, + { + "epoch": 3.96, + "learning_rate": 2.605702672742949e-06, + "loss": 0.559, + "step": 10721 + }, + { + "epoch": 3.96, + "learning_rate": 2.6056657223796036e-06, + "loss": 0.6174, + "step": 10722 + }, + { + "epoch": 3.96, + "learning_rate": 2.6056287720162582e-06, + "loss": 0.5621, + "step": 10723 + }, + { + "epoch": 3.96, + "learning_rate": 2.605591821652913e-06, + "loss": 0.646, + "step": 10724 + }, + { + "epoch": 3.96, + "learning_rate": 2.6055548712895675e-06, + "loss": 0.6452, + "step": 10725 + }, + { + "epoch": 3.96, + "learning_rate": 2.6055179209262226e-06, + "loss": 0.6016, + "step": 10726 + }, + { + "epoch": 3.96, + "learning_rate": 2.605480970562877e-06, + "loss": 0.5337, + "step": 10727 + }, + { + "epoch": 3.96, + "learning_rate": 2.6054440201995323e-06, + "loss": 0.5744, + "step": 10728 + }, + { + "epoch": 3.96, + "learning_rate": 2.605407069836187e-06, + "loss": 0.5947, + "step": 10729 + }, + { + "epoch": 3.96, + "learning_rate": 2.6053701194728415e-06, + "loss": 0.5359, + "step": 10730 + }, + { + "epoch": 3.96, + "learning_rate": 2.605333169109496e-06, + "loss": 0.4467, + "step": 10731 + }, + { + "epoch": 3.96, + "learning_rate": 2.6052962187461512e-06, + "loss": 0.6273, + "step": 10732 + }, + { + "epoch": 3.96, + "learning_rate": 2.605259268382806e-06, + "loss": 0.5177, + "step": 10733 + }, + { + "epoch": 3.96, + "learning_rate": 2.6052223180194605e-06, + "loss": 0.5224, + "step": 10734 + }, + { + "epoch": 3.96, + "learning_rate": 2.6051853676561156e-06, + "loss": 0.5023, + "step": 10735 + }, + { + "epoch": 3.96, + "learning_rate": 2.6051484172927698e-06, + "loss": 0.5246, + "step": 10736 + }, + { + "epoch": 3.96, + "learning_rate": 2.605111466929425e-06, + "loss": 0.5298, + "step": 10737 + }, + { + "epoch": 3.97, + "learning_rate": 2.6050745165660795e-06, + "loss": 0.5192, + "step": 10738 + }, + { + "epoch": 3.97, + "learning_rate": 2.6050375662027345e-06, + "loss": 0.6894, + "step": 10739 + }, + { + "epoch": 3.97, + "learning_rate": 2.605000615839389e-06, + "loss": 0.5292, + "step": 10740 + }, + { + "epoch": 3.97, + "learning_rate": 2.6049636654760442e-06, + "loss": 0.5605, + "step": 10741 + }, + { + "epoch": 3.97, + "learning_rate": 2.6049267151126984e-06, + "loss": 0.5466, + "step": 10742 + }, + { + "epoch": 3.97, + "learning_rate": 2.6048897647493535e-06, + "loss": 0.5528, + "step": 10743 + }, + { + "epoch": 3.97, + "learning_rate": 2.604852814386008e-06, + "loss": 0.5542, + "step": 10744 + }, + { + "epoch": 3.97, + "learning_rate": 2.6048158640226628e-06, + "loss": 0.6416, + "step": 10745 + }, + { + "epoch": 3.97, + "learning_rate": 2.604778913659318e-06, + "loss": 0.4495, + "step": 10746 + }, + { + "epoch": 3.97, + "learning_rate": 2.6047419632959725e-06, + "loss": 0.5456, + "step": 10747 + }, + { + "epoch": 3.97, + "learning_rate": 2.604705012932627e-06, + "loss": 0.5601, + "step": 10748 + }, + { + "epoch": 3.97, + "learning_rate": 2.6046680625692817e-06, + "loss": 0.5858, + "step": 10749 + }, + { + "epoch": 3.97, + "learning_rate": 2.604631112205937e-06, + "loss": 0.6104, + "step": 10750 + }, + { + "epoch": 3.97, + "learning_rate": 2.6045941618425914e-06, + "loss": 0.4834, + "step": 10751 + }, + { + "epoch": 3.97, + "learning_rate": 2.6045572114792465e-06, + "loss": 0.4818, + "step": 10752 + }, + { + "epoch": 3.97, + "learning_rate": 2.604520261115901e-06, + "loss": 0.5698, + "step": 10753 + }, + { + "epoch": 3.97, + "learning_rate": 2.6044833107525558e-06, + "loss": 0.495, + "step": 10754 + }, + { + "epoch": 3.97, + "learning_rate": 2.6044463603892104e-06, + "loss": 0.5409, + "step": 10755 + }, + { + "epoch": 3.97, + "learning_rate": 2.604409410025865e-06, + "loss": 0.4464, + "step": 10756 + }, + { + "epoch": 3.97, + "learning_rate": 2.60437245966252e-06, + "loss": 0.4867, + "step": 10757 + }, + { + "epoch": 3.97, + "learning_rate": 2.6043355092991747e-06, + "loss": 0.4756, + "step": 10758 + }, + { + "epoch": 3.97, + "learning_rate": 2.60429855893583e-06, + "loss": 0.4818, + "step": 10759 + }, + { + "epoch": 3.97, + "learning_rate": 2.6042616085724844e-06, + "loss": 0.5295, + "step": 10760 + }, + { + "epoch": 3.97, + "learning_rate": 2.604224658209139e-06, + "loss": 0.5369, + "step": 10761 + }, + { + "epoch": 3.97, + "learning_rate": 2.6041877078457937e-06, + "loss": 0.5521, + "step": 10762 + }, + { + "epoch": 3.97, + "learning_rate": 2.6041507574824488e-06, + "loss": 0.5342, + "step": 10763 + }, + { + "epoch": 3.97, + "learning_rate": 2.6041138071191034e-06, + "loss": 0.5807, + "step": 10764 + }, + { + "epoch": 3.98, + "learning_rate": 2.604076856755758e-06, + "loss": 0.4564, + "step": 10765 + }, + { + "epoch": 3.98, + "learning_rate": 2.604039906392413e-06, + "loss": 0.519, + "step": 10766 + }, + { + "epoch": 3.98, + "learning_rate": 2.6040029560290677e-06, + "loss": 0.6013, + "step": 10767 + }, + { + "epoch": 3.98, + "learning_rate": 2.6039660056657224e-06, + "loss": 0.4881, + "step": 10768 + }, + { + "epoch": 3.98, + "learning_rate": 2.603929055302377e-06, + "loss": 0.4861, + "step": 10769 + }, + { + "epoch": 3.98, + "learning_rate": 2.603892104939032e-06, + "loss": 0.4779, + "step": 10770 + }, + { + "epoch": 3.98, + "learning_rate": 2.6038551545756867e-06, + "loss": 0.6267, + "step": 10771 + }, + { + "epoch": 3.98, + "learning_rate": 2.6038182042123418e-06, + "loss": 0.5345, + "step": 10772 + }, + { + "epoch": 3.98, + "learning_rate": 2.6037812538489964e-06, + "loss": 0.4615, + "step": 10773 + }, + { + "epoch": 3.98, + "learning_rate": 2.603744303485651e-06, + "loss": 0.5497, + "step": 10774 + }, + { + "epoch": 3.98, + "learning_rate": 2.6037073531223057e-06, + "loss": 0.5024, + "step": 10775 + }, + { + "epoch": 3.98, + "learning_rate": 2.6036704027589603e-06, + "loss": 0.5212, + "step": 10776 + }, + { + "epoch": 3.98, + "learning_rate": 2.6036334523956154e-06, + "loss": 0.4574, + "step": 10777 + }, + { + "epoch": 3.98, + "learning_rate": 2.60359650203227e-06, + "loss": 0.5401, + "step": 10778 + }, + { + "epoch": 3.98, + "learning_rate": 2.603559551668925e-06, + "loss": 0.4657, + "step": 10779 + }, + { + "epoch": 3.98, + "learning_rate": 2.6035226013055797e-06, + "loss": 0.6005, + "step": 10780 + }, + { + "epoch": 3.98, + "learning_rate": 2.6034856509422343e-06, + "loss": 0.5825, + "step": 10781 + }, + { + "epoch": 3.98, + "learning_rate": 2.603448700578889e-06, + "loss": 0.4535, + "step": 10782 + }, + { + "epoch": 3.98, + "learning_rate": 2.603411750215544e-06, + "loss": 0.4627, + "step": 10783 + }, + { + "epoch": 3.98, + "learning_rate": 2.6033747998521987e-06, + "loss": 0.5868, + "step": 10784 + }, + { + "epoch": 3.98, + "learning_rate": 2.6033378494888533e-06, + "loss": 0.4825, + "step": 10785 + }, + { + "epoch": 3.98, + "learning_rate": 2.6033008991255084e-06, + "loss": 0.4908, + "step": 10786 + }, + { + "epoch": 3.98, + "learning_rate": 2.6032639487621626e-06, + "loss": 0.4976, + "step": 10787 + }, + { + "epoch": 3.98, + "learning_rate": 2.6032269983988176e-06, + "loss": 0.4994, + "step": 10788 + }, + { + "epoch": 3.98, + "learning_rate": 2.6031900480354723e-06, + "loss": 0.5441, + "step": 10789 + }, + { + "epoch": 3.98, + "learning_rate": 2.6031530976721273e-06, + "loss": 0.4383, + "step": 10790 + }, + { + "epoch": 3.98, + "learning_rate": 2.603116147308782e-06, + "loss": 0.5614, + "step": 10791 + }, + { + "epoch": 3.99, + "learning_rate": 2.603079196945437e-06, + "loss": 0.5514, + "step": 10792 + }, + { + "epoch": 3.99, + "learning_rate": 2.6030422465820912e-06, + "loss": 0.4635, + "step": 10793 + }, + { + "epoch": 3.99, + "learning_rate": 2.603005296218746e-06, + "loss": 0.4604, + "step": 10794 + }, + { + "epoch": 3.99, + "learning_rate": 2.602968345855401e-06, + "loss": 0.6591, + "step": 10795 + }, + { + "epoch": 3.99, + "learning_rate": 2.6029313954920556e-06, + "loss": 0.5546, + "step": 10796 + }, + { + "epoch": 3.99, + "learning_rate": 2.6028944451287106e-06, + "loss": 0.5608, + "step": 10797 + }, + { + "epoch": 3.99, + "learning_rate": 2.6028574947653653e-06, + "loss": 0.6119, + "step": 10798 + }, + { + "epoch": 3.99, + "learning_rate": 2.60282054440202e-06, + "loss": 0.5797, + "step": 10799 + }, + { + "epoch": 3.99, + "learning_rate": 2.6027835940386745e-06, + "loss": 0.513, + "step": 10800 + }, + { + "epoch": 3.99, + "learning_rate": 2.6027466436753296e-06, + "loss": 0.562, + "step": 10801 + }, + { + "epoch": 3.99, + "learning_rate": 2.6027096933119842e-06, + "loss": 0.4966, + "step": 10802 + }, + { + "epoch": 3.99, + "learning_rate": 2.6026727429486393e-06, + "loss": 0.5383, + "step": 10803 + }, + { + "epoch": 3.99, + "learning_rate": 2.602635792585294e-06, + "loss": 0.5888, + "step": 10804 + }, + { + "epoch": 3.99, + "learning_rate": 2.6025988422219486e-06, + "loss": 0.4739, + "step": 10805 + }, + { + "epoch": 3.99, + "learning_rate": 2.6025618918586032e-06, + "loss": 0.4474, + "step": 10806 + }, + { + "epoch": 3.99, + "learning_rate": 2.602524941495258e-06, + "loss": 0.4591, + "step": 10807 + }, + { + "epoch": 3.99, + "learning_rate": 2.602487991131913e-06, + "loss": 0.4938, + "step": 10808 + }, + { + "epoch": 3.99, + "learning_rate": 2.6024510407685675e-06, + "loss": 0.5424, + "step": 10809 + }, + { + "epoch": 3.99, + "learning_rate": 2.6024140904052226e-06, + "loss": 0.592, + "step": 10810 + }, + { + "epoch": 3.99, + "learning_rate": 2.6023771400418772e-06, + "loss": 0.4813, + "step": 10811 + }, + { + "epoch": 3.99, + "learning_rate": 2.602340189678532e-06, + "loss": 0.5961, + "step": 10812 + }, + { + "epoch": 3.99, + "learning_rate": 2.6023032393151865e-06, + "loss": 0.6089, + "step": 10813 + }, + { + "epoch": 3.99, + "learning_rate": 2.602266288951841e-06, + "loss": 0.5478, + "step": 10814 + }, + { + "epoch": 3.99, + "learning_rate": 2.6022293385884962e-06, + "loss": 0.5401, + "step": 10815 + }, + { + "epoch": 3.99, + "learning_rate": 2.602192388225151e-06, + "loss": 0.4764, + "step": 10816 + }, + { + "epoch": 3.99, + "learning_rate": 2.602155437861806e-06, + "loss": 0.5666, + "step": 10817 + }, + { + "epoch": 3.99, + "learning_rate": 2.6021184874984605e-06, + "loss": 0.5717, + "step": 10818 + }, + { + "epoch": 4.0, + "learning_rate": 2.602081537135115e-06, + "loss": 0.4992, + "step": 10819 + }, + { + "epoch": 4.0, + "learning_rate": 2.60204458677177e-06, + "loss": 0.522, + "step": 10820 + }, + { + "epoch": 4.0, + "learning_rate": 2.602007636408425e-06, + "loss": 0.5644, + "step": 10821 + }, + { + "epoch": 4.0, + "learning_rate": 2.6019706860450795e-06, + "loss": 0.4025, + "step": 10822 + }, + { + "epoch": 4.0, + "learning_rate": 2.6019337356817346e-06, + "loss": 0.4421, + "step": 10823 + }, + { + "epoch": 4.0, + "learning_rate": 2.6018967853183892e-06, + "loss": 0.5141, + "step": 10824 + }, + { + "epoch": 4.0, + "learning_rate": 2.601859834955044e-06, + "loss": 0.5258, + "step": 10825 + }, + { + "epoch": 4.0, + "learning_rate": 2.6018228845916985e-06, + "loss": 0.5101, + "step": 10826 + }, + { + "epoch": 4.0, + "learning_rate": 2.601785934228353e-06, + "loss": 0.6086, + "step": 10827 + }, + { + "epoch": 4.0, + "learning_rate": 2.601748983865008e-06, + "loss": 0.5886, + "step": 10828 + }, + { + "epoch": 4.0, + "learning_rate": 2.601712033501663e-06, + "loss": 0.5077, + "step": 10829 + }, + { + "epoch": 4.0, + "learning_rate": 2.601675083138318e-06, + "loss": 0.6703, + "step": 10830 + }, + { + "epoch": 4.0, + "learning_rate": 2.6016381327749725e-06, + "loss": 0.451, + "step": 10831 + }, + { + "epoch": 4.0, + "learning_rate": 2.601601182411627e-06, + "loss": 0.2898, + "step": 10832 + }, + { + "epoch": 4.0, + "eval_accuracy": 0.7183358729279217, + "eval_loss": 0.5513486266136169, + "eval_runtime": 25.4611, + "eval_samples_per_second": 1701.182, + "eval_steps_per_second": 26.59, + "step": 10832 + }, + { + "epoch": 4.0, + "learning_rate": 2.6015642320482818e-06, + "loss": 0.5078, + "step": 10833 + }, + { + "epoch": 4.0, + "learning_rate": 2.6015272816849364e-06, + "loss": 0.5423, + "step": 10834 + }, + { + "epoch": 4.0, + "learning_rate": 2.6014903313215915e-06, + "loss": 0.4367, + "step": 10835 + }, + { + "epoch": 4.0, + "learning_rate": 2.601453380958246e-06, + "loss": 0.5495, + "step": 10836 + }, + { + "epoch": 4.0, + "learning_rate": 2.601416430594901e-06, + "loss": 0.5824, + "step": 10837 + }, + { + "epoch": 4.0, + "learning_rate": 2.6013794802315554e-06, + "loss": 0.6176, + "step": 10838 + }, + { + "epoch": 4.0, + "learning_rate": 2.6013425298682105e-06, + "loss": 0.5231, + "step": 10839 + }, + { + "epoch": 4.0, + "learning_rate": 2.601305579504865e-06, + "loss": 0.5111, + "step": 10840 + }, + { + "epoch": 4.0, + "learning_rate": 2.60126862914152e-06, + "loss": 0.504, + "step": 10841 + }, + { + "epoch": 4.0, + "learning_rate": 2.6012316787781748e-06, + "loss": 0.5717, + "step": 10842 + }, + { + "epoch": 4.0, + "learning_rate": 2.60119472841483e-06, + "loss": 0.5892, + "step": 10843 + }, + { + "epoch": 4.0, + "learning_rate": 2.601157778051484e-06, + "loss": 0.4733, + "step": 10844 + }, + { + "epoch": 4.0, + "learning_rate": 2.6011208276881387e-06, + "loss": 0.4803, + "step": 10845 + }, + { + "epoch": 4.01, + "learning_rate": 2.6010838773247938e-06, + "loss": 0.5492, + "step": 10846 + }, + { + "epoch": 4.01, + "learning_rate": 2.6010469269614484e-06, + "loss": 0.4877, + "step": 10847 + }, + { + "epoch": 4.01, + "learning_rate": 2.6010099765981035e-06, + "loss": 0.5775, + "step": 10848 + }, + { + "epoch": 4.01, + "learning_rate": 2.600973026234758e-06, + "loss": 0.6614, + "step": 10849 + }, + { + "epoch": 4.01, + "learning_rate": 2.6009360758714127e-06, + "loss": 0.5383, + "step": 10850 + }, + { + "epoch": 4.01, + "learning_rate": 2.6008991255080674e-06, + "loss": 0.4993, + "step": 10851 + }, + { + "epoch": 4.01, + "learning_rate": 2.6008621751447224e-06, + "loss": 0.6101, + "step": 10852 + }, + { + "epoch": 4.01, + "learning_rate": 2.600825224781377e-06, + "loss": 0.5182, + "step": 10853 + }, + { + "epoch": 4.01, + "learning_rate": 2.6007882744180317e-06, + "loss": 0.5871, + "step": 10854 + }, + { + "epoch": 4.01, + "learning_rate": 2.6007513240546868e-06, + "loss": 0.4803, + "step": 10855 + }, + { + "epoch": 4.01, + "learning_rate": 2.6007143736913414e-06, + "loss": 0.5703, + "step": 10856 + }, + { + "epoch": 4.01, + "learning_rate": 2.600677423327996e-06, + "loss": 0.4687, + "step": 10857 + }, + { + "epoch": 4.01, + "learning_rate": 2.6006404729646507e-06, + "loss": 0.5777, + "step": 10858 + }, + { + "epoch": 4.01, + "learning_rate": 2.6006035226013057e-06, + "loss": 0.5152, + "step": 10859 + }, + { + "epoch": 4.01, + "learning_rate": 2.6005665722379604e-06, + "loss": 0.505, + "step": 10860 + }, + { + "epoch": 4.01, + "learning_rate": 2.6005296218746154e-06, + "loss": 0.4888, + "step": 10861 + }, + { + "epoch": 4.01, + "learning_rate": 2.60049267151127e-06, + "loss": 0.5625, + "step": 10862 + }, + { + "epoch": 4.01, + "learning_rate": 2.6004557211479247e-06, + "loss": 0.4995, + "step": 10863 + }, + { + "epoch": 4.01, + "learning_rate": 2.6004187707845793e-06, + "loss": 0.5832, + "step": 10864 + }, + { + "epoch": 4.01, + "learning_rate": 2.600381820421234e-06, + "loss": 0.5832, + "step": 10865 + }, + { + "epoch": 4.01, + "learning_rate": 2.600344870057889e-06, + "loss": 0.4426, + "step": 10866 + }, + { + "epoch": 4.01, + "learning_rate": 2.6003079196945437e-06, + "loss": 0.5627, + "step": 10867 + }, + { + "epoch": 4.01, + "learning_rate": 2.6002709693311987e-06, + "loss": 0.4435, + "step": 10868 + }, + { + "epoch": 4.01, + "learning_rate": 2.6002340189678534e-06, + "loss": 0.5223, + "step": 10869 + }, + { + "epoch": 4.01, + "learning_rate": 2.600197068604508e-06, + "loss": 0.4711, + "step": 10870 + }, + { + "epoch": 4.01, + "learning_rate": 2.6001601182411626e-06, + "loss": 0.578, + "step": 10871 + }, + { + "epoch": 4.01, + "learning_rate": 2.6001231678778177e-06, + "loss": 0.4491, + "step": 10872 + }, + { + "epoch": 4.02, + "learning_rate": 2.6000862175144723e-06, + "loss": 0.44, + "step": 10873 + }, + { + "epoch": 4.02, + "learning_rate": 2.600049267151127e-06, + "loss": 0.5334, + "step": 10874 + }, + { + "epoch": 4.02, + "learning_rate": 2.600012316787782e-06, + "loss": 0.4933, + "step": 10875 + }, + { + "epoch": 4.02, + "learning_rate": 2.5999753664244367e-06, + "loss": 0.6113, + "step": 10876 + }, + { + "epoch": 4.02, + "learning_rate": 2.5999384160610913e-06, + "loss": 0.5166, + "step": 10877 + }, + { + "epoch": 4.02, + "learning_rate": 2.599901465697746e-06, + "loss": 0.514, + "step": 10878 + }, + { + "epoch": 4.02, + "learning_rate": 2.599864515334401e-06, + "loss": 0.4761, + "step": 10879 + }, + { + "epoch": 4.02, + "learning_rate": 2.5998275649710556e-06, + "loss": 0.5817, + "step": 10880 + }, + { + "epoch": 4.02, + "learning_rate": 2.5997906146077107e-06, + "loss": 0.5647, + "step": 10881 + }, + { + "epoch": 4.02, + "learning_rate": 2.5997536642443653e-06, + "loss": 0.4181, + "step": 10882 + }, + { + "epoch": 4.02, + "learning_rate": 2.59971671388102e-06, + "loss": 0.5021, + "step": 10883 + }, + { + "epoch": 4.02, + "learning_rate": 2.5996797635176746e-06, + "loss": 0.4966, + "step": 10884 + }, + { + "epoch": 4.02, + "learning_rate": 2.5996428131543292e-06, + "loss": 0.4939, + "step": 10885 + }, + { + "epoch": 4.02, + "learning_rate": 2.5996058627909843e-06, + "loss": 0.5741, + "step": 10886 + }, + { + "epoch": 4.02, + "learning_rate": 2.599568912427639e-06, + "loss": 0.5698, + "step": 10887 + }, + { + "epoch": 4.02, + "learning_rate": 2.5995319620642936e-06, + "loss": 0.4194, + "step": 10888 + }, + { + "epoch": 4.02, + "learning_rate": 2.599495011700948e-06, + "loss": 0.489, + "step": 10889 + }, + { + "epoch": 4.02, + "learning_rate": 2.5994580613376033e-06, + "loss": 0.566, + "step": 10890 + }, + { + "epoch": 4.02, + "learning_rate": 2.599421110974258e-06, + "loss": 0.5702, + "step": 10891 + }, + { + "epoch": 4.02, + "learning_rate": 2.599384160610913e-06, + "loss": 0.4336, + "step": 10892 + }, + { + "epoch": 4.02, + "learning_rate": 2.5993472102475676e-06, + "loss": 0.5639, + "step": 10893 + }, + { + "epoch": 4.02, + "learning_rate": 2.5993102598842222e-06, + "loss": 0.5714, + "step": 10894 + }, + { + "epoch": 4.02, + "learning_rate": 2.599273309520877e-06, + "loss": 0.5528, + "step": 10895 + }, + { + "epoch": 4.02, + "learning_rate": 2.5992363591575315e-06, + "loss": 0.6274, + "step": 10896 + }, + { + "epoch": 4.02, + "learning_rate": 2.5991994087941866e-06, + "loss": 0.4982, + "step": 10897 + }, + { + "epoch": 4.02, + "learning_rate": 2.599162458430841e-06, + "loss": 0.6429, + "step": 10898 + }, + { + "epoch": 4.02, + "learning_rate": 2.5991255080674963e-06, + "loss": 0.4752, + "step": 10899 + }, + { + "epoch": 4.03, + "learning_rate": 2.599088557704151e-06, + "loss": 0.3966, + "step": 10900 + }, + { + "epoch": 4.03, + "learning_rate": 2.5990516073408055e-06, + "loss": 0.4511, + "step": 10901 + }, + { + "epoch": 4.03, + "learning_rate": 2.59901465697746e-06, + "loss": 0.538, + "step": 10902 + }, + { + "epoch": 4.03, + "learning_rate": 2.5989777066141152e-06, + "loss": 0.5801, + "step": 10903 + }, + { + "epoch": 4.03, + "learning_rate": 2.59894075625077e-06, + "loss": 0.4237, + "step": 10904 + }, + { + "epoch": 4.03, + "learning_rate": 2.5989038058874245e-06, + "loss": 0.4202, + "step": 10905 + }, + { + "epoch": 4.03, + "learning_rate": 2.5988668555240796e-06, + "loss": 0.6582, + "step": 10906 + }, + { + "epoch": 4.03, + "learning_rate": 2.598829905160734e-06, + "loss": 0.5859, + "step": 10907 + }, + { + "epoch": 4.03, + "learning_rate": 2.598792954797389e-06, + "loss": 0.509, + "step": 10908 + }, + { + "epoch": 4.03, + "learning_rate": 2.5987560044340435e-06, + "loss": 0.3801, + "step": 10909 + }, + { + "epoch": 4.03, + "learning_rate": 2.5987190540706985e-06, + "loss": 0.5323, + "step": 10910 + }, + { + "epoch": 4.03, + "learning_rate": 2.598682103707353e-06, + "loss": 0.5453, + "step": 10911 + }, + { + "epoch": 4.03, + "learning_rate": 2.5986451533440082e-06, + "loss": 0.5868, + "step": 10912 + }, + { + "epoch": 4.03, + "learning_rate": 2.598608202980663e-06, + "loss": 0.6097, + "step": 10913 + }, + { + "epoch": 4.03, + "learning_rate": 2.5985712526173175e-06, + "loss": 0.448, + "step": 10914 + }, + { + "epoch": 4.03, + "learning_rate": 2.598534302253972e-06, + "loss": 0.5739, + "step": 10915 + }, + { + "epoch": 4.03, + "learning_rate": 2.5984973518906268e-06, + "loss": 0.5151, + "step": 10916 + }, + { + "epoch": 4.03, + "learning_rate": 2.598460401527282e-06, + "loss": 0.6179, + "step": 10917 + }, + { + "epoch": 4.03, + "learning_rate": 2.5984234511639365e-06, + "loss": 0.5827, + "step": 10918 + }, + { + "epoch": 4.03, + "learning_rate": 2.5983865008005915e-06, + "loss": 0.4984, + "step": 10919 + }, + { + "epoch": 4.03, + "learning_rate": 2.598349550437246e-06, + "loss": 0.6134, + "step": 10920 + }, + { + "epoch": 4.03, + "learning_rate": 2.598312600073901e-06, + "loss": 0.592, + "step": 10921 + }, + { + "epoch": 4.03, + "learning_rate": 2.5982756497105554e-06, + "loss": 0.5355, + "step": 10922 + }, + { + "epoch": 4.03, + "learning_rate": 2.5982386993472105e-06, + "loss": 0.6198, + "step": 10923 + }, + { + "epoch": 4.03, + "learning_rate": 2.598201748983865e-06, + "loss": 0.4994, + "step": 10924 + }, + { + "epoch": 4.03, + "learning_rate": 2.5981647986205198e-06, + "loss": 0.531, + "step": 10925 + }, + { + "epoch": 4.03, + "learning_rate": 2.598127848257175e-06, + "loss": 0.5397, + "step": 10926 + }, + { + "epoch": 4.04, + "learning_rate": 2.598090897893829e-06, + "loss": 0.4949, + "step": 10927 + }, + { + "epoch": 4.04, + "learning_rate": 2.598053947530484e-06, + "loss": 0.4756, + "step": 10928 + }, + { + "epoch": 4.04, + "learning_rate": 2.5980169971671387e-06, + "loss": 0.4797, + "step": 10929 + }, + { + "epoch": 4.04, + "learning_rate": 2.597980046803794e-06, + "loss": 0.6013, + "step": 10930 + }, + { + "epoch": 4.04, + "learning_rate": 2.5979430964404484e-06, + "loss": 0.5575, + "step": 10931 + }, + { + "epoch": 4.04, + "learning_rate": 2.5979061460771035e-06, + "loss": 0.5869, + "step": 10932 + }, + { + "epoch": 4.04, + "learning_rate": 2.5978691957137577e-06, + "loss": 0.5811, + "step": 10933 + }, + { + "epoch": 4.04, + "learning_rate": 2.5978322453504123e-06, + "loss": 0.4793, + "step": 10934 + }, + { + "epoch": 4.04, + "learning_rate": 2.5977952949870674e-06, + "loss": 0.5137, + "step": 10935 + }, + { + "epoch": 4.04, + "learning_rate": 2.597758344623722e-06, + "loss": 0.627, + "step": 10936 + }, + { + "epoch": 4.04, + "learning_rate": 2.597721394260377e-06, + "loss": 0.5602, + "step": 10937 + }, + { + "epoch": 4.04, + "learning_rate": 2.5976844438970317e-06, + "loss": 0.5207, + "step": 10938 + }, + { + "epoch": 4.04, + "learning_rate": 2.5976474935336864e-06, + "loss": 0.5185, + "step": 10939 + }, + { + "epoch": 4.04, + "learning_rate": 2.597610543170341e-06, + "loss": 0.7132, + "step": 10940 + }, + { + "epoch": 4.04, + "learning_rate": 2.597573592806996e-06, + "loss": 0.5352, + "step": 10941 + }, + { + "epoch": 4.04, + "learning_rate": 2.5975366424436507e-06, + "loss": 0.4592, + "step": 10942 + }, + { + "epoch": 4.04, + "learning_rate": 2.5974996920803058e-06, + "loss": 0.559, + "step": 10943 + }, + { + "epoch": 4.04, + "learning_rate": 2.5974627417169604e-06, + "loss": 0.4727, + "step": 10944 + }, + { + "epoch": 4.04, + "learning_rate": 2.597425791353615e-06, + "loss": 0.4839, + "step": 10945 + }, + { + "epoch": 4.04, + "learning_rate": 2.5973888409902697e-06, + "loss": 0.5773, + "step": 10946 + }, + { + "epoch": 4.04, + "learning_rate": 2.5973518906269243e-06, + "loss": 0.5901, + "step": 10947 + }, + { + "epoch": 4.04, + "learning_rate": 2.5973149402635794e-06, + "loss": 0.486, + "step": 10948 + }, + { + "epoch": 4.04, + "learning_rate": 2.597277989900234e-06, + "loss": 0.5146, + "step": 10949 + }, + { + "epoch": 4.04, + "learning_rate": 2.597241039536889e-06, + "loss": 0.4518, + "step": 10950 + }, + { + "epoch": 4.04, + "learning_rate": 2.5972040891735437e-06, + "loss": 0.5947, + "step": 10951 + }, + { + "epoch": 4.04, + "learning_rate": 2.5971671388101983e-06, + "loss": 0.5506, + "step": 10952 + }, + { + "epoch": 4.04, + "learning_rate": 2.597130188446853e-06, + "loss": 0.5183, + "step": 10953 + }, + { + "epoch": 4.05, + "learning_rate": 2.5970932380835076e-06, + "loss": 0.5429, + "step": 10954 + }, + { + "epoch": 4.05, + "learning_rate": 2.5970562877201627e-06, + "loss": 0.5394, + "step": 10955 + }, + { + "epoch": 4.05, + "learning_rate": 2.5970193373568173e-06, + "loss": 0.5294, + "step": 10956 + }, + { + "epoch": 4.05, + "learning_rate": 2.5969823869934724e-06, + "loss": 0.6092, + "step": 10957 + }, + { + "epoch": 4.05, + "learning_rate": 2.596945436630127e-06, + "loss": 0.4536, + "step": 10958 + }, + { + "epoch": 4.05, + "learning_rate": 2.5969084862667816e-06, + "loss": 0.6558, + "step": 10959 + }, + { + "epoch": 4.05, + "learning_rate": 2.5968715359034363e-06, + "loss": 0.579, + "step": 10960 + }, + { + "epoch": 4.05, + "learning_rate": 2.5968345855400913e-06, + "loss": 0.44, + "step": 10961 + }, + { + "epoch": 4.05, + "learning_rate": 2.596797635176746e-06, + "loss": 0.5189, + "step": 10962 + }, + { + "epoch": 4.05, + "learning_rate": 2.5967606848134006e-06, + "loss": 0.6338, + "step": 10963 + }, + { + "epoch": 4.05, + "learning_rate": 2.5967237344500557e-06, + "loss": 0.5516, + "step": 10964 + }, + { + "epoch": 4.05, + "learning_rate": 2.5966867840867103e-06, + "loss": 0.5414, + "step": 10965 + }, + { + "epoch": 4.05, + "learning_rate": 2.596649833723365e-06, + "loss": 0.5078, + "step": 10966 + }, + { + "epoch": 4.05, + "learning_rate": 2.5966128833600196e-06, + "loss": 0.5062, + "step": 10967 + }, + { + "epoch": 4.05, + "learning_rate": 2.5965759329966746e-06, + "loss": 0.615, + "step": 10968 + }, + { + "epoch": 4.05, + "learning_rate": 2.5965389826333293e-06, + "loss": 0.5096, + "step": 10969 + }, + { + "epoch": 4.05, + "learning_rate": 2.5965020322699843e-06, + "loss": 0.5602, + "step": 10970 + }, + { + "epoch": 4.05, + "learning_rate": 2.596465081906639e-06, + "loss": 0.4804, + "step": 10971 + }, + { + "epoch": 4.05, + "learning_rate": 2.5964281315432936e-06, + "loss": 0.507, + "step": 10972 + }, + { + "epoch": 4.05, + "learning_rate": 2.5963911811799483e-06, + "loss": 0.5426, + "step": 10973 + }, + { + "epoch": 4.05, + "learning_rate": 2.596354230816603e-06, + "loss": 0.6748, + "step": 10974 + }, + { + "epoch": 4.05, + "learning_rate": 2.596317280453258e-06, + "loss": 0.5219, + "step": 10975 + }, + { + "epoch": 4.05, + "learning_rate": 2.5962803300899126e-06, + "loss": 0.5085, + "step": 10976 + }, + { + "epoch": 4.05, + "learning_rate": 2.5962433797265676e-06, + "loss": 0.5476, + "step": 10977 + }, + { + "epoch": 4.05, + "learning_rate": 2.596206429363222e-06, + "loss": 0.524, + "step": 10978 + }, + { + "epoch": 4.05, + "learning_rate": 2.596169478999877e-06, + "loss": 0.4604, + "step": 10979 + }, + { + "epoch": 4.05, + "learning_rate": 2.5961325286365316e-06, + "loss": 0.508, + "step": 10980 + }, + { + "epoch": 4.06, + "learning_rate": 2.5960955782731866e-06, + "loss": 0.6131, + "step": 10981 + }, + { + "epoch": 4.06, + "learning_rate": 2.5960586279098413e-06, + "loss": 0.5077, + "step": 10982 + }, + { + "epoch": 4.06, + "learning_rate": 2.596021677546496e-06, + "loss": 0.548, + "step": 10983 + }, + { + "epoch": 4.06, + "learning_rate": 2.5959847271831505e-06, + "loss": 0.5729, + "step": 10984 + }, + { + "epoch": 4.06, + "learning_rate": 2.595947776819805e-06, + "loss": 0.532, + "step": 10985 + }, + { + "epoch": 4.06, + "learning_rate": 2.5959108264564602e-06, + "loss": 0.5378, + "step": 10986 + }, + { + "epoch": 4.06, + "learning_rate": 2.595873876093115e-06, + "loss": 0.546, + "step": 10987 + }, + { + "epoch": 4.06, + "learning_rate": 2.59583692572977e-06, + "loss": 0.578, + "step": 10988 + }, + { + "epoch": 4.06, + "learning_rate": 2.5957999753664246e-06, + "loss": 0.5174, + "step": 10989 + }, + { + "epoch": 4.06, + "learning_rate": 2.595763025003079e-06, + "loss": 0.5233, + "step": 10990 + }, + { + "epoch": 4.06, + "learning_rate": 2.595726074639734e-06, + "loss": 0.427, + "step": 10991 + }, + { + "epoch": 4.06, + "learning_rate": 2.595689124276389e-06, + "loss": 0.5523, + "step": 10992 + }, + { + "epoch": 4.06, + "learning_rate": 2.5956521739130435e-06, + "loss": 0.5304, + "step": 10993 + }, + { + "epoch": 4.06, + "learning_rate": 2.595615223549698e-06, + "loss": 0.5762, + "step": 10994 + }, + { + "epoch": 4.06, + "learning_rate": 2.5955782731863532e-06, + "loss": 0.4978, + "step": 10995 + }, + { + "epoch": 4.06, + "learning_rate": 2.595541322823008e-06, + "loss": 0.5427, + "step": 10996 + }, + { + "epoch": 4.06, + "learning_rate": 2.5955043724596625e-06, + "loss": 0.4956, + "step": 10997 + }, + { + "epoch": 4.06, + "learning_rate": 2.595467422096317e-06, + "loss": 0.4854, + "step": 10998 + }, + { + "epoch": 4.06, + "learning_rate": 2.595430471732972e-06, + "loss": 0.5771, + "step": 10999 + }, + { + "epoch": 4.06, + "learning_rate": 2.595393521369627e-06, + "loss": 0.4974, + "step": 11000 + }, + { + "epoch": 4.06, + "learning_rate": 2.595356571006282e-06, + "loss": 0.5771, + "step": 11001 + }, + { + "epoch": 4.06, + "learning_rate": 2.5953196206429365e-06, + "loss": 0.4441, + "step": 11002 + }, + { + "epoch": 4.06, + "learning_rate": 2.595282670279591e-06, + "loss": 0.4984, + "step": 11003 + }, + { + "epoch": 4.06, + "learning_rate": 2.595245719916246e-06, + "loss": 0.5, + "step": 11004 + }, + { + "epoch": 4.06, + "learning_rate": 2.5952087695529004e-06, + "loss": 0.4694, + "step": 11005 + }, + { + "epoch": 4.06, + "learning_rate": 2.5951718191895555e-06, + "loss": 0.5245, + "step": 11006 + }, + { + "epoch": 4.06, + "learning_rate": 2.59513486882621e-06, + "loss": 0.513, + "step": 11007 + }, + { + "epoch": 4.06, + "learning_rate": 2.595097918462865e-06, + "loss": 0.5265, + "step": 11008 + }, + { + "epoch": 4.07, + "learning_rate": 2.59506096809952e-06, + "loss": 0.5669, + "step": 11009 + }, + { + "epoch": 4.07, + "learning_rate": 2.5950240177361745e-06, + "loss": 0.4465, + "step": 11010 + }, + { + "epoch": 4.07, + "learning_rate": 2.594987067372829e-06, + "loss": 0.5191, + "step": 11011 + }, + { + "epoch": 4.07, + "learning_rate": 2.594950117009484e-06, + "loss": 0.527, + "step": 11012 + }, + { + "epoch": 4.07, + "learning_rate": 2.594913166646139e-06, + "loss": 0.54, + "step": 11013 + }, + { + "epoch": 4.07, + "learning_rate": 2.5948762162827934e-06, + "loss": 0.5268, + "step": 11014 + }, + { + "epoch": 4.07, + "learning_rate": 2.5948392659194485e-06, + "loss": 0.4943, + "step": 11015 + }, + { + "epoch": 4.07, + "learning_rate": 2.594802315556103e-06, + "loss": 0.4548, + "step": 11016 + }, + { + "epoch": 4.07, + "learning_rate": 2.5947653651927578e-06, + "loss": 0.6392, + "step": 11017 + }, + { + "epoch": 4.07, + "learning_rate": 2.5947284148294124e-06, + "loss": 0.4947, + "step": 11018 + }, + { + "epoch": 4.07, + "learning_rate": 2.5946914644660675e-06, + "loss": 0.568, + "step": 11019 + }, + { + "epoch": 4.07, + "learning_rate": 2.594654514102722e-06, + "loss": 0.5287, + "step": 11020 + }, + { + "epoch": 4.07, + "learning_rate": 2.594617563739377e-06, + "loss": 0.4876, + "step": 11021 + }, + { + "epoch": 4.07, + "learning_rate": 2.594580613376032e-06, + "loss": 0.4798, + "step": 11022 + }, + { + "epoch": 4.07, + "learning_rate": 2.594543663012686e-06, + "loss": 0.4625, + "step": 11023 + }, + { + "epoch": 4.07, + "learning_rate": 2.594506712649341e-06, + "loss": 0.5372, + "step": 11024 + }, + { + "epoch": 4.07, + "learning_rate": 2.5944697622859957e-06, + "loss": 0.5989, + "step": 11025 + }, + { + "epoch": 4.07, + "learning_rate": 2.5944328119226508e-06, + "loss": 0.565, + "step": 11026 + }, + { + "epoch": 4.07, + "learning_rate": 2.5943958615593054e-06, + "loss": 0.3936, + "step": 11027 + }, + { + "epoch": 4.07, + "learning_rate": 2.5943589111959605e-06, + "loss": 0.4254, + "step": 11028 + }, + { + "epoch": 4.07, + "learning_rate": 2.5943219608326147e-06, + "loss": 0.6242, + "step": 11029 + }, + { + "epoch": 4.07, + "learning_rate": 2.5942850104692697e-06, + "loss": 0.5225, + "step": 11030 + }, + { + "epoch": 4.07, + "learning_rate": 2.5942480601059244e-06, + "loss": 0.5114, + "step": 11031 + }, + { + "epoch": 4.07, + "learning_rate": 2.5942111097425794e-06, + "loss": 0.5224, + "step": 11032 + }, + { + "epoch": 4.07, + "learning_rate": 2.594174159379234e-06, + "loss": 0.6026, + "step": 11033 + }, + { + "epoch": 4.07, + "learning_rate": 2.5941372090158887e-06, + "loss": 0.405, + "step": 11034 + }, + { + "epoch": 4.07, + "learning_rate": 2.5941002586525433e-06, + "loss": 0.4749, + "step": 11035 + }, + { + "epoch": 4.08, + "learning_rate": 2.594063308289198e-06, + "loss": 0.5768, + "step": 11036 + }, + { + "epoch": 4.08, + "learning_rate": 2.594026357925853e-06, + "loss": 0.4103, + "step": 11037 + }, + { + "epoch": 4.08, + "learning_rate": 2.5939894075625077e-06, + "loss": 0.5068, + "step": 11038 + }, + { + "epoch": 4.08, + "learning_rate": 2.5939524571991627e-06, + "loss": 0.6401, + "step": 11039 + }, + { + "epoch": 4.08, + "learning_rate": 2.5939155068358174e-06, + "loss": 0.567, + "step": 11040 + }, + { + "epoch": 4.08, + "learning_rate": 2.593878556472472e-06, + "loss": 0.4934, + "step": 11041 + }, + { + "epoch": 4.08, + "learning_rate": 2.5938416061091266e-06, + "loss": 0.4903, + "step": 11042 + }, + { + "epoch": 4.08, + "learning_rate": 2.5938046557457813e-06, + "loss": 0.4634, + "step": 11043 + }, + { + "epoch": 4.08, + "learning_rate": 2.5937677053824363e-06, + "loss": 0.4128, + "step": 11044 + }, + { + "epoch": 4.08, + "learning_rate": 2.593730755019091e-06, + "loss": 0.3633, + "step": 11045 + }, + { + "epoch": 4.08, + "learning_rate": 2.593693804655746e-06, + "loss": 0.5058, + "step": 11046 + }, + { + "epoch": 4.08, + "learning_rate": 2.5936568542924007e-06, + "loss": 0.523, + "step": 11047 + }, + { + "epoch": 4.08, + "learning_rate": 2.5936199039290553e-06, + "loss": 0.462, + "step": 11048 + }, + { + "epoch": 4.08, + "learning_rate": 2.59358295356571e-06, + "loss": 0.4155, + "step": 11049 + }, + { + "epoch": 4.08, + "learning_rate": 2.593546003202365e-06, + "loss": 0.6034, + "step": 11050 + }, + { + "epoch": 4.08, + "learning_rate": 2.5935090528390196e-06, + "loss": 0.5722, + "step": 11051 + }, + { + "epoch": 4.08, + "learning_rate": 2.5934721024756747e-06, + "loss": 0.4456, + "step": 11052 + }, + { + "epoch": 4.08, + "learning_rate": 2.5934351521123293e-06, + "loss": 0.5193, + "step": 11053 + }, + { + "epoch": 4.08, + "learning_rate": 2.593398201748984e-06, + "loss": 0.6419, + "step": 11054 + }, + { + "epoch": 4.08, + "learning_rate": 2.5933612513856386e-06, + "loss": 0.4145, + "step": 11055 + }, + { + "epoch": 4.08, + "learning_rate": 2.5933243010222932e-06, + "loss": 0.5439, + "step": 11056 + }, + { + "epoch": 4.08, + "learning_rate": 2.5932873506589483e-06, + "loss": 0.4523, + "step": 11057 + }, + { + "epoch": 4.08, + "learning_rate": 2.593250400295603e-06, + "loss": 0.4995, + "step": 11058 + }, + { + "epoch": 4.08, + "learning_rate": 2.593213449932258e-06, + "loss": 0.4642, + "step": 11059 + }, + { + "epoch": 4.08, + "learning_rate": 2.5931764995689126e-06, + "loss": 0.5424, + "step": 11060 + }, + { + "epoch": 4.08, + "learning_rate": 2.5931395492055673e-06, + "loss": 0.4686, + "step": 11061 + }, + { + "epoch": 4.08, + "learning_rate": 2.593102598842222e-06, + "loss": 0.5305, + "step": 11062 + }, + { + "epoch": 4.09, + "learning_rate": 2.5930656484788765e-06, + "loss": 0.5779, + "step": 11063 + }, + { + "epoch": 4.09, + "learning_rate": 2.5930286981155316e-06, + "loss": 0.5117, + "step": 11064 + }, + { + "epoch": 4.09, + "learning_rate": 2.5929917477521862e-06, + "loss": 0.5078, + "step": 11065 + }, + { + "epoch": 4.09, + "learning_rate": 2.5929547973888413e-06, + "loss": 0.5659, + "step": 11066 + }, + { + "epoch": 4.09, + "learning_rate": 2.592917847025496e-06, + "loss": 0.5147, + "step": 11067 + }, + { + "epoch": 4.09, + "learning_rate": 2.5928808966621506e-06, + "loss": 0.4432, + "step": 11068 + }, + { + "epoch": 4.09, + "learning_rate": 2.592843946298805e-06, + "loss": 0.4696, + "step": 11069 + }, + { + "epoch": 4.09, + "learning_rate": 2.5928069959354603e-06, + "loss": 0.5562, + "step": 11070 + }, + { + "epoch": 4.09, + "learning_rate": 2.592770045572115e-06, + "loss": 0.5556, + "step": 11071 + }, + { + "epoch": 4.09, + "learning_rate": 2.59273309520877e-06, + "loss": 0.4955, + "step": 11072 + }, + { + "epoch": 4.09, + "learning_rate": 2.592696144845424e-06, + "loss": 0.5236, + "step": 11073 + }, + { + "epoch": 4.09, + "learning_rate": 2.592659194482079e-06, + "loss": 0.591, + "step": 11074 + }, + { + "epoch": 4.09, + "learning_rate": 2.592622244118734e-06, + "loss": 0.4222, + "step": 11075 + }, + { + "epoch": 4.09, + "learning_rate": 2.5925852937553885e-06, + "loss": 0.5901, + "step": 11076 + }, + { + "epoch": 4.09, + "learning_rate": 2.5925483433920436e-06, + "loss": 0.5016, + "step": 11077 + }, + { + "epoch": 4.09, + "learning_rate": 2.592511393028698e-06, + "loss": 0.5383, + "step": 11078 + }, + { + "epoch": 4.09, + "learning_rate": 2.592474442665353e-06, + "loss": 0.5739, + "step": 11079 + }, + { + "epoch": 4.09, + "learning_rate": 2.5924374923020075e-06, + "loss": 0.4689, + "step": 11080 + }, + { + "epoch": 4.09, + "learning_rate": 2.5924005419386625e-06, + "loss": 0.4871, + "step": 11081 + }, + { + "epoch": 4.09, + "learning_rate": 2.592363591575317e-06, + "loss": 0.4569, + "step": 11082 + }, + { + "epoch": 4.09, + "learning_rate": 2.592326641211972e-06, + "loss": 0.4954, + "step": 11083 + }, + { + "epoch": 4.09, + "learning_rate": 2.592289690848627e-06, + "loss": 0.47, + "step": 11084 + }, + { + "epoch": 4.09, + "learning_rate": 2.5922527404852815e-06, + "loss": 0.5701, + "step": 11085 + }, + { + "epoch": 4.09, + "learning_rate": 2.592215790121936e-06, + "loss": 0.4619, + "step": 11086 + }, + { + "epoch": 4.09, + "learning_rate": 2.5921788397585908e-06, + "loss": 0.4152, + "step": 11087 + }, + { + "epoch": 4.09, + "learning_rate": 2.592141889395246e-06, + "loss": 0.6699, + "step": 11088 + }, + { + "epoch": 4.09, + "learning_rate": 2.5921049390319005e-06, + "loss": 0.5803, + "step": 11089 + }, + { + "epoch": 4.1, + "learning_rate": 2.5920679886685555e-06, + "loss": 0.446, + "step": 11090 + }, + { + "epoch": 4.1, + "learning_rate": 2.59203103830521e-06, + "loss": 0.5239, + "step": 11091 + }, + { + "epoch": 4.1, + "learning_rate": 2.591994087941865e-06, + "loss": 0.4729, + "step": 11092 + }, + { + "epoch": 4.1, + "learning_rate": 2.5919571375785194e-06, + "loss": 0.5393, + "step": 11093 + }, + { + "epoch": 4.1, + "learning_rate": 2.591920187215174e-06, + "loss": 0.5156, + "step": 11094 + }, + { + "epoch": 4.1, + "learning_rate": 2.591883236851829e-06, + "loss": 0.5953, + "step": 11095 + }, + { + "epoch": 4.1, + "learning_rate": 2.5918462864884838e-06, + "loss": 0.4195, + "step": 11096 + }, + { + "epoch": 4.1, + "learning_rate": 2.591809336125139e-06, + "loss": 0.4817, + "step": 11097 + }, + { + "epoch": 4.1, + "learning_rate": 2.5917723857617935e-06, + "loss": 0.475, + "step": 11098 + }, + { + "epoch": 4.1, + "learning_rate": 2.591735435398448e-06, + "loss": 0.5126, + "step": 11099 + }, + { + "epoch": 4.1, + "learning_rate": 2.5916984850351027e-06, + "loss": 0.5696, + "step": 11100 + }, + { + "epoch": 4.1, + "learning_rate": 2.591661534671758e-06, + "loss": 0.5533, + "step": 11101 + }, + { + "epoch": 4.1, + "learning_rate": 2.5916245843084124e-06, + "loss": 0.3829, + "step": 11102 + }, + { + "epoch": 4.1, + "learning_rate": 2.591587633945067e-06, + "loss": 0.5724, + "step": 11103 + }, + { + "epoch": 4.1, + "learning_rate": 2.591550683581722e-06, + "loss": 0.4304, + "step": 11104 + }, + { + "epoch": 4.1, + "learning_rate": 2.5915137332183768e-06, + "loss": 0.4743, + "step": 11105 + }, + { + "epoch": 4.1, + "learning_rate": 2.5914767828550314e-06, + "loss": 0.5941, + "step": 11106 + }, + { + "epoch": 4.1, + "learning_rate": 2.591439832491686e-06, + "loss": 0.4349, + "step": 11107 + }, + { + "epoch": 4.1, + "learning_rate": 2.591402882128341e-06, + "loss": 0.5039, + "step": 11108 + }, + { + "epoch": 4.1, + "learning_rate": 2.5913659317649957e-06, + "loss": 0.5664, + "step": 11109 + }, + { + "epoch": 4.1, + "learning_rate": 2.591328981401651e-06, + "loss": 0.5389, + "step": 11110 + }, + { + "epoch": 4.1, + "learning_rate": 2.5912920310383054e-06, + "loss": 0.4825, + "step": 11111 + }, + { + "epoch": 4.1, + "learning_rate": 2.59125508067496e-06, + "loss": 0.507, + "step": 11112 + }, + { + "epoch": 4.1, + "learning_rate": 2.5912181303116147e-06, + "loss": 0.5528, + "step": 11113 + }, + { + "epoch": 4.1, + "learning_rate": 2.5911811799482694e-06, + "loss": 0.4755, + "step": 11114 + }, + { + "epoch": 4.1, + "learning_rate": 2.5911442295849244e-06, + "loss": 0.6133, + "step": 11115 + }, + { + "epoch": 4.1, + "learning_rate": 2.591107279221579e-06, + "loss": 0.4049, + "step": 11116 + }, + { + "epoch": 4.11, + "learning_rate": 2.591070328858234e-06, + "loss": 0.5577, + "step": 11117 + }, + { + "epoch": 4.11, + "learning_rate": 2.5910333784948883e-06, + "loss": 0.5551, + "step": 11118 + }, + { + "epoch": 4.11, + "learning_rate": 2.5909964281315434e-06, + "loss": 0.5227, + "step": 11119 + }, + { + "epoch": 4.11, + "learning_rate": 2.590959477768198e-06, + "loss": 0.5475, + "step": 11120 + }, + { + "epoch": 4.11, + "learning_rate": 2.590922527404853e-06, + "loss": 0.4706, + "step": 11121 + }, + { + "epoch": 4.11, + "learning_rate": 2.5908855770415077e-06, + "loss": 0.4887, + "step": 11122 + }, + { + "epoch": 4.11, + "learning_rate": 2.5908486266781624e-06, + "loss": 0.6742, + "step": 11123 + }, + { + "epoch": 4.11, + "learning_rate": 2.590811676314817e-06, + "loss": 0.5526, + "step": 11124 + }, + { + "epoch": 4.11, + "learning_rate": 2.5907747259514716e-06, + "loss": 0.5231, + "step": 11125 + }, + { + "epoch": 4.11, + "learning_rate": 2.5907377755881267e-06, + "loss": 0.5077, + "step": 11126 + }, + { + "epoch": 4.11, + "learning_rate": 2.5907008252247813e-06, + "loss": 0.5177, + "step": 11127 + }, + { + "epoch": 4.11, + "learning_rate": 2.5906638748614364e-06, + "loss": 0.524, + "step": 11128 + }, + { + "epoch": 4.11, + "learning_rate": 2.590626924498091e-06, + "loss": 0.4856, + "step": 11129 + }, + { + "epoch": 4.11, + "learning_rate": 2.5905899741347457e-06, + "loss": 0.6102, + "step": 11130 + }, + { + "epoch": 4.11, + "learning_rate": 2.5905530237714003e-06, + "loss": 0.5436, + "step": 11131 + }, + { + "epoch": 4.11, + "learning_rate": 2.5905160734080554e-06, + "loss": 0.6112, + "step": 11132 + }, + { + "epoch": 4.11, + "learning_rate": 2.59047912304471e-06, + "loss": 0.5726, + "step": 11133 + }, + { + "epoch": 4.11, + "learning_rate": 2.5904421726813646e-06, + "loss": 0.5799, + "step": 11134 + }, + { + "epoch": 4.11, + "learning_rate": 2.5904052223180197e-06, + "loss": 0.4841, + "step": 11135 + }, + { + "epoch": 4.11, + "learning_rate": 2.5903682719546743e-06, + "loss": 0.5745, + "step": 11136 + }, + { + "epoch": 4.11, + "learning_rate": 2.590331321591329e-06, + "loss": 0.4881, + "step": 11137 + }, + { + "epoch": 4.11, + "learning_rate": 2.5902943712279836e-06, + "loss": 0.6128, + "step": 11138 + }, + { + "epoch": 4.11, + "learning_rate": 2.5902574208646387e-06, + "loss": 0.5115, + "step": 11139 + }, + { + "epoch": 4.11, + "learning_rate": 2.5902204705012933e-06, + "loss": 0.4854, + "step": 11140 + }, + { + "epoch": 4.11, + "learning_rate": 2.5901835201379483e-06, + "loss": 0.5893, + "step": 11141 + }, + { + "epoch": 4.11, + "learning_rate": 2.590146569774603e-06, + "loss": 0.4872, + "step": 11142 + }, + { + "epoch": 4.11, + "learning_rate": 2.5901096194112576e-06, + "loss": 0.4801, + "step": 11143 + }, + { + "epoch": 4.12, + "learning_rate": 2.5900726690479123e-06, + "loss": 0.533, + "step": 11144 + }, + { + "epoch": 4.12, + "learning_rate": 2.590035718684567e-06, + "loss": 0.6318, + "step": 11145 + }, + { + "epoch": 4.12, + "learning_rate": 2.589998768321222e-06, + "loss": 0.4368, + "step": 11146 + }, + { + "epoch": 4.12, + "learning_rate": 2.5899618179578766e-06, + "loss": 0.541, + "step": 11147 + }, + { + "epoch": 4.12, + "learning_rate": 2.5899248675945317e-06, + "loss": 0.459, + "step": 11148 + }, + { + "epoch": 4.12, + "learning_rate": 2.5898879172311863e-06, + "loss": 0.6345, + "step": 11149 + }, + { + "epoch": 4.12, + "learning_rate": 2.589850966867841e-06, + "loss": 0.426, + "step": 11150 + }, + { + "epoch": 4.12, + "learning_rate": 2.5898140165044956e-06, + "loss": 0.6217, + "step": 11151 + }, + { + "epoch": 4.12, + "learning_rate": 2.5897770661411506e-06, + "loss": 0.5912, + "step": 11152 + }, + { + "epoch": 4.12, + "learning_rate": 2.5897401157778053e-06, + "loss": 0.4708, + "step": 11153 + }, + { + "epoch": 4.12, + "learning_rate": 2.58970316541446e-06, + "loss": 0.4921, + "step": 11154 + }, + { + "epoch": 4.12, + "learning_rate": 2.589666215051115e-06, + "loss": 0.5079, + "step": 11155 + }, + { + "epoch": 4.12, + "learning_rate": 2.5896292646877696e-06, + "loss": 0.5666, + "step": 11156 + }, + { + "epoch": 4.12, + "learning_rate": 2.5895923143244242e-06, + "loss": 0.5331, + "step": 11157 + }, + { + "epoch": 4.12, + "learning_rate": 2.589555363961079e-06, + "loss": 0.455, + "step": 11158 + }, + { + "epoch": 4.12, + "learning_rate": 2.589518413597734e-06, + "loss": 0.3944, + "step": 11159 + }, + { + "epoch": 4.12, + "learning_rate": 2.5894814632343886e-06, + "loss": 0.4598, + "step": 11160 + }, + { + "epoch": 4.12, + "learning_rate": 2.5894445128710436e-06, + "loss": 0.5965, + "step": 11161 + }, + { + "epoch": 4.12, + "learning_rate": 2.5894075625076983e-06, + "loss": 0.6335, + "step": 11162 + }, + { + "epoch": 4.12, + "learning_rate": 2.5893706121443525e-06, + "loss": 0.4843, + "step": 11163 + }, + { + "epoch": 4.12, + "learning_rate": 2.5893336617810075e-06, + "loss": 0.5604, + "step": 11164 + }, + { + "epoch": 4.12, + "learning_rate": 2.589296711417662e-06, + "loss": 0.4814, + "step": 11165 + }, + { + "epoch": 4.12, + "learning_rate": 2.5892597610543172e-06, + "loss": 0.5111, + "step": 11166 + }, + { + "epoch": 4.12, + "learning_rate": 2.589222810690972e-06, + "loss": 0.4954, + "step": 11167 + }, + { + "epoch": 4.12, + "learning_rate": 2.589185860327627e-06, + "loss": 0.5082, + "step": 11168 + }, + { + "epoch": 4.12, + "learning_rate": 2.589148909964281e-06, + "loss": 0.6726, + "step": 11169 + }, + { + "epoch": 4.12, + "learning_rate": 2.589111959600936e-06, + "loss": 0.6436, + "step": 11170 + }, + { + "epoch": 4.13, + "learning_rate": 2.589075009237591e-06, + "loss": 0.5404, + "step": 11171 + }, + { + "epoch": 4.13, + "learning_rate": 2.589038058874246e-06, + "loss": 0.6052, + "step": 11172 + }, + { + "epoch": 4.13, + "learning_rate": 2.5890011085109005e-06, + "loss": 0.4738, + "step": 11173 + }, + { + "epoch": 4.13, + "learning_rate": 2.588964158147555e-06, + "loss": 0.54, + "step": 11174 + }, + { + "epoch": 4.13, + "learning_rate": 2.58892720778421e-06, + "loss": 0.4196, + "step": 11175 + }, + { + "epoch": 4.13, + "learning_rate": 2.5888902574208644e-06, + "loss": 0.5529, + "step": 11176 + }, + { + "epoch": 4.13, + "learning_rate": 2.5888533070575195e-06, + "loss": 0.4419, + "step": 11177 + }, + { + "epoch": 4.13, + "learning_rate": 2.588816356694174e-06, + "loss": 0.618, + "step": 11178 + }, + { + "epoch": 4.13, + "learning_rate": 2.588779406330829e-06, + "loss": 0.467, + "step": 11179 + }, + { + "epoch": 4.13, + "learning_rate": 2.588742455967484e-06, + "loss": 0.6667, + "step": 11180 + }, + { + "epoch": 4.13, + "learning_rate": 2.5887055056041385e-06, + "loss": 0.5953, + "step": 11181 + }, + { + "epoch": 4.13, + "learning_rate": 2.588668555240793e-06, + "loss": 0.4234, + "step": 11182 + }, + { + "epoch": 4.13, + "learning_rate": 2.5886316048774477e-06, + "loss": 0.4371, + "step": 11183 + }, + { + "epoch": 4.13, + "learning_rate": 2.588594654514103e-06, + "loss": 0.5591, + "step": 11184 + }, + { + "epoch": 4.13, + "learning_rate": 2.5885577041507574e-06, + "loss": 0.5149, + "step": 11185 + }, + { + "epoch": 4.13, + "learning_rate": 2.5885207537874125e-06, + "loss": 0.5957, + "step": 11186 + }, + { + "epoch": 4.13, + "learning_rate": 2.588483803424067e-06, + "loss": 0.4679, + "step": 11187 + }, + { + "epoch": 4.13, + "learning_rate": 2.5884468530607218e-06, + "loss": 0.4364, + "step": 11188 + }, + { + "epoch": 4.13, + "learning_rate": 2.5884099026973764e-06, + "loss": 0.5951, + "step": 11189 + }, + { + "epoch": 4.13, + "learning_rate": 2.5883729523340315e-06, + "loss": 0.5675, + "step": 11190 + }, + { + "epoch": 4.13, + "learning_rate": 2.588336001970686e-06, + "loss": 0.5434, + "step": 11191 + }, + { + "epoch": 4.13, + "learning_rate": 2.5882990516073407e-06, + "loss": 0.5863, + "step": 11192 + }, + { + "epoch": 4.13, + "learning_rate": 2.588262101243996e-06, + "loss": 0.5161, + "step": 11193 + }, + { + "epoch": 4.13, + "learning_rate": 2.5882251508806504e-06, + "loss": 0.5015, + "step": 11194 + }, + { + "epoch": 4.13, + "learning_rate": 2.588188200517305e-06, + "loss": 0.6077, + "step": 11195 + }, + { + "epoch": 4.13, + "learning_rate": 2.5881512501539597e-06, + "loss": 0.4451, + "step": 11196 + }, + { + "epoch": 4.13, + "learning_rate": 2.5881142997906148e-06, + "loss": 0.4619, + "step": 11197 + }, + { + "epoch": 4.14, + "learning_rate": 2.5880773494272694e-06, + "loss": 0.5293, + "step": 11198 + }, + { + "epoch": 4.14, + "learning_rate": 2.5880403990639245e-06, + "loss": 0.551, + "step": 11199 + }, + { + "epoch": 4.14, + "learning_rate": 2.588003448700579e-06, + "loss": 0.385, + "step": 11200 + }, + { + "epoch": 4.14, + "learning_rate": 2.5879664983372337e-06, + "loss": 0.4764, + "step": 11201 + }, + { + "epoch": 4.14, + "learning_rate": 2.5879295479738884e-06, + "loss": 0.5966, + "step": 11202 + }, + { + "epoch": 4.14, + "learning_rate": 2.587892597610543e-06, + "loss": 0.5043, + "step": 11203 + }, + { + "epoch": 4.14, + "learning_rate": 2.587855647247198e-06, + "loss": 0.4585, + "step": 11204 + }, + { + "epoch": 4.14, + "learning_rate": 2.5878186968838527e-06, + "loss": 0.5296, + "step": 11205 + }, + { + "epoch": 4.14, + "learning_rate": 2.5877817465205078e-06, + "loss": 0.5333, + "step": 11206 + }, + { + "epoch": 4.14, + "learning_rate": 2.5877447961571624e-06, + "loss": 0.5633, + "step": 11207 + }, + { + "epoch": 4.14, + "learning_rate": 2.587707845793817e-06, + "loss": 0.4316, + "step": 11208 + }, + { + "epoch": 4.14, + "learning_rate": 2.5876708954304717e-06, + "loss": 0.6422, + "step": 11209 + }, + { + "epoch": 4.14, + "learning_rate": 2.5876339450671267e-06, + "loss": 0.5908, + "step": 11210 + }, + { + "epoch": 4.14, + "learning_rate": 2.5875969947037814e-06, + "loss": 0.6096, + "step": 11211 + }, + { + "epoch": 4.14, + "learning_rate": 2.587560044340436e-06, + "loss": 0.6188, + "step": 11212 + }, + { + "epoch": 4.14, + "learning_rate": 2.587523093977091e-06, + "loss": 0.4406, + "step": 11213 + }, + { + "epoch": 4.14, + "learning_rate": 2.5874861436137453e-06, + "loss": 0.5402, + "step": 11214 + }, + { + "epoch": 4.14, + "learning_rate": 2.5874491932504003e-06, + "loss": 0.6015, + "step": 11215 + }, + { + "epoch": 4.14, + "learning_rate": 2.587412242887055e-06, + "loss": 0.4129, + "step": 11216 + }, + { + "epoch": 4.14, + "learning_rate": 2.58737529252371e-06, + "loss": 0.6082, + "step": 11217 + }, + { + "epoch": 4.14, + "learning_rate": 2.5873383421603647e-06, + "loss": 0.5446, + "step": 11218 + }, + { + "epoch": 4.14, + "learning_rate": 2.5873013917970197e-06, + "loss": 0.5063, + "step": 11219 + }, + { + "epoch": 4.14, + "learning_rate": 2.587264441433674e-06, + "loss": 0.4554, + "step": 11220 + }, + { + "epoch": 4.14, + "learning_rate": 2.587227491070329e-06, + "loss": 0.5888, + "step": 11221 + }, + { + "epoch": 4.14, + "learning_rate": 2.5871905407069836e-06, + "loss": 0.4796, + "step": 11222 + }, + { + "epoch": 4.14, + "learning_rate": 2.5871535903436383e-06, + "loss": 0.5408, + "step": 11223 + }, + { + "epoch": 4.14, + "learning_rate": 2.5871166399802933e-06, + "loss": 0.6133, + "step": 11224 + }, + { + "epoch": 4.15, + "learning_rate": 2.587079689616948e-06, + "loss": 0.5376, + "step": 11225 + }, + { + "epoch": 4.15, + "learning_rate": 2.5870427392536026e-06, + "loss": 0.5962, + "step": 11226 + }, + { + "epoch": 4.15, + "learning_rate": 2.5870057888902572e-06, + "loss": 0.5373, + "step": 11227 + }, + { + "epoch": 4.15, + "learning_rate": 2.5869688385269123e-06, + "loss": 0.577, + "step": 11228 + }, + { + "epoch": 4.15, + "learning_rate": 2.586931888163567e-06, + "loss": 0.4729, + "step": 11229 + }, + { + "epoch": 4.15, + "learning_rate": 2.586894937800222e-06, + "loss": 0.5229, + "step": 11230 + }, + { + "epoch": 4.15, + "learning_rate": 2.5868579874368766e-06, + "loss": 0.5196, + "step": 11231 + }, + { + "epoch": 4.15, + "learning_rate": 2.5868210370735313e-06, + "loss": 0.5093, + "step": 11232 + }, + { + "epoch": 4.15, + "learning_rate": 2.586784086710186e-06, + "loss": 0.4533, + "step": 11233 + }, + { + "epoch": 4.15, + "learning_rate": 2.5867471363468405e-06, + "loss": 0.5203, + "step": 11234 + }, + { + "epoch": 4.15, + "learning_rate": 2.5867101859834956e-06, + "loss": 0.4807, + "step": 11235 + }, + { + "epoch": 4.15, + "learning_rate": 2.5866732356201502e-06, + "loss": 0.4617, + "step": 11236 + }, + { + "epoch": 4.15, + "learning_rate": 2.5866362852568053e-06, + "loss": 0.5986, + "step": 11237 + }, + { + "epoch": 4.15, + "learning_rate": 2.58659933489346e-06, + "loss": 0.4706, + "step": 11238 + }, + { + "epoch": 4.15, + "learning_rate": 2.5865623845301146e-06, + "loss": 0.588, + "step": 11239 + }, + { + "epoch": 4.15, + "learning_rate": 2.5865254341667692e-06, + "loss": 0.4222, + "step": 11240 + }, + { + "epoch": 4.15, + "learning_rate": 2.5864884838034243e-06, + "loss": 0.7157, + "step": 11241 + }, + { + "epoch": 4.15, + "learning_rate": 2.586451533440079e-06, + "loss": 0.5357, + "step": 11242 + }, + { + "epoch": 4.15, + "learning_rate": 2.5864145830767335e-06, + "loss": 0.5038, + "step": 11243 + }, + { + "epoch": 4.15, + "learning_rate": 2.5863776327133886e-06, + "loss": 0.4539, + "step": 11244 + }, + { + "epoch": 4.15, + "learning_rate": 2.5863406823500432e-06, + "loss": 0.5507, + "step": 11245 + }, + { + "epoch": 4.15, + "learning_rate": 2.586303731986698e-06, + "loss": 0.5436, + "step": 11246 + }, + { + "epoch": 4.15, + "learning_rate": 2.5862667816233525e-06, + "loss": 0.4698, + "step": 11247 + }, + { + "epoch": 4.15, + "learning_rate": 2.5862298312600076e-06, + "loss": 0.4403, + "step": 11248 + }, + { + "epoch": 4.15, + "learning_rate": 2.5861928808966622e-06, + "loss": 0.5588, + "step": 11249 + }, + { + "epoch": 4.15, + "learning_rate": 2.5861559305333173e-06, + "loss": 0.6758, + "step": 11250 + }, + { + "epoch": 4.15, + "learning_rate": 2.586118980169972e-06, + "loss": 0.4377, + "step": 11251 + }, + { + "epoch": 4.16, + "learning_rate": 2.5860820298066265e-06, + "loss": 0.5111, + "step": 11252 + }, + { + "epoch": 4.16, + "learning_rate": 2.586045079443281e-06, + "loss": 0.6088, + "step": 11253 + }, + { + "epoch": 4.16, + "learning_rate": 2.586008129079936e-06, + "loss": 0.5156, + "step": 11254 + }, + { + "epoch": 4.16, + "learning_rate": 2.585971178716591e-06, + "loss": 0.3671, + "step": 11255 + }, + { + "epoch": 4.16, + "learning_rate": 2.5859342283532455e-06, + "loss": 0.4833, + "step": 11256 + }, + { + "epoch": 4.16, + "learning_rate": 2.5858972779899006e-06, + "loss": 0.615, + "step": 11257 + }, + { + "epoch": 4.16, + "learning_rate": 2.5858603276265548e-06, + "loss": 0.4441, + "step": 11258 + }, + { + "epoch": 4.16, + "learning_rate": 2.58582337726321e-06, + "loss": 0.4431, + "step": 11259 + }, + { + "epoch": 4.16, + "learning_rate": 2.5857864268998645e-06, + "loss": 0.4984, + "step": 11260 + }, + { + "epoch": 4.16, + "learning_rate": 2.5857494765365195e-06, + "loss": 0.5611, + "step": 11261 + }, + { + "epoch": 4.16, + "learning_rate": 2.585712526173174e-06, + "loss": 0.4857, + "step": 11262 + }, + { + "epoch": 4.16, + "learning_rate": 2.585675575809829e-06, + "loss": 0.5152, + "step": 11263 + }, + { + "epoch": 4.16, + "learning_rate": 2.5856386254464835e-06, + "loss": 0.5317, + "step": 11264 + }, + { + "epoch": 4.16, + "learning_rate": 2.585601675083138e-06, + "loss": 0.5372, + "step": 11265 + }, + { + "epoch": 4.16, + "learning_rate": 2.585564724719793e-06, + "loss": 0.6149, + "step": 11266 + }, + { + "epoch": 4.16, + "learning_rate": 2.5855277743564478e-06, + "loss": 0.496, + "step": 11267 + }, + { + "epoch": 4.16, + "learning_rate": 2.585490823993103e-06, + "loss": 0.5111, + "step": 11268 + }, + { + "epoch": 4.16, + "learning_rate": 2.5854538736297575e-06, + "loss": 0.73, + "step": 11269 + }, + { + "epoch": 4.16, + "learning_rate": 2.585416923266412e-06, + "loss": 0.4994, + "step": 11270 + }, + { + "epoch": 4.16, + "learning_rate": 2.5853799729030668e-06, + "loss": 0.5302, + "step": 11271 + }, + { + "epoch": 4.16, + "learning_rate": 2.5853430225397214e-06, + "loss": 0.5286, + "step": 11272 + }, + { + "epoch": 4.16, + "learning_rate": 2.5853060721763765e-06, + "loss": 0.5386, + "step": 11273 + }, + { + "epoch": 4.16, + "learning_rate": 2.585269121813031e-06, + "loss": 0.52, + "step": 11274 + }, + { + "epoch": 4.16, + "learning_rate": 2.585232171449686e-06, + "loss": 0.5286, + "step": 11275 + }, + { + "epoch": 4.16, + "learning_rate": 2.5851952210863408e-06, + "loss": 0.6265, + "step": 11276 + }, + { + "epoch": 4.16, + "learning_rate": 2.5851582707229954e-06, + "loss": 0.3788, + "step": 11277 + }, + { + "epoch": 4.16, + "learning_rate": 2.58512132035965e-06, + "loss": 0.4985, + "step": 11278 + }, + { + "epoch": 4.17, + "learning_rate": 2.585084369996305e-06, + "loss": 0.4777, + "step": 11279 + }, + { + "epoch": 4.17, + "learning_rate": 2.5850474196329598e-06, + "loss": 0.4949, + "step": 11280 + }, + { + "epoch": 4.17, + "learning_rate": 2.585010469269615e-06, + "loss": 0.587, + "step": 11281 + }, + { + "epoch": 4.17, + "learning_rate": 2.5849735189062695e-06, + "loss": 0.5487, + "step": 11282 + }, + { + "epoch": 4.17, + "learning_rate": 2.584936568542924e-06, + "loss": 0.5511, + "step": 11283 + }, + { + "epoch": 4.17, + "learning_rate": 2.5848996181795787e-06, + "loss": 0.5417, + "step": 11284 + }, + { + "epoch": 4.17, + "learning_rate": 2.5848626678162334e-06, + "loss": 0.5701, + "step": 11285 + }, + { + "epoch": 4.17, + "learning_rate": 2.5848257174528884e-06, + "loss": 0.5132, + "step": 11286 + }, + { + "epoch": 4.17, + "learning_rate": 2.584788767089543e-06, + "loss": 0.6653, + "step": 11287 + }, + { + "epoch": 4.17, + "learning_rate": 2.584751816726198e-06, + "loss": 0.561, + "step": 11288 + }, + { + "epoch": 4.17, + "learning_rate": 2.5847148663628528e-06, + "loss": 0.4573, + "step": 11289 + }, + { + "epoch": 4.17, + "learning_rate": 2.5846779159995074e-06, + "loss": 0.5849, + "step": 11290 + }, + { + "epoch": 4.17, + "learning_rate": 2.584640965636162e-06, + "loss": 0.4956, + "step": 11291 + }, + { + "epoch": 4.17, + "learning_rate": 2.5846040152728167e-06, + "loss": 0.4845, + "step": 11292 + }, + { + "epoch": 4.17, + "learning_rate": 2.5845670649094717e-06, + "loss": 0.6613, + "step": 11293 + }, + { + "epoch": 4.17, + "learning_rate": 2.5845301145461264e-06, + "loss": 0.5755, + "step": 11294 + }, + { + "epoch": 4.17, + "learning_rate": 2.5844931641827814e-06, + "loss": 0.5408, + "step": 11295 + }, + { + "epoch": 4.17, + "learning_rate": 2.584456213819436e-06, + "loss": 0.4806, + "step": 11296 + }, + { + "epoch": 4.17, + "learning_rate": 2.5844192634560907e-06, + "loss": 0.5051, + "step": 11297 + }, + { + "epoch": 4.17, + "learning_rate": 2.5843823130927453e-06, + "loss": 0.5951, + "step": 11298 + }, + { + "epoch": 4.17, + "learning_rate": 2.5843453627294004e-06, + "loss": 0.4464, + "step": 11299 + }, + { + "epoch": 4.17, + "learning_rate": 2.584308412366055e-06, + "loss": 0.48, + "step": 11300 + }, + { + "epoch": 4.17, + "learning_rate": 2.58427146200271e-06, + "loss": 0.6601, + "step": 11301 + }, + { + "epoch": 4.17, + "learning_rate": 2.5842345116393647e-06, + "loss": 0.568, + "step": 11302 + }, + { + "epoch": 4.17, + "learning_rate": 2.584197561276019e-06, + "loss": 0.4911, + "step": 11303 + }, + { + "epoch": 4.17, + "learning_rate": 2.584160610912674e-06, + "loss": 0.5171, + "step": 11304 + }, + { + "epoch": 4.17, + "learning_rate": 2.5841236605493286e-06, + "loss": 0.4518, + "step": 11305 + }, + { + "epoch": 4.18, + "learning_rate": 2.5840867101859837e-06, + "loss": 0.4356, + "step": 11306 + }, + { + "epoch": 4.18, + "learning_rate": 2.5840497598226383e-06, + "loss": 0.6261, + "step": 11307 + }, + { + "epoch": 4.18, + "learning_rate": 2.5840128094592934e-06, + "loss": 0.7089, + "step": 11308 + }, + { + "epoch": 4.18, + "learning_rate": 2.5839758590959476e-06, + "loss": 0.4861, + "step": 11309 + }, + { + "epoch": 4.18, + "learning_rate": 2.5839389087326027e-06, + "loss": 0.5433, + "step": 11310 + }, + { + "epoch": 4.18, + "learning_rate": 2.5839019583692573e-06, + "loss": 0.5243, + "step": 11311 + }, + { + "epoch": 4.18, + "learning_rate": 2.583865008005912e-06, + "loss": 0.4862, + "step": 11312 + }, + { + "epoch": 4.18, + "learning_rate": 2.583828057642567e-06, + "loss": 0.4697, + "step": 11313 + }, + { + "epoch": 4.18, + "learning_rate": 2.5837911072792216e-06, + "loss": 0.513, + "step": 11314 + }, + { + "epoch": 4.18, + "learning_rate": 2.5837541569158763e-06, + "loss": 0.5031, + "step": 11315 + }, + { + "epoch": 4.18, + "learning_rate": 2.583717206552531e-06, + "loss": 0.4902, + "step": 11316 + }, + { + "epoch": 4.18, + "learning_rate": 2.583680256189186e-06, + "loss": 0.607, + "step": 11317 + }, + { + "epoch": 4.18, + "learning_rate": 2.5836433058258406e-06, + "loss": 0.4299, + "step": 11318 + }, + { + "epoch": 4.18, + "learning_rate": 2.5836063554624957e-06, + "loss": 0.583, + "step": 11319 + }, + { + "epoch": 4.18, + "learning_rate": 2.5835694050991503e-06, + "loss": 0.5505, + "step": 11320 + }, + { + "epoch": 4.18, + "learning_rate": 2.583532454735805e-06, + "loss": 0.5232, + "step": 11321 + }, + { + "epoch": 4.18, + "learning_rate": 2.5834955043724596e-06, + "loss": 0.605, + "step": 11322 + }, + { + "epoch": 4.18, + "learning_rate": 2.583458554009114e-06, + "loss": 0.3759, + "step": 11323 + }, + { + "epoch": 4.18, + "learning_rate": 2.5834216036457693e-06, + "loss": 0.6106, + "step": 11324 + }, + { + "epoch": 4.18, + "learning_rate": 2.583384653282424e-06, + "loss": 0.5159, + "step": 11325 + }, + { + "epoch": 4.18, + "learning_rate": 2.583347702919079e-06, + "loss": 0.5801, + "step": 11326 + }, + { + "epoch": 4.18, + "learning_rate": 2.5833107525557336e-06, + "loss": 0.5029, + "step": 11327 + }, + { + "epoch": 4.18, + "learning_rate": 2.5832738021923882e-06, + "loss": 0.5422, + "step": 11328 + }, + { + "epoch": 4.18, + "learning_rate": 2.583236851829043e-06, + "loss": 0.6026, + "step": 11329 + }, + { + "epoch": 4.18, + "learning_rate": 2.583199901465698e-06, + "loss": 0.6395, + "step": 11330 + }, + { + "epoch": 4.18, + "learning_rate": 2.5831629511023526e-06, + "loss": 0.5708, + "step": 11331 + }, + { + "epoch": 4.18, + "learning_rate": 2.583126000739007e-06, + "loss": 0.496, + "step": 11332 + }, + { + "epoch": 4.19, + "learning_rate": 2.5830890503756623e-06, + "loss": 0.4521, + "step": 11333 + }, + { + "epoch": 4.19, + "learning_rate": 2.583052100012317e-06, + "loss": 0.5667, + "step": 11334 + }, + { + "epoch": 4.19, + "learning_rate": 2.5830151496489715e-06, + "loss": 0.5035, + "step": 11335 + }, + { + "epoch": 4.19, + "learning_rate": 2.582978199285626e-06, + "loss": 0.5007, + "step": 11336 + }, + { + "epoch": 4.19, + "learning_rate": 2.5829412489222812e-06, + "loss": 0.5085, + "step": 11337 + }, + { + "epoch": 4.19, + "learning_rate": 2.582904298558936e-06, + "loss": 0.6136, + "step": 11338 + }, + { + "epoch": 4.19, + "learning_rate": 2.582867348195591e-06, + "loss": 0.4734, + "step": 11339 + }, + { + "epoch": 4.19, + "learning_rate": 2.5828303978322456e-06, + "loss": 0.494, + "step": 11340 + }, + { + "epoch": 4.19, + "learning_rate": 2.5827934474689e-06, + "loss": 0.5265, + "step": 11341 + }, + { + "epoch": 4.19, + "learning_rate": 2.582756497105555e-06, + "loss": 0.4443, + "step": 11342 + }, + { + "epoch": 4.19, + "learning_rate": 2.5827195467422095e-06, + "loss": 0.4263, + "step": 11343 + }, + { + "epoch": 4.19, + "learning_rate": 2.5826825963788645e-06, + "loss": 0.4945, + "step": 11344 + }, + { + "epoch": 4.19, + "learning_rate": 2.582645646015519e-06, + "loss": 0.6831, + "step": 11345 + }, + { + "epoch": 4.19, + "learning_rate": 2.5826086956521742e-06, + "loss": 0.4734, + "step": 11346 + }, + { + "epoch": 4.19, + "learning_rate": 2.582571745288829e-06, + "loss": 0.3874, + "step": 11347 + }, + { + "epoch": 4.19, + "learning_rate": 2.5825347949254835e-06, + "loss": 0.4411, + "step": 11348 + }, + { + "epoch": 4.19, + "learning_rate": 2.582497844562138e-06, + "loss": 0.5176, + "step": 11349 + }, + { + "epoch": 4.19, + "learning_rate": 2.582460894198793e-06, + "loss": 0.4886, + "step": 11350 + }, + { + "epoch": 4.19, + "learning_rate": 2.582423943835448e-06, + "loss": 0.5098, + "step": 11351 + }, + { + "epoch": 4.19, + "learning_rate": 2.5823869934721025e-06, + "loss": 0.628, + "step": 11352 + }, + { + "epoch": 4.19, + "learning_rate": 2.5823500431087575e-06, + "loss": 0.506, + "step": 11353 + }, + { + "epoch": 4.19, + "learning_rate": 2.5823130927454117e-06, + "loss": 0.5782, + "step": 11354 + }, + { + "epoch": 4.19, + "learning_rate": 2.582276142382067e-06, + "loss": 0.508, + "step": 11355 + }, + { + "epoch": 4.19, + "learning_rate": 2.5822391920187214e-06, + "loss": 0.4777, + "step": 11356 + }, + { + "epoch": 4.19, + "learning_rate": 2.5822022416553765e-06, + "loss": 0.4728, + "step": 11357 + }, + { + "epoch": 4.19, + "learning_rate": 2.582165291292031e-06, + "loss": 0.538, + "step": 11358 + }, + { + "epoch": 4.19, + "learning_rate": 2.582128340928686e-06, + "loss": 0.5596, + "step": 11359 + }, + { + "epoch": 4.19, + "learning_rate": 2.5820913905653404e-06, + "loss": 0.5639, + "step": 11360 + }, + { + "epoch": 4.2, + "learning_rate": 2.5820544402019955e-06, + "loss": 0.5235, + "step": 11361 + }, + { + "epoch": 4.2, + "learning_rate": 2.58201748983865e-06, + "loss": 0.5455, + "step": 11362 + }, + { + "epoch": 4.2, + "learning_rate": 2.5819805394753047e-06, + "loss": 0.4893, + "step": 11363 + }, + { + "epoch": 4.2, + "learning_rate": 2.58194358911196e-06, + "loss": 0.4921, + "step": 11364 + }, + { + "epoch": 4.2, + "learning_rate": 2.5819066387486144e-06, + "loss": 0.5764, + "step": 11365 + }, + { + "epoch": 4.2, + "learning_rate": 2.581869688385269e-06, + "loss": 0.4391, + "step": 11366 + }, + { + "epoch": 4.2, + "learning_rate": 2.5818327380219237e-06, + "loss": 0.4996, + "step": 11367 + }, + { + "epoch": 4.2, + "learning_rate": 2.5817957876585788e-06, + "loss": 0.5444, + "step": 11368 + }, + { + "epoch": 4.2, + "learning_rate": 2.5817588372952334e-06, + "loss": 0.3749, + "step": 11369 + }, + { + "epoch": 4.2, + "learning_rate": 2.5817218869318885e-06, + "loss": 0.4394, + "step": 11370 + }, + { + "epoch": 4.2, + "learning_rate": 2.581684936568543e-06, + "loss": 0.6205, + "step": 11371 + }, + { + "epoch": 4.2, + "learning_rate": 2.5816479862051977e-06, + "loss": 0.5076, + "step": 11372 + }, + { + "epoch": 4.2, + "learning_rate": 2.5816110358418524e-06, + "loss": 0.4504, + "step": 11373 + }, + { + "epoch": 4.2, + "learning_rate": 2.581574085478507e-06, + "loss": 0.5878, + "step": 11374 + }, + { + "epoch": 4.2, + "learning_rate": 2.581537135115162e-06, + "loss": 0.5258, + "step": 11375 + }, + { + "epoch": 4.2, + "learning_rate": 2.5815001847518167e-06, + "loss": 0.5989, + "step": 11376 + }, + { + "epoch": 4.2, + "learning_rate": 2.5814632343884718e-06, + "loss": 0.6584, + "step": 11377 + }, + { + "epoch": 4.2, + "learning_rate": 2.5814262840251264e-06, + "loss": 0.5807, + "step": 11378 + }, + { + "epoch": 4.2, + "learning_rate": 2.581389333661781e-06, + "loss": 0.5227, + "step": 11379 + }, + { + "epoch": 4.2, + "learning_rate": 2.5813523832984357e-06, + "loss": 0.4104, + "step": 11380 + }, + { + "epoch": 4.2, + "learning_rate": 2.5813154329350907e-06, + "loss": 0.5218, + "step": 11381 + }, + { + "epoch": 4.2, + "learning_rate": 2.5812784825717454e-06, + "loss": 0.4841, + "step": 11382 + }, + { + "epoch": 4.2, + "learning_rate": 2.5812415322084e-06, + "loss": 0.4884, + "step": 11383 + }, + { + "epoch": 4.2, + "learning_rate": 2.581204581845055e-06, + "loss": 0.7171, + "step": 11384 + }, + { + "epoch": 4.2, + "learning_rate": 2.5811676314817097e-06, + "loss": 0.5605, + "step": 11385 + }, + { + "epoch": 4.2, + "learning_rate": 2.5811306811183643e-06, + "loss": 0.5615, + "step": 11386 + }, + { + "epoch": 4.2, + "learning_rate": 2.581093730755019e-06, + "loss": 0.4388, + "step": 11387 + }, + { + "epoch": 4.21, + "learning_rate": 2.581056780391674e-06, + "loss": 0.5276, + "step": 11388 + }, + { + "epoch": 4.21, + "learning_rate": 2.5810198300283287e-06, + "loss": 0.5215, + "step": 11389 + }, + { + "epoch": 4.21, + "learning_rate": 2.5809828796649837e-06, + "loss": 0.5143, + "step": 11390 + }, + { + "epoch": 4.21, + "learning_rate": 2.5809459293016384e-06, + "loss": 0.469, + "step": 11391 + }, + { + "epoch": 4.21, + "learning_rate": 2.580908978938293e-06, + "loss": 0.4757, + "step": 11392 + }, + { + "epoch": 4.21, + "learning_rate": 2.5808720285749476e-06, + "loss": 0.5905, + "step": 11393 + }, + { + "epoch": 4.21, + "learning_rate": 2.5808350782116023e-06, + "loss": 0.446, + "step": 11394 + }, + { + "epoch": 4.21, + "learning_rate": 2.5807981278482573e-06, + "loss": 0.5399, + "step": 11395 + }, + { + "epoch": 4.21, + "learning_rate": 2.580761177484912e-06, + "loss": 0.4804, + "step": 11396 + }, + { + "epoch": 4.21, + "learning_rate": 2.580724227121567e-06, + "loss": 0.3972, + "step": 11397 + }, + { + "epoch": 4.21, + "learning_rate": 2.5806872767582217e-06, + "loss": 0.4845, + "step": 11398 + }, + { + "epoch": 4.21, + "learning_rate": 2.5806503263948763e-06, + "loss": 0.5681, + "step": 11399 + }, + { + "epoch": 4.21, + "learning_rate": 2.580613376031531e-06, + "loss": 0.4456, + "step": 11400 + }, + { + "epoch": 4.21, + "learning_rate": 2.580576425668186e-06, + "loss": 0.5122, + "step": 11401 + }, + { + "epoch": 4.21, + "learning_rate": 2.5805394753048406e-06, + "loss": 0.5198, + "step": 11402 + }, + { + "epoch": 4.21, + "learning_rate": 2.5805025249414953e-06, + "loss": 0.5816, + "step": 11403 + }, + { + "epoch": 4.21, + "learning_rate": 2.5804655745781503e-06, + "loss": 0.5047, + "step": 11404 + }, + { + "epoch": 4.21, + "learning_rate": 2.5804286242148046e-06, + "loss": 0.5264, + "step": 11405 + }, + { + "epoch": 4.21, + "learning_rate": 2.5803916738514596e-06, + "loss": 0.6068, + "step": 11406 + }, + { + "epoch": 4.21, + "learning_rate": 2.5803547234881143e-06, + "loss": 0.5031, + "step": 11407 + }, + { + "epoch": 4.21, + "learning_rate": 2.5803177731247693e-06, + "loss": 0.55, + "step": 11408 + }, + { + "epoch": 4.21, + "learning_rate": 2.580280822761424e-06, + "loss": 0.4789, + "step": 11409 + }, + { + "epoch": 4.21, + "learning_rate": 2.580243872398079e-06, + "loss": 0.5312, + "step": 11410 + }, + { + "epoch": 4.21, + "learning_rate": 2.5802069220347332e-06, + "loss": 0.5105, + "step": 11411 + }, + { + "epoch": 4.21, + "learning_rate": 2.580169971671388e-06, + "loss": 0.4319, + "step": 11412 + }, + { + "epoch": 4.21, + "learning_rate": 2.580133021308043e-06, + "loss": 0.4648, + "step": 11413 + }, + { + "epoch": 4.21, + "learning_rate": 2.5800960709446976e-06, + "loss": 0.5362, + "step": 11414 + }, + { + "epoch": 4.22, + "learning_rate": 2.5800591205813526e-06, + "loss": 0.4945, + "step": 11415 + }, + { + "epoch": 4.22, + "learning_rate": 2.5800221702180072e-06, + "loss": 0.5053, + "step": 11416 + }, + { + "epoch": 4.22, + "learning_rate": 2.579985219854662e-06, + "loss": 0.6048, + "step": 11417 + }, + { + "epoch": 4.22, + "learning_rate": 2.5799482694913165e-06, + "loss": 0.628, + "step": 11418 + }, + { + "epoch": 4.22, + "learning_rate": 2.5799113191279716e-06, + "loss": 0.5446, + "step": 11419 + }, + { + "epoch": 4.22, + "learning_rate": 2.5798743687646262e-06, + "loss": 0.5887, + "step": 11420 + }, + { + "epoch": 4.22, + "learning_rate": 2.579837418401281e-06, + "loss": 0.5305, + "step": 11421 + }, + { + "epoch": 4.22, + "learning_rate": 2.579800468037936e-06, + "loss": 0.567, + "step": 11422 + }, + { + "epoch": 4.22, + "learning_rate": 2.5797635176745906e-06, + "loss": 0.4317, + "step": 11423 + }, + { + "epoch": 4.22, + "learning_rate": 2.579726567311245e-06, + "loss": 0.5674, + "step": 11424 + }, + { + "epoch": 4.22, + "learning_rate": 2.5796896169479e-06, + "loss": 0.4791, + "step": 11425 + }, + { + "epoch": 4.22, + "learning_rate": 2.579652666584555e-06, + "loss": 0.5192, + "step": 11426 + }, + { + "epoch": 4.22, + "learning_rate": 2.5796157162212095e-06, + "loss": 0.4974, + "step": 11427 + }, + { + "epoch": 4.22, + "learning_rate": 2.5795787658578646e-06, + "loss": 0.5936, + "step": 11428 + }, + { + "epoch": 4.22, + "learning_rate": 2.5795418154945192e-06, + "loss": 0.6146, + "step": 11429 + }, + { + "epoch": 4.22, + "learning_rate": 2.579504865131174e-06, + "loss": 0.4575, + "step": 11430 + }, + { + "epoch": 4.22, + "learning_rate": 2.5794679147678285e-06, + "loss": 0.477, + "step": 11431 + }, + { + "epoch": 4.22, + "learning_rate": 2.579430964404483e-06, + "loss": 0.5834, + "step": 11432 + }, + { + "epoch": 4.22, + "learning_rate": 2.579394014041138e-06, + "loss": 0.4525, + "step": 11433 + }, + { + "epoch": 4.22, + "learning_rate": 2.579357063677793e-06, + "loss": 0.532, + "step": 11434 + }, + { + "epoch": 4.22, + "learning_rate": 2.579320113314448e-06, + "loss": 0.5403, + "step": 11435 + }, + { + "epoch": 4.22, + "learning_rate": 2.5792831629511025e-06, + "loss": 0.4767, + "step": 11436 + }, + { + "epoch": 4.22, + "learning_rate": 2.579246212587757e-06, + "loss": 0.5465, + "step": 11437 + }, + { + "epoch": 4.22, + "learning_rate": 2.579209262224412e-06, + "loss": 0.6037, + "step": 11438 + }, + { + "epoch": 4.22, + "learning_rate": 2.579172311861067e-06, + "loss": 0.5757, + "step": 11439 + }, + { + "epoch": 4.22, + "learning_rate": 2.5791353614977215e-06, + "loss": 0.5164, + "step": 11440 + }, + { + "epoch": 4.22, + "learning_rate": 2.579098411134376e-06, + "loss": 0.4796, + "step": 11441 + }, + { + "epoch": 4.23, + "learning_rate": 2.579061460771031e-06, + "loss": 0.5516, + "step": 11442 + }, + { + "epoch": 4.23, + "learning_rate": 2.5790245104076854e-06, + "loss": 0.3666, + "step": 11443 + }, + { + "epoch": 4.23, + "learning_rate": 2.5789875600443405e-06, + "loss": 0.4747, + "step": 11444 + }, + { + "epoch": 4.23, + "learning_rate": 2.578950609680995e-06, + "loss": 0.5655, + "step": 11445 + }, + { + "epoch": 4.23, + "learning_rate": 2.57891365931765e-06, + "loss": 0.6211, + "step": 11446 + }, + { + "epoch": 4.23, + "learning_rate": 2.578876708954305e-06, + "loss": 0.5732, + "step": 11447 + }, + { + "epoch": 4.23, + "learning_rate": 2.57883975859096e-06, + "loss": 0.5345, + "step": 11448 + }, + { + "epoch": 4.23, + "learning_rate": 2.578802808227614e-06, + "loss": 0.6075, + "step": 11449 + }, + { + "epoch": 4.23, + "learning_rate": 2.578765857864269e-06, + "loss": 0.5563, + "step": 11450 + }, + { + "epoch": 4.23, + "learning_rate": 2.5787289075009238e-06, + "loss": 0.5019, + "step": 11451 + }, + { + "epoch": 4.23, + "learning_rate": 2.5786919571375784e-06, + "loss": 0.44, + "step": 11452 + }, + { + "epoch": 4.23, + "learning_rate": 2.5786550067742335e-06, + "loss": 0.4992, + "step": 11453 + }, + { + "epoch": 4.23, + "learning_rate": 2.578618056410888e-06, + "loss": 0.4418, + "step": 11454 + }, + { + "epoch": 4.23, + "learning_rate": 2.5785811060475427e-06, + "loss": 0.5255, + "step": 11455 + }, + { + "epoch": 4.23, + "learning_rate": 2.5785441556841974e-06, + "loss": 0.4975, + "step": 11456 + }, + { + "epoch": 4.23, + "learning_rate": 2.5785072053208524e-06, + "loss": 0.4397, + "step": 11457 + }, + { + "epoch": 4.23, + "learning_rate": 2.578470254957507e-06, + "loss": 0.5799, + "step": 11458 + }, + { + "epoch": 4.23, + "learning_rate": 2.578433304594162e-06, + "loss": 0.5955, + "step": 11459 + }, + { + "epoch": 4.23, + "learning_rate": 2.5783963542308168e-06, + "loss": 0.5473, + "step": 11460 + }, + { + "epoch": 4.23, + "learning_rate": 2.5783594038674714e-06, + "loss": 0.6372, + "step": 11461 + }, + { + "epoch": 4.23, + "learning_rate": 2.578322453504126e-06, + "loss": 0.5103, + "step": 11462 + }, + { + "epoch": 4.23, + "learning_rate": 2.5782855031407807e-06, + "loss": 0.4632, + "step": 11463 + }, + { + "epoch": 4.23, + "learning_rate": 2.5782485527774357e-06, + "loss": 0.4504, + "step": 11464 + }, + { + "epoch": 4.23, + "learning_rate": 2.5782116024140904e-06, + "loss": 0.4667, + "step": 11465 + }, + { + "epoch": 4.23, + "learning_rate": 2.5781746520507454e-06, + "loss": 0.5424, + "step": 11466 + }, + { + "epoch": 4.23, + "learning_rate": 2.5781377016874e-06, + "loss": 0.5199, + "step": 11467 + }, + { + "epoch": 4.23, + "learning_rate": 2.5781007513240547e-06, + "loss": 0.5861, + "step": 11468 + }, + { + "epoch": 4.24, + "learning_rate": 2.5780638009607093e-06, + "loss": 0.559, + "step": 11469 + }, + { + "epoch": 4.24, + "learning_rate": 2.5780268505973644e-06, + "loss": 0.5193, + "step": 11470 + }, + { + "epoch": 4.24, + "learning_rate": 2.577989900234019e-06, + "loss": 0.412, + "step": 11471 + }, + { + "epoch": 4.24, + "learning_rate": 2.5779529498706737e-06, + "loss": 0.5304, + "step": 11472 + }, + { + "epoch": 4.24, + "learning_rate": 2.5779159995073287e-06, + "loss": 0.4446, + "step": 11473 + }, + { + "epoch": 4.24, + "learning_rate": 2.5778790491439834e-06, + "loss": 0.5609, + "step": 11474 + }, + { + "epoch": 4.24, + "learning_rate": 2.577842098780638e-06, + "loss": 0.6009, + "step": 11475 + }, + { + "epoch": 4.24, + "learning_rate": 2.5778051484172926e-06, + "loss": 0.6624, + "step": 11476 + }, + { + "epoch": 4.24, + "learning_rate": 2.5777681980539477e-06, + "loss": 0.5028, + "step": 11477 + }, + { + "epoch": 4.24, + "learning_rate": 2.5777312476906023e-06, + "loss": 0.6201, + "step": 11478 + }, + { + "epoch": 4.24, + "learning_rate": 2.5776942973272574e-06, + "loss": 0.5144, + "step": 11479 + }, + { + "epoch": 4.24, + "learning_rate": 2.577657346963912e-06, + "loss": 0.5523, + "step": 11480 + }, + { + "epoch": 4.24, + "learning_rate": 2.5776203966005667e-06, + "loss": 0.4967, + "step": 11481 + }, + { + "epoch": 4.24, + "learning_rate": 2.5775834462372213e-06, + "loss": 0.443, + "step": 11482 + }, + { + "epoch": 4.24, + "learning_rate": 2.577546495873876e-06, + "loss": 0.3852, + "step": 11483 + }, + { + "epoch": 4.24, + "learning_rate": 2.577509545510531e-06, + "loss": 0.4762, + "step": 11484 + }, + { + "epoch": 4.24, + "learning_rate": 2.5774725951471856e-06, + "loss": 0.5298, + "step": 11485 + }, + { + "epoch": 4.24, + "learning_rate": 2.5774356447838407e-06, + "loss": 0.4639, + "step": 11486 + }, + { + "epoch": 4.24, + "learning_rate": 2.5773986944204953e-06, + "loss": 0.5712, + "step": 11487 + }, + { + "epoch": 4.24, + "learning_rate": 2.57736174405715e-06, + "loss": 0.4746, + "step": 11488 + }, + { + "epoch": 4.24, + "learning_rate": 2.5773247936938046e-06, + "loss": 0.5533, + "step": 11489 + }, + { + "epoch": 4.24, + "learning_rate": 2.5772878433304597e-06, + "loss": 0.4323, + "step": 11490 + }, + { + "epoch": 4.24, + "learning_rate": 2.5772508929671143e-06, + "loss": 0.5102, + "step": 11491 + }, + { + "epoch": 4.24, + "learning_rate": 2.577213942603769e-06, + "loss": 0.4802, + "step": 11492 + }, + { + "epoch": 4.24, + "learning_rate": 2.577176992240424e-06, + "loss": 0.5754, + "step": 11493 + }, + { + "epoch": 4.24, + "learning_rate": 2.577140041877078e-06, + "loss": 0.6024, + "step": 11494 + }, + { + "epoch": 4.24, + "learning_rate": 2.5771030915137333e-06, + "loss": 0.5269, + "step": 11495 + }, + { + "epoch": 4.25, + "learning_rate": 2.577066141150388e-06, + "loss": 0.474, + "step": 11496 + }, + { + "epoch": 4.25, + "learning_rate": 2.577029190787043e-06, + "loss": 0.5133, + "step": 11497 + }, + { + "epoch": 4.25, + "learning_rate": 2.5769922404236976e-06, + "loss": 0.5726, + "step": 11498 + }, + { + "epoch": 4.25, + "learning_rate": 2.5769552900603527e-06, + "loss": 0.4847, + "step": 11499 + }, + { + "epoch": 4.25, + "learning_rate": 2.576918339697007e-06, + "loss": 0.4905, + "step": 11500 + }, + { + "epoch": 4.25, + "learning_rate": 2.5768813893336615e-06, + "loss": 0.485, + "step": 11501 + }, + { + "epoch": 4.25, + "learning_rate": 2.5768444389703166e-06, + "loss": 0.5733, + "step": 11502 + }, + { + "epoch": 4.25, + "learning_rate": 2.576807488606971e-06, + "loss": 0.591, + "step": 11503 + }, + { + "epoch": 4.25, + "learning_rate": 2.5767705382436263e-06, + "loss": 0.5312, + "step": 11504 + }, + { + "epoch": 4.25, + "learning_rate": 2.576733587880281e-06, + "loss": 0.6274, + "step": 11505 + }, + { + "epoch": 4.25, + "learning_rate": 2.5766966375169355e-06, + "loss": 0.4707, + "step": 11506 + }, + { + "epoch": 4.25, + "learning_rate": 2.57665968715359e-06, + "loss": 0.4587, + "step": 11507 + }, + { + "epoch": 4.25, + "learning_rate": 2.5766227367902452e-06, + "loss": 0.5215, + "step": 11508 + }, + { + "epoch": 4.25, + "learning_rate": 2.5765857864269e-06, + "loss": 0.5284, + "step": 11509 + }, + { + "epoch": 4.25, + "learning_rate": 2.576548836063555e-06, + "loss": 0.5574, + "step": 11510 + }, + { + "epoch": 4.25, + "learning_rate": 2.5765118857002096e-06, + "loss": 0.578, + "step": 11511 + }, + { + "epoch": 4.25, + "learning_rate": 2.576474935336864e-06, + "loss": 0.5543, + "step": 11512 + }, + { + "epoch": 4.25, + "learning_rate": 2.576437984973519e-06, + "loss": 0.6725, + "step": 11513 + }, + { + "epoch": 4.25, + "learning_rate": 2.5764010346101735e-06, + "loss": 0.5872, + "step": 11514 + }, + { + "epoch": 4.25, + "learning_rate": 2.5763640842468285e-06, + "loss": 0.5971, + "step": 11515 + }, + { + "epoch": 4.25, + "learning_rate": 2.576327133883483e-06, + "loss": 0.4585, + "step": 11516 + }, + { + "epoch": 4.25, + "learning_rate": 2.5762901835201382e-06, + "loss": 0.554, + "step": 11517 + }, + { + "epoch": 4.25, + "learning_rate": 2.576253233156793e-06, + "loss": 0.5687, + "step": 11518 + }, + { + "epoch": 4.25, + "learning_rate": 2.5762162827934475e-06, + "loss": 0.528, + "step": 11519 + }, + { + "epoch": 4.25, + "learning_rate": 2.576179332430102e-06, + "loss": 0.5414, + "step": 11520 + }, + { + "epoch": 4.25, + "learning_rate": 2.5761423820667568e-06, + "loss": 0.467, + "step": 11521 + }, + { + "epoch": 4.25, + "learning_rate": 2.576105431703412e-06, + "loss": 0.5032, + "step": 11522 + }, + { + "epoch": 4.26, + "learning_rate": 2.5760684813400665e-06, + "loss": 0.4944, + "step": 11523 + }, + { + "epoch": 4.26, + "learning_rate": 2.5760315309767215e-06, + "loss": 0.4749, + "step": 11524 + }, + { + "epoch": 4.26, + "learning_rate": 2.575994580613376e-06, + "loss": 0.4279, + "step": 11525 + }, + { + "epoch": 4.26, + "learning_rate": 2.575957630250031e-06, + "loss": 0.5075, + "step": 11526 + }, + { + "epoch": 4.26, + "learning_rate": 2.5759206798866854e-06, + "loss": 0.3733, + "step": 11527 + }, + { + "epoch": 4.26, + "learning_rate": 2.5758837295233405e-06, + "loss": 0.4863, + "step": 11528 + }, + { + "epoch": 4.26, + "learning_rate": 2.575846779159995e-06, + "loss": 0.5648, + "step": 11529 + }, + { + "epoch": 4.26, + "learning_rate": 2.57580982879665e-06, + "loss": 0.5738, + "step": 11530 + }, + { + "epoch": 4.26, + "learning_rate": 2.575772878433305e-06, + "loss": 0.4881, + "step": 11531 + }, + { + "epoch": 4.26, + "learning_rate": 2.5757359280699595e-06, + "loss": 0.584, + "step": 11532 + }, + { + "epoch": 4.26, + "learning_rate": 2.575698977706614e-06, + "loss": 0.4919, + "step": 11533 + }, + { + "epoch": 4.26, + "learning_rate": 2.5756620273432687e-06, + "loss": 0.4409, + "step": 11534 + }, + { + "epoch": 4.26, + "learning_rate": 2.575625076979924e-06, + "loss": 0.4597, + "step": 11535 + }, + { + "epoch": 4.26, + "learning_rate": 2.5755881266165784e-06, + "loss": 0.4595, + "step": 11536 + }, + { + "epoch": 4.26, + "learning_rate": 2.5755511762532335e-06, + "loss": 0.4949, + "step": 11537 + }, + { + "epoch": 4.26, + "learning_rate": 2.575514225889888e-06, + "loss": 0.4921, + "step": 11538 + }, + { + "epoch": 4.26, + "learning_rate": 2.5754772755265428e-06, + "loss": 0.5223, + "step": 11539 + }, + { + "epoch": 4.26, + "learning_rate": 2.5754403251631974e-06, + "loss": 0.6553, + "step": 11540 + }, + { + "epoch": 4.26, + "learning_rate": 2.575403374799852e-06, + "loss": 0.5033, + "step": 11541 + }, + { + "epoch": 4.26, + "learning_rate": 2.575366424436507e-06, + "loss": 0.4863, + "step": 11542 + }, + { + "epoch": 4.26, + "learning_rate": 2.5753294740731617e-06, + "loss": 0.4165, + "step": 11543 + }, + { + "epoch": 4.26, + "learning_rate": 2.575292523709817e-06, + "loss": 0.5035, + "step": 11544 + }, + { + "epoch": 4.26, + "learning_rate": 2.575255573346471e-06, + "loss": 0.488, + "step": 11545 + }, + { + "epoch": 4.26, + "learning_rate": 2.575218622983126e-06, + "loss": 0.4344, + "step": 11546 + }, + { + "epoch": 4.26, + "learning_rate": 2.5751816726197807e-06, + "loss": 0.5837, + "step": 11547 + }, + { + "epoch": 4.26, + "learning_rate": 2.5751447222564358e-06, + "loss": 0.4584, + "step": 11548 + }, + { + "epoch": 4.26, + "learning_rate": 2.5751077718930904e-06, + "loss": 0.5134, + "step": 11549 + }, + { + "epoch": 4.27, + "learning_rate": 2.5750708215297455e-06, + "loss": 0.6245, + "step": 11550 + }, + { + "epoch": 4.27, + "learning_rate": 2.5750338711663997e-06, + "loss": 0.59, + "step": 11551 + }, + { + "epoch": 4.27, + "learning_rate": 2.5749969208030543e-06, + "loss": 0.4421, + "step": 11552 + }, + { + "epoch": 4.27, + "learning_rate": 2.5749599704397094e-06, + "loss": 0.5738, + "step": 11553 + }, + { + "epoch": 4.27, + "learning_rate": 2.574923020076364e-06, + "loss": 0.6629, + "step": 11554 + }, + { + "epoch": 4.27, + "learning_rate": 2.574886069713019e-06, + "loss": 0.5246, + "step": 11555 + }, + { + "epoch": 4.27, + "learning_rate": 2.5748491193496737e-06, + "loss": 0.5185, + "step": 11556 + }, + { + "epoch": 4.27, + "learning_rate": 2.5748121689863284e-06, + "loss": 0.466, + "step": 11557 + }, + { + "epoch": 4.27, + "learning_rate": 2.574775218622983e-06, + "loss": 0.3997, + "step": 11558 + }, + { + "epoch": 4.27, + "learning_rate": 2.574738268259638e-06, + "loss": 0.6152, + "step": 11559 + }, + { + "epoch": 4.27, + "learning_rate": 2.5747013178962927e-06, + "loss": 0.457, + "step": 11560 + }, + { + "epoch": 4.27, + "learning_rate": 2.5746643675329473e-06, + "loss": 0.5051, + "step": 11561 + }, + { + "epoch": 4.27, + "learning_rate": 2.5746274171696024e-06, + "loss": 0.5332, + "step": 11562 + }, + { + "epoch": 4.27, + "learning_rate": 2.574590466806257e-06, + "loss": 0.4942, + "step": 11563 + }, + { + "epoch": 4.27, + "learning_rate": 2.5745535164429117e-06, + "loss": 0.5534, + "step": 11564 + }, + { + "epoch": 4.27, + "learning_rate": 2.5745165660795663e-06, + "loss": 0.5479, + "step": 11565 + }, + { + "epoch": 4.27, + "learning_rate": 2.5744796157162213e-06, + "loss": 0.5639, + "step": 11566 + }, + { + "epoch": 4.27, + "learning_rate": 2.574442665352876e-06, + "loss": 0.561, + "step": 11567 + }, + { + "epoch": 4.27, + "learning_rate": 2.574405714989531e-06, + "loss": 0.5933, + "step": 11568 + }, + { + "epoch": 4.27, + "learning_rate": 2.5743687646261857e-06, + "loss": 0.5311, + "step": 11569 + }, + { + "epoch": 4.27, + "learning_rate": 2.5743318142628403e-06, + "loss": 0.603, + "step": 11570 + }, + { + "epoch": 4.27, + "learning_rate": 2.574294863899495e-06, + "loss": 0.5863, + "step": 11571 + }, + { + "epoch": 4.27, + "learning_rate": 2.5742579135361496e-06, + "loss": 0.6447, + "step": 11572 + }, + { + "epoch": 4.27, + "learning_rate": 2.5742209631728047e-06, + "loss": 0.4927, + "step": 11573 + }, + { + "epoch": 4.27, + "learning_rate": 2.5741840128094593e-06, + "loss": 0.5115, + "step": 11574 + }, + { + "epoch": 4.27, + "learning_rate": 2.5741470624461143e-06, + "loss": 0.5332, + "step": 11575 + }, + { + "epoch": 4.27, + "learning_rate": 2.574110112082769e-06, + "loss": 0.6164, + "step": 11576 + }, + { + "epoch": 4.28, + "learning_rate": 2.5740731617194236e-06, + "loss": 0.4416, + "step": 11577 + }, + { + "epoch": 4.28, + "learning_rate": 2.5740362113560783e-06, + "loss": 0.4384, + "step": 11578 + }, + { + "epoch": 4.28, + "learning_rate": 2.5739992609927333e-06, + "loss": 0.5338, + "step": 11579 + }, + { + "epoch": 4.28, + "learning_rate": 2.573962310629388e-06, + "loss": 0.5219, + "step": 11580 + }, + { + "epoch": 4.28, + "learning_rate": 2.5739253602660426e-06, + "loss": 0.5224, + "step": 11581 + }, + { + "epoch": 4.28, + "learning_rate": 2.5738884099026977e-06, + "loss": 0.4773, + "step": 11582 + }, + { + "epoch": 4.28, + "learning_rate": 2.5738514595393523e-06, + "loss": 0.4593, + "step": 11583 + }, + { + "epoch": 4.28, + "learning_rate": 2.573814509176007e-06, + "loss": 0.5138, + "step": 11584 + }, + { + "epoch": 4.28, + "learning_rate": 2.5737775588126616e-06, + "loss": 0.4985, + "step": 11585 + }, + { + "epoch": 4.28, + "learning_rate": 2.5737406084493166e-06, + "loss": 0.4411, + "step": 11586 + }, + { + "epoch": 4.28, + "learning_rate": 2.5737036580859713e-06, + "loss": 0.5047, + "step": 11587 + }, + { + "epoch": 4.28, + "learning_rate": 2.5736667077226263e-06, + "loss": 0.6119, + "step": 11588 + }, + { + "epoch": 4.28, + "learning_rate": 2.573629757359281e-06, + "loss": 0.5912, + "step": 11589 + }, + { + "epoch": 4.28, + "learning_rate": 2.5735928069959356e-06, + "loss": 0.5339, + "step": 11590 + }, + { + "epoch": 4.28, + "learning_rate": 2.5735558566325902e-06, + "loss": 0.4623, + "step": 11591 + }, + { + "epoch": 4.28, + "learning_rate": 2.573518906269245e-06, + "loss": 0.6351, + "step": 11592 + }, + { + "epoch": 4.28, + "learning_rate": 2.5734819559059e-06, + "loss": 0.6047, + "step": 11593 + }, + { + "epoch": 4.28, + "learning_rate": 2.5734450055425546e-06, + "loss": 0.6058, + "step": 11594 + }, + { + "epoch": 4.28, + "learning_rate": 2.5734080551792096e-06, + "loss": 0.5214, + "step": 11595 + }, + { + "epoch": 4.28, + "learning_rate": 2.573371104815864e-06, + "loss": 0.4927, + "step": 11596 + }, + { + "epoch": 4.28, + "learning_rate": 2.573334154452519e-06, + "loss": 0.4879, + "step": 11597 + }, + { + "epoch": 4.28, + "learning_rate": 2.5732972040891735e-06, + "loss": 0.5089, + "step": 11598 + }, + { + "epoch": 4.28, + "learning_rate": 2.5732602537258286e-06, + "loss": 0.4735, + "step": 11599 + }, + { + "epoch": 4.28, + "learning_rate": 2.5732233033624832e-06, + "loss": 0.426, + "step": 11600 + }, + { + "epoch": 4.28, + "learning_rate": 2.573186352999138e-06, + "loss": 0.5873, + "step": 11601 + }, + { + "epoch": 4.28, + "learning_rate": 2.5731494026357925e-06, + "loss": 0.5358, + "step": 11602 + }, + { + "epoch": 4.28, + "learning_rate": 2.573112452272447e-06, + "loss": 0.5226, + "step": 11603 + }, + { + "epoch": 4.29, + "learning_rate": 2.573075501909102e-06, + "loss": 0.5216, + "step": 11604 + }, + { + "epoch": 4.29, + "learning_rate": 2.573038551545757e-06, + "loss": 0.5476, + "step": 11605 + }, + { + "epoch": 4.29, + "learning_rate": 2.573001601182412e-06, + "loss": 0.4764, + "step": 11606 + }, + { + "epoch": 4.29, + "learning_rate": 2.5729646508190665e-06, + "loss": 0.5305, + "step": 11607 + }, + { + "epoch": 4.29, + "learning_rate": 2.572927700455721e-06, + "loss": 0.5379, + "step": 11608 + }, + { + "epoch": 4.29, + "learning_rate": 2.572890750092376e-06, + "loss": 0.6184, + "step": 11609 + }, + { + "epoch": 4.29, + "learning_rate": 2.572853799729031e-06, + "loss": 0.4914, + "step": 11610 + }, + { + "epoch": 4.29, + "learning_rate": 2.5728168493656855e-06, + "loss": 0.4953, + "step": 11611 + }, + { + "epoch": 4.29, + "learning_rate": 2.57277989900234e-06, + "loss": 0.4357, + "step": 11612 + }, + { + "epoch": 4.29, + "learning_rate": 2.572742948638995e-06, + "loss": 0.5914, + "step": 11613 + }, + { + "epoch": 4.29, + "learning_rate": 2.57270599827565e-06, + "loss": 0.4945, + "step": 11614 + }, + { + "epoch": 4.29, + "learning_rate": 2.5726690479123045e-06, + "loss": 0.5849, + "step": 11615 + }, + { + "epoch": 4.29, + "learning_rate": 2.572632097548959e-06, + "loss": 0.4558, + "step": 11616 + }, + { + "epoch": 4.29, + "learning_rate": 2.572595147185614e-06, + "loss": 0.4549, + "step": 11617 + }, + { + "epoch": 4.29, + "learning_rate": 2.572558196822269e-06, + "loss": 0.556, + "step": 11618 + }, + { + "epoch": 4.29, + "learning_rate": 2.572521246458924e-06, + "loss": 0.5905, + "step": 11619 + }, + { + "epoch": 4.29, + "learning_rate": 2.5724842960955785e-06, + "loss": 0.5698, + "step": 11620 + }, + { + "epoch": 4.29, + "learning_rate": 2.572447345732233e-06, + "loss": 0.5093, + "step": 11621 + }, + { + "epoch": 4.29, + "learning_rate": 2.5724103953688878e-06, + "loss": 0.5459, + "step": 11622 + }, + { + "epoch": 4.29, + "learning_rate": 2.5723734450055424e-06, + "loss": 0.6105, + "step": 11623 + }, + { + "epoch": 4.29, + "learning_rate": 2.5723364946421975e-06, + "loss": 0.4576, + "step": 11624 + }, + { + "epoch": 4.29, + "learning_rate": 2.572299544278852e-06, + "loss": 0.5682, + "step": 11625 + }, + { + "epoch": 4.29, + "learning_rate": 2.572262593915507e-06, + "loss": 0.4743, + "step": 11626 + }, + { + "epoch": 4.29, + "learning_rate": 2.572225643552162e-06, + "loss": 0.4674, + "step": 11627 + }, + { + "epoch": 4.29, + "learning_rate": 2.5721886931888164e-06, + "loss": 0.4364, + "step": 11628 + }, + { + "epoch": 4.29, + "learning_rate": 2.572151742825471e-06, + "loss": 0.452, + "step": 11629 + }, + { + "epoch": 4.29, + "learning_rate": 2.572114792462126e-06, + "loss": 0.5014, + "step": 11630 + }, + { + "epoch": 4.3, + "learning_rate": 2.5720778420987808e-06, + "loss": 0.6017, + "step": 11631 + }, + { + "epoch": 4.3, + "learning_rate": 2.5720408917354354e-06, + "loss": 0.5131, + "step": 11632 + }, + { + "epoch": 4.3, + "learning_rate": 2.5720039413720905e-06, + "loss": 0.5011, + "step": 11633 + }, + { + "epoch": 4.3, + "learning_rate": 2.5719669910087447e-06, + "loss": 0.5363, + "step": 11634 + }, + { + "epoch": 4.3, + "learning_rate": 2.5719300406453997e-06, + "loss": 0.5658, + "step": 11635 + }, + { + "epoch": 4.3, + "learning_rate": 2.5718930902820544e-06, + "loss": 0.5067, + "step": 11636 + }, + { + "epoch": 4.3, + "learning_rate": 2.5718561399187094e-06, + "loss": 0.5526, + "step": 11637 + }, + { + "epoch": 4.3, + "learning_rate": 2.571819189555364e-06, + "loss": 0.4518, + "step": 11638 + }, + { + "epoch": 4.3, + "learning_rate": 2.571782239192019e-06, + "loss": 0.5159, + "step": 11639 + }, + { + "epoch": 4.3, + "learning_rate": 2.5717452888286733e-06, + "loss": 0.5, + "step": 11640 + }, + { + "epoch": 4.3, + "learning_rate": 2.571708338465328e-06, + "loss": 0.5254, + "step": 11641 + }, + { + "epoch": 4.3, + "learning_rate": 2.571671388101983e-06, + "loss": 0.3762, + "step": 11642 + }, + { + "epoch": 4.3, + "learning_rate": 2.5716344377386377e-06, + "loss": 0.4858, + "step": 11643 + }, + { + "epoch": 4.3, + "learning_rate": 2.5715974873752927e-06, + "loss": 0.4356, + "step": 11644 + }, + { + "epoch": 4.3, + "learning_rate": 2.5715605370119474e-06, + "loss": 0.5356, + "step": 11645 + }, + { + "epoch": 4.3, + "learning_rate": 2.571523586648602e-06, + "loss": 0.5421, + "step": 11646 + }, + { + "epoch": 4.3, + "learning_rate": 2.5714866362852566e-06, + "loss": 0.452, + "step": 11647 + }, + { + "epoch": 4.3, + "learning_rate": 2.5714496859219117e-06, + "loss": 0.5127, + "step": 11648 + }, + { + "epoch": 4.3, + "learning_rate": 2.5714127355585663e-06, + "loss": 0.49, + "step": 11649 + }, + { + "epoch": 4.3, + "learning_rate": 2.571375785195221e-06, + "loss": 0.4968, + "step": 11650 + }, + { + "epoch": 4.3, + "learning_rate": 2.571338834831876e-06, + "loss": 0.5786, + "step": 11651 + }, + { + "epoch": 4.3, + "learning_rate": 2.5713018844685307e-06, + "loss": 0.5431, + "step": 11652 + }, + { + "epoch": 4.3, + "learning_rate": 2.5712649341051853e-06, + "loss": 0.5541, + "step": 11653 + }, + { + "epoch": 4.3, + "learning_rate": 2.57122798374184e-06, + "loss": 0.5638, + "step": 11654 + }, + { + "epoch": 4.3, + "learning_rate": 2.571191033378495e-06, + "loss": 0.4891, + "step": 11655 + }, + { + "epoch": 4.3, + "learning_rate": 2.5711540830151496e-06, + "loss": 0.5401, + "step": 11656 + }, + { + "epoch": 4.3, + "learning_rate": 2.5711171326518047e-06, + "loss": 0.5617, + "step": 11657 + }, + { + "epoch": 4.31, + "learning_rate": 2.5710801822884593e-06, + "loss": 0.5857, + "step": 11658 + }, + { + "epoch": 4.31, + "learning_rate": 2.571043231925114e-06, + "loss": 0.4978, + "step": 11659 + }, + { + "epoch": 4.31, + "learning_rate": 2.5710062815617686e-06, + "loss": 0.5646, + "step": 11660 + }, + { + "epoch": 4.31, + "learning_rate": 2.5709693311984232e-06, + "loss": 0.6195, + "step": 11661 + }, + { + "epoch": 4.31, + "learning_rate": 2.5709323808350783e-06, + "loss": 0.4644, + "step": 11662 + }, + { + "epoch": 4.31, + "learning_rate": 2.570895430471733e-06, + "loss": 0.4426, + "step": 11663 + }, + { + "epoch": 4.31, + "learning_rate": 2.570858480108388e-06, + "loss": 0.5093, + "step": 11664 + }, + { + "epoch": 4.31, + "learning_rate": 2.5708215297450426e-06, + "loss": 0.5885, + "step": 11665 + }, + { + "epoch": 4.31, + "learning_rate": 2.5707845793816973e-06, + "loss": 0.4361, + "step": 11666 + }, + { + "epoch": 4.31, + "learning_rate": 2.570747629018352e-06, + "loss": 0.4413, + "step": 11667 + }, + { + "epoch": 4.31, + "learning_rate": 2.570710678655007e-06, + "loss": 0.5432, + "step": 11668 + }, + { + "epoch": 4.31, + "learning_rate": 2.5706737282916616e-06, + "loss": 0.5157, + "step": 11669 + }, + { + "epoch": 4.31, + "learning_rate": 2.5706367779283162e-06, + "loss": 0.5259, + "step": 11670 + }, + { + "epoch": 4.31, + "learning_rate": 2.5705998275649713e-06, + "loss": 0.5509, + "step": 11671 + }, + { + "epoch": 4.31, + "learning_rate": 2.570562877201626e-06, + "loss": 0.4968, + "step": 11672 + }, + { + "epoch": 4.31, + "learning_rate": 2.5705259268382806e-06, + "loss": 0.4485, + "step": 11673 + }, + { + "epoch": 4.31, + "learning_rate": 2.5704889764749352e-06, + "loss": 0.4547, + "step": 11674 + }, + { + "epoch": 4.31, + "learning_rate": 2.5704520261115903e-06, + "loss": 0.609, + "step": 11675 + }, + { + "epoch": 4.31, + "learning_rate": 2.570415075748245e-06, + "loss": 0.5105, + "step": 11676 + }, + { + "epoch": 4.31, + "learning_rate": 2.5703781253849e-06, + "loss": 0.455, + "step": 11677 + }, + { + "epoch": 4.31, + "learning_rate": 2.5703411750215546e-06, + "loss": 0.4904, + "step": 11678 + }, + { + "epoch": 4.31, + "learning_rate": 2.5703042246582092e-06, + "loss": 0.5221, + "step": 11679 + }, + { + "epoch": 4.31, + "learning_rate": 2.570267274294864e-06, + "loss": 0.6388, + "step": 11680 + }, + { + "epoch": 4.31, + "learning_rate": 2.5702303239315185e-06, + "loss": 0.6501, + "step": 11681 + }, + { + "epoch": 4.31, + "learning_rate": 2.5701933735681736e-06, + "loss": 0.5344, + "step": 11682 + }, + { + "epoch": 4.31, + "learning_rate": 2.5701564232048282e-06, + "loss": 0.5019, + "step": 11683 + }, + { + "epoch": 4.31, + "learning_rate": 2.5701194728414833e-06, + "loss": 0.4899, + "step": 11684 + }, + { + "epoch": 4.31, + "learning_rate": 2.5700825224781375e-06, + "loss": 0.6355, + "step": 11685 + }, + { + "epoch": 4.32, + "learning_rate": 2.5700455721147925e-06, + "loss": 0.5751, + "step": 11686 + }, + { + "epoch": 4.32, + "learning_rate": 2.570008621751447e-06, + "loss": 0.6153, + "step": 11687 + }, + { + "epoch": 4.32, + "learning_rate": 2.5699716713881022e-06, + "loss": 0.5452, + "step": 11688 + }, + { + "epoch": 4.32, + "learning_rate": 2.569934721024757e-06, + "loss": 0.4656, + "step": 11689 + }, + { + "epoch": 4.32, + "learning_rate": 2.5698977706614115e-06, + "loss": 0.3469, + "step": 11690 + }, + { + "epoch": 4.32, + "learning_rate": 2.569860820298066e-06, + "loss": 0.5298, + "step": 11691 + }, + { + "epoch": 4.32, + "learning_rate": 2.5698238699347208e-06, + "loss": 0.4025, + "step": 11692 + }, + { + "epoch": 4.32, + "learning_rate": 2.569786919571376e-06, + "loss": 0.5959, + "step": 11693 + }, + { + "epoch": 4.32, + "learning_rate": 2.5697499692080305e-06, + "loss": 0.4497, + "step": 11694 + }, + { + "epoch": 4.32, + "learning_rate": 2.5697130188446855e-06, + "loss": 0.4521, + "step": 11695 + }, + { + "epoch": 4.32, + "learning_rate": 2.56967606848134e-06, + "loss": 0.5296, + "step": 11696 + }, + { + "epoch": 4.32, + "learning_rate": 2.569639118117995e-06, + "loss": 0.5151, + "step": 11697 + }, + { + "epoch": 4.32, + "learning_rate": 2.5696021677546495e-06, + "loss": 0.4224, + "step": 11698 + }, + { + "epoch": 4.32, + "learning_rate": 2.5695652173913045e-06, + "loss": 0.6454, + "step": 11699 + }, + { + "epoch": 4.32, + "learning_rate": 2.569528267027959e-06, + "loss": 0.4836, + "step": 11700 + }, + { + "epoch": 4.32, + "learning_rate": 2.5694913166646138e-06, + "loss": 0.4433, + "step": 11701 + }, + { + "epoch": 4.32, + "learning_rate": 2.569454366301269e-06, + "loss": 0.464, + "step": 11702 + }, + { + "epoch": 4.32, + "learning_rate": 2.5694174159379235e-06, + "loss": 0.5128, + "step": 11703 + }, + { + "epoch": 4.32, + "learning_rate": 2.569380465574578e-06, + "loss": 0.574, + "step": 11704 + }, + { + "epoch": 4.32, + "learning_rate": 2.5693435152112328e-06, + "loss": 0.5055, + "step": 11705 + }, + { + "epoch": 4.32, + "learning_rate": 2.569306564847888e-06, + "loss": 0.5522, + "step": 11706 + }, + { + "epoch": 4.32, + "learning_rate": 2.5692696144845425e-06, + "loss": 0.5591, + "step": 11707 + }, + { + "epoch": 4.32, + "learning_rate": 2.5692326641211975e-06, + "loss": 0.5485, + "step": 11708 + }, + { + "epoch": 4.32, + "learning_rate": 2.569195713757852e-06, + "loss": 0.5604, + "step": 11709 + }, + { + "epoch": 4.32, + "learning_rate": 2.5691587633945068e-06, + "loss": 0.4558, + "step": 11710 + }, + { + "epoch": 4.32, + "learning_rate": 2.5691218130311614e-06, + "loss": 0.4815, + "step": 11711 + }, + { + "epoch": 4.32, + "learning_rate": 2.569084862667816e-06, + "loss": 0.6238, + "step": 11712 + }, + { + "epoch": 4.33, + "learning_rate": 2.569047912304471e-06, + "loss": 0.5844, + "step": 11713 + }, + { + "epoch": 4.33, + "learning_rate": 2.5690109619411258e-06, + "loss": 0.5209, + "step": 11714 + }, + { + "epoch": 4.33, + "learning_rate": 2.568974011577781e-06, + "loss": 0.5967, + "step": 11715 + }, + { + "epoch": 4.33, + "learning_rate": 2.5689370612144354e-06, + "loss": 0.5939, + "step": 11716 + }, + { + "epoch": 4.33, + "learning_rate": 2.56890011085109e-06, + "loss": 0.4921, + "step": 11717 + }, + { + "epoch": 4.33, + "learning_rate": 2.5688631604877447e-06, + "loss": 0.5056, + "step": 11718 + }, + { + "epoch": 4.33, + "learning_rate": 2.5688262101243998e-06, + "loss": 0.5057, + "step": 11719 + }, + { + "epoch": 4.33, + "learning_rate": 2.5687892597610544e-06, + "loss": 0.5623, + "step": 11720 + }, + { + "epoch": 4.33, + "learning_rate": 2.568752309397709e-06, + "loss": 0.4219, + "step": 11721 + }, + { + "epoch": 4.33, + "learning_rate": 2.568715359034364e-06, + "loss": 0.6205, + "step": 11722 + }, + { + "epoch": 4.33, + "learning_rate": 2.5686784086710188e-06, + "loss": 0.5414, + "step": 11723 + }, + { + "epoch": 4.33, + "learning_rate": 2.5686414583076734e-06, + "loss": 0.482, + "step": 11724 + }, + { + "epoch": 4.33, + "learning_rate": 2.568604507944328e-06, + "loss": 0.6168, + "step": 11725 + }, + { + "epoch": 4.33, + "learning_rate": 2.568567557580983e-06, + "loss": 0.6185, + "step": 11726 + }, + { + "epoch": 4.33, + "learning_rate": 2.5685306072176377e-06, + "loss": 0.6209, + "step": 11727 + }, + { + "epoch": 4.33, + "learning_rate": 2.5684936568542928e-06, + "loss": 0.5163, + "step": 11728 + }, + { + "epoch": 4.33, + "learning_rate": 2.5684567064909474e-06, + "loss": 0.5976, + "step": 11729 + }, + { + "epoch": 4.33, + "learning_rate": 2.5684197561276016e-06, + "loss": 0.5449, + "step": 11730 + }, + { + "epoch": 4.33, + "learning_rate": 2.5683828057642567e-06, + "loss": 0.4765, + "step": 11731 + }, + { + "epoch": 4.33, + "learning_rate": 2.5683458554009113e-06, + "loss": 0.5448, + "step": 11732 + }, + { + "epoch": 4.33, + "learning_rate": 2.5683089050375664e-06, + "loss": 0.4897, + "step": 11733 + }, + { + "epoch": 4.33, + "learning_rate": 2.568271954674221e-06, + "loss": 0.5156, + "step": 11734 + }, + { + "epoch": 4.33, + "learning_rate": 2.568235004310876e-06, + "loss": 0.4092, + "step": 11735 + }, + { + "epoch": 4.33, + "learning_rate": 2.5681980539475303e-06, + "loss": 0.5151, + "step": 11736 + }, + { + "epoch": 4.33, + "learning_rate": 2.5681611035841854e-06, + "loss": 0.4798, + "step": 11737 + }, + { + "epoch": 4.33, + "learning_rate": 2.56812415322084e-06, + "loss": 0.4957, + "step": 11738 + }, + { + "epoch": 4.33, + "learning_rate": 2.568087202857495e-06, + "loss": 0.5097, + "step": 11739 + }, + { + "epoch": 4.34, + "learning_rate": 2.5680502524941497e-06, + "loss": 0.5043, + "step": 11740 + }, + { + "epoch": 4.34, + "learning_rate": 2.5680133021308043e-06, + "loss": 0.4659, + "step": 11741 + }, + { + "epoch": 4.34, + "learning_rate": 2.567976351767459e-06, + "loss": 0.5243, + "step": 11742 + }, + { + "epoch": 4.34, + "learning_rate": 2.5679394014041136e-06, + "loss": 0.6143, + "step": 11743 + }, + { + "epoch": 4.34, + "learning_rate": 2.5679024510407687e-06, + "loss": 0.5229, + "step": 11744 + }, + { + "epoch": 4.34, + "learning_rate": 2.5678655006774233e-06, + "loss": 0.6449, + "step": 11745 + }, + { + "epoch": 4.34, + "learning_rate": 2.5678285503140784e-06, + "loss": 0.497, + "step": 11746 + }, + { + "epoch": 4.34, + "learning_rate": 2.567791599950733e-06, + "loss": 0.4719, + "step": 11747 + }, + { + "epoch": 4.34, + "learning_rate": 2.5677546495873876e-06, + "loss": 0.5918, + "step": 11748 + }, + { + "epoch": 4.34, + "learning_rate": 2.5677176992240423e-06, + "loss": 0.5717, + "step": 11749 + }, + { + "epoch": 4.34, + "learning_rate": 2.567680748860697e-06, + "loss": 0.6214, + "step": 11750 + }, + { + "epoch": 4.34, + "learning_rate": 2.567643798497352e-06, + "loss": 0.5484, + "step": 11751 + }, + { + "epoch": 4.34, + "learning_rate": 2.5676068481340066e-06, + "loss": 0.4391, + "step": 11752 + }, + { + "epoch": 4.34, + "learning_rate": 2.5675698977706617e-06, + "loss": 0.5736, + "step": 11753 + }, + { + "epoch": 4.34, + "learning_rate": 2.5675329474073163e-06, + "loss": 0.6666, + "step": 11754 + }, + { + "epoch": 4.34, + "learning_rate": 2.567495997043971e-06, + "loss": 0.4408, + "step": 11755 + }, + { + "epoch": 4.34, + "learning_rate": 2.5674590466806256e-06, + "loss": 0.5202, + "step": 11756 + }, + { + "epoch": 4.34, + "learning_rate": 2.5674220963172806e-06, + "loss": 0.6735, + "step": 11757 + }, + { + "epoch": 4.34, + "learning_rate": 2.5673851459539353e-06, + "loss": 0.4835, + "step": 11758 + }, + { + "epoch": 4.34, + "learning_rate": 2.5673481955905903e-06, + "loss": 0.6197, + "step": 11759 + }, + { + "epoch": 4.34, + "learning_rate": 2.567311245227245e-06, + "loss": 0.5241, + "step": 11760 + }, + { + "epoch": 4.34, + "learning_rate": 2.5672742948638996e-06, + "loss": 0.5375, + "step": 11761 + }, + { + "epoch": 4.34, + "learning_rate": 2.5672373445005542e-06, + "loss": 0.5494, + "step": 11762 + }, + { + "epoch": 4.34, + "learning_rate": 2.567200394137209e-06, + "loss": 0.4667, + "step": 11763 + }, + { + "epoch": 4.34, + "learning_rate": 2.567163443773864e-06, + "loss": 0.5436, + "step": 11764 + }, + { + "epoch": 4.34, + "learning_rate": 2.5671264934105186e-06, + "loss": 0.5108, + "step": 11765 + }, + { + "epoch": 4.34, + "learning_rate": 2.5670895430471736e-06, + "loss": 0.5099, + "step": 11766 + }, + { + "epoch": 4.35, + "learning_rate": 2.5670525926838283e-06, + "loss": 0.4966, + "step": 11767 + }, + { + "epoch": 4.35, + "learning_rate": 2.567015642320483e-06, + "loss": 0.4703, + "step": 11768 + }, + { + "epoch": 4.35, + "learning_rate": 2.5669786919571375e-06, + "loss": 0.4626, + "step": 11769 + }, + { + "epoch": 4.35, + "learning_rate": 2.566941741593792e-06, + "loss": 0.4585, + "step": 11770 + }, + { + "epoch": 4.35, + "learning_rate": 2.5669047912304472e-06, + "loss": 0.5296, + "step": 11771 + }, + { + "epoch": 4.35, + "learning_rate": 2.566867840867102e-06, + "loss": 0.5708, + "step": 11772 + }, + { + "epoch": 4.35, + "learning_rate": 2.566830890503757e-06, + "loss": 0.5815, + "step": 11773 + }, + { + "epoch": 4.35, + "learning_rate": 2.5667939401404116e-06, + "loss": 0.5458, + "step": 11774 + }, + { + "epoch": 4.35, + "learning_rate": 2.566756989777066e-06, + "loss": 0.5714, + "step": 11775 + }, + { + "epoch": 4.35, + "learning_rate": 2.566720039413721e-06, + "loss": 0.5571, + "step": 11776 + }, + { + "epoch": 4.35, + "learning_rate": 2.566683089050376e-06, + "loss": 0.4922, + "step": 11777 + }, + { + "epoch": 4.35, + "learning_rate": 2.5666461386870305e-06, + "loss": 0.5457, + "step": 11778 + }, + { + "epoch": 4.35, + "learning_rate": 2.5666091883236856e-06, + "loss": 0.4281, + "step": 11779 + }, + { + "epoch": 4.35, + "learning_rate": 2.5665722379603402e-06, + "loss": 0.5076, + "step": 11780 + }, + { + "epoch": 4.35, + "learning_rate": 2.5665352875969944e-06, + "loss": 0.5925, + "step": 11781 + }, + { + "epoch": 4.35, + "learning_rate": 2.5664983372336495e-06, + "loss": 0.5682, + "step": 11782 + }, + { + "epoch": 4.35, + "learning_rate": 2.566461386870304e-06, + "loss": 0.5534, + "step": 11783 + }, + { + "epoch": 4.35, + "learning_rate": 2.566424436506959e-06, + "loss": 0.5033, + "step": 11784 + }, + { + "epoch": 4.35, + "learning_rate": 2.566387486143614e-06, + "loss": 0.4655, + "step": 11785 + }, + { + "epoch": 4.35, + "learning_rate": 2.5663505357802685e-06, + "loss": 0.5272, + "step": 11786 + }, + { + "epoch": 4.35, + "learning_rate": 2.566313585416923e-06, + "loss": 0.397, + "step": 11787 + }, + { + "epoch": 4.35, + "learning_rate": 2.566276635053578e-06, + "loss": 0.5142, + "step": 11788 + }, + { + "epoch": 4.35, + "learning_rate": 2.566239684690233e-06, + "loss": 0.5135, + "step": 11789 + }, + { + "epoch": 4.35, + "learning_rate": 2.5662027343268874e-06, + "loss": 0.5207, + "step": 11790 + }, + { + "epoch": 4.35, + "learning_rate": 2.5661657839635425e-06, + "loss": 0.5237, + "step": 11791 + }, + { + "epoch": 4.35, + "learning_rate": 2.566128833600197e-06, + "loss": 0.5504, + "step": 11792 + }, + { + "epoch": 4.35, + "learning_rate": 2.5660918832368518e-06, + "loss": 0.4937, + "step": 11793 + }, + { + "epoch": 4.36, + "learning_rate": 2.5660549328735064e-06, + "loss": 0.5133, + "step": 11794 + }, + { + "epoch": 4.36, + "learning_rate": 2.5660179825101615e-06, + "loss": 0.488, + "step": 11795 + }, + { + "epoch": 4.36, + "learning_rate": 2.565981032146816e-06, + "loss": 0.5531, + "step": 11796 + }, + { + "epoch": 4.36, + "learning_rate": 2.565944081783471e-06, + "loss": 0.4535, + "step": 11797 + }, + { + "epoch": 4.36, + "learning_rate": 2.565907131420126e-06, + "loss": 0.5208, + "step": 11798 + }, + { + "epoch": 4.36, + "learning_rate": 2.5658701810567804e-06, + "loss": 0.561, + "step": 11799 + }, + { + "epoch": 4.36, + "learning_rate": 2.565833230693435e-06, + "loss": 0.581, + "step": 11800 + }, + { + "epoch": 4.36, + "learning_rate": 2.5657962803300897e-06, + "loss": 0.5929, + "step": 11801 + }, + { + "epoch": 4.36, + "learning_rate": 2.5657593299667448e-06, + "loss": 0.6097, + "step": 11802 + }, + { + "epoch": 4.36, + "learning_rate": 2.5657223796033994e-06, + "loss": 0.5641, + "step": 11803 + }, + { + "epoch": 4.36, + "learning_rate": 2.5656854292400545e-06, + "loss": 0.5622, + "step": 11804 + }, + { + "epoch": 4.36, + "learning_rate": 2.565648478876709e-06, + "loss": 0.4987, + "step": 11805 + }, + { + "epoch": 4.36, + "learning_rate": 2.5656115285133637e-06, + "loss": 0.5217, + "step": 11806 + }, + { + "epoch": 4.36, + "learning_rate": 2.5655745781500184e-06, + "loss": 0.5413, + "step": 11807 + }, + { + "epoch": 4.36, + "learning_rate": 2.5655376277866734e-06, + "loss": 0.5071, + "step": 11808 + }, + { + "epoch": 4.36, + "learning_rate": 2.565500677423328e-06, + "loss": 0.4938, + "step": 11809 + }, + { + "epoch": 4.36, + "learning_rate": 2.5654637270599827e-06, + "loss": 0.5783, + "step": 11810 + }, + { + "epoch": 4.36, + "learning_rate": 2.5654267766966378e-06, + "loss": 0.5133, + "step": 11811 + }, + { + "epoch": 4.36, + "learning_rate": 2.5653898263332924e-06, + "loss": 0.6035, + "step": 11812 + }, + { + "epoch": 4.36, + "learning_rate": 2.565352875969947e-06, + "loss": 0.5089, + "step": 11813 + }, + { + "epoch": 4.36, + "learning_rate": 2.5653159256066017e-06, + "loss": 0.5356, + "step": 11814 + }, + { + "epoch": 4.36, + "learning_rate": 2.5652789752432567e-06, + "loss": 0.4148, + "step": 11815 + }, + { + "epoch": 4.36, + "learning_rate": 2.5652420248799114e-06, + "loss": 0.4731, + "step": 11816 + }, + { + "epoch": 4.36, + "learning_rate": 2.5652050745165664e-06, + "loss": 0.5126, + "step": 11817 + }, + { + "epoch": 4.36, + "learning_rate": 2.565168124153221e-06, + "loss": 0.4096, + "step": 11818 + }, + { + "epoch": 4.36, + "learning_rate": 2.5651311737898757e-06, + "loss": 0.511, + "step": 11819 + }, + { + "epoch": 4.36, + "learning_rate": 2.5650942234265303e-06, + "loss": 0.5145, + "step": 11820 + }, + { + "epoch": 4.37, + "learning_rate": 2.565057273063185e-06, + "loss": 0.4607, + "step": 11821 + }, + { + "epoch": 4.37, + "learning_rate": 2.56502032269984e-06, + "loss": 0.4755, + "step": 11822 + }, + { + "epoch": 4.37, + "learning_rate": 2.5649833723364947e-06, + "loss": 0.4952, + "step": 11823 + }, + { + "epoch": 4.37, + "learning_rate": 2.5649464219731497e-06, + "loss": 0.5825, + "step": 11824 + }, + { + "epoch": 4.37, + "learning_rate": 2.564909471609804e-06, + "loss": 0.5421, + "step": 11825 + }, + { + "epoch": 4.37, + "learning_rate": 2.564872521246459e-06, + "loss": 0.4442, + "step": 11826 + }, + { + "epoch": 4.37, + "learning_rate": 2.5648355708831136e-06, + "loss": 0.5165, + "step": 11827 + }, + { + "epoch": 4.37, + "learning_rate": 2.5647986205197687e-06, + "loss": 0.5119, + "step": 11828 + }, + { + "epoch": 4.37, + "learning_rate": 2.5647616701564233e-06, + "loss": 0.4635, + "step": 11829 + }, + { + "epoch": 4.37, + "learning_rate": 2.564724719793078e-06, + "loss": 0.49, + "step": 11830 + }, + { + "epoch": 4.37, + "learning_rate": 2.5646877694297326e-06, + "loss": 0.4834, + "step": 11831 + }, + { + "epoch": 4.37, + "learning_rate": 2.5646508190663873e-06, + "loss": 0.4965, + "step": 11832 + }, + { + "epoch": 4.37, + "learning_rate": 2.5646138687030423e-06, + "loss": 0.6216, + "step": 11833 + }, + { + "epoch": 4.37, + "learning_rate": 2.564576918339697e-06, + "loss": 0.5755, + "step": 11834 + }, + { + "epoch": 4.37, + "learning_rate": 2.564539967976352e-06, + "loss": 0.4741, + "step": 11835 + }, + { + "epoch": 4.37, + "learning_rate": 2.5645030176130066e-06, + "loss": 0.474, + "step": 11836 + }, + { + "epoch": 4.37, + "learning_rate": 2.5644660672496613e-06, + "loss": 0.5546, + "step": 11837 + }, + { + "epoch": 4.37, + "learning_rate": 2.564429116886316e-06, + "loss": 0.5583, + "step": 11838 + }, + { + "epoch": 4.37, + "learning_rate": 2.564392166522971e-06, + "loss": 0.4594, + "step": 11839 + }, + { + "epoch": 4.37, + "learning_rate": 2.5643552161596256e-06, + "loss": 0.5744, + "step": 11840 + }, + { + "epoch": 4.37, + "learning_rate": 2.5643182657962803e-06, + "loss": 0.4357, + "step": 11841 + }, + { + "epoch": 4.37, + "learning_rate": 2.5642813154329353e-06, + "loss": 0.5161, + "step": 11842 + }, + { + "epoch": 4.37, + "learning_rate": 2.56424436506959e-06, + "loss": 0.4809, + "step": 11843 + }, + { + "epoch": 4.37, + "learning_rate": 2.5642074147062446e-06, + "loss": 0.5237, + "step": 11844 + }, + { + "epoch": 4.37, + "learning_rate": 2.5641704643428992e-06, + "loss": 0.5634, + "step": 11845 + }, + { + "epoch": 4.37, + "learning_rate": 2.5641335139795543e-06, + "loss": 0.5789, + "step": 11846 + }, + { + "epoch": 4.37, + "learning_rate": 2.564096563616209e-06, + "loss": 0.4316, + "step": 11847 + }, + { + "epoch": 4.38, + "learning_rate": 2.564059613252864e-06, + "loss": 0.4974, + "step": 11848 + }, + { + "epoch": 4.38, + "learning_rate": 2.5640226628895186e-06, + "loss": 0.6853, + "step": 11849 + }, + { + "epoch": 4.38, + "learning_rate": 2.5639857125261732e-06, + "loss": 0.5741, + "step": 11850 + }, + { + "epoch": 4.38, + "learning_rate": 2.563948762162828e-06, + "loss": 0.4437, + "step": 11851 + }, + { + "epoch": 4.38, + "learning_rate": 2.5639118117994825e-06, + "loss": 0.5608, + "step": 11852 + }, + { + "epoch": 4.38, + "learning_rate": 2.5638748614361376e-06, + "loss": 0.5803, + "step": 11853 + }, + { + "epoch": 4.38, + "learning_rate": 2.5638379110727922e-06, + "loss": 0.4573, + "step": 11854 + }, + { + "epoch": 4.38, + "learning_rate": 2.5638009607094473e-06, + "loss": 0.4878, + "step": 11855 + }, + { + "epoch": 4.38, + "learning_rate": 2.563764010346102e-06, + "loss": 0.4312, + "step": 11856 + }, + { + "epoch": 4.38, + "learning_rate": 2.5637270599827566e-06, + "loss": 0.4647, + "step": 11857 + }, + { + "epoch": 4.38, + "learning_rate": 2.563690109619411e-06, + "loss": 0.633, + "step": 11858 + }, + { + "epoch": 4.38, + "learning_rate": 2.5636531592560662e-06, + "loss": 0.5024, + "step": 11859 + }, + { + "epoch": 4.38, + "learning_rate": 2.563616208892721e-06, + "loss": 0.5648, + "step": 11860 + }, + { + "epoch": 4.38, + "learning_rate": 2.5635792585293755e-06, + "loss": 0.4834, + "step": 11861 + }, + { + "epoch": 4.38, + "learning_rate": 2.5635423081660306e-06, + "loss": 0.5238, + "step": 11862 + }, + { + "epoch": 4.38, + "learning_rate": 2.5635053578026852e-06, + "loss": 0.6211, + "step": 11863 + }, + { + "epoch": 4.38, + "learning_rate": 2.56346840743934e-06, + "loss": 0.592, + "step": 11864 + }, + { + "epoch": 4.38, + "learning_rate": 2.5634314570759945e-06, + "loss": 0.5571, + "step": 11865 + }, + { + "epoch": 4.38, + "learning_rate": 2.5633945067126495e-06, + "loss": 0.6193, + "step": 11866 + }, + { + "epoch": 4.38, + "learning_rate": 2.563357556349304e-06, + "loss": 0.388, + "step": 11867 + }, + { + "epoch": 4.38, + "learning_rate": 2.5633206059859592e-06, + "loss": 0.5947, + "step": 11868 + }, + { + "epoch": 4.38, + "learning_rate": 2.563283655622614e-06, + "loss": 0.595, + "step": 11869 + }, + { + "epoch": 4.38, + "learning_rate": 2.563246705259268e-06, + "loss": 0.5485, + "step": 11870 + }, + { + "epoch": 4.38, + "learning_rate": 2.563209754895923e-06, + "loss": 0.6071, + "step": 11871 + }, + { + "epoch": 4.38, + "learning_rate": 2.563172804532578e-06, + "loss": 0.6364, + "step": 11872 + }, + { + "epoch": 4.38, + "learning_rate": 2.563135854169233e-06, + "loss": 0.5247, + "step": 11873 + }, + { + "epoch": 4.38, + "learning_rate": 2.5630989038058875e-06, + "loss": 0.4628, + "step": 11874 + }, + { + "epoch": 4.39, + "learning_rate": 2.5630619534425425e-06, + "loss": 0.5684, + "step": 11875 + }, + { + "epoch": 4.39, + "learning_rate": 2.5630250030791968e-06, + "loss": 0.5314, + "step": 11876 + }, + { + "epoch": 4.39, + "learning_rate": 2.562988052715852e-06, + "loss": 0.5973, + "step": 11877 + }, + { + "epoch": 4.39, + "learning_rate": 2.5629511023525065e-06, + "loss": 0.499, + "step": 11878 + }, + { + "epoch": 4.39, + "learning_rate": 2.5629141519891615e-06, + "loss": 0.5303, + "step": 11879 + }, + { + "epoch": 4.39, + "learning_rate": 2.562877201625816e-06, + "loss": 0.5485, + "step": 11880 + }, + { + "epoch": 4.39, + "learning_rate": 2.562840251262471e-06, + "loss": 0.4777, + "step": 11881 + }, + { + "epoch": 4.39, + "learning_rate": 2.5628033008991254e-06, + "loss": 0.5282, + "step": 11882 + }, + { + "epoch": 4.39, + "learning_rate": 2.56276635053578e-06, + "loss": 0.4862, + "step": 11883 + }, + { + "epoch": 4.39, + "learning_rate": 2.562729400172435e-06, + "loss": 0.5671, + "step": 11884 + }, + { + "epoch": 4.39, + "learning_rate": 2.5626924498090898e-06, + "loss": 0.4709, + "step": 11885 + }, + { + "epoch": 4.39, + "learning_rate": 2.562655499445745e-06, + "loss": 0.4825, + "step": 11886 + }, + { + "epoch": 4.39, + "learning_rate": 2.5626185490823995e-06, + "loss": 0.4889, + "step": 11887 + }, + { + "epoch": 4.39, + "learning_rate": 2.562581598719054e-06, + "loss": 0.4538, + "step": 11888 + }, + { + "epoch": 4.39, + "learning_rate": 2.5625446483557087e-06, + "loss": 0.4625, + "step": 11889 + }, + { + "epoch": 4.39, + "learning_rate": 2.5625076979923634e-06, + "loss": 0.6611, + "step": 11890 + }, + { + "epoch": 4.39, + "learning_rate": 2.5624707476290184e-06, + "loss": 0.5166, + "step": 11891 + }, + { + "epoch": 4.39, + "learning_rate": 2.562433797265673e-06, + "loss": 0.5528, + "step": 11892 + }, + { + "epoch": 4.39, + "learning_rate": 2.562396846902328e-06, + "loss": 0.5537, + "step": 11893 + }, + { + "epoch": 4.39, + "learning_rate": 2.5623598965389828e-06, + "loss": 0.6292, + "step": 11894 + }, + { + "epoch": 4.39, + "learning_rate": 2.5623229461756374e-06, + "loss": 0.5287, + "step": 11895 + }, + { + "epoch": 4.39, + "learning_rate": 2.562285995812292e-06, + "loss": 0.5015, + "step": 11896 + }, + { + "epoch": 4.39, + "learning_rate": 2.562249045448947e-06, + "loss": 0.5794, + "step": 11897 + }, + { + "epoch": 4.39, + "learning_rate": 2.5622120950856017e-06, + "loss": 0.4218, + "step": 11898 + }, + { + "epoch": 4.39, + "learning_rate": 2.5621751447222564e-06, + "loss": 0.5612, + "step": 11899 + }, + { + "epoch": 4.39, + "learning_rate": 2.5621381943589114e-06, + "loss": 0.5151, + "step": 11900 + }, + { + "epoch": 4.39, + "learning_rate": 2.562101243995566e-06, + "loss": 0.4812, + "step": 11901 + }, + { + "epoch": 4.4, + "learning_rate": 2.5620642936322207e-06, + "loss": 0.5215, + "step": 11902 + }, + { + "epoch": 4.4, + "learning_rate": 2.5620273432688753e-06, + "loss": 0.5434, + "step": 11903 + }, + { + "epoch": 4.4, + "learning_rate": 2.5619903929055304e-06, + "loss": 0.5945, + "step": 11904 + }, + { + "epoch": 4.4, + "learning_rate": 2.561953442542185e-06, + "loss": 0.5508, + "step": 11905 + }, + { + "epoch": 4.4, + "learning_rate": 2.56191649217884e-06, + "loss": 0.5535, + "step": 11906 + }, + { + "epoch": 4.4, + "learning_rate": 2.5618795418154947e-06, + "loss": 0.5407, + "step": 11907 + }, + { + "epoch": 4.4, + "learning_rate": 2.5618425914521494e-06, + "loss": 0.6304, + "step": 11908 + }, + { + "epoch": 4.4, + "learning_rate": 2.561805641088804e-06, + "loss": 0.5339, + "step": 11909 + }, + { + "epoch": 4.4, + "learning_rate": 2.5617686907254586e-06, + "loss": 0.5969, + "step": 11910 + }, + { + "epoch": 4.4, + "learning_rate": 2.5617317403621137e-06, + "loss": 0.5634, + "step": 11911 + }, + { + "epoch": 4.4, + "learning_rate": 2.5616947899987683e-06, + "loss": 0.4544, + "step": 11912 + }, + { + "epoch": 4.4, + "learning_rate": 2.5616578396354234e-06, + "loss": 0.5722, + "step": 11913 + }, + { + "epoch": 4.4, + "learning_rate": 2.561620889272078e-06, + "loss": 0.54, + "step": 11914 + }, + { + "epoch": 4.4, + "learning_rate": 2.5615839389087327e-06, + "loss": 0.5222, + "step": 11915 + }, + { + "epoch": 4.4, + "learning_rate": 2.5615469885453873e-06, + "loss": 0.5218, + "step": 11916 + }, + { + "epoch": 4.4, + "learning_rate": 2.5615100381820424e-06, + "loss": 0.4376, + "step": 11917 + }, + { + "epoch": 4.4, + "learning_rate": 2.561473087818697e-06, + "loss": 0.5861, + "step": 11918 + }, + { + "epoch": 4.4, + "learning_rate": 2.5614361374553516e-06, + "loss": 0.5222, + "step": 11919 + }, + { + "epoch": 4.4, + "learning_rate": 2.5613991870920067e-06, + "loss": 0.4187, + "step": 11920 + }, + { + "epoch": 4.4, + "learning_rate": 2.561362236728661e-06, + "loss": 0.5851, + "step": 11921 + }, + { + "epoch": 4.4, + "learning_rate": 2.561325286365316e-06, + "loss": 0.4494, + "step": 11922 + }, + { + "epoch": 4.4, + "learning_rate": 2.5612883360019706e-06, + "loss": 0.4658, + "step": 11923 + }, + { + "epoch": 4.4, + "learning_rate": 2.5612513856386257e-06, + "loss": 0.4429, + "step": 11924 + }, + { + "epoch": 4.4, + "learning_rate": 2.5612144352752803e-06, + "loss": 0.4985, + "step": 11925 + }, + { + "epoch": 4.4, + "learning_rate": 2.5611774849119354e-06, + "loss": 0.4985, + "step": 11926 + }, + { + "epoch": 4.4, + "learning_rate": 2.5611405345485896e-06, + "loss": 0.583, + "step": 11927 + }, + { + "epoch": 4.4, + "learning_rate": 2.5611035841852446e-06, + "loss": 0.6984, + "step": 11928 + }, + { + "epoch": 4.41, + "learning_rate": 2.5610666338218993e-06, + "loss": 0.4973, + "step": 11929 + }, + { + "epoch": 4.41, + "learning_rate": 2.561029683458554e-06, + "loss": 0.3918, + "step": 11930 + }, + { + "epoch": 4.41, + "learning_rate": 2.560992733095209e-06, + "loss": 0.4745, + "step": 11931 + }, + { + "epoch": 4.41, + "learning_rate": 2.5609557827318636e-06, + "loss": 0.5482, + "step": 11932 + }, + { + "epoch": 4.41, + "learning_rate": 2.5609188323685182e-06, + "loss": 0.6141, + "step": 11933 + }, + { + "epoch": 4.41, + "learning_rate": 2.560881882005173e-06, + "loss": 0.5599, + "step": 11934 + }, + { + "epoch": 4.41, + "learning_rate": 2.560844931641828e-06, + "loss": 0.5083, + "step": 11935 + }, + { + "epoch": 4.41, + "learning_rate": 2.5608079812784826e-06, + "loss": 0.4881, + "step": 11936 + }, + { + "epoch": 4.41, + "learning_rate": 2.5607710309151376e-06, + "loss": 0.6073, + "step": 11937 + }, + { + "epoch": 4.41, + "learning_rate": 2.5607340805517923e-06, + "loss": 0.4182, + "step": 11938 + }, + { + "epoch": 4.41, + "learning_rate": 2.560697130188447e-06, + "loss": 0.5406, + "step": 11939 + }, + { + "epoch": 4.41, + "learning_rate": 2.5606601798251015e-06, + "loss": 0.498, + "step": 11940 + }, + { + "epoch": 4.41, + "learning_rate": 2.560623229461756e-06, + "loss": 0.5487, + "step": 11941 + }, + { + "epoch": 4.41, + "learning_rate": 2.5605862790984112e-06, + "loss": 0.4294, + "step": 11942 + }, + { + "epoch": 4.41, + "learning_rate": 2.560549328735066e-06, + "loss": 0.5966, + "step": 11943 + }, + { + "epoch": 4.41, + "learning_rate": 2.560512378371721e-06, + "loss": 0.4486, + "step": 11944 + }, + { + "epoch": 4.41, + "learning_rate": 2.5604754280083756e-06, + "loss": 0.4858, + "step": 11945 + }, + { + "epoch": 4.41, + "learning_rate": 2.56043847764503e-06, + "loss": 0.534, + "step": 11946 + }, + { + "epoch": 4.41, + "learning_rate": 2.560401527281685e-06, + "loss": 0.6897, + "step": 11947 + }, + { + "epoch": 4.41, + "learning_rate": 2.56036457691834e-06, + "loss": 0.4989, + "step": 11948 + }, + { + "epoch": 4.41, + "learning_rate": 2.5603276265549945e-06, + "loss": 0.5224, + "step": 11949 + }, + { + "epoch": 4.41, + "learning_rate": 2.560290676191649e-06, + "loss": 0.5486, + "step": 11950 + }, + { + "epoch": 4.41, + "learning_rate": 2.5602537258283042e-06, + "loss": 0.5668, + "step": 11951 + }, + { + "epoch": 4.41, + "learning_rate": 2.560216775464959e-06, + "loss": 0.5639, + "step": 11952 + }, + { + "epoch": 4.41, + "learning_rate": 2.5601798251016135e-06, + "loss": 0.5136, + "step": 11953 + }, + { + "epoch": 4.41, + "learning_rate": 2.560142874738268e-06, + "loss": 0.6316, + "step": 11954 + }, + { + "epoch": 4.41, + "learning_rate": 2.560105924374923e-06, + "loss": 0.4461, + "step": 11955 + }, + { + "epoch": 4.42, + "learning_rate": 2.560068974011578e-06, + "loss": 0.5613, + "step": 11956 + }, + { + "epoch": 4.42, + "learning_rate": 2.560032023648233e-06, + "loss": 0.6311, + "step": 11957 + }, + { + "epoch": 4.42, + "learning_rate": 2.5599950732848875e-06, + "loss": 0.3781, + "step": 11958 + }, + { + "epoch": 4.42, + "learning_rate": 2.559958122921542e-06, + "loss": 0.5269, + "step": 11959 + }, + { + "epoch": 4.42, + "learning_rate": 2.559921172558197e-06, + "loss": 0.5456, + "step": 11960 + }, + { + "epoch": 4.42, + "learning_rate": 2.5598842221948514e-06, + "loss": 0.563, + "step": 11961 + }, + { + "epoch": 4.42, + "learning_rate": 2.5598472718315065e-06, + "loss": 0.6038, + "step": 11962 + }, + { + "epoch": 4.42, + "learning_rate": 2.559810321468161e-06, + "loss": 0.5185, + "step": 11963 + }, + { + "epoch": 4.42, + "learning_rate": 2.559773371104816e-06, + "loss": 0.4846, + "step": 11964 + }, + { + "epoch": 4.42, + "learning_rate": 2.559736420741471e-06, + "loss": 0.5515, + "step": 11965 + }, + { + "epoch": 4.42, + "learning_rate": 2.5596994703781255e-06, + "loss": 0.4682, + "step": 11966 + }, + { + "epoch": 4.42, + "learning_rate": 2.55966252001478e-06, + "loss": 0.5373, + "step": 11967 + }, + { + "epoch": 4.42, + "learning_rate": 2.559625569651435e-06, + "loss": 0.454, + "step": 11968 + }, + { + "epoch": 4.42, + "learning_rate": 2.55958861928809e-06, + "loss": 0.5941, + "step": 11969 + }, + { + "epoch": 4.42, + "learning_rate": 2.5595516689247444e-06, + "loss": 0.4691, + "step": 11970 + }, + { + "epoch": 4.42, + "learning_rate": 2.559514718561399e-06, + "loss": 0.5336, + "step": 11971 + }, + { + "epoch": 4.42, + "learning_rate": 2.5594777681980537e-06, + "loss": 0.524, + "step": 11972 + }, + { + "epoch": 4.42, + "learning_rate": 2.5594408178347088e-06, + "loss": 0.5234, + "step": 11973 + }, + { + "epoch": 4.42, + "learning_rate": 2.5594038674713634e-06, + "loss": 0.5252, + "step": 11974 + }, + { + "epoch": 4.42, + "learning_rate": 2.5593669171080185e-06, + "loss": 0.6607, + "step": 11975 + }, + { + "epoch": 4.42, + "learning_rate": 2.559329966744673e-06, + "loss": 0.5144, + "step": 11976 + }, + { + "epoch": 4.42, + "learning_rate": 2.5592930163813277e-06, + "loss": 0.4671, + "step": 11977 + }, + { + "epoch": 4.42, + "learning_rate": 2.5592560660179824e-06, + "loss": 0.5235, + "step": 11978 + }, + { + "epoch": 4.42, + "learning_rate": 2.559219115654637e-06, + "loss": 0.4075, + "step": 11979 + }, + { + "epoch": 4.42, + "learning_rate": 2.559182165291292e-06, + "loss": 0.46, + "step": 11980 + }, + { + "epoch": 4.42, + "learning_rate": 2.5591452149279467e-06, + "loss": 0.5207, + "step": 11981 + }, + { + "epoch": 4.42, + "learning_rate": 2.5591082645646018e-06, + "loss": 0.5824, + "step": 11982 + }, + { + "epoch": 4.43, + "learning_rate": 2.5590713142012564e-06, + "loss": 0.6286, + "step": 11983 + }, + { + "epoch": 4.43, + "learning_rate": 2.559034363837911e-06, + "loss": 0.5433, + "step": 11984 + }, + { + "epoch": 4.43, + "learning_rate": 2.5589974134745657e-06, + "loss": 0.5107, + "step": 11985 + }, + { + "epoch": 4.43, + "learning_rate": 2.5589604631112207e-06, + "loss": 0.494, + "step": 11986 + }, + { + "epoch": 4.43, + "learning_rate": 2.5589235127478754e-06, + "loss": 0.5047, + "step": 11987 + }, + { + "epoch": 4.43, + "learning_rate": 2.5588865623845304e-06, + "loss": 0.4499, + "step": 11988 + }, + { + "epoch": 4.43, + "learning_rate": 2.558849612021185e-06, + "loss": 0.5949, + "step": 11989 + }, + { + "epoch": 4.43, + "learning_rate": 2.5588126616578397e-06, + "loss": 0.5644, + "step": 11990 + }, + { + "epoch": 4.43, + "learning_rate": 2.5587757112944944e-06, + "loss": 0.5569, + "step": 11991 + }, + { + "epoch": 4.43, + "learning_rate": 2.558738760931149e-06, + "loss": 0.5039, + "step": 11992 + }, + { + "epoch": 4.43, + "learning_rate": 2.558701810567804e-06, + "loss": 0.4872, + "step": 11993 + }, + { + "epoch": 4.43, + "learning_rate": 2.5586648602044587e-06, + "loss": 0.5793, + "step": 11994 + }, + { + "epoch": 4.43, + "learning_rate": 2.5586279098411137e-06, + "loss": 0.5607, + "step": 11995 + }, + { + "epoch": 4.43, + "learning_rate": 2.5585909594777684e-06, + "loss": 0.5428, + "step": 11996 + }, + { + "epoch": 4.43, + "learning_rate": 2.558554009114423e-06, + "loss": 0.5231, + "step": 11997 + }, + { + "epoch": 4.43, + "learning_rate": 2.5585170587510777e-06, + "loss": 0.5037, + "step": 11998 + }, + { + "epoch": 4.43, + "learning_rate": 2.5584801083877323e-06, + "loss": 0.4684, + "step": 11999 + }, + { + "epoch": 4.43, + "learning_rate": 2.5584431580243873e-06, + "loss": 0.485, + "step": 12000 + }, + { + "epoch": 4.43, + "learning_rate": 2.558406207661042e-06, + "loss": 0.5712, + "step": 12001 + }, + { + "epoch": 4.43, + "learning_rate": 2.558369257297697e-06, + "loss": 0.5208, + "step": 12002 + }, + { + "epoch": 4.43, + "learning_rate": 2.5583323069343517e-06, + "loss": 0.4911, + "step": 12003 + }, + { + "epoch": 4.43, + "learning_rate": 2.5582953565710063e-06, + "loss": 0.5986, + "step": 12004 + }, + { + "epoch": 4.43, + "learning_rate": 2.558258406207661e-06, + "loss": 0.5007, + "step": 12005 + }, + { + "epoch": 4.43, + "learning_rate": 2.558221455844316e-06, + "loss": 0.7133, + "step": 12006 + }, + { + "epoch": 4.43, + "learning_rate": 2.5581845054809707e-06, + "loss": 0.5912, + "step": 12007 + }, + { + "epoch": 4.43, + "learning_rate": 2.5581475551176257e-06, + "loss": 0.5125, + "step": 12008 + }, + { + "epoch": 4.43, + "learning_rate": 2.5581106047542803e-06, + "loss": 0.602, + "step": 12009 + }, + { + "epoch": 4.44, + "learning_rate": 2.5580736543909346e-06, + "loss": 0.4832, + "step": 12010 + }, + { + "epoch": 4.44, + "learning_rate": 2.5580367040275896e-06, + "loss": 0.5533, + "step": 12011 + }, + { + "epoch": 4.44, + "learning_rate": 2.5579997536642443e-06, + "loss": 0.5704, + "step": 12012 + }, + { + "epoch": 4.44, + "learning_rate": 2.5579628033008993e-06, + "loss": 0.5075, + "step": 12013 + }, + { + "epoch": 4.44, + "learning_rate": 2.557925852937554e-06, + "loss": 0.4698, + "step": 12014 + }, + { + "epoch": 4.44, + "learning_rate": 2.557888902574209e-06, + "loss": 0.5182, + "step": 12015 + }, + { + "epoch": 4.44, + "learning_rate": 2.5578519522108632e-06, + "loss": 0.5154, + "step": 12016 + }, + { + "epoch": 4.44, + "learning_rate": 2.5578150018475183e-06, + "loss": 0.4924, + "step": 12017 + }, + { + "epoch": 4.44, + "learning_rate": 2.557778051484173e-06, + "loss": 0.4934, + "step": 12018 + }, + { + "epoch": 4.44, + "learning_rate": 2.5577411011208276e-06, + "loss": 0.5852, + "step": 12019 + }, + { + "epoch": 4.44, + "learning_rate": 2.5577041507574826e-06, + "loss": 0.4409, + "step": 12020 + }, + { + "epoch": 4.44, + "learning_rate": 2.5576672003941373e-06, + "loss": 0.4304, + "step": 12021 + }, + { + "epoch": 4.44, + "learning_rate": 2.557630250030792e-06, + "loss": 0.5054, + "step": 12022 + }, + { + "epoch": 4.44, + "learning_rate": 2.5575932996674465e-06, + "loss": 0.5253, + "step": 12023 + }, + { + "epoch": 4.44, + "learning_rate": 2.5575563493041016e-06, + "loss": 0.4429, + "step": 12024 + }, + { + "epoch": 4.44, + "learning_rate": 2.5575193989407562e-06, + "loss": 0.4802, + "step": 12025 + }, + { + "epoch": 4.44, + "learning_rate": 2.5574824485774113e-06, + "loss": 0.5739, + "step": 12026 + }, + { + "epoch": 4.44, + "learning_rate": 2.557445498214066e-06, + "loss": 0.5116, + "step": 12027 + }, + { + "epoch": 4.44, + "learning_rate": 2.5574085478507206e-06, + "loss": 0.6074, + "step": 12028 + }, + { + "epoch": 4.44, + "learning_rate": 2.557371597487375e-06, + "loss": 0.532, + "step": 12029 + }, + { + "epoch": 4.44, + "learning_rate": 2.55733464712403e-06, + "loss": 0.5137, + "step": 12030 + }, + { + "epoch": 4.44, + "learning_rate": 2.557297696760685e-06, + "loss": 0.5236, + "step": 12031 + }, + { + "epoch": 4.44, + "learning_rate": 2.5572607463973395e-06, + "loss": 0.6189, + "step": 12032 + }, + { + "epoch": 4.44, + "learning_rate": 2.5572237960339946e-06, + "loss": 0.5238, + "step": 12033 + }, + { + "epoch": 4.44, + "learning_rate": 2.5571868456706492e-06, + "loss": 0.4309, + "step": 12034 + }, + { + "epoch": 4.44, + "learning_rate": 2.557149895307304e-06, + "loss": 0.4984, + "step": 12035 + }, + { + "epoch": 4.44, + "learning_rate": 2.5571129449439585e-06, + "loss": 0.4683, + "step": 12036 + }, + { + "epoch": 4.44, + "learning_rate": 2.5570759945806136e-06, + "loss": 0.5974, + "step": 12037 + }, + { + "epoch": 4.45, + "learning_rate": 2.557039044217268e-06, + "loss": 0.5131, + "step": 12038 + }, + { + "epoch": 4.45, + "learning_rate": 2.557002093853923e-06, + "loss": 0.6305, + "step": 12039 + }, + { + "epoch": 4.45, + "learning_rate": 2.556965143490578e-06, + "loss": 0.5086, + "step": 12040 + }, + { + "epoch": 4.45, + "learning_rate": 2.5569281931272325e-06, + "loss": 0.6074, + "step": 12041 + }, + { + "epoch": 4.45, + "learning_rate": 2.556891242763887e-06, + "loss": 0.5548, + "step": 12042 + }, + { + "epoch": 4.45, + "learning_rate": 2.556854292400542e-06, + "loss": 0.5016, + "step": 12043 + }, + { + "epoch": 4.45, + "learning_rate": 2.556817342037197e-06, + "loss": 0.5451, + "step": 12044 + }, + { + "epoch": 4.45, + "learning_rate": 2.5567803916738515e-06, + "loss": 0.6119, + "step": 12045 + }, + { + "epoch": 4.45, + "learning_rate": 2.5567434413105066e-06, + "loss": 0.4604, + "step": 12046 + }, + { + "epoch": 4.45, + "learning_rate": 2.556706490947161e-06, + "loss": 0.4772, + "step": 12047 + }, + { + "epoch": 4.45, + "learning_rate": 2.556669540583816e-06, + "loss": 0.4691, + "step": 12048 + }, + { + "epoch": 4.45, + "learning_rate": 2.5566325902204705e-06, + "loss": 0.5088, + "step": 12049 + }, + { + "epoch": 4.45, + "learning_rate": 2.556595639857125e-06, + "loss": 0.5761, + "step": 12050 + }, + { + "epoch": 4.45, + "learning_rate": 2.55655868949378e-06, + "loss": 0.4883, + "step": 12051 + }, + { + "epoch": 4.45, + "learning_rate": 2.556521739130435e-06, + "loss": 0.525, + "step": 12052 + }, + { + "epoch": 4.45, + "learning_rate": 2.55648478876709e-06, + "loss": 0.5191, + "step": 12053 + }, + { + "epoch": 4.45, + "learning_rate": 2.5564478384037445e-06, + "loss": 0.5211, + "step": 12054 + }, + { + "epoch": 4.45, + "learning_rate": 2.556410888040399e-06, + "loss": 0.4942, + "step": 12055 + }, + { + "epoch": 4.45, + "learning_rate": 2.5563739376770538e-06, + "loss": 0.4761, + "step": 12056 + }, + { + "epoch": 4.45, + "learning_rate": 2.556336987313709e-06, + "loss": 0.4874, + "step": 12057 + }, + { + "epoch": 4.45, + "learning_rate": 2.5563000369503635e-06, + "loss": 0.5417, + "step": 12058 + }, + { + "epoch": 4.45, + "learning_rate": 2.556263086587018e-06, + "loss": 0.4527, + "step": 12059 + }, + { + "epoch": 4.45, + "learning_rate": 2.556226136223673e-06, + "loss": 0.5263, + "step": 12060 + }, + { + "epoch": 4.45, + "learning_rate": 2.5561891858603274e-06, + "loss": 0.534, + "step": 12061 + }, + { + "epoch": 4.45, + "learning_rate": 2.5561522354969824e-06, + "loss": 0.4165, + "step": 12062 + }, + { + "epoch": 4.45, + "learning_rate": 2.556115285133637e-06, + "loss": 0.4989, + "step": 12063 + }, + { + "epoch": 4.45, + "learning_rate": 2.556078334770292e-06, + "loss": 0.4322, + "step": 12064 + }, + { + "epoch": 4.46, + "learning_rate": 2.5560413844069468e-06, + "loss": 0.6851, + "step": 12065 + }, + { + "epoch": 4.46, + "learning_rate": 2.556004434043602e-06, + "loss": 0.6199, + "step": 12066 + }, + { + "epoch": 4.46, + "learning_rate": 2.555967483680256e-06, + "loss": 0.6309, + "step": 12067 + }, + { + "epoch": 4.46, + "learning_rate": 2.555930533316911e-06, + "loss": 0.6311, + "step": 12068 + }, + { + "epoch": 4.46, + "learning_rate": 2.5558935829535657e-06, + "loss": 0.487, + "step": 12069 + }, + { + "epoch": 4.46, + "learning_rate": 2.5558566325902204e-06, + "loss": 0.5266, + "step": 12070 + }, + { + "epoch": 4.46, + "learning_rate": 2.5558196822268754e-06, + "loss": 0.4797, + "step": 12071 + }, + { + "epoch": 4.46, + "learning_rate": 2.55578273186353e-06, + "loss": 0.5995, + "step": 12072 + }, + { + "epoch": 4.46, + "learning_rate": 2.5557457815001847e-06, + "loss": 0.4464, + "step": 12073 + }, + { + "epoch": 4.46, + "learning_rate": 2.5557088311368393e-06, + "loss": 0.4435, + "step": 12074 + }, + { + "epoch": 4.46, + "learning_rate": 2.5556718807734944e-06, + "loss": 0.4291, + "step": 12075 + }, + { + "epoch": 4.46, + "learning_rate": 2.555634930410149e-06, + "loss": 0.4794, + "step": 12076 + }, + { + "epoch": 4.46, + "learning_rate": 2.555597980046804e-06, + "loss": 0.585, + "step": 12077 + }, + { + "epoch": 4.46, + "learning_rate": 2.5555610296834587e-06, + "loss": 0.4962, + "step": 12078 + }, + { + "epoch": 4.46, + "learning_rate": 2.5555240793201134e-06, + "loss": 0.4005, + "step": 12079 + }, + { + "epoch": 4.46, + "learning_rate": 2.555487128956768e-06, + "loss": 0.4662, + "step": 12080 + }, + { + "epoch": 4.46, + "learning_rate": 2.5554501785934226e-06, + "loss": 0.5095, + "step": 12081 + }, + { + "epoch": 4.46, + "learning_rate": 2.5554132282300777e-06, + "loss": 0.5261, + "step": 12082 + }, + { + "epoch": 4.46, + "learning_rate": 2.5553762778667323e-06, + "loss": 0.4588, + "step": 12083 + }, + { + "epoch": 4.46, + "learning_rate": 2.5553393275033874e-06, + "loss": 0.6301, + "step": 12084 + }, + { + "epoch": 4.46, + "learning_rate": 2.555302377140042e-06, + "loss": 0.5007, + "step": 12085 + }, + { + "epoch": 4.46, + "learning_rate": 2.5552654267766967e-06, + "loss": 0.4912, + "step": 12086 + }, + { + "epoch": 4.46, + "learning_rate": 2.5552284764133513e-06, + "loss": 0.4193, + "step": 12087 + }, + { + "epoch": 4.46, + "learning_rate": 2.5551915260500064e-06, + "loss": 0.592, + "step": 12088 + }, + { + "epoch": 4.46, + "learning_rate": 2.555154575686661e-06, + "loss": 0.5419, + "step": 12089 + }, + { + "epoch": 4.46, + "learning_rate": 2.5551176253233156e-06, + "loss": 0.4798, + "step": 12090 + }, + { + "epoch": 4.46, + "learning_rate": 2.5550806749599707e-06, + "loss": 0.4698, + "step": 12091 + }, + { + "epoch": 4.47, + "learning_rate": 2.5550437245966253e-06, + "loss": 0.4177, + "step": 12092 + }, + { + "epoch": 4.47, + "learning_rate": 2.55500677423328e-06, + "loss": 0.5597, + "step": 12093 + }, + { + "epoch": 4.47, + "learning_rate": 2.5549698238699346e-06, + "loss": 0.5665, + "step": 12094 + }, + { + "epoch": 4.47, + "learning_rate": 2.5549328735065897e-06, + "loss": 0.4809, + "step": 12095 + }, + { + "epoch": 4.47, + "learning_rate": 2.5548959231432443e-06, + "loss": 0.5575, + "step": 12096 + }, + { + "epoch": 4.47, + "learning_rate": 2.5548589727798994e-06, + "loss": 0.5538, + "step": 12097 + }, + { + "epoch": 4.47, + "learning_rate": 2.554822022416554e-06, + "loss": 0.4585, + "step": 12098 + }, + { + "epoch": 4.47, + "learning_rate": 2.5547850720532086e-06, + "loss": 0.5215, + "step": 12099 + }, + { + "epoch": 4.47, + "learning_rate": 2.5547481216898633e-06, + "loss": 0.5606, + "step": 12100 + }, + { + "epoch": 4.47, + "learning_rate": 2.554711171326518e-06, + "loss": 0.5131, + "step": 12101 + }, + { + "epoch": 4.47, + "learning_rate": 2.554674220963173e-06, + "loss": 0.457, + "step": 12102 + }, + { + "epoch": 4.47, + "learning_rate": 2.5546372705998276e-06, + "loss": 0.5408, + "step": 12103 + }, + { + "epoch": 4.47, + "learning_rate": 2.5546003202364827e-06, + "loss": 0.501, + "step": 12104 + }, + { + "epoch": 4.47, + "learning_rate": 2.5545633698731373e-06, + "loss": 0.5873, + "step": 12105 + }, + { + "epoch": 4.47, + "learning_rate": 2.554526419509792e-06, + "loss": 0.4893, + "step": 12106 + }, + { + "epoch": 4.47, + "learning_rate": 2.5544894691464466e-06, + "loss": 0.583, + "step": 12107 + }, + { + "epoch": 4.47, + "learning_rate": 2.5544525187831016e-06, + "loss": 0.5691, + "step": 12108 + }, + { + "epoch": 4.47, + "learning_rate": 2.5544155684197563e-06, + "loss": 0.5522, + "step": 12109 + }, + { + "epoch": 4.47, + "learning_rate": 2.554378618056411e-06, + "loss": 0.4531, + "step": 12110 + }, + { + "epoch": 4.47, + "learning_rate": 2.554341667693066e-06, + "loss": 0.5427, + "step": 12111 + }, + { + "epoch": 4.47, + "learning_rate": 2.55430471732972e-06, + "loss": 0.6117, + "step": 12112 + }, + { + "epoch": 4.47, + "learning_rate": 2.5542677669663752e-06, + "loss": 0.435, + "step": 12113 + }, + { + "epoch": 4.47, + "learning_rate": 2.55423081660303e-06, + "loss": 0.5513, + "step": 12114 + }, + { + "epoch": 4.47, + "learning_rate": 2.554193866239685e-06, + "loss": 0.4578, + "step": 12115 + }, + { + "epoch": 4.47, + "learning_rate": 2.5541569158763396e-06, + "loss": 0.4753, + "step": 12116 + }, + { + "epoch": 4.47, + "learning_rate": 2.5541199655129946e-06, + "loss": 0.5313, + "step": 12117 + }, + { + "epoch": 4.47, + "learning_rate": 2.554083015149649e-06, + "loss": 0.454, + "step": 12118 + }, + { + "epoch": 4.48, + "learning_rate": 2.5540460647863035e-06, + "loss": 0.4835, + "step": 12119 + }, + { + "epoch": 4.48, + "learning_rate": 2.5540091144229585e-06, + "loss": 0.5684, + "step": 12120 + }, + { + "epoch": 4.48, + "learning_rate": 2.553972164059613e-06, + "loss": 0.5246, + "step": 12121 + }, + { + "epoch": 4.48, + "learning_rate": 2.5539352136962682e-06, + "loss": 0.5493, + "step": 12122 + }, + { + "epoch": 4.48, + "learning_rate": 2.553898263332923e-06, + "loss": 0.6047, + "step": 12123 + }, + { + "epoch": 4.48, + "learning_rate": 2.5538613129695775e-06, + "loss": 0.4667, + "step": 12124 + }, + { + "epoch": 4.48, + "learning_rate": 2.553824362606232e-06, + "loss": 0.4637, + "step": 12125 + }, + { + "epoch": 4.48, + "learning_rate": 2.553787412242887e-06, + "loss": 0.46, + "step": 12126 + }, + { + "epoch": 4.48, + "learning_rate": 2.553750461879542e-06, + "loss": 0.5065, + "step": 12127 + }, + { + "epoch": 4.48, + "learning_rate": 2.5537135115161965e-06, + "loss": 0.4829, + "step": 12128 + }, + { + "epoch": 4.48, + "learning_rate": 2.5536765611528515e-06, + "loss": 0.525, + "step": 12129 + }, + { + "epoch": 4.48, + "learning_rate": 2.553639610789506e-06, + "loss": 0.5902, + "step": 12130 + }, + { + "epoch": 4.48, + "learning_rate": 2.553602660426161e-06, + "loss": 0.5464, + "step": 12131 + }, + { + "epoch": 4.48, + "learning_rate": 2.5535657100628155e-06, + "loss": 0.4355, + "step": 12132 + }, + { + "epoch": 4.48, + "learning_rate": 2.5535287596994705e-06, + "loss": 0.5428, + "step": 12133 + }, + { + "epoch": 4.48, + "learning_rate": 2.553491809336125e-06, + "loss": 0.4315, + "step": 12134 + }, + { + "epoch": 4.48, + "learning_rate": 2.55345485897278e-06, + "loss": 0.529, + "step": 12135 + }, + { + "epoch": 4.48, + "learning_rate": 2.553417908609435e-06, + "loss": 0.4968, + "step": 12136 + }, + { + "epoch": 4.48, + "learning_rate": 2.5533809582460895e-06, + "loss": 0.4988, + "step": 12137 + }, + { + "epoch": 4.48, + "learning_rate": 2.553344007882744e-06, + "loss": 0.5359, + "step": 12138 + }, + { + "epoch": 4.48, + "learning_rate": 2.5533070575193988e-06, + "loss": 0.5103, + "step": 12139 + }, + { + "epoch": 4.48, + "learning_rate": 2.553270107156054e-06, + "loss": 0.4992, + "step": 12140 + }, + { + "epoch": 4.48, + "learning_rate": 2.5532331567927084e-06, + "loss": 0.5396, + "step": 12141 + }, + { + "epoch": 4.48, + "learning_rate": 2.5531962064293635e-06, + "loss": 0.6264, + "step": 12142 + }, + { + "epoch": 4.48, + "learning_rate": 2.553159256066018e-06, + "loss": 0.5923, + "step": 12143 + }, + { + "epoch": 4.48, + "learning_rate": 2.5531223057026728e-06, + "loss": 0.5337, + "step": 12144 + }, + { + "epoch": 4.48, + "learning_rate": 2.5530853553393274e-06, + "loss": 0.4792, + "step": 12145 + }, + { + "epoch": 4.49, + "learning_rate": 2.5530484049759825e-06, + "loss": 0.576, + "step": 12146 + }, + { + "epoch": 4.49, + "learning_rate": 2.553011454612637e-06, + "loss": 0.4001, + "step": 12147 + }, + { + "epoch": 4.49, + "learning_rate": 2.5529745042492918e-06, + "loss": 0.4929, + "step": 12148 + }, + { + "epoch": 4.49, + "learning_rate": 2.552937553885947e-06, + "loss": 0.5212, + "step": 12149 + }, + { + "epoch": 4.49, + "learning_rate": 2.5529006035226014e-06, + "loss": 0.4877, + "step": 12150 + }, + { + "epoch": 4.49, + "learning_rate": 2.552863653159256e-06, + "loss": 0.428, + "step": 12151 + }, + { + "epoch": 4.49, + "learning_rate": 2.5528267027959107e-06, + "loss": 0.5024, + "step": 12152 + }, + { + "epoch": 4.49, + "learning_rate": 2.5527897524325658e-06, + "loss": 0.5734, + "step": 12153 + }, + { + "epoch": 4.49, + "learning_rate": 2.5527528020692204e-06, + "loss": 0.5568, + "step": 12154 + }, + { + "epoch": 4.49, + "learning_rate": 2.5527158517058755e-06, + "loss": 0.5758, + "step": 12155 + }, + { + "epoch": 4.49, + "learning_rate": 2.55267890134253e-06, + "loss": 0.4804, + "step": 12156 + }, + { + "epoch": 4.49, + "learning_rate": 2.5526419509791848e-06, + "loss": 0.5649, + "step": 12157 + }, + { + "epoch": 4.49, + "learning_rate": 2.5526050006158394e-06, + "loss": 0.5047, + "step": 12158 + }, + { + "epoch": 4.49, + "learning_rate": 2.552568050252494e-06, + "loss": 0.5054, + "step": 12159 + }, + { + "epoch": 4.49, + "learning_rate": 2.552531099889149e-06, + "loss": 0.5088, + "step": 12160 + }, + { + "epoch": 4.49, + "learning_rate": 2.5524941495258037e-06, + "loss": 0.4657, + "step": 12161 + }, + { + "epoch": 4.49, + "learning_rate": 2.5524571991624584e-06, + "loss": 0.4659, + "step": 12162 + }, + { + "epoch": 4.49, + "learning_rate": 2.552420248799113e-06, + "loss": 0.572, + "step": 12163 + }, + { + "epoch": 4.49, + "learning_rate": 2.552383298435768e-06, + "loss": 0.5903, + "step": 12164 + }, + { + "epoch": 4.49, + "learning_rate": 2.5523463480724227e-06, + "loss": 0.5418, + "step": 12165 + }, + { + "epoch": 4.49, + "learning_rate": 2.5523093977090777e-06, + "loss": 0.457, + "step": 12166 + }, + { + "epoch": 4.49, + "learning_rate": 2.5522724473457324e-06, + "loss": 0.5019, + "step": 12167 + }, + { + "epoch": 4.49, + "learning_rate": 2.552235496982387e-06, + "loss": 0.4232, + "step": 12168 + }, + { + "epoch": 4.49, + "learning_rate": 2.5521985466190417e-06, + "loss": 0.5865, + "step": 12169 + }, + { + "epoch": 4.49, + "learning_rate": 2.5521615962556963e-06, + "loss": 0.5791, + "step": 12170 + }, + { + "epoch": 4.49, + "learning_rate": 2.5521246458923514e-06, + "loss": 0.5614, + "step": 12171 + }, + { + "epoch": 4.49, + "learning_rate": 2.552087695529006e-06, + "loss": 0.5247, + "step": 12172 + }, + { + "epoch": 4.5, + "learning_rate": 2.552050745165661e-06, + "loss": 0.4656, + "step": 12173 + }, + { + "epoch": 4.5, + "learning_rate": 2.5520137948023157e-06, + "loss": 0.5943, + "step": 12174 + }, + { + "epoch": 4.5, + "learning_rate": 2.5519768444389703e-06, + "loss": 0.5218, + "step": 12175 + }, + { + "epoch": 4.5, + "learning_rate": 2.551939894075625e-06, + "loss": 0.3985, + "step": 12176 + }, + { + "epoch": 4.5, + "learning_rate": 2.55190294371228e-06, + "loss": 0.6943, + "step": 12177 + }, + { + "epoch": 4.5, + "learning_rate": 2.5518659933489347e-06, + "loss": 0.4893, + "step": 12178 + }, + { + "epoch": 4.5, + "learning_rate": 2.5518290429855893e-06, + "loss": 0.6479, + "step": 12179 + }, + { + "epoch": 4.5, + "learning_rate": 2.5517920926222444e-06, + "loss": 0.6022, + "step": 12180 + }, + { + "epoch": 4.5, + "learning_rate": 2.551755142258899e-06, + "loss": 0.5612, + "step": 12181 + }, + { + "epoch": 4.5, + "learning_rate": 2.5517181918955536e-06, + "loss": 0.4704, + "step": 12182 + }, + { + "epoch": 4.5, + "learning_rate": 2.5516812415322083e-06, + "loss": 0.6074, + "step": 12183 + }, + { + "epoch": 4.5, + "learning_rate": 2.5516442911688633e-06, + "loss": 0.5326, + "step": 12184 + }, + { + "epoch": 4.5, + "learning_rate": 2.551607340805518e-06, + "loss": 0.5294, + "step": 12185 + }, + { + "epoch": 4.5, + "learning_rate": 2.551570390442173e-06, + "loss": 0.5068, + "step": 12186 + }, + { + "epoch": 4.5, + "learning_rate": 2.5515334400788277e-06, + "loss": 0.5257, + "step": 12187 + }, + { + "epoch": 4.5, + "learning_rate": 2.5514964897154823e-06, + "loss": 0.6101, + "step": 12188 + }, + { + "epoch": 4.5, + "learning_rate": 2.551459539352137e-06, + "loss": 0.5644, + "step": 12189 + }, + { + "epoch": 4.5, + "learning_rate": 2.5514225889887916e-06, + "loss": 0.5923, + "step": 12190 + }, + { + "epoch": 4.5, + "learning_rate": 2.5513856386254466e-06, + "loss": 0.5377, + "step": 12191 + }, + { + "epoch": 4.5, + "learning_rate": 2.5513486882621013e-06, + "loss": 0.4325, + "step": 12192 + }, + { + "epoch": 4.5, + "learning_rate": 2.5513117378987563e-06, + "loss": 0.4519, + "step": 12193 + }, + { + "epoch": 4.5, + "learning_rate": 2.551274787535411e-06, + "loss": 0.5194, + "step": 12194 + }, + { + "epoch": 4.5, + "learning_rate": 2.5512378371720656e-06, + "loss": 0.4893, + "step": 12195 + }, + { + "epoch": 4.5, + "learning_rate": 2.5512008868087202e-06, + "loss": 0.6222, + "step": 12196 + }, + { + "epoch": 4.5, + "learning_rate": 2.5511639364453753e-06, + "loss": 0.5343, + "step": 12197 + }, + { + "epoch": 4.5, + "learning_rate": 2.55112698608203e-06, + "loss": 0.5576, + "step": 12198 + }, + { + "epoch": 4.5, + "learning_rate": 2.5510900357186846e-06, + "loss": 0.539, + "step": 12199 + }, + { + "epoch": 4.51, + "learning_rate": 2.5510530853553396e-06, + "loss": 0.5209, + "step": 12200 + }, + { + "epoch": 4.51, + "learning_rate": 2.551016134991994e-06, + "loss": 0.4517, + "step": 12201 + }, + { + "epoch": 4.51, + "learning_rate": 2.550979184628649e-06, + "loss": 0.4497, + "step": 12202 + }, + { + "epoch": 4.51, + "learning_rate": 2.5509422342653035e-06, + "loss": 0.4552, + "step": 12203 + }, + { + "epoch": 4.51, + "learning_rate": 2.5509052839019586e-06, + "loss": 0.5121, + "step": 12204 + }, + { + "epoch": 4.51, + "learning_rate": 2.5508683335386132e-06, + "loss": 0.479, + "step": 12205 + }, + { + "epoch": 4.51, + "learning_rate": 2.5508313831752683e-06, + "loss": 0.4196, + "step": 12206 + }, + { + "epoch": 4.51, + "learning_rate": 2.5507944328119225e-06, + "loss": 0.4603, + "step": 12207 + }, + { + "epoch": 4.51, + "learning_rate": 2.550757482448577e-06, + "loss": 0.4996, + "step": 12208 + }, + { + "epoch": 4.51, + "learning_rate": 2.550720532085232e-06, + "loss": 0.4849, + "step": 12209 + }, + { + "epoch": 4.51, + "learning_rate": 2.550683581721887e-06, + "loss": 0.6055, + "step": 12210 + }, + { + "epoch": 4.51, + "learning_rate": 2.550646631358542e-06, + "loss": 0.5598, + "step": 12211 + }, + { + "epoch": 4.51, + "learning_rate": 2.5506096809951965e-06, + "loss": 0.4979, + "step": 12212 + }, + { + "epoch": 4.51, + "learning_rate": 2.550572730631851e-06, + "loss": 0.4198, + "step": 12213 + }, + { + "epoch": 4.51, + "learning_rate": 2.550535780268506e-06, + "loss": 0.4369, + "step": 12214 + }, + { + "epoch": 4.51, + "learning_rate": 2.550498829905161e-06, + "loss": 0.4211, + "step": 12215 + }, + { + "epoch": 4.51, + "learning_rate": 2.5504618795418155e-06, + "loss": 0.5576, + "step": 12216 + }, + { + "epoch": 4.51, + "learning_rate": 2.5504249291784706e-06, + "loss": 0.5158, + "step": 12217 + }, + { + "epoch": 4.51, + "learning_rate": 2.550387978815125e-06, + "loss": 0.5832, + "step": 12218 + }, + { + "epoch": 4.51, + "learning_rate": 2.55035102845178e-06, + "loss": 0.4668, + "step": 12219 + }, + { + "epoch": 4.51, + "learning_rate": 2.5503140780884345e-06, + "loss": 0.5445, + "step": 12220 + }, + { + "epoch": 4.51, + "learning_rate": 2.550277127725089e-06, + "loss": 0.5359, + "step": 12221 + }, + { + "epoch": 4.51, + "learning_rate": 2.550240177361744e-06, + "loss": 0.5422, + "step": 12222 + }, + { + "epoch": 4.51, + "learning_rate": 2.550203226998399e-06, + "loss": 0.4834, + "step": 12223 + }, + { + "epoch": 4.51, + "learning_rate": 2.550166276635054e-06, + "loss": 0.586, + "step": 12224 + }, + { + "epoch": 4.51, + "learning_rate": 2.5501293262717085e-06, + "loss": 0.4944, + "step": 12225 + }, + { + "epoch": 4.51, + "learning_rate": 2.550092375908363e-06, + "loss": 0.4824, + "step": 12226 + }, + { + "epoch": 4.52, + "learning_rate": 2.5500554255450178e-06, + "loss": 0.5447, + "step": 12227 + }, + { + "epoch": 4.52, + "learning_rate": 2.5500184751816724e-06, + "loss": 0.5376, + "step": 12228 + }, + { + "epoch": 4.52, + "learning_rate": 2.5499815248183275e-06, + "loss": 0.4793, + "step": 12229 + }, + { + "epoch": 4.52, + "learning_rate": 2.549944574454982e-06, + "loss": 0.6271, + "step": 12230 + }, + { + "epoch": 4.52, + "learning_rate": 2.549907624091637e-06, + "loss": 0.6046, + "step": 12231 + }, + { + "epoch": 4.52, + "learning_rate": 2.549870673728292e-06, + "loss": 0.5758, + "step": 12232 + }, + { + "epoch": 4.52, + "learning_rate": 2.5498337233649464e-06, + "loss": 0.6042, + "step": 12233 + }, + { + "epoch": 4.52, + "learning_rate": 2.549796773001601e-06, + "loss": 0.5788, + "step": 12234 + }, + { + "epoch": 4.52, + "learning_rate": 2.549759822638256e-06, + "loss": 0.6521, + "step": 12235 + }, + { + "epoch": 4.52, + "learning_rate": 2.5497228722749108e-06, + "loss": 0.4633, + "step": 12236 + }, + { + "epoch": 4.52, + "learning_rate": 2.549685921911566e-06, + "loss": 0.5584, + "step": 12237 + }, + { + "epoch": 4.52, + "learning_rate": 2.5496489715482205e-06, + "loss": 0.4249, + "step": 12238 + }, + { + "epoch": 4.52, + "learning_rate": 2.549612021184875e-06, + "loss": 0.5858, + "step": 12239 + }, + { + "epoch": 4.52, + "learning_rate": 2.5495750708215297e-06, + "loss": 0.5013, + "step": 12240 + }, + { + "epoch": 4.52, + "learning_rate": 2.5495381204581844e-06, + "loss": 0.456, + "step": 12241 + }, + { + "epoch": 4.52, + "learning_rate": 2.5495011700948394e-06, + "loss": 0.5242, + "step": 12242 + }, + { + "epoch": 4.52, + "learning_rate": 2.549464219731494e-06, + "loss": 0.4494, + "step": 12243 + }, + { + "epoch": 4.52, + "learning_rate": 2.549427269368149e-06, + "loss": 0.5342, + "step": 12244 + }, + { + "epoch": 4.52, + "learning_rate": 2.5493903190048038e-06, + "loss": 0.6252, + "step": 12245 + }, + { + "epoch": 4.52, + "learning_rate": 2.5493533686414584e-06, + "loss": 0.5334, + "step": 12246 + }, + { + "epoch": 4.52, + "learning_rate": 2.549316418278113e-06, + "loss": 0.5188, + "step": 12247 + }, + { + "epoch": 4.52, + "learning_rate": 2.5492794679147677e-06, + "loss": 0.5403, + "step": 12248 + }, + { + "epoch": 4.52, + "learning_rate": 2.5492425175514227e-06, + "loss": 0.5328, + "step": 12249 + }, + { + "epoch": 4.52, + "learning_rate": 2.5492055671880774e-06, + "loss": 0.5726, + "step": 12250 + }, + { + "epoch": 4.52, + "learning_rate": 2.5491686168247324e-06, + "loss": 0.6196, + "step": 12251 + }, + { + "epoch": 4.52, + "learning_rate": 2.5491316664613866e-06, + "loss": 0.567, + "step": 12252 + }, + { + "epoch": 4.52, + "learning_rate": 2.5490947160980417e-06, + "loss": 0.4678, + "step": 12253 + }, + { + "epoch": 4.53, + "learning_rate": 2.5490577657346963e-06, + "loss": 0.4491, + "step": 12254 + }, + { + "epoch": 4.53, + "learning_rate": 2.5490208153713514e-06, + "loss": 0.4169, + "step": 12255 + }, + { + "epoch": 4.53, + "learning_rate": 2.548983865008006e-06, + "loss": 0.4667, + "step": 12256 + }, + { + "epoch": 4.53, + "learning_rate": 2.548946914644661e-06, + "loss": 0.4628, + "step": 12257 + }, + { + "epoch": 4.53, + "learning_rate": 2.5489099642813153e-06, + "loss": 0.5128, + "step": 12258 + }, + { + "epoch": 4.53, + "learning_rate": 2.54887301391797e-06, + "loss": 0.4478, + "step": 12259 + }, + { + "epoch": 4.53, + "learning_rate": 2.548836063554625e-06, + "loss": 0.5156, + "step": 12260 + }, + { + "epoch": 4.53, + "learning_rate": 2.5487991131912796e-06, + "loss": 0.5342, + "step": 12261 + }, + { + "epoch": 4.53, + "learning_rate": 2.5487621628279347e-06, + "loss": 0.5045, + "step": 12262 + }, + { + "epoch": 4.53, + "learning_rate": 2.5487252124645893e-06, + "loss": 0.6143, + "step": 12263 + }, + { + "epoch": 4.53, + "learning_rate": 2.548688262101244e-06, + "loss": 0.5374, + "step": 12264 + }, + { + "epoch": 4.53, + "learning_rate": 2.5486513117378986e-06, + "loss": 0.4871, + "step": 12265 + }, + { + "epoch": 4.53, + "learning_rate": 2.5486143613745537e-06, + "loss": 0.4376, + "step": 12266 + }, + { + "epoch": 4.53, + "learning_rate": 2.5485774110112083e-06, + "loss": 0.4531, + "step": 12267 + }, + { + "epoch": 4.53, + "learning_rate": 2.548540460647863e-06, + "loss": 0.469, + "step": 12268 + }, + { + "epoch": 4.53, + "learning_rate": 2.548503510284518e-06, + "loss": 0.513, + "step": 12269 + }, + { + "epoch": 4.53, + "learning_rate": 2.5484665599211726e-06, + "loss": 0.4989, + "step": 12270 + }, + { + "epoch": 4.53, + "learning_rate": 2.5484296095578273e-06, + "loss": 0.5078, + "step": 12271 + }, + { + "epoch": 4.53, + "learning_rate": 2.548392659194482e-06, + "loss": 0.4981, + "step": 12272 + }, + { + "epoch": 4.53, + "learning_rate": 2.548355708831137e-06, + "loss": 0.5155, + "step": 12273 + }, + { + "epoch": 4.53, + "learning_rate": 2.5483187584677916e-06, + "loss": 0.5201, + "step": 12274 + }, + { + "epoch": 4.53, + "learning_rate": 2.5482818081044467e-06, + "loss": 0.4849, + "step": 12275 + }, + { + "epoch": 4.53, + "learning_rate": 2.5482448577411013e-06, + "loss": 0.582, + "step": 12276 + }, + { + "epoch": 4.53, + "learning_rate": 2.548207907377756e-06, + "loss": 0.4984, + "step": 12277 + }, + { + "epoch": 4.53, + "learning_rate": 2.5481709570144106e-06, + "loss": 0.4907, + "step": 12278 + }, + { + "epoch": 4.53, + "learning_rate": 2.5481340066510652e-06, + "loss": 0.5723, + "step": 12279 + }, + { + "epoch": 4.53, + "learning_rate": 2.5480970562877203e-06, + "loss": 0.578, + "step": 12280 + }, + { + "epoch": 4.54, + "learning_rate": 2.548060105924375e-06, + "loss": 0.4843, + "step": 12281 + }, + { + "epoch": 4.54, + "learning_rate": 2.54802315556103e-06, + "loss": 0.5896, + "step": 12282 + }, + { + "epoch": 4.54, + "learning_rate": 2.5479862051976846e-06, + "loss": 0.5386, + "step": 12283 + }, + { + "epoch": 4.54, + "learning_rate": 2.5479492548343392e-06, + "loss": 0.5029, + "step": 12284 + }, + { + "epoch": 4.54, + "learning_rate": 2.547912304470994e-06, + "loss": 0.5498, + "step": 12285 + }, + { + "epoch": 4.54, + "learning_rate": 2.547875354107649e-06, + "loss": 0.4507, + "step": 12286 + }, + { + "epoch": 4.54, + "learning_rate": 2.5478384037443036e-06, + "loss": 0.5913, + "step": 12287 + }, + { + "epoch": 4.54, + "learning_rate": 2.5478014533809582e-06, + "loss": 0.4909, + "step": 12288 + }, + { + "epoch": 4.54, + "learning_rate": 2.5477645030176133e-06, + "loss": 0.424, + "step": 12289 + }, + { + "epoch": 4.54, + "learning_rate": 2.547727552654268e-06, + "loss": 0.4849, + "step": 12290 + }, + { + "epoch": 4.54, + "learning_rate": 2.5476906022909225e-06, + "loss": 0.5562, + "step": 12291 + }, + { + "epoch": 4.54, + "learning_rate": 2.547653651927577e-06, + "loss": 0.4971, + "step": 12292 + }, + { + "epoch": 4.54, + "learning_rate": 2.5476167015642322e-06, + "loss": 0.6014, + "step": 12293 + }, + { + "epoch": 4.54, + "learning_rate": 2.547579751200887e-06, + "loss": 0.4268, + "step": 12294 + }, + { + "epoch": 4.54, + "learning_rate": 2.547542800837542e-06, + "loss": 0.5125, + "step": 12295 + }, + { + "epoch": 4.54, + "learning_rate": 2.5475058504741966e-06, + "loss": 0.5263, + "step": 12296 + }, + { + "epoch": 4.54, + "learning_rate": 2.5474689001108512e-06, + "loss": 0.5132, + "step": 12297 + }, + { + "epoch": 4.54, + "learning_rate": 2.547431949747506e-06, + "loss": 0.362, + "step": 12298 + }, + { + "epoch": 4.54, + "learning_rate": 2.5473949993841605e-06, + "loss": 0.4469, + "step": 12299 + }, + { + "epoch": 4.54, + "learning_rate": 2.5473580490208155e-06, + "loss": 0.4856, + "step": 12300 + }, + { + "epoch": 4.54, + "learning_rate": 2.54732109865747e-06, + "loss": 0.5305, + "step": 12301 + }, + { + "epoch": 4.54, + "learning_rate": 2.5472841482941252e-06, + "loss": 0.5105, + "step": 12302 + }, + { + "epoch": 4.54, + "learning_rate": 2.5472471979307795e-06, + "loss": 0.5559, + "step": 12303 + }, + { + "epoch": 4.54, + "learning_rate": 2.5472102475674345e-06, + "loss": 0.5648, + "step": 12304 + }, + { + "epoch": 4.54, + "learning_rate": 2.547173297204089e-06, + "loss": 0.5479, + "step": 12305 + }, + { + "epoch": 4.54, + "learning_rate": 2.5471363468407442e-06, + "loss": 0.5583, + "step": 12306 + }, + { + "epoch": 4.54, + "learning_rate": 2.547099396477399e-06, + "loss": 0.4869, + "step": 12307 + }, + { + "epoch": 4.55, + "learning_rate": 2.5470624461140535e-06, + "loss": 0.4603, + "step": 12308 + }, + { + "epoch": 4.55, + "learning_rate": 2.547025495750708e-06, + "loss": 0.569, + "step": 12309 + }, + { + "epoch": 4.55, + "learning_rate": 2.5469885453873628e-06, + "loss": 0.6268, + "step": 12310 + }, + { + "epoch": 4.55, + "learning_rate": 2.546951595024018e-06, + "loss": 0.4876, + "step": 12311 + }, + { + "epoch": 4.55, + "learning_rate": 2.5469146446606725e-06, + "loss": 0.5444, + "step": 12312 + }, + { + "epoch": 4.55, + "learning_rate": 2.5468776942973275e-06, + "loss": 0.4632, + "step": 12313 + }, + { + "epoch": 4.55, + "learning_rate": 2.546840743933982e-06, + "loss": 0.4145, + "step": 12314 + }, + { + "epoch": 4.55, + "learning_rate": 2.546803793570637e-06, + "loss": 0.5786, + "step": 12315 + }, + { + "epoch": 4.55, + "learning_rate": 2.5467668432072914e-06, + "loss": 0.5359, + "step": 12316 + }, + { + "epoch": 4.55, + "learning_rate": 2.5467298928439465e-06, + "loss": 0.4853, + "step": 12317 + }, + { + "epoch": 4.55, + "learning_rate": 2.546692942480601e-06, + "loss": 0.4542, + "step": 12318 + }, + { + "epoch": 4.55, + "learning_rate": 2.5466559921172558e-06, + "loss": 0.5478, + "step": 12319 + }, + { + "epoch": 4.55, + "learning_rate": 2.546619041753911e-06, + "loss": 0.5446, + "step": 12320 + }, + { + "epoch": 4.55, + "learning_rate": 2.5465820913905655e-06, + "loss": 0.4938, + "step": 12321 + }, + { + "epoch": 4.55, + "learning_rate": 2.54654514102722e-06, + "loss": 0.4118, + "step": 12322 + }, + { + "epoch": 4.55, + "learning_rate": 2.5465081906638747e-06, + "loss": 0.5246, + "step": 12323 + }, + { + "epoch": 4.55, + "learning_rate": 2.5464712403005298e-06, + "loss": 0.4214, + "step": 12324 + }, + { + "epoch": 4.55, + "learning_rate": 2.5464342899371844e-06, + "loss": 0.5969, + "step": 12325 + }, + { + "epoch": 4.55, + "learning_rate": 2.5463973395738395e-06, + "loss": 0.5475, + "step": 12326 + }, + { + "epoch": 4.55, + "learning_rate": 2.546360389210494e-06, + "loss": 0.4162, + "step": 12327 + }, + { + "epoch": 4.55, + "learning_rate": 2.5463234388471488e-06, + "loss": 0.5082, + "step": 12328 + }, + { + "epoch": 4.55, + "learning_rate": 2.5462864884838034e-06, + "loss": 0.5402, + "step": 12329 + }, + { + "epoch": 4.55, + "learning_rate": 2.546249538120458e-06, + "loss": 0.4814, + "step": 12330 + }, + { + "epoch": 4.55, + "learning_rate": 2.546212587757113e-06, + "loss": 0.5142, + "step": 12331 + }, + { + "epoch": 4.55, + "learning_rate": 2.5461756373937677e-06, + "loss": 0.473, + "step": 12332 + }, + { + "epoch": 4.55, + "learning_rate": 2.5461386870304228e-06, + "loss": 0.536, + "step": 12333 + }, + { + "epoch": 4.55, + "learning_rate": 2.5461017366670774e-06, + "loss": 0.3904, + "step": 12334 + }, + { + "epoch": 4.56, + "learning_rate": 2.546064786303732e-06, + "loss": 0.4963, + "step": 12335 + }, + { + "epoch": 4.56, + "learning_rate": 2.5460278359403867e-06, + "loss": 0.4392, + "step": 12336 + }, + { + "epoch": 4.56, + "learning_rate": 2.5459908855770418e-06, + "loss": 0.4888, + "step": 12337 + }, + { + "epoch": 4.56, + "learning_rate": 2.5459539352136964e-06, + "loss": 0.5902, + "step": 12338 + }, + { + "epoch": 4.56, + "learning_rate": 2.545916984850351e-06, + "loss": 0.3869, + "step": 12339 + }, + { + "epoch": 4.56, + "learning_rate": 2.545880034487006e-06, + "loss": 0.4805, + "step": 12340 + }, + { + "epoch": 4.56, + "learning_rate": 2.5458430841236607e-06, + "loss": 0.5714, + "step": 12341 + }, + { + "epoch": 4.56, + "learning_rate": 2.5458061337603154e-06, + "loss": 0.4102, + "step": 12342 + }, + { + "epoch": 4.56, + "learning_rate": 2.54576918339697e-06, + "loss": 0.4186, + "step": 12343 + }, + { + "epoch": 4.56, + "learning_rate": 2.545732233033625e-06, + "loss": 0.412, + "step": 12344 + }, + { + "epoch": 4.56, + "learning_rate": 2.5456952826702797e-06, + "loss": 0.5385, + "step": 12345 + }, + { + "epoch": 4.56, + "learning_rate": 2.5456583323069348e-06, + "loss": 0.5895, + "step": 12346 + }, + { + "epoch": 4.56, + "learning_rate": 2.545621381943589e-06, + "loss": 0.4521, + "step": 12347 + }, + { + "epoch": 4.56, + "learning_rate": 2.5455844315802436e-06, + "loss": 0.4723, + "step": 12348 + }, + { + "epoch": 4.56, + "learning_rate": 2.5455474812168987e-06, + "loss": 0.504, + "step": 12349 + }, + { + "epoch": 4.56, + "learning_rate": 2.5455105308535533e-06, + "loss": 0.507, + "step": 12350 + }, + { + "epoch": 4.56, + "learning_rate": 2.5454735804902084e-06, + "loss": 0.6316, + "step": 12351 + }, + { + "epoch": 4.56, + "learning_rate": 2.545436630126863e-06, + "loss": 0.5508, + "step": 12352 + }, + { + "epoch": 4.56, + "learning_rate": 2.5453996797635176e-06, + "loss": 0.5366, + "step": 12353 + }, + { + "epoch": 4.56, + "learning_rate": 2.5453627294001723e-06, + "loss": 0.3834, + "step": 12354 + }, + { + "epoch": 4.56, + "learning_rate": 2.5453257790368273e-06, + "loss": 0.4331, + "step": 12355 + }, + { + "epoch": 4.56, + "learning_rate": 2.545288828673482e-06, + "loss": 0.6167, + "step": 12356 + }, + { + "epoch": 4.56, + "learning_rate": 2.5452518783101366e-06, + "loss": 0.5702, + "step": 12357 + }, + { + "epoch": 4.56, + "learning_rate": 2.5452149279467917e-06, + "loss": 0.4923, + "step": 12358 + }, + { + "epoch": 4.56, + "learning_rate": 2.5451779775834463e-06, + "loss": 0.4595, + "step": 12359 + }, + { + "epoch": 4.56, + "learning_rate": 2.545141027220101e-06, + "loss": 0.4104, + "step": 12360 + }, + { + "epoch": 4.56, + "learning_rate": 2.5451040768567556e-06, + "loss": 0.4426, + "step": 12361 + }, + { + "epoch": 4.56, + "learning_rate": 2.5450671264934106e-06, + "loss": 0.4923, + "step": 12362 + }, + { + "epoch": 4.57, + "learning_rate": 2.5450301761300653e-06, + "loss": 0.3883, + "step": 12363 + }, + { + "epoch": 4.57, + "learning_rate": 2.5449932257667203e-06, + "loss": 0.5859, + "step": 12364 + }, + { + "epoch": 4.57, + "learning_rate": 2.544956275403375e-06, + "loss": 0.4174, + "step": 12365 + }, + { + "epoch": 4.57, + "learning_rate": 2.5449193250400296e-06, + "loss": 0.4677, + "step": 12366 + }, + { + "epoch": 4.57, + "learning_rate": 2.5448823746766842e-06, + "loss": 0.5839, + "step": 12367 + }, + { + "epoch": 4.57, + "learning_rate": 2.544845424313339e-06, + "loss": 0.463, + "step": 12368 + }, + { + "epoch": 4.57, + "learning_rate": 2.544808473949994e-06, + "loss": 0.5823, + "step": 12369 + }, + { + "epoch": 4.57, + "learning_rate": 2.5447715235866486e-06, + "loss": 0.4974, + "step": 12370 + }, + { + "epoch": 4.57, + "learning_rate": 2.5447345732233036e-06, + "loss": 0.5628, + "step": 12371 + }, + { + "epoch": 4.57, + "learning_rate": 2.5446976228599583e-06, + "loss": 0.5817, + "step": 12372 + }, + { + "epoch": 4.57, + "learning_rate": 2.544660672496613e-06, + "loss": 0.4452, + "step": 12373 + }, + { + "epoch": 4.57, + "learning_rate": 2.5446237221332675e-06, + "loss": 0.4886, + "step": 12374 + }, + { + "epoch": 4.57, + "learning_rate": 2.5445867717699226e-06, + "loss": 0.551, + "step": 12375 + }, + { + "epoch": 4.57, + "learning_rate": 2.5445498214065772e-06, + "loss": 0.4295, + "step": 12376 + }, + { + "epoch": 4.57, + "learning_rate": 2.544512871043232e-06, + "loss": 0.5887, + "step": 12377 + }, + { + "epoch": 4.57, + "learning_rate": 2.544475920679887e-06, + "loss": 0.4343, + "step": 12378 + }, + { + "epoch": 4.57, + "learning_rate": 2.5444389703165416e-06, + "loss": 0.5682, + "step": 12379 + }, + { + "epoch": 4.57, + "learning_rate": 2.544402019953196e-06, + "loss": 0.4204, + "step": 12380 + }, + { + "epoch": 4.57, + "learning_rate": 2.544365069589851e-06, + "loss": 0.5452, + "step": 12381 + }, + { + "epoch": 4.57, + "learning_rate": 2.544328119226506e-06, + "loss": 0.503, + "step": 12382 + }, + { + "epoch": 4.57, + "learning_rate": 2.5442911688631605e-06, + "loss": 0.5754, + "step": 12383 + }, + { + "epoch": 4.57, + "learning_rate": 2.5442542184998156e-06, + "loss": 0.547, + "step": 12384 + }, + { + "epoch": 4.57, + "learning_rate": 2.5442172681364702e-06, + "loss": 0.5882, + "step": 12385 + }, + { + "epoch": 4.57, + "learning_rate": 2.544180317773125e-06, + "loss": 0.5248, + "step": 12386 + }, + { + "epoch": 4.57, + "learning_rate": 2.5441433674097795e-06, + "loss": 0.5583, + "step": 12387 + }, + { + "epoch": 4.57, + "learning_rate": 2.544106417046434e-06, + "loss": 0.6084, + "step": 12388 + }, + { + "epoch": 4.57, + "learning_rate": 2.544069466683089e-06, + "loss": 0.4811, + "step": 12389 + }, + { + "epoch": 4.58, + "learning_rate": 2.544032516319744e-06, + "loss": 0.5425, + "step": 12390 + }, + { + "epoch": 4.58, + "learning_rate": 2.543995565956399e-06, + "loss": 0.4783, + "step": 12391 + }, + { + "epoch": 4.58, + "learning_rate": 2.543958615593053e-06, + "loss": 0.6117, + "step": 12392 + }, + { + "epoch": 4.58, + "learning_rate": 2.543921665229708e-06, + "loss": 0.7165, + "step": 12393 + }, + { + "epoch": 4.58, + "learning_rate": 2.543884714866363e-06, + "loss": 0.467, + "step": 12394 + }, + { + "epoch": 4.58, + "learning_rate": 2.543847764503018e-06, + "loss": 0.5098, + "step": 12395 + }, + { + "epoch": 4.58, + "learning_rate": 2.5438108141396725e-06, + "loss": 0.5298, + "step": 12396 + }, + { + "epoch": 4.58, + "learning_rate": 2.543773863776327e-06, + "loss": 0.5731, + "step": 12397 + }, + { + "epoch": 4.58, + "learning_rate": 2.5437369134129818e-06, + "loss": 0.6425, + "step": 12398 + }, + { + "epoch": 4.58, + "learning_rate": 2.5436999630496364e-06, + "loss": 0.5389, + "step": 12399 + }, + { + "epoch": 4.58, + "learning_rate": 2.5436630126862915e-06, + "loss": 0.5288, + "step": 12400 + }, + { + "epoch": 4.58, + "learning_rate": 2.543626062322946e-06, + "loss": 0.4509, + "step": 12401 + }, + { + "epoch": 4.58, + "learning_rate": 2.543589111959601e-06, + "loss": 0.44, + "step": 12402 + }, + { + "epoch": 4.58, + "learning_rate": 2.543552161596256e-06, + "loss": 0.5152, + "step": 12403 + }, + { + "epoch": 4.58, + "learning_rate": 2.5435152112329104e-06, + "loss": 0.6061, + "step": 12404 + }, + { + "epoch": 4.58, + "learning_rate": 2.543478260869565e-06, + "loss": 0.4762, + "step": 12405 + }, + { + "epoch": 4.58, + "learning_rate": 2.54344131050622e-06, + "loss": 0.4608, + "step": 12406 + }, + { + "epoch": 4.58, + "learning_rate": 2.5434043601428748e-06, + "loss": 0.5265, + "step": 12407 + }, + { + "epoch": 4.58, + "learning_rate": 2.5433674097795294e-06, + "loss": 0.5395, + "step": 12408 + }, + { + "epoch": 4.58, + "learning_rate": 2.5433304594161845e-06, + "loss": 0.46, + "step": 12409 + }, + { + "epoch": 4.58, + "learning_rate": 2.543293509052839e-06, + "loss": 0.7182, + "step": 12410 + }, + { + "epoch": 4.58, + "learning_rate": 2.5432565586894937e-06, + "loss": 0.6632, + "step": 12411 + }, + { + "epoch": 4.58, + "learning_rate": 2.5432196083261484e-06, + "loss": 0.5738, + "step": 12412 + }, + { + "epoch": 4.58, + "learning_rate": 2.5431826579628034e-06, + "loss": 0.4964, + "step": 12413 + }, + { + "epoch": 4.58, + "learning_rate": 2.543145707599458e-06, + "loss": 0.6465, + "step": 12414 + }, + { + "epoch": 4.58, + "learning_rate": 2.543108757236113e-06, + "loss": 0.585, + "step": 12415 + }, + { + "epoch": 4.58, + "learning_rate": 2.5430718068727678e-06, + "loss": 0.6163, + "step": 12416 + }, + { + "epoch": 4.59, + "learning_rate": 2.5430348565094224e-06, + "loss": 0.552, + "step": 12417 + }, + { + "epoch": 4.59, + "learning_rate": 2.542997906146077e-06, + "loss": 0.5924, + "step": 12418 + }, + { + "epoch": 4.59, + "learning_rate": 2.5429609557827317e-06, + "loss": 0.4854, + "step": 12419 + }, + { + "epoch": 4.59, + "learning_rate": 2.5429240054193867e-06, + "loss": 0.573, + "step": 12420 + }, + { + "epoch": 4.59, + "learning_rate": 2.5428870550560414e-06, + "loss": 0.4491, + "step": 12421 + }, + { + "epoch": 4.59, + "learning_rate": 2.5428501046926964e-06, + "loss": 0.5901, + "step": 12422 + }, + { + "epoch": 4.59, + "learning_rate": 2.542813154329351e-06, + "loss": 0.5644, + "step": 12423 + }, + { + "epoch": 4.59, + "learning_rate": 2.5427762039660057e-06, + "loss": 0.585, + "step": 12424 + }, + { + "epoch": 4.59, + "learning_rate": 2.5427392536026603e-06, + "loss": 0.4837, + "step": 12425 + }, + { + "epoch": 4.59, + "learning_rate": 2.5427023032393154e-06, + "loss": 0.5891, + "step": 12426 + }, + { + "epoch": 4.59, + "learning_rate": 2.54266535287597e-06, + "loss": 0.6116, + "step": 12427 + }, + { + "epoch": 4.59, + "learning_rate": 2.5426284025126247e-06, + "loss": 0.5487, + "step": 12428 + }, + { + "epoch": 4.59, + "learning_rate": 2.5425914521492797e-06, + "loss": 0.6427, + "step": 12429 + }, + { + "epoch": 4.59, + "learning_rate": 2.5425545017859344e-06, + "loss": 0.4893, + "step": 12430 + }, + { + "epoch": 4.59, + "learning_rate": 2.542517551422589e-06, + "loss": 0.4597, + "step": 12431 + }, + { + "epoch": 4.59, + "learning_rate": 2.5424806010592437e-06, + "loss": 0.4944, + "step": 12432 + }, + { + "epoch": 4.59, + "learning_rate": 2.5424436506958987e-06, + "loss": 0.579, + "step": 12433 + }, + { + "epoch": 4.59, + "learning_rate": 2.5424067003325533e-06, + "loss": 0.5192, + "step": 12434 + }, + { + "epoch": 4.59, + "learning_rate": 2.5423697499692084e-06, + "loss": 0.4515, + "step": 12435 + }, + { + "epoch": 4.59, + "learning_rate": 2.542332799605863e-06, + "loss": 0.5077, + "step": 12436 + }, + { + "epoch": 4.59, + "learning_rate": 2.5422958492425173e-06, + "loss": 0.4838, + "step": 12437 + }, + { + "epoch": 4.59, + "learning_rate": 2.5422588988791723e-06, + "loss": 0.5012, + "step": 12438 + }, + { + "epoch": 4.59, + "learning_rate": 2.542221948515827e-06, + "loss": 0.485, + "step": 12439 + }, + { + "epoch": 4.59, + "learning_rate": 2.542184998152482e-06, + "loss": 0.4586, + "step": 12440 + }, + { + "epoch": 4.59, + "learning_rate": 2.5421480477891366e-06, + "loss": 0.4481, + "step": 12441 + }, + { + "epoch": 4.59, + "learning_rate": 2.5421110974257917e-06, + "loss": 0.4156, + "step": 12442 + }, + { + "epoch": 4.59, + "learning_rate": 2.542074147062446e-06, + "loss": 0.6447, + "step": 12443 + }, + { + "epoch": 4.6, + "learning_rate": 2.542037196699101e-06, + "loss": 0.4343, + "step": 12444 + }, + { + "epoch": 4.6, + "learning_rate": 2.5420002463357556e-06, + "loss": 0.5012, + "step": 12445 + }, + { + "epoch": 4.6, + "learning_rate": 2.5419632959724107e-06, + "loss": 0.5503, + "step": 12446 + }, + { + "epoch": 4.6, + "learning_rate": 2.5419263456090653e-06, + "loss": 0.5124, + "step": 12447 + }, + { + "epoch": 4.6, + "learning_rate": 2.54188939524572e-06, + "loss": 0.4492, + "step": 12448 + }, + { + "epoch": 4.6, + "learning_rate": 2.5418524448823746e-06, + "loss": 0.4787, + "step": 12449 + }, + { + "epoch": 4.6, + "learning_rate": 2.5418154945190292e-06, + "loss": 0.5263, + "step": 12450 + }, + { + "epoch": 4.6, + "learning_rate": 2.5417785441556843e-06, + "loss": 0.3947, + "step": 12451 + }, + { + "epoch": 4.6, + "learning_rate": 2.541741593792339e-06, + "loss": 0.5745, + "step": 12452 + }, + { + "epoch": 4.6, + "learning_rate": 2.541704643428994e-06, + "loss": 0.5301, + "step": 12453 + }, + { + "epoch": 4.6, + "learning_rate": 2.5416676930656486e-06, + "loss": 0.5128, + "step": 12454 + }, + { + "epoch": 4.6, + "learning_rate": 2.5416307427023033e-06, + "loss": 0.5091, + "step": 12455 + }, + { + "epoch": 4.6, + "learning_rate": 2.541593792338958e-06, + "loss": 0.4975, + "step": 12456 + }, + { + "epoch": 4.6, + "learning_rate": 2.5415568419756125e-06, + "loss": 0.5881, + "step": 12457 + }, + { + "epoch": 4.6, + "learning_rate": 2.5415198916122676e-06, + "loss": 0.5763, + "step": 12458 + }, + { + "epoch": 4.6, + "learning_rate": 2.5414829412489222e-06, + "loss": 0.4611, + "step": 12459 + }, + { + "epoch": 4.6, + "learning_rate": 2.5414459908855773e-06, + "loss": 0.5647, + "step": 12460 + }, + { + "epoch": 4.6, + "learning_rate": 2.541409040522232e-06, + "loss": 0.5282, + "step": 12461 + }, + { + "epoch": 4.6, + "learning_rate": 2.5413720901588866e-06, + "loss": 0.5975, + "step": 12462 + }, + { + "epoch": 4.6, + "learning_rate": 2.541335139795541e-06, + "loss": 0.5073, + "step": 12463 + }, + { + "epoch": 4.6, + "learning_rate": 2.5412981894321963e-06, + "loss": 0.4041, + "step": 12464 + }, + { + "epoch": 4.6, + "learning_rate": 2.541261239068851e-06, + "loss": 0.4715, + "step": 12465 + }, + { + "epoch": 4.6, + "learning_rate": 2.541224288705506e-06, + "loss": 0.4823, + "step": 12466 + }, + { + "epoch": 4.6, + "learning_rate": 2.5411873383421606e-06, + "loss": 0.5126, + "step": 12467 + }, + { + "epoch": 4.6, + "learning_rate": 2.5411503879788152e-06, + "loss": 0.5345, + "step": 12468 + }, + { + "epoch": 4.6, + "learning_rate": 2.54111343761547e-06, + "loss": 0.5045, + "step": 12469 + }, + { + "epoch": 4.6, + "learning_rate": 2.5410764872521245e-06, + "loss": 0.5312, + "step": 12470 + }, + { + "epoch": 4.61, + "learning_rate": 2.5410395368887796e-06, + "loss": 0.5872, + "step": 12471 + }, + { + "epoch": 4.61, + "learning_rate": 2.541002586525434e-06, + "loss": 0.5829, + "step": 12472 + }, + { + "epoch": 4.61, + "learning_rate": 2.5409656361620893e-06, + "loss": 0.4735, + "step": 12473 + }, + { + "epoch": 4.61, + "learning_rate": 2.540928685798744e-06, + "loss": 0.4984, + "step": 12474 + }, + { + "epoch": 4.61, + "learning_rate": 2.5408917354353985e-06, + "loss": 0.6558, + "step": 12475 + }, + { + "epoch": 4.61, + "learning_rate": 2.540854785072053e-06, + "loss": 0.605, + "step": 12476 + }, + { + "epoch": 4.61, + "learning_rate": 2.540817834708708e-06, + "loss": 0.4398, + "step": 12477 + }, + { + "epoch": 4.61, + "learning_rate": 2.540780884345363e-06, + "loss": 0.5663, + "step": 12478 + }, + { + "epoch": 4.61, + "learning_rate": 2.5407439339820175e-06, + "loss": 0.4977, + "step": 12479 + }, + { + "epoch": 4.61, + "learning_rate": 2.5407069836186726e-06, + "loss": 0.5023, + "step": 12480 + }, + { + "epoch": 4.61, + "learning_rate": 2.540670033255327e-06, + "loss": 0.4909, + "step": 12481 + }, + { + "epoch": 4.61, + "learning_rate": 2.540633082891982e-06, + "loss": 0.593, + "step": 12482 + }, + { + "epoch": 4.61, + "learning_rate": 2.5405961325286365e-06, + "loss": 0.5349, + "step": 12483 + }, + { + "epoch": 4.61, + "learning_rate": 2.5405591821652915e-06, + "loss": 0.515, + "step": 12484 + }, + { + "epoch": 4.61, + "learning_rate": 2.540522231801946e-06, + "loss": 0.5061, + "step": 12485 + }, + { + "epoch": 4.61, + "learning_rate": 2.5404852814386012e-06, + "loss": 0.5951, + "step": 12486 + }, + { + "epoch": 4.61, + "learning_rate": 2.540448331075256e-06, + "loss": 0.4433, + "step": 12487 + }, + { + "epoch": 4.61, + "learning_rate": 2.54041138071191e-06, + "loss": 0.456, + "step": 12488 + }, + { + "epoch": 4.61, + "learning_rate": 2.540374430348565e-06, + "loss": 0.5688, + "step": 12489 + }, + { + "epoch": 4.61, + "learning_rate": 2.5403374799852198e-06, + "loss": 0.5234, + "step": 12490 + }, + { + "epoch": 4.61, + "learning_rate": 2.540300529621875e-06, + "loss": 0.4936, + "step": 12491 + }, + { + "epoch": 4.61, + "learning_rate": 2.5402635792585295e-06, + "loss": 0.4695, + "step": 12492 + }, + { + "epoch": 4.61, + "learning_rate": 2.5402266288951845e-06, + "loss": 0.5568, + "step": 12493 + }, + { + "epoch": 4.61, + "learning_rate": 2.5401896785318387e-06, + "loss": 0.6608, + "step": 12494 + }, + { + "epoch": 4.61, + "learning_rate": 2.540152728168494e-06, + "loss": 0.5221, + "step": 12495 + }, + { + "epoch": 4.61, + "learning_rate": 2.5401157778051484e-06, + "loss": 0.5998, + "step": 12496 + }, + { + "epoch": 4.61, + "learning_rate": 2.540078827441803e-06, + "loss": 0.5664, + "step": 12497 + }, + { + "epoch": 4.62, + "learning_rate": 2.540041877078458e-06, + "loss": 0.6161, + "step": 12498 + }, + { + "epoch": 4.62, + "learning_rate": 2.5400049267151128e-06, + "loss": 0.479, + "step": 12499 + }, + { + "epoch": 4.62, + "learning_rate": 2.5399679763517674e-06, + "loss": 0.5708, + "step": 12500 + }, + { + "epoch": 4.62, + "learning_rate": 2.539931025988422e-06, + "loss": 0.5584, + "step": 12501 + }, + { + "epoch": 4.62, + "learning_rate": 2.539894075625077e-06, + "loss": 0.62, + "step": 12502 + }, + { + "epoch": 4.62, + "learning_rate": 2.5398571252617317e-06, + "loss": 0.5277, + "step": 12503 + }, + { + "epoch": 4.62, + "learning_rate": 2.539820174898387e-06, + "loss": 0.527, + "step": 12504 + }, + { + "epoch": 4.62, + "learning_rate": 2.5397832245350414e-06, + "loss": 0.4753, + "step": 12505 + }, + { + "epoch": 4.62, + "learning_rate": 2.539746274171696e-06, + "loss": 0.4161, + "step": 12506 + }, + { + "epoch": 4.62, + "learning_rate": 2.5397093238083507e-06, + "loss": 0.4603, + "step": 12507 + }, + { + "epoch": 4.62, + "learning_rate": 2.5396723734450053e-06, + "loss": 0.4344, + "step": 12508 + }, + { + "epoch": 4.62, + "learning_rate": 2.5396354230816604e-06, + "loss": 0.6464, + "step": 12509 + }, + { + "epoch": 4.62, + "learning_rate": 2.539598472718315e-06, + "loss": 0.6658, + "step": 12510 + }, + { + "epoch": 4.62, + "learning_rate": 2.53956152235497e-06, + "loss": 0.4671, + "step": 12511 + }, + { + "epoch": 4.62, + "learning_rate": 2.5395245719916247e-06, + "loss": 0.612, + "step": 12512 + }, + { + "epoch": 4.62, + "learning_rate": 2.5394876216282794e-06, + "loss": 0.6315, + "step": 12513 + }, + { + "epoch": 4.62, + "learning_rate": 2.539450671264934e-06, + "loss": 0.5402, + "step": 12514 + }, + { + "epoch": 4.62, + "learning_rate": 2.539413720901589e-06, + "loss": 0.6224, + "step": 12515 + }, + { + "epoch": 4.62, + "learning_rate": 2.5393767705382437e-06, + "loss": 0.5706, + "step": 12516 + }, + { + "epoch": 4.62, + "learning_rate": 2.5393398201748983e-06, + "loss": 0.5583, + "step": 12517 + }, + { + "epoch": 4.62, + "learning_rate": 2.5393028698115534e-06, + "loss": 0.4531, + "step": 12518 + }, + { + "epoch": 4.62, + "learning_rate": 2.539265919448208e-06, + "loss": 0.4956, + "step": 12519 + }, + { + "epoch": 4.62, + "learning_rate": 2.5392289690848627e-06, + "loss": 0.4601, + "step": 12520 + }, + { + "epoch": 4.62, + "learning_rate": 2.5391920187215173e-06, + "loss": 0.5002, + "step": 12521 + }, + { + "epoch": 4.62, + "learning_rate": 2.5391550683581724e-06, + "loss": 0.5247, + "step": 12522 + }, + { + "epoch": 4.62, + "learning_rate": 2.539118117994827e-06, + "loss": 0.53, + "step": 12523 + }, + { + "epoch": 4.62, + "learning_rate": 2.539081167631482e-06, + "loss": 0.4759, + "step": 12524 + }, + { + "epoch": 4.63, + "learning_rate": 2.5390442172681367e-06, + "loss": 0.4622, + "step": 12525 + }, + { + "epoch": 4.63, + "learning_rate": 2.5390072669047913e-06, + "loss": 0.5054, + "step": 12526 + }, + { + "epoch": 4.63, + "learning_rate": 2.538970316541446e-06, + "loss": 0.5798, + "step": 12527 + }, + { + "epoch": 4.63, + "learning_rate": 2.5389333661781006e-06, + "loss": 0.4257, + "step": 12528 + }, + { + "epoch": 4.63, + "learning_rate": 2.5388964158147557e-06, + "loss": 0.6651, + "step": 12529 + }, + { + "epoch": 4.63, + "learning_rate": 2.5388594654514103e-06, + "loss": 0.508, + "step": 12530 + }, + { + "epoch": 4.63, + "learning_rate": 2.5388225150880654e-06, + "loss": 0.591, + "step": 12531 + }, + { + "epoch": 4.63, + "learning_rate": 2.5387855647247196e-06, + "loss": 0.4637, + "step": 12532 + }, + { + "epoch": 4.63, + "learning_rate": 2.5387486143613746e-06, + "loss": 0.5397, + "step": 12533 + }, + { + "epoch": 4.63, + "learning_rate": 2.5387116639980293e-06, + "loss": 0.5736, + "step": 12534 + }, + { + "epoch": 4.63, + "learning_rate": 2.5386747136346843e-06, + "loss": 0.478, + "step": 12535 + }, + { + "epoch": 4.63, + "learning_rate": 2.538637763271339e-06, + "loss": 0.5851, + "step": 12536 + }, + { + "epoch": 4.63, + "learning_rate": 2.5386008129079936e-06, + "loss": 0.46, + "step": 12537 + }, + { + "epoch": 4.63, + "learning_rate": 2.5385638625446482e-06, + "loss": 0.5124, + "step": 12538 + }, + { + "epoch": 4.63, + "learning_rate": 2.538526912181303e-06, + "loss": 0.5156, + "step": 12539 + }, + { + "epoch": 4.63, + "learning_rate": 2.538489961817958e-06, + "loss": 0.5591, + "step": 12540 + }, + { + "epoch": 4.63, + "learning_rate": 2.5384530114546126e-06, + "loss": 0.5238, + "step": 12541 + }, + { + "epoch": 4.63, + "learning_rate": 2.5384160610912676e-06, + "loss": 0.5753, + "step": 12542 + }, + { + "epoch": 4.63, + "learning_rate": 2.5383791107279223e-06, + "loss": 0.5579, + "step": 12543 + }, + { + "epoch": 4.63, + "learning_rate": 2.538342160364577e-06, + "loss": 0.4986, + "step": 12544 + }, + { + "epoch": 4.63, + "learning_rate": 2.5383052100012315e-06, + "loss": 0.616, + "step": 12545 + }, + { + "epoch": 4.63, + "learning_rate": 2.5382682596378866e-06, + "loss": 0.6073, + "step": 12546 + }, + { + "epoch": 4.63, + "learning_rate": 2.5382313092745412e-06, + "loss": 0.5186, + "step": 12547 + }, + { + "epoch": 4.63, + "learning_rate": 2.538194358911196e-06, + "loss": 0.4993, + "step": 12548 + }, + { + "epoch": 4.63, + "learning_rate": 2.538157408547851e-06, + "loss": 0.5965, + "step": 12549 + }, + { + "epoch": 4.63, + "learning_rate": 2.5381204581845056e-06, + "loss": 0.4077, + "step": 12550 + }, + { + "epoch": 4.63, + "learning_rate": 2.53808350782116e-06, + "loss": 0.527, + "step": 12551 + }, + { + "epoch": 4.64, + "learning_rate": 2.538046557457815e-06, + "loss": 0.5164, + "step": 12552 + }, + { + "epoch": 4.64, + "learning_rate": 2.53800960709447e-06, + "loss": 0.536, + "step": 12553 + }, + { + "epoch": 4.64, + "learning_rate": 2.5379726567311245e-06, + "loss": 0.4492, + "step": 12554 + }, + { + "epoch": 4.64, + "learning_rate": 2.5379357063677796e-06, + "loss": 0.4805, + "step": 12555 + }, + { + "epoch": 4.64, + "learning_rate": 2.5378987560044342e-06, + "loss": 0.4128, + "step": 12556 + }, + { + "epoch": 4.64, + "learning_rate": 2.537861805641089e-06, + "loss": 0.5136, + "step": 12557 + }, + { + "epoch": 4.64, + "learning_rate": 2.5378248552777435e-06, + "loss": 0.4687, + "step": 12558 + }, + { + "epoch": 4.64, + "learning_rate": 2.537787904914398e-06, + "loss": 0.4823, + "step": 12559 + }, + { + "epoch": 4.64, + "learning_rate": 2.537750954551053e-06, + "loss": 0.4954, + "step": 12560 + }, + { + "epoch": 4.64, + "learning_rate": 2.537714004187708e-06, + "loss": 0.4506, + "step": 12561 + }, + { + "epoch": 4.64, + "learning_rate": 2.537677053824363e-06, + "loss": 0.5125, + "step": 12562 + }, + { + "epoch": 4.64, + "learning_rate": 2.5376401034610175e-06, + "loss": 0.5352, + "step": 12563 + }, + { + "epoch": 4.64, + "learning_rate": 2.537603153097672e-06, + "loss": 0.5279, + "step": 12564 + }, + { + "epoch": 4.64, + "learning_rate": 2.537566202734327e-06, + "loss": 0.5161, + "step": 12565 + }, + { + "epoch": 4.64, + "learning_rate": 2.537529252370982e-06, + "loss": 0.58, + "step": 12566 + }, + { + "epoch": 4.64, + "learning_rate": 2.5374923020076365e-06, + "loss": 0.4474, + "step": 12567 + }, + { + "epoch": 4.64, + "learning_rate": 2.537455351644291e-06, + "loss": 0.5682, + "step": 12568 + }, + { + "epoch": 4.64, + "learning_rate": 2.537418401280946e-06, + "loss": 0.5189, + "step": 12569 + }, + { + "epoch": 4.64, + "learning_rate": 2.537381450917601e-06, + "loss": 0.5266, + "step": 12570 + }, + { + "epoch": 4.64, + "learning_rate": 2.5373445005542555e-06, + "loss": 0.4767, + "step": 12571 + }, + { + "epoch": 4.64, + "learning_rate": 2.53730755019091e-06, + "loss": 0.5746, + "step": 12572 + }, + { + "epoch": 4.64, + "learning_rate": 2.537270599827565e-06, + "loss": 0.6353, + "step": 12573 + }, + { + "epoch": 4.64, + "learning_rate": 2.53723364946422e-06, + "loss": 0.4828, + "step": 12574 + }, + { + "epoch": 4.64, + "learning_rate": 2.537196699100875e-06, + "loss": 0.4846, + "step": 12575 + }, + { + "epoch": 4.64, + "learning_rate": 2.5371597487375295e-06, + "loss": 0.4712, + "step": 12576 + }, + { + "epoch": 4.64, + "learning_rate": 2.5371227983741837e-06, + "loss": 0.5646, + "step": 12577 + }, + { + "epoch": 4.64, + "learning_rate": 2.5370858480108388e-06, + "loss": 0.5768, + "step": 12578 + }, + { + "epoch": 4.65, + "learning_rate": 2.5370488976474934e-06, + "loss": 0.7022, + "step": 12579 + }, + { + "epoch": 4.65, + "learning_rate": 2.5370119472841485e-06, + "loss": 0.4416, + "step": 12580 + }, + { + "epoch": 4.65, + "learning_rate": 2.536974996920803e-06, + "loss": 0.6244, + "step": 12581 + }, + { + "epoch": 4.65, + "learning_rate": 2.536938046557458e-06, + "loss": 0.483, + "step": 12582 + }, + { + "epoch": 4.65, + "learning_rate": 2.5369010961941124e-06, + "loss": 0.5403, + "step": 12583 + }, + { + "epoch": 4.65, + "learning_rate": 2.5368641458307674e-06, + "loss": 0.6299, + "step": 12584 + }, + { + "epoch": 4.65, + "learning_rate": 2.536827195467422e-06, + "loss": 0.5286, + "step": 12585 + }, + { + "epoch": 4.65, + "learning_rate": 2.5367902451040767e-06, + "loss": 0.409, + "step": 12586 + }, + { + "epoch": 4.65, + "learning_rate": 2.5367532947407318e-06, + "loss": 0.5802, + "step": 12587 + }, + { + "epoch": 4.65, + "learning_rate": 2.5367163443773864e-06, + "loss": 0.481, + "step": 12588 + }, + { + "epoch": 4.65, + "learning_rate": 2.536679394014041e-06, + "loss": 0.5372, + "step": 12589 + }, + { + "epoch": 4.65, + "learning_rate": 2.5366424436506957e-06, + "loss": 0.4793, + "step": 12590 + }, + { + "epoch": 4.65, + "learning_rate": 2.5366054932873507e-06, + "loss": 0.4732, + "step": 12591 + }, + { + "epoch": 4.65, + "learning_rate": 2.5365685429240054e-06, + "loss": 0.6514, + "step": 12592 + }, + { + "epoch": 4.65, + "learning_rate": 2.5365315925606604e-06, + "loss": 0.4998, + "step": 12593 + }, + { + "epoch": 4.65, + "learning_rate": 2.536494642197315e-06, + "loss": 0.5694, + "step": 12594 + }, + { + "epoch": 4.65, + "learning_rate": 2.5364576918339697e-06, + "loss": 0.4631, + "step": 12595 + }, + { + "epoch": 4.65, + "learning_rate": 2.5364207414706244e-06, + "loss": 0.5713, + "step": 12596 + }, + { + "epoch": 4.65, + "learning_rate": 2.536383791107279e-06, + "loss": 0.4752, + "step": 12597 + }, + { + "epoch": 4.65, + "learning_rate": 2.536346840743934e-06, + "loss": 0.631, + "step": 12598 + }, + { + "epoch": 4.65, + "learning_rate": 2.5363098903805887e-06, + "loss": 0.4866, + "step": 12599 + }, + { + "epoch": 4.65, + "learning_rate": 2.5362729400172437e-06, + "loss": 0.4486, + "step": 12600 + }, + { + "epoch": 4.65, + "learning_rate": 2.5362359896538984e-06, + "loss": 0.6149, + "step": 12601 + }, + { + "epoch": 4.65, + "learning_rate": 2.536199039290553e-06, + "loss": 0.5613, + "step": 12602 + }, + { + "epoch": 4.65, + "learning_rate": 2.5361620889272077e-06, + "loss": 0.5087, + "step": 12603 + }, + { + "epoch": 4.65, + "learning_rate": 2.5361251385638627e-06, + "loss": 0.4828, + "step": 12604 + }, + { + "epoch": 4.65, + "learning_rate": 2.5360881882005174e-06, + "loss": 0.6604, + "step": 12605 + }, + { + "epoch": 4.66, + "learning_rate": 2.536051237837172e-06, + "loss": 0.6364, + "step": 12606 + }, + { + "epoch": 4.66, + "learning_rate": 2.536014287473827e-06, + "loss": 0.5879, + "step": 12607 + }, + { + "epoch": 4.66, + "learning_rate": 2.5359773371104817e-06, + "loss": 0.4604, + "step": 12608 + }, + { + "epoch": 4.66, + "learning_rate": 2.5359403867471363e-06, + "loss": 0.4339, + "step": 12609 + }, + { + "epoch": 4.66, + "learning_rate": 2.535903436383791e-06, + "loss": 0.5479, + "step": 12610 + }, + { + "epoch": 4.66, + "learning_rate": 2.535866486020446e-06, + "loss": 0.5538, + "step": 12611 + }, + { + "epoch": 4.66, + "learning_rate": 2.5358295356571007e-06, + "loss": 0.6961, + "step": 12612 + }, + { + "epoch": 4.66, + "learning_rate": 2.5357925852937557e-06, + "loss": 0.4658, + "step": 12613 + }, + { + "epoch": 4.66, + "learning_rate": 2.5357556349304104e-06, + "loss": 0.5084, + "step": 12614 + }, + { + "epoch": 4.66, + "learning_rate": 2.535718684567065e-06, + "loss": 0.4704, + "step": 12615 + }, + { + "epoch": 4.66, + "learning_rate": 2.5356817342037196e-06, + "loss": 0.694, + "step": 12616 + }, + { + "epoch": 4.66, + "learning_rate": 2.5356447838403743e-06, + "loss": 0.4659, + "step": 12617 + }, + { + "epoch": 4.66, + "learning_rate": 2.5356078334770293e-06, + "loss": 0.4338, + "step": 12618 + }, + { + "epoch": 4.66, + "learning_rate": 2.535570883113684e-06, + "loss": 0.5014, + "step": 12619 + }, + { + "epoch": 4.66, + "learning_rate": 2.535533932750339e-06, + "loss": 0.4957, + "step": 12620 + }, + { + "epoch": 4.66, + "learning_rate": 2.5354969823869937e-06, + "loss": 0.484, + "step": 12621 + }, + { + "epoch": 4.66, + "learning_rate": 2.5354600320236483e-06, + "loss": 0.5847, + "step": 12622 + }, + { + "epoch": 4.66, + "learning_rate": 2.535423081660303e-06, + "loss": 0.4975, + "step": 12623 + }, + { + "epoch": 4.66, + "learning_rate": 2.535386131296958e-06, + "loss": 0.5161, + "step": 12624 + }, + { + "epoch": 4.66, + "learning_rate": 2.5353491809336126e-06, + "loss": 0.5069, + "step": 12625 + }, + { + "epoch": 4.66, + "learning_rate": 2.5353122305702673e-06, + "loss": 0.4469, + "step": 12626 + }, + { + "epoch": 4.66, + "learning_rate": 2.5352752802069223e-06, + "loss": 0.3797, + "step": 12627 + }, + { + "epoch": 4.66, + "learning_rate": 2.5352383298435765e-06, + "loss": 0.4784, + "step": 12628 + }, + { + "epoch": 4.66, + "learning_rate": 2.5352013794802316e-06, + "loss": 0.5072, + "step": 12629 + }, + { + "epoch": 4.66, + "learning_rate": 2.5351644291168862e-06, + "loss": 0.5618, + "step": 12630 + }, + { + "epoch": 4.66, + "learning_rate": 2.5351274787535413e-06, + "loss": 0.4143, + "step": 12631 + }, + { + "epoch": 4.66, + "learning_rate": 2.535090528390196e-06, + "loss": 0.5542, + "step": 12632 + }, + { + "epoch": 4.67, + "learning_rate": 2.535053578026851e-06, + "loss": 0.4373, + "step": 12633 + }, + { + "epoch": 4.67, + "learning_rate": 2.535016627663505e-06, + "loss": 0.4135, + "step": 12634 + }, + { + "epoch": 4.67, + "learning_rate": 2.5349796773001603e-06, + "loss": 0.4893, + "step": 12635 + }, + { + "epoch": 4.67, + "learning_rate": 2.534942726936815e-06, + "loss": 0.5246, + "step": 12636 + }, + { + "epoch": 4.67, + "learning_rate": 2.5349057765734695e-06, + "loss": 0.4063, + "step": 12637 + }, + { + "epoch": 4.67, + "learning_rate": 2.5348688262101246e-06, + "loss": 0.4782, + "step": 12638 + }, + { + "epoch": 4.67, + "learning_rate": 2.5348318758467792e-06, + "loss": 0.3979, + "step": 12639 + }, + { + "epoch": 4.67, + "learning_rate": 2.534794925483434e-06, + "loss": 0.557, + "step": 12640 + }, + { + "epoch": 4.67, + "learning_rate": 2.5347579751200885e-06, + "loss": 0.482, + "step": 12641 + }, + { + "epoch": 4.67, + "learning_rate": 2.5347210247567436e-06, + "loss": 0.6085, + "step": 12642 + }, + { + "epoch": 4.67, + "learning_rate": 2.534684074393398e-06, + "loss": 0.4451, + "step": 12643 + }, + { + "epoch": 4.67, + "learning_rate": 2.5346471240300533e-06, + "loss": 0.458, + "step": 12644 + }, + { + "epoch": 4.67, + "learning_rate": 2.534610173666708e-06, + "loss": 0.5952, + "step": 12645 + }, + { + "epoch": 4.67, + "learning_rate": 2.5345732233033625e-06, + "loss": 0.5133, + "step": 12646 + }, + { + "epoch": 4.67, + "learning_rate": 2.534536272940017e-06, + "loss": 0.4368, + "step": 12647 + }, + { + "epoch": 4.67, + "learning_rate": 2.534499322576672e-06, + "loss": 0.6015, + "step": 12648 + }, + { + "epoch": 4.67, + "learning_rate": 2.534462372213327e-06, + "loss": 0.4074, + "step": 12649 + }, + { + "epoch": 4.67, + "learning_rate": 2.5344254218499815e-06, + "loss": 0.57, + "step": 12650 + }, + { + "epoch": 4.67, + "learning_rate": 2.5343884714866366e-06, + "loss": 0.5024, + "step": 12651 + }, + { + "epoch": 4.67, + "learning_rate": 2.534351521123291e-06, + "loss": 0.6091, + "step": 12652 + }, + { + "epoch": 4.67, + "learning_rate": 2.534314570759946e-06, + "loss": 0.4676, + "step": 12653 + }, + { + "epoch": 4.67, + "learning_rate": 2.5342776203966005e-06, + "loss": 0.579, + "step": 12654 + }, + { + "epoch": 4.67, + "learning_rate": 2.5342406700332555e-06, + "loss": 0.4955, + "step": 12655 + }, + { + "epoch": 4.67, + "learning_rate": 2.53420371966991e-06, + "loss": 0.628, + "step": 12656 + }, + { + "epoch": 4.67, + "learning_rate": 2.534166769306565e-06, + "loss": 0.5917, + "step": 12657 + }, + { + "epoch": 4.67, + "learning_rate": 2.53412981894322e-06, + "loss": 0.5212, + "step": 12658 + }, + { + "epoch": 4.67, + "learning_rate": 2.5340928685798745e-06, + "loss": 0.5328, + "step": 12659 + }, + { + "epoch": 4.68, + "learning_rate": 2.534055918216529e-06, + "loss": 0.5735, + "step": 12660 + }, + { + "epoch": 4.68, + "learning_rate": 2.5340189678531838e-06, + "loss": 0.5319, + "step": 12661 + }, + { + "epoch": 4.68, + "learning_rate": 2.533982017489839e-06, + "loss": 0.5274, + "step": 12662 + }, + { + "epoch": 4.68, + "learning_rate": 2.5339450671264935e-06, + "loss": 0.5365, + "step": 12663 + }, + { + "epoch": 4.68, + "learning_rate": 2.5339081167631485e-06, + "loss": 0.4977, + "step": 12664 + }, + { + "epoch": 4.68, + "learning_rate": 2.533871166399803e-06, + "loss": 0.5466, + "step": 12665 + }, + { + "epoch": 4.68, + "learning_rate": 2.533834216036458e-06, + "loss": 0.4077, + "step": 12666 + }, + { + "epoch": 4.68, + "learning_rate": 2.5337972656731124e-06, + "loss": 0.5089, + "step": 12667 + }, + { + "epoch": 4.68, + "learning_rate": 2.533760315309767e-06, + "loss": 0.4063, + "step": 12668 + }, + { + "epoch": 4.68, + "learning_rate": 2.533723364946422e-06, + "loss": 0.5958, + "step": 12669 + }, + { + "epoch": 4.68, + "learning_rate": 2.5336864145830768e-06, + "loss": 0.4092, + "step": 12670 + }, + { + "epoch": 4.68, + "learning_rate": 2.533649464219732e-06, + "loss": 0.5749, + "step": 12671 + }, + { + "epoch": 4.68, + "learning_rate": 2.5336125138563865e-06, + "loss": 0.539, + "step": 12672 + }, + { + "epoch": 4.68, + "learning_rate": 2.533575563493041e-06, + "loss": 0.511, + "step": 12673 + }, + { + "epoch": 4.68, + "learning_rate": 2.5335386131296957e-06, + "loss": 0.5566, + "step": 12674 + }, + { + "epoch": 4.68, + "learning_rate": 2.533501662766351e-06, + "loss": 0.4881, + "step": 12675 + }, + { + "epoch": 4.68, + "learning_rate": 2.5334647124030054e-06, + "loss": 0.4595, + "step": 12676 + }, + { + "epoch": 4.68, + "learning_rate": 2.53342776203966e-06, + "loss": 0.5088, + "step": 12677 + }, + { + "epoch": 4.68, + "learning_rate": 2.533390811676315e-06, + "loss": 0.4849, + "step": 12678 + }, + { + "epoch": 4.68, + "learning_rate": 2.5333538613129693e-06, + "loss": 0.4461, + "step": 12679 + }, + { + "epoch": 4.68, + "learning_rate": 2.5333169109496244e-06, + "loss": 0.5112, + "step": 12680 + }, + { + "epoch": 4.68, + "learning_rate": 2.533279960586279e-06, + "loss": 0.5203, + "step": 12681 + }, + { + "epoch": 4.68, + "learning_rate": 2.533243010222934e-06, + "loss": 0.6077, + "step": 12682 + }, + { + "epoch": 4.68, + "learning_rate": 2.5332060598595887e-06, + "loss": 0.6469, + "step": 12683 + }, + { + "epoch": 4.68, + "learning_rate": 2.533169109496244e-06, + "loss": 0.5569, + "step": 12684 + }, + { + "epoch": 4.68, + "learning_rate": 2.533132159132898e-06, + "loss": 0.5665, + "step": 12685 + }, + { + "epoch": 4.68, + "learning_rate": 2.5330952087695526e-06, + "loss": 0.5175, + "step": 12686 + }, + { + "epoch": 4.69, + "learning_rate": 2.5330582584062077e-06, + "loss": 0.5987, + "step": 12687 + }, + { + "epoch": 4.69, + "learning_rate": 2.5330213080428623e-06, + "loss": 0.599, + "step": 12688 + }, + { + "epoch": 4.69, + "learning_rate": 2.5329843576795174e-06, + "loss": 0.5489, + "step": 12689 + }, + { + "epoch": 4.69, + "learning_rate": 2.532947407316172e-06, + "loss": 0.4435, + "step": 12690 + }, + { + "epoch": 4.69, + "learning_rate": 2.5329104569528267e-06, + "loss": 0.4831, + "step": 12691 + }, + { + "epoch": 4.69, + "learning_rate": 2.5328735065894813e-06, + "loss": 0.4834, + "step": 12692 + }, + { + "epoch": 4.69, + "learning_rate": 2.5328365562261364e-06, + "loss": 0.5199, + "step": 12693 + }, + { + "epoch": 4.69, + "learning_rate": 2.532799605862791e-06, + "loss": 0.5302, + "step": 12694 + }, + { + "epoch": 4.69, + "learning_rate": 2.532762655499446e-06, + "loss": 0.495, + "step": 12695 + }, + { + "epoch": 4.69, + "learning_rate": 2.5327257051361007e-06, + "loss": 0.5711, + "step": 12696 + }, + { + "epoch": 4.69, + "learning_rate": 2.5326887547727553e-06, + "loss": 0.5455, + "step": 12697 + }, + { + "epoch": 4.69, + "learning_rate": 2.53265180440941e-06, + "loss": 0.5839, + "step": 12698 + }, + { + "epoch": 4.69, + "learning_rate": 2.5326148540460646e-06, + "loss": 0.4806, + "step": 12699 + }, + { + "epoch": 4.69, + "learning_rate": 2.5325779036827197e-06, + "loss": 0.4895, + "step": 12700 + }, + { + "epoch": 4.69, + "learning_rate": 2.5325409533193743e-06, + "loss": 0.494, + "step": 12701 + }, + { + "epoch": 4.69, + "learning_rate": 2.5325040029560294e-06, + "loss": 0.5762, + "step": 12702 + }, + { + "epoch": 4.69, + "learning_rate": 2.532467052592684e-06, + "loss": 0.4939, + "step": 12703 + }, + { + "epoch": 4.69, + "learning_rate": 2.5324301022293386e-06, + "loss": 0.5343, + "step": 12704 + }, + { + "epoch": 4.69, + "learning_rate": 2.5323931518659933e-06, + "loss": 0.4211, + "step": 12705 + }, + { + "epoch": 4.69, + "learning_rate": 2.532356201502648e-06, + "loss": 0.5223, + "step": 12706 + }, + { + "epoch": 4.69, + "learning_rate": 2.532319251139303e-06, + "loss": 0.5023, + "step": 12707 + }, + { + "epoch": 4.69, + "learning_rate": 2.5322823007759576e-06, + "loss": 0.5026, + "step": 12708 + }, + { + "epoch": 4.69, + "learning_rate": 2.5322453504126127e-06, + "loss": 0.5579, + "step": 12709 + }, + { + "epoch": 4.69, + "learning_rate": 2.5322084000492673e-06, + "loss": 0.4583, + "step": 12710 + }, + { + "epoch": 4.69, + "learning_rate": 2.532171449685922e-06, + "loss": 0.5765, + "step": 12711 + }, + { + "epoch": 4.69, + "learning_rate": 2.5321344993225766e-06, + "loss": 0.472, + "step": 12712 + }, + { + "epoch": 4.69, + "learning_rate": 2.5320975489592316e-06, + "loss": 0.4065, + "step": 12713 + }, + { + "epoch": 4.69, + "learning_rate": 2.5320605985958863e-06, + "loss": 0.5153, + "step": 12714 + }, + { + "epoch": 4.7, + "learning_rate": 2.5320236482325413e-06, + "loss": 0.4769, + "step": 12715 + }, + { + "epoch": 4.7, + "learning_rate": 2.531986697869196e-06, + "loss": 0.5906, + "step": 12716 + }, + { + "epoch": 4.7, + "learning_rate": 2.5319497475058506e-06, + "loss": 0.4436, + "step": 12717 + }, + { + "epoch": 4.7, + "learning_rate": 2.5319127971425052e-06, + "loss": 0.5682, + "step": 12718 + }, + { + "epoch": 4.7, + "learning_rate": 2.53187584677916e-06, + "loss": 0.5193, + "step": 12719 + }, + { + "epoch": 4.7, + "learning_rate": 2.531838896415815e-06, + "loss": 0.4919, + "step": 12720 + }, + { + "epoch": 4.7, + "learning_rate": 2.5318019460524696e-06, + "loss": 0.4838, + "step": 12721 + }, + { + "epoch": 4.7, + "learning_rate": 2.5317649956891246e-06, + "loss": 0.5013, + "step": 12722 + }, + { + "epoch": 4.7, + "learning_rate": 2.531728045325779e-06, + "loss": 0.5091, + "step": 12723 + }, + { + "epoch": 4.7, + "learning_rate": 2.531691094962434e-06, + "loss": 0.4627, + "step": 12724 + }, + { + "epoch": 4.7, + "learning_rate": 2.5316541445990885e-06, + "loss": 0.5843, + "step": 12725 + }, + { + "epoch": 4.7, + "learning_rate": 2.531617194235743e-06, + "loss": 0.4947, + "step": 12726 + }, + { + "epoch": 4.7, + "learning_rate": 2.5315802438723982e-06, + "loss": 0.5384, + "step": 12727 + }, + { + "epoch": 4.7, + "learning_rate": 2.531543293509053e-06, + "loss": 0.5594, + "step": 12728 + }, + { + "epoch": 4.7, + "learning_rate": 2.5315063431457075e-06, + "loss": 0.502, + "step": 12729 + }, + { + "epoch": 4.7, + "learning_rate": 2.531469392782362e-06, + "loss": 0.4414, + "step": 12730 + }, + { + "epoch": 4.7, + "learning_rate": 2.5314324424190172e-06, + "loss": 0.5412, + "step": 12731 + }, + { + "epoch": 4.7, + "learning_rate": 2.531395492055672e-06, + "loss": 0.4668, + "step": 12732 + }, + { + "epoch": 4.7, + "learning_rate": 2.531358541692327e-06, + "loss": 0.5427, + "step": 12733 + }, + { + "epoch": 4.7, + "learning_rate": 2.5313215913289815e-06, + "loss": 0.5866, + "step": 12734 + }, + { + "epoch": 4.7, + "learning_rate": 2.531284640965636e-06, + "loss": 0.4792, + "step": 12735 + }, + { + "epoch": 4.7, + "learning_rate": 2.531247690602291e-06, + "loss": 0.5011, + "step": 12736 + }, + { + "epoch": 4.7, + "learning_rate": 2.5312107402389455e-06, + "loss": 0.5206, + "step": 12737 + }, + { + "epoch": 4.7, + "learning_rate": 2.5311737898756005e-06, + "loss": 0.5489, + "step": 12738 + }, + { + "epoch": 4.7, + "learning_rate": 2.531136839512255e-06, + "loss": 0.5089, + "step": 12739 + }, + { + "epoch": 4.7, + "learning_rate": 2.5310998891489102e-06, + "loss": 0.5237, + "step": 12740 + }, + { + "epoch": 4.7, + "learning_rate": 2.531062938785565e-06, + "loss": 0.4879, + "step": 12741 + }, + { + "epoch": 4.71, + "learning_rate": 2.5310259884222195e-06, + "loss": 0.4491, + "step": 12742 + }, + { + "epoch": 4.71, + "learning_rate": 2.530989038058874e-06, + "loss": 0.4921, + "step": 12743 + }, + { + "epoch": 4.71, + "learning_rate": 2.530952087695529e-06, + "loss": 0.5627, + "step": 12744 + }, + { + "epoch": 4.71, + "learning_rate": 2.530915137332184e-06, + "loss": 0.5256, + "step": 12745 + }, + { + "epoch": 4.71, + "learning_rate": 2.5308781869688385e-06, + "loss": 0.4599, + "step": 12746 + }, + { + "epoch": 4.71, + "learning_rate": 2.5308412366054935e-06, + "loss": 0.5325, + "step": 12747 + }, + { + "epoch": 4.71, + "learning_rate": 2.530804286242148e-06, + "loss": 0.5628, + "step": 12748 + }, + { + "epoch": 4.71, + "learning_rate": 2.5307673358788028e-06, + "loss": 0.4608, + "step": 12749 + }, + { + "epoch": 4.71, + "learning_rate": 2.5307303855154574e-06, + "loss": 0.4731, + "step": 12750 + }, + { + "epoch": 4.71, + "learning_rate": 2.5306934351521125e-06, + "loss": 0.6495, + "step": 12751 + }, + { + "epoch": 4.71, + "learning_rate": 2.530656484788767e-06, + "loss": 0.503, + "step": 12752 + }, + { + "epoch": 4.71, + "learning_rate": 2.530619534425422e-06, + "loss": 0.5323, + "step": 12753 + }, + { + "epoch": 4.71, + "learning_rate": 2.530582584062077e-06, + "loss": 0.5566, + "step": 12754 + }, + { + "epoch": 4.71, + "learning_rate": 2.5305456336987315e-06, + "loss": 0.5696, + "step": 12755 + }, + { + "epoch": 4.71, + "learning_rate": 2.530508683335386e-06, + "loss": 0.5117, + "step": 12756 + }, + { + "epoch": 4.71, + "learning_rate": 2.5304717329720407e-06, + "loss": 0.4218, + "step": 12757 + }, + { + "epoch": 4.71, + "learning_rate": 2.5304347826086958e-06, + "loss": 0.4957, + "step": 12758 + }, + { + "epoch": 4.71, + "learning_rate": 2.5303978322453504e-06, + "loss": 0.4652, + "step": 12759 + }, + { + "epoch": 4.71, + "learning_rate": 2.5303608818820055e-06, + "loss": 0.4679, + "step": 12760 + }, + { + "epoch": 4.71, + "learning_rate": 2.53032393151866e-06, + "loss": 0.4105, + "step": 12761 + }, + { + "epoch": 4.71, + "learning_rate": 2.5302869811553148e-06, + "loss": 0.5232, + "step": 12762 + }, + { + "epoch": 4.71, + "learning_rate": 2.5302500307919694e-06, + "loss": 0.4502, + "step": 12763 + }, + { + "epoch": 4.71, + "learning_rate": 2.5302130804286245e-06, + "loss": 0.4887, + "step": 12764 + }, + { + "epoch": 4.71, + "learning_rate": 2.530176130065279e-06, + "loss": 0.3617, + "step": 12765 + }, + { + "epoch": 4.71, + "learning_rate": 2.5301391797019337e-06, + "loss": 0.497, + "step": 12766 + }, + { + "epoch": 4.71, + "learning_rate": 2.5301022293385888e-06, + "loss": 0.4695, + "step": 12767 + }, + { + "epoch": 4.71, + "learning_rate": 2.530065278975243e-06, + "loss": 0.5256, + "step": 12768 + }, + { + "epoch": 4.72, + "learning_rate": 2.530028328611898e-06, + "loss": 0.5021, + "step": 12769 + }, + { + "epoch": 4.72, + "learning_rate": 2.5299913782485527e-06, + "loss": 0.6046, + "step": 12770 + }, + { + "epoch": 4.72, + "learning_rate": 2.5299544278852078e-06, + "loss": 0.5181, + "step": 12771 + }, + { + "epoch": 4.72, + "learning_rate": 2.5299174775218624e-06, + "loss": 0.5562, + "step": 12772 + }, + { + "epoch": 4.72, + "learning_rate": 2.5298805271585175e-06, + "loss": 0.5572, + "step": 12773 + }, + { + "epoch": 4.72, + "learning_rate": 2.5298435767951717e-06, + "loss": 0.5594, + "step": 12774 + }, + { + "epoch": 4.72, + "learning_rate": 2.5298066264318267e-06, + "loss": 0.4831, + "step": 12775 + }, + { + "epoch": 4.72, + "learning_rate": 2.5297696760684814e-06, + "loss": 0.4672, + "step": 12776 + }, + { + "epoch": 4.72, + "learning_rate": 2.529732725705136e-06, + "loss": 0.4666, + "step": 12777 + }, + { + "epoch": 4.72, + "learning_rate": 2.529695775341791e-06, + "loss": 0.3841, + "step": 12778 + }, + { + "epoch": 4.72, + "learning_rate": 2.5296588249784457e-06, + "loss": 0.5483, + "step": 12779 + }, + { + "epoch": 4.72, + "learning_rate": 2.5296218746151003e-06, + "loss": 0.4245, + "step": 12780 + }, + { + "epoch": 4.72, + "learning_rate": 2.529584924251755e-06, + "loss": 0.3994, + "step": 12781 + }, + { + "epoch": 4.72, + "learning_rate": 2.52954797388841e-06, + "loss": 0.6438, + "step": 12782 + }, + { + "epoch": 4.72, + "learning_rate": 2.5295110235250647e-06, + "loss": 0.4982, + "step": 12783 + }, + { + "epoch": 4.72, + "learning_rate": 2.5294740731617197e-06, + "loss": 0.5543, + "step": 12784 + }, + { + "epoch": 4.72, + "learning_rate": 2.5294371227983744e-06, + "loss": 0.5051, + "step": 12785 + }, + { + "epoch": 4.72, + "learning_rate": 2.529400172435029e-06, + "loss": 0.3901, + "step": 12786 + }, + { + "epoch": 4.72, + "learning_rate": 2.5293632220716836e-06, + "loss": 0.5286, + "step": 12787 + }, + { + "epoch": 4.72, + "learning_rate": 2.5293262717083383e-06, + "loss": 0.4952, + "step": 12788 + }, + { + "epoch": 4.72, + "learning_rate": 2.5292893213449933e-06, + "loss": 0.5113, + "step": 12789 + }, + { + "epoch": 4.72, + "learning_rate": 2.529252370981648e-06, + "loss": 0.4912, + "step": 12790 + }, + { + "epoch": 4.72, + "learning_rate": 2.529215420618303e-06, + "loss": 0.5881, + "step": 12791 + }, + { + "epoch": 4.72, + "learning_rate": 2.5291784702549577e-06, + "loss": 0.561, + "step": 12792 + }, + { + "epoch": 4.72, + "learning_rate": 2.5291415198916123e-06, + "loss": 0.4615, + "step": 12793 + }, + { + "epoch": 4.72, + "learning_rate": 2.529104569528267e-06, + "loss": 0.5822, + "step": 12794 + }, + { + "epoch": 4.72, + "learning_rate": 2.529067619164922e-06, + "loss": 0.3762, + "step": 12795 + }, + { + "epoch": 4.73, + "learning_rate": 2.5290306688015766e-06, + "loss": 0.4205, + "step": 12796 + }, + { + "epoch": 4.73, + "learning_rate": 2.5289937184382313e-06, + "loss": 0.4613, + "step": 12797 + }, + { + "epoch": 4.73, + "learning_rate": 2.5289567680748863e-06, + "loss": 0.5062, + "step": 12798 + }, + { + "epoch": 4.73, + "learning_rate": 2.528919817711541e-06, + "loss": 0.5532, + "step": 12799 + }, + { + "epoch": 4.73, + "learning_rate": 2.5288828673481956e-06, + "loss": 0.5351, + "step": 12800 + }, + { + "epoch": 4.73, + "learning_rate": 2.5288459169848502e-06, + "loss": 0.5657, + "step": 12801 + }, + { + "epoch": 4.73, + "learning_rate": 2.5288089666215053e-06, + "loss": 0.5183, + "step": 12802 + }, + { + "epoch": 4.73, + "learning_rate": 2.52877201625816e-06, + "loss": 0.5734, + "step": 12803 + }, + { + "epoch": 4.73, + "learning_rate": 2.528735065894815e-06, + "loss": 0.5562, + "step": 12804 + }, + { + "epoch": 4.73, + "learning_rate": 2.5286981155314696e-06, + "loss": 0.544, + "step": 12805 + }, + { + "epoch": 4.73, + "learning_rate": 2.5286611651681243e-06, + "loss": 0.4519, + "step": 12806 + }, + { + "epoch": 4.73, + "learning_rate": 2.528624214804779e-06, + "loss": 0.5115, + "step": 12807 + }, + { + "epoch": 4.73, + "learning_rate": 2.5285872644414335e-06, + "loss": 0.5511, + "step": 12808 + }, + { + "epoch": 4.73, + "learning_rate": 2.5285503140780886e-06, + "loss": 0.5958, + "step": 12809 + }, + { + "epoch": 4.73, + "learning_rate": 2.5285133637147432e-06, + "loss": 0.4984, + "step": 12810 + }, + { + "epoch": 4.73, + "learning_rate": 2.5284764133513983e-06, + "loss": 0.431, + "step": 12811 + }, + { + "epoch": 4.73, + "learning_rate": 2.528439462988053e-06, + "loss": 0.5028, + "step": 12812 + }, + { + "epoch": 4.73, + "learning_rate": 2.5284025126247076e-06, + "loss": 0.6545, + "step": 12813 + }, + { + "epoch": 4.73, + "learning_rate": 2.528365562261362e-06, + "loss": 0.5212, + "step": 12814 + }, + { + "epoch": 4.73, + "learning_rate": 2.528328611898017e-06, + "loss": 0.5727, + "step": 12815 + }, + { + "epoch": 4.73, + "learning_rate": 2.528291661534672e-06, + "loss": 0.4462, + "step": 12816 + }, + { + "epoch": 4.73, + "learning_rate": 2.5282547111713265e-06, + "loss": 0.535, + "step": 12817 + }, + { + "epoch": 4.73, + "learning_rate": 2.5282177608079816e-06, + "loss": 0.5141, + "step": 12818 + }, + { + "epoch": 4.73, + "learning_rate": 2.528180810444636e-06, + "loss": 0.5476, + "step": 12819 + }, + { + "epoch": 4.73, + "learning_rate": 2.528143860081291e-06, + "loss": 0.5331, + "step": 12820 + }, + { + "epoch": 4.73, + "learning_rate": 2.5281069097179455e-06, + "loss": 0.4918, + "step": 12821 + }, + { + "epoch": 4.73, + "learning_rate": 2.5280699593546006e-06, + "loss": 0.488, + "step": 12822 + }, + { + "epoch": 4.74, + "learning_rate": 2.528033008991255e-06, + "loss": 0.5004, + "step": 12823 + }, + { + "epoch": 4.74, + "learning_rate": 2.5279960586279103e-06, + "loss": 0.572, + "step": 12824 + }, + { + "epoch": 4.74, + "learning_rate": 2.5279591082645645e-06, + "loss": 0.4617, + "step": 12825 + }, + { + "epoch": 4.74, + "learning_rate": 2.527922157901219e-06, + "loss": 0.4901, + "step": 12826 + }, + { + "epoch": 4.74, + "learning_rate": 2.527885207537874e-06, + "loss": 0.6172, + "step": 12827 + }, + { + "epoch": 4.74, + "learning_rate": 2.527848257174529e-06, + "loss": 0.5156, + "step": 12828 + }, + { + "epoch": 4.74, + "learning_rate": 2.527811306811184e-06, + "loss": 0.4538, + "step": 12829 + }, + { + "epoch": 4.74, + "learning_rate": 2.5277743564478385e-06, + "loss": 0.4452, + "step": 12830 + }, + { + "epoch": 4.74, + "learning_rate": 2.527737406084493e-06, + "loss": 0.4889, + "step": 12831 + }, + { + "epoch": 4.74, + "learning_rate": 2.5277004557211478e-06, + "loss": 0.4845, + "step": 12832 + }, + { + "epoch": 4.74, + "learning_rate": 2.527663505357803e-06, + "loss": 0.433, + "step": 12833 + }, + { + "epoch": 4.74, + "learning_rate": 2.5276265549944575e-06, + "loss": 0.6378, + "step": 12834 + }, + { + "epoch": 4.74, + "learning_rate": 2.527589604631112e-06, + "loss": 0.4962, + "step": 12835 + }, + { + "epoch": 4.74, + "learning_rate": 2.527552654267767e-06, + "loss": 0.492, + "step": 12836 + }, + { + "epoch": 4.74, + "learning_rate": 2.527515703904422e-06, + "loss": 0.5122, + "step": 12837 + }, + { + "epoch": 4.74, + "learning_rate": 2.5274787535410764e-06, + "loss": 0.3784, + "step": 12838 + }, + { + "epoch": 4.74, + "learning_rate": 2.527441803177731e-06, + "loss": 0.5777, + "step": 12839 + }, + { + "epoch": 4.74, + "learning_rate": 2.527404852814386e-06, + "loss": 0.4231, + "step": 12840 + }, + { + "epoch": 4.74, + "learning_rate": 2.5273679024510408e-06, + "loss": 0.6385, + "step": 12841 + }, + { + "epoch": 4.74, + "learning_rate": 2.527330952087696e-06, + "loss": 0.532, + "step": 12842 + }, + { + "epoch": 4.74, + "learning_rate": 2.5272940017243505e-06, + "loss": 0.4291, + "step": 12843 + }, + { + "epoch": 4.74, + "learning_rate": 2.527257051361005e-06, + "loss": 0.4571, + "step": 12844 + }, + { + "epoch": 4.74, + "learning_rate": 2.5272201009976597e-06, + "loss": 0.4584, + "step": 12845 + }, + { + "epoch": 4.74, + "learning_rate": 2.5271831506343144e-06, + "loss": 0.4491, + "step": 12846 + }, + { + "epoch": 4.74, + "learning_rate": 2.5271462002709694e-06, + "loss": 0.4749, + "step": 12847 + }, + { + "epoch": 4.74, + "learning_rate": 2.527109249907624e-06, + "loss": 0.5132, + "step": 12848 + }, + { + "epoch": 4.74, + "learning_rate": 2.527072299544279e-06, + "loss": 0.4843, + "step": 12849 + }, + { + "epoch": 4.75, + "learning_rate": 2.5270353491809338e-06, + "loss": 0.4294, + "step": 12850 + }, + { + "epoch": 4.75, + "learning_rate": 2.5269983988175884e-06, + "loss": 0.5341, + "step": 12851 + }, + { + "epoch": 4.75, + "learning_rate": 2.526961448454243e-06, + "loss": 0.3907, + "step": 12852 + }, + { + "epoch": 4.75, + "learning_rate": 2.526924498090898e-06, + "loss": 0.3911, + "step": 12853 + }, + { + "epoch": 4.75, + "learning_rate": 2.5268875477275527e-06, + "loss": 0.4735, + "step": 12854 + }, + { + "epoch": 4.75, + "learning_rate": 2.5268505973642074e-06, + "loss": 0.4012, + "step": 12855 + }, + { + "epoch": 4.75, + "learning_rate": 2.5268136470008624e-06, + "loss": 0.5724, + "step": 12856 + }, + { + "epoch": 4.75, + "learning_rate": 2.526776696637517e-06, + "loss": 0.5838, + "step": 12857 + }, + { + "epoch": 4.75, + "learning_rate": 2.5267397462741717e-06, + "loss": 0.5821, + "step": 12858 + }, + { + "epoch": 4.75, + "learning_rate": 2.5267027959108263e-06, + "loss": 0.5218, + "step": 12859 + }, + { + "epoch": 4.75, + "learning_rate": 2.5266658455474814e-06, + "loss": 0.5583, + "step": 12860 + }, + { + "epoch": 4.75, + "learning_rate": 2.526628895184136e-06, + "loss": 0.5111, + "step": 12861 + }, + { + "epoch": 4.75, + "learning_rate": 2.526591944820791e-06, + "loss": 0.5228, + "step": 12862 + }, + { + "epoch": 4.75, + "learning_rate": 2.5265549944574457e-06, + "loss": 0.5316, + "step": 12863 + }, + { + "epoch": 4.75, + "learning_rate": 2.5265180440941004e-06, + "loss": 0.5149, + "step": 12864 + }, + { + "epoch": 4.75, + "learning_rate": 2.526481093730755e-06, + "loss": 0.5333, + "step": 12865 + }, + { + "epoch": 4.75, + "learning_rate": 2.5264441433674097e-06, + "loss": 0.6522, + "step": 12866 + }, + { + "epoch": 4.75, + "learning_rate": 2.5264071930040647e-06, + "loss": 0.4516, + "step": 12867 + }, + { + "epoch": 4.75, + "learning_rate": 2.5263702426407193e-06, + "loss": 0.4836, + "step": 12868 + }, + { + "epoch": 4.75, + "learning_rate": 2.5263332922773744e-06, + "loss": 0.476, + "step": 12869 + }, + { + "epoch": 4.75, + "learning_rate": 2.5262963419140286e-06, + "loss": 0.5844, + "step": 12870 + }, + { + "epoch": 4.75, + "learning_rate": 2.5262593915506837e-06, + "loss": 0.5979, + "step": 12871 + }, + { + "epoch": 4.75, + "learning_rate": 2.5262224411873383e-06, + "loss": 0.5575, + "step": 12872 + }, + { + "epoch": 4.75, + "learning_rate": 2.5261854908239934e-06, + "loss": 0.5259, + "step": 12873 + }, + { + "epoch": 4.75, + "learning_rate": 2.526148540460648e-06, + "loss": 0.5645, + "step": 12874 + }, + { + "epoch": 4.75, + "learning_rate": 2.5261115900973026e-06, + "loss": 0.4139, + "step": 12875 + }, + { + "epoch": 4.75, + "learning_rate": 2.5260746397339573e-06, + "loss": 0.453, + "step": 12876 + }, + { + "epoch": 4.76, + "learning_rate": 2.526037689370612e-06, + "loss": 0.4782, + "step": 12877 + }, + { + "epoch": 4.76, + "learning_rate": 2.526000739007267e-06, + "loss": 0.4541, + "step": 12878 + }, + { + "epoch": 4.76, + "learning_rate": 2.5259637886439216e-06, + "loss": 0.5957, + "step": 12879 + }, + { + "epoch": 4.76, + "learning_rate": 2.5259268382805767e-06, + "loss": 0.5877, + "step": 12880 + }, + { + "epoch": 4.76, + "learning_rate": 2.5258898879172313e-06, + "loss": 0.4566, + "step": 12881 + }, + { + "epoch": 4.76, + "learning_rate": 2.525852937553886e-06, + "loss": 0.4037, + "step": 12882 + }, + { + "epoch": 4.76, + "learning_rate": 2.5258159871905406e-06, + "loss": 0.512, + "step": 12883 + }, + { + "epoch": 4.76, + "learning_rate": 2.5257790368271956e-06, + "loss": 0.5063, + "step": 12884 + }, + { + "epoch": 4.76, + "learning_rate": 2.5257420864638503e-06, + "loss": 0.5404, + "step": 12885 + }, + { + "epoch": 4.76, + "learning_rate": 2.525705136100505e-06, + "loss": 0.4401, + "step": 12886 + }, + { + "epoch": 4.76, + "learning_rate": 2.52566818573716e-06, + "loss": 0.4014, + "step": 12887 + }, + { + "epoch": 4.76, + "learning_rate": 2.5256312353738146e-06, + "loss": 0.4185, + "step": 12888 + }, + { + "epoch": 4.76, + "learning_rate": 2.5255942850104693e-06, + "loss": 0.4935, + "step": 12889 + }, + { + "epoch": 4.76, + "learning_rate": 2.525557334647124e-06, + "loss": 0.5562, + "step": 12890 + }, + { + "epoch": 4.76, + "learning_rate": 2.525520384283779e-06, + "loss": 0.489, + "step": 12891 + }, + { + "epoch": 4.76, + "learning_rate": 2.5254834339204336e-06, + "loss": 0.4848, + "step": 12892 + }, + { + "epoch": 4.76, + "learning_rate": 2.5254464835570886e-06, + "loss": 0.3291, + "step": 12893 + }, + { + "epoch": 4.76, + "learning_rate": 2.5254095331937433e-06, + "loss": 0.4819, + "step": 12894 + }, + { + "epoch": 4.76, + "learning_rate": 2.525372582830398e-06, + "loss": 0.4821, + "step": 12895 + }, + { + "epoch": 4.76, + "learning_rate": 2.5253356324670526e-06, + "loss": 0.4248, + "step": 12896 + }, + { + "epoch": 4.76, + "learning_rate": 2.525298682103707e-06, + "loss": 0.4808, + "step": 12897 + }, + { + "epoch": 4.76, + "learning_rate": 2.5252617317403623e-06, + "loss": 0.5134, + "step": 12898 + }, + { + "epoch": 4.76, + "learning_rate": 2.525224781377017e-06, + "loss": 0.5559, + "step": 12899 + }, + { + "epoch": 4.76, + "learning_rate": 2.525187831013672e-06, + "loss": 0.5877, + "step": 12900 + }, + { + "epoch": 4.76, + "learning_rate": 2.5251508806503266e-06, + "loss": 0.5676, + "step": 12901 + }, + { + "epoch": 4.76, + "learning_rate": 2.5251139302869812e-06, + "loss": 0.4081, + "step": 12902 + }, + { + "epoch": 4.76, + "learning_rate": 2.525076979923636e-06, + "loss": 0.4718, + "step": 12903 + }, + { + "epoch": 4.77, + "learning_rate": 2.525040029560291e-06, + "loss": 0.444, + "step": 12904 + }, + { + "epoch": 4.77, + "learning_rate": 2.5250030791969456e-06, + "loss": 0.4334, + "step": 12905 + }, + { + "epoch": 4.77, + "learning_rate": 2.5249661288336e-06, + "loss": 0.4958, + "step": 12906 + }, + { + "epoch": 4.77, + "learning_rate": 2.5249291784702552e-06, + "loss": 0.4661, + "step": 12907 + }, + { + "epoch": 4.77, + "learning_rate": 2.5248922281069095e-06, + "loss": 0.4593, + "step": 12908 + }, + { + "epoch": 4.77, + "learning_rate": 2.5248552777435645e-06, + "loss": 0.4299, + "step": 12909 + }, + { + "epoch": 4.77, + "learning_rate": 2.524818327380219e-06, + "loss": 0.3599, + "step": 12910 + }, + { + "epoch": 4.77, + "learning_rate": 2.5247813770168742e-06, + "loss": 0.6254, + "step": 12911 + }, + { + "epoch": 4.77, + "learning_rate": 2.524744426653529e-06, + "loss": 0.4868, + "step": 12912 + }, + { + "epoch": 4.77, + "learning_rate": 2.524707476290184e-06, + "loss": 0.4994, + "step": 12913 + }, + { + "epoch": 4.77, + "learning_rate": 2.524670525926838e-06, + "loss": 0.6007, + "step": 12914 + }, + { + "epoch": 4.77, + "learning_rate": 2.5246335755634928e-06, + "loss": 0.5412, + "step": 12915 + }, + { + "epoch": 4.77, + "learning_rate": 2.524596625200148e-06, + "loss": 0.4349, + "step": 12916 + }, + { + "epoch": 4.77, + "learning_rate": 2.5245596748368025e-06, + "loss": 0.6, + "step": 12917 + }, + { + "epoch": 4.77, + "learning_rate": 2.5245227244734575e-06, + "loss": 0.4435, + "step": 12918 + }, + { + "epoch": 4.77, + "learning_rate": 2.524485774110112e-06, + "loss": 0.51, + "step": 12919 + }, + { + "epoch": 4.77, + "learning_rate": 2.524448823746767e-06, + "loss": 0.5471, + "step": 12920 + }, + { + "epoch": 4.77, + "learning_rate": 2.5244118733834214e-06, + "loss": 0.5131, + "step": 12921 + }, + { + "epoch": 4.77, + "learning_rate": 2.5243749230200765e-06, + "loss": 0.4175, + "step": 12922 + }, + { + "epoch": 4.77, + "learning_rate": 2.524337972656731e-06, + "loss": 0.4788, + "step": 12923 + }, + { + "epoch": 4.77, + "learning_rate": 2.524301022293386e-06, + "loss": 0.463, + "step": 12924 + }, + { + "epoch": 4.77, + "learning_rate": 2.524264071930041e-06, + "loss": 0.5177, + "step": 12925 + }, + { + "epoch": 4.77, + "learning_rate": 2.5242271215666955e-06, + "loss": 0.6768, + "step": 12926 + }, + { + "epoch": 4.77, + "learning_rate": 2.52419017120335e-06, + "loss": 0.4917, + "step": 12927 + }, + { + "epoch": 4.77, + "learning_rate": 2.5241532208400047e-06, + "loss": 0.4994, + "step": 12928 + }, + { + "epoch": 4.77, + "learning_rate": 2.52411627047666e-06, + "loss": 0.484, + "step": 12929 + }, + { + "epoch": 4.77, + "learning_rate": 2.5240793201133144e-06, + "loss": 0.6779, + "step": 12930 + }, + { + "epoch": 4.78, + "learning_rate": 2.5240423697499695e-06, + "loss": 0.5605, + "step": 12931 + }, + { + "epoch": 4.78, + "learning_rate": 2.524005419386624e-06, + "loss": 0.532, + "step": 12932 + }, + { + "epoch": 4.78, + "learning_rate": 2.5239684690232788e-06, + "loss": 0.4842, + "step": 12933 + }, + { + "epoch": 4.78, + "learning_rate": 2.5239315186599334e-06, + "loss": 0.4979, + "step": 12934 + }, + { + "epoch": 4.78, + "learning_rate": 2.523894568296588e-06, + "loss": 0.548, + "step": 12935 + }, + { + "epoch": 4.78, + "learning_rate": 2.523857617933243e-06, + "loss": 0.4778, + "step": 12936 + }, + { + "epoch": 4.78, + "learning_rate": 2.5238206675698977e-06, + "loss": 0.4148, + "step": 12937 + }, + { + "epoch": 4.78, + "learning_rate": 2.523783717206553e-06, + "loss": 0.4984, + "step": 12938 + }, + { + "epoch": 4.78, + "learning_rate": 2.5237467668432074e-06, + "loss": 0.5791, + "step": 12939 + }, + { + "epoch": 4.78, + "learning_rate": 2.523709816479862e-06, + "loss": 0.4901, + "step": 12940 + }, + { + "epoch": 4.78, + "learning_rate": 2.5236728661165167e-06, + "loss": 0.5551, + "step": 12941 + }, + { + "epoch": 4.78, + "learning_rate": 2.5236359157531718e-06, + "loss": 0.5493, + "step": 12942 + }, + { + "epoch": 4.78, + "learning_rate": 2.5235989653898264e-06, + "loss": 0.4943, + "step": 12943 + }, + { + "epoch": 4.78, + "learning_rate": 2.5235620150264815e-06, + "loss": 0.5081, + "step": 12944 + }, + { + "epoch": 4.78, + "learning_rate": 2.523525064663136e-06, + "loss": 0.513, + "step": 12945 + }, + { + "epoch": 4.78, + "learning_rate": 2.5234881142997907e-06, + "loss": 0.4534, + "step": 12946 + }, + { + "epoch": 4.78, + "learning_rate": 2.5234511639364454e-06, + "loss": 0.5287, + "step": 12947 + }, + { + "epoch": 4.78, + "learning_rate": 2.5234142135731e-06, + "loss": 0.4759, + "step": 12948 + }, + { + "epoch": 4.78, + "learning_rate": 2.523377263209755e-06, + "loss": 0.5017, + "step": 12949 + }, + { + "epoch": 4.78, + "learning_rate": 2.5233403128464097e-06, + "loss": 0.5472, + "step": 12950 + }, + { + "epoch": 4.78, + "learning_rate": 2.5233033624830648e-06, + "loss": 0.4829, + "step": 12951 + }, + { + "epoch": 4.78, + "learning_rate": 2.5232664121197194e-06, + "loss": 0.5401, + "step": 12952 + }, + { + "epoch": 4.78, + "learning_rate": 2.523229461756374e-06, + "loss": 0.4437, + "step": 12953 + }, + { + "epoch": 4.78, + "learning_rate": 2.5231925113930287e-06, + "loss": 0.5817, + "step": 12954 + }, + { + "epoch": 4.78, + "learning_rate": 2.5231555610296833e-06, + "loss": 0.4985, + "step": 12955 + }, + { + "epoch": 4.78, + "learning_rate": 2.5231186106663384e-06, + "loss": 0.4565, + "step": 12956 + }, + { + "epoch": 4.78, + "learning_rate": 2.523081660302993e-06, + "loss": 0.4801, + "step": 12957 + }, + { + "epoch": 4.79, + "learning_rate": 2.523044709939648e-06, + "loss": 0.5754, + "step": 12958 + }, + { + "epoch": 4.79, + "learning_rate": 2.5230077595763023e-06, + "loss": 0.4972, + "step": 12959 + }, + { + "epoch": 4.79, + "learning_rate": 2.5229708092129573e-06, + "loss": 0.4799, + "step": 12960 + }, + { + "epoch": 4.79, + "learning_rate": 2.522933858849612e-06, + "loss": 0.428, + "step": 12961 + }, + { + "epoch": 4.79, + "learning_rate": 2.522896908486267e-06, + "loss": 0.4712, + "step": 12962 + }, + { + "epoch": 4.79, + "learning_rate": 2.5228599581229217e-06, + "loss": 0.4901, + "step": 12963 + }, + { + "epoch": 4.79, + "learning_rate": 2.5228230077595767e-06, + "loss": 0.5675, + "step": 12964 + }, + { + "epoch": 4.79, + "learning_rate": 2.522786057396231e-06, + "loss": 0.4768, + "step": 12965 + }, + { + "epoch": 4.79, + "learning_rate": 2.5227491070328856e-06, + "loss": 0.4837, + "step": 12966 + }, + { + "epoch": 4.79, + "learning_rate": 2.5227121566695406e-06, + "loss": 0.5437, + "step": 12967 + }, + { + "epoch": 4.79, + "learning_rate": 2.5226752063061953e-06, + "loss": 0.4345, + "step": 12968 + }, + { + "epoch": 4.79, + "learning_rate": 2.5226382559428503e-06, + "loss": 0.5722, + "step": 12969 + }, + { + "epoch": 4.79, + "learning_rate": 2.522601305579505e-06, + "loss": 0.4419, + "step": 12970 + }, + { + "epoch": 4.79, + "learning_rate": 2.5225643552161596e-06, + "loss": 0.4519, + "step": 12971 + }, + { + "epoch": 4.79, + "learning_rate": 2.5225274048528142e-06, + "loss": 0.4919, + "step": 12972 + }, + { + "epoch": 4.79, + "learning_rate": 2.5224904544894693e-06, + "loss": 0.5449, + "step": 12973 + }, + { + "epoch": 4.79, + "learning_rate": 2.522453504126124e-06, + "loss": 0.4209, + "step": 12974 + }, + { + "epoch": 4.79, + "learning_rate": 2.5224165537627786e-06, + "loss": 0.5089, + "step": 12975 + }, + { + "epoch": 4.79, + "learning_rate": 2.5223796033994336e-06, + "loss": 0.6642, + "step": 12976 + }, + { + "epoch": 4.79, + "learning_rate": 2.5223426530360883e-06, + "loss": 0.5141, + "step": 12977 + }, + { + "epoch": 4.79, + "learning_rate": 2.522305702672743e-06, + "loss": 0.3841, + "step": 12978 + }, + { + "epoch": 4.79, + "learning_rate": 2.5222687523093975e-06, + "loss": 0.592, + "step": 12979 + }, + { + "epoch": 4.79, + "learning_rate": 2.5222318019460526e-06, + "loss": 0.5174, + "step": 12980 + }, + { + "epoch": 4.79, + "learning_rate": 2.5221948515827072e-06, + "loss": 0.4843, + "step": 12981 + }, + { + "epoch": 4.79, + "learning_rate": 2.5221579012193623e-06, + "loss": 0.63, + "step": 12982 + }, + { + "epoch": 4.79, + "learning_rate": 2.522120950856017e-06, + "loss": 0.3826, + "step": 12983 + }, + { + "epoch": 4.79, + "learning_rate": 2.5220840004926716e-06, + "loss": 0.4651, + "step": 12984 + }, + { + "epoch": 4.8, + "learning_rate": 2.522047050129326e-06, + "loss": 0.4264, + "step": 12985 + }, + { + "epoch": 4.8, + "learning_rate": 2.522010099765981e-06, + "loss": 0.741, + "step": 12986 + }, + { + "epoch": 4.8, + "learning_rate": 2.521973149402636e-06, + "loss": 0.5116, + "step": 12987 + }, + { + "epoch": 4.8, + "learning_rate": 2.5219361990392905e-06, + "loss": 0.5984, + "step": 12988 + }, + { + "epoch": 4.8, + "learning_rate": 2.5218992486759456e-06, + "loss": 0.601, + "step": 12989 + }, + { + "epoch": 4.8, + "learning_rate": 2.5218622983126002e-06, + "loss": 0.5728, + "step": 12990 + }, + { + "epoch": 4.8, + "learning_rate": 2.521825347949255e-06, + "loss": 0.4666, + "step": 12991 + }, + { + "epoch": 4.8, + "learning_rate": 2.5217883975859095e-06, + "loss": 0.5656, + "step": 12992 + }, + { + "epoch": 4.8, + "learning_rate": 2.5217514472225646e-06, + "loss": 0.5376, + "step": 12993 + }, + { + "epoch": 4.8, + "learning_rate": 2.521714496859219e-06, + "loss": 0.4738, + "step": 12994 + }, + { + "epoch": 4.8, + "learning_rate": 2.521677546495874e-06, + "loss": 0.4781, + "step": 12995 + }, + { + "epoch": 4.8, + "learning_rate": 2.521640596132529e-06, + "loss": 0.6949, + "step": 12996 + }, + { + "epoch": 4.8, + "learning_rate": 2.5216036457691835e-06, + "loss": 0.4537, + "step": 12997 + }, + { + "epoch": 4.8, + "learning_rate": 2.521566695405838e-06, + "loss": 0.4419, + "step": 12998 + }, + { + "epoch": 4.8, + "learning_rate": 2.521529745042493e-06, + "loss": 0.5494, + "step": 12999 + }, + { + "epoch": 4.8, + "learning_rate": 2.521492794679148e-06, + "loss": 0.4355, + "step": 13000 + }, + { + "epoch": 4.8, + "learning_rate": 2.5214558443158025e-06, + "loss": 0.5233, + "step": 13001 + }, + { + "epoch": 4.8, + "learning_rate": 2.5214188939524576e-06, + "loss": 0.576, + "step": 13002 + }, + { + "epoch": 4.8, + "learning_rate": 2.521381943589112e-06, + "loss": 0.5395, + "step": 13003 + }, + { + "epoch": 4.8, + "learning_rate": 2.521344993225767e-06, + "loss": 0.4484, + "step": 13004 + }, + { + "epoch": 4.8, + "learning_rate": 2.5213080428624215e-06, + "loss": 0.4716, + "step": 13005 + }, + { + "epoch": 4.8, + "learning_rate": 2.521271092499076e-06, + "loss": 0.4738, + "step": 13006 + }, + { + "epoch": 4.8, + "learning_rate": 2.521234142135731e-06, + "loss": 0.4667, + "step": 13007 + }, + { + "epoch": 4.8, + "learning_rate": 2.521197191772386e-06, + "loss": 0.4874, + "step": 13008 + }, + { + "epoch": 4.8, + "learning_rate": 2.521160241409041e-06, + "loss": 0.5045, + "step": 13009 + }, + { + "epoch": 4.8, + "learning_rate": 2.521123291045695e-06, + "loss": 0.4727, + "step": 13010 + }, + { + "epoch": 4.8, + "learning_rate": 2.52108634068235e-06, + "loss": 0.4925, + "step": 13011 + }, + { + "epoch": 4.81, + "learning_rate": 2.5210493903190048e-06, + "loss": 0.4075, + "step": 13012 + }, + { + "epoch": 4.81, + "learning_rate": 2.52101243995566e-06, + "loss": 0.6457, + "step": 13013 + }, + { + "epoch": 4.81, + "learning_rate": 2.5209754895923145e-06, + "loss": 0.5207, + "step": 13014 + }, + { + "epoch": 4.81, + "learning_rate": 2.520938539228969e-06, + "loss": 0.5175, + "step": 13015 + }, + { + "epoch": 4.81, + "learning_rate": 2.5209015888656238e-06, + "loss": 0.5127, + "step": 13016 + }, + { + "epoch": 4.81, + "learning_rate": 2.5208646385022784e-06, + "loss": 0.5201, + "step": 13017 + }, + { + "epoch": 4.81, + "learning_rate": 2.5208276881389334e-06, + "loss": 0.6509, + "step": 13018 + }, + { + "epoch": 4.81, + "learning_rate": 2.520790737775588e-06, + "loss": 0.5269, + "step": 13019 + }, + { + "epoch": 4.81, + "learning_rate": 2.520753787412243e-06, + "loss": 0.4213, + "step": 13020 + }, + { + "epoch": 4.81, + "learning_rate": 2.5207168370488978e-06, + "loss": 0.485, + "step": 13021 + }, + { + "epoch": 4.81, + "learning_rate": 2.5206798866855524e-06, + "loss": 0.538, + "step": 13022 + }, + { + "epoch": 4.81, + "learning_rate": 2.520642936322207e-06, + "loss": 0.4861, + "step": 13023 + }, + { + "epoch": 4.81, + "learning_rate": 2.520605985958862e-06, + "loss": 0.5701, + "step": 13024 + }, + { + "epoch": 4.81, + "learning_rate": 2.5205690355955167e-06, + "loss": 0.5273, + "step": 13025 + }, + { + "epoch": 4.81, + "learning_rate": 2.5205320852321714e-06, + "loss": 0.6766, + "step": 13026 + }, + { + "epoch": 4.81, + "learning_rate": 2.5204951348688264e-06, + "loss": 0.5148, + "step": 13027 + }, + { + "epoch": 4.81, + "learning_rate": 2.520458184505481e-06, + "loss": 0.5831, + "step": 13028 + }, + { + "epoch": 4.81, + "learning_rate": 2.5204212341421357e-06, + "loss": 0.4852, + "step": 13029 + }, + { + "epoch": 4.81, + "learning_rate": 2.5203842837787904e-06, + "loss": 0.4838, + "step": 13030 + }, + { + "epoch": 4.81, + "learning_rate": 2.5203473334154454e-06, + "loss": 0.5652, + "step": 13031 + }, + { + "epoch": 4.81, + "learning_rate": 2.5203103830521e-06, + "loss": 0.4659, + "step": 13032 + }, + { + "epoch": 4.81, + "learning_rate": 2.520273432688755e-06, + "loss": 0.4248, + "step": 13033 + }, + { + "epoch": 4.81, + "learning_rate": 2.5202364823254097e-06, + "loss": 0.5208, + "step": 13034 + }, + { + "epoch": 4.81, + "learning_rate": 2.5201995319620644e-06, + "loss": 0.4191, + "step": 13035 + }, + { + "epoch": 4.81, + "learning_rate": 2.520162581598719e-06, + "loss": 0.461, + "step": 13036 + }, + { + "epoch": 4.81, + "learning_rate": 2.5201256312353737e-06, + "loss": 0.5143, + "step": 13037 + }, + { + "epoch": 4.81, + "learning_rate": 2.5200886808720287e-06, + "loss": 0.4214, + "step": 13038 + }, + { + "epoch": 4.81, + "learning_rate": 2.5200517305086834e-06, + "loss": 0.4633, + "step": 13039 + }, + { + "epoch": 4.82, + "learning_rate": 2.5200147801453384e-06, + "loss": 0.5275, + "step": 13040 + }, + { + "epoch": 4.82, + "learning_rate": 2.519977829781993e-06, + "loss": 0.4617, + "step": 13041 + }, + { + "epoch": 4.82, + "learning_rate": 2.5199408794186477e-06, + "loss": 0.4994, + "step": 13042 + }, + { + "epoch": 4.82, + "learning_rate": 2.5199039290553023e-06, + "loss": 0.5178, + "step": 13043 + }, + { + "epoch": 4.82, + "learning_rate": 2.519866978691957e-06, + "loss": 0.4981, + "step": 13044 + }, + { + "epoch": 4.82, + "learning_rate": 2.519830028328612e-06, + "loss": 0.4373, + "step": 13045 + }, + { + "epoch": 4.82, + "learning_rate": 2.5197930779652667e-06, + "loss": 0.5316, + "step": 13046 + }, + { + "epoch": 4.82, + "learning_rate": 2.5197561276019217e-06, + "loss": 0.4898, + "step": 13047 + }, + { + "epoch": 4.82, + "learning_rate": 2.5197191772385764e-06, + "loss": 0.585, + "step": 13048 + }, + { + "epoch": 4.82, + "learning_rate": 2.519682226875231e-06, + "loss": 0.4866, + "step": 13049 + }, + { + "epoch": 4.82, + "learning_rate": 2.5196452765118856e-06, + "loss": 0.5243, + "step": 13050 + }, + { + "epoch": 4.82, + "learning_rate": 2.5196083261485407e-06, + "loss": 0.4898, + "step": 13051 + }, + { + "epoch": 4.82, + "learning_rate": 2.5195713757851953e-06, + "loss": 0.5417, + "step": 13052 + }, + { + "epoch": 4.82, + "learning_rate": 2.5195344254218504e-06, + "loss": 0.5213, + "step": 13053 + }, + { + "epoch": 4.82, + "learning_rate": 2.519497475058505e-06, + "loss": 0.4509, + "step": 13054 + }, + { + "epoch": 4.82, + "learning_rate": 2.5194605246951592e-06, + "loss": 0.5002, + "step": 13055 + }, + { + "epoch": 4.82, + "learning_rate": 2.5194235743318143e-06, + "loss": 0.5372, + "step": 13056 + }, + { + "epoch": 4.82, + "learning_rate": 2.519386623968469e-06, + "loss": 0.588, + "step": 13057 + }, + { + "epoch": 4.82, + "learning_rate": 2.519349673605124e-06, + "loss": 0.4867, + "step": 13058 + }, + { + "epoch": 4.82, + "learning_rate": 2.5193127232417786e-06, + "loss": 0.4718, + "step": 13059 + }, + { + "epoch": 4.82, + "learning_rate": 2.5192757728784333e-06, + "loss": 0.4684, + "step": 13060 + }, + { + "epoch": 4.82, + "learning_rate": 2.519238822515088e-06, + "loss": 0.5497, + "step": 13061 + }, + { + "epoch": 4.82, + "learning_rate": 2.519201872151743e-06, + "loss": 0.6071, + "step": 13062 + }, + { + "epoch": 4.82, + "learning_rate": 2.5191649217883976e-06, + "loss": 0.4607, + "step": 13063 + }, + { + "epoch": 4.82, + "learning_rate": 2.5191279714250522e-06, + "loss": 0.5561, + "step": 13064 + }, + { + "epoch": 4.82, + "learning_rate": 2.5190910210617073e-06, + "loss": 0.5313, + "step": 13065 + }, + { + "epoch": 4.82, + "learning_rate": 2.519054070698362e-06, + "loss": 0.5241, + "step": 13066 + }, + { + "epoch": 4.83, + "learning_rate": 2.5190171203350166e-06, + "loss": 0.4869, + "step": 13067 + }, + { + "epoch": 4.83, + "learning_rate": 2.518980169971671e-06, + "loss": 0.4887, + "step": 13068 + }, + { + "epoch": 4.83, + "learning_rate": 2.5189432196083263e-06, + "loss": 0.479, + "step": 13069 + }, + { + "epoch": 4.83, + "learning_rate": 2.518906269244981e-06, + "loss": 0.5721, + "step": 13070 + }, + { + "epoch": 4.83, + "learning_rate": 2.518869318881636e-06, + "loss": 0.5028, + "step": 13071 + }, + { + "epoch": 4.83, + "learning_rate": 2.5188323685182906e-06, + "loss": 0.4579, + "step": 13072 + }, + { + "epoch": 4.83, + "learning_rate": 2.5187954181549452e-06, + "loss": 0.4474, + "step": 13073 + }, + { + "epoch": 4.83, + "learning_rate": 2.5187584677916e-06, + "loss": 0.5255, + "step": 13074 + }, + { + "epoch": 4.83, + "learning_rate": 2.5187215174282545e-06, + "loss": 0.5199, + "step": 13075 + }, + { + "epoch": 4.83, + "learning_rate": 2.5186845670649096e-06, + "loss": 0.5434, + "step": 13076 + }, + { + "epoch": 4.83, + "learning_rate": 2.518647616701564e-06, + "loss": 0.4682, + "step": 13077 + }, + { + "epoch": 4.83, + "learning_rate": 2.5186106663382193e-06, + "loss": 0.4628, + "step": 13078 + }, + { + "epoch": 4.83, + "learning_rate": 2.518573715974874e-06, + "loss": 0.48, + "step": 13079 + }, + { + "epoch": 4.83, + "learning_rate": 2.5185367656115285e-06, + "loss": 0.4438, + "step": 13080 + }, + { + "epoch": 4.83, + "learning_rate": 2.518499815248183e-06, + "loss": 0.5736, + "step": 13081 + }, + { + "epoch": 4.83, + "learning_rate": 2.5184628648848382e-06, + "loss": 0.6706, + "step": 13082 + }, + { + "epoch": 4.83, + "learning_rate": 2.518425914521493e-06, + "loss": 0.6099, + "step": 13083 + }, + { + "epoch": 4.83, + "learning_rate": 2.5183889641581475e-06, + "loss": 0.5403, + "step": 13084 + }, + { + "epoch": 4.83, + "learning_rate": 2.5183520137948026e-06, + "loss": 0.485, + "step": 13085 + }, + { + "epoch": 4.83, + "learning_rate": 2.518315063431457e-06, + "loss": 0.5039, + "step": 13086 + }, + { + "epoch": 4.83, + "learning_rate": 2.518278113068112e-06, + "loss": 0.5608, + "step": 13087 + }, + { + "epoch": 4.83, + "learning_rate": 2.5182411627047665e-06, + "loss": 0.6826, + "step": 13088 + }, + { + "epoch": 4.83, + "learning_rate": 2.5182042123414215e-06, + "loss": 0.5792, + "step": 13089 + }, + { + "epoch": 4.83, + "learning_rate": 2.518167261978076e-06, + "loss": 0.4825, + "step": 13090 + }, + { + "epoch": 4.83, + "learning_rate": 2.5181303116147312e-06, + "loss": 0.4912, + "step": 13091 + }, + { + "epoch": 4.83, + "learning_rate": 2.518093361251386e-06, + "loss": 0.4387, + "step": 13092 + }, + { + "epoch": 4.83, + "learning_rate": 2.5180564108880405e-06, + "loss": 0.5146, + "step": 13093 + }, + { + "epoch": 4.84, + "learning_rate": 2.518019460524695e-06, + "loss": 0.4624, + "step": 13094 + }, + { + "epoch": 4.84, + "learning_rate": 2.5179825101613498e-06, + "loss": 0.5534, + "step": 13095 + }, + { + "epoch": 4.84, + "learning_rate": 2.517945559798005e-06, + "loss": 0.5485, + "step": 13096 + }, + { + "epoch": 4.84, + "learning_rate": 2.5179086094346595e-06, + "loss": 0.6863, + "step": 13097 + }, + { + "epoch": 4.84, + "learning_rate": 2.5178716590713145e-06, + "loss": 0.523, + "step": 13098 + }, + { + "epoch": 4.84, + "learning_rate": 2.5178347087079687e-06, + "loss": 0.4419, + "step": 13099 + }, + { + "epoch": 4.84, + "learning_rate": 2.517797758344624e-06, + "loss": 0.4725, + "step": 13100 + }, + { + "epoch": 4.84, + "learning_rate": 2.5177608079812784e-06, + "loss": 0.4756, + "step": 13101 + }, + { + "epoch": 4.84, + "learning_rate": 2.5177238576179335e-06, + "loss": 0.6292, + "step": 13102 + }, + { + "epoch": 4.84, + "learning_rate": 2.517686907254588e-06, + "loss": 0.5478, + "step": 13103 + }, + { + "epoch": 4.84, + "learning_rate": 2.5176499568912428e-06, + "loss": 0.5037, + "step": 13104 + }, + { + "epoch": 4.84, + "learning_rate": 2.5176130065278974e-06, + "loss": 0.5455, + "step": 13105 + }, + { + "epoch": 4.84, + "learning_rate": 2.517576056164552e-06, + "loss": 0.5925, + "step": 13106 + }, + { + "epoch": 4.84, + "learning_rate": 2.517539105801207e-06, + "loss": 0.5545, + "step": 13107 + }, + { + "epoch": 4.84, + "learning_rate": 2.5175021554378617e-06, + "loss": 0.4982, + "step": 13108 + }, + { + "epoch": 4.84, + "learning_rate": 2.517465205074517e-06, + "loss": 0.4544, + "step": 13109 + }, + { + "epoch": 4.84, + "learning_rate": 2.5174282547111714e-06, + "loss": 0.4656, + "step": 13110 + }, + { + "epoch": 4.84, + "learning_rate": 2.517391304347826e-06, + "loss": 0.5683, + "step": 13111 + }, + { + "epoch": 4.84, + "learning_rate": 2.5173543539844807e-06, + "loss": 0.5131, + "step": 13112 + }, + { + "epoch": 4.84, + "learning_rate": 2.5173174036211358e-06, + "loss": 0.6273, + "step": 13113 + }, + { + "epoch": 4.84, + "learning_rate": 2.5172804532577904e-06, + "loss": 0.5197, + "step": 13114 + }, + { + "epoch": 4.84, + "learning_rate": 2.517243502894445e-06, + "loss": 0.512, + "step": 13115 + }, + { + "epoch": 4.84, + "learning_rate": 2.5172065525311e-06, + "loss": 0.5765, + "step": 13116 + }, + { + "epoch": 4.84, + "learning_rate": 2.5171696021677547e-06, + "loss": 0.5197, + "step": 13117 + }, + { + "epoch": 4.84, + "learning_rate": 2.5171326518044094e-06, + "loss": 0.4331, + "step": 13118 + }, + { + "epoch": 4.84, + "learning_rate": 2.517095701441064e-06, + "loss": 0.4741, + "step": 13119 + }, + { + "epoch": 4.84, + "learning_rate": 2.517058751077719e-06, + "loss": 0.4782, + "step": 13120 + }, + { + "epoch": 4.85, + "learning_rate": 2.5170218007143737e-06, + "loss": 0.5365, + "step": 13121 + }, + { + "epoch": 4.85, + "learning_rate": 2.5169848503510288e-06, + "loss": 0.5512, + "step": 13122 + }, + { + "epoch": 4.85, + "learning_rate": 2.5169478999876834e-06, + "loss": 0.4578, + "step": 13123 + }, + { + "epoch": 4.85, + "learning_rate": 2.516910949624338e-06, + "loss": 0.5548, + "step": 13124 + }, + { + "epoch": 4.85, + "learning_rate": 2.5168739992609927e-06, + "loss": 0.5207, + "step": 13125 + }, + { + "epoch": 4.85, + "learning_rate": 2.5168370488976473e-06, + "loss": 0.5881, + "step": 13126 + }, + { + "epoch": 4.85, + "learning_rate": 2.5168000985343024e-06, + "loss": 0.537, + "step": 13127 + }, + { + "epoch": 4.85, + "learning_rate": 2.516763148170957e-06, + "loss": 0.5195, + "step": 13128 + }, + { + "epoch": 4.85, + "learning_rate": 2.516726197807612e-06, + "loss": 0.5177, + "step": 13129 + }, + { + "epoch": 4.85, + "learning_rate": 2.5166892474442667e-06, + "loss": 0.4342, + "step": 13130 + }, + { + "epoch": 4.85, + "learning_rate": 2.5166522970809213e-06, + "loss": 0.5241, + "step": 13131 + }, + { + "epoch": 4.85, + "learning_rate": 2.516615346717576e-06, + "loss": 0.5521, + "step": 13132 + }, + { + "epoch": 4.85, + "learning_rate": 2.516578396354231e-06, + "loss": 0.6574, + "step": 13133 + }, + { + "epoch": 4.85, + "learning_rate": 2.5165414459908857e-06, + "loss": 0.4945, + "step": 13134 + }, + { + "epoch": 4.85, + "learning_rate": 2.5165044956275403e-06, + "loss": 0.5848, + "step": 13135 + }, + { + "epoch": 4.85, + "learning_rate": 2.5164675452641954e-06, + "loss": 0.4626, + "step": 13136 + }, + { + "epoch": 4.85, + "learning_rate": 2.51643059490085e-06, + "loss": 0.4904, + "step": 13137 + }, + { + "epoch": 4.85, + "learning_rate": 2.5163936445375046e-06, + "loss": 0.5111, + "step": 13138 + }, + { + "epoch": 4.85, + "learning_rate": 2.5163566941741593e-06, + "loss": 0.5709, + "step": 13139 + }, + { + "epoch": 4.85, + "learning_rate": 2.5163197438108143e-06, + "loss": 0.5682, + "step": 13140 + }, + { + "epoch": 4.85, + "learning_rate": 2.516282793447469e-06, + "loss": 0.4277, + "step": 13141 + }, + { + "epoch": 4.85, + "learning_rate": 2.516245843084124e-06, + "loss": 0.5145, + "step": 13142 + }, + { + "epoch": 4.85, + "learning_rate": 2.5162088927207787e-06, + "loss": 0.4085, + "step": 13143 + }, + { + "epoch": 4.85, + "learning_rate": 2.516171942357433e-06, + "loss": 0.5497, + "step": 13144 + }, + { + "epoch": 4.85, + "learning_rate": 2.516134991994088e-06, + "loss": 0.5501, + "step": 13145 + }, + { + "epoch": 4.85, + "learning_rate": 2.5160980416307426e-06, + "loss": 0.5708, + "step": 13146 + }, + { + "epoch": 4.85, + "learning_rate": 2.5160610912673976e-06, + "loss": 0.55, + "step": 13147 + }, + { + "epoch": 4.86, + "learning_rate": 2.5160241409040523e-06, + "loss": 0.5197, + "step": 13148 + }, + { + "epoch": 4.86, + "learning_rate": 2.5159871905407073e-06, + "loss": 0.531, + "step": 13149 + }, + { + "epoch": 4.86, + "learning_rate": 2.5159502401773615e-06, + "loss": 0.5574, + "step": 13150 + }, + { + "epoch": 4.86, + "learning_rate": 2.5159132898140166e-06, + "loss": 0.6587, + "step": 13151 + }, + { + "epoch": 4.86, + "learning_rate": 2.5158763394506712e-06, + "loss": 0.5446, + "step": 13152 + }, + { + "epoch": 4.86, + "learning_rate": 2.5158393890873263e-06, + "loss": 0.4334, + "step": 13153 + }, + { + "epoch": 4.86, + "learning_rate": 2.515802438723981e-06, + "loss": 0.5635, + "step": 13154 + }, + { + "epoch": 4.86, + "learning_rate": 2.5157654883606356e-06, + "loss": 0.4564, + "step": 13155 + }, + { + "epoch": 4.86, + "learning_rate": 2.5157285379972902e-06, + "loss": 0.6296, + "step": 13156 + }, + { + "epoch": 4.86, + "learning_rate": 2.515691587633945e-06, + "loss": 0.5684, + "step": 13157 + }, + { + "epoch": 4.86, + "learning_rate": 2.5156546372706e-06, + "loss": 0.4056, + "step": 13158 + }, + { + "epoch": 4.86, + "learning_rate": 2.5156176869072545e-06, + "loss": 0.4581, + "step": 13159 + }, + { + "epoch": 4.86, + "learning_rate": 2.5155807365439096e-06, + "loss": 0.4454, + "step": 13160 + }, + { + "epoch": 4.86, + "learning_rate": 2.5155437861805642e-06, + "loss": 0.5377, + "step": 13161 + }, + { + "epoch": 4.86, + "learning_rate": 2.515506835817219e-06, + "loss": 0.5768, + "step": 13162 + }, + { + "epoch": 4.86, + "learning_rate": 2.5154698854538735e-06, + "loss": 0.3893, + "step": 13163 + }, + { + "epoch": 4.86, + "learning_rate": 2.515432935090528e-06, + "loss": 0.5213, + "step": 13164 + }, + { + "epoch": 4.86, + "learning_rate": 2.5153959847271832e-06, + "loss": 0.3646, + "step": 13165 + }, + { + "epoch": 4.86, + "learning_rate": 2.515359034363838e-06, + "loss": 0.5075, + "step": 13166 + }, + { + "epoch": 4.86, + "learning_rate": 2.515322084000493e-06, + "loss": 0.5412, + "step": 13167 + }, + { + "epoch": 4.86, + "learning_rate": 2.5152851336371475e-06, + "loss": 0.4188, + "step": 13168 + }, + { + "epoch": 4.86, + "learning_rate": 2.515248183273802e-06, + "loss": 0.4335, + "step": 13169 + }, + { + "epoch": 4.86, + "learning_rate": 2.515211232910457e-06, + "loss": 0.5192, + "step": 13170 + }, + { + "epoch": 4.86, + "learning_rate": 2.515174282547112e-06, + "loss": 0.4347, + "step": 13171 + }, + { + "epoch": 4.86, + "learning_rate": 2.5151373321837665e-06, + "loss": 0.4853, + "step": 13172 + }, + { + "epoch": 4.86, + "learning_rate": 2.5151003818204216e-06, + "loss": 0.5331, + "step": 13173 + }, + { + "epoch": 4.86, + "learning_rate": 2.5150634314570762e-06, + "loss": 0.4805, + "step": 13174 + }, + { + "epoch": 4.87, + "learning_rate": 2.515026481093731e-06, + "loss": 0.5049, + "step": 13175 + }, + { + "epoch": 4.87, + "learning_rate": 2.5149895307303855e-06, + "loss": 0.4975, + "step": 13176 + }, + { + "epoch": 4.87, + "learning_rate": 2.51495258036704e-06, + "loss": 0.439, + "step": 13177 + }, + { + "epoch": 4.87, + "learning_rate": 2.514915630003695e-06, + "loss": 0.5547, + "step": 13178 + }, + { + "epoch": 4.87, + "learning_rate": 2.51487867964035e-06, + "loss": 0.4545, + "step": 13179 + }, + { + "epoch": 4.87, + "learning_rate": 2.514841729277005e-06, + "loss": 0.5707, + "step": 13180 + }, + { + "epoch": 4.87, + "learning_rate": 2.5148047789136595e-06, + "loss": 0.441, + "step": 13181 + }, + { + "epoch": 4.87, + "learning_rate": 2.514767828550314e-06, + "loss": 0.4647, + "step": 13182 + }, + { + "epoch": 4.87, + "learning_rate": 2.5147308781869688e-06, + "loss": 0.5827, + "step": 13183 + }, + { + "epoch": 4.87, + "learning_rate": 2.5146939278236234e-06, + "loss": 0.5075, + "step": 13184 + }, + { + "epoch": 4.87, + "learning_rate": 2.5146569774602785e-06, + "loss": 0.4959, + "step": 13185 + }, + { + "epoch": 4.87, + "learning_rate": 2.514620027096933e-06, + "loss": 0.5208, + "step": 13186 + }, + { + "epoch": 4.87, + "learning_rate": 2.514583076733588e-06, + "loss": 0.4626, + "step": 13187 + }, + { + "epoch": 4.87, + "learning_rate": 2.514546126370243e-06, + "loss": 0.6231, + "step": 13188 + }, + { + "epoch": 4.87, + "learning_rate": 2.5145091760068975e-06, + "loss": 0.3987, + "step": 13189 + }, + { + "epoch": 4.87, + "learning_rate": 2.514472225643552e-06, + "loss": 0.515, + "step": 13190 + }, + { + "epoch": 4.87, + "learning_rate": 2.514435275280207e-06, + "loss": 0.6003, + "step": 13191 + }, + { + "epoch": 4.87, + "learning_rate": 2.5143983249168618e-06, + "loss": 0.5661, + "step": 13192 + }, + { + "epoch": 4.87, + "learning_rate": 2.514361374553517e-06, + "loss": 0.488, + "step": 13193 + }, + { + "epoch": 4.87, + "learning_rate": 2.5143244241901715e-06, + "loss": 0.5475, + "step": 13194 + }, + { + "epoch": 4.87, + "learning_rate": 2.5142874738268257e-06, + "loss": 0.4522, + "step": 13195 + }, + { + "epoch": 4.87, + "learning_rate": 2.5142505234634808e-06, + "loss": 0.5282, + "step": 13196 + }, + { + "epoch": 4.87, + "learning_rate": 2.5142135731001354e-06, + "loss": 0.4755, + "step": 13197 + }, + { + "epoch": 4.87, + "learning_rate": 2.5141766227367905e-06, + "loss": 0.5744, + "step": 13198 + }, + { + "epoch": 4.87, + "learning_rate": 2.514139672373445e-06, + "loss": 0.5734, + "step": 13199 + }, + { + "epoch": 4.87, + "learning_rate": 2.5141027220101e-06, + "loss": 0.4878, + "step": 13200 + }, + { + "epoch": 4.87, + "learning_rate": 2.5140657716467544e-06, + "loss": 0.5474, + "step": 13201 + }, + { + "epoch": 4.88, + "learning_rate": 2.5140288212834094e-06, + "loss": 0.5044, + "step": 13202 + }, + { + "epoch": 4.88, + "learning_rate": 2.513991870920064e-06, + "loss": 0.5146, + "step": 13203 + }, + { + "epoch": 4.88, + "learning_rate": 2.5139549205567187e-06, + "loss": 0.5428, + "step": 13204 + }, + { + "epoch": 4.88, + "learning_rate": 2.5139179701933738e-06, + "loss": 0.4917, + "step": 13205 + }, + { + "epoch": 4.88, + "learning_rate": 2.5138810198300284e-06, + "loss": 0.4607, + "step": 13206 + }, + { + "epoch": 4.88, + "learning_rate": 2.513844069466683e-06, + "loss": 0.4741, + "step": 13207 + }, + { + "epoch": 4.88, + "learning_rate": 2.5138071191033377e-06, + "loss": 0.3727, + "step": 13208 + }, + { + "epoch": 4.88, + "learning_rate": 2.5137701687399927e-06, + "loss": 0.4848, + "step": 13209 + }, + { + "epoch": 4.88, + "learning_rate": 2.5137332183766474e-06, + "loss": 0.5203, + "step": 13210 + }, + { + "epoch": 4.88, + "learning_rate": 2.5136962680133024e-06, + "loss": 0.5383, + "step": 13211 + }, + { + "epoch": 4.88, + "learning_rate": 2.513659317649957e-06, + "loss": 0.5321, + "step": 13212 + }, + { + "epoch": 4.88, + "learning_rate": 2.5136223672866117e-06, + "loss": 0.4369, + "step": 13213 + }, + { + "epoch": 4.88, + "learning_rate": 2.5135854169232663e-06, + "loss": 0.5652, + "step": 13214 + }, + { + "epoch": 4.88, + "learning_rate": 2.513548466559921e-06, + "loss": 0.5168, + "step": 13215 + }, + { + "epoch": 4.88, + "learning_rate": 2.513511516196576e-06, + "loss": 0.4911, + "step": 13216 + }, + { + "epoch": 4.88, + "learning_rate": 2.5134745658332307e-06, + "loss": 0.5753, + "step": 13217 + }, + { + "epoch": 4.88, + "learning_rate": 2.5134376154698857e-06, + "loss": 0.5877, + "step": 13218 + }, + { + "epoch": 4.88, + "learning_rate": 2.5134006651065404e-06, + "loss": 0.5236, + "step": 13219 + }, + { + "epoch": 4.88, + "learning_rate": 2.513363714743195e-06, + "loss": 0.5346, + "step": 13220 + }, + { + "epoch": 4.88, + "learning_rate": 2.5133267643798496e-06, + "loss": 0.5039, + "step": 13221 + }, + { + "epoch": 4.88, + "learning_rate": 2.5132898140165047e-06, + "loss": 0.558, + "step": 13222 + }, + { + "epoch": 4.88, + "learning_rate": 2.5132528636531593e-06, + "loss": 0.5001, + "step": 13223 + }, + { + "epoch": 4.88, + "learning_rate": 2.513215913289814e-06, + "loss": 0.6099, + "step": 13224 + }, + { + "epoch": 4.88, + "learning_rate": 2.513178962926469e-06, + "loss": 0.5915, + "step": 13225 + }, + { + "epoch": 4.88, + "learning_rate": 2.5131420125631237e-06, + "loss": 0.4881, + "step": 13226 + }, + { + "epoch": 4.88, + "learning_rate": 2.5131050621997783e-06, + "loss": 0.526, + "step": 13227 + }, + { + "epoch": 4.88, + "learning_rate": 2.513068111836433e-06, + "loss": 0.5161, + "step": 13228 + }, + { + "epoch": 4.89, + "learning_rate": 2.513031161473088e-06, + "loss": 0.5831, + "step": 13229 + }, + { + "epoch": 4.89, + "learning_rate": 2.5129942111097426e-06, + "loss": 0.4494, + "step": 13230 + }, + { + "epoch": 4.89, + "learning_rate": 2.5129572607463977e-06, + "loss": 0.5871, + "step": 13231 + }, + { + "epoch": 4.89, + "learning_rate": 2.5129203103830523e-06, + "loss": 0.5026, + "step": 13232 + }, + { + "epoch": 4.89, + "learning_rate": 2.512883360019707e-06, + "loss": 0.4089, + "step": 13233 + }, + { + "epoch": 4.89, + "learning_rate": 2.5128464096563616e-06, + "loss": 0.5067, + "step": 13234 + }, + { + "epoch": 4.89, + "learning_rate": 2.5128094592930162e-06, + "loss": 0.4793, + "step": 13235 + }, + { + "epoch": 4.89, + "learning_rate": 2.5127725089296713e-06, + "loss": 0.6129, + "step": 13236 + }, + { + "epoch": 4.89, + "learning_rate": 2.512735558566326e-06, + "loss": 0.5227, + "step": 13237 + }, + { + "epoch": 4.89, + "learning_rate": 2.512698608202981e-06, + "loss": 0.4521, + "step": 13238 + }, + { + "epoch": 4.89, + "learning_rate": 2.5126616578396356e-06, + "loss": 0.3979, + "step": 13239 + }, + { + "epoch": 4.89, + "learning_rate": 2.5126247074762903e-06, + "loss": 0.5332, + "step": 13240 + }, + { + "epoch": 4.89, + "learning_rate": 2.512587757112945e-06, + "loss": 0.5393, + "step": 13241 + }, + { + "epoch": 4.89, + "learning_rate": 2.5125508067496e-06, + "loss": 0.5223, + "step": 13242 + }, + { + "epoch": 4.89, + "learning_rate": 2.5125138563862546e-06, + "loss": 0.5261, + "step": 13243 + }, + { + "epoch": 4.89, + "learning_rate": 2.5124769060229092e-06, + "loss": 0.5255, + "step": 13244 + }, + { + "epoch": 4.89, + "learning_rate": 2.512439955659564e-06, + "loss": 0.5049, + "step": 13245 + }, + { + "epoch": 4.89, + "learning_rate": 2.5124030052962185e-06, + "loss": 0.4686, + "step": 13246 + }, + { + "epoch": 4.89, + "learning_rate": 2.5123660549328736e-06, + "loss": 0.5063, + "step": 13247 + }, + { + "epoch": 4.89, + "learning_rate": 2.512329104569528e-06, + "loss": 0.4963, + "step": 13248 + }, + { + "epoch": 4.89, + "learning_rate": 2.5122921542061833e-06, + "loss": 0.4947, + "step": 13249 + }, + { + "epoch": 4.89, + "learning_rate": 2.512255203842838e-06, + "loss": 0.4266, + "step": 13250 + }, + { + "epoch": 4.89, + "learning_rate": 2.5122182534794925e-06, + "loss": 0.5627, + "step": 13251 + }, + { + "epoch": 4.89, + "learning_rate": 2.512181303116147e-06, + "loss": 0.4182, + "step": 13252 + }, + { + "epoch": 4.89, + "learning_rate": 2.5121443527528022e-06, + "loss": 0.4639, + "step": 13253 + }, + { + "epoch": 4.89, + "learning_rate": 2.512107402389457e-06, + "loss": 0.2968, + "step": 13254 + }, + { + "epoch": 4.89, + "learning_rate": 2.5120704520261115e-06, + "loss": 0.5969, + "step": 13255 + }, + { + "epoch": 4.9, + "learning_rate": 2.5120335016627666e-06, + "loss": 0.5907, + "step": 13256 + }, + { + "epoch": 4.9, + "learning_rate": 2.511996551299421e-06, + "loss": 0.4124, + "step": 13257 + }, + { + "epoch": 4.9, + "learning_rate": 2.511959600936076e-06, + "loss": 0.6192, + "step": 13258 + }, + { + "epoch": 4.9, + "learning_rate": 2.5119226505727305e-06, + "loss": 0.6019, + "step": 13259 + }, + { + "epoch": 4.9, + "learning_rate": 2.5118857002093855e-06, + "loss": 0.4342, + "step": 13260 + }, + { + "epoch": 4.9, + "learning_rate": 2.51184874984604e-06, + "loss": 0.5606, + "step": 13261 + }, + { + "epoch": 4.9, + "learning_rate": 2.5118117994826952e-06, + "loss": 0.4532, + "step": 13262 + }, + { + "epoch": 4.9, + "learning_rate": 2.51177484911935e-06, + "loss": 0.4213, + "step": 13263 + }, + { + "epoch": 4.9, + "learning_rate": 2.5117378987560045e-06, + "loss": 0.4861, + "step": 13264 + }, + { + "epoch": 4.9, + "learning_rate": 2.511700948392659e-06, + "loss": 0.4448, + "step": 13265 + }, + { + "epoch": 4.9, + "learning_rate": 2.5116639980293138e-06, + "loss": 0.4945, + "step": 13266 + }, + { + "epoch": 4.9, + "learning_rate": 2.511627047665969e-06, + "loss": 0.4606, + "step": 13267 + }, + { + "epoch": 4.9, + "learning_rate": 2.5115900973026235e-06, + "loss": 0.4768, + "step": 13268 + }, + { + "epoch": 4.9, + "learning_rate": 2.5115531469392785e-06, + "loss": 0.6601, + "step": 13269 + }, + { + "epoch": 4.9, + "learning_rate": 2.511516196575933e-06, + "loss": 0.5612, + "step": 13270 + }, + { + "epoch": 4.9, + "learning_rate": 2.511479246212588e-06, + "loss": 0.5158, + "step": 13271 + }, + { + "epoch": 4.9, + "learning_rate": 2.5114422958492424e-06, + "loss": 0.5582, + "step": 13272 + }, + { + "epoch": 4.9, + "learning_rate": 2.511405345485897e-06, + "loss": 0.5734, + "step": 13273 + }, + { + "epoch": 4.9, + "learning_rate": 2.511368395122552e-06, + "loss": 0.4923, + "step": 13274 + }, + { + "epoch": 4.9, + "learning_rate": 2.5113314447592068e-06, + "loss": 0.404, + "step": 13275 + }, + { + "epoch": 4.9, + "learning_rate": 2.511294494395862e-06, + "loss": 0.6036, + "step": 13276 + }, + { + "epoch": 4.9, + "learning_rate": 2.5112575440325165e-06, + "loss": 0.551, + "step": 13277 + }, + { + "epoch": 4.9, + "learning_rate": 2.511220593669171e-06, + "loss": 0.4598, + "step": 13278 + }, + { + "epoch": 4.9, + "learning_rate": 2.5111836433058257e-06, + "loss": 0.7226, + "step": 13279 + }, + { + "epoch": 4.9, + "learning_rate": 2.511146692942481e-06, + "loss": 0.6347, + "step": 13280 + }, + { + "epoch": 4.9, + "learning_rate": 2.5111097425791354e-06, + "loss": 0.6151, + "step": 13281 + }, + { + "epoch": 4.9, + "learning_rate": 2.5110727922157905e-06, + "loss": 0.5114, + "step": 13282 + }, + { + "epoch": 4.91, + "learning_rate": 2.511035841852445e-06, + "loss": 0.5439, + "step": 13283 + }, + { + "epoch": 4.91, + "learning_rate": 2.5109988914890993e-06, + "loss": 0.4764, + "step": 13284 + }, + { + "epoch": 4.91, + "learning_rate": 2.5109619411257544e-06, + "loss": 0.4269, + "step": 13285 + }, + { + "epoch": 4.91, + "learning_rate": 2.510924990762409e-06, + "loss": 0.5004, + "step": 13286 + }, + { + "epoch": 4.91, + "learning_rate": 2.510888040399064e-06, + "loss": 0.5498, + "step": 13287 + }, + { + "epoch": 4.91, + "learning_rate": 2.5108510900357187e-06, + "loss": 0.3782, + "step": 13288 + }, + { + "epoch": 4.91, + "learning_rate": 2.510814139672374e-06, + "loss": 0.5442, + "step": 13289 + }, + { + "epoch": 4.91, + "learning_rate": 2.510777189309028e-06, + "loss": 0.4822, + "step": 13290 + }, + { + "epoch": 4.91, + "learning_rate": 2.510740238945683e-06, + "loss": 0.4745, + "step": 13291 + }, + { + "epoch": 4.91, + "learning_rate": 2.5107032885823377e-06, + "loss": 0.6113, + "step": 13292 + }, + { + "epoch": 4.91, + "learning_rate": 2.5106663382189923e-06, + "loss": 0.505, + "step": 13293 + }, + { + "epoch": 4.91, + "learning_rate": 2.5106293878556474e-06, + "loss": 0.4878, + "step": 13294 + }, + { + "epoch": 4.91, + "learning_rate": 2.510592437492302e-06, + "loss": 0.489, + "step": 13295 + }, + { + "epoch": 4.91, + "learning_rate": 2.5105554871289567e-06, + "loss": 0.5047, + "step": 13296 + }, + { + "epoch": 4.91, + "learning_rate": 2.5105185367656113e-06, + "loss": 0.5106, + "step": 13297 + }, + { + "epoch": 4.91, + "learning_rate": 2.5104815864022664e-06, + "loss": 0.5261, + "step": 13298 + }, + { + "epoch": 4.91, + "learning_rate": 2.510444636038921e-06, + "loss": 0.6229, + "step": 13299 + }, + { + "epoch": 4.91, + "learning_rate": 2.510407685675576e-06, + "loss": 0.5548, + "step": 13300 + }, + { + "epoch": 4.91, + "learning_rate": 2.5103707353122307e-06, + "loss": 0.475, + "step": 13301 + }, + { + "epoch": 4.91, + "learning_rate": 2.5103337849488853e-06, + "loss": 0.5638, + "step": 13302 + }, + { + "epoch": 4.91, + "learning_rate": 2.51029683458554e-06, + "loss": 0.6006, + "step": 13303 + }, + { + "epoch": 4.91, + "learning_rate": 2.5102598842221946e-06, + "loss": 0.4991, + "step": 13304 + }, + { + "epoch": 4.91, + "learning_rate": 2.5102229338588497e-06, + "loss": 0.4174, + "step": 13305 + }, + { + "epoch": 4.91, + "learning_rate": 2.5101859834955043e-06, + "loss": 0.7251, + "step": 13306 + }, + { + "epoch": 4.91, + "learning_rate": 2.5101490331321594e-06, + "loss": 0.5792, + "step": 13307 + }, + { + "epoch": 4.91, + "learning_rate": 2.510112082768814e-06, + "loss": 0.4979, + "step": 13308 + }, + { + "epoch": 4.91, + "learning_rate": 2.5100751324054686e-06, + "loss": 0.6845, + "step": 13309 + }, + { + "epoch": 4.92, + "learning_rate": 2.5100381820421233e-06, + "loss": 0.5508, + "step": 13310 + }, + { + "epoch": 4.92, + "learning_rate": 2.5100012316787783e-06, + "loss": 0.4981, + "step": 13311 + }, + { + "epoch": 4.92, + "learning_rate": 2.509964281315433e-06, + "loss": 0.4451, + "step": 13312 + }, + { + "epoch": 4.92, + "learning_rate": 2.5099273309520876e-06, + "loss": 0.5543, + "step": 13313 + }, + { + "epoch": 4.92, + "learning_rate": 2.5098903805887427e-06, + "loss": 0.4959, + "step": 13314 + }, + { + "epoch": 4.92, + "learning_rate": 2.5098534302253973e-06, + "loss": 0.4912, + "step": 13315 + }, + { + "epoch": 4.92, + "learning_rate": 2.509816479862052e-06, + "loss": 0.6252, + "step": 13316 + }, + { + "epoch": 4.92, + "learning_rate": 2.5097795294987066e-06, + "loss": 0.5026, + "step": 13317 + }, + { + "epoch": 4.92, + "learning_rate": 2.5097425791353616e-06, + "loss": 0.5464, + "step": 13318 + }, + { + "epoch": 4.92, + "learning_rate": 2.5097056287720163e-06, + "loss": 0.5039, + "step": 13319 + }, + { + "epoch": 4.92, + "learning_rate": 2.5096686784086713e-06, + "loss": 0.4721, + "step": 13320 + }, + { + "epoch": 4.92, + "learning_rate": 2.509631728045326e-06, + "loss": 0.5986, + "step": 13321 + }, + { + "epoch": 4.92, + "learning_rate": 2.5095947776819806e-06, + "loss": 0.5217, + "step": 13322 + }, + { + "epoch": 4.92, + "learning_rate": 2.5095578273186353e-06, + "loss": 0.5894, + "step": 13323 + }, + { + "epoch": 4.92, + "learning_rate": 2.50952087695529e-06, + "loss": 0.4331, + "step": 13324 + }, + { + "epoch": 4.92, + "learning_rate": 2.509483926591945e-06, + "loss": 0.4953, + "step": 13325 + }, + { + "epoch": 4.92, + "learning_rate": 2.5094469762285996e-06, + "loss": 0.5291, + "step": 13326 + }, + { + "epoch": 4.92, + "learning_rate": 2.5094100258652546e-06, + "loss": 0.5141, + "step": 13327 + }, + { + "epoch": 4.92, + "learning_rate": 2.5093730755019093e-06, + "loss": 0.3943, + "step": 13328 + }, + { + "epoch": 4.92, + "learning_rate": 2.509336125138564e-06, + "loss": 0.489, + "step": 13329 + }, + { + "epoch": 4.92, + "learning_rate": 2.5092991747752186e-06, + "loss": 0.4784, + "step": 13330 + }, + { + "epoch": 4.92, + "learning_rate": 2.5092622244118736e-06, + "loss": 0.4612, + "step": 13331 + }, + { + "epoch": 4.92, + "learning_rate": 2.5092252740485283e-06, + "loss": 0.4676, + "step": 13332 + }, + { + "epoch": 4.92, + "learning_rate": 2.509188323685183e-06, + "loss": 0.4536, + "step": 13333 + }, + { + "epoch": 4.92, + "learning_rate": 2.509151373321838e-06, + "loss": 0.5252, + "step": 13334 + }, + { + "epoch": 4.92, + "learning_rate": 2.509114422958492e-06, + "loss": 0.5541, + "step": 13335 + }, + { + "epoch": 4.92, + "learning_rate": 2.5090774725951472e-06, + "loss": 0.5984, + "step": 13336 + }, + { + "epoch": 4.93, + "learning_rate": 2.509040522231802e-06, + "loss": 0.4361, + "step": 13337 + }, + { + "epoch": 4.93, + "learning_rate": 2.509003571868457e-06, + "loss": 0.478, + "step": 13338 + }, + { + "epoch": 4.93, + "learning_rate": 2.5089666215051116e-06, + "loss": 0.4601, + "step": 13339 + }, + { + "epoch": 4.93, + "learning_rate": 2.5089296711417666e-06, + "loss": 0.4426, + "step": 13340 + }, + { + "epoch": 4.93, + "learning_rate": 2.508892720778421e-06, + "loss": 0.4701, + "step": 13341 + }, + { + "epoch": 4.93, + "learning_rate": 2.508855770415076e-06, + "loss": 0.6036, + "step": 13342 + }, + { + "epoch": 4.93, + "learning_rate": 2.5088188200517305e-06, + "loss": 0.6127, + "step": 13343 + }, + { + "epoch": 4.93, + "learning_rate": 2.508781869688385e-06, + "loss": 0.4755, + "step": 13344 + }, + { + "epoch": 4.93, + "learning_rate": 2.5087449193250402e-06, + "loss": 0.5559, + "step": 13345 + }, + { + "epoch": 4.93, + "learning_rate": 2.508707968961695e-06, + "loss": 0.5075, + "step": 13346 + }, + { + "epoch": 4.93, + "learning_rate": 2.5086710185983495e-06, + "loss": 0.5567, + "step": 13347 + }, + { + "epoch": 4.93, + "learning_rate": 2.508634068235004e-06, + "loss": 0.535, + "step": 13348 + }, + { + "epoch": 4.93, + "learning_rate": 2.508597117871659e-06, + "loss": 0.4524, + "step": 13349 + }, + { + "epoch": 4.93, + "learning_rate": 2.508560167508314e-06, + "loss": 0.4889, + "step": 13350 + }, + { + "epoch": 4.93, + "learning_rate": 2.508523217144969e-06, + "loss": 0.4601, + "step": 13351 + }, + { + "epoch": 4.93, + "learning_rate": 2.5084862667816235e-06, + "loss": 0.5101, + "step": 13352 + }, + { + "epoch": 4.93, + "learning_rate": 2.508449316418278e-06, + "loss": 0.5733, + "step": 13353 + }, + { + "epoch": 4.93, + "learning_rate": 2.508412366054933e-06, + "loss": 0.611, + "step": 13354 + }, + { + "epoch": 4.93, + "learning_rate": 2.5083754156915874e-06, + "loss": 0.4407, + "step": 13355 + }, + { + "epoch": 4.93, + "learning_rate": 2.5083384653282425e-06, + "loss": 0.5297, + "step": 13356 + }, + { + "epoch": 4.93, + "learning_rate": 2.508301514964897e-06, + "loss": 0.5546, + "step": 13357 + }, + { + "epoch": 4.93, + "learning_rate": 2.508264564601552e-06, + "loss": 0.3914, + "step": 13358 + }, + { + "epoch": 4.93, + "learning_rate": 2.508227614238207e-06, + "loss": 0.4802, + "step": 13359 + }, + { + "epoch": 4.93, + "learning_rate": 2.5081906638748615e-06, + "loss": 0.5551, + "step": 13360 + }, + { + "epoch": 4.93, + "learning_rate": 2.508153713511516e-06, + "loss": 0.5274, + "step": 13361 + }, + { + "epoch": 4.93, + "learning_rate": 2.508116763148171e-06, + "loss": 0.4397, + "step": 13362 + }, + { + "epoch": 4.93, + "learning_rate": 2.508079812784826e-06, + "loss": 0.5135, + "step": 13363 + }, + { + "epoch": 4.94, + "learning_rate": 2.5080428624214804e-06, + "loss": 0.4008, + "step": 13364 + }, + { + "epoch": 4.94, + "learning_rate": 2.5080059120581355e-06, + "loss": 0.438, + "step": 13365 + }, + { + "epoch": 4.94, + "learning_rate": 2.50796896169479e-06, + "loss": 0.5244, + "step": 13366 + }, + { + "epoch": 4.94, + "learning_rate": 2.5079320113314448e-06, + "loss": 0.5833, + "step": 13367 + }, + { + "epoch": 4.94, + "learning_rate": 2.5078950609680994e-06, + "loss": 0.5107, + "step": 13368 + }, + { + "epoch": 4.94, + "learning_rate": 2.5078581106047545e-06, + "loss": 0.5121, + "step": 13369 + }, + { + "epoch": 4.94, + "learning_rate": 2.507821160241409e-06, + "loss": 0.4904, + "step": 13370 + }, + { + "epoch": 4.94, + "learning_rate": 2.507784209878064e-06, + "loss": 0.595, + "step": 13371 + }, + { + "epoch": 4.94, + "learning_rate": 2.507747259514719e-06, + "loss": 0.4962, + "step": 13372 + }, + { + "epoch": 4.94, + "learning_rate": 2.5077103091513734e-06, + "loss": 0.55, + "step": 13373 + }, + { + "epoch": 4.94, + "learning_rate": 2.507673358788028e-06, + "loss": 0.474, + "step": 13374 + }, + { + "epoch": 4.94, + "learning_rate": 2.5076364084246827e-06, + "loss": 0.4949, + "step": 13375 + }, + { + "epoch": 4.94, + "learning_rate": 2.5075994580613378e-06, + "loss": 0.5581, + "step": 13376 + }, + { + "epoch": 4.94, + "learning_rate": 2.5075625076979924e-06, + "loss": 0.5021, + "step": 13377 + }, + { + "epoch": 4.94, + "learning_rate": 2.5075255573346475e-06, + "loss": 0.434, + "step": 13378 + }, + { + "epoch": 4.94, + "learning_rate": 2.507488606971302e-06, + "loss": 0.4385, + "step": 13379 + }, + { + "epoch": 4.94, + "learning_rate": 2.5074516566079567e-06, + "loss": 0.5186, + "step": 13380 + }, + { + "epoch": 4.94, + "learning_rate": 2.5074147062446114e-06, + "loss": 0.5083, + "step": 13381 + }, + { + "epoch": 4.94, + "learning_rate": 2.5073777558812664e-06, + "loss": 0.4744, + "step": 13382 + }, + { + "epoch": 4.94, + "learning_rate": 2.507340805517921e-06, + "loss": 0.5842, + "step": 13383 + }, + { + "epoch": 4.94, + "learning_rate": 2.5073038551545757e-06, + "loss": 0.597, + "step": 13384 + }, + { + "epoch": 4.94, + "learning_rate": 2.5072669047912308e-06, + "loss": 0.4604, + "step": 13385 + }, + { + "epoch": 4.94, + "learning_rate": 2.507229954427885e-06, + "loss": 0.46, + "step": 13386 + }, + { + "epoch": 4.94, + "learning_rate": 2.50719300406454e-06, + "loss": 0.5086, + "step": 13387 + }, + { + "epoch": 4.94, + "learning_rate": 2.5071560537011947e-06, + "loss": 0.5429, + "step": 13388 + }, + { + "epoch": 4.94, + "learning_rate": 2.5071191033378497e-06, + "loss": 0.545, + "step": 13389 + }, + { + "epoch": 4.94, + "learning_rate": 2.5070821529745044e-06, + "loss": 0.4668, + "step": 13390 + }, + { + "epoch": 4.94, + "learning_rate": 2.5070452026111594e-06, + "loss": 0.5386, + "step": 13391 + }, + { + "epoch": 4.95, + "learning_rate": 2.5070082522478136e-06, + "loss": 0.3824, + "step": 13392 + }, + { + "epoch": 4.95, + "learning_rate": 2.5069713018844683e-06, + "loss": 0.6245, + "step": 13393 + }, + { + "epoch": 4.95, + "learning_rate": 2.5069343515211233e-06, + "loss": 0.5289, + "step": 13394 + }, + { + "epoch": 4.95, + "learning_rate": 2.506897401157778e-06, + "loss": 0.4844, + "step": 13395 + }, + { + "epoch": 4.95, + "learning_rate": 2.506860450794433e-06, + "loss": 0.4229, + "step": 13396 + }, + { + "epoch": 4.95, + "learning_rate": 2.5068235004310877e-06, + "loss": 0.5375, + "step": 13397 + }, + { + "epoch": 4.95, + "learning_rate": 2.5067865500677423e-06, + "loss": 0.5566, + "step": 13398 + }, + { + "epoch": 4.95, + "learning_rate": 2.506749599704397e-06, + "loss": 0.5337, + "step": 13399 + }, + { + "epoch": 4.95, + "learning_rate": 2.506712649341052e-06, + "loss": 0.5395, + "step": 13400 + }, + { + "epoch": 4.95, + "learning_rate": 2.5066756989777066e-06, + "loss": 0.4878, + "step": 13401 + }, + { + "epoch": 4.95, + "learning_rate": 2.5066387486143617e-06, + "loss": 0.5429, + "step": 13402 + }, + { + "epoch": 4.95, + "learning_rate": 2.5066017982510163e-06, + "loss": 0.5358, + "step": 13403 + }, + { + "epoch": 4.95, + "learning_rate": 2.506564847887671e-06, + "loss": 0.5608, + "step": 13404 + }, + { + "epoch": 4.95, + "learning_rate": 2.5065278975243256e-06, + "loss": 0.5566, + "step": 13405 + }, + { + "epoch": 4.95, + "learning_rate": 2.5064909471609802e-06, + "loss": 0.495, + "step": 13406 + }, + { + "epoch": 4.95, + "learning_rate": 2.5064539967976353e-06, + "loss": 0.4801, + "step": 13407 + }, + { + "epoch": 4.95, + "learning_rate": 2.50641704643429e-06, + "loss": 0.5573, + "step": 13408 + }, + { + "epoch": 4.95, + "learning_rate": 2.506380096070945e-06, + "loss": 0.4577, + "step": 13409 + }, + { + "epoch": 4.95, + "learning_rate": 2.5063431457075996e-06, + "loss": 0.5005, + "step": 13410 + }, + { + "epoch": 4.95, + "learning_rate": 2.5063061953442543e-06, + "loss": 0.5221, + "step": 13411 + }, + { + "epoch": 4.95, + "learning_rate": 2.506269244980909e-06, + "loss": 0.5333, + "step": 13412 + }, + { + "epoch": 4.95, + "learning_rate": 2.5062322946175635e-06, + "loss": 0.4601, + "step": 13413 + }, + { + "epoch": 4.95, + "learning_rate": 2.5061953442542186e-06, + "loss": 0.4886, + "step": 13414 + }, + { + "epoch": 4.95, + "learning_rate": 2.5061583938908732e-06, + "loss": 0.3695, + "step": 13415 + }, + { + "epoch": 4.95, + "learning_rate": 2.5061214435275283e-06, + "loss": 0.536, + "step": 13416 + }, + { + "epoch": 4.95, + "learning_rate": 2.506084493164183e-06, + "loss": 0.5121, + "step": 13417 + }, + { + "epoch": 4.95, + "learning_rate": 2.5060475428008376e-06, + "loss": 0.601, + "step": 13418 + }, + { + "epoch": 4.96, + "learning_rate": 2.506010592437492e-06, + "loss": 0.432, + "step": 13419 + }, + { + "epoch": 4.96, + "learning_rate": 2.5059736420741473e-06, + "loss": 0.5039, + "step": 13420 + }, + { + "epoch": 4.96, + "learning_rate": 2.505936691710802e-06, + "loss": 0.5223, + "step": 13421 + }, + { + "epoch": 4.96, + "learning_rate": 2.505899741347457e-06, + "loss": 0.4442, + "step": 13422 + }, + { + "epoch": 4.96, + "learning_rate": 2.5058627909841116e-06, + "loss": 0.4485, + "step": 13423 + }, + { + "epoch": 4.96, + "learning_rate": 2.5058258406207662e-06, + "loss": 0.3981, + "step": 13424 + }, + { + "epoch": 4.96, + "learning_rate": 2.505788890257421e-06, + "loss": 0.4435, + "step": 13425 + }, + { + "epoch": 4.96, + "learning_rate": 2.5057519398940755e-06, + "loss": 0.5014, + "step": 13426 + }, + { + "epoch": 4.96, + "learning_rate": 2.5057149895307306e-06, + "loss": 0.5923, + "step": 13427 + }, + { + "epoch": 4.96, + "learning_rate": 2.505678039167385e-06, + "loss": 0.4549, + "step": 13428 + }, + { + "epoch": 4.96, + "learning_rate": 2.5056410888040403e-06, + "loss": 0.5473, + "step": 13429 + }, + { + "epoch": 4.96, + "learning_rate": 2.505604138440695e-06, + "loss": 0.4908, + "step": 13430 + }, + { + "epoch": 4.96, + "learning_rate": 2.5055671880773495e-06, + "loss": 0.4954, + "step": 13431 + }, + { + "epoch": 4.96, + "learning_rate": 2.505530237714004e-06, + "loss": 0.5812, + "step": 13432 + }, + { + "epoch": 4.96, + "learning_rate": 2.505493287350659e-06, + "loss": 0.5318, + "step": 13433 + }, + { + "epoch": 4.96, + "learning_rate": 2.505456336987314e-06, + "loss": 0.6301, + "step": 13434 + }, + { + "epoch": 4.96, + "learning_rate": 2.5054193866239685e-06, + "loss": 0.4279, + "step": 13435 + }, + { + "epoch": 4.96, + "learning_rate": 2.505382436260623e-06, + "loss": 0.4223, + "step": 13436 + }, + { + "epoch": 4.96, + "learning_rate": 2.5053454858972778e-06, + "loss": 0.4112, + "step": 13437 + }, + { + "epoch": 4.96, + "learning_rate": 2.505308535533933e-06, + "loss": 0.6201, + "step": 13438 + }, + { + "epoch": 4.96, + "learning_rate": 2.5052715851705875e-06, + "loss": 0.554, + "step": 13439 + }, + { + "epoch": 4.96, + "learning_rate": 2.5052346348072425e-06, + "loss": 0.4626, + "step": 13440 + }, + { + "epoch": 4.96, + "learning_rate": 2.505197684443897e-06, + "loss": 0.5238, + "step": 13441 + }, + { + "epoch": 4.96, + "learning_rate": 2.505160734080552e-06, + "loss": 0.5002, + "step": 13442 + }, + { + "epoch": 4.96, + "learning_rate": 2.5051237837172064e-06, + "loss": 0.5551, + "step": 13443 + }, + { + "epoch": 4.96, + "learning_rate": 2.505086833353861e-06, + "loss": 0.4068, + "step": 13444 + }, + { + "epoch": 4.96, + "learning_rate": 2.505049882990516e-06, + "loss": 0.4951, + "step": 13445 + }, + { + "epoch": 4.97, + "learning_rate": 2.5050129326271708e-06, + "loss": 0.5722, + "step": 13446 + }, + { + "epoch": 4.97, + "learning_rate": 2.504975982263826e-06, + "loss": 0.4386, + "step": 13447 + }, + { + "epoch": 4.97, + "learning_rate": 2.5049390319004805e-06, + "loss": 0.5594, + "step": 13448 + }, + { + "epoch": 4.97, + "learning_rate": 2.504902081537135e-06, + "loss": 0.4148, + "step": 13449 + }, + { + "epoch": 4.97, + "learning_rate": 2.5048651311737897e-06, + "loss": 0.5324, + "step": 13450 + }, + { + "epoch": 4.97, + "learning_rate": 2.504828180810445e-06, + "loss": 0.513, + "step": 13451 + }, + { + "epoch": 4.97, + "learning_rate": 2.5047912304470994e-06, + "loss": 0.5259, + "step": 13452 + }, + { + "epoch": 4.97, + "learning_rate": 2.504754280083754e-06, + "loss": 0.3808, + "step": 13453 + }, + { + "epoch": 4.97, + "learning_rate": 2.504717329720409e-06, + "loss": 0.4949, + "step": 13454 + }, + { + "epoch": 4.97, + "learning_rate": 2.5046803793570638e-06, + "loss": 0.5127, + "step": 13455 + }, + { + "epoch": 4.97, + "learning_rate": 2.5046434289937184e-06, + "loss": 0.6627, + "step": 13456 + }, + { + "epoch": 4.97, + "learning_rate": 2.504606478630373e-06, + "loss": 0.4661, + "step": 13457 + }, + { + "epoch": 4.97, + "learning_rate": 2.504569528267028e-06, + "loss": 0.4851, + "step": 13458 + }, + { + "epoch": 4.97, + "learning_rate": 2.5045325779036827e-06, + "loss": 0.4976, + "step": 13459 + }, + { + "epoch": 4.97, + "learning_rate": 2.504495627540338e-06, + "loss": 0.5995, + "step": 13460 + }, + { + "epoch": 4.97, + "learning_rate": 2.5044586771769924e-06, + "loss": 0.4782, + "step": 13461 + }, + { + "epoch": 4.97, + "learning_rate": 2.504421726813647e-06, + "loss": 0.4845, + "step": 13462 + }, + { + "epoch": 4.97, + "learning_rate": 2.5043847764503017e-06, + "loss": 0.5361, + "step": 13463 + }, + { + "epoch": 4.97, + "learning_rate": 2.5043478260869564e-06, + "loss": 0.5367, + "step": 13464 + }, + { + "epoch": 4.97, + "learning_rate": 2.5043108757236114e-06, + "loss": 0.4469, + "step": 13465 + }, + { + "epoch": 4.97, + "learning_rate": 2.504273925360266e-06, + "loss": 0.4951, + "step": 13466 + }, + { + "epoch": 4.97, + "learning_rate": 2.504236974996921e-06, + "loss": 0.4579, + "step": 13467 + }, + { + "epoch": 4.97, + "learning_rate": 2.5042000246335757e-06, + "loss": 0.478, + "step": 13468 + }, + { + "epoch": 4.97, + "learning_rate": 2.5041630742702304e-06, + "loss": 0.5438, + "step": 13469 + }, + { + "epoch": 4.97, + "learning_rate": 2.504126123906885e-06, + "loss": 0.5296, + "step": 13470 + }, + { + "epoch": 4.97, + "learning_rate": 2.50408917354354e-06, + "loss": 0.4469, + "step": 13471 + }, + { + "epoch": 4.97, + "learning_rate": 2.5040522231801947e-06, + "loss": 0.472, + "step": 13472 + }, + { + "epoch": 4.98, + "learning_rate": 2.5040152728168494e-06, + "loss": 0.4574, + "step": 13473 + }, + { + "epoch": 4.98, + "learning_rate": 2.5039783224535044e-06, + "loss": 0.6194, + "step": 13474 + }, + { + "epoch": 4.98, + "learning_rate": 2.5039413720901586e-06, + "loss": 0.4236, + "step": 13475 + }, + { + "epoch": 4.98, + "learning_rate": 2.5039044217268137e-06, + "loss": 0.4773, + "step": 13476 + }, + { + "epoch": 4.98, + "learning_rate": 2.5038674713634683e-06, + "loss": 0.4613, + "step": 13477 + }, + { + "epoch": 4.98, + "learning_rate": 2.5038305210001234e-06, + "loss": 0.5103, + "step": 13478 + }, + { + "epoch": 4.98, + "learning_rate": 2.503793570636778e-06, + "loss": 0.505, + "step": 13479 + }, + { + "epoch": 4.98, + "learning_rate": 2.503756620273433e-06, + "loss": 0.4156, + "step": 13480 + }, + { + "epoch": 4.98, + "learning_rate": 2.5037196699100873e-06, + "loss": 0.5116, + "step": 13481 + }, + { + "epoch": 4.98, + "learning_rate": 2.5036827195467424e-06, + "loss": 0.4908, + "step": 13482 + }, + { + "epoch": 4.98, + "learning_rate": 2.503645769183397e-06, + "loss": 0.41, + "step": 13483 + }, + { + "epoch": 4.98, + "learning_rate": 2.5036088188200516e-06, + "loss": 0.4835, + "step": 13484 + }, + { + "epoch": 4.98, + "learning_rate": 2.5035718684567067e-06, + "loss": 0.5972, + "step": 13485 + }, + { + "epoch": 4.98, + "learning_rate": 2.5035349180933613e-06, + "loss": 0.5204, + "step": 13486 + }, + { + "epoch": 4.98, + "learning_rate": 2.503497967730016e-06, + "loss": 0.518, + "step": 13487 + }, + { + "epoch": 4.98, + "learning_rate": 2.5034610173666706e-06, + "loss": 0.5269, + "step": 13488 + }, + { + "epoch": 4.98, + "learning_rate": 2.5034240670033257e-06, + "loss": 0.5819, + "step": 13489 + }, + { + "epoch": 4.98, + "learning_rate": 2.5033871166399803e-06, + "loss": 0.5493, + "step": 13490 + }, + { + "epoch": 4.98, + "learning_rate": 2.5033501662766353e-06, + "loss": 0.5007, + "step": 13491 + }, + { + "epoch": 4.98, + "learning_rate": 2.50331321591329e-06, + "loss": 0.5655, + "step": 13492 + }, + { + "epoch": 4.98, + "learning_rate": 2.5032762655499446e-06, + "loss": 0.4954, + "step": 13493 + }, + { + "epoch": 4.98, + "learning_rate": 2.5032393151865993e-06, + "loss": 0.5216, + "step": 13494 + }, + { + "epoch": 4.98, + "learning_rate": 2.503202364823254e-06, + "loss": 0.6086, + "step": 13495 + }, + { + "epoch": 4.98, + "learning_rate": 2.503165414459909e-06, + "loss": 0.4027, + "step": 13496 + }, + { + "epoch": 4.98, + "learning_rate": 2.5031284640965636e-06, + "loss": 0.5337, + "step": 13497 + }, + { + "epoch": 4.98, + "learning_rate": 2.5030915137332187e-06, + "loss": 0.5636, + "step": 13498 + }, + { + "epoch": 4.98, + "learning_rate": 2.5030545633698733e-06, + "loss": 0.4464, + "step": 13499 + }, + { + "epoch": 4.99, + "learning_rate": 2.503017613006528e-06, + "loss": 0.5893, + "step": 13500 + }, + { + "epoch": 4.99, + "learning_rate": 2.5029806626431826e-06, + "loss": 0.5647, + "step": 13501 + }, + { + "epoch": 4.99, + "learning_rate": 2.5029437122798376e-06, + "loss": 0.5163, + "step": 13502 + }, + { + "epoch": 4.99, + "learning_rate": 2.5029067619164923e-06, + "loss": 0.5573, + "step": 13503 + }, + { + "epoch": 4.99, + "learning_rate": 2.502869811553147e-06, + "loss": 0.4736, + "step": 13504 + }, + { + "epoch": 4.99, + "learning_rate": 2.502832861189802e-06, + "loss": 0.4581, + "step": 13505 + }, + { + "epoch": 4.99, + "learning_rate": 2.5027959108264566e-06, + "loss": 0.4832, + "step": 13506 + }, + { + "epoch": 4.99, + "learning_rate": 2.5027589604631112e-06, + "loss": 0.5609, + "step": 13507 + }, + { + "epoch": 4.99, + "learning_rate": 2.502722010099766e-06, + "loss": 0.5424, + "step": 13508 + }, + { + "epoch": 4.99, + "learning_rate": 2.502685059736421e-06, + "loss": 0.452, + "step": 13509 + }, + { + "epoch": 4.99, + "learning_rate": 2.5026481093730756e-06, + "loss": 0.5064, + "step": 13510 + }, + { + "epoch": 4.99, + "learning_rate": 2.5026111590097306e-06, + "loss": 0.4351, + "step": 13511 + }, + { + "epoch": 4.99, + "learning_rate": 2.5025742086463853e-06, + "loss": 0.4914, + "step": 13512 + }, + { + "epoch": 4.99, + "learning_rate": 2.50253725828304e-06, + "loss": 0.3799, + "step": 13513 + }, + { + "epoch": 4.99, + "learning_rate": 2.5025003079196945e-06, + "loss": 0.5321, + "step": 13514 + }, + { + "epoch": 4.99, + "learning_rate": 2.502463357556349e-06, + "loss": 0.4901, + "step": 13515 + }, + { + "epoch": 4.99, + "learning_rate": 2.5024264071930042e-06, + "loss": 0.5271, + "step": 13516 + }, + { + "epoch": 4.99, + "learning_rate": 2.502389456829659e-06, + "loss": 0.4902, + "step": 13517 + }, + { + "epoch": 4.99, + "learning_rate": 2.502352506466314e-06, + "loss": 0.4827, + "step": 13518 + }, + { + "epoch": 4.99, + "learning_rate": 2.5023155561029686e-06, + "loss": 0.592, + "step": 13519 + }, + { + "epoch": 4.99, + "learning_rate": 2.502278605739623e-06, + "loss": 0.3904, + "step": 13520 + }, + { + "epoch": 4.99, + "learning_rate": 2.502241655376278e-06, + "loss": 0.5394, + "step": 13521 + }, + { + "epoch": 4.99, + "learning_rate": 2.5022047050129325e-06, + "loss": 0.5412, + "step": 13522 + }, + { + "epoch": 4.99, + "learning_rate": 2.5021677546495875e-06, + "loss": 0.4146, + "step": 13523 + }, + { + "epoch": 4.99, + "learning_rate": 2.502130804286242e-06, + "loss": 0.4747, + "step": 13524 + }, + { + "epoch": 4.99, + "learning_rate": 2.5020938539228972e-06, + "loss": 0.564, + "step": 13525 + }, + { + "epoch": 4.99, + "learning_rate": 2.5020569035595514e-06, + "loss": 0.4625, + "step": 13526 + }, + { + "epoch": 5.0, + "learning_rate": 2.5020199531962065e-06, + "loss": 0.3634, + "step": 13527 + }, + { + "epoch": 5.0, + "learning_rate": 2.501983002832861e-06, + "loss": 0.4703, + "step": 13528 + }, + { + "epoch": 5.0, + "learning_rate": 2.501946052469516e-06, + "loss": 0.4437, + "step": 13529 + }, + { + "epoch": 5.0, + "learning_rate": 2.501909102106171e-06, + "loss": 0.4987, + "step": 13530 + }, + { + "epoch": 5.0, + "learning_rate": 2.501872151742826e-06, + "loss": 0.6148, + "step": 13531 + }, + { + "epoch": 5.0, + "learning_rate": 2.50183520137948e-06, + "loss": 0.4698, + "step": 13532 + }, + { + "epoch": 5.0, + "learning_rate": 2.5017982510161347e-06, + "loss": 0.4964, + "step": 13533 + }, + { + "epoch": 5.0, + "learning_rate": 2.50176130065279e-06, + "loss": 0.4809, + "step": 13534 + }, + { + "epoch": 5.0, + "learning_rate": 2.5017243502894444e-06, + "loss": 0.5156, + "step": 13535 + }, + { + "epoch": 5.0, + "learning_rate": 2.5016873999260995e-06, + "loss": 0.4924, + "step": 13536 + }, + { + "epoch": 5.0, + "learning_rate": 2.501650449562754e-06, + "loss": 0.5045, + "step": 13537 + }, + { + "epoch": 5.0, + "learning_rate": 2.5016134991994088e-06, + "loss": 0.3921, + "step": 13538 + }, + { + "epoch": 5.0, + "learning_rate": 2.5015765488360634e-06, + "loss": 0.5571, + "step": 13539 + }, + { + "epoch": 5.0, + "learning_rate": 2.5015395984727185e-06, + "loss": 0.4047, + "step": 13540 + }, + { + "epoch": 5.0, + "eval_accuracy": 0.7352126333287159, + "eval_loss": 0.5278155207633972, + "eval_runtime": 25.1789, + "eval_samples_per_second": 1720.247, + "eval_steps_per_second": 26.888, + "step": 13540 + }, + { + "epoch": 5.0, + "learning_rate": 2.501502648109373e-06, + "loss": 0.5569, + "step": 13541 + }, + { + "epoch": 5.0, + "learning_rate": 2.5014656977460277e-06, + "loss": 0.4412, + "step": 13542 + }, + { + "epoch": 5.0, + "learning_rate": 2.501428747382683e-06, + "loss": 0.503, + "step": 13543 + }, + { + "epoch": 5.0, + "learning_rate": 2.5013917970193374e-06, + "loss": 0.4316, + "step": 13544 + }, + { + "epoch": 5.0, + "learning_rate": 2.501354846655992e-06, + "loss": 0.6787, + "step": 13545 + }, + { + "epoch": 5.0, + "learning_rate": 2.5013178962926467e-06, + "loss": 0.4709, + "step": 13546 + }, + { + "epoch": 5.0, + "learning_rate": 2.5012809459293018e-06, + "loss": 0.4695, + "step": 13547 + }, + { + "epoch": 5.0, + "learning_rate": 2.5012439955659564e-06, + "loss": 0.4315, + "step": 13548 + }, + { + "epoch": 5.0, + "learning_rate": 2.5012070452026115e-06, + "loss": 0.4975, + "step": 13549 + }, + { + "epoch": 5.0, + "learning_rate": 2.501170094839266e-06, + "loss": 0.4869, + "step": 13550 + }, + { + "epoch": 5.0, + "learning_rate": 2.5011331444759207e-06, + "loss": 0.5149, + "step": 13551 + }, + { + "epoch": 5.0, + "learning_rate": 2.5010961941125754e-06, + "loss": 0.6198, + "step": 13552 + }, + { + "epoch": 5.0, + "learning_rate": 2.50105924374923e-06, + "loss": 0.4955, + "step": 13553 + }, + { + "epoch": 5.01, + "learning_rate": 2.501022293385885e-06, + "loss": 0.4263, + "step": 13554 + }, + { + "epoch": 5.01, + "learning_rate": 2.5009853430225397e-06, + "loss": 0.4065, + "step": 13555 + }, + { + "epoch": 5.01, + "learning_rate": 2.5009483926591948e-06, + "loss": 0.6126, + "step": 13556 + }, + { + "epoch": 5.01, + "learning_rate": 2.5009114422958494e-06, + "loss": 0.4697, + "step": 13557 + }, + { + "epoch": 5.01, + "learning_rate": 2.500874491932504e-06, + "loss": 0.5448, + "step": 13558 + }, + { + "epoch": 5.01, + "learning_rate": 2.5008375415691587e-06, + "loss": 0.5114, + "step": 13559 + }, + { + "epoch": 5.01, + "learning_rate": 2.5008005912058137e-06, + "loss": 0.4199, + "step": 13560 + }, + { + "epoch": 5.01, + "learning_rate": 2.5007636408424684e-06, + "loss": 0.4897, + "step": 13561 + }, + { + "epoch": 5.01, + "learning_rate": 2.500726690479123e-06, + "loss": 0.569, + "step": 13562 + }, + { + "epoch": 5.01, + "learning_rate": 2.500689740115778e-06, + "loss": 0.3355, + "step": 13563 + }, + { + "epoch": 5.01, + "learning_rate": 2.5006527897524327e-06, + "loss": 0.5087, + "step": 13564 + }, + { + "epoch": 5.01, + "learning_rate": 2.5006158393890873e-06, + "loss": 0.4681, + "step": 13565 + }, + { + "epoch": 5.01, + "learning_rate": 2.500578889025742e-06, + "loss": 0.4508, + "step": 13566 + }, + { + "epoch": 5.01, + "learning_rate": 2.500541938662397e-06, + "loss": 0.5397, + "step": 13567 + }, + { + "epoch": 5.01, + "learning_rate": 2.5005049882990517e-06, + "loss": 0.5532, + "step": 13568 + }, + { + "epoch": 5.01, + "learning_rate": 2.5004680379357067e-06, + "loss": 0.3666, + "step": 13569 + }, + { + "epoch": 5.01, + "learning_rate": 2.5004310875723614e-06, + "loss": 0.4682, + "step": 13570 + }, + { + "epoch": 5.01, + "learning_rate": 2.500394137209016e-06, + "loss": 0.5191, + "step": 13571 + }, + { + "epoch": 5.01, + "learning_rate": 2.5003571868456706e-06, + "loss": 0.5398, + "step": 13572 + }, + { + "epoch": 5.01, + "learning_rate": 2.5003202364823253e-06, + "loss": 0.4802, + "step": 13573 + }, + { + "epoch": 5.01, + "learning_rate": 2.5002832861189803e-06, + "loss": 0.4809, + "step": 13574 + }, + { + "epoch": 5.01, + "learning_rate": 2.500246335755635e-06, + "loss": 0.5201, + "step": 13575 + }, + { + "epoch": 5.01, + "learning_rate": 2.50020938539229e-06, + "loss": 0.4797, + "step": 13576 + }, + { + "epoch": 5.01, + "learning_rate": 2.5001724350289442e-06, + "loss": 0.434, + "step": 13577 + }, + { + "epoch": 5.01, + "learning_rate": 2.5001354846655993e-06, + "loss": 0.4032, + "step": 13578 + }, + { + "epoch": 5.01, + "learning_rate": 2.500098534302254e-06, + "loss": 0.6311, + "step": 13579 + }, + { + "epoch": 5.01, + "learning_rate": 2.500061583938909e-06, + "loss": 0.3729, + "step": 13580 + }, + { + "epoch": 5.02, + "learning_rate": 2.5000246335755636e-06, + "loss": 0.503, + "step": 13581 + }, + { + "epoch": 5.02, + "learning_rate": 2.4999876832122183e-06, + "loss": 0.4965, + "step": 13582 + }, + { + "epoch": 5.02, + "learning_rate": 2.499950732848873e-06, + "loss": 0.4748, + "step": 13583 + }, + { + "epoch": 5.02, + "learning_rate": 2.4999137824855275e-06, + "loss": 0.474, + "step": 13584 + }, + { + "epoch": 5.02, + "learning_rate": 2.4998768321221826e-06, + "loss": 0.3752, + "step": 13585 + }, + { + "epoch": 5.02, + "learning_rate": 2.4998398817588372e-06, + "loss": 0.526, + "step": 13586 + }, + { + "epoch": 5.02, + "learning_rate": 2.4998029313954923e-06, + "loss": 0.4077, + "step": 13587 + }, + { + "epoch": 5.02, + "learning_rate": 2.499765981032147e-06, + "loss": 0.5448, + "step": 13588 + }, + { + "epoch": 5.02, + "learning_rate": 2.4997290306688016e-06, + "loss": 0.5091, + "step": 13589 + }, + { + "epoch": 5.02, + "learning_rate": 2.4996920803054562e-06, + "loss": 0.5294, + "step": 13590 + }, + { + "epoch": 5.02, + "learning_rate": 2.4996551299421113e-06, + "loss": 0.4707, + "step": 13591 + }, + { + "epoch": 5.02, + "learning_rate": 2.499618179578766e-06, + "loss": 0.3472, + "step": 13592 + }, + { + "epoch": 5.02, + "learning_rate": 2.4995812292154205e-06, + "loss": 0.4556, + "step": 13593 + }, + { + "epoch": 5.02, + "learning_rate": 2.4995442788520756e-06, + "loss": 0.4522, + "step": 13594 + }, + { + "epoch": 5.02, + "learning_rate": 2.4995073284887302e-06, + "loss": 0.5449, + "step": 13595 + }, + { + "epoch": 5.02, + "learning_rate": 2.499470378125385e-06, + "loss": 0.3985, + "step": 13596 + }, + { + "epoch": 5.02, + "learning_rate": 2.4994334277620395e-06, + "loss": 0.4603, + "step": 13597 + }, + { + "epoch": 5.02, + "learning_rate": 2.4993964773986946e-06, + "loss": 0.479, + "step": 13598 + }, + { + "epoch": 5.02, + "learning_rate": 2.4993595270353492e-06, + "loss": 0.4618, + "step": 13599 + }, + { + "epoch": 5.02, + "learning_rate": 2.4993225766720043e-06, + "loss": 0.4539, + "step": 13600 + }, + { + "epoch": 5.02, + "learning_rate": 2.499285626308659e-06, + "loss": 0.5232, + "step": 13601 + }, + { + "epoch": 5.02, + "learning_rate": 2.4992486759453135e-06, + "loss": 0.4196, + "step": 13602 + }, + { + "epoch": 5.02, + "learning_rate": 2.499211725581968e-06, + "loss": 0.502, + "step": 13603 + }, + { + "epoch": 5.02, + "learning_rate": 2.499174775218623e-06, + "loss": 0.5, + "step": 13604 + }, + { + "epoch": 5.02, + "learning_rate": 2.499137824855278e-06, + "loss": 0.5486, + "step": 13605 + }, + { + "epoch": 5.02, + "learning_rate": 2.4991008744919325e-06, + "loss": 0.4549, + "step": 13606 + }, + { + "epoch": 5.02, + "learning_rate": 2.4990639241285876e-06, + "loss": 0.5361, + "step": 13607 + }, + { + "epoch": 5.03, + "learning_rate": 2.499026973765242e-06, + "loss": 0.58, + "step": 13608 + }, + { + "epoch": 5.03, + "learning_rate": 2.498990023401897e-06, + "loss": 0.4582, + "step": 13609 + }, + { + "epoch": 5.03, + "learning_rate": 2.4989530730385515e-06, + "loss": 0.4992, + "step": 13610 + }, + { + "epoch": 5.03, + "learning_rate": 2.4989161226752065e-06, + "loss": 0.4036, + "step": 13611 + }, + { + "epoch": 5.03, + "learning_rate": 2.498879172311861e-06, + "loss": 0.4883, + "step": 13612 + }, + { + "epoch": 5.03, + "learning_rate": 2.498842221948516e-06, + "loss": 0.5624, + "step": 13613 + }, + { + "epoch": 5.03, + "learning_rate": 2.498805271585171e-06, + "loss": 0.4419, + "step": 13614 + }, + { + "epoch": 5.03, + "learning_rate": 2.4987683212218255e-06, + "loss": 0.647, + "step": 13615 + }, + { + "epoch": 5.03, + "learning_rate": 2.49873137085848e-06, + "loss": 0.4599, + "step": 13616 + }, + { + "epoch": 5.03, + "learning_rate": 2.4986944204951348e-06, + "loss": 0.5152, + "step": 13617 + }, + { + "epoch": 5.03, + "learning_rate": 2.49865747013179e-06, + "loss": 0.5663, + "step": 13618 + }, + { + "epoch": 5.03, + "learning_rate": 2.4986205197684445e-06, + "loss": 0.4944, + "step": 13619 + }, + { + "epoch": 5.03, + "learning_rate": 2.4985835694050995e-06, + "loss": 0.5407, + "step": 13620 + }, + { + "epoch": 5.03, + "learning_rate": 2.4985466190417538e-06, + "loss": 0.5421, + "step": 13621 + }, + { + "epoch": 5.03, + "learning_rate": 2.4985096686784084e-06, + "loss": 0.5089, + "step": 13622 + }, + { + "epoch": 5.03, + "learning_rate": 2.4984727183150635e-06, + "loss": 0.4273, + "step": 13623 + }, + { + "epoch": 5.03, + "learning_rate": 2.498435767951718e-06, + "loss": 0.5237, + "step": 13624 + }, + { + "epoch": 5.03, + "learning_rate": 2.498398817588373e-06, + "loss": 0.4887, + "step": 13625 + }, + { + "epoch": 5.03, + "learning_rate": 2.4983618672250278e-06, + "loss": 0.5257, + "step": 13626 + }, + { + "epoch": 5.03, + "learning_rate": 2.4983249168616824e-06, + "loss": 0.5716, + "step": 13627 + }, + { + "epoch": 5.03, + "learning_rate": 2.498287966498337e-06, + "loss": 0.4455, + "step": 13628 + }, + { + "epoch": 5.03, + "learning_rate": 2.498251016134992e-06, + "loss": 0.4596, + "step": 13629 + }, + { + "epoch": 5.03, + "learning_rate": 2.4982140657716468e-06, + "loss": 0.4499, + "step": 13630 + }, + { + "epoch": 5.03, + "learning_rate": 2.498177115408302e-06, + "loss": 0.4743, + "step": 13631 + }, + { + "epoch": 5.03, + "learning_rate": 2.4981401650449564e-06, + "loss": 0.531, + "step": 13632 + }, + { + "epoch": 5.03, + "learning_rate": 2.498103214681611e-06, + "loss": 0.4085, + "step": 13633 + }, + { + "epoch": 5.03, + "learning_rate": 2.4980662643182657e-06, + "loss": 0.5108, + "step": 13634 + }, + { + "epoch": 5.04, + "learning_rate": 2.4980293139549204e-06, + "loss": 0.4806, + "step": 13635 + }, + { + "epoch": 5.04, + "learning_rate": 2.4979923635915754e-06, + "loss": 0.4629, + "step": 13636 + }, + { + "epoch": 5.04, + "learning_rate": 2.49795541322823e-06, + "loss": 0.4284, + "step": 13637 + }, + { + "epoch": 5.04, + "learning_rate": 2.497918462864885e-06, + "loss": 0.461, + "step": 13638 + }, + { + "epoch": 5.04, + "learning_rate": 2.4978815125015398e-06, + "loss": 0.3731, + "step": 13639 + }, + { + "epoch": 5.04, + "learning_rate": 2.4978445621381944e-06, + "loss": 0.603, + "step": 13640 + }, + { + "epoch": 5.04, + "learning_rate": 2.497807611774849e-06, + "loss": 0.5258, + "step": 13641 + }, + { + "epoch": 5.04, + "learning_rate": 2.4977706614115037e-06, + "loss": 0.4352, + "step": 13642 + }, + { + "epoch": 5.04, + "learning_rate": 2.4977337110481587e-06, + "loss": 0.5511, + "step": 13643 + }, + { + "epoch": 5.04, + "learning_rate": 2.4976967606848134e-06, + "loss": 0.5205, + "step": 13644 + }, + { + "epoch": 5.04, + "learning_rate": 2.4976598103214684e-06, + "loss": 0.5361, + "step": 13645 + }, + { + "epoch": 5.04, + "learning_rate": 2.497622859958123e-06, + "loss": 0.4829, + "step": 13646 + }, + { + "epoch": 5.04, + "learning_rate": 2.4975859095947777e-06, + "loss": 0.5435, + "step": 13647 + }, + { + "epoch": 5.04, + "learning_rate": 2.4975489592314323e-06, + "loss": 0.427, + "step": 13648 + }, + { + "epoch": 5.04, + "learning_rate": 2.4975120088680874e-06, + "loss": 0.4305, + "step": 13649 + }, + { + "epoch": 5.04, + "learning_rate": 2.497475058504742e-06, + "loss": 0.3939, + "step": 13650 + }, + { + "epoch": 5.04, + "learning_rate": 2.497438108141397e-06, + "loss": 0.6016, + "step": 13651 + }, + { + "epoch": 5.04, + "learning_rate": 2.4974011577780517e-06, + "loss": 0.5137, + "step": 13652 + }, + { + "epoch": 5.04, + "learning_rate": 2.4973642074147064e-06, + "loss": 0.5977, + "step": 13653 + }, + { + "epoch": 5.04, + "learning_rate": 2.497327257051361e-06, + "loss": 0.4637, + "step": 13654 + }, + { + "epoch": 5.04, + "learning_rate": 2.4972903066880156e-06, + "loss": 0.5527, + "step": 13655 + }, + { + "epoch": 5.04, + "learning_rate": 2.4972533563246707e-06, + "loss": 0.4149, + "step": 13656 + }, + { + "epoch": 5.04, + "learning_rate": 2.4972164059613253e-06, + "loss": 0.5136, + "step": 13657 + }, + { + "epoch": 5.04, + "learning_rate": 2.4971794555979804e-06, + "loss": 0.4217, + "step": 13658 + }, + { + "epoch": 5.04, + "learning_rate": 2.497142505234635e-06, + "loss": 0.5084, + "step": 13659 + }, + { + "epoch": 5.04, + "learning_rate": 2.4971055548712897e-06, + "loss": 0.3701, + "step": 13660 + }, + { + "epoch": 5.04, + "learning_rate": 2.4970686045079443e-06, + "loss": 0.5684, + "step": 13661 + }, + { + "epoch": 5.05, + "learning_rate": 2.497031654144599e-06, + "loss": 0.4216, + "step": 13662 + }, + { + "epoch": 5.05, + "learning_rate": 2.496994703781254e-06, + "loss": 0.3993, + "step": 13663 + }, + { + "epoch": 5.05, + "learning_rate": 2.4969577534179086e-06, + "loss": 0.5302, + "step": 13664 + }, + { + "epoch": 5.05, + "learning_rate": 2.4969208030545637e-06, + "loss": 0.5398, + "step": 13665 + }, + { + "epoch": 5.05, + "learning_rate": 2.496883852691218e-06, + "loss": 0.5186, + "step": 13666 + }, + { + "epoch": 5.05, + "learning_rate": 2.496846902327873e-06, + "loss": 0.5829, + "step": 13667 + }, + { + "epoch": 5.05, + "learning_rate": 2.4968099519645276e-06, + "loss": 0.3935, + "step": 13668 + }, + { + "epoch": 5.05, + "learning_rate": 2.4967730016011827e-06, + "loss": 0.394, + "step": 13669 + }, + { + "epoch": 5.05, + "learning_rate": 2.4967360512378373e-06, + "loss": 0.531, + "step": 13670 + }, + { + "epoch": 5.05, + "learning_rate": 2.4966991008744924e-06, + "loss": 0.445, + "step": 13671 + }, + { + "epoch": 5.05, + "learning_rate": 2.4966621505111466e-06, + "loss": 0.4718, + "step": 13672 + }, + { + "epoch": 5.05, + "learning_rate": 2.496625200147801e-06, + "loss": 0.5023, + "step": 13673 + }, + { + "epoch": 5.05, + "learning_rate": 2.4965882497844563e-06, + "loss": 0.454, + "step": 13674 + }, + { + "epoch": 5.05, + "learning_rate": 2.496551299421111e-06, + "loss": 0.4773, + "step": 13675 + }, + { + "epoch": 5.05, + "learning_rate": 2.496514349057766e-06, + "loss": 0.6079, + "step": 13676 + }, + { + "epoch": 5.05, + "learning_rate": 2.4964773986944206e-06, + "loss": 0.5194, + "step": 13677 + }, + { + "epoch": 5.05, + "learning_rate": 2.4964404483310752e-06, + "loss": 0.6375, + "step": 13678 + }, + { + "epoch": 5.05, + "learning_rate": 2.49640349796773e-06, + "loss": 0.4646, + "step": 13679 + }, + { + "epoch": 5.05, + "learning_rate": 2.496366547604385e-06, + "loss": 0.4346, + "step": 13680 + }, + { + "epoch": 5.05, + "learning_rate": 2.4963295972410396e-06, + "loss": 0.4863, + "step": 13681 + }, + { + "epoch": 5.05, + "learning_rate": 2.496292646877694e-06, + "loss": 0.5499, + "step": 13682 + }, + { + "epoch": 5.05, + "learning_rate": 2.4962556965143493e-06, + "loss": 0.608, + "step": 13683 + }, + { + "epoch": 5.05, + "learning_rate": 2.496218746151004e-06, + "loss": 0.4915, + "step": 13684 + }, + { + "epoch": 5.05, + "learning_rate": 2.4961817957876585e-06, + "loss": 0.4576, + "step": 13685 + }, + { + "epoch": 5.05, + "learning_rate": 2.496144845424313e-06, + "loss": 0.4947, + "step": 13686 + }, + { + "epoch": 5.05, + "learning_rate": 2.4961078950609682e-06, + "loss": 0.4239, + "step": 13687 + }, + { + "epoch": 5.05, + "learning_rate": 2.496070944697623e-06, + "loss": 0.493, + "step": 13688 + }, + { + "epoch": 5.06, + "learning_rate": 2.496033994334278e-06, + "loss": 0.4409, + "step": 13689 + }, + { + "epoch": 5.06, + "learning_rate": 2.4959970439709326e-06, + "loss": 0.458, + "step": 13690 + }, + { + "epoch": 5.06, + "learning_rate": 2.495960093607587e-06, + "loss": 0.3591, + "step": 13691 + }, + { + "epoch": 5.06, + "learning_rate": 2.495923143244242e-06, + "loss": 0.3879, + "step": 13692 + }, + { + "epoch": 5.06, + "learning_rate": 2.4958861928808965e-06, + "loss": 0.5121, + "step": 13693 + }, + { + "epoch": 5.06, + "learning_rate": 2.4958492425175515e-06, + "loss": 0.5112, + "step": 13694 + }, + { + "epoch": 5.06, + "learning_rate": 2.495812292154206e-06, + "loss": 0.5738, + "step": 13695 + }, + { + "epoch": 5.06, + "learning_rate": 2.4957753417908612e-06, + "loss": 0.456, + "step": 13696 + }, + { + "epoch": 5.06, + "learning_rate": 2.495738391427516e-06, + "loss": 0.5312, + "step": 13697 + }, + { + "epoch": 5.06, + "learning_rate": 2.4957014410641705e-06, + "loss": 0.4362, + "step": 13698 + }, + { + "epoch": 5.06, + "learning_rate": 2.495664490700825e-06, + "loss": 0.4872, + "step": 13699 + }, + { + "epoch": 5.06, + "learning_rate": 2.49562754033748e-06, + "loss": 0.4394, + "step": 13700 + }, + { + "epoch": 5.06, + "learning_rate": 2.495590589974135e-06, + "loss": 0.5135, + "step": 13701 + }, + { + "epoch": 5.06, + "learning_rate": 2.4955536396107895e-06, + "loss": 0.5361, + "step": 13702 + }, + { + "epoch": 5.06, + "learning_rate": 2.4955166892474445e-06, + "loss": 0.4915, + "step": 13703 + }, + { + "epoch": 5.06, + "learning_rate": 2.495479738884099e-06, + "loss": 0.5982, + "step": 13704 + }, + { + "epoch": 5.06, + "learning_rate": 2.495442788520754e-06, + "loss": 0.5731, + "step": 13705 + }, + { + "epoch": 5.06, + "learning_rate": 2.4954058381574084e-06, + "loss": 0.5933, + "step": 13706 + }, + { + "epoch": 5.06, + "learning_rate": 2.4953688877940635e-06, + "loss": 0.5339, + "step": 13707 + }, + { + "epoch": 5.06, + "learning_rate": 2.495331937430718e-06, + "loss": 0.5199, + "step": 13708 + }, + { + "epoch": 5.06, + "learning_rate": 2.495294987067373e-06, + "loss": 0.5393, + "step": 13709 + }, + { + "epoch": 5.06, + "learning_rate": 2.495258036704028e-06, + "loss": 0.4001, + "step": 13710 + }, + { + "epoch": 5.06, + "learning_rate": 2.4952210863406825e-06, + "loss": 0.4177, + "step": 13711 + }, + { + "epoch": 5.06, + "learning_rate": 2.495184135977337e-06, + "loss": 0.5796, + "step": 13712 + }, + { + "epoch": 5.06, + "learning_rate": 2.4951471856139917e-06, + "loss": 0.4847, + "step": 13713 + }, + { + "epoch": 5.06, + "learning_rate": 2.495110235250647e-06, + "loss": 0.4432, + "step": 13714 + }, + { + "epoch": 5.06, + "learning_rate": 2.4950732848873014e-06, + "loss": 0.4977, + "step": 13715 + }, + { + "epoch": 5.06, + "learning_rate": 2.4950363345239565e-06, + "loss": 0.5627, + "step": 13716 + }, + { + "epoch": 5.07, + "learning_rate": 2.4949993841606107e-06, + "loss": 0.5158, + "step": 13717 + }, + { + "epoch": 5.07, + "learning_rate": 2.4949624337972658e-06, + "loss": 0.552, + "step": 13718 + }, + { + "epoch": 5.07, + "learning_rate": 2.4949254834339204e-06, + "loss": 0.5684, + "step": 13719 + }, + { + "epoch": 5.07, + "learning_rate": 2.4948885330705755e-06, + "loss": 0.506, + "step": 13720 + }, + { + "epoch": 5.07, + "learning_rate": 2.49485158270723e-06, + "loss": 0.495, + "step": 13721 + }, + { + "epoch": 5.07, + "learning_rate": 2.4948146323438847e-06, + "loss": 0.5582, + "step": 13722 + }, + { + "epoch": 5.07, + "learning_rate": 2.4947776819805394e-06, + "loss": 0.4212, + "step": 13723 + }, + { + "epoch": 5.07, + "learning_rate": 2.494740731617194e-06, + "loss": 0.5517, + "step": 13724 + }, + { + "epoch": 5.07, + "learning_rate": 2.494703781253849e-06, + "loss": 0.4561, + "step": 13725 + }, + { + "epoch": 5.07, + "learning_rate": 2.4946668308905037e-06, + "loss": 0.5525, + "step": 13726 + }, + { + "epoch": 5.07, + "learning_rate": 2.4946298805271588e-06, + "loss": 0.4701, + "step": 13727 + }, + { + "epoch": 5.07, + "learning_rate": 2.4945929301638134e-06, + "loss": 0.4917, + "step": 13728 + }, + { + "epoch": 5.07, + "learning_rate": 2.494555979800468e-06, + "loss": 0.5009, + "step": 13729 + }, + { + "epoch": 5.07, + "learning_rate": 2.4945190294371227e-06, + "loss": 0.4889, + "step": 13730 + }, + { + "epoch": 5.07, + "learning_rate": 2.4944820790737777e-06, + "loss": 0.5012, + "step": 13731 + }, + { + "epoch": 5.07, + "learning_rate": 2.4944451287104324e-06, + "loss": 0.5807, + "step": 13732 + }, + { + "epoch": 5.07, + "learning_rate": 2.494408178347087e-06, + "loss": 0.5998, + "step": 13733 + }, + { + "epoch": 5.07, + "learning_rate": 2.494371227983742e-06, + "loss": 0.4477, + "step": 13734 + }, + { + "epoch": 5.07, + "learning_rate": 2.4943342776203967e-06, + "loss": 0.528, + "step": 13735 + }, + { + "epoch": 5.07, + "learning_rate": 2.4942973272570513e-06, + "loss": 0.442, + "step": 13736 + }, + { + "epoch": 5.07, + "learning_rate": 2.494260376893706e-06, + "loss": 0.5576, + "step": 13737 + }, + { + "epoch": 5.07, + "learning_rate": 2.494223426530361e-06, + "loss": 0.4207, + "step": 13738 + }, + { + "epoch": 5.07, + "learning_rate": 2.4941864761670157e-06, + "loss": 0.4358, + "step": 13739 + }, + { + "epoch": 5.07, + "learning_rate": 2.4941495258036707e-06, + "loss": 0.5489, + "step": 13740 + }, + { + "epoch": 5.07, + "learning_rate": 2.4941125754403254e-06, + "loss": 0.5321, + "step": 13741 + }, + { + "epoch": 5.07, + "learning_rate": 2.49407562507698e-06, + "loss": 0.5434, + "step": 13742 + }, + { + "epoch": 5.07, + "learning_rate": 2.4940386747136346e-06, + "loss": 0.5665, + "step": 13743 + }, + { + "epoch": 5.08, + "learning_rate": 2.4940017243502893e-06, + "loss": 0.527, + "step": 13744 + }, + { + "epoch": 5.08, + "learning_rate": 2.4939647739869443e-06, + "loss": 0.4071, + "step": 13745 + }, + { + "epoch": 5.08, + "learning_rate": 2.493927823623599e-06, + "loss": 0.5117, + "step": 13746 + }, + { + "epoch": 5.08, + "learning_rate": 2.493890873260254e-06, + "loss": 0.5, + "step": 13747 + }, + { + "epoch": 5.08, + "learning_rate": 2.4938539228969087e-06, + "loss": 0.4082, + "step": 13748 + }, + { + "epoch": 5.08, + "learning_rate": 2.4938169725335633e-06, + "loss": 0.5015, + "step": 13749 + }, + { + "epoch": 5.08, + "learning_rate": 2.493780022170218e-06, + "loss": 0.4195, + "step": 13750 + }, + { + "epoch": 5.08, + "learning_rate": 2.4937430718068726e-06, + "loss": 0.4706, + "step": 13751 + }, + { + "epoch": 5.08, + "learning_rate": 2.4937061214435276e-06, + "loss": 0.4721, + "step": 13752 + }, + { + "epoch": 5.08, + "learning_rate": 2.4936691710801823e-06, + "loss": 0.4824, + "step": 13753 + }, + { + "epoch": 5.08, + "learning_rate": 2.4936322207168373e-06, + "loss": 0.4508, + "step": 13754 + }, + { + "epoch": 5.08, + "learning_rate": 2.493595270353492e-06, + "loss": 0.4644, + "step": 13755 + }, + { + "epoch": 5.08, + "learning_rate": 2.4935583199901466e-06, + "loss": 0.4849, + "step": 13756 + }, + { + "epoch": 5.08, + "learning_rate": 2.4935213696268013e-06, + "loss": 0.5317, + "step": 13757 + }, + { + "epoch": 5.08, + "learning_rate": 2.4934844192634563e-06, + "loss": 0.4289, + "step": 13758 + }, + { + "epoch": 5.08, + "learning_rate": 2.493447468900111e-06, + "loss": 0.4851, + "step": 13759 + }, + { + "epoch": 5.08, + "learning_rate": 2.493410518536766e-06, + "loss": 0.5374, + "step": 13760 + }, + { + "epoch": 5.08, + "learning_rate": 2.4933735681734206e-06, + "loss": 0.5077, + "step": 13761 + }, + { + "epoch": 5.08, + "learning_rate": 2.493336617810075e-06, + "loss": 0.4315, + "step": 13762 + }, + { + "epoch": 5.08, + "learning_rate": 2.49329966744673e-06, + "loss": 0.4402, + "step": 13763 + }, + { + "epoch": 5.08, + "learning_rate": 2.4932627170833846e-06, + "loss": 0.486, + "step": 13764 + }, + { + "epoch": 5.08, + "learning_rate": 2.4932257667200396e-06, + "loss": 0.3755, + "step": 13765 + }, + { + "epoch": 5.08, + "learning_rate": 2.4931888163566942e-06, + "loss": 0.4766, + "step": 13766 + }, + { + "epoch": 5.08, + "learning_rate": 2.4931518659933493e-06, + "loss": 0.4757, + "step": 13767 + }, + { + "epoch": 5.08, + "learning_rate": 2.4931149156300035e-06, + "loss": 0.4371, + "step": 13768 + }, + { + "epoch": 5.08, + "learning_rate": 2.4930779652666586e-06, + "loss": 0.514, + "step": 13769 + }, + { + "epoch": 5.08, + "learning_rate": 2.4930410149033132e-06, + "loss": 0.5048, + "step": 13770 + }, + { + "epoch": 5.09, + "learning_rate": 2.493004064539968e-06, + "loss": 0.5597, + "step": 13771 + }, + { + "epoch": 5.09, + "learning_rate": 2.492967114176623e-06, + "loss": 0.5653, + "step": 13772 + }, + { + "epoch": 5.09, + "learning_rate": 2.4929301638132776e-06, + "loss": 0.5503, + "step": 13773 + }, + { + "epoch": 5.09, + "learning_rate": 2.492893213449932e-06, + "loss": 0.5883, + "step": 13774 + }, + { + "epoch": 5.09, + "learning_rate": 2.492856263086587e-06, + "loss": 0.4407, + "step": 13775 + }, + { + "epoch": 5.09, + "learning_rate": 2.492819312723242e-06, + "loss": 0.3705, + "step": 13776 + }, + { + "epoch": 5.09, + "learning_rate": 2.4927823623598965e-06, + "loss": 0.4683, + "step": 13777 + }, + { + "epoch": 5.09, + "learning_rate": 2.4927454119965516e-06, + "loss": 0.4858, + "step": 13778 + }, + { + "epoch": 5.09, + "learning_rate": 2.4927084616332062e-06, + "loss": 0.528, + "step": 13779 + }, + { + "epoch": 5.09, + "learning_rate": 2.492671511269861e-06, + "loss": 0.5092, + "step": 13780 + }, + { + "epoch": 5.09, + "learning_rate": 2.4926345609065155e-06, + "loss": 0.56, + "step": 13781 + }, + { + "epoch": 5.09, + "learning_rate": 2.49259761054317e-06, + "loss": 0.5409, + "step": 13782 + }, + { + "epoch": 5.09, + "learning_rate": 2.492560660179825e-06, + "loss": 0.4206, + "step": 13783 + }, + { + "epoch": 5.09, + "learning_rate": 2.49252370981648e-06, + "loss": 0.4492, + "step": 13784 + }, + { + "epoch": 5.09, + "learning_rate": 2.492486759453135e-06, + "loss": 0.4264, + "step": 13785 + }, + { + "epoch": 5.09, + "learning_rate": 2.4924498090897895e-06, + "loss": 0.4944, + "step": 13786 + }, + { + "epoch": 5.09, + "learning_rate": 2.492412858726444e-06, + "loss": 0.4457, + "step": 13787 + }, + { + "epoch": 5.09, + "learning_rate": 2.492375908363099e-06, + "loss": 0.4958, + "step": 13788 + }, + { + "epoch": 5.09, + "learning_rate": 2.492338957999754e-06, + "loss": 0.4639, + "step": 13789 + }, + { + "epoch": 5.09, + "learning_rate": 2.4923020076364085e-06, + "loss": 0.3069, + "step": 13790 + }, + { + "epoch": 5.09, + "learning_rate": 2.492265057273063e-06, + "loss": 0.5734, + "step": 13791 + }, + { + "epoch": 5.09, + "learning_rate": 2.492228106909718e-06, + "loss": 0.5651, + "step": 13792 + }, + { + "epoch": 5.09, + "learning_rate": 2.492191156546373e-06, + "loss": 0.5558, + "step": 13793 + }, + { + "epoch": 5.09, + "learning_rate": 2.4921542061830275e-06, + "loss": 0.5343, + "step": 13794 + }, + { + "epoch": 5.09, + "learning_rate": 2.492117255819682e-06, + "loss": 0.5033, + "step": 13795 + }, + { + "epoch": 5.09, + "learning_rate": 2.492080305456337e-06, + "loss": 0.5015, + "step": 13796 + }, + { + "epoch": 5.09, + "learning_rate": 2.492043355092992e-06, + "loss": 0.4329, + "step": 13797 + }, + { + "epoch": 5.1, + "learning_rate": 2.492006404729647e-06, + "loss": 0.4049, + "step": 13798 + }, + { + "epoch": 5.1, + "learning_rate": 2.4919694543663015e-06, + "loss": 0.426, + "step": 13799 + }, + { + "epoch": 5.1, + "learning_rate": 2.491932504002956e-06, + "loss": 0.4629, + "step": 13800 + }, + { + "epoch": 5.1, + "learning_rate": 2.4918955536396108e-06, + "loss": 0.5331, + "step": 13801 + }, + { + "epoch": 5.1, + "learning_rate": 2.4918586032762654e-06, + "loss": 0.5235, + "step": 13802 + }, + { + "epoch": 5.1, + "learning_rate": 2.4918216529129205e-06, + "loss": 0.3877, + "step": 13803 + }, + { + "epoch": 5.1, + "learning_rate": 2.491784702549575e-06, + "loss": 0.5109, + "step": 13804 + }, + { + "epoch": 5.1, + "learning_rate": 2.49174775218623e-06, + "loss": 0.3858, + "step": 13805 + }, + { + "epoch": 5.1, + "learning_rate": 2.4917108018228844e-06, + "loss": 0.4778, + "step": 13806 + }, + { + "epoch": 5.1, + "learning_rate": 2.4916738514595394e-06, + "loss": 0.5817, + "step": 13807 + }, + { + "epoch": 5.1, + "learning_rate": 2.491636901096194e-06, + "loss": 0.4302, + "step": 13808 + }, + { + "epoch": 5.1, + "learning_rate": 2.491599950732849e-06, + "loss": 0.5624, + "step": 13809 + }, + { + "epoch": 5.1, + "learning_rate": 2.4915630003695038e-06, + "loss": 0.4427, + "step": 13810 + }, + { + "epoch": 5.1, + "learning_rate": 2.4915260500061584e-06, + "loss": 0.5964, + "step": 13811 + }, + { + "epoch": 5.1, + "learning_rate": 2.491489099642813e-06, + "loss": 0.4306, + "step": 13812 + }, + { + "epoch": 5.1, + "learning_rate": 2.4914521492794677e-06, + "loss": 0.5729, + "step": 13813 + }, + { + "epoch": 5.1, + "learning_rate": 2.4914151989161227e-06, + "loss": 0.5522, + "step": 13814 + }, + { + "epoch": 5.1, + "learning_rate": 2.4913782485527774e-06, + "loss": 0.4586, + "step": 13815 + }, + { + "epoch": 5.1, + "learning_rate": 2.4913412981894324e-06, + "loss": 0.4084, + "step": 13816 + }, + { + "epoch": 5.1, + "learning_rate": 2.491304347826087e-06, + "loss": 0.4104, + "step": 13817 + }, + { + "epoch": 5.1, + "learning_rate": 2.4912673974627417e-06, + "loss": 0.4228, + "step": 13818 + }, + { + "epoch": 5.1, + "learning_rate": 2.4912304470993963e-06, + "loss": 0.4578, + "step": 13819 + }, + { + "epoch": 5.1, + "learning_rate": 2.4911934967360514e-06, + "loss": 0.436, + "step": 13820 + }, + { + "epoch": 5.1, + "learning_rate": 2.491156546372706e-06, + "loss": 0.5779, + "step": 13821 + }, + { + "epoch": 5.1, + "learning_rate": 2.4911195960093607e-06, + "loss": 0.4029, + "step": 13822 + }, + { + "epoch": 5.1, + "learning_rate": 2.4910826456460157e-06, + "loss": 0.5935, + "step": 13823 + }, + { + "epoch": 5.1, + "learning_rate": 2.4910456952826704e-06, + "loss": 0.4652, + "step": 13824 + }, + { + "epoch": 5.11, + "learning_rate": 2.491008744919325e-06, + "loss": 0.4244, + "step": 13825 + }, + { + "epoch": 5.11, + "learning_rate": 2.4909717945559796e-06, + "loss": 0.4914, + "step": 13826 + }, + { + "epoch": 5.11, + "learning_rate": 2.4909348441926347e-06, + "loss": 0.4274, + "step": 13827 + }, + { + "epoch": 5.11, + "learning_rate": 2.4908978938292893e-06, + "loss": 0.5504, + "step": 13828 + }, + { + "epoch": 5.11, + "learning_rate": 2.4908609434659444e-06, + "loss": 0.4597, + "step": 13829 + }, + { + "epoch": 5.11, + "learning_rate": 2.490823993102599e-06, + "loss": 0.4292, + "step": 13830 + }, + { + "epoch": 5.11, + "learning_rate": 2.4907870427392537e-06, + "loss": 0.3283, + "step": 13831 + }, + { + "epoch": 5.11, + "learning_rate": 2.4907500923759083e-06, + "loss": 0.4437, + "step": 13832 + }, + { + "epoch": 5.11, + "learning_rate": 2.490713142012563e-06, + "loss": 0.4386, + "step": 13833 + }, + { + "epoch": 5.11, + "learning_rate": 2.490676191649218e-06, + "loss": 0.4612, + "step": 13834 + }, + { + "epoch": 5.11, + "learning_rate": 2.4906392412858726e-06, + "loss": 0.4795, + "step": 13835 + }, + { + "epoch": 5.11, + "learning_rate": 2.4906022909225277e-06, + "loss": 0.5344, + "step": 13836 + }, + { + "epoch": 5.11, + "learning_rate": 2.4905653405591823e-06, + "loss": 0.5301, + "step": 13837 + }, + { + "epoch": 5.11, + "learning_rate": 2.490528390195837e-06, + "loss": 0.5783, + "step": 13838 + }, + { + "epoch": 5.11, + "learning_rate": 2.4904914398324916e-06, + "loss": 0.4008, + "step": 13839 + }, + { + "epoch": 5.11, + "learning_rate": 2.4904544894691467e-06, + "loss": 0.4309, + "step": 13840 + }, + { + "epoch": 5.11, + "learning_rate": 2.4904175391058013e-06, + "loss": 0.3688, + "step": 13841 + }, + { + "epoch": 5.11, + "learning_rate": 2.490380588742456e-06, + "loss": 0.4074, + "step": 13842 + }, + { + "epoch": 5.11, + "learning_rate": 2.490343638379111e-06, + "loss": 0.3714, + "step": 13843 + }, + { + "epoch": 5.11, + "learning_rate": 2.4903066880157656e-06, + "loss": 0.4179, + "step": 13844 + }, + { + "epoch": 5.11, + "learning_rate": 2.4902697376524203e-06, + "loss": 0.6478, + "step": 13845 + }, + { + "epoch": 5.11, + "learning_rate": 2.490232787289075e-06, + "loss": 0.4917, + "step": 13846 + }, + { + "epoch": 5.11, + "learning_rate": 2.49019583692573e-06, + "loss": 0.4245, + "step": 13847 + }, + { + "epoch": 5.11, + "learning_rate": 2.4901588865623846e-06, + "loss": 0.4858, + "step": 13848 + }, + { + "epoch": 5.11, + "learning_rate": 2.4901219361990397e-06, + "loss": 0.5154, + "step": 13849 + }, + { + "epoch": 5.11, + "learning_rate": 2.4900849858356943e-06, + "loss": 0.4568, + "step": 13850 + }, + { + "epoch": 5.11, + "learning_rate": 2.4900480354723485e-06, + "loss": 0.4409, + "step": 13851 + }, + { + "epoch": 5.12, + "learning_rate": 2.4900110851090036e-06, + "loss": 0.5624, + "step": 13852 + }, + { + "epoch": 5.12, + "learning_rate": 2.489974134745658e-06, + "loss": 0.3338, + "step": 13853 + }, + { + "epoch": 5.12, + "learning_rate": 2.4899371843823133e-06, + "loss": 0.5737, + "step": 13854 + }, + { + "epoch": 5.12, + "learning_rate": 2.489900234018968e-06, + "loss": 0.5645, + "step": 13855 + }, + { + "epoch": 5.12, + "learning_rate": 2.489863283655623e-06, + "loss": 0.5625, + "step": 13856 + }, + { + "epoch": 5.12, + "learning_rate": 2.489826333292277e-06, + "loss": 0.4582, + "step": 13857 + }, + { + "epoch": 5.12, + "learning_rate": 2.4897893829289322e-06, + "loss": 0.5365, + "step": 13858 + }, + { + "epoch": 5.12, + "learning_rate": 2.489752432565587e-06, + "loss": 0.7677, + "step": 13859 + }, + { + "epoch": 5.12, + "learning_rate": 2.489715482202242e-06, + "loss": 0.4568, + "step": 13860 + }, + { + "epoch": 5.12, + "learning_rate": 2.4896785318388966e-06, + "loss": 0.3933, + "step": 13861 + }, + { + "epoch": 5.12, + "learning_rate": 2.489641581475551e-06, + "loss": 0.4668, + "step": 13862 + }, + { + "epoch": 5.12, + "learning_rate": 2.489604631112206e-06, + "loss": 0.4398, + "step": 13863 + }, + { + "epoch": 5.12, + "learning_rate": 2.4895676807488605e-06, + "loss": 0.4476, + "step": 13864 + }, + { + "epoch": 5.12, + "learning_rate": 2.4895307303855155e-06, + "loss": 0.3942, + "step": 13865 + }, + { + "epoch": 5.12, + "learning_rate": 2.48949378002217e-06, + "loss": 0.4673, + "step": 13866 + }, + { + "epoch": 5.12, + "learning_rate": 2.4894568296588252e-06, + "loss": 0.4776, + "step": 13867 + }, + { + "epoch": 5.12, + "learning_rate": 2.48941987929548e-06, + "loss": 0.4893, + "step": 13868 + }, + { + "epoch": 5.12, + "learning_rate": 2.4893829289321345e-06, + "loss": 0.5165, + "step": 13869 + }, + { + "epoch": 5.12, + "learning_rate": 2.489345978568789e-06, + "loss": 0.5076, + "step": 13870 + }, + { + "epoch": 5.12, + "learning_rate": 2.4893090282054438e-06, + "loss": 0.4656, + "step": 13871 + }, + { + "epoch": 5.12, + "learning_rate": 2.489272077842099e-06, + "loss": 0.501, + "step": 13872 + }, + { + "epoch": 5.12, + "learning_rate": 2.4892351274787535e-06, + "loss": 0.5437, + "step": 13873 + }, + { + "epoch": 5.12, + "learning_rate": 2.4891981771154085e-06, + "loss": 0.4771, + "step": 13874 + }, + { + "epoch": 5.12, + "learning_rate": 2.489161226752063e-06, + "loss": 0.462, + "step": 13875 + }, + { + "epoch": 5.12, + "learning_rate": 2.489124276388718e-06, + "loss": 0.5558, + "step": 13876 + }, + { + "epoch": 5.12, + "learning_rate": 2.4890873260253724e-06, + "loss": 0.5715, + "step": 13877 + }, + { + "epoch": 5.12, + "learning_rate": 2.4890503756620275e-06, + "loss": 0.3676, + "step": 13878 + }, + { + "epoch": 5.13, + "learning_rate": 2.489013425298682e-06, + "loss": 0.4437, + "step": 13879 + }, + { + "epoch": 5.13, + "learning_rate": 2.488976474935337e-06, + "loss": 0.5304, + "step": 13880 + }, + { + "epoch": 5.13, + "learning_rate": 2.488939524571992e-06, + "loss": 0.5165, + "step": 13881 + }, + { + "epoch": 5.13, + "learning_rate": 2.4889025742086465e-06, + "loss": 0.5339, + "step": 13882 + }, + { + "epoch": 5.13, + "learning_rate": 2.488865623845301e-06, + "loss": 0.5148, + "step": 13883 + }, + { + "epoch": 5.13, + "learning_rate": 2.4888286734819557e-06, + "loss": 0.5678, + "step": 13884 + }, + { + "epoch": 5.13, + "learning_rate": 2.488791723118611e-06, + "loss": 0.5874, + "step": 13885 + }, + { + "epoch": 5.13, + "learning_rate": 2.4887547727552654e-06, + "loss": 0.5339, + "step": 13886 + }, + { + "epoch": 5.13, + "learning_rate": 2.4887178223919205e-06, + "loss": 0.5847, + "step": 13887 + }, + { + "epoch": 5.13, + "learning_rate": 2.488680872028575e-06, + "loss": 0.4841, + "step": 13888 + }, + { + "epoch": 5.13, + "learning_rate": 2.4886439216652298e-06, + "loss": 0.5606, + "step": 13889 + }, + { + "epoch": 5.13, + "learning_rate": 2.4886069713018844e-06, + "loss": 0.4893, + "step": 13890 + }, + { + "epoch": 5.13, + "learning_rate": 2.488570020938539e-06, + "loss": 0.5299, + "step": 13891 + }, + { + "epoch": 5.13, + "learning_rate": 2.488533070575194e-06, + "loss": 0.4663, + "step": 13892 + }, + { + "epoch": 5.13, + "learning_rate": 2.4884961202118487e-06, + "loss": 0.4554, + "step": 13893 + }, + { + "epoch": 5.13, + "learning_rate": 2.488459169848504e-06, + "loss": 0.6153, + "step": 13894 + }, + { + "epoch": 5.13, + "learning_rate": 2.4884222194851584e-06, + "loss": 0.4115, + "step": 13895 + }, + { + "epoch": 5.13, + "learning_rate": 2.488385269121813e-06, + "loss": 0.5688, + "step": 13896 + }, + { + "epoch": 5.13, + "learning_rate": 2.4883483187584677e-06, + "loss": 0.3654, + "step": 13897 + }, + { + "epoch": 5.13, + "learning_rate": 2.4883113683951228e-06, + "loss": 0.5044, + "step": 13898 + }, + { + "epoch": 5.13, + "learning_rate": 2.4882744180317774e-06, + "loss": 0.5527, + "step": 13899 + }, + { + "epoch": 5.13, + "learning_rate": 2.4882374676684325e-06, + "loss": 0.427, + "step": 13900 + }, + { + "epoch": 5.13, + "learning_rate": 2.488200517305087e-06, + "loss": 0.377, + "step": 13901 + }, + { + "epoch": 5.13, + "learning_rate": 2.4881635669417413e-06, + "loss": 0.4658, + "step": 13902 + }, + { + "epoch": 5.13, + "learning_rate": 2.4881266165783964e-06, + "loss": 0.5191, + "step": 13903 + }, + { + "epoch": 5.13, + "learning_rate": 2.488089666215051e-06, + "loss": 0.5329, + "step": 13904 + }, + { + "epoch": 5.13, + "learning_rate": 2.488052715851706e-06, + "loss": 0.5623, + "step": 13905 + }, + { + "epoch": 5.14, + "learning_rate": 2.4880157654883607e-06, + "loss": 0.3503, + "step": 13906 + }, + { + "epoch": 5.14, + "learning_rate": 2.4879788151250158e-06, + "loss": 0.5005, + "step": 13907 + }, + { + "epoch": 5.14, + "learning_rate": 2.48794186476167e-06, + "loss": 0.4222, + "step": 13908 + }, + { + "epoch": 5.14, + "learning_rate": 2.487904914398325e-06, + "loss": 0.4932, + "step": 13909 + }, + { + "epoch": 5.14, + "learning_rate": 2.4878679640349797e-06, + "loss": 0.5068, + "step": 13910 + }, + { + "epoch": 5.14, + "learning_rate": 2.4878310136716343e-06, + "loss": 0.5997, + "step": 13911 + }, + { + "epoch": 5.14, + "learning_rate": 2.4877940633082894e-06, + "loss": 0.479, + "step": 13912 + }, + { + "epoch": 5.14, + "learning_rate": 2.487757112944944e-06, + "loss": 0.4398, + "step": 13913 + }, + { + "epoch": 5.14, + "learning_rate": 2.4877201625815987e-06, + "loss": 0.4649, + "step": 13914 + }, + { + "epoch": 5.14, + "learning_rate": 2.4876832122182533e-06, + "loss": 0.4989, + "step": 13915 + }, + { + "epoch": 5.14, + "learning_rate": 2.4876462618549083e-06, + "loss": 0.4974, + "step": 13916 + }, + { + "epoch": 5.14, + "learning_rate": 2.487609311491563e-06, + "loss": 0.5613, + "step": 13917 + }, + { + "epoch": 5.14, + "learning_rate": 2.487572361128218e-06, + "loss": 0.4864, + "step": 13918 + }, + { + "epoch": 5.14, + "learning_rate": 2.4875354107648727e-06, + "loss": 0.47, + "step": 13919 + }, + { + "epoch": 5.14, + "learning_rate": 2.4874984604015273e-06, + "loss": 0.5419, + "step": 13920 + }, + { + "epoch": 5.14, + "learning_rate": 2.487461510038182e-06, + "loss": 0.5253, + "step": 13921 + }, + { + "epoch": 5.14, + "learning_rate": 2.4874245596748366e-06, + "loss": 0.4578, + "step": 13922 + }, + { + "epoch": 5.14, + "learning_rate": 2.4873876093114917e-06, + "loss": 0.3773, + "step": 13923 + }, + { + "epoch": 5.14, + "learning_rate": 2.4873506589481463e-06, + "loss": 0.4466, + "step": 13924 + }, + { + "epoch": 5.14, + "learning_rate": 2.4873137085848013e-06, + "loss": 0.4865, + "step": 13925 + }, + { + "epoch": 5.14, + "learning_rate": 2.487276758221456e-06, + "loss": 0.3808, + "step": 13926 + }, + { + "epoch": 5.14, + "learning_rate": 2.4872398078581106e-06, + "loss": 0.4037, + "step": 13927 + }, + { + "epoch": 5.14, + "learning_rate": 2.4872028574947653e-06, + "loss": 0.4408, + "step": 13928 + }, + { + "epoch": 5.14, + "learning_rate": 2.4871659071314203e-06, + "loss": 0.5002, + "step": 13929 + }, + { + "epoch": 5.14, + "learning_rate": 2.487128956768075e-06, + "loss": 0.4419, + "step": 13930 + }, + { + "epoch": 5.14, + "learning_rate": 2.4870920064047296e-06, + "loss": 0.5034, + "step": 13931 + }, + { + "epoch": 5.14, + "learning_rate": 2.4870550560413846e-06, + "loss": 0.444, + "step": 13932 + }, + { + "epoch": 5.15, + "learning_rate": 2.4870181056780393e-06, + "loss": 0.5485, + "step": 13933 + }, + { + "epoch": 5.15, + "learning_rate": 2.486981155314694e-06, + "loss": 0.5177, + "step": 13934 + }, + { + "epoch": 5.15, + "learning_rate": 2.4869442049513486e-06, + "loss": 0.4493, + "step": 13935 + }, + { + "epoch": 5.15, + "learning_rate": 2.4869072545880036e-06, + "loss": 0.5577, + "step": 13936 + }, + { + "epoch": 5.15, + "learning_rate": 2.4868703042246583e-06, + "loss": 0.5085, + "step": 13937 + }, + { + "epoch": 5.15, + "learning_rate": 2.4868333538613133e-06, + "loss": 0.4874, + "step": 13938 + }, + { + "epoch": 5.15, + "learning_rate": 2.486796403497968e-06, + "loss": 0.5665, + "step": 13939 + }, + { + "epoch": 5.15, + "learning_rate": 2.4867594531346226e-06, + "loss": 0.4976, + "step": 13940 + }, + { + "epoch": 5.15, + "learning_rate": 2.4867225027712772e-06, + "loss": 0.544, + "step": 13941 + }, + { + "epoch": 5.15, + "learning_rate": 2.486685552407932e-06, + "loss": 0.4326, + "step": 13942 + }, + { + "epoch": 5.15, + "learning_rate": 2.486648602044587e-06, + "loss": 0.5639, + "step": 13943 + }, + { + "epoch": 5.15, + "learning_rate": 2.4866116516812416e-06, + "loss": 0.4192, + "step": 13944 + }, + { + "epoch": 5.15, + "learning_rate": 2.4865747013178966e-06, + "loss": 0.5736, + "step": 13945 + }, + { + "epoch": 5.15, + "learning_rate": 2.4865377509545513e-06, + "loss": 0.4539, + "step": 13946 + }, + { + "epoch": 5.15, + "learning_rate": 2.486500800591206e-06, + "loss": 0.3682, + "step": 13947 + }, + { + "epoch": 5.15, + "learning_rate": 2.4864638502278605e-06, + "loss": 0.4039, + "step": 13948 + }, + { + "epoch": 5.15, + "learning_rate": 2.4864268998645156e-06, + "loss": 0.4413, + "step": 13949 + }, + { + "epoch": 5.15, + "learning_rate": 2.4863899495011702e-06, + "loss": 0.4152, + "step": 13950 + }, + { + "epoch": 5.15, + "learning_rate": 2.486352999137825e-06, + "loss": 0.5131, + "step": 13951 + }, + { + "epoch": 5.15, + "learning_rate": 2.48631604877448e-06, + "loss": 0.4601, + "step": 13952 + }, + { + "epoch": 5.15, + "learning_rate": 2.486279098411134e-06, + "loss": 0.5127, + "step": 13953 + }, + { + "epoch": 5.15, + "learning_rate": 2.486242148047789e-06, + "loss": 0.4357, + "step": 13954 + }, + { + "epoch": 5.15, + "learning_rate": 2.486205197684444e-06, + "loss": 0.4775, + "step": 13955 + }, + { + "epoch": 5.15, + "learning_rate": 2.486168247321099e-06, + "loss": 0.4981, + "step": 13956 + }, + { + "epoch": 5.15, + "learning_rate": 2.4861312969577535e-06, + "loss": 0.4815, + "step": 13957 + }, + { + "epoch": 5.15, + "learning_rate": 2.4860943465944086e-06, + "loss": 0.3837, + "step": 13958 + }, + { + "epoch": 5.15, + "learning_rate": 2.486057396231063e-06, + "loss": 0.5535, + "step": 13959 + }, + { + "epoch": 5.16, + "learning_rate": 2.486020445867718e-06, + "loss": 0.3882, + "step": 13960 + }, + { + "epoch": 5.16, + "learning_rate": 2.4859834955043725e-06, + "loss": 0.372, + "step": 13961 + }, + { + "epoch": 5.16, + "learning_rate": 2.485946545141027e-06, + "loss": 0.5362, + "step": 13962 + }, + { + "epoch": 5.16, + "learning_rate": 2.485909594777682e-06, + "loss": 0.5494, + "step": 13963 + }, + { + "epoch": 5.16, + "learning_rate": 2.485872644414337e-06, + "loss": 0.4617, + "step": 13964 + }, + { + "epoch": 5.16, + "learning_rate": 2.4858356940509915e-06, + "loss": 0.5547, + "step": 13965 + }, + { + "epoch": 5.16, + "learning_rate": 2.485798743687646e-06, + "loss": 0.4158, + "step": 13966 + }, + { + "epoch": 5.16, + "learning_rate": 2.485761793324301e-06, + "loss": 0.4266, + "step": 13967 + }, + { + "epoch": 5.16, + "learning_rate": 2.485724842960956e-06, + "loss": 0.4117, + "step": 13968 + }, + { + "epoch": 5.16, + "learning_rate": 2.485687892597611e-06, + "loss": 0.5689, + "step": 13969 + }, + { + "epoch": 5.16, + "learning_rate": 2.4856509422342655e-06, + "loss": 0.533, + "step": 13970 + }, + { + "epoch": 5.16, + "learning_rate": 2.48561399187092e-06, + "loss": 0.4768, + "step": 13971 + }, + { + "epoch": 5.16, + "learning_rate": 2.4855770415075748e-06, + "loss": 0.5582, + "step": 13972 + }, + { + "epoch": 5.16, + "learning_rate": 2.4855400911442294e-06, + "loss": 0.4095, + "step": 13973 + }, + { + "epoch": 5.16, + "learning_rate": 2.4855031407808845e-06, + "loss": 0.5729, + "step": 13974 + }, + { + "epoch": 5.16, + "learning_rate": 2.485466190417539e-06, + "loss": 0.5209, + "step": 13975 + }, + { + "epoch": 5.16, + "learning_rate": 2.485429240054194e-06, + "loss": 0.4573, + "step": 13976 + }, + { + "epoch": 5.16, + "learning_rate": 2.485392289690849e-06, + "loss": 0.4579, + "step": 13977 + }, + { + "epoch": 5.16, + "learning_rate": 2.4853553393275034e-06, + "loss": 0.548, + "step": 13978 + }, + { + "epoch": 5.16, + "learning_rate": 2.485318388964158e-06, + "loss": 0.4604, + "step": 13979 + }, + { + "epoch": 5.16, + "learning_rate": 2.4852814386008127e-06, + "loss": 0.4732, + "step": 13980 + }, + { + "epoch": 5.16, + "learning_rate": 2.4852444882374678e-06, + "loss": 0.5344, + "step": 13981 + }, + { + "epoch": 5.16, + "learning_rate": 2.4852075378741224e-06, + "loss": 0.5774, + "step": 13982 + }, + { + "epoch": 5.16, + "learning_rate": 2.4851705875107775e-06, + "loss": 0.4808, + "step": 13983 + }, + { + "epoch": 5.16, + "learning_rate": 2.485133637147432e-06, + "loss": 0.4821, + "step": 13984 + }, + { + "epoch": 5.16, + "learning_rate": 2.4850966867840867e-06, + "loss": 0.4952, + "step": 13985 + }, + { + "epoch": 5.16, + "learning_rate": 2.4850597364207414e-06, + "loss": 0.3962, + "step": 13986 + }, + { + "epoch": 5.17, + "learning_rate": 2.4850227860573964e-06, + "loss": 0.5079, + "step": 13987 + }, + { + "epoch": 5.17, + "learning_rate": 2.484985835694051e-06, + "loss": 0.3985, + "step": 13988 + }, + { + "epoch": 5.17, + "learning_rate": 2.484948885330706e-06, + "loss": 0.4873, + "step": 13989 + }, + { + "epoch": 5.17, + "learning_rate": 2.4849119349673608e-06, + "loss": 0.5323, + "step": 13990 + }, + { + "epoch": 5.17, + "learning_rate": 2.4848749846040154e-06, + "loss": 0.5027, + "step": 13991 + }, + { + "epoch": 5.17, + "learning_rate": 2.48483803424067e-06, + "loss": 0.5576, + "step": 13992 + }, + { + "epoch": 5.17, + "learning_rate": 2.4848010838773247e-06, + "loss": 0.5715, + "step": 13993 + }, + { + "epoch": 5.17, + "learning_rate": 2.4847641335139797e-06, + "loss": 0.4222, + "step": 13994 + }, + { + "epoch": 5.17, + "learning_rate": 2.4847271831506344e-06, + "loss": 0.5162, + "step": 13995 + }, + { + "epoch": 5.17, + "learning_rate": 2.4846902327872894e-06, + "loss": 0.5714, + "step": 13996 + }, + { + "epoch": 5.17, + "learning_rate": 2.4846532824239436e-06, + "loss": 0.4446, + "step": 13997 + }, + { + "epoch": 5.17, + "learning_rate": 2.4846163320605987e-06, + "loss": 0.5531, + "step": 13998 + }, + { + "epoch": 5.17, + "learning_rate": 2.4845793816972533e-06, + "loss": 0.5396, + "step": 13999 + }, + { + "epoch": 5.17, + "learning_rate": 2.484542431333908e-06, + "loss": 0.5841, + "step": 14000 + }, + { + "epoch": 5.17, + "learning_rate": 2.484505480970563e-06, + "loss": 0.5142, + "step": 14001 + }, + { + "epoch": 5.17, + "learning_rate": 2.4844685306072177e-06, + "loss": 0.5714, + "step": 14002 + }, + { + "epoch": 5.17, + "learning_rate": 2.4844315802438723e-06, + "loss": 0.4232, + "step": 14003 + }, + { + "epoch": 5.17, + "learning_rate": 2.484394629880527e-06, + "loss": 0.556, + "step": 14004 + }, + { + "epoch": 5.17, + "learning_rate": 2.484357679517182e-06, + "loss": 0.4563, + "step": 14005 + }, + { + "epoch": 5.17, + "learning_rate": 2.4843207291538366e-06, + "loss": 0.5121, + "step": 14006 + }, + { + "epoch": 5.17, + "learning_rate": 2.4842837787904917e-06, + "loss": 0.5683, + "step": 14007 + }, + { + "epoch": 5.17, + "learning_rate": 2.4842468284271463e-06, + "loss": 0.5689, + "step": 14008 + }, + { + "epoch": 5.17, + "learning_rate": 2.484209878063801e-06, + "loss": 0.5208, + "step": 14009 + }, + { + "epoch": 5.17, + "learning_rate": 2.4841729277004556e-06, + "loss": 0.5388, + "step": 14010 + }, + { + "epoch": 5.17, + "learning_rate": 2.4841359773371102e-06, + "loss": 0.508, + "step": 14011 + }, + { + "epoch": 5.17, + "learning_rate": 2.4840990269737653e-06, + "loss": 0.5639, + "step": 14012 + }, + { + "epoch": 5.17, + "learning_rate": 2.48406207661042e-06, + "loss": 0.372, + "step": 14013 + }, + { + "epoch": 5.18, + "learning_rate": 2.484025126247075e-06, + "loss": 0.5508, + "step": 14014 + }, + { + "epoch": 5.18, + "learning_rate": 2.4839881758837296e-06, + "loss": 0.5037, + "step": 14015 + }, + { + "epoch": 5.18, + "learning_rate": 2.4839512255203843e-06, + "loss": 0.515, + "step": 14016 + }, + { + "epoch": 5.18, + "learning_rate": 2.483914275157039e-06, + "loss": 0.5594, + "step": 14017 + }, + { + "epoch": 5.18, + "learning_rate": 2.483877324793694e-06, + "loss": 0.5241, + "step": 14018 + }, + { + "epoch": 5.18, + "learning_rate": 2.4838403744303486e-06, + "loss": 0.4988, + "step": 14019 + }, + { + "epoch": 5.18, + "learning_rate": 2.4838034240670032e-06, + "loss": 0.4657, + "step": 14020 + }, + { + "epoch": 5.18, + "learning_rate": 2.4837664737036583e-06, + "loss": 0.4627, + "step": 14021 + }, + { + "epoch": 5.18, + "learning_rate": 2.483729523340313e-06, + "loss": 0.3918, + "step": 14022 + }, + { + "epoch": 5.18, + "learning_rate": 2.4836925729769676e-06, + "loss": 0.5547, + "step": 14023 + }, + { + "epoch": 5.18, + "learning_rate": 2.4836556226136222e-06, + "loss": 0.5066, + "step": 14024 + }, + { + "epoch": 5.18, + "learning_rate": 2.4836186722502773e-06, + "loss": 0.5734, + "step": 14025 + }, + { + "epoch": 5.18, + "learning_rate": 2.483581721886932e-06, + "loss": 0.436, + "step": 14026 + }, + { + "epoch": 5.18, + "learning_rate": 2.483544771523587e-06, + "loss": 0.4702, + "step": 14027 + }, + { + "epoch": 5.18, + "learning_rate": 2.4835078211602416e-06, + "loss": 0.4796, + "step": 14028 + }, + { + "epoch": 5.18, + "learning_rate": 2.4834708707968962e-06, + "loss": 0.4026, + "step": 14029 + }, + { + "epoch": 5.18, + "learning_rate": 2.483433920433551e-06, + "loss": 0.5077, + "step": 14030 + }, + { + "epoch": 5.18, + "learning_rate": 2.4833969700702055e-06, + "loss": 0.592, + "step": 14031 + }, + { + "epoch": 5.18, + "learning_rate": 2.4833600197068606e-06, + "loss": 0.5998, + "step": 14032 + }, + { + "epoch": 5.18, + "learning_rate": 2.483323069343515e-06, + "loss": 0.48, + "step": 14033 + }, + { + "epoch": 5.18, + "learning_rate": 2.4832861189801703e-06, + "loss": 0.4332, + "step": 14034 + }, + { + "epoch": 5.18, + "learning_rate": 2.483249168616825e-06, + "loss": 0.4695, + "step": 14035 + }, + { + "epoch": 5.18, + "learning_rate": 2.4832122182534795e-06, + "loss": 0.4354, + "step": 14036 + }, + { + "epoch": 5.18, + "learning_rate": 2.483175267890134e-06, + "loss": 0.4746, + "step": 14037 + }, + { + "epoch": 5.18, + "learning_rate": 2.4831383175267892e-06, + "loss": 0.5591, + "step": 14038 + }, + { + "epoch": 5.18, + "learning_rate": 2.483101367163444e-06, + "loss": 0.3541, + "step": 14039 + }, + { + "epoch": 5.18, + "learning_rate": 2.4830644168000985e-06, + "loss": 0.4709, + "step": 14040 + }, + { + "epoch": 5.19, + "learning_rate": 2.4830274664367536e-06, + "loss": 0.4648, + "step": 14041 + }, + { + "epoch": 5.19, + "learning_rate": 2.4829905160734078e-06, + "loss": 0.5082, + "step": 14042 + }, + { + "epoch": 5.19, + "learning_rate": 2.482953565710063e-06, + "loss": 0.4954, + "step": 14043 + }, + { + "epoch": 5.19, + "learning_rate": 2.4829166153467175e-06, + "loss": 0.4189, + "step": 14044 + }, + { + "epoch": 5.19, + "learning_rate": 2.4828796649833725e-06, + "loss": 0.4877, + "step": 14045 + }, + { + "epoch": 5.19, + "learning_rate": 2.482842714620027e-06, + "loss": 0.4549, + "step": 14046 + }, + { + "epoch": 5.19, + "learning_rate": 2.4828057642566822e-06, + "loss": 0.5826, + "step": 14047 + }, + { + "epoch": 5.19, + "learning_rate": 2.4827688138933365e-06, + "loss": 0.5695, + "step": 14048 + }, + { + "epoch": 5.19, + "learning_rate": 2.4827318635299915e-06, + "loss": 0.4885, + "step": 14049 + }, + { + "epoch": 5.19, + "learning_rate": 2.482694913166646e-06, + "loss": 0.4956, + "step": 14050 + }, + { + "epoch": 5.19, + "learning_rate": 2.4826579628033008e-06, + "loss": 0.5554, + "step": 14051 + }, + { + "epoch": 5.19, + "learning_rate": 2.482621012439956e-06, + "loss": 0.61, + "step": 14052 + }, + { + "epoch": 5.19, + "learning_rate": 2.4825840620766105e-06, + "loss": 0.6091, + "step": 14053 + }, + { + "epoch": 5.19, + "learning_rate": 2.482547111713265e-06, + "loss": 0.4902, + "step": 14054 + }, + { + "epoch": 5.19, + "learning_rate": 2.4825101613499198e-06, + "loss": 0.4574, + "step": 14055 + }, + { + "epoch": 5.19, + "learning_rate": 2.482473210986575e-06, + "loss": 0.4296, + "step": 14056 + }, + { + "epoch": 5.19, + "learning_rate": 2.4824362606232295e-06, + "loss": 0.4918, + "step": 14057 + }, + { + "epoch": 5.19, + "learning_rate": 2.4823993102598845e-06, + "loss": 0.5644, + "step": 14058 + }, + { + "epoch": 5.19, + "learning_rate": 2.482362359896539e-06, + "loss": 0.5257, + "step": 14059 + }, + { + "epoch": 5.19, + "learning_rate": 2.4823254095331938e-06, + "loss": 0.5734, + "step": 14060 + }, + { + "epoch": 5.19, + "learning_rate": 2.4822884591698484e-06, + "loss": 0.5234, + "step": 14061 + }, + { + "epoch": 5.19, + "learning_rate": 2.482251508806503e-06, + "loss": 0.4976, + "step": 14062 + }, + { + "epoch": 5.19, + "learning_rate": 2.482214558443158e-06, + "loss": 0.4784, + "step": 14063 + }, + { + "epoch": 5.19, + "learning_rate": 2.4821776080798128e-06, + "loss": 0.4996, + "step": 14064 + }, + { + "epoch": 5.19, + "learning_rate": 2.482140657716468e-06, + "loss": 0.4025, + "step": 14065 + }, + { + "epoch": 5.19, + "learning_rate": 2.4821037073531224e-06, + "loss": 0.528, + "step": 14066 + }, + { + "epoch": 5.19, + "learning_rate": 2.482066756989777e-06, + "loss": 0.3642, + "step": 14067 + }, + { + "epoch": 5.19, + "learning_rate": 2.4820298066264317e-06, + "loss": 0.4891, + "step": 14068 + }, + { + "epoch": 5.2, + "learning_rate": 2.4819928562630868e-06, + "loss": 0.4944, + "step": 14069 + }, + { + "epoch": 5.2, + "learning_rate": 2.4819559058997414e-06, + "loss": 0.5672, + "step": 14070 + }, + { + "epoch": 5.2, + "learning_rate": 2.481918955536396e-06, + "loss": 0.3487, + "step": 14071 + }, + { + "epoch": 5.2, + "learning_rate": 2.481882005173051e-06, + "loss": 0.5314, + "step": 14072 + }, + { + "epoch": 5.2, + "learning_rate": 2.4818450548097058e-06, + "loss": 0.3692, + "step": 14073 + }, + { + "epoch": 5.2, + "learning_rate": 2.4818081044463604e-06, + "loss": 0.4478, + "step": 14074 + }, + { + "epoch": 5.2, + "learning_rate": 2.481771154083015e-06, + "loss": 0.4487, + "step": 14075 + }, + { + "epoch": 5.2, + "learning_rate": 2.48173420371967e-06, + "loss": 0.4437, + "step": 14076 + }, + { + "epoch": 5.2, + "learning_rate": 2.4816972533563247e-06, + "loss": 0.4675, + "step": 14077 + }, + { + "epoch": 5.2, + "learning_rate": 2.4816603029929798e-06, + "loss": 0.4966, + "step": 14078 + }, + { + "epoch": 5.2, + "learning_rate": 2.4816233526296344e-06, + "loss": 0.4332, + "step": 14079 + }, + { + "epoch": 5.2, + "learning_rate": 2.481586402266289e-06, + "loss": 0.651, + "step": 14080 + }, + { + "epoch": 5.2, + "learning_rate": 2.4815494519029437e-06, + "loss": 0.5223, + "step": 14081 + }, + { + "epoch": 5.2, + "learning_rate": 2.4815125015395983e-06, + "loss": 0.5151, + "step": 14082 + }, + { + "epoch": 5.2, + "learning_rate": 2.4814755511762534e-06, + "loss": 0.4407, + "step": 14083 + }, + { + "epoch": 5.2, + "learning_rate": 2.481438600812908e-06, + "loss": 0.473, + "step": 14084 + }, + { + "epoch": 5.2, + "learning_rate": 2.481401650449563e-06, + "loss": 0.4316, + "step": 14085 + }, + { + "epoch": 5.2, + "learning_rate": 2.4813647000862177e-06, + "loss": 0.4448, + "step": 14086 + }, + { + "epoch": 5.2, + "learning_rate": 2.4813277497228724e-06, + "loss": 0.4825, + "step": 14087 + }, + { + "epoch": 5.2, + "learning_rate": 2.481290799359527e-06, + "loss": 0.464, + "step": 14088 + }, + { + "epoch": 5.2, + "learning_rate": 2.481253848996182e-06, + "loss": 0.5362, + "step": 14089 + }, + { + "epoch": 5.2, + "learning_rate": 2.4812168986328367e-06, + "loss": 0.4803, + "step": 14090 + }, + { + "epoch": 5.2, + "learning_rate": 2.4811799482694913e-06, + "loss": 0.6307, + "step": 14091 + }, + { + "epoch": 5.2, + "learning_rate": 2.4811429979061464e-06, + "loss": 0.5117, + "step": 14092 + }, + { + "epoch": 5.2, + "learning_rate": 2.4811060475428006e-06, + "loss": 0.4442, + "step": 14093 + }, + { + "epoch": 5.2, + "learning_rate": 2.4810690971794557e-06, + "loss": 0.4105, + "step": 14094 + }, + { + "epoch": 5.2, + "learning_rate": 2.4810321468161103e-06, + "loss": 0.4917, + "step": 14095 + }, + { + "epoch": 5.21, + "learning_rate": 2.4809951964527654e-06, + "loss": 0.4196, + "step": 14096 + }, + { + "epoch": 5.21, + "learning_rate": 2.48095824608942e-06, + "loss": 0.4959, + "step": 14097 + }, + { + "epoch": 5.21, + "learning_rate": 2.480921295726075e-06, + "loss": 0.4617, + "step": 14098 + }, + { + "epoch": 5.21, + "learning_rate": 2.4808843453627293e-06, + "loss": 0.4327, + "step": 14099 + }, + { + "epoch": 5.21, + "learning_rate": 2.480847394999384e-06, + "loss": 0.5841, + "step": 14100 + }, + { + "epoch": 5.21, + "learning_rate": 2.480810444636039e-06, + "loss": 0.5878, + "step": 14101 + }, + { + "epoch": 5.21, + "learning_rate": 2.4807734942726936e-06, + "loss": 0.5653, + "step": 14102 + }, + { + "epoch": 5.21, + "learning_rate": 2.4807365439093487e-06, + "loss": 0.4001, + "step": 14103 + }, + { + "epoch": 5.21, + "learning_rate": 2.4806995935460033e-06, + "loss": 0.4248, + "step": 14104 + }, + { + "epoch": 5.21, + "learning_rate": 2.480662643182658e-06, + "loss": 0.4824, + "step": 14105 + }, + { + "epoch": 5.21, + "learning_rate": 2.4806256928193126e-06, + "loss": 0.5736, + "step": 14106 + }, + { + "epoch": 5.21, + "learning_rate": 2.4805887424559676e-06, + "loss": 0.4303, + "step": 14107 + }, + { + "epoch": 5.21, + "learning_rate": 2.4805517920926223e-06, + "loss": 0.4554, + "step": 14108 + }, + { + "epoch": 5.21, + "learning_rate": 2.4805148417292773e-06, + "loss": 0.5818, + "step": 14109 + }, + { + "epoch": 5.21, + "learning_rate": 2.480477891365932e-06, + "loss": 0.6054, + "step": 14110 + }, + { + "epoch": 5.21, + "learning_rate": 2.4804409410025866e-06, + "loss": 0.4877, + "step": 14111 + }, + { + "epoch": 5.21, + "learning_rate": 2.4804039906392412e-06, + "loss": 0.5094, + "step": 14112 + }, + { + "epoch": 5.21, + "learning_rate": 2.480367040275896e-06, + "loss": 0.4701, + "step": 14113 + }, + { + "epoch": 5.21, + "learning_rate": 2.480330089912551e-06, + "loss": 0.4136, + "step": 14114 + }, + { + "epoch": 5.21, + "learning_rate": 2.4802931395492056e-06, + "loss": 0.5228, + "step": 14115 + }, + { + "epoch": 5.21, + "learning_rate": 2.4802561891858606e-06, + "loss": 0.5688, + "step": 14116 + }, + { + "epoch": 5.21, + "learning_rate": 2.4802192388225153e-06, + "loss": 0.5135, + "step": 14117 + }, + { + "epoch": 5.21, + "learning_rate": 2.48018228845917e-06, + "loss": 0.4566, + "step": 14118 + }, + { + "epoch": 5.21, + "learning_rate": 2.4801453380958245e-06, + "loss": 0.5499, + "step": 14119 + }, + { + "epoch": 5.21, + "learning_rate": 2.480108387732479e-06, + "loss": 0.4188, + "step": 14120 + }, + { + "epoch": 5.21, + "learning_rate": 2.4800714373691342e-06, + "loss": 0.4349, + "step": 14121 + }, + { + "epoch": 5.21, + "learning_rate": 2.480034487005789e-06, + "loss": 0.4584, + "step": 14122 + }, + { + "epoch": 5.22, + "learning_rate": 2.479997536642444e-06, + "loss": 0.4696, + "step": 14123 + }, + { + "epoch": 5.22, + "learning_rate": 2.4799605862790986e-06, + "loss": 0.4481, + "step": 14124 + }, + { + "epoch": 5.22, + "learning_rate": 2.479923635915753e-06, + "loss": 0.5631, + "step": 14125 + }, + { + "epoch": 5.22, + "learning_rate": 2.479886685552408e-06, + "loss": 0.4059, + "step": 14126 + }, + { + "epoch": 5.22, + "learning_rate": 2.479849735189063e-06, + "loss": 0.5335, + "step": 14127 + }, + { + "epoch": 5.22, + "learning_rate": 2.4798127848257175e-06, + "loss": 0.5795, + "step": 14128 + }, + { + "epoch": 5.22, + "learning_rate": 2.4797758344623726e-06, + "loss": 0.3943, + "step": 14129 + }, + { + "epoch": 5.22, + "learning_rate": 2.4797388840990272e-06, + "loss": 0.5944, + "step": 14130 + }, + { + "epoch": 5.22, + "learning_rate": 2.479701933735682e-06, + "loss": 0.5043, + "step": 14131 + }, + { + "epoch": 5.22, + "learning_rate": 2.4796649833723365e-06, + "loss": 0.4507, + "step": 14132 + }, + { + "epoch": 5.22, + "learning_rate": 2.479628033008991e-06, + "loss": 0.5393, + "step": 14133 + }, + { + "epoch": 5.22, + "learning_rate": 2.479591082645646e-06, + "loss": 0.5106, + "step": 14134 + }, + { + "epoch": 5.22, + "learning_rate": 2.479554132282301e-06, + "loss": 0.4851, + "step": 14135 + }, + { + "epoch": 5.22, + "learning_rate": 2.479517181918956e-06, + "loss": 0.5721, + "step": 14136 + }, + { + "epoch": 5.22, + "learning_rate": 2.4794802315556105e-06, + "loss": 0.4423, + "step": 14137 + }, + { + "epoch": 5.22, + "learning_rate": 2.479443281192265e-06, + "loss": 0.5224, + "step": 14138 + }, + { + "epoch": 5.22, + "learning_rate": 2.47940633082892e-06, + "loss": 0.5099, + "step": 14139 + }, + { + "epoch": 5.22, + "learning_rate": 2.4793693804655744e-06, + "loss": 0.5956, + "step": 14140 + }, + { + "epoch": 5.22, + "learning_rate": 2.4793324301022295e-06, + "loss": 0.3914, + "step": 14141 + }, + { + "epoch": 5.22, + "learning_rate": 2.479295479738884e-06, + "loss": 0.5228, + "step": 14142 + }, + { + "epoch": 5.22, + "learning_rate": 2.479258529375539e-06, + "loss": 0.4534, + "step": 14143 + }, + { + "epoch": 5.22, + "learning_rate": 2.4792215790121934e-06, + "loss": 0.4566, + "step": 14144 + }, + { + "epoch": 5.22, + "learning_rate": 2.4791846286488485e-06, + "loss": 0.4474, + "step": 14145 + }, + { + "epoch": 5.22, + "learning_rate": 2.479147678285503e-06, + "loss": 0.4986, + "step": 14146 + }, + { + "epoch": 5.22, + "learning_rate": 2.479110727922158e-06, + "loss": 0.4878, + "step": 14147 + }, + { + "epoch": 5.22, + "learning_rate": 2.479073777558813e-06, + "loss": 0.5827, + "step": 14148 + }, + { + "epoch": 5.22, + "learning_rate": 2.4790368271954674e-06, + "loss": 0.4229, + "step": 14149 + }, + { + "epoch": 5.23, + "learning_rate": 2.478999876832122e-06, + "loss": 0.5496, + "step": 14150 + }, + { + "epoch": 5.23, + "learning_rate": 2.4789629264687767e-06, + "loss": 0.4508, + "step": 14151 + }, + { + "epoch": 5.23, + "learning_rate": 2.4789259761054318e-06, + "loss": 0.4855, + "step": 14152 + }, + { + "epoch": 5.23, + "learning_rate": 2.4788890257420864e-06, + "loss": 0.4782, + "step": 14153 + }, + { + "epoch": 5.23, + "learning_rate": 2.4788520753787415e-06, + "loss": 0.4325, + "step": 14154 + }, + { + "epoch": 5.23, + "learning_rate": 2.478815125015396e-06, + "loss": 0.4899, + "step": 14155 + }, + { + "epoch": 5.23, + "learning_rate": 2.4787781746520507e-06, + "loss": 0.5248, + "step": 14156 + }, + { + "epoch": 5.23, + "learning_rate": 2.4787412242887054e-06, + "loss": 0.4421, + "step": 14157 + }, + { + "epoch": 5.23, + "learning_rate": 2.4787042739253604e-06, + "loss": 0.457, + "step": 14158 + }, + { + "epoch": 5.23, + "learning_rate": 2.478667323562015e-06, + "loss": 0.4592, + "step": 14159 + }, + { + "epoch": 5.23, + "learning_rate": 2.4786303731986697e-06, + "loss": 0.4734, + "step": 14160 + }, + { + "epoch": 5.23, + "learning_rate": 2.4785934228353248e-06, + "loss": 0.3739, + "step": 14161 + }, + { + "epoch": 5.23, + "learning_rate": 2.4785564724719794e-06, + "loss": 0.5207, + "step": 14162 + }, + { + "epoch": 5.23, + "learning_rate": 2.478519522108634e-06, + "loss": 0.515, + "step": 14163 + }, + { + "epoch": 5.23, + "learning_rate": 2.4784825717452887e-06, + "loss": 0.49, + "step": 14164 + }, + { + "epoch": 5.23, + "learning_rate": 2.4784456213819437e-06, + "loss": 0.4567, + "step": 14165 + }, + { + "epoch": 5.23, + "learning_rate": 2.4784086710185984e-06, + "loss": 0.4693, + "step": 14166 + }, + { + "epoch": 5.23, + "learning_rate": 2.4783717206552534e-06, + "loss": 0.5726, + "step": 14167 + }, + { + "epoch": 5.23, + "learning_rate": 2.478334770291908e-06, + "loss": 0.5797, + "step": 14168 + }, + { + "epoch": 5.23, + "learning_rate": 2.4782978199285627e-06, + "loss": 0.5555, + "step": 14169 + }, + { + "epoch": 5.23, + "learning_rate": 2.4782608695652173e-06, + "loss": 0.4241, + "step": 14170 + }, + { + "epoch": 5.23, + "learning_rate": 2.478223919201872e-06, + "loss": 0.4708, + "step": 14171 + }, + { + "epoch": 5.23, + "learning_rate": 2.478186968838527e-06, + "loss": 0.5092, + "step": 14172 + }, + { + "epoch": 5.23, + "learning_rate": 2.4781500184751817e-06, + "loss": 0.3252, + "step": 14173 + }, + { + "epoch": 5.23, + "learning_rate": 2.4781130681118367e-06, + "loss": 0.5152, + "step": 14174 + }, + { + "epoch": 5.23, + "learning_rate": 2.4780761177484914e-06, + "loss": 0.4526, + "step": 14175 + }, + { + "epoch": 5.23, + "learning_rate": 2.478039167385146e-06, + "loss": 0.5783, + "step": 14176 + }, + { + "epoch": 5.24, + "learning_rate": 2.4780022170218006e-06, + "loss": 0.4583, + "step": 14177 + }, + { + "epoch": 5.24, + "learning_rate": 2.4779652666584557e-06, + "loss": 0.4194, + "step": 14178 + }, + { + "epoch": 5.24, + "learning_rate": 2.4779283162951103e-06, + "loss": 0.4703, + "step": 14179 + }, + { + "epoch": 5.24, + "learning_rate": 2.477891365931765e-06, + "loss": 0.4932, + "step": 14180 + }, + { + "epoch": 5.24, + "learning_rate": 2.47785441556842e-06, + "loss": 0.5146, + "step": 14181 + }, + { + "epoch": 5.24, + "learning_rate": 2.4778174652050743e-06, + "loss": 0.6271, + "step": 14182 + }, + { + "epoch": 5.24, + "learning_rate": 2.4777805148417293e-06, + "loss": 0.6571, + "step": 14183 + }, + { + "epoch": 5.24, + "learning_rate": 2.477743564478384e-06, + "loss": 0.5301, + "step": 14184 + }, + { + "epoch": 5.24, + "learning_rate": 2.477706614115039e-06, + "loss": 0.4234, + "step": 14185 + }, + { + "epoch": 5.24, + "learning_rate": 2.4776696637516936e-06, + "loss": 0.5012, + "step": 14186 + }, + { + "epoch": 5.24, + "learning_rate": 2.4776327133883487e-06, + "loss": 0.4216, + "step": 14187 + }, + { + "epoch": 5.24, + "learning_rate": 2.477595763025003e-06, + "loss": 0.4906, + "step": 14188 + }, + { + "epoch": 5.24, + "learning_rate": 2.477558812661658e-06, + "loss": 0.5203, + "step": 14189 + }, + { + "epoch": 5.24, + "learning_rate": 2.4775218622983126e-06, + "loss": 0.4092, + "step": 14190 + }, + { + "epoch": 5.24, + "learning_rate": 2.4774849119349672e-06, + "loss": 0.3865, + "step": 14191 + }, + { + "epoch": 5.24, + "learning_rate": 2.4774479615716223e-06, + "loss": 0.5004, + "step": 14192 + }, + { + "epoch": 5.24, + "learning_rate": 2.477411011208277e-06, + "loss": 0.4891, + "step": 14193 + }, + { + "epoch": 5.24, + "learning_rate": 2.4773740608449316e-06, + "loss": 0.4787, + "step": 14194 + }, + { + "epoch": 5.24, + "learning_rate": 2.4773371104815862e-06, + "loss": 0.4198, + "step": 14195 + }, + { + "epoch": 5.24, + "learning_rate": 2.4773001601182413e-06, + "loss": 0.6159, + "step": 14196 + }, + { + "epoch": 5.24, + "learning_rate": 2.477263209754896e-06, + "loss": 0.4515, + "step": 14197 + }, + { + "epoch": 5.24, + "learning_rate": 2.477226259391551e-06, + "loss": 0.6078, + "step": 14198 + }, + { + "epoch": 5.24, + "learning_rate": 2.4771893090282056e-06, + "loss": 0.4382, + "step": 14199 + }, + { + "epoch": 5.24, + "learning_rate": 2.4771523586648602e-06, + "loss": 0.4512, + "step": 14200 + }, + { + "epoch": 5.24, + "learning_rate": 2.477115408301515e-06, + "loss": 0.3971, + "step": 14201 + }, + { + "epoch": 5.24, + "learning_rate": 2.4770784579381695e-06, + "loss": 0.4509, + "step": 14202 + }, + { + "epoch": 5.24, + "learning_rate": 2.4770415075748246e-06, + "loss": 0.5332, + "step": 14203 + }, + { + "epoch": 5.25, + "learning_rate": 2.4770045572114792e-06, + "loss": 0.3475, + "step": 14204 + }, + { + "epoch": 5.25, + "learning_rate": 2.4769676068481343e-06, + "loss": 0.5685, + "step": 14205 + }, + { + "epoch": 5.25, + "learning_rate": 2.476930656484789e-06, + "loss": 0.3984, + "step": 14206 + }, + { + "epoch": 5.25, + "learning_rate": 2.4768937061214436e-06, + "loss": 0.4565, + "step": 14207 + }, + { + "epoch": 5.25, + "learning_rate": 2.476856755758098e-06, + "loss": 0.4954, + "step": 14208 + }, + { + "epoch": 5.25, + "learning_rate": 2.476819805394753e-06, + "loss": 0.5027, + "step": 14209 + }, + { + "epoch": 5.25, + "learning_rate": 2.476782855031408e-06, + "loss": 0.4832, + "step": 14210 + }, + { + "epoch": 5.25, + "learning_rate": 2.4767459046680625e-06, + "loss": 0.4449, + "step": 14211 + }, + { + "epoch": 5.25, + "learning_rate": 2.4767089543047176e-06, + "loss": 0.4438, + "step": 14212 + }, + { + "epoch": 5.25, + "learning_rate": 2.4766720039413722e-06, + "loss": 0.4369, + "step": 14213 + }, + { + "epoch": 5.25, + "learning_rate": 2.476635053578027e-06, + "loss": 0.4817, + "step": 14214 + }, + { + "epoch": 5.25, + "learning_rate": 2.4765981032146815e-06, + "loss": 0.418, + "step": 14215 + }, + { + "epoch": 5.25, + "learning_rate": 2.4765611528513365e-06, + "loss": 0.6268, + "step": 14216 + }, + { + "epoch": 5.25, + "learning_rate": 2.476524202487991e-06, + "loss": 0.4558, + "step": 14217 + }, + { + "epoch": 5.25, + "learning_rate": 2.4764872521246462e-06, + "loss": 0.6155, + "step": 14218 + }, + { + "epoch": 5.25, + "learning_rate": 2.476450301761301e-06, + "loss": 0.5257, + "step": 14219 + }, + { + "epoch": 5.25, + "learning_rate": 2.4764133513979555e-06, + "loss": 0.4686, + "step": 14220 + }, + { + "epoch": 5.25, + "learning_rate": 2.47637640103461e-06, + "loss": 0.5295, + "step": 14221 + }, + { + "epoch": 5.25, + "learning_rate": 2.476339450671265e-06, + "loss": 0.5693, + "step": 14222 + }, + { + "epoch": 5.25, + "learning_rate": 2.47630250030792e-06, + "loss": 0.4179, + "step": 14223 + }, + { + "epoch": 5.25, + "learning_rate": 2.4762655499445745e-06, + "loss": 0.5408, + "step": 14224 + }, + { + "epoch": 5.25, + "learning_rate": 2.4762285995812295e-06, + "loss": 0.5277, + "step": 14225 + }, + { + "epoch": 5.25, + "learning_rate": 2.476191649217884e-06, + "loss": 0.5099, + "step": 14226 + }, + { + "epoch": 5.25, + "learning_rate": 2.476154698854539e-06, + "loss": 0.542, + "step": 14227 + }, + { + "epoch": 5.25, + "learning_rate": 2.4761177484911935e-06, + "loss": 0.4958, + "step": 14228 + }, + { + "epoch": 5.25, + "learning_rate": 2.476080798127848e-06, + "loss": 0.4846, + "step": 14229 + }, + { + "epoch": 5.25, + "learning_rate": 2.476043847764503e-06, + "loss": 0.5617, + "step": 14230 + }, + { + "epoch": 5.26, + "learning_rate": 2.476006897401158e-06, + "loss": 0.5259, + "step": 14231 + }, + { + "epoch": 5.26, + "learning_rate": 2.475969947037813e-06, + "loss": 0.5861, + "step": 14232 + }, + { + "epoch": 5.26, + "learning_rate": 2.475932996674467e-06, + "loss": 0.347, + "step": 14233 + }, + { + "epoch": 5.26, + "learning_rate": 2.475896046311122e-06, + "loss": 0.4256, + "step": 14234 + }, + { + "epoch": 5.26, + "learning_rate": 2.4758590959477768e-06, + "loss": 0.5109, + "step": 14235 + }, + { + "epoch": 5.26, + "learning_rate": 2.475822145584432e-06, + "loss": 0.4948, + "step": 14236 + }, + { + "epoch": 5.26, + "learning_rate": 2.4757851952210865e-06, + "loss": 0.4263, + "step": 14237 + }, + { + "epoch": 5.26, + "learning_rate": 2.4757482448577415e-06, + "loss": 0.4265, + "step": 14238 + }, + { + "epoch": 5.26, + "learning_rate": 2.4757112944943957e-06, + "loss": 0.521, + "step": 14239 + }, + { + "epoch": 5.26, + "learning_rate": 2.4756743441310504e-06, + "loss": 0.533, + "step": 14240 + }, + { + "epoch": 5.26, + "learning_rate": 2.4756373937677054e-06, + "loss": 0.6233, + "step": 14241 + }, + { + "epoch": 5.26, + "learning_rate": 2.47560044340436e-06, + "loss": 0.5043, + "step": 14242 + }, + { + "epoch": 5.26, + "learning_rate": 2.475563493041015e-06, + "loss": 0.4545, + "step": 14243 + }, + { + "epoch": 5.26, + "learning_rate": 2.4755265426776698e-06, + "loss": 0.5435, + "step": 14244 + }, + { + "epoch": 5.26, + "learning_rate": 2.4754895923143244e-06, + "loss": 0.3907, + "step": 14245 + }, + { + "epoch": 5.26, + "learning_rate": 2.475452641950979e-06, + "loss": 0.4563, + "step": 14246 + }, + { + "epoch": 5.26, + "learning_rate": 2.475415691587634e-06, + "loss": 0.6078, + "step": 14247 + }, + { + "epoch": 5.26, + "learning_rate": 2.4753787412242887e-06, + "loss": 0.4216, + "step": 14248 + }, + { + "epoch": 5.26, + "learning_rate": 2.4753417908609434e-06, + "loss": 0.5242, + "step": 14249 + }, + { + "epoch": 5.26, + "learning_rate": 2.4753048404975984e-06, + "loss": 0.4557, + "step": 14250 + }, + { + "epoch": 5.26, + "learning_rate": 2.475267890134253e-06, + "loss": 0.4635, + "step": 14251 + }, + { + "epoch": 5.26, + "learning_rate": 2.4752309397709077e-06, + "loss": 0.5611, + "step": 14252 + }, + { + "epoch": 5.26, + "learning_rate": 2.4751939894075623e-06, + "loss": 0.5246, + "step": 14253 + }, + { + "epoch": 5.26, + "learning_rate": 2.4751570390442174e-06, + "loss": 0.5528, + "step": 14254 + }, + { + "epoch": 5.26, + "learning_rate": 2.475120088680872e-06, + "loss": 0.4533, + "step": 14255 + }, + { + "epoch": 5.26, + "learning_rate": 2.475083138317527e-06, + "loss": 0.4719, + "step": 14256 + }, + { + "epoch": 5.26, + "learning_rate": 2.4750461879541817e-06, + "loss": 0.5723, + "step": 14257 + }, + { + "epoch": 5.27, + "learning_rate": 2.4750092375908364e-06, + "loss": 0.4728, + "step": 14258 + }, + { + "epoch": 5.27, + "learning_rate": 2.474972287227491e-06, + "loss": 0.5748, + "step": 14259 + }, + { + "epoch": 5.27, + "learning_rate": 2.4749353368641456e-06, + "loss": 0.4918, + "step": 14260 + }, + { + "epoch": 5.27, + "learning_rate": 2.4748983865008007e-06, + "loss": 0.3944, + "step": 14261 + }, + { + "epoch": 5.27, + "learning_rate": 2.4748614361374553e-06, + "loss": 0.4099, + "step": 14262 + }, + { + "epoch": 5.27, + "learning_rate": 2.4748244857741104e-06, + "loss": 0.4798, + "step": 14263 + }, + { + "epoch": 5.27, + "learning_rate": 2.474787535410765e-06, + "loss": 0.5347, + "step": 14264 + }, + { + "epoch": 5.27, + "learning_rate": 2.4747505850474197e-06, + "loss": 0.5079, + "step": 14265 + }, + { + "epoch": 5.27, + "learning_rate": 2.4747136346840743e-06, + "loss": 0.609, + "step": 14266 + }, + { + "epoch": 5.27, + "learning_rate": 2.4746766843207294e-06, + "loss": 0.5359, + "step": 14267 + }, + { + "epoch": 5.27, + "learning_rate": 2.474639733957384e-06, + "loss": 0.458, + "step": 14268 + }, + { + "epoch": 5.27, + "learning_rate": 2.4746027835940386e-06, + "loss": 0.4631, + "step": 14269 + }, + { + "epoch": 5.27, + "learning_rate": 2.4745658332306937e-06, + "loss": 0.482, + "step": 14270 + }, + { + "epoch": 5.27, + "learning_rate": 2.4745288828673483e-06, + "loss": 0.4136, + "step": 14271 + }, + { + "epoch": 5.27, + "learning_rate": 2.474491932504003e-06, + "loss": 0.4005, + "step": 14272 + }, + { + "epoch": 5.27, + "learning_rate": 2.4744549821406576e-06, + "loss": 0.4635, + "step": 14273 + }, + { + "epoch": 5.27, + "learning_rate": 2.4744180317773127e-06, + "loss": 0.5127, + "step": 14274 + }, + { + "epoch": 5.27, + "learning_rate": 2.4743810814139673e-06, + "loss": 0.4969, + "step": 14275 + }, + { + "epoch": 5.27, + "learning_rate": 2.4743441310506224e-06, + "loss": 0.4537, + "step": 14276 + }, + { + "epoch": 5.27, + "learning_rate": 2.474307180687277e-06, + "loss": 0.6351, + "step": 14277 + }, + { + "epoch": 5.27, + "learning_rate": 2.4742702303239316e-06, + "loss": 0.4165, + "step": 14278 + }, + { + "epoch": 5.27, + "learning_rate": 2.4742332799605863e-06, + "loss": 0.4279, + "step": 14279 + }, + { + "epoch": 5.27, + "learning_rate": 2.474196329597241e-06, + "loss": 0.5797, + "step": 14280 + }, + { + "epoch": 5.27, + "learning_rate": 2.474159379233896e-06, + "loss": 0.4398, + "step": 14281 + }, + { + "epoch": 5.27, + "learning_rate": 2.4741224288705506e-06, + "loss": 0.4791, + "step": 14282 + }, + { + "epoch": 5.27, + "learning_rate": 2.4740854785072057e-06, + "loss": 0.3953, + "step": 14283 + }, + { + "epoch": 5.27, + "learning_rate": 2.47404852814386e-06, + "loss": 0.5649, + "step": 14284 + }, + { + "epoch": 5.28, + "learning_rate": 2.474011577780515e-06, + "loss": 0.3811, + "step": 14285 + }, + { + "epoch": 5.28, + "learning_rate": 2.4739746274171696e-06, + "loss": 0.3781, + "step": 14286 + }, + { + "epoch": 5.28, + "learning_rate": 2.4739376770538246e-06, + "loss": 0.4862, + "step": 14287 + }, + { + "epoch": 5.28, + "learning_rate": 2.4739007266904793e-06, + "loss": 0.6262, + "step": 14288 + }, + { + "epoch": 5.28, + "learning_rate": 2.473863776327134e-06, + "loss": 0.5068, + "step": 14289 + }, + { + "epoch": 5.28, + "learning_rate": 2.4738268259637885e-06, + "loss": 0.5211, + "step": 14290 + }, + { + "epoch": 5.28, + "learning_rate": 2.473789875600443e-06, + "loss": 0.6135, + "step": 14291 + }, + { + "epoch": 5.28, + "learning_rate": 2.4737529252370982e-06, + "loss": 0.517, + "step": 14292 + }, + { + "epoch": 5.28, + "learning_rate": 2.473715974873753e-06, + "loss": 0.4548, + "step": 14293 + }, + { + "epoch": 5.28, + "learning_rate": 2.473679024510408e-06, + "loss": 0.531, + "step": 14294 + }, + { + "epoch": 5.28, + "learning_rate": 2.4736420741470626e-06, + "loss": 0.4577, + "step": 14295 + }, + { + "epoch": 5.28, + "learning_rate": 2.473605123783717e-06, + "loss": 0.4406, + "step": 14296 + }, + { + "epoch": 5.28, + "learning_rate": 2.473568173420372e-06, + "loss": 0.5182, + "step": 14297 + }, + { + "epoch": 5.28, + "learning_rate": 2.473531223057027e-06, + "loss": 0.3772, + "step": 14298 + }, + { + "epoch": 5.28, + "learning_rate": 2.4734942726936815e-06, + "loss": 0.493, + "step": 14299 + }, + { + "epoch": 5.28, + "learning_rate": 2.473457322330336e-06, + "loss": 0.4699, + "step": 14300 + }, + { + "epoch": 5.28, + "learning_rate": 2.4734203719669912e-06, + "loss": 0.466, + "step": 14301 + }, + { + "epoch": 5.28, + "learning_rate": 2.473383421603646e-06, + "loss": 0.4655, + "step": 14302 + }, + { + "epoch": 5.28, + "learning_rate": 2.4733464712403005e-06, + "loss": 0.5359, + "step": 14303 + }, + { + "epoch": 5.28, + "learning_rate": 2.473309520876955e-06, + "loss": 0.3845, + "step": 14304 + }, + { + "epoch": 5.28, + "learning_rate": 2.47327257051361e-06, + "loss": 0.5278, + "step": 14305 + }, + { + "epoch": 5.28, + "learning_rate": 2.473235620150265e-06, + "loss": 0.5646, + "step": 14306 + }, + { + "epoch": 5.28, + "learning_rate": 2.47319866978692e-06, + "loss": 0.4179, + "step": 14307 + }, + { + "epoch": 5.28, + "learning_rate": 2.4731617194235745e-06, + "loss": 0.5743, + "step": 14308 + }, + { + "epoch": 5.28, + "learning_rate": 2.473124769060229e-06, + "loss": 0.5308, + "step": 14309 + }, + { + "epoch": 5.28, + "learning_rate": 2.473087818696884e-06, + "loss": 0.5852, + "step": 14310 + }, + { + "epoch": 5.28, + "learning_rate": 2.4730508683335384e-06, + "loss": 0.4812, + "step": 14311 + }, + { + "epoch": 5.29, + "learning_rate": 2.4730139179701935e-06, + "loss": 0.6111, + "step": 14312 + }, + { + "epoch": 5.29, + "learning_rate": 2.472976967606848e-06, + "loss": 0.5236, + "step": 14313 + }, + { + "epoch": 5.29, + "learning_rate": 2.472940017243503e-06, + "loss": 0.5768, + "step": 14314 + }, + { + "epoch": 5.29, + "learning_rate": 2.472903066880158e-06, + "loss": 0.4041, + "step": 14315 + }, + { + "epoch": 5.29, + "learning_rate": 2.4728661165168125e-06, + "loss": 0.4893, + "step": 14316 + }, + { + "epoch": 5.29, + "learning_rate": 2.472829166153467e-06, + "loss": 0.5168, + "step": 14317 + }, + { + "epoch": 5.29, + "learning_rate": 2.472792215790122e-06, + "loss": 0.548, + "step": 14318 + }, + { + "epoch": 5.29, + "learning_rate": 2.472755265426777e-06, + "loss": 0.4024, + "step": 14319 + }, + { + "epoch": 5.29, + "learning_rate": 2.4727183150634314e-06, + "loss": 0.6447, + "step": 14320 + }, + { + "epoch": 5.29, + "learning_rate": 2.4726813647000865e-06, + "loss": 0.4452, + "step": 14321 + }, + { + "epoch": 5.29, + "learning_rate": 2.472644414336741e-06, + "loss": 0.4376, + "step": 14322 + }, + { + "epoch": 5.29, + "learning_rate": 2.4726074639733958e-06, + "loss": 0.4646, + "step": 14323 + }, + { + "epoch": 5.29, + "learning_rate": 2.4725705136100504e-06, + "loss": 0.5581, + "step": 14324 + }, + { + "epoch": 5.29, + "learning_rate": 2.4725335632467055e-06, + "loss": 0.4746, + "step": 14325 + }, + { + "epoch": 5.29, + "learning_rate": 2.47249661288336e-06, + "loss": 0.5234, + "step": 14326 + }, + { + "epoch": 5.29, + "learning_rate": 2.472459662520015e-06, + "loss": 0.4527, + "step": 14327 + }, + { + "epoch": 5.29, + "learning_rate": 2.47242271215667e-06, + "loss": 0.4471, + "step": 14328 + }, + { + "epoch": 5.29, + "learning_rate": 2.472385761793324e-06, + "loss": 0.477, + "step": 14329 + }, + { + "epoch": 5.29, + "learning_rate": 2.472348811429979e-06, + "loss": 0.5068, + "step": 14330 + }, + { + "epoch": 5.29, + "learning_rate": 2.4723118610666337e-06, + "loss": 0.4812, + "step": 14331 + }, + { + "epoch": 5.29, + "learning_rate": 2.4722749107032888e-06, + "loss": 0.5133, + "step": 14332 + }, + { + "epoch": 5.29, + "learning_rate": 2.4722379603399434e-06, + "loss": 0.561, + "step": 14333 + }, + { + "epoch": 5.29, + "learning_rate": 2.472201009976598e-06, + "loss": 0.4855, + "step": 14334 + }, + { + "epoch": 5.29, + "learning_rate": 2.4721640596132527e-06, + "loss": 0.4864, + "step": 14335 + }, + { + "epoch": 5.29, + "learning_rate": 2.4721271092499077e-06, + "loss": 0.4373, + "step": 14336 + }, + { + "epoch": 5.29, + "learning_rate": 2.4720901588865624e-06, + "loss": 0.4889, + "step": 14337 + }, + { + "epoch": 5.29, + "learning_rate": 2.4720532085232174e-06, + "loss": 0.5677, + "step": 14338 + }, + { + "epoch": 5.3, + "learning_rate": 2.472016258159872e-06, + "loss": 0.4991, + "step": 14339 + }, + { + "epoch": 5.3, + "learning_rate": 2.4719793077965267e-06, + "loss": 0.4285, + "step": 14340 + }, + { + "epoch": 5.3, + "learning_rate": 2.4719423574331813e-06, + "loss": 0.5039, + "step": 14341 + }, + { + "epoch": 5.3, + "learning_rate": 2.471905407069836e-06, + "loss": 0.4617, + "step": 14342 + }, + { + "epoch": 5.3, + "learning_rate": 2.471868456706491e-06, + "loss": 0.5004, + "step": 14343 + }, + { + "epoch": 5.3, + "learning_rate": 2.4718315063431457e-06, + "loss": 0.5515, + "step": 14344 + }, + { + "epoch": 5.3, + "learning_rate": 2.4717945559798007e-06, + "loss": 0.4988, + "step": 14345 + }, + { + "epoch": 5.3, + "learning_rate": 2.4717576056164554e-06, + "loss": 0.4517, + "step": 14346 + }, + { + "epoch": 5.3, + "learning_rate": 2.47172065525311e-06, + "loss": 0.5198, + "step": 14347 + }, + { + "epoch": 5.3, + "learning_rate": 2.4716837048897647e-06, + "loss": 0.5532, + "step": 14348 + }, + { + "epoch": 5.3, + "learning_rate": 2.4716467545264193e-06, + "loss": 0.5426, + "step": 14349 + }, + { + "epoch": 5.3, + "learning_rate": 2.4716098041630743e-06, + "loss": 0.3901, + "step": 14350 + }, + { + "epoch": 5.3, + "learning_rate": 2.471572853799729e-06, + "loss": 0.408, + "step": 14351 + }, + { + "epoch": 5.3, + "learning_rate": 2.471535903436384e-06, + "loss": 0.3999, + "step": 14352 + }, + { + "epoch": 5.3, + "learning_rate": 2.4714989530730387e-06, + "loss": 0.528, + "step": 14353 + }, + { + "epoch": 5.3, + "learning_rate": 2.4714620027096933e-06, + "loss": 0.6188, + "step": 14354 + }, + { + "epoch": 5.3, + "learning_rate": 2.471425052346348e-06, + "loss": 0.4559, + "step": 14355 + }, + { + "epoch": 5.3, + "learning_rate": 2.471388101983003e-06, + "loss": 0.6589, + "step": 14356 + }, + { + "epoch": 5.3, + "learning_rate": 2.4713511516196577e-06, + "loss": 0.59, + "step": 14357 + }, + { + "epoch": 5.3, + "learning_rate": 2.4713142012563127e-06, + "loss": 0.4349, + "step": 14358 + }, + { + "epoch": 5.3, + "learning_rate": 2.4712772508929673e-06, + "loss": 0.4482, + "step": 14359 + }, + { + "epoch": 5.3, + "learning_rate": 2.471240300529622e-06, + "loss": 0.5188, + "step": 14360 + }, + { + "epoch": 5.3, + "learning_rate": 2.4712033501662766e-06, + "loss": 0.488, + "step": 14361 + }, + { + "epoch": 5.3, + "learning_rate": 2.4711663998029313e-06, + "loss": 0.4055, + "step": 14362 + }, + { + "epoch": 5.3, + "learning_rate": 2.4711294494395863e-06, + "loss": 0.4618, + "step": 14363 + }, + { + "epoch": 5.3, + "learning_rate": 2.471092499076241e-06, + "loss": 0.3826, + "step": 14364 + }, + { + "epoch": 5.3, + "learning_rate": 2.471055548712896e-06, + "loss": 0.5126, + "step": 14365 + }, + { + "epoch": 5.31, + "learning_rate": 2.4710185983495506e-06, + "loss": 0.5044, + "step": 14366 + }, + { + "epoch": 5.31, + "learning_rate": 2.4709816479862053e-06, + "loss": 0.5046, + "step": 14367 + }, + { + "epoch": 5.31, + "learning_rate": 2.47094469762286e-06, + "loss": 0.54, + "step": 14368 + }, + { + "epoch": 5.31, + "learning_rate": 2.4709077472595146e-06, + "loss": 0.4467, + "step": 14369 + }, + { + "epoch": 5.31, + "learning_rate": 2.4708707968961696e-06, + "loss": 0.4327, + "step": 14370 + }, + { + "epoch": 5.31, + "learning_rate": 2.4708338465328243e-06, + "loss": 0.579, + "step": 14371 + }, + { + "epoch": 5.31, + "learning_rate": 2.4707968961694793e-06, + "loss": 0.5014, + "step": 14372 + }, + { + "epoch": 5.31, + "learning_rate": 2.4707599458061335e-06, + "loss": 0.4575, + "step": 14373 + }, + { + "epoch": 5.31, + "learning_rate": 2.4707229954427886e-06, + "loss": 0.3657, + "step": 14374 + }, + { + "epoch": 5.31, + "learning_rate": 2.4706860450794432e-06, + "loss": 0.4435, + "step": 14375 + }, + { + "epoch": 5.31, + "learning_rate": 2.4706490947160983e-06, + "loss": 0.565, + "step": 14376 + }, + { + "epoch": 5.31, + "learning_rate": 2.470612144352753e-06, + "loss": 0.4459, + "step": 14377 + }, + { + "epoch": 5.31, + "learning_rate": 2.470575193989408e-06, + "loss": 0.498, + "step": 14378 + }, + { + "epoch": 5.31, + "learning_rate": 2.470538243626062e-06, + "loss": 0.4494, + "step": 14379 + }, + { + "epoch": 5.31, + "learning_rate": 2.470501293262717e-06, + "loss": 0.3777, + "step": 14380 + }, + { + "epoch": 5.31, + "learning_rate": 2.470464342899372e-06, + "loss": 0.4822, + "step": 14381 + }, + { + "epoch": 5.31, + "learning_rate": 2.4704273925360265e-06, + "loss": 0.5032, + "step": 14382 + }, + { + "epoch": 5.31, + "learning_rate": 2.4703904421726816e-06, + "loss": 0.505, + "step": 14383 + }, + { + "epoch": 5.31, + "learning_rate": 2.4703534918093362e-06, + "loss": 0.4318, + "step": 14384 + }, + { + "epoch": 5.31, + "learning_rate": 2.470316541445991e-06, + "loss": 0.5244, + "step": 14385 + }, + { + "epoch": 5.31, + "learning_rate": 2.4702795910826455e-06, + "loss": 0.3578, + "step": 14386 + }, + { + "epoch": 5.31, + "learning_rate": 2.4702426407193006e-06, + "loss": 0.601, + "step": 14387 + }, + { + "epoch": 5.31, + "learning_rate": 2.470205690355955e-06, + "loss": 0.4596, + "step": 14388 + }, + { + "epoch": 5.31, + "learning_rate": 2.47016873999261e-06, + "loss": 0.4672, + "step": 14389 + }, + { + "epoch": 5.31, + "learning_rate": 2.470131789629265e-06, + "loss": 0.4296, + "step": 14390 + }, + { + "epoch": 5.31, + "learning_rate": 2.4700948392659195e-06, + "loss": 0.4505, + "step": 14391 + }, + { + "epoch": 5.31, + "learning_rate": 2.470057888902574e-06, + "loss": 0.5351, + "step": 14392 + }, + { + "epoch": 5.31, + "learning_rate": 2.470020938539229e-06, + "loss": 0.4659, + "step": 14393 + }, + { + "epoch": 5.32, + "learning_rate": 2.469983988175884e-06, + "loss": 0.5027, + "step": 14394 + }, + { + "epoch": 5.32, + "learning_rate": 2.4699470378125385e-06, + "loss": 0.5791, + "step": 14395 + }, + { + "epoch": 5.32, + "learning_rate": 2.4699100874491936e-06, + "loss": 0.3806, + "step": 14396 + }, + { + "epoch": 5.32, + "learning_rate": 2.469873137085848e-06, + "loss": 0.5975, + "step": 14397 + }, + { + "epoch": 5.32, + "learning_rate": 2.469836186722503e-06, + "loss": 0.423, + "step": 14398 + }, + { + "epoch": 5.32, + "learning_rate": 2.4697992363591575e-06, + "loss": 0.5142, + "step": 14399 + }, + { + "epoch": 5.32, + "learning_rate": 2.469762285995812e-06, + "loss": 0.4534, + "step": 14400 + }, + { + "epoch": 5.32, + "learning_rate": 2.469725335632467e-06, + "loss": 0.3935, + "step": 14401 + }, + { + "epoch": 5.32, + "learning_rate": 2.469688385269122e-06, + "loss": 0.4347, + "step": 14402 + }, + { + "epoch": 5.32, + "learning_rate": 2.469651434905777e-06, + "loss": 0.5013, + "step": 14403 + }, + { + "epoch": 5.32, + "learning_rate": 2.4696144845424315e-06, + "loss": 0.4971, + "step": 14404 + }, + { + "epoch": 5.32, + "learning_rate": 2.469577534179086e-06, + "loss": 0.5081, + "step": 14405 + }, + { + "epoch": 5.32, + "learning_rate": 2.4695405838157408e-06, + "loss": 0.6526, + "step": 14406 + }, + { + "epoch": 5.32, + "learning_rate": 2.469503633452396e-06, + "loss": 0.5149, + "step": 14407 + }, + { + "epoch": 5.32, + "learning_rate": 2.4694666830890505e-06, + "loss": 0.4318, + "step": 14408 + }, + { + "epoch": 5.32, + "learning_rate": 2.469429732725705e-06, + "loss": 0.622, + "step": 14409 + }, + { + "epoch": 5.32, + "learning_rate": 2.46939278236236e-06, + "loss": 0.545, + "step": 14410 + }, + { + "epoch": 5.32, + "learning_rate": 2.469355831999015e-06, + "loss": 0.4431, + "step": 14411 + }, + { + "epoch": 5.32, + "learning_rate": 2.4693188816356694e-06, + "loss": 0.3727, + "step": 14412 + }, + { + "epoch": 5.32, + "learning_rate": 2.469281931272324e-06, + "loss": 0.5431, + "step": 14413 + }, + { + "epoch": 5.32, + "learning_rate": 2.469244980908979e-06, + "loss": 0.6012, + "step": 14414 + }, + { + "epoch": 5.32, + "learning_rate": 2.4692080305456338e-06, + "loss": 0.5042, + "step": 14415 + }, + { + "epoch": 5.32, + "learning_rate": 2.469171080182289e-06, + "loss": 0.4173, + "step": 14416 + }, + { + "epoch": 5.32, + "learning_rate": 2.4691341298189435e-06, + "loss": 0.4589, + "step": 14417 + }, + { + "epoch": 5.32, + "learning_rate": 2.469097179455598e-06, + "loss": 0.5458, + "step": 14418 + }, + { + "epoch": 5.32, + "learning_rate": 2.4690602290922527e-06, + "loss": 0.5411, + "step": 14419 + }, + { + "epoch": 5.32, + "learning_rate": 2.4690232787289074e-06, + "loss": 0.4779, + "step": 14420 + }, + { + "epoch": 5.33, + "learning_rate": 2.4689863283655624e-06, + "loss": 0.5357, + "step": 14421 + }, + { + "epoch": 5.33, + "learning_rate": 2.468949378002217e-06, + "loss": 0.505, + "step": 14422 + }, + { + "epoch": 5.33, + "learning_rate": 2.468912427638872e-06, + "loss": 0.5602, + "step": 14423 + }, + { + "epoch": 5.33, + "learning_rate": 2.4688754772755263e-06, + "loss": 0.5573, + "step": 14424 + }, + { + "epoch": 5.33, + "learning_rate": 2.4688385269121814e-06, + "loss": 0.5451, + "step": 14425 + }, + { + "epoch": 5.33, + "learning_rate": 2.468801576548836e-06, + "loss": 0.4457, + "step": 14426 + }, + { + "epoch": 5.33, + "learning_rate": 2.468764626185491e-06, + "loss": 0.4649, + "step": 14427 + }, + { + "epoch": 5.33, + "learning_rate": 2.4687276758221457e-06, + "loss": 0.5022, + "step": 14428 + }, + { + "epoch": 5.33, + "learning_rate": 2.4686907254588004e-06, + "loss": 0.3311, + "step": 14429 + }, + { + "epoch": 5.33, + "learning_rate": 2.468653775095455e-06, + "loss": 0.5708, + "step": 14430 + }, + { + "epoch": 5.33, + "learning_rate": 2.4686168247321096e-06, + "loss": 0.5948, + "step": 14431 + }, + { + "epoch": 5.33, + "learning_rate": 2.4685798743687647e-06, + "loss": 0.4221, + "step": 14432 + }, + { + "epoch": 5.33, + "learning_rate": 2.4685429240054193e-06, + "loss": 0.4127, + "step": 14433 + }, + { + "epoch": 5.33, + "learning_rate": 2.4685059736420744e-06, + "loss": 0.4958, + "step": 14434 + }, + { + "epoch": 5.33, + "learning_rate": 2.468469023278729e-06, + "loss": 0.5268, + "step": 14435 + }, + { + "epoch": 5.33, + "learning_rate": 2.4684320729153837e-06, + "loss": 0.5976, + "step": 14436 + }, + { + "epoch": 5.33, + "learning_rate": 2.4683951225520383e-06, + "loss": 0.4984, + "step": 14437 + }, + { + "epoch": 5.33, + "learning_rate": 2.468358172188693e-06, + "loss": 0.4103, + "step": 14438 + }, + { + "epoch": 5.33, + "learning_rate": 2.468321221825348e-06, + "loss": 0.5833, + "step": 14439 + }, + { + "epoch": 5.33, + "learning_rate": 2.4682842714620026e-06, + "loss": 0.4052, + "step": 14440 + }, + { + "epoch": 5.33, + "learning_rate": 2.4682473210986577e-06, + "loss": 0.4998, + "step": 14441 + }, + { + "epoch": 5.33, + "learning_rate": 2.4682103707353123e-06, + "loss": 0.4874, + "step": 14442 + }, + { + "epoch": 5.33, + "learning_rate": 2.468173420371967e-06, + "loss": 0.4835, + "step": 14443 + }, + { + "epoch": 5.33, + "learning_rate": 2.4681364700086216e-06, + "loss": 0.3839, + "step": 14444 + }, + { + "epoch": 5.33, + "learning_rate": 2.4680995196452767e-06, + "loss": 0.564, + "step": 14445 + }, + { + "epoch": 5.33, + "learning_rate": 2.4680625692819313e-06, + "loss": 0.51, + "step": 14446 + }, + { + "epoch": 5.33, + "learning_rate": 2.4680256189185864e-06, + "loss": 0.4663, + "step": 14447 + }, + { + "epoch": 5.34, + "learning_rate": 2.467988668555241e-06, + "loss": 0.4164, + "step": 14448 + }, + { + "epoch": 5.34, + "learning_rate": 2.4679517181918956e-06, + "loss": 0.4348, + "step": 14449 + }, + { + "epoch": 5.34, + "learning_rate": 2.4679147678285503e-06, + "loss": 0.5756, + "step": 14450 + }, + { + "epoch": 5.34, + "learning_rate": 2.467877817465205e-06, + "loss": 0.4675, + "step": 14451 + }, + { + "epoch": 5.34, + "learning_rate": 2.46784086710186e-06, + "loss": 0.5369, + "step": 14452 + }, + { + "epoch": 5.34, + "learning_rate": 2.4678039167385146e-06, + "loss": 0.4103, + "step": 14453 + }, + { + "epoch": 5.34, + "learning_rate": 2.4677669663751697e-06, + "loss": 0.4914, + "step": 14454 + }, + { + "epoch": 5.34, + "learning_rate": 2.4677300160118243e-06, + "loss": 0.4589, + "step": 14455 + }, + { + "epoch": 5.34, + "learning_rate": 2.467693065648479e-06, + "loss": 0.5243, + "step": 14456 + }, + { + "epoch": 5.34, + "learning_rate": 2.4676561152851336e-06, + "loss": 0.4004, + "step": 14457 + }, + { + "epoch": 5.34, + "learning_rate": 2.4676191649217882e-06, + "loss": 0.5096, + "step": 14458 + }, + { + "epoch": 5.34, + "learning_rate": 2.4675822145584433e-06, + "loss": 0.4194, + "step": 14459 + }, + { + "epoch": 5.34, + "learning_rate": 2.467545264195098e-06, + "loss": 0.4705, + "step": 14460 + }, + { + "epoch": 5.34, + "learning_rate": 2.467508313831753e-06, + "loss": 0.4521, + "step": 14461 + }, + { + "epoch": 5.34, + "learning_rate": 2.4674713634684076e-06, + "loss": 0.5103, + "step": 14462 + }, + { + "epoch": 5.34, + "learning_rate": 2.4674344131050622e-06, + "loss": 0.4725, + "step": 14463 + }, + { + "epoch": 5.34, + "learning_rate": 2.467397462741717e-06, + "loss": 0.5232, + "step": 14464 + }, + { + "epoch": 5.34, + "learning_rate": 2.467360512378372e-06, + "loss": 0.5543, + "step": 14465 + }, + { + "epoch": 5.34, + "learning_rate": 2.4673235620150266e-06, + "loss": 0.556, + "step": 14466 + }, + { + "epoch": 5.34, + "learning_rate": 2.4672866116516816e-06, + "loss": 0.5713, + "step": 14467 + }, + { + "epoch": 5.34, + "learning_rate": 2.4672496612883363e-06, + "loss": 0.3933, + "step": 14468 + }, + { + "epoch": 5.34, + "learning_rate": 2.4672127109249905e-06, + "loss": 0.4069, + "step": 14469 + }, + { + "epoch": 5.34, + "learning_rate": 2.4671757605616455e-06, + "loss": 0.5213, + "step": 14470 + }, + { + "epoch": 5.34, + "learning_rate": 2.4671388101983e-06, + "loss": 0.5204, + "step": 14471 + }, + { + "epoch": 5.34, + "learning_rate": 2.4671018598349552e-06, + "loss": 0.5049, + "step": 14472 + }, + { + "epoch": 5.34, + "learning_rate": 2.46706490947161e-06, + "loss": 0.5201, + "step": 14473 + }, + { + "epoch": 5.34, + "learning_rate": 2.467027959108265e-06, + "loss": 0.4845, + "step": 14474 + }, + { + "epoch": 5.35, + "learning_rate": 2.466991008744919e-06, + "loss": 0.4869, + "step": 14475 + }, + { + "epoch": 5.35, + "learning_rate": 2.466954058381574e-06, + "loss": 0.5639, + "step": 14476 + }, + { + "epoch": 5.35, + "learning_rate": 2.466917108018229e-06, + "loss": 0.4233, + "step": 14477 + }, + { + "epoch": 5.35, + "learning_rate": 2.4668801576548835e-06, + "loss": 0.4326, + "step": 14478 + }, + { + "epoch": 5.35, + "learning_rate": 2.4668432072915385e-06, + "loss": 0.5519, + "step": 14479 + }, + { + "epoch": 5.35, + "learning_rate": 2.466806256928193e-06, + "loss": 0.4711, + "step": 14480 + }, + { + "epoch": 5.35, + "learning_rate": 2.466769306564848e-06, + "loss": 0.4313, + "step": 14481 + }, + { + "epoch": 5.35, + "learning_rate": 2.4667323562015025e-06, + "loss": 0.5168, + "step": 14482 + }, + { + "epoch": 5.35, + "learning_rate": 2.4666954058381575e-06, + "loss": 0.4742, + "step": 14483 + }, + { + "epoch": 5.35, + "learning_rate": 2.466658455474812e-06, + "loss": 0.4415, + "step": 14484 + }, + { + "epoch": 5.35, + "learning_rate": 2.466621505111467e-06, + "loss": 0.4465, + "step": 14485 + }, + { + "epoch": 5.35, + "learning_rate": 2.466584554748122e-06, + "loss": 0.4784, + "step": 14486 + }, + { + "epoch": 5.35, + "learning_rate": 2.4665476043847765e-06, + "loss": 0.5325, + "step": 14487 + }, + { + "epoch": 5.35, + "learning_rate": 2.466510654021431e-06, + "loss": 0.4701, + "step": 14488 + }, + { + "epoch": 5.35, + "learning_rate": 2.4664737036580858e-06, + "loss": 0.5977, + "step": 14489 + }, + { + "epoch": 5.35, + "learning_rate": 2.466436753294741e-06, + "loss": 0.3984, + "step": 14490 + }, + { + "epoch": 5.35, + "learning_rate": 2.4663998029313954e-06, + "loss": 0.4248, + "step": 14491 + }, + { + "epoch": 5.35, + "learning_rate": 2.4663628525680505e-06, + "loss": 0.4465, + "step": 14492 + }, + { + "epoch": 5.35, + "learning_rate": 2.466325902204705e-06, + "loss": 0.416, + "step": 14493 + }, + { + "epoch": 5.35, + "learning_rate": 2.4662889518413598e-06, + "loss": 0.4621, + "step": 14494 + }, + { + "epoch": 5.35, + "learning_rate": 2.4662520014780144e-06, + "loss": 0.4906, + "step": 14495 + }, + { + "epoch": 5.35, + "learning_rate": 2.4662150511146695e-06, + "loss": 0.5314, + "step": 14496 + }, + { + "epoch": 5.35, + "learning_rate": 2.466178100751324e-06, + "loss": 0.4909, + "step": 14497 + }, + { + "epoch": 5.35, + "learning_rate": 2.4661411503879788e-06, + "loss": 0.5651, + "step": 14498 + }, + { + "epoch": 5.35, + "learning_rate": 2.466104200024634e-06, + "loss": 0.6509, + "step": 14499 + }, + { + "epoch": 5.35, + "learning_rate": 2.4660672496612884e-06, + "loss": 0.4839, + "step": 14500 + }, + { + "epoch": 5.35, + "learning_rate": 2.466030299297943e-06, + "loss": 0.4401, + "step": 14501 + }, + { + "epoch": 5.36, + "learning_rate": 2.4659933489345977e-06, + "loss": 0.5747, + "step": 14502 + }, + { + "epoch": 5.36, + "learning_rate": 2.4659563985712528e-06, + "loss": 0.4539, + "step": 14503 + }, + { + "epoch": 5.36, + "learning_rate": 2.4659194482079074e-06, + "loss": 0.5937, + "step": 14504 + }, + { + "epoch": 5.36, + "learning_rate": 2.4658824978445625e-06, + "loss": 0.6276, + "step": 14505 + }, + { + "epoch": 5.36, + "learning_rate": 2.465845547481217e-06, + "loss": 0.5365, + "step": 14506 + }, + { + "epoch": 5.36, + "learning_rate": 2.4658085971178718e-06, + "loss": 0.4299, + "step": 14507 + }, + { + "epoch": 5.36, + "learning_rate": 2.4657716467545264e-06, + "loss": 0.4683, + "step": 14508 + }, + { + "epoch": 5.36, + "learning_rate": 2.465734696391181e-06, + "loss": 0.4943, + "step": 14509 + }, + { + "epoch": 5.36, + "learning_rate": 2.465697746027836e-06, + "loss": 0.472, + "step": 14510 + }, + { + "epoch": 5.36, + "learning_rate": 2.4656607956644907e-06, + "loss": 0.5623, + "step": 14511 + }, + { + "epoch": 5.36, + "learning_rate": 2.4656238453011458e-06, + "loss": 0.5192, + "step": 14512 + }, + { + "epoch": 5.36, + "learning_rate": 2.4655868949378004e-06, + "loss": 0.4721, + "step": 14513 + }, + { + "epoch": 5.36, + "learning_rate": 2.465549944574455e-06, + "loss": 0.3572, + "step": 14514 + }, + { + "epoch": 5.36, + "learning_rate": 2.4655129942111097e-06, + "loss": 0.5427, + "step": 14515 + }, + { + "epoch": 5.36, + "learning_rate": 2.4654760438477647e-06, + "loss": 0.4426, + "step": 14516 + }, + { + "epoch": 5.36, + "learning_rate": 2.4654390934844194e-06, + "loss": 0.4745, + "step": 14517 + }, + { + "epoch": 5.36, + "learning_rate": 2.465402143121074e-06, + "loss": 0.5196, + "step": 14518 + }, + { + "epoch": 5.36, + "learning_rate": 2.465365192757729e-06, + "loss": 0.4462, + "step": 14519 + }, + { + "epoch": 5.36, + "learning_rate": 2.4653282423943833e-06, + "loss": 0.459, + "step": 14520 + }, + { + "epoch": 5.36, + "learning_rate": 2.4652912920310384e-06, + "loss": 0.4966, + "step": 14521 + }, + { + "epoch": 5.36, + "learning_rate": 2.465254341667693e-06, + "loss": 0.5607, + "step": 14522 + }, + { + "epoch": 5.36, + "learning_rate": 2.465217391304348e-06, + "loss": 0.4563, + "step": 14523 + }, + { + "epoch": 5.36, + "learning_rate": 2.4651804409410027e-06, + "loss": 0.5142, + "step": 14524 + }, + { + "epoch": 5.36, + "learning_rate": 2.4651434905776573e-06, + "loss": 0.5593, + "step": 14525 + }, + { + "epoch": 5.36, + "learning_rate": 2.465106540214312e-06, + "loss": 0.5256, + "step": 14526 + }, + { + "epoch": 5.36, + "learning_rate": 2.465069589850967e-06, + "loss": 0.4568, + "step": 14527 + }, + { + "epoch": 5.36, + "learning_rate": 2.4650326394876217e-06, + "loss": 0.4635, + "step": 14528 + }, + { + "epoch": 5.37, + "learning_rate": 2.4649956891242763e-06, + "loss": 0.4069, + "step": 14529 + }, + { + "epoch": 5.37, + "learning_rate": 2.4649587387609314e-06, + "loss": 0.4576, + "step": 14530 + }, + { + "epoch": 5.37, + "learning_rate": 2.464921788397586e-06, + "loss": 0.553, + "step": 14531 + }, + { + "epoch": 5.37, + "learning_rate": 2.4648848380342406e-06, + "loss": 0.4505, + "step": 14532 + }, + { + "epoch": 5.37, + "learning_rate": 2.4648478876708953e-06, + "loss": 0.582, + "step": 14533 + }, + { + "epoch": 5.37, + "learning_rate": 2.4648109373075503e-06, + "loss": 0.4604, + "step": 14534 + }, + { + "epoch": 5.37, + "learning_rate": 2.464773986944205e-06, + "loss": 0.4959, + "step": 14535 + }, + { + "epoch": 5.37, + "learning_rate": 2.46473703658086e-06, + "loss": 0.4711, + "step": 14536 + }, + { + "epoch": 5.37, + "learning_rate": 2.4647000862175147e-06, + "loss": 0.5332, + "step": 14537 + }, + { + "epoch": 5.37, + "learning_rate": 2.4646631358541693e-06, + "loss": 0.4806, + "step": 14538 + }, + { + "epoch": 5.37, + "learning_rate": 2.464626185490824e-06, + "loss": 0.4971, + "step": 14539 + }, + { + "epoch": 5.37, + "learning_rate": 2.4645892351274786e-06, + "loss": 0.4933, + "step": 14540 + }, + { + "epoch": 5.37, + "learning_rate": 2.4645522847641336e-06, + "loss": 0.4689, + "step": 14541 + }, + { + "epoch": 5.37, + "learning_rate": 2.4645153344007883e-06, + "loss": 0.4059, + "step": 14542 + }, + { + "epoch": 5.37, + "learning_rate": 2.4644783840374433e-06, + "loss": 0.5948, + "step": 14543 + }, + { + "epoch": 5.37, + "learning_rate": 2.464441433674098e-06, + "loss": 0.4737, + "step": 14544 + }, + { + "epoch": 5.37, + "learning_rate": 2.4644044833107526e-06, + "loss": 0.3768, + "step": 14545 + }, + { + "epoch": 5.37, + "learning_rate": 2.4643675329474072e-06, + "loss": 0.5667, + "step": 14546 + }, + { + "epoch": 5.37, + "learning_rate": 2.4643305825840623e-06, + "loss": 0.4358, + "step": 14547 + }, + { + "epoch": 5.37, + "learning_rate": 2.464293632220717e-06, + "loss": 0.5127, + "step": 14548 + }, + { + "epoch": 5.37, + "learning_rate": 2.4642566818573716e-06, + "loss": 0.4634, + "step": 14549 + }, + { + "epoch": 5.37, + "learning_rate": 2.4642197314940266e-06, + "loss": 0.4233, + "step": 14550 + }, + { + "epoch": 5.37, + "learning_rate": 2.4641827811306813e-06, + "loss": 0.5204, + "step": 14551 + }, + { + "epoch": 5.37, + "learning_rate": 2.464145830767336e-06, + "loss": 0.5853, + "step": 14552 + }, + { + "epoch": 5.37, + "learning_rate": 2.4641088804039905e-06, + "loss": 0.4109, + "step": 14553 + }, + { + "epoch": 5.37, + "learning_rate": 2.4640719300406456e-06, + "loss": 0.417, + "step": 14554 + }, + { + "epoch": 5.37, + "learning_rate": 2.4640349796773002e-06, + "loss": 0.457, + "step": 14555 + }, + { + "epoch": 5.38, + "learning_rate": 2.4639980293139553e-06, + "loss": 0.3671, + "step": 14556 + }, + { + "epoch": 5.38, + "learning_rate": 2.46396107895061e-06, + "loss": 0.4875, + "step": 14557 + }, + { + "epoch": 5.38, + "learning_rate": 2.463924128587264e-06, + "loss": 0.7059, + "step": 14558 + }, + { + "epoch": 5.38, + "learning_rate": 2.463887178223919e-06, + "loss": 0.4544, + "step": 14559 + }, + { + "epoch": 5.38, + "learning_rate": 2.463850227860574e-06, + "loss": 0.5961, + "step": 14560 + }, + { + "epoch": 5.38, + "learning_rate": 2.463813277497229e-06, + "loss": 0.541, + "step": 14561 + }, + { + "epoch": 5.38, + "learning_rate": 2.4637763271338835e-06, + "loss": 0.4278, + "step": 14562 + }, + { + "epoch": 5.38, + "learning_rate": 2.4637393767705386e-06, + "loss": 0.4522, + "step": 14563 + }, + { + "epoch": 5.38, + "learning_rate": 2.463702426407193e-06, + "loss": 0.47, + "step": 14564 + }, + { + "epoch": 5.38, + "learning_rate": 2.463665476043848e-06, + "loss": 0.6505, + "step": 14565 + }, + { + "epoch": 5.38, + "learning_rate": 2.4636285256805025e-06, + "loss": 0.3683, + "step": 14566 + }, + { + "epoch": 5.38, + "learning_rate": 2.4635915753171576e-06, + "loss": 0.5453, + "step": 14567 + }, + { + "epoch": 5.38, + "learning_rate": 2.463554624953812e-06, + "loss": 0.5411, + "step": 14568 + }, + { + "epoch": 5.38, + "learning_rate": 2.463517674590467e-06, + "loss": 0.4604, + "step": 14569 + }, + { + "epoch": 5.38, + "learning_rate": 2.4634807242271215e-06, + "loss": 0.4851, + "step": 14570 + }, + { + "epoch": 5.38, + "learning_rate": 2.463443773863776e-06, + "loss": 0.3969, + "step": 14571 + }, + { + "epoch": 5.38, + "learning_rate": 2.463406823500431e-06, + "loss": 0.5052, + "step": 14572 + }, + { + "epoch": 5.38, + "learning_rate": 2.463369873137086e-06, + "loss": 0.4426, + "step": 14573 + }, + { + "epoch": 5.38, + "learning_rate": 2.463332922773741e-06, + "loss": 0.4033, + "step": 14574 + }, + { + "epoch": 5.38, + "learning_rate": 2.4632959724103955e-06, + "loss": 0.4944, + "step": 14575 + }, + { + "epoch": 5.38, + "learning_rate": 2.46325902204705e-06, + "loss": 0.5924, + "step": 14576 + }, + { + "epoch": 5.38, + "learning_rate": 2.4632220716837048e-06, + "loss": 0.5462, + "step": 14577 + }, + { + "epoch": 5.38, + "learning_rate": 2.4631851213203594e-06, + "loss": 0.5425, + "step": 14578 + }, + { + "epoch": 5.38, + "learning_rate": 2.4631481709570145e-06, + "loss": 0.443, + "step": 14579 + }, + { + "epoch": 5.38, + "learning_rate": 2.463111220593669e-06, + "loss": 0.3936, + "step": 14580 + }, + { + "epoch": 5.38, + "learning_rate": 2.463074270230324e-06, + "loss": 0.5278, + "step": 14581 + }, + { + "epoch": 5.38, + "learning_rate": 2.463037319866979e-06, + "loss": 0.4442, + "step": 14582 + }, + { + "epoch": 5.39, + "learning_rate": 2.4630003695036334e-06, + "loss": 0.5183, + "step": 14583 + }, + { + "epoch": 5.39, + "learning_rate": 2.462963419140288e-06, + "loss": 0.4806, + "step": 14584 + }, + { + "epoch": 5.39, + "learning_rate": 2.462926468776943e-06, + "loss": 0.6045, + "step": 14585 + }, + { + "epoch": 5.39, + "learning_rate": 2.4628895184135978e-06, + "loss": 0.4712, + "step": 14586 + }, + { + "epoch": 5.39, + "learning_rate": 2.462852568050253e-06, + "loss": 0.5323, + "step": 14587 + }, + { + "epoch": 5.39, + "learning_rate": 2.4628156176869075e-06, + "loss": 0.399, + "step": 14588 + }, + { + "epoch": 5.39, + "learning_rate": 2.462778667323562e-06, + "loss": 0.5342, + "step": 14589 + }, + { + "epoch": 5.39, + "learning_rate": 2.4627417169602167e-06, + "loss": 0.5098, + "step": 14590 + }, + { + "epoch": 5.39, + "learning_rate": 2.4627047665968714e-06, + "loss": 0.4797, + "step": 14591 + }, + { + "epoch": 5.39, + "learning_rate": 2.4626678162335264e-06, + "loss": 0.5182, + "step": 14592 + }, + { + "epoch": 5.39, + "learning_rate": 2.462630865870181e-06, + "loss": 0.4529, + "step": 14593 + }, + { + "epoch": 5.39, + "learning_rate": 2.462593915506836e-06, + "loss": 0.5519, + "step": 14594 + }, + { + "epoch": 5.39, + "learning_rate": 2.4625569651434908e-06, + "loss": 0.4881, + "step": 14595 + }, + { + "epoch": 5.39, + "learning_rate": 2.4625200147801454e-06, + "loss": 0.4321, + "step": 14596 + }, + { + "epoch": 5.39, + "learning_rate": 2.4624830644168e-06, + "loss": 0.4532, + "step": 14597 + }, + { + "epoch": 5.39, + "learning_rate": 2.4624461140534547e-06, + "loss": 0.5323, + "step": 14598 + }, + { + "epoch": 5.39, + "learning_rate": 2.4624091636901097e-06, + "loss": 0.5431, + "step": 14599 + }, + { + "epoch": 5.39, + "learning_rate": 2.4623722133267644e-06, + "loss": 0.4272, + "step": 14600 + }, + { + "epoch": 5.39, + "learning_rate": 2.4623352629634194e-06, + "loss": 0.5147, + "step": 14601 + }, + { + "epoch": 5.39, + "learning_rate": 2.462298312600074e-06, + "loss": 0.3982, + "step": 14602 + }, + { + "epoch": 5.39, + "learning_rate": 2.4622613622367287e-06, + "loss": 0.5031, + "step": 14603 + }, + { + "epoch": 5.39, + "learning_rate": 2.4622244118733833e-06, + "loss": 0.5446, + "step": 14604 + }, + { + "epoch": 5.39, + "learning_rate": 2.4621874615100384e-06, + "loss": 0.5778, + "step": 14605 + }, + { + "epoch": 5.39, + "learning_rate": 2.462150511146693e-06, + "loss": 0.4796, + "step": 14606 + }, + { + "epoch": 5.39, + "learning_rate": 2.462113560783348e-06, + "loss": 0.5105, + "step": 14607 + }, + { + "epoch": 5.39, + "learning_rate": 2.4620766104200027e-06, + "loss": 0.5869, + "step": 14608 + }, + { + "epoch": 5.39, + "learning_rate": 2.462039660056657e-06, + "loss": 0.4009, + "step": 14609 + }, + { + "epoch": 5.4, + "learning_rate": 2.462002709693312e-06, + "loss": 0.5484, + "step": 14610 + }, + { + "epoch": 5.4, + "learning_rate": 2.4619657593299666e-06, + "loss": 0.5259, + "step": 14611 + }, + { + "epoch": 5.4, + "learning_rate": 2.4619288089666217e-06, + "loss": 0.5087, + "step": 14612 + }, + { + "epoch": 5.4, + "learning_rate": 2.4618918586032763e-06, + "loss": 0.47, + "step": 14613 + }, + { + "epoch": 5.4, + "learning_rate": 2.4618549082399314e-06, + "loss": 0.5171, + "step": 14614 + }, + { + "epoch": 5.4, + "learning_rate": 2.4618179578765856e-06, + "loss": 0.4494, + "step": 14615 + }, + { + "epoch": 5.4, + "learning_rate": 2.4617810075132407e-06, + "loss": 0.5493, + "step": 14616 + }, + { + "epoch": 5.4, + "learning_rate": 2.4617440571498953e-06, + "loss": 0.4143, + "step": 14617 + }, + { + "epoch": 5.4, + "learning_rate": 2.46170710678655e-06, + "loss": 0.4075, + "step": 14618 + }, + { + "epoch": 5.4, + "learning_rate": 2.461670156423205e-06, + "loss": 0.4994, + "step": 14619 + }, + { + "epoch": 5.4, + "learning_rate": 2.4616332060598596e-06, + "loss": 0.5254, + "step": 14620 + }, + { + "epoch": 5.4, + "learning_rate": 2.4615962556965143e-06, + "loss": 0.4011, + "step": 14621 + }, + { + "epoch": 5.4, + "learning_rate": 2.461559305333169e-06, + "loss": 0.6036, + "step": 14622 + }, + { + "epoch": 5.4, + "learning_rate": 2.461522354969824e-06, + "loss": 0.5314, + "step": 14623 + }, + { + "epoch": 5.4, + "learning_rate": 2.4614854046064786e-06, + "loss": 0.4291, + "step": 14624 + }, + { + "epoch": 5.4, + "learning_rate": 2.4614484542431337e-06, + "loss": 0.4203, + "step": 14625 + }, + { + "epoch": 5.4, + "learning_rate": 2.4614115038797883e-06, + "loss": 0.3914, + "step": 14626 + }, + { + "epoch": 5.4, + "learning_rate": 2.461374553516443e-06, + "loss": 0.4526, + "step": 14627 + }, + { + "epoch": 5.4, + "learning_rate": 2.4613376031530976e-06, + "loss": 0.4347, + "step": 14628 + }, + { + "epoch": 5.4, + "learning_rate": 2.4613006527897522e-06, + "loss": 0.4992, + "step": 14629 + }, + { + "epoch": 5.4, + "learning_rate": 2.4612637024264073e-06, + "loss": 0.5951, + "step": 14630 + }, + { + "epoch": 5.4, + "learning_rate": 2.461226752063062e-06, + "loss": 0.5034, + "step": 14631 + }, + { + "epoch": 5.4, + "learning_rate": 2.461189801699717e-06, + "loss": 0.5386, + "step": 14632 + }, + { + "epoch": 5.4, + "learning_rate": 2.4611528513363716e-06, + "loss": 0.4411, + "step": 14633 + }, + { + "epoch": 5.4, + "learning_rate": 2.4611159009730262e-06, + "loss": 0.4669, + "step": 14634 + }, + { + "epoch": 5.4, + "learning_rate": 2.461078950609681e-06, + "loss": 0.4615, + "step": 14635 + }, + { + "epoch": 5.4, + "learning_rate": 2.461042000246336e-06, + "loss": 0.4095, + "step": 14636 + }, + { + "epoch": 5.41, + "learning_rate": 2.4610050498829906e-06, + "loss": 0.4978, + "step": 14637 + }, + { + "epoch": 5.41, + "learning_rate": 2.4609680995196452e-06, + "loss": 0.4643, + "step": 14638 + }, + { + "epoch": 5.41, + "learning_rate": 2.4609311491563003e-06, + "loss": 0.4464, + "step": 14639 + }, + { + "epoch": 5.41, + "learning_rate": 2.460894198792955e-06, + "loss": 0.4657, + "step": 14640 + }, + { + "epoch": 5.41, + "learning_rate": 2.4608572484296095e-06, + "loss": 0.6353, + "step": 14641 + }, + { + "epoch": 5.41, + "learning_rate": 2.460820298066264e-06, + "loss": 0.4353, + "step": 14642 + }, + { + "epoch": 5.41, + "learning_rate": 2.4607833477029192e-06, + "loss": 0.5028, + "step": 14643 + }, + { + "epoch": 5.41, + "learning_rate": 2.460746397339574e-06, + "loss": 0.5682, + "step": 14644 + }, + { + "epoch": 5.41, + "learning_rate": 2.460709446976229e-06, + "loss": 0.4537, + "step": 14645 + }, + { + "epoch": 5.41, + "learning_rate": 2.4606724966128836e-06, + "loss": 0.4651, + "step": 14646 + }, + { + "epoch": 5.41, + "learning_rate": 2.4606355462495382e-06, + "loss": 0.5143, + "step": 14647 + }, + { + "epoch": 5.41, + "learning_rate": 2.460598595886193e-06, + "loss": 0.5306, + "step": 14648 + }, + { + "epoch": 5.41, + "learning_rate": 2.4605616455228475e-06, + "loss": 0.4949, + "step": 14649 + }, + { + "epoch": 5.41, + "learning_rate": 2.4605246951595025e-06, + "loss": 0.4609, + "step": 14650 + }, + { + "epoch": 5.41, + "learning_rate": 2.460487744796157e-06, + "loss": 0.559, + "step": 14651 + }, + { + "epoch": 5.41, + "learning_rate": 2.4604507944328122e-06, + "loss": 0.401, + "step": 14652 + }, + { + "epoch": 5.41, + "learning_rate": 2.460413844069467e-06, + "loss": 0.5015, + "step": 14653 + }, + { + "epoch": 5.41, + "learning_rate": 2.4603768937061215e-06, + "loss": 0.4903, + "step": 14654 + }, + { + "epoch": 5.41, + "learning_rate": 2.460339943342776e-06, + "loss": 0.4086, + "step": 14655 + }, + { + "epoch": 5.41, + "learning_rate": 2.4603029929794312e-06, + "loss": 0.5078, + "step": 14656 + }, + { + "epoch": 5.41, + "learning_rate": 2.460266042616086e-06, + "loss": 0.4567, + "step": 14657 + }, + { + "epoch": 5.41, + "learning_rate": 2.4602290922527405e-06, + "loss": 0.4118, + "step": 14658 + }, + { + "epoch": 5.41, + "learning_rate": 2.4601921418893955e-06, + "loss": 0.5846, + "step": 14659 + }, + { + "epoch": 5.41, + "learning_rate": 2.4601551915260498e-06, + "loss": 0.3969, + "step": 14660 + }, + { + "epoch": 5.41, + "learning_rate": 2.460118241162705e-06, + "loss": 0.5605, + "step": 14661 + }, + { + "epoch": 5.41, + "learning_rate": 2.4600812907993595e-06, + "loss": 0.5808, + "step": 14662 + }, + { + "epoch": 5.41, + "learning_rate": 2.4600443404360145e-06, + "loss": 0.4509, + "step": 14663 + }, + { + "epoch": 5.42, + "learning_rate": 2.460007390072669e-06, + "loss": 0.4951, + "step": 14664 + }, + { + "epoch": 5.42, + "learning_rate": 2.4599704397093242e-06, + "loss": 0.5258, + "step": 14665 + }, + { + "epoch": 5.42, + "learning_rate": 2.4599334893459784e-06, + "loss": 0.5282, + "step": 14666 + }, + { + "epoch": 5.42, + "learning_rate": 2.459896538982633e-06, + "loss": 0.5676, + "step": 14667 + }, + { + "epoch": 5.42, + "learning_rate": 2.459859588619288e-06, + "loss": 0.4853, + "step": 14668 + }, + { + "epoch": 5.42, + "learning_rate": 2.4598226382559428e-06, + "loss": 0.5142, + "step": 14669 + }, + { + "epoch": 5.42, + "learning_rate": 2.459785687892598e-06, + "loss": 0.4889, + "step": 14670 + }, + { + "epoch": 5.42, + "learning_rate": 2.4597487375292525e-06, + "loss": 0.5639, + "step": 14671 + }, + { + "epoch": 5.42, + "learning_rate": 2.459711787165907e-06, + "loss": 0.4002, + "step": 14672 + }, + { + "epoch": 5.42, + "learning_rate": 2.4596748368025617e-06, + "loss": 0.4357, + "step": 14673 + }, + { + "epoch": 5.42, + "learning_rate": 2.4596378864392168e-06, + "loss": 0.5615, + "step": 14674 + }, + { + "epoch": 5.42, + "learning_rate": 2.4596009360758714e-06, + "loss": 0.5921, + "step": 14675 + }, + { + "epoch": 5.42, + "learning_rate": 2.4595639857125265e-06, + "loss": 0.5413, + "step": 14676 + }, + { + "epoch": 5.42, + "learning_rate": 2.459527035349181e-06, + "loss": 0.4177, + "step": 14677 + }, + { + "epoch": 5.42, + "learning_rate": 2.4594900849858358e-06, + "loss": 0.4733, + "step": 14678 + }, + { + "epoch": 5.42, + "learning_rate": 2.4594531346224904e-06, + "loss": 0.3841, + "step": 14679 + }, + { + "epoch": 5.42, + "learning_rate": 2.459416184259145e-06, + "loss": 0.5225, + "step": 14680 + }, + { + "epoch": 5.42, + "learning_rate": 2.4593792338958e-06, + "loss": 0.5465, + "step": 14681 + }, + { + "epoch": 5.42, + "learning_rate": 2.4593422835324547e-06, + "loss": 0.4777, + "step": 14682 + }, + { + "epoch": 5.42, + "learning_rate": 2.4593053331691098e-06, + "loss": 0.4971, + "step": 14683 + }, + { + "epoch": 5.42, + "learning_rate": 2.4592683828057644e-06, + "loss": 0.5361, + "step": 14684 + }, + { + "epoch": 5.42, + "learning_rate": 2.459231432442419e-06, + "loss": 0.5061, + "step": 14685 + }, + { + "epoch": 5.42, + "learning_rate": 2.4591944820790737e-06, + "loss": 0.4705, + "step": 14686 + }, + { + "epoch": 5.42, + "learning_rate": 2.4591575317157283e-06, + "loss": 0.4932, + "step": 14687 + }, + { + "epoch": 5.42, + "learning_rate": 2.4591205813523834e-06, + "loss": 0.5117, + "step": 14688 + }, + { + "epoch": 5.42, + "learning_rate": 2.459083630989038e-06, + "loss": 0.499, + "step": 14689 + }, + { + "epoch": 5.42, + "learning_rate": 2.459046680625693e-06, + "loss": 0.4021, + "step": 14690 + }, + { + "epoch": 5.43, + "learning_rate": 2.4590097302623477e-06, + "loss": 0.4779, + "step": 14691 + }, + { + "epoch": 5.43, + "learning_rate": 2.4589727798990024e-06, + "loss": 0.4649, + "step": 14692 + }, + { + "epoch": 5.43, + "learning_rate": 2.458935829535657e-06, + "loss": 0.442, + "step": 14693 + }, + { + "epoch": 5.43, + "learning_rate": 2.458898879172312e-06, + "loss": 0.3966, + "step": 14694 + }, + { + "epoch": 5.43, + "learning_rate": 2.4588619288089667e-06, + "loss": 0.4347, + "step": 14695 + }, + { + "epoch": 5.43, + "learning_rate": 2.4588249784456218e-06, + "loss": 0.4895, + "step": 14696 + }, + { + "epoch": 5.43, + "learning_rate": 2.4587880280822764e-06, + "loss": 0.4723, + "step": 14697 + }, + { + "epoch": 5.43, + "learning_rate": 2.458751077718931e-06, + "loss": 0.4464, + "step": 14698 + }, + { + "epoch": 5.43, + "learning_rate": 2.4587141273555857e-06, + "loss": 0.5396, + "step": 14699 + }, + { + "epoch": 5.43, + "learning_rate": 2.4586771769922403e-06, + "loss": 0.6322, + "step": 14700 + }, + { + "epoch": 5.43, + "learning_rate": 2.4586402266288954e-06, + "loss": 0.4851, + "step": 14701 + }, + { + "epoch": 5.43, + "learning_rate": 2.45860327626555e-06, + "loss": 0.5463, + "step": 14702 + }, + { + "epoch": 5.43, + "learning_rate": 2.458566325902205e-06, + "loss": 0.4528, + "step": 14703 + }, + { + "epoch": 5.43, + "learning_rate": 2.4585293755388597e-06, + "loss": 0.4694, + "step": 14704 + }, + { + "epoch": 5.43, + "learning_rate": 2.4584924251755143e-06, + "loss": 0.5796, + "step": 14705 + }, + { + "epoch": 5.43, + "learning_rate": 2.458455474812169e-06, + "loss": 0.4799, + "step": 14706 + }, + { + "epoch": 5.43, + "learning_rate": 2.4584185244488236e-06, + "loss": 0.4931, + "step": 14707 + }, + { + "epoch": 5.43, + "learning_rate": 2.4583815740854787e-06, + "loss": 0.5714, + "step": 14708 + }, + { + "epoch": 5.43, + "learning_rate": 2.4583446237221333e-06, + "loss": 0.4775, + "step": 14709 + }, + { + "epoch": 5.43, + "learning_rate": 2.458307673358788e-06, + "loss": 0.5102, + "step": 14710 + }, + { + "epoch": 5.43, + "learning_rate": 2.4582707229954426e-06, + "loss": 0.536, + "step": 14711 + }, + { + "epoch": 5.43, + "learning_rate": 2.4582337726320976e-06, + "loss": 0.4789, + "step": 14712 + }, + { + "epoch": 5.43, + "learning_rate": 2.4581968222687523e-06, + "loss": 0.4219, + "step": 14713 + }, + { + "epoch": 5.43, + "learning_rate": 2.4581598719054073e-06, + "loss": 0.3892, + "step": 14714 + }, + { + "epoch": 5.43, + "learning_rate": 2.458122921542062e-06, + "loss": 0.5311, + "step": 14715 + }, + { + "epoch": 5.43, + "learning_rate": 2.4580859711787166e-06, + "loss": 0.4438, + "step": 14716 + }, + { + "epoch": 5.43, + "learning_rate": 2.4580490208153712e-06, + "loss": 0.4502, + "step": 14717 + }, + { + "epoch": 5.44, + "learning_rate": 2.458012070452026e-06, + "loss": 0.4264, + "step": 14718 + }, + { + "epoch": 5.44, + "learning_rate": 2.457975120088681e-06, + "loss": 0.4746, + "step": 14719 + }, + { + "epoch": 5.44, + "learning_rate": 2.4579381697253356e-06, + "loss": 0.4778, + "step": 14720 + }, + { + "epoch": 5.44, + "learning_rate": 2.4579012193619906e-06, + "loss": 0.383, + "step": 14721 + }, + { + "epoch": 5.44, + "learning_rate": 2.4578642689986453e-06, + "loss": 0.6149, + "step": 14722 + }, + { + "epoch": 5.44, + "learning_rate": 2.4578273186353e-06, + "loss": 0.3604, + "step": 14723 + }, + { + "epoch": 5.44, + "learning_rate": 2.4577903682719545e-06, + "loss": 0.469, + "step": 14724 + }, + { + "epoch": 5.44, + "learning_rate": 2.4577534179086096e-06, + "loss": 0.5289, + "step": 14725 + }, + { + "epoch": 5.44, + "learning_rate": 2.4577164675452642e-06, + "loss": 0.5002, + "step": 14726 + }, + { + "epoch": 5.44, + "learning_rate": 2.457679517181919e-06, + "loss": 0.4156, + "step": 14727 + }, + { + "epoch": 5.44, + "learning_rate": 2.457642566818574e-06, + "loss": 0.5485, + "step": 14728 + }, + { + "epoch": 5.44, + "learning_rate": 2.4576056164552286e-06, + "loss": 0.464, + "step": 14729 + }, + { + "epoch": 5.44, + "learning_rate": 2.457568666091883e-06, + "loss": 0.4758, + "step": 14730 + }, + { + "epoch": 5.44, + "learning_rate": 2.457531715728538e-06, + "loss": 0.4769, + "step": 14731 + }, + { + "epoch": 5.44, + "learning_rate": 2.457494765365193e-06, + "loss": 0.6324, + "step": 14732 + }, + { + "epoch": 5.44, + "learning_rate": 2.4574578150018475e-06, + "loss": 0.4439, + "step": 14733 + }, + { + "epoch": 5.44, + "learning_rate": 2.4574208646385026e-06, + "loss": 0.4461, + "step": 14734 + }, + { + "epoch": 5.44, + "learning_rate": 2.4573839142751572e-06, + "loss": 0.5428, + "step": 14735 + }, + { + "epoch": 5.44, + "learning_rate": 2.457346963911812e-06, + "loss": 0.6056, + "step": 14736 + }, + { + "epoch": 5.44, + "learning_rate": 2.4573100135484665e-06, + "loss": 0.4345, + "step": 14737 + }, + { + "epoch": 5.44, + "learning_rate": 2.457273063185121e-06, + "loss": 0.6155, + "step": 14738 + }, + { + "epoch": 5.44, + "learning_rate": 2.457236112821776e-06, + "loss": 0.3964, + "step": 14739 + }, + { + "epoch": 5.44, + "learning_rate": 2.457199162458431e-06, + "loss": 0.4033, + "step": 14740 + }, + { + "epoch": 5.44, + "learning_rate": 2.457162212095086e-06, + "loss": 0.458, + "step": 14741 + }, + { + "epoch": 5.44, + "learning_rate": 2.4571252617317405e-06, + "loss": 0.6085, + "step": 14742 + }, + { + "epoch": 5.44, + "learning_rate": 2.457088311368395e-06, + "loss": 0.444, + "step": 14743 + }, + { + "epoch": 5.44, + "learning_rate": 2.45705136100505e-06, + "loss": 0.3961, + "step": 14744 + }, + { + "epoch": 5.44, + "learning_rate": 2.457014410641705e-06, + "loss": 0.5669, + "step": 14745 + }, + { + "epoch": 5.45, + "learning_rate": 2.4569774602783595e-06, + "loss": 0.4483, + "step": 14746 + }, + { + "epoch": 5.45, + "learning_rate": 2.456940509915014e-06, + "loss": 0.5042, + "step": 14747 + }, + { + "epoch": 5.45, + "learning_rate": 2.456903559551669e-06, + "loss": 0.4095, + "step": 14748 + }, + { + "epoch": 5.45, + "learning_rate": 2.4568666091883234e-06, + "loss": 0.4862, + "step": 14749 + }, + { + "epoch": 5.45, + "learning_rate": 2.4568296588249785e-06, + "loss": 0.493, + "step": 14750 + }, + { + "epoch": 5.45, + "learning_rate": 2.456792708461633e-06, + "loss": 0.5264, + "step": 14751 + }, + { + "epoch": 5.45, + "learning_rate": 2.456755758098288e-06, + "loss": 0.3993, + "step": 14752 + }, + { + "epoch": 5.45, + "learning_rate": 2.456718807734943e-06, + "loss": 0.3989, + "step": 14753 + }, + { + "epoch": 5.45, + "learning_rate": 2.456681857371598e-06, + "loss": 0.4552, + "step": 14754 + }, + { + "epoch": 5.45, + "learning_rate": 2.456644907008252e-06, + "loss": 0.4933, + "step": 14755 + }, + { + "epoch": 5.45, + "learning_rate": 2.456607956644907e-06, + "loss": 0.5834, + "step": 14756 + }, + { + "epoch": 5.45, + "learning_rate": 2.4565710062815618e-06, + "loss": 0.4609, + "step": 14757 + }, + { + "epoch": 5.45, + "learning_rate": 2.4565340559182164e-06, + "loss": 0.4962, + "step": 14758 + }, + { + "epoch": 5.45, + "learning_rate": 2.4564971055548715e-06, + "loss": 0.4445, + "step": 14759 + }, + { + "epoch": 5.45, + "learning_rate": 2.456460155191526e-06, + "loss": 0.4692, + "step": 14760 + }, + { + "epoch": 5.45, + "learning_rate": 2.4564232048281807e-06, + "loss": 0.499, + "step": 14761 + }, + { + "epoch": 5.45, + "learning_rate": 2.4563862544648354e-06, + "loss": 0.5611, + "step": 14762 + }, + { + "epoch": 5.45, + "learning_rate": 2.4563493041014904e-06, + "loss": 0.5377, + "step": 14763 + }, + { + "epoch": 5.45, + "learning_rate": 2.456312353738145e-06, + "loss": 0.4489, + "step": 14764 + }, + { + "epoch": 5.45, + "learning_rate": 2.4562754033748e-06, + "loss": 0.5241, + "step": 14765 + }, + { + "epoch": 5.45, + "learning_rate": 2.4562384530114548e-06, + "loss": 0.4586, + "step": 14766 + }, + { + "epoch": 5.45, + "learning_rate": 2.4562015026481094e-06, + "loss": 0.5198, + "step": 14767 + }, + { + "epoch": 5.45, + "learning_rate": 2.456164552284764e-06, + "loss": 0.5439, + "step": 14768 + }, + { + "epoch": 5.45, + "learning_rate": 2.4561276019214187e-06, + "loss": 0.3912, + "step": 14769 + }, + { + "epoch": 5.45, + "learning_rate": 2.4560906515580737e-06, + "loss": 0.6815, + "step": 14770 + }, + { + "epoch": 5.45, + "learning_rate": 2.4560537011947284e-06, + "loss": 0.4873, + "step": 14771 + }, + { + "epoch": 5.45, + "learning_rate": 2.4560167508313834e-06, + "loss": 0.3999, + "step": 14772 + }, + { + "epoch": 5.46, + "learning_rate": 2.455979800468038e-06, + "loss": 0.487, + "step": 14773 + }, + { + "epoch": 5.46, + "learning_rate": 2.4559428501046927e-06, + "loss": 0.4494, + "step": 14774 + }, + { + "epoch": 5.46, + "learning_rate": 2.4559058997413473e-06, + "loss": 0.5379, + "step": 14775 + }, + { + "epoch": 5.46, + "learning_rate": 2.4558689493780024e-06, + "loss": 0.4151, + "step": 14776 + }, + { + "epoch": 5.46, + "learning_rate": 2.455831999014657e-06, + "loss": 0.479, + "step": 14777 + }, + { + "epoch": 5.46, + "learning_rate": 2.4557950486513117e-06, + "loss": 0.4689, + "step": 14778 + }, + { + "epoch": 5.46, + "learning_rate": 2.4557580982879667e-06, + "loss": 0.4404, + "step": 14779 + }, + { + "epoch": 5.46, + "learning_rate": 2.4557211479246214e-06, + "loss": 0.4129, + "step": 14780 + }, + { + "epoch": 5.46, + "learning_rate": 2.455684197561276e-06, + "loss": 0.4303, + "step": 14781 + }, + { + "epoch": 5.46, + "learning_rate": 2.4556472471979307e-06, + "loss": 0.4147, + "step": 14782 + }, + { + "epoch": 5.46, + "learning_rate": 2.4556102968345857e-06, + "loss": 0.4406, + "step": 14783 + }, + { + "epoch": 5.46, + "learning_rate": 2.4555733464712403e-06, + "loss": 0.3422, + "step": 14784 + }, + { + "epoch": 5.46, + "learning_rate": 2.4555363961078954e-06, + "loss": 0.4398, + "step": 14785 + }, + { + "epoch": 5.46, + "learning_rate": 2.45549944574455e-06, + "loss": 0.4831, + "step": 14786 + }, + { + "epoch": 5.46, + "learning_rate": 2.4554624953812047e-06, + "loss": 0.5651, + "step": 14787 + }, + { + "epoch": 5.46, + "learning_rate": 2.4554255450178593e-06, + "loss": 0.55, + "step": 14788 + }, + { + "epoch": 5.46, + "learning_rate": 2.455388594654514e-06, + "loss": 0.4756, + "step": 14789 + }, + { + "epoch": 5.46, + "learning_rate": 2.455351644291169e-06, + "loss": 0.3826, + "step": 14790 + }, + { + "epoch": 5.46, + "learning_rate": 2.4553146939278236e-06, + "loss": 0.5217, + "step": 14791 + }, + { + "epoch": 5.46, + "learning_rate": 2.4552777435644787e-06, + "loss": 0.488, + "step": 14792 + }, + { + "epoch": 5.46, + "learning_rate": 2.4552407932011333e-06, + "loss": 0.5296, + "step": 14793 + }, + { + "epoch": 5.46, + "learning_rate": 2.455203842837788e-06, + "loss": 0.4966, + "step": 14794 + }, + { + "epoch": 5.46, + "learning_rate": 2.4551668924744426e-06, + "loss": 0.4973, + "step": 14795 + }, + { + "epoch": 5.46, + "learning_rate": 2.4551299421110977e-06, + "loss": 0.5815, + "step": 14796 + }, + { + "epoch": 5.46, + "learning_rate": 2.4550929917477523e-06, + "loss": 0.4917, + "step": 14797 + }, + { + "epoch": 5.46, + "learning_rate": 2.455056041384407e-06, + "loss": 0.5613, + "step": 14798 + }, + { + "epoch": 5.46, + "learning_rate": 2.455019091021062e-06, + "loss": 0.4388, + "step": 14799 + }, + { + "epoch": 5.47, + "learning_rate": 2.4549821406577162e-06, + "loss": 0.5101, + "step": 14800 + }, + { + "epoch": 5.47, + "learning_rate": 2.4549451902943713e-06, + "loss": 0.4094, + "step": 14801 + }, + { + "epoch": 5.47, + "learning_rate": 2.454908239931026e-06, + "loss": 0.372, + "step": 14802 + }, + { + "epoch": 5.47, + "learning_rate": 2.454871289567681e-06, + "loss": 0.5182, + "step": 14803 + }, + { + "epoch": 5.47, + "learning_rate": 2.4548343392043356e-06, + "loss": 0.5033, + "step": 14804 + }, + { + "epoch": 5.47, + "learning_rate": 2.4547973888409907e-06, + "loss": 0.5318, + "step": 14805 + }, + { + "epoch": 5.47, + "learning_rate": 2.454760438477645e-06, + "loss": 0.4245, + "step": 14806 + }, + { + "epoch": 5.47, + "learning_rate": 2.4547234881142995e-06, + "loss": 0.586, + "step": 14807 + }, + { + "epoch": 5.47, + "learning_rate": 2.4546865377509546e-06, + "loss": 0.441, + "step": 14808 + }, + { + "epoch": 5.47, + "learning_rate": 2.4546495873876092e-06, + "loss": 0.4456, + "step": 14809 + }, + { + "epoch": 5.47, + "learning_rate": 2.4546126370242643e-06, + "loss": 0.583, + "step": 14810 + }, + { + "epoch": 5.47, + "learning_rate": 2.454575686660919e-06, + "loss": 0.4929, + "step": 14811 + }, + { + "epoch": 5.47, + "learning_rate": 2.4545387362975736e-06, + "loss": 0.4528, + "step": 14812 + }, + { + "epoch": 5.47, + "learning_rate": 2.454501785934228e-06, + "loss": 0.5575, + "step": 14813 + }, + { + "epoch": 5.47, + "learning_rate": 2.4544648355708833e-06, + "loss": 0.5484, + "step": 14814 + }, + { + "epoch": 5.47, + "learning_rate": 2.454427885207538e-06, + "loss": 0.502, + "step": 14815 + }, + { + "epoch": 5.47, + "learning_rate": 2.454390934844193e-06, + "loss": 0.6339, + "step": 14816 + }, + { + "epoch": 5.47, + "learning_rate": 2.4543539844808476e-06, + "loss": 0.4692, + "step": 14817 + }, + { + "epoch": 5.47, + "learning_rate": 2.4543170341175022e-06, + "loss": 0.5776, + "step": 14818 + }, + { + "epoch": 5.47, + "learning_rate": 2.454280083754157e-06, + "loss": 0.5647, + "step": 14819 + }, + { + "epoch": 5.47, + "learning_rate": 2.4542431333908115e-06, + "loss": 0.5061, + "step": 14820 + }, + { + "epoch": 5.47, + "learning_rate": 2.4542061830274666e-06, + "loss": 0.4521, + "step": 14821 + }, + { + "epoch": 5.47, + "learning_rate": 2.454169232664121e-06, + "loss": 0.571, + "step": 14822 + }, + { + "epoch": 5.47, + "learning_rate": 2.4541322823007763e-06, + "loss": 0.4198, + "step": 14823 + }, + { + "epoch": 5.47, + "learning_rate": 2.454095331937431e-06, + "loss": 0.4161, + "step": 14824 + }, + { + "epoch": 5.47, + "learning_rate": 2.4540583815740855e-06, + "loss": 0.4427, + "step": 14825 + }, + { + "epoch": 5.47, + "learning_rate": 2.45402143121074e-06, + "loss": 0.5391, + "step": 14826 + }, + { + "epoch": 5.48, + "learning_rate": 2.453984480847395e-06, + "loss": 0.5568, + "step": 14827 + }, + { + "epoch": 5.48, + "learning_rate": 2.45394753048405e-06, + "loss": 0.4078, + "step": 14828 + }, + { + "epoch": 5.48, + "learning_rate": 2.4539105801207045e-06, + "loss": 0.3751, + "step": 14829 + }, + { + "epoch": 5.48, + "learning_rate": 2.4538736297573596e-06, + "loss": 0.4913, + "step": 14830 + }, + { + "epoch": 5.48, + "learning_rate": 2.453836679394014e-06, + "loss": 0.5223, + "step": 14831 + }, + { + "epoch": 5.48, + "learning_rate": 2.453799729030669e-06, + "loss": 0.4557, + "step": 14832 + }, + { + "epoch": 5.48, + "learning_rate": 2.4537627786673235e-06, + "loss": 0.5328, + "step": 14833 + }, + { + "epoch": 5.48, + "learning_rate": 2.4537258283039785e-06, + "loss": 0.5331, + "step": 14834 + }, + { + "epoch": 5.48, + "learning_rate": 2.453688877940633e-06, + "loss": 0.5007, + "step": 14835 + }, + { + "epoch": 5.48, + "learning_rate": 2.4536519275772882e-06, + "loss": 0.4627, + "step": 14836 + }, + { + "epoch": 5.48, + "learning_rate": 2.453614977213943e-06, + "loss": 0.5309, + "step": 14837 + }, + { + "epoch": 5.48, + "learning_rate": 2.4535780268505975e-06, + "loss": 0.49, + "step": 14838 + }, + { + "epoch": 5.48, + "learning_rate": 2.453541076487252e-06, + "loss": 0.5418, + "step": 14839 + }, + { + "epoch": 5.48, + "learning_rate": 2.4535041261239068e-06, + "loss": 0.5572, + "step": 14840 + }, + { + "epoch": 5.48, + "learning_rate": 2.453467175760562e-06, + "loss": 0.4716, + "step": 14841 + }, + { + "epoch": 5.48, + "learning_rate": 2.4534302253972165e-06, + "loss": 0.5154, + "step": 14842 + }, + { + "epoch": 5.48, + "learning_rate": 2.4533932750338715e-06, + "loss": 0.571, + "step": 14843 + }, + { + "epoch": 5.48, + "learning_rate": 2.453356324670526e-06, + "loss": 0.4128, + "step": 14844 + }, + { + "epoch": 5.48, + "learning_rate": 2.453319374307181e-06, + "loss": 0.6258, + "step": 14845 + }, + { + "epoch": 5.48, + "learning_rate": 2.4532824239438354e-06, + "loss": 0.508, + "step": 14846 + }, + { + "epoch": 5.48, + "learning_rate": 2.45324547358049e-06, + "loss": 0.4715, + "step": 14847 + }, + { + "epoch": 5.48, + "learning_rate": 2.453208523217145e-06, + "loss": 0.4583, + "step": 14848 + }, + { + "epoch": 5.48, + "learning_rate": 2.4531715728537998e-06, + "loss": 0.4359, + "step": 14849 + }, + { + "epoch": 5.48, + "learning_rate": 2.453134622490455e-06, + "loss": 0.4835, + "step": 14850 + }, + { + "epoch": 5.48, + "learning_rate": 2.453097672127109e-06, + "loss": 0.4846, + "step": 14851 + }, + { + "epoch": 5.48, + "learning_rate": 2.453060721763764e-06, + "loss": 0.4619, + "step": 14852 + }, + { + "epoch": 5.48, + "learning_rate": 2.4530237714004187e-06, + "loss": 0.5831, + "step": 14853 + }, + { + "epoch": 5.49, + "learning_rate": 2.452986821037074e-06, + "loss": 0.5108, + "step": 14854 + }, + { + "epoch": 5.49, + "learning_rate": 2.4529498706737284e-06, + "loss": 0.4599, + "step": 14855 + }, + { + "epoch": 5.49, + "learning_rate": 2.4529129203103835e-06, + "loss": 0.3889, + "step": 14856 + }, + { + "epoch": 5.49, + "learning_rate": 2.4528759699470377e-06, + "loss": 0.4565, + "step": 14857 + }, + { + "epoch": 5.49, + "learning_rate": 2.4528390195836923e-06, + "loss": 0.5655, + "step": 14858 + }, + { + "epoch": 5.49, + "learning_rate": 2.4528020692203474e-06, + "loss": 0.5679, + "step": 14859 + }, + { + "epoch": 5.49, + "learning_rate": 2.452765118857002e-06, + "loss": 0.525, + "step": 14860 + }, + { + "epoch": 5.49, + "learning_rate": 2.452728168493657e-06, + "loss": 0.5011, + "step": 14861 + }, + { + "epoch": 5.49, + "learning_rate": 2.4526912181303117e-06, + "loss": 0.4377, + "step": 14862 + }, + { + "epoch": 5.49, + "learning_rate": 2.4526542677669664e-06, + "loss": 0.4755, + "step": 14863 + }, + { + "epoch": 5.49, + "learning_rate": 2.452617317403621e-06, + "loss": 0.5907, + "step": 14864 + }, + { + "epoch": 5.49, + "learning_rate": 2.452580367040276e-06, + "loss": 0.4271, + "step": 14865 + }, + { + "epoch": 5.49, + "learning_rate": 2.4525434166769307e-06, + "loss": 0.4758, + "step": 14866 + }, + { + "epoch": 5.49, + "learning_rate": 2.4525064663135853e-06, + "loss": 0.4813, + "step": 14867 + }, + { + "epoch": 5.49, + "learning_rate": 2.4524695159502404e-06, + "loss": 0.5384, + "step": 14868 + }, + { + "epoch": 5.49, + "learning_rate": 2.452432565586895e-06, + "loss": 0.5338, + "step": 14869 + }, + { + "epoch": 5.49, + "learning_rate": 2.4523956152235497e-06, + "loss": 0.4569, + "step": 14870 + }, + { + "epoch": 5.49, + "learning_rate": 2.4523586648602043e-06, + "loss": 0.4979, + "step": 14871 + }, + { + "epoch": 5.49, + "learning_rate": 2.4523217144968594e-06, + "loss": 0.5184, + "step": 14872 + }, + { + "epoch": 5.49, + "learning_rate": 2.452284764133514e-06, + "loss": 0.4052, + "step": 14873 + }, + { + "epoch": 5.49, + "learning_rate": 2.452247813770169e-06, + "loss": 0.3627, + "step": 14874 + }, + { + "epoch": 5.49, + "learning_rate": 2.4522108634068237e-06, + "loss": 0.6294, + "step": 14875 + }, + { + "epoch": 5.49, + "learning_rate": 2.4521739130434783e-06, + "loss": 0.4262, + "step": 14876 + }, + { + "epoch": 5.49, + "learning_rate": 2.452136962680133e-06, + "loss": 0.3959, + "step": 14877 + }, + { + "epoch": 5.49, + "learning_rate": 2.4521000123167876e-06, + "loss": 0.3998, + "step": 14878 + }, + { + "epoch": 5.49, + "learning_rate": 2.4520630619534427e-06, + "loss": 0.4775, + "step": 14879 + }, + { + "epoch": 5.49, + "learning_rate": 2.4520261115900973e-06, + "loss": 0.4415, + "step": 14880 + }, + { + "epoch": 5.5, + "learning_rate": 2.4519891612267524e-06, + "loss": 0.4419, + "step": 14881 + }, + { + "epoch": 5.5, + "learning_rate": 2.451952210863407e-06, + "loss": 0.4917, + "step": 14882 + }, + { + "epoch": 5.5, + "learning_rate": 2.4519152605000616e-06, + "loss": 0.3844, + "step": 14883 + }, + { + "epoch": 5.5, + "learning_rate": 2.4518783101367163e-06, + "loss": 0.6036, + "step": 14884 + }, + { + "epoch": 5.5, + "learning_rate": 2.4518413597733713e-06, + "loss": 0.5186, + "step": 14885 + }, + { + "epoch": 5.5, + "learning_rate": 2.451804409410026e-06, + "loss": 0.6305, + "step": 14886 + }, + { + "epoch": 5.5, + "learning_rate": 2.4517674590466806e-06, + "loss": 0.3669, + "step": 14887 + }, + { + "epoch": 5.5, + "learning_rate": 2.4517305086833357e-06, + "loss": 0.503, + "step": 14888 + }, + { + "epoch": 5.5, + "learning_rate": 2.4516935583199903e-06, + "loss": 0.5038, + "step": 14889 + }, + { + "epoch": 5.5, + "learning_rate": 2.451656607956645e-06, + "loss": 0.6202, + "step": 14890 + }, + { + "epoch": 5.5, + "learning_rate": 2.4516196575932996e-06, + "loss": 0.4152, + "step": 14891 + }, + { + "epoch": 5.5, + "learning_rate": 2.4515827072299546e-06, + "loss": 0.3672, + "step": 14892 + }, + { + "epoch": 5.5, + "learning_rate": 2.4515457568666093e-06, + "loss": 0.5219, + "step": 14893 + }, + { + "epoch": 5.5, + "learning_rate": 2.4515088065032643e-06, + "loss": 0.5043, + "step": 14894 + }, + { + "epoch": 5.5, + "learning_rate": 2.4514718561399185e-06, + "loss": 0.4048, + "step": 14895 + }, + { + "epoch": 5.5, + "learning_rate": 2.4514349057765736e-06, + "loss": 0.5167, + "step": 14896 + }, + { + "epoch": 5.5, + "learning_rate": 2.4513979554132282e-06, + "loss": 0.5322, + "step": 14897 + }, + { + "epoch": 5.5, + "learning_rate": 2.451361005049883e-06, + "loss": 0.6732, + "step": 14898 + }, + { + "epoch": 5.5, + "learning_rate": 2.451324054686538e-06, + "loss": 0.4536, + "step": 14899 + }, + { + "epoch": 5.5, + "learning_rate": 2.4512871043231926e-06, + "loss": 0.4423, + "step": 14900 + }, + { + "epoch": 5.5, + "learning_rate": 2.451250153959847e-06, + "loss": 0.5218, + "step": 14901 + }, + { + "epoch": 5.5, + "learning_rate": 2.451213203596502e-06, + "loss": 0.4335, + "step": 14902 + }, + { + "epoch": 5.5, + "learning_rate": 2.451176253233157e-06, + "loss": 0.3923, + "step": 14903 + }, + { + "epoch": 5.5, + "learning_rate": 2.4511393028698115e-06, + "loss": 0.4704, + "step": 14904 + }, + { + "epoch": 5.5, + "learning_rate": 2.4511023525064666e-06, + "loss": 0.4032, + "step": 14905 + }, + { + "epoch": 5.5, + "learning_rate": 2.4510654021431212e-06, + "loss": 0.4193, + "step": 14906 + }, + { + "epoch": 5.5, + "learning_rate": 2.451028451779776e-06, + "loss": 0.4638, + "step": 14907 + }, + { + "epoch": 5.51, + "learning_rate": 2.4509915014164305e-06, + "loss": 0.4691, + "step": 14908 + }, + { + "epoch": 5.51, + "learning_rate": 2.450954551053085e-06, + "loss": 0.4981, + "step": 14909 + }, + { + "epoch": 5.51, + "learning_rate": 2.45091760068974e-06, + "loss": 0.5932, + "step": 14910 + }, + { + "epoch": 5.51, + "learning_rate": 2.450880650326395e-06, + "loss": 0.5222, + "step": 14911 + }, + { + "epoch": 5.51, + "learning_rate": 2.45084369996305e-06, + "loss": 0.4732, + "step": 14912 + }, + { + "epoch": 5.51, + "learning_rate": 2.4508067495997045e-06, + "loss": 0.4651, + "step": 14913 + }, + { + "epoch": 5.51, + "learning_rate": 2.450769799236359e-06, + "loss": 0.4842, + "step": 14914 + }, + { + "epoch": 5.51, + "learning_rate": 2.450732848873014e-06, + "loss": 0.5585, + "step": 14915 + }, + { + "epoch": 5.51, + "learning_rate": 2.4506958985096685e-06, + "loss": 0.5102, + "step": 14916 + }, + { + "epoch": 5.51, + "learning_rate": 2.4506589481463235e-06, + "loss": 0.4776, + "step": 14917 + }, + { + "epoch": 5.51, + "learning_rate": 2.450621997782978e-06, + "loss": 0.534, + "step": 14918 + }, + { + "epoch": 5.51, + "learning_rate": 2.450585047419633e-06, + "loss": 0.4891, + "step": 14919 + }, + { + "epoch": 5.51, + "learning_rate": 2.450548097056288e-06, + "loss": 0.572, + "step": 14920 + }, + { + "epoch": 5.51, + "learning_rate": 2.4505111466929425e-06, + "loss": 0.4901, + "step": 14921 + }, + { + "epoch": 5.51, + "learning_rate": 2.450474196329597e-06, + "loss": 0.4637, + "step": 14922 + }, + { + "epoch": 5.51, + "learning_rate": 2.450437245966252e-06, + "loss": 0.5028, + "step": 14923 + }, + { + "epoch": 5.51, + "learning_rate": 2.450400295602907e-06, + "loss": 0.4247, + "step": 14924 + }, + { + "epoch": 5.51, + "learning_rate": 2.450363345239562e-06, + "loss": 0.5558, + "step": 14925 + }, + { + "epoch": 5.51, + "learning_rate": 2.4503263948762165e-06, + "loss": 0.4384, + "step": 14926 + }, + { + "epoch": 5.51, + "learning_rate": 2.450289444512871e-06, + "loss": 0.644, + "step": 14927 + }, + { + "epoch": 5.51, + "learning_rate": 2.4502524941495258e-06, + "loss": 0.4811, + "step": 14928 + }, + { + "epoch": 5.51, + "learning_rate": 2.4502155437861804e-06, + "loss": 0.4446, + "step": 14929 + }, + { + "epoch": 5.51, + "learning_rate": 2.4501785934228355e-06, + "loss": 0.4295, + "step": 14930 + }, + { + "epoch": 5.51, + "learning_rate": 2.45014164305949e-06, + "loss": 0.5509, + "step": 14931 + }, + { + "epoch": 5.51, + "learning_rate": 2.450104692696145e-06, + "loss": 0.4208, + "step": 14932 + }, + { + "epoch": 5.51, + "learning_rate": 2.4500677423328e-06, + "loss": 0.4962, + "step": 14933 + }, + { + "epoch": 5.51, + "learning_rate": 2.4500307919694544e-06, + "loss": 0.4649, + "step": 14934 + }, + { + "epoch": 5.52, + "learning_rate": 2.449993841606109e-06, + "loss": 0.4687, + "step": 14935 + }, + { + "epoch": 5.52, + "learning_rate": 2.4499568912427637e-06, + "loss": 0.3828, + "step": 14936 + }, + { + "epoch": 5.52, + "learning_rate": 2.4499199408794188e-06, + "loss": 0.5104, + "step": 14937 + }, + { + "epoch": 5.52, + "learning_rate": 2.4498829905160734e-06, + "loss": 0.5072, + "step": 14938 + }, + { + "epoch": 5.52, + "learning_rate": 2.4498460401527285e-06, + "loss": 0.4753, + "step": 14939 + }, + { + "epoch": 5.52, + "learning_rate": 2.4498090897893827e-06, + "loss": 0.5987, + "step": 14940 + }, + { + "epoch": 5.52, + "learning_rate": 2.4497721394260377e-06, + "loss": 0.3449, + "step": 14941 + }, + { + "epoch": 5.52, + "learning_rate": 2.4497351890626924e-06, + "loss": 0.4875, + "step": 14942 + }, + { + "epoch": 5.52, + "learning_rate": 2.4496982386993474e-06, + "loss": 0.5787, + "step": 14943 + }, + { + "epoch": 5.52, + "learning_rate": 2.449661288336002e-06, + "loss": 0.6378, + "step": 14944 + }, + { + "epoch": 5.52, + "learning_rate": 2.449624337972657e-06, + "loss": 0.4213, + "step": 14945 + }, + { + "epoch": 5.52, + "learning_rate": 2.4495873876093114e-06, + "loss": 0.4895, + "step": 14946 + }, + { + "epoch": 5.52, + "learning_rate": 2.449550437245966e-06, + "loss": 0.4272, + "step": 14947 + }, + { + "epoch": 5.52, + "learning_rate": 2.449513486882621e-06, + "loss": 0.4592, + "step": 14948 + }, + { + "epoch": 5.52, + "learning_rate": 2.4494765365192757e-06, + "loss": 0.4874, + "step": 14949 + }, + { + "epoch": 5.52, + "learning_rate": 2.4494395861559307e-06, + "loss": 0.4764, + "step": 14950 + }, + { + "epoch": 5.52, + "learning_rate": 2.4494026357925854e-06, + "loss": 0.5282, + "step": 14951 + }, + { + "epoch": 5.52, + "learning_rate": 2.44936568542924e-06, + "loss": 0.5797, + "step": 14952 + }, + { + "epoch": 5.52, + "learning_rate": 2.4493287350658947e-06, + "loss": 0.486, + "step": 14953 + }, + { + "epoch": 5.52, + "learning_rate": 2.4492917847025497e-06, + "loss": 0.4943, + "step": 14954 + }, + { + "epoch": 5.52, + "learning_rate": 2.4492548343392044e-06, + "loss": 0.5623, + "step": 14955 + }, + { + "epoch": 5.52, + "learning_rate": 2.449217883975859e-06, + "loss": 0.5447, + "step": 14956 + }, + { + "epoch": 5.52, + "learning_rate": 2.449180933612514e-06, + "loss": 0.6301, + "step": 14957 + }, + { + "epoch": 5.52, + "learning_rate": 2.4491439832491687e-06, + "loss": 0.5004, + "step": 14958 + }, + { + "epoch": 5.52, + "learning_rate": 2.4491070328858233e-06, + "loss": 0.5003, + "step": 14959 + }, + { + "epoch": 5.52, + "learning_rate": 2.449070082522478e-06, + "loss": 0.4745, + "step": 14960 + }, + { + "epoch": 5.52, + "learning_rate": 2.449033132159133e-06, + "loss": 0.4903, + "step": 14961 + }, + { + "epoch": 5.53, + "learning_rate": 2.4489961817957877e-06, + "loss": 0.4354, + "step": 14962 + }, + { + "epoch": 5.53, + "learning_rate": 2.4489592314324427e-06, + "loss": 0.368, + "step": 14963 + }, + { + "epoch": 5.53, + "learning_rate": 2.4489222810690974e-06, + "loss": 0.4323, + "step": 14964 + }, + { + "epoch": 5.53, + "learning_rate": 2.448885330705752e-06, + "loss": 0.3606, + "step": 14965 + }, + { + "epoch": 5.53, + "learning_rate": 2.4488483803424066e-06, + "loss": 0.3638, + "step": 14966 + }, + { + "epoch": 5.53, + "learning_rate": 2.4488114299790613e-06, + "loss": 0.4787, + "step": 14967 + }, + { + "epoch": 5.53, + "learning_rate": 2.4487744796157163e-06, + "loss": 0.4627, + "step": 14968 + }, + { + "epoch": 5.53, + "learning_rate": 2.448737529252371e-06, + "loss": 0.4049, + "step": 14969 + }, + { + "epoch": 5.53, + "learning_rate": 2.448700578889026e-06, + "loss": 0.4835, + "step": 14970 + }, + { + "epoch": 5.53, + "learning_rate": 2.4486636285256807e-06, + "loss": 0.5028, + "step": 14971 + }, + { + "epoch": 5.53, + "learning_rate": 2.4486266781623353e-06, + "loss": 0.4571, + "step": 14972 + }, + { + "epoch": 5.53, + "learning_rate": 2.44858972779899e-06, + "loss": 0.4826, + "step": 14973 + }, + { + "epoch": 5.53, + "learning_rate": 2.448552777435645e-06, + "loss": 0.5192, + "step": 14974 + }, + { + "epoch": 5.53, + "learning_rate": 2.4485158270722996e-06, + "loss": 0.522, + "step": 14975 + }, + { + "epoch": 5.53, + "learning_rate": 2.4484788767089543e-06, + "loss": 0.4098, + "step": 14976 + }, + { + "epoch": 5.53, + "learning_rate": 2.4484419263456093e-06, + "loss": 0.4141, + "step": 14977 + }, + { + "epoch": 5.53, + "learning_rate": 2.448404975982264e-06, + "loss": 0.2947, + "step": 14978 + }, + { + "epoch": 5.53, + "learning_rate": 2.4483680256189186e-06, + "loss": 0.5064, + "step": 14979 + }, + { + "epoch": 5.53, + "learning_rate": 2.4483310752555732e-06, + "loss": 0.4608, + "step": 14980 + }, + { + "epoch": 5.53, + "learning_rate": 2.4482941248922283e-06, + "loss": 0.425, + "step": 14981 + }, + { + "epoch": 5.53, + "learning_rate": 2.448257174528883e-06, + "loss": 0.3837, + "step": 14982 + }, + { + "epoch": 5.53, + "learning_rate": 2.448220224165538e-06, + "loss": 0.4176, + "step": 14983 + }, + { + "epoch": 5.53, + "learning_rate": 2.4481832738021926e-06, + "loss": 0.5377, + "step": 14984 + }, + { + "epoch": 5.53, + "learning_rate": 2.4481463234388473e-06, + "loss": 0.4851, + "step": 14985 + }, + { + "epoch": 5.53, + "learning_rate": 2.448109373075502e-06, + "loss": 0.5113, + "step": 14986 + }, + { + "epoch": 5.53, + "learning_rate": 2.4480724227121565e-06, + "loss": 0.4299, + "step": 14987 + }, + { + "epoch": 5.53, + "learning_rate": 2.4480354723488116e-06, + "loss": 0.4114, + "step": 14988 + }, + { + "epoch": 5.54, + "learning_rate": 2.4479985219854662e-06, + "loss": 0.5002, + "step": 14989 + }, + { + "epoch": 5.54, + "learning_rate": 2.4479615716221213e-06, + "loss": 0.4024, + "step": 14990 + }, + { + "epoch": 5.54, + "learning_rate": 2.4479246212587755e-06, + "loss": 0.674, + "step": 14991 + }, + { + "epoch": 5.54, + "learning_rate": 2.4478876708954306e-06, + "loss": 0.5365, + "step": 14992 + }, + { + "epoch": 5.54, + "learning_rate": 2.447850720532085e-06, + "loss": 0.4407, + "step": 14993 + }, + { + "epoch": 5.54, + "learning_rate": 2.4478137701687403e-06, + "loss": 0.3906, + "step": 14994 + }, + { + "epoch": 5.54, + "learning_rate": 2.447776819805395e-06, + "loss": 0.4899, + "step": 14995 + }, + { + "epoch": 5.54, + "learning_rate": 2.4477398694420495e-06, + "loss": 0.49, + "step": 14996 + }, + { + "epoch": 5.54, + "learning_rate": 2.447702919078704e-06, + "loss": 0.4716, + "step": 14997 + }, + { + "epoch": 5.54, + "learning_rate": 2.447665968715359e-06, + "loss": 0.4599, + "step": 14998 + }, + { + "epoch": 5.54, + "learning_rate": 2.447629018352014e-06, + "loss": 0.5932, + "step": 14999 + }, + { + "epoch": 5.54, + "learning_rate": 2.4475920679886685e-06, + "loss": 0.5075, + "step": 15000 + }, + { + "epoch": 5.54, + "learning_rate": 2.4475551176253236e-06, + "loss": 0.5885, + "step": 15001 + }, + { + "epoch": 5.54, + "learning_rate": 2.447518167261978e-06, + "loss": 0.4789, + "step": 15002 + }, + { + "epoch": 5.54, + "learning_rate": 2.447481216898633e-06, + "loss": 0.5096, + "step": 15003 + }, + { + "epoch": 5.54, + "learning_rate": 2.4474442665352875e-06, + "loss": 0.4265, + "step": 15004 + }, + { + "epoch": 5.54, + "learning_rate": 2.4474073161719425e-06, + "loss": 0.4897, + "step": 15005 + }, + { + "epoch": 5.54, + "learning_rate": 2.447370365808597e-06, + "loss": 0.5495, + "step": 15006 + }, + { + "epoch": 5.54, + "learning_rate": 2.447333415445252e-06, + "loss": 0.5266, + "step": 15007 + }, + { + "epoch": 5.54, + "learning_rate": 2.447296465081907e-06, + "loss": 0.4318, + "step": 15008 + }, + { + "epoch": 5.54, + "learning_rate": 2.4472595147185615e-06, + "loss": 0.5236, + "step": 15009 + }, + { + "epoch": 5.54, + "learning_rate": 2.447222564355216e-06, + "loss": 0.5746, + "step": 15010 + }, + { + "epoch": 5.54, + "learning_rate": 2.4471856139918708e-06, + "loss": 0.4001, + "step": 15011 + }, + { + "epoch": 5.54, + "learning_rate": 2.447148663628526e-06, + "loss": 0.5899, + "step": 15012 + }, + { + "epoch": 5.54, + "learning_rate": 2.4471117132651805e-06, + "loss": 0.5341, + "step": 15013 + }, + { + "epoch": 5.54, + "learning_rate": 2.4470747629018355e-06, + "loss": 0.427, + "step": 15014 + }, + { + "epoch": 5.54, + "learning_rate": 2.44703781253849e-06, + "loss": 0.4866, + "step": 15015 + }, + { + "epoch": 5.55, + "learning_rate": 2.447000862175145e-06, + "loss": 0.4435, + "step": 15016 + }, + { + "epoch": 5.55, + "learning_rate": 2.4469639118117994e-06, + "loss": 0.5342, + "step": 15017 + }, + { + "epoch": 5.55, + "learning_rate": 2.446926961448454e-06, + "loss": 0.6884, + "step": 15018 + }, + { + "epoch": 5.55, + "learning_rate": 2.446890011085109e-06, + "loss": 0.5462, + "step": 15019 + }, + { + "epoch": 5.55, + "learning_rate": 2.4468530607217638e-06, + "loss": 0.5293, + "step": 15020 + }, + { + "epoch": 5.55, + "learning_rate": 2.446816110358419e-06, + "loss": 0.479, + "step": 15021 + }, + { + "epoch": 5.55, + "learning_rate": 2.4467791599950735e-06, + "loss": 0.4918, + "step": 15022 + }, + { + "epoch": 5.55, + "learning_rate": 2.446742209631728e-06, + "loss": 0.5965, + "step": 15023 + }, + { + "epoch": 5.55, + "learning_rate": 2.4467052592683827e-06, + "loss": 0.4588, + "step": 15024 + }, + { + "epoch": 5.55, + "learning_rate": 2.446668308905038e-06, + "loss": 0.4797, + "step": 15025 + }, + { + "epoch": 5.55, + "learning_rate": 2.4466313585416924e-06, + "loss": 0.4389, + "step": 15026 + }, + { + "epoch": 5.55, + "learning_rate": 2.446594408178347e-06, + "loss": 0.5477, + "step": 15027 + }, + { + "epoch": 5.55, + "learning_rate": 2.446557457815002e-06, + "loss": 0.6097, + "step": 15028 + }, + { + "epoch": 5.55, + "learning_rate": 2.4465205074516568e-06, + "loss": 0.5481, + "step": 15029 + }, + { + "epoch": 5.55, + "learning_rate": 2.4464835570883114e-06, + "loss": 0.4488, + "step": 15030 + }, + { + "epoch": 5.55, + "learning_rate": 2.446446606724966e-06, + "loss": 0.4015, + "step": 15031 + }, + { + "epoch": 5.55, + "learning_rate": 2.446409656361621e-06, + "loss": 0.439, + "step": 15032 + }, + { + "epoch": 5.55, + "learning_rate": 2.4463727059982757e-06, + "loss": 0.4284, + "step": 15033 + }, + { + "epoch": 5.55, + "learning_rate": 2.446335755634931e-06, + "loss": 0.4531, + "step": 15034 + }, + { + "epoch": 5.55, + "learning_rate": 2.4462988052715854e-06, + "loss": 0.553, + "step": 15035 + }, + { + "epoch": 5.55, + "learning_rate": 2.4462618549082396e-06, + "loss": 0.434, + "step": 15036 + }, + { + "epoch": 5.55, + "learning_rate": 2.4462249045448947e-06, + "loss": 0.4743, + "step": 15037 + }, + { + "epoch": 5.55, + "learning_rate": 2.4461879541815493e-06, + "loss": 0.5815, + "step": 15038 + }, + { + "epoch": 5.55, + "learning_rate": 2.4461510038182044e-06, + "loss": 0.4668, + "step": 15039 + }, + { + "epoch": 5.55, + "learning_rate": 2.446114053454859e-06, + "loss": 0.4902, + "step": 15040 + }, + { + "epoch": 5.55, + "learning_rate": 2.446077103091514e-06, + "loss": 0.5688, + "step": 15041 + }, + { + "epoch": 5.55, + "learning_rate": 2.4460401527281683e-06, + "loss": 0.5157, + "step": 15042 + }, + { + "epoch": 5.56, + "learning_rate": 2.4460032023648234e-06, + "loss": 0.4087, + "step": 15043 + }, + { + "epoch": 5.56, + "learning_rate": 2.445966252001478e-06, + "loss": 0.4957, + "step": 15044 + }, + { + "epoch": 5.56, + "learning_rate": 2.445929301638133e-06, + "loss": 0.5549, + "step": 15045 + }, + { + "epoch": 5.56, + "learning_rate": 2.4458923512747877e-06, + "loss": 0.525, + "step": 15046 + }, + { + "epoch": 5.56, + "learning_rate": 2.4458554009114423e-06, + "loss": 0.4919, + "step": 15047 + }, + { + "epoch": 5.56, + "learning_rate": 2.445818450548097e-06, + "loss": 0.4573, + "step": 15048 + }, + { + "epoch": 5.56, + "learning_rate": 2.4457815001847516e-06, + "loss": 0.471, + "step": 15049 + }, + { + "epoch": 5.56, + "learning_rate": 2.4457445498214067e-06, + "loss": 0.4759, + "step": 15050 + }, + { + "epoch": 5.56, + "learning_rate": 2.4457075994580613e-06, + "loss": 0.428, + "step": 15051 + }, + { + "epoch": 5.56, + "learning_rate": 2.4456706490947164e-06, + "loss": 0.5815, + "step": 15052 + }, + { + "epoch": 5.56, + "learning_rate": 2.445633698731371e-06, + "loss": 0.3973, + "step": 15053 + }, + { + "epoch": 5.56, + "learning_rate": 2.4455967483680256e-06, + "loss": 0.5105, + "step": 15054 + }, + { + "epoch": 5.56, + "learning_rate": 2.4455597980046803e-06, + "loss": 0.5396, + "step": 15055 + }, + { + "epoch": 5.56, + "learning_rate": 2.445522847641335e-06, + "loss": 0.5327, + "step": 15056 + }, + { + "epoch": 5.56, + "learning_rate": 2.44548589727799e-06, + "loss": 0.3973, + "step": 15057 + }, + { + "epoch": 5.56, + "learning_rate": 2.4454489469146446e-06, + "loss": 0.4829, + "step": 15058 + }, + { + "epoch": 5.56, + "learning_rate": 2.4454119965512997e-06, + "loss": 0.4949, + "step": 15059 + }, + { + "epoch": 5.56, + "learning_rate": 2.4453750461879543e-06, + "loss": 0.4913, + "step": 15060 + }, + { + "epoch": 5.56, + "learning_rate": 2.445338095824609e-06, + "loss": 0.4845, + "step": 15061 + }, + { + "epoch": 5.56, + "learning_rate": 2.4453011454612636e-06, + "loss": 0.4667, + "step": 15062 + }, + { + "epoch": 5.56, + "learning_rate": 2.4452641950979186e-06, + "loss": 0.4387, + "step": 15063 + }, + { + "epoch": 5.56, + "learning_rate": 2.4452272447345733e-06, + "loss": 0.3938, + "step": 15064 + }, + { + "epoch": 5.56, + "learning_rate": 2.4451902943712283e-06, + "loss": 0.4381, + "step": 15065 + }, + { + "epoch": 5.56, + "learning_rate": 2.445153344007883e-06, + "loss": 0.412, + "step": 15066 + }, + { + "epoch": 5.56, + "learning_rate": 2.4451163936445376e-06, + "loss": 0.3988, + "step": 15067 + }, + { + "epoch": 5.56, + "learning_rate": 2.4450794432811922e-06, + "loss": 0.6081, + "step": 15068 + }, + { + "epoch": 5.56, + "learning_rate": 2.445042492917847e-06, + "loss": 0.5158, + "step": 15069 + }, + { + "epoch": 5.56, + "learning_rate": 2.445005542554502e-06, + "loss": 0.5161, + "step": 15070 + }, + { + "epoch": 5.57, + "learning_rate": 2.4449685921911566e-06, + "loss": 0.4422, + "step": 15071 + }, + { + "epoch": 5.57, + "learning_rate": 2.4449316418278116e-06, + "loss": 0.4466, + "step": 15072 + }, + { + "epoch": 5.57, + "learning_rate": 2.4448946914644663e-06, + "loss": 0.5278, + "step": 15073 + }, + { + "epoch": 5.57, + "learning_rate": 2.444857741101121e-06, + "loss": 0.5374, + "step": 15074 + }, + { + "epoch": 5.57, + "learning_rate": 2.4448207907377755e-06, + "loss": 0.5547, + "step": 15075 + }, + { + "epoch": 5.57, + "learning_rate": 2.44478384037443e-06, + "loss": 0.436, + "step": 15076 + }, + { + "epoch": 5.57, + "learning_rate": 2.4447468900110852e-06, + "loss": 0.561, + "step": 15077 + }, + { + "epoch": 5.57, + "learning_rate": 2.44470993964774e-06, + "loss": 0.5022, + "step": 15078 + }, + { + "epoch": 5.57, + "learning_rate": 2.444672989284395e-06, + "loss": 0.3952, + "step": 15079 + }, + { + "epoch": 5.57, + "learning_rate": 2.444636038921049e-06, + "loss": 0.5411, + "step": 15080 + }, + { + "epoch": 5.57, + "learning_rate": 2.4445990885577042e-06, + "loss": 0.4242, + "step": 15081 + }, + { + "epoch": 5.57, + "learning_rate": 2.444562138194359e-06, + "loss": 0.644, + "step": 15082 + }, + { + "epoch": 5.57, + "learning_rate": 2.444525187831014e-06, + "loss": 0.4825, + "step": 15083 + }, + { + "epoch": 5.57, + "learning_rate": 2.4444882374676685e-06, + "loss": 0.5028, + "step": 15084 + }, + { + "epoch": 5.57, + "learning_rate": 2.4444512871043236e-06, + "loss": 0.4831, + "step": 15085 + }, + { + "epoch": 5.57, + "learning_rate": 2.444414336740978e-06, + "loss": 0.4483, + "step": 15086 + }, + { + "epoch": 5.57, + "learning_rate": 2.4443773863776325e-06, + "loss": 0.5509, + "step": 15087 + }, + { + "epoch": 5.57, + "learning_rate": 2.4443404360142875e-06, + "loss": 0.5502, + "step": 15088 + }, + { + "epoch": 5.57, + "learning_rate": 2.444303485650942e-06, + "loss": 0.5519, + "step": 15089 + }, + { + "epoch": 5.57, + "learning_rate": 2.4442665352875972e-06, + "loss": 0.4424, + "step": 15090 + }, + { + "epoch": 5.57, + "learning_rate": 2.444229584924252e-06, + "loss": 0.5492, + "step": 15091 + }, + { + "epoch": 5.57, + "learning_rate": 2.4441926345609065e-06, + "loss": 0.4016, + "step": 15092 + }, + { + "epoch": 5.57, + "learning_rate": 2.444155684197561e-06, + "loss": 0.5334, + "step": 15093 + }, + { + "epoch": 5.57, + "learning_rate": 2.444118733834216e-06, + "loss": 0.4983, + "step": 15094 + }, + { + "epoch": 5.57, + "learning_rate": 2.444081783470871e-06, + "loss": 0.6294, + "step": 15095 + }, + { + "epoch": 5.57, + "learning_rate": 2.4440448331075255e-06, + "loss": 0.5049, + "step": 15096 + }, + { + "epoch": 5.57, + "learning_rate": 2.4440078827441805e-06, + "loss": 0.4686, + "step": 15097 + }, + { + "epoch": 5.58, + "learning_rate": 2.443970932380835e-06, + "loss": 0.4595, + "step": 15098 + }, + { + "epoch": 5.58, + "learning_rate": 2.4439339820174898e-06, + "loss": 0.5421, + "step": 15099 + }, + { + "epoch": 5.58, + "learning_rate": 2.4438970316541444e-06, + "loss": 0.3847, + "step": 15100 + }, + { + "epoch": 5.58, + "learning_rate": 2.4438600812907995e-06, + "loss": 0.4788, + "step": 15101 + }, + { + "epoch": 5.58, + "learning_rate": 2.443823130927454e-06, + "loss": 0.4445, + "step": 15102 + }, + { + "epoch": 5.58, + "learning_rate": 2.443786180564109e-06, + "loss": 0.4775, + "step": 15103 + }, + { + "epoch": 5.58, + "learning_rate": 2.443749230200764e-06, + "loss": 0.3864, + "step": 15104 + }, + { + "epoch": 5.58, + "learning_rate": 2.4437122798374185e-06, + "loss": 0.5537, + "step": 15105 + }, + { + "epoch": 5.58, + "learning_rate": 2.443675329474073e-06, + "loss": 0.4379, + "step": 15106 + }, + { + "epoch": 5.58, + "learning_rate": 2.4436383791107277e-06, + "loss": 0.4994, + "step": 15107 + }, + { + "epoch": 5.58, + "learning_rate": 2.4436014287473828e-06, + "loss": 0.3866, + "step": 15108 + }, + { + "epoch": 5.58, + "learning_rate": 2.4435644783840374e-06, + "loss": 0.4889, + "step": 15109 + }, + { + "epoch": 5.58, + "learning_rate": 2.4435275280206925e-06, + "loss": 0.4301, + "step": 15110 + }, + { + "epoch": 5.58, + "learning_rate": 2.443490577657347e-06, + "loss": 0.432, + "step": 15111 + }, + { + "epoch": 5.58, + "learning_rate": 2.4434536272940018e-06, + "loss": 0.4402, + "step": 15112 + }, + { + "epoch": 5.58, + "learning_rate": 2.4434166769306564e-06, + "loss": 0.3835, + "step": 15113 + }, + { + "epoch": 5.58, + "learning_rate": 2.4433797265673115e-06, + "loss": 0.4416, + "step": 15114 + }, + { + "epoch": 5.58, + "learning_rate": 2.443342776203966e-06, + "loss": 0.4232, + "step": 15115 + }, + { + "epoch": 5.58, + "learning_rate": 2.4433058258406207e-06, + "loss": 0.4319, + "step": 15116 + }, + { + "epoch": 5.58, + "learning_rate": 2.4432688754772758e-06, + "loss": 0.4759, + "step": 15117 + }, + { + "epoch": 5.58, + "learning_rate": 2.4432319251139304e-06, + "loss": 0.4774, + "step": 15118 + }, + { + "epoch": 5.58, + "learning_rate": 2.443194974750585e-06, + "loss": 0.4077, + "step": 15119 + }, + { + "epoch": 5.58, + "learning_rate": 2.4431580243872397e-06, + "loss": 0.4971, + "step": 15120 + }, + { + "epoch": 5.58, + "learning_rate": 2.4431210740238948e-06, + "loss": 0.4478, + "step": 15121 + }, + { + "epoch": 5.58, + "learning_rate": 2.4430841236605494e-06, + "loss": 0.5298, + "step": 15122 + }, + { + "epoch": 5.58, + "learning_rate": 2.4430471732972044e-06, + "loss": 0.4787, + "step": 15123 + }, + { + "epoch": 5.58, + "learning_rate": 2.443010222933859e-06, + "loss": 0.4989, + "step": 15124 + }, + { + "epoch": 5.59, + "learning_rate": 2.4429732725705137e-06, + "loss": 0.4871, + "step": 15125 + }, + { + "epoch": 5.59, + "learning_rate": 2.4429363222071684e-06, + "loss": 0.4986, + "step": 15126 + }, + { + "epoch": 5.59, + "learning_rate": 2.442899371843823e-06, + "loss": 0.4566, + "step": 15127 + }, + { + "epoch": 5.59, + "learning_rate": 2.442862421480478e-06, + "loss": 0.4452, + "step": 15128 + }, + { + "epoch": 5.59, + "learning_rate": 2.4428254711171327e-06, + "loss": 0.5269, + "step": 15129 + }, + { + "epoch": 5.59, + "learning_rate": 2.4427885207537878e-06, + "loss": 0.545, + "step": 15130 + }, + { + "epoch": 5.59, + "learning_rate": 2.442751570390442e-06, + "loss": 0.5595, + "step": 15131 + }, + { + "epoch": 5.59, + "learning_rate": 2.442714620027097e-06, + "loss": 0.4722, + "step": 15132 + }, + { + "epoch": 5.59, + "learning_rate": 2.4426776696637517e-06, + "loss": 0.4895, + "step": 15133 + }, + { + "epoch": 5.59, + "learning_rate": 2.4426407193004067e-06, + "loss": 0.5521, + "step": 15134 + }, + { + "epoch": 5.59, + "learning_rate": 2.4426037689370614e-06, + "loss": 0.3859, + "step": 15135 + }, + { + "epoch": 5.59, + "learning_rate": 2.442566818573716e-06, + "loss": 0.6154, + "step": 15136 + }, + { + "epoch": 5.59, + "learning_rate": 2.4425298682103706e-06, + "loss": 0.496, + "step": 15137 + }, + { + "epoch": 5.59, + "learning_rate": 2.4424929178470253e-06, + "loss": 0.5611, + "step": 15138 + }, + { + "epoch": 5.59, + "learning_rate": 2.4424559674836803e-06, + "loss": 0.5227, + "step": 15139 + }, + { + "epoch": 5.59, + "learning_rate": 2.442419017120335e-06, + "loss": 0.4924, + "step": 15140 + }, + { + "epoch": 5.59, + "learning_rate": 2.44238206675699e-06, + "loss": 0.3985, + "step": 15141 + }, + { + "epoch": 5.59, + "learning_rate": 2.4423451163936447e-06, + "loss": 0.4482, + "step": 15142 + }, + { + "epoch": 5.59, + "learning_rate": 2.4423081660302993e-06, + "loss": 0.4599, + "step": 15143 + }, + { + "epoch": 5.59, + "learning_rate": 2.442271215666954e-06, + "loss": 0.4936, + "step": 15144 + }, + { + "epoch": 5.59, + "learning_rate": 2.4422342653036086e-06, + "loss": 0.4755, + "step": 15145 + }, + { + "epoch": 5.59, + "learning_rate": 2.4421973149402636e-06, + "loss": 0.4551, + "step": 15146 + }, + { + "epoch": 5.59, + "learning_rate": 2.4421603645769183e-06, + "loss": 0.3816, + "step": 15147 + }, + { + "epoch": 5.59, + "learning_rate": 2.4421234142135733e-06, + "loss": 0.4318, + "step": 15148 + }, + { + "epoch": 5.59, + "learning_rate": 2.442086463850228e-06, + "loss": 0.5834, + "step": 15149 + }, + { + "epoch": 5.59, + "learning_rate": 2.4420495134868826e-06, + "loss": 0.4476, + "step": 15150 + }, + { + "epoch": 5.59, + "learning_rate": 2.4420125631235372e-06, + "loss": 0.4492, + "step": 15151 + }, + { + "epoch": 5.6, + "learning_rate": 2.4419756127601923e-06, + "loss": 0.436, + "step": 15152 + }, + { + "epoch": 5.6, + "learning_rate": 2.441938662396847e-06, + "loss": 0.5596, + "step": 15153 + }, + { + "epoch": 5.6, + "learning_rate": 2.441901712033502e-06, + "loss": 0.4566, + "step": 15154 + }, + { + "epoch": 5.6, + "learning_rate": 2.4418647616701566e-06, + "loss": 0.4573, + "step": 15155 + }, + { + "epoch": 5.6, + "learning_rate": 2.4418278113068113e-06, + "loss": 0.4218, + "step": 15156 + }, + { + "epoch": 5.6, + "learning_rate": 2.441790860943466e-06, + "loss": 0.4134, + "step": 15157 + }, + { + "epoch": 5.6, + "learning_rate": 2.4417539105801205e-06, + "loss": 0.4532, + "step": 15158 + }, + { + "epoch": 5.6, + "learning_rate": 2.4417169602167756e-06, + "loss": 0.5191, + "step": 15159 + }, + { + "epoch": 5.6, + "learning_rate": 2.4416800098534302e-06, + "loss": 0.5346, + "step": 15160 + }, + { + "epoch": 5.6, + "learning_rate": 2.4416430594900853e-06, + "loss": 0.4627, + "step": 15161 + }, + { + "epoch": 5.6, + "learning_rate": 2.44160610912674e-06, + "loss": 0.4717, + "step": 15162 + }, + { + "epoch": 5.6, + "learning_rate": 2.4415691587633946e-06, + "loss": 0.5362, + "step": 15163 + }, + { + "epoch": 5.6, + "learning_rate": 2.441532208400049e-06, + "loss": 0.5818, + "step": 15164 + }, + { + "epoch": 5.6, + "learning_rate": 2.441495258036704e-06, + "loss": 0.533, + "step": 15165 + }, + { + "epoch": 5.6, + "learning_rate": 2.441458307673359e-06, + "loss": 0.442, + "step": 15166 + }, + { + "epoch": 5.6, + "learning_rate": 2.4414213573100135e-06, + "loss": 0.5308, + "step": 15167 + }, + { + "epoch": 5.6, + "learning_rate": 2.4413844069466686e-06, + "loss": 0.5123, + "step": 15168 + }, + { + "epoch": 5.6, + "learning_rate": 2.4413474565833232e-06, + "loss": 0.3468, + "step": 15169 + }, + { + "epoch": 5.6, + "learning_rate": 2.441310506219978e-06, + "loss": 0.4149, + "step": 15170 + }, + { + "epoch": 5.6, + "learning_rate": 2.4412735558566325e-06, + "loss": 0.4514, + "step": 15171 + }, + { + "epoch": 5.6, + "learning_rate": 2.4412366054932876e-06, + "loss": 0.4876, + "step": 15172 + }, + { + "epoch": 5.6, + "learning_rate": 2.441199655129942e-06, + "loss": 0.5256, + "step": 15173 + }, + { + "epoch": 5.6, + "learning_rate": 2.4411627047665973e-06, + "loss": 0.4839, + "step": 15174 + }, + { + "epoch": 5.6, + "learning_rate": 2.441125754403252e-06, + "loss": 0.517, + "step": 15175 + }, + { + "epoch": 5.6, + "learning_rate": 2.441088804039906e-06, + "loss": 0.4674, + "step": 15176 + }, + { + "epoch": 5.6, + "learning_rate": 2.441051853676561e-06, + "loss": 0.5028, + "step": 15177 + }, + { + "epoch": 5.6, + "learning_rate": 2.441014903313216e-06, + "loss": 0.4362, + "step": 15178 + }, + { + "epoch": 5.61, + "learning_rate": 2.440977952949871e-06, + "loss": 0.456, + "step": 15179 + }, + { + "epoch": 5.61, + "learning_rate": 2.4409410025865255e-06, + "loss": 0.4573, + "step": 15180 + }, + { + "epoch": 5.61, + "learning_rate": 2.4409040522231806e-06, + "loss": 0.4505, + "step": 15181 + }, + { + "epoch": 5.61, + "learning_rate": 2.4408671018598348e-06, + "loss": 0.6261, + "step": 15182 + }, + { + "epoch": 5.61, + "learning_rate": 2.44083015149649e-06, + "loss": 0.5016, + "step": 15183 + }, + { + "epoch": 5.61, + "learning_rate": 2.4407932011331445e-06, + "loss": 0.5272, + "step": 15184 + }, + { + "epoch": 5.61, + "learning_rate": 2.440756250769799e-06, + "loss": 0.5419, + "step": 15185 + }, + { + "epoch": 5.61, + "learning_rate": 2.440719300406454e-06, + "loss": 0.5641, + "step": 15186 + }, + { + "epoch": 5.61, + "learning_rate": 2.440682350043109e-06, + "loss": 0.3799, + "step": 15187 + }, + { + "epoch": 5.61, + "learning_rate": 2.4406453996797634e-06, + "loss": 0.4128, + "step": 15188 + }, + { + "epoch": 5.61, + "learning_rate": 2.440608449316418e-06, + "loss": 0.5058, + "step": 15189 + }, + { + "epoch": 5.61, + "learning_rate": 2.440571498953073e-06, + "loss": 0.4182, + "step": 15190 + }, + { + "epoch": 5.61, + "learning_rate": 2.4405345485897278e-06, + "loss": 0.4614, + "step": 15191 + }, + { + "epoch": 5.61, + "learning_rate": 2.440497598226383e-06, + "loss": 0.5257, + "step": 15192 + }, + { + "epoch": 5.61, + "learning_rate": 2.4404606478630375e-06, + "loss": 0.5179, + "step": 15193 + }, + { + "epoch": 5.61, + "learning_rate": 2.440423697499692e-06, + "loss": 0.5388, + "step": 15194 + }, + { + "epoch": 5.61, + "learning_rate": 2.4403867471363467e-06, + "loss": 0.4472, + "step": 15195 + }, + { + "epoch": 5.61, + "learning_rate": 2.4403497967730014e-06, + "loss": 0.5173, + "step": 15196 + }, + { + "epoch": 5.61, + "learning_rate": 2.4403128464096564e-06, + "loss": 0.6001, + "step": 15197 + }, + { + "epoch": 5.61, + "learning_rate": 2.440275896046311e-06, + "loss": 0.4346, + "step": 15198 + }, + { + "epoch": 5.61, + "learning_rate": 2.440238945682966e-06, + "loss": 0.529, + "step": 15199 + }, + { + "epoch": 5.61, + "learning_rate": 2.4402019953196208e-06, + "loss": 0.4278, + "step": 15200 + }, + { + "epoch": 5.61, + "learning_rate": 2.4401650449562754e-06, + "loss": 0.5801, + "step": 15201 + }, + { + "epoch": 5.61, + "learning_rate": 2.44012809459293e-06, + "loss": 0.5691, + "step": 15202 + }, + { + "epoch": 5.61, + "learning_rate": 2.440091144229585e-06, + "loss": 0.448, + "step": 15203 + }, + { + "epoch": 5.61, + "learning_rate": 2.4400541938662397e-06, + "loss": 0.5712, + "step": 15204 + }, + { + "epoch": 5.61, + "learning_rate": 2.4400172435028944e-06, + "loss": 0.5103, + "step": 15205 + }, + { + "epoch": 5.62, + "learning_rate": 2.4399802931395494e-06, + "loss": 0.5045, + "step": 15206 + }, + { + "epoch": 5.62, + "learning_rate": 2.439943342776204e-06, + "loss": 0.4843, + "step": 15207 + }, + { + "epoch": 5.62, + "learning_rate": 2.4399063924128587e-06, + "loss": 0.4424, + "step": 15208 + }, + { + "epoch": 5.62, + "learning_rate": 2.4398694420495133e-06, + "loss": 0.4159, + "step": 15209 + }, + { + "epoch": 5.62, + "learning_rate": 2.4398324916861684e-06, + "loss": 0.4493, + "step": 15210 + }, + { + "epoch": 5.62, + "learning_rate": 2.439795541322823e-06, + "loss": 0.5036, + "step": 15211 + }, + { + "epoch": 5.62, + "learning_rate": 2.439758590959478e-06, + "loss": 0.5112, + "step": 15212 + }, + { + "epoch": 5.62, + "learning_rate": 2.4397216405961327e-06, + "loss": 0.4888, + "step": 15213 + }, + { + "epoch": 5.62, + "learning_rate": 2.4396846902327874e-06, + "loss": 0.5273, + "step": 15214 + }, + { + "epoch": 5.62, + "learning_rate": 2.439647739869442e-06, + "loss": 0.5399, + "step": 15215 + }, + { + "epoch": 5.62, + "learning_rate": 2.4396107895060966e-06, + "loss": 0.4734, + "step": 15216 + }, + { + "epoch": 5.62, + "learning_rate": 2.4395738391427517e-06, + "loss": 0.4095, + "step": 15217 + }, + { + "epoch": 5.62, + "learning_rate": 2.4395368887794063e-06, + "loss": 0.4761, + "step": 15218 + }, + { + "epoch": 5.62, + "learning_rate": 2.4394999384160614e-06, + "loss": 0.4076, + "step": 15219 + }, + { + "epoch": 5.62, + "learning_rate": 2.439462988052716e-06, + "loss": 0.4038, + "step": 15220 + }, + { + "epoch": 5.62, + "learning_rate": 2.4394260376893707e-06, + "loss": 0.5396, + "step": 15221 + }, + { + "epoch": 5.62, + "learning_rate": 2.4393890873260253e-06, + "loss": 0.4314, + "step": 15222 + }, + { + "epoch": 5.62, + "learning_rate": 2.4393521369626804e-06, + "loss": 0.4826, + "step": 15223 + }, + { + "epoch": 5.62, + "learning_rate": 2.439315186599335e-06, + "loss": 0.5612, + "step": 15224 + }, + { + "epoch": 5.62, + "learning_rate": 2.4392782362359896e-06, + "loss": 0.6163, + "step": 15225 + }, + { + "epoch": 5.62, + "learning_rate": 2.4392412858726447e-06, + "loss": 0.4452, + "step": 15226 + }, + { + "epoch": 5.62, + "learning_rate": 2.439204335509299e-06, + "loss": 0.4175, + "step": 15227 + }, + { + "epoch": 5.62, + "learning_rate": 2.439167385145954e-06, + "loss": 0.3781, + "step": 15228 + }, + { + "epoch": 5.62, + "learning_rate": 2.4391304347826086e-06, + "loss": 0.3914, + "step": 15229 + }, + { + "epoch": 5.62, + "learning_rate": 2.4390934844192637e-06, + "loss": 0.6163, + "step": 15230 + }, + { + "epoch": 5.62, + "learning_rate": 2.4390565340559183e-06, + "loss": 0.4604, + "step": 15231 + }, + { + "epoch": 5.62, + "learning_rate": 2.4390195836925734e-06, + "loss": 0.5137, + "step": 15232 + }, + { + "epoch": 5.63, + "learning_rate": 2.4389826333292276e-06, + "loss": 0.4315, + "step": 15233 + }, + { + "epoch": 5.63, + "learning_rate": 2.4389456829658826e-06, + "loss": 0.4641, + "step": 15234 + }, + { + "epoch": 5.63, + "learning_rate": 2.4389087326025373e-06, + "loss": 0.4856, + "step": 15235 + }, + { + "epoch": 5.63, + "learning_rate": 2.438871782239192e-06, + "loss": 0.392, + "step": 15236 + }, + { + "epoch": 5.63, + "learning_rate": 2.438834831875847e-06, + "loss": 0.5149, + "step": 15237 + }, + { + "epoch": 5.63, + "learning_rate": 2.4387978815125016e-06, + "loss": 0.5963, + "step": 15238 + }, + { + "epoch": 5.63, + "learning_rate": 2.4387609311491563e-06, + "loss": 0.6154, + "step": 15239 + }, + { + "epoch": 5.63, + "learning_rate": 2.438723980785811e-06, + "loss": 0.4207, + "step": 15240 + }, + { + "epoch": 5.63, + "learning_rate": 2.438687030422466e-06, + "loss": 0.4239, + "step": 15241 + }, + { + "epoch": 5.63, + "learning_rate": 2.4386500800591206e-06, + "loss": 0.5248, + "step": 15242 + }, + { + "epoch": 5.63, + "learning_rate": 2.4386131296957756e-06, + "loss": 0.4812, + "step": 15243 + }, + { + "epoch": 5.63, + "learning_rate": 2.4385761793324303e-06, + "loss": 0.525, + "step": 15244 + }, + { + "epoch": 5.63, + "learning_rate": 2.438539228969085e-06, + "loss": 0.4616, + "step": 15245 + }, + { + "epoch": 5.63, + "learning_rate": 2.4385022786057396e-06, + "loss": 0.4047, + "step": 15246 + }, + { + "epoch": 5.63, + "learning_rate": 2.438465328242394e-06, + "loss": 0.5552, + "step": 15247 + }, + { + "epoch": 5.63, + "learning_rate": 2.4384283778790493e-06, + "loss": 0.4813, + "step": 15248 + }, + { + "epoch": 5.63, + "learning_rate": 2.438391427515704e-06, + "loss": 0.5377, + "step": 15249 + }, + { + "epoch": 5.63, + "learning_rate": 2.438354477152359e-06, + "loss": 0.4764, + "step": 15250 + }, + { + "epoch": 5.63, + "learning_rate": 2.4383175267890136e-06, + "loss": 0.5233, + "step": 15251 + }, + { + "epoch": 5.63, + "learning_rate": 2.4382805764256682e-06, + "loss": 0.5231, + "step": 15252 + }, + { + "epoch": 5.63, + "learning_rate": 2.438243626062323e-06, + "loss": 0.5071, + "step": 15253 + }, + { + "epoch": 5.63, + "learning_rate": 2.438206675698978e-06, + "loss": 0.5426, + "step": 15254 + }, + { + "epoch": 5.63, + "learning_rate": 2.4381697253356326e-06, + "loss": 0.3976, + "step": 15255 + }, + { + "epoch": 5.63, + "learning_rate": 2.438132774972287e-06, + "loss": 0.5518, + "step": 15256 + }, + { + "epoch": 5.63, + "learning_rate": 2.4380958246089422e-06, + "loss": 0.5627, + "step": 15257 + }, + { + "epoch": 5.63, + "learning_rate": 2.438058874245597e-06, + "loss": 0.4897, + "step": 15258 + }, + { + "epoch": 5.63, + "learning_rate": 2.4380219238822515e-06, + "loss": 0.4608, + "step": 15259 + }, + { + "epoch": 5.64, + "learning_rate": 2.437984973518906e-06, + "loss": 0.3885, + "step": 15260 + }, + { + "epoch": 5.64, + "learning_rate": 2.4379480231555612e-06, + "loss": 0.4878, + "step": 15261 + }, + { + "epoch": 5.64, + "learning_rate": 2.437911072792216e-06, + "loss": 0.4698, + "step": 15262 + }, + { + "epoch": 5.64, + "learning_rate": 2.437874122428871e-06, + "loss": 0.581, + "step": 15263 + }, + { + "epoch": 5.64, + "learning_rate": 2.4378371720655256e-06, + "loss": 0.5016, + "step": 15264 + }, + { + "epoch": 5.64, + "learning_rate": 2.43780022170218e-06, + "loss": 0.4836, + "step": 15265 + }, + { + "epoch": 5.64, + "learning_rate": 2.437763271338835e-06, + "loss": 0.4679, + "step": 15266 + }, + { + "epoch": 5.64, + "learning_rate": 2.4377263209754895e-06, + "loss": 0.4136, + "step": 15267 + }, + { + "epoch": 5.64, + "learning_rate": 2.4376893706121445e-06, + "loss": 0.3661, + "step": 15268 + }, + { + "epoch": 5.64, + "learning_rate": 2.437652420248799e-06, + "loss": 0.3686, + "step": 15269 + }, + { + "epoch": 5.64, + "learning_rate": 2.4376154698854542e-06, + "loss": 0.5431, + "step": 15270 + }, + { + "epoch": 5.64, + "learning_rate": 2.4375785195221084e-06, + "loss": 0.522, + "step": 15271 + }, + { + "epoch": 5.64, + "learning_rate": 2.4375415691587635e-06, + "loss": 0.4511, + "step": 15272 + }, + { + "epoch": 5.64, + "learning_rate": 2.437504618795418e-06, + "loss": 0.4433, + "step": 15273 + }, + { + "epoch": 5.64, + "learning_rate": 2.437467668432073e-06, + "loss": 0.4531, + "step": 15274 + }, + { + "epoch": 5.64, + "learning_rate": 2.437430718068728e-06, + "loss": 0.4996, + "step": 15275 + }, + { + "epoch": 5.64, + "learning_rate": 2.4373937677053825e-06, + "loss": 0.4362, + "step": 15276 + }, + { + "epoch": 5.64, + "learning_rate": 2.437356817342037e-06, + "loss": 0.4615, + "step": 15277 + }, + { + "epoch": 5.64, + "learning_rate": 2.4373198669786917e-06, + "loss": 0.4702, + "step": 15278 + }, + { + "epoch": 5.64, + "learning_rate": 2.437282916615347e-06, + "loss": 0.5047, + "step": 15279 + }, + { + "epoch": 5.64, + "learning_rate": 2.4372459662520014e-06, + "loss": 0.459, + "step": 15280 + }, + { + "epoch": 5.64, + "learning_rate": 2.4372090158886565e-06, + "loss": 0.475, + "step": 15281 + }, + { + "epoch": 5.64, + "learning_rate": 2.437172065525311e-06, + "loss": 0.4306, + "step": 15282 + }, + { + "epoch": 5.64, + "learning_rate": 2.4371351151619658e-06, + "loss": 0.3802, + "step": 15283 + }, + { + "epoch": 5.64, + "learning_rate": 2.4370981647986204e-06, + "loss": 0.3367, + "step": 15284 + }, + { + "epoch": 5.64, + "learning_rate": 2.437061214435275e-06, + "loss": 0.5691, + "step": 15285 + }, + { + "epoch": 5.64, + "learning_rate": 2.43702426407193e-06, + "loss": 0.5027, + "step": 15286 + }, + { + "epoch": 5.65, + "learning_rate": 2.4369873137085847e-06, + "loss": 0.4025, + "step": 15287 + }, + { + "epoch": 5.65, + "learning_rate": 2.43695036334524e-06, + "loss": 0.4153, + "step": 15288 + }, + { + "epoch": 5.65, + "learning_rate": 2.4369134129818944e-06, + "loss": 0.3448, + "step": 15289 + }, + { + "epoch": 5.65, + "learning_rate": 2.436876462618549e-06, + "loss": 0.386, + "step": 15290 + }, + { + "epoch": 5.65, + "learning_rate": 2.4368395122552037e-06, + "loss": 0.5806, + "step": 15291 + }, + { + "epoch": 5.65, + "learning_rate": 2.4368025618918588e-06, + "loss": 0.4101, + "step": 15292 + }, + { + "epoch": 5.65, + "learning_rate": 2.4367656115285134e-06, + "loss": 0.4989, + "step": 15293 + }, + { + "epoch": 5.65, + "learning_rate": 2.4367286611651685e-06, + "loss": 0.4397, + "step": 15294 + }, + { + "epoch": 5.65, + "learning_rate": 2.436691710801823e-06, + "loss": 0.3803, + "step": 15295 + }, + { + "epoch": 5.65, + "learning_rate": 2.4366547604384777e-06, + "loss": 0.5145, + "step": 15296 + }, + { + "epoch": 5.65, + "learning_rate": 2.4366178100751324e-06, + "loss": 0.4362, + "step": 15297 + }, + { + "epoch": 5.65, + "learning_rate": 2.436580859711787e-06, + "loss": 0.4875, + "step": 15298 + }, + { + "epoch": 5.65, + "learning_rate": 2.436543909348442e-06, + "loss": 0.4991, + "step": 15299 + }, + { + "epoch": 5.65, + "learning_rate": 2.4365069589850967e-06, + "loss": 0.432, + "step": 15300 + }, + { + "epoch": 5.65, + "learning_rate": 2.4364700086217518e-06, + "loss": 0.3862, + "step": 15301 + }, + { + "epoch": 5.65, + "learning_rate": 2.4364330582584064e-06, + "loss": 0.5146, + "step": 15302 + }, + { + "epoch": 5.65, + "learning_rate": 2.436396107895061e-06, + "loss": 0.4062, + "step": 15303 + }, + { + "epoch": 5.65, + "learning_rate": 2.4363591575317157e-06, + "loss": 0.5081, + "step": 15304 + }, + { + "epoch": 5.65, + "learning_rate": 2.4363222071683703e-06, + "loss": 0.5692, + "step": 15305 + }, + { + "epoch": 5.65, + "learning_rate": 2.4362852568050254e-06, + "loss": 0.564, + "step": 15306 + }, + { + "epoch": 5.65, + "learning_rate": 2.43624830644168e-06, + "loss": 0.3808, + "step": 15307 + }, + { + "epoch": 5.65, + "learning_rate": 2.436211356078335e-06, + "loss": 0.4373, + "step": 15308 + }, + { + "epoch": 5.65, + "learning_rate": 2.4361744057149897e-06, + "loss": 0.5855, + "step": 15309 + }, + { + "epoch": 5.65, + "learning_rate": 2.4361374553516443e-06, + "loss": 0.4734, + "step": 15310 + }, + { + "epoch": 5.65, + "learning_rate": 2.436100504988299e-06, + "loss": 0.4176, + "step": 15311 + }, + { + "epoch": 5.65, + "learning_rate": 2.436063554624954e-06, + "loss": 0.4057, + "step": 15312 + }, + { + "epoch": 5.65, + "learning_rate": 2.4360266042616087e-06, + "loss": 0.4444, + "step": 15313 + }, + { + "epoch": 5.66, + "learning_rate": 2.4359896538982637e-06, + "loss": 0.359, + "step": 15314 + }, + { + "epoch": 5.66, + "learning_rate": 2.4359527035349184e-06, + "loss": 0.4107, + "step": 15315 + }, + { + "epoch": 5.66, + "learning_rate": 2.4359157531715726e-06, + "loss": 0.5613, + "step": 15316 + }, + { + "epoch": 5.66, + "learning_rate": 2.4358788028082276e-06, + "loss": 0.3738, + "step": 15317 + }, + { + "epoch": 5.66, + "learning_rate": 2.4358418524448823e-06, + "loss": 0.4582, + "step": 15318 + }, + { + "epoch": 5.66, + "learning_rate": 2.4358049020815373e-06, + "loss": 0.417, + "step": 15319 + }, + { + "epoch": 5.66, + "learning_rate": 2.435767951718192e-06, + "loss": 0.4005, + "step": 15320 + }, + { + "epoch": 5.66, + "learning_rate": 2.435731001354847e-06, + "loss": 0.5375, + "step": 15321 + }, + { + "epoch": 5.66, + "learning_rate": 2.4356940509915012e-06, + "loss": 0.4794, + "step": 15322 + }, + { + "epoch": 5.66, + "learning_rate": 2.4356571006281563e-06, + "loss": 0.6054, + "step": 15323 + }, + { + "epoch": 5.66, + "learning_rate": 2.435620150264811e-06, + "loss": 0.4406, + "step": 15324 + }, + { + "epoch": 5.66, + "learning_rate": 2.4355831999014656e-06, + "loss": 0.3889, + "step": 15325 + }, + { + "epoch": 5.66, + "learning_rate": 2.4355462495381206e-06, + "loss": 0.5557, + "step": 15326 + }, + { + "epoch": 5.66, + "learning_rate": 2.4355092991747753e-06, + "loss": 0.4734, + "step": 15327 + }, + { + "epoch": 5.66, + "learning_rate": 2.43547234881143e-06, + "loss": 0.4833, + "step": 15328 + }, + { + "epoch": 5.66, + "learning_rate": 2.4354353984480845e-06, + "loss": 0.4607, + "step": 15329 + }, + { + "epoch": 5.66, + "learning_rate": 2.4353984480847396e-06, + "loss": 0.6013, + "step": 15330 + }, + { + "epoch": 5.66, + "learning_rate": 2.4353614977213942e-06, + "loss": 0.4547, + "step": 15331 + }, + { + "epoch": 5.66, + "learning_rate": 2.4353245473580493e-06, + "loss": 0.3502, + "step": 15332 + }, + { + "epoch": 5.66, + "learning_rate": 2.435287596994704e-06, + "loss": 0.4169, + "step": 15333 + }, + { + "epoch": 5.66, + "learning_rate": 2.4352506466313586e-06, + "loss": 0.5336, + "step": 15334 + }, + { + "epoch": 5.66, + "learning_rate": 2.435213696268013e-06, + "loss": 0.47, + "step": 15335 + }, + { + "epoch": 5.66, + "learning_rate": 2.435176745904668e-06, + "loss": 0.4456, + "step": 15336 + }, + { + "epoch": 5.66, + "learning_rate": 2.435139795541323e-06, + "loss": 0.4361, + "step": 15337 + }, + { + "epoch": 5.66, + "learning_rate": 2.4351028451779775e-06, + "loss": 0.5134, + "step": 15338 + }, + { + "epoch": 5.66, + "learning_rate": 2.4350658948146326e-06, + "loss": 0.6062, + "step": 15339 + }, + { + "epoch": 5.66, + "learning_rate": 2.4350289444512872e-06, + "loss": 0.5696, + "step": 15340 + }, + { + "epoch": 5.67, + "learning_rate": 2.434991994087942e-06, + "loss": 0.4817, + "step": 15341 + }, + { + "epoch": 5.67, + "learning_rate": 2.4349550437245965e-06, + "loss": 0.441, + "step": 15342 + }, + { + "epoch": 5.67, + "learning_rate": 2.4349180933612516e-06, + "loss": 0.521, + "step": 15343 + }, + { + "epoch": 5.67, + "learning_rate": 2.434881142997906e-06, + "loss": 0.3964, + "step": 15344 + }, + { + "epoch": 5.67, + "learning_rate": 2.434844192634561e-06, + "loss": 0.5516, + "step": 15345 + }, + { + "epoch": 5.67, + "learning_rate": 2.434807242271216e-06, + "loss": 0.4857, + "step": 15346 + }, + { + "epoch": 5.67, + "learning_rate": 2.4347702919078705e-06, + "loss": 0.554, + "step": 15347 + }, + { + "epoch": 5.67, + "learning_rate": 2.434733341544525e-06, + "loss": 0.4446, + "step": 15348 + }, + { + "epoch": 5.67, + "learning_rate": 2.43469639118118e-06, + "loss": 0.52, + "step": 15349 + }, + { + "epoch": 5.67, + "learning_rate": 2.434659440817835e-06, + "loss": 0.4522, + "step": 15350 + }, + { + "epoch": 5.67, + "learning_rate": 2.4346224904544895e-06, + "loss": 0.5478, + "step": 15351 + }, + { + "epoch": 5.67, + "learning_rate": 2.4345855400911446e-06, + "loss": 0.4788, + "step": 15352 + }, + { + "epoch": 5.67, + "learning_rate": 2.434548589727799e-06, + "loss": 0.4746, + "step": 15353 + }, + { + "epoch": 5.67, + "learning_rate": 2.434511639364454e-06, + "loss": 0.563, + "step": 15354 + }, + { + "epoch": 5.67, + "learning_rate": 2.4344746890011085e-06, + "loss": 0.4562, + "step": 15355 + }, + { + "epoch": 5.67, + "learning_rate": 2.434437738637763e-06, + "loss": 0.4409, + "step": 15356 + }, + { + "epoch": 5.67, + "learning_rate": 2.434400788274418e-06, + "loss": 0.5435, + "step": 15357 + }, + { + "epoch": 5.67, + "learning_rate": 2.434363837911073e-06, + "loss": 0.4009, + "step": 15358 + }, + { + "epoch": 5.67, + "learning_rate": 2.434326887547728e-06, + "loss": 0.4405, + "step": 15359 + }, + { + "epoch": 5.67, + "learning_rate": 2.4342899371843825e-06, + "loss": 0.4725, + "step": 15360 + }, + { + "epoch": 5.67, + "learning_rate": 2.434252986821037e-06, + "loss": 0.4699, + "step": 15361 + }, + { + "epoch": 5.67, + "learning_rate": 2.4342160364576918e-06, + "loss": 0.4319, + "step": 15362 + }, + { + "epoch": 5.67, + "learning_rate": 2.434179086094347e-06, + "loss": 0.4142, + "step": 15363 + }, + { + "epoch": 5.67, + "learning_rate": 2.4341421357310015e-06, + "loss": 0.4642, + "step": 15364 + }, + { + "epoch": 5.67, + "learning_rate": 2.434105185367656e-06, + "loss": 0.4403, + "step": 15365 + }, + { + "epoch": 5.67, + "learning_rate": 2.434068235004311e-06, + "loss": 0.5278, + "step": 15366 + }, + { + "epoch": 5.67, + "learning_rate": 2.4340312846409654e-06, + "loss": 0.4703, + "step": 15367 + }, + { + "epoch": 5.68, + "learning_rate": 2.4339943342776204e-06, + "loss": 0.4834, + "step": 15368 + }, + { + "epoch": 5.68, + "learning_rate": 2.433957383914275e-06, + "loss": 0.4828, + "step": 15369 + }, + { + "epoch": 5.68, + "learning_rate": 2.43392043355093e-06, + "loss": 0.4772, + "step": 15370 + }, + { + "epoch": 5.68, + "learning_rate": 2.4338834831875848e-06, + "loss": 0.5066, + "step": 15371 + }, + { + "epoch": 5.68, + "learning_rate": 2.43384653282424e-06, + "loss": 0.4861, + "step": 15372 + }, + { + "epoch": 5.68, + "learning_rate": 2.433809582460894e-06, + "loss": 0.4491, + "step": 15373 + }, + { + "epoch": 5.68, + "learning_rate": 2.4337726320975487e-06, + "loss": 0.3652, + "step": 15374 + }, + { + "epoch": 5.68, + "learning_rate": 2.4337356817342037e-06, + "loss": 0.5276, + "step": 15375 + }, + { + "epoch": 5.68, + "learning_rate": 2.4336987313708584e-06, + "loss": 0.4707, + "step": 15376 + }, + { + "epoch": 5.68, + "learning_rate": 2.4336617810075134e-06, + "loss": 0.4809, + "step": 15377 + }, + { + "epoch": 5.68, + "learning_rate": 2.433624830644168e-06, + "loss": 0.5018, + "step": 15378 + }, + { + "epoch": 5.68, + "learning_rate": 2.4335878802808227e-06, + "loss": 0.4575, + "step": 15379 + }, + { + "epoch": 5.68, + "learning_rate": 2.4335509299174774e-06, + "loss": 0.4602, + "step": 15380 + }, + { + "epoch": 5.68, + "learning_rate": 2.4335139795541324e-06, + "loss": 0.6296, + "step": 15381 + }, + { + "epoch": 5.68, + "learning_rate": 2.433477029190787e-06, + "loss": 0.4757, + "step": 15382 + }, + { + "epoch": 5.68, + "learning_rate": 2.433440078827442e-06, + "loss": 0.4758, + "step": 15383 + }, + { + "epoch": 5.68, + "learning_rate": 2.4334031284640967e-06, + "loss": 0.4689, + "step": 15384 + }, + { + "epoch": 5.68, + "learning_rate": 2.4333661781007514e-06, + "loss": 0.5334, + "step": 15385 + }, + { + "epoch": 5.68, + "learning_rate": 2.433329227737406e-06, + "loss": 0.5425, + "step": 15386 + }, + { + "epoch": 5.68, + "learning_rate": 2.4332922773740607e-06, + "loss": 0.5301, + "step": 15387 + }, + { + "epoch": 5.68, + "learning_rate": 2.4332553270107157e-06, + "loss": 0.484, + "step": 15388 + }, + { + "epoch": 5.68, + "learning_rate": 2.4332183766473704e-06, + "loss": 0.5095, + "step": 15389 + }, + { + "epoch": 5.68, + "learning_rate": 2.4331814262840254e-06, + "loss": 0.3611, + "step": 15390 + }, + { + "epoch": 5.68, + "learning_rate": 2.43314447592068e-06, + "loss": 0.5241, + "step": 15391 + }, + { + "epoch": 5.68, + "learning_rate": 2.4331075255573347e-06, + "loss": 0.5665, + "step": 15392 + }, + { + "epoch": 5.68, + "learning_rate": 2.4330705751939893e-06, + "loss": 0.4818, + "step": 15393 + }, + { + "epoch": 5.68, + "learning_rate": 2.433033624830644e-06, + "loss": 0.4108, + "step": 15394 + }, + { + "epoch": 5.69, + "learning_rate": 2.432996674467299e-06, + "loss": 0.4066, + "step": 15395 + }, + { + "epoch": 5.69, + "learning_rate": 2.4329597241039537e-06, + "loss": 0.5233, + "step": 15396 + }, + { + "epoch": 5.69, + "learning_rate": 2.4329227737406087e-06, + "loss": 0.4935, + "step": 15397 + }, + { + "epoch": 5.69, + "learning_rate": 2.4328858233772634e-06, + "loss": 0.4817, + "step": 15398 + }, + { + "epoch": 5.69, + "learning_rate": 2.432848873013918e-06, + "loss": 0.5299, + "step": 15399 + }, + { + "epoch": 5.69, + "learning_rate": 2.4328119226505726e-06, + "loss": 0.5393, + "step": 15400 + }, + { + "epoch": 5.69, + "learning_rate": 2.4327749722872277e-06, + "loss": 0.5264, + "step": 15401 + }, + { + "epoch": 5.69, + "learning_rate": 2.4327380219238823e-06, + "loss": 0.5262, + "step": 15402 + }, + { + "epoch": 5.69, + "learning_rate": 2.4327010715605374e-06, + "loss": 0.586, + "step": 15403 + }, + { + "epoch": 5.69, + "learning_rate": 2.432664121197192e-06, + "loss": 0.5098, + "step": 15404 + }, + { + "epoch": 5.69, + "learning_rate": 2.4326271708338467e-06, + "loss": 0.4162, + "step": 15405 + }, + { + "epoch": 5.69, + "learning_rate": 2.4325902204705013e-06, + "loss": 0.5072, + "step": 15406 + }, + { + "epoch": 5.69, + "learning_rate": 2.432553270107156e-06, + "loss": 0.4265, + "step": 15407 + }, + { + "epoch": 5.69, + "learning_rate": 2.432516319743811e-06, + "loss": 0.3901, + "step": 15408 + }, + { + "epoch": 5.69, + "learning_rate": 2.4324793693804656e-06, + "loss": 0.4167, + "step": 15409 + }, + { + "epoch": 5.69, + "learning_rate": 2.4324424190171207e-06, + "loss": 0.3818, + "step": 15410 + }, + { + "epoch": 5.69, + "learning_rate": 2.4324054686537753e-06, + "loss": 0.4785, + "step": 15411 + }, + { + "epoch": 5.69, + "learning_rate": 2.43236851829043e-06, + "loss": 0.5117, + "step": 15412 + }, + { + "epoch": 5.69, + "learning_rate": 2.4323315679270846e-06, + "loss": 0.3471, + "step": 15413 + }, + { + "epoch": 5.69, + "learning_rate": 2.4322946175637392e-06, + "loss": 0.4715, + "step": 15414 + }, + { + "epoch": 5.69, + "learning_rate": 2.4322576672003943e-06, + "loss": 0.4798, + "step": 15415 + }, + { + "epoch": 5.69, + "learning_rate": 2.432220716837049e-06, + "loss": 0.4766, + "step": 15416 + }, + { + "epoch": 5.69, + "learning_rate": 2.432183766473704e-06, + "loss": 0.4586, + "step": 15417 + }, + { + "epoch": 5.69, + "learning_rate": 2.432146816110358e-06, + "loss": 0.4642, + "step": 15418 + }, + { + "epoch": 5.69, + "learning_rate": 2.4321098657470133e-06, + "loss": 0.3849, + "step": 15419 + }, + { + "epoch": 5.69, + "learning_rate": 2.432072915383668e-06, + "loss": 0.4407, + "step": 15420 + }, + { + "epoch": 5.69, + "learning_rate": 2.432035965020323e-06, + "loss": 0.3196, + "step": 15421 + }, + { + "epoch": 5.69, + "learning_rate": 2.4319990146569776e-06, + "loss": 0.5253, + "step": 15422 + }, + { + "epoch": 5.7, + "learning_rate": 2.4319620642936322e-06, + "loss": 0.5588, + "step": 15423 + }, + { + "epoch": 5.7, + "learning_rate": 2.431925113930287e-06, + "loss": 0.5466, + "step": 15424 + }, + { + "epoch": 5.7, + "learning_rate": 2.4318881635669415e-06, + "loss": 0.5156, + "step": 15425 + }, + { + "epoch": 5.7, + "learning_rate": 2.4318512132035966e-06, + "loss": 0.398, + "step": 15426 + }, + { + "epoch": 5.7, + "learning_rate": 2.431814262840251e-06, + "loss": 0.4723, + "step": 15427 + }, + { + "epoch": 5.7, + "learning_rate": 2.4317773124769063e-06, + "loss": 0.4552, + "step": 15428 + }, + { + "epoch": 5.7, + "learning_rate": 2.431740362113561e-06, + "loss": 0.418, + "step": 15429 + }, + { + "epoch": 5.7, + "learning_rate": 2.4317034117502155e-06, + "loss": 0.421, + "step": 15430 + }, + { + "epoch": 5.7, + "learning_rate": 2.43166646138687e-06, + "loss": 0.4828, + "step": 15431 + }, + { + "epoch": 5.7, + "learning_rate": 2.4316295110235252e-06, + "loss": 0.4154, + "step": 15432 + }, + { + "epoch": 5.7, + "learning_rate": 2.43159256066018e-06, + "loss": 0.6013, + "step": 15433 + }, + { + "epoch": 5.7, + "learning_rate": 2.4315556102968345e-06, + "loss": 0.4452, + "step": 15434 + }, + { + "epoch": 5.7, + "learning_rate": 2.4315186599334896e-06, + "loss": 0.5821, + "step": 15435 + }, + { + "epoch": 5.7, + "learning_rate": 2.431481709570144e-06, + "loss": 0.3858, + "step": 15436 + }, + { + "epoch": 5.7, + "learning_rate": 2.431444759206799e-06, + "loss": 0.3732, + "step": 15437 + }, + { + "epoch": 5.7, + "learning_rate": 2.4314078088434535e-06, + "loss": 0.4859, + "step": 15438 + }, + { + "epoch": 5.7, + "learning_rate": 2.4313708584801085e-06, + "loss": 0.4098, + "step": 15439 + }, + { + "epoch": 5.7, + "learning_rate": 2.431333908116763e-06, + "loss": 0.5791, + "step": 15440 + }, + { + "epoch": 5.7, + "learning_rate": 2.4312969577534182e-06, + "loss": 0.4584, + "step": 15441 + }, + { + "epoch": 5.7, + "learning_rate": 2.431260007390073e-06, + "loss": 0.4597, + "step": 15442 + }, + { + "epoch": 5.7, + "learning_rate": 2.4312230570267275e-06, + "loss": 0.5107, + "step": 15443 + }, + { + "epoch": 5.7, + "learning_rate": 2.431186106663382e-06, + "loss": 0.4605, + "step": 15444 + }, + { + "epoch": 5.7, + "learning_rate": 2.4311491563000368e-06, + "loss": 0.5102, + "step": 15445 + }, + { + "epoch": 5.7, + "learning_rate": 2.431112205936692e-06, + "loss": 0.4495, + "step": 15446 + }, + { + "epoch": 5.7, + "learning_rate": 2.4310752555733465e-06, + "loss": 0.3908, + "step": 15447 + }, + { + "epoch": 5.7, + "learning_rate": 2.4310383052100015e-06, + "loss": 0.4304, + "step": 15448 + }, + { + "epoch": 5.7, + "learning_rate": 2.431001354846656e-06, + "loss": 0.4746, + "step": 15449 + }, + { + "epoch": 5.71, + "learning_rate": 2.430964404483311e-06, + "loss": 0.3763, + "step": 15450 + }, + { + "epoch": 5.71, + "learning_rate": 2.4309274541199654e-06, + "loss": 0.3756, + "step": 15451 + }, + { + "epoch": 5.71, + "learning_rate": 2.4308905037566205e-06, + "loss": 0.4948, + "step": 15452 + }, + { + "epoch": 5.71, + "learning_rate": 2.430853553393275e-06, + "loss": 0.6361, + "step": 15453 + }, + { + "epoch": 5.71, + "learning_rate": 2.4308166030299298e-06, + "loss": 0.6094, + "step": 15454 + }, + { + "epoch": 5.71, + "learning_rate": 2.430779652666585e-06, + "loss": 0.4836, + "step": 15455 + }, + { + "epoch": 5.71, + "learning_rate": 2.430742702303239e-06, + "loss": 0.4559, + "step": 15456 + }, + { + "epoch": 5.71, + "learning_rate": 2.430705751939894e-06, + "loss": 0.5491, + "step": 15457 + }, + { + "epoch": 5.71, + "learning_rate": 2.4306688015765487e-06, + "loss": 0.4504, + "step": 15458 + }, + { + "epoch": 5.71, + "learning_rate": 2.430631851213204e-06, + "loss": 0.5316, + "step": 15459 + }, + { + "epoch": 5.71, + "learning_rate": 2.4305949008498584e-06, + "loss": 0.4954, + "step": 15460 + }, + { + "epoch": 5.71, + "learning_rate": 2.4305579504865135e-06, + "loss": 0.5267, + "step": 15461 + }, + { + "epoch": 5.71, + "learning_rate": 2.4305210001231677e-06, + "loss": 0.4464, + "step": 15462 + }, + { + "epoch": 5.71, + "learning_rate": 2.4304840497598228e-06, + "loss": 0.4802, + "step": 15463 + }, + { + "epoch": 5.71, + "learning_rate": 2.4304470993964774e-06, + "loss": 0.5936, + "step": 15464 + }, + { + "epoch": 5.71, + "learning_rate": 2.430410149033132e-06, + "loss": 0.3874, + "step": 15465 + }, + { + "epoch": 5.71, + "learning_rate": 2.430373198669787e-06, + "loss": 0.5368, + "step": 15466 + }, + { + "epoch": 5.71, + "learning_rate": 2.4303362483064417e-06, + "loss": 0.4662, + "step": 15467 + }, + { + "epoch": 5.71, + "learning_rate": 2.4302992979430964e-06, + "loss": 0.4425, + "step": 15468 + }, + { + "epoch": 5.71, + "learning_rate": 2.430262347579751e-06, + "loss": 0.4009, + "step": 15469 + }, + { + "epoch": 5.71, + "learning_rate": 2.430225397216406e-06, + "loss": 0.4514, + "step": 15470 + }, + { + "epoch": 5.71, + "learning_rate": 2.4301884468530607e-06, + "loss": 0.507, + "step": 15471 + }, + { + "epoch": 5.71, + "learning_rate": 2.4301514964897158e-06, + "loss": 0.4796, + "step": 15472 + }, + { + "epoch": 5.71, + "learning_rate": 2.4301145461263704e-06, + "loss": 0.4394, + "step": 15473 + }, + { + "epoch": 5.71, + "learning_rate": 2.430077595763025e-06, + "loss": 0.4369, + "step": 15474 + }, + { + "epoch": 5.71, + "learning_rate": 2.4300406453996797e-06, + "loss": 0.611, + "step": 15475 + }, + { + "epoch": 5.71, + "learning_rate": 2.4300036950363343e-06, + "loss": 0.5905, + "step": 15476 + }, + { + "epoch": 5.72, + "learning_rate": 2.4299667446729894e-06, + "loss": 0.5545, + "step": 15477 + }, + { + "epoch": 5.72, + "learning_rate": 2.429929794309644e-06, + "loss": 0.4671, + "step": 15478 + }, + { + "epoch": 5.72, + "learning_rate": 2.429892843946299e-06, + "loss": 0.4268, + "step": 15479 + }, + { + "epoch": 5.72, + "learning_rate": 2.4298558935829537e-06, + "loss": 0.5193, + "step": 15480 + }, + { + "epoch": 5.72, + "learning_rate": 2.4298189432196083e-06, + "loss": 0.5104, + "step": 15481 + }, + { + "epoch": 5.72, + "learning_rate": 2.429781992856263e-06, + "loss": 0.4405, + "step": 15482 + }, + { + "epoch": 5.72, + "learning_rate": 2.429745042492918e-06, + "loss": 0.471, + "step": 15483 + }, + { + "epoch": 5.72, + "learning_rate": 2.4297080921295727e-06, + "loss": 0.5285, + "step": 15484 + }, + { + "epoch": 5.72, + "learning_rate": 2.4296711417662273e-06, + "loss": 0.5357, + "step": 15485 + }, + { + "epoch": 5.72, + "learning_rate": 2.4296341914028824e-06, + "loss": 0.4293, + "step": 15486 + }, + { + "epoch": 5.72, + "learning_rate": 2.429597241039537e-06, + "loss": 0.4481, + "step": 15487 + }, + { + "epoch": 5.72, + "learning_rate": 2.4295602906761916e-06, + "loss": 0.4657, + "step": 15488 + }, + { + "epoch": 5.72, + "learning_rate": 2.4295233403128463e-06, + "loss": 0.4282, + "step": 15489 + }, + { + "epoch": 5.72, + "learning_rate": 2.4294863899495013e-06, + "loss": 0.4726, + "step": 15490 + }, + { + "epoch": 5.72, + "learning_rate": 2.429449439586156e-06, + "loss": 0.3885, + "step": 15491 + }, + { + "epoch": 5.72, + "learning_rate": 2.429412489222811e-06, + "loss": 0.4461, + "step": 15492 + }, + { + "epoch": 5.72, + "learning_rate": 2.4293755388594657e-06, + "loss": 0.5401, + "step": 15493 + }, + { + "epoch": 5.72, + "learning_rate": 2.4293385884961203e-06, + "loss": 0.539, + "step": 15494 + }, + { + "epoch": 5.72, + "learning_rate": 2.429301638132775e-06, + "loss": 0.4508, + "step": 15495 + }, + { + "epoch": 5.72, + "learning_rate": 2.4292646877694296e-06, + "loss": 0.4824, + "step": 15496 + }, + { + "epoch": 5.72, + "learning_rate": 2.4292277374060846e-06, + "loss": 0.4052, + "step": 15497 + }, + { + "epoch": 5.72, + "learning_rate": 2.4291907870427393e-06, + "loss": 0.5589, + "step": 15498 + }, + { + "epoch": 5.72, + "learning_rate": 2.4291538366793943e-06, + "loss": 0.4139, + "step": 15499 + }, + { + "epoch": 5.72, + "learning_rate": 2.429116886316049e-06, + "loss": 0.5161, + "step": 15500 + }, + { + "epoch": 5.72, + "learning_rate": 2.4290799359527036e-06, + "loss": 0.4525, + "step": 15501 + }, + { + "epoch": 5.72, + "learning_rate": 2.4290429855893582e-06, + "loss": 0.4292, + "step": 15502 + }, + { + "epoch": 5.72, + "learning_rate": 2.4290060352260133e-06, + "loss": 0.4744, + "step": 15503 + }, + { + "epoch": 5.73, + "learning_rate": 2.428969084862668e-06, + "loss": 0.4394, + "step": 15504 + }, + { + "epoch": 5.73, + "learning_rate": 2.4289321344993226e-06, + "loss": 0.3589, + "step": 15505 + }, + { + "epoch": 5.73, + "learning_rate": 2.4288951841359776e-06, + "loss": 0.3426, + "step": 15506 + }, + { + "epoch": 5.73, + "learning_rate": 2.428858233772632e-06, + "loss": 0.5186, + "step": 15507 + }, + { + "epoch": 5.73, + "learning_rate": 2.428821283409287e-06, + "loss": 0.4673, + "step": 15508 + }, + { + "epoch": 5.73, + "learning_rate": 2.4287843330459415e-06, + "loss": 0.558, + "step": 15509 + }, + { + "epoch": 5.73, + "learning_rate": 2.4287473826825966e-06, + "loss": 0.4792, + "step": 15510 + }, + { + "epoch": 5.73, + "learning_rate": 2.4287104323192512e-06, + "loss": 0.4202, + "step": 15511 + }, + { + "epoch": 5.73, + "learning_rate": 2.4286734819559063e-06, + "loss": 0.4418, + "step": 15512 + }, + { + "epoch": 5.73, + "learning_rate": 2.4286365315925605e-06, + "loss": 0.5052, + "step": 15513 + }, + { + "epoch": 5.73, + "learning_rate": 2.428599581229215e-06, + "loss": 0.3453, + "step": 15514 + }, + { + "epoch": 5.73, + "learning_rate": 2.4285626308658702e-06, + "loss": 0.3926, + "step": 15515 + }, + { + "epoch": 5.73, + "learning_rate": 2.428525680502525e-06, + "loss": 0.421, + "step": 15516 + }, + { + "epoch": 5.73, + "learning_rate": 2.42848873013918e-06, + "loss": 0.4127, + "step": 15517 + }, + { + "epoch": 5.73, + "learning_rate": 2.4284517797758345e-06, + "loss": 0.4777, + "step": 15518 + }, + { + "epoch": 5.73, + "learning_rate": 2.428414829412489e-06, + "loss": 0.5294, + "step": 15519 + }, + { + "epoch": 5.73, + "learning_rate": 2.428377879049144e-06, + "loss": 0.4787, + "step": 15520 + }, + { + "epoch": 5.73, + "learning_rate": 2.428340928685799e-06, + "loss": 0.5049, + "step": 15521 + }, + { + "epoch": 5.73, + "learning_rate": 2.4283039783224535e-06, + "loss": 0.4819, + "step": 15522 + }, + { + "epoch": 5.73, + "learning_rate": 2.4282670279591086e-06, + "loss": 0.4846, + "step": 15523 + }, + { + "epoch": 5.73, + "learning_rate": 2.428230077595763e-06, + "loss": 0.4378, + "step": 15524 + }, + { + "epoch": 5.73, + "learning_rate": 2.428193127232418e-06, + "loss": 0.4863, + "step": 15525 + }, + { + "epoch": 5.73, + "learning_rate": 2.4281561768690725e-06, + "loss": 0.4356, + "step": 15526 + }, + { + "epoch": 5.73, + "learning_rate": 2.428119226505727e-06, + "loss": 0.4225, + "step": 15527 + }, + { + "epoch": 5.73, + "learning_rate": 2.428082276142382e-06, + "loss": 0.5045, + "step": 15528 + }, + { + "epoch": 5.73, + "learning_rate": 2.428045325779037e-06, + "loss": 0.4432, + "step": 15529 + }, + { + "epoch": 5.73, + "learning_rate": 2.428008375415692e-06, + "loss": 0.4974, + "step": 15530 + }, + { + "epoch": 5.74, + "learning_rate": 2.4279714250523465e-06, + "loss": 0.4335, + "step": 15531 + }, + { + "epoch": 5.74, + "learning_rate": 2.427934474689001e-06, + "loss": 0.4798, + "step": 15532 + }, + { + "epoch": 5.74, + "learning_rate": 2.4278975243256558e-06, + "loss": 0.478, + "step": 15533 + }, + { + "epoch": 5.74, + "learning_rate": 2.4278605739623104e-06, + "loss": 0.6068, + "step": 15534 + }, + { + "epoch": 5.74, + "learning_rate": 2.4278236235989655e-06, + "loss": 0.4906, + "step": 15535 + }, + { + "epoch": 5.74, + "learning_rate": 2.42778667323562e-06, + "loss": 0.4392, + "step": 15536 + }, + { + "epoch": 5.74, + "learning_rate": 2.427749722872275e-06, + "loss": 0.4789, + "step": 15537 + }, + { + "epoch": 5.74, + "learning_rate": 2.42771277250893e-06, + "loss": 0.5481, + "step": 15538 + }, + { + "epoch": 5.74, + "learning_rate": 2.4276758221455845e-06, + "loss": 0.6036, + "step": 15539 + }, + { + "epoch": 5.74, + "learning_rate": 2.427638871782239e-06, + "loss": 0.4888, + "step": 15540 + }, + { + "epoch": 5.74, + "learning_rate": 2.427601921418894e-06, + "loss": 0.5696, + "step": 15541 + }, + { + "epoch": 5.74, + "learning_rate": 2.4275649710555488e-06, + "loss": 0.431, + "step": 15542 + }, + { + "epoch": 5.74, + "learning_rate": 2.427528020692204e-06, + "loss": 0.441, + "step": 15543 + }, + { + "epoch": 5.74, + "learning_rate": 2.4274910703288585e-06, + "loss": 0.4062, + "step": 15544 + }, + { + "epoch": 5.74, + "learning_rate": 2.427454119965513e-06, + "loss": 0.4037, + "step": 15545 + }, + { + "epoch": 5.74, + "learning_rate": 2.4274171696021678e-06, + "loss": 0.4491, + "step": 15546 + }, + { + "epoch": 5.74, + "learning_rate": 2.4273802192388224e-06, + "loss": 0.4567, + "step": 15547 + }, + { + "epoch": 5.74, + "learning_rate": 2.4273432688754775e-06, + "loss": 0.4876, + "step": 15548 + }, + { + "epoch": 5.74, + "learning_rate": 2.427306318512132e-06, + "loss": 0.4509, + "step": 15549 + }, + { + "epoch": 5.74, + "learning_rate": 2.427269368148787e-06, + "loss": 0.4982, + "step": 15550 + }, + { + "epoch": 5.74, + "learning_rate": 2.4272324177854418e-06, + "loss": 0.5719, + "step": 15551 + }, + { + "epoch": 5.74, + "learning_rate": 2.4271954674220964e-06, + "loss": 0.4545, + "step": 15552 + }, + { + "epoch": 5.74, + "learning_rate": 2.427158517058751e-06, + "loss": 0.4747, + "step": 15553 + }, + { + "epoch": 5.74, + "learning_rate": 2.4271215666954057e-06, + "loss": 0.4926, + "step": 15554 + }, + { + "epoch": 5.74, + "learning_rate": 2.4270846163320608e-06, + "loss": 0.4629, + "step": 15555 + }, + { + "epoch": 5.74, + "learning_rate": 2.4270476659687154e-06, + "loss": 0.5508, + "step": 15556 + }, + { + "epoch": 5.74, + "learning_rate": 2.4270107156053704e-06, + "loss": 0.4682, + "step": 15557 + }, + { + "epoch": 5.75, + "learning_rate": 2.4269737652420247e-06, + "loss": 0.4552, + "step": 15558 + }, + { + "epoch": 5.75, + "learning_rate": 2.4269368148786797e-06, + "loss": 0.5209, + "step": 15559 + }, + { + "epoch": 5.75, + "learning_rate": 2.4268998645153344e-06, + "loss": 0.4185, + "step": 15560 + }, + { + "epoch": 5.75, + "learning_rate": 2.4268629141519894e-06, + "loss": 0.5721, + "step": 15561 + }, + { + "epoch": 5.75, + "learning_rate": 2.426825963788644e-06, + "loss": 0.5347, + "step": 15562 + }, + { + "epoch": 5.75, + "learning_rate": 2.426789013425299e-06, + "loss": 0.4186, + "step": 15563 + }, + { + "epoch": 5.75, + "learning_rate": 2.4267520630619533e-06, + "loss": 0.5073, + "step": 15564 + }, + { + "epoch": 5.75, + "learning_rate": 2.426715112698608e-06, + "loss": 0.4895, + "step": 15565 + }, + { + "epoch": 5.75, + "learning_rate": 2.426678162335263e-06, + "loss": 0.4605, + "step": 15566 + }, + { + "epoch": 5.75, + "learning_rate": 2.4266412119719177e-06, + "loss": 0.4132, + "step": 15567 + }, + { + "epoch": 5.75, + "learning_rate": 2.4266042616085727e-06, + "loss": 0.3615, + "step": 15568 + }, + { + "epoch": 5.75, + "learning_rate": 2.4265673112452274e-06, + "loss": 0.4565, + "step": 15569 + }, + { + "epoch": 5.75, + "learning_rate": 2.426530360881882e-06, + "loss": 0.4453, + "step": 15570 + }, + { + "epoch": 5.75, + "learning_rate": 2.4264934105185366e-06, + "loss": 0.468, + "step": 15571 + }, + { + "epoch": 5.75, + "learning_rate": 2.4264564601551917e-06, + "loss": 0.4551, + "step": 15572 + }, + { + "epoch": 5.75, + "learning_rate": 2.4264195097918463e-06, + "loss": 0.4703, + "step": 15573 + }, + { + "epoch": 5.75, + "learning_rate": 2.426382559428501e-06, + "loss": 0.4156, + "step": 15574 + }, + { + "epoch": 5.75, + "learning_rate": 2.426345609065156e-06, + "loss": 0.5368, + "step": 15575 + }, + { + "epoch": 5.75, + "learning_rate": 2.4263086587018107e-06, + "loss": 0.4538, + "step": 15576 + }, + { + "epoch": 5.75, + "learning_rate": 2.4262717083384653e-06, + "loss": 0.4836, + "step": 15577 + }, + { + "epoch": 5.75, + "learning_rate": 2.42623475797512e-06, + "loss": 0.6039, + "step": 15578 + }, + { + "epoch": 5.75, + "learning_rate": 2.426197807611775e-06, + "loss": 0.4175, + "step": 15579 + }, + { + "epoch": 5.75, + "learning_rate": 2.4261608572484296e-06, + "loss": 0.4012, + "step": 15580 + }, + { + "epoch": 5.75, + "learning_rate": 2.4261239068850847e-06, + "loss": 0.4239, + "step": 15581 + }, + { + "epoch": 5.75, + "learning_rate": 2.4260869565217393e-06, + "loss": 0.5246, + "step": 15582 + }, + { + "epoch": 5.75, + "learning_rate": 2.426050006158394e-06, + "loss": 0.4498, + "step": 15583 + }, + { + "epoch": 5.75, + "learning_rate": 2.4260130557950486e-06, + "loss": 0.4236, + "step": 15584 + }, + { + "epoch": 5.76, + "learning_rate": 2.4259761054317032e-06, + "loss": 0.632, + "step": 15585 + }, + { + "epoch": 5.76, + "learning_rate": 2.4259391550683583e-06, + "loss": 0.4926, + "step": 15586 + }, + { + "epoch": 5.76, + "learning_rate": 2.425902204705013e-06, + "loss": 0.444, + "step": 15587 + }, + { + "epoch": 5.76, + "learning_rate": 2.425865254341668e-06, + "loss": 0.4335, + "step": 15588 + }, + { + "epoch": 5.76, + "learning_rate": 2.4258283039783226e-06, + "loss": 0.4096, + "step": 15589 + }, + { + "epoch": 5.76, + "learning_rate": 2.4257913536149773e-06, + "loss": 0.503, + "step": 15590 + }, + { + "epoch": 5.76, + "learning_rate": 2.425754403251632e-06, + "loss": 0.5732, + "step": 15591 + }, + { + "epoch": 5.76, + "learning_rate": 2.425717452888287e-06, + "loss": 0.4478, + "step": 15592 + }, + { + "epoch": 5.76, + "learning_rate": 2.4256805025249416e-06, + "loss": 0.4768, + "step": 15593 + }, + { + "epoch": 5.76, + "learning_rate": 2.4256435521615962e-06, + "loss": 0.4081, + "step": 15594 + }, + { + "epoch": 5.76, + "learning_rate": 2.4256066017982513e-06, + "loss": 0.4839, + "step": 15595 + }, + { + "epoch": 5.76, + "learning_rate": 2.425569651434906e-06, + "loss": 0.4989, + "step": 15596 + }, + { + "epoch": 5.76, + "learning_rate": 2.4255327010715606e-06, + "loss": 0.5269, + "step": 15597 + }, + { + "epoch": 5.76, + "learning_rate": 2.425495750708215e-06, + "loss": 0.4272, + "step": 15598 + }, + { + "epoch": 5.76, + "learning_rate": 2.4254588003448703e-06, + "loss": 0.4702, + "step": 15599 + }, + { + "epoch": 5.76, + "learning_rate": 2.425421849981525e-06, + "loss": 0.463, + "step": 15600 + }, + { + "epoch": 5.76, + "learning_rate": 2.42538489961818e-06, + "loss": 0.4383, + "step": 15601 + }, + { + "epoch": 5.76, + "learning_rate": 2.4253479492548346e-06, + "loss": 0.4489, + "step": 15602 + }, + { + "epoch": 5.76, + "learning_rate": 2.425310998891489e-06, + "loss": 0.6003, + "step": 15603 + }, + { + "epoch": 5.76, + "learning_rate": 2.425274048528144e-06, + "loss": 0.4786, + "step": 15604 + }, + { + "epoch": 5.76, + "learning_rate": 2.4252370981647985e-06, + "loss": 0.4449, + "step": 15605 + }, + { + "epoch": 5.76, + "learning_rate": 2.4252001478014536e-06, + "loss": 0.4319, + "step": 15606 + }, + { + "epoch": 5.76, + "learning_rate": 2.425163197438108e-06, + "loss": 0.4438, + "step": 15607 + }, + { + "epoch": 5.76, + "learning_rate": 2.425126247074763e-06, + "loss": 0.4306, + "step": 15608 + }, + { + "epoch": 5.76, + "learning_rate": 2.4250892967114175e-06, + "loss": 0.3749, + "step": 15609 + }, + { + "epoch": 5.76, + "learning_rate": 2.4250523463480725e-06, + "loss": 0.616, + "step": 15610 + }, + { + "epoch": 5.76, + "learning_rate": 2.425015395984727e-06, + "loss": 0.4764, + "step": 15611 + }, + { + "epoch": 5.77, + "learning_rate": 2.4249784456213822e-06, + "loss": 0.4046, + "step": 15612 + }, + { + "epoch": 5.77, + "learning_rate": 2.424941495258037e-06, + "loss": 0.2997, + "step": 15613 + }, + { + "epoch": 5.77, + "learning_rate": 2.4249045448946915e-06, + "loss": 0.4777, + "step": 15614 + }, + { + "epoch": 5.77, + "learning_rate": 2.424867594531346e-06, + "loss": 0.5554, + "step": 15615 + }, + { + "epoch": 5.77, + "learning_rate": 2.4248306441680008e-06, + "loss": 0.5527, + "step": 15616 + }, + { + "epoch": 5.77, + "learning_rate": 2.424793693804656e-06, + "loss": 0.5306, + "step": 15617 + }, + { + "epoch": 5.77, + "learning_rate": 2.4247567434413105e-06, + "loss": 0.3234, + "step": 15618 + }, + { + "epoch": 5.77, + "learning_rate": 2.4247197930779655e-06, + "loss": 0.5867, + "step": 15619 + }, + { + "epoch": 5.77, + "learning_rate": 2.42468284271462e-06, + "loss": 0.5337, + "step": 15620 + }, + { + "epoch": 5.77, + "learning_rate": 2.424645892351275e-06, + "loss": 0.5051, + "step": 15621 + }, + { + "epoch": 5.77, + "learning_rate": 2.4246089419879294e-06, + "loss": 0.5884, + "step": 15622 + }, + { + "epoch": 5.77, + "learning_rate": 2.424571991624584e-06, + "loss": 0.5763, + "step": 15623 + }, + { + "epoch": 5.77, + "learning_rate": 2.424535041261239e-06, + "loss": 0.5499, + "step": 15624 + }, + { + "epoch": 5.77, + "learning_rate": 2.4244980908978938e-06, + "loss": 0.4676, + "step": 15625 + }, + { + "epoch": 5.77, + "learning_rate": 2.424461140534549e-06, + "loss": 0.4737, + "step": 15626 + }, + { + "epoch": 5.77, + "learning_rate": 2.4244241901712035e-06, + "loss": 0.5374, + "step": 15627 + }, + { + "epoch": 5.77, + "learning_rate": 2.424387239807858e-06, + "loss": 0.4734, + "step": 15628 + }, + { + "epoch": 5.77, + "learning_rate": 2.4243502894445127e-06, + "loss": 0.5157, + "step": 15629 + }, + { + "epoch": 5.77, + "learning_rate": 2.424313339081168e-06, + "loss": 0.4926, + "step": 15630 + }, + { + "epoch": 5.77, + "learning_rate": 2.4242763887178224e-06, + "loss": 0.3834, + "step": 15631 + }, + { + "epoch": 5.77, + "learning_rate": 2.4242394383544775e-06, + "loss": 0.4927, + "step": 15632 + }, + { + "epoch": 5.77, + "learning_rate": 2.424202487991132e-06, + "loss": 0.4705, + "step": 15633 + }, + { + "epoch": 5.77, + "learning_rate": 2.4241655376277868e-06, + "loss": 0.4489, + "step": 15634 + }, + { + "epoch": 5.77, + "learning_rate": 2.4241285872644414e-06, + "loss": 0.6331, + "step": 15635 + }, + { + "epoch": 5.77, + "learning_rate": 2.424091636901096e-06, + "loss": 0.4367, + "step": 15636 + }, + { + "epoch": 5.77, + "learning_rate": 2.424054686537751e-06, + "loss": 0.4685, + "step": 15637 + }, + { + "epoch": 5.77, + "learning_rate": 2.4240177361744057e-06, + "loss": 0.4603, + "step": 15638 + }, + { + "epoch": 5.78, + "learning_rate": 2.423980785811061e-06, + "loss": 0.4603, + "step": 15639 + }, + { + "epoch": 5.78, + "learning_rate": 2.4239438354477154e-06, + "loss": 0.4738, + "step": 15640 + }, + { + "epoch": 5.78, + "learning_rate": 2.42390688508437e-06, + "loss": 0.4463, + "step": 15641 + }, + { + "epoch": 5.78, + "learning_rate": 2.4238699347210247e-06, + "loss": 0.498, + "step": 15642 + }, + { + "epoch": 5.78, + "learning_rate": 2.4238329843576793e-06, + "loss": 0.4728, + "step": 15643 + }, + { + "epoch": 5.78, + "learning_rate": 2.4237960339943344e-06, + "loss": 0.4774, + "step": 15644 + }, + { + "epoch": 5.78, + "learning_rate": 2.423759083630989e-06, + "loss": 0.4072, + "step": 15645 + }, + { + "epoch": 5.78, + "learning_rate": 2.423722133267644e-06, + "loss": 0.4992, + "step": 15646 + }, + { + "epoch": 5.78, + "learning_rate": 2.4236851829042983e-06, + "loss": 0.4333, + "step": 15647 + }, + { + "epoch": 5.78, + "learning_rate": 2.4236482325409534e-06, + "loss": 0.5045, + "step": 15648 + }, + { + "epoch": 5.78, + "learning_rate": 2.423611282177608e-06, + "loss": 0.5184, + "step": 15649 + }, + { + "epoch": 5.78, + "learning_rate": 2.423574331814263e-06, + "loss": 0.442, + "step": 15650 + }, + { + "epoch": 5.78, + "learning_rate": 2.4235373814509177e-06, + "loss": 0.4384, + "step": 15651 + }, + { + "epoch": 5.78, + "learning_rate": 2.4235004310875728e-06, + "loss": 0.3824, + "step": 15652 + }, + { + "epoch": 5.78, + "learning_rate": 2.423463480724227e-06, + "loss": 0.3631, + "step": 15653 + }, + { + "epoch": 5.78, + "learning_rate": 2.4234265303608816e-06, + "loss": 0.4181, + "step": 15654 + }, + { + "epoch": 5.78, + "learning_rate": 2.4233895799975367e-06, + "loss": 0.3877, + "step": 15655 + }, + { + "epoch": 5.78, + "learning_rate": 2.4233526296341913e-06, + "loss": 0.5979, + "step": 15656 + }, + { + "epoch": 5.78, + "learning_rate": 2.4233156792708464e-06, + "loss": 0.5241, + "step": 15657 + }, + { + "epoch": 5.78, + "learning_rate": 2.423278728907501e-06, + "loss": 0.432, + "step": 15658 + }, + { + "epoch": 5.78, + "learning_rate": 2.4232417785441556e-06, + "loss": 0.4928, + "step": 15659 + }, + { + "epoch": 5.78, + "learning_rate": 2.4232048281808103e-06, + "loss": 0.4265, + "step": 15660 + }, + { + "epoch": 5.78, + "learning_rate": 2.4231678778174653e-06, + "loss": 0.4705, + "step": 15661 + }, + { + "epoch": 5.78, + "learning_rate": 2.42313092745412e-06, + "loss": 0.4757, + "step": 15662 + }, + { + "epoch": 5.78, + "learning_rate": 2.4230939770907746e-06, + "loss": 0.5138, + "step": 15663 + }, + { + "epoch": 5.78, + "learning_rate": 2.4230570267274297e-06, + "loss": 0.4909, + "step": 15664 + }, + { + "epoch": 5.78, + "learning_rate": 2.4230200763640843e-06, + "loss": 0.386, + "step": 15665 + }, + { + "epoch": 5.79, + "learning_rate": 2.422983126000739e-06, + "loss": 0.454, + "step": 15666 + }, + { + "epoch": 5.79, + "learning_rate": 2.4229461756373936e-06, + "loss": 0.4844, + "step": 15667 + }, + { + "epoch": 5.79, + "learning_rate": 2.4229092252740486e-06, + "loss": 0.4287, + "step": 15668 + }, + { + "epoch": 5.79, + "learning_rate": 2.4228722749107033e-06, + "loss": 0.5475, + "step": 15669 + }, + { + "epoch": 5.79, + "learning_rate": 2.4228353245473583e-06, + "loss": 0.6206, + "step": 15670 + }, + { + "epoch": 5.79, + "learning_rate": 2.422798374184013e-06, + "loss": 0.4238, + "step": 15671 + }, + { + "epoch": 5.79, + "learning_rate": 2.4227614238206676e-06, + "loss": 0.4654, + "step": 15672 + }, + { + "epoch": 5.79, + "learning_rate": 2.4227244734573223e-06, + "loss": 0.5061, + "step": 15673 + }, + { + "epoch": 5.79, + "learning_rate": 2.422687523093977e-06, + "loss": 0.4963, + "step": 15674 + }, + { + "epoch": 5.79, + "learning_rate": 2.422650572730632e-06, + "loss": 0.5156, + "step": 15675 + }, + { + "epoch": 5.79, + "learning_rate": 2.4226136223672866e-06, + "loss": 0.4811, + "step": 15676 + }, + { + "epoch": 5.79, + "learning_rate": 2.4225766720039416e-06, + "loss": 0.44, + "step": 15677 + }, + { + "epoch": 5.79, + "learning_rate": 2.4225397216405963e-06, + "loss": 0.5136, + "step": 15678 + }, + { + "epoch": 5.79, + "learning_rate": 2.422502771277251e-06, + "loss": 0.4487, + "step": 15679 + }, + { + "epoch": 5.79, + "learning_rate": 2.4224658209139056e-06, + "loss": 0.5199, + "step": 15680 + }, + { + "epoch": 5.79, + "learning_rate": 2.4224288705505606e-06, + "loss": 0.5126, + "step": 15681 + }, + { + "epoch": 5.79, + "learning_rate": 2.4223919201872152e-06, + "loss": 0.4831, + "step": 15682 + }, + { + "epoch": 5.79, + "learning_rate": 2.42235496982387e-06, + "loss": 0.4436, + "step": 15683 + }, + { + "epoch": 5.79, + "learning_rate": 2.422318019460525e-06, + "loss": 0.5234, + "step": 15684 + }, + { + "epoch": 5.79, + "learning_rate": 2.4222810690971796e-06, + "loss": 0.4281, + "step": 15685 + }, + { + "epoch": 5.79, + "learning_rate": 2.4222441187338342e-06, + "loss": 0.4193, + "step": 15686 + }, + { + "epoch": 5.79, + "learning_rate": 2.422207168370489e-06, + "loss": 0.3778, + "step": 15687 + }, + { + "epoch": 5.79, + "learning_rate": 2.422170218007144e-06, + "loss": 0.4327, + "step": 15688 + }, + { + "epoch": 5.79, + "learning_rate": 2.4221332676437986e-06, + "loss": 0.4642, + "step": 15689 + }, + { + "epoch": 5.79, + "learning_rate": 2.4220963172804536e-06, + "loss": 0.4351, + "step": 15690 + }, + { + "epoch": 5.79, + "learning_rate": 2.4220593669171082e-06, + "loss": 0.5327, + "step": 15691 + }, + { + "epoch": 5.79, + "learning_rate": 2.422022416553763e-06, + "loss": 0.3689, + "step": 15692 + }, + { + "epoch": 5.8, + "learning_rate": 2.4219854661904175e-06, + "loss": 0.4639, + "step": 15693 + }, + { + "epoch": 5.8, + "learning_rate": 2.421948515827072e-06, + "loss": 0.4338, + "step": 15694 + }, + { + "epoch": 5.8, + "learning_rate": 2.4219115654637272e-06, + "loss": 0.6062, + "step": 15695 + }, + { + "epoch": 5.8, + "learning_rate": 2.421874615100382e-06, + "loss": 0.4209, + "step": 15696 + }, + { + "epoch": 5.8, + "learning_rate": 2.421837664737037e-06, + "loss": 0.4794, + "step": 15697 + }, + { + "epoch": 5.8, + "learning_rate": 2.421800714373691e-06, + "loss": 0.507, + "step": 15698 + }, + { + "epoch": 5.8, + "learning_rate": 2.421763764010346e-06, + "loss": 0.4333, + "step": 15699 + }, + { + "epoch": 5.8, + "learning_rate": 2.421726813647001e-06, + "loss": 0.4462, + "step": 15700 + }, + { + "epoch": 5.8, + "learning_rate": 2.421689863283656e-06, + "loss": 0.507, + "step": 15701 + }, + { + "epoch": 5.8, + "learning_rate": 2.4216529129203105e-06, + "loss": 0.642, + "step": 15702 + }, + { + "epoch": 5.8, + "learning_rate": 2.421615962556965e-06, + "loss": 0.5465, + "step": 15703 + }, + { + "epoch": 5.8, + "learning_rate": 2.42157901219362e-06, + "loss": 0.5521, + "step": 15704 + }, + { + "epoch": 5.8, + "learning_rate": 2.4215420618302744e-06, + "loss": 0.5611, + "step": 15705 + }, + { + "epoch": 5.8, + "learning_rate": 2.4215051114669295e-06, + "loss": 0.5188, + "step": 15706 + }, + { + "epoch": 5.8, + "learning_rate": 2.421468161103584e-06, + "loss": 0.5041, + "step": 15707 + }, + { + "epoch": 5.8, + "learning_rate": 2.421431210740239e-06, + "loss": 0.5021, + "step": 15708 + }, + { + "epoch": 5.8, + "learning_rate": 2.421394260376894e-06, + "loss": 0.4063, + "step": 15709 + }, + { + "epoch": 5.8, + "learning_rate": 2.4213573100135485e-06, + "loss": 0.4555, + "step": 15710 + }, + { + "epoch": 5.8, + "learning_rate": 2.421320359650203e-06, + "loss": 0.4064, + "step": 15711 + }, + { + "epoch": 5.8, + "learning_rate": 2.421283409286858e-06, + "loss": 0.5727, + "step": 15712 + }, + { + "epoch": 5.8, + "learning_rate": 2.421246458923513e-06, + "loss": 0.4921, + "step": 15713 + }, + { + "epoch": 5.8, + "learning_rate": 2.4212095085601674e-06, + "loss": 0.5064, + "step": 15714 + }, + { + "epoch": 5.8, + "learning_rate": 2.4211725581968225e-06, + "loss": 0.5936, + "step": 15715 + }, + { + "epoch": 5.8, + "learning_rate": 2.421135607833477e-06, + "loss": 0.4486, + "step": 15716 + }, + { + "epoch": 5.8, + "learning_rate": 2.4210986574701318e-06, + "loss": 0.4943, + "step": 15717 + }, + { + "epoch": 5.8, + "learning_rate": 2.4210617071067864e-06, + "loss": 0.4584, + "step": 15718 + }, + { + "epoch": 5.8, + "learning_rate": 2.4210247567434415e-06, + "loss": 0.4401, + "step": 15719 + }, + { + "epoch": 5.81, + "learning_rate": 2.420987806380096e-06, + "loss": 0.5271, + "step": 15720 + }, + { + "epoch": 5.81, + "learning_rate": 2.420950856016751e-06, + "loss": 0.4494, + "step": 15721 + }, + { + "epoch": 5.81, + "learning_rate": 2.420913905653406e-06, + "loss": 0.4491, + "step": 15722 + }, + { + "epoch": 5.81, + "learning_rate": 2.4208769552900604e-06, + "loss": 0.5366, + "step": 15723 + }, + { + "epoch": 5.81, + "learning_rate": 2.420840004926715e-06, + "loss": 0.4783, + "step": 15724 + }, + { + "epoch": 5.81, + "learning_rate": 2.4208030545633697e-06, + "loss": 0.5297, + "step": 15725 + }, + { + "epoch": 5.81, + "learning_rate": 2.4207661042000248e-06, + "loss": 0.4612, + "step": 15726 + }, + { + "epoch": 5.81, + "learning_rate": 2.4207291538366794e-06, + "loss": 0.3964, + "step": 15727 + }, + { + "epoch": 5.81, + "learning_rate": 2.4206922034733345e-06, + "loss": 0.5212, + "step": 15728 + }, + { + "epoch": 5.81, + "learning_rate": 2.420655253109989e-06, + "loss": 0.6242, + "step": 15729 + }, + { + "epoch": 5.81, + "learning_rate": 2.4206183027466437e-06, + "loss": 0.5729, + "step": 15730 + }, + { + "epoch": 5.81, + "learning_rate": 2.4205813523832984e-06, + "loss": 0.3965, + "step": 15731 + }, + { + "epoch": 5.81, + "learning_rate": 2.4205444020199534e-06, + "loss": 0.3664, + "step": 15732 + }, + { + "epoch": 5.81, + "learning_rate": 2.420507451656608e-06, + "loss": 0.4399, + "step": 15733 + }, + { + "epoch": 5.81, + "learning_rate": 2.4204705012932627e-06, + "loss": 0.49, + "step": 15734 + }, + { + "epoch": 5.81, + "learning_rate": 2.4204335509299178e-06, + "loss": 0.6244, + "step": 15735 + }, + { + "epoch": 5.81, + "learning_rate": 2.4203966005665724e-06, + "loss": 0.4568, + "step": 15736 + }, + { + "epoch": 5.81, + "learning_rate": 2.420359650203227e-06, + "loss": 0.4458, + "step": 15737 + }, + { + "epoch": 5.81, + "learning_rate": 2.4203226998398817e-06, + "loss": 0.3856, + "step": 15738 + }, + { + "epoch": 5.81, + "learning_rate": 2.4202857494765367e-06, + "loss": 0.5026, + "step": 15739 + }, + { + "epoch": 5.81, + "learning_rate": 2.4202487991131914e-06, + "loss": 0.4731, + "step": 15740 + }, + { + "epoch": 5.81, + "learning_rate": 2.4202118487498464e-06, + "loss": 0.474, + "step": 15741 + }, + { + "epoch": 5.81, + "learning_rate": 2.420174898386501e-06, + "loss": 0.4211, + "step": 15742 + }, + { + "epoch": 5.81, + "learning_rate": 2.4201379480231553e-06, + "loss": 0.407, + "step": 15743 + }, + { + "epoch": 5.81, + "learning_rate": 2.4201009976598103e-06, + "loss": 0.536, + "step": 15744 + }, + { + "epoch": 5.81, + "learning_rate": 2.420064047296465e-06, + "loss": 0.5239, + "step": 15745 + }, + { + "epoch": 5.81, + "learning_rate": 2.42002709693312e-06, + "loss": 0.4289, + "step": 15746 + }, + { + "epoch": 5.81, + "learning_rate": 2.4199901465697747e-06, + "loss": 0.4889, + "step": 15747 + }, + { + "epoch": 5.82, + "learning_rate": 2.4199531962064297e-06, + "loss": 0.5008, + "step": 15748 + }, + { + "epoch": 5.82, + "learning_rate": 2.419916245843084e-06, + "loss": 0.6093, + "step": 15749 + }, + { + "epoch": 5.82, + "learning_rate": 2.419879295479739e-06, + "loss": 0.5854, + "step": 15750 + }, + { + "epoch": 5.82, + "learning_rate": 2.4198423451163936e-06, + "loss": 0.4235, + "step": 15751 + }, + { + "epoch": 5.82, + "learning_rate": 2.4198053947530487e-06, + "loss": 0.4481, + "step": 15752 + }, + { + "epoch": 5.82, + "learning_rate": 2.4197684443897033e-06, + "loss": 0.61, + "step": 15753 + }, + { + "epoch": 5.82, + "learning_rate": 2.419731494026358e-06, + "loss": 0.5653, + "step": 15754 + }, + { + "epoch": 5.82, + "learning_rate": 2.4196945436630126e-06, + "loss": 0.5649, + "step": 15755 + }, + { + "epoch": 5.82, + "learning_rate": 2.4196575932996672e-06, + "loss": 0.4712, + "step": 15756 + }, + { + "epoch": 5.82, + "learning_rate": 2.4196206429363223e-06, + "loss": 0.6034, + "step": 15757 + }, + { + "epoch": 5.82, + "learning_rate": 2.419583692572977e-06, + "loss": 0.4509, + "step": 15758 + }, + { + "epoch": 5.82, + "learning_rate": 2.419546742209632e-06, + "loss": 0.4465, + "step": 15759 + }, + { + "epoch": 5.82, + "learning_rate": 2.4195097918462866e-06, + "loss": 0.473, + "step": 15760 + }, + { + "epoch": 5.82, + "learning_rate": 2.4194728414829413e-06, + "loss": 0.468, + "step": 15761 + }, + { + "epoch": 5.82, + "learning_rate": 2.419435891119596e-06, + "loss": 0.4655, + "step": 15762 + }, + { + "epoch": 5.82, + "learning_rate": 2.4193989407562505e-06, + "loss": 0.4416, + "step": 15763 + }, + { + "epoch": 5.82, + "learning_rate": 2.4193619903929056e-06, + "loss": 0.455, + "step": 15764 + }, + { + "epoch": 5.82, + "learning_rate": 2.4193250400295602e-06, + "loss": 0.3844, + "step": 15765 + }, + { + "epoch": 5.82, + "learning_rate": 2.4192880896662153e-06, + "loss": 0.4358, + "step": 15766 + }, + { + "epoch": 5.82, + "learning_rate": 2.41925113930287e-06, + "loss": 0.4719, + "step": 15767 + }, + { + "epoch": 5.82, + "learning_rate": 2.4192141889395246e-06, + "loss": 0.3545, + "step": 15768 + }, + { + "epoch": 5.82, + "learning_rate": 2.419177238576179e-06, + "loss": 0.4942, + "step": 15769 + }, + { + "epoch": 5.82, + "learning_rate": 2.4191402882128343e-06, + "loss": 0.4285, + "step": 15770 + }, + { + "epoch": 5.82, + "learning_rate": 2.419103337849489e-06, + "loss": 0.4709, + "step": 15771 + }, + { + "epoch": 5.82, + "learning_rate": 2.419066387486144e-06, + "loss": 0.5405, + "step": 15772 + }, + { + "epoch": 5.82, + "learning_rate": 2.4190294371227986e-06, + "loss": 0.4761, + "step": 15773 + }, + { + "epoch": 5.82, + "learning_rate": 2.4189924867594532e-06, + "loss": 0.4307, + "step": 15774 + }, + { + "epoch": 5.83, + "learning_rate": 2.418955536396108e-06, + "loss": 0.4766, + "step": 15775 + }, + { + "epoch": 5.83, + "learning_rate": 2.4189185860327625e-06, + "loss": 0.4827, + "step": 15776 + }, + { + "epoch": 5.83, + "learning_rate": 2.4188816356694176e-06, + "loss": 0.4346, + "step": 15777 + }, + { + "epoch": 5.83, + "learning_rate": 2.418844685306072e-06, + "loss": 0.5597, + "step": 15778 + }, + { + "epoch": 5.83, + "learning_rate": 2.4188077349427273e-06, + "loss": 0.5835, + "step": 15779 + }, + { + "epoch": 5.83, + "learning_rate": 2.418770784579382e-06, + "loss": 0.5708, + "step": 15780 + }, + { + "epoch": 5.83, + "learning_rate": 2.4187338342160365e-06, + "loss": 0.4068, + "step": 15781 + }, + { + "epoch": 5.83, + "learning_rate": 2.418696883852691e-06, + "loss": 0.4523, + "step": 15782 + }, + { + "epoch": 5.83, + "learning_rate": 2.418659933489346e-06, + "loss": 0.5358, + "step": 15783 + }, + { + "epoch": 5.83, + "learning_rate": 2.418622983126001e-06, + "loss": 0.4554, + "step": 15784 + }, + { + "epoch": 5.83, + "learning_rate": 2.4185860327626555e-06, + "loss": 0.5984, + "step": 15785 + }, + { + "epoch": 5.83, + "learning_rate": 2.4185490823993106e-06, + "loss": 0.4156, + "step": 15786 + }, + { + "epoch": 5.83, + "learning_rate": 2.418512132035965e-06, + "loss": 0.4181, + "step": 15787 + }, + { + "epoch": 5.83, + "learning_rate": 2.41847518167262e-06, + "loss": 0.5765, + "step": 15788 + }, + { + "epoch": 5.83, + "learning_rate": 2.4184382313092745e-06, + "loss": 0.4321, + "step": 15789 + }, + { + "epoch": 5.83, + "learning_rate": 2.4184012809459295e-06, + "loss": 0.6374, + "step": 15790 + }, + { + "epoch": 5.83, + "learning_rate": 2.418364330582584e-06, + "loss": 0.4709, + "step": 15791 + }, + { + "epoch": 5.83, + "learning_rate": 2.4183273802192392e-06, + "loss": 0.4244, + "step": 15792 + }, + { + "epoch": 5.83, + "learning_rate": 2.418290429855894e-06, + "loss": 0.5166, + "step": 15793 + }, + { + "epoch": 5.83, + "learning_rate": 2.418253479492548e-06, + "loss": 0.573, + "step": 15794 + }, + { + "epoch": 5.83, + "learning_rate": 2.418216529129203e-06, + "loss": 0.5017, + "step": 15795 + }, + { + "epoch": 5.83, + "learning_rate": 2.4181795787658578e-06, + "loss": 0.4105, + "step": 15796 + }, + { + "epoch": 5.83, + "learning_rate": 2.418142628402513e-06, + "loss": 0.5554, + "step": 15797 + }, + { + "epoch": 5.83, + "learning_rate": 2.4181056780391675e-06, + "loss": 0.4792, + "step": 15798 + }, + { + "epoch": 5.83, + "learning_rate": 2.418068727675822e-06, + "loss": 0.4753, + "step": 15799 + }, + { + "epoch": 5.83, + "learning_rate": 2.4180317773124767e-06, + "loss": 0.5359, + "step": 15800 + }, + { + "epoch": 5.83, + "learning_rate": 2.417994826949132e-06, + "loss": 0.486, + "step": 15801 + }, + { + "epoch": 5.84, + "learning_rate": 2.4179578765857864e-06, + "loss": 0.478, + "step": 15802 + }, + { + "epoch": 5.84, + "learning_rate": 2.417920926222441e-06, + "loss": 0.4889, + "step": 15803 + }, + { + "epoch": 5.84, + "learning_rate": 2.417883975859096e-06, + "loss": 0.4555, + "step": 15804 + }, + { + "epoch": 5.84, + "learning_rate": 2.4178470254957508e-06, + "loss": 0.4442, + "step": 15805 + }, + { + "epoch": 5.84, + "learning_rate": 2.4178100751324054e-06, + "loss": 0.427, + "step": 15806 + }, + { + "epoch": 5.84, + "learning_rate": 2.41777312476906e-06, + "loss": 0.4369, + "step": 15807 + }, + { + "epoch": 5.84, + "learning_rate": 2.417736174405715e-06, + "loss": 0.411, + "step": 15808 + }, + { + "epoch": 5.84, + "learning_rate": 2.4176992240423697e-06, + "loss": 0.4314, + "step": 15809 + }, + { + "epoch": 5.84, + "learning_rate": 2.417662273679025e-06, + "loss": 0.3782, + "step": 15810 + }, + { + "epoch": 5.84, + "learning_rate": 2.4176253233156794e-06, + "loss": 0.4591, + "step": 15811 + }, + { + "epoch": 5.84, + "learning_rate": 2.417588372952334e-06, + "loss": 0.5125, + "step": 15812 + }, + { + "epoch": 5.84, + "learning_rate": 2.4175514225889887e-06, + "loss": 0.4565, + "step": 15813 + }, + { + "epoch": 5.84, + "learning_rate": 2.4175144722256434e-06, + "loss": 0.4308, + "step": 15814 + }, + { + "epoch": 5.84, + "learning_rate": 2.4174775218622984e-06, + "loss": 0.3903, + "step": 15815 + }, + { + "epoch": 5.84, + "learning_rate": 2.417440571498953e-06, + "loss": 0.4519, + "step": 15816 + }, + { + "epoch": 5.84, + "learning_rate": 2.417403621135608e-06, + "loss": 0.376, + "step": 15817 + }, + { + "epoch": 5.84, + "learning_rate": 2.4173666707722627e-06, + "loss": 0.44, + "step": 15818 + }, + { + "epoch": 5.84, + "learning_rate": 2.4173297204089174e-06, + "loss": 0.5333, + "step": 15819 + }, + { + "epoch": 5.84, + "learning_rate": 2.417292770045572e-06, + "loss": 0.4764, + "step": 15820 + }, + { + "epoch": 5.84, + "learning_rate": 2.417255819682227e-06, + "loss": 0.4884, + "step": 15821 + }, + { + "epoch": 5.84, + "learning_rate": 2.4172188693188817e-06, + "loss": 0.5499, + "step": 15822 + }, + { + "epoch": 5.84, + "learning_rate": 2.4171819189555364e-06, + "loss": 0.3939, + "step": 15823 + }, + { + "epoch": 5.84, + "learning_rate": 2.4171449685921914e-06, + "loss": 0.4766, + "step": 15824 + }, + { + "epoch": 5.84, + "learning_rate": 2.417108018228846e-06, + "loss": 0.5241, + "step": 15825 + }, + { + "epoch": 5.84, + "learning_rate": 2.4170710678655007e-06, + "loss": 0.3859, + "step": 15826 + }, + { + "epoch": 5.84, + "learning_rate": 2.4170341175021553e-06, + "loss": 0.5594, + "step": 15827 + }, + { + "epoch": 5.84, + "learning_rate": 2.4169971671388104e-06, + "loss": 0.5413, + "step": 15828 + }, + { + "epoch": 5.85, + "learning_rate": 2.416960216775465e-06, + "loss": 0.4193, + "step": 15829 + }, + { + "epoch": 5.85, + "learning_rate": 2.41692326641212e-06, + "loss": 0.3598, + "step": 15830 + }, + { + "epoch": 5.85, + "learning_rate": 2.4168863160487747e-06, + "loss": 0.3931, + "step": 15831 + }, + { + "epoch": 5.85, + "learning_rate": 2.416849365685429e-06, + "loss": 0.4641, + "step": 15832 + }, + { + "epoch": 5.85, + "learning_rate": 2.416812415322084e-06, + "loss": 0.4291, + "step": 15833 + }, + { + "epoch": 5.85, + "learning_rate": 2.4167754649587386e-06, + "loss": 0.4657, + "step": 15834 + }, + { + "epoch": 5.85, + "learning_rate": 2.4167385145953937e-06, + "loss": 0.5223, + "step": 15835 + }, + { + "epoch": 5.85, + "learning_rate": 2.4167015642320483e-06, + "loss": 0.3814, + "step": 15836 + }, + { + "epoch": 5.85, + "learning_rate": 2.4166646138687034e-06, + "loss": 0.49, + "step": 15837 + }, + { + "epoch": 5.85, + "learning_rate": 2.4166276635053576e-06, + "loss": 0.4461, + "step": 15838 + }, + { + "epoch": 5.85, + "learning_rate": 2.4165907131420127e-06, + "loss": 0.3706, + "step": 15839 + }, + { + "epoch": 5.85, + "learning_rate": 2.4165537627786673e-06, + "loss": 0.3813, + "step": 15840 + }, + { + "epoch": 5.85, + "learning_rate": 2.4165168124153223e-06, + "loss": 0.567, + "step": 15841 + }, + { + "epoch": 5.85, + "learning_rate": 2.416479862051977e-06, + "loss": 0.3663, + "step": 15842 + }, + { + "epoch": 5.85, + "learning_rate": 2.4164429116886316e-06, + "loss": 0.4354, + "step": 15843 + }, + { + "epoch": 5.85, + "learning_rate": 2.4164059613252863e-06, + "loss": 0.513, + "step": 15844 + }, + { + "epoch": 5.85, + "learning_rate": 2.416369010961941e-06, + "loss": 0.5357, + "step": 15845 + }, + { + "epoch": 5.85, + "learning_rate": 2.416332060598596e-06, + "loss": 0.3729, + "step": 15846 + }, + { + "epoch": 5.85, + "learning_rate": 2.4162951102352506e-06, + "loss": 0.5312, + "step": 15847 + }, + { + "epoch": 5.85, + "learning_rate": 2.4162581598719057e-06, + "loss": 0.5013, + "step": 15848 + }, + { + "epoch": 5.85, + "learning_rate": 2.4162212095085603e-06, + "loss": 0.4237, + "step": 15849 + }, + { + "epoch": 5.85, + "learning_rate": 2.416184259145215e-06, + "loss": 0.4956, + "step": 15850 + }, + { + "epoch": 5.85, + "learning_rate": 2.4161473087818696e-06, + "loss": 0.538, + "step": 15851 + }, + { + "epoch": 5.85, + "learning_rate": 2.416110358418524e-06, + "loss": 0.3448, + "step": 15852 + }, + { + "epoch": 5.85, + "learning_rate": 2.4160734080551793e-06, + "loss": 0.4812, + "step": 15853 + }, + { + "epoch": 5.85, + "learning_rate": 2.416036457691834e-06, + "loss": 0.4085, + "step": 15854 + }, + { + "epoch": 5.85, + "learning_rate": 2.415999507328489e-06, + "loss": 0.4616, + "step": 15855 + }, + { + "epoch": 5.86, + "learning_rate": 2.4159625569651436e-06, + "loss": 0.4674, + "step": 15856 + }, + { + "epoch": 5.86, + "learning_rate": 2.4159256066017982e-06, + "loss": 0.5087, + "step": 15857 + }, + { + "epoch": 5.86, + "learning_rate": 2.415888656238453e-06, + "loss": 0.4715, + "step": 15858 + }, + { + "epoch": 5.86, + "learning_rate": 2.415851705875108e-06, + "loss": 0.5061, + "step": 15859 + }, + { + "epoch": 5.86, + "learning_rate": 2.4158147555117626e-06, + "loss": 0.4026, + "step": 15860 + }, + { + "epoch": 5.86, + "learning_rate": 2.4157778051484176e-06, + "loss": 0.3626, + "step": 15861 + }, + { + "epoch": 5.86, + "learning_rate": 2.4157408547850723e-06, + "loss": 0.5183, + "step": 15862 + }, + { + "epoch": 5.86, + "learning_rate": 2.415703904421727e-06, + "loss": 0.4419, + "step": 15863 + }, + { + "epoch": 5.86, + "learning_rate": 2.4156669540583815e-06, + "loss": 0.5034, + "step": 15864 + }, + { + "epoch": 5.86, + "learning_rate": 2.415630003695036e-06, + "loss": 0.4109, + "step": 15865 + }, + { + "epoch": 5.86, + "learning_rate": 2.4155930533316912e-06, + "loss": 0.5248, + "step": 15866 + }, + { + "epoch": 5.86, + "learning_rate": 2.415556102968346e-06, + "loss": 0.4689, + "step": 15867 + }, + { + "epoch": 5.86, + "learning_rate": 2.415519152605001e-06, + "loss": 0.4782, + "step": 15868 + }, + { + "epoch": 5.86, + "learning_rate": 2.4154822022416556e-06, + "loss": 0.572, + "step": 15869 + }, + { + "epoch": 5.86, + "learning_rate": 2.41544525187831e-06, + "loss": 0.4835, + "step": 15870 + }, + { + "epoch": 5.86, + "learning_rate": 2.415408301514965e-06, + "loss": 0.4159, + "step": 15871 + }, + { + "epoch": 5.86, + "learning_rate": 2.4153713511516195e-06, + "loss": 0.4773, + "step": 15872 + }, + { + "epoch": 5.86, + "learning_rate": 2.4153344007882745e-06, + "loss": 0.4265, + "step": 15873 + }, + { + "epoch": 5.86, + "learning_rate": 2.415297450424929e-06, + "loss": 0.4084, + "step": 15874 + }, + { + "epoch": 5.86, + "learning_rate": 2.4152605000615842e-06, + "loss": 0.5216, + "step": 15875 + }, + { + "epoch": 5.86, + "learning_rate": 2.415223549698239e-06, + "loss": 0.4547, + "step": 15876 + }, + { + "epoch": 5.86, + "learning_rate": 2.4151865993348935e-06, + "loss": 0.5024, + "step": 15877 + }, + { + "epoch": 5.86, + "learning_rate": 2.415149648971548e-06, + "loss": 0.4907, + "step": 15878 + }, + { + "epoch": 5.86, + "learning_rate": 2.415112698608203e-06, + "loss": 0.4164, + "step": 15879 + }, + { + "epoch": 5.86, + "learning_rate": 2.415075748244858e-06, + "loss": 0.5598, + "step": 15880 + }, + { + "epoch": 5.86, + "learning_rate": 2.415038797881513e-06, + "loss": 0.527, + "step": 15881 + }, + { + "epoch": 5.86, + "learning_rate": 2.4150018475181675e-06, + "loss": 0.4702, + "step": 15882 + }, + { + "epoch": 5.87, + "learning_rate": 2.4149648971548217e-06, + "loss": 0.4221, + "step": 15883 + }, + { + "epoch": 5.87, + "learning_rate": 2.414927946791477e-06, + "loss": 0.4371, + "step": 15884 + }, + { + "epoch": 5.87, + "learning_rate": 2.4148909964281314e-06, + "loss": 0.5727, + "step": 15885 + }, + { + "epoch": 5.87, + "learning_rate": 2.4148540460647865e-06, + "loss": 0.4513, + "step": 15886 + }, + { + "epoch": 5.87, + "learning_rate": 2.414817095701441e-06, + "loss": 0.5189, + "step": 15887 + }, + { + "epoch": 5.87, + "learning_rate": 2.414780145338096e-06, + "loss": 0.4932, + "step": 15888 + }, + { + "epoch": 5.87, + "learning_rate": 2.4147431949747504e-06, + "loss": 0.4468, + "step": 15889 + }, + { + "epoch": 5.87, + "learning_rate": 2.4147062446114055e-06, + "loss": 0.4207, + "step": 15890 + }, + { + "epoch": 5.87, + "learning_rate": 2.41466929424806e-06, + "loss": 0.5029, + "step": 15891 + }, + { + "epoch": 5.87, + "learning_rate": 2.4146323438847147e-06, + "loss": 0.5446, + "step": 15892 + }, + { + "epoch": 5.87, + "learning_rate": 2.41459539352137e-06, + "loss": 0.4823, + "step": 15893 + }, + { + "epoch": 5.87, + "learning_rate": 2.4145584431580244e-06, + "loss": 0.4091, + "step": 15894 + }, + { + "epoch": 5.87, + "learning_rate": 2.414521492794679e-06, + "loss": 0.521, + "step": 15895 + }, + { + "epoch": 5.87, + "learning_rate": 2.4144845424313337e-06, + "loss": 0.3363, + "step": 15896 + }, + { + "epoch": 5.87, + "learning_rate": 2.4144475920679888e-06, + "loss": 0.3516, + "step": 15897 + }, + { + "epoch": 5.87, + "learning_rate": 2.4144106417046434e-06, + "loss": 0.5294, + "step": 15898 + }, + { + "epoch": 5.87, + "learning_rate": 2.4143736913412985e-06, + "loss": 0.515, + "step": 15899 + }, + { + "epoch": 5.87, + "learning_rate": 2.414336740977953e-06, + "loss": 0.4917, + "step": 15900 + }, + { + "epoch": 5.87, + "learning_rate": 2.4142997906146077e-06, + "loss": 0.4001, + "step": 15901 + }, + { + "epoch": 5.87, + "learning_rate": 2.4142628402512624e-06, + "loss": 0.4395, + "step": 15902 + }, + { + "epoch": 5.87, + "learning_rate": 2.414225889887917e-06, + "loss": 0.3714, + "step": 15903 + }, + { + "epoch": 5.87, + "learning_rate": 2.414188939524572e-06, + "loss": 0.4499, + "step": 15904 + }, + { + "epoch": 5.87, + "learning_rate": 2.4141519891612267e-06, + "loss": 0.5076, + "step": 15905 + }, + { + "epoch": 5.87, + "learning_rate": 2.4141150387978818e-06, + "loss": 0.296, + "step": 15906 + }, + { + "epoch": 5.87, + "learning_rate": 2.4140780884345364e-06, + "loss": 0.4666, + "step": 15907 + }, + { + "epoch": 5.87, + "learning_rate": 2.414041138071191e-06, + "loss": 0.443, + "step": 15908 + }, + { + "epoch": 5.87, + "learning_rate": 2.4140041877078457e-06, + "loss": 0.5384, + "step": 15909 + }, + { + "epoch": 5.88, + "learning_rate": 2.4139672373445007e-06, + "loss": 0.4157, + "step": 15910 + }, + { + "epoch": 5.88, + "learning_rate": 2.4139302869811554e-06, + "loss": 0.5354, + "step": 15911 + }, + { + "epoch": 5.88, + "learning_rate": 2.41389333661781e-06, + "loss": 0.5061, + "step": 15912 + }, + { + "epoch": 5.88, + "learning_rate": 2.413856386254465e-06, + "loss": 0.5661, + "step": 15913 + }, + { + "epoch": 5.88, + "learning_rate": 2.4138194358911197e-06, + "loss": 0.5038, + "step": 15914 + }, + { + "epoch": 5.88, + "learning_rate": 2.4137824855277743e-06, + "loss": 0.47, + "step": 15915 + }, + { + "epoch": 5.88, + "learning_rate": 2.413745535164429e-06, + "loss": 0.5839, + "step": 15916 + }, + { + "epoch": 5.88, + "learning_rate": 2.413708584801084e-06, + "loss": 0.3484, + "step": 15917 + }, + { + "epoch": 5.88, + "learning_rate": 2.4136716344377387e-06, + "loss": 0.4937, + "step": 15918 + }, + { + "epoch": 5.88, + "learning_rate": 2.4136346840743937e-06, + "loss": 0.5615, + "step": 15919 + }, + { + "epoch": 5.88, + "learning_rate": 2.4135977337110484e-06, + "loss": 0.4525, + "step": 15920 + }, + { + "epoch": 5.88, + "learning_rate": 2.413560783347703e-06, + "loss": 0.5404, + "step": 15921 + }, + { + "epoch": 5.88, + "learning_rate": 2.4135238329843576e-06, + "loss": 0.7021, + "step": 15922 + }, + { + "epoch": 5.88, + "learning_rate": 2.4134868826210123e-06, + "loss": 0.4677, + "step": 15923 + }, + { + "epoch": 5.88, + "learning_rate": 2.4134499322576673e-06, + "loss": 0.46, + "step": 15924 + }, + { + "epoch": 5.88, + "learning_rate": 2.413412981894322e-06, + "loss": 0.3819, + "step": 15925 + }, + { + "epoch": 5.88, + "learning_rate": 2.413376031530977e-06, + "loss": 0.5433, + "step": 15926 + }, + { + "epoch": 5.88, + "learning_rate": 2.4133390811676317e-06, + "loss": 0.535, + "step": 15927 + }, + { + "epoch": 5.88, + "learning_rate": 2.4133021308042863e-06, + "loss": 0.4765, + "step": 15928 + }, + { + "epoch": 5.88, + "learning_rate": 2.413265180440941e-06, + "loss": 0.5699, + "step": 15929 + }, + { + "epoch": 5.88, + "learning_rate": 2.413228230077596e-06, + "loss": 0.4716, + "step": 15930 + }, + { + "epoch": 5.88, + "learning_rate": 2.4131912797142506e-06, + "loss": 0.507, + "step": 15931 + }, + { + "epoch": 5.88, + "learning_rate": 2.4131543293509053e-06, + "loss": 0.6382, + "step": 15932 + }, + { + "epoch": 5.88, + "learning_rate": 2.4131173789875603e-06, + "loss": 0.4476, + "step": 15933 + }, + { + "epoch": 5.88, + "learning_rate": 2.4130804286242145e-06, + "loss": 0.5103, + "step": 15934 + }, + { + "epoch": 5.88, + "learning_rate": 2.4130434782608696e-06, + "loss": 0.4126, + "step": 15935 + }, + { + "epoch": 5.88, + "learning_rate": 2.4130065278975242e-06, + "loss": 0.4079, + "step": 15936 + }, + { + "epoch": 5.89, + "learning_rate": 2.4129695775341793e-06, + "loss": 0.4819, + "step": 15937 + }, + { + "epoch": 5.89, + "learning_rate": 2.412932627170834e-06, + "loss": 0.4258, + "step": 15938 + }, + { + "epoch": 5.89, + "learning_rate": 2.412895676807489e-06, + "loss": 0.4828, + "step": 15939 + }, + { + "epoch": 5.89, + "learning_rate": 2.4128587264441432e-06, + "loss": 0.5737, + "step": 15940 + }, + { + "epoch": 5.89, + "learning_rate": 2.4128217760807983e-06, + "loss": 0.4671, + "step": 15941 + }, + { + "epoch": 5.89, + "learning_rate": 2.412784825717453e-06, + "loss": 0.4778, + "step": 15942 + }, + { + "epoch": 5.89, + "learning_rate": 2.4127478753541075e-06, + "loss": 0.5053, + "step": 15943 + }, + { + "epoch": 5.89, + "learning_rate": 2.4127109249907626e-06, + "loss": 0.5816, + "step": 15944 + }, + { + "epoch": 5.89, + "learning_rate": 2.4126739746274172e-06, + "loss": 0.5232, + "step": 15945 + }, + { + "epoch": 5.89, + "learning_rate": 2.412637024264072e-06, + "loss": 0.4109, + "step": 15946 + }, + { + "epoch": 5.89, + "learning_rate": 2.4126000739007265e-06, + "loss": 0.4316, + "step": 15947 + }, + { + "epoch": 5.89, + "learning_rate": 2.4125631235373816e-06, + "loss": 0.44, + "step": 15948 + }, + { + "epoch": 5.89, + "learning_rate": 2.4125261731740362e-06, + "loss": 0.4656, + "step": 15949 + }, + { + "epoch": 5.89, + "learning_rate": 2.4124892228106913e-06, + "loss": 0.5192, + "step": 15950 + }, + { + "epoch": 5.89, + "learning_rate": 2.412452272447346e-06, + "loss": 0.3567, + "step": 15951 + }, + { + "epoch": 5.89, + "learning_rate": 2.4124153220840005e-06, + "loss": 0.5656, + "step": 15952 + }, + { + "epoch": 5.89, + "learning_rate": 2.412378371720655e-06, + "loss": 0.4912, + "step": 15953 + }, + { + "epoch": 5.89, + "learning_rate": 2.41234142135731e-06, + "loss": 0.5328, + "step": 15954 + }, + { + "epoch": 5.89, + "learning_rate": 2.412304470993965e-06, + "loss": 0.4785, + "step": 15955 + }, + { + "epoch": 5.89, + "learning_rate": 2.4122675206306195e-06, + "loss": 0.4769, + "step": 15956 + }, + { + "epoch": 5.89, + "learning_rate": 2.4122305702672746e-06, + "loss": 0.5679, + "step": 15957 + }, + { + "epoch": 5.89, + "learning_rate": 2.412193619903929e-06, + "loss": 0.4586, + "step": 15958 + }, + { + "epoch": 5.89, + "learning_rate": 2.412156669540584e-06, + "loss": 0.5702, + "step": 15959 + }, + { + "epoch": 5.89, + "learning_rate": 2.4121197191772385e-06, + "loss": 0.5204, + "step": 15960 + }, + { + "epoch": 5.89, + "learning_rate": 2.4120827688138935e-06, + "loss": 0.5955, + "step": 15961 + }, + { + "epoch": 5.89, + "learning_rate": 2.412045818450548e-06, + "loss": 0.4249, + "step": 15962 + }, + { + "epoch": 5.89, + "learning_rate": 2.412008868087203e-06, + "loss": 0.522, + "step": 15963 + }, + { + "epoch": 5.9, + "learning_rate": 2.411971917723858e-06, + "loss": 0.4177, + "step": 15964 + }, + { + "epoch": 5.9, + "learning_rate": 2.4119349673605125e-06, + "loss": 0.501, + "step": 15965 + }, + { + "epoch": 5.9, + "learning_rate": 2.411898016997167e-06, + "loss": 0.5466, + "step": 15966 + }, + { + "epoch": 5.9, + "learning_rate": 2.4118610666338218e-06, + "loss": 0.4682, + "step": 15967 + }, + { + "epoch": 5.9, + "learning_rate": 2.411824116270477e-06, + "loss": 0.5109, + "step": 15968 + }, + { + "epoch": 5.9, + "learning_rate": 2.4117871659071315e-06, + "loss": 0.4079, + "step": 15969 + }, + { + "epoch": 5.9, + "learning_rate": 2.4117502155437865e-06, + "loss": 0.5757, + "step": 15970 + }, + { + "epoch": 5.9, + "learning_rate": 2.411713265180441e-06, + "loss": 0.4994, + "step": 15971 + }, + { + "epoch": 5.9, + "learning_rate": 2.411676314817096e-06, + "loss": 0.4854, + "step": 15972 + }, + { + "epoch": 5.9, + "learning_rate": 2.4116393644537505e-06, + "loss": 0.4767, + "step": 15973 + }, + { + "epoch": 5.9, + "learning_rate": 2.411602414090405e-06, + "loss": 0.5374, + "step": 15974 + }, + { + "epoch": 5.9, + "learning_rate": 2.41156546372706e-06, + "loss": 0.5012, + "step": 15975 + }, + { + "epoch": 5.9, + "learning_rate": 2.4115285133637148e-06, + "loss": 0.5478, + "step": 15976 + }, + { + "epoch": 5.9, + "learning_rate": 2.41149156300037e-06, + "loss": 0.4613, + "step": 15977 + }, + { + "epoch": 5.9, + "learning_rate": 2.4114546126370245e-06, + "loss": 0.4795, + "step": 15978 + }, + { + "epoch": 5.9, + "learning_rate": 2.411417662273679e-06, + "loss": 0.4641, + "step": 15979 + }, + { + "epoch": 5.9, + "learning_rate": 2.4113807119103338e-06, + "loss": 0.4193, + "step": 15980 + }, + { + "epoch": 5.9, + "learning_rate": 2.411343761546989e-06, + "loss": 0.4238, + "step": 15981 + }, + { + "epoch": 5.9, + "learning_rate": 2.4113068111836434e-06, + "loss": 0.5381, + "step": 15982 + }, + { + "epoch": 5.9, + "learning_rate": 2.411269860820298e-06, + "loss": 0.4324, + "step": 15983 + }, + { + "epoch": 5.9, + "learning_rate": 2.4112329104569527e-06, + "loss": 0.4904, + "step": 15984 + }, + { + "epoch": 5.9, + "learning_rate": 2.4111959600936074e-06, + "loss": 0.4451, + "step": 15985 + }, + { + "epoch": 5.9, + "learning_rate": 2.4111590097302624e-06, + "loss": 0.4315, + "step": 15986 + }, + { + "epoch": 5.9, + "learning_rate": 2.411122059366917e-06, + "loss": 0.4419, + "step": 15987 + }, + { + "epoch": 5.9, + "learning_rate": 2.411085109003572e-06, + "loss": 0.5235, + "step": 15988 + }, + { + "epoch": 5.9, + "learning_rate": 2.4110481586402268e-06, + "loss": 0.4468, + "step": 15989 + }, + { + "epoch": 5.9, + "learning_rate": 2.4110112082768814e-06, + "loss": 0.3975, + "step": 15990 + }, + { + "epoch": 5.91, + "learning_rate": 2.410974257913536e-06, + "loss": 0.5532, + "step": 15991 + }, + { + "epoch": 5.91, + "learning_rate": 2.4109373075501907e-06, + "loss": 0.443, + "step": 15992 + }, + { + "epoch": 5.91, + "learning_rate": 2.4109003571868457e-06, + "loss": 0.5023, + "step": 15993 + }, + { + "epoch": 5.91, + "learning_rate": 2.4108634068235004e-06, + "loss": 0.4162, + "step": 15994 + }, + { + "epoch": 5.91, + "learning_rate": 2.4108264564601554e-06, + "loss": 0.5282, + "step": 15995 + }, + { + "epoch": 5.91, + "learning_rate": 2.41078950609681e-06, + "loss": 0.4438, + "step": 15996 + }, + { + "epoch": 5.91, + "learning_rate": 2.4107525557334647e-06, + "loss": 0.3854, + "step": 15997 + }, + { + "epoch": 5.91, + "learning_rate": 2.4107156053701193e-06, + "loss": 0.3754, + "step": 15998 + }, + { + "epoch": 5.91, + "learning_rate": 2.4106786550067744e-06, + "loss": 0.5517, + "step": 15999 + }, + { + "epoch": 5.91, + "learning_rate": 2.410641704643429e-06, + "loss": 0.466, + "step": 16000 + }, + { + "epoch": 5.91, + "learning_rate": 2.410604754280084e-06, + "loss": 0.4723, + "step": 16001 + }, + { + "epoch": 5.91, + "learning_rate": 2.4105678039167387e-06, + "loss": 0.4874, + "step": 16002 + }, + { + "epoch": 5.91, + "learning_rate": 2.4105308535533934e-06, + "loss": 0.4363, + "step": 16003 + }, + { + "epoch": 5.91, + "learning_rate": 2.410493903190048e-06, + "loss": 0.4859, + "step": 16004 + }, + { + "epoch": 5.91, + "learning_rate": 2.4104569528267026e-06, + "loss": 0.413, + "step": 16005 + }, + { + "epoch": 5.91, + "learning_rate": 2.4104200024633577e-06, + "loss": 0.441, + "step": 16006 + }, + { + "epoch": 5.91, + "learning_rate": 2.4103830521000123e-06, + "loss": 0.4896, + "step": 16007 + }, + { + "epoch": 5.91, + "learning_rate": 2.4103461017366674e-06, + "loss": 0.4631, + "step": 16008 + }, + { + "epoch": 5.91, + "learning_rate": 2.410309151373322e-06, + "loss": 0.4549, + "step": 16009 + }, + { + "epoch": 5.91, + "learning_rate": 2.4102722010099767e-06, + "loss": 0.5978, + "step": 16010 + }, + { + "epoch": 5.91, + "learning_rate": 2.4102352506466313e-06, + "loss": 0.4275, + "step": 16011 + }, + { + "epoch": 5.91, + "learning_rate": 2.410198300283286e-06, + "loss": 0.4903, + "step": 16012 + }, + { + "epoch": 5.91, + "learning_rate": 2.410161349919941e-06, + "loss": 0.431, + "step": 16013 + }, + { + "epoch": 5.91, + "learning_rate": 2.4101243995565956e-06, + "loss": 0.5703, + "step": 16014 + }, + { + "epoch": 5.91, + "learning_rate": 2.4100874491932507e-06, + "loss": 0.4006, + "step": 16015 + }, + { + "epoch": 5.91, + "learning_rate": 2.4100504988299053e-06, + "loss": 0.5557, + "step": 16016 + }, + { + "epoch": 5.91, + "learning_rate": 2.41001354846656e-06, + "loss": 0.5434, + "step": 16017 + }, + { + "epoch": 5.92, + "learning_rate": 2.4099765981032146e-06, + "loss": 0.4235, + "step": 16018 + }, + { + "epoch": 5.92, + "learning_rate": 2.4099396477398697e-06, + "loss": 0.4561, + "step": 16019 + }, + { + "epoch": 5.92, + "learning_rate": 2.4099026973765243e-06, + "loss": 0.4781, + "step": 16020 + }, + { + "epoch": 5.92, + "learning_rate": 2.4098657470131794e-06, + "loss": 0.5029, + "step": 16021 + }, + { + "epoch": 5.92, + "learning_rate": 2.409828796649834e-06, + "loss": 0.5465, + "step": 16022 + }, + { + "epoch": 5.92, + "learning_rate": 2.409791846286488e-06, + "loss": 0.5111, + "step": 16023 + }, + { + "epoch": 5.92, + "learning_rate": 2.4097548959231433e-06, + "loss": 0.5174, + "step": 16024 + }, + { + "epoch": 5.92, + "learning_rate": 2.409717945559798e-06, + "loss": 0.4, + "step": 16025 + }, + { + "epoch": 5.92, + "learning_rate": 2.409680995196453e-06, + "loss": 0.579, + "step": 16026 + }, + { + "epoch": 5.92, + "learning_rate": 2.4096440448331076e-06, + "loss": 0.4336, + "step": 16027 + }, + { + "epoch": 5.92, + "learning_rate": 2.4096070944697627e-06, + "loss": 0.5982, + "step": 16028 + }, + { + "epoch": 5.92, + "learning_rate": 2.409570144106417e-06, + "loss": 0.5633, + "step": 16029 + }, + { + "epoch": 5.92, + "learning_rate": 2.409533193743072e-06, + "loss": 0.5443, + "step": 16030 + }, + { + "epoch": 5.92, + "learning_rate": 2.4094962433797266e-06, + "loss": 0.4499, + "step": 16031 + }, + { + "epoch": 5.92, + "learning_rate": 2.409459293016381e-06, + "loss": 0.4502, + "step": 16032 + }, + { + "epoch": 5.92, + "learning_rate": 2.4094223426530363e-06, + "loss": 0.614, + "step": 16033 + }, + { + "epoch": 5.92, + "learning_rate": 2.409385392289691e-06, + "loss": 0.5906, + "step": 16034 + }, + { + "epoch": 5.92, + "learning_rate": 2.4093484419263455e-06, + "loss": 0.5382, + "step": 16035 + }, + { + "epoch": 5.92, + "learning_rate": 2.409311491563e-06, + "loss": 0.5585, + "step": 16036 + }, + { + "epoch": 5.92, + "learning_rate": 2.4092745411996552e-06, + "loss": 0.3836, + "step": 16037 + }, + { + "epoch": 5.92, + "learning_rate": 2.40923759083631e-06, + "loss": 0.5568, + "step": 16038 + }, + { + "epoch": 5.92, + "learning_rate": 2.409200640472965e-06, + "loss": 0.4201, + "step": 16039 + }, + { + "epoch": 5.92, + "learning_rate": 2.4091636901096196e-06, + "loss": 0.4722, + "step": 16040 + }, + { + "epoch": 5.92, + "learning_rate": 2.409126739746274e-06, + "loss": 0.486, + "step": 16041 + }, + { + "epoch": 5.92, + "learning_rate": 2.409089789382929e-06, + "loss": 0.4263, + "step": 16042 + }, + { + "epoch": 5.92, + "learning_rate": 2.4090528390195835e-06, + "loss": 0.3517, + "step": 16043 + }, + { + "epoch": 5.92, + "learning_rate": 2.4090158886562385e-06, + "loss": 0.4813, + "step": 16044 + }, + { + "epoch": 5.93, + "learning_rate": 2.408978938292893e-06, + "loss": 0.4999, + "step": 16045 + }, + { + "epoch": 5.93, + "learning_rate": 2.4089419879295482e-06, + "loss": 0.5047, + "step": 16046 + }, + { + "epoch": 5.93, + "learning_rate": 2.408905037566203e-06, + "loss": 0.5141, + "step": 16047 + }, + { + "epoch": 5.93, + "learning_rate": 2.4088680872028575e-06, + "loss": 0.5012, + "step": 16048 + }, + { + "epoch": 5.93, + "learning_rate": 2.408831136839512e-06, + "loss": 0.5509, + "step": 16049 + }, + { + "epoch": 5.93, + "learning_rate": 2.408794186476167e-06, + "loss": 0.5882, + "step": 16050 + }, + { + "epoch": 5.93, + "learning_rate": 2.408757236112822e-06, + "loss": 0.4223, + "step": 16051 + }, + { + "epoch": 5.93, + "learning_rate": 2.4087202857494765e-06, + "loss": 0.4735, + "step": 16052 + }, + { + "epoch": 5.93, + "learning_rate": 2.4086833353861315e-06, + "loss": 0.4605, + "step": 16053 + }, + { + "epoch": 5.93, + "learning_rate": 2.408646385022786e-06, + "loss": 0.3889, + "step": 16054 + }, + { + "epoch": 5.93, + "learning_rate": 2.408609434659441e-06, + "loss": 0.4374, + "step": 16055 + }, + { + "epoch": 5.93, + "learning_rate": 2.4085724842960954e-06, + "loss": 0.4426, + "step": 16056 + }, + { + "epoch": 5.93, + "learning_rate": 2.4085355339327505e-06, + "loss": 0.3942, + "step": 16057 + }, + { + "epoch": 5.93, + "learning_rate": 2.408498583569405e-06, + "loss": 0.4529, + "step": 16058 + }, + { + "epoch": 5.93, + "learning_rate": 2.40846163320606e-06, + "loss": 0.5173, + "step": 16059 + }, + { + "epoch": 5.93, + "learning_rate": 2.408424682842715e-06, + "loss": 0.4653, + "step": 16060 + }, + { + "epoch": 5.93, + "learning_rate": 2.4083877324793695e-06, + "loss": 0.4839, + "step": 16061 + }, + { + "epoch": 5.93, + "learning_rate": 2.408350782116024e-06, + "loss": 0.5123, + "step": 16062 + }, + { + "epoch": 5.93, + "learning_rate": 2.4083138317526787e-06, + "loss": 0.572, + "step": 16063 + }, + { + "epoch": 5.93, + "learning_rate": 2.408276881389334e-06, + "loss": 0.5584, + "step": 16064 + }, + { + "epoch": 5.93, + "learning_rate": 2.4082399310259884e-06, + "loss": 0.4347, + "step": 16065 + }, + { + "epoch": 5.93, + "learning_rate": 2.4082029806626435e-06, + "loss": 0.433, + "step": 16066 + }, + { + "epoch": 5.93, + "learning_rate": 2.408166030299298e-06, + "loss": 0.5181, + "step": 16067 + }, + { + "epoch": 5.93, + "learning_rate": 2.4081290799359528e-06, + "loss": 0.4792, + "step": 16068 + }, + { + "epoch": 5.93, + "learning_rate": 2.4080921295726074e-06, + "loss": 0.4401, + "step": 16069 + }, + { + "epoch": 5.93, + "learning_rate": 2.4080551792092625e-06, + "loss": 0.5431, + "step": 16070 + }, + { + "epoch": 5.93, + "learning_rate": 2.408018228845917e-06, + "loss": 0.5349, + "step": 16071 + }, + { + "epoch": 5.94, + "learning_rate": 2.4079812784825717e-06, + "loss": 0.4552, + "step": 16072 + }, + { + "epoch": 5.94, + "learning_rate": 2.407944328119227e-06, + "loss": 0.5227, + "step": 16073 + }, + { + "epoch": 5.94, + "learning_rate": 2.407907377755881e-06, + "loss": 0.428, + "step": 16074 + }, + { + "epoch": 5.94, + "learning_rate": 2.407870427392536e-06, + "loss": 0.4722, + "step": 16075 + }, + { + "epoch": 5.94, + "learning_rate": 2.4078334770291907e-06, + "loss": 0.3787, + "step": 16076 + }, + { + "epoch": 5.94, + "learning_rate": 2.4077965266658458e-06, + "loss": 0.5295, + "step": 16077 + }, + { + "epoch": 5.94, + "learning_rate": 2.4077595763025004e-06, + "loss": 0.4907, + "step": 16078 + }, + { + "epoch": 5.94, + "learning_rate": 2.4077226259391555e-06, + "loss": 0.4526, + "step": 16079 + }, + { + "epoch": 5.94, + "learning_rate": 2.4076856755758097e-06, + "loss": 0.5099, + "step": 16080 + }, + { + "epoch": 5.94, + "learning_rate": 2.4076487252124643e-06, + "loss": 0.4484, + "step": 16081 + }, + { + "epoch": 5.94, + "learning_rate": 2.4076117748491194e-06, + "loss": 0.4135, + "step": 16082 + }, + { + "epoch": 5.94, + "learning_rate": 2.407574824485774e-06, + "loss": 0.3829, + "step": 16083 + }, + { + "epoch": 5.94, + "learning_rate": 2.407537874122429e-06, + "loss": 0.6012, + "step": 16084 + }, + { + "epoch": 5.94, + "learning_rate": 2.4075009237590837e-06, + "loss": 0.529, + "step": 16085 + }, + { + "epoch": 5.94, + "learning_rate": 2.4074639733957383e-06, + "loss": 0.4784, + "step": 16086 + }, + { + "epoch": 5.94, + "learning_rate": 2.407427023032393e-06, + "loss": 0.3418, + "step": 16087 + }, + { + "epoch": 5.94, + "learning_rate": 2.407390072669048e-06, + "loss": 0.4493, + "step": 16088 + }, + { + "epoch": 5.94, + "learning_rate": 2.4073531223057027e-06, + "loss": 0.4484, + "step": 16089 + }, + { + "epoch": 5.94, + "learning_rate": 2.4073161719423577e-06, + "loss": 0.5657, + "step": 16090 + }, + { + "epoch": 5.94, + "learning_rate": 2.4072792215790124e-06, + "loss": 0.3717, + "step": 16091 + }, + { + "epoch": 5.94, + "learning_rate": 2.407242271215667e-06, + "loss": 0.5024, + "step": 16092 + }, + { + "epoch": 5.94, + "learning_rate": 2.4072053208523216e-06, + "loss": 0.5665, + "step": 16093 + }, + { + "epoch": 5.94, + "learning_rate": 2.4071683704889763e-06, + "loss": 0.5402, + "step": 16094 + }, + { + "epoch": 5.94, + "learning_rate": 2.4071314201256313e-06, + "loss": 0.4006, + "step": 16095 + }, + { + "epoch": 5.94, + "learning_rate": 2.407094469762286e-06, + "loss": 0.6256, + "step": 16096 + }, + { + "epoch": 5.94, + "learning_rate": 2.407057519398941e-06, + "loss": 0.4003, + "step": 16097 + }, + { + "epoch": 5.94, + "learning_rate": 2.4070205690355957e-06, + "loss": 0.4756, + "step": 16098 + }, + { + "epoch": 5.94, + "learning_rate": 2.4069836186722503e-06, + "loss": 0.3706, + "step": 16099 + }, + { + "epoch": 5.95, + "learning_rate": 2.406946668308905e-06, + "loss": 0.5276, + "step": 16100 + }, + { + "epoch": 5.95, + "learning_rate": 2.4069097179455596e-06, + "loss": 0.5117, + "step": 16101 + }, + { + "epoch": 5.95, + "learning_rate": 2.4068727675822146e-06, + "loss": 0.5682, + "step": 16102 + }, + { + "epoch": 5.95, + "learning_rate": 2.4068358172188693e-06, + "loss": 0.3836, + "step": 16103 + }, + { + "epoch": 5.95, + "learning_rate": 2.4067988668555243e-06, + "loss": 0.5203, + "step": 16104 + }, + { + "epoch": 5.95, + "learning_rate": 2.406761916492179e-06, + "loss": 0.4155, + "step": 16105 + }, + { + "epoch": 5.95, + "learning_rate": 2.4067249661288336e-06, + "loss": 0.367, + "step": 16106 + }, + { + "epoch": 5.95, + "learning_rate": 2.4066880157654883e-06, + "loss": 0.5619, + "step": 16107 + }, + { + "epoch": 5.95, + "learning_rate": 2.4066510654021433e-06, + "loss": 0.5391, + "step": 16108 + }, + { + "epoch": 5.95, + "learning_rate": 2.406614115038798e-06, + "loss": 0.3912, + "step": 16109 + }, + { + "epoch": 5.95, + "learning_rate": 2.406577164675453e-06, + "loss": 0.5439, + "step": 16110 + }, + { + "epoch": 5.95, + "learning_rate": 2.4065402143121076e-06, + "loss": 0.4591, + "step": 16111 + }, + { + "epoch": 5.95, + "learning_rate": 2.4065032639487623e-06, + "loss": 0.4662, + "step": 16112 + }, + { + "epoch": 5.95, + "learning_rate": 2.406466313585417e-06, + "loss": 0.49, + "step": 16113 + }, + { + "epoch": 5.95, + "learning_rate": 2.4064293632220716e-06, + "loss": 0.4364, + "step": 16114 + }, + { + "epoch": 5.95, + "learning_rate": 2.4063924128587266e-06, + "loss": 0.567, + "step": 16115 + }, + { + "epoch": 5.95, + "learning_rate": 2.4063554624953812e-06, + "loss": 0.444, + "step": 16116 + }, + { + "epoch": 5.95, + "learning_rate": 2.4063185121320363e-06, + "loss": 0.4987, + "step": 16117 + }, + { + "epoch": 5.95, + "learning_rate": 2.406281561768691e-06, + "loss": 0.4388, + "step": 16118 + }, + { + "epoch": 5.95, + "learning_rate": 2.4062446114053456e-06, + "loss": 0.3438, + "step": 16119 + }, + { + "epoch": 5.95, + "learning_rate": 2.4062076610420002e-06, + "loss": 0.3865, + "step": 16120 + }, + { + "epoch": 5.95, + "learning_rate": 2.406170710678655e-06, + "loss": 0.4237, + "step": 16121 + }, + { + "epoch": 5.95, + "learning_rate": 2.40613376031531e-06, + "loss": 0.4638, + "step": 16122 + }, + { + "epoch": 5.95, + "learning_rate": 2.4060968099519646e-06, + "loss": 0.3508, + "step": 16123 + }, + { + "epoch": 5.95, + "learning_rate": 2.4060598595886196e-06, + "loss": 0.4248, + "step": 16124 + }, + { + "epoch": 5.95, + "learning_rate": 2.406022909225274e-06, + "loss": 0.5484, + "step": 16125 + }, + { + "epoch": 5.95, + "learning_rate": 2.405985958861929e-06, + "loss": 0.4426, + "step": 16126 + }, + { + "epoch": 5.96, + "learning_rate": 2.4059490084985835e-06, + "loss": 0.4043, + "step": 16127 + }, + { + "epoch": 5.96, + "learning_rate": 2.4059120581352386e-06, + "loss": 0.6151, + "step": 16128 + }, + { + "epoch": 5.96, + "learning_rate": 2.4058751077718932e-06, + "loss": 0.4735, + "step": 16129 + }, + { + "epoch": 5.96, + "learning_rate": 2.4058381574085483e-06, + "loss": 0.4683, + "step": 16130 + }, + { + "epoch": 5.96, + "learning_rate": 2.4058012070452025e-06, + "loss": 0.4976, + "step": 16131 + }, + { + "epoch": 5.96, + "learning_rate": 2.405764256681857e-06, + "loss": 0.5198, + "step": 16132 + }, + { + "epoch": 5.96, + "learning_rate": 2.405727306318512e-06, + "loss": 0.4357, + "step": 16133 + }, + { + "epoch": 5.96, + "learning_rate": 2.405690355955167e-06, + "loss": 0.6598, + "step": 16134 + }, + { + "epoch": 5.96, + "learning_rate": 2.405653405591822e-06, + "loss": 0.4965, + "step": 16135 + }, + { + "epoch": 5.96, + "learning_rate": 2.4056164552284765e-06, + "loss": 0.4699, + "step": 16136 + }, + { + "epoch": 5.96, + "learning_rate": 2.405579504865131e-06, + "loss": 0.5353, + "step": 16137 + }, + { + "epoch": 5.96, + "learning_rate": 2.405542554501786e-06, + "loss": 0.4996, + "step": 16138 + }, + { + "epoch": 5.96, + "learning_rate": 2.405505604138441e-06, + "loss": 0.518, + "step": 16139 + }, + { + "epoch": 5.96, + "learning_rate": 2.4054686537750955e-06, + "loss": 0.4615, + "step": 16140 + }, + { + "epoch": 5.96, + "learning_rate": 2.40543170341175e-06, + "loss": 0.4354, + "step": 16141 + }, + { + "epoch": 5.96, + "learning_rate": 2.405394753048405e-06, + "loss": 0.5736, + "step": 16142 + }, + { + "epoch": 5.96, + "learning_rate": 2.40535780268506e-06, + "loss": 0.6244, + "step": 16143 + }, + { + "epoch": 5.96, + "learning_rate": 2.4053208523217145e-06, + "loss": 0.4852, + "step": 16144 + }, + { + "epoch": 5.96, + "learning_rate": 2.405283901958369e-06, + "loss": 0.4984, + "step": 16145 + }, + { + "epoch": 5.96, + "learning_rate": 2.405246951595024e-06, + "loss": 0.4984, + "step": 16146 + }, + { + "epoch": 5.96, + "learning_rate": 2.405210001231679e-06, + "loss": 0.402, + "step": 16147 + }, + { + "epoch": 5.96, + "learning_rate": 2.405173050868334e-06, + "loss": 0.4922, + "step": 16148 + }, + { + "epoch": 5.96, + "learning_rate": 2.4051361005049885e-06, + "loss": 0.6325, + "step": 16149 + }, + { + "epoch": 5.96, + "learning_rate": 2.405099150141643e-06, + "loss": 0.5039, + "step": 16150 + }, + { + "epoch": 5.96, + "learning_rate": 2.4050621997782978e-06, + "loss": 0.5203, + "step": 16151 + }, + { + "epoch": 5.96, + "learning_rate": 2.4050252494149524e-06, + "loss": 0.5391, + "step": 16152 + }, + { + "epoch": 5.96, + "learning_rate": 2.4049882990516075e-06, + "loss": 0.496, + "step": 16153 + }, + { + "epoch": 5.97, + "learning_rate": 2.404951348688262e-06, + "loss": 0.5052, + "step": 16154 + }, + { + "epoch": 5.97, + "learning_rate": 2.404914398324917e-06, + "loss": 0.5251, + "step": 16155 + }, + { + "epoch": 5.97, + "learning_rate": 2.404877447961572e-06, + "loss": 0.56, + "step": 16156 + }, + { + "epoch": 5.97, + "learning_rate": 2.4048404975982264e-06, + "loss": 0.4722, + "step": 16157 + }, + { + "epoch": 5.97, + "learning_rate": 2.404803547234881e-06, + "loss": 0.6368, + "step": 16158 + }, + { + "epoch": 5.97, + "learning_rate": 2.404766596871536e-06, + "loss": 0.4716, + "step": 16159 + }, + { + "epoch": 5.97, + "learning_rate": 2.4047296465081908e-06, + "loss": 0.4937, + "step": 16160 + }, + { + "epoch": 5.97, + "learning_rate": 2.4046926961448454e-06, + "loss": 0.5074, + "step": 16161 + }, + { + "epoch": 5.97, + "learning_rate": 2.4046557457815005e-06, + "loss": 0.4997, + "step": 16162 + }, + { + "epoch": 5.97, + "learning_rate": 2.404618795418155e-06, + "loss": 0.4201, + "step": 16163 + }, + { + "epoch": 5.97, + "learning_rate": 2.4045818450548097e-06, + "loss": 0.3264, + "step": 16164 + }, + { + "epoch": 5.97, + "learning_rate": 2.4045448946914644e-06, + "loss": 0.4695, + "step": 16165 + }, + { + "epoch": 5.97, + "learning_rate": 2.4045079443281194e-06, + "loss": 0.5139, + "step": 16166 + }, + { + "epoch": 5.97, + "learning_rate": 2.404470993964774e-06, + "loss": 0.3774, + "step": 16167 + }, + { + "epoch": 5.97, + "learning_rate": 2.404434043601429e-06, + "loss": 0.4428, + "step": 16168 + }, + { + "epoch": 5.97, + "learning_rate": 2.4043970932380833e-06, + "loss": 0.5044, + "step": 16169 + }, + { + "epoch": 5.97, + "learning_rate": 2.4043601428747384e-06, + "loss": 0.4571, + "step": 16170 + }, + { + "epoch": 5.97, + "learning_rate": 2.404323192511393e-06, + "loss": 0.4831, + "step": 16171 + }, + { + "epoch": 5.97, + "learning_rate": 2.4042862421480477e-06, + "loss": 0.4164, + "step": 16172 + }, + { + "epoch": 5.97, + "learning_rate": 2.4042492917847027e-06, + "loss": 0.4753, + "step": 16173 + }, + { + "epoch": 5.97, + "learning_rate": 2.4042123414213574e-06, + "loss": 0.5457, + "step": 16174 + }, + { + "epoch": 5.97, + "learning_rate": 2.404175391058012e-06, + "loss": 0.4566, + "step": 16175 + }, + { + "epoch": 5.97, + "learning_rate": 2.4041384406946666e-06, + "loss": 0.4271, + "step": 16176 + }, + { + "epoch": 5.97, + "learning_rate": 2.4041014903313217e-06, + "loss": 0.4833, + "step": 16177 + }, + { + "epoch": 5.97, + "learning_rate": 2.4040645399679763e-06, + "loss": 0.5587, + "step": 16178 + }, + { + "epoch": 5.97, + "learning_rate": 2.4040275896046314e-06, + "loss": 0.4854, + "step": 16179 + }, + { + "epoch": 5.97, + "learning_rate": 2.403990639241286e-06, + "loss": 0.4947, + "step": 16180 + }, + { + "epoch": 5.98, + "learning_rate": 2.4039536888779407e-06, + "loss": 0.3566, + "step": 16181 + }, + { + "epoch": 5.98, + "learning_rate": 2.4039167385145953e-06, + "loss": 0.534, + "step": 16182 + }, + { + "epoch": 5.98, + "learning_rate": 2.40387978815125e-06, + "loss": 0.4201, + "step": 16183 + }, + { + "epoch": 5.98, + "learning_rate": 2.403842837787905e-06, + "loss": 0.4456, + "step": 16184 + }, + { + "epoch": 5.98, + "learning_rate": 2.4038058874245596e-06, + "loss": 0.5917, + "step": 16185 + }, + { + "epoch": 5.98, + "learning_rate": 2.4037689370612147e-06, + "loss": 0.5245, + "step": 16186 + }, + { + "epoch": 5.98, + "learning_rate": 2.4037319866978693e-06, + "loss": 0.3685, + "step": 16187 + }, + { + "epoch": 5.98, + "learning_rate": 2.403695036334524e-06, + "loss": 0.4906, + "step": 16188 + }, + { + "epoch": 5.98, + "learning_rate": 2.4036580859711786e-06, + "loss": 0.4281, + "step": 16189 + }, + { + "epoch": 5.98, + "learning_rate": 2.4036211356078337e-06, + "loss": 0.4243, + "step": 16190 + }, + { + "epoch": 5.98, + "learning_rate": 2.4035841852444883e-06, + "loss": 0.4266, + "step": 16191 + }, + { + "epoch": 5.98, + "learning_rate": 2.403547234881143e-06, + "loss": 0.4601, + "step": 16192 + }, + { + "epoch": 5.98, + "learning_rate": 2.403510284517798e-06, + "loss": 0.4228, + "step": 16193 + }, + { + "epoch": 5.98, + "learning_rate": 2.4034733341544526e-06, + "loss": 0.5815, + "step": 16194 + }, + { + "epoch": 5.98, + "learning_rate": 2.4034363837911073e-06, + "loss": 0.407, + "step": 16195 + }, + { + "epoch": 5.98, + "learning_rate": 2.403399433427762e-06, + "loss": 0.396, + "step": 16196 + }, + { + "epoch": 5.98, + "learning_rate": 2.403362483064417e-06, + "loss": 0.3697, + "step": 16197 + }, + { + "epoch": 5.98, + "learning_rate": 2.4033255327010716e-06, + "loss": 0.3988, + "step": 16198 + }, + { + "epoch": 5.98, + "learning_rate": 2.4032885823377267e-06, + "loss": 0.4156, + "step": 16199 + }, + { + "epoch": 5.98, + "learning_rate": 2.4032516319743813e-06, + "loss": 0.3943, + "step": 16200 + }, + { + "epoch": 5.98, + "learning_rate": 2.403214681611036e-06, + "loss": 0.4697, + "step": 16201 + }, + { + "epoch": 5.98, + "learning_rate": 2.4031777312476906e-06, + "loss": 0.3309, + "step": 16202 + }, + { + "epoch": 5.98, + "learning_rate": 2.403140780884345e-06, + "loss": 0.3273, + "step": 16203 + }, + { + "epoch": 5.98, + "learning_rate": 2.4031038305210003e-06, + "loss": 0.4288, + "step": 16204 + }, + { + "epoch": 5.98, + "learning_rate": 2.403066880157655e-06, + "loss": 0.5201, + "step": 16205 + }, + { + "epoch": 5.98, + "learning_rate": 2.40302992979431e-06, + "loss": 0.4269, + "step": 16206 + }, + { + "epoch": 5.98, + "learning_rate": 2.4029929794309646e-06, + "loss": 0.5468, + "step": 16207 + }, + { + "epoch": 5.99, + "learning_rate": 2.4029560290676192e-06, + "loss": 0.4601, + "step": 16208 + }, + { + "epoch": 5.99, + "learning_rate": 2.402919078704274e-06, + "loss": 0.5964, + "step": 16209 + }, + { + "epoch": 5.99, + "learning_rate": 2.402882128340929e-06, + "loss": 0.4429, + "step": 16210 + }, + { + "epoch": 5.99, + "learning_rate": 2.4028451779775836e-06, + "loss": 0.5862, + "step": 16211 + }, + { + "epoch": 5.99, + "learning_rate": 2.402808227614238e-06, + "loss": 0.4421, + "step": 16212 + }, + { + "epoch": 5.99, + "learning_rate": 2.4027712772508933e-06, + "loss": 0.3269, + "step": 16213 + }, + { + "epoch": 5.99, + "learning_rate": 2.4027343268875475e-06, + "loss": 0.3691, + "step": 16214 + }, + { + "epoch": 5.99, + "learning_rate": 2.4026973765242025e-06, + "loss": 0.373, + "step": 16215 + }, + { + "epoch": 5.99, + "learning_rate": 2.402660426160857e-06, + "loss": 0.6885, + "step": 16216 + }, + { + "epoch": 5.99, + "learning_rate": 2.4026234757975122e-06, + "loss": 0.5063, + "step": 16217 + }, + { + "epoch": 5.99, + "learning_rate": 2.402586525434167e-06, + "loss": 0.5344, + "step": 16218 + }, + { + "epoch": 5.99, + "learning_rate": 2.402549575070822e-06, + "loss": 0.4854, + "step": 16219 + }, + { + "epoch": 5.99, + "learning_rate": 2.402512624707476e-06, + "loss": 0.3688, + "step": 16220 + }, + { + "epoch": 5.99, + "learning_rate": 2.4024756743441308e-06, + "loss": 0.5521, + "step": 16221 + }, + { + "epoch": 5.99, + "learning_rate": 2.402438723980786e-06, + "loss": 0.5142, + "step": 16222 + }, + { + "epoch": 5.99, + "learning_rate": 2.4024017736174405e-06, + "loss": 0.4962, + "step": 16223 + }, + { + "epoch": 5.99, + "learning_rate": 2.4023648232540955e-06, + "loss": 0.3658, + "step": 16224 + }, + { + "epoch": 5.99, + "learning_rate": 2.40232787289075e-06, + "loss": 0.4409, + "step": 16225 + }, + { + "epoch": 5.99, + "learning_rate": 2.402290922527405e-06, + "loss": 0.4603, + "step": 16226 + }, + { + "epoch": 5.99, + "learning_rate": 2.4022539721640594e-06, + "loss": 0.4289, + "step": 16227 + }, + { + "epoch": 5.99, + "learning_rate": 2.4022170218007145e-06, + "loss": 0.4604, + "step": 16228 + }, + { + "epoch": 5.99, + "learning_rate": 2.402180071437369e-06, + "loss": 0.4162, + "step": 16229 + }, + { + "epoch": 5.99, + "learning_rate": 2.402143121074024e-06, + "loss": 0.433, + "step": 16230 + }, + { + "epoch": 5.99, + "learning_rate": 2.402106170710679e-06, + "loss": 0.5125, + "step": 16231 + }, + { + "epoch": 5.99, + "learning_rate": 2.4020692203473335e-06, + "loss": 0.5532, + "step": 16232 + }, + { + "epoch": 5.99, + "learning_rate": 2.402032269983988e-06, + "loss": 0.4138, + "step": 16233 + }, + { + "epoch": 5.99, + "learning_rate": 2.4019953196206427e-06, + "loss": 0.4261, + "step": 16234 + }, + { + "epoch": 6.0, + "learning_rate": 2.401958369257298e-06, + "loss": 0.4191, + "step": 16235 + }, + { + "epoch": 6.0, + "learning_rate": 2.4019214188939524e-06, + "loss": 0.5417, + "step": 16236 + }, + { + "epoch": 6.0, + "learning_rate": 2.4018844685306075e-06, + "loss": 0.5491, + "step": 16237 + }, + { + "epoch": 6.0, + "learning_rate": 2.401847518167262e-06, + "loss": 0.4906, + "step": 16238 + }, + { + "epoch": 6.0, + "learning_rate": 2.4018105678039168e-06, + "loss": 0.3837, + "step": 16239 + }, + { + "epoch": 6.0, + "learning_rate": 2.4017736174405714e-06, + "loss": 0.4739, + "step": 16240 + }, + { + "epoch": 6.0, + "learning_rate": 2.401736667077226e-06, + "loss": 0.4377, + "step": 16241 + }, + { + "epoch": 6.0, + "learning_rate": 2.401699716713881e-06, + "loss": 0.4532, + "step": 16242 + }, + { + "epoch": 6.0, + "learning_rate": 2.4016627663505357e-06, + "loss": 0.4802, + "step": 16243 + }, + { + "epoch": 6.0, + "learning_rate": 2.401625815987191e-06, + "loss": 0.6144, + "step": 16244 + }, + { + "epoch": 6.0, + "learning_rate": 2.4015888656238454e-06, + "loss": 0.4835, + "step": 16245 + }, + { + "epoch": 6.0, + "learning_rate": 2.4015519152605e-06, + "loss": 0.4403, + "step": 16246 + }, + { + "epoch": 6.0, + "learning_rate": 2.4015149648971547e-06, + "loss": 0.5036, + "step": 16247 + }, + { + "epoch": 6.0, + "learning_rate": 2.4014780145338098e-06, + "loss": 0.5445, + "step": 16248 + }, + { + "epoch": 6.0, + "eval_accuracy": 0.7613935448123009, + "eval_loss": 0.49804672598838806, + "eval_runtime": 25.4167, + "eval_samples_per_second": 1704.158, + "eval_steps_per_second": 26.636, + "step": 16248 + }, + { + "epoch": 6.0, + "learning_rate": 2.4014410641704644e-06, + "loss": 0.4423, + "step": 16249 + }, + { + "epoch": 6.0, + "learning_rate": 2.4014041138071195e-06, + "loss": 0.4122, + "step": 16250 + }, + { + "epoch": 6.0, + "learning_rate": 2.401367163443774e-06, + "loss": 0.4867, + "step": 16251 + }, + { + "epoch": 6.0, + "learning_rate": 2.4013302130804287e-06, + "loss": 0.3995, + "step": 16252 + }, + { + "epoch": 6.0, + "learning_rate": 2.4012932627170834e-06, + "loss": 0.4687, + "step": 16253 + }, + { + "epoch": 6.0, + "learning_rate": 2.401256312353738e-06, + "loss": 0.4527, + "step": 16254 + }, + { + "epoch": 6.0, + "learning_rate": 2.401219361990393e-06, + "loss": 0.4136, + "step": 16255 + }, + { + "epoch": 6.0, + "learning_rate": 2.4011824116270477e-06, + "loss": 0.3543, + "step": 16256 + }, + { + "epoch": 6.0, + "learning_rate": 2.4011454612637028e-06, + "loss": 0.3853, + "step": 16257 + }, + { + "epoch": 6.0, + "learning_rate": 2.4011085109003574e-06, + "loss": 0.5257, + "step": 16258 + }, + { + "epoch": 6.0, + "learning_rate": 2.401071560537012e-06, + "loss": 0.3899, + "step": 16259 + }, + { + "epoch": 6.0, + "learning_rate": 2.4010346101736667e-06, + "loss": 0.5016, + "step": 16260 + }, + { + "epoch": 6.0, + "learning_rate": 2.4009976598103213e-06, + "loss": 0.5088, + "step": 16261 + }, + { + "epoch": 6.01, + "learning_rate": 2.4009607094469764e-06, + "loss": 0.3722, + "step": 16262 + }, + { + "epoch": 6.01, + "learning_rate": 2.400923759083631e-06, + "loss": 0.442, + "step": 16263 + }, + { + "epoch": 6.01, + "learning_rate": 2.400886808720286e-06, + "loss": 0.4377, + "step": 16264 + }, + { + "epoch": 6.01, + "learning_rate": 2.4008498583569403e-06, + "loss": 0.5773, + "step": 16265 + }, + { + "epoch": 6.01, + "learning_rate": 2.4008129079935953e-06, + "loss": 0.4184, + "step": 16266 + }, + { + "epoch": 6.01, + "learning_rate": 2.40077595763025e-06, + "loss": 0.4003, + "step": 16267 + }, + { + "epoch": 6.01, + "learning_rate": 2.400739007266905e-06, + "loss": 0.4214, + "step": 16268 + }, + { + "epoch": 6.01, + "learning_rate": 2.4007020569035597e-06, + "loss": 0.4307, + "step": 16269 + }, + { + "epoch": 6.01, + "learning_rate": 2.4006651065402147e-06, + "loss": 0.4877, + "step": 16270 + }, + { + "epoch": 6.01, + "learning_rate": 2.400628156176869e-06, + "loss": 0.4391, + "step": 16271 + }, + { + "epoch": 6.01, + "learning_rate": 2.4005912058135236e-06, + "loss": 0.4209, + "step": 16272 + }, + { + "epoch": 6.01, + "learning_rate": 2.4005542554501787e-06, + "loss": 0.3317, + "step": 16273 + }, + { + "epoch": 6.01, + "learning_rate": 2.4005173050868333e-06, + "loss": 0.3509, + "step": 16274 + }, + { + "epoch": 6.01, + "learning_rate": 2.4004803547234883e-06, + "loss": 0.5, + "step": 16275 + }, + { + "epoch": 6.01, + "learning_rate": 2.400443404360143e-06, + "loss": 0.4475, + "step": 16276 + }, + { + "epoch": 6.01, + "learning_rate": 2.4004064539967976e-06, + "loss": 0.4351, + "step": 16277 + }, + { + "epoch": 6.01, + "learning_rate": 2.4003695036334523e-06, + "loss": 0.5654, + "step": 16278 + }, + { + "epoch": 6.01, + "learning_rate": 2.4003325532701073e-06, + "loss": 0.4479, + "step": 16279 + }, + { + "epoch": 6.01, + "learning_rate": 2.400295602906762e-06, + "loss": 0.4804, + "step": 16280 + }, + { + "epoch": 6.01, + "learning_rate": 2.4002586525434166e-06, + "loss": 0.389, + "step": 16281 + }, + { + "epoch": 6.01, + "learning_rate": 2.4002217021800716e-06, + "loss": 0.4025, + "step": 16282 + }, + { + "epoch": 6.01, + "learning_rate": 2.4001847518167263e-06, + "loss": 0.53, + "step": 16283 + }, + { + "epoch": 6.01, + "learning_rate": 2.400147801453381e-06, + "loss": 0.4803, + "step": 16284 + }, + { + "epoch": 6.01, + "learning_rate": 2.4001108510900356e-06, + "loss": 0.3944, + "step": 16285 + }, + { + "epoch": 6.01, + "learning_rate": 2.4000739007266906e-06, + "loss": 0.4801, + "step": 16286 + }, + { + "epoch": 6.01, + "learning_rate": 2.4000369503633453e-06, + "loss": 0.5156, + "step": 16287 + }, + { + "epoch": 6.01, + "learning_rate": 2.4000000000000003e-06, + "loss": 0.4405, + "step": 16288 + }, + { + "epoch": 6.02, + "learning_rate": 2.399963049636655e-06, + "loss": 0.4513, + "step": 16289 + }, + { + "epoch": 6.02, + "learning_rate": 2.3999260992733096e-06, + "loss": 0.397, + "step": 16290 + }, + { + "epoch": 6.02, + "learning_rate": 2.3998891489099642e-06, + "loss": 0.4067, + "step": 16291 + }, + { + "epoch": 6.02, + "learning_rate": 2.399852198546619e-06, + "loss": 0.4455, + "step": 16292 + }, + { + "epoch": 6.02, + "learning_rate": 2.399815248183274e-06, + "loss": 0.3514, + "step": 16293 + }, + { + "epoch": 6.02, + "learning_rate": 2.3997782978199286e-06, + "loss": 0.464, + "step": 16294 + }, + { + "epoch": 6.02, + "learning_rate": 2.3997413474565836e-06, + "loss": 0.5211, + "step": 16295 + }, + { + "epoch": 6.02, + "learning_rate": 2.3997043970932383e-06, + "loss": 0.4605, + "step": 16296 + }, + { + "epoch": 6.02, + "learning_rate": 2.399667446729893e-06, + "loss": 0.4893, + "step": 16297 + }, + { + "epoch": 6.02, + "learning_rate": 2.3996304963665475e-06, + "loss": 0.5394, + "step": 16298 + }, + { + "epoch": 6.02, + "learning_rate": 2.3995935460032026e-06, + "loss": 0.5009, + "step": 16299 + }, + { + "epoch": 6.02, + "learning_rate": 2.3995565956398572e-06, + "loss": 0.4752, + "step": 16300 + }, + { + "epoch": 6.02, + "learning_rate": 2.399519645276512e-06, + "loss": 0.4082, + "step": 16301 + }, + { + "epoch": 6.02, + "learning_rate": 2.399482694913167e-06, + "loss": 0.4537, + "step": 16302 + }, + { + "epoch": 6.02, + "learning_rate": 2.3994457445498216e-06, + "loss": 0.3805, + "step": 16303 + }, + { + "epoch": 6.02, + "learning_rate": 2.399408794186476e-06, + "loss": 0.5799, + "step": 16304 + }, + { + "epoch": 6.02, + "learning_rate": 2.399371843823131e-06, + "loss": 0.3934, + "step": 16305 + }, + { + "epoch": 6.02, + "learning_rate": 2.399334893459786e-06, + "loss": 0.5177, + "step": 16306 + }, + { + "epoch": 6.02, + "learning_rate": 2.3992979430964405e-06, + "loss": 0.5216, + "step": 16307 + }, + { + "epoch": 6.02, + "learning_rate": 2.3992609927330956e-06, + "loss": 0.4824, + "step": 16308 + }, + { + "epoch": 6.02, + "learning_rate": 2.3992240423697502e-06, + "loss": 0.3628, + "step": 16309 + }, + { + "epoch": 6.02, + "learning_rate": 2.3991870920064044e-06, + "loss": 0.4845, + "step": 16310 + }, + { + "epoch": 6.02, + "learning_rate": 2.3991501416430595e-06, + "loss": 0.5006, + "step": 16311 + }, + { + "epoch": 6.02, + "learning_rate": 2.399113191279714e-06, + "loss": 0.4924, + "step": 16312 + }, + { + "epoch": 6.02, + "learning_rate": 2.399076240916369e-06, + "loss": 0.4145, + "step": 16313 + }, + { + "epoch": 6.02, + "learning_rate": 2.399039290553024e-06, + "loss": 0.4091, + "step": 16314 + }, + { + "epoch": 6.02, + "learning_rate": 2.399002340189679e-06, + "loss": 0.4245, + "step": 16315 + }, + { + "epoch": 6.03, + "learning_rate": 2.398965389826333e-06, + "loss": 0.4752, + "step": 16316 + }, + { + "epoch": 6.03, + "learning_rate": 2.398928439462988e-06, + "loss": 0.4523, + "step": 16317 + }, + { + "epoch": 6.03, + "learning_rate": 2.398891489099643e-06, + "loss": 0.4259, + "step": 16318 + }, + { + "epoch": 6.03, + "learning_rate": 2.398854538736298e-06, + "loss": 0.4603, + "step": 16319 + }, + { + "epoch": 6.03, + "learning_rate": 2.3988175883729525e-06, + "loss": 0.4579, + "step": 16320 + }, + { + "epoch": 6.03, + "learning_rate": 2.398780638009607e-06, + "loss": 0.4894, + "step": 16321 + }, + { + "epoch": 6.03, + "learning_rate": 2.3987436876462618e-06, + "loss": 0.3341, + "step": 16322 + }, + { + "epoch": 6.03, + "learning_rate": 2.3987067372829164e-06, + "loss": 0.3803, + "step": 16323 + }, + { + "epoch": 6.03, + "learning_rate": 2.3986697869195715e-06, + "loss": 0.4747, + "step": 16324 + }, + { + "epoch": 6.03, + "learning_rate": 2.398632836556226e-06, + "loss": 0.3969, + "step": 16325 + }, + { + "epoch": 6.03, + "learning_rate": 2.398595886192881e-06, + "loss": 0.4357, + "step": 16326 + }, + { + "epoch": 6.03, + "learning_rate": 2.398558935829536e-06, + "loss": 0.4919, + "step": 16327 + }, + { + "epoch": 6.03, + "learning_rate": 2.3985219854661904e-06, + "loss": 0.4175, + "step": 16328 + }, + { + "epoch": 6.03, + "learning_rate": 2.398485035102845e-06, + "loss": 0.4154, + "step": 16329 + }, + { + "epoch": 6.03, + "learning_rate": 2.3984480847394997e-06, + "loss": 0.4948, + "step": 16330 + }, + { + "epoch": 6.03, + "learning_rate": 2.3984111343761548e-06, + "loss": 0.4945, + "step": 16331 + }, + { + "epoch": 6.03, + "learning_rate": 2.3983741840128094e-06, + "loss": 0.4268, + "step": 16332 + }, + { + "epoch": 6.03, + "learning_rate": 2.3983372336494645e-06, + "loss": 0.4014, + "step": 16333 + }, + { + "epoch": 6.03, + "learning_rate": 2.398300283286119e-06, + "loss": 0.3147, + "step": 16334 + }, + { + "epoch": 6.03, + "learning_rate": 2.3982633329227737e-06, + "loss": 0.4768, + "step": 16335 + }, + { + "epoch": 6.03, + "learning_rate": 2.3982263825594284e-06, + "loss": 0.5239, + "step": 16336 + }, + { + "epoch": 6.03, + "learning_rate": 2.3981894321960834e-06, + "loss": 0.4846, + "step": 16337 + }, + { + "epoch": 6.03, + "learning_rate": 2.398152481832738e-06, + "loss": 0.4473, + "step": 16338 + }, + { + "epoch": 6.03, + "learning_rate": 2.398115531469393e-06, + "loss": 0.5405, + "step": 16339 + }, + { + "epoch": 6.03, + "learning_rate": 2.3980785811060478e-06, + "loss": 0.4045, + "step": 16340 + }, + { + "epoch": 6.03, + "learning_rate": 2.3980416307427024e-06, + "loss": 0.4713, + "step": 16341 + }, + { + "epoch": 6.03, + "learning_rate": 2.398004680379357e-06, + "loss": 0.5499, + "step": 16342 + }, + { + "epoch": 6.04, + "learning_rate": 2.3979677300160117e-06, + "loss": 0.3774, + "step": 16343 + }, + { + "epoch": 6.04, + "learning_rate": 2.3979307796526667e-06, + "loss": 0.4323, + "step": 16344 + }, + { + "epoch": 6.04, + "learning_rate": 2.3978938292893214e-06, + "loss": 0.4422, + "step": 16345 + }, + { + "epoch": 6.04, + "learning_rate": 2.3978568789259764e-06, + "loss": 0.5211, + "step": 16346 + }, + { + "epoch": 6.04, + "learning_rate": 2.397819928562631e-06, + "loss": 0.4543, + "step": 16347 + }, + { + "epoch": 6.04, + "learning_rate": 2.3977829781992857e-06, + "loss": 0.4623, + "step": 16348 + }, + { + "epoch": 6.04, + "learning_rate": 2.3977460278359403e-06, + "loss": 0.3879, + "step": 16349 + }, + { + "epoch": 6.04, + "learning_rate": 2.397709077472595e-06, + "loss": 0.5035, + "step": 16350 + }, + { + "epoch": 6.04, + "learning_rate": 2.39767212710925e-06, + "loss": 0.5505, + "step": 16351 + }, + { + "epoch": 6.04, + "learning_rate": 2.3976351767459047e-06, + "loss": 0.5392, + "step": 16352 + }, + { + "epoch": 6.04, + "learning_rate": 2.3975982263825597e-06, + "loss": 0.5, + "step": 16353 + }, + { + "epoch": 6.04, + "learning_rate": 2.397561276019214e-06, + "loss": 0.4839, + "step": 16354 + }, + { + "epoch": 6.04, + "learning_rate": 2.397524325655869e-06, + "loss": 0.6056, + "step": 16355 + }, + { + "epoch": 6.04, + "learning_rate": 2.3974873752925236e-06, + "loss": 0.3738, + "step": 16356 + }, + { + "epoch": 6.04, + "learning_rate": 2.3974504249291787e-06, + "loss": 0.3572, + "step": 16357 + }, + { + "epoch": 6.04, + "learning_rate": 2.3974134745658333e-06, + "loss": 0.4918, + "step": 16358 + }, + { + "epoch": 6.04, + "learning_rate": 2.3973765242024884e-06, + "loss": 0.5035, + "step": 16359 + }, + { + "epoch": 6.04, + "learning_rate": 2.3973395738391426e-06, + "loss": 0.4971, + "step": 16360 + }, + { + "epoch": 6.04, + "learning_rate": 2.3973026234757972e-06, + "loss": 0.5136, + "step": 16361 + }, + { + "epoch": 6.04, + "learning_rate": 2.3972656731124523e-06, + "loss": 0.5178, + "step": 16362 + }, + { + "epoch": 6.04, + "learning_rate": 2.397228722749107e-06, + "loss": 0.6285, + "step": 16363 + }, + { + "epoch": 6.04, + "learning_rate": 2.397191772385762e-06, + "loss": 0.4647, + "step": 16364 + }, + { + "epoch": 6.04, + "learning_rate": 2.3971548220224166e-06, + "loss": 0.3991, + "step": 16365 + }, + { + "epoch": 6.04, + "learning_rate": 2.3971178716590713e-06, + "loss": 0.5768, + "step": 16366 + }, + { + "epoch": 6.04, + "learning_rate": 2.397080921295726e-06, + "loss": 0.363, + "step": 16367 + }, + { + "epoch": 6.04, + "learning_rate": 2.397043970932381e-06, + "loss": 0.5629, + "step": 16368 + }, + { + "epoch": 6.04, + "learning_rate": 2.3970070205690356e-06, + "loss": 0.4208, + "step": 16369 + }, + { + "epoch": 6.05, + "learning_rate": 2.3969700702056902e-06, + "loss": 0.4469, + "step": 16370 + }, + { + "epoch": 6.05, + "learning_rate": 2.3969331198423453e-06, + "loss": 0.3085, + "step": 16371 + }, + { + "epoch": 6.05, + "learning_rate": 2.396896169479e-06, + "loss": 0.4291, + "step": 16372 + }, + { + "epoch": 6.05, + "learning_rate": 2.3968592191156546e-06, + "loss": 0.5093, + "step": 16373 + }, + { + "epoch": 6.05, + "learning_rate": 2.3968222687523092e-06, + "loss": 0.5482, + "step": 16374 + }, + { + "epoch": 6.05, + "learning_rate": 2.3967853183889643e-06, + "loss": 0.4804, + "step": 16375 + }, + { + "epoch": 6.05, + "learning_rate": 2.396748368025619e-06, + "loss": 0.3826, + "step": 16376 + }, + { + "epoch": 6.05, + "learning_rate": 2.396711417662274e-06, + "loss": 0.398, + "step": 16377 + }, + { + "epoch": 6.05, + "learning_rate": 2.3966744672989286e-06, + "loss": 0.4157, + "step": 16378 + }, + { + "epoch": 6.05, + "learning_rate": 2.3966375169355832e-06, + "loss": 0.4479, + "step": 16379 + }, + { + "epoch": 6.05, + "learning_rate": 2.396600566572238e-06, + "loss": 0.4416, + "step": 16380 + }, + { + "epoch": 6.05, + "learning_rate": 2.3965636162088925e-06, + "loss": 0.4465, + "step": 16381 + }, + { + "epoch": 6.05, + "learning_rate": 2.3965266658455476e-06, + "loss": 0.4782, + "step": 16382 + }, + { + "epoch": 6.05, + "learning_rate": 2.396489715482202e-06, + "loss": 0.437, + "step": 16383 + }, + { + "epoch": 6.05, + "learning_rate": 2.3964527651188573e-06, + "loss": 0.4351, + "step": 16384 + }, + { + "epoch": 6.05, + "learning_rate": 2.396415814755512e-06, + "loss": 0.4575, + "step": 16385 + }, + { + "epoch": 6.05, + "learning_rate": 2.3963788643921665e-06, + "loss": 0.4326, + "step": 16386 + }, + { + "epoch": 6.05, + "learning_rate": 2.396341914028821e-06, + "loss": 0.4066, + "step": 16387 + }, + { + "epoch": 6.05, + "learning_rate": 2.3963049636654762e-06, + "loss": 0.41, + "step": 16388 + }, + { + "epoch": 6.05, + "learning_rate": 2.396268013302131e-06, + "loss": 0.5583, + "step": 16389 + }, + { + "epoch": 6.05, + "learning_rate": 2.3962310629387855e-06, + "loss": 0.4816, + "step": 16390 + }, + { + "epoch": 6.05, + "learning_rate": 2.3961941125754406e-06, + "loss": 0.423, + "step": 16391 + }, + { + "epoch": 6.05, + "learning_rate": 2.396157162212095e-06, + "loss": 0.3565, + "step": 16392 + }, + { + "epoch": 6.05, + "learning_rate": 2.39612021184875e-06, + "loss": 0.4143, + "step": 16393 + }, + { + "epoch": 6.05, + "learning_rate": 2.3960832614854045e-06, + "loss": 0.5509, + "step": 16394 + }, + { + "epoch": 6.05, + "learning_rate": 2.3960463111220595e-06, + "loss": 0.4886, + "step": 16395 + }, + { + "epoch": 6.05, + "learning_rate": 2.396009360758714e-06, + "loss": 0.4383, + "step": 16396 + }, + { + "epoch": 6.06, + "learning_rate": 2.3959724103953692e-06, + "loss": 0.4539, + "step": 16397 + }, + { + "epoch": 6.06, + "learning_rate": 2.395935460032024e-06, + "loss": 0.4127, + "step": 16398 + }, + { + "epoch": 6.06, + "learning_rate": 2.3958985096686785e-06, + "loss": 0.4948, + "step": 16399 + }, + { + "epoch": 6.06, + "learning_rate": 2.395861559305333e-06, + "loss": 0.4261, + "step": 16400 + }, + { + "epoch": 6.06, + "learning_rate": 2.3958246089419878e-06, + "loss": 0.4743, + "step": 16401 + }, + { + "epoch": 6.06, + "learning_rate": 2.395787658578643e-06, + "loss": 0.4768, + "step": 16402 + }, + { + "epoch": 6.06, + "learning_rate": 2.3957507082152975e-06, + "loss": 0.3745, + "step": 16403 + }, + { + "epoch": 6.06, + "learning_rate": 2.3957137578519525e-06, + "loss": 0.4603, + "step": 16404 + }, + { + "epoch": 6.06, + "learning_rate": 2.3956768074886068e-06, + "loss": 0.4458, + "step": 16405 + }, + { + "epoch": 6.06, + "learning_rate": 2.395639857125262e-06, + "loss": 0.3639, + "step": 16406 + }, + { + "epoch": 6.06, + "learning_rate": 2.3956029067619165e-06, + "loss": 0.3532, + "step": 16407 + }, + { + "epoch": 6.06, + "learning_rate": 2.3955659563985715e-06, + "loss": 0.371, + "step": 16408 + }, + { + "epoch": 6.06, + "learning_rate": 2.395529006035226e-06, + "loss": 0.5083, + "step": 16409 + }, + { + "epoch": 6.06, + "learning_rate": 2.3954920556718808e-06, + "loss": 0.4601, + "step": 16410 + }, + { + "epoch": 6.06, + "learning_rate": 2.3954551053085354e-06, + "loss": 0.3763, + "step": 16411 + }, + { + "epoch": 6.06, + "learning_rate": 2.39541815494519e-06, + "loss": 0.453, + "step": 16412 + }, + { + "epoch": 6.06, + "learning_rate": 2.395381204581845e-06, + "loss": 0.4413, + "step": 16413 + }, + { + "epoch": 6.06, + "learning_rate": 2.3953442542184998e-06, + "loss": 0.481, + "step": 16414 + }, + { + "epoch": 6.06, + "learning_rate": 2.395307303855155e-06, + "loss": 0.4633, + "step": 16415 + }, + { + "epoch": 6.06, + "learning_rate": 2.3952703534918094e-06, + "loss": 0.513, + "step": 16416 + }, + { + "epoch": 6.06, + "learning_rate": 2.395233403128464e-06, + "loss": 0.3394, + "step": 16417 + }, + { + "epoch": 6.06, + "learning_rate": 2.3951964527651187e-06, + "loss": 0.3905, + "step": 16418 + }, + { + "epoch": 6.06, + "learning_rate": 2.3951595024017738e-06, + "loss": 0.5184, + "step": 16419 + }, + { + "epoch": 6.06, + "learning_rate": 2.3951225520384284e-06, + "loss": 0.4098, + "step": 16420 + }, + { + "epoch": 6.06, + "learning_rate": 2.395085601675083e-06, + "loss": 0.4517, + "step": 16421 + }, + { + "epoch": 6.06, + "learning_rate": 2.395048651311738e-06, + "loss": 0.4645, + "step": 16422 + }, + { + "epoch": 6.06, + "learning_rate": 2.3950117009483928e-06, + "loss": 0.4617, + "step": 16423 + }, + { + "epoch": 6.06, + "learning_rate": 2.3949747505850474e-06, + "loss": 0.461, + "step": 16424 + }, + { + "epoch": 6.07, + "learning_rate": 2.394937800221702e-06, + "loss": 0.4115, + "step": 16425 + }, + { + "epoch": 6.07, + "learning_rate": 2.394900849858357e-06, + "loss": 0.4179, + "step": 16426 + }, + { + "epoch": 6.07, + "learning_rate": 2.3948638994950117e-06, + "loss": 0.467, + "step": 16427 + }, + { + "epoch": 6.07, + "learning_rate": 2.3948269491316668e-06, + "loss": 0.5071, + "step": 16428 + }, + { + "epoch": 6.07, + "learning_rate": 2.3947899987683214e-06, + "loss": 0.4626, + "step": 16429 + }, + { + "epoch": 6.07, + "learning_rate": 2.394753048404976e-06, + "loss": 0.5382, + "step": 16430 + }, + { + "epoch": 6.07, + "learning_rate": 2.3947160980416307e-06, + "loss": 0.4586, + "step": 16431 + }, + { + "epoch": 6.07, + "learning_rate": 2.3946791476782853e-06, + "loss": 0.4074, + "step": 16432 + }, + { + "epoch": 6.07, + "learning_rate": 2.3946421973149404e-06, + "loss": 0.4361, + "step": 16433 + }, + { + "epoch": 6.07, + "learning_rate": 2.394605246951595e-06, + "loss": 0.5613, + "step": 16434 + }, + { + "epoch": 6.07, + "learning_rate": 2.39456829658825e-06, + "loss": 0.4996, + "step": 16435 + }, + { + "epoch": 6.07, + "learning_rate": 2.3945313462249047e-06, + "loss": 0.4457, + "step": 16436 + }, + { + "epoch": 6.07, + "learning_rate": 2.3944943958615594e-06, + "loss": 0.546, + "step": 16437 + }, + { + "epoch": 6.07, + "learning_rate": 2.394457445498214e-06, + "loss": 0.6002, + "step": 16438 + }, + { + "epoch": 6.07, + "learning_rate": 2.394420495134869e-06, + "loss": 0.543, + "step": 16439 + }, + { + "epoch": 6.07, + "learning_rate": 2.3943835447715237e-06, + "loss": 0.4078, + "step": 16440 + }, + { + "epoch": 6.07, + "learning_rate": 2.3943465944081783e-06, + "loss": 0.35, + "step": 16441 + }, + { + "epoch": 6.07, + "learning_rate": 2.3943096440448334e-06, + "loss": 0.4092, + "step": 16442 + }, + { + "epoch": 6.07, + "learning_rate": 2.394272693681488e-06, + "loss": 0.3527, + "step": 16443 + }, + { + "epoch": 6.07, + "learning_rate": 2.3942357433181427e-06, + "loss": 0.3624, + "step": 16444 + }, + { + "epoch": 6.07, + "learning_rate": 2.3941987929547973e-06, + "loss": 0.4009, + "step": 16445 + }, + { + "epoch": 6.07, + "learning_rate": 2.3941618425914524e-06, + "loss": 0.5295, + "step": 16446 + }, + { + "epoch": 6.07, + "learning_rate": 2.394124892228107e-06, + "loss": 0.4955, + "step": 16447 + }, + { + "epoch": 6.07, + "learning_rate": 2.394087941864762e-06, + "loss": 0.4382, + "step": 16448 + }, + { + "epoch": 6.07, + "learning_rate": 2.3940509915014167e-06, + "loss": 0.4951, + "step": 16449 + }, + { + "epoch": 6.07, + "learning_rate": 2.394014041138071e-06, + "loss": 0.3589, + "step": 16450 + }, + { + "epoch": 6.07, + "learning_rate": 2.393977090774726e-06, + "loss": 0.3174, + "step": 16451 + }, + { + "epoch": 6.08, + "learning_rate": 2.3939401404113806e-06, + "loss": 0.4094, + "step": 16452 + }, + { + "epoch": 6.08, + "learning_rate": 2.3939031900480357e-06, + "loss": 0.4775, + "step": 16453 + }, + { + "epoch": 6.08, + "learning_rate": 2.3938662396846903e-06, + "loss": 0.5277, + "step": 16454 + }, + { + "epoch": 6.08, + "learning_rate": 2.3938292893213454e-06, + "loss": 0.5213, + "step": 16455 + }, + { + "epoch": 6.08, + "learning_rate": 2.3937923389579996e-06, + "loss": 0.4277, + "step": 16456 + }, + { + "epoch": 6.08, + "learning_rate": 2.3937553885946546e-06, + "loss": 0.3416, + "step": 16457 + }, + { + "epoch": 6.08, + "learning_rate": 2.3937184382313093e-06, + "loss": 0.512, + "step": 16458 + }, + { + "epoch": 6.08, + "learning_rate": 2.3936814878679643e-06, + "loss": 0.5668, + "step": 16459 + }, + { + "epoch": 6.08, + "learning_rate": 2.393644537504619e-06, + "loss": 0.3806, + "step": 16460 + }, + { + "epoch": 6.08, + "learning_rate": 2.3936075871412736e-06, + "loss": 0.6065, + "step": 16461 + }, + { + "epoch": 6.08, + "learning_rate": 2.3935706367779282e-06, + "loss": 0.5055, + "step": 16462 + }, + { + "epoch": 6.08, + "learning_rate": 2.393533686414583e-06, + "loss": 0.4366, + "step": 16463 + }, + { + "epoch": 6.08, + "learning_rate": 2.393496736051238e-06, + "loss": 0.4949, + "step": 16464 + }, + { + "epoch": 6.08, + "learning_rate": 2.3934597856878926e-06, + "loss": 0.6041, + "step": 16465 + }, + { + "epoch": 6.08, + "learning_rate": 2.3934228353245476e-06, + "loss": 0.5104, + "step": 16466 + }, + { + "epoch": 6.08, + "learning_rate": 2.3933858849612023e-06, + "loss": 0.3355, + "step": 16467 + }, + { + "epoch": 6.08, + "learning_rate": 2.393348934597857e-06, + "loss": 0.552, + "step": 16468 + }, + { + "epoch": 6.08, + "learning_rate": 2.3933119842345115e-06, + "loss": 0.4307, + "step": 16469 + }, + { + "epoch": 6.08, + "learning_rate": 2.393275033871166e-06, + "loss": 0.4874, + "step": 16470 + }, + { + "epoch": 6.08, + "learning_rate": 2.3932380835078212e-06, + "loss": 0.3952, + "step": 16471 + }, + { + "epoch": 6.08, + "learning_rate": 2.393201133144476e-06, + "loss": 0.3815, + "step": 16472 + }, + { + "epoch": 6.08, + "learning_rate": 2.393164182781131e-06, + "loss": 0.4732, + "step": 16473 + }, + { + "epoch": 6.08, + "learning_rate": 2.3931272324177856e-06, + "loss": 0.6878, + "step": 16474 + }, + { + "epoch": 6.08, + "learning_rate": 2.39309028205444e-06, + "loss": 0.4732, + "step": 16475 + }, + { + "epoch": 6.08, + "learning_rate": 2.393053331691095e-06, + "loss": 0.4258, + "step": 16476 + }, + { + "epoch": 6.08, + "learning_rate": 2.39301638132775e-06, + "loss": 0.4494, + "step": 16477 + }, + { + "epoch": 6.08, + "learning_rate": 2.3929794309644045e-06, + "loss": 0.3911, + "step": 16478 + }, + { + "epoch": 6.09, + "learning_rate": 2.3929424806010596e-06, + "loss": 0.4494, + "step": 16479 + }, + { + "epoch": 6.09, + "learning_rate": 2.3929055302377142e-06, + "loss": 0.4746, + "step": 16480 + }, + { + "epoch": 6.09, + "learning_rate": 2.392868579874369e-06, + "loss": 0.5405, + "step": 16481 + }, + { + "epoch": 6.09, + "learning_rate": 2.3928316295110235e-06, + "loss": 0.3358, + "step": 16482 + }, + { + "epoch": 6.09, + "learning_rate": 2.392794679147678e-06, + "loss": 0.4834, + "step": 16483 + }, + { + "epoch": 6.09, + "learning_rate": 2.392757728784333e-06, + "loss": 0.4955, + "step": 16484 + }, + { + "epoch": 6.09, + "learning_rate": 2.392720778420988e-06, + "loss": 0.448, + "step": 16485 + }, + { + "epoch": 6.09, + "learning_rate": 2.392683828057643e-06, + "loss": 0.4126, + "step": 16486 + }, + { + "epoch": 6.09, + "learning_rate": 2.3926468776942975e-06, + "loss": 0.631, + "step": 16487 + }, + { + "epoch": 6.09, + "learning_rate": 2.392609927330952e-06, + "loss": 0.4217, + "step": 16488 + }, + { + "epoch": 6.09, + "learning_rate": 2.392572976967607e-06, + "loss": 0.4673, + "step": 16489 + }, + { + "epoch": 6.09, + "learning_rate": 2.3925360266042614e-06, + "loss": 0.4888, + "step": 16490 + }, + { + "epoch": 6.09, + "learning_rate": 2.3924990762409165e-06, + "loss": 0.329, + "step": 16491 + }, + { + "epoch": 6.09, + "learning_rate": 2.392462125877571e-06, + "loss": 0.5351, + "step": 16492 + }, + { + "epoch": 6.09, + "learning_rate": 2.392425175514226e-06, + "loss": 0.3796, + "step": 16493 + }, + { + "epoch": 6.09, + "learning_rate": 2.392388225150881e-06, + "loss": 0.5049, + "step": 16494 + }, + { + "epoch": 6.09, + "learning_rate": 2.3923512747875355e-06, + "loss": 0.4428, + "step": 16495 + }, + { + "epoch": 6.09, + "learning_rate": 2.39231432442419e-06, + "loss": 0.5713, + "step": 16496 + }, + { + "epoch": 6.09, + "learning_rate": 2.392277374060845e-06, + "loss": 0.3563, + "step": 16497 + }, + { + "epoch": 6.09, + "learning_rate": 2.3922404236975e-06, + "loss": 0.4185, + "step": 16498 + }, + { + "epoch": 6.09, + "learning_rate": 2.392203473334155e-06, + "loss": 0.4236, + "step": 16499 + }, + { + "epoch": 6.09, + "learning_rate": 2.3921665229708095e-06, + "loss": 0.5308, + "step": 16500 + }, + { + "epoch": 6.09, + "learning_rate": 2.3921295726074637e-06, + "loss": 0.4667, + "step": 16501 + }, + { + "epoch": 6.09, + "learning_rate": 2.3920926222441188e-06, + "loss": 0.5065, + "step": 16502 + }, + { + "epoch": 6.09, + "learning_rate": 2.3920556718807734e-06, + "loss": 0.4621, + "step": 16503 + }, + { + "epoch": 6.09, + "learning_rate": 2.3920187215174285e-06, + "loss": 0.4994, + "step": 16504 + }, + { + "epoch": 6.09, + "learning_rate": 2.391981771154083e-06, + "loss": 0.4601, + "step": 16505 + }, + { + "epoch": 6.1, + "learning_rate": 2.391944820790738e-06, + "loss": 0.513, + "step": 16506 + }, + { + "epoch": 6.1, + "learning_rate": 2.3919078704273924e-06, + "loss": 0.449, + "step": 16507 + }, + { + "epoch": 6.1, + "learning_rate": 2.3918709200640474e-06, + "loss": 0.4552, + "step": 16508 + }, + { + "epoch": 6.1, + "learning_rate": 2.391833969700702e-06, + "loss": 0.5568, + "step": 16509 + }, + { + "epoch": 6.1, + "learning_rate": 2.3917970193373567e-06, + "loss": 0.4697, + "step": 16510 + }, + { + "epoch": 6.1, + "learning_rate": 2.3917600689740118e-06, + "loss": 0.4861, + "step": 16511 + }, + { + "epoch": 6.1, + "learning_rate": 2.3917231186106664e-06, + "loss": 0.3651, + "step": 16512 + }, + { + "epoch": 6.1, + "learning_rate": 2.391686168247321e-06, + "loss": 0.4168, + "step": 16513 + }, + { + "epoch": 6.1, + "learning_rate": 2.3916492178839757e-06, + "loss": 0.5203, + "step": 16514 + }, + { + "epoch": 6.1, + "learning_rate": 2.3916122675206307e-06, + "loss": 0.437, + "step": 16515 + }, + { + "epoch": 6.1, + "learning_rate": 2.3915753171572854e-06, + "loss": 0.443, + "step": 16516 + }, + { + "epoch": 6.1, + "learning_rate": 2.3915383667939404e-06, + "loss": 0.3696, + "step": 16517 + }, + { + "epoch": 6.1, + "learning_rate": 2.391501416430595e-06, + "loss": 0.5168, + "step": 16518 + }, + { + "epoch": 6.1, + "learning_rate": 2.3914644660672497e-06, + "loss": 0.3891, + "step": 16519 + }, + { + "epoch": 6.1, + "learning_rate": 2.3914275157039043e-06, + "loss": 0.3844, + "step": 16520 + }, + { + "epoch": 6.1, + "learning_rate": 2.391390565340559e-06, + "loss": 0.4475, + "step": 16521 + }, + { + "epoch": 6.1, + "learning_rate": 2.391353614977214e-06, + "loss": 0.4513, + "step": 16522 + }, + { + "epoch": 6.1, + "learning_rate": 2.3913166646138687e-06, + "loss": 0.3678, + "step": 16523 + }, + { + "epoch": 6.1, + "learning_rate": 2.3912797142505237e-06, + "loss": 0.4834, + "step": 16524 + }, + { + "epoch": 6.1, + "learning_rate": 2.3912427638871784e-06, + "loss": 0.5448, + "step": 16525 + }, + { + "epoch": 6.1, + "learning_rate": 2.391205813523833e-06, + "loss": 0.4457, + "step": 16526 + }, + { + "epoch": 6.1, + "learning_rate": 2.3911688631604876e-06, + "loss": 0.4299, + "step": 16527 + }, + { + "epoch": 6.1, + "learning_rate": 2.3911319127971427e-06, + "loss": 0.4346, + "step": 16528 + }, + { + "epoch": 6.1, + "learning_rate": 2.3910949624337973e-06, + "loss": 0.498, + "step": 16529 + }, + { + "epoch": 6.1, + "learning_rate": 2.391058012070452e-06, + "loss": 0.432, + "step": 16530 + }, + { + "epoch": 6.1, + "learning_rate": 2.391021061707107e-06, + "loss": 0.4796, + "step": 16531 + }, + { + "epoch": 6.1, + "learning_rate": 2.3909841113437617e-06, + "loss": 0.5074, + "step": 16532 + }, + { + "epoch": 6.11, + "learning_rate": 2.3909471609804163e-06, + "loss": 0.5418, + "step": 16533 + }, + { + "epoch": 6.11, + "learning_rate": 2.390910210617071e-06, + "loss": 0.3766, + "step": 16534 + }, + { + "epoch": 6.11, + "learning_rate": 2.390873260253726e-06, + "loss": 0.4275, + "step": 16535 + }, + { + "epoch": 6.11, + "learning_rate": 2.3908363098903806e-06, + "loss": 0.4919, + "step": 16536 + }, + { + "epoch": 6.11, + "learning_rate": 2.3907993595270357e-06, + "loss": 0.6401, + "step": 16537 + }, + { + "epoch": 6.11, + "learning_rate": 2.3907624091636903e-06, + "loss": 0.3723, + "step": 16538 + }, + { + "epoch": 6.11, + "learning_rate": 2.390725458800345e-06, + "loss": 0.2925, + "step": 16539 + }, + { + "epoch": 6.11, + "learning_rate": 2.3906885084369996e-06, + "loss": 0.4761, + "step": 16540 + }, + { + "epoch": 6.11, + "learning_rate": 2.3906515580736542e-06, + "loss": 0.5277, + "step": 16541 + }, + { + "epoch": 6.11, + "learning_rate": 2.3906146077103093e-06, + "loss": 0.5682, + "step": 16542 + }, + { + "epoch": 6.11, + "learning_rate": 2.390577657346964e-06, + "loss": 0.4748, + "step": 16543 + }, + { + "epoch": 6.11, + "learning_rate": 2.390540706983619e-06, + "loss": 0.4483, + "step": 16544 + }, + { + "epoch": 6.11, + "learning_rate": 2.3905037566202732e-06, + "loss": 0.5582, + "step": 16545 + }, + { + "epoch": 6.11, + "learning_rate": 2.3904668062569283e-06, + "loss": 0.4502, + "step": 16546 + }, + { + "epoch": 6.11, + "learning_rate": 2.390429855893583e-06, + "loss": 0.4206, + "step": 16547 + }, + { + "epoch": 6.11, + "learning_rate": 2.390392905530238e-06, + "loss": 0.4566, + "step": 16548 + }, + { + "epoch": 6.11, + "learning_rate": 2.3903559551668926e-06, + "loss": 0.4489, + "step": 16549 + }, + { + "epoch": 6.11, + "learning_rate": 2.3903190048035472e-06, + "loss": 0.4502, + "step": 16550 + }, + { + "epoch": 6.11, + "learning_rate": 2.390282054440202e-06, + "loss": 0.3854, + "step": 16551 + }, + { + "epoch": 6.11, + "learning_rate": 2.3902451040768565e-06, + "loss": 0.4756, + "step": 16552 + }, + { + "epoch": 6.11, + "learning_rate": 2.3902081537135116e-06, + "loss": 0.486, + "step": 16553 + }, + { + "epoch": 6.11, + "learning_rate": 2.3901712033501662e-06, + "loss": 0.5054, + "step": 16554 + }, + { + "epoch": 6.11, + "learning_rate": 2.3901342529868213e-06, + "loss": 0.511, + "step": 16555 + }, + { + "epoch": 6.11, + "learning_rate": 2.390097302623476e-06, + "loss": 0.4725, + "step": 16556 + }, + { + "epoch": 6.11, + "learning_rate": 2.3900603522601305e-06, + "loss": 0.4935, + "step": 16557 + }, + { + "epoch": 6.11, + "learning_rate": 2.390023401896785e-06, + "loss": 0.4348, + "step": 16558 + }, + { + "epoch": 6.11, + "learning_rate": 2.38998645153344e-06, + "loss": 0.4847, + "step": 16559 + }, + { + "epoch": 6.12, + "learning_rate": 2.389949501170095e-06, + "loss": 0.5545, + "step": 16560 + }, + { + "epoch": 6.12, + "learning_rate": 2.3899125508067495e-06, + "loss": 0.375, + "step": 16561 + }, + { + "epoch": 6.12, + "learning_rate": 2.3898756004434046e-06, + "loss": 0.5644, + "step": 16562 + }, + { + "epoch": 6.12, + "learning_rate": 2.3898386500800592e-06, + "loss": 0.5221, + "step": 16563 + }, + { + "epoch": 6.12, + "learning_rate": 2.389801699716714e-06, + "loss": 0.5331, + "step": 16564 + }, + { + "epoch": 6.12, + "learning_rate": 2.3897647493533685e-06, + "loss": 0.467, + "step": 16565 + }, + { + "epoch": 6.12, + "learning_rate": 2.3897277989900235e-06, + "loss": 0.4112, + "step": 16566 + }, + { + "epoch": 6.12, + "learning_rate": 2.389690848626678e-06, + "loss": 0.3757, + "step": 16567 + }, + { + "epoch": 6.12, + "learning_rate": 2.3896538982633332e-06, + "loss": 0.4378, + "step": 16568 + }, + { + "epoch": 6.12, + "learning_rate": 2.389616947899988e-06, + "loss": 0.428, + "step": 16569 + }, + { + "epoch": 6.12, + "learning_rate": 2.3895799975366425e-06, + "loss": 0.445, + "step": 16570 + }, + { + "epoch": 6.12, + "learning_rate": 2.389543047173297e-06, + "loss": 0.4069, + "step": 16571 + }, + { + "epoch": 6.12, + "learning_rate": 2.389506096809952e-06, + "loss": 0.4963, + "step": 16572 + }, + { + "epoch": 6.12, + "learning_rate": 2.389469146446607e-06, + "loss": 0.329, + "step": 16573 + }, + { + "epoch": 6.12, + "learning_rate": 2.3894321960832615e-06, + "loss": 0.5677, + "step": 16574 + }, + { + "epoch": 6.12, + "learning_rate": 2.3893952457199165e-06, + "loss": 0.3816, + "step": 16575 + }, + { + "epoch": 6.12, + "learning_rate": 2.389358295356571e-06, + "loss": 0.4793, + "step": 16576 + }, + { + "epoch": 6.12, + "learning_rate": 2.389321344993226e-06, + "loss": 0.4821, + "step": 16577 + }, + { + "epoch": 6.12, + "learning_rate": 2.3892843946298805e-06, + "loss": 0.3898, + "step": 16578 + }, + { + "epoch": 6.12, + "learning_rate": 2.389247444266535e-06, + "loss": 0.499, + "step": 16579 + }, + { + "epoch": 6.12, + "learning_rate": 2.38921049390319e-06, + "loss": 0.4316, + "step": 16580 + }, + { + "epoch": 6.12, + "learning_rate": 2.389173543539845e-06, + "loss": 0.4015, + "step": 16581 + }, + { + "epoch": 6.12, + "learning_rate": 2.3891365931765e-06, + "loss": 0.3815, + "step": 16582 + }, + { + "epoch": 6.12, + "learning_rate": 2.3890996428131545e-06, + "loss": 0.3212, + "step": 16583 + }, + { + "epoch": 6.12, + "learning_rate": 2.389062692449809e-06, + "loss": 0.5156, + "step": 16584 + }, + { + "epoch": 6.12, + "learning_rate": 2.3890257420864638e-06, + "loss": 0.4262, + "step": 16585 + }, + { + "epoch": 6.12, + "learning_rate": 2.388988791723119e-06, + "loss": 0.4808, + "step": 16586 + }, + { + "epoch": 6.13, + "learning_rate": 2.3889518413597735e-06, + "loss": 0.3761, + "step": 16587 + }, + { + "epoch": 6.13, + "learning_rate": 2.3889148909964285e-06, + "loss": 0.423, + "step": 16588 + }, + { + "epoch": 6.13, + "learning_rate": 2.388877940633083e-06, + "loss": 0.4327, + "step": 16589 + }, + { + "epoch": 6.13, + "learning_rate": 2.3888409902697374e-06, + "loss": 0.5545, + "step": 16590 + }, + { + "epoch": 6.13, + "learning_rate": 2.3888040399063924e-06, + "loss": 0.4331, + "step": 16591 + }, + { + "epoch": 6.13, + "learning_rate": 2.388767089543047e-06, + "loss": 0.419, + "step": 16592 + }, + { + "epoch": 6.13, + "learning_rate": 2.388730139179702e-06, + "loss": 0.3721, + "step": 16593 + }, + { + "epoch": 6.13, + "learning_rate": 2.3886931888163568e-06, + "loss": 0.4859, + "step": 16594 + }, + { + "epoch": 6.13, + "learning_rate": 2.388656238453012e-06, + "loss": 0.3798, + "step": 16595 + }, + { + "epoch": 6.13, + "learning_rate": 2.388619288089666e-06, + "loss": 0.4268, + "step": 16596 + }, + { + "epoch": 6.13, + "learning_rate": 2.388582337726321e-06, + "loss": 0.5173, + "step": 16597 + }, + { + "epoch": 6.13, + "learning_rate": 2.3885453873629757e-06, + "loss": 0.5606, + "step": 16598 + }, + { + "epoch": 6.13, + "learning_rate": 2.3885084369996304e-06, + "loss": 0.3744, + "step": 16599 + }, + { + "epoch": 6.13, + "learning_rate": 2.3884714866362854e-06, + "loss": 0.451, + "step": 16600 + }, + { + "epoch": 6.13, + "learning_rate": 2.38843453627294e-06, + "loss": 0.4334, + "step": 16601 + }, + { + "epoch": 6.13, + "learning_rate": 2.3883975859095947e-06, + "loss": 0.533, + "step": 16602 + }, + { + "epoch": 6.13, + "learning_rate": 2.3883606355462493e-06, + "loss": 0.4615, + "step": 16603 + }, + { + "epoch": 6.13, + "learning_rate": 2.3883236851829044e-06, + "loss": 0.3939, + "step": 16604 + }, + { + "epoch": 6.13, + "learning_rate": 2.388286734819559e-06, + "loss": 0.415, + "step": 16605 + }, + { + "epoch": 6.13, + "learning_rate": 2.388249784456214e-06, + "loss": 0.5355, + "step": 16606 + }, + { + "epoch": 6.13, + "learning_rate": 2.3882128340928687e-06, + "loss": 0.6756, + "step": 16607 + }, + { + "epoch": 6.13, + "learning_rate": 2.3881758837295234e-06, + "loss": 0.5023, + "step": 16608 + }, + { + "epoch": 6.13, + "learning_rate": 2.388138933366178e-06, + "loss": 0.4669, + "step": 16609 + }, + { + "epoch": 6.13, + "learning_rate": 2.3881019830028326e-06, + "loss": 0.503, + "step": 16610 + }, + { + "epoch": 6.13, + "learning_rate": 2.3880650326394877e-06, + "loss": 0.5253, + "step": 16611 + }, + { + "epoch": 6.13, + "learning_rate": 2.3880280822761423e-06, + "loss": 0.5738, + "step": 16612 + }, + { + "epoch": 6.13, + "learning_rate": 2.3879911319127974e-06, + "loss": 0.378, + "step": 16613 + }, + { + "epoch": 6.14, + "learning_rate": 2.387954181549452e-06, + "loss": 0.5068, + "step": 16614 + }, + { + "epoch": 6.14, + "learning_rate": 2.3879172311861067e-06, + "loss": 0.4159, + "step": 16615 + }, + { + "epoch": 6.14, + "learning_rate": 2.3878802808227613e-06, + "loss": 0.5311, + "step": 16616 + }, + { + "epoch": 6.14, + "learning_rate": 2.3878433304594164e-06, + "loss": 0.4207, + "step": 16617 + }, + { + "epoch": 6.14, + "learning_rate": 2.387806380096071e-06, + "loss": 0.368, + "step": 16618 + }, + { + "epoch": 6.14, + "learning_rate": 2.3877694297327256e-06, + "loss": 0.5151, + "step": 16619 + }, + { + "epoch": 6.14, + "learning_rate": 2.3877324793693807e-06, + "loss": 0.3811, + "step": 16620 + }, + { + "epoch": 6.14, + "learning_rate": 2.3876955290060353e-06, + "loss": 0.4082, + "step": 16621 + }, + { + "epoch": 6.14, + "learning_rate": 2.38765857864269e-06, + "loss": 0.3481, + "step": 16622 + }, + { + "epoch": 6.14, + "learning_rate": 2.3876216282793446e-06, + "loss": 0.4109, + "step": 16623 + }, + { + "epoch": 6.14, + "learning_rate": 2.3875846779159997e-06, + "loss": 0.4397, + "step": 16624 + }, + { + "epoch": 6.14, + "learning_rate": 2.3875477275526543e-06, + "loss": 0.3619, + "step": 16625 + }, + { + "epoch": 6.14, + "learning_rate": 2.3875107771893094e-06, + "loss": 0.2827, + "step": 16626 + }, + { + "epoch": 6.14, + "learning_rate": 2.387473826825964e-06, + "loss": 0.3218, + "step": 16627 + }, + { + "epoch": 6.14, + "learning_rate": 2.3874368764626186e-06, + "loss": 0.3394, + "step": 16628 + }, + { + "epoch": 6.14, + "learning_rate": 2.3873999260992733e-06, + "loss": 0.5431, + "step": 16629 + }, + { + "epoch": 6.14, + "learning_rate": 2.387362975735928e-06, + "loss": 0.4204, + "step": 16630 + }, + { + "epoch": 6.14, + "learning_rate": 2.387326025372583e-06, + "loss": 0.3434, + "step": 16631 + }, + { + "epoch": 6.14, + "learning_rate": 2.3872890750092376e-06, + "loss": 0.5187, + "step": 16632 + }, + { + "epoch": 6.14, + "learning_rate": 2.3872521246458927e-06, + "loss": 0.4261, + "step": 16633 + }, + { + "epoch": 6.14, + "learning_rate": 2.3872151742825473e-06, + "loss": 0.336, + "step": 16634 + }, + { + "epoch": 6.14, + "learning_rate": 2.387178223919202e-06, + "loss": 0.4571, + "step": 16635 + }, + { + "epoch": 6.14, + "learning_rate": 2.3871412735558566e-06, + "loss": 0.5568, + "step": 16636 + }, + { + "epoch": 6.14, + "learning_rate": 2.3871043231925116e-06, + "loss": 0.5325, + "step": 16637 + }, + { + "epoch": 6.14, + "learning_rate": 2.3870673728291663e-06, + "loss": 0.4278, + "step": 16638 + }, + { + "epoch": 6.14, + "learning_rate": 2.387030422465821e-06, + "loss": 0.4814, + "step": 16639 + }, + { + "epoch": 6.14, + "learning_rate": 2.386993472102476e-06, + "loss": 0.5283, + "step": 16640 + }, + { + "epoch": 6.15, + "learning_rate": 2.38695652173913e-06, + "loss": 0.3477, + "step": 16641 + }, + { + "epoch": 6.15, + "learning_rate": 2.3869195713757852e-06, + "loss": 0.5141, + "step": 16642 + }, + { + "epoch": 6.15, + "learning_rate": 2.38688262101244e-06, + "loss": 0.4363, + "step": 16643 + }, + { + "epoch": 6.15, + "learning_rate": 2.386845670649095e-06, + "loss": 0.5025, + "step": 16644 + }, + { + "epoch": 6.15, + "learning_rate": 2.3868087202857496e-06, + "loss": 0.4743, + "step": 16645 + }, + { + "epoch": 6.15, + "learning_rate": 2.3867717699224046e-06, + "loss": 0.4899, + "step": 16646 + }, + { + "epoch": 6.15, + "learning_rate": 2.386734819559059e-06, + "loss": 0.3901, + "step": 16647 + }, + { + "epoch": 6.15, + "learning_rate": 2.386697869195714e-06, + "loss": 0.391, + "step": 16648 + }, + { + "epoch": 6.15, + "learning_rate": 2.3866609188323685e-06, + "loss": 0.4374, + "step": 16649 + }, + { + "epoch": 6.15, + "learning_rate": 2.386623968469023e-06, + "loss": 0.4711, + "step": 16650 + }, + { + "epoch": 6.15, + "learning_rate": 2.3865870181056782e-06, + "loss": 0.4682, + "step": 16651 + }, + { + "epoch": 6.15, + "learning_rate": 2.386550067742333e-06, + "loss": 0.5101, + "step": 16652 + }, + { + "epoch": 6.15, + "learning_rate": 2.3865131173789875e-06, + "loss": 0.4863, + "step": 16653 + }, + { + "epoch": 6.15, + "learning_rate": 2.386476167015642e-06, + "loss": 0.341, + "step": 16654 + }, + { + "epoch": 6.15, + "learning_rate": 2.386439216652297e-06, + "loss": 0.4383, + "step": 16655 + }, + { + "epoch": 6.15, + "learning_rate": 2.386402266288952e-06, + "loss": 0.5629, + "step": 16656 + }, + { + "epoch": 6.15, + "learning_rate": 2.386365315925607e-06, + "loss": 0.4261, + "step": 16657 + }, + { + "epoch": 6.15, + "learning_rate": 2.3863283655622615e-06, + "loss": 0.4528, + "step": 16658 + }, + { + "epoch": 6.15, + "learning_rate": 2.386291415198916e-06, + "loss": 0.4763, + "step": 16659 + }, + { + "epoch": 6.15, + "learning_rate": 2.386254464835571e-06, + "loss": 0.5895, + "step": 16660 + }, + { + "epoch": 6.15, + "learning_rate": 2.3862175144722254e-06, + "loss": 0.4924, + "step": 16661 + }, + { + "epoch": 6.15, + "learning_rate": 2.3861805641088805e-06, + "loss": 0.4491, + "step": 16662 + }, + { + "epoch": 6.15, + "learning_rate": 2.386143613745535e-06, + "loss": 0.5265, + "step": 16663 + }, + { + "epoch": 6.15, + "learning_rate": 2.38610666338219e-06, + "loss": 0.5146, + "step": 16664 + }, + { + "epoch": 6.15, + "learning_rate": 2.386069713018845e-06, + "loss": 0.5779, + "step": 16665 + }, + { + "epoch": 6.15, + "learning_rate": 2.3860327626554995e-06, + "loss": 0.438, + "step": 16666 + }, + { + "epoch": 6.15, + "learning_rate": 2.385995812292154e-06, + "loss": 0.3741, + "step": 16667 + }, + { + "epoch": 6.16, + "learning_rate": 2.385958861928809e-06, + "loss": 0.3553, + "step": 16668 + }, + { + "epoch": 6.16, + "learning_rate": 2.385921911565464e-06, + "loss": 0.4397, + "step": 16669 + }, + { + "epoch": 6.16, + "learning_rate": 2.3858849612021184e-06, + "loss": 0.5541, + "step": 16670 + }, + { + "epoch": 6.16, + "learning_rate": 2.3858480108387735e-06, + "loss": 0.4303, + "step": 16671 + }, + { + "epoch": 6.16, + "learning_rate": 2.385811060475428e-06, + "loss": 0.4236, + "step": 16672 + }, + { + "epoch": 6.16, + "learning_rate": 2.3857741101120828e-06, + "loss": 0.4747, + "step": 16673 + }, + { + "epoch": 6.16, + "learning_rate": 2.3857371597487374e-06, + "loss": 0.5147, + "step": 16674 + }, + { + "epoch": 6.16, + "learning_rate": 2.3857002093853925e-06, + "loss": 0.4548, + "step": 16675 + }, + { + "epoch": 6.16, + "learning_rate": 2.385663259022047e-06, + "loss": 0.5017, + "step": 16676 + }, + { + "epoch": 6.16, + "learning_rate": 2.385626308658702e-06, + "loss": 0.5006, + "step": 16677 + }, + { + "epoch": 6.16, + "learning_rate": 2.385589358295357e-06, + "loss": 0.3492, + "step": 16678 + }, + { + "epoch": 6.16, + "learning_rate": 2.3855524079320114e-06, + "loss": 0.4452, + "step": 16679 + }, + { + "epoch": 6.16, + "learning_rate": 2.385515457568666e-06, + "loss": 0.4056, + "step": 16680 + }, + { + "epoch": 6.16, + "learning_rate": 2.3854785072053207e-06, + "loss": 0.5622, + "step": 16681 + }, + { + "epoch": 6.16, + "learning_rate": 2.3854415568419758e-06, + "loss": 0.4642, + "step": 16682 + }, + { + "epoch": 6.16, + "learning_rate": 2.3854046064786304e-06, + "loss": 0.3454, + "step": 16683 + }, + { + "epoch": 6.16, + "learning_rate": 2.3853676561152855e-06, + "loss": 0.379, + "step": 16684 + }, + { + "epoch": 6.16, + "learning_rate": 2.38533070575194e-06, + "loss": 0.4731, + "step": 16685 + }, + { + "epoch": 6.16, + "learning_rate": 2.3852937553885947e-06, + "loss": 0.4372, + "step": 16686 + }, + { + "epoch": 6.16, + "learning_rate": 2.3852568050252494e-06, + "loss": 0.4185, + "step": 16687 + }, + { + "epoch": 6.16, + "learning_rate": 2.3852198546619044e-06, + "loss": 0.4973, + "step": 16688 + }, + { + "epoch": 6.16, + "learning_rate": 2.385182904298559e-06, + "loss": 0.5673, + "step": 16689 + }, + { + "epoch": 6.16, + "learning_rate": 2.3851459539352137e-06, + "loss": 0.4441, + "step": 16690 + }, + { + "epoch": 6.16, + "learning_rate": 2.3851090035718688e-06, + "loss": 0.4273, + "step": 16691 + }, + { + "epoch": 6.16, + "learning_rate": 2.385072053208523e-06, + "loss": 0.3926, + "step": 16692 + }, + { + "epoch": 6.16, + "learning_rate": 2.385035102845178e-06, + "loss": 0.3802, + "step": 16693 + }, + { + "epoch": 6.16, + "learning_rate": 2.3849981524818327e-06, + "loss": 0.5296, + "step": 16694 + }, + { + "epoch": 6.17, + "learning_rate": 2.3849612021184877e-06, + "loss": 0.4923, + "step": 16695 + }, + { + "epoch": 6.17, + "learning_rate": 2.3849242517551424e-06, + "loss": 0.3954, + "step": 16696 + }, + { + "epoch": 6.17, + "learning_rate": 2.384887301391797e-06, + "loss": 0.3861, + "step": 16697 + }, + { + "epoch": 6.17, + "learning_rate": 2.3848503510284517e-06, + "loss": 0.4759, + "step": 16698 + }, + { + "epoch": 6.17, + "learning_rate": 2.3848134006651063e-06, + "loss": 0.4215, + "step": 16699 + }, + { + "epoch": 6.17, + "learning_rate": 2.3847764503017613e-06, + "loss": 0.3944, + "step": 16700 + }, + { + "epoch": 6.17, + "learning_rate": 2.384739499938416e-06, + "loss": 0.41, + "step": 16701 + }, + { + "epoch": 6.17, + "learning_rate": 2.384702549575071e-06, + "loss": 0.4981, + "step": 16702 + }, + { + "epoch": 6.17, + "learning_rate": 2.3846655992117257e-06, + "loss": 0.5354, + "step": 16703 + }, + { + "epoch": 6.17, + "learning_rate": 2.3846286488483803e-06, + "loss": 0.5331, + "step": 16704 + }, + { + "epoch": 6.17, + "learning_rate": 2.384591698485035e-06, + "loss": 0.4563, + "step": 16705 + }, + { + "epoch": 6.17, + "learning_rate": 2.38455474812169e-06, + "loss": 0.4919, + "step": 16706 + }, + { + "epoch": 6.17, + "learning_rate": 2.3845177977583446e-06, + "loss": 0.3142, + "step": 16707 + }, + { + "epoch": 6.17, + "learning_rate": 2.3844808473949997e-06, + "loss": 0.5144, + "step": 16708 + }, + { + "epoch": 6.17, + "learning_rate": 2.3844438970316543e-06, + "loss": 0.4363, + "step": 16709 + }, + { + "epoch": 6.17, + "learning_rate": 2.384406946668309e-06, + "loss": 0.4757, + "step": 16710 + }, + { + "epoch": 6.17, + "learning_rate": 2.3843699963049636e-06, + "loss": 0.476, + "step": 16711 + }, + { + "epoch": 6.17, + "learning_rate": 2.3843330459416183e-06, + "loss": 0.4262, + "step": 16712 + }, + { + "epoch": 6.17, + "learning_rate": 2.3842960955782733e-06, + "loss": 0.3715, + "step": 16713 + }, + { + "epoch": 6.17, + "learning_rate": 2.384259145214928e-06, + "loss": 0.4121, + "step": 16714 + }, + { + "epoch": 6.17, + "learning_rate": 2.384222194851583e-06, + "loss": 0.3995, + "step": 16715 + }, + { + "epoch": 6.17, + "learning_rate": 2.3841852444882376e-06, + "loss": 0.4131, + "step": 16716 + }, + { + "epoch": 6.17, + "learning_rate": 2.3841482941248923e-06, + "loss": 0.4361, + "step": 16717 + }, + { + "epoch": 6.17, + "learning_rate": 2.384111343761547e-06, + "loss": 0.3281, + "step": 16718 + }, + { + "epoch": 6.17, + "learning_rate": 2.3840743933982016e-06, + "loss": 0.4089, + "step": 16719 + }, + { + "epoch": 6.17, + "learning_rate": 2.3840374430348566e-06, + "loss": 0.5013, + "step": 16720 + }, + { + "epoch": 6.17, + "learning_rate": 2.3840004926715113e-06, + "loss": 0.4975, + "step": 16721 + }, + { + "epoch": 6.18, + "learning_rate": 2.3839635423081663e-06, + "loss": 0.5452, + "step": 16722 + }, + { + "epoch": 6.18, + "learning_rate": 2.383926591944821e-06, + "loss": 0.3973, + "step": 16723 + }, + { + "epoch": 6.18, + "learning_rate": 2.3838896415814756e-06, + "loss": 0.4666, + "step": 16724 + }, + { + "epoch": 6.18, + "learning_rate": 2.3838526912181302e-06, + "loss": 0.3256, + "step": 16725 + }, + { + "epoch": 6.18, + "learning_rate": 2.3838157408547853e-06, + "loss": 0.5581, + "step": 16726 + }, + { + "epoch": 6.18, + "learning_rate": 2.38377879049144e-06, + "loss": 0.4314, + "step": 16727 + }, + { + "epoch": 6.18, + "learning_rate": 2.383741840128095e-06, + "loss": 0.4637, + "step": 16728 + }, + { + "epoch": 6.18, + "learning_rate": 2.3837048897647496e-06, + "loss": 0.5165, + "step": 16729 + }, + { + "epoch": 6.18, + "learning_rate": 2.383667939401404e-06, + "loss": 0.4111, + "step": 16730 + }, + { + "epoch": 6.18, + "learning_rate": 2.383630989038059e-06, + "loss": 0.5064, + "step": 16731 + }, + { + "epoch": 6.18, + "learning_rate": 2.3835940386747135e-06, + "loss": 0.542, + "step": 16732 + }, + { + "epoch": 6.18, + "learning_rate": 2.3835570883113686e-06, + "loss": 0.6277, + "step": 16733 + }, + { + "epoch": 6.18, + "learning_rate": 2.3835201379480232e-06, + "loss": 0.4935, + "step": 16734 + }, + { + "epoch": 6.18, + "learning_rate": 2.3834831875846783e-06, + "loss": 0.4902, + "step": 16735 + }, + { + "epoch": 6.18, + "learning_rate": 2.3834462372213325e-06, + "loss": 0.4295, + "step": 16736 + }, + { + "epoch": 6.18, + "learning_rate": 2.3834092868579876e-06, + "loss": 0.4503, + "step": 16737 + }, + { + "epoch": 6.18, + "learning_rate": 2.383372336494642e-06, + "loss": 0.4426, + "step": 16738 + }, + { + "epoch": 6.18, + "learning_rate": 2.383335386131297e-06, + "loss": 0.4197, + "step": 16739 + }, + { + "epoch": 6.18, + "learning_rate": 2.383298435767952e-06, + "loss": 0.4165, + "step": 16740 + }, + { + "epoch": 6.18, + "learning_rate": 2.3832614854046065e-06, + "loss": 0.5156, + "step": 16741 + }, + { + "epoch": 6.18, + "learning_rate": 2.383224535041261e-06, + "loss": 0.3869, + "step": 16742 + }, + { + "epoch": 6.18, + "learning_rate": 2.383187584677916e-06, + "loss": 0.4157, + "step": 16743 + }, + { + "epoch": 6.18, + "learning_rate": 2.383150634314571e-06, + "loss": 0.3325, + "step": 16744 + }, + { + "epoch": 6.18, + "learning_rate": 2.3831136839512255e-06, + "loss": 0.466, + "step": 16745 + }, + { + "epoch": 6.18, + "learning_rate": 2.3830767335878806e-06, + "loss": 0.2771, + "step": 16746 + }, + { + "epoch": 6.18, + "learning_rate": 2.383039783224535e-06, + "loss": 0.4128, + "step": 16747 + }, + { + "epoch": 6.18, + "learning_rate": 2.38300283286119e-06, + "loss": 0.4447, + "step": 16748 + }, + { + "epoch": 6.19, + "learning_rate": 2.3829658824978445e-06, + "loss": 0.5372, + "step": 16749 + }, + { + "epoch": 6.19, + "learning_rate": 2.382928932134499e-06, + "loss": 0.4802, + "step": 16750 + }, + { + "epoch": 6.19, + "learning_rate": 2.382891981771154e-06, + "loss": 0.4872, + "step": 16751 + }, + { + "epoch": 6.19, + "learning_rate": 2.382855031407809e-06, + "loss": 0.4075, + "step": 16752 + }, + { + "epoch": 6.19, + "learning_rate": 2.382818081044464e-06, + "loss": 0.4422, + "step": 16753 + }, + { + "epoch": 6.19, + "learning_rate": 2.3827811306811185e-06, + "loss": 0.4213, + "step": 16754 + }, + { + "epoch": 6.19, + "learning_rate": 2.382744180317773e-06, + "loss": 0.5063, + "step": 16755 + }, + { + "epoch": 6.19, + "learning_rate": 2.3827072299544278e-06, + "loss": 0.378, + "step": 16756 + }, + { + "epoch": 6.19, + "learning_rate": 2.382670279591083e-06, + "loss": 0.3079, + "step": 16757 + }, + { + "epoch": 6.19, + "learning_rate": 2.3826333292277375e-06, + "loss": 0.3773, + "step": 16758 + }, + { + "epoch": 6.19, + "learning_rate": 2.382596378864392e-06, + "loss": 0.518, + "step": 16759 + }, + { + "epoch": 6.19, + "learning_rate": 2.382559428501047e-06, + "loss": 0.499, + "step": 16760 + }, + { + "epoch": 6.19, + "learning_rate": 2.382522478137702e-06, + "loss": 0.6052, + "step": 16761 + }, + { + "epoch": 6.19, + "learning_rate": 2.3824855277743564e-06, + "loss": 0.4573, + "step": 16762 + }, + { + "epoch": 6.19, + "learning_rate": 2.382448577411011e-06, + "loss": 0.5219, + "step": 16763 + }, + { + "epoch": 6.19, + "learning_rate": 2.382411627047666e-06, + "loss": 0.4988, + "step": 16764 + }, + { + "epoch": 6.19, + "learning_rate": 2.3823746766843208e-06, + "loss": 0.5381, + "step": 16765 + }, + { + "epoch": 6.19, + "learning_rate": 2.382337726320976e-06, + "loss": 0.3227, + "step": 16766 + }, + { + "epoch": 6.19, + "learning_rate": 2.3823007759576305e-06, + "loss": 0.5123, + "step": 16767 + }, + { + "epoch": 6.19, + "learning_rate": 2.382263825594285e-06, + "loss": 0.5168, + "step": 16768 + }, + { + "epoch": 6.19, + "learning_rate": 2.3822268752309397e-06, + "loss": 0.371, + "step": 16769 + }, + { + "epoch": 6.19, + "learning_rate": 2.3821899248675944e-06, + "loss": 0.4948, + "step": 16770 + }, + { + "epoch": 6.19, + "learning_rate": 2.3821529745042494e-06, + "loss": 0.4863, + "step": 16771 + }, + { + "epoch": 6.19, + "learning_rate": 2.382116024140904e-06, + "loss": 0.5041, + "step": 16772 + }, + { + "epoch": 6.19, + "learning_rate": 2.382079073777559e-06, + "loss": 0.5179, + "step": 16773 + }, + { + "epoch": 6.19, + "learning_rate": 2.3820421234142138e-06, + "loss": 0.5605, + "step": 16774 + }, + { + "epoch": 6.19, + "learning_rate": 2.3820051730508684e-06, + "loss": 0.5198, + "step": 16775 + }, + { + "epoch": 6.19, + "learning_rate": 2.381968222687523e-06, + "loss": 0.574, + "step": 16776 + }, + { + "epoch": 6.2, + "learning_rate": 2.381931272324178e-06, + "loss": 0.5497, + "step": 16777 + }, + { + "epoch": 6.2, + "learning_rate": 2.3818943219608327e-06, + "loss": 0.4031, + "step": 16778 + }, + { + "epoch": 6.2, + "learning_rate": 2.3818573715974874e-06, + "loss": 0.5832, + "step": 16779 + }, + { + "epoch": 6.2, + "learning_rate": 2.3818204212341424e-06, + "loss": 0.4208, + "step": 16780 + }, + { + "epoch": 6.2, + "learning_rate": 2.3817834708707966e-06, + "loss": 0.4095, + "step": 16781 + }, + { + "epoch": 6.2, + "learning_rate": 2.3817465205074517e-06, + "loss": 0.4832, + "step": 16782 + }, + { + "epoch": 6.2, + "learning_rate": 2.3817095701441063e-06, + "loss": 0.5731, + "step": 16783 + }, + { + "epoch": 6.2, + "learning_rate": 2.3816726197807614e-06, + "loss": 0.5311, + "step": 16784 + }, + { + "epoch": 6.2, + "learning_rate": 2.381635669417416e-06, + "loss": 0.4478, + "step": 16785 + }, + { + "epoch": 6.2, + "learning_rate": 2.381598719054071e-06, + "loss": 0.3801, + "step": 16786 + }, + { + "epoch": 6.2, + "learning_rate": 2.3815617686907253e-06, + "loss": 0.492, + "step": 16787 + }, + { + "epoch": 6.2, + "learning_rate": 2.38152481832738e-06, + "loss": 0.4225, + "step": 16788 + }, + { + "epoch": 6.2, + "learning_rate": 2.381487867964035e-06, + "loss": 0.4826, + "step": 16789 + }, + { + "epoch": 6.2, + "learning_rate": 2.3814509176006896e-06, + "loss": 0.3902, + "step": 16790 + }, + { + "epoch": 6.2, + "learning_rate": 2.3814139672373447e-06, + "loss": 0.4164, + "step": 16791 + }, + { + "epoch": 6.2, + "learning_rate": 2.3813770168739993e-06, + "loss": 0.4733, + "step": 16792 + }, + { + "epoch": 6.2, + "learning_rate": 2.381340066510654e-06, + "loss": 0.4552, + "step": 16793 + }, + { + "epoch": 6.2, + "learning_rate": 2.3813031161473086e-06, + "loss": 0.4479, + "step": 16794 + }, + { + "epoch": 6.2, + "learning_rate": 2.3812661657839637e-06, + "loss": 0.539, + "step": 16795 + }, + { + "epoch": 6.2, + "learning_rate": 2.3812292154206183e-06, + "loss": 0.4349, + "step": 16796 + }, + { + "epoch": 6.2, + "learning_rate": 2.3811922650572734e-06, + "loss": 0.357, + "step": 16797 + }, + { + "epoch": 6.2, + "learning_rate": 2.381155314693928e-06, + "loss": 0.5292, + "step": 16798 + }, + { + "epoch": 6.2, + "learning_rate": 2.3811183643305826e-06, + "loss": 0.4524, + "step": 16799 + }, + { + "epoch": 6.2, + "learning_rate": 2.3810814139672373e-06, + "loss": 0.5791, + "step": 16800 + }, + { + "epoch": 6.2, + "learning_rate": 2.381044463603892e-06, + "loss": 0.5713, + "step": 16801 + }, + { + "epoch": 6.2, + "learning_rate": 2.381007513240547e-06, + "loss": 0.4585, + "step": 16802 + }, + { + "epoch": 6.2, + "learning_rate": 2.3809705628772016e-06, + "loss": 0.4252, + "step": 16803 + }, + { + "epoch": 6.21, + "learning_rate": 2.3809336125138567e-06, + "loss": 0.4123, + "step": 16804 + }, + { + "epoch": 6.21, + "learning_rate": 2.3808966621505113e-06, + "loss": 0.4071, + "step": 16805 + }, + { + "epoch": 6.21, + "learning_rate": 2.380859711787166e-06, + "loss": 0.4637, + "step": 16806 + }, + { + "epoch": 6.21, + "learning_rate": 2.3808227614238206e-06, + "loss": 0.4917, + "step": 16807 + }, + { + "epoch": 6.21, + "learning_rate": 2.380785811060475e-06, + "loss": 0.4768, + "step": 16808 + }, + { + "epoch": 6.21, + "learning_rate": 2.3807488606971303e-06, + "loss": 0.429, + "step": 16809 + }, + { + "epoch": 6.21, + "learning_rate": 2.380711910333785e-06, + "loss": 0.4017, + "step": 16810 + }, + { + "epoch": 6.21, + "learning_rate": 2.38067495997044e-06, + "loss": 0.3524, + "step": 16811 + }, + { + "epoch": 6.21, + "learning_rate": 2.3806380096070946e-06, + "loss": 0.5041, + "step": 16812 + }, + { + "epoch": 6.21, + "learning_rate": 2.3806010592437492e-06, + "loss": 0.4928, + "step": 16813 + }, + { + "epoch": 6.21, + "learning_rate": 2.380564108880404e-06, + "loss": 0.6055, + "step": 16814 + }, + { + "epoch": 6.21, + "learning_rate": 2.380527158517059e-06, + "loss": 0.5143, + "step": 16815 + }, + { + "epoch": 6.21, + "learning_rate": 2.3804902081537136e-06, + "loss": 0.4772, + "step": 16816 + }, + { + "epoch": 6.21, + "learning_rate": 2.3804532577903686e-06, + "loss": 0.3602, + "step": 16817 + }, + { + "epoch": 6.21, + "learning_rate": 2.3804163074270233e-06, + "loss": 0.6246, + "step": 16818 + }, + { + "epoch": 6.21, + "learning_rate": 2.380379357063678e-06, + "loss": 0.4276, + "step": 16819 + }, + { + "epoch": 6.21, + "learning_rate": 2.3803424067003325e-06, + "loss": 0.4647, + "step": 16820 + }, + { + "epoch": 6.21, + "learning_rate": 2.380305456336987e-06, + "loss": 0.456, + "step": 16821 + }, + { + "epoch": 6.21, + "learning_rate": 2.3802685059736422e-06, + "loss": 0.315, + "step": 16822 + }, + { + "epoch": 6.21, + "learning_rate": 2.380231555610297e-06, + "loss": 0.4692, + "step": 16823 + }, + { + "epoch": 6.21, + "learning_rate": 2.380194605246952e-06, + "loss": 0.3654, + "step": 16824 + }, + { + "epoch": 6.21, + "learning_rate": 2.3801576548836066e-06, + "loss": 0.4329, + "step": 16825 + }, + { + "epoch": 6.21, + "learning_rate": 2.380120704520261e-06, + "loss": 0.4163, + "step": 16826 + }, + { + "epoch": 6.21, + "learning_rate": 2.380083754156916e-06, + "loss": 0.3941, + "step": 16827 + }, + { + "epoch": 6.21, + "learning_rate": 2.3800468037935705e-06, + "loss": 0.465, + "step": 16828 + }, + { + "epoch": 6.21, + "learning_rate": 2.3800098534302255e-06, + "loss": 0.4243, + "step": 16829 + }, + { + "epoch": 6.21, + "learning_rate": 2.37997290306688e-06, + "loss": 0.3917, + "step": 16830 + }, + { + "epoch": 6.22, + "learning_rate": 2.3799359527035352e-06, + "loss": 0.5382, + "step": 16831 + }, + { + "epoch": 6.22, + "learning_rate": 2.3798990023401895e-06, + "loss": 0.4486, + "step": 16832 + }, + { + "epoch": 6.22, + "learning_rate": 2.3798620519768445e-06, + "loss": 0.5057, + "step": 16833 + }, + { + "epoch": 6.22, + "learning_rate": 2.379825101613499e-06, + "loss": 0.4286, + "step": 16834 + }, + { + "epoch": 6.22, + "learning_rate": 2.379788151250154e-06, + "loss": 0.3925, + "step": 16835 + }, + { + "epoch": 6.22, + "learning_rate": 2.379751200886809e-06, + "loss": 0.4035, + "step": 16836 + }, + { + "epoch": 6.22, + "learning_rate": 2.379714250523464e-06, + "loss": 0.4982, + "step": 16837 + }, + { + "epoch": 6.22, + "learning_rate": 2.379677300160118e-06, + "loss": 0.5387, + "step": 16838 + }, + { + "epoch": 6.22, + "learning_rate": 2.3796403497967728e-06, + "loss": 0.439, + "step": 16839 + }, + { + "epoch": 6.22, + "learning_rate": 2.379603399433428e-06, + "loss": 0.5141, + "step": 16840 + }, + { + "epoch": 6.22, + "learning_rate": 2.3795664490700824e-06, + "loss": 0.5034, + "step": 16841 + }, + { + "epoch": 6.22, + "learning_rate": 2.3795294987067375e-06, + "loss": 0.5521, + "step": 16842 + }, + { + "epoch": 6.22, + "learning_rate": 2.379492548343392e-06, + "loss": 0.3763, + "step": 16843 + }, + { + "epoch": 6.22, + "learning_rate": 2.3794555979800468e-06, + "loss": 0.4105, + "step": 16844 + }, + { + "epoch": 6.22, + "learning_rate": 2.3794186476167014e-06, + "loss": 0.4266, + "step": 16845 + }, + { + "epoch": 6.22, + "learning_rate": 2.3793816972533565e-06, + "loss": 0.5217, + "step": 16846 + }, + { + "epoch": 6.22, + "learning_rate": 2.379344746890011e-06, + "loss": 0.4189, + "step": 16847 + }, + { + "epoch": 6.22, + "learning_rate": 2.3793077965266658e-06, + "loss": 0.4062, + "step": 16848 + }, + { + "epoch": 6.22, + "learning_rate": 2.379270846163321e-06, + "loss": 0.4405, + "step": 16849 + }, + { + "epoch": 6.22, + "learning_rate": 2.3792338957999754e-06, + "loss": 0.5486, + "step": 16850 + }, + { + "epoch": 6.22, + "learning_rate": 2.37919694543663e-06, + "loss": 0.3965, + "step": 16851 + }, + { + "epoch": 6.22, + "learning_rate": 2.3791599950732847e-06, + "loss": 0.4361, + "step": 16852 + }, + { + "epoch": 6.22, + "learning_rate": 2.3791230447099398e-06, + "loss": 0.4199, + "step": 16853 + }, + { + "epoch": 6.22, + "learning_rate": 2.3790860943465944e-06, + "loss": 0.3777, + "step": 16854 + }, + { + "epoch": 6.22, + "learning_rate": 2.3790491439832495e-06, + "loss": 0.4211, + "step": 16855 + }, + { + "epoch": 6.22, + "learning_rate": 2.379012193619904e-06, + "loss": 0.4569, + "step": 16856 + }, + { + "epoch": 6.22, + "learning_rate": 2.3789752432565587e-06, + "loss": 0.6187, + "step": 16857 + }, + { + "epoch": 6.23, + "learning_rate": 2.3789382928932134e-06, + "loss": 0.3965, + "step": 16858 + }, + { + "epoch": 6.23, + "learning_rate": 2.378901342529868e-06, + "loss": 0.391, + "step": 16859 + }, + { + "epoch": 6.23, + "learning_rate": 2.378864392166523e-06, + "loss": 0.4685, + "step": 16860 + }, + { + "epoch": 6.23, + "learning_rate": 2.3788274418031777e-06, + "loss": 0.3814, + "step": 16861 + }, + { + "epoch": 6.23, + "learning_rate": 2.3787904914398328e-06, + "loss": 0.4091, + "step": 16862 + }, + { + "epoch": 6.23, + "learning_rate": 2.3787535410764874e-06, + "loss": 0.6284, + "step": 16863 + }, + { + "epoch": 6.23, + "learning_rate": 2.378716590713142e-06, + "loss": 0.5691, + "step": 16864 + }, + { + "epoch": 6.23, + "learning_rate": 2.3786796403497967e-06, + "loss": 0.4177, + "step": 16865 + }, + { + "epoch": 6.23, + "learning_rate": 2.3786426899864517e-06, + "loss": 0.3986, + "step": 16866 + }, + { + "epoch": 6.23, + "learning_rate": 2.3786057396231064e-06, + "loss": 0.4092, + "step": 16867 + }, + { + "epoch": 6.23, + "learning_rate": 2.378568789259761e-06, + "loss": 0.5316, + "step": 16868 + }, + { + "epoch": 6.23, + "learning_rate": 2.378531838896416e-06, + "loss": 0.4366, + "step": 16869 + }, + { + "epoch": 6.23, + "learning_rate": 2.3784948885330707e-06, + "loss": 0.5083, + "step": 16870 + }, + { + "epoch": 6.23, + "learning_rate": 2.3784579381697254e-06, + "loss": 0.3766, + "step": 16871 + }, + { + "epoch": 6.23, + "learning_rate": 2.37842098780638e-06, + "loss": 0.4991, + "step": 16872 + }, + { + "epoch": 6.23, + "learning_rate": 2.378384037443035e-06, + "loss": 0.4477, + "step": 16873 + }, + { + "epoch": 6.23, + "learning_rate": 2.3783470870796897e-06, + "loss": 0.4741, + "step": 16874 + }, + { + "epoch": 6.23, + "learning_rate": 2.3783101367163447e-06, + "loss": 0.5591, + "step": 16875 + }, + { + "epoch": 6.23, + "learning_rate": 2.3782731863529994e-06, + "loss": 0.5061, + "step": 16876 + }, + { + "epoch": 6.23, + "learning_rate": 2.378236235989654e-06, + "loss": 0.5051, + "step": 16877 + }, + { + "epoch": 6.23, + "learning_rate": 2.3781992856263087e-06, + "loss": 0.4251, + "step": 16878 + }, + { + "epoch": 6.23, + "learning_rate": 2.3781623352629633e-06, + "loss": 0.4195, + "step": 16879 + }, + { + "epoch": 6.23, + "learning_rate": 2.3781253848996184e-06, + "loss": 0.3502, + "step": 16880 + }, + { + "epoch": 6.23, + "learning_rate": 2.378088434536273e-06, + "loss": 0.4227, + "step": 16881 + }, + { + "epoch": 6.23, + "learning_rate": 2.3780514841729276e-06, + "loss": 0.4409, + "step": 16882 + }, + { + "epoch": 6.23, + "learning_rate": 2.3780145338095823e-06, + "loss": 0.4701, + "step": 16883 + }, + { + "epoch": 6.23, + "learning_rate": 2.3779775834462373e-06, + "loss": 0.4824, + "step": 16884 + }, + { + "epoch": 6.24, + "learning_rate": 2.377940633082892e-06, + "loss": 0.5285, + "step": 16885 + }, + { + "epoch": 6.24, + "learning_rate": 2.377903682719547e-06, + "loss": 0.3362, + "step": 16886 + }, + { + "epoch": 6.24, + "learning_rate": 2.3778667323562017e-06, + "loss": 0.4772, + "step": 16887 + }, + { + "epoch": 6.24, + "learning_rate": 2.3778297819928563e-06, + "loss": 0.4888, + "step": 16888 + }, + { + "epoch": 6.24, + "learning_rate": 2.377792831629511e-06, + "loss": 0.322, + "step": 16889 + }, + { + "epoch": 6.24, + "learning_rate": 2.3777558812661656e-06, + "loss": 0.5605, + "step": 16890 + }, + { + "epoch": 6.24, + "learning_rate": 2.3777189309028206e-06, + "loss": 0.4393, + "step": 16891 + }, + { + "epoch": 6.24, + "learning_rate": 2.3776819805394753e-06, + "loss": 0.4166, + "step": 16892 + }, + { + "epoch": 6.24, + "learning_rate": 2.3776450301761303e-06, + "loss": 0.3159, + "step": 16893 + }, + { + "epoch": 6.24, + "learning_rate": 2.377608079812785e-06, + "loss": 0.4763, + "step": 16894 + }, + { + "epoch": 6.24, + "learning_rate": 2.3775711294494396e-06, + "loss": 0.5003, + "step": 16895 + }, + { + "epoch": 6.24, + "learning_rate": 2.3775341790860942e-06, + "loss": 0.4176, + "step": 16896 + }, + { + "epoch": 6.24, + "learning_rate": 2.3774972287227493e-06, + "loss": 0.4085, + "step": 16897 + }, + { + "epoch": 6.24, + "learning_rate": 2.377460278359404e-06, + "loss": 0.4859, + "step": 16898 + }, + { + "epoch": 6.24, + "learning_rate": 2.3774233279960586e-06, + "loss": 0.4686, + "step": 16899 + }, + { + "epoch": 6.24, + "learning_rate": 2.3773863776327136e-06, + "loss": 0.5291, + "step": 16900 + }, + { + "epoch": 6.24, + "learning_rate": 2.3773494272693683e-06, + "loss": 0.5119, + "step": 16901 + }, + { + "epoch": 6.24, + "learning_rate": 2.377312476906023e-06, + "loss": 0.383, + "step": 16902 + }, + { + "epoch": 6.24, + "learning_rate": 2.3772755265426775e-06, + "loss": 0.347, + "step": 16903 + }, + { + "epoch": 6.24, + "learning_rate": 2.3772385761793326e-06, + "loss": 0.5352, + "step": 16904 + }, + { + "epoch": 6.24, + "learning_rate": 2.3772016258159872e-06, + "loss": 0.4153, + "step": 16905 + }, + { + "epoch": 6.24, + "learning_rate": 2.3771646754526423e-06, + "loss": 0.4301, + "step": 16906 + }, + { + "epoch": 6.24, + "learning_rate": 2.377127725089297e-06, + "loss": 0.5241, + "step": 16907 + }, + { + "epoch": 6.24, + "learning_rate": 2.3770907747259516e-06, + "loss": 0.3413, + "step": 16908 + }, + { + "epoch": 6.24, + "learning_rate": 2.377053824362606e-06, + "loss": 0.4441, + "step": 16909 + }, + { + "epoch": 6.24, + "learning_rate": 2.377016873999261e-06, + "loss": 0.4867, + "step": 16910 + }, + { + "epoch": 6.24, + "learning_rate": 2.376979923635916e-06, + "loss": 0.5236, + "step": 16911 + }, + { + "epoch": 6.25, + "learning_rate": 2.3769429732725705e-06, + "loss": 0.3322, + "step": 16912 + }, + { + "epoch": 6.25, + "learning_rate": 2.3769060229092256e-06, + "loss": 0.4989, + "step": 16913 + }, + { + "epoch": 6.25, + "learning_rate": 2.3768690725458802e-06, + "loss": 0.4205, + "step": 16914 + }, + { + "epoch": 6.25, + "learning_rate": 2.376832122182535e-06, + "loss": 0.456, + "step": 16915 + }, + { + "epoch": 6.25, + "learning_rate": 2.3767951718191895e-06, + "loss": 0.3156, + "step": 16916 + }, + { + "epoch": 6.25, + "learning_rate": 2.3767582214558446e-06, + "loss": 0.5281, + "step": 16917 + }, + { + "epoch": 6.25, + "learning_rate": 2.376721271092499e-06, + "loss": 0.3419, + "step": 16918 + }, + { + "epoch": 6.25, + "learning_rate": 2.376684320729154e-06, + "loss": 0.458, + "step": 16919 + }, + { + "epoch": 6.25, + "learning_rate": 2.376647370365809e-06, + "loss": 0.4935, + "step": 16920 + }, + { + "epoch": 6.25, + "learning_rate": 2.376610420002463e-06, + "loss": 0.297, + "step": 16921 + }, + { + "epoch": 6.25, + "learning_rate": 2.376573469639118e-06, + "loss": 0.4542, + "step": 16922 + }, + { + "epoch": 6.25, + "learning_rate": 2.376536519275773e-06, + "loss": 0.4216, + "step": 16923 + }, + { + "epoch": 6.25, + "learning_rate": 2.376499568912428e-06, + "loss": 0.4111, + "step": 16924 + }, + { + "epoch": 6.25, + "learning_rate": 2.3764626185490825e-06, + "loss": 0.5619, + "step": 16925 + }, + { + "epoch": 6.25, + "learning_rate": 2.3764256681857376e-06, + "loss": 0.3504, + "step": 16926 + }, + { + "epoch": 6.25, + "learning_rate": 2.3763887178223918e-06, + "loss": 0.5311, + "step": 16927 + }, + { + "epoch": 6.25, + "learning_rate": 2.3763517674590464e-06, + "loss": 0.5075, + "step": 16928 + }, + { + "epoch": 6.25, + "learning_rate": 2.3763148170957015e-06, + "loss": 0.4363, + "step": 16929 + }, + { + "epoch": 6.25, + "learning_rate": 2.376277866732356e-06, + "loss": 0.472, + "step": 16930 + }, + { + "epoch": 6.25, + "learning_rate": 2.376240916369011e-06, + "loss": 0.4701, + "step": 16931 + }, + { + "epoch": 6.25, + "learning_rate": 2.376203966005666e-06, + "loss": 0.4029, + "step": 16932 + }, + { + "epoch": 6.25, + "learning_rate": 2.3761670156423204e-06, + "loss": 0.3569, + "step": 16933 + }, + { + "epoch": 6.25, + "learning_rate": 2.376130065278975e-06, + "loss": 0.4879, + "step": 16934 + }, + { + "epoch": 6.25, + "learning_rate": 2.37609311491563e-06, + "loss": 0.4078, + "step": 16935 + }, + { + "epoch": 6.25, + "learning_rate": 2.3760561645522848e-06, + "loss": 0.4028, + "step": 16936 + }, + { + "epoch": 6.25, + "learning_rate": 2.37601921418894e-06, + "loss": 0.3769, + "step": 16937 + }, + { + "epoch": 6.25, + "learning_rate": 2.3759822638255945e-06, + "loss": 0.4959, + "step": 16938 + }, + { + "epoch": 6.26, + "learning_rate": 2.375945313462249e-06, + "loss": 0.4431, + "step": 16939 + }, + { + "epoch": 6.26, + "learning_rate": 2.3759083630989037e-06, + "loss": 0.4602, + "step": 16940 + }, + { + "epoch": 6.26, + "learning_rate": 2.3758714127355584e-06, + "loss": 0.4455, + "step": 16941 + }, + { + "epoch": 6.26, + "learning_rate": 2.3758344623722134e-06, + "loss": 0.4768, + "step": 16942 + }, + { + "epoch": 6.26, + "learning_rate": 2.375797512008868e-06, + "loss": 0.388, + "step": 16943 + }, + { + "epoch": 6.26, + "learning_rate": 2.375760561645523e-06, + "loss": 0.4543, + "step": 16944 + }, + { + "epoch": 6.26, + "learning_rate": 2.3757236112821778e-06, + "loss": 0.4307, + "step": 16945 + }, + { + "epoch": 6.26, + "learning_rate": 2.3756866609188324e-06, + "loss": 0.4751, + "step": 16946 + }, + { + "epoch": 6.26, + "learning_rate": 2.375649710555487e-06, + "loss": 0.3937, + "step": 16947 + }, + { + "epoch": 6.26, + "learning_rate": 2.3756127601921417e-06, + "loss": 0.3706, + "step": 16948 + }, + { + "epoch": 6.26, + "learning_rate": 2.3755758098287967e-06, + "loss": 0.4115, + "step": 16949 + }, + { + "epoch": 6.26, + "learning_rate": 2.3755388594654514e-06, + "loss": 0.4949, + "step": 16950 + }, + { + "epoch": 6.26, + "learning_rate": 2.3755019091021064e-06, + "loss": 0.4686, + "step": 16951 + }, + { + "epoch": 6.26, + "learning_rate": 2.375464958738761e-06, + "loss": 0.4266, + "step": 16952 + }, + { + "epoch": 6.26, + "learning_rate": 2.3754280083754157e-06, + "loss": 0.3851, + "step": 16953 + }, + { + "epoch": 6.26, + "learning_rate": 2.3753910580120703e-06, + "loss": 0.3437, + "step": 16954 + }, + { + "epoch": 6.26, + "learning_rate": 2.3753541076487254e-06, + "loss": 0.323, + "step": 16955 + }, + { + "epoch": 6.26, + "learning_rate": 2.37531715728538e-06, + "loss": 0.5086, + "step": 16956 + }, + { + "epoch": 6.26, + "learning_rate": 2.375280206922035e-06, + "loss": 0.3519, + "step": 16957 + }, + { + "epoch": 6.26, + "learning_rate": 2.3752432565586897e-06, + "loss": 0.515, + "step": 16958 + }, + { + "epoch": 6.26, + "learning_rate": 2.3752063061953444e-06, + "loss": 0.3882, + "step": 16959 + }, + { + "epoch": 6.26, + "learning_rate": 2.375169355831999e-06, + "loss": 0.4882, + "step": 16960 + }, + { + "epoch": 6.26, + "learning_rate": 2.3751324054686536e-06, + "loss": 0.4604, + "step": 16961 + }, + { + "epoch": 6.26, + "learning_rate": 2.3750954551053087e-06, + "loss": 0.4906, + "step": 16962 + }, + { + "epoch": 6.26, + "learning_rate": 2.3750585047419633e-06, + "loss": 0.5129, + "step": 16963 + }, + { + "epoch": 6.26, + "learning_rate": 2.3750215543786184e-06, + "loss": 0.5457, + "step": 16964 + }, + { + "epoch": 6.26, + "learning_rate": 2.374984604015273e-06, + "loss": 0.4045, + "step": 16965 + }, + { + "epoch": 6.27, + "learning_rate": 2.3749476536519277e-06, + "loss": 0.4109, + "step": 16966 + }, + { + "epoch": 6.27, + "learning_rate": 2.3749107032885823e-06, + "loss": 0.52, + "step": 16967 + }, + { + "epoch": 6.27, + "learning_rate": 2.374873752925237e-06, + "loss": 0.4098, + "step": 16968 + }, + { + "epoch": 6.27, + "learning_rate": 2.374836802561892e-06, + "loss": 0.3218, + "step": 16969 + }, + { + "epoch": 6.27, + "learning_rate": 2.3747998521985466e-06, + "loss": 0.4889, + "step": 16970 + }, + { + "epoch": 6.27, + "learning_rate": 2.3747629018352017e-06, + "loss": 0.5105, + "step": 16971 + }, + { + "epoch": 6.27, + "learning_rate": 2.374725951471856e-06, + "loss": 0.4994, + "step": 16972 + }, + { + "epoch": 6.27, + "learning_rate": 2.374689001108511e-06, + "loss": 0.4326, + "step": 16973 + }, + { + "epoch": 6.27, + "learning_rate": 2.3746520507451656e-06, + "loss": 0.4448, + "step": 16974 + }, + { + "epoch": 6.27, + "learning_rate": 2.3746151003818207e-06, + "loss": 0.4368, + "step": 16975 + }, + { + "epoch": 6.27, + "learning_rate": 2.3745781500184753e-06, + "loss": 0.4253, + "step": 16976 + }, + { + "epoch": 6.27, + "learning_rate": 2.3745411996551304e-06, + "loss": 0.4761, + "step": 16977 + }, + { + "epoch": 6.27, + "learning_rate": 2.3745042492917846e-06, + "loss": 0.551, + "step": 16978 + }, + { + "epoch": 6.27, + "learning_rate": 2.3744672989284392e-06, + "loss": 0.3764, + "step": 16979 + }, + { + "epoch": 6.27, + "learning_rate": 2.3744303485650943e-06, + "loss": 0.4211, + "step": 16980 + }, + { + "epoch": 6.27, + "learning_rate": 2.374393398201749e-06, + "loss": 0.4376, + "step": 16981 + }, + { + "epoch": 6.27, + "learning_rate": 2.374356447838404e-06, + "loss": 0.4346, + "step": 16982 + }, + { + "epoch": 6.27, + "learning_rate": 2.3743194974750586e-06, + "loss": 0.4583, + "step": 16983 + }, + { + "epoch": 6.27, + "learning_rate": 2.3742825471117132e-06, + "loss": 0.3662, + "step": 16984 + }, + { + "epoch": 6.27, + "learning_rate": 2.374245596748368e-06, + "loss": 0.4229, + "step": 16985 + }, + { + "epoch": 6.27, + "learning_rate": 2.374208646385023e-06, + "loss": 0.5453, + "step": 16986 + }, + { + "epoch": 6.27, + "learning_rate": 2.3741716960216776e-06, + "loss": 0.4725, + "step": 16987 + }, + { + "epoch": 6.27, + "learning_rate": 2.3741347456583322e-06, + "loss": 0.433, + "step": 16988 + }, + { + "epoch": 6.27, + "learning_rate": 2.3740977952949873e-06, + "loss": 0.3751, + "step": 16989 + }, + { + "epoch": 6.27, + "learning_rate": 2.374060844931642e-06, + "loss": 0.4868, + "step": 16990 + }, + { + "epoch": 6.27, + "learning_rate": 2.3740238945682965e-06, + "loss": 0.4366, + "step": 16991 + }, + { + "epoch": 6.27, + "learning_rate": 2.373986944204951e-06, + "loss": 0.4574, + "step": 16992 + }, + { + "epoch": 6.28, + "learning_rate": 2.3739499938416062e-06, + "loss": 0.4986, + "step": 16993 + }, + { + "epoch": 6.28, + "learning_rate": 2.373913043478261e-06, + "loss": 0.4784, + "step": 16994 + }, + { + "epoch": 6.28, + "learning_rate": 2.373876093114916e-06, + "loss": 0.5717, + "step": 16995 + }, + { + "epoch": 6.28, + "learning_rate": 2.3738391427515706e-06, + "loss": 0.5566, + "step": 16996 + }, + { + "epoch": 6.28, + "learning_rate": 2.3738021923882252e-06, + "loss": 0.3715, + "step": 16997 + }, + { + "epoch": 6.28, + "learning_rate": 2.37376524202488e-06, + "loss": 0.4958, + "step": 16998 + }, + { + "epoch": 6.28, + "learning_rate": 2.3737282916615345e-06, + "loss": 0.4193, + "step": 16999 + }, + { + "epoch": 6.28, + "learning_rate": 2.3736913412981895e-06, + "loss": 0.4693, + "step": 17000 + }, + { + "epoch": 6.28, + "learning_rate": 2.373654390934844e-06, + "loss": 0.5113, + "step": 17001 + }, + { + "epoch": 6.28, + "learning_rate": 2.3736174405714992e-06, + "loss": 0.3975, + "step": 17002 + }, + { + "epoch": 6.28, + "learning_rate": 2.373580490208154e-06, + "loss": 0.2852, + "step": 17003 + }, + { + "epoch": 6.28, + "learning_rate": 2.3735435398448085e-06, + "loss": 0.3907, + "step": 17004 + }, + { + "epoch": 6.28, + "learning_rate": 2.373506589481463e-06, + "loss": 0.4787, + "step": 17005 + }, + { + "epoch": 6.28, + "learning_rate": 2.3734696391181182e-06, + "loss": 0.4882, + "step": 17006 + }, + { + "epoch": 6.28, + "learning_rate": 2.373432688754773e-06, + "loss": 0.4341, + "step": 17007 + }, + { + "epoch": 6.28, + "learning_rate": 2.3733957383914275e-06, + "loss": 0.4682, + "step": 17008 + }, + { + "epoch": 6.28, + "learning_rate": 2.3733587880280825e-06, + "loss": 0.5286, + "step": 17009 + }, + { + "epoch": 6.28, + "learning_rate": 2.373321837664737e-06, + "loss": 0.526, + "step": 17010 + }, + { + "epoch": 6.28, + "learning_rate": 2.373284887301392e-06, + "loss": 0.4567, + "step": 17011 + }, + { + "epoch": 6.28, + "learning_rate": 2.3732479369380465e-06, + "loss": 0.4047, + "step": 17012 + }, + { + "epoch": 6.28, + "learning_rate": 2.3732109865747015e-06, + "loss": 0.4411, + "step": 17013 + }, + { + "epoch": 6.28, + "learning_rate": 2.373174036211356e-06, + "loss": 0.3834, + "step": 17014 + }, + { + "epoch": 6.28, + "learning_rate": 2.373137085848011e-06, + "loss": 0.3598, + "step": 17015 + }, + { + "epoch": 6.28, + "learning_rate": 2.373100135484666e-06, + "loss": 0.4203, + "step": 17016 + }, + { + "epoch": 6.28, + "learning_rate": 2.37306318512132e-06, + "loss": 0.4913, + "step": 17017 + }, + { + "epoch": 6.28, + "learning_rate": 2.373026234757975e-06, + "loss": 0.4778, + "step": 17018 + }, + { + "epoch": 6.28, + "learning_rate": 2.3729892843946298e-06, + "loss": 0.4778, + "step": 17019 + }, + { + "epoch": 6.29, + "learning_rate": 2.372952334031285e-06, + "loss": 0.4145, + "step": 17020 + }, + { + "epoch": 6.29, + "learning_rate": 2.3729153836679395e-06, + "loss": 0.3977, + "step": 17021 + }, + { + "epoch": 6.29, + "learning_rate": 2.3728784333045945e-06, + "loss": 0.4762, + "step": 17022 + }, + { + "epoch": 6.29, + "learning_rate": 2.3728414829412487e-06, + "loss": 0.4696, + "step": 17023 + }, + { + "epoch": 6.29, + "learning_rate": 2.3728045325779038e-06, + "loss": 0.5331, + "step": 17024 + }, + { + "epoch": 6.29, + "learning_rate": 2.3727675822145584e-06, + "loss": 0.4808, + "step": 17025 + }, + { + "epoch": 6.29, + "learning_rate": 2.3727306318512135e-06, + "loss": 0.503, + "step": 17026 + }, + { + "epoch": 6.29, + "learning_rate": 2.372693681487868e-06, + "loss": 0.4996, + "step": 17027 + }, + { + "epoch": 6.29, + "learning_rate": 2.3726567311245228e-06, + "loss": 0.444, + "step": 17028 + }, + { + "epoch": 6.29, + "learning_rate": 2.3726197807611774e-06, + "loss": 0.3986, + "step": 17029 + }, + { + "epoch": 6.29, + "learning_rate": 2.372582830397832e-06, + "loss": 0.4091, + "step": 17030 + }, + { + "epoch": 6.29, + "learning_rate": 2.372545880034487e-06, + "loss": 0.5297, + "step": 17031 + }, + { + "epoch": 6.29, + "learning_rate": 2.3725089296711417e-06, + "loss": 0.4577, + "step": 17032 + }, + { + "epoch": 6.29, + "learning_rate": 2.3724719793077968e-06, + "loss": 0.42, + "step": 17033 + }, + { + "epoch": 6.29, + "learning_rate": 2.3724350289444514e-06, + "loss": 0.5637, + "step": 17034 + }, + { + "epoch": 6.29, + "learning_rate": 2.372398078581106e-06, + "loss": 0.5388, + "step": 17035 + }, + { + "epoch": 6.29, + "learning_rate": 2.3723611282177607e-06, + "loss": 0.4587, + "step": 17036 + }, + { + "epoch": 6.29, + "learning_rate": 2.3723241778544153e-06, + "loss": 0.421, + "step": 17037 + }, + { + "epoch": 6.29, + "learning_rate": 2.3722872274910704e-06, + "loss": 0.4067, + "step": 17038 + }, + { + "epoch": 6.29, + "learning_rate": 2.372250277127725e-06, + "loss": 0.429, + "step": 17039 + }, + { + "epoch": 6.29, + "learning_rate": 2.37221332676438e-06, + "loss": 0.3815, + "step": 17040 + }, + { + "epoch": 6.29, + "learning_rate": 2.3721763764010347e-06, + "loss": 0.4803, + "step": 17041 + }, + { + "epoch": 6.29, + "learning_rate": 2.3721394260376894e-06, + "loss": 0.4099, + "step": 17042 + }, + { + "epoch": 6.29, + "learning_rate": 2.372102475674344e-06, + "loss": 0.3901, + "step": 17043 + }, + { + "epoch": 6.29, + "learning_rate": 2.372065525310999e-06, + "loss": 0.4437, + "step": 17044 + }, + { + "epoch": 6.29, + "learning_rate": 2.3720285749476537e-06, + "loss": 0.4909, + "step": 17045 + }, + { + "epoch": 6.29, + "learning_rate": 2.3719916245843088e-06, + "loss": 0.3675, + "step": 17046 + }, + { + "epoch": 6.3, + "learning_rate": 2.3719546742209634e-06, + "loss": 0.5241, + "step": 17047 + }, + { + "epoch": 6.3, + "learning_rate": 2.371917723857618e-06, + "loss": 0.4902, + "step": 17048 + }, + { + "epoch": 6.3, + "learning_rate": 2.3718807734942727e-06, + "loss": 0.4522, + "step": 17049 + }, + { + "epoch": 6.3, + "learning_rate": 2.3718438231309273e-06, + "loss": 0.4579, + "step": 17050 + }, + { + "epoch": 6.3, + "learning_rate": 2.3718068727675824e-06, + "loss": 0.4227, + "step": 17051 + }, + { + "epoch": 6.3, + "learning_rate": 2.371769922404237e-06, + "loss": 0.4111, + "step": 17052 + }, + { + "epoch": 6.3, + "learning_rate": 2.371732972040892e-06, + "loss": 0.4244, + "step": 17053 + }, + { + "epoch": 6.3, + "learning_rate": 2.3716960216775467e-06, + "loss": 0.5969, + "step": 17054 + }, + { + "epoch": 6.3, + "learning_rate": 2.3716590713142013e-06, + "loss": 0.4909, + "step": 17055 + }, + { + "epoch": 6.3, + "learning_rate": 2.371622120950856e-06, + "loss": 0.5562, + "step": 17056 + }, + { + "epoch": 6.3, + "learning_rate": 2.3715851705875106e-06, + "loss": 0.5495, + "step": 17057 + }, + { + "epoch": 6.3, + "learning_rate": 2.3715482202241657e-06, + "loss": 0.3677, + "step": 17058 + }, + { + "epoch": 6.3, + "learning_rate": 2.3715112698608203e-06, + "loss": 0.3882, + "step": 17059 + }, + { + "epoch": 6.3, + "learning_rate": 2.3714743194974754e-06, + "loss": 0.5066, + "step": 17060 + }, + { + "epoch": 6.3, + "learning_rate": 2.37143736913413e-06, + "loss": 0.4571, + "step": 17061 + }, + { + "epoch": 6.3, + "learning_rate": 2.3714004187707846e-06, + "loss": 0.4849, + "step": 17062 + }, + { + "epoch": 6.3, + "learning_rate": 2.3713634684074393e-06, + "loss": 0.5482, + "step": 17063 + }, + { + "epoch": 6.3, + "learning_rate": 2.3713265180440943e-06, + "loss": 0.5281, + "step": 17064 + }, + { + "epoch": 6.3, + "learning_rate": 2.371289567680749e-06, + "loss": 0.4654, + "step": 17065 + }, + { + "epoch": 6.3, + "learning_rate": 2.371252617317404e-06, + "loss": 0.5294, + "step": 17066 + }, + { + "epoch": 6.3, + "learning_rate": 2.3712156669540587e-06, + "loss": 0.399, + "step": 17067 + }, + { + "epoch": 6.3, + "learning_rate": 2.371178716590713e-06, + "loss": 0.4269, + "step": 17068 + }, + { + "epoch": 6.3, + "learning_rate": 2.371141766227368e-06, + "loss": 0.5076, + "step": 17069 + }, + { + "epoch": 6.3, + "learning_rate": 2.3711048158640226e-06, + "loss": 0.3982, + "step": 17070 + }, + { + "epoch": 6.3, + "learning_rate": 2.3710678655006776e-06, + "loss": 0.619, + "step": 17071 + }, + { + "epoch": 6.3, + "learning_rate": 2.3710309151373323e-06, + "loss": 0.3575, + "step": 17072 + }, + { + "epoch": 6.3, + "learning_rate": 2.370993964773987e-06, + "loss": 0.514, + "step": 17073 + }, + { + "epoch": 6.31, + "learning_rate": 2.3709570144106415e-06, + "loss": 0.5883, + "step": 17074 + }, + { + "epoch": 6.31, + "learning_rate": 2.3709200640472966e-06, + "loss": 0.3701, + "step": 17075 + }, + { + "epoch": 6.31, + "learning_rate": 2.3708831136839512e-06, + "loss": 0.4854, + "step": 17076 + }, + { + "epoch": 6.31, + "learning_rate": 2.370846163320606e-06, + "loss": 0.4448, + "step": 17077 + }, + { + "epoch": 6.31, + "learning_rate": 2.370809212957261e-06, + "loss": 0.3843, + "step": 17078 + }, + { + "epoch": 6.31, + "learning_rate": 2.3707722625939156e-06, + "loss": 0.392, + "step": 17079 + }, + { + "epoch": 6.31, + "learning_rate": 2.37073531223057e-06, + "loss": 0.3203, + "step": 17080 + }, + { + "epoch": 6.31, + "learning_rate": 2.370698361867225e-06, + "loss": 0.5179, + "step": 17081 + }, + { + "epoch": 6.31, + "learning_rate": 2.37066141150388e-06, + "loss": 0.4382, + "step": 17082 + }, + { + "epoch": 6.31, + "learning_rate": 2.3706244611405345e-06, + "loss": 0.363, + "step": 17083 + }, + { + "epoch": 6.31, + "learning_rate": 2.3705875107771896e-06, + "loss": 0.4787, + "step": 17084 + }, + { + "epoch": 6.31, + "learning_rate": 2.3705505604138442e-06, + "loss": 0.3569, + "step": 17085 + }, + { + "epoch": 6.31, + "learning_rate": 2.370513610050499e-06, + "loss": 0.4189, + "step": 17086 + }, + { + "epoch": 6.31, + "learning_rate": 2.3704766596871535e-06, + "loss": 0.4909, + "step": 17087 + }, + { + "epoch": 6.31, + "learning_rate": 2.370439709323808e-06, + "loss": 0.4996, + "step": 17088 + }, + { + "epoch": 6.31, + "learning_rate": 2.370402758960463e-06, + "loss": 0.4899, + "step": 17089 + }, + { + "epoch": 6.31, + "learning_rate": 2.370365808597118e-06, + "loss": 0.4075, + "step": 17090 + }, + { + "epoch": 6.31, + "learning_rate": 2.370328858233773e-06, + "loss": 0.4191, + "step": 17091 + }, + { + "epoch": 6.31, + "learning_rate": 2.3702919078704275e-06, + "loss": 0.4727, + "step": 17092 + }, + { + "epoch": 6.31, + "learning_rate": 2.370254957507082e-06, + "loss": 0.5382, + "step": 17093 + }, + { + "epoch": 6.31, + "learning_rate": 2.370218007143737e-06, + "loss": 0.4802, + "step": 17094 + }, + { + "epoch": 6.31, + "learning_rate": 2.370181056780392e-06, + "loss": 0.3581, + "step": 17095 + }, + { + "epoch": 6.31, + "learning_rate": 2.3701441064170465e-06, + "loss": 0.4478, + "step": 17096 + }, + { + "epoch": 6.31, + "learning_rate": 2.370107156053701e-06, + "loss": 0.3775, + "step": 17097 + }, + { + "epoch": 6.31, + "learning_rate": 2.370070205690356e-06, + "loss": 0.466, + "step": 17098 + }, + { + "epoch": 6.31, + "learning_rate": 2.370033255327011e-06, + "loss": 0.4728, + "step": 17099 + }, + { + "epoch": 6.31, + "learning_rate": 2.3699963049636655e-06, + "loss": 0.4562, + "step": 17100 + }, + { + "epoch": 6.31, + "learning_rate": 2.36995935460032e-06, + "loss": 0.4855, + "step": 17101 + }, + { + "epoch": 6.32, + "learning_rate": 2.369922404236975e-06, + "loss": 0.5026, + "step": 17102 + }, + { + "epoch": 6.32, + "learning_rate": 2.36988545387363e-06, + "loss": 0.5286, + "step": 17103 + }, + { + "epoch": 6.32, + "learning_rate": 2.369848503510285e-06, + "loss": 0.5693, + "step": 17104 + }, + { + "epoch": 6.32, + "learning_rate": 2.3698115531469395e-06, + "loss": 0.576, + "step": 17105 + }, + { + "epoch": 6.32, + "learning_rate": 2.369774602783594e-06, + "loss": 0.5801, + "step": 17106 + }, + { + "epoch": 6.32, + "learning_rate": 2.3697376524202488e-06, + "loss": 0.4693, + "step": 17107 + }, + { + "epoch": 6.32, + "learning_rate": 2.3697007020569034e-06, + "loss": 0.4962, + "step": 17108 + }, + { + "epoch": 6.32, + "learning_rate": 2.3696637516935585e-06, + "loss": 0.358, + "step": 17109 + }, + { + "epoch": 6.32, + "learning_rate": 2.369626801330213e-06, + "loss": 0.343, + "step": 17110 + }, + { + "epoch": 6.32, + "learning_rate": 2.369589850966868e-06, + "loss": 0.497, + "step": 17111 + }, + { + "epoch": 6.32, + "learning_rate": 2.3695529006035224e-06, + "loss": 0.3522, + "step": 17112 + }, + { + "epoch": 6.32, + "learning_rate": 2.3695159502401774e-06, + "loss": 0.4418, + "step": 17113 + }, + { + "epoch": 6.32, + "learning_rate": 2.369478999876832e-06, + "loss": 0.4638, + "step": 17114 + }, + { + "epoch": 6.32, + "learning_rate": 2.369442049513487e-06, + "loss": 0.4431, + "step": 17115 + }, + { + "epoch": 6.32, + "learning_rate": 2.3694050991501418e-06, + "loss": 0.5612, + "step": 17116 + }, + { + "epoch": 6.32, + "learning_rate": 2.3693681487867964e-06, + "loss": 0.3726, + "step": 17117 + }, + { + "epoch": 6.32, + "learning_rate": 2.369331198423451e-06, + "loss": 0.4272, + "step": 17118 + }, + { + "epoch": 6.32, + "learning_rate": 2.3692942480601057e-06, + "loss": 0.3163, + "step": 17119 + }, + { + "epoch": 6.32, + "learning_rate": 2.3692572976967607e-06, + "loss": 0.4776, + "step": 17120 + }, + { + "epoch": 6.32, + "learning_rate": 2.3692203473334154e-06, + "loss": 0.4794, + "step": 17121 + }, + { + "epoch": 6.32, + "learning_rate": 2.3691833969700704e-06, + "loss": 0.5092, + "step": 17122 + }, + { + "epoch": 6.32, + "learning_rate": 2.369146446606725e-06, + "loss": 0.4308, + "step": 17123 + }, + { + "epoch": 6.32, + "learning_rate": 2.3691094962433797e-06, + "loss": 0.4159, + "step": 17124 + }, + { + "epoch": 6.32, + "learning_rate": 2.3690725458800343e-06, + "loss": 0.3587, + "step": 17125 + }, + { + "epoch": 6.32, + "learning_rate": 2.3690355955166894e-06, + "loss": 0.4657, + "step": 17126 + }, + { + "epoch": 6.32, + "learning_rate": 2.368998645153344e-06, + "loss": 0.4378, + "step": 17127 + }, + { + "epoch": 6.32, + "learning_rate": 2.3689616947899987e-06, + "loss": 0.5007, + "step": 17128 + }, + { + "epoch": 6.33, + "learning_rate": 2.3689247444266537e-06, + "loss": 0.4273, + "step": 17129 + }, + { + "epoch": 6.33, + "learning_rate": 2.3688877940633084e-06, + "loss": 0.4587, + "step": 17130 + }, + { + "epoch": 6.33, + "learning_rate": 2.368850843699963e-06, + "loss": 0.5865, + "step": 17131 + }, + { + "epoch": 6.33, + "learning_rate": 2.3688138933366177e-06, + "loss": 0.4945, + "step": 17132 + }, + { + "epoch": 6.33, + "learning_rate": 2.3687769429732727e-06, + "loss": 0.4225, + "step": 17133 + }, + { + "epoch": 6.33, + "learning_rate": 2.3687399926099273e-06, + "loss": 0.4754, + "step": 17134 + }, + { + "epoch": 6.33, + "learning_rate": 2.3687030422465824e-06, + "loss": 0.4482, + "step": 17135 + }, + { + "epoch": 6.33, + "learning_rate": 2.368666091883237e-06, + "loss": 0.5304, + "step": 17136 + }, + { + "epoch": 6.33, + "learning_rate": 2.3686291415198917e-06, + "loss": 0.374, + "step": 17137 + }, + { + "epoch": 6.33, + "learning_rate": 2.3685921911565463e-06, + "loss": 0.4633, + "step": 17138 + }, + { + "epoch": 6.33, + "learning_rate": 2.368555240793201e-06, + "loss": 0.5865, + "step": 17139 + }, + { + "epoch": 6.33, + "learning_rate": 2.368518290429856e-06, + "loss": 0.4444, + "step": 17140 + }, + { + "epoch": 6.33, + "learning_rate": 2.3684813400665106e-06, + "loss": 0.4452, + "step": 17141 + }, + { + "epoch": 6.33, + "learning_rate": 2.3684443897031657e-06, + "loss": 0.4704, + "step": 17142 + }, + { + "epoch": 6.33, + "learning_rate": 2.3684074393398203e-06, + "loss": 0.3991, + "step": 17143 + }, + { + "epoch": 6.33, + "learning_rate": 2.368370488976475e-06, + "loss": 0.447, + "step": 17144 + }, + { + "epoch": 6.33, + "learning_rate": 2.3683335386131296e-06, + "loss": 0.3633, + "step": 17145 + }, + { + "epoch": 6.33, + "learning_rate": 2.3682965882497847e-06, + "loss": 0.4263, + "step": 17146 + }, + { + "epoch": 6.33, + "learning_rate": 2.3682596378864393e-06, + "loss": 0.4304, + "step": 17147 + }, + { + "epoch": 6.33, + "learning_rate": 2.368222687523094e-06, + "loss": 0.4191, + "step": 17148 + }, + { + "epoch": 6.33, + "learning_rate": 2.368185737159749e-06, + "loss": 0.4586, + "step": 17149 + }, + { + "epoch": 6.33, + "learning_rate": 2.3681487867964036e-06, + "loss": 0.4874, + "step": 17150 + }, + { + "epoch": 6.33, + "learning_rate": 2.3681118364330583e-06, + "loss": 0.3726, + "step": 17151 + }, + { + "epoch": 6.33, + "learning_rate": 2.368074886069713e-06, + "loss": 0.4358, + "step": 17152 + }, + { + "epoch": 6.33, + "learning_rate": 2.368037935706368e-06, + "loss": 0.4278, + "step": 17153 + }, + { + "epoch": 6.33, + "learning_rate": 2.3680009853430226e-06, + "loss": 0.5109, + "step": 17154 + }, + { + "epoch": 6.33, + "learning_rate": 2.3679640349796777e-06, + "loss": 0.3899, + "step": 17155 + }, + { + "epoch": 6.34, + "learning_rate": 2.3679270846163323e-06, + "loss": 0.3886, + "step": 17156 + }, + { + "epoch": 6.34, + "learning_rate": 2.3678901342529865e-06, + "loss": 0.5394, + "step": 17157 + }, + { + "epoch": 6.34, + "learning_rate": 2.3678531838896416e-06, + "loss": 0.3877, + "step": 17158 + }, + { + "epoch": 6.34, + "learning_rate": 2.3678162335262962e-06, + "loss": 0.4467, + "step": 17159 + }, + { + "epoch": 6.34, + "learning_rate": 2.3677792831629513e-06, + "loss": 0.4028, + "step": 17160 + }, + { + "epoch": 6.34, + "learning_rate": 2.367742332799606e-06, + "loss": 0.501, + "step": 17161 + }, + { + "epoch": 6.34, + "learning_rate": 2.367705382436261e-06, + "loss": 0.4177, + "step": 17162 + }, + { + "epoch": 6.34, + "learning_rate": 2.367668432072915e-06, + "loss": 0.5122, + "step": 17163 + }, + { + "epoch": 6.34, + "learning_rate": 2.3676314817095703e-06, + "loss": 0.4505, + "step": 17164 + }, + { + "epoch": 6.34, + "learning_rate": 2.367594531346225e-06, + "loss": 0.4129, + "step": 17165 + }, + { + "epoch": 6.34, + "learning_rate": 2.36755758098288e-06, + "loss": 0.4013, + "step": 17166 + }, + { + "epoch": 6.34, + "learning_rate": 2.3675206306195346e-06, + "loss": 0.4252, + "step": 17167 + }, + { + "epoch": 6.34, + "learning_rate": 2.3674836802561892e-06, + "loss": 0.416, + "step": 17168 + }, + { + "epoch": 6.34, + "learning_rate": 2.367446729892844e-06, + "loss": 0.511, + "step": 17169 + }, + { + "epoch": 6.34, + "learning_rate": 2.3674097795294985e-06, + "loss": 0.5164, + "step": 17170 + }, + { + "epoch": 6.34, + "learning_rate": 2.3673728291661536e-06, + "loss": 0.5555, + "step": 17171 + }, + { + "epoch": 6.34, + "learning_rate": 2.367335878802808e-06, + "loss": 0.433, + "step": 17172 + }, + { + "epoch": 6.34, + "learning_rate": 2.3672989284394632e-06, + "loss": 0.3903, + "step": 17173 + }, + { + "epoch": 6.34, + "learning_rate": 2.367261978076118e-06, + "loss": 0.4403, + "step": 17174 + }, + { + "epoch": 6.34, + "learning_rate": 2.3672250277127725e-06, + "loss": 0.474, + "step": 17175 + }, + { + "epoch": 6.34, + "learning_rate": 2.367188077349427e-06, + "loss": 0.4908, + "step": 17176 + }, + { + "epoch": 6.34, + "learning_rate": 2.367151126986082e-06, + "loss": 0.4444, + "step": 17177 + }, + { + "epoch": 6.34, + "learning_rate": 2.367114176622737e-06, + "loss": 0.4935, + "step": 17178 + }, + { + "epoch": 6.34, + "learning_rate": 2.3670772262593915e-06, + "loss": 0.3627, + "step": 17179 + }, + { + "epoch": 6.34, + "learning_rate": 2.3670402758960466e-06, + "loss": 0.4781, + "step": 17180 + }, + { + "epoch": 6.34, + "learning_rate": 2.367003325532701e-06, + "loss": 0.3962, + "step": 17181 + }, + { + "epoch": 6.34, + "learning_rate": 2.366966375169356e-06, + "loss": 0.5065, + "step": 17182 + }, + { + "epoch": 6.35, + "learning_rate": 2.3669294248060105e-06, + "loss": 0.3374, + "step": 17183 + }, + { + "epoch": 6.35, + "learning_rate": 2.3668924744426655e-06, + "loss": 0.4492, + "step": 17184 + }, + { + "epoch": 6.35, + "learning_rate": 2.36685552407932e-06, + "loss": 0.4947, + "step": 17185 + }, + { + "epoch": 6.35, + "learning_rate": 2.3668185737159752e-06, + "loss": 0.4576, + "step": 17186 + }, + { + "epoch": 6.35, + "learning_rate": 2.36678162335263e-06, + "loss": 0.428, + "step": 17187 + }, + { + "epoch": 6.35, + "learning_rate": 2.3667446729892845e-06, + "loss": 0.4416, + "step": 17188 + }, + { + "epoch": 6.35, + "learning_rate": 2.366707722625939e-06, + "loss": 0.4173, + "step": 17189 + }, + { + "epoch": 6.35, + "learning_rate": 2.3666707722625938e-06, + "loss": 0.5469, + "step": 17190 + }, + { + "epoch": 6.35, + "learning_rate": 2.366633821899249e-06, + "loss": 0.4885, + "step": 17191 + }, + { + "epoch": 6.35, + "learning_rate": 2.3665968715359035e-06, + "loss": 0.4034, + "step": 17192 + }, + { + "epoch": 6.35, + "learning_rate": 2.3665599211725585e-06, + "loss": 0.4112, + "step": 17193 + }, + { + "epoch": 6.35, + "learning_rate": 2.366522970809213e-06, + "loss": 0.4385, + "step": 17194 + }, + { + "epoch": 6.35, + "learning_rate": 2.366486020445868e-06, + "loss": 0.5121, + "step": 17195 + }, + { + "epoch": 6.35, + "learning_rate": 2.3664490700825224e-06, + "loss": 0.5258, + "step": 17196 + }, + { + "epoch": 6.35, + "learning_rate": 2.366412119719177e-06, + "loss": 0.4797, + "step": 17197 + }, + { + "epoch": 6.35, + "learning_rate": 2.366375169355832e-06, + "loss": 0.3895, + "step": 17198 + }, + { + "epoch": 6.35, + "learning_rate": 2.3663382189924868e-06, + "loss": 0.5161, + "step": 17199 + }, + { + "epoch": 6.35, + "learning_rate": 2.366301268629142e-06, + "loss": 0.5047, + "step": 17200 + }, + { + "epoch": 6.35, + "learning_rate": 2.3662643182657965e-06, + "loss": 0.4755, + "step": 17201 + }, + { + "epoch": 6.35, + "learning_rate": 2.366227367902451e-06, + "loss": 0.4288, + "step": 17202 + }, + { + "epoch": 6.35, + "learning_rate": 2.3661904175391057e-06, + "loss": 0.4451, + "step": 17203 + }, + { + "epoch": 6.35, + "learning_rate": 2.366153467175761e-06, + "loss": 0.5266, + "step": 17204 + }, + { + "epoch": 6.35, + "learning_rate": 2.3661165168124154e-06, + "loss": 0.381, + "step": 17205 + }, + { + "epoch": 6.35, + "learning_rate": 2.3660795664490705e-06, + "loss": 0.445, + "step": 17206 + }, + { + "epoch": 6.35, + "learning_rate": 2.366042616085725e-06, + "loss": 0.4528, + "step": 17207 + }, + { + "epoch": 6.35, + "learning_rate": 2.3660056657223793e-06, + "loss": 0.4589, + "step": 17208 + }, + { + "epoch": 6.35, + "learning_rate": 2.3659687153590344e-06, + "loss": 0.3797, + "step": 17209 + }, + { + "epoch": 6.36, + "learning_rate": 2.365931764995689e-06, + "loss": 0.4608, + "step": 17210 + }, + { + "epoch": 6.36, + "learning_rate": 2.365894814632344e-06, + "loss": 0.4615, + "step": 17211 + }, + { + "epoch": 6.36, + "learning_rate": 2.3658578642689987e-06, + "loss": 0.3681, + "step": 17212 + }, + { + "epoch": 6.36, + "learning_rate": 2.365820913905654e-06, + "loss": 0.5138, + "step": 17213 + }, + { + "epoch": 6.36, + "learning_rate": 2.365783963542308e-06, + "loss": 0.3266, + "step": 17214 + }, + { + "epoch": 6.36, + "learning_rate": 2.365747013178963e-06, + "loss": 0.3729, + "step": 17215 + }, + { + "epoch": 6.36, + "learning_rate": 2.3657100628156177e-06, + "loss": 0.4556, + "step": 17216 + }, + { + "epoch": 6.36, + "learning_rate": 2.3656731124522723e-06, + "loss": 0.4229, + "step": 17217 + }, + { + "epoch": 6.36, + "learning_rate": 2.3656361620889274e-06, + "loss": 0.4987, + "step": 17218 + }, + { + "epoch": 6.36, + "learning_rate": 2.365599211725582e-06, + "loss": 0.3615, + "step": 17219 + }, + { + "epoch": 6.36, + "learning_rate": 2.3655622613622367e-06, + "loss": 0.4285, + "step": 17220 + }, + { + "epoch": 6.36, + "learning_rate": 2.3655253109988913e-06, + "loss": 0.5653, + "step": 17221 + }, + { + "epoch": 6.36, + "learning_rate": 2.3654883606355464e-06, + "loss": 0.4558, + "step": 17222 + }, + { + "epoch": 6.36, + "learning_rate": 2.365451410272201e-06, + "loss": 0.3836, + "step": 17223 + }, + { + "epoch": 6.36, + "learning_rate": 2.365414459908856e-06, + "loss": 0.416, + "step": 17224 + }, + { + "epoch": 6.36, + "learning_rate": 2.3653775095455107e-06, + "loss": 0.494, + "step": 17225 + }, + { + "epoch": 6.36, + "learning_rate": 2.3653405591821653e-06, + "loss": 0.534, + "step": 17226 + }, + { + "epoch": 6.36, + "learning_rate": 2.36530360881882e-06, + "loss": 0.3696, + "step": 17227 + }, + { + "epoch": 6.36, + "learning_rate": 2.3652666584554746e-06, + "loss": 0.4754, + "step": 17228 + }, + { + "epoch": 6.36, + "learning_rate": 2.3652297080921297e-06, + "loss": 0.5882, + "step": 17229 + }, + { + "epoch": 6.36, + "learning_rate": 2.3651927577287843e-06, + "loss": 0.5235, + "step": 17230 + }, + { + "epoch": 6.36, + "learning_rate": 2.3651558073654394e-06, + "loss": 0.5553, + "step": 17231 + }, + { + "epoch": 6.36, + "learning_rate": 2.365118857002094e-06, + "loss": 0.3708, + "step": 17232 + }, + { + "epoch": 6.36, + "learning_rate": 2.3650819066387486e-06, + "loss": 0.4726, + "step": 17233 + }, + { + "epoch": 6.36, + "learning_rate": 2.3650449562754033e-06, + "loss": 0.4144, + "step": 17234 + }, + { + "epoch": 6.36, + "learning_rate": 2.3650080059120583e-06, + "loss": 0.4879, + "step": 17235 + }, + { + "epoch": 6.36, + "learning_rate": 2.364971055548713e-06, + "loss": 0.4268, + "step": 17236 + }, + { + "epoch": 6.37, + "learning_rate": 2.3649341051853676e-06, + "loss": 0.5087, + "step": 17237 + }, + { + "epoch": 6.37, + "learning_rate": 2.3648971548220227e-06, + "loss": 0.5997, + "step": 17238 + }, + { + "epoch": 6.37, + "learning_rate": 2.3648602044586773e-06, + "loss": 0.5133, + "step": 17239 + }, + { + "epoch": 6.37, + "learning_rate": 2.364823254095332e-06, + "loss": 0.5723, + "step": 17240 + }, + { + "epoch": 6.37, + "learning_rate": 2.3647863037319866e-06, + "loss": 0.5305, + "step": 17241 + }, + { + "epoch": 6.37, + "learning_rate": 2.3647493533686416e-06, + "loss": 0.4287, + "step": 17242 + }, + { + "epoch": 6.37, + "learning_rate": 2.3647124030052963e-06, + "loss": 0.5252, + "step": 17243 + }, + { + "epoch": 6.37, + "learning_rate": 2.3646754526419513e-06, + "loss": 0.4681, + "step": 17244 + }, + { + "epoch": 6.37, + "learning_rate": 2.364638502278606e-06, + "loss": 0.5071, + "step": 17245 + }, + { + "epoch": 6.37, + "learning_rate": 2.3646015519152606e-06, + "loss": 0.4079, + "step": 17246 + }, + { + "epoch": 6.37, + "learning_rate": 2.3645646015519152e-06, + "loss": 0.5081, + "step": 17247 + }, + { + "epoch": 6.37, + "learning_rate": 2.36452765118857e-06, + "loss": 0.3898, + "step": 17248 + }, + { + "epoch": 6.37, + "learning_rate": 2.364490700825225e-06, + "loss": 0.4064, + "step": 17249 + }, + { + "epoch": 6.37, + "learning_rate": 2.3644537504618796e-06, + "loss": 0.4308, + "step": 17250 + }, + { + "epoch": 6.37, + "learning_rate": 2.3644168000985346e-06, + "loss": 0.5177, + "step": 17251 + }, + { + "epoch": 6.37, + "learning_rate": 2.3643798497351893e-06, + "loss": 0.567, + "step": 17252 + }, + { + "epoch": 6.37, + "learning_rate": 2.364342899371844e-06, + "loss": 0.4583, + "step": 17253 + }, + { + "epoch": 6.37, + "learning_rate": 2.3643059490084985e-06, + "loss": 0.3846, + "step": 17254 + }, + { + "epoch": 6.37, + "learning_rate": 2.3642689986451536e-06, + "loss": 0.3891, + "step": 17255 + }, + { + "epoch": 6.37, + "learning_rate": 2.3642320482818082e-06, + "loss": 0.3977, + "step": 17256 + }, + { + "epoch": 6.37, + "learning_rate": 2.364195097918463e-06, + "loss": 0.4708, + "step": 17257 + }, + { + "epoch": 6.37, + "learning_rate": 2.3641581475551175e-06, + "loss": 0.5057, + "step": 17258 + }, + { + "epoch": 6.37, + "learning_rate": 2.364121197191772e-06, + "loss": 0.4804, + "step": 17259 + }, + { + "epoch": 6.37, + "learning_rate": 2.364084246828427e-06, + "loss": 0.5107, + "step": 17260 + }, + { + "epoch": 6.37, + "learning_rate": 2.364047296465082e-06, + "loss": 0.4559, + "step": 17261 + }, + { + "epoch": 6.37, + "learning_rate": 2.364010346101737e-06, + "loss": 0.4252, + "step": 17262 + }, + { + "epoch": 6.37, + "learning_rate": 2.3639733957383915e-06, + "loss": 0.4535, + "step": 17263 + }, + { + "epoch": 6.38, + "learning_rate": 2.363936445375046e-06, + "loss": 0.3758, + "step": 17264 + }, + { + "epoch": 6.38, + "learning_rate": 2.363899495011701e-06, + "loss": 0.4518, + "step": 17265 + }, + { + "epoch": 6.38, + "learning_rate": 2.3638625446483554e-06, + "loss": 0.376, + "step": 17266 + }, + { + "epoch": 6.38, + "learning_rate": 2.3638255942850105e-06, + "loss": 0.3458, + "step": 17267 + }, + { + "epoch": 6.38, + "learning_rate": 2.363788643921665e-06, + "loss": 0.4159, + "step": 17268 + }, + { + "epoch": 6.38, + "learning_rate": 2.36375169355832e-06, + "loss": 0.4697, + "step": 17269 + }, + { + "epoch": 6.38, + "learning_rate": 2.363714743194975e-06, + "loss": 0.4141, + "step": 17270 + }, + { + "epoch": 6.38, + "learning_rate": 2.3636777928316295e-06, + "loss": 0.564, + "step": 17271 + }, + { + "epoch": 6.38, + "learning_rate": 2.363640842468284e-06, + "loss": 0.5781, + "step": 17272 + }, + { + "epoch": 6.38, + "learning_rate": 2.363603892104939e-06, + "loss": 0.3602, + "step": 17273 + }, + { + "epoch": 6.38, + "learning_rate": 2.363566941741594e-06, + "loss": 0.3741, + "step": 17274 + }, + { + "epoch": 6.38, + "learning_rate": 2.363529991378249e-06, + "loss": 0.4162, + "step": 17275 + }, + { + "epoch": 6.38, + "learning_rate": 2.3634930410149035e-06, + "loss": 0.419, + "step": 17276 + }, + { + "epoch": 6.38, + "learning_rate": 2.363456090651558e-06, + "loss": 0.4293, + "step": 17277 + }, + { + "epoch": 6.38, + "learning_rate": 2.3634191402882128e-06, + "loss": 0.4983, + "step": 17278 + }, + { + "epoch": 6.38, + "learning_rate": 2.3633821899248674e-06, + "loss": 0.4458, + "step": 17279 + }, + { + "epoch": 6.38, + "learning_rate": 2.3633452395615225e-06, + "loss": 0.4356, + "step": 17280 + }, + { + "epoch": 6.38, + "learning_rate": 2.363308289198177e-06, + "loss": 0.4999, + "step": 17281 + }, + { + "epoch": 6.38, + "learning_rate": 2.363271338834832e-06, + "loss": 0.4136, + "step": 17282 + }, + { + "epoch": 6.38, + "learning_rate": 2.363234388471487e-06, + "loss": 0.3971, + "step": 17283 + }, + { + "epoch": 6.38, + "learning_rate": 2.3631974381081414e-06, + "loss": 0.4834, + "step": 17284 + }, + { + "epoch": 6.38, + "learning_rate": 2.363160487744796e-06, + "loss": 0.4265, + "step": 17285 + }, + { + "epoch": 6.38, + "learning_rate": 2.3631235373814507e-06, + "loss": 0.5539, + "step": 17286 + }, + { + "epoch": 6.38, + "learning_rate": 2.3630865870181058e-06, + "loss": 0.4987, + "step": 17287 + }, + { + "epoch": 6.38, + "learning_rate": 2.3630496366547604e-06, + "loss": 0.4952, + "step": 17288 + }, + { + "epoch": 6.38, + "learning_rate": 2.3630126862914155e-06, + "loss": 0.427, + "step": 17289 + }, + { + "epoch": 6.38, + "learning_rate": 2.36297573592807e-06, + "loss": 0.3185, + "step": 17290 + }, + { + "epoch": 6.39, + "learning_rate": 2.3629387855647247e-06, + "loss": 0.4521, + "step": 17291 + }, + { + "epoch": 6.39, + "learning_rate": 2.3629018352013794e-06, + "loss": 0.5237, + "step": 17292 + }, + { + "epoch": 6.39, + "learning_rate": 2.3628648848380344e-06, + "loss": 0.4634, + "step": 17293 + }, + { + "epoch": 6.39, + "learning_rate": 2.362827934474689e-06, + "loss": 0.4207, + "step": 17294 + }, + { + "epoch": 6.39, + "learning_rate": 2.362790984111344e-06, + "loss": 0.5516, + "step": 17295 + }, + { + "epoch": 6.39, + "learning_rate": 2.3627540337479988e-06, + "loss": 0.4673, + "step": 17296 + }, + { + "epoch": 6.39, + "learning_rate": 2.362717083384653e-06, + "loss": 0.3478, + "step": 17297 + }, + { + "epoch": 6.39, + "learning_rate": 2.362680133021308e-06, + "loss": 0.456, + "step": 17298 + }, + { + "epoch": 6.39, + "learning_rate": 2.3626431826579627e-06, + "loss": 0.5258, + "step": 17299 + }, + { + "epoch": 6.39, + "learning_rate": 2.3626062322946177e-06, + "loss": 0.4054, + "step": 17300 + }, + { + "epoch": 6.39, + "learning_rate": 2.3625692819312724e-06, + "loss": 0.295, + "step": 17301 + }, + { + "epoch": 6.39, + "learning_rate": 2.3625323315679274e-06, + "loss": 0.5195, + "step": 17302 + }, + { + "epoch": 6.39, + "learning_rate": 2.3624953812045817e-06, + "loss": 0.368, + "step": 17303 + }, + { + "epoch": 6.39, + "learning_rate": 2.3624584308412367e-06, + "loss": 0.4798, + "step": 17304 + }, + { + "epoch": 6.39, + "learning_rate": 2.3624214804778914e-06, + "loss": 0.4101, + "step": 17305 + }, + { + "epoch": 6.39, + "learning_rate": 2.362384530114546e-06, + "loss": 0.4708, + "step": 17306 + }, + { + "epoch": 6.39, + "learning_rate": 2.362347579751201e-06, + "loss": 0.3608, + "step": 17307 + }, + { + "epoch": 6.39, + "learning_rate": 2.3623106293878557e-06, + "loss": 0.4562, + "step": 17308 + }, + { + "epoch": 6.39, + "learning_rate": 2.3622736790245103e-06, + "loss": 0.429, + "step": 17309 + }, + { + "epoch": 6.39, + "learning_rate": 2.362236728661165e-06, + "loss": 0.5311, + "step": 17310 + }, + { + "epoch": 6.39, + "learning_rate": 2.36219977829782e-06, + "loss": 0.3417, + "step": 17311 + }, + { + "epoch": 6.39, + "learning_rate": 2.3621628279344747e-06, + "loss": 0.5126, + "step": 17312 + }, + { + "epoch": 6.39, + "learning_rate": 2.3621258775711297e-06, + "loss": 0.3664, + "step": 17313 + }, + { + "epoch": 6.39, + "learning_rate": 2.3620889272077844e-06, + "loss": 0.4605, + "step": 17314 + }, + { + "epoch": 6.39, + "learning_rate": 2.362051976844439e-06, + "loss": 0.4428, + "step": 17315 + }, + { + "epoch": 6.39, + "learning_rate": 2.3620150264810936e-06, + "loss": 0.4582, + "step": 17316 + }, + { + "epoch": 6.39, + "learning_rate": 2.3619780761177483e-06, + "loss": 0.3852, + "step": 17317 + }, + { + "epoch": 6.4, + "learning_rate": 2.3619411257544033e-06, + "loss": 0.4645, + "step": 17318 + }, + { + "epoch": 6.4, + "learning_rate": 2.361904175391058e-06, + "loss": 0.3697, + "step": 17319 + }, + { + "epoch": 6.4, + "learning_rate": 2.361867225027713e-06, + "loss": 0.3779, + "step": 17320 + }, + { + "epoch": 6.4, + "learning_rate": 2.3618302746643677e-06, + "loss": 0.296, + "step": 17321 + }, + { + "epoch": 6.4, + "learning_rate": 2.3617933243010223e-06, + "loss": 0.4569, + "step": 17322 + }, + { + "epoch": 6.4, + "learning_rate": 2.361756373937677e-06, + "loss": 0.5395, + "step": 17323 + }, + { + "epoch": 6.4, + "learning_rate": 2.361719423574332e-06, + "loss": 0.4089, + "step": 17324 + }, + { + "epoch": 6.4, + "learning_rate": 2.3616824732109866e-06, + "loss": 0.4155, + "step": 17325 + }, + { + "epoch": 6.4, + "learning_rate": 2.3616455228476413e-06, + "loss": 0.4715, + "step": 17326 + }, + { + "epoch": 6.4, + "learning_rate": 2.3616085724842963e-06, + "loss": 0.4553, + "step": 17327 + }, + { + "epoch": 6.4, + "learning_rate": 2.361571622120951e-06, + "loss": 0.5507, + "step": 17328 + }, + { + "epoch": 6.4, + "learning_rate": 2.3615346717576056e-06, + "loss": 0.3808, + "step": 17329 + }, + { + "epoch": 6.4, + "learning_rate": 2.3614977213942602e-06, + "loss": 0.4331, + "step": 17330 + }, + { + "epoch": 6.4, + "learning_rate": 2.3614607710309153e-06, + "loss": 0.4656, + "step": 17331 + }, + { + "epoch": 6.4, + "learning_rate": 2.36142382066757e-06, + "loss": 0.3928, + "step": 17332 + }, + { + "epoch": 6.4, + "learning_rate": 2.361386870304225e-06, + "loss": 0.3288, + "step": 17333 + }, + { + "epoch": 6.4, + "learning_rate": 2.3613499199408796e-06, + "loss": 0.5049, + "step": 17334 + }, + { + "epoch": 6.4, + "learning_rate": 2.3613129695775343e-06, + "loss": 0.435, + "step": 17335 + }, + { + "epoch": 6.4, + "learning_rate": 2.361276019214189e-06, + "loss": 0.4015, + "step": 17336 + }, + { + "epoch": 6.4, + "learning_rate": 2.3612390688508435e-06, + "loss": 0.4827, + "step": 17337 + }, + { + "epoch": 6.4, + "learning_rate": 2.3612021184874986e-06, + "loss": 0.4553, + "step": 17338 + }, + { + "epoch": 6.4, + "learning_rate": 2.3611651681241532e-06, + "loss": 0.3919, + "step": 17339 + }, + { + "epoch": 6.4, + "learning_rate": 2.3611282177608083e-06, + "loss": 0.5057, + "step": 17340 + }, + { + "epoch": 6.4, + "learning_rate": 2.361091267397463e-06, + "loss": 0.4816, + "step": 17341 + }, + { + "epoch": 6.4, + "learning_rate": 2.3610543170341176e-06, + "loss": 0.4613, + "step": 17342 + }, + { + "epoch": 6.4, + "learning_rate": 2.361017366670772e-06, + "loss": 0.4434, + "step": 17343 + }, + { + "epoch": 6.4, + "learning_rate": 2.3609804163074273e-06, + "loss": 0.5066, + "step": 17344 + }, + { + "epoch": 6.41, + "learning_rate": 2.360943465944082e-06, + "loss": 0.4206, + "step": 17345 + }, + { + "epoch": 6.41, + "learning_rate": 2.3609065155807365e-06, + "loss": 0.4373, + "step": 17346 + }, + { + "epoch": 6.41, + "learning_rate": 2.3608695652173916e-06, + "loss": 0.3951, + "step": 17347 + }, + { + "epoch": 6.41, + "learning_rate": 2.360832614854046e-06, + "loss": 0.4481, + "step": 17348 + }, + { + "epoch": 6.41, + "learning_rate": 2.360795664490701e-06, + "loss": 0.5079, + "step": 17349 + }, + { + "epoch": 6.41, + "learning_rate": 2.3607587141273555e-06, + "loss": 0.4392, + "step": 17350 + }, + { + "epoch": 6.41, + "learning_rate": 2.3607217637640106e-06, + "loss": 0.6246, + "step": 17351 + }, + { + "epoch": 6.41, + "learning_rate": 2.360684813400665e-06, + "loss": 0.4344, + "step": 17352 + }, + { + "epoch": 6.41, + "learning_rate": 2.3606478630373203e-06, + "loss": 0.4363, + "step": 17353 + }, + { + "epoch": 6.41, + "learning_rate": 2.3606109126739745e-06, + "loss": 0.4357, + "step": 17354 + }, + { + "epoch": 6.41, + "learning_rate": 2.3605739623106295e-06, + "loss": 0.431, + "step": 17355 + }, + { + "epoch": 6.41, + "learning_rate": 2.360537011947284e-06, + "loss": 0.4118, + "step": 17356 + }, + { + "epoch": 6.41, + "learning_rate": 2.360500061583939e-06, + "loss": 0.5846, + "step": 17357 + }, + { + "epoch": 6.41, + "learning_rate": 2.360463111220594e-06, + "loss": 0.4344, + "step": 17358 + }, + { + "epoch": 6.41, + "learning_rate": 2.3604261608572485e-06, + "loss": 0.3654, + "step": 17359 + }, + { + "epoch": 6.41, + "learning_rate": 2.360389210493903e-06, + "loss": 0.5672, + "step": 17360 + }, + { + "epoch": 6.41, + "learning_rate": 2.3603522601305578e-06, + "loss": 0.426, + "step": 17361 + }, + { + "epoch": 6.41, + "learning_rate": 2.360315309767213e-06, + "loss": 0.5642, + "step": 17362 + }, + { + "epoch": 6.41, + "learning_rate": 2.3602783594038675e-06, + "loss": 0.4111, + "step": 17363 + }, + { + "epoch": 6.41, + "learning_rate": 2.3602414090405225e-06, + "loss": 0.5517, + "step": 17364 + }, + { + "epoch": 6.41, + "learning_rate": 2.360204458677177e-06, + "loss": 0.4178, + "step": 17365 + }, + { + "epoch": 6.41, + "learning_rate": 2.360167508313832e-06, + "loss": 0.3697, + "step": 17366 + }, + { + "epoch": 6.41, + "learning_rate": 2.3601305579504864e-06, + "loss": 0.451, + "step": 17367 + }, + { + "epoch": 6.41, + "learning_rate": 2.360093607587141e-06, + "loss": 0.4749, + "step": 17368 + }, + { + "epoch": 6.41, + "learning_rate": 2.360056657223796e-06, + "loss": 0.4839, + "step": 17369 + }, + { + "epoch": 6.41, + "learning_rate": 2.3600197068604508e-06, + "loss": 0.4593, + "step": 17370 + }, + { + "epoch": 6.41, + "learning_rate": 2.359982756497106e-06, + "loss": 0.3974, + "step": 17371 + }, + { + "epoch": 6.42, + "learning_rate": 2.3599458061337605e-06, + "loss": 0.4498, + "step": 17372 + }, + { + "epoch": 6.42, + "learning_rate": 2.359908855770415e-06, + "loss": 0.4929, + "step": 17373 + }, + { + "epoch": 6.42, + "learning_rate": 2.3598719054070697e-06, + "loss": 0.4235, + "step": 17374 + }, + { + "epoch": 6.42, + "learning_rate": 2.359834955043725e-06, + "loss": 0.4805, + "step": 17375 + }, + { + "epoch": 6.42, + "learning_rate": 2.3597980046803794e-06, + "loss": 0.3325, + "step": 17376 + }, + { + "epoch": 6.42, + "learning_rate": 2.359761054317034e-06, + "loss": 0.5871, + "step": 17377 + }, + { + "epoch": 6.42, + "learning_rate": 2.359724103953689e-06, + "loss": 0.4871, + "step": 17378 + }, + { + "epoch": 6.42, + "learning_rate": 2.3596871535903438e-06, + "loss": 0.4293, + "step": 17379 + }, + { + "epoch": 6.42, + "learning_rate": 2.3596502032269984e-06, + "loss": 0.4688, + "step": 17380 + }, + { + "epoch": 6.42, + "learning_rate": 2.359613252863653e-06, + "loss": 0.4555, + "step": 17381 + }, + { + "epoch": 6.42, + "learning_rate": 2.359576302500308e-06, + "loss": 0.4424, + "step": 17382 + }, + { + "epoch": 6.42, + "learning_rate": 2.3595393521369627e-06, + "loss": 0.3788, + "step": 17383 + }, + { + "epoch": 6.42, + "learning_rate": 2.359502401773618e-06, + "loss": 0.3658, + "step": 17384 + }, + { + "epoch": 6.42, + "learning_rate": 2.3594654514102724e-06, + "loss": 0.5593, + "step": 17385 + }, + { + "epoch": 6.42, + "learning_rate": 2.359428501046927e-06, + "loss": 0.3977, + "step": 17386 + }, + { + "epoch": 6.42, + "learning_rate": 2.3593915506835817e-06, + "loss": 0.4366, + "step": 17387 + }, + { + "epoch": 6.42, + "learning_rate": 2.3593546003202363e-06, + "loss": 0.4999, + "step": 17388 + }, + { + "epoch": 6.42, + "learning_rate": 2.3593176499568914e-06, + "loss": 0.382, + "step": 17389 + }, + { + "epoch": 6.42, + "learning_rate": 2.359280699593546e-06, + "loss": 0.3785, + "step": 17390 + }, + { + "epoch": 6.42, + "learning_rate": 2.359243749230201e-06, + "loss": 0.394, + "step": 17391 + }, + { + "epoch": 6.42, + "learning_rate": 2.3592067988668557e-06, + "loss": 0.6046, + "step": 17392 + }, + { + "epoch": 6.42, + "learning_rate": 2.3591698485035104e-06, + "loss": 0.4941, + "step": 17393 + }, + { + "epoch": 6.42, + "learning_rate": 2.359132898140165e-06, + "loss": 0.4837, + "step": 17394 + }, + { + "epoch": 6.42, + "learning_rate": 2.35909594777682e-06, + "loss": 0.4562, + "step": 17395 + }, + { + "epoch": 6.42, + "learning_rate": 2.3590589974134747e-06, + "loss": 0.6168, + "step": 17396 + }, + { + "epoch": 6.42, + "learning_rate": 2.3590220470501293e-06, + "loss": 0.415, + "step": 17397 + }, + { + "epoch": 6.42, + "learning_rate": 2.3589850966867844e-06, + "loss": 0.418, + "step": 17398 + }, + { + "epoch": 6.43, + "learning_rate": 2.3589481463234386e-06, + "loss": 0.4879, + "step": 17399 + }, + { + "epoch": 6.43, + "learning_rate": 2.3589111959600937e-06, + "loss": 0.3319, + "step": 17400 + }, + { + "epoch": 6.43, + "learning_rate": 2.3588742455967483e-06, + "loss": 0.4415, + "step": 17401 + }, + { + "epoch": 6.43, + "learning_rate": 2.3588372952334034e-06, + "loss": 0.3244, + "step": 17402 + }, + { + "epoch": 6.43, + "learning_rate": 2.358800344870058e-06, + "loss": 0.3647, + "step": 17403 + }, + { + "epoch": 6.43, + "learning_rate": 2.358763394506713e-06, + "loss": 0.369, + "step": 17404 + }, + { + "epoch": 6.43, + "learning_rate": 2.3587264441433673e-06, + "loss": 0.4274, + "step": 17405 + }, + { + "epoch": 6.43, + "learning_rate": 2.358689493780022e-06, + "loss": 0.3887, + "step": 17406 + }, + { + "epoch": 6.43, + "learning_rate": 2.358652543416677e-06, + "loss": 0.473, + "step": 17407 + }, + { + "epoch": 6.43, + "learning_rate": 2.3586155930533316e-06, + "loss": 0.3755, + "step": 17408 + }, + { + "epoch": 6.43, + "learning_rate": 2.3585786426899867e-06, + "loss": 0.4522, + "step": 17409 + }, + { + "epoch": 6.43, + "learning_rate": 2.3585416923266413e-06, + "loss": 0.415, + "step": 17410 + }, + { + "epoch": 6.43, + "learning_rate": 2.358504741963296e-06, + "loss": 0.6547, + "step": 17411 + }, + { + "epoch": 6.43, + "learning_rate": 2.3584677915999506e-06, + "loss": 0.4058, + "step": 17412 + }, + { + "epoch": 6.43, + "learning_rate": 2.3584308412366056e-06, + "loss": 0.4653, + "step": 17413 + }, + { + "epoch": 6.43, + "learning_rate": 2.3583938908732603e-06, + "loss": 0.4875, + "step": 17414 + }, + { + "epoch": 6.43, + "learning_rate": 2.3583569405099153e-06, + "loss": 0.3888, + "step": 17415 + }, + { + "epoch": 6.43, + "learning_rate": 2.35831999014657e-06, + "loss": 0.4314, + "step": 17416 + }, + { + "epoch": 6.43, + "learning_rate": 2.3582830397832246e-06, + "loss": 0.4186, + "step": 17417 + }, + { + "epoch": 6.43, + "learning_rate": 2.3582460894198792e-06, + "loss": 0.5535, + "step": 17418 + }, + { + "epoch": 6.43, + "learning_rate": 2.358209139056534e-06, + "loss": 0.5141, + "step": 17419 + }, + { + "epoch": 6.43, + "learning_rate": 2.358172188693189e-06, + "loss": 0.3959, + "step": 17420 + }, + { + "epoch": 6.43, + "learning_rate": 2.3581352383298436e-06, + "loss": 0.4102, + "step": 17421 + }, + { + "epoch": 6.43, + "learning_rate": 2.3580982879664986e-06, + "loss": 0.3736, + "step": 17422 + }, + { + "epoch": 6.43, + "learning_rate": 2.3580613376031533e-06, + "loss": 0.4598, + "step": 17423 + }, + { + "epoch": 6.43, + "learning_rate": 2.358024387239808e-06, + "loss": 0.4553, + "step": 17424 + }, + { + "epoch": 6.43, + "learning_rate": 2.3579874368764625e-06, + "loss": 0.4174, + "step": 17425 + }, + { + "epoch": 6.44, + "learning_rate": 2.357950486513117e-06, + "loss": 0.4306, + "step": 17426 + }, + { + "epoch": 6.44, + "learning_rate": 2.3579135361497722e-06, + "loss": 0.4407, + "step": 17427 + }, + { + "epoch": 6.44, + "learning_rate": 2.357876585786427e-06, + "loss": 0.4424, + "step": 17428 + }, + { + "epoch": 6.44, + "learning_rate": 2.357839635423082e-06, + "loss": 0.5103, + "step": 17429 + }, + { + "epoch": 6.44, + "learning_rate": 2.3578026850597366e-06, + "loss": 0.4603, + "step": 17430 + }, + { + "epoch": 6.44, + "learning_rate": 2.3577657346963912e-06, + "loss": 0.4251, + "step": 17431 + }, + { + "epoch": 6.44, + "learning_rate": 2.357728784333046e-06, + "loss": 0.466, + "step": 17432 + }, + { + "epoch": 6.44, + "learning_rate": 2.357691833969701e-06, + "loss": 0.4586, + "step": 17433 + }, + { + "epoch": 6.44, + "learning_rate": 2.3576548836063555e-06, + "loss": 0.4875, + "step": 17434 + }, + { + "epoch": 6.44, + "learning_rate": 2.3576179332430106e-06, + "loss": 0.5852, + "step": 17435 + }, + { + "epoch": 6.44, + "learning_rate": 2.3575809828796652e-06, + "loss": 0.4459, + "step": 17436 + }, + { + "epoch": 6.44, + "learning_rate": 2.35754403251632e-06, + "loss": 0.4264, + "step": 17437 + }, + { + "epoch": 6.44, + "learning_rate": 2.3575070821529745e-06, + "loss": 0.385, + "step": 17438 + }, + { + "epoch": 6.44, + "learning_rate": 2.357470131789629e-06, + "loss": 0.4022, + "step": 17439 + }, + { + "epoch": 6.44, + "learning_rate": 2.3574331814262842e-06, + "loss": 0.3062, + "step": 17440 + }, + { + "epoch": 6.44, + "learning_rate": 2.357396231062939e-06, + "loss": 0.4365, + "step": 17441 + }, + { + "epoch": 6.44, + "learning_rate": 2.357359280699594e-06, + "loss": 0.4585, + "step": 17442 + }, + { + "epoch": 6.44, + "learning_rate": 2.357322330336248e-06, + "loss": 0.4121, + "step": 17443 + }, + { + "epoch": 6.44, + "learning_rate": 2.357285379972903e-06, + "loss": 0.5037, + "step": 17444 + }, + { + "epoch": 6.44, + "learning_rate": 2.357248429609558e-06, + "loss": 0.5065, + "step": 17445 + }, + { + "epoch": 6.44, + "learning_rate": 2.3572114792462125e-06, + "loss": 0.3792, + "step": 17446 + }, + { + "epoch": 6.44, + "learning_rate": 2.3571745288828675e-06, + "loss": 0.4412, + "step": 17447 + }, + { + "epoch": 6.44, + "learning_rate": 2.357137578519522e-06, + "loss": 0.4588, + "step": 17448 + }, + { + "epoch": 6.44, + "learning_rate": 2.3571006281561768e-06, + "loss": 0.5166, + "step": 17449 + }, + { + "epoch": 6.44, + "learning_rate": 2.3570636777928314e-06, + "loss": 0.5622, + "step": 17450 + }, + { + "epoch": 6.44, + "learning_rate": 2.3570267274294865e-06, + "loss": 0.4134, + "step": 17451 + }, + { + "epoch": 6.44, + "learning_rate": 2.356989777066141e-06, + "loss": 0.4963, + "step": 17452 + }, + { + "epoch": 6.44, + "learning_rate": 2.356952826702796e-06, + "loss": 0.4415, + "step": 17453 + }, + { + "epoch": 6.45, + "learning_rate": 2.356915876339451e-06, + "loss": 0.4229, + "step": 17454 + }, + { + "epoch": 6.45, + "learning_rate": 2.3568789259761055e-06, + "loss": 0.366, + "step": 17455 + }, + { + "epoch": 6.45, + "learning_rate": 2.35684197561276e-06, + "loss": 0.4563, + "step": 17456 + }, + { + "epoch": 6.45, + "learning_rate": 2.3568050252494147e-06, + "loss": 0.4975, + "step": 17457 + }, + { + "epoch": 6.45, + "learning_rate": 2.3567680748860698e-06, + "loss": 0.4291, + "step": 17458 + }, + { + "epoch": 6.45, + "learning_rate": 2.3567311245227244e-06, + "loss": 0.5949, + "step": 17459 + }, + { + "epoch": 6.45, + "learning_rate": 2.3566941741593795e-06, + "loss": 0.4577, + "step": 17460 + }, + { + "epoch": 6.45, + "learning_rate": 2.356657223796034e-06, + "loss": 0.5409, + "step": 17461 + }, + { + "epoch": 6.45, + "learning_rate": 2.3566202734326888e-06, + "loss": 0.352, + "step": 17462 + }, + { + "epoch": 6.45, + "learning_rate": 2.3565833230693434e-06, + "loss": 0.6124, + "step": 17463 + }, + { + "epoch": 6.45, + "learning_rate": 2.3565463727059985e-06, + "loss": 0.2754, + "step": 17464 + }, + { + "epoch": 6.45, + "learning_rate": 2.356509422342653e-06, + "loss": 0.4582, + "step": 17465 + }, + { + "epoch": 6.45, + "learning_rate": 2.3564724719793077e-06, + "loss": 0.4015, + "step": 17466 + }, + { + "epoch": 6.45, + "learning_rate": 2.3564355216159628e-06, + "loss": 0.5935, + "step": 17467 + }, + { + "epoch": 6.45, + "learning_rate": 2.3563985712526174e-06, + "loss": 0.4032, + "step": 17468 + }, + { + "epoch": 6.45, + "learning_rate": 2.356361620889272e-06, + "loss": 0.3663, + "step": 17469 + }, + { + "epoch": 6.45, + "learning_rate": 2.3563246705259267e-06, + "loss": 0.441, + "step": 17470 + }, + { + "epoch": 6.45, + "learning_rate": 2.3562877201625818e-06, + "loss": 0.5242, + "step": 17471 + }, + { + "epoch": 6.45, + "learning_rate": 2.3562507697992364e-06, + "loss": 0.3399, + "step": 17472 + }, + { + "epoch": 6.45, + "learning_rate": 2.3562138194358914e-06, + "loss": 0.5175, + "step": 17473 + }, + { + "epoch": 6.45, + "learning_rate": 2.356176869072546e-06, + "loss": 0.4355, + "step": 17474 + }, + { + "epoch": 6.45, + "learning_rate": 2.3561399187092007e-06, + "loss": 0.392, + "step": 17475 + }, + { + "epoch": 6.45, + "learning_rate": 2.3561029683458554e-06, + "loss": 0.4313, + "step": 17476 + }, + { + "epoch": 6.45, + "learning_rate": 2.35606601798251e-06, + "loss": 0.5132, + "step": 17477 + }, + { + "epoch": 6.45, + "learning_rate": 2.356029067619165e-06, + "loss": 0.4815, + "step": 17478 + }, + { + "epoch": 6.45, + "learning_rate": 2.3559921172558197e-06, + "loss": 0.5559, + "step": 17479 + }, + { + "epoch": 6.45, + "learning_rate": 2.3559551668924748e-06, + "loss": 0.3135, + "step": 17480 + }, + { + "epoch": 6.46, + "learning_rate": 2.3559182165291294e-06, + "loss": 0.4511, + "step": 17481 + }, + { + "epoch": 6.46, + "learning_rate": 2.355881266165784e-06, + "loss": 0.4405, + "step": 17482 + }, + { + "epoch": 6.46, + "learning_rate": 2.3558443158024387e-06, + "loss": 0.4259, + "step": 17483 + }, + { + "epoch": 6.46, + "learning_rate": 2.3558073654390937e-06, + "loss": 0.4541, + "step": 17484 + }, + { + "epoch": 6.46, + "learning_rate": 2.3557704150757484e-06, + "loss": 0.3348, + "step": 17485 + }, + { + "epoch": 6.46, + "learning_rate": 2.355733464712403e-06, + "loss": 0.4575, + "step": 17486 + }, + { + "epoch": 6.46, + "learning_rate": 2.355696514349058e-06, + "loss": 0.4679, + "step": 17487 + }, + { + "epoch": 6.46, + "learning_rate": 2.3556595639857123e-06, + "loss": 0.4573, + "step": 17488 + }, + { + "epoch": 6.46, + "learning_rate": 2.3556226136223673e-06, + "loss": 0.5503, + "step": 17489 + }, + { + "epoch": 6.46, + "learning_rate": 2.355585663259022e-06, + "loss": 0.5116, + "step": 17490 + }, + { + "epoch": 6.46, + "learning_rate": 2.355548712895677e-06, + "loss": 0.5328, + "step": 17491 + }, + { + "epoch": 6.46, + "learning_rate": 2.3555117625323317e-06, + "loss": 0.4511, + "step": 17492 + }, + { + "epoch": 6.46, + "learning_rate": 2.3554748121689867e-06, + "loss": 0.4033, + "step": 17493 + }, + { + "epoch": 6.46, + "learning_rate": 2.355437861805641e-06, + "loss": 0.3338, + "step": 17494 + }, + { + "epoch": 6.46, + "learning_rate": 2.3554009114422956e-06, + "loss": 0.4839, + "step": 17495 + }, + { + "epoch": 6.46, + "learning_rate": 2.3553639610789506e-06, + "loss": 0.3393, + "step": 17496 + }, + { + "epoch": 6.46, + "learning_rate": 2.3553270107156053e-06, + "loss": 0.5338, + "step": 17497 + }, + { + "epoch": 6.46, + "learning_rate": 2.3552900603522603e-06, + "loss": 0.5139, + "step": 17498 + }, + { + "epoch": 6.46, + "learning_rate": 2.355253109988915e-06, + "loss": 0.535, + "step": 17499 + }, + { + "epoch": 6.46, + "learning_rate": 2.3552161596255696e-06, + "loss": 0.3383, + "step": 17500 + }, + { + "epoch": 6.46, + "learning_rate": 2.3551792092622242e-06, + "loss": 0.5511, + "step": 17501 + }, + { + "epoch": 6.46, + "learning_rate": 2.3551422588988793e-06, + "loss": 0.4088, + "step": 17502 + }, + { + "epoch": 6.46, + "learning_rate": 2.355105308535534e-06, + "loss": 0.4429, + "step": 17503 + }, + { + "epoch": 6.46, + "learning_rate": 2.355068358172189e-06, + "loss": 0.5694, + "step": 17504 + }, + { + "epoch": 6.46, + "learning_rate": 2.3550314078088436e-06, + "loss": 0.5164, + "step": 17505 + }, + { + "epoch": 6.46, + "learning_rate": 2.3549944574454983e-06, + "loss": 0.6025, + "step": 17506 + }, + { + "epoch": 6.46, + "learning_rate": 2.354957507082153e-06, + "loss": 0.4619, + "step": 17507 + }, + { + "epoch": 6.47, + "learning_rate": 2.3549205567188075e-06, + "loss": 0.3297, + "step": 17508 + }, + { + "epoch": 6.47, + "learning_rate": 2.3548836063554626e-06, + "loss": 0.4936, + "step": 17509 + }, + { + "epoch": 6.47, + "learning_rate": 2.3548466559921172e-06, + "loss": 0.5244, + "step": 17510 + }, + { + "epoch": 6.47, + "learning_rate": 2.3548097056287723e-06, + "loss": 0.5229, + "step": 17511 + }, + { + "epoch": 6.47, + "learning_rate": 2.354772755265427e-06, + "loss": 0.4314, + "step": 17512 + }, + { + "epoch": 6.47, + "learning_rate": 2.3547358049020816e-06, + "loss": 0.4567, + "step": 17513 + }, + { + "epoch": 6.47, + "learning_rate": 2.354698854538736e-06, + "loss": 0.4604, + "step": 17514 + }, + { + "epoch": 6.47, + "learning_rate": 2.354661904175391e-06, + "loss": 0.4792, + "step": 17515 + }, + { + "epoch": 6.47, + "learning_rate": 2.354624953812046e-06, + "loss": 0.4182, + "step": 17516 + }, + { + "epoch": 6.47, + "learning_rate": 2.3545880034487005e-06, + "loss": 0.5123, + "step": 17517 + }, + { + "epoch": 6.47, + "learning_rate": 2.3545510530853556e-06, + "loss": 0.3204, + "step": 17518 + }, + { + "epoch": 6.47, + "learning_rate": 2.3545141027220102e-06, + "loss": 0.4672, + "step": 17519 + }, + { + "epoch": 6.47, + "learning_rate": 2.354477152358665e-06, + "loss": 0.4057, + "step": 17520 + }, + { + "epoch": 6.47, + "learning_rate": 2.3544402019953195e-06, + "loss": 0.4091, + "step": 17521 + }, + { + "epoch": 6.47, + "learning_rate": 2.3544032516319746e-06, + "loss": 0.4865, + "step": 17522 + }, + { + "epoch": 6.47, + "learning_rate": 2.354366301268629e-06, + "loss": 0.4765, + "step": 17523 + }, + { + "epoch": 6.47, + "learning_rate": 2.3543293509052843e-06, + "loss": 0.4272, + "step": 17524 + }, + { + "epoch": 6.47, + "learning_rate": 2.354292400541939e-06, + "loss": 0.4246, + "step": 17525 + }, + { + "epoch": 6.47, + "learning_rate": 2.3542554501785935e-06, + "loss": 0.4714, + "step": 17526 + }, + { + "epoch": 6.47, + "learning_rate": 2.354218499815248e-06, + "loss": 0.4326, + "step": 17527 + }, + { + "epoch": 6.47, + "learning_rate": 2.354181549451903e-06, + "loss": 0.5659, + "step": 17528 + }, + { + "epoch": 6.47, + "learning_rate": 2.354144599088558e-06, + "loss": 0.4584, + "step": 17529 + }, + { + "epoch": 6.47, + "learning_rate": 2.3541076487252125e-06, + "loss": 0.5272, + "step": 17530 + }, + { + "epoch": 6.47, + "learning_rate": 2.3540706983618676e-06, + "loss": 0.4433, + "step": 17531 + }, + { + "epoch": 6.47, + "learning_rate": 2.354033747998522e-06, + "loss": 0.4713, + "step": 17532 + }, + { + "epoch": 6.47, + "learning_rate": 2.353996797635177e-06, + "loss": 0.4743, + "step": 17533 + }, + { + "epoch": 6.47, + "learning_rate": 2.3539598472718315e-06, + "loss": 0.6357, + "step": 17534 + }, + { + "epoch": 6.48, + "learning_rate": 2.353922896908486e-06, + "loss": 0.502, + "step": 17535 + }, + { + "epoch": 6.48, + "learning_rate": 2.353885946545141e-06, + "loss": 0.434, + "step": 17536 + }, + { + "epoch": 6.48, + "learning_rate": 2.353848996181796e-06, + "loss": 0.4398, + "step": 17537 + }, + { + "epoch": 6.48, + "learning_rate": 2.353812045818451e-06, + "loss": 0.4817, + "step": 17538 + }, + { + "epoch": 6.48, + "learning_rate": 2.353775095455105e-06, + "loss": 0.4562, + "step": 17539 + }, + { + "epoch": 6.48, + "learning_rate": 2.35373814509176e-06, + "loss": 0.5196, + "step": 17540 + }, + { + "epoch": 6.48, + "learning_rate": 2.3537011947284148e-06, + "loss": 0.4733, + "step": 17541 + }, + { + "epoch": 6.48, + "learning_rate": 2.35366424436507e-06, + "loss": 0.5734, + "step": 17542 + }, + { + "epoch": 6.48, + "learning_rate": 2.3536272940017245e-06, + "loss": 0.5261, + "step": 17543 + }, + { + "epoch": 6.48, + "learning_rate": 2.3535903436383795e-06, + "loss": 0.4443, + "step": 17544 + }, + { + "epoch": 6.48, + "learning_rate": 2.3535533932750337e-06, + "loss": 0.5253, + "step": 17545 + }, + { + "epoch": 6.48, + "learning_rate": 2.3535164429116884e-06, + "loss": 0.3843, + "step": 17546 + }, + { + "epoch": 6.48, + "learning_rate": 2.3534794925483434e-06, + "loss": 0.4997, + "step": 17547 + }, + { + "epoch": 6.48, + "learning_rate": 2.353442542184998e-06, + "loss": 0.3469, + "step": 17548 + }, + { + "epoch": 6.48, + "learning_rate": 2.353405591821653e-06, + "loss": 0.5389, + "step": 17549 + }, + { + "epoch": 6.48, + "learning_rate": 2.3533686414583078e-06, + "loss": 0.5562, + "step": 17550 + }, + { + "epoch": 6.48, + "learning_rate": 2.3533316910949624e-06, + "loss": 0.3893, + "step": 17551 + }, + { + "epoch": 6.48, + "learning_rate": 2.353294740731617e-06, + "loss": 0.4776, + "step": 17552 + }, + { + "epoch": 6.48, + "learning_rate": 2.353257790368272e-06, + "loss": 0.4637, + "step": 17553 + }, + { + "epoch": 6.48, + "learning_rate": 2.3532208400049267e-06, + "loss": 0.439, + "step": 17554 + }, + { + "epoch": 6.48, + "learning_rate": 2.3531838896415814e-06, + "loss": 0.3865, + "step": 17555 + }, + { + "epoch": 6.48, + "learning_rate": 2.3531469392782364e-06, + "loss": 0.4779, + "step": 17556 + }, + { + "epoch": 6.48, + "learning_rate": 2.353109988914891e-06, + "loss": 0.5006, + "step": 17557 + }, + { + "epoch": 6.48, + "learning_rate": 2.3530730385515457e-06, + "loss": 0.5236, + "step": 17558 + }, + { + "epoch": 6.48, + "learning_rate": 2.3530360881882003e-06, + "loss": 0.4915, + "step": 17559 + }, + { + "epoch": 6.48, + "learning_rate": 2.3529991378248554e-06, + "loss": 0.4474, + "step": 17560 + }, + { + "epoch": 6.48, + "learning_rate": 2.35296218746151e-06, + "loss": 0.4183, + "step": 17561 + }, + { + "epoch": 6.49, + "learning_rate": 2.352925237098165e-06, + "loss": 0.4269, + "step": 17562 + }, + { + "epoch": 6.49, + "learning_rate": 2.3528882867348197e-06, + "loss": 0.3433, + "step": 17563 + }, + { + "epoch": 6.49, + "learning_rate": 2.3528513363714744e-06, + "loss": 0.5942, + "step": 17564 + }, + { + "epoch": 6.49, + "learning_rate": 2.352814386008129e-06, + "loss": 0.3898, + "step": 17565 + }, + { + "epoch": 6.49, + "learning_rate": 2.3527774356447836e-06, + "loss": 0.4446, + "step": 17566 + }, + { + "epoch": 6.49, + "learning_rate": 2.3527404852814387e-06, + "loss": 0.4202, + "step": 17567 + }, + { + "epoch": 6.49, + "learning_rate": 2.3527035349180933e-06, + "loss": 0.4019, + "step": 17568 + }, + { + "epoch": 6.49, + "learning_rate": 2.3526665845547484e-06, + "loss": 0.5018, + "step": 17569 + }, + { + "epoch": 6.49, + "learning_rate": 2.352629634191403e-06, + "loss": 0.5229, + "step": 17570 + }, + { + "epoch": 6.49, + "learning_rate": 2.3525926838280577e-06, + "loss": 0.4971, + "step": 17571 + }, + { + "epoch": 6.49, + "learning_rate": 2.3525557334647123e-06, + "loss": 0.4117, + "step": 17572 + }, + { + "epoch": 6.49, + "learning_rate": 2.3525187831013674e-06, + "loss": 0.4572, + "step": 17573 + }, + { + "epoch": 6.49, + "learning_rate": 2.352481832738022e-06, + "loss": 0.388, + "step": 17574 + }, + { + "epoch": 6.49, + "learning_rate": 2.3524448823746766e-06, + "loss": 0.5667, + "step": 17575 + }, + { + "epoch": 6.49, + "learning_rate": 2.3524079320113317e-06, + "loss": 0.4031, + "step": 17576 + }, + { + "epoch": 6.49, + "learning_rate": 2.3523709816479863e-06, + "loss": 0.3852, + "step": 17577 + }, + { + "epoch": 6.49, + "learning_rate": 2.352334031284641e-06, + "loss": 0.5209, + "step": 17578 + }, + { + "epoch": 6.49, + "learning_rate": 2.3522970809212956e-06, + "loss": 0.3913, + "step": 17579 + }, + { + "epoch": 6.49, + "learning_rate": 2.3522601305579507e-06, + "loss": 0.4774, + "step": 17580 + }, + { + "epoch": 6.49, + "learning_rate": 2.3522231801946053e-06, + "loss": 0.3704, + "step": 17581 + }, + { + "epoch": 6.49, + "learning_rate": 2.3521862298312604e-06, + "loss": 0.4699, + "step": 17582 + }, + { + "epoch": 6.49, + "learning_rate": 2.352149279467915e-06, + "loss": 0.5051, + "step": 17583 + }, + { + "epoch": 6.49, + "learning_rate": 2.3521123291045696e-06, + "loss": 0.4409, + "step": 17584 + }, + { + "epoch": 6.49, + "learning_rate": 2.3520753787412243e-06, + "loss": 0.5042, + "step": 17585 + }, + { + "epoch": 6.49, + "learning_rate": 2.352038428377879e-06, + "loss": 0.4239, + "step": 17586 + }, + { + "epoch": 6.49, + "learning_rate": 2.352001478014534e-06, + "loss": 0.448, + "step": 17587 + }, + { + "epoch": 6.49, + "learning_rate": 2.3519645276511886e-06, + "loss": 0.3473, + "step": 17588 + }, + { + "epoch": 6.5, + "learning_rate": 2.3519275772878437e-06, + "loss": 0.3396, + "step": 17589 + }, + { + "epoch": 6.5, + "learning_rate": 2.351890626924498e-06, + "loss": 0.467, + "step": 17590 + }, + { + "epoch": 6.5, + "learning_rate": 2.351853676561153e-06, + "loss": 0.3744, + "step": 17591 + }, + { + "epoch": 6.5, + "learning_rate": 2.3518167261978076e-06, + "loss": 0.4968, + "step": 17592 + }, + { + "epoch": 6.5, + "learning_rate": 2.3517797758344626e-06, + "loss": 0.3807, + "step": 17593 + }, + { + "epoch": 6.5, + "learning_rate": 2.3517428254711173e-06, + "loss": 0.4014, + "step": 17594 + }, + { + "epoch": 6.5, + "learning_rate": 2.351705875107772e-06, + "loss": 0.5238, + "step": 17595 + }, + { + "epoch": 6.5, + "learning_rate": 2.3516689247444266e-06, + "loss": 0.401, + "step": 17596 + }, + { + "epoch": 6.5, + "learning_rate": 2.351631974381081e-06, + "loss": 0.3713, + "step": 17597 + }, + { + "epoch": 6.5, + "learning_rate": 2.3515950240177363e-06, + "loss": 0.5117, + "step": 17598 + }, + { + "epoch": 6.5, + "learning_rate": 2.351558073654391e-06, + "loss": 0.492, + "step": 17599 + }, + { + "epoch": 6.5, + "learning_rate": 2.351521123291046e-06, + "loss": 0.3657, + "step": 17600 + }, + { + "epoch": 6.5, + "learning_rate": 2.3514841729277006e-06, + "loss": 0.4635, + "step": 17601 + }, + { + "epoch": 6.5, + "learning_rate": 2.3514472225643552e-06, + "loss": 0.5247, + "step": 17602 + }, + { + "epoch": 6.5, + "learning_rate": 2.35141027220101e-06, + "loss": 0.4456, + "step": 17603 + }, + { + "epoch": 6.5, + "learning_rate": 2.351373321837665e-06, + "loss": 0.5609, + "step": 17604 + }, + { + "epoch": 6.5, + "learning_rate": 2.3513363714743196e-06, + "loss": 0.5116, + "step": 17605 + }, + { + "epoch": 6.5, + "learning_rate": 2.351299421110974e-06, + "loss": 0.428, + "step": 17606 + }, + { + "epoch": 6.5, + "learning_rate": 2.3512624707476292e-06, + "loss": 0.3585, + "step": 17607 + }, + { + "epoch": 6.5, + "learning_rate": 2.351225520384284e-06, + "loss": 0.5551, + "step": 17608 + }, + { + "epoch": 6.5, + "learning_rate": 2.3511885700209385e-06, + "loss": 0.4989, + "step": 17609 + }, + { + "epoch": 6.5, + "learning_rate": 2.351151619657593e-06, + "loss": 0.443, + "step": 17610 + }, + { + "epoch": 6.5, + "learning_rate": 2.3511146692942482e-06, + "loss": 0.404, + "step": 17611 + }, + { + "epoch": 6.5, + "learning_rate": 2.351077718930903e-06, + "loss": 0.3558, + "step": 17612 + }, + { + "epoch": 6.5, + "learning_rate": 2.351040768567558e-06, + "loss": 0.464, + "step": 17613 + }, + { + "epoch": 6.5, + "learning_rate": 2.3510038182042126e-06, + "loss": 0.4394, + "step": 17614 + }, + { + "epoch": 6.5, + "learning_rate": 2.350966867840867e-06, + "loss": 0.4555, + "step": 17615 + }, + { + "epoch": 6.51, + "learning_rate": 2.350929917477522e-06, + "loss": 0.3307, + "step": 17616 + }, + { + "epoch": 6.51, + "learning_rate": 2.3508929671141765e-06, + "loss": 0.4875, + "step": 17617 + }, + { + "epoch": 6.51, + "learning_rate": 2.3508560167508315e-06, + "loss": 0.3513, + "step": 17618 + }, + { + "epoch": 6.51, + "learning_rate": 2.350819066387486e-06, + "loss": 0.3226, + "step": 17619 + }, + { + "epoch": 6.51, + "learning_rate": 2.3507821160241412e-06, + "loss": 0.3875, + "step": 17620 + }, + { + "epoch": 6.51, + "learning_rate": 2.350745165660796e-06, + "loss": 0.5067, + "step": 17621 + }, + { + "epoch": 6.51, + "learning_rate": 2.3507082152974505e-06, + "loss": 0.5288, + "step": 17622 + }, + { + "epoch": 6.51, + "learning_rate": 2.350671264934105e-06, + "loss": 0.4552, + "step": 17623 + }, + { + "epoch": 6.51, + "learning_rate": 2.35063431457076e-06, + "loss": 0.4357, + "step": 17624 + }, + { + "epoch": 6.51, + "learning_rate": 2.350597364207415e-06, + "loss": 0.4801, + "step": 17625 + }, + { + "epoch": 6.51, + "learning_rate": 2.3505604138440695e-06, + "loss": 0.5296, + "step": 17626 + }, + { + "epoch": 6.51, + "learning_rate": 2.3505234634807245e-06, + "loss": 0.4941, + "step": 17627 + }, + { + "epoch": 6.51, + "learning_rate": 2.350486513117379e-06, + "loss": 0.4931, + "step": 17628 + }, + { + "epoch": 6.51, + "learning_rate": 2.350449562754034e-06, + "loss": 0.4396, + "step": 17629 + }, + { + "epoch": 6.51, + "learning_rate": 2.3504126123906884e-06, + "loss": 0.4719, + "step": 17630 + }, + { + "epoch": 6.51, + "learning_rate": 2.3503756620273435e-06, + "loss": 0.3961, + "step": 17631 + }, + { + "epoch": 6.51, + "learning_rate": 2.350338711663998e-06, + "loss": 0.3514, + "step": 17632 + }, + { + "epoch": 6.51, + "learning_rate": 2.350301761300653e-06, + "loss": 0.2975, + "step": 17633 + }, + { + "epoch": 6.51, + "learning_rate": 2.3502648109373074e-06, + "loss": 0.4349, + "step": 17634 + }, + { + "epoch": 6.51, + "learning_rate": 2.350227860573962e-06, + "loss": 0.5009, + "step": 17635 + }, + { + "epoch": 6.51, + "learning_rate": 2.350190910210617e-06, + "loss": 0.3715, + "step": 17636 + }, + { + "epoch": 6.51, + "learning_rate": 2.3501539598472717e-06, + "loss": 0.3825, + "step": 17637 + }, + { + "epoch": 6.51, + "learning_rate": 2.350117009483927e-06, + "loss": 0.502, + "step": 17638 + }, + { + "epoch": 6.51, + "learning_rate": 2.3500800591205814e-06, + "loss": 0.467, + "step": 17639 + }, + { + "epoch": 6.51, + "learning_rate": 2.350043108757236e-06, + "loss": 0.4345, + "step": 17640 + }, + { + "epoch": 6.51, + "learning_rate": 2.3500061583938907e-06, + "loss": 0.4169, + "step": 17641 + }, + { + "epoch": 6.51, + "learning_rate": 2.3499692080305458e-06, + "loss": 0.4776, + "step": 17642 + }, + { + "epoch": 6.52, + "learning_rate": 2.3499322576672004e-06, + "loss": 0.4358, + "step": 17643 + }, + { + "epoch": 6.52, + "learning_rate": 2.3498953073038555e-06, + "loss": 0.3454, + "step": 17644 + }, + { + "epoch": 6.52, + "learning_rate": 2.34985835694051e-06, + "loss": 0.3962, + "step": 17645 + }, + { + "epoch": 6.52, + "learning_rate": 2.3498214065771647e-06, + "loss": 0.3315, + "step": 17646 + }, + { + "epoch": 6.52, + "learning_rate": 2.3497844562138194e-06, + "loss": 0.5142, + "step": 17647 + }, + { + "epoch": 6.52, + "learning_rate": 2.349747505850474e-06, + "loss": 0.4181, + "step": 17648 + }, + { + "epoch": 6.52, + "learning_rate": 2.349710555487129e-06, + "loss": 0.5492, + "step": 17649 + }, + { + "epoch": 6.52, + "learning_rate": 2.3496736051237837e-06, + "loss": 0.3825, + "step": 17650 + }, + { + "epoch": 6.52, + "learning_rate": 2.3496366547604388e-06, + "loss": 0.5568, + "step": 17651 + }, + { + "epoch": 6.52, + "learning_rate": 2.3495997043970934e-06, + "loss": 0.3777, + "step": 17652 + }, + { + "epoch": 6.52, + "learning_rate": 2.349562754033748e-06, + "loss": 0.4361, + "step": 17653 + }, + { + "epoch": 6.52, + "learning_rate": 2.3495258036704027e-06, + "loss": 0.4005, + "step": 17654 + }, + { + "epoch": 6.52, + "learning_rate": 2.3494888533070573e-06, + "loss": 0.5522, + "step": 17655 + }, + { + "epoch": 6.52, + "learning_rate": 2.3494519029437124e-06, + "loss": 0.3427, + "step": 17656 + }, + { + "epoch": 6.52, + "learning_rate": 2.349414952580367e-06, + "loss": 0.4567, + "step": 17657 + }, + { + "epoch": 6.52, + "learning_rate": 2.349378002217022e-06, + "loss": 0.4788, + "step": 17658 + }, + { + "epoch": 6.52, + "learning_rate": 2.3493410518536767e-06, + "loss": 0.4828, + "step": 17659 + }, + { + "epoch": 6.52, + "learning_rate": 2.3493041014903313e-06, + "loss": 0.5265, + "step": 17660 + }, + { + "epoch": 6.52, + "learning_rate": 2.349267151126986e-06, + "loss": 0.407, + "step": 17661 + }, + { + "epoch": 6.52, + "learning_rate": 2.349230200763641e-06, + "loss": 0.4046, + "step": 17662 + }, + { + "epoch": 6.52, + "learning_rate": 2.3491932504002957e-06, + "loss": 0.5403, + "step": 17663 + }, + { + "epoch": 6.52, + "learning_rate": 2.3491563000369507e-06, + "loss": 0.4427, + "step": 17664 + }, + { + "epoch": 6.52, + "learning_rate": 2.3491193496736054e-06, + "loss": 0.5352, + "step": 17665 + }, + { + "epoch": 6.52, + "learning_rate": 2.34908239931026e-06, + "loss": 0.457, + "step": 17666 + }, + { + "epoch": 6.52, + "learning_rate": 2.3490454489469146e-06, + "loss": 0.3836, + "step": 17667 + }, + { + "epoch": 6.52, + "learning_rate": 2.3490084985835693e-06, + "loss": 0.3731, + "step": 17668 + }, + { + "epoch": 6.52, + "learning_rate": 2.3489715482202243e-06, + "loss": 0.3912, + "step": 17669 + }, + { + "epoch": 6.53, + "learning_rate": 2.348934597856879e-06, + "loss": 0.3773, + "step": 17670 + }, + { + "epoch": 6.53, + "learning_rate": 2.348897647493534e-06, + "loss": 0.4215, + "step": 17671 + }, + { + "epoch": 6.53, + "learning_rate": 2.3488606971301887e-06, + "loss": 0.5169, + "step": 17672 + }, + { + "epoch": 6.53, + "learning_rate": 2.3488237467668433e-06, + "loss": 0.3599, + "step": 17673 + }, + { + "epoch": 6.53, + "learning_rate": 2.348786796403498e-06, + "loss": 0.4726, + "step": 17674 + }, + { + "epoch": 6.53, + "learning_rate": 2.3487498460401526e-06, + "loss": 0.4878, + "step": 17675 + }, + { + "epoch": 6.53, + "learning_rate": 2.3487128956768076e-06, + "loss": 0.3709, + "step": 17676 + }, + { + "epoch": 6.53, + "learning_rate": 2.3486759453134623e-06, + "loss": 0.5661, + "step": 17677 + }, + { + "epoch": 6.53, + "learning_rate": 2.3486389949501173e-06, + "loss": 0.4022, + "step": 17678 + }, + { + "epoch": 6.53, + "learning_rate": 2.3486020445867715e-06, + "loss": 0.3933, + "step": 17679 + }, + { + "epoch": 6.53, + "learning_rate": 2.3485650942234266e-06, + "loss": 0.4166, + "step": 17680 + }, + { + "epoch": 6.53, + "learning_rate": 2.3485281438600812e-06, + "loss": 0.4949, + "step": 17681 + }, + { + "epoch": 6.53, + "learning_rate": 2.3484911934967363e-06, + "loss": 0.4391, + "step": 17682 + }, + { + "epoch": 6.53, + "learning_rate": 2.348454243133391e-06, + "loss": 0.3591, + "step": 17683 + }, + { + "epoch": 6.53, + "learning_rate": 2.3484172927700456e-06, + "loss": 0.5116, + "step": 17684 + }, + { + "epoch": 6.53, + "learning_rate": 2.3483803424067e-06, + "loss": 0.4019, + "step": 17685 + }, + { + "epoch": 6.53, + "learning_rate": 2.348343392043355e-06, + "loss": 0.4966, + "step": 17686 + }, + { + "epoch": 6.53, + "learning_rate": 2.34830644168001e-06, + "loss": 0.4433, + "step": 17687 + }, + { + "epoch": 6.53, + "learning_rate": 2.3482694913166645e-06, + "loss": 0.4457, + "step": 17688 + }, + { + "epoch": 6.53, + "learning_rate": 2.3482325409533196e-06, + "loss": 0.4681, + "step": 17689 + }, + { + "epoch": 6.53, + "learning_rate": 2.3481955905899742e-06, + "loss": 0.4615, + "step": 17690 + }, + { + "epoch": 6.53, + "learning_rate": 2.348158640226629e-06, + "loss": 0.4067, + "step": 17691 + }, + { + "epoch": 6.53, + "learning_rate": 2.3481216898632835e-06, + "loss": 0.4069, + "step": 17692 + }, + { + "epoch": 6.53, + "learning_rate": 2.3480847394999386e-06, + "loss": 0.3543, + "step": 17693 + }, + { + "epoch": 6.53, + "learning_rate": 2.348047789136593e-06, + "loss": 0.5513, + "step": 17694 + }, + { + "epoch": 6.53, + "learning_rate": 2.348010838773248e-06, + "loss": 0.4811, + "step": 17695 + }, + { + "epoch": 6.53, + "learning_rate": 2.347973888409903e-06, + "loss": 0.413, + "step": 17696 + }, + { + "epoch": 6.54, + "learning_rate": 2.3479369380465575e-06, + "loss": 0.4571, + "step": 17697 + }, + { + "epoch": 6.54, + "learning_rate": 2.347899987683212e-06, + "loss": 0.4655, + "step": 17698 + }, + { + "epoch": 6.54, + "learning_rate": 2.347863037319867e-06, + "loss": 0.4915, + "step": 17699 + }, + { + "epoch": 6.54, + "learning_rate": 2.347826086956522e-06, + "loss": 0.4317, + "step": 17700 + }, + { + "epoch": 6.54, + "learning_rate": 2.3477891365931765e-06, + "loss": 0.413, + "step": 17701 + }, + { + "epoch": 6.54, + "learning_rate": 2.3477521862298316e-06, + "loss": 0.4913, + "step": 17702 + }, + { + "epoch": 6.54, + "learning_rate": 2.347715235866486e-06, + "loss": 0.481, + "step": 17703 + }, + { + "epoch": 6.54, + "learning_rate": 2.347678285503141e-06, + "loss": 0.3451, + "step": 17704 + }, + { + "epoch": 6.54, + "learning_rate": 2.3476413351397955e-06, + "loss": 0.472, + "step": 17705 + }, + { + "epoch": 6.54, + "learning_rate": 2.34760438477645e-06, + "loss": 0.4767, + "step": 17706 + }, + { + "epoch": 6.54, + "learning_rate": 2.347567434413105e-06, + "loss": 0.4621, + "step": 17707 + }, + { + "epoch": 6.54, + "learning_rate": 2.34753048404976e-06, + "loss": 0.3942, + "step": 17708 + }, + { + "epoch": 6.54, + "learning_rate": 2.347493533686415e-06, + "loss": 0.2863, + "step": 17709 + }, + { + "epoch": 6.54, + "learning_rate": 2.3474565833230695e-06, + "loss": 0.3571, + "step": 17710 + }, + { + "epoch": 6.54, + "learning_rate": 2.347419632959724e-06, + "loss": 0.4649, + "step": 17711 + }, + { + "epoch": 6.54, + "learning_rate": 2.3473826825963788e-06, + "loss": 0.5161, + "step": 17712 + }, + { + "epoch": 6.54, + "learning_rate": 2.347345732233034e-06, + "loss": 0.4431, + "step": 17713 + }, + { + "epoch": 6.54, + "learning_rate": 2.3473087818696885e-06, + "loss": 0.5173, + "step": 17714 + }, + { + "epoch": 6.54, + "learning_rate": 2.347271831506343e-06, + "loss": 0.4036, + "step": 17715 + }, + { + "epoch": 6.54, + "learning_rate": 2.347234881142998e-06, + "loss": 0.4413, + "step": 17716 + }, + { + "epoch": 6.54, + "learning_rate": 2.347197930779653e-06, + "loss": 0.4101, + "step": 17717 + }, + { + "epoch": 6.54, + "learning_rate": 2.3471609804163074e-06, + "loss": 0.3809, + "step": 17718 + }, + { + "epoch": 6.54, + "learning_rate": 2.347124030052962e-06, + "loss": 0.3311, + "step": 17719 + }, + { + "epoch": 6.54, + "learning_rate": 2.347087079689617e-06, + "loss": 0.4899, + "step": 17720 + }, + { + "epoch": 6.54, + "learning_rate": 2.3470501293262718e-06, + "loss": 0.4279, + "step": 17721 + }, + { + "epoch": 6.54, + "learning_rate": 2.347013178962927e-06, + "loss": 0.5454, + "step": 17722 + }, + { + "epoch": 6.54, + "learning_rate": 2.3469762285995815e-06, + "loss": 0.4863, + "step": 17723 + }, + { + "epoch": 6.55, + "learning_rate": 2.3469392782362357e-06, + "loss": 0.4854, + "step": 17724 + }, + { + "epoch": 6.55, + "learning_rate": 2.3469023278728907e-06, + "loss": 0.4932, + "step": 17725 + }, + { + "epoch": 6.55, + "learning_rate": 2.3468653775095454e-06, + "loss": 0.4749, + "step": 17726 + }, + { + "epoch": 6.55, + "learning_rate": 2.3468284271462004e-06, + "loss": 0.4847, + "step": 17727 + }, + { + "epoch": 6.55, + "learning_rate": 2.346791476782855e-06, + "loss": 0.4212, + "step": 17728 + }, + { + "epoch": 6.55, + "learning_rate": 2.34675452641951e-06, + "loss": 0.3986, + "step": 17729 + }, + { + "epoch": 6.55, + "learning_rate": 2.3467175760561644e-06, + "loss": 0.5351, + "step": 17730 + }, + { + "epoch": 6.55, + "learning_rate": 2.3466806256928194e-06, + "loss": 0.4836, + "step": 17731 + }, + { + "epoch": 6.55, + "learning_rate": 2.346643675329474e-06, + "loss": 0.4071, + "step": 17732 + }, + { + "epoch": 6.55, + "learning_rate": 2.346606724966129e-06, + "loss": 0.5064, + "step": 17733 + }, + { + "epoch": 6.55, + "learning_rate": 2.3465697746027837e-06, + "loss": 0.4381, + "step": 17734 + }, + { + "epoch": 6.55, + "learning_rate": 2.3465328242394384e-06, + "loss": 0.462, + "step": 17735 + }, + { + "epoch": 6.55, + "learning_rate": 2.346495873876093e-06, + "loss": 0.4692, + "step": 17736 + }, + { + "epoch": 6.55, + "learning_rate": 2.3464589235127477e-06, + "loss": 0.4493, + "step": 17737 + }, + { + "epoch": 6.55, + "learning_rate": 2.3464219731494027e-06, + "loss": 0.5134, + "step": 17738 + }, + { + "epoch": 6.55, + "learning_rate": 2.3463850227860574e-06, + "loss": 0.3892, + "step": 17739 + }, + { + "epoch": 6.55, + "learning_rate": 2.3463480724227124e-06, + "loss": 0.5082, + "step": 17740 + }, + { + "epoch": 6.55, + "learning_rate": 2.346311122059367e-06, + "loss": 0.4125, + "step": 17741 + }, + { + "epoch": 6.55, + "learning_rate": 2.3462741716960217e-06, + "loss": 0.4649, + "step": 17742 + }, + { + "epoch": 6.55, + "learning_rate": 2.3462372213326763e-06, + "loss": 0.5136, + "step": 17743 + }, + { + "epoch": 6.55, + "learning_rate": 2.346200270969331e-06, + "loss": 0.5181, + "step": 17744 + }, + { + "epoch": 6.55, + "learning_rate": 2.346163320605986e-06, + "loss": 0.5399, + "step": 17745 + }, + { + "epoch": 6.55, + "learning_rate": 2.3461263702426407e-06, + "loss": 0.4272, + "step": 17746 + }, + { + "epoch": 6.55, + "learning_rate": 2.3460894198792957e-06, + "loss": 0.5, + "step": 17747 + }, + { + "epoch": 6.55, + "learning_rate": 2.3460524695159504e-06, + "loss": 0.3563, + "step": 17748 + }, + { + "epoch": 6.55, + "learning_rate": 2.346015519152605e-06, + "loss": 0.5376, + "step": 17749 + }, + { + "epoch": 6.55, + "learning_rate": 2.3459785687892596e-06, + "loss": 0.3728, + "step": 17750 + }, + { + "epoch": 6.56, + "learning_rate": 2.3459416184259147e-06, + "loss": 0.3878, + "step": 17751 + }, + { + "epoch": 6.56, + "learning_rate": 2.3459046680625693e-06, + "loss": 0.4162, + "step": 17752 + }, + { + "epoch": 6.56, + "learning_rate": 2.3458677176992244e-06, + "loss": 0.4413, + "step": 17753 + }, + { + "epoch": 6.56, + "learning_rate": 2.345830767335879e-06, + "loss": 0.3796, + "step": 17754 + }, + { + "epoch": 6.56, + "learning_rate": 2.3457938169725337e-06, + "loss": 0.4465, + "step": 17755 + }, + { + "epoch": 6.56, + "learning_rate": 2.3457568666091883e-06, + "loss": 0.5468, + "step": 17756 + }, + { + "epoch": 6.56, + "learning_rate": 2.345719916245843e-06, + "loss": 0.4822, + "step": 17757 + }, + { + "epoch": 6.56, + "learning_rate": 2.345682965882498e-06, + "loss": 0.5048, + "step": 17758 + }, + { + "epoch": 6.56, + "learning_rate": 2.3456460155191526e-06, + "loss": 0.49, + "step": 17759 + }, + { + "epoch": 6.56, + "learning_rate": 2.3456090651558077e-06, + "loss": 0.4046, + "step": 17760 + }, + { + "epoch": 6.56, + "learning_rate": 2.3455721147924623e-06, + "loss": 0.4283, + "step": 17761 + }, + { + "epoch": 6.56, + "learning_rate": 2.345535164429117e-06, + "loss": 0.4757, + "step": 17762 + }, + { + "epoch": 6.56, + "learning_rate": 2.3454982140657716e-06, + "loss": 0.3156, + "step": 17763 + }, + { + "epoch": 6.56, + "learning_rate": 2.3454612637024262e-06, + "loss": 0.4265, + "step": 17764 + }, + { + "epoch": 6.56, + "learning_rate": 2.3454243133390813e-06, + "loss": 0.4489, + "step": 17765 + }, + { + "epoch": 6.56, + "learning_rate": 2.345387362975736e-06, + "loss": 0.4374, + "step": 17766 + }, + { + "epoch": 6.56, + "learning_rate": 2.345350412612391e-06, + "loss": 0.4182, + "step": 17767 + }, + { + "epoch": 6.56, + "learning_rate": 2.3453134622490456e-06, + "loss": 0.3935, + "step": 17768 + }, + { + "epoch": 6.56, + "learning_rate": 2.3452765118857003e-06, + "loss": 0.4886, + "step": 17769 + }, + { + "epoch": 6.56, + "learning_rate": 2.345239561522355e-06, + "loss": 0.611, + "step": 17770 + }, + { + "epoch": 6.56, + "learning_rate": 2.34520261115901e-06, + "loss": 0.3675, + "step": 17771 + }, + { + "epoch": 6.56, + "learning_rate": 2.3451656607956646e-06, + "loss": 0.5223, + "step": 17772 + }, + { + "epoch": 6.56, + "learning_rate": 2.3451287104323196e-06, + "loss": 0.4454, + "step": 17773 + }, + { + "epoch": 6.56, + "learning_rate": 2.3450917600689743e-06, + "loss": 0.446, + "step": 17774 + }, + { + "epoch": 6.56, + "learning_rate": 2.3450548097056285e-06, + "loss": 0.4513, + "step": 17775 + }, + { + "epoch": 6.56, + "learning_rate": 2.3450178593422836e-06, + "loss": 0.3325, + "step": 17776 + }, + { + "epoch": 6.56, + "learning_rate": 2.344980908978938e-06, + "loss": 0.3886, + "step": 17777 + }, + { + "epoch": 6.56, + "learning_rate": 2.3449439586155933e-06, + "loss": 0.3757, + "step": 17778 + }, + { + "epoch": 6.57, + "learning_rate": 2.344907008252248e-06, + "loss": 0.483, + "step": 17779 + }, + { + "epoch": 6.57, + "learning_rate": 2.344870057888903e-06, + "loss": 0.4963, + "step": 17780 + }, + { + "epoch": 6.57, + "learning_rate": 2.344833107525557e-06, + "loss": 0.4845, + "step": 17781 + }, + { + "epoch": 6.57, + "learning_rate": 2.3447961571622122e-06, + "loss": 0.4368, + "step": 17782 + }, + { + "epoch": 6.57, + "learning_rate": 2.344759206798867e-06, + "loss": 0.3228, + "step": 17783 + }, + { + "epoch": 6.57, + "learning_rate": 2.3447222564355215e-06, + "loss": 0.4479, + "step": 17784 + }, + { + "epoch": 6.57, + "learning_rate": 2.3446853060721766e-06, + "loss": 0.3422, + "step": 17785 + }, + { + "epoch": 6.57, + "learning_rate": 2.344648355708831e-06, + "loss": 0.6575, + "step": 17786 + }, + { + "epoch": 6.57, + "learning_rate": 2.344611405345486e-06, + "loss": 0.466, + "step": 17787 + }, + { + "epoch": 6.57, + "learning_rate": 2.3445744549821405e-06, + "loss": 0.3716, + "step": 17788 + }, + { + "epoch": 6.57, + "learning_rate": 2.3445375046187955e-06, + "loss": 0.4494, + "step": 17789 + }, + { + "epoch": 6.57, + "learning_rate": 2.34450055425545e-06, + "loss": 0.4519, + "step": 17790 + }, + { + "epoch": 6.57, + "learning_rate": 2.3444636038921052e-06, + "loss": 0.3866, + "step": 17791 + }, + { + "epoch": 6.57, + "learning_rate": 2.34442665352876e-06, + "loss": 0.5508, + "step": 17792 + }, + { + "epoch": 6.57, + "learning_rate": 2.3443897031654145e-06, + "loss": 0.4435, + "step": 17793 + }, + { + "epoch": 6.57, + "learning_rate": 2.344352752802069e-06, + "loss": 0.4737, + "step": 17794 + }, + { + "epoch": 6.57, + "learning_rate": 2.3443158024387238e-06, + "loss": 0.4576, + "step": 17795 + }, + { + "epoch": 6.57, + "learning_rate": 2.344278852075379e-06, + "loss": 0.5042, + "step": 17796 + }, + { + "epoch": 6.57, + "learning_rate": 2.3442419017120335e-06, + "loss": 0.3154, + "step": 17797 + }, + { + "epoch": 6.57, + "learning_rate": 2.3442049513486885e-06, + "loss": 0.5631, + "step": 17798 + }, + { + "epoch": 6.57, + "learning_rate": 2.344168000985343e-06, + "loss": 0.4945, + "step": 17799 + }, + { + "epoch": 6.57, + "learning_rate": 2.344131050621998e-06, + "loss": 0.4644, + "step": 17800 + }, + { + "epoch": 6.57, + "learning_rate": 2.3440941002586524e-06, + "loss": 0.427, + "step": 17801 + }, + { + "epoch": 6.57, + "learning_rate": 2.3440571498953075e-06, + "loss": 0.5004, + "step": 17802 + }, + { + "epoch": 6.57, + "learning_rate": 2.344020199531962e-06, + "loss": 0.5043, + "step": 17803 + }, + { + "epoch": 6.57, + "learning_rate": 2.3439832491686168e-06, + "loss": 0.4711, + "step": 17804 + }, + { + "epoch": 6.57, + "learning_rate": 2.343946298805272e-06, + "loss": 0.4577, + "step": 17805 + }, + { + "epoch": 6.58, + "learning_rate": 2.3439093484419265e-06, + "loss": 0.5361, + "step": 17806 + }, + { + "epoch": 6.58, + "learning_rate": 2.343872398078581e-06, + "loss": 0.4061, + "step": 17807 + }, + { + "epoch": 6.58, + "learning_rate": 2.3438354477152357e-06, + "loss": 0.4226, + "step": 17808 + }, + { + "epoch": 6.58, + "learning_rate": 2.343798497351891e-06, + "loss": 0.5881, + "step": 17809 + }, + { + "epoch": 6.58, + "learning_rate": 2.3437615469885454e-06, + "loss": 0.4023, + "step": 17810 + }, + { + "epoch": 6.58, + "learning_rate": 2.3437245966252005e-06, + "loss": 0.4349, + "step": 17811 + }, + { + "epoch": 6.58, + "learning_rate": 2.343687646261855e-06, + "loss": 0.3713, + "step": 17812 + }, + { + "epoch": 6.58, + "learning_rate": 2.3436506958985098e-06, + "loss": 0.5088, + "step": 17813 + }, + { + "epoch": 6.58, + "learning_rate": 2.3436137455351644e-06, + "loss": 0.3439, + "step": 17814 + }, + { + "epoch": 6.58, + "learning_rate": 2.343576795171819e-06, + "loss": 0.5448, + "step": 17815 + }, + { + "epoch": 6.58, + "learning_rate": 2.343539844808474e-06, + "loss": 0.4376, + "step": 17816 + }, + { + "epoch": 6.58, + "learning_rate": 2.3435028944451287e-06, + "loss": 0.3832, + "step": 17817 + }, + { + "epoch": 6.58, + "learning_rate": 2.343465944081784e-06, + "loss": 0.4216, + "step": 17818 + }, + { + "epoch": 6.58, + "learning_rate": 2.343428993718438e-06, + "loss": 0.4928, + "step": 17819 + }, + { + "epoch": 6.58, + "learning_rate": 2.343392043355093e-06, + "loss": 0.414, + "step": 17820 + }, + { + "epoch": 6.58, + "learning_rate": 2.3433550929917477e-06, + "loss": 0.4854, + "step": 17821 + }, + { + "epoch": 6.58, + "learning_rate": 2.3433181426284028e-06, + "loss": 0.4171, + "step": 17822 + }, + { + "epoch": 6.58, + "learning_rate": 2.3432811922650574e-06, + "loss": 0.4486, + "step": 17823 + }, + { + "epoch": 6.58, + "learning_rate": 2.343244241901712e-06, + "loss": 0.5109, + "step": 17824 + }, + { + "epoch": 6.58, + "learning_rate": 2.3432072915383667e-06, + "loss": 0.4426, + "step": 17825 + }, + { + "epoch": 6.58, + "learning_rate": 2.3431703411750213e-06, + "loss": 0.3277, + "step": 17826 + }, + { + "epoch": 6.58, + "learning_rate": 2.3431333908116764e-06, + "loss": 0.4635, + "step": 17827 + }, + { + "epoch": 6.58, + "learning_rate": 2.343096440448331e-06, + "loss": 0.3774, + "step": 17828 + }, + { + "epoch": 6.58, + "learning_rate": 2.343059490084986e-06, + "loss": 0.3864, + "step": 17829 + }, + { + "epoch": 6.58, + "learning_rate": 2.3430225397216407e-06, + "loss": 0.4552, + "step": 17830 + }, + { + "epoch": 6.58, + "learning_rate": 2.3429855893582953e-06, + "loss": 0.3466, + "step": 17831 + }, + { + "epoch": 6.58, + "learning_rate": 2.34294863899495e-06, + "loss": 0.5136, + "step": 17832 + }, + { + "epoch": 6.59, + "learning_rate": 2.342911688631605e-06, + "loss": 0.4184, + "step": 17833 + }, + { + "epoch": 6.59, + "learning_rate": 2.3428747382682597e-06, + "loss": 0.448, + "step": 17834 + }, + { + "epoch": 6.59, + "learning_rate": 2.3428377879049143e-06, + "loss": 0.5186, + "step": 17835 + }, + { + "epoch": 6.59, + "learning_rate": 2.3428008375415694e-06, + "loss": 0.4779, + "step": 17836 + }, + { + "epoch": 6.59, + "learning_rate": 2.342763887178224e-06, + "loss": 0.3966, + "step": 17837 + }, + { + "epoch": 6.59, + "learning_rate": 2.3427269368148786e-06, + "loss": 0.451, + "step": 17838 + }, + { + "epoch": 6.59, + "learning_rate": 2.3426899864515333e-06, + "loss": 0.452, + "step": 17839 + }, + { + "epoch": 6.59, + "learning_rate": 2.3426530360881883e-06, + "loss": 0.4312, + "step": 17840 + }, + { + "epoch": 6.59, + "learning_rate": 2.342616085724843e-06, + "loss": 0.5444, + "step": 17841 + }, + { + "epoch": 6.59, + "learning_rate": 2.342579135361498e-06, + "loss": 0.4559, + "step": 17842 + }, + { + "epoch": 6.59, + "learning_rate": 2.3425421849981527e-06, + "loss": 0.4475, + "step": 17843 + }, + { + "epoch": 6.59, + "learning_rate": 2.3425052346348073e-06, + "loss": 0.4878, + "step": 17844 + }, + { + "epoch": 6.59, + "learning_rate": 2.342468284271462e-06, + "loss": 0.4336, + "step": 17845 + }, + { + "epoch": 6.59, + "learning_rate": 2.3424313339081166e-06, + "loss": 0.4735, + "step": 17846 + }, + { + "epoch": 6.59, + "learning_rate": 2.3423943835447716e-06, + "loss": 0.4963, + "step": 17847 + }, + { + "epoch": 6.59, + "learning_rate": 2.3423574331814263e-06, + "loss": 0.4508, + "step": 17848 + }, + { + "epoch": 6.59, + "learning_rate": 2.3423204828180813e-06, + "loss": 0.39, + "step": 17849 + }, + { + "epoch": 6.59, + "learning_rate": 2.342283532454736e-06, + "loss": 0.5527, + "step": 17850 + }, + { + "epoch": 6.59, + "learning_rate": 2.3422465820913906e-06, + "loss": 0.5039, + "step": 17851 + }, + { + "epoch": 6.59, + "learning_rate": 2.3422096317280452e-06, + "loss": 0.4548, + "step": 17852 + }, + { + "epoch": 6.59, + "learning_rate": 2.3421726813647003e-06, + "loss": 0.4717, + "step": 17853 + }, + { + "epoch": 6.59, + "learning_rate": 2.342135731001355e-06, + "loss": 0.4443, + "step": 17854 + }, + { + "epoch": 6.59, + "learning_rate": 2.3420987806380096e-06, + "loss": 0.3812, + "step": 17855 + }, + { + "epoch": 6.59, + "learning_rate": 2.3420618302746646e-06, + "loss": 0.5675, + "step": 17856 + }, + { + "epoch": 6.59, + "learning_rate": 2.3420248799113193e-06, + "loss": 0.3767, + "step": 17857 + }, + { + "epoch": 6.59, + "learning_rate": 2.341987929547974e-06, + "loss": 0.4892, + "step": 17858 + }, + { + "epoch": 6.59, + "learning_rate": 2.3419509791846285e-06, + "loss": 0.6231, + "step": 17859 + }, + { + "epoch": 6.6, + "learning_rate": 2.3419140288212836e-06, + "loss": 0.4308, + "step": 17860 + }, + { + "epoch": 6.6, + "learning_rate": 2.3418770784579382e-06, + "loss": 0.3558, + "step": 17861 + }, + { + "epoch": 6.6, + "learning_rate": 2.3418401280945933e-06, + "loss": 0.4773, + "step": 17862 + }, + { + "epoch": 6.6, + "learning_rate": 2.341803177731248e-06, + "loss": 0.3917, + "step": 17863 + }, + { + "epoch": 6.6, + "learning_rate": 2.341766227367902e-06, + "loss": 0.437, + "step": 17864 + }, + { + "epoch": 6.6, + "learning_rate": 2.3417292770045572e-06, + "loss": 0.3556, + "step": 17865 + }, + { + "epoch": 6.6, + "learning_rate": 2.341692326641212e-06, + "loss": 0.434, + "step": 17866 + }, + { + "epoch": 6.6, + "learning_rate": 2.341655376277867e-06, + "loss": 0.5068, + "step": 17867 + }, + { + "epoch": 6.6, + "learning_rate": 2.3416184259145215e-06, + "loss": 0.3907, + "step": 17868 + }, + { + "epoch": 6.6, + "learning_rate": 2.3415814755511766e-06, + "loss": 0.5083, + "step": 17869 + }, + { + "epoch": 6.6, + "learning_rate": 2.341544525187831e-06, + "loss": 0.5697, + "step": 17870 + }, + { + "epoch": 6.6, + "learning_rate": 2.341507574824486e-06, + "loss": 0.3361, + "step": 17871 + }, + { + "epoch": 6.6, + "learning_rate": 2.3414706244611405e-06, + "loss": 0.5383, + "step": 17872 + }, + { + "epoch": 6.6, + "learning_rate": 2.3414336740977956e-06, + "loss": 0.5195, + "step": 17873 + }, + { + "epoch": 6.6, + "learning_rate": 2.34139672373445e-06, + "loss": 0.4842, + "step": 17874 + }, + { + "epoch": 6.6, + "learning_rate": 2.341359773371105e-06, + "loss": 0.4787, + "step": 17875 + }, + { + "epoch": 6.6, + "learning_rate": 2.3413228230077595e-06, + "loss": 0.3499, + "step": 17876 + }, + { + "epoch": 6.6, + "learning_rate": 2.341285872644414e-06, + "loss": 0.5468, + "step": 17877 + }, + { + "epoch": 6.6, + "learning_rate": 2.341248922281069e-06, + "loss": 0.4396, + "step": 17878 + }, + { + "epoch": 6.6, + "learning_rate": 2.341211971917724e-06, + "loss": 0.3843, + "step": 17879 + }, + { + "epoch": 6.6, + "learning_rate": 2.341175021554379e-06, + "loss": 0.3951, + "step": 17880 + }, + { + "epoch": 6.6, + "learning_rate": 2.3411380711910335e-06, + "loss": 0.3834, + "step": 17881 + }, + { + "epoch": 6.6, + "learning_rate": 2.341101120827688e-06, + "loss": 0.3764, + "step": 17882 + }, + { + "epoch": 6.6, + "learning_rate": 2.3410641704643428e-06, + "loss": 0.4373, + "step": 17883 + }, + { + "epoch": 6.6, + "learning_rate": 2.3410272201009974e-06, + "loss": 0.3859, + "step": 17884 + }, + { + "epoch": 6.6, + "learning_rate": 2.3409902697376525e-06, + "loss": 0.4078, + "step": 17885 + }, + { + "epoch": 6.6, + "learning_rate": 2.340953319374307e-06, + "loss": 0.4856, + "step": 17886 + }, + { + "epoch": 6.61, + "learning_rate": 2.340916369010962e-06, + "loss": 0.4841, + "step": 17887 + }, + { + "epoch": 6.61, + "learning_rate": 2.340879418647617e-06, + "loss": 0.444, + "step": 17888 + }, + { + "epoch": 6.61, + "learning_rate": 2.3408424682842715e-06, + "loss": 0.501, + "step": 17889 + }, + { + "epoch": 6.61, + "learning_rate": 2.340805517920926e-06, + "loss": 0.3825, + "step": 17890 + }, + { + "epoch": 6.61, + "learning_rate": 2.340768567557581e-06, + "loss": 0.4399, + "step": 17891 + }, + { + "epoch": 6.61, + "learning_rate": 2.3407316171942358e-06, + "loss": 0.4565, + "step": 17892 + }, + { + "epoch": 6.61, + "learning_rate": 2.340694666830891e-06, + "loss": 0.3872, + "step": 17893 + }, + { + "epoch": 6.61, + "learning_rate": 2.3406577164675455e-06, + "loss": 0.3854, + "step": 17894 + }, + { + "epoch": 6.61, + "learning_rate": 2.3406207661042e-06, + "loss": 0.5366, + "step": 17895 + }, + { + "epoch": 6.61, + "learning_rate": 2.3405838157408548e-06, + "loss": 0.4673, + "step": 17896 + }, + { + "epoch": 6.61, + "learning_rate": 2.3405468653775094e-06, + "loss": 0.3976, + "step": 17897 + }, + { + "epoch": 6.61, + "learning_rate": 2.3405099150141645e-06, + "loss": 0.4798, + "step": 17898 + }, + { + "epoch": 6.61, + "learning_rate": 2.340472964650819e-06, + "loss": 0.4895, + "step": 17899 + }, + { + "epoch": 6.61, + "learning_rate": 2.340436014287474e-06, + "loss": 0.3483, + "step": 17900 + }, + { + "epoch": 6.61, + "learning_rate": 2.3403990639241288e-06, + "loss": 0.474, + "step": 17901 + }, + { + "epoch": 6.61, + "learning_rate": 2.3403621135607834e-06, + "loss": 0.4685, + "step": 17902 + }, + { + "epoch": 6.61, + "learning_rate": 2.340325163197438e-06, + "loss": 0.4018, + "step": 17903 + }, + { + "epoch": 6.61, + "learning_rate": 2.3402882128340927e-06, + "loss": 0.4014, + "step": 17904 + }, + { + "epoch": 6.61, + "learning_rate": 2.3402512624707478e-06, + "loss": 0.4341, + "step": 17905 + }, + { + "epoch": 6.61, + "learning_rate": 2.3402143121074024e-06, + "loss": 0.5287, + "step": 17906 + }, + { + "epoch": 6.61, + "learning_rate": 2.3401773617440574e-06, + "loss": 0.5639, + "step": 17907 + }, + { + "epoch": 6.61, + "learning_rate": 2.340140411380712e-06, + "loss": 0.4662, + "step": 17908 + }, + { + "epoch": 6.61, + "learning_rate": 2.3401034610173667e-06, + "loss": 0.4672, + "step": 17909 + }, + { + "epoch": 6.61, + "learning_rate": 2.3400665106540214e-06, + "loss": 0.3928, + "step": 17910 + }, + { + "epoch": 6.61, + "learning_rate": 2.3400295602906764e-06, + "loss": 0.4414, + "step": 17911 + }, + { + "epoch": 6.61, + "learning_rate": 2.339992609927331e-06, + "loss": 0.4064, + "step": 17912 + }, + { + "epoch": 6.61, + "learning_rate": 2.3399556595639857e-06, + "loss": 0.4259, + "step": 17913 + }, + { + "epoch": 6.62, + "learning_rate": 2.3399187092006408e-06, + "loss": 0.3723, + "step": 17914 + }, + { + "epoch": 6.62, + "learning_rate": 2.339881758837295e-06, + "loss": 0.5749, + "step": 17915 + }, + { + "epoch": 6.62, + "learning_rate": 2.33984480847395e-06, + "loss": 0.3892, + "step": 17916 + }, + { + "epoch": 6.62, + "learning_rate": 2.3398078581106047e-06, + "loss": 0.3385, + "step": 17917 + }, + { + "epoch": 6.62, + "learning_rate": 2.3397709077472597e-06, + "loss": 0.552, + "step": 17918 + }, + { + "epoch": 6.62, + "learning_rate": 2.3397339573839144e-06, + "loss": 0.476, + "step": 17919 + }, + { + "epoch": 6.62, + "learning_rate": 2.3396970070205694e-06, + "loss": 0.4445, + "step": 17920 + }, + { + "epoch": 6.62, + "learning_rate": 2.3396600566572236e-06, + "loss": 0.4835, + "step": 17921 + }, + { + "epoch": 6.62, + "learning_rate": 2.3396231062938787e-06, + "loss": 0.5418, + "step": 17922 + }, + { + "epoch": 6.62, + "learning_rate": 2.3395861559305333e-06, + "loss": 0.5724, + "step": 17923 + }, + { + "epoch": 6.62, + "learning_rate": 2.339549205567188e-06, + "loss": 0.488, + "step": 17924 + }, + { + "epoch": 6.62, + "learning_rate": 2.339512255203843e-06, + "loss": 0.3274, + "step": 17925 + }, + { + "epoch": 6.62, + "learning_rate": 2.3394753048404977e-06, + "loss": 0.4905, + "step": 17926 + }, + { + "epoch": 6.62, + "learning_rate": 2.3394383544771523e-06, + "loss": 0.5224, + "step": 17927 + }, + { + "epoch": 6.62, + "learning_rate": 2.339401404113807e-06, + "loss": 0.4953, + "step": 17928 + }, + { + "epoch": 6.62, + "learning_rate": 2.339364453750462e-06, + "loss": 0.4668, + "step": 17929 + }, + { + "epoch": 6.62, + "learning_rate": 2.3393275033871166e-06, + "loss": 0.3001, + "step": 17930 + }, + { + "epoch": 6.62, + "learning_rate": 2.3392905530237717e-06, + "loss": 0.4362, + "step": 17931 + }, + { + "epoch": 6.62, + "learning_rate": 2.3392536026604263e-06, + "loss": 0.5065, + "step": 17932 + }, + { + "epoch": 6.62, + "learning_rate": 2.339216652297081e-06, + "loss": 0.319, + "step": 17933 + }, + { + "epoch": 6.62, + "learning_rate": 2.3391797019337356e-06, + "loss": 0.5134, + "step": 17934 + }, + { + "epoch": 6.62, + "learning_rate": 2.3391427515703902e-06, + "loss": 0.5603, + "step": 17935 + }, + { + "epoch": 6.62, + "learning_rate": 2.3391058012070453e-06, + "loss": 0.495, + "step": 17936 + }, + { + "epoch": 6.62, + "learning_rate": 2.3390688508437e-06, + "loss": 0.3792, + "step": 17937 + }, + { + "epoch": 6.62, + "learning_rate": 2.339031900480355e-06, + "loss": 0.3641, + "step": 17938 + }, + { + "epoch": 6.62, + "learning_rate": 2.3389949501170096e-06, + "loss": 0.4483, + "step": 17939 + }, + { + "epoch": 6.62, + "learning_rate": 2.3389579997536643e-06, + "loss": 0.3936, + "step": 17940 + }, + { + "epoch": 6.63, + "learning_rate": 2.338921049390319e-06, + "loss": 0.425, + "step": 17941 + }, + { + "epoch": 6.63, + "learning_rate": 2.338884099026974e-06, + "loss": 0.4609, + "step": 17942 + }, + { + "epoch": 6.63, + "learning_rate": 2.3388471486636286e-06, + "loss": 0.508, + "step": 17943 + }, + { + "epoch": 6.63, + "learning_rate": 2.3388101983002832e-06, + "loss": 0.5699, + "step": 17944 + }, + { + "epoch": 6.63, + "learning_rate": 2.3387732479369383e-06, + "loss": 0.5254, + "step": 17945 + }, + { + "epoch": 6.63, + "learning_rate": 2.338736297573593e-06, + "loss": 0.4429, + "step": 17946 + }, + { + "epoch": 6.63, + "learning_rate": 2.3386993472102476e-06, + "loss": 0.3907, + "step": 17947 + }, + { + "epoch": 6.63, + "learning_rate": 2.338662396846902e-06, + "loss": 0.5642, + "step": 17948 + }, + { + "epoch": 6.63, + "learning_rate": 2.3386254464835573e-06, + "loss": 0.3791, + "step": 17949 + }, + { + "epoch": 6.63, + "learning_rate": 2.338588496120212e-06, + "loss": 0.49, + "step": 17950 + }, + { + "epoch": 6.63, + "learning_rate": 2.338551545756867e-06, + "loss": 0.3805, + "step": 17951 + }, + { + "epoch": 6.63, + "learning_rate": 2.3385145953935216e-06, + "loss": 0.4354, + "step": 17952 + }, + { + "epoch": 6.63, + "learning_rate": 2.3384776450301762e-06, + "loss": 0.3727, + "step": 17953 + }, + { + "epoch": 6.63, + "learning_rate": 2.338440694666831e-06, + "loss": 0.4333, + "step": 17954 + }, + { + "epoch": 6.63, + "learning_rate": 2.3384037443034855e-06, + "loss": 0.5652, + "step": 17955 + }, + { + "epoch": 6.63, + "learning_rate": 2.3383667939401406e-06, + "loss": 0.4687, + "step": 17956 + }, + { + "epoch": 6.63, + "learning_rate": 2.338329843576795e-06, + "loss": 0.334, + "step": 17957 + }, + { + "epoch": 6.63, + "learning_rate": 2.3382928932134503e-06, + "loss": 0.4186, + "step": 17958 + }, + { + "epoch": 6.63, + "learning_rate": 2.338255942850105e-06, + "loss": 0.4938, + "step": 17959 + }, + { + "epoch": 6.63, + "learning_rate": 2.3382189924867595e-06, + "loss": 0.4815, + "step": 17960 + }, + { + "epoch": 6.63, + "learning_rate": 2.338182042123414e-06, + "loss": 0.4985, + "step": 17961 + }, + { + "epoch": 6.63, + "learning_rate": 2.3381450917600692e-06, + "loss": 0.4437, + "step": 17962 + }, + { + "epoch": 6.63, + "learning_rate": 2.338108141396724e-06, + "loss": 0.4944, + "step": 17963 + }, + { + "epoch": 6.63, + "learning_rate": 2.3380711910333785e-06, + "loss": 0.3438, + "step": 17964 + }, + { + "epoch": 6.63, + "learning_rate": 2.3380342406700336e-06, + "loss": 0.4346, + "step": 17965 + }, + { + "epoch": 6.63, + "learning_rate": 2.3379972903066878e-06, + "loss": 0.4722, + "step": 17966 + }, + { + "epoch": 6.63, + "learning_rate": 2.337960339943343e-06, + "loss": 0.4334, + "step": 17967 + }, + { + "epoch": 6.64, + "learning_rate": 2.3379233895799975e-06, + "loss": 0.4109, + "step": 17968 + }, + { + "epoch": 6.64, + "learning_rate": 2.3378864392166525e-06, + "loss": 0.401, + "step": 17969 + }, + { + "epoch": 6.64, + "learning_rate": 2.337849488853307e-06, + "loss": 0.349, + "step": 17970 + }, + { + "epoch": 6.64, + "learning_rate": 2.337812538489962e-06, + "loss": 0.5137, + "step": 17971 + }, + { + "epoch": 6.64, + "learning_rate": 2.3377755881266164e-06, + "loss": 0.5119, + "step": 17972 + }, + { + "epoch": 6.64, + "learning_rate": 2.337738637763271e-06, + "loss": 0.5785, + "step": 17973 + }, + { + "epoch": 6.64, + "learning_rate": 2.337701687399926e-06, + "loss": 0.3765, + "step": 17974 + }, + { + "epoch": 6.64, + "learning_rate": 2.3376647370365808e-06, + "loss": 0.5065, + "step": 17975 + }, + { + "epoch": 6.64, + "learning_rate": 2.337627786673236e-06, + "loss": 0.5414, + "step": 17976 + }, + { + "epoch": 6.64, + "learning_rate": 2.3375908363098905e-06, + "loss": 0.4381, + "step": 17977 + }, + { + "epoch": 6.64, + "learning_rate": 2.337553885946545e-06, + "loss": 0.4471, + "step": 17978 + }, + { + "epoch": 6.64, + "learning_rate": 2.3375169355831997e-06, + "loss": 0.521, + "step": 17979 + }, + { + "epoch": 6.64, + "learning_rate": 2.337479985219855e-06, + "loss": 0.3325, + "step": 17980 + }, + { + "epoch": 6.64, + "learning_rate": 2.3374430348565094e-06, + "loss": 0.4572, + "step": 17981 + }, + { + "epoch": 6.64, + "learning_rate": 2.3374060844931645e-06, + "loss": 0.4341, + "step": 17982 + }, + { + "epoch": 6.64, + "learning_rate": 2.337369134129819e-06, + "loss": 0.4366, + "step": 17983 + }, + { + "epoch": 6.64, + "learning_rate": 2.3373321837664738e-06, + "loss": 0.4756, + "step": 17984 + }, + { + "epoch": 6.64, + "learning_rate": 2.3372952334031284e-06, + "loss": 0.428, + "step": 17985 + }, + { + "epoch": 6.64, + "learning_rate": 2.337258283039783e-06, + "loss": 0.4883, + "step": 17986 + }, + { + "epoch": 6.64, + "learning_rate": 2.337221332676438e-06, + "loss": 0.4673, + "step": 17987 + }, + { + "epoch": 6.64, + "learning_rate": 2.3371843823130927e-06, + "loss": 0.4145, + "step": 17988 + }, + { + "epoch": 6.64, + "learning_rate": 2.337147431949748e-06, + "loss": 0.4509, + "step": 17989 + }, + { + "epoch": 6.64, + "learning_rate": 2.3371104815864024e-06, + "loss": 0.4021, + "step": 17990 + }, + { + "epoch": 6.64, + "learning_rate": 2.337073531223057e-06, + "loss": 0.4102, + "step": 17991 + }, + { + "epoch": 6.64, + "learning_rate": 2.3370365808597117e-06, + "loss": 0.3683, + "step": 17992 + }, + { + "epoch": 6.64, + "learning_rate": 2.3369996304963663e-06, + "loss": 0.5611, + "step": 17993 + }, + { + "epoch": 6.64, + "learning_rate": 2.3369626801330214e-06, + "loss": 0.3941, + "step": 17994 + }, + { + "epoch": 6.65, + "learning_rate": 2.336925729769676e-06, + "loss": 0.414, + "step": 17995 + }, + { + "epoch": 6.65, + "learning_rate": 2.336888779406331e-06, + "loss": 0.5463, + "step": 17996 + }, + { + "epoch": 6.65, + "learning_rate": 2.3368518290429857e-06, + "loss": 0.6064, + "step": 17997 + }, + { + "epoch": 6.65, + "learning_rate": 2.3368148786796404e-06, + "loss": 0.3442, + "step": 17998 + }, + { + "epoch": 6.65, + "learning_rate": 2.336777928316295e-06, + "loss": 0.5178, + "step": 17999 + }, + { + "epoch": 6.65, + "learning_rate": 2.33674097795295e-06, + "loss": 0.5367, + "step": 18000 + }, + { + "epoch": 6.65, + "learning_rate": 2.3367040275896047e-06, + "loss": 0.5137, + "step": 18001 + }, + { + "epoch": 6.65, + "learning_rate": 2.3366670772262598e-06, + "loss": 0.4851, + "step": 18002 + }, + { + "epoch": 6.65, + "learning_rate": 2.3366301268629144e-06, + "loss": 0.4763, + "step": 18003 + }, + { + "epoch": 6.65, + "learning_rate": 2.3365931764995686e-06, + "loss": 0.4318, + "step": 18004 + }, + { + "epoch": 6.65, + "learning_rate": 2.3365562261362237e-06, + "loss": 0.3709, + "step": 18005 + }, + { + "epoch": 6.65, + "learning_rate": 2.3365192757728783e-06, + "loss": 0.4408, + "step": 18006 + }, + { + "epoch": 6.65, + "learning_rate": 2.3364823254095334e-06, + "loss": 0.3667, + "step": 18007 + }, + { + "epoch": 6.65, + "learning_rate": 2.336445375046188e-06, + "loss": 0.5035, + "step": 18008 + }, + { + "epoch": 6.65, + "learning_rate": 2.336408424682843e-06, + "loss": 0.4206, + "step": 18009 + }, + { + "epoch": 6.65, + "learning_rate": 2.3363714743194973e-06, + "loss": 0.5705, + "step": 18010 + }, + { + "epoch": 6.65, + "learning_rate": 2.3363345239561523e-06, + "loss": 0.4499, + "step": 18011 + }, + { + "epoch": 6.65, + "learning_rate": 2.336297573592807e-06, + "loss": 0.5256, + "step": 18012 + }, + { + "epoch": 6.65, + "learning_rate": 2.3362606232294616e-06, + "loss": 0.4024, + "step": 18013 + }, + { + "epoch": 6.65, + "learning_rate": 2.3362236728661167e-06, + "loss": 0.3533, + "step": 18014 + }, + { + "epoch": 6.65, + "learning_rate": 2.3361867225027713e-06, + "loss": 0.4196, + "step": 18015 + }, + { + "epoch": 6.65, + "learning_rate": 2.336149772139426e-06, + "loss": 0.4966, + "step": 18016 + }, + { + "epoch": 6.65, + "learning_rate": 2.3361128217760806e-06, + "loss": 0.4604, + "step": 18017 + }, + { + "epoch": 6.65, + "learning_rate": 2.3360758714127356e-06, + "loss": 0.4072, + "step": 18018 + }, + { + "epoch": 6.65, + "learning_rate": 2.3360389210493903e-06, + "loss": 0.4754, + "step": 18019 + }, + { + "epoch": 6.65, + "learning_rate": 2.3360019706860453e-06, + "loss": 0.4017, + "step": 18020 + }, + { + "epoch": 6.65, + "learning_rate": 2.3359650203227e-06, + "loss": 0.5241, + "step": 18021 + }, + { + "epoch": 6.66, + "learning_rate": 2.3359280699593546e-06, + "loss": 0.34, + "step": 18022 + }, + { + "epoch": 6.66, + "learning_rate": 2.3358911195960093e-06, + "loss": 0.4594, + "step": 18023 + }, + { + "epoch": 6.66, + "learning_rate": 2.335854169232664e-06, + "loss": 0.5465, + "step": 18024 + }, + { + "epoch": 6.66, + "learning_rate": 2.335817218869319e-06, + "loss": 0.5813, + "step": 18025 + }, + { + "epoch": 6.66, + "learning_rate": 2.3357802685059736e-06, + "loss": 0.4495, + "step": 18026 + }, + { + "epoch": 6.66, + "learning_rate": 2.3357433181426286e-06, + "loss": 0.3304, + "step": 18027 + }, + { + "epoch": 6.66, + "learning_rate": 2.3357063677792833e-06, + "loss": 0.4824, + "step": 18028 + }, + { + "epoch": 6.66, + "learning_rate": 2.335669417415938e-06, + "loss": 0.4722, + "step": 18029 + }, + { + "epoch": 6.66, + "learning_rate": 2.3356324670525926e-06, + "loss": 0.3961, + "step": 18030 + }, + { + "epoch": 6.66, + "learning_rate": 2.3355955166892476e-06, + "loss": 0.4991, + "step": 18031 + }, + { + "epoch": 6.66, + "learning_rate": 2.3355585663259022e-06, + "loss": 0.4494, + "step": 18032 + }, + { + "epoch": 6.66, + "learning_rate": 2.335521615962557e-06, + "loss": 0.4809, + "step": 18033 + }, + { + "epoch": 6.66, + "learning_rate": 2.335484665599212e-06, + "loss": 0.4673, + "step": 18034 + }, + { + "epoch": 6.66, + "learning_rate": 2.3354477152358666e-06, + "loss": 0.491, + "step": 18035 + }, + { + "epoch": 6.66, + "learning_rate": 2.3354107648725212e-06, + "loss": 0.3779, + "step": 18036 + }, + { + "epoch": 6.66, + "learning_rate": 2.335373814509176e-06, + "loss": 0.4366, + "step": 18037 + }, + { + "epoch": 6.66, + "learning_rate": 2.335336864145831e-06, + "loss": 0.3942, + "step": 18038 + }, + { + "epoch": 6.66, + "learning_rate": 2.3352999137824856e-06, + "loss": 0.3018, + "step": 18039 + }, + { + "epoch": 6.66, + "learning_rate": 2.3352629634191406e-06, + "loss": 0.3981, + "step": 18040 + }, + { + "epoch": 6.66, + "learning_rate": 2.3352260130557952e-06, + "loss": 0.4937, + "step": 18041 + }, + { + "epoch": 6.66, + "learning_rate": 2.33518906269245e-06, + "loss": 0.5726, + "step": 18042 + }, + { + "epoch": 6.66, + "learning_rate": 2.3351521123291045e-06, + "loss": 0.4971, + "step": 18043 + }, + { + "epoch": 6.66, + "learning_rate": 2.335115161965759e-06, + "loss": 0.6565, + "step": 18044 + }, + { + "epoch": 6.66, + "learning_rate": 2.3350782116024142e-06, + "loss": 0.4312, + "step": 18045 + }, + { + "epoch": 6.66, + "learning_rate": 2.335041261239069e-06, + "loss": 0.3528, + "step": 18046 + }, + { + "epoch": 6.66, + "learning_rate": 2.335004310875724e-06, + "loss": 0.6103, + "step": 18047 + }, + { + "epoch": 6.66, + "learning_rate": 2.3349673605123786e-06, + "loss": 0.4644, + "step": 18048 + }, + { + "epoch": 6.67, + "learning_rate": 2.334930410149033e-06, + "loss": 0.4753, + "step": 18049 + }, + { + "epoch": 6.67, + "learning_rate": 2.334893459785688e-06, + "loss": 0.5222, + "step": 18050 + }, + { + "epoch": 6.67, + "learning_rate": 2.334856509422343e-06, + "loss": 0.3924, + "step": 18051 + }, + { + "epoch": 6.67, + "learning_rate": 2.3348195590589975e-06, + "loss": 0.5846, + "step": 18052 + }, + { + "epoch": 6.67, + "learning_rate": 2.334782608695652e-06, + "loss": 0.3945, + "step": 18053 + }, + { + "epoch": 6.67, + "learning_rate": 2.3347456583323072e-06, + "loss": 0.5146, + "step": 18054 + }, + { + "epoch": 6.67, + "learning_rate": 2.3347087079689614e-06, + "loss": 0.3961, + "step": 18055 + }, + { + "epoch": 6.67, + "learning_rate": 2.3346717576056165e-06, + "loss": 0.531, + "step": 18056 + }, + { + "epoch": 6.67, + "learning_rate": 2.334634807242271e-06, + "loss": 0.5055, + "step": 18057 + }, + { + "epoch": 6.67, + "learning_rate": 2.334597856878926e-06, + "loss": 0.4495, + "step": 18058 + }, + { + "epoch": 6.67, + "learning_rate": 2.334560906515581e-06, + "loss": 0.4656, + "step": 18059 + }, + { + "epoch": 6.67, + "learning_rate": 2.334523956152236e-06, + "loss": 0.5423, + "step": 18060 + }, + { + "epoch": 6.67, + "learning_rate": 2.33448700578889e-06, + "loss": 0.4597, + "step": 18061 + }, + { + "epoch": 6.67, + "learning_rate": 2.334450055425545e-06, + "loss": 0.4427, + "step": 18062 + }, + { + "epoch": 6.67, + "learning_rate": 2.3344131050622e-06, + "loss": 0.4677, + "step": 18063 + }, + { + "epoch": 6.67, + "learning_rate": 2.3343761546988544e-06, + "loss": 0.5108, + "step": 18064 + }, + { + "epoch": 6.67, + "learning_rate": 2.3343392043355095e-06, + "loss": 0.4468, + "step": 18065 + }, + { + "epoch": 6.67, + "learning_rate": 2.334302253972164e-06, + "loss": 0.5456, + "step": 18066 + }, + { + "epoch": 6.67, + "learning_rate": 2.3342653036088188e-06, + "loss": 0.5101, + "step": 18067 + }, + { + "epoch": 6.67, + "learning_rate": 2.3342283532454734e-06, + "loss": 0.5705, + "step": 18068 + }, + { + "epoch": 6.67, + "learning_rate": 2.3341914028821285e-06, + "loss": 0.4759, + "step": 18069 + }, + { + "epoch": 6.67, + "learning_rate": 2.334154452518783e-06, + "loss": 0.431, + "step": 18070 + }, + { + "epoch": 6.67, + "learning_rate": 2.334117502155438e-06, + "loss": 0.4537, + "step": 18071 + }, + { + "epoch": 6.67, + "learning_rate": 2.334080551792093e-06, + "loss": 0.4662, + "step": 18072 + }, + { + "epoch": 6.67, + "learning_rate": 2.3340436014287474e-06, + "loss": 0.4339, + "step": 18073 + }, + { + "epoch": 6.67, + "learning_rate": 2.334006651065402e-06, + "loss": 0.3824, + "step": 18074 + }, + { + "epoch": 6.67, + "learning_rate": 2.3339697007020567e-06, + "loss": 0.5125, + "step": 18075 + }, + { + "epoch": 6.68, + "learning_rate": 2.3339327503387118e-06, + "loss": 0.5318, + "step": 18076 + }, + { + "epoch": 6.68, + "learning_rate": 2.3338957999753664e-06, + "loss": 0.4507, + "step": 18077 + }, + { + "epoch": 6.68, + "learning_rate": 2.3338588496120215e-06, + "loss": 0.4775, + "step": 18078 + }, + { + "epoch": 6.68, + "learning_rate": 2.333821899248676e-06, + "loss": 0.3805, + "step": 18079 + }, + { + "epoch": 6.68, + "learning_rate": 2.3337849488853307e-06, + "loss": 0.4063, + "step": 18080 + }, + { + "epoch": 6.68, + "learning_rate": 2.3337479985219854e-06, + "loss": 0.5243, + "step": 18081 + }, + { + "epoch": 6.68, + "learning_rate": 2.3337110481586404e-06, + "loss": 0.4781, + "step": 18082 + }, + { + "epoch": 6.68, + "learning_rate": 2.333674097795295e-06, + "loss": 0.7039, + "step": 18083 + }, + { + "epoch": 6.68, + "learning_rate": 2.3336371474319497e-06, + "loss": 0.4734, + "step": 18084 + }, + { + "epoch": 6.68, + "learning_rate": 2.3336001970686048e-06, + "loss": 0.4888, + "step": 18085 + }, + { + "epoch": 6.68, + "learning_rate": 2.3335632467052594e-06, + "loss": 0.5075, + "step": 18086 + }, + { + "epoch": 6.68, + "learning_rate": 2.333526296341914e-06, + "loss": 0.4753, + "step": 18087 + }, + { + "epoch": 6.68, + "learning_rate": 2.3334893459785687e-06, + "loss": 0.4669, + "step": 18088 + }, + { + "epoch": 6.68, + "learning_rate": 2.3334523956152237e-06, + "loss": 0.4662, + "step": 18089 + }, + { + "epoch": 6.68, + "learning_rate": 2.3334154452518784e-06, + "loss": 0.4311, + "step": 18090 + }, + { + "epoch": 6.68, + "learning_rate": 2.3333784948885334e-06, + "loss": 0.397, + "step": 18091 + }, + { + "epoch": 6.68, + "learning_rate": 2.333341544525188e-06, + "loss": 0.3748, + "step": 18092 + }, + { + "epoch": 6.68, + "learning_rate": 2.3333045941618427e-06, + "loss": 0.2937, + "step": 18093 + }, + { + "epoch": 6.68, + "learning_rate": 2.3332676437984973e-06, + "loss": 0.3961, + "step": 18094 + }, + { + "epoch": 6.68, + "learning_rate": 2.333230693435152e-06, + "loss": 0.4106, + "step": 18095 + }, + { + "epoch": 6.68, + "learning_rate": 2.333193743071807e-06, + "loss": 0.5298, + "step": 18096 + }, + { + "epoch": 6.68, + "learning_rate": 2.3331567927084617e-06, + "loss": 0.4764, + "step": 18097 + }, + { + "epoch": 6.68, + "learning_rate": 2.3331198423451167e-06, + "loss": 0.458, + "step": 18098 + }, + { + "epoch": 6.68, + "learning_rate": 2.3330828919817714e-06, + "loss": 0.5908, + "step": 18099 + }, + { + "epoch": 6.68, + "learning_rate": 2.333045941618426e-06, + "loss": 0.462, + "step": 18100 + }, + { + "epoch": 6.68, + "learning_rate": 2.3330089912550806e-06, + "loss": 0.3524, + "step": 18101 + }, + { + "epoch": 6.68, + "learning_rate": 2.3329720408917357e-06, + "loss": 0.4233, + "step": 18102 + }, + { + "epoch": 6.69, + "learning_rate": 2.3329350905283903e-06, + "loss": 0.4801, + "step": 18103 + }, + { + "epoch": 6.69, + "learning_rate": 2.332898140165045e-06, + "loss": 0.3489, + "step": 18104 + }, + { + "epoch": 6.69, + "learning_rate": 2.3328611898017e-06, + "loss": 0.4382, + "step": 18105 + }, + { + "epoch": 6.69, + "learning_rate": 2.3328242394383542e-06, + "loss": 0.3605, + "step": 18106 + }, + { + "epoch": 6.69, + "learning_rate": 2.3327872890750093e-06, + "loss": 0.391, + "step": 18107 + }, + { + "epoch": 6.69, + "learning_rate": 2.332750338711664e-06, + "loss": 0.3279, + "step": 18108 + }, + { + "epoch": 6.69, + "learning_rate": 2.332713388348319e-06, + "loss": 0.5035, + "step": 18109 + }, + { + "epoch": 6.69, + "learning_rate": 2.3326764379849736e-06, + "loss": 0.3499, + "step": 18110 + }, + { + "epoch": 6.69, + "learning_rate": 2.3326394876216287e-06, + "loss": 0.3969, + "step": 18111 + }, + { + "epoch": 6.69, + "learning_rate": 2.332602537258283e-06, + "loss": 0.4729, + "step": 18112 + }, + { + "epoch": 6.69, + "learning_rate": 2.3325655868949375e-06, + "loss": 0.4199, + "step": 18113 + }, + { + "epoch": 6.69, + "learning_rate": 2.3325286365315926e-06, + "loss": 0.55, + "step": 18114 + }, + { + "epoch": 6.69, + "learning_rate": 2.3324916861682472e-06, + "loss": 0.4471, + "step": 18115 + }, + { + "epoch": 6.69, + "learning_rate": 2.3324547358049023e-06, + "loss": 0.4201, + "step": 18116 + }, + { + "epoch": 6.69, + "learning_rate": 2.332417785441557e-06, + "loss": 0.3663, + "step": 18117 + }, + { + "epoch": 6.69, + "learning_rate": 2.3323808350782116e-06, + "loss": 0.4865, + "step": 18118 + }, + { + "epoch": 6.69, + "learning_rate": 2.332343884714866e-06, + "loss": 0.3599, + "step": 18119 + }, + { + "epoch": 6.69, + "learning_rate": 2.3323069343515213e-06, + "loss": 0.4508, + "step": 18120 + }, + { + "epoch": 6.69, + "learning_rate": 2.332269983988176e-06, + "loss": 0.4901, + "step": 18121 + }, + { + "epoch": 6.69, + "learning_rate": 2.332233033624831e-06, + "loss": 0.3521, + "step": 18122 + }, + { + "epoch": 6.69, + "learning_rate": 2.3321960832614856e-06, + "loss": 0.4242, + "step": 18123 + }, + { + "epoch": 6.69, + "learning_rate": 2.3321591328981402e-06, + "loss": 0.4832, + "step": 18124 + }, + { + "epoch": 6.69, + "learning_rate": 2.332122182534795e-06, + "loss": 0.4567, + "step": 18125 + }, + { + "epoch": 6.69, + "learning_rate": 2.3320852321714495e-06, + "loss": 0.3718, + "step": 18126 + }, + { + "epoch": 6.69, + "learning_rate": 2.3320482818081046e-06, + "loss": 0.3554, + "step": 18127 + }, + { + "epoch": 6.69, + "learning_rate": 2.332011331444759e-06, + "loss": 0.4483, + "step": 18128 + }, + { + "epoch": 6.69, + "learning_rate": 2.3319743810814143e-06, + "loss": 0.5366, + "step": 18129 + }, + { + "epoch": 6.69, + "learning_rate": 2.331937430718069e-06, + "loss": 0.4993, + "step": 18130 + }, + { + "epoch": 6.7, + "learning_rate": 2.3319004803547235e-06, + "loss": 0.4373, + "step": 18131 + }, + { + "epoch": 6.7, + "learning_rate": 2.331863529991378e-06, + "loss": 0.4832, + "step": 18132 + }, + { + "epoch": 6.7, + "learning_rate": 2.331826579628033e-06, + "loss": 0.5121, + "step": 18133 + }, + { + "epoch": 6.7, + "learning_rate": 2.331789629264688e-06, + "loss": 0.4773, + "step": 18134 + }, + { + "epoch": 6.7, + "learning_rate": 2.3317526789013425e-06, + "loss": 0.3531, + "step": 18135 + }, + { + "epoch": 6.7, + "learning_rate": 2.3317157285379976e-06, + "loss": 0.4655, + "step": 18136 + }, + { + "epoch": 6.7, + "learning_rate": 2.331678778174652e-06, + "loss": 0.4889, + "step": 18137 + }, + { + "epoch": 6.7, + "learning_rate": 2.331641827811307e-06, + "loss": 0.4297, + "step": 18138 + }, + { + "epoch": 6.7, + "learning_rate": 2.3316048774479615e-06, + "loss": 0.4583, + "step": 18139 + }, + { + "epoch": 6.7, + "learning_rate": 2.3315679270846165e-06, + "loss": 0.4291, + "step": 18140 + }, + { + "epoch": 6.7, + "learning_rate": 2.331530976721271e-06, + "loss": 0.4434, + "step": 18141 + }, + { + "epoch": 6.7, + "learning_rate": 2.3314940263579262e-06, + "loss": 0.3778, + "step": 18142 + }, + { + "epoch": 6.7, + "learning_rate": 2.331457075994581e-06, + "loss": 0.3733, + "step": 18143 + }, + { + "epoch": 6.7, + "learning_rate": 2.3314201256312355e-06, + "loss": 0.505, + "step": 18144 + }, + { + "epoch": 6.7, + "learning_rate": 2.33138317526789e-06, + "loss": 0.5028, + "step": 18145 + }, + { + "epoch": 6.7, + "learning_rate": 2.3313462249045448e-06, + "loss": 0.3873, + "step": 18146 + }, + { + "epoch": 6.7, + "learning_rate": 2.3313092745412e-06, + "loss": 0.3938, + "step": 18147 + }, + { + "epoch": 6.7, + "learning_rate": 2.3312723241778545e-06, + "loss": 0.3849, + "step": 18148 + }, + { + "epoch": 6.7, + "learning_rate": 2.3312353738145095e-06, + "loss": 0.4481, + "step": 18149 + }, + { + "epoch": 6.7, + "learning_rate": 2.331198423451164e-06, + "loss": 0.4272, + "step": 18150 + }, + { + "epoch": 6.7, + "learning_rate": 2.331161473087819e-06, + "loss": 0.4131, + "step": 18151 + }, + { + "epoch": 6.7, + "learning_rate": 2.3311245227244734e-06, + "loss": 0.5786, + "step": 18152 + }, + { + "epoch": 6.7, + "learning_rate": 2.331087572361128e-06, + "loss": 0.4457, + "step": 18153 + }, + { + "epoch": 6.7, + "learning_rate": 2.331050621997783e-06, + "loss": 0.5037, + "step": 18154 + }, + { + "epoch": 6.7, + "learning_rate": 2.3310136716344378e-06, + "loss": 0.3862, + "step": 18155 + }, + { + "epoch": 6.7, + "learning_rate": 2.3309767212710924e-06, + "loss": 0.491, + "step": 18156 + }, + { + "epoch": 6.7, + "learning_rate": 2.330939770907747e-06, + "loss": 0.4542, + "step": 18157 + }, + { + "epoch": 6.71, + "learning_rate": 2.330902820544402e-06, + "loss": 0.4783, + "step": 18158 + }, + { + "epoch": 6.71, + "learning_rate": 2.3308658701810567e-06, + "loss": 0.5502, + "step": 18159 + }, + { + "epoch": 6.71, + "learning_rate": 2.330828919817712e-06, + "loss": 0.4244, + "step": 18160 + }, + { + "epoch": 6.71, + "learning_rate": 2.3307919694543664e-06, + "loss": 0.4606, + "step": 18161 + }, + { + "epoch": 6.71, + "learning_rate": 2.330755019091021e-06, + "loss": 0.5607, + "step": 18162 + }, + { + "epoch": 6.71, + "learning_rate": 2.3307180687276757e-06, + "loss": 0.4425, + "step": 18163 + }, + { + "epoch": 6.71, + "learning_rate": 2.3306811183643304e-06, + "loss": 0.4735, + "step": 18164 + }, + { + "epoch": 6.71, + "learning_rate": 2.3306441680009854e-06, + "loss": 0.3718, + "step": 18165 + }, + { + "epoch": 6.71, + "learning_rate": 2.33060721763764e-06, + "loss": 0.324, + "step": 18166 + }, + { + "epoch": 6.71, + "learning_rate": 2.330570267274295e-06, + "loss": 0.379, + "step": 18167 + }, + { + "epoch": 6.71, + "learning_rate": 2.3305333169109497e-06, + "loss": 0.3883, + "step": 18168 + }, + { + "epoch": 6.71, + "learning_rate": 2.3304963665476044e-06, + "loss": 0.3269, + "step": 18169 + }, + { + "epoch": 6.71, + "learning_rate": 2.330459416184259e-06, + "loss": 0.4588, + "step": 18170 + }, + { + "epoch": 6.71, + "learning_rate": 2.330422465820914e-06, + "loss": 0.3743, + "step": 18171 + }, + { + "epoch": 6.71, + "learning_rate": 2.3303855154575687e-06, + "loss": 0.358, + "step": 18172 + }, + { + "epoch": 6.71, + "learning_rate": 2.3303485650942234e-06, + "loss": 0.4025, + "step": 18173 + }, + { + "epoch": 6.71, + "learning_rate": 2.3303116147308784e-06, + "loss": 0.4416, + "step": 18174 + }, + { + "epoch": 6.71, + "learning_rate": 2.330274664367533e-06, + "loss": 0.5214, + "step": 18175 + }, + { + "epoch": 6.71, + "learning_rate": 2.3302377140041877e-06, + "loss": 0.6426, + "step": 18176 + }, + { + "epoch": 6.71, + "learning_rate": 2.3302007636408423e-06, + "loss": 0.4672, + "step": 18177 + }, + { + "epoch": 6.71, + "learning_rate": 2.3301638132774974e-06, + "loss": 0.5724, + "step": 18178 + }, + { + "epoch": 6.71, + "learning_rate": 2.330126862914152e-06, + "loss": 0.3427, + "step": 18179 + }, + { + "epoch": 6.71, + "learning_rate": 2.330089912550807e-06, + "loss": 0.3958, + "step": 18180 + }, + { + "epoch": 6.71, + "learning_rate": 2.3300529621874617e-06, + "loss": 0.5454, + "step": 18181 + }, + { + "epoch": 6.71, + "learning_rate": 2.3300160118241163e-06, + "loss": 0.5863, + "step": 18182 + }, + { + "epoch": 6.71, + "learning_rate": 2.329979061460771e-06, + "loss": 0.3939, + "step": 18183 + }, + { + "epoch": 6.71, + "learning_rate": 2.3299421110974256e-06, + "loss": 0.3984, + "step": 18184 + }, + { + "epoch": 6.72, + "learning_rate": 2.3299051607340807e-06, + "loss": 0.4798, + "step": 18185 + }, + { + "epoch": 6.72, + "learning_rate": 2.3298682103707353e-06, + "loss": 0.4962, + "step": 18186 + }, + { + "epoch": 6.72, + "learning_rate": 2.3298312600073904e-06, + "loss": 0.5075, + "step": 18187 + }, + { + "epoch": 6.72, + "learning_rate": 2.329794309644045e-06, + "loss": 0.5051, + "step": 18188 + }, + { + "epoch": 6.72, + "learning_rate": 2.3297573592806997e-06, + "loss": 0.485, + "step": 18189 + }, + { + "epoch": 6.72, + "learning_rate": 2.3297204089173543e-06, + "loss": 0.4185, + "step": 18190 + }, + { + "epoch": 6.72, + "learning_rate": 2.3296834585540093e-06, + "loss": 0.3924, + "step": 18191 + }, + { + "epoch": 6.72, + "learning_rate": 2.329646508190664e-06, + "loss": 0.4762, + "step": 18192 + }, + { + "epoch": 6.72, + "learning_rate": 2.3296095578273186e-06, + "loss": 0.4019, + "step": 18193 + }, + { + "epoch": 6.72, + "learning_rate": 2.3295726074639737e-06, + "loss": 0.3691, + "step": 18194 + }, + { + "epoch": 6.72, + "learning_rate": 2.329535657100628e-06, + "loss": 0.4243, + "step": 18195 + }, + { + "epoch": 6.72, + "learning_rate": 2.329498706737283e-06, + "loss": 0.4876, + "step": 18196 + }, + { + "epoch": 6.72, + "learning_rate": 2.3294617563739376e-06, + "loss": 0.4738, + "step": 18197 + }, + { + "epoch": 6.72, + "learning_rate": 2.3294248060105926e-06, + "loss": 0.402, + "step": 18198 + }, + { + "epoch": 6.72, + "learning_rate": 2.3293878556472473e-06, + "loss": 0.2973, + "step": 18199 + }, + { + "epoch": 6.72, + "learning_rate": 2.3293509052839023e-06, + "loss": 0.5099, + "step": 18200 + }, + { + "epoch": 6.72, + "learning_rate": 2.3293139549205566e-06, + "loss": 0.434, + "step": 18201 + }, + { + "epoch": 6.72, + "learning_rate": 2.329277004557211e-06, + "loss": 0.4623, + "step": 18202 + }, + { + "epoch": 6.72, + "learning_rate": 2.3292400541938663e-06, + "loss": 0.5033, + "step": 18203 + }, + { + "epoch": 6.72, + "learning_rate": 2.329203103830521e-06, + "loss": 0.5688, + "step": 18204 + }, + { + "epoch": 6.72, + "learning_rate": 2.329166153467176e-06, + "loss": 0.3114, + "step": 18205 + }, + { + "epoch": 6.72, + "learning_rate": 2.3291292031038306e-06, + "loss": 0.431, + "step": 18206 + }, + { + "epoch": 6.72, + "learning_rate": 2.3290922527404852e-06, + "loss": 0.4323, + "step": 18207 + }, + { + "epoch": 6.72, + "learning_rate": 2.32905530237714e-06, + "loss": 0.3951, + "step": 18208 + }, + { + "epoch": 6.72, + "learning_rate": 2.329018352013795e-06, + "loss": 0.4791, + "step": 18209 + }, + { + "epoch": 6.72, + "learning_rate": 2.3289814016504496e-06, + "loss": 0.4151, + "step": 18210 + }, + { + "epoch": 6.72, + "learning_rate": 2.3289444512871046e-06, + "loss": 0.4515, + "step": 18211 + }, + { + "epoch": 6.73, + "learning_rate": 2.3289075009237593e-06, + "loss": 0.4308, + "step": 18212 + }, + { + "epoch": 6.73, + "learning_rate": 2.328870550560414e-06, + "loss": 0.3994, + "step": 18213 + }, + { + "epoch": 6.73, + "learning_rate": 2.3288336001970685e-06, + "loss": 0.402, + "step": 18214 + }, + { + "epoch": 6.73, + "learning_rate": 2.328796649833723e-06, + "loss": 0.579, + "step": 18215 + }, + { + "epoch": 6.73, + "learning_rate": 2.3287596994703782e-06, + "loss": 0.4058, + "step": 18216 + }, + { + "epoch": 6.73, + "learning_rate": 2.328722749107033e-06, + "loss": 0.4739, + "step": 18217 + }, + { + "epoch": 6.73, + "learning_rate": 2.328685798743688e-06, + "loss": 0.419, + "step": 18218 + }, + { + "epoch": 6.73, + "learning_rate": 2.3286488483803426e-06, + "loss": 0.4968, + "step": 18219 + }, + { + "epoch": 6.73, + "learning_rate": 2.328611898016997e-06, + "loss": 0.5012, + "step": 18220 + }, + { + "epoch": 6.73, + "learning_rate": 2.328574947653652e-06, + "loss": 0.5132, + "step": 18221 + }, + { + "epoch": 6.73, + "learning_rate": 2.3285379972903065e-06, + "loss": 0.3843, + "step": 18222 + }, + { + "epoch": 6.73, + "learning_rate": 2.3285010469269615e-06, + "loss": 0.4044, + "step": 18223 + }, + { + "epoch": 6.73, + "learning_rate": 2.328464096563616e-06, + "loss": 0.5329, + "step": 18224 + }, + { + "epoch": 6.73, + "learning_rate": 2.3284271462002712e-06, + "loss": 0.4719, + "step": 18225 + }, + { + "epoch": 6.73, + "learning_rate": 2.328390195836926e-06, + "loss": 0.4131, + "step": 18226 + }, + { + "epoch": 6.73, + "learning_rate": 2.3283532454735805e-06, + "loss": 0.3301, + "step": 18227 + }, + { + "epoch": 6.73, + "learning_rate": 2.328316295110235e-06, + "loss": 0.3859, + "step": 18228 + }, + { + "epoch": 6.73, + "learning_rate": 2.32827934474689e-06, + "loss": 0.5237, + "step": 18229 + }, + { + "epoch": 6.73, + "learning_rate": 2.328242394383545e-06, + "loss": 0.363, + "step": 18230 + }, + { + "epoch": 6.73, + "learning_rate": 2.3282054440202e-06, + "loss": 0.4013, + "step": 18231 + }, + { + "epoch": 6.73, + "learning_rate": 2.3281684936568545e-06, + "loss": 0.496, + "step": 18232 + }, + { + "epoch": 6.73, + "learning_rate": 2.328131543293509e-06, + "loss": 0.4169, + "step": 18233 + }, + { + "epoch": 6.73, + "learning_rate": 2.328094592930164e-06, + "loss": 0.5352, + "step": 18234 + }, + { + "epoch": 6.73, + "learning_rate": 2.3280576425668184e-06, + "loss": 0.3712, + "step": 18235 + }, + { + "epoch": 6.73, + "learning_rate": 2.3280206922034735e-06, + "loss": 0.3607, + "step": 18236 + }, + { + "epoch": 6.73, + "learning_rate": 2.327983741840128e-06, + "loss": 0.4205, + "step": 18237 + }, + { + "epoch": 6.73, + "learning_rate": 2.327946791476783e-06, + "loss": 0.5328, + "step": 18238 + }, + { + "epoch": 6.74, + "learning_rate": 2.327909841113438e-06, + "loss": 0.4408, + "step": 18239 + }, + { + "epoch": 6.74, + "learning_rate": 2.3278728907500925e-06, + "loss": 0.4293, + "step": 18240 + }, + { + "epoch": 6.74, + "learning_rate": 2.327835940386747e-06, + "loss": 0.536, + "step": 18241 + }, + { + "epoch": 6.74, + "learning_rate": 2.3277989900234017e-06, + "loss": 0.4743, + "step": 18242 + }, + { + "epoch": 6.74, + "learning_rate": 2.327762039660057e-06, + "loss": 0.5309, + "step": 18243 + }, + { + "epoch": 6.74, + "learning_rate": 2.3277250892967114e-06, + "loss": 0.5176, + "step": 18244 + }, + { + "epoch": 6.74, + "learning_rate": 2.3276881389333665e-06, + "loss": 0.3155, + "step": 18245 + }, + { + "epoch": 6.74, + "learning_rate": 2.3276511885700207e-06, + "loss": 0.4767, + "step": 18246 + }, + { + "epoch": 6.74, + "learning_rate": 2.3276142382066758e-06, + "loss": 0.5949, + "step": 18247 + }, + { + "epoch": 6.74, + "learning_rate": 2.3275772878433304e-06, + "loss": 0.5349, + "step": 18248 + }, + { + "epoch": 6.74, + "learning_rate": 2.3275403374799855e-06, + "loss": 0.4909, + "step": 18249 + }, + { + "epoch": 6.74, + "learning_rate": 2.32750338711664e-06, + "loss": 0.4711, + "step": 18250 + }, + { + "epoch": 6.74, + "learning_rate": 2.327466436753295e-06, + "loss": 0.3723, + "step": 18251 + }, + { + "epoch": 6.74, + "learning_rate": 2.3274294863899494e-06, + "loss": 0.5113, + "step": 18252 + }, + { + "epoch": 6.74, + "learning_rate": 2.327392536026604e-06, + "loss": 0.5688, + "step": 18253 + }, + { + "epoch": 6.74, + "learning_rate": 2.327355585663259e-06, + "loss": 0.5084, + "step": 18254 + }, + { + "epoch": 6.74, + "learning_rate": 2.3273186352999137e-06, + "loss": 0.3385, + "step": 18255 + }, + { + "epoch": 6.74, + "learning_rate": 2.3272816849365688e-06, + "loss": 0.4551, + "step": 18256 + }, + { + "epoch": 6.74, + "learning_rate": 2.3272447345732234e-06, + "loss": 0.5259, + "step": 18257 + }, + { + "epoch": 6.74, + "learning_rate": 2.327207784209878e-06, + "loss": 0.4259, + "step": 18258 + }, + { + "epoch": 6.74, + "learning_rate": 2.3271708338465327e-06, + "loss": 0.2976, + "step": 18259 + }, + { + "epoch": 6.74, + "learning_rate": 2.3271338834831877e-06, + "loss": 0.4169, + "step": 18260 + }, + { + "epoch": 6.74, + "learning_rate": 2.3270969331198424e-06, + "loss": 0.4459, + "step": 18261 + }, + { + "epoch": 6.74, + "learning_rate": 2.327059982756497e-06, + "loss": 0.5242, + "step": 18262 + }, + { + "epoch": 6.74, + "learning_rate": 2.327023032393152e-06, + "loss": 0.5989, + "step": 18263 + }, + { + "epoch": 6.74, + "learning_rate": 2.3269860820298067e-06, + "loss": 0.4553, + "step": 18264 + }, + { + "epoch": 6.74, + "learning_rate": 2.3269491316664613e-06, + "loss": 0.5152, + "step": 18265 + }, + { + "epoch": 6.75, + "learning_rate": 2.326912181303116e-06, + "loss": 0.401, + "step": 18266 + }, + { + "epoch": 6.75, + "learning_rate": 2.326875230939771e-06, + "loss": 0.4938, + "step": 18267 + }, + { + "epoch": 6.75, + "learning_rate": 2.3268382805764257e-06, + "loss": 0.3713, + "step": 18268 + }, + { + "epoch": 6.75, + "learning_rate": 2.3268013302130807e-06, + "loss": 0.51, + "step": 18269 + }, + { + "epoch": 6.75, + "learning_rate": 2.3267643798497354e-06, + "loss": 0.4706, + "step": 18270 + }, + { + "epoch": 6.75, + "learning_rate": 2.32672742948639e-06, + "loss": 0.4814, + "step": 18271 + }, + { + "epoch": 6.75, + "learning_rate": 2.3266904791230446e-06, + "loss": 0.5581, + "step": 18272 + }, + { + "epoch": 6.75, + "learning_rate": 2.3266535287596993e-06, + "loss": 0.4003, + "step": 18273 + }, + { + "epoch": 6.75, + "learning_rate": 2.3266165783963543e-06, + "loss": 0.4183, + "step": 18274 + }, + { + "epoch": 6.75, + "learning_rate": 2.326579628033009e-06, + "loss": 0.3458, + "step": 18275 + }, + { + "epoch": 6.75, + "learning_rate": 2.326542677669664e-06, + "loss": 0.3644, + "step": 18276 + }, + { + "epoch": 6.75, + "learning_rate": 2.3265057273063187e-06, + "loss": 0.4174, + "step": 18277 + }, + { + "epoch": 6.75, + "learning_rate": 2.3264687769429733e-06, + "loss": 0.3859, + "step": 18278 + }, + { + "epoch": 6.75, + "learning_rate": 2.326431826579628e-06, + "loss": 0.4526, + "step": 18279 + }, + { + "epoch": 6.75, + "learning_rate": 2.326394876216283e-06, + "loss": 0.3025, + "step": 18280 + }, + { + "epoch": 6.75, + "learning_rate": 2.3263579258529376e-06, + "loss": 0.3491, + "step": 18281 + }, + { + "epoch": 6.75, + "learning_rate": 2.3263209754895923e-06, + "loss": 0.5077, + "step": 18282 + }, + { + "epoch": 6.75, + "learning_rate": 2.3262840251262473e-06, + "loss": 0.4911, + "step": 18283 + }, + { + "epoch": 6.75, + "learning_rate": 2.326247074762902e-06, + "loss": 0.4673, + "step": 18284 + }, + { + "epoch": 6.75, + "learning_rate": 2.3262101243995566e-06, + "loss": 0.331, + "step": 18285 + }, + { + "epoch": 6.75, + "learning_rate": 2.3261731740362112e-06, + "loss": 0.4895, + "step": 18286 + }, + { + "epoch": 6.75, + "learning_rate": 2.3261362236728663e-06, + "loss": 0.4763, + "step": 18287 + }, + { + "epoch": 6.75, + "learning_rate": 2.326099273309521e-06, + "loss": 0.5101, + "step": 18288 + }, + { + "epoch": 6.75, + "learning_rate": 2.326062322946176e-06, + "loss": 0.3825, + "step": 18289 + }, + { + "epoch": 6.75, + "learning_rate": 2.3260253725828306e-06, + "loss": 0.4581, + "step": 18290 + }, + { + "epoch": 6.75, + "learning_rate": 2.3259884222194853e-06, + "loss": 0.3988, + "step": 18291 + }, + { + "epoch": 6.75, + "learning_rate": 2.32595147185614e-06, + "loss": 0.4578, + "step": 18292 + }, + { + "epoch": 6.76, + "learning_rate": 2.3259145214927945e-06, + "loss": 0.4955, + "step": 18293 + }, + { + "epoch": 6.76, + "learning_rate": 2.3258775711294496e-06, + "loss": 0.4251, + "step": 18294 + }, + { + "epoch": 6.76, + "learning_rate": 2.3258406207661042e-06, + "loss": 0.4611, + "step": 18295 + }, + { + "epoch": 6.76, + "learning_rate": 2.3258036704027593e-06, + "loss": 0.5523, + "step": 18296 + }, + { + "epoch": 6.76, + "learning_rate": 2.3257667200394135e-06, + "loss": 0.4688, + "step": 18297 + }, + { + "epoch": 6.76, + "learning_rate": 2.3257297696760686e-06, + "loss": 0.4402, + "step": 18298 + }, + { + "epoch": 6.76, + "learning_rate": 2.325692819312723e-06, + "loss": 0.396, + "step": 18299 + }, + { + "epoch": 6.76, + "learning_rate": 2.3256558689493783e-06, + "loss": 0.4243, + "step": 18300 + }, + { + "epoch": 6.76, + "learning_rate": 2.325618918586033e-06, + "loss": 0.4545, + "step": 18301 + }, + { + "epoch": 6.76, + "learning_rate": 2.3255819682226875e-06, + "loss": 0.4599, + "step": 18302 + }, + { + "epoch": 6.76, + "learning_rate": 2.325545017859342e-06, + "loss": 0.4983, + "step": 18303 + }, + { + "epoch": 6.76, + "learning_rate": 2.325508067495997e-06, + "loss": 0.555, + "step": 18304 + }, + { + "epoch": 6.76, + "learning_rate": 2.325471117132652e-06, + "loss": 0.4544, + "step": 18305 + }, + { + "epoch": 6.76, + "learning_rate": 2.3254341667693065e-06, + "loss": 0.44, + "step": 18306 + }, + { + "epoch": 6.76, + "learning_rate": 2.3253972164059616e-06, + "loss": 0.4612, + "step": 18307 + }, + { + "epoch": 6.76, + "learning_rate": 2.325360266042616e-06, + "loss": 0.5459, + "step": 18308 + }, + { + "epoch": 6.76, + "learning_rate": 2.325323315679271e-06, + "loss": 0.3229, + "step": 18309 + }, + { + "epoch": 6.76, + "learning_rate": 2.3252863653159255e-06, + "loss": 0.378, + "step": 18310 + }, + { + "epoch": 6.76, + "learning_rate": 2.3252494149525805e-06, + "loss": 0.4532, + "step": 18311 + }, + { + "epoch": 6.76, + "learning_rate": 2.325212464589235e-06, + "loss": 0.3601, + "step": 18312 + }, + { + "epoch": 6.76, + "learning_rate": 2.32517551422589e-06, + "loss": 0.4503, + "step": 18313 + }, + { + "epoch": 6.76, + "learning_rate": 2.325138563862545e-06, + "loss": 0.4012, + "step": 18314 + }, + { + "epoch": 6.76, + "learning_rate": 2.3251016134991995e-06, + "loss": 0.3938, + "step": 18315 + }, + { + "epoch": 6.76, + "learning_rate": 2.325064663135854e-06, + "loss": 0.3296, + "step": 18316 + }, + { + "epoch": 6.76, + "learning_rate": 2.3250277127725088e-06, + "loss": 0.3112, + "step": 18317 + }, + { + "epoch": 6.76, + "learning_rate": 2.324990762409164e-06, + "loss": 0.377, + "step": 18318 + }, + { + "epoch": 6.76, + "learning_rate": 2.3249538120458185e-06, + "loss": 0.4481, + "step": 18319 + }, + { + "epoch": 6.77, + "learning_rate": 2.3249168616824735e-06, + "loss": 0.4485, + "step": 18320 + }, + { + "epoch": 6.77, + "learning_rate": 2.324879911319128e-06, + "loss": 0.4535, + "step": 18321 + }, + { + "epoch": 6.77, + "learning_rate": 2.324842960955783e-06, + "loss": 0.4551, + "step": 18322 + }, + { + "epoch": 6.77, + "learning_rate": 2.3248060105924375e-06, + "loss": 0.3828, + "step": 18323 + }, + { + "epoch": 6.77, + "learning_rate": 2.324769060229092e-06, + "loss": 0.4881, + "step": 18324 + }, + { + "epoch": 6.77, + "learning_rate": 2.324732109865747e-06, + "loss": 0.4297, + "step": 18325 + }, + { + "epoch": 6.77, + "learning_rate": 2.3246951595024018e-06, + "loss": 0.3739, + "step": 18326 + }, + { + "epoch": 6.77, + "learning_rate": 2.324658209139057e-06, + "loss": 0.3934, + "step": 18327 + }, + { + "epoch": 6.77, + "learning_rate": 2.3246212587757115e-06, + "loss": 0.6004, + "step": 18328 + }, + { + "epoch": 6.77, + "learning_rate": 2.324584308412366e-06, + "loss": 0.4322, + "step": 18329 + }, + { + "epoch": 6.77, + "learning_rate": 2.3245473580490208e-06, + "loss": 0.3505, + "step": 18330 + }, + { + "epoch": 6.77, + "learning_rate": 2.324510407685676e-06, + "loss": 0.5653, + "step": 18331 + }, + { + "epoch": 6.77, + "learning_rate": 2.3244734573223304e-06, + "loss": 0.4957, + "step": 18332 + }, + { + "epoch": 6.77, + "learning_rate": 2.324436506958985e-06, + "loss": 0.3753, + "step": 18333 + }, + { + "epoch": 6.77, + "learning_rate": 2.32439955659564e-06, + "loss": 0.3669, + "step": 18334 + }, + { + "epoch": 6.77, + "learning_rate": 2.3243626062322948e-06, + "loss": 0.3313, + "step": 18335 + }, + { + "epoch": 6.77, + "learning_rate": 2.3243256558689494e-06, + "loss": 0.4447, + "step": 18336 + }, + { + "epoch": 6.77, + "learning_rate": 2.324288705505604e-06, + "loss": 0.4015, + "step": 18337 + }, + { + "epoch": 6.77, + "learning_rate": 2.324251755142259e-06, + "loss": 0.4542, + "step": 18338 + }, + { + "epoch": 6.77, + "learning_rate": 2.3242148047789138e-06, + "loss": 0.5237, + "step": 18339 + }, + { + "epoch": 6.77, + "learning_rate": 2.324177854415569e-06, + "loss": 0.3679, + "step": 18340 + }, + { + "epoch": 6.77, + "learning_rate": 2.3241409040522234e-06, + "loss": 0.3386, + "step": 18341 + }, + { + "epoch": 6.77, + "learning_rate": 2.3241039536888777e-06, + "loss": 0.5085, + "step": 18342 + }, + { + "epoch": 6.77, + "learning_rate": 2.3240670033255327e-06, + "loss": 0.5394, + "step": 18343 + }, + { + "epoch": 6.77, + "learning_rate": 2.3240300529621874e-06, + "loss": 0.5264, + "step": 18344 + }, + { + "epoch": 6.77, + "learning_rate": 2.3239931025988424e-06, + "loss": 0.6051, + "step": 18345 + }, + { + "epoch": 6.77, + "learning_rate": 2.323956152235497e-06, + "loss": 0.3588, + "step": 18346 + }, + { + "epoch": 6.78, + "learning_rate": 2.3239192018721517e-06, + "loss": 0.5617, + "step": 18347 + }, + { + "epoch": 6.78, + "learning_rate": 2.3238822515088063e-06, + "loss": 0.5427, + "step": 18348 + }, + { + "epoch": 6.78, + "learning_rate": 2.3238453011454614e-06, + "loss": 0.4493, + "step": 18349 + }, + { + "epoch": 6.78, + "learning_rate": 2.323808350782116e-06, + "loss": 0.3745, + "step": 18350 + }, + { + "epoch": 6.78, + "learning_rate": 2.323771400418771e-06, + "loss": 0.5456, + "step": 18351 + }, + { + "epoch": 6.78, + "learning_rate": 2.3237344500554257e-06, + "loss": 0.3237, + "step": 18352 + }, + { + "epoch": 6.78, + "learning_rate": 2.3236974996920804e-06, + "loss": 0.3691, + "step": 18353 + }, + { + "epoch": 6.78, + "learning_rate": 2.323660549328735e-06, + "loss": 0.4862, + "step": 18354 + }, + { + "epoch": 6.78, + "learning_rate": 2.3236235989653896e-06, + "loss": 0.5073, + "step": 18355 + }, + { + "epoch": 6.78, + "learning_rate": 2.3235866486020447e-06, + "loss": 0.3499, + "step": 18356 + }, + { + "epoch": 6.78, + "learning_rate": 2.3235496982386993e-06, + "loss": 0.539, + "step": 18357 + }, + { + "epoch": 6.78, + "learning_rate": 2.3235127478753544e-06, + "loss": 0.5564, + "step": 18358 + }, + { + "epoch": 6.78, + "learning_rate": 2.323475797512009e-06, + "loss": 0.4574, + "step": 18359 + }, + { + "epoch": 6.78, + "learning_rate": 2.3234388471486637e-06, + "loss": 0.5217, + "step": 18360 + }, + { + "epoch": 6.78, + "learning_rate": 2.3234018967853183e-06, + "loss": 0.4462, + "step": 18361 + }, + { + "epoch": 6.78, + "learning_rate": 2.323364946421973e-06, + "loss": 0.4382, + "step": 18362 + }, + { + "epoch": 6.78, + "learning_rate": 2.323327996058628e-06, + "loss": 0.4991, + "step": 18363 + }, + { + "epoch": 6.78, + "learning_rate": 2.3232910456952826e-06, + "loss": 0.3773, + "step": 18364 + }, + { + "epoch": 6.78, + "learning_rate": 2.3232540953319377e-06, + "loss": 0.5198, + "step": 18365 + }, + { + "epoch": 6.78, + "learning_rate": 2.3232171449685923e-06, + "loss": 0.4965, + "step": 18366 + }, + { + "epoch": 6.78, + "learning_rate": 2.323180194605247e-06, + "loss": 0.375, + "step": 18367 + }, + { + "epoch": 6.78, + "learning_rate": 2.3231432442419016e-06, + "loss": 0.5348, + "step": 18368 + }, + { + "epoch": 6.78, + "learning_rate": 2.3231062938785567e-06, + "loss": 0.444, + "step": 18369 + }, + { + "epoch": 6.78, + "learning_rate": 2.3230693435152113e-06, + "loss": 0.4271, + "step": 18370 + }, + { + "epoch": 6.78, + "learning_rate": 2.3230323931518664e-06, + "loss": 0.2648, + "step": 18371 + }, + { + "epoch": 6.78, + "learning_rate": 2.322995442788521e-06, + "loss": 0.3738, + "step": 18372 + }, + { + "epoch": 6.78, + "learning_rate": 2.3229584924251756e-06, + "loss": 0.4705, + "step": 18373 + }, + { + "epoch": 6.79, + "learning_rate": 2.3229215420618303e-06, + "loss": 0.5039, + "step": 18374 + }, + { + "epoch": 6.79, + "learning_rate": 2.322884591698485e-06, + "loss": 0.4289, + "step": 18375 + }, + { + "epoch": 6.79, + "learning_rate": 2.32284764133514e-06, + "loss": 0.5554, + "step": 18376 + }, + { + "epoch": 6.79, + "learning_rate": 2.3228106909717946e-06, + "loss": 0.4258, + "step": 18377 + }, + { + "epoch": 6.79, + "learning_rate": 2.3227737406084497e-06, + "loss": 0.3775, + "step": 18378 + }, + { + "epoch": 6.79, + "learning_rate": 2.3227367902451043e-06, + "loss": 0.3683, + "step": 18379 + }, + { + "epoch": 6.79, + "learning_rate": 2.322699839881759e-06, + "loss": 0.6625, + "step": 18380 + }, + { + "epoch": 6.79, + "learning_rate": 2.3226628895184136e-06, + "loss": 0.4085, + "step": 18381 + }, + { + "epoch": 6.79, + "learning_rate": 2.322625939155068e-06, + "loss": 0.5453, + "step": 18382 + }, + { + "epoch": 6.79, + "learning_rate": 2.3225889887917233e-06, + "loss": 0.4838, + "step": 18383 + }, + { + "epoch": 6.79, + "learning_rate": 2.322552038428378e-06, + "loss": 0.4755, + "step": 18384 + }, + { + "epoch": 6.79, + "learning_rate": 2.322515088065033e-06, + "loss": 0.5194, + "step": 18385 + }, + { + "epoch": 6.79, + "learning_rate": 2.322478137701687e-06, + "loss": 0.4631, + "step": 18386 + }, + { + "epoch": 6.79, + "learning_rate": 2.3224411873383422e-06, + "loss": 0.4311, + "step": 18387 + }, + { + "epoch": 6.79, + "learning_rate": 2.322404236974997e-06, + "loss": 0.383, + "step": 18388 + }, + { + "epoch": 6.79, + "learning_rate": 2.322367286611652e-06, + "loss": 0.437, + "step": 18389 + }, + { + "epoch": 6.79, + "learning_rate": 2.3223303362483066e-06, + "loss": 0.417, + "step": 18390 + }, + { + "epoch": 6.79, + "learning_rate": 2.322293385884961e-06, + "loss": 0.4401, + "step": 18391 + }, + { + "epoch": 6.79, + "learning_rate": 2.322256435521616e-06, + "loss": 0.4618, + "step": 18392 + }, + { + "epoch": 6.79, + "learning_rate": 2.3222194851582705e-06, + "loss": 0.5182, + "step": 18393 + }, + { + "epoch": 6.79, + "learning_rate": 2.3221825347949255e-06, + "loss": 0.4173, + "step": 18394 + }, + { + "epoch": 6.79, + "learning_rate": 2.32214558443158e-06, + "loss": 0.5295, + "step": 18395 + }, + { + "epoch": 6.79, + "learning_rate": 2.3221086340682352e-06, + "loss": 0.4737, + "step": 18396 + }, + { + "epoch": 6.79, + "learning_rate": 2.32207168370489e-06, + "loss": 0.5928, + "step": 18397 + }, + { + "epoch": 6.79, + "learning_rate": 2.3220347333415445e-06, + "loss": 0.4179, + "step": 18398 + }, + { + "epoch": 6.79, + "learning_rate": 2.321997782978199e-06, + "loss": 0.4422, + "step": 18399 + }, + { + "epoch": 6.79, + "learning_rate": 2.321960832614854e-06, + "loss": 0.3953, + "step": 18400 + }, + { + "epoch": 6.8, + "learning_rate": 2.321923882251509e-06, + "loss": 0.4745, + "step": 18401 + }, + { + "epoch": 6.8, + "learning_rate": 2.3218869318881635e-06, + "loss": 0.4314, + "step": 18402 + }, + { + "epoch": 6.8, + "learning_rate": 2.3218499815248185e-06, + "loss": 0.4163, + "step": 18403 + }, + { + "epoch": 6.8, + "learning_rate": 2.321813031161473e-06, + "loss": 0.3525, + "step": 18404 + }, + { + "epoch": 6.8, + "learning_rate": 2.321776080798128e-06, + "loss": 0.4323, + "step": 18405 + }, + { + "epoch": 6.8, + "learning_rate": 2.3217391304347824e-06, + "loss": 0.4794, + "step": 18406 + }, + { + "epoch": 6.8, + "learning_rate": 2.3217021800714375e-06, + "loss": 0.4979, + "step": 18407 + }, + { + "epoch": 6.8, + "learning_rate": 2.321665229708092e-06, + "loss": 0.4943, + "step": 18408 + }, + { + "epoch": 6.8, + "learning_rate": 2.321628279344747e-06, + "loss": 0.5019, + "step": 18409 + }, + { + "epoch": 6.8, + "learning_rate": 2.321591328981402e-06, + "loss": 0.4056, + "step": 18410 + }, + { + "epoch": 6.8, + "learning_rate": 2.3215543786180565e-06, + "loss": 0.5011, + "step": 18411 + }, + { + "epoch": 6.8, + "learning_rate": 2.321517428254711e-06, + "loss": 0.4978, + "step": 18412 + }, + { + "epoch": 6.8, + "learning_rate": 2.3214804778913657e-06, + "loss": 0.447, + "step": 18413 + }, + { + "epoch": 6.8, + "learning_rate": 2.321443527528021e-06, + "loss": 0.4295, + "step": 18414 + }, + { + "epoch": 6.8, + "learning_rate": 2.3214065771646754e-06, + "loss": 0.4447, + "step": 18415 + }, + { + "epoch": 6.8, + "learning_rate": 2.3213696268013305e-06, + "loss": 0.4601, + "step": 18416 + }, + { + "epoch": 6.8, + "learning_rate": 2.321332676437985e-06, + "loss": 0.38, + "step": 18417 + }, + { + "epoch": 6.8, + "learning_rate": 2.3212957260746398e-06, + "loss": 0.4132, + "step": 18418 + }, + { + "epoch": 6.8, + "learning_rate": 2.3212587757112944e-06, + "loss": 0.4757, + "step": 18419 + }, + { + "epoch": 6.8, + "learning_rate": 2.3212218253479495e-06, + "loss": 0.4128, + "step": 18420 + }, + { + "epoch": 6.8, + "learning_rate": 2.321184874984604e-06, + "loss": 0.3968, + "step": 18421 + }, + { + "epoch": 6.8, + "learning_rate": 2.3211479246212587e-06, + "loss": 0.3791, + "step": 18422 + }, + { + "epoch": 6.8, + "learning_rate": 2.321110974257914e-06, + "loss": 0.4162, + "step": 18423 + }, + { + "epoch": 6.8, + "learning_rate": 2.3210740238945684e-06, + "loss": 0.4437, + "step": 18424 + }, + { + "epoch": 6.8, + "learning_rate": 2.321037073531223e-06, + "loss": 0.5057, + "step": 18425 + }, + { + "epoch": 6.8, + "learning_rate": 2.3210001231678777e-06, + "loss": 0.4143, + "step": 18426 + }, + { + "epoch": 6.8, + "learning_rate": 2.3209631728045328e-06, + "loss": 0.395, + "step": 18427 + }, + { + "epoch": 6.81, + "learning_rate": 2.3209262224411874e-06, + "loss": 0.4258, + "step": 18428 + }, + { + "epoch": 6.81, + "learning_rate": 2.3208892720778425e-06, + "loss": 0.3273, + "step": 18429 + }, + { + "epoch": 6.81, + "learning_rate": 2.320852321714497e-06, + "loss": 0.4979, + "step": 18430 + }, + { + "epoch": 6.81, + "learning_rate": 2.3208153713511513e-06, + "loss": 0.5354, + "step": 18431 + }, + { + "epoch": 6.81, + "learning_rate": 2.3207784209878064e-06, + "loss": 0.4032, + "step": 18432 + }, + { + "epoch": 6.81, + "learning_rate": 2.320741470624461e-06, + "loss": 0.3993, + "step": 18433 + }, + { + "epoch": 6.81, + "learning_rate": 2.320704520261116e-06, + "loss": 0.5257, + "step": 18434 + }, + { + "epoch": 6.81, + "learning_rate": 2.3206675698977707e-06, + "loss": 0.4181, + "step": 18435 + }, + { + "epoch": 6.81, + "learning_rate": 2.3206306195344258e-06, + "loss": 0.4911, + "step": 18436 + }, + { + "epoch": 6.81, + "learning_rate": 2.32059366917108e-06, + "loss": 0.513, + "step": 18437 + }, + { + "epoch": 6.81, + "learning_rate": 2.320556718807735e-06, + "loss": 0.4115, + "step": 18438 + }, + { + "epoch": 6.81, + "learning_rate": 2.3205197684443897e-06, + "loss": 0.4526, + "step": 18439 + }, + { + "epoch": 6.81, + "learning_rate": 2.3204828180810447e-06, + "loss": 0.5603, + "step": 18440 + }, + { + "epoch": 6.81, + "learning_rate": 2.3204458677176994e-06, + "loss": 0.4811, + "step": 18441 + }, + { + "epoch": 6.81, + "learning_rate": 2.320408917354354e-06, + "loss": 0.3744, + "step": 18442 + }, + { + "epoch": 6.81, + "learning_rate": 2.3203719669910086e-06, + "loss": 0.4254, + "step": 18443 + }, + { + "epoch": 6.81, + "learning_rate": 2.3203350166276633e-06, + "loss": 0.3738, + "step": 18444 + }, + { + "epoch": 6.81, + "learning_rate": 2.3202980662643183e-06, + "loss": 0.4066, + "step": 18445 + }, + { + "epoch": 6.81, + "learning_rate": 2.320261115900973e-06, + "loss": 0.4233, + "step": 18446 + }, + { + "epoch": 6.81, + "learning_rate": 2.320224165537628e-06, + "loss": 0.3827, + "step": 18447 + }, + { + "epoch": 6.81, + "learning_rate": 2.3201872151742827e-06, + "loss": 0.4835, + "step": 18448 + }, + { + "epoch": 6.81, + "learning_rate": 2.3201502648109373e-06, + "loss": 0.4314, + "step": 18449 + }, + { + "epoch": 6.81, + "learning_rate": 2.320113314447592e-06, + "loss": 0.4897, + "step": 18450 + }, + { + "epoch": 6.81, + "learning_rate": 2.3200763640842466e-06, + "loss": 0.5226, + "step": 18451 + }, + { + "epoch": 6.81, + "learning_rate": 2.3200394137209016e-06, + "loss": 0.3326, + "step": 18452 + }, + { + "epoch": 6.81, + "learning_rate": 2.3200024633575563e-06, + "loss": 0.3054, + "step": 18453 + }, + { + "epoch": 6.81, + "learning_rate": 2.3199655129942113e-06, + "loss": 0.484, + "step": 18454 + }, + { + "epoch": 6.81, + "learning_rate": 2.319928562630866e-06, + "loss": 0.438, + "step": 18455 + }, + { + "epoch": 6.82, + "learning_rate": 2.3198916122675206e-06, + "loss": 0.5591, + "step": 18456 + }, + { + "epoch": 6.82, + "learning_rate": 2.3198546619041753e-06, + "loss": 0.3841, + "step": 18457 + }, + { + "epoch": 6.82, + "learning_rate": 2.3198177115408303e-06, + "loss": 0.3413, + "step": 18458 + }, + { + "epoch": 6.82, + "learning_rate": 2.319780761177485e-06, + "loss": 0.5191, + "step": 18459 + }, + { + "epoch": 6.82, + "learning_rate": 2.31974381081414e-06, + "loss": 0.3591, + "step": 18460 + }, + { + "epoch": 6.82, + "learning_rate": 2.3197068604507946e-06, + "loss": 0.4248, + "step": 18461 + }, + { + "epoch": 6.82, + "learning_rate": 2.3196699100874493e-06, + "loss": 0.4547, + "step": 18462 + }, + { + "epoch": 6.82, + "learning_rate": 2.319632959724104e-06, + "loss": 0.4366, + "step": 18463 + }, + { + "epoch": 6.82, + "learning_rate": 2.3195960093607586e-06, + "loss": 0.6302, + "step": 18464 + }, + { + "epoch": 6.82, + "learning_rate": 2.3195590589974136e-06, + "loss": 0.4062, + "step": 18465 + }, + { + "epoch": 6.82, + "learning_rate": 2.3195221086340682e-06, + "loss": 0.4899, + "step": 18466 + }, + { + "epoch": 6.82, + "learning_rate": 2.3194851582707233e-06, + "loss": 0.4033, + "step": 18467 + }, + { + "epoch": 6.82, + "learning_rate": 2.319448207907378e-06, + "loss": 0.5199, + "step": 18468 + }, + { + "epoch": 6.82, + "learning_rate": 2.3194112575440326e-06, + "loss": 0.5863, + "step": 18469 + }, + { + "epoch": 6.82, + "learning_rate": 2.3193743071806872e-06, + "loss": 0.4512, + "step": 18470 + }, + { + "epoch": 6.82, + "learning_rate": 2.319337356817342e-06, + "loss": 0.5377, + "step": 18471 + }, + { + "epoch": 6.82, + "learning_rate": 2.319300406453997e-06, + "loss": 0.5361, + "step": 18472 + }, + { + "epoch": 6.82, + "learning_rate": 2.3192634560906516e-06, + "loss": 0.5403, + "step": 18473 + }, + { + "epoch": 6.82, + "learning_rate": 2.3192265057273066e-06, + "loss": 0.3439, + "step": 18474 + }, + { + "epoch": 6.82, + "learning_rate": 2.3191895553639612e-06, + "loss": 0.5349, + "step": 18475 + }, + { + "epoch": 6.82, + "learning_rate": 2.319152605000616e-06, + "loss": 0.4005, + "step": 18476 + }, + { + "epoch": 6.82, + "learning_rate": 2.3191156546372705e-06, + "loss": 0.3459, + "step": 18477 + }, + { + "epoch": 6.82, + "learning_rate": 2.3190787042739256e-06, + "loss": 0.6171, + "step": 18478 + }, + { + "epoch": 6.82, + "learning_rate": 2.3190417539105802e-06, + "loss": 0.4016, + "step": 18479 + }, + { + "epoch": 6.82, + "learning_rate": 2.3190048035472353e-06, + "loss": 0.3917, + "step": 18480 + }, + { + "epoch": 6.82, + "learning_rate": 2.31896785318389e-06, + "loss": 0.508, + "step": 18481 + }, + { + "epoch": 6.82, + "learning_rate": 2.318930902820544e-06, + "loss": 0.5179, + "step": 18482 + }, + { + "epoch": 6.83, + "learning_rate": 2.318893952457199e-06, + "loss": 0.4288, + "step": 18483 + }, + { + "epoch": 6.83, + "learning_rate": 2.318857002093854e-06, + "loss": 0.6188, + "step": 18484 + }, + { + "epoch": 6.83, + "learning_rate": 2.318820051730509e-06, + "loss": 0.5135, + "step": 18485 + }, + { + "epoch": 6.83, + "learning_rate": 2.3187831013671635e-06, + "loss": 0.5194, + "step": 18486 + }, + { + "epoch": 6.83, + "learning_rate": 2.3187461510038186e-06, + "loss": 0.452, + "step": 18487 + }, + { + "epoch": 6.83, + "learning_rate": 2.318709200640473e-06, + "loss": 0.4833, + "step": 18488 + }, + { + "epoch": 6.83, + "learning_rate": 2.318672250277128e-06, + "loss": 0.4226, + "step": 18489 + }, + { + "epoch": 6.83, + "learning_rate": 2.3186352999137825e-06, + "loss": 0.3752, + "step": 18490 + }, + { + "epoch": 6.83, + "learning_rate": 2.318598349550437e-06, + "loss": 0.2827, + "step": 18491 + }, + { + "epoch": 6.83, + "learning_rate": 2.318561399187092e-06, + "loss": 0.4576, + "step": 18492 + }, + { + "epoch": 6.83, + "learning_rate": 2.318524448823747e-06, + "loss": 0.4465, + "step": 18493 + }, + { + "epoch": 6.83, + "learning_rate": 2.3184874984604015e-06, + "loss": 0.3911, + "step": 18494 + }, + { + "epoch": 6.83, + "learning_rate": 2.318450548097056e-06, + "loss": 0.3521, + "step": 18495 + }, + { + "epoch": 6.83, + "learning_rate": 2.318413597733711e-06, + "loss": 0.3579, + "step": 18496 + }, + { + "epoch": 6.83, + "learning_rate": 2.318376647370366e-06, + "loss": 0.6091, + "step": 18497 + }, + { + "epoch": 6.83, + "learning_rate": 2.318339697007021e-06, + "loss": 0.3427, + "step": 18498 + }, + { + "epoch": 6.83, + "learning_rate": 2.3183027466436755e-06, + "loss": 0.5425, + "step": 18499 + }, + { + "epoch": 6.83, + "learning_rate": 2.31826579628033e-06, + "loss": 0.4163, + "step": 18500 + }, + { + "epoch": 6.83, + "learning_rate": 2.3182288459169848e-06, + "loss": 0.3823, + "step": 18501 + }, + { + "epoch": 6.83, + "learning_rate": 2.3181918955536394e-06, + "loss": 0.4718, + "step": 18502 + }, + { + "epoch": 6.83, + "learning_rate": 2.3181549451902945e-06, + "loss": 0.3281, + "step": 18503 + }, + { + "epoch": 6.83, + "learning_rate": 2.318117994826949e-06, + "loss": 0.4665, + "step": 18504 + }, + { + "epoch": 6.83, + "learning_rate": 2.318081044463604e-06, + "loss": 0.3816, + "step": 18505 + }, + { + "epoch": 6.83, + "learning_rate": 2.3180440941002588e-06, + "loss": 0.5011, + "step": 18506 + }, + { + "epoch": 6.83, + "learning_rate": 2.3180071437369134e-06, + "loss": 0.5077, + "step": 18507 + }, + { + "epoch": 6.83, + "learning_rate": 2.317970193373568e-06, + "loss": 0.3982, + "step": 18508 + }, + { + "epoch": 6.83, + "learning_rate": 2.317933243010223e-06, + "loss": 0.3392, + "step": 18509 + }, + { + "epoch": 6.84, + "learning_rate": 2.3178962926468778e-06, + "loss": 0.5334, + "step": 18510 + }, + { + "epoch": 6.84, + "learning_rate": 2.3178593422835324e-06, + "loss": 0.3942, + "step": 18511 + }, + { + "epoch": 6.84, + "learning_rate": 2.3178223919201875e-06, + "loss": 0.3901, + "step": 18512 + }, + { + "epoch": 6.84, + "learning_rate": 2.317785441556842e-06, + "loss": 0.4123, + "step": 18513 + }, + { + "epoch": 6.84, + "learning_rate": 2.3177484911934967e-06, + "loss": 0.4252, + "step": 18514 + }, + { + "epoch": 6.84, + "learning_rate": 2.3177115408301514e-06, + "loss": 0.3871, + "step": 18515 + }, + { + "epoch": 6.84, + "learning_rate": 2.3176745904668064e-06, + "loss": 0.5291, + "step": 18516 + }, + { + "epoch": 6.84, + "learning_rate": 2.317637640103461e-06, + "loss": 0.4546, + "step": 18517 + }, + { + "epoch": 6.84, + "learning_rate": 2.317600689740116e-06, + "loss": 0.5294, + "step": 18518 + }, + { + "epoch": 6.84, + "learning_rate": 2.3175637393767708e-06, + "loss": 0.4361, + "step": 18519 + }, + { + "epoch": 6.84, + "learning_rate": 2.3175267890134254e-06, + "loss": 0.3402, + "step": 18520 + }, + { + "epoch": 6.84, + "learning_rate": 2.31748983865008e-06, + "loss": 0.582, + "step": 18521 + }, + { + "epoch": 6.84, + "learning_rate": 2.3174528882867347e-06, + "loss": 0.4779, + "step": 18522 + }, + { + "epoch": 6.84, + "learning_rate": 2.3174159379233897e-06, + "loss": 0.4044, + "step": 18523 + }, + { + "epoch": 6.84, + "learning_rate": 2.3173789875600444e-06, + "loss": 0.4918, + "step": 18524 + }, + { + "epoch": 6.84, + "learning_rate": 2.3173420371966994e-06, + "loss": 0.2985, + "step": 18525 + }, + { + "epoch": 6.84, + "learning_rate": 2.317305086833354e-06, + "loss": 0.423, + "step": 18526 + }, + { + "epoch": 6.84, + "learning_rate": 2.3172681364700087e-06, + "loss": 0.4061, + "step": 18527 + }, + { + "epoch": 6.84, + "learning_rate": 2.3172311861066633e-06, + "loss": 0.4141, + "step": 18528 + }, + { + "epoch": 6.84, + "learning_rate": 2.3171942357433184e-06, + "loss": 0.5093, + "step": 18529 + }, + { + "epoch": 6.84, + "learning_rate": 2.317157285379973e-06, + "loss": 0.4193, + "step": 18530 + }, + { + "epoch": 6.84, + "learning_rate": 2.3171203350166277e-06, + "loss": 0.2557, + "step": 18531 + }, + { + "epoch": 6.84, + "learning_rate": 2.3170833846532823e-06, + "loss": 0.3893, + "step": 18532 + }, + { + "epoch": 6.84, + "learning_rate": 2.317046434289937e-06, + "loss": 0.5546, + "step": 18533 + }, + { + "epoch": 6.84, + "learning_rate": 2.317009483926592e-06, + "loss": 0.3986, + "step": 18534 + }, + { + "epoch": 6.84, + "learning_rate": 2.3169725335632466e-06, + "loss": 0.4678, + "step": 18535 + }, + { + "epoch": 6.84, + "learning_rate": 2.3169355831999017e-06, + "loss": 0.4197, + "step": 18536 + }, + { + "epoch": 6.85, + "learning_rate": 2.3168986328365563e-06, + "loss": 0.4205, + "step": 18537 + }, + { + "epoch": 6.85, + "learning_rate": 2.316861682473211e-06, + "loss": 0.4731, + "step": 18538 + }, + { + "epoch": 6.85, + "learning_rate": 2.3168247321098656e-06, + "loss": 0.439, + "step": 18539 + }, + { + "epoch": 6.85, + "learning_rate": 2.3167877817465207e-06, + "loss": 0.5859, + "step": 18540 + }, + { + "epoch": 6.85, + "learning_rate": 2.3167508313831753e-06, + "loss": 0.2887, + "step": 18541 + }, + { + "epoch": 6.85, + "learning_rate": 2.31671388101983e-06, + "loss": 0.5253, + "step": 18542 + }, + { + "epoch": 6.85, + "learning_rate": 2.316676930656485e-06, + "loss": 0.3746, + "step": 18543 + }, + { + "epoch": 6.85, + "learning_rate": 2.3166399802931396e-06, + "loss": 0.4562, + "step": 18544 + }, + { + "epoch": 6.85, + "learning_rate": 2.3166030299297943e-06, + "loss": 0.5029, + "step": 18545 + }, + { + "epoch": 6.85, + "learning_rate": 2.316566079566449e-06, + "loss": 0.5552, + "step": 18546 + }, + { + "epoch": 6.85, + "learning_rate": 2.316529129203104e-06, + "loss": 0.4371, + "step": 18547 + }, + { + "epoch": 6.85, + "learning_rate": 2.3164921788397586e-06, + "loss": 0.3657, + "step": 18548 + }, + { + "epoch": 6.85, + "learning_rate": 2.3164552284764137e-06, + "loss": 0.5022, + "step": 18549 + }, + { + "epoch": 6.85, + "learning_rate": 2.3164182781130683e-06, + "loss": 0.4584, + "step": 18550 + }, + { + "epoch": 6.85, + "learning_rate": 2.316381327749723e-06, + "loss": 0.407, + "step": 18551 + }, + { + "epoch": 6.85, + "learning_rate": 2.3163443773863776e-06, + "loss": 0.4978, + "step": 18552 + }, + { + "epoch": 6.85, + "learning_rate": 2.316307427023032e-06, + "loss": 0.4824, + "step": 18553 + }, + { + "epoch": 6.85, + "learning_rate": 2.3162704766596873e-06, + "loss": 0.5063, + "step": 18554 + }, + { + "epoch": 6.85, + "learning_rate": 2.316233526296342e-06, + "loss": 0.4359, + "step": 18555 + }, + { + "epoch": 6.85, + "learning_rate": 2.316196575932997e-06, + "loss": 0.5028, + "step": 18556 + }, + { + "epoch": 6.85, + "learning_rate": 2.3161596255696516e-06, + "loss": 0.3738, + "step": 18557 + }, + { + "epoch": 6.85, + "learning_rate": 2.3161226752063062e-06, + "loss": 0.3329, + "step": 18558 + }, + { + "epoch": 6.85, + "learning_rate": 2.316085724842961e-06, + "loss": 0.3841, + "step": 18559 + }, + { + "epoch": 6.85, + "learning_rate": 2.316048774479616e-06, + "loss": 0.3414, + "step": 18560 + }, + { + "epoch": 6.85, + "learning_rate": 2.3160118241162706e-06, + "loss": 0.3915, + "step": 18561 + }, + { + "epoch": 6.85, + "learning_rate": 2.315974873752925e-06, + "loss": 0.4516, + "step": 18562 + }, + { + "epoch": 6.85, + "learning_rate": 2.3159379233895803e-06, + "loss": 0.4298, + "step": 18563 + }, + { + "epoch": 6.86, + "learning_rate": 2.315900973026235e-06, + "loss": 0.4494, + "step": 18564 + }, + { + "epoch": 6.86, + "learning_rate": 2.3158640226628895e-06, + "loss": 0.4734, + "step": 18565 + }, + { + "epoch": 6.86, + "learning_rate": 2.315827072299544e-06, + "loss": 0.4485, + "step": 18566 + }, + { + "epoch": 6.86, + "learning_rate": 2.3157901219361992e-06, + "loss": 0.5396, + "step": 18567 + }, + { + "epoch": 6.86, + "learning_rate": 2.315753171572854e-06, + "loss": 0.3866, + "step": 18568 + }, + { + "epoch": 6.86, + "learning_rate": 2.315716221209509e-06, + "loss": 0.3818, + "step": 18569 + }, + { + "epoch": 6.86, + "learning_rate": 2.3156792708461636e-06, + "loss": 0.5051, + "step": 18570 + }, + { + "epoch": 6.86, + "learning_rate": 2.3156423204828178e-06, + "loss": 0.4561, + "step": 18571 + }, + { + "epoch": 6.86, + "learning_rate": 2.315605370119473e-06, + "loss": 0.4969, + "step": 18572 + }, + { + "epoch": 6.86, + "learning_rate": 2.3155684197561275e-06, + "loss": 0.4631, + "step": 18573 + }, + { + "epoch": 6.86, + "learning_rate": 2.3155314693927825e-06, + "loss": 0.4091, + "step": 18574 + }, + { + "epoch": 6.86, + "learning_rate": 2.315494519029437e-06, + "loss": 0.4903, + "step": 18575 + }, + { + "epoch": 6.86, + "learning_rate": 2.3154575686660922e-06, + "loss": 0.3711, + "step": 18576 + }, + { + "epoch": 6.86, + "learning_rate": 2.3154206183027464e-06, + "loss": 0.5965, + "step": 18577 + }, + { + "epoch": 6.86, + "learning_rate": 2.3153836679394015e-06, + "loss": 0.5305, + "step": 18578 + }, + { + "epoch": 6.86, + "learning_rate": 2.315346717576056e-06, + "loss": 0.3971, + "step": 18579 + }, + { + "epoch": 6.86, + "learning_rate": 2.315309767212711e-06, + "loss": 0.5099, + "step": 18580 + }, + { + "epoch": 6.86, + "learning_rate": 2.315272816849366e-06, + "loss": 0.4216, + "step": 18581 + }, + { + "epoch": 6.86, + "learning_rate": 2.3152358664860205e-06, + "loss": 0.4507, + "step": 18582 + }, + { + "epoch": 6.86, + "learning_rate": 2.315198916122675e-06, + "loss": 0.4769, + "step": 18583 + }, + { + "epoch": 6.86, + "learning_rate": 2.3151619657593297e-06, + "loss": 0.4154, + "step": 18584 + }, + { + "epoch": 6.86, + "learning_rate": 2.315125015395985e-06, + "loss": 0.5981, + "step": 18585 + }, + { + "epoch": 6.86, + "learning_rate": 2.3150880650326394e-06, + "loss": 0.4801, + "step": 18586 + }, + { + "epoch": 6.86, + "learning_rate": 2.3150511146692945e-06, + "loss": 0.4245, + "step": 18587 + }, + { + "epoch": 6.86, + "learning_rate": 2.315014164305949e-06, + "loss": 0.3448, + "step": 18588 + }, + { + "epoch": 6.86, + "learning_rate": 2.3149772139426038e-06, + "loss": 0.3876, + "step": 18589 + }, + { + "epoch": 6.86, + "learning_rate": 2.3149402635792584e-06, + "loss": 0.5094, + "step": 18590 + }, + { + "epoch": 6.87, + "learning_rate": 2.314903313215913e-06, + "loss": 0.5601, + "step": 18591 + }, + { + "epoch": 6.87, + "learning_rate": 2.314866362852568e-06, + "loss": 0.4622, + "step": 18592 + }, + { + "epoch": 6.87, + "learning_rate": 2.3148294124892227e-06, + "loss": 0.4829, + "step": 18593 + }, + { + "epoch": 6.87, + "learning_rate": 2.314792462125878e-06, + "loss": 0.5602, + "step": 18594 + }, + { + "epoch": 6.87, + "learning_rate": 2.3147555117625324e-06, + "loss": 0.3548, + "step": 18595 + }, + { + "epoch": 6.87, + "learning_rate": 2.314718561399187e-06, + "loss": 0.5275, + "step": 18596 + }, + { + "epoch": 6.87, + "learning_rate": 2.3146816110358417e-06, + "loss": 0.3725, + "step": 18597 + }, + { + "epoch": 6.87, + "learning_rate": 2.3146446606724968e-06, + "loss": 0.4647, + "step": 18598 + }, + { + "epoch": 6.87, + "learning_rate": 2.3146077103091514e-06, + "loss": 0.4542, + "step": 18599 + }, + { + "epoch": 6.87, + "learning_rate": 2.3145707599458065e-06, + "loss": 0.5321, + "step": 18600 + }, + { + "epoch": 6.87, + "learning_rate": 2.314533809582461e-06, + "loss": 0.3621, + "step": 18601 + }, + { + "epoch": 6.87, + "learning_rate": 2.3144968592191157e-06, + "loss": 0.5376, + "step": 18602 + }, + { + "epoch": 6.87, + "learning_rate": 2.3144599088557704e-06, + "loss": 0.4295, + "step": 18603 + }, + { + "epoch": 6.87, + "learning_rate": 2.314422958492425e-06, + "loss": 0.4275, + "step": 18604 + }, + { + "epoch": 6.87, + "learning_rate": 2.31438600812908e-06, + "loss": 0.5032, + "step": 18605 + }, + { + "epoch": 6.87, + "learning_rate": 2.3143490577657347e-06, + "loss": 0.4399, + "step": 18606 + }, + { + "epoch": 6.87, + "learning_rate": 2.3143121074023898e-06, + "loss": 0.3405, + "step": 18607 + }, + { + "epoch": 6.87, + "learning_rate": 2.3142751570390444e-06, + "loss": 0.3073, + "step": 18608 + }, + { + "epoch": 6.87, + "learning_rate": 2.314238206675699e-06, + "loss": 0.5019, + "step": 18609 + }, + { + "epoch": 6.87, + "learning_rate": 2.3142012563123537e-06, + "loss": 0.3844, + "step": 18610 + }, + { + "epoch": 6.87, + "learning_rate": 2.3141643059490083e-06, + "loss": 0.4341, + "step": 18611 + }, + { + "epoch": 6.87, + "learning_rate": 2.3141273555856634e-06, + "loss": 0.4, + "step": 18612 + }, + { + "epoch": 6.87, + "learning_rate": 2.314090405222318e-06, + "loss": 0.3922, + "step": 18613 + }, + { + "epoch": 6.87, + "learning_rate": 2.314053454858973e-06, + "loss": 0.3806, + "step": 18614 + }, + { + "epoch": 6.87, + "learning_rate": 2.3140165044956277e-06, + "loss": 0.4405, + "step": 18615 + }, + { + "epoch": 6.87, + "learning_rate": 2.3139795541322823e-06, + "loss": 0.5021, + "step": 18616 + }, + { + "epoch": 6.87, + "learning_rate": 2.313942603768937e-06, + "loss": 0.5277, + "step": 18617 + }, + { + "epoch": 6.88, + "learning_rate": 2.313905653405592e-06, + "loss": 0.4442, + "step": 18618 + }, + { + "epoch": 6.88, + "learning_rate": 2.3138687030422467e-06, + "loss": 0.4336, + "step": 18619 + }, + { + "epoch": 6.88, + "learning_rate": 2.3138317526789013e-06, + "loss": 0.4669, + "step": 18620 + }, + { + "epoch": 6.88, + "learning_rate": 2.3137948023155564e-06, + "loss": 0.4225, + "step": 18621 + }, + { + "epoch": 6.88, + "learning_rate": 2.3137578519522106e-06, + "loss": 0.4292, + "step": 18622 + }, + { + "epoch": 6.88, + "learning_rate": 2.3137209015888657e-06, + "loss": 0.5732, + "step": 18623 + }, + { + "epoch": 6.88, + "learning_rate": 2.3136839512255203e-06, + "loss": 0.4894, + "step": 18624 + }, + { + "epoch": 6.88, + "learning_rate": 2.3136470008621753e-06, + "loss": 0.4227, + "step": 18625 + }, + { + "epoch": 6.88, + "learning_rate": 2.31361005049883e-06, + "loss": 0.438, + "step": 18626 + }, + { + "epoch": 6.88, + "learning_rate": 2.313573100135485e-06, + "loss": 0.4037, + "step": 18627 + }, + { + "epoch": 6.88, + "learning_rate": 2.3135361497721393e-06, + "loss": 0.3024, + "step": 18628 + }, + { + "epoch": 6.88, + "learning_rate": 2.3134991994087943e-06, + "loss": 0.3279, + "step": 18629 + }, + { + "epoch": 6.88, + "learning_rate": 2.313462249045449e-06, + "loss": 0.4912, + "step": 18630 + }, + { + "epoch": 6.88, + "learning_rate": 2.3134252986821036e-06, + "loss": 0.4569, + "step": 18631 + }, + { + "epoch": 6.88, + "learning_rate": 2.3133883483187586e-06, + "loss": 0.3898, + "step": 18632 + }, + { + "epoch": 6.88, + "learning_rate": 2.3133513979554133e-06, + "loss": 0.4465, + "step": 18633 + }, + { + "epoch": 6.88, + "learning_rate": 2.313314447592068e-06, + "loss": 0.3936, + "step": 18634 + }, + { + "epoch": 6.88, + "learning_rate": 2.3132774972287226e-06, + "loss": 0.4969, + "step": 18635 + }, + { + "epoch": 6.88, + "learning_rate": 2.3132405468653776e-06, + "loss": 0.4577, + "step": 18636 + }, + { + "epoch": 6.88, + "learning_rate": 2.3132035965020323e-06, + "loss": 0.444, + "step": 18637 + }, + { + "epoch": 6.88, + "learning_rate": 2.3131666461386873e-06, + "loss": 0.5139, + "step": 18638 + }, + { + "epoch": 6.88, + "learning_rate": 2.313129695775342e-06, + "loss": 0.5914, + "step": 18639 + }, + { + "epoch": 6.88, + "learning_rate": 2.3130927454119966e-06, + "loss": 0.3467, + "step": 18640 + }, + { + "epoch": 6.88, + "learning_rate": 2.3130557950486512e-06, + "loss": 0.5936, + "step": 18641 + }, + { + "epoch": 6.88, + "learning_rate": 2.313018844685306e-06, + "loss": 0.5578, + "step": 18642 + }, + { + "epoch": 6.88, + "learning_rate": 2.312981894321961e-06, + "loss": 0.4126, + "step": 18643 + }, + { + "epoch": 6.88, + "learning_rate": 2.3129449439586156e-06, + "loss": 0.4019, + "step": 18644 + }, + { + "epoch": 6.89, + "learning_rate": 2.3129079935952706e-06, + "loss": 0.3626, + "step": 18645 + }, + { + "epoch": 6.89, + "learning_rate": 2.3128710432319253e-06, + "loss": 0.5588, + "step": 18646 + }, + { + "epoch": 6.89, + "learning_rate": 2.31283409286858e-06, + "loss": 0.4896, + "step": 18647 + }, + { + "epoch": 6.89, + "learning_rate": 2.3127971425052345e-06, + "loss": 0.3417, + "step": 18648 + }, + { + "epoch": 6.89, + "learning_rate": 2.3127601921418896e-06, + "loss": 0.391, + "step": 18649 + }, + { + "epoch": 6.89, + "learning_rate": 2.3127232417785442e-06, + "loss": 0.4832, + "step": 18650 + }, + { + "epoch": 6.89, + "learning_rate": 2.312686291415199e-06, + "loss": 0.4132, + "step": 18651 + }, + { + "epoch": 6.89, + "learning_rate": 2.312649341051854e-06, + "loss": 0.3703, + "step": 18652 + }, + { + "epoch": 6.89, + "learning_rate": 2.3126123906885086e-06, + "loss": 0.4379, + "step": 18653 + }, + { + "epoch": 6.89, + "learning_rate": 2.312575440325163e-06, + "loss": 0.4666, + "step": 18654 + }, + { + "epoch": 6.89, + "learning_rate": 2.312538489961818e-06, + "loss": 0.3476, + "step": 18655 + }, + { + "epoch": 6.89, + "learning_rate": 2.312501539598473e-06, + "loss": 0.5025, + "step": 18656 + }, + { + "epoch": 6.89, + "learning_rate": 2.3124645892351275e-06, + "loss": 0.5576, + "step": 18657 + }, + { + "epoch": 6.89, + "learning_rate": 2.3124276388717826e-06, + "loss": 0.467, + "step": 18658 + }, + { + "epoch": 6.89, + "learning_rate": 2.3123906885084372e-06, + "loss": 0.3463, + "step": 18659 + }, + { + "epoch": 6.89, + "learning_rate": 2.312353738145092e-06, + "loss": 0.4577, + "step": 18660 + }, + { + "epoch": 6.89, + "learning_rate": 2.3123167877817465e-06, + "loss": 0.469, + "step": 18661 + }, + { + "epoch": 6.89, + "learning_rate": 2.312279837418401e-06, + "loss": 0.4227, + "step": 18662 + }, + { + "epoch": 6.89, + "learning_rate": 2.312242887055056e-06, + "loss": 0.3711, + "step": 18663 + }, + { + "epoch": 6.89, + "learning_rate": 2.312205936691711e-06, + "loss": 0.4987, + "step": 18664 + }, + { + "epoch": 6.89, + "learning_rate": 2.312168986328366e-06, + "loss": 0.4409, + "step": 18665 + }, + { + "epoch": 6.89, + "learning_rate": 2.3121320359650205e-06, + "loss": 0.5538, + "step": 18666 + }, + { + "epoch": 6.89, + "learning_rate": 2.312095085601675e-06, + "loss": 0.4635, + "step": 18667 + }, + { + "epoch": 6.89, + "learning_rate": 2.31205813523833e-06, + "loss": 0.4593, + "step": 18668 + }, + { + "epoch": 6.89, + "learning_rate": 2.312021184874985e-06, + "loss": 0.5378, + "step": 18669 + }, + { + "epoch": 6.89, + "learning_rate": 2.3119842345116395e-06, + "loss": 0.4624, + "step": 18670 + }, + { + "epoch": 6.89, + "learning_rate": 2.311947284148294e-06, + "loss": 0.4262, + "step": 18671 + }, + { + "epoch": 6.9, + "learning_rate": 2.311910333784949e-06, + "loss": 0.3969, + "step": 18672 + }, + { + "epoch": 6.9, + "learning_rate": 2.3118733834216034e-06, + "loss": 0.3729, + "step": 18673 + }, + { + "epoch": 6.9, + "learning_rate": 2.3118364330582585e-06, + "loss": 0.4437, + "step": 18674 + }, + { + "epoch": 6.9, + "learning_rate": 2.311799482694913e-06, + "loss": 0.4611, + "step": 18675 + }, + { + "epoch": 6.9, + "learning_rate": 2.311762532331568e-06, + "loss": 0.4713, + "step": 18676 + }, + { + "epoch": 6.9, + "learning_rate": 2.311725581968223e-06, + "loss": 0.4021, + "step": 18677 + }, + { + "epoch": 6.9, + "learning_rate": 2.311688631604878e-06, + "loss": 0.4268, + "step": 18678 + }, + { + "epoch": 6.9, + "learning_rate": 2.311651681241532e-06, + "loss": 0.4283, + "step": 18679 + }, + { + "epoch": 6.9, + "learning_rate": 2.3116147308781867e-06, + "loss": 0.5021, + "step": 18680 + }, + { + "epoch": 6.9, + "learning_rate": 2.3115777805148418e-06, + "loss": 0.3478, + "step": 18681 + }, + { + "epoch": 6.9, + "learning_rate": 2.3115408301514964e-06, + "loss": 0.4643, + "step": 18682 + }, + { + "epoch": 6.9, + "learning_rate": 2.3115038797881515e-06, + "loss": 0.4346, + "step": 18683 + }, + { + "epoch": 6.9, + "learning_rate": 2.311466929424806e-06, + "loss": 0.4602, + "step": 18684 + }, + { + "epoch": 6.9, + "learning_rate": 2.3114299790614607e-06, + "loss": 0.4658, + "step": 18685 + }, + { + "epoch": 6.9, + "learning_rate": 2.3113930286981154e-06, + "loss": 0.3432, + "step": 18686 + }, + { + "epoch": 6.9, + "learning_rate": 2.3113560783347704e-06, + "loss": 0.3854, + "step": 18687 + }, + { + "epoch": 6.9, + "learning_rate": 2.311319127971425e-06, + "loss": 0.6031, + "step": 18688 + }, + { + "epoch": 6.9, + "learning_rate": 2.31128217760808e-06, + "loss": 0.4115, + "step": 18689 + }, + { + "epoch": 6.9, + "learning_rate": 2.3112452272447348e-06, + "loss": 0.3991, + "step": 18690 + }, + { + "epoch": 6.9, + "learning_rate": 2.3112082768813894e-06, + "loss": 0.5083, + "step": 18691 + }, + { + "epoch": 6.9, + "learning_rate": 2.311171326518044e-06, + "loss": 0.5129, + "step": 18692 + }, + { + "epoch": 6.9, + "learning_rate": 2.3111343761546987e-06, + "loss": 0.3939, + "step": 18693 + }, + { + "epoch": 6.9, + "learning_rate": 2.3110974257913537e-06, + "loss": 0.5021, + "step": 18694 + }, + { + "epoch": 6.9, + "learning_rate": 2.3110604754280084e-06, + "loss": 0.5383, + "step": 18695 + }, + { + "epoch": 6.9, + "learning_rate": 2.3110235250646634e-06, + "loss": 0.5372, + "step": 18696 + }, + { + "epoch": 6.9, + "learning_rate": 2.310986574701318e-06, + "loss": 0.307, + "step": 18697 + }, + { + "epoch": 6.9, + "learning_rate": 2.3109496243379727e-06, + "loss": 0.4024, + "step": 18698 + }, + { + "epoch": 6.91, + "learning_rate": 2.3109126739746273e-06, + "loss": 0.4734, + "step": 18699 + }, + { + "epoch": 6.91, + "learning_rate": 2.310875723611282e-06, + "loss": 0.4084, + "step": 18700 + }, + { + "epoch": 6.91, + "learning_rate": 2.310838773247937e-06, + "loss": 0.3634, + "step": 18701 + }, + { + "epoch": 6.91, + "learning_rate": 2.3108018228845917e-06, + "loss": 0.2767, + "step": 18702 + }, + { + "epoch": 6.91, + "learning_rate": 2.3107648725212467e-06, + "loss": 0.4315, + "step": 18703 + }, + { + "epoch": 6.91, + "learning_rate": 2.3107279221579014e-06, + "loss": 0.41, + "step": 18704 + }, + { + "epoch": 6.91, + "learning_rate": 2.310690971794556e-06, + "loss": 0.4751, + "step": 18705 + }, + { + "epoch": 6.91, + "learning_rate": 2.3106540214312106e-06, + "loss": 0.4088, + "step": 18706 + }, + { + "epoch": 6.91, + "learning_rate": 2.3106170710678657e-06, + "loss": 0.336, + "step": 18707 + }, + { + "epoch": 6.91, + "learning_rate": 2.3105801207045203e-06, + "loss": 0.4199, + "step": 18708 + }, + { + "epoch": 6.91, + "learning_rate": 2.3105431703411754e-06, + "loss": 0.4506, + "step": 18709 + }, + { + "epoch": 6.91, + "learning_rate": 2.31050621997783e-06, + "loss": 0.3831, + "step": 18710 + }, + { + "epoch": 6.91, + "learning_rate": 2.3104692696144847e-06, + "loss": 0.5521, + "step": 18711 + }, + { + "epoch": 6.91, + "learning_rate": 2.3104323192511393e-06, + "loss": 0.5709, + "step": 18712 + }, + { + "epoch": 6.91, + "learning_rate": 2.310395368887794e-06, + "loss": 0.6186, + "step": 18713 + }, + { + "epoch": 6.91, + "learning_rate": 2.310358418524449e-06, + "loss": 0.4075, + "step": 18714 + }, + { + "epoch": 6.91, + "learning_rate": 2.3103214681611036e-06, + "loss": 0.5833, + "step": 18715 + }, + { + "epoch": 6.91, + "learning_rate": 2.3102845177977587e-06, + "loss": 0.5284, + "step": 18716 + }, + { + "epoch": 6.91, + "learning_rate": 2.310247567434413e-06, + "loss": 0.3758, + "step": 18717 + }, + { + "epoch": 6.91, + "learning_rate": 2.310210617071068e-06, + "loss": 0.3156, + "step": 18718 + }, + { + "epoch": 6.91, + "learning_rate": 2.3101736667077226e-06, + "loss": 0.4965, + "step": 18719 + }, + { + "epoch": 6.91, + "learning_rate": 2.3101367163443772e-06, + "loss": 0.3976, + "step": 18720 + }, + { + "epoch": 6.91, + "learning_rate": 2.3100997659810323e-06, + "loss": 0.4093, + "step": 18721 + }, + { + "epoch": 6.91, + "learning_rate": 2.310062815617687e-06, + "loss": 0.4234, + "step": 18722 + }, + { + "epoch": 6.91, + "learning_rate": 2.3100258652543416e-06, + "loss": 0.3468, + "step": 18723 + }, + { + "epoch": 6.91, + "learning_rate": 2.3099889148909962e-06, + "loss": 0.4355, + "step": 18724 + }, + { + "epoch": 6.91, + "learning_rate": 2.3099519645276513e-06, + "loss": 0.5026, + "step": 18725 + }, + { + "epoch": 6.92, + "learning_rate": 2.309915014164306e-06, + "loss": 0.3938, + "step": 18726 + }, + { + "epoch": 6.92, + "learning_rate": 2.309878063800961e-06, + "loss": 0.411, + "step": 18727 + }, + { + "epoch": 6.92, + "learning_rate": 2.3098411134376156e-06, + "loss": 0.5479, + "step": 18728 + }, + { + "epoch": 6.92, + "learning_rate": 2.3098041630742702e-06, + "loss": 0.5043, + "step": 18729 + }, + { + "epoch": 6.92, + "learning_rate": 2.309767212710925e-06, + "loss": 0.535, + "step": 18730 + }, + { + "epoch": 6.92, + "learning_rate": 2.3097302623475795e-06, + "loss": 0.4354, + "step": 18731 + }, + { + "epoch": 6.92, + "learning_rate": 2.3096933119842346e-06, + "loss": 0.4086, + "step": 18732 + }, + { + "epoch": 6.92, + "learning_rate": 2.309656361620889e-06, + "loss": 0.4138, + "step": 18733 + }, + { + "epoch": 6.92, + "learning_rate": 2.3096194112575443e-06, + "loss": 0.4436, + "step": 18734 + }, + { + "epoch": 6.92, + "learning_rate": 2.309582460894199e-06, + "loss": 0.3058, + "step": 18735 + }, + { + "epoch": 6.92, + "learning_rate": 2.3095455105308535e-06, + "loss": 0.3908, + "step": 18736 + }, + { + "epoch": 6.92, + "learning_rate": 2.309508560167508e-06, + "loss": 0.4219, + "step": 18737 + }, + { + "epoch": 6.92, + "learning_rate": 2.3094716098041632e-06, + "loss": 0.4877, + "step": 18738 + }, + { + "epoch": 6.92, + "learning_rate": 2.309434659440818e-06, + "loss": 0.3875, + "step": 18739 + }, + { + "epoch": 6.92, + "learning_rate": 2.3093977090774725e-06, + "loss": 0.4942, + "step": 18740 + }, + { + "epoch": 6.92, + "learning_rate": 2.3093607587141276e-06, + "loss": 0.4082, + "step": 18741 + }, + { + "epoch": 6.92, + "learning_rate": 2.309323808350782e-06, + "loss": 0.3756, + "step": 18742 + }, + { + "epoch": 6.92, + "learning_rate": 2.309286857987437e-06, + "loss": 0.3323, + "step": 18743 + }, + { + "epoch": 6.92, + "learning_rate": 2.3092499076240915e-06, + "loss": 0.4472, + "step": 18744 + }, + { + "epoch": 6.92, + "learning_rate": 2.3092129572607465e-06, + "loss": 0.5336, + "step": 18745 + }, + { + "epoch": 6.92, + "learning_rate": 2.309176006897401e-06, + "loss": 0.4682, + "step": 18746 + }, + { + "epoch": 6.92, + "learning_rate": 2.3091390565340562e-06, + "loss": 0.5105, + "step": 18747 + }, + { + "epoch": 6.92, + "learning_rate": 2.309102106170711e-06, + "loss": 0.424, + "step": 18748 + }, + { + "epoch": 6.92, + "learning_rate": 2.3090651558073655e-06, + "loss": 0.461, + "step": 18749 + }, + { + "epoch": 6.92, + "learning_rate": 2.30902820544402e-06, + "loss": 0.5026, + "step": 18750 + }, + { + "epoch": 6.92, + "learning_rate": 2.3089912550806748e-06, + "loss": 0.4839, + "step": 18751 + }, + { + "epoch": 6.92, + "learning_rate": 2.30895430471733e-06, + "loss": 0.5196, + "step": 18752 + }, + { + "epoch": 6.93, + "learning_rate": 2.3089173543539845e-06, + "loss": 0.4344, + "step": 18753 + }, + { + "epoch": 6.93, + "learning_rate": 2.3088804039906395e-06, + "loss": 0.4914, + "step": 18754 + }, + { + "epoch": 6.93, + "learning_rate": 2.308843453627294e-06, + "loss": 0.652, + "step": 18755 + }, + { + "epoch": 6.93, + "learning_rate": 2.308806503263949e-06, + "loss": 0.4202, + "step": 18756 + }, + { + "epoch": 6.93, + "learning_rate": 2.3087695529006034e-06, + "loss": 0.5299, + "step": 18757 + }, + { + "epoch": 6.93, + "learning_rate": 2.3087326025372585e-06, + "loss": 0.394, + "step": 18758 + }, + { + "epoch": 6.93, + "learning_rate": 2.308695652173913e-06, + "loss": 0.4094, + "step": 18759 + }, + { + "epoch": 6.93, + "learning_rate": 2.3086587018105678e-06, + "loss": 0.4511, + "step": 18760 + }, + { + "epoch": 6.93, + "learning_rate": 2.308621751447223e-06, + "loss": 0.461, + "step": 18761 + }, + { + "epoch": 6.93, + "learning_rate": 2.308584801083877e-06, + "loss": 0.3554, + "step": 18762 + }, + { + "epoch": 6.93, + "learning_rate": 2.308547850720532e-06, + "loss": 0.5337, + "step": 18763 + }, + { + "epoch": 6.93, + "learning_rate": 2.3085109003571868e-06, + "loss": 0.3935, + "step": 18764 + }, + { + "epoch": 6.93, + "learning_rate": 2.308473949993842e-06, + "loss": 0.4415, + "step": 18765 + }, + { + "epoch": 6.93, + "learning_rate": 2.3084369996304964e-06, + "loss": 0.3489, + "step": 18766 + }, + { + "epoch": 6.93, + "learning_rate": 2.3084000492671515e-06, + "loss": 0.3737, + "step": 18767 + }, + { + "epoch": 6.93, + "learning_rate": 2.3083630989038057e-06, + "loss": 0.3307, + "step": 18768 + }, + { + "epoch": 6.93, + "learning_rate": 2.3083261485404608e-06, + "loss": 0.4608, + "step": 18769 + }, + { + "epoch": 6.93, + "learning_rate": 2.3082891981771154e-06, + "loss": 0.4013, + "step": 18770 + }, + { + "epoch": 6.93, + "learning_rate": 2.30825224781377e-06, + "loss": 0.3887, + "step": 18771 + }, + { + "epoch": 6.93, + "learning_rate": 2.308215297450425e-06, + "loss": 0.5047, + "step": 18772 + }, + { + "epoch": 6.93, + "learning_rate": 2.3081783470870798e-06, + "loss": 0.3772, + "step": 18773 + }, + { + "epoch": 6.93, + "learning_rate": 2.3081413967237344e-06, + "loss": 0.4249, + "step": 18774 + }, + { + "epoch": 6.93, + "learning_rate": 2.308104446360389e-06, + "loss": 0.4081, + "step": 18775 + }, + { + "epoch": 6.93, + "learning_rate": 2.308067495997044e-06, + "loss": 0.38, + "step": 18776 + }, + { + "epoch": 6.93, + "learning_rate": 2.3080305456336987e-06, + "loss": 0.5178, + "step": 18777 + }, + { + "epoch": 6.93, + "learning_rate": 2.3079935952703538e-06, + "loss": 0.4173, + "step": 18778 + }, + { + "epoch": 6.93, + "learning_rate": 2.3079566449070084e-06, + "loss": 0.5913, + "step": 18779 + }, + { + "epoch": 6.94, + "learning_rate": 2.307919694543663e-06, + "loss": 0.4911, + "step": 18780 + }, + { + "epoch": 6.94, + "learning_rate": 2.3078827441803177e-06, + "loss": 0.4254, + "step": 18781 + }, + { + "epoch": 6.94, + "learning_rate": 2.3078457938169723e-06, + "loss": 0.4927, + "step": 18782 + }, + { + "epoch": 6.94, + "learning_rate": 2.3078088434536274e-06, + "loss": 0.4549, + "step": 18783 + }, + { + "epoch": 6.94, + "learning_rate": 2.307771893090282e-06, + "loss": 0.4584, + "step": 18784 + }, + { + "epoch": 6.94, + "learning_rate": 2.307734942726937e-06, + "loss": 0.3828, + "step": 18785 + }, + { + "epoch": 6.94, + "learning_rate": 2.3076979923635917e-06, + "loss": 0.3098, + "step": 18786 + }, + { + "epoch": 6.94, + "learning_rate": 2.3076610420002464e-06, + "loss": 0.4278, + "step": 18787 + }, + { + "epoch": 6.94, + "learning_rate": 2.307624091636901e-06, + "loss": 0.3678, + "step": 18788 + }, + { + "epoch": 6.94, + "learning_rate": 2.307587141273556e-06, + "loss": 0.4338, + "step": 18789 + }, + { + "epoch": 6.94, + "learning_rate": 2.3075501909102107e-06, + "loss": 0.3877, + "step": 18790 + }, + { + "epoch": 6.94, + "learning_rate": 2.3075132405468653e-06, + "loss": 0.4423, + "step": 18791 + }, + { + "epoch": 6.94, + "learning_rate": 2.3074762901835204e-06, + "loss": 0.4499, + "step": 18792 + }, + { + "epoch": 6.94, + "learning_rate": 2.307439339820175e-06, + "loss": 0.4875, + "step": 18793 + }, + { + "epoch": 6.94, + "learning_rate": 2.3074023894568297e-06, + "loss": 0.5541, + "step": 18794 + }, + { + "epoch": 6.94, + "learning_rate": 2.3073654390934843e-06, + "loss": 0.5629, + "step": 18795 + }, + { + "epoch": 6.94, + "learning_rate": 2.3073284887301394e-06, + "loss": 0.445, + "step": 18796 + }, + { + "epoch": 6.94, + "learning_rate": 2.307291538366794e-06, + "loss": 0.3391, + "step": 18797 + }, + { + "epoch": 6.94, + "learning_rate": 2.307254588003449e-06, + "loss": 0.4989, + "step": 18798 + }, + { + "epoch": 6.94, + "learning_rate": 2.3072176376401037e-06, + "loss": 0.4683, + "step": 18799 + }, + { + "epoch": 6.94, + "learning_rate": 2.3071806872767583e-06, + "loss": 0.4726, + "step": 18800 + }, + { + "epoch": 6.94, + "learning_rate": 2.307143736913413e-06, + "loss": 0.4779, + "step": 18801 + }, + { + "epoch": 6.94, + "learning_rate": 2.3071067865500676e-06, + "loss": 0.3412, + "step": 18802 + }, + { + "epoch": 6.94, + "learning_rate": 2.3070698361867227e-06, + "loss": 0.3863, + "step": 18803 + }, + { + "epoch": 6.94, + "learning_rate": 2.3070328858233773e-06, + "loss": 0.5807, + "step": 18804 + }, + { + "epoch": 6.94, + "learning_rate": 2.3069959354600324e-06, + "loss": 0.3436, + "step": 18805 + }, + { + "epoch": 6.94, + "learning_rate": 2.306958985096687e-06, + "loss": 0.5536, + "step": 18806 + }, + { + "epoch": 6.94, + "learning_rate": 2.3069220347333416e-06, + "loss": 0.3874, + "step": 18807 + }, + { + "epoch": 6.95, + "learning_rate": 2.3068850843699963e-06, + "loss": 0.4987, + "step": 18808 + }, + { + "epoch": 6.95, + "learning_rate": 2.3068481340066513e-06, + "loss": 0.432, + "step": 18809 + }, + { + "epoch": 6.95, + "learning_rate": 2.306811183643306e-06, + "loss": 0.5309, + "step": 18810 + }, + { + "epoch": 6.95, + "learning_rate": 2.3067742332799606e-06, + "loss": 0.4244, + "step": 18811 + }, + { + "epoch": 6.95, + "learning_rate": 2.3067372829166157e-06, + "loss": 0.4907, + "step": 18812 + }, + { + "epoch": 6.95, + "learning_rate": 2.30670033255327e-06, + "loss": 0.406, + "step": 18813 + }, + { + "epoch": 6.95, + "learning_rate": 2.306663382189925e-06, + "loss": 0.4446, + "step": 18814 + }, + { + "epoch": 6.95, + "learning_rate": 2.3066264318265796e-06, + "loss": 0.3129, + "step": 18815 + }, + { + "epoch": 6.95, + "learning_rate": 2.3065894814632346e-06, + "loss": 0.4695, + "step": 18816 + }, + { + "epoch": 6.95, + "learning_rate": 2.3065525310998893e-06, + "loss": 0.4811, + "step": 18817 + }, + { + "epoch": 6.95, + "learning_rate": 2.3065155807365443e-06, + "loss": 0.4129, + "step": 18818 + }, + { + "epoch": 6.95, + "learning_rate": 2.3064786303731985e-06, + "loss": 0.3867, + "step": 18819 + }, + { + "epoch": 6.95, + "learning_rate": 2.306441680009853e-06, + "loss": 0.4772, + "step": 18820 + }, + { + "epoch": 6.95, + "learning_rate": 2.3064047296465082e-06, + "loss": 0.52, + "step": 18821 + }, + { + "epoch": 6.95, + "learning_rate": 2.306367779283163e-06, + "loss": 0.4517, + "step": 18822 + }, + { + "epoch": 6.95, + "learning_rate": 2.306330828919818e-06, + "loss": 0.4141, + "step": 18823 + }, + { + "epoch": 6.95, + "learning_rate": 2.3062938785564726e-06, + "loss": 0.4324, + "step": 18824 + }, + { + "epoch": 6.95, + "learning_rate": 2.306256928193127e-06, + "loss": 0.3938, + "step": 18825 + }, + { + "epoch": 6.95, + "learning_rate": 2.306219977829782e-06, + "loss": 0.363, + "step": 18826 + }, + { + "epoch": 6.95, + "learning_rate": 2.306183027466437e-06, + "loss": 0.4595, + "step": 18827 + }, + { + "epoch": 6.95, + "learning_rate": 2.3061460771030915e-06, + "loss": 0.4524, + "step": 18828 + }, + { + "epoch": 6.95, + "learning_rate": 2.3061091267397466e-06, + "loss": 0.4113, + "step": 18829 + }, + { + "epoch": 6.95, + "learning_rate": 2.3060721763764012e-06, + "loss": 0.5118, + "step": 18830 + }, + { + "epoch": 6.95, + "learning_rate": 2.306035226013056e-06, + "loss": 0.3654, + "step": 18831 + }, + { + "epoch": 6.95, + "learning_rate": 2.3059982756497105e-06, + "loss": 0.3947, + "step": 18832 + }, + { + "epoch": 6.95, + "learning_rate": 2.305961325286365e-06, + "loss": 0.4348, + "step": 18833 + }, + { + "epoch": 6.95, + "learning_rate": 2.30592437492302e-06, + "loss": 0.3501, + "step": 18834 + }, + { + "epoch": 6.96, + "learning_rate": 2.305887424559675e-06, + "loss": 0.4469, + "step": 18835 + }, + { + "epoch": 6.96, + "learning_rate": 2.30585047419633e-06, + "loss": 0.3666, + "step": 18836 + }, + { + "epoch": 6.96, + "learning_rate": 2.3058135238329845e-06, + "loss": 0.4729, + "step": 18837 + }, + { + "epoch": 6.96, + "learning_rate": 2.305776573469639e-06, + "loss": 0.441, + "step": 18838 + }, + { + "epoch": 6.96, + "learning_rate": 2.305739623106294e-06, + "loss": 0.331, + "step": 18839 + }, + { + "epoch": 6.96, + "learning_rate": 2.3057026727429484e-06, + "loss": 0.3136, + "step": 18840 + }, + { + "epoch": 6.96, + "learning_rate": 2.3056657223796035e-06, + "loss": 0.3742, + "step": 18841 + }, + { + "epoch": 6.96, + "learning_rate": 2.305628772016258e-06, + "loss": 0.388, + "step": 18842 + }, + { + "epoch": 6.96, + "learning_rate": 2.305591821652913e-06, + "loss": 0.4219, + "step": 18843 + }, + { + "epoch": 6.96, + "learning_rate": 2.305554871289568e-06, + "loss": 0.4835, + "step": 18844 + }, + { + "epoch": 6.96, + "learning_rate": 2.3055179209262225e-06, + "loss": 0.3648, + "step": 18845 + }, + { + "epoch": 6.96, + "learning_rate": 2.305480970562877e-06, + "loss": 0.4731, + "step": 18846 + }, + { + "epoch": 6.96, + "learning_rate": 2.305444020199532e-06, + "loss": 0.4624, + "step": 18847 + }, + { + "epoch": 6.96, + "learning_rate": 2.305407069836187e-06, + "loss": 0.4387, + "step": 18848 + }, + { + "epoch": 6.96, + "learning_rate": 2.3053701194728414e-06, + "loss": 0.4732, + "step": 18849 + }, + { + "epoch": 6.96, + "learning_rate": 2.3053331691094965e-06, + "loss": 0.6275, + "step": 18850 + }, + { + "epoch": 6.96, + "learning_rate": 2.305296218746151e-06, + "loss": 0.4152, + "step": 18851 + }, + { + "epoch": 6.96, + "learning_rate": 2.3052592683828058e-06, + "loss": 0.4725, + "step": 18852 + }, + { + "epoch": 6.96, + "learning_rate": 2.3052223180194604e-06, + "loss": 0.4281, + "step": 18853 + }, + { + "epoch": 6.96, + "learning_rate": 2.3051853676561155e-06, + "loss": 0.5457, + "step": 18854 + }, + { + "epoch": 6.96, + "learning_rate": 2.30514841729277e-06, + "loss": 0.4238, + "step": 18855 + }, + { + "epoch": 6.96, + "learning_rate": 2.305111466929425e-06, + "loss": 0.4819, + "step": 18856 + }, + { + "epoch": 6.96, + "learning_rate": 2.30507451656608e-06, + "loss": 0.4128, + "step": 18857 + }, + { + "epoch": 6.96, + "learning_rate": 2.3050375662027344e-06, + "loss": 0.3206, + "step": 18858 + }, + { + "epoch": 6.96, + "learning_rate": 2.305000615839389e-06, + "loss": 0.4485, + "step": 18859 + }, + { + "epoch": 6.96, + "learning_rate": 2.3049636654760437e-06, + "loss": 0.4585, + "step": 18860 + }, + { + "epoch": 6.96, + "learning_rate": 2.3049267151126988e-06, + "loss": 0.4083, + "step": 18861 + }, + { + "epoch": 6.97, + "learning_rate": 2.3048897647493534e-06, + "loss": 0.4692, + "step": 18862 + }, + { + "epoch": 6.97, + "learning_rate": 2.3048528143860085e-06, + "loss": 0.3525, + "step": 18863 + }, + { + "epoch": 6.97, + "learning_rate": 2.3048158640226627e-06, + "loss": 0.3811, + "step": 18864 + }, + { + "epoch": 6.97, + "learning_rate": 2.3047789136593177e-06, + "loss": 0.4748, + "step": 18865 + }, + { + "epoch": 6.97, + "learning_rate": 2.3047419632959724e-06, + "loss": 0.4551, + "step": 18866 + }, + { + "epoch": 6.97, + "learning_rate": 2.3047050129326274e-06, + "loss": 0.426, + "step": 18867 + }, + { + "epoch": 6.97, + "learning_rate": 2.304668062569282e-06, + "loss": 0.5107, + "step": 18868 + }, + { + "epoch": 6.97, + "learning_rate": 2.3046311122059367e-06, + "loss": 0.5496, + "step": 18869 + }, + { + "epoch": 6.97, + "learning_rate": 2.3045941618425913e-06, + "loss": 0.3752, + "step": 18870 + }, + { + "epoch": 6.97, + "learning_rate": 2.304557211479246e-06, + "loss": 0.5544, + "step": 18871 + }, + { + "epoch": 6.97, + "learning_rate": 2.304520261115901e-06, + "loss": 0.5268, + "step": 18872 + }, + { + "epoch": 6.97, + "learning_rate": 2.3044833107525557e-06, + "loss": 0.4946, + "step": 18873 + }, + { + "epoch": 6.97, + "learning_rate": 2.3044463603892107e-06, + "loss": 0.4555, + "step": 18874 + }, + { + "epoch": 6.97, + "learning_rate": 2.3044094100258654e-06, + "loss": 0.461, + "step": 18875 + }, + { + "epoch": 6.97, + "learning_rate": 2.30437245966252e-06, + "loss": 0.3568, + "step": 18876 + }, + { + "epoch": 6.97, + "learning_rate": 2.3043355092991746e-06, + "loss": 0.4919, + "step": 18877 + }, + { + "epoch": 6.97, + "learning_rate": 2.3042985589358297e-06, + "loss": 0.4261, + "step": 18878 + }, + { + "epoch": 6.97, + "learning_rate": 2.3042616085724843e-06, + "loss": 0.4912, + "step": 18879 + }, + { + "epoch": 6.97, + "learning_rate": 2.304224658209139e-06, + "loss": 0.4577, + "step": 18880 + }, + { + "epoch": 6.97, + "learning_rate": 2.304187707845794e-06, + "loss": 0.5412, + "step": 18881 + }, + { + "epoch": 6.97, + "learning_rate": 2.3041507574824487e-06, + "loss": 0.5339, + "step": 18882 + }, + { + "epoch": 6.97, + "learning_rate": 2.3041138071191033e-06, + "loss": 0.4954, + "step": 18883 + }, + { + "epoch": 6.97, + "learning_rate": 2.304076856755758e-06, + "loss": 0.359, + "step": 18884 + }, + { + "epoch": 6.97, + "learning_rate": 2.304039906392413e-06, + "loss": 0.4463, + "step": 18885 + }, + { + "epoch": 6.97, + "learning_rate": 2.3040029560290676e-06, + "loss": 0.3944, + "step": 18886 + }, + { + "epoch": 6.97, + "learning_rate": 2.3039660056657227e-06, + "loss": 0.4454, + "step": 18887 + }, + { + "epoch": 6.97, + "learning_rate": 2.3039290553023773e-06, + "loss": 0.579, + "step": 18888 + }, + { + "epoch": 6.98, + "learning_rate": 2.303892104939032e-06, + "loss": 0.3572, + "step": 18889 + }, + { + "epoch": 6.98, + "learning_rate": 2.3038551545756866e-06, + "loss": 0.2895, + "step": 18890 + }, + { + "epoch": 6.98, + "learning_rate": 2.3038182042123412e-06, + "loss": 0.5208, + "step": 18891 + }, + { + "epoch": 6.98, + "learning_rate": 2.3037812538489963e-06, + "loss": 0.3692, + "step": 18892 + }, + { + "epoch": 6.98, + "learning_rate": 2.303744303485651e-06, + "loss": 0.494, + "step": 18893 + }, + { + "epoch": 6.98, + "learning_rate": 2.303707353122306e-06, + "loss": 0.5965, + "step": 18894 + }, + { + "epoch": 6.98, + "learning_rate": 2.3036704027589606e-06, + "loss": 0.3745, + "step": 18895 + }, + { + "epoch": 6.98, + "learning_rate": 2.3036334523956153e-06, + "loss": 0.3826, + "step": 18896 + }, + { + "epoch": 6.98, + "learning_rate": 2.30359650203227e-06, + "loss": 0.4089, + "step": 18897 + }, + { + "epoch": 6.98, + "learning_rate": 2.303559551668925e-06, + "loss": 0.3748, + "step": 18898 + }, + { + "epoch": 6.98, + "learning_rate": 2.3035226013055796e-06, + "loss": 0.4634, + "step": 18899 + }, + { + "epoch": 6.98, + "learning_rate": 2.3034856509422342e-06, + "loss": 0.5641, + "step": 18900 + }, + { + "epoch": 6.98, + "learning_rate": 2.3034487005788893e-06, + "loss": 0.4962, + "step": 18901 + }, + { + "epoch": 6.98, + "learning_rate": 2.303411750215544e-06, + "loss": 0.3977, + "step": 18902 + }, + { + "epoch": 6.98, + "learning_rate": 2.3033747998521986e-06, + "loss": 0.4125, + "step": 18903 + }, + { + "epoch": 6.98, + "learning_rate": 2.3033378494888532e-06, + "loss": 0.4855, + "step": 18904 + }, + { + "epoch": 6.98, + "learning_rate": 2.3033008991255083e-06, + "loss": 0.4619, + "step": 18905 + }, + { + "epoch": 6.98, + "learning_rate": 2.303263948762163e-06, + "loss": 0.3707, + "step": 18906 + }, + { + "epoch": 6.98, + "learning_rate": 2.303226998398818e-06, + "loss": 0.4307, + "step": 18907 + }, + { + "epoch": 6.98, + "learning_rate": 2.303190048035472e-06, + "loss": 0.3662, + "step": 18908 + }, + { + "epoch": 6.98, + "learning_rate": 2.303153097672127e-06, + "loss": 0.5426, + "step": 18909 + }, + { + "epoch": 6.98, + "learning_rate": 2.303116147308782e-06, + "loss": 0.5038, + "step": 18910 + }, + { + "epoch": 6.98, + "learning_rate": 2.3030791969454365e-06, + "loss": 0.522, + "step": 18911 + }, + { + "epoch": 6.98, + "learning_rate": 2.3030422465820916e-06, + "loss": 0.3902, + "step": 18912 + }, + { + "epoch": 6.98, + "learning_rate": 2.3030052962187462e-06, + "loss": 0.3834, + "step": 18913 + }, + { + "epoch": 6.98, + "learning_rate": 2.302968345855401e-06, + "loss": 0.3885, + "step": 18914 + }, + { + "epoch": 6.98, + "learning_rate": 2.3029313954920555e-06, + "loss": 0.4574, + "step": 18915 + }, + { + "epoch": 6.99, + "learning_rate": 2.3028944451287105e-06, + "loss": 0.5239, + "step": 18916 + }, + { + "epoch": 6.99, + "learning_rate": 2.302857494765365e-06, + "loss": 0.4714, + "step": 18917 + }, + { + "epoch": 6.99, + "learning_rate": 2.3028205444020202e-06, + "loss": 0.4413, + "step": 18918 + }, + { + "epoch": 6.99, + "learning_rate": 2.302783594038675e-06, + "loss": 0.3843, + "step": 18919 + }, + { + "epoch": 6.99, + "learning_rate": 2.3027466436753295e-06, + "loss": 0.4123, + "step": 18920 + }, + { + "epoch": 6.99, + "learning_rate": 2.302709693311984e-06, + "loss": 0.4696, + "step": 18921 + }, + { + "epoch": 6.99, + "learning_rate": 2.302672742948639e-06, + "loss": 0.3383, + "step": 18922 + }, + { + "epoch": 6.99, + "learning_rate": 2.302635792585294e-06, + "loss": 0.4382, + "step": 18923 + }, + { + "epoch": 6.99, + "learning_rate": 2.3025988422219485e-06, + "loss": 0.4666, + "step": 18924 + }, + { + "epoch": 6.99, + "learning_rate": 2.3025618918586035e-06, + "loss": 0.5301, + "step": 18925 + }, + { + "epoch": 6.99, + "learning_rate": 2.302524941495258e-06, + "loss": 0.4895, + "step": 18926 + }, + { + "epoch": 6.99, + "learning_rate": 2.302487991131913e-06, + "loss": 0.5793, + "step": 18927 + }, + { + "epoch": 6.99, + "learning_rate": 2.3024510407685675e-06, + "loss": 0.487, + "step": 18928 + }, + { + "epoch": 6.99, + "learning_rate": 2.302414090405222e-06, + "loss": 0.4742, + "step": 18929 + }, + { + "epoch": 6.99, + "learning_rate": 2.302377140041877e-06, + "loss": 0.4329, + "step": 18930 + }, + { + "epoch": 6.99, + "learning_rate": 2.302340189678532e-06, + "loss": 0.4509, + "step": 18931 + }, + { + "epoch": 6.99, + "learning_rate": 2.302303239315187e-06, + "loss": 0.4886, + "step": 18932 + }, + { + "epoch": 6.99, + "learning_rate": 2.3022662889518415e-06, + "loss": 0.5528, + "step": 18933 + }, + { + "epoch": 6.99, + "learning_rate": 2.302229338588496e-06, + "loss": 0.4568, + "step": 18934 + }, + { + "epoch": 6.99, + "learning_rate": 2.3021923882251508e-06, + "loss": 0.4889, + "step": 18935 + }, + { + "epoch": 6.99, + "learning_rate": 2.302155437861806e-06, + "loss": 0.4836, + "step": 18936 + }, + { + "epoch": 6.99, + "learning_rate": 2.3021184874984605e-06, + "loss": 0.402, + "step": 18937 + }, + { + "epoch": 6.99, + "learning_rate": 2.3020815371351155e-06, + "loss": 0.4532, + "step": 18938 + }, + { + "epoch": 6.99, + "learning_rate": 2.30204458677177e-06, + "loss": 0.4048, + "step": 18939 + }, + { + "epoch": 6.99, + "learning_rate": 2.3020076364084248e-06, + "loss": 0.4426, + "step": 18940 + }, + { + "epoch": 6.99, + "learning_rate": 2.3019706860450794e-06, + "loss": 0.3752, + "step": 18941 + }, + { + "epoch": 6.99, + "learning_rate": 2.301933735681734e-06, + "loss": 0.593, + "step": 18942 + }, + { + "epoch": 7.0, + "learning_rate": 2.301896785318389e-06, + "loss": 0.4323, + "step": 18943 + }, + { + "epoch": 7.0, + "learning_rate": 2.3018598349550438e-06, + "loss": 0.3992, + "step": 18944 + }, + { + "epoch": 7.0, + "learning_rate": 2.301822884591699e-06, + "loss": 0.4479, + "step": 18945 + }, + { + "epoch": 7.0, + "learning_rate": 2.3017859342283535e-06, + "loss": 0.4045, + "step": 18946 + }, + { + "epoch": 7.0, + "learning_rate": 2.301748983865008e-06, + "loss": 0.3946, + "step": 18947 + }, + { + "epoch": 7.0, + "learning_rate": 2.3017120335016627e-06, + "loss": 0.4598, + "step": 18948 + }, + { + "epoch": 7.0, + "learning_rate": 2.3016750831383174e-06, + "loss": 0.5166, + "step": 18949 + }, + { + "epoch": 7.0, + "learning_rate": 2.3016381327749724e-06, + "loss": 0.4498, + "step": 18950 + }, + { + "epoch": 7.0, + "learning_rate": 2.301601182411627e-06, + "loss": 0.3544, + "step": 18951 + }, + { + "epoch": 7.0, + "learning_rate": 2.301564232048282e-06, + "loss": 0.4745, + "step": 18952 + }, + { + "epoch": 7.0, + "learning_rate": 2.3015272816849363e-06, + "loss": 0.4543, + "step": 18953 + }, + { + "epoch": 7.0, + "learning_rate": 2.3014903313215914e-06, + "loss": 0.5206, + "step": 18954 + }, + { + "epoch": 7.0, + "learning_rate": 2.301453380958246e-06, + "loss": 0.504, + "step": 18955 + }, + { + "epoch": 7.0, + "learning_rate": 2.301416430594901e-06, + "loss": 0.5081, + "step": 18956 + }, + { + "epoch": 7.0, + "eval_accuracy": 0.7711594403657016, + "eval_loss": 0.4822496771812439, + "eval_runtime": 25.179, + "eval_samples_per_second": 1720.246, + "eval_steps_per_second": 26.888, + "step": 18956 + }, + { + "epoch": 7.0, + "learning_rate": 2.3013794802315557e-06, + "loss": 0.3942, + "step": 18957 + }, + { + "epoch": 7.0, + "learning_rate": 2.3013425298682108e-06, + "loss": 0.3667, + "step": 18958 + }, + { + "epoch": 7.0, + "learning_rate": 2.301305579504865e-06, + "loss": 0.4456, + "step": 18959 + }, + { + "epoch": 7.0, + "learning_rate": 2.3012686291415196e-06, + "loss": 0.3589, + "step": 18960 + }, + { + "epoch": 7.0, + "learning_rate": 2.3012316787781747e-06, + "loss": 0.4583, + "step": 18961 + }, + { + "epoch": 7.0, + "learning_rate": 2.3011947284148293e-06, + "loss": 0.47, + "step": 18962 + }, + { + "epoch": 7.0, + "learning_rate": 2.3011577780514844e-06, + "loss": 0.4159, + "step": 18963 + }, + { + "epoch": 7.0, + "learning_rate": 2.301120827688139e-06, + "loss": 0.5245, + "step": 18964 + }, + { + "epoch": 7.0, + "learning_rate": 2.3010838773247937e-06, + "loss": 0.4351, + "step": 18965 + }, + { + "epoch": 7.0, + "learning_rate": 2.3010469269614483e-06, + "loss": 0.4713, + "step": 18966 + }, + { + "epoch": 7.0, + "learning_rate": 2.3010099765981034e-06, + "loss": 0.397, + "step": 18967 + }, + { + "epoch": 7.0, + "learning_rate": 2.300973026234758e-06, + "loss": 0.4287, + "step": 18968 + }, + { + "epoch": 7.0, + "learning_rate": 2.3009360758714126e-06, + "loss": 0.401, + "step": 18969 + }, + { + "epoch": 7.01, + "learning_rate": 2.3008991255080677e-06, + "loss": 0.4084, + "step": 18970 + }, + { + "epoch": 7.01, + "learning_rate": 2.3008621751447223e-06, + "loss": 0.4258, + "step": 18971 + }, + { + "epoch": 7.01, + "learning_rate": 2.300825224781377e-06, + "loss": 0.3794, + "step": 18972 + }, + { + "epoch": 7.01, + "learning_rate": 2.3007882744180316e-06, + "loss": 0.5053, + "step": 18973 + }, + { + "epoch": 7.01, + "learning_rate": 2.3007513240546867e-06, + "loss": 0.4339, + "step": 18974 + }, + { + "epoch": 7.01, + "learning_rate": 2.3007143736913413e-06, + "loss": 0.5361, + "step": 18975 + }, + { + "epoch": 7.01, + "learning_rate": 2.3006774233279964e-06, + "loss": 0.3568, + "step": 18976 + }, + { + "epoch": 7.01, + "learning_rate": 2.300640472964651e-06, + "loss": 0.3852, + "step": 18977 + }, + { + "epoch": 7.01, + "learning_rate": 2.3006035226013056e-06, + "loss": 0.3927, + "step": 18978 + }, + { + "epoch": 7.01, + "learning_rate": 2.3005665722379603e-06, + "loss": 0.49, + "step": 18979 + }, + { + "epoch": 7.01, + "learning_rate": 2.300529621874615e-06, + "loss": 0.3995, + "step": 18980 + }, + { + "epoch": 7.01, + "learning_rate": 2.30049267151127e-06, + "loss": 0.4254, + "step": 18981 + }, + { + "epoch": 7.01, + "learning_rate": 2.3004557211479246e-06, + "loss": 0.5437, + "step": 18982 + }, + { + "epoch": 7.01, + "learning_rate": 2.3004187707845797e-06, + "loss": 0.4103, + "step": 18983 + }, + { + "epoch": 7.01, + "learning_rate": 2.3003818204212343e-06, + "loss": 0.4825, + "step": 18984 + }, + { + "epoch": 7.01, + "learning_rate": 2.300344870057889e-06, + "loss": 0.4686, + "step": 18985 + }, + { + "epoch": 7.01, + "learning_rate": 2.3003079196945436e-06, + "loss": 0.49, + "step": 18986 + }, + { + "epoch": 7.01, + "learning_rate": 2.3002709693311986e-06, + "loss": 0.409, + "step": 18987 + }, + { + "epoch": 7.01, + "learning_rate": 2.3002340189678533e-06, + "loss": 0.3526, + "step": 18988 + }, + { + "epoch": 7.01, + "learning_rate": 2.300197068604508e-06, + "loss": 0.3945, + "step": 18989 + }, + { + "epoch": 7.01, + "learning_rate": 2.300160118241163e-06, + "loss": 0.3877, + "step": 18990 + }, + { + "epoch": 7.01, + "learning_rate": 2.3001231678778176e-06, + "loss": 0.2951, + "step": 18991 + }, + { + "epoch": 7.01, + "learning_rate": 2.3000862175144722e-06, + "loss": 0.3926, + "step": 18992 + }, + { + "epoch": 7.01, + "learning_rate": 2.300049267151127e-06, + "loss": 0.3962, + "step": 18993 + }, + { + "epoch": 7.01, + "learning_rate": 2.300012316787782e-06, + "loss": 0.4734, + "step": 18994 + }, + { + "epoch": 7.01, + "learning_rate": 2.2999753664244366e-06, + "loss": 0.4836, + "step": 18995 + }, + { + "epoch": 7.01, + "learning_rate": 2.2999384160610916e-06, + "loss": 0.4241, + "step": 18996 + }, + { + "epoch": 7.02, + "learning_rate": 2.2999014656977463e-06, + "loss": 0.3897, + "step": 18997 + }, + { + "epoch": 7.02, + "learning_rate": 2.299864515334401e-06, + "loss": 0.4534, + "step": 18998 + }, + { + "epoch": 7.02, + "learning_rate": 2.2998275649710555e-06, + "loss": 0.4197, + "step": 18999 + }, + { + "epoch": 7.02, + "learning_rate": 2.29979061460771e-06, + "loss": 0.4252, + "step": 19000 + }, + { + "epoch": 7.02, + "learning_rate": 2.2997536642443652e-06, + "loss": 0.3137, + "step": 19001 + }, + { + "epoch": 7.02, + "learning_rate": 2.29971671388102e-06, + "loss": 0.4097, + "step": 19002 + }, + { + "epoch": 7.02, + "learning_rate": 2.299679763517675e-06, + "loss": 0.3292, + "step": 19003 + }, + { + "epoch": 7.02, + "learning_rate": 2.299642813154329e-06, + "loss": 0.4619, + "step": 19004 + }, + { + "epoch": 7.02, + "learning_rate": 2.299605862790984e-06, + "loss": 0.3729, + "step": 19005 + }, + { + "epoch": 7.02, + "learning_rate": 2.299568912427639e-06, + "loss": 0.4452, + "step": 19006 + }, + { + "epoch": 7.02, + "learning_rate": 2.299531962064294e-06, + "loss": 0.393, + "step": 19007 + }, + { + "epoch": 7.02, + "learning_rate": 2.2994950117009485e-06, + "loss": 0.3933, + "step": 19008 + }, + { + "epoch": 7.02, + "learning_rate": 2.299458061337603e-06, + "loss": 0.4646, + "step": 19009 + }, + { + "epoch": 7.02, + "learning_rate": 2.299421110974258e-06, + "loss": 0.5876, + "step": 19010 + }, + { + "epoch": 7.02, + "learning_rate": 2.2993841606109124e-06, + "loss": 0.4528, + "step": 19011 + }, + { + "epoch": 7.02, + "learning_rate": 2.2993472102475675e-06, + "loss": 0.4346, + "step": 19012 + }, + { + "epoch": 7.02, + "learning_rate": 2.299310259884222e-06, + "loss": 0.4163, + "step": 19013 + }, + { + "epoch": 7.02, + "learning_rate": 2.299273309520877e-06, + "loss": 0.5255, + "step": 19014 + }, + { + "epoch": 7.02, + "learning_rate": 2.299236359157532e-06, + "loss": 0.3747, + "step": 19015 + }, + { + "epoch": 7.02, + "learning_rate": 2.2991994087941865e-06, + "loss": 0.4326, + "step": 19016 + }, + { + "epoch": 7.02, + "learning_rate": 2.299162458430841e-06, + "loss": 0.383, + "step": 19017 + }, + { + "epoch": 7.02, + "learning_rate": 2.299125508067496e-06, + "loss": 0.3887, + "step": 19018 + }, + { + "epoch": 7.02, + "learning_rate": 2.299088557704151e-06, + "loss": 0.389, + "step": 19019 + }, + { + "epoch": 7.02, + "learning_rate": 2.2990516073408054e-06, + "loss": 0.5153, + "step": 19020 + }, + { + "epoch": 7.02, + "learning_rate": 2.2990146569774605e-06, + "loss": 0.3926, + "step": 19021 + }, + { + "epoch": 7.02, + "learning_rate": 2.298977706614115e-06, + "loss": 0.4654, + "step": 19022 + }, + { + "epoch": 7.02, + "learning_rate": 2.2989407562507698e-06, + "loss": 0.4207, + "step": 19023 + }, + { + "epoch": 7.03, + "learning_rate": 2.2989038058874244e-06, + "loss": 0.5207, + "step": 19024 + }, + { + "epoch": 7.03, + "learning_rate": 2.2988668555240795e-06, + "loss": 0.4498, + "step": 19025 + }, + { + "epoch": 7.03, + "learning_rate": 2.298829905160734e-06, + "loss": 0.5101, + "step": 19026 + }, + { + "epoch": 7.03, + "learning_rate": 2.298792954797389e-06, + "loss": 0.4674, + "step": 19027 + }, + { + "epoch": 7.03, + "learning_rate": 2.298756004434044e-06, + "loss": 0.5396, + "step": 19028 + }, + { + "epoch": 7.03, + "learning_rate": 2.2987190540706984e-06, + "loss": 0.5235, + "step": 19029 + }, + { + "epoch": 7.03, + "learning_rate": 2.298682103707353e-06, + "loss": 0.3297, + "step": 19030 + }, + { + "epoch": 7.03, + "learning_rate": 2.2986451533440077e-06, + "loss": 0.5406, + "step": 19031 + }, + { + "epoch": 7.03, + "learning_rate": 2.2986082029806628e-06, + "loss": 0.2949, + "step": 19032 + }, + { + "epoch": 7.03, + "learning_rate": 2.2985712526173174e-06, + "loss": 0.3395, + "step": 19033 + }, + { + "epoch": 7.03, + "learning_rate": 2.2985343022539725e-06, + "loss": 0.3937, + "step": 19034 + }, + { + "epoch": 7.03, + "learning_rate": 2.298497351890627e-06, + "loss": 0.4574, + "step": 19035 + }, + { + "epoch": 7.03, + "learning_rate": 2.2984604015272817e-06, + "loss": 0.581, + "step": 19036 + }, + { + "epoch": 7.03, + "learning_rate": 2.2984234511639364e-06, + "loss": 0.4043, + "step": 19037 + }, + { + "epoch": 7.03, + "learning_rate": 2.2983865008005914e-06, + "loss": 0.3138, + "step": 19038 + }, + { + "epoch": 7.03, + "learning_rate": 2.298349550437246e-06, + "loss": 0.5351, + "step": 19039 + }, + { + "epoch": 7.03, + "learning_rate": 2.2983126000739007e-06, + "loss": 0.4131, + "step": 19040 + }, + { + "epoch": 7.03, + "learning_rate": 2.2982756497105558e-06, + "loss": 0.3405, + "step": 19041 + }, + { + "epoch": 7.03, + "learning_rate": 2.2982386993472104e-06, + "loss": 0.4518, + "step": 19042 + }, + { + "epoch": 7.03, + "learning_rate": 2.298201748983865e-06, + "loss": 0.3125, + "step": 19043 + }, + { + "epoch": 7.03, + "learning_rate": 2.2981647986205197e-06, + "loss": 0.3954, + "step": 19044 + }, + { + "epoch": 7.03, + "learning_rate": 2.2981278482571747e-06, + "loss": 0.4913, + "step": 19045 + }, + { + "epoch": 7.03, + "learning_rate": 2.2980908978938294e-06, + "loss": 0.4386, + "step": 19046 + }, + { + "epoch": 7.03, + "learning_rate": 2.2980539475304844e-06, + "loss": 0.4801, + "step": 19047 + }, + { + "epoch": 7.03, + "learning_rate": 2.298016997167139e-06, + "loss": 0.5839, + "step": 19048 + }, + { + "epoch": 7.03, + "learning_rate": 2.2979800468037933e-06, + "loss": 0.4719, + "step": 19049 + }, + { + "epoch": 7.03, + "learning_rate": 2.2979430964404483e-06, + "loss": 0.5328, + "step": 19050 + }, + { + "epoch": 7.04, + "learning_rate": 2.297906146077103e-06, + "loss": 0.503, + "step": 19051 + }, + { + "epoch": 7.04, + "learning_rate": 2.297869195713758e-06, + "loss": 0.3761, + "step": 19052 + }, + { + "epoch": 7.04, + "learning_rate": 2.2978322453504127e-06, + "loss": 0.4687, + "step": 19053 + }, + { + "epoch": 7.04, + "learning_rate": 2.2977952949870677e-06, + "loss": 0.5142, + "step": 19054 + }, + { + "epoch": 7.04, + "learning_rate": 2.297758344623722e-06, + "loss": 0.5059, + "step": 19055 + }, + { + "epoch": 7.04, + "learning_rate": 2.297721394260377e-06, + "loss": 0.4645, + "step": 19056 + }, + { + "epoch": 7.04, + "learning_rate": 2.2976844438970316e-06, + "loss": 0.4004, + "step": 19057 + }, + { + "epoch": 7.04, + "learning_rate": 2.2976474935336867e-06, + "loss": 0.4727, + "step": 19058 + }, + { + "epoch": 7.04, + "learning_rate": 2.2976105431703413e-06, + "loss": 0.4505, + "step": 19059 + }, + { + "epoch": 7.04, + "learning_rate": 2.297573592806996e-06, + "loss": 0.413, + "step": 19060 + }, + { + "epoch": 7.04, + "learning_rate": 2.2975366424436506e-06, + "loss": 0.334, + "step": 19061 + }, + { + "epoch": 7.04, + "learning_rate": 2.2974996920803053e-06, + "loss": 0.513, + "step": 19062 + }, + { + "epoch": 7.04, + "learning_rate": 2.2974627417169603e-06, + "loss": 0.396, + "step": 19063 + }, + { + "epoch": 7.04, + "learning_rate": 2.297425791353615e-06, + "loss": 0.428, + "step": 19064 + }, + { + "epoch": 7.04, + "learning_rate": 2.29738884099027e-06, + "loss": 0.3457, + "step": 19065 + }, + { + "epoch": 7.04, + "learning_rate": 2.2973518906269246e-06, + "loss": 0.3326, + "step": 19066 + }, + { + "epoch": 7.04, + "learning_rate": 2.2973149402635793e-06, + "loss": 0.482, + "step": 19067 + }, + { + "epoch": 7.04, + "learning_rate": 2.297277989900234e-06, + "loss": 0.4793, + "step": 19068 + }, + { + "epoch": 7.04, + "learning_rate": 2.2972410395368886e-06, + "loss": 0.383, + "step": 19069 + }, + { + "epoch": 7.04, + "learning_rate": 2.2972040891735436e-06, + "loss": 0.3816, + "step": 19070 + }, + { + "epoch": 7.04, + "learning_rate": 2.2971671388101983e-06, + "loss": 0.4015, + "step": 19071 + }, + { + "epoch": 7.04, + "learning_rate": 2.2971301884468533e-06, + "loss": 0.4582, + "step": 19072 + }, + { + "epoch": 7.04, + "learning_rate": 2.297093238083508e-06, + "loss": 0.4043, + "step": 19073 + }, + { + "epoch": 7.04, + "learning_rate": 2.2970562877201626e-06, + "loss": 0.4382, + "step": 19074 + }, + { + "epoch": 7.04, + "learning_rate": 2.2970193373568172e-06, + "loss": 0.374, + "step": 19075 + }, + { + "epoch": 7.04, + "learning_rate": 2.2969823869934723e-06, + "loss": 0.4115, + "step": 19076 + }, + { + "epoch": 7.04, + "learning_rate": 2.296945436630127e-06, + "loss": 0.4768, + "step": 19077 + }, + { + "epoch": 7.05, + "learning_rate": 2.2969084862667816e-06, + "loss": 0.4497, + "step": 19078 + }, + { + "epoch": 7.05, + "learning_rate": 2.2968715359034366e-06, + "loss": 0.3345, + "step": 19079 + }, + { + "epoch": 7.05, + "learning_rate": 2.2968345855400913e-06, + "loss": 0.3512, + "step": 19080 + }, + { + "epoch": 7.05, + "learning_rate": 2.296797635176746e-06, + "loss": 0.4953, + "step": 19081 + }, + { + "epoch": 7.05, + "learning_rate": 2.2967606848134005e-06, + "loss": 0.3724, + "step": 19082 + }, + { + "epoch": 7.05, + "learning_rate": 2.2967237344500556e-06, + "loss": 0.2606, + "step": 19083 + }, + { + "epoch": 7.05, + "learning_rate": 2.2966867840867102e-06, + "loss": 0.5071, + "step": 19084 + }, + { + "epoch": 7.05, + "learning_rate": 2.2966498337233653e-06, + "loss": 0.2997, + "step": 19085 + }, + { + "epoch": 7.05, + "learning_rate": 2.29661288336002e-06, + "loss": 0.5459, + "step": 19086 + }, + { + "epoch": 7.05, + "learning_rate": 2.2965759329966746e-06, + "loss": 0.336, + "step": 19087 + }, + { + "epoch": 7.05, + "learning_rate": 2.296538982633329e-06, + "loss": 0.4225, + "step": 19088 + }, + { + "epoch": 7.05, + "learning_rate": 2.296502032269984e-06, + "loss": 0.4, + "step": 19089 + }, + { + "epoch": 7.05, + "learning_rate": 2.296465081906639e-06, + "loss": 0.4267, + "step": 19090 + }, + { + "epoch": 7.05, + "learning_rate": 2.2964281315432935e-06, + "loss": 0.5213, + "step": 19091 + }, + { + "epoch": 7.05, + "learning_rate": 2.2963911811799486e-06, + "loss": 0.3458, + "step": 19092 + }, + { + "epoch": 7.05, + "learning_rate": 2.296354230816603e-06, + "loss": 0.3688, + "step": 19093 + }, + { + "epoch": 7.05, + "learning_rate": 2.296317280453258e-06, + "loss": 0.4143, + "step": 19094 + }, + { + "epoch": 7.05, + "learning_rate": 2.2962803300899125e-06, + "loss": 0.4434, + "step": 19095 + }, + { + "epoch": 7.05, + "learning_rate": 2.2962433797265676e-06, + "loss": 0.4253, + "step": 19096 + }, + { + "epoch": 7.05, + "learning_rate": 2.296206429363222e-06, + "loss": 0.44, + "step": 19097 + }, + { + "epoch": 7.05, + "learning_rate": 2.296169478999877e-06, + "loss": 0.4042, + "step": 19098 + }, + { + "epoch": 7.05, + "learning_rate": 2.2961325286365315e-06, + "loss": 0.3881, + "step": 19099 + }, + { + "epoch": 7.05, + "learning_rate": 2.296095578273186e-06, + "loss": 0.434, + "step": 19100 + }, + { + "epoch": 7.05, + "learning_rate": 2.296058627909841e-06, + "loss": 0.4428, + "step": 19101 + }, + { + "epoch": 7.05, + "learning_rate": 2.296021677546496e-06, + "loss": 0.3621, + "step": 19102 + }, + { + "epoch": 7.05, + "learning_rate": 2.295984727183151e-06, + "loss": 0.4163, + "step": 19103 + }, + { + "epoch": 7.05, + "learning_rate": 2.2959477768198055e-06, + "loss": 0.3281, + "step": 19104 + }, + { + "epoch": 7.06, + "learning_rate": 2.29591082645646e-06, + "loss": 0.3813, + "step": 19105 + }, + { + "epoch": 7.06, + "learning_rate": 2.2958738760931148e-06, + "loss": 0.372, + "step": 19106 + }, + { + "epoch": 7.06, + "learning_rate": 2.29583692572977e-06, + "loss": 0.3997, + "step": 19107 + }, + { + "epoch": 7.06, + "learning_rate": 2.2957999753664245e-06, + "loss": 0.4219, + "step": 19108 + }, + { + "epoch": 7.06, + "learning_rate": 2.295763025003079e-06, + "loss": 0.3903, + "step": 19109 + }, + { + "epoch": 7.06, + "learning_rate": 2.295726074639734e-06, + "loss": 0.3115, + "step": 19110 + }, + { + "epoch": 7.06, + "learning_rate": 2.295689124276389e-06, + "loss": 0.497, + "step": 19111 + }, + { + "epoch": 7.06, + "learning_rate": 2.2956521739130434e-06, + "loss": 0.4562, + "step": 19112 + }, + { + "epoch": 7.06, + "learning_rate": 2.295615223549698e-06, + "loss": 0.4228, + "step": 19113 + }, + { + "epoch": 7.06, + "learning_rate": 2.295578273186353e-06, + "loss": 0.4103, + "step": 19114 + }, + { + "epoch": 7.06, + "learning_rate": 2.2955413228230078e-06, + "loss": 0.3654, + "step": 19115 + }, + { + "epoch": 7.06, + "learning_rate": 2.295504372459663e-06, + "loss": 0.5554, + "step": 19116 + }, + { + "epoch": 7.06, + "learning_rate": 2.2954674220963175e-06, + "loss": 0.4925, + "step": 19117 + }, + { + "epoch": 7.06, + "learning_rate": 2.295430471732972e-06, + "loss": 0.3298, + "step": 19118 + }, + { + "epoch": 7.06, + "learning_rate": 2.2953935213696267e-06, + "loss": 0.4302, + "step": 19119 + }, + { + "epoch": 7.06, + "learning_rate": 2.2953565710062814e-06, + "loss": 0.4628, + "step": 19120 + }, + { + "epoch": 7.06, + "learning_rate": 2.2953196206429364e-06, + "loss": 0.4226, + "step": 19121 + }, + { + "epoch": 7.06, + "learning_rate": 2.295282670279591e-06, + "loss": 0.4934, + "step": 19122 + }, + { + "epoch": 7.06, + "learning_rate": 2.295245719916246e-06, + "loss": 0.3568, + "step": 19123 + }, + { + "epoch": 7.06, + "learning_rate": 2.2952087695529008e-06, + "loss": 0.37, + "step": 19124 + }, + { + "epoch": 7.06, + "learning_rate": 2.2951718191895554e-06, + "loss": 0.336, + "step": 19125 + }, + { + "epoch": 7.06, + "learning_rate": 2.29513486882621e-06, + "loss": 0.4094, + "step": 19126 + }, + { + "epoch": 7.06, + "learning_rate": 2.295097918462865e-06, + "loss": 0.4649, + "step": 19127 + }, + { + "epoch": 7.06, + "learning_rate": 2.2950609680995197e-06, + "loss": 0.4252, + "step": 19128 + }, + { + "epoch": 7.06, + "learning_rate": 2.2950240177361744e-06, + "loss": 0.3531, + "step": 19129 + }, + { + "epoch": 7.06, + "learning_rate": 2.2949870673728294e-06, + "loss": 0.4685, + "step": 19130 + }, + { + "epoch": 7.06, + "learning_rate": 2.294950117009484e-06, + "loss": 0.3975, + "step": 19131 + }, + { + "epoch": 7.06, + "learning_rate": 2.2949131666461387e-06, + "loss": 0.3999, + "step": 19132 + }, + { + "epoch": 7.07, + "learning_rate": 2.2948762162827933e-06, + "loss": 0.2678, + "step": 19133 + }, + { + "epoch": 7.07, + "learning_rate": 2.2948392659194484e-06, + "loss": 0.3143, + "step": 19134 + }, + { + "epoch": 7.07, + "learning_rate": 2.294802315556103e-06, + "loss": 0.4507, + "step": 19135 + }, + { + "epoch": 7.07, + "learning_rate": 2.294765365192758e-06, + "loss": 0.3745, + "step": 19136 + }, + { + "epoch": 7.07, + "learning_rate": 2.2947284148294127e-06, + "loss": 0.332, + "step": 19137 + }, + { + "epoch": 7.07, + "learning_rate": 2.294691464466067e-06, + "loss": 0.3182, + "step": 19138 + }, + { + "epoch": 7.07, + "learning_rate": 2.294654514102722e-06, + "loss": 0.3496, + "step": 19139 + }, + { + "epoch": 7.07, + "learning_rate": 2.2946175637393766e-06, + "loss": 0.5335, + "step": 19140 + }, + { + "epoch": 7.07, + "learning_rate": 2.2945806133760317e-06, + "loss": 0.383, + "step": 19141 + }, + { + "epoch": 7.07, + "learning_rate": 2.2945436630126863e-06, + "loss": 0.3402, + "step": 19142 + }, + { + "epoch": 7.07, + "learning_rate": 2.2945067126493414e-06, + "loss": 0.3616, + "step": 19143 + }, + { + "epoch": 7.07, + "learning_rate": 2.2944697622859956e-06, + "loss": 0.4109, + "step": 19144 + }, + { + "epoch": 7.07, + "learning_rate": 2.2944328119226507e-06, + "loss": 0.4738, + "step": 19145 + }, + { + "epoch": 7.07, + "learning_rate": 2.2943958615593053e-06, + "loss": 0.4946, + "step": 19146 + }, + { + "epoch": 7.07, + "learning_rate": 2.2943589111959604e-06, + "loss": 0.3783, + "step": 19147 + }, + { + "epoch": 7.07, + "learning_rate": 2.294321960832615e-06, + "loss": 0.4695, + "step": 19148 + }, + { + "epoch": 7.07, + "learning_rate": 2.2942850104692696e-06, + "loss": 0.4825, + "step": 19149 + }, + { + "epoch": 7.07, + "learning_rate": 2.2942480601059243e-06, + "loss": 0.6781, + "step": 19150 + }, + { + "epoch": 7.07, + "learning_rate": 2.294211109742579e-06, + "loss": 0.454, + "step": 19151 + }, + { + "epoch": 7.07, + "learning_rate": 2.294174159379234e-06, + "loss": 0.348, + "step": 19152 + }, + { + "epoch": 7.07, + "learning_rate": 2.2941372090158886e-06, + "loss": 0.4348, + "step": 19153 + }, + { + "epoch": 7.07, + "learning_rate": 2.2941002586525437e-06, + "loss": 0.3759, + "step": 19154 + }, + { + "epoch": 7.07, + "learning_rate": 2.2940633082891983e-06, + "loss": 0.4052, + "step": 19155 + }, + { + "epoch": 7.07, + "learning_rate": 2.294026357925853e-06, + "loss": 0.4724, + "step": 19156 + }, + { + "epoch": 7.07, + "learning_rate": 2.2939894075625076e-06, + "loss": 0.4303, + "step": 19157 + }, + { + "epoch": 7.07, + "learning_rate": 2.293952457199162e-06, + "loss": 0.5093, + "step": 19158 + }, + { + "epoch": 7.07, + "learning_rate": 2.2939155068358173e-06, + "loss": 0.3374, + "step": 19159 + }, + { + "epoch": 7.08, + "learning_rate": 2.293878556472472e-06, + "loss": 0.4714, + "step": 19160 + }, + { + "epoch": 7.08, + "learning_rate": 2.293841606109127e-06, + "loss": 0.5623, + "step": 19161 + }, + { + "epoch": 7.08, + "learning_rate": 2.2938046557457816e-06, + "loss": 0.3446, + "step": 19162 + }, + { + "epoch": 7.08, + "learning_rate": 2.2937677053824362e-06, + "loss": 0.4463, + "step": 19163 + }, + { + "epoch": 7.08, + "learning_rate": 2.293730755019091e-06, + "loss": 0.4053, + "step": 19164 + }, + { + "epoch": 7.08, + "learning_rate": 2.293693804655746e-06, + "loss": 0.4369, + "step": 19165 + }, + { + "epoch": 7.08, + "learning_rate": 2.2936568542924006e-06, + "loss": 0.4476, + "step": 19166 + }, + { + "epoch": 7.08, + "learning_rate": 2.2936199039290556e-06, + "loss": 0.5027, + "step": 19167 + }, + { + "epoch": 7.08, + "learning_rate": 2.2935829535657103e-06, + "loss": 0.5296, + "step": 19168 + }, + { + "epoch": 7.08, + "learning_rate": 2.293546003202365e-06, + "loss": 0.3609, + "step": 19169 + }, + { + "epoch": 7.08, + "learning_rate": 2.2935090528390195e-06, + "loss": 0.4463, + "step": 19170 + }, + { + "epoch": 7.08, + "learning_rate": 2.293472102475674e-06, + "loss": 0.3183, + "step": 19171 + }, + { + "epoch": 7.08, + "learning_rate": 2.2934351521123292e-06, + "loss": 0.4665, + "step": 19172 + }, + { + "epoch": 7.08, + "learning_rate": 2.293398201748984e-06, + "loss": 0.3404, + "step": 19173 + }, + { + "epoch": 7.08, + "learning_rate": 2.293361251385639e-06, + "loss": 0.3596, + "step": 19174 + }, + { + "epoch": 7.08, + "learning_rate": 2.2933243010222936e-06, + "loss": 0.4277, + "step": 19175 + }, + { + "epoch": 7.08, + "learning_rate": 2.293287350658948e-06, + "loss": 0.3713, + "step": 19176 + }, + { + "epoch": 7.08, + "learning_rate": 2.293250400295603e-06, + "loss": 0.3728, + "step": 19177 + }, + { + "epoch": 7.08, + "learning_rate": 2.2932134499322575e-06, + "loss": 0.479, + "step": 19178 + }, + { + "epoch": 7.08, + "learning_rate": 2.2931764995689125e-06, + "loss": 0.5818, + "step": 19179 + }, + { + "epoch": 7.08, + "learning_rate": 2.293139549205567e-06, + "loss": 0.3102, + "step": 19180 + }, + { + "epoch": 7.08, + "learning_rate": 2.2931025988422222e-06, + "loss": 0.4777, + "step": 19181 + }, + { + "epoch": 7.08, + "learning_rate": 2.293065648478877e-06, + "loss": 0.48, + "step": 19182 + }, + { + "epoch": 7.08, + "learning_rate": 2.2930286981155315e-06, + "loss": 0.4223, + "step": 19183 + }, + { + "epoch": 7.08, + "learning_rate": 2.292991747752186e-06, + "loss": 0.4357, + "step": 19184 + }, + { + "epoch": 7.08, + "learning_rate": 2.292954797388841e-06, + "loss": 0.4229, + "step": 19185 + }, + { + "epoch": 7.08, + "learning_rate": 2.292917847025496e-06, + "loss": 0.4997, + "step": 19186 + }, + { + "epoch": 7.09, + "learning_rate": 2.292880896662151e-06, + "loss": 0.3207, + "step": 19187 + }, + { + "epoch": 7.09, + "learning_rate": 2.2928439462988055e-06, + "loss": 0.395, + "step": 19188 + }, + { + "epoch": 7.09, + "learning_rate": 2.2928069959354598e-06, + "loss": 0.3851, + "step": 19189 + }, + { + "epoch": 7.09, + "learning_rate": 2.292770045572115e-06, + "loss": 0.3812, + "step": 19190 + }, + { + "epoch": 7.09, + "learning_rate": 2.2927330952087694e-06, + "loss": 0.418, + "step": 19191 + }, + { + "epoch": 7.09, + "learning_rate": 2.2926961448454245e-06, + "loss": 0.3825, + "step": 19192 + }, + { + "epoch": 7.09, + "learning_rate": 2.292659194482079e-06, + "loss": 0.4297, + "step": 19193 + }, + { + "epoch": 7.09, + "learning_rate": 2.292622244118734e-06, + "loss": 0.3833, + "step": 19194 + }, + { + "epoch": 7.09, + "learning_rate": 2.2925852937553884e-06, + "loss": 0.4347, + "step": 19195 + }, + { + "epoch": 7.09, + "learning_rate": 2.2925483433920435e-06, + "loss": 0.4303, + "step": 19196 + }, + { + "epoch": 7.09, + "learning_rate": 2.292511393028698e-06, + "loss": 0.4731, + "step": 19197 + }, + { + "epoch": 7.09, + "learning_rate": 2.2924744426653528e-06, + "loss": 0.3272, + "step": 19198 + }, + { + "epoch": 7.09, + "learning_rate": 2.292437492302008e-06, + "loss": 0.3038, + "step": 19199 + }, + { + "epoch": 7.09, + "learning_rate": 2.2924005419386624e-06, + "loss": 0.4092, + "step": 19200 + }, + { + "epoch": 7.09, + "learning_rate": 2.292363591575317e-06, + "loss": 0.4359, + "step": 19201 + }, + { + "epoch": 7.09, + "learning_rate": 2.2923266412119717e-06, + "loss": 0.5008, + "step": 19202 + }, + { + "epoch": 7.09, + "learning_rate": 2.2922896908486268e-06, + "loss": 0.3831, + "step": 19203 + }, + { + "epoch": 7.09, + "learning_rate": 2.2922527404852814e-06, + "loss": 0.5218, + "step": 19204 + }, + { + "epoch": 7.09, + "learning_rate": 2.2922157901219365e-06, + "loss": 0.3804, + "step": 19205 + }, + { + "epoch": 7.09, + "learning_rate": 2.292178839758591e-06, + "loss": 0.335, + "step": 19206 + }, + { + "epoch": 7.09, + "learning_rate": 2.2921418893952457e-06, + "loss": 0.3717, + "step": 19207 + }, + { + "epoch": 7.09, + "learning_rate": 2.2921049390319004e-06, + "loss": 0.5511, + "step": 19208 + }, + { + "epoch": 7.09, + "learning_rate": 2.292067988668555e-06, + "loss": 0.472, + "step": 19209 + }, + { + "epoch": 7.09, + "learning_rate": 2.29203103830521e-06, + "loss": 0.3601, + "step": 19210 + }, + { + "epoch": 7.09, + "learning_rate": 2.2919940879418647e-06, + "loss": 0.4093, + "step": 19211 + }, + { + "epoch": 7.09, + "learning_rate": 2.2919571375785198e-06, + "loss": 0.5084, + "step": 19212 + }, + { + "epoch": 7.09, + "learning_rate": 2.2919201872151744e-06, + "loss": 0.5399, + "step": 19213 + }, + { + "epoch": 7.1, + "learning_rate": 2.291883236851829e-06, + "loss": 0.4648, + "step": 19214 + }, + { + "epoch": 7.1, + "learning_rate": 2.2918462864884837e-06, + "loss": 0.3153, + "step": 19215 + }, + { + "epoch": 7.1, + "learning_rate": 2.2918093361251387e-06, + "loss": 0.3684, + "step": 19216 + }, + { + "epoch": 7.1, + "learning_rate": 2.2917723857617934e-06, + "loss": 0.4182, + "step": 19217 + }, + { + "epoch": 7.1, + "learning_rate": 2.291735435398448e-06, + "loss": 0.4536, + "step": 19218 + }, + { + "epoch": 7.1, + "learning_rate": 2.291698485035103e-06, + "loss": 0.4834, + "step": 19219 + }, + { + "epoch": 7.1, + "learning_rate": 2.2916615346717577e-06, + "loss": 0.2715, + "step": 19220 + }, + { + "epoch": 7.1, + "learning_rate": 2.2916245843084124e-06, + "loss": 0.5374, + "step": 19221 + }, + { + "epoch": 7.1, + "learning_rate": 2.291587633945067e-06, + "loss": 0.425, + "step": 19222 + }, + { + "epoch": 7.1, + "learning_rate": 2.291550683581722e-06, + "loss": 0.5412, + "step": 19223 + }, + { + "epoch": 7.1, + "learning_rate": 2.2915137332183767e-06, + "loss": 0.4667, + "step": 19224 + }, + { + "epoch": 7.1, + "learning_rate": 2.2914767828550317e-06, + "loss": 0.3941, + "step": 19225 + }, + { + "epoch": 7.1, + "learning_rate": 2.2914398324916864e-06, + "loss": 0.3502, + "step": 19226 + }, + { + "epoch": 7.1, + "learning_rate": 2.291402882128341e-06, + "loss": 0.4352, + "step": 19227 + }, + { + "epoch": 7.1, + "learning_rate": 2.2913659317649957e-06, + "loss": 0.4554, + "step": 19228 + }, + { + "epoch": 7.1, + "learning_rate": 2.2913289814016503e-06, + "loss": 0.4097, + "step": 19229 + }, + { + "epoch": 7.1, + "learning_rate": 2.2912920310383054e-06, + "loss": 0.4959, + "step": 19230 + }, + { + "epoch": 7.1, + "learning_rate": 2.29125508067496e-06, + "loss": 0.4293, + "step": 19231 + }, + { + "epoch": 7.1, + "learning_rate": 2.291218130311615e-06, + "loss": 0.4665, + "step": 19232 + }, + { + "epoch": 7.1, + "learning_rate": 2.2911811799482697e-06, + "loss": 0.2785, + "step": 19233 + }, + { + "epoch": 7.1, + "learning_rate": 2.2911442295849243e-06, + "loss": 0.4238, + "step": 19234 + }, + { + "epoch": 7.1, + "learning_rate": 2.291107279221579e-06, + "loss": 0.393, + "step": 19235 + }, + { + "epoch": 7.1, + "learning_rate": 2.291070328858234e-06, + "loss": 0.4325, + "step": 19236 + }, + { + "epoch": 7.1, + "learning_rate": 2.2910333784948887e-06, + "loss": 0.3326, + "step": 19237 + }, + { + "epoch": 7.1, + "learning_rate": 2.2909964281315433e-06, + "loss": 0.5087, + "step": 19238 + }, + { + "epoch": 7.1, + "learning_rate": 2.2909594777681984e-06, + "loss": 0.3597, + "step": 19239 + }, + { + "epoch": 7.1, + "learning_rate": 2.2909225274048526e-06, + "loss": 0.6277, + "step": 19240 + }, + { + "epoch": 7.11, + "learning_rate": 2.2908855770415076e-06, + "loss": 0.4607, + "step": 19241 + }, + { + "epoch": 7.11, + "learning_rate": 2.2908486266781623e-06, + "loss": 0.4381, + "step": 19242 + }, + { + "epoch": 7.11, + "learning_rate": 2.2908116763148173e-06, + "loss": 0.388, + "step": 19243 + }, + { + "epoch": 7.11, + "learning_rate": 2.290774725951472e-06, + "loss": 0.4421, + "step": 19244 + }, + { + "epoch": 7.11, + "learning_rate": 2.2907377755881266e-06, + "loss": 0.3412, + "step": 19245 + }, + { + "epoch": 7.11, + "learning_rate": 2.2907008252247812e-06, + "loss": 0.4154, + "step": 19246 + }, + { + "epoch": 7.11, + "learning_rate": 2.2906638748614363e-06, + "loss": 0.3957, + "step": 19247 + }, + { + "epoch": 7.11, + "learning_rate": 2.290626924498091e-06, + "loss": 0.5648, + "step": 19248 + }, + { + "epoch": 7.11, + "learning_rate": 2.2905899741347456e-06, + "loss": 0.3191, + "step": 19249 + }, + { + "epoch": 7.11, + "learning_rate": 2.2905530237714006e-06, + "loss": 0.3834, + "step": 19250 + }, + { + "epoch": 7.11, + "learning_rate": 2.2905160734080553e-06, + "loss": 0.3767, + "step": 19251 + }, + { + "epoch": 7.11, + "learning_rate": 2.29047912304471e-06, + "loss": 0.2761, + "step": 19252 + }, + { + "epoch": 7.11, + "learning_rate": 2.2904421726813645e-06, + "loss": 0.3346, + "step": 19253 + }, + { + "epoch": 7.11, + "learning_rate": 2.2904052223180196e-06, + "loss": 0.4094, + "step": 19254 + }, + { + "epoch": 7.11, + "learning_rate": 2.2903682719546742e-06, + "loss": 0.4346, + "step": 19255 + }, + { + "epoch": 7.11, + "learning_rate": 2.2903313215913293e-06, + "loss": 0.3923, + "step": 19256 + }, + { + "epoch": 7.11, + "learning_rate": 2.290294371227984e-06, + "loss": 0.4404, + "step": 19257 + }, + { + "epoch": 7.11, + "learning_rate": 2.2902574208646386e-06, + "loss": 0.4596, + "step": 19258 + }, + { + "epoch": 7.11, + "learning_rate": 2.290220470501293e-06, + "loss": 0.2787, + "step": 19259 + }, + { + "epoch": 7.11, + "learning_rate": 2.290183520137948e-06, + "loss": 0.5542, + "step": 19260 + }, + { + "epoch": 7.11, + "learning_rate": 2.290146569774603e-06, + "loss": 0.4815, + "step": 19261 + }, + { + "epoch": 7.11, + "learning_rate": 2.2901096194112575e-06, + "loss": 0.3881, + "step": 19262 + }, + { + "epoch": 7.11, + "learning_rate": 2.2900726690479126e-06, + "loss": 0.4361, + "step": 19263 + }, + { + "epoch": 7.11, + "learning_rate": 2.2900357186845672e-06, + "loss": 0.4439, + "step": 19264 + }, + { + "epoch": 7.11, + "learning_rate": 2.289998768321222e-06, + "loss": 0.3841, + "step": 19265 + }, + { + "epoch": 7.11, + "learning_rate": 2.2899618179578765e-06, + "loss": 0.324, + "step": 19266 + }, + { + "epoch": 7.11, + "learning_rate": 2.2899248675945316e-06, + "loss": 0.3844, + "step": 19267 + }, + { + "epoch": 7.12, + "learning_rate": 2.289887917231186e-06, + "loss": 0.3901, + "step": 19268 + }, + { + "epoch": 7.12, + "learning_rate": 2.289850966867841e-06, + "loss": 0.5499, + "step": 19269 + }, + { + "epoch": 7.12, + "learning_rate": 2.289814016504496e-06, + "loss": 0.3699, + "step": 19270 + }, + { + "epoch": 7.12, + "learning_rate": 2.2897770661411505e-06, + "loss": 0.4978, + "step": 19271 + }, + { + "epoch": 7.12, + "learning_rate": 2.289740115777805e-06, + "loss": 0.4656, + "step": 19272 + }, + { + "epoch": 7.12, + "learning_rate": 2.28970316541446e-06, + "loss": 0.3405, + "step": 19273 + }, + { + "epoch": 7.12, + "learning_rate": 2.289666215051115e-06, + "loss": 0.4611, + "step": 19274 + }, + { + "epoch": 7.12, + "learning_rate": 2.2896292646877695e-06, + "loss": 0.4277, + "step": 19275 + }, + { + "epoch": 7.12, + "learning_rate": 2.2895923143244246e-06, + "loss": 0.615, + "step": 19276 + }, + { + "epoch": 7.12, + "learning_rate": 2.289555363961079e-06, + "loss": 0.2855, + "step": 19277 + }, + { + "epoch": 7.12, + "learning_rate": 2.2895184135977334e-06, + "loss": 0.5231, + "step": 19278 + }, + { + "epoch": 7.12, + "learning_rate": 2.2894814632343885e-06, + "loss": 0.4005, + "step": 19279 + }, + { + "epoch": 7.12, + "learning_rate": 2.289444512871043e-06, + "loss": 0.4238, + "step": 19280 + }, + { + "epoch": 7.12, + "learning_rate": 2.289407562507698e-06, + "loss": 0.4805, + "step": 19281 + }, + { + "epoch": 7.12, + "learning_rate": 2.289370612144353e-06, + "loss": 0.4409, + "step": 19282 + }, + { + "epoch": 7.12, + "learning_rate": 2.289333661781008e-06, + "loss": 0.2811, + "step": 19283 + }, + { + "epoch": 7.12, + "learning_rate": 2.289296711417662e-06, + "loss": 0.4036, + "step": 19284 + }, + { + "epoch": 7.12, + "learning_rate": 2.289259761054317e-06, + "loss": 0.4668, + "step": 19285 + }, + { + "epoch": 7.12, + "learning_rate": 2.2892228106909718e-06, + "loss": 0.3975, + "step": 19286 + }, + { + "epoch": 7.12, + "learning_rate": 2.289185860327627e-06, + "loss": 0.4686, + "step": 19287 + }, + { + "epoch": 7.12, + "learning_rate": 2.2891489099642815e-06, + "loss": 0.4178, + "step": 19288 + }, + { + "epoch": 7.12, + "learning_rate": 2.289111959600936e-06, + "loss": 0.5322, + "step": 19289 + }, + { + "epoch": 7.12, + "learning_rate": 2.2890750092375907e-06, + "loss": 0.3152, + "step": 19290 + }, + { + "epoch": 7.12, + "learning_rate": 2.2890380588742454e-06, + "loss": 0.3655, + "step": 19291 + }, + { + "epoch": 7.12, + "learning_rate": 2.2890011085109004e-06, + "loss": 0.4342, + "step": 19292 + }, + { + "epoch": 7.12, + "learning_rate": 2.288964158147555e-06, + "loss": 0.4618, + "step": 19293 + }, + { + "epoch": 7.12, + "learning_rate": 2.28892720778421e-06, + "loss": 0.4672, + "step": 19294 + }, + { + "epoch": 7.13, + "learning_rate": 2.2888902574208648e-06, + "loss": 0.3931, + "step": 19295 + }, + { + "epoch": 7.13, + "learning_rate": 2.2888533070575194e-06, + "loss": 0.4366, + "step": 19296 + }, + { + "epoch": 7.13, + "learning_rate": 2.288816356694174e-06, + "loss": 0.4022, + "step": 19297 + }, + { + "epoch": 7.13, + "learning_rate": 2.2887794063308287e-06, + "loss": 0.2784, + "step": 19298 + }, + { + "epoch": 7.13, + "learning_rate": 2.2887424559674837e-06, + "loss": 0.3262, + "step": 19299 + }, + { + "epoch": 7.13, + "learning_rate": 2.2887055056041384e-06, + "loss": 0.2807, + "step": 19300 + }, + { + "epoch": 7.13, + "learning_rate": 2.2886685552407934e-06, + "loss": 0.3246, + "step": 19301 + }, + { + "epoch": 7.13, + "learning_rate": 2.288631604877448e-06, + "loss": 0.2891, + "step": 19302 + }, + { + "epoch": 7.13, + "learning_rate": 2.2885946545141027e-06, + "loss": 0.4741, + "step": 19303 + }, + { + "epoch": 7.13, + "learning_rate": 2.2885577041507573e-06, + "loss": 0.5195, + "step": 19304 + }, + { + "epoch": 7.13, + "learning_rate": 2.2885207537874124e-06, + "loss": 0.3608, + "step": 19305 + }, + { + "epoch": 7.13, + "learning_rate": 2.288483803424067e-06, + "loss": 0.3719, + "step": 19306 + }, + { + "epoch": 7.13, + "learning_rate": 2.2884468530607217e-06, + "loss": 0.4936, + "step": 19307 + }, + { + "epoch": 7.13, + "learning_rate": 2.2884099026973767e-06, + "loss": 0.3732, + "step": 19308 + }, + { + "epoch": 7.13, + "learning_rate": 2.2883729523340314e-06, + "loss": 0.4011, + "step": 19309 + }, + { + "epoch": 7.13, + "learning_rate": 2.288336001970686e-06, + "loss": 0.4001, + "step": 19310 + }, + { + "epoch": 7.13, + "learning_rate": 2.2882990516073406e-06, + "loss": 0.4596, + "step": 19311 + }, + { + "epoch": 7.13, + "learning_rate": 2.2882621012439957e-06, + "loss": 0.534, + "step": 19312 + }, + { + "epoch": 7.13, + "learning_rate": 2.2882251508806503e-06, + "loss": 0.4796, + "step": 19313 + }, + { + "epoch": 7.13, + "learning_rate": 2.2881882005173054e-06, + "loss": 0.3957, + "step": 19314 + }, + { + "epoch": 7.13, + "learning_rate": 2.28815125015396e-06, + "loss": 0.4476, + "step": 19315 + }, + { + "epoch": 7.13, + "learning_rate": 2.2881142997906147e-06, + "loss": 0.449, + "step": 19316 + }, + { + "epoch": 7.13, + "learning_rate": 2.2880773494272693e-06, + "loss": 0.3767, + "step": 19317 + }, + { + "epoch": 7.13, + "learning_rate": 2.288040399063924e-06, + "loss": 0.4524, + "step": 19318 + }, + { + "epoch": 7.13, + "learning_rate": 2.288003448700579e-06, + "loss": 0.3778, + "step": 19319 + }, + { + "epoch": 7.13, + "learning_rate": 2.2879664983372336e-06, + "loss": 0.2905, + "step": 19320 + }, + { + "epoch": 7.13, + "learning_rate": 2.2879295479738887e-06, + "loss": 0.346, + "step": 19321 + }, + { + "epoch": 7.14, + "learning_rate": 2.2878925976105433e-06, + "loss": 0.4386, + "step": 19322 + }, + { + "epoch": 7.14, + "learning_rate": 2.287855647247198e-06, + "loss": 0.3739, + "step": 19323 + }, + { + "epoch": 7.14, + "learning_rate": 2.2878186968838526e-06, + "loss": 0.3999, + "step": 19324 + }, + { + "epoch": 7.14, + "learning_rate": 2.2877817465205077e-06, + "loss": 0.3275, + "step": 19325 + }, + { + "epoch": 7.14, + "learning_rate": 2.2877447961571623e-06, + "loss": 0.4439, + "step": 19326 + }, + { + "epoch": 7.14, + "learning_rate": 2.287707845793817e-06, + "loss": 0.4733, + "step": 19327 + }, + { + "epoch": 7.14, + "learning_rate": 2.287670895430472e-06, + "loss": 0.4297, + "step": 19328 + }, + { + "epoch": 7.14, + "learning_rate": 2.2876339450671262e-06, + "loss": 0.523, + "step": 19329 + }, + { + "epoch": 7.14, + "learning_rate": 2.2875969947037813e-06, + "loss": 0.5887, + "step": 19330 + }, + { + "epoch": 7.14, + "learning_rate": 2.287560044340436e-06, + "loss": 0.5265, + "step": 19331 + }, + { + "epoch": 7.14, + "learning_rate": 2.287523093977091e-06, + "loss": 0.4334, + "step": 19332 + }, + { + "epoch": 7.14, + "learning_rate": 2.2874861436137456e-06, + "loss": 0.3876, + "step": 19333 + }, + { + "epoch": 7.14, + "learning_rate": 2.2874491932504007e-06, + "loss": 0.3499, + "step": 19334 + }, + { + "epoch": 7.14, + "learning_rate": 2.287412242887055e-06, + "loss": 0.2269, + "step": 19335 + }, + { + "epoch": 7.14, + "learning_rate": 2.28737529252371e-06, + "loss": 0.4467, + "step": 19336 + }, + { + "epoch": 7.14, + "learning_rate": 2.2873383421603646e-06, + "loss": 0.3834, + "step": 19337 + }, + { + "epoch": 7.14, + "learning_rate": 2.2873013917970192e-06, + "loss": 0.4484, + "step": 19338 + }, + { + "epoch": 7.14, + "learning_rate": 2.2872644414336743e-06, + "loss": 0.5704, + "step": 19339 + }, + { + "epoch": 7.14, + "learning_rate": 2.287227491070329e-06, + "loss": 0.4128, + "step": 19340 + }, + { + "epoch": 7.14, + "learning_rate": 2.2871905407069835e-06, + "loss": 0.3343, + "step": 19341 + }, + { + "epoch": 7.14, + "learning_rate": 2.287153590343638e-06, + "loss": 0.4087, + "step": 19342 + }, + { + "epoch": 7.14, + "learning_rate": 2.2871166399802932e-06, + "loss": 0.4983, + "step": 19343 + }, + { + "epoch": 7.14, + "learning_rate": 2.287079689616948e-06, + "loss": 0.4801, + "step": 19344 + }, + { + "epoch": 7.14, + "learning_rate": 2.287042739253603e-06, + "loss": 0.309, + "step": 19345 + }, + { + "epoch": 7.14, + "learning_rate": 2.2870057888902576e-06, + "loss": 0.5095, + "step": 19346 + }, + { + "epoch": 7.14, + "learning_rate": 2.2869688385269122e-06, + "loss": 0.4913, + "step": 19347 + }, + { + "epoch": 7.14, + "learning_rate": 2.286931888163567e-06, + "loss": 0.2841, + "step": 19348 + }, + { + "epoch": 7.15, + "learning_rate": 2.2868949378002215e-06, + "loss": 0.4848, + "step": 19349 + }, + { + "epoch": 7.15, + "learning_rate": 2.2868579874368765e-06, + "loss": 0.4434, + "step": 19350 + }, + { + "epoch": 7.15, + "learning_rate": 2.286821037073531e-06, + "loss": 0.5216, + "step": 19351 + }, + { + "epoch": 7.15, + "learning_rate": 2.2867840867101862e-06, + "loss": 0.4461, + "step": 19352 + }, + { + "epoch": 7.15, + "learning_rate": 2.286747136346841e-06, + "loss": 0.3495, + "step": 19353 + }, + { + "epoch": 7.15, + "learning_rate": 2.2867101859834955e-06, + "loss": 0.4155, + "step": 19354 + }, + { + "epoch": 7.15, + "learning_rate": 2.28667323562015e-06, + "loss": 0.3451, + "step": 19355 + }, + { + "epoch": 7.15, + "learning_rate": 2.2866362852568052e-06, + "loss": 0.3604, + "step": 19356 + }, + { + "epoch": 7.15, + "learning_rate": 2.28659933489346e-06, + "loss": 0.4898, + "step": 19357 + }, + { + "epoch": 7.15, + "learning_rate": 2.2865623845301145e-06, + "loss": 0.4518, + "step": 19358 + }, + { + "epoch": 7.15, + "learning_rate": 2.2865254341667695e-06, + "loss": 0.3886, + "step": 19359 + }, + { + "epoch": 7.15, + "learning_rate": 2.286488483803424e-06, + "loss": 0.4186, + "step": 19360 + }, + { + "epoch": 7.15, + "learning_rate": 2.286451533440079e-06, + "loss": 0.3158, + "step": 19361 + }, + { + "epoch": 7.15, + "learning_rate": 2.2864145830767335e-06, + "loss": 0.631, + "step": 19362 + }, + { + "epoch": 7.15, + "learning_rate": 2.2863776327133885e-06, + "loss": 0.3818, + "step": 19363 + }, + { + "epoch": 7.15, + "learning_rate": 2.286340682350043e-06, + "loss": 0.4212, + "step": 19364 + }, + { + "epoch": 7.15, + "learning_rate": 2.286303731986698e-06, + "loss": 0.342, + "step": 19365 + }, + { + "epoch": 7.15, + "learning_rate": 2.286266781623353e-06, + "loss": 0.4274, + "step": 19366 + }, + { + "epoch": 7.15, + "learning_rate": 2.2862298312600075e-06, + "loss": 0.4277, + "step": 19367 + }, + { + "epoch": 7.15, + "learning_rate": 2.286192880896662e-06, + "loss": 0.4352, + "step": 19368 + }, + { + "epoch": 7.15, + "learning_rate": 2.2861559305333168e-06, + "loss": 0.4043, + "step": 19369 + }, + { + "epoch": 7.15, + "learning_rate": 2.286118980169972e-06, + "loss": 0.3638, + "step": 19370 + }, + { + "epoch": 7.15, + "learning_rate": 2.2860820298066265e-06, + "loss": 0.4136, + "step": 19371 + }, + { + "epoch": 7.15, + "learning_rate": 2.2860450794432815e-06, + "loss": 0.2821, + "step": 19372 + }, + { + "epoch": 7.15, + "learning_rate": 2.286008129079936e-06, + "loss": 0.3644, + "step": 19373 + }, + { + "epoch": 7.15, + "learning_rate": 2.2859711787165908e-06, + "loss": 0.361, + "step": 19374 + }, + { + "epoch": 7.15, + "learning_rate": 2.2859342283532454e-06, + "loss": 0.4769, + "step": 19375 + }, + { + "epoch": 7.16, + "learning_rate": 2.2858972779899005e-06, + "loss": 0.4211, + "step": 19376 + }, + { + "epoch": 7.16, + "learning_rate": 2.285860327626555e-06, + "loss": 0.4282, + "step": 19377 + }, + { + "epoch": 7.16, + "learning_rate": 2.2858233772632098e-06, + "loss": 0.3535, + "step": 19378 + }, + { + "epoch": 7.16, + "learning_rate": 2.285786426899865e-06, + "loss": 0.4391, + "step": 19379 + }, + { + "epoch": 7.16, + "learning_rate": 2.285749476536519e-06, + "loss": 0.496, + "step": 19380 + }, + { + "epoch": 7.16, + "learning_rate": 2.285712526173174e-06, + "loss": 0.3683, + "step": 19381 + }, + { + "epoch": 7.16, + "learning_rate": 2.2856755758098287e-06, + "loss": 0.5055, + "step": 19382 + }, + { + "epoch": 7.16, + "learning_rate": 2.2856386254464838e-06, + "loss": 0.5306, + "step": 19383 + }, + { + "epoch": 7.16, + "learning_rate": 2.2856016750831384e-06, + "loss": 0.4819, + "step": 19384 + }, + { + "epoch": 7.16, + "learning_rate": 2.2855647247197935e-06, + "loss": 0.3901, + "step": 19385 + }, + { + "epoch": 7.16, + "learning_rate": 2.2855277743564477e-06, + "loss": 0.3356, + "step": 19386 + }, + { + "epoch": 7.16, + "learning_rate": 2.2854908239931023e-06, + "loss": 0.3518, + "step": 19387 + }, + { + "epoch": 7.16, + "learning_rate": 2.2854538736297574e-06, + "loss": 0.3777, + "step": 19388 + }, + { + "epoch": 7.16, + "learning_rate": 2.285416923266412e-06, + "loss": 0.4713, + "step": 19389 + }, + { + "epoch": 7.16, + "learning_rate": 2.285379972903067e-06, + "loss": 0.4563, + "step": 19390 + }, + { + "epoch": 7.16, + "learning_rate": 2.2853430225397217e-06, + "loss": 0.3936, + "step": 19391 + }, + { + "epoch": 7.16, + "learning_rate": 2.2853060721763764e-06, + "loss": 0.4876, + "step": 19392 + }, + { + "epoch": 7.16, + "learning_rate": 2.285269121813031e-06, + "loss": 0.5072, + "step": 19393 + }, + { + "epoch": 7.16, + "learning_rate": 2.285232171449686e-06, + "loss": 0.3926, + "step": 19394 + }, + { + "epoch": 7.16, + "learning_rate": 2.2851952210863407e-06, + "loss": 0.3457, + "step": 19395 + }, + { + "epoch": 7.16, + "learning_rate": 2.2851582707229958e-06, + "loss": 0.2819, + "step": 19396 + }, + { + "epoch": 7.16, + "learning_rate": 2.2851213203596504e-06, + "loss": 0.422, + "step": 19397 + }, + { + "epoch": 7.16, + "learning_rate": 2.285084369996305e-06, + "loss": 0.3978, + "step": 19398 + }, + { + "epoch": 7.16, + "learning_rate": 2.2850474196329597e-06, + "loss": 0.4546, + "step": 19399 + }, + { + "epoch": 7.16, + "learning_rate": 2.2850104692696143e-06, + "loss": 0.3736, + "step": 19400 + }, + { + "epoch": 7.16, + "learning_rate": 2.2849735189062694e-06, + "loss": 0.5472, + "step": 19401 + }, + { + "epoch": 7.16, + "learning_rate": 2.284936568542924e-06, + "loss": 0.4398, + "step": 19402 + }, + { + "epoch": 7.17, + "learning_rate": 2.284899618179579e-06, + "loss": 0.5831, + "step": 19403 + }, + { + "epoch": 7.17, + "learning_rate": 2.2848626678162337e-06, + "loss": 0.4049, + "step": 19404 + }, + { + "epoch": 7.17, + "learning_rate": 2.2848257174528883e-06, + "loss": 0.405, + "step": 19405 + }, + { + "epoch": 7.17, + "learning_rate": 2.284788767089543e-06, + "loss": 0.4036, + "step": 19406 + }, + { + "epoch": 7.17, + "learning_rate": 2.2847518167261976e-06, + "loss": 0.4539, + "step": 19407 + }, + { + "epoch": 7.17, + "learning_rate": 2.2847148663628527e-06, + "loss": 0.3967, + "step": 19408 + }, + { + "epoch": 7.17, + "learning_rate": 2.2846779159995073e-06, + "loss": 0.3684, + "step": 19409 + }, + { + "epoch": 7.17, + "learning_rate": 2.2846409656361624e-06, + "loss": 0.4281, + "step": 19410 + }, + { + "epoch": 7.17, + "learning_rate": 2.284604015272817e-06, + "loss": 0.4391, + "step": 19411 + }, + { + "epoch": 7.17, + "learning_rate": 2.2845670649094716e-06, + "loss": 0.397, + "step": 19412 + }, + { + "epoch": 7.17, + "learning_rate": 2.2845301145461263e-06, + "loss": 0.3795, + "step": 19413 + }, + { + "epoch": 7.17, + "learning_rate": 2.2844931641827813e-06, + "loss": 0.3759, + "step": 19414 + }, + { + "epoch": 7.17, + "learning_rate": 2.284456213819436e-06, + "loss": 0.4075, + "step": 19415 + }, + { + "epoch": 7.17, + "learning_rate": 2.284419263456091e-06, + "loss": 0.4313, + "step": 19416 + }, + { + "epoch": 7.17, + "learning_rate": 2.2843823130927457e-06, + "loss": 0.4426, + "step": 19417 + }, + { + "epoch": 7.17, + "learning_rate": 2.2843453627294003e-06, + "loss": 0.4408, + "step": 19418 + }, + { + "epoch": 7.17, + "learning_rate": 2.284308412366055e-06, + "loss": 0.3725, + "step": 19419 + }, + { + "epoch": 7.17, + "learning_rate": 2.2842714620027096e-06, + "loss": 0.3289, + "step": 19420 + }, + { + "epoch": 7.17, + "learning_rate": 2.2842345116393646e-06, + "loss": 0.3455, + "step": 19421 + }, + { + "epoch": 7.17, + "learning_rate": 2.2841975612760193e-06, + "loss": 0.5642, + "step": 19422 + }, + { + "epoch": 7.17, + "learning_rate": 2.2841606109126743e-06, + "loss": 0.4468, + "step": 19423 + }, + { + "epoch": 7.17, + "learning_rate": 2.284123660549329e-06, + "loss": 0.3259, + "step": 19424 + }, + { + "epoch": 7.17, + "learning_rate": 2.2840867101859836e-06, + "loss": 0.4512, + "step": 19425 + }, + { + "epoch": 7.17, + "learning_rate": 2.2840497598226382e-06, + "loss": 0.4835, + "step": 19426 + }, + { + "epoch": 7.17, + "learning_rate": 2.284012809459293e-06, + "loss": 0.4446, + "step": 19427 + }, + { + "epoch": 7.17, + "learning_rate": 2.283975859095948e-06, + "loss": 0.3772, + "step": 19428 + }, + { + "epoch": 7.17, + "learning_rate": 2.2839389087326026e-06, + "loss": 0.4982, + "step": 19429 + }, + { + "epoch": 7.18, + "learning_rate": 2.2839019583692576e-06, + "loss": 0.3707, + "step": 19430 + }, + { + "epoch": 7.18, + "learning_rate": 2.283865008005912e-06, + "loss": 0.2811, + "step": 19431 + }, + { + "epoch": 7.18, + "learning_rate": 2.283828057642567e-06, + "loss": 0.5135, + "step": 19432 + }, + { + "epoch": 7.18, + "learning_rate": 2.2837911072792215e-06, + "loss": 0.5192, + "step": 19433 + }, + { + "epoch": 7.18, + "learning_rate": 2.2837541569158766e-06, + "loss": 0.5025, + "step": 19434 + }, + { + "epoch": 7.18, + "learning_rate": 2.2837172065525312e-06, + "loss": 0.389, + "step": 19435 + }, + { + "epoch": 7.18, + "learning_rate": 2.283680256189186e-06, + "loss": 0.3184, + "step": 19436 + }, + { + "epoch": 7.18, + "learning_rate": 2.2836433058258405e-06, + "loss": 0.4816, + "step": 19437 + }, + { + "epoch": 7.18, + "learning_rate": 2.283606355462495e-06, + "loss": 0.4573, + "step": 19438 + }, + { + "epoch": 7.18, + "learning_rate": 2.28356940509915e-06, + "loss": 0.3527, + "step": 19439 + }, + { + "epoch": 7.18, + "learning_rate": 2.283532454735805e-06, + "loss": 0.3898, + "step": 19440 + }, + { + "epoch": 7.18, + "learning_rate": 2.28349550437246e-06, + "loss": 0.5384, + "step": 19441 + }, + { + "epoch": 7.18, + "learning_rate": 2.2834585540091145e-06, + "loss": 0.354, + "step": 19442 + }, + { + "epoch": 7.18, + "learning_rate": 2.283421603645769e-06, + "loss": 0.3956, + "step": 19443 + }, + { + "epoch": 7.18, + "learning_rate": 2.283384653282424e-06, + "loss": 0.3454, + "step": 19444 + }, + { + "epoch": 7.18, + "learning_rate": 2.283347702919079e-06, + "loss": 0.2923, + "step": 19445 + }, + { + "epoch": 7.18, + "learning_rate": 2.2833107525557335e-06, + "loss": 0.4433, + "step": 19446 + }, + { + "epoch": 7.18, + "learning_rate": 2.283273802192388e-06, + "loss": 0.4954, + "step": 19447 + }, + { + "epoch": 7.18, + "learning_rate": 2.283236851829043e-06, + "loss": 0.3961, + "step": 19448 + }, + { + "epoch": 7.18, + "learning_rate": 2.283199901465698e-06, + "loss": 0.3918, + "step": 19449 + }, + { + "epoch": 7.18, + "learning_rate": 2.2831629511023525e-06, + "loss": 0.2314, + "step": 19450 + }, + { + "epoch": 7.18, + "learning_rate": 2.283126000739007e-06, + "loss": 0.3959, + "step": 19451 + }, + { + "epoch": 7.18, + "learning_rate": 2.283089050375662e-06, + "loss": 0.3148, + "step": 19452 + }, + { + "epoch": 7.18, + "learning_rate": 2.283052100012317e-06, + "loss": 0.4847, + "step": 19453 + }, + { + "epoch": 7.18, + "learning_rate": 2.283015149648972e-06, + "loss": 0.4913, + "step": 19454 + }, + { + "epoch": 7.18, + "learning_rate": 2.2829781992856265e-06, + "loss": 0.4412, + "step": 19455 + }, + { + "epoch": 7.18, + "learning_rate": 2.282941248922281e-06, + "loss": 0.3933, + "step": 19456 + }, + { + "epoch": 7.19, + "learning_rate": 2.2829042985589358e-06, + "loss": 0.4022, + "step": 19457 + }, + { + "epoch": 7.19, + "learning_rate": 2.2828673481955904e-06, + "loss": 0.3202, + "step": 19458 + }, + { + "epoch": 7.19, + "learning_rate": 2.2828303978322455e-06, + "loss": 0.3581, + "step": 19459 + }, + { + "epoch": 7.19, + "learning_rate": 2.2827934474689e-06, + "loss": 0.3301, + "step": 19460 + }, + { + "epoch": 7.19, + "learning_rate": 2.282756497105555e-06, + "loss": 0.442, + "step": 19461 + }, + { + "epoch": 7.19, + "learning_rate": 2.28271954674221e-06, + "loss": 0.2776, + "step": 19462 + }, + { + "epoch": 7.19, + "learning_rate": 2.2826825963788644e-06, + "loss": 0.4128, + "step": 19463 + }, + { + "epoch": 7.19, + "learning_rate": 2.282645646015519e-06, + "loss": 0.4693, + "step": 19464 + }, + { + "epoch": 7.19, + "learning_rate": 2.282608695652174e-06, + "loss": 0.5367, + "step": 19465 + }, + { + "epoch": 7.19, + "learning_rate": 2.2825717452888288e-06, + "loss": 0.3126, + "step": 19466 + }, + { + "epoch": 7.19, + "learning_rate": 2.2825347949254834e-06, + "loss": 0.481, + "step": 19467 + }, + { + "epoch": 7.19, + "learning_rate": 2.2824978445621385e-06, + "loss": 0.4389, + "step": 19468 + }, + { + "epoch": 7.19, + "learning_rate": 2.2824608941987927e-06, + "loss": 0.5008, + "step": 19469 + }, + { + "epoch": 7.19, + "learning_rate": 2.2824239438354477e-06, + "loss": 0.5915, + "step": 19470 + }, + { + "epoch": 7.19, + "learning_rate": 2.2823869934721024e-06, + "loss": 0.4023, + "step": 19471 + }, + { + "epoch": 7.19, + "learning_rate": 2.2823500431087574e-06, + "loss": 0.3253, + "step": 19472 + }, + { + "epoch": 7.19, + "learning_rate": 2.282313092745412e-06, + "loss": 0.5328, + "step": 19473 + }, + { + "epoch": 7.19, + "learning_rate": 2.282276142382067e-06, + "loss": 0.4462, + "step": 19474 + }, + { + "epoch": 7.19, + "learning_rate": 2.2822391920187213e-06, + "loss": 0.3475, + "step": 19475 + }, + { + "epoch": 7.19, + "learning_rate": 2.2822022416553764e-06, + "loss": 0.402, + "step": 19476 + }, + { + "epoch": 7.19, + "learning_rate": 2.282165291292031e-06, + "loss": 0.4356, + "step": 19477 + }, + { + "epoch": 7.19, + "learning_rate": 2.2821283409286857e-06, + "loss": 0.3329, + "step": 19478 + }, + { + "epoch": 7.19, + "learning_rate": 2.2820913905653407e-06, + "loss": 0.3958, + "step": 19479 + }, + { + "epoch": 7.19, + "learning_rate": 2.2820544402019954e-06, + "loss": 0.3558, + "step": 19480 + }, + { + "epoch": 7.19, + "learning_rate": 2.28201748983865e-06, + "loss": 0.4261, + "step": 19481 + }, + { + "epoch": 7.19, + "learning_rate": 2.2819805394753047e-06, + "loss": 0.3326, + "step": 19482 + }, + { + "epoch": 7.19, + "learning_rate": 2.2819435891119597e-06, + "loss": 0.3665, + "step": 19483 + }, + { + "epoch": 7.19, + "learning_rate": 2.2819066387486143e-06, + "loss": 0.3729, + "step": 19484 + }, + { + "epoch": 7.2, + "learning_rate": 2.2818696883852694e-06, + "loss": 0.4296, + "step": 19485 + }, + { + "epoch": 7.2, + "learning_rate": 2.281832738021924e-06, + "loss": 0.4714, + "step": 19486 + }, + { + "epoch": 7.2, + "learning_rate": 2.2817957876585787e-06, + "loss": 0.3866, + "step": 19487 + }, + { + "epoch": 7.2, + "learning_rate": 2.2817588372952333e-06, + "loss": 0.5623, + "step": 19488 + }, + { + "epoch": 7.2, + "learning_rate": 2.281721886931888e-06, + "loss": 0.3293, + "step": 19489 + }, + { + "epoch": 7.2, + "learning_rate": 2.281684936568543e-06, + "loss": 0.3821, + "step": 19490 + }, + { + "epoch": 7.2, + "learning_rate": 2.2816479862051976e-06, + "loss": 0.3967, + "step": 19491 + }, + { + "epoch": 7.2, + "learning_rate": 2.2816110358418527e-06, + "loss": 0.4371, + "step": 19492 + }, + { + "epoch": 7.2, + "learning_rate": 2.2815740854785073e-06, + "loss": 0.3766, + "step": 19493 + }, + { + "epoch": 7.2, + "learning_rate": 2.281537135115162e-06, + "loss": 0.4809, + "step": 19494 + }, + { + "epoch": 7.2, + "learning_rate": 2.2815001847518166e-06, + "loss": 0.4314, + "step": 19495 + }, + { + "epoch": 7.2, + "learning_rate": 2.2814632343884717e-06, + "loss": 0.2791, + "step": 19496 + }, + { + "epoch": 7.2, + "learning_rate": 2.2814262840251263e-06, + "loss": 0.4279, + "step": 19497 + }, + { + "epoch": 7.2, + "learning_rate": 2.281389333661781e-06, + "loss": 0.4272, + "step": 19498 + }, + { + "epoch": 7.2, + "learning_rate": 2.281352383298436e-06, + "loss": 0.3963, + "step": 19499 + }, + { + "epoch": 7.2, + "learning_rate": 2.2813154329350906e-06, + "loss": 0.3354, + "step": 19500 + }, + { + "epoch": 7.2, + "learning_rate": 2.2812784825717453e-06, + "loss": 0.4167, + "step": 19501 + }, + { + "epoch": 7.2, + "learning_rate": 2.2812415322084e-06, + "loss": 0.3553, + "step": 19502 + }, + { + "epoch": 7.2, + "learning_rate": 2.281204581845055e-06, + "loss": 0.4168, + "step": 19503 + }, + { + "epoch": 7.2, + "learning_rate": 2.2811676314817096e-06, + "loss": 0.3351, + "step": 19504 + }, + { + "epoch": 7.2, + "learning_rate": 2.2811306811183647e-06, + "loss": 0.452, + "step": 19505 + }, + { + "epoch": 7.2, + "learning_rate": 2.2810937307550193e-06, + "loss": 0.4198, + "step": 19506 + }, + { + "epoch": 7.2, + "learning_rate": 2.281056780391674e-06, + "loss": 0.3781, + "step": 19507 + }, + { + "epoch": 7.2, + "learning_rate": 2.2810198300283286e-06, + "loss": 0.3905, + "step": 19508 + }, + { + "epoch": 7.2, + "learning_rate": 2.2809828796649832e-06, + "loss": 0.3053, + "step": 19509 + }, + { + "epoch": 7.2, + "learning_rate": 2.2809459293016383e-06, + "loss": 0.3579, + "step": 19510 + }, + { + "epoch": 7.2, + "learning_rate": 2.280908978938293e-06, + "loss": 0.3936, + "step": 19511 + }, + { + "epoch": 7.21, + "learning_rate": 2.280872028574948e-06, + "loss": 0.4308, + "step": 19512 + }, + { + "epoch": 7.21, + "learning_rate": 2.2808350782116026e-06, + "loss": 0.2778, + "step": 19513 + }, + { + "epoch": 7.21, + "learning_rate": 2.2807981278482573e-06, + "loss": 0.5636, + "step": 19514 + }, + { + "epoch": 7.21, + "learning_rate": 2.280761177484912e-06, + "loss": 0.3942, + "step": 19515 + }, + { + "epoch": 7.21, + "learning_rate": 2.280724227121567e-06, + "loss": 0.2646, + "step": 19516 + }, + { + "epoch": 7.21, + "learning_rate": 2.2806872767582216e-06, + "loss": 0.4436, + "step": 19517 + }, + { + "epoch": 7.21, + "learning_rate": 2.2806503263948762e-06, + "loss": 0.4405, + "step": 19518 + }, + { + "epoch": 7.21, + "learning_rate": 2.2806133760315313e-06, + "loss": 0.4009, + "step": 19519 + }, + { + "epoch": 7.21, + "learning_rate": 2.2805764256681855e-06, + "loss": 0.5361, + "step": 19520 + }, + { + "epoch": 7.21, + "learning_rate": 2.2805394753048406e-06, + "loss": 0.5518, + "step": 19521 + }, + { + "epoch": 7.21, + "learning_rate": 2.280502524941495e-06, + "loss": 0.4295, + "step": 19522 + }, + { + "epoch": 7.21, + "learning_rate": 2.2804655745781502e-06, + "loss": 0.4998, + "step": 19523 + }, + { + "epoch": 7.21, + "learning_rate": 2.280428624214805e-06, + "loss": 0.4019, + "step": 19524 + }, + { + "epoch": 7.21, + "learning_rate": 2.28039167385146e-06, + "loss": 0.3965, + "step": 19525 + }, + { + "epoch": 7.21, + "learning_rate": 2.280354723488114e-06, + "loss": 0.3502, + "step": 19526 + }, + { + "epoch": 7.21, + "learning_rate": 2.280317773124769e-06, + "loss": 0.374, + "step": 19527 + }, + { + "epoch": 7.21, + "learning_rate": 2.280280822761424e-06, + "loss": 0.5246, + "step": 19528 + }, + { + "epoch": 7.21, + "learning_rate": 2.2802438723980785e-06, + "loss": 0.3631, + "step": 19529 + }, + { + "epoch": 7.21, + "learning_rate": 2.2802069220347336e-06, + "loss": 0.2879, + "step": 19530 + }, + { + "epoch": 7.21, + "learning_rate": 2.280169971671388e-06, + "loss": 0.551, + "step": 19531 + }, + { + "epoch": 7.21, + "learning_rate": 2.280133021308043e-06, + "loss": 0.4017, + "step": 19532 + }, + { + "epoch": 7.21, + "learning_rate": 2.2800960709446975e-06, + "loss": 0.3765, + "step": 19533 + }, + { + "epoch": 7.21, + "learning_rate": 2.2800591205813525e-06, + "loss": 0.3364, + "step": 19534 + }, + { + "epoch": 7.21, + "learning_rate": 2.280022170218007e-06, + "loss": 0.3842, + "step": 19535 + }, + { + "epoch": 7.21, + "learning_rate": 2.279985219854662e-06, + "loss": 0.501, + "step": 19536 + }, + { + "epoch": 7.21, + "learning_rate": 2.279948269491317e-06, + "loss": 0.4425, + "step": 19537 + }, + { + "epoch": 7.21, + "learning_rate": 2.2799113191279715e-06, + "loss": 0.4296, + "step": 19538 + }, + { + "epoch": 7.22, + "learning_rate": 2.279874368764626e-06, + "loss": 0.3383, + "step": 19539 + }, + { + "epoch": 7.22, + "learning_rate": 2.2798374184012808e-06, + "loss": 0.25, + "step": 19540 + }, + { + "epoch": 7.22, + "learning_rate": 2.279800468037936e-06, + "loss": 0.3887, + "step": 19541 + }, + { + "epoch": 7.22, + "learning_rate": 2.2797635176745905e-06, + "loss": 0.3649, + "step": 19542 + }, + { + "epoch": 7.22, + "learning_rate": 2.2797265673112455e-06, + "loss": 0.4423, + "step": 19543 + }, + { + "epoch": 7.22, + "learning_rate": 2.2796896169479e-06, + "loss": 0.3182, + "step": 19544 + }, + { + "epoch": 7.22, + "learning_rate": 2.279652666584555e-06, + "loss": 0.3285, + "step": 19545 + }, + { + "epoch": 7.22, + "learning_rate": 2.2796157162212094e-06, + "loss": 0.4744, + "step": 19546 + }, + { + "epoch": 7.22, + "learning_rate": 2.279578765857864e-06, + "loss": 0.4643, + "step": 19547 + }, + { + "epoch": 7.22, + "learning_rate": 2.279541815494519e-06, + "loss": 0.3206, + "step": 19548 + }, + { + "epoch": 7.22, + "learning_rate": 2.2795048651311738e-06, + "loss": 0.3727, + "step": 19549 + }, + { + "epoch": 7.22, + "learning_rate": 2.279467914767829e-06, + "loss": 0.4609, + "step": 19550 + }, + { + "epoch": 7.22, + "learning_rate": 2.2794309644044835e-06, + "loss": 0.4345, + "step": 19551 + }, + { + "epoch": 7.22, + "learning_rate": 2.279394014041138e-06, + "loss": 0.4199, + "step": 19552 + }, + { + "epoch": 7.22, + "learning_rate": 2.2793570636777927e-06, + "loss": 0.4654, + "step": 19553 + }, + { + "epoch": 7.22, + "learning_rate": 2.279320113314448e-06, + "loss": 0.4078, + "step": 19554 + }, + { + "epoch": 7.22, + "learning_rate": 2.2792831629511024e-06, + "loss": 0.3957, + "step": 19555 + }, + { + "epoch": 7.22, + "learning_rate": 2.279246212587757e-06, + "loss": 0.4496, + "step": 19556 + }, + { + "epoch": 7.22, + "learning_rate": 2.279209262224412e-06, + "loss": 0.357, + "step": 19557 + }, + { + "epoch": 7.22, + "learning_rate": 2.2791723118610668e-06, + "loss": 0.3547, + "step": 19558 + }, + { + "epoch": 7.22, + "learning_rate": 2.2791353614977214e-06, + "loss": 0.328, + "step": 19559 + }, + { + "epoch": 7.22, + "learning_rate": 2.279098411134376e-06, + "loss": 0.5031, + "step": 19560 + }, + { + "epoch": 7.22, + "learning_rate": 2.279061460771031e-06, + "loss": 0.4684, + "step": 19561 + }, + { + "epoch": 7.22, + "learning_rate": 2.2790245104076857e-06, + "loss": 0.5152, + "step": 19562 + }, + { + "epoch": 7.22, + "learning_rate": 2.278987560044341e-06, + "loss": 0.5385, + "step": 19563 + }, + { + "epoch": 7.22, + "learning_rate": 2.2789506096809954e-06, + "loss": 0.462, + "step": 19564 + }, + { + "epoch": 7.22, + "learning_rate": 2.27891365931765e-06, + "loss": 0.3817, + "step": 19565 + }, + { + "epoch": 7.23, + "learning_rate": 2.2788767089543047e-06, + "loss": 0.4345, + "step": 19566 + }, + { + "epoch": 7.23, + "learning_rate": 2.2788397585909593e-06, + "loss": 0.4863, + "step": 19567 + }, + { + "epoch": 7.23, + "learning_rate": 2.2788028082276144e-06, + "loss": 0.4051, + "step": 19568 + }, + { + "epoch": 7.23, + "learning_rate": 2.278765857864269e-06, + "loss": 0.4103, + "step": 19569 + }, + { + "epoch": 7.23, + "learning_rate": 2.278728907500924e-06, + "loss": 0.4683, + "step": 19570 + }, + { + "epoch": 7.23, + "learning_rate": 2.2786919571375783e-06, + "loss": 0.3868, + "step": 19571 + }, + { + "epoch": 7.23, + "learning_rate": 2.2786550067742334e-06, + "loss": 0.3676, + "step": 19572 + }, + { + "epoch": 7.23, + "learning_rate": 2.278618056410888e-06, + "loss": 0.4037, + "step": 19573 + }, + { + "epoch": 7.23, + "learning_rate": 2.278581106047543e-06, + "loss": 0.5049, + "step": 19574 + }, + { + "epoch": 7.23, + "learning_rate": 2.2785441556841977e-06, + "loss": 0.5374, + "step": 19575 + }, + { + "epoch": 7.23, + "learning_rate": 2.2785072053208523e-06, + "loss": 0.42, + "step": 19576 + }, + { + "epoch": 7.23, + "learning_rate": 2.278470254957507e-06, + "loss": 0.3206, + "step": 19577 + }, + { + "epoch": 7.23, + "learning_rate": 2.2784333045941616e-06, + "loss": 0.3905, + "step": 19578 + }, + { + "epoch": 7.23, + "learning_rate": 2.2783963542308167e-06, + "loss": 0.3909, + "step": 19579 + }, + { + "epoch": 7.23, + "learning_rate": 2.2783594038674713e-06, + "loss": 0.4404, + "step": 19580 + }, + { + "epoch": 7.23, + "learning_rate": 2.2783224535041264e-06, + "loss": 0.4379, + "step": 19581 + }, + { + "epoch": 7.23, + "learning_rate": 2.278285503140781e-06, + "loss": 0.4361, + "step": 19582 + }, + { + "epoch": 7.23, + "learning_rate": 2.2782485527774356e-06, + "loss": 0.5026, + "step": 19583 + }, + { + "epoch": 7.23, + "learning_rate": 2.2782116024140903e-06, + "loss": 0.5197, + "step": 19584 + }, + { + "epoch": 7.23, + "learning_rate": 2.2781746520507453e-06, + "loss": 0.4805, + "step": 19585 + }, + { + "epoch": 7.23, + "learning_rate": 2.2781377016874e-06, + "loss": 0.4423, + "step": 19586 + }, + { + "epoch": 7.23, + "learning_rate": 2.2781007513240546e-06, + "loss": 0.3753, + "step": 19587 + }, + { + "epoch": 7.23, + "learning_rate": 2.2780638009607097e-06, + "loss": 0.498, + "step": 19588 + }, + { + "epoch": 7.23, + "learning_rate": 2.2780268505973643e-06, + "loss": 0.5068, + "step": 19589 + }, + { + "epoch": 7.23, + "learning_rate": 2.277989900234019e-06, + "loss": 0.3531, + "step": 19590 + }, + { + "epoch": 7.23, + "learning_rate": 2.2779529498706736e-06, + "loss": 0.3986, + "step": 19591 + }, + { + "epoch": 7.23, + "learning_rate": 2.2779159995073286e-06, + "loss": 0.3695, + "step": 19592 + }, + { + "epoch": 7.24, + "learning_rate": 2.2778790491439833e-06, + "loss": 0.3936, + "step": 19593 + }, + { + "epoch": 7.24, + "learning_rate": 2.2778420987806383e-06, + "loss": 0.2502, + "step": 19594 + }, + { + "epoch": 7.24, + "learning_rate": 2.277805148417293e-06, + "loss": 0.4083, + "step": 19595 + }, + { + "epoch": 7.24, + "learning_rate": 2.2777681980539476e-06, + "loss": 0.3818, + "step": 19596 + }, + { + "epoch": 7.24, + "learning_rate": 2.2777312476906022e-06, + "loss": 0.4885, + "step": 19597 + }, + { + "epoch": 7.24, + "learning_rate": 2.277694297327257e-06, + "loss": 0.4467, + "step": 19598 + }, + { + "epoch": 7.24, + "learning_rate": 2.277657346963912e-06, + "loss": 0.4323, + "step": 19599 + }, + { + "epoch": 7.24, + "learning_rate": 2.2776203966005666e-06, + "loss": 0.568, + "step": 19600 + }, + { + "epoch": 7.24, + "learning_rate": 2.2775834462372216e-06, + "loss": 0.4496, + "step": 19601 + }, + { + "epoch": 7.24, + "learning_rate": 2.2775464958738763e-06, + "loss": 0.3712, + "step": 19602 + }, + { + "epoch": 7.24, + "learning_rate": 2.277509545510531e-06, + "loss": 0.4825, + "step": 19603 + }, + { + "epoch": 7.24, + "learning_rate": 2.2774725951471855e-06, + "loss": 0.4724, + "step": 19604 + }, + { + "epoch": 7.24, + "learning_rate": 2.2774356447838406e-06, + "loss": 0.5057, + "step": 19605 + }, + { + "epoch": 7.24, + "learning_rate": 2.2773986944204952e-06, + "loss": 0.5242, + "step": 19606 + }, + { + "epoch": 7.24, + "learning_rate": 2.27736174405715e-06, + "loss": 0.3818, + "step": 19607 + }, + { + "epoch": 7.24, + "learning_rate": 2.277324793693805e-06, + "loss": 0.4362, + "step": 19608 + }, + { + "epoch": 7.24, + "learning_rate": 2.2772878433304596e-06, + "loss": 0.3451, + "step": 19609 + }, + { + "epoch": 7.24, + "learning_rate": 2.277250892967114e-06, + "loss": 0.4551, + "step": 19610 + }, + { + "epoch": 7.24, + "learning_rate": 2.277213942603769e-06, + "loss": 0.5102, + "step": 19611 + }, + { + "epoch": 7.24, + "learning_rate": 2.277176992240424e-06, + "loss": 0.3843, + "step": 19612 + }, + { + "epoch": 7.24, + "learning_rate": 2.2771400418770785e-06, + "loss": 0.2788, + "step": 19613 + }, + { + "epoch": 7.24, + "learning_rate": 2.2771030915137336e-06, + "loss": 0.3099, + "step": 19614 + }, + { + "epoch": 7.24, + "learning_rate": 2.2770661411503882e-06, + "loss": 0.3877, + "step": 19615 + }, + { + "epoch": 7.24, + "learning_rate": 2.2770291907870424e-06, + "loss": 0.4282, + "step": 19616 + }, + { + "epoch": 7.24, + "learning_rate": 2.2769922404236975e-06, + "loss": 0.5409, + "step": 19617 + }, + { + "epoch": 7.24, + "learning_rate": 2.276955290060352e-06, + "loss": 0.4515, + "step": 19618 + }, + { + "epoch": 7.24, + "learning_rate": 2.276918339697007e-06, + "loss": 0.3957, + "step": 19619 + }, + { + "epoch": 7.25, + "learning_rate": 2.276881389333662e-06, + "loss": 0.5762, + "step": 19620 + }, + { + "epoch": 7.25, + "learning_rate": 2.2768444389703165e-06, + "loss": 0.4306, + "step": 19621 + }, + { + "epoch": 7.25, + "learning_rate": 2.276807488606971e-06, + "loss": 0.4726, + "step": 19622 + }, + { + "epoch": 7.25, + "learning_rate": 2.276770538243626e-06, + "loss": 0.4792, + "step": 19623 + }, + { + "epoch": 7.25, + "learning_rate": 2.276733587880281e-06, + "loss": 0.4369, + "step": 19624 + }, + { + "epoch": 7.25, + "learning_rate": 2.276696637516936e-06, + "loss": 0.4427, + "step": 19625 + }, + { + "epoch": 7.25, + "learning_rate": 2.2766596871535905e-06, + "loss": 0.3617, + "step": 19626 + }, + { + "epoch": 7.25, + "learning_rate": 2.276622736790245e-06, + "loss": 0.437, + "step": 19627 + }, + { + "epoch": 7.25, + "learning_rate": 2.2765857864268998e-06, + "loss": 0.4562, + "step": 19628 + }, + { + "epoch": 7.25, + "learning_rate": 2.2765488360635544e-06, + "loss": 0.3545, + "step": 19629 + }, + { + "epoch": 7.25, + "learning_rate": 2.2765118857002095e-06, + "loss": 0.5058, + "step": 19630 + }, + { + "epoch": 7.25, + "learning_rate": 2.276474935336864e-06, + "loss": 0.3878, + "step": 19631 + }, + { + "epoch": 7.25, + "learning_rate": 2.276437984973519e-06, + "loss": 0.4908, + "step": 19632 + }, + { + "epoch": 7.25, + "learning_rate": 2.276401034610174e-06, + "loss": 0.3753, + "step": 19633 + }, + { + "epoch": 7.25, + "learning_rate": 2.2763640842468284e-06, + "loss": 0.3923, + "step": 19634 + }, + { + "epoch": 7.25, + "learning_rate": 2.276327133883483e-06, + "loss": 0.3954, + "step": 19635 + }, + { + "epoch": 7.25, + "learning_rate": 2.2762901835201377e-06, + "loss": 0.4536, + "step": 19636 + }, + { + "epoch": 7.25, + "learning_rate": 2.2762532331567928e-06, + "loss": 0.3814, + "step": 19637 + }, + { + "epoch": 7.25, + "learning_rate": 2.2762162827934474e-06, + "loss": 0.3807, + "step": 19638 + }, + { + "epoch": 7.25, + "learning_rate": 2.2761793324301025e-06, + "loss": 0.5013, + "step": 19639 + }, + { + "epoch": 7.25, + "learning_rate": 2.276142382066757e-06, + "loss": 0.4224, + "step": 19640 + }, + { + "epoch": 7.25, + "learning_rate": 2.2761054317034117e-06, + "loss": 0.367, + "step": 19641 + }, + { + "epoch": 7.25, + "learning_rate": 2.2760684813400664e-06, + "loss": 0.3922, + "step": 19642 + }, + { + "epoch": 7.25, + "learning_rate": 2.2760315309767214e-06, + "loss": 0.3632, + "step": 19643 + }, + { + "epoch": 7.25, + "learning_rate": 2.275994580613376e-06, + "loss": 0.5028, + "step": 19644 + }, + { + "epoch": 7.25, + "learning_rate": 2.275957630250031e-06, + "loss": 0.3729, + "step": 19645 + }, + { + "epoch": 7.25, + "learning_rate": 2.2759206798866858e-06, + "loss": 0.3077, + "step": 19646 + }, + { + "epoch": 7.26, + "learning_rate": 2.2758837295233404e-06, + "loss": 0.499, + "step": 19647 + }, + { + "epoch": 7.26, + "learning_rate": 2.275846779159995e-06, + "loss": 0.4476, + "step": 19648 + }, + { + "epoch": 7.26, + "learning_rate": 2.2758098287966497e-06, + "loss": 0.4244, + "step": 19649 + }, + { + "epoch": 7.26, + "learning_rate": 2.2757728784333047e-06, + "loss": 0.4721, + "step": 19650 + }, + { + "epoch": 7.26, + "learning_rate": 2.2757359280699594e-06, + "loss": 0.378, + "step": 19651 + }, + { + "epoch": 7.26, + "learning_rate": 2.2756989777066144e-06, + "loss": 0.5078, + "step": 19652 + }, + { + "epoch": 7.26, + "learning_rate": 2.275662027343269e-06, + "loss": 0.5056, + "step": 19653 + }, + { + "epoch": 7.26, + "learning_rate": 2.2756250769799237e-06, + "loss": 0.3936, + "step": 19654 + }, + { + "epoch": 7.26, + "learning_rate": 2.2755881266165784e-06, + "loss": 0.3888, + "step": 19655 + }, + { + "epoch": 7.26, + "learning_rate": 2.275551176253233e-06, + "loss": 0.5028, + "step": 19656 + }, + { + "epoch": 7.26, + "learning_rate": 2.275514225889888e-06, + "loss": 0.4999, + "step": 19657 + }, + { + "epoch": 7.26, + "learning_rate": 2.2754772755265427e-06, + "loss": 0.4322, + "step": 19658 + }, + { + "epoch": 7.26, + "learning_rate": 2.2754403251631977e-06, + "loss": 0.4735, + "step": 19659 + }, + { + "epoch": 7.26, + "learning_rate": 2.275403374799852e-06, + "loss": 0.4623, + "step": 19660 + }, + { + "epoch": 7.26, + "learning_rate": 2.275366424436507e-06, + "loss": 0.4438, + "step": 19661 + }, + { + "epoch": 7.26, + "learning_rate": 2.2753294740731617e-06, + "loss": 0.5425, + "step": 19662 + }, + { + "epoch": 7.26, + "learning_rate": 2.2752925237098167e-06, + "loss": 0.6162, + "step": 19663 + }, + { + "epoch": 7.26, + "learning_rate": 2.2752555733464714e-06, + "loss": 0.3318, + "step": 19664 + }, + { + "epoch": 7.26, + "learning_rate": 2.2752186229831264e-06, + "loss": 0.4782, + "step": 19665 + }, + { + "epoch": 7.26, + "learning_rate": 2.2751816726197806e-06, + "loss": 0.3403, + "step": 19666 + }, + { + "epoch": 7.26, + "learning_rate": 2.2751447222564353e-06, + "loss": 0.5419, + "step": 19667 + }, + { + "epoch": 7.26, + "learning_rate": 2.2751077718930903e-06, + "loss": 0.4341, + "step": 19668 + }, + { + "epoch": 7.26, + "learning_rate": 2.275070821529745e-06, + "loss": 0.381, + "step": 19669 + }, + { + "epoch": 7.26, + "learning_rate": 2.2750338711664e-06, + "loss": 0.2465, + "step": 19670 + }, + { + "epoch": 7.26, + "learning_rate": 2.2749969208030547e-06, + "loss": 0.4516, + "step": 19671 + }, + { + "epoch": 7.26, + "learning_rate": 2.2749599704397093e-06, + "loss": 0.3637, + "step": 19672 + }, + { + "epoch": 7.26, + "learning_rate": 2.274923020076364e-06, + "loss": 0.3681, + "step": 19673 + }, + { + "epoch": 7.27, + "learning_rate": 2.274886069713019e-06, + "loss": 0.4607, + "step": 19674 + }, + { + "epoch": 7.27, + "learning_rate": 2.2748491193496736e-06, + "loss": 0.3584, + "step": 19675 + }, + { + "epoch": 7.27, + "learning_rate": 2.2748121689863283e-06, + "loss": 0.4103, + "step": 19676 + }, + { + "epoch": 7.27, + "learning_rate": 2.2747752186229833e-06, + "loss": 0.4647, + "step": 19677 + }, + { + "epoch": 7.27, + "learning_rate": 2.274738268259638e-06, + "loss": 0.4058, + "step": 19678 + }, + { + "epoch": 7.27, + "learning_rate": 2.2747013178962926e-06, + "loss": 0.3407, + "step": 19679 + }, + { + "epoch": 7.27, + "learning_rate": 2.2746643675329472e-06, + "loss": 0.3842, + "step": 19680 + }, + { + "epoch": 7.27, + "learning_rate": 2.2746274171696023e-06, + "loss": 0.3623, + "step": 19681 + }, + { + "epoch": 7.27, + "learning_rate": 2.274590466806257e-06, + "loss": 0.371, + "step": 19682 + }, + { + "epoch": 7.27, + "learning_rate": 2.274553516442912e-06, + "loss": 0.4419, + "step": 19683 + }, + { + "epoch": 7.27, + "learning_rate": 2.2745165660795666e-06, + "loss": 0.3639, + "step": 19684 + }, + { + "epoch": 7.27, + "learning_rate": 2.2744796157162213e-06, + "loss": 0.4317, + "step": 19685 + }, + { + "epoch": 7.27, + "learning_rate": 2.274442665352876e-06, + "loss": 0.4685, + "step": 19686 + }, + { + "epoch": 7.27, + "learning_rate": 2.2744057149895305e-06, + "loss": 0.6055, + "step": 19687 + }, + { + "epoch": 7.27, + "learning_rate": 2.2743687646261856e-06, + "loss": 0.4221, + "step": 19688 + }, + { + "epoch": 7.27, + "learning_rate": 2.2743318142628402e-06, + "loss": 0.3627, + "step": 19689 + }, + { + "epoch": 7.27, + "learning_rate": 2.2742948638994953e-06, + "loss": 0.4866, + "step": 19690 + }, + { + "epoch": 7.27, + "learning_rate": 2.27425791353615e-06, + "loss": 0.2857, + "step": 19691 + }, + { + "epoch": 7.27, + "learning_rate": 2.2742209631728046e-06, + "loss": 0.485, + "step": 19692 + }, + { + "epoch": 7.27, + "learning_rate": 2.274184012809459e-06, + "loss": 0.4612, + "step": 19693 + }, + { + "epoch": 7.27, + "learning_rate": 2.2741470624461143e-06, + "loss": 0.5673, + "step": 19694 + }, + { + "epoch": 7.27, + "learning_rate": 2.274110112082769e-06, + "loss": 0.3305, + "step": 19695 + }, + { + "epoch": 7.27, + "learning_rate": 2.2740731617194235e-06, + "loss": 0.3379, + "step": 19696 + }, + { + "epoch": 7.27, + "learning_rate": 2.2740362113560786e-06, + "loss": 0.4391, + "step": 19697 + }, + { + "epoch": 7.27, + "learning_rate": 2.2739992609927332e-06, + "loss": 0.4215, + "step": 19698 + }, + { + "epoch": 7.27, + "learning_rate": 2.273962310629388e-06, + "loss": 0.4686, + "step": 19699 + }, + { + "epoch": 7.27, + "learning_rate": 2.2739253602660425e-06, + "loss": 0.4072, + "step": 19700 + }, + { + "epoch": 7.28, + "learning_rate": 2.2738884099026976e-06, + "loss": 0.497, + "step": 19701 + }, + { + "epoch": 7.28, + "learning_rate": 2.273851459539352e-06, + "loss": 0.412, + "step": 19702 + }, + { + "epoch": 7.28, + "learning_rate": 2.2738145091760073e-06, + "loss": 0.4707, + "step": 19703 + }, + { + "epoch": 7.28, + "learning_rate": 2.273777558812662e-06, + "loss": 0.3901, + "step": 19704 + }, + { + "epoch": 7.28, + "learning_rate": 2.2737406084493165e-06, + "loss": 0.4586, + "step": 19705 + }, + { + "epoch": 7.28, + "learning_rate": 2.273703658085971e-06, + "loss": 0.407, + "step": 19706 + }, + { + "epoch": 7.28, + "learning_rate": 2.273666707722626e-06, + "loss": 0.3404, + "step": 19707 + }, + { + "epoch": 7.28, + "learning_rate": 2.273629757359281e-06, + "loss": 0.3217, + "step": 19708 + }, + { + "epoch": 7.28, + "learning_rate": 2.2735928069959355e-06, + "loss": 0.3733, + "step": 19709 + }, + { + "epoch": 7.28, + "learning_rate": 2.2735558566325906e-06, + "loss": 0.4102, + "step": 19710 + }, + { + "epoch": 7.28, + "learning_rate": 2.2735189062692448e-06, + "loss": 0.3866, + "step": 19711 + }, + { + "epoch": 7.28, + "learning_rate": 2.2734819559059e-06, + "loss": 0.3804, + "step": 19712 + }, + { + "epoch": 7.28, + "learning_rate": 2.2734450055425545e-06, + "loss": 0.6043, + "step": 19713 + }, + { + "epoch": 7.28, + "learning_rate": 2.2734080551792095e-06, + "loss": 0.5325, + "step": 19714 + }, + { + "epoch": 7.28, + "learning_rate": 2.273371104815864e-06, + "loss": 0.459, + "step": 19715 + }, + { + "epoch": 7.28, + "learning_rate": 2.273334154452519e-06, + "loss": 0.4396, + "step": 19716 + }, + { + "epoch": 7.28, + "learning_rate": 2.2732972040891734e-06, + "loss": 0.586, + "step": 19717 + }, + { + "epoch": 7.28, + "learning_rate": 2.273260253725828e-06, + "loss": 0.5043, + "step": 19718 + }, + { + "epoch": 7.28, + "learning_rate": 2.273223303362483e-06, + "loss": 0.5026, + "step": 19719 + }, + { + "epoch": 7.28, + "learning_rate": 2.2731863529991378e-06, + "loss": 0.3774, + "step": 19720 + }, + { + "epoch": 7.28, + "learning_rate": 2.273149402635793e-06, + "loss": 0.4202, + "step": 19721 + }, + { + "epoch": 7.28, + "learning_rate": 2.2731124522724475e-06, + "loss": 0.3739, + "step": 19722 + }, + { + "epoch": 7.28, + "learning_rate": 2.273075501909102e-06, + "loss": 0.3488, + "step": 19723 + }, + { + "epoch": 7.28, + "learning_rate": 2.2730385515457567e-06, + "loss": 0.3753, + "step": 19724 + }, + { + "epoch": 7.28, + "learning_rate": 2.273001601182412e-06, + "loss": 0.5857, + "step": 19725 + }, + { + "epoch": 7.28, + "learning_rate": 2.2729646508190664e-06, + "loss": 0.3854, + "step": 19726 + }, + { + "epoch": 7.28, + "learning_rate": 2.272927700455721e-06, + "loss": 0.4792, + "step": 19727 + }, + { + "epoch": 7.29, + "learning_rate": 2.272890750092376e-06, + "loss": 0.4263, + "step": 19728 + }, + { + "epoch": 7.29, + "learning_rate": 2.2728537997290308e-06, + "loss": 0.4033, + "step": 19729 + }, + { + "epoch": 7.29, + "learning_rate": 2.2728168493656854e-06, + "loss": 0.3939, + "step": 19730 + }, + { + "epoch": 7.29, + "learning_rate": 2.27277989900234e-06, + "loss": 0.431, + "step": 19731 + }, + { + "epoch": 7.29, + "learning_rate": 2.272742948638995e-06, + "loss": 0.39, + "step": 19732 + }, + { + "epoch": 7.29, + "learning_rate": 2.2727059982756497e-06, + "loss": 0.478, + "step": 19733 + }, + { + "epoch": 7.29, + "learning_rate": 2.272669047912305e-06, + "loss": 0.367, + "step": 19734 + }, + { + "epoch": 7.29, + "learning_rate": 2.2726320975489594e-06, + "loss": 0.4385, + "step": 19735 + }, + { + "epoch": 7.29, + "learning_rate": 2.272595147185614e-06, + "loss": 0.3489, + "step": 19736 + }, + { + "epoch": 7.29, + "learning_rate": 2.2725581968222687e-06, + "loss": 0.3865, + "step": 19737 + }, + { + "epoch": 7.29, + "learning_rate": 2.2725212464589233e-06, + "loss": 0.4867, + "step": 19738 + }, + { + "epoch": 7.29, + "learning_rate": 2.2724842960955784e-06, + "loss": 0.491, + "step": 19739 + }, + { + "epoch": 7.29, + "learning_rate": 2.272447345732233e-06, + "loss": 0.4641, + "step": 19740 + }, + { + "epoch": 7.29, + "learning_rate": 2.272410395368888e-06, + "loss": 0.3551, + "step": 19741 + }, + { + "epoch": 7.29, + "learning_rate": 2.2723734450055427e-06, + "loss": 0.4129, + "step": 19742 + }, + { + "epoch": 7.29, + "learning_rate": 2.2723364946421974e-06, + "loss": 0.4489, + "step": 19743 + }, + { + "epoch": 7.29, + "learning_rate": 2.272299544278852e-06, + "loss": 0.4298, + "step": 19744 + }, + { + "epoch": 7.29, + "learning_rate": 2.272262593915507e-06, + "loss": 0.2976, + "step": 19745 + }, + { + "epoch": 7.29, + "learning_rate": 2.2722256435521617e-06, + "loss": 0.498, + "step": 19746 + }, + { + "epoch": 7.29, + "learning_rate": 2.2721886931888163e-06, + "loss": 0.3376, + "step": 19747 + }, + { + "epoch": 7.29, + "learning_rate": 2.2721517428254714e-06, + "loss": 0.3845, + "step": 19748 + }, + { + "epoch": 7.29, + "learning_rate": 2.272114792462126e-06, + "loss": 0.4165, + "step": 19749 + }, + { + "epoch": 7.29, + "learning_rate": 2.2720778420987807e-06, + "loss": 0.446, + "step": 19750 + }, + { + "epoch": 7.29, + "learning_rate": 2.2720408917354353e-06, + "loss": 0.402, + "step": 19751 + }, + { + "epoch": 7.29, + "learning_rate": 2.2720039413720904e-06, + "loss": 0.3988, + "step": 19752 + }, + { + "epoch": 7.29, + "learning_rate": 2.271966991008745e-06, + "loss": 0.3297, + "step": 19753 + }, + { + "epoch": 7.29, + "learning_rate": 2.2719300406454e-06, + "loss": 0.3128, + "step": 19754 + }, + { + "epoch": 7.3, + "learning_rate": 2.2718930902820547e-06, + "loss": 0.3094, + "step": 19755 + }, + { + "epoch": 7.3, + "learning_rate": 2.271856139918709e-06, + "loss": 0.5301, + "step": 19756 + }, + { + "epoch": 7.3, + "learning_rate": 2.271819189555364e-06, + "loss": 0.497, + "step": 19757 + }, + { + "epoch": 7.3, + "learning_rate": 2.2717822391920186e-06, + "loss": 0.462, + "step": 19758 + }, + { + "epoch": 7.3, + "learning_rate": 2.2717452888286737e-06, + "loss": 0.3702, + "step": 19759 + }, + { + "epoch": 7.3, + "learning_rate": 2.2717083384653283e-06, + "loss": 0.3711, + "step": 19760 + }, + { + "epoch": 7.3, + "learning_rate": 2.2716713881019834e-06, + "loss": 0.3945, + "step": 19761 + }, + { + "epoch": 7.3, + "learning_rate": 2.2716344377386376e-06, + "loss": 0.5241, + "step": 19762 + }, + { + "epoch": 7.3, + "learning_rate": 2.2715974873752926e-06, + "loss": 0.4766, + "step": 19763 + }, + { + "epoch": 7.3, + "learning_rate": 2.2715605370119473e-06, + "loss": 0.2843, + "step": 19764 + }, + { + "epoch": 7.3, + "learning_rate": 2.2715235866486023e-06, + "loss": 0.4114, + "step": 19765 + }, + { + "epoch": 7.3, + "learning_rate": 2.271486636285257e-06, + "loss": 0.5119, + "step": 19766 + }, + { + "epoch": 7.3, + "learning_rate": 2.2714496859219116e-06, + "loss": 0.377, + "step": 19767 + }, + { + "epoch": 7.3, + "learning_rate": 2.2714127355585662e-06, + "loss": 0.4019, + "step": 19768 + }, + { + "epoch": 7.3, + "learning_rate": 2.271375785195221e-06, + "loss": 0.4798, + "step": 19769 + }, + { + "epoch": 7.3, + "learning_rate": 2.271338834831876e-06, + "loss": 0.4947, + "step": 19770 + }, + { + "epoch": 7.3, + "learning_rate": 2.2713018844685306e-06, + "loss": 0.3909, + "step": 19771 + }, + { + "epoch": 7.3, + "learning_rate": 2.2712649341051856e-06, + "loss": 0.3985, + "step": 19772 + }, + { + "epoch": 7.3, + "learning_rate": 2.2712279837418403e-06, + "loss": 0.2581, + "step": 19773 + }, + { + "epoch": 7.3, + "learning_rate": 2.271191033378495e-06, + "loss": 0.435, + "step": 19774 + }, + { + "epoch": 7.3, + "learning_rate": 2.2711540830151495e-06, + "loss": 0.5163, + "step": 19775 + }, + { + "epoch": 7.3, + "learning_rate": 2.271117132651804e-06, + "loss": 0.48, + "step": 19776 + }, + { + "epoch": 7.3, + "learning_rate": 2.2710801822884592e-06, + "loss": 0.4041, + "step": 19777 + }, + { + "epoch": 7.3, + "learning_rate": 2.271043231925114e-06, + "loss": 0.4122, + "step": 19778 + }, + { + "epoch": 7.3, + "learning_rate": 2.271006281561769e-06, + "loss": 0.411, + "step": 19779 + }, + { + "epoch": 7.3, + "learning_rate": 2.2709693311984236e-06, + "loss": 0.4814, + "step": 19780 + }, + { + "epoch": 7.3, + "learning_rate": 2.2709323808350782e-06, + "loss": 0.3254, + "step": 19781 + }, + { + "epoch": 7.31, + "learning_rate": 2.270895430471733e-06, + "loss": 0.4321, + "step": 19782 + }, + { + "epoch": 7.31, + "learning_rate": 2.270858480108388e-06, + "loss": 0.4973, + "step": 19783 + }, + { + "epoch": 7.31, + "learning_rate": 2.2708215297450425e-06, + "loss": 0.3723, + "step": 19784 + }, + { + "epoch": 7.31, + "learning_rate": 2.270784579381697e-06, + "loss": 0.366, + "step": 19785 + }, + { + "epoch": 7.31, + "learning_rate": 2.2707476290183522e-06, + "loss": 0.324, + "step": 19786 + }, + { + "epoch": 7.31, + "learning_rate": 2.270710678655007e-06, + "loss": 0.4327, + "step": 19787 + }, + { + "epoch": 7.31, + "learning_rate": 2.2706737282916615e-06, + "loss": 0.43, + "step": 19788 + }, + { + "epoch": 7.31, + "learning_rate": 2.270636777928316e-06, + "loss": 0.4184, + "step": 19789 + }, + { + "epoch": 7.31, + "learning_rate": 2.2705998275649712e-06, + "loss": 0.376, + "step": 19790 + }, + { + "epoch": 7.31, + "learning_rate": 2.270562877201626e-06, + "loss": 0.4291, + "step": 19791 + }, + { + "epoch": 7.31, + "learning_rate": 2.270525926838281e-06, + "loss": 0.4815, + "step": 19792 + }, + { + "epoch": 7.31, + "learning_rate": 2.2704889764749355e-06, + "loss": 0.4964, + "step": 19793 + }, + { + "epoch": 7.31, + "learning_rate": 2.27045202611159e-06, + "loss": 0.3458, + "step": 19794 + }, + { + "epoch": 7.31, + "learning_rate": 2.270415075748245e-06, + "loss": 0.398, + "step": 19795 + }, + { + "epoch": 7.31, + "learning_rate": 2.2703781253848995e-06, + "loss": 0.4618, + "step": 19796 + }, + { + "epoch": 7.31, + "learning_rate": 2.2703411750215545e-06, + "loss": 0.5162, + "step": 19797 + }, + { + "epoch": 7.31, + "learning_rate": 2.270304224658209e-06, + "loss": 0.5325, + "step": 19798 + }, + { + "epoch": 7.31, + "learning_rate": 2.270267274294864e-06, + "loss": 0.5226, + "step": 19799 + }, + { + "epoch": 7.31, + "learning_rate": 2.270230323931519e-06, + "loss": 0.4185, + "step": 19800 + }, + { + "epoch": 7.31, + "learning_rate": 2.2701933735681735e-06, + "loss": 0.3731, + "step": 19801 + }, + { + "epoch": 7.31, + "learning_rate": 2.270156423204828e-06, + "loss": 0.4546, + "step": 19802 + }, + { + "epoch": 7.31, + "learning_rate": 2.270119472841483e-06, + "loss": 0.5359, + "step": 19803 + }, + { + "epoch": 7.31, + "learning_rate": 2.270082522478138e-06, + "loss": 0.3724, + "step": 19804 + }, + { + "epoch": 7.31, + "learning_rate": 2.2700455721147925e-06, + "loss": 0.5084, + "step": 19805 + }, + { + "epoch": 7.31, + "learning_rate": 2.270008621751447e-06, + "loss": 0.4405, + "step": 19806 + }, + { + "epoch": 7.31, + "learning_rate": 2.2699716713881017e-06, + "loss": 0.4049, + "step": 19807 + }, + { + "epoch": 7.31, + "learning_rate": 2.2699347210247568e-06, + "loss": 0.3634, + "step": 19808 + }, + { + "epoch": 7.31, + "learning_rate": 2.2698977706614114e-06, + "loss": 0.3834, + "step": 19809 + }, + { + "epoch": 7.32, + "learning_rate": 2.2698608202980665e-06, + "loss": 0.4396, + "step": 19810 + }, + { + "epoch": 7.32, + "learning_rate": 2.269823869934721e-06, + "loss": 0.3384, + "step": 19811 + }, + { + "epoch": 7.32, + "learning_rate": 2.2697869195713758e-06, + "loss": 0.5403, + "step": 19812 + }, + { + "epoch": 7.32, + "learning_rate": 2.2697499692080304e-06, + "loss": 0.3603, + "step": 19813 + }, + { + "epoch": 7.32, + "learning_rate": 2.2697130188446855e-06, + "loss": 0.3763, + "step": 19814 + }, + { + "epoch": 7.32, + "learning_rate": 2.26967606848134e-06, + "loss": 0.4282, + "step": 19815 + }, + { + "epoch": 7.32, + "learning_rate": 2.2696391181179947e-06, + "loss": 0.4966, + "step": 19816 + }, + { + "epoch": 7.32, + "learning_rate": 2.2696021677546498e-06, + "loss": 0.3428, + "step": 19817 + }, + { + "epoch": 7.32, + "learning_rate": 2.2695652173913044e-06, + "loss": 0.6525, + "step": 19818 + }, + { + "epoch": 7.32, + "learning_rate": 2.269528267027959e-06, + "loss": 0.38, + "step": 19819 + }, + { + "epoch": 7.32, + "learning_rate": 2.2694913166646137e-06, + "loss": 0.413, + "step": 19820 + }, + { + "epoch": 7.32, + "learning_rate": 2.2694543663012688e-06, + "loss": 0.406, + "step": 19821 + }, + { + "epoch": 7.32, + "learning_rate": 2.2694174159379234e-06, + "loss": 0.2843, + "step": 19822 + }, + { + "epoch": 7.32, + "learning_rate": 2.2693804655745784e-06, + "loss": 0.3885, + "step": 19823 + }, + { + "epoch": 7.32, + "learning_rate": 2.269343515211233e-06, + "loss": 0.4489, + "step": 19824 + }, + { + "epoch": 7.32, + "learning_rate": 2.2693065648478877e-06, + "loss": 0.3918, + "step": 19825 + }, + { + "epoch": 7.32, + "learning_rate": 2.2692696144845424e-06, + "loss": 0.4699, + "step": 19826 + }, + { + "epoch": 7.32, + "learning_rate": 2.269232664121197e-06, + "loss": 0.4681, + "step": 19827 + }, + { + "epoch": 7.32, + "learning_rate": 2.269195713757852e-06, + "loss": 0.4299, + "step": 19828 + }, + { + "epoch": 7.32, + "learning_rate": 2.2691587633945067e-06, + "loss": 0.4063, + "step": 19829 + }, + { + "epoch": 7.32, + "learning_rate": 2.2691218130311618e-06, + "loss": 0.4561, + "step": 19830 + }, + { + "epoch": 7.32, + "learning_rate": 2.2690848626678164e-06, + "loss": 0.54, + "step": 19831 + }, + { + "epoch": 7.32, + "learning_rate": 2.269047912304471e-06, + "loss": 0.3784, + "step": 19832 + }, + { + "epoch": 7.32, + "learning_rate": 2.2690109619411257e-06, + "loss": 0.3774, + "step": 19833 + }, + { + "epoch": 7.32, + "learning_rate": 2.2689740115777807e-06, + "loss": 0.4883, + "step": 19834 + }, + { + "epoch": 7.32, + "learning_rate": 2.2689370612144354e-06, + "loss": 0.3874, + "step": 19835 + }, + { + "epoch": 7.32, + "learning_rate": 2.26890011085109e-06, + "loss": 0.4786, + "step": 19836 + }, + { + "epoch": 7.33, + "learning_rate": 2.268863160487745e-06, + "loss": 0.4023, + "step": 19837 + }, + { + "epoch": 7.33, + "learning_rate": 2.2688262101243997e-06, + "loss": 0.4121, + "step": 19838 + }, + { + "epoch": 7.33, + "learning_rate": 2.2687892597610543e-06, + "loss": 0.4176, + "step": 19839 + }, + { + "epoch": 7.33, + "learning_rate": 2.268752309397709e-06, + "loss": 0.3592, + "step": 19840 + }, + { + "epoch": 7.33, + "learning_rate": 2.268715359034364e-06, + "loss": 0.4277, + "step": 19841 + }, + { + "epoch": 7.33, + "learning_rate": 2.2686784086710187e-06, + "loss": 0.4548, + "step": 19842 + }, + { + "epoch": 7.33, + "learning_rate": 2.2686414583076737e-06, + "loss": 0.4243, + "step": 19843 + }, + { + "epoch": 7.33, + "learning_rate": 2.2686045079443284e-06, + "loss": 0.4149, + "step": 19844 + }, + { + "epoch": 7.33, + "learning_rate": 2.2685675575809826e-06, + "loss": 0.4543, + "step": 19845 + }, + { + "epoch": 7.33, + "learning_rate": 2.2685306072176376e-06, + "loss": 0.4871, + "step": 19846 + }, + { + "epoch": 7.33, + "learning_rate": 2.2684936568542923e-06, + "loss": 0.4464, + "step": 19847 + }, + { + "epoch": 7.33, + "learning_rate": 2.2684567064909473e-06, + "loss": 0.3124, + "step": 19848 + }, + { + "epoch": 7.33, + "learning_rate": 2.268419756127602e-06, + "loss": 0.4336, + "step": 19849 + }, + { + "epoch": 7.33, + "learning_rate": 2.268382805764257e-06, + "loss": 0.4414, + "step": 19850 + }, + { + "epoch": 7.33, + "learning_rate": 2.2683458554009112e-06, + "loss": 0.4737, + "step": 19851 + }, + { + "epoch": 7.33, + "learning_rate": 2.2683089050375663e-06, + "loss": 0.3145, + "step": 19852 + }, + { + "epoch": 7.33, + "learning_rate": 2.268271954674221e-06, + "loss": 0.38, + "step": 19853 + }, + { + "epoch": 7.33, + "learning_rate": 2.268235004310876e-06, + "loss": 0.4232, + "step": 19854 + }, + { + "epoch": 7.33, + "learning_rate": 2.2681980539475306e-06, + "loss": 0.3695, + "step": 19855 + }, + { + "epoch": 7.33, + "learning_rate": 2.2681611035841853e-06, + "loss": 0.38, + "step": 19856 + }, + { + "epoch": 7.33, + "learning_rate": 2.26812415322084e-06, + "loss": 0.4464, + "step": 19857 + }, + { + "epoch": 7.33, + "learning_rate": 2.2680872028574945e-06, + "loss": 0.3255, + "step": 19858 + }, + { + "epoch": 7.33, + "learning_rate": 2.2680502524941496e-06, + "loss": 0.4767, + "step": 19859 + }, + { + "epoch": 7.33, + "learning_rate": 2.2680133021308042e-06, + "loss": 0.5007, + "step": 19860 + }, + { + "epoch": 7.33, + "learning_rate": 2.2679763517674593e-06, + "loss": 0.2885, + "step": 19861 + }, + { + "epoch": 7.33, + "learning_rate": 2.267939401404114e-06, + "loss": 0.5064, + "step": 19862 + }, + { + "epoch": 7.33, + "learning_rate": 2.2679024510407686e-06, + "loss": 0.3678, + "step": 19863 + }, + { + "epoch": 7.34, + "learning_rate": 2.267865500677423e-06, + "loss": 0.3791, + "step": 19864 + }, + { + "epoch": 7.34, + "learning_rate": 2.267828550314078e-06, + "loss": 0.477, + "step": 19865 + }, + { + "epoch": 7.34, + "learning_rate": 2.267791599950733e-06, + "loss": 0.3062, + "step": 19866 + }, + { + "epoch": 7.34, + "learning_rate": 2.2677546495873875e-06, + "loss": 0.4051, + "step": 19867 + }, + { + "epoch": 7.34, + "learning_rate": 2.2677176992240426e-06, + "loss": 0.4971, + "step": 19868 + }, + { + "epoch": 7.34, + "learning_rate": 2.2676807488606972e-06, + "loss": 0.3585, + "step": 19869 + }, + { + "epoch": 7.34, + "learning_rate": 2.267643798497352e-06, + "loss": 0.4243, + "step": 19870 + }, + { + "epoch": 7.34, + "learning_rate": 2.2676068481340065e-06, + "loss": 0.4366, + "step": 19871 + }, + { + "epoch": 7.34, + "learning_rate": 2.2675698977706616e-06, + "loss": 0.4357, + "step": 19872 + }, + { + "epoch": 7.34, + "learning_rate": 2.267532947407316e-06, + "loss": 0.3372, + "step": 19873 + }, + { + "epoch": 7.34, + "learning_rate": 2.2674959970439713e-06, + "loss": 0.4221, + "step": 19874 + }, + { + "epoch": 7.34, + "learning_rate": 2.267459046680626e-06, + "loss": 0.3967, + "step": 19875 + }, + { + "epoch": 7.34, + "learning_rate": 2.2674220963172805e-06, + "loss": 0.5591, + "step": 19876 + }, + { + "epoch": 7.34, + "learning_rate": 2.267385145953935e-06, + "loss": 0.551, + "step": 19877 + }, + { + "epoch": 7.34, + "learning_rate": 2.26734819559059e-06, + "loss": 0.4848, + "step": 19878 + }, + { + "epoch": 7.34, + "learning_rate": 2.267311245227245e-06, + "loss": 0.3203, + "step": 19879 + }, + { + "epoch": 7.34, + "learning_rate": 2.2672742948638995e-06, + "loss": 0.4553, + "step": 19880 + }, + { + "epoch": 7.34, + "learning_rate": 2.2672373445005546e-06, + "loss": 0.4479, + "step": 19881 + }, + { + "epoch": 7.34, + "learning_rate": 2.267200394137209e-06, + "loss": 0.3336, + "step": 19882 + }, + { + "epoch": 7.34, + "learning_rate": 2.267163443773864e-06, + "loss": 0.4486, + "step": 19883 + }, + { + "epoch": 7.34, + "learning_rate": 2.2671264934105185e-06, + "loss": 0.4082, + "step": 19884 + }, + { + "epoch": 7.34, + "learning_rate": 2.267089543047173e-06, + "loss": 0.4353, + "step": 19885 + }, + { + "epoch": 7.34, + "learning_rate": 2.267052592683828e-06, + "loss": 0.3907, + "step": 19886 + }, + { + "epoch": 7.34, + "learning_rate": 2.267015642320483e-06, + "loss": 0.4078, + "step": 19887 + }, + { + "epoch": 7.34, + "learning_rate": 2.266978691957138e-06, + "loss": 0.471, + "step": 19888 + }, + { + "epoch": 7.34, + "learning_rate": 2.2669417415937925e-06, + "loss": 0.4408, + "step": 19889 + }, + { + "epoch": 7.34, + "learning_rate": 2.266904791230447e-06, + "loss": 0.4988, + "step": 19890 + }, + { + "epoch": 7.35, + "learning_rate": 2.2668678408671018e-06, + "loss": 0.4681, + "step": 19891 + }, + { + "epoch": 7.35, + "learning_rate": 2.266830890503757e-06, + "loss": 0.3369, + "step": 19892 + }, + { + "epoch": 7.35, + "learning_rate": 2.2667939401404115e-06, + "loss": 0.4423, + "step": 19893 + }, + { + "epoch": 7.35, + "learning_rate": 2.2667569897770665e-06, + "loss": 0.4604, + "step": 19894 + }, + { + "epoch": 7.35, + "learning_rate": 2.266720039413721e-06, + "loss": 0.4958, + "step": 19895 + }, + { + "epoch": 7.35, + "learning_rate": 2.2666830890503754e-06, + "loss": 0.289, + "step": 19896 + }, + { + "epoch": 7.35, + "learning_rate": 2.2666461386870304e-06, + "loss": 0.418, + "step": 19897 + }, + { + "epoch": 7.35, + "learning_rate": 2.266609188323685e-06, + "loss": 0.4032, + "step": 19898 + }, + { + "epoch": 7.35, + "learning_rate": 2.26657223796034e-06, + "loss": 0.4156, + "step": 19899 + }, + { + "epoch": 7.35, + "learning_rate": 2.2665352875969948e-06, + "loss": 0.5127, + "step": 19900 + }, + { + "epoch": 7.35, + "learning_rate": 2.26649833723365e-06, + "loss": 0.4548, + "step": 19901 + }, + { + "epoch": 7.35, + "learning_rate": 2.266461386870304e-06, + "loss": 0.4198, + "step": 19902 + }, + { + "epoch": 7.35, + "learning_rate": 2.266424436506959e-06, + "loss": 0.4702, + "step": 19903 + }, + { + "epoch": 7.35, + "learning_rate": 2.2663874861436137e-06, + "loss": 0.3432, + "step": 19904 + }, + { + "epoch": 7.35, + "learning_rate": 2.2663505357802684e-06, + "loss": 0.4719, + "step": 19905 + }, + { + "epoch": 7.35, + "learning_rate": 2.2663135854169234e-06, + "loss": 0.5833, + "step": 19906 + }, + { + "epoch": 7.35, + "learning_rate": 2.266276635053578e-06, + "loss": 0.3589, + "step": 19907 + }, + { + "epoch": 7.35, + "learning_rate": 2.2662396846902327e-06, + "loss": 0.5082, + "step": 19908 + }, + { + "epoch": 7.35, + "learning_rate": 2.2662027343268873e-06, + "loss": 0.4691, + "step": 19909 + }, + { + "epoch": 7.35, + "learning_rate": 2.2661657839635424e-06, + "loss": 0.5414, + "step": 19910 + }, + { + "epoch": 7.35, + "learning_rate": 2.266128833600197e-06, + "loss": 0.4192, + "step": 19911 + }, + { + "epoch": 7.35, + "learning_rate": 2.266091883236852e-06, + "loss": 0.3646, + "step": 19912 + }, + { + "epoch": 7.35, + "learning_rate": 2.2660549328735067e-06, + "loss": 0.4663, + "step": 19913 + }, + { + "epoch": 7.35, + "learning_rate": 2.2660179825101614e-06, + "loss": 0.3762, + "step": 19914 + }, + { + "epoch": 7.35, + "learning_rate": 2.265981032146816e-06, + "loss": 0.4939, + "step": 19915 + }, + { + "epoch": 7.35, + "learning_rate": 2.2659440817834706e-06, + "loss": 0.3914, + "step": 19916 + }, + { + "epoch": 7.35, + "learning_rate": 2.2659071314201257e-06, + "loss": 0.408, + "step": 19917 + }, + { + "epoch": 7.36, + "learning_rate": 2.2658701810567803e-06, + "loss": 0.4264, + "step": 19918 + }, + { + "epoch": 7.36, + "learning_rate": 2.2658332306934354e-06, + "loss": 0.4754, + "step": 19919 + }, + { + "epoch": 7.36, + "learning_rate": 2.26579628033009e-06, + "loss": 0.3935, + "step": 19920 + }, + { + "epoch": 7.36, + "learning_rate": 2.2657593299667447e-06, + "loss": 0.3792, + "step": 19921 + }, + { + "epoch": 7.36, + "learning_rate": 2.2657223796033993e-06, + "loss": 0.4148, + "step": 19922 + }, + { + "epoch": 7.36, + "learning_rate": 2.2656854292400544e-06, + "loss": 0.3614, + "step": 19923 + }, + { + "epoch": 7.36, + "learning_rate": 2.265648478876709e-06, + "loss": 0.4213, + "step": 19924 + }, + { + "epoch": 7.36, + "learning_rate": 2.2656115285133636e-06, + "loss": 0.4146, + "step": 19925 + }, + { + "epoch": 7.36, + "learning_rate": 2.2655745781500187e-06, + "loss": 0.4107, + "step": 19926 + }, + { + "epoch": 7.36, + "learning_rate": 2.2655376277866733e-06, + "loss": 0.4839, + "step": 19927 + }, + { + "epoch": 7.36, + "learning_rate": 2.265500677423328e-06, + "loss": 0.3761, + "step": 19928 + }, + { + "epoch": 7.36, + "learning_rate": 2.2654637270599826e-06, + "loss": 0.4151, + "step": 19929 + }, + { + "epoch": 7.36, + "learning_rate": 2.2654267766966377e-06, + "loss": 0.4207, + "step": 19930 + }, + { + "epoch": 7.36, + "learning_rate": 2.2653898263332923e-06, + "loss": 0.4728, + "step": 19931 + }, + { + "epoch": 7.36, + "learning_rate": 2.2653528759699474e-06, + "loss": 0.5071, + "step": 19932 + }, + { + "epoch": 7.36, + "learning_rate": 2.265315925606602e-06, + "loss": 0.4533, + "step": 19933 + }, + { + "epoch": 7.36, + "learning_rate": 2.2652789752432566e-06, + "loss": 0.42, + "step": 19934 + }, + { + "epoch": 7.36, + "learning_rate": 2.2652420248799113e-06, + "loss": 0.5017, + "step": 19935 + }, + { + "epoch": 7.36, + "learning_rate": 2.265205074516566e-06, + "loss": 0.4807, + "step": 19936 + }, + { + "epoch": 7.36, + "learning_rate": 2.265168124153221e-06, + "loss": 0.4783, + "step": 19937 + }, + { + "epoch": 7.36, + "learning_rate": 2.2651311737898756e-06, + "loss": 0.3417, + "step": 19938 + }, + { + "epoch": 7.36, + "learning_rate": 2.2650942234265307e-06, + "loss": 0.2844, + "step": 19939 + }, + { + "epoch": 7.36, + "learning_rate": 2.2650572730631853e-06, + "loss": 0.4434, + "step": 19940 + }, + { + "epoch": 7.36, + "learning_rate": 2.26502032269984e-06, + "loss": 0.3657, + "step": 19941 + }, + { + "epoch": 7.36, + "learning_rate": 2.2649833723364946e-06, + "loss": 0.4043, + "step": 19942 + }, + { + "epoch": 7.36, + "learning_rate": 2.2649464219731496e-06, + "loss": 0.3447, + "step": 19943 + }, + { + "epoch": 7.36, + "learning_rate": 2.2649094716098043e-06, + "loss": 0.3397, + "step": 19944 + }, + { + "epoch": 7.37, + "learning_rate": 2.264872521246459e-06, + "loss": 0.3744, + "step": 19945 + }, + { + "epoch": 7.37, + "learning_rate": 2.264835570883114e-06, + "loss": 0.4212, + "step": 19946 + }, + { + "epoch": 7.37, + "learning_rate": 2.264798620519768e-06, + "loss": 0.4233, + "step": 19947 + }, + { + "epoch": 7.37, + "learning_rate": 2.2647616701564233e-06, + "loss": 0.5524, + "step": 19948 + }, + { + "epoch": 7.37, + "learning_rate": 2.264724719793078e-06, + "loss": 0.3889, + "step": 19949 + }, + { + "epoch": 7.37, + "learning_rate": 2.264687769429733e-06, + "loss": 0.3835, + "step": 19950 + }, + { + "epoch": 7.37, + "learning_rate": 2.2646508190663876e-06, + "loss": 0.4091, + "step": 19951 + }, + { + "epoch": 7.37, + "learning_rate": 2.2646138687030426e-06, + "loss": 0.4159, + "step": 19952 + }, + { + "epoch": 7.37, + "learning_rate": 2.264576918339697e-06, + "loss": 0.5117, + "step": 19953 + }, + { + "epoch": 7.37, + "learning_rate": 2.264539967976352e-06, + "loss": 0.4525, + "step": 19954 + }, + { + "epoch": 7.37, + "learning_rate": 2.2645030176130066e-06, + "loss": 0.6446, + "step": 19955 + }, + { + "epoch": 7.37, + "learning_rate": 2.264466067249661e-06, + "loss": 0.3711, + "step": 19956 + }, + { + "epoch": 7.37, + "learning_rate": 2.2644291168863162e-06, + "loss": 0.4068, + "step": 19957 + }, + { + "epoch": 7.37, + "learning_rate": 2.264392166522971e-06, + "loss": 0.3913, + "step": 19958 + }, + { + "epoch": 7.37, + "learning_rate": 2.2643552161596255e-06, + "loss": 0.4036, + "step": 19959 + }, + { + "epoch": 7.37, + "learning_rate": 2.26431826579628e-06, + "loss": 0.3892, + "step": 19960 + }, + { + "epoch": 7.37, + "learning_rate": 2.2642813154329352e-06, + "loss": 0.4652, + "step": 19961 + }, + { + "epoch": 7.37, + "learning_rate": 2.26424436506959e-06, + "loss": 0.3279, + "step": 19962 + }, + { + "epoch": 7.37, + "learning_rate": 2.264207414706245e-06, + "loss": 0.4516, + "step": 19963 + }, + { + "epoch": 7.37, + "learning_rate": 2.2641704643428996e-06, + "loss": 0.3469, + "step": 19964 + }, + { + "epoch": 7.37, + "learning_rate": 2.264133513979554e-06, + "loss": 0.3649, + "step": 19965 + }, + { + "epoch": 7.37, + "learning_rate": 2.264096563616209e-06, + "loss": 0.4191, + "step": 19966 + }, + { + "epoch": 7.37, + "learning_rate": 2.2640596132528635e-06, + "loss": 0.5399, + "step": 19967 + }, + { + "epoch": 7.37, + "learning_rate": 2.2640226628895185e-06, + "loss": 0.3698, + "step": 19968 + }, + { + "epoch": 7.37, + "learning_rate": 2.263985712526173e-06, + "loss": 0.4484, + "step": 19969 + }, + { + "epoch": 7.37, + "learning_rate": 2.2639487621628282e-06, + "loss": 0.4434, + "step": 19970 + }, + { + "epoch": 7.37, + "learning_rate": 2.263911811799483e-06, + "loss": 0.4719, + "step": 19971 + }, + { + "epoch": 7.38, + "learning_rate": 2.2638748614361375e-06, + "loss": 0.4949, + "step": 19972 + }, + { + "epoch": 7.38, + "learning_rate": 2.263837911072792e-06, + "loss": 0.4634, + "step": 19973 + }, + { + "epoch": 7.38, + "learning_rate": 2.263800960709447e-06, + "loss": 0.4637, + "step": 19974 + }, + { + "epoch": 7.38, + "learning_rate": 2.263764010346102e-06, + "loss": 0.5661, + "step": 19975 + }, + { + "epoch": 7.38, + "learning_rate": 2.2637270599827565e-06, + "loss": 0.2905, + "step": 19976 + }, + { + "epoch": 7.38, + "learning_rate": 2.2636901096194115e-06, + "loss": 0.4018, + "step": 19977 + }, + { + "epoch": 7.38, + "learning_rate": 2.263653159256066e-06, + "loss": 0.4491, + "step": 19978 + }, + { + "epoch": 7.38, + "learning_rate": 2.263616208892721e-06, + "loss": 0.4446, + "step": 19979 + }, + { + "epoch": 7.38, + "learning_rate": 2.2635792585293754e-06, + "loss": 0.5339, + "step": 19980 + }, + { + "epoch": 7.38, + "learning_rate": 2.2635423081660305e-06, + "loss": 0.4827, + "step": 19981 + }, + { + "epoch": 7.38, + "learning_rate": 2.263505357802685e-06, + "loss": 0.3958, + "step": 19982 + }, + { + "epoch": 7.38, + "learning_rate": 2.26346840743934e-06, + "loss": 0.5915, + "step": 19983 + }, + { + "epoch": 7.38, + "learning_rate": 2.263431457075995e-06, + "loss": 0.305, + "step": 19984 + }, + { + "epoch": 7.38, + "learning_rate": 2.2633945067126495e-06, + "loss": 0.5139, + "step": 19985 + }, + { + "epoch": 7.38, + "learning_rate": 2.263357556349304e-06, + "loss": 0.54, + "step": 19986 + }, + { + "epoch": 7.38, + "learning_rate": 2.2633206059859587e-06, + "loss": 0.3712, + "step": 19987 + }, + { + "epoch": 7.38, + "learning_rate": 2.263283655622614e-06, + "loss": 0.3537, + "step": 19988 + }, + { + "epoch": 7.38, + "learning_rate": 2.2632467052592684e-06, + "loss": 0.4671, + "step": 19989 + }, + { + "epoch": 7.38, + "learning_rate": 2.2632097548959235e-06, + "loss": 0.4761, + "step": 19990 + }, + { + "epoch": 7.38, + "learning_rate": 2.2631728045325777e-06, + "loss": 0.4522, + "step": 19991 + }, + { + "epoch": 7.38, + "learning_rate": 2.2631358541692328e-06, + "loss": 0.4802, + "step": 19992 + }, + { + "epoch": 7.38, + "learning_rate": 2.2630989038058874e-06, + "loss": 0.4133, + "step": 19993 + }, + { + "epoch": 7.38, + "learning_rate": 2.2630619534425425e-06, + "loss": 0.427, + "step": 19994 + }, + { + "epoch": 7.38, + "learning_rate": 2.263025003079197e-06, + "loss": 0.526, + "step": 19995 + }, + { + "epoch": 7.38, + "learning_rate": 2.2629880527158517e-06, + "loss": 0.4115, + "step": 19996 + }, + { + "epoch": 7.38, + "learning_rate": 2.2629511023525064e-06, + "loss": 0.3376, + "step": 19997 + }, + { + "epoch": 7.38, + "learning_rate": 2.262914151989161e-06, + "loss": 0.3495, + "step": 19998 + }, + { + "epoch": 7.39, + "learning_rate": 2.262877201625816e-06, + "loss": 0.4438, + "step": 19999 + }, + { + "epoch": 7.39, + "learning_rate": 2.2628402512624707e-06, + "loss": 0.5024, + "step": 20000 + }, + { + "epoch": 7.39, + "learning_rate": 2.2628033008991258e-06, + "loss": 0.3512, + "step": 20001 + }, + { + "epoch": 7.39, + "learning_rate": 2.2627663505357804e-06, + "loss": 0.4669, + "step": 20002 + }, + { + "epoch": 7.39, + "learning_rate": 2.262729400172435e-06, + "loss": 0.3527, + "step": 20003 + }, + { + "epoch": 7.39, + "learning_rate": 2.2626924498090897e-06, + "loss": 0.4515, + "step": 20004 + }, + { + "epoch": 7.39, + "learning_rate": 2.2626554994457443e-06, + "loss": 0.4894, + "step": 20005 + }, + { + "epoch": 7.39, + "learning_rate": 2.2626185490823994e-06, + "loss": 0.4391, + "step": 20006 + }, + { + "epoch": 7.39, + "learning_rate": 2.262581598719054e-06, + "loss": 0.3509, + "step": 20007 + }, + { + "epoch": 7.39, + "learning_rate": 2.262544648355709e-06, + "loss": 0.4362, + "step": 20008 + }, + { + "epoch": 7.39, + "learning_rate": 2.2625076979923637e-06, + "loss": 0.3999, + "step": 20009 + }, + { + "epoch": 7.39, + "learning_rate": 2.2624707476290183e-06, + "loss": 0.4856, + "step": 20010 + }, + { + "epoch": 7.39, + "learning_rate": 2.262433797265673e-06, + "loss": 0.5153, + "step": 20011 + }, + { + "epoch": 7.39, + "learning_rate": 2.262396846902328e-06, + "loss": 0.467, + "step": 20012 + }, + { + "epoch": 7.39, + "learning_rate": 2.2623598965389827e-06, + "loss": 0.4201, + "step": 20013 + }, + { + "epoch": 7.39, + "learning_rate": 2.2623229461756373e-06, + "loss": 0.399, + "step": 20014 + }, + { + "epoch": 7.39, + "learning_rate": 2.2622859958122924e-06, + "loss": 0.614, + "step": 20015 + }, + { + "epoch": 7.39, + "learning_rate": 2.262249045448947e-06, + "loss": 0.3841, + "step": 20016 + }, + { + "epoch": 7.39, + "learning_rate": 2.2622120950856016e-06, + "loss": 0.3255, + "step": 20017 + }, + { + "epoch": 7.39, + "learning_rate": 2.2621751447222563e-06, + "loss": 0.4144, + "step": 20018 + }, + { + "epoch": 7.39, + "learning_rate": 2.2621381943589113e-06, + "loss": 0.4349, + "step": 20019 + }, + { + "epoch": 7.39, + "learning_rate": 2.262101243995566e-06, + "loss": 0.4802, + "step": 20020 + }, + { + "epoch": 7.39, + "learning_rate": 2.262064293632221e-06, + "loss": 0.3588, + "step": 20021 + }, + { + "epoch": 7.39, + "learning_rate": 2.2620273432688757e-06, + "loss": 0.3456, + "step": 20022 + }, + { + "epoch": 7.39, + "learning_rate": 2.2619903929055303e-06, + "loss": 0.3261, + "step": 20023 + }, + { + "epoch": 7.39, + "learning_rate": 2.261953442542185e-06, + "loss": 0.3997, + "step": 20024 + }, + { + "epoch": 7.39, + "learning_rate": 2.2619164921788396e-06, + "loss": 0.3689, + "step": 20025 + }, + { + "epoch": 7.4, + "learning_rate": 2.2618795418154946e-06, + "loss": 0.4106, + "step": 20026 + }, + { + "epoch": 7.4, + "learning_rate": 2.2618425914521493e-06, + "loss": 0.4342, + "step": 20027 + }, + { + "epoch": 7.4, + "learning_rate": 2.2618056410888043e-06, + "loss": 0.4816, + "step": 20028 + }, + { + "epoch": 7.4, + "learning_rate": 2.261768690725459e-06, + "loss": 0.3744, + "step": 20029 + }, + { + "epoch": 7.4, + "learning_rate": 2.2617317403621136e-06, + "loss": 0.3369, + "step": 20030 + }, + { + "epoch": 7.4, + "learning_rate": 2.2616947899987682e-06, + "loss": 0.4242, + "step": 20031 + }, + { + "epoch": 7.4, + "learning_rate": 2.2616578396354233e-06, + "loss": 0.4146, + "step": 20032 + }, + { + "epoch": 7.4, + "learning_rate": 2.261620889272078e-06, + "loss": 0.3781, + "step": 20033 + }, + { + "epoch": 7.4, + "learning_rate": 2.2615839389087326e-06, + "loss": 0.5186, + "step": 20034 + }, + { + "epoch": 7.4, + "learning_rate": 2.2615469885453876e-06, + "loss": 0.5692, + "step": 20035 + }, + { + "epoch": 7.4, + "learning_rate": 2.261510038182042e-06, + "loss": 0.379, + "step": 20036 + }, + { + "epoch": 7.4, + "learning_rate": 2.261473087818697e-06, + "loss": 0.3939, + "step": 20037 + }, + { + "epoch": 7.4, + "learning_rate": 2.2614361374553515e-06, + "loss": 0.5864, + "step": 20038 + }, + { + "epoch": 7.4, + "learning_rate": 2.2613991870920066e-06, + "loss": 0.4716, + "step": 20039 + }, + { + "epoch": 7.4, + "learning_rate": 2.2613622367286612e-06, + "loss": 0.2647, + "step": 20040 + }, + { + "epoch": 7.4, + "learning_rate": 2.2613252863653163e-06, + "loss": 0.4665, + "step": 20041 + }, + { + "epoch": 7.4, + "learning_rate": 2.2612883360019705e-06, + "loss": 0.3607, + "step": 20042 + }, + { + "epoch": 7.4, + "learning_rate": 2.2612513856386256e-06, + "loss": 0.6283, + "step": 20043 + }, + { + "epoch": 7.4, + "learning_rate": 2.26121443527528e-06, + "loss": 0.3784, + "step": 20044 + }, + { + "epoch": 7.4, + "learning_rate": 2.261177484911935e-06, + "loss": 0.3798, + "step": 20045 + }, + { + "epoch": 7.4, + "learning_rate": 2.26114053454859e-06, + "loss": 0.3747, + "step": 20046 + }, + { + "epoch": 7.4, + "learning_rate": 2.2611035841852445e-06, + "loss": 0.4007, + "step": 20047 + }, + { + "epoch": 7.4, + "learning_rate": 2.261066633821899e-06, + "loss": 0.3838, + "step": 20048 + }, + { + "epoch": 7.4, + "learning_rate": 2.261029683458554e-06, + "loss": 0.3693, + "step": 20049 + }, + { + "epoch": 7.4, + "learning_rate": 2.260992733095209e-06, + "loss": 0.3318, + "step": 20050 + }, + { + "epoch": 7.4, + "learning_rate": 2.2609557827318635e-06, + "loss": 0.4794, + "step": 20051 + }, + { + "epoch": 7.4, + "learning_rate": 2.2609188323685186e-06, + "loss": 0.459, + "step": 20052 + }, + { + "epoch": 7.41, + "learning_rate": 2.260881882005173e-06, + "loss": 0.3552, + "step": 20053 + }, + { + "epoch": 7.41, + "learning_rate": 2.260844931641828e-06, + "loss": 0.3511, + "step": 20054 + }, + { + "epoch": 7.41, + "learning_rate": 2.2608079812784825e-06, + "loss": 0.4011, + "step": 20055 + }, + { + "epoch": 7.41, + "learning_rate": 2.260771030915137e-06, + "loss": 0.3054, + "step": 20056 + }, + { + "epoch": 7.41, + "learning_rate": 2.260734080551792e-06, + "loss": 0.364, + "step": 20057 + }, + { + "epoch": 7.41, + "learning_rate": 2.260697130188447e-06, + "loss": 0.4524, + "step": 20058 + }, + { + "epoch": 7.41, + "learning_rate": 2.260660179825102e-06, + "loss": 0.3152, + "step": 20059 + }, + { + "epoch": 7.41, + "learning_rate": 2.2606232294617565e-06, + "loss": 0.4408, + "step": 20060 + }, + { + "epoch": 7.41, + "learning_rate": 2.260586279098411e-06, + "loss": 0.4518, + "step": 20061 + }, + { + "epoch": 7.41, + "learning_rate": 2.2605493287350658e-06, + "loss": 0.4027, + "step": 20062 + }, + { + "epoch": 7.41, + "learning_rate": 2.260512378371721e-06, + "loss": 0.431, + "step": 20063 + }, + { + "epoch": 7.41, + "learning_rate": 2.2604754280083755e-06, + "loss": 0.4342, + "step": 20064 + }, + { + "epoch": 7.41, + "learning_rate": 2.26043847764503e-06, + "loss": 0.4327, + "step": 20065 + }, + { + "epoch": 7.41, + "learning_rate": 2.260401527281685e-06, + "loss": 0.3807, + "step": 20066 + }, + { + "epoch": 7.41, + "learning_rate": 2.26036457691834e-06, + "loss": 0.3725, + "step": 20067 + }, + { + "epoch": 7.41, + "learning_rate": 2.2603276265549944e-06, + "loss": 0.5254, + "step": 20068 + }, + { + "epoch": 7.41, + "learning_rate": 2.260290676191649e-06, + "loss": 0.4187, + "step": 20069 + }, + { + "epoch": 7.41, + "learning_rate": 2.260253725828304e-06, + "loss": 0.3391, + "step": 20070 + }, + { + "epoch": 7.41, + "learning_rate": 2.2602167754649588e-06, + "loss": 0.3322, + "step": 20071 + }, + { + "epoch": 7.41, + "learning_rate": 2.260179825101614e-06, + "loss": 0.4186, + "step": 20072 + }, + { + "epoch": 7.41, + "learning_rate": 2.2601428747382685e-06, + "loss": 0.5577, + "step": 20073 + }, + { + "epoch": 7.41, + "learning_rate": 2.260105924374923e-06, + "loss": 0.4358, + "step": 20074 + }, + { + "epoch": 7.41, + "learning_rate": 2.2600689740115777e-06, + "loss": 0.3724, + "step": 20075 + }, + { + "epoch": 7.41, + "learning_rate": 2.2600320236482324e-06, + "loss": 0.4149, + "step": 20076 + }, + { + "epoch": 7.41, + "learning_rate": 2.2599950732848874e-06, + "loss": 0.3566, + "step": 20077 + }, + { + "epoch": 7.41, + "learning_rate": 2.259958122921542e-06, + "loss": 0.4943, + "step": 20078 + }, + { + "epoch": 7.41, + "learning_rate": 2.259921172558197e-06, + "loss": 0.4014, + "step": 20079 + }, + { + "epoch": 7.42, + "learning_rate": 2.2598842221948518e-06, + "loss": 0.349, + "step": 20080 + }, + { + "epoch": 7.42, + "learning_rate": 2.2598472718315064e-06, + "loss": 0.4778, + "step": 20081 + }, + { + "epoch": 7.42, + "learning_rate": 2.259810321468161e-06, + "loss": 0.3871, + "step": 20082 + }, + { + "epoch": 7.42, + "learning_rate": 2.259773371104816e-06, + "loss": 0.4401, + "step": 20083 + }, + { + "epoch": 7.42, + "learning_rate": 2.2597364207414707e-06, + "loss": 0.5469, + "step": 20084 + }, + { + "epoch": 7.42, + "learning_rate": 2.2596994703781254e-06, + "loss": 0.4848, + "step": 20085 + }, + { + "epoch": 7.42, + "learning_rate": 2.2596625200147804e-06, + "loss": 0.3329, + "step": 20086 + }, + { + "epoch": 7.42, + "learning_rate": 2.2596255696514347e-06, + "loss": 0.3695, + "step": 20087 + }, + { + "epoch": 7.42, + "learning_rate": 2.2595886192880897e-06, + "loss": 0.5103, + "step": 20088 + }, + { + "epoch": 7.42, + "learning_rate": 2.2595516689247444e-06, + "loss": 0.2669, + "step": 20089 + }, + { + "epoch": 7.42, + "learning_rate": 2.2595147185613994e-06, + "loss": 0.4125, + "step": 20090 + }, + { + "epoch": 7.42, + "learning_rate": 2.259477768198054e-06, + "loss": 0.4303, + "step": 20091 + }, + { + "epoch": 7.42, + "learning_rate": 2.259440817834709e-06, + "loss": 0.5693, + "step": 20092 + }, + { + "epoch": 7.42, + "learning_rate": 2.2594038674713633e-06, + "loss": 0.3458, + "step": 20093 + }, + { + "epoch": 7.42, + "learning_rate": 2.259366917108018e-06, + "loss": 0.4932, + "step": 20094 + }, + { + "epoch": 7.42, + "learning_rate": 2.259329966744673e-06, + "loss": 0.4388, + "step": 20095 + }, + { + "epoch": 7.42, + "learning_rate": 2.2592930163813277e-06, + "loss": 0.3946, + "step": 20096 + }, + { + "epoch": 7.42, + "learning_rate": 2.2592560660179827e-06, + "loss": 0.3297, + "step": 20097 + }, + { + "epoch": 7.42, + "learning_rate": 2.2592191156546373e-06, + "loss": 0.5735, + "step": 20098 + }, + { + "epoch": 7.42, + "learning_rate": 2.259182165291292e-06, + "loss": 0.4352, + "step": 20099 + }, + { + "epoch": 7.42, + "learning_rate": 2.2591452149279466e-06, + "loss": 0.3746, + "step": 20100 + }, + { + "epoch": 7.42, + "learning_rate": 2.2591082645646017e-06, + "loss": 0.4745, + "step": 20101 + }, + { + "epoch": 7.42, + "learning_rate": 2.2590713142012563e-06, + "loss": 0.4484, + "step": 20102 + }, + { + "epoch": 7.42, + "learning_rate": 2.2590343638379114e-06, + "loss": 0.3433, + "step": 20103 + }, + { + "epoch": 7.42, + "learning_rate": 2.258997413474566e-06, + "loss": 0.4188, + "step": 20104 + }, + { + "epoch": 7.42, + "learning_rate": 2.2589604631112207e-06, + "loss": 0.3477, + "step": 20105 + }, + { + "epoch": 7.42, + "learning_rate": 2.2589235127478753e-06, + "loss": 0.5212, + "step": 20106 + }, + { + "epoch": 7.43, + "learning_rate": 2.25888656238453e-06, + "loss": 0.5812, + "step": 20107 + }, + { + "epoch": 7.43, + "learning_rate": 2.258849612021185e-06, + "loss": 0.5314, + "step": 20108 + }, + { + "epoch": 7.43, + "learning_rate": 2.2588126616578396e-06, + "loss": 0.3678, + "step": 20109 + }, + { + "epoch": 7.43, + "learning_rate": 2.2587757112944947e-06, + "loss": 0.4392, + "step": 20110 + }, + { + "epoch": 7.43, + "learning_rate": 2.2587387609311493e-06, + "loss": 0.4247, + "step": 20111 + }, + { + "epoch": 7.43, + "learning_rate": 2.258701810567804e-06, + "loss": 0.5077, + "step": 20112 + }, + { + "epoch": 7.43, + "learning_rate": 2.2586648602044586e-06, + "loss": 0.3552, + "step": 20113 + }, + { + "epoch": 7.43, + "learning_rate": 2.2586279098411132e-06, + "loss": 0.4118, + "step": 20114 + }, + { + "epoch": 7.43, + "learning_rate": 2.2585909594777683e-06, + "loss": 0.5582, + "step": 20115 + }, + { + "epoch": 7.43, + "learning_rate": 2.258554009114423e-06, + "loss": 0.4875, + "step": 20116 + }, + { + "epoch": 7.43, + "learning_rate": 2.258517058751078e-06, + "loss": 0.5145, + "step": 20117 + }, + { + "epoch": 7.43, + "learning_rate": 2.2584801083877326e-06, + "loss": 0.5122, + "step": 20118 + }, + { + "epoch": 7.43, + "learning_rate": 2.2584431580243873e-06, + "loss": 0.5943, + "step": 20119 + }, + { + "epoch": 7.43, + "learning_rate": 2.258406207661042e-06, + "loss": 0.4804, + "step": 20120 + }, + { + "epoch": 7.43, + "learning_rate": 2.258369257297697e-06, + "loss": 0.4328, + "step": 20121 + }, + { + "epoch": 7.43, + "learning_rate": 2.2583323069343516e-06, + "loss": 0.3485, + "step": 20122 + }, + { + "epoch": 7.43, + "learning_rate": 2.2582953565710066e-06, + "loss": 0.4617, + "step": 20123 + }, + { + "epoch": 7.43, + "learning_rate": 2.2582584062076613e-06, + "loss": 0.3232, + "step": 20124 + }, + { + "epoch": 7.43, + "learning_rate": 2.258221455844316e-06, + "loss": 0.385, + "step": 20125 + }, + { + "epoch": 7.43, + "learning_rate": 2.2581845054809706e-06, + "loss": 0.426, + "step": 20126 + }, + { + "epoch": 7.43, + "learning_rate": 2.258147555117625e-06, + "loss": 0.5139, + "step": 20127 + }, + { + "epoch": 7.43, + "learning_rate": 2.2581106047542803e-06, + "loss": 0.3633, + "step": 20128 + }, + { + "epoch": 7.43, + "learning_rate": 2.258073654390935e-06, + "loss": 0.4681, + "step": 20129 + }, + { + "epoch": 7.43, + "learning_rate": 2.25803670402759e-06, + "loss": 0.3935, + "step": 20130 + }, + { + "epoch": 7.43, + "learning_rate": 2.2579997536642446e-06, + "loss": 0.3471, + "step": 20131 + }, + { + "epoch": 7.43, + "learning_rate": 2.2579628033008992e-06, + "loss": 0.3318, + "step": 20132 + }, + { + "epoch": 7.43, + "learning_rate": 2.257925852937554e-06, + "loss": 0.3985, + "step": 20133 + }, + { + "epoch": 7.44, + "learning_rate": 2.2578889025742085e-06, + "loss": 0.3732, + "step": 20134 + }, + { + "epoch": 7.44, + "learning_rate": 2.2578519522108636e-06, + "loss": 0.454, + "step": 20135 + }, + { + "epoch": 7.44, + "learning_rate": 2.257815001847518e-06, + "loss": 0.4431, + "step": 20136 + }, + { + "epoch": 7.44, + "learning_rate": 2.2577780514841733e-06, + "loss": 0.4776, + "step": 20137 + }, + { + "epoch": 7.44, + "learning_rate": 2.2577411011208275e-06, + "loss": 0.719, + "step": 20138 + }, + { + "epoch": 7.44, + "learning_rate": 2.2577041507574825e-06, + "loss": 0.4103, + "step": 20139 + }, + { + "epoch": 7.44, + "learning_rate": 2.257667200394137e-06, + "loss": 0.4815, + "step": 20140 + }, + { + "epoch": 7.44, + "learning_rate": 2.2576302500307922e-06, + "loss": 0.3145, + "step": 20141 + }, + { + "epoch": 7.44, + "learning_rate": 2.257593299667447e-06, + "loss": 0.5826, + "step": 20142 + }, + { + "epoch": 7.44, + "learning_rate": 2.257556349304102e-06, + "loss": 0.405, + "step": 20143 + }, + { + "epoch": 7.44, + "learning_rate": 2.257519398940756e-06, + "loss": 0.3629, + "step": 20144 + }, + { + "epoch": 7.44, + "learning_rate": 2.2574824485774108e-06, + "loss": 0.3915, + "step": 20145 + }, + { + "epoch": 7.44, + "learning_rate": 2.257445498214066e-06, + "loss": 0.3565, + "step": 20146 + }, + { + "epoch": 7.44, + "learning_rate": 2.2574085478507205e-06, + "loss": 0.391, + "step": 20147 + }, + { + "epoch": 7.44, + "learning_rate": 2.2573715974873755e-06, + "loss": 0.4158, + "step": 20148 + }, + { + "epoch": 7.44, + "learning_rate": 2.25733464712403e-06, + "loss": 0.3053, + "step": 20149 + }, + { + "epoch": 7.44, + "learning_rate": 2.257297696760685e-06, + "loss": 0.4586, + "step": 20150 + }, + { + "epoch": 7.44, + "learning_rate": 2.2572607463973394e-06, + "loss": 0.54, + "step": 20151 + }, + { + "epoch": 7.44, + "learning_rate": 2.2572237960339945e-06, + "loss": 0.3756, + "step": 20152 + }, + { + "epoch": 7.44, + "learning_rate": 2.257186845670649e-06, + "loss": 0.5696, + "step": 20153 + }, + { + "epoch": 7.44, + "learning_rate": 2.2571498953073038e-06, + "loss": 0.466, + "step": 20154 + }, + { + "epoch": 7.44, + "learning_rate": 2.257112944943959e-06, + "loss": 0.4339, + "step": 20155 + }, + { + "epoch": 7.44, + "learning_rate": 2.2570759945806135e-06, + "loss": 0.3361, + "step": 20156 + }, + { + "epoch": 7.44, + "learning_rate": 2.257039044217268e-06, + "loss": 0.4719, + "step": 20157 + }, + { + "epoch": 7.44, + "learning_rate": 2.2570020938539227e-06, + "loss": 0.3748, + "step": 20158 + }, + { + "epoch": 7.44, + "learning_rate": 2.256965143490578e-06, + "loss": 0.4764, + "step": 20159 + }, + { + "epoch": 7.44, + "learning_rate": 2.2569281931272324e-06, + "loss": 0.5537, + "step": 20160 + }, + { + "epoch": 7.44, + "learning_rate": 2.2568912427638875e-06, + "loss": 0.4816, + "step": 20161 + }, + { + "epoch": 7.45, + "learning_rate": 2.256854292400542e-06, + "loss": 0.4167, + "step": 20162 + }, + { + "epoch": 7.45, + "learning_rate": 2.2568173420371968e-06, + "loss": 0.4949, + "step": 20163 + }, + { + "epoch": 7.45, + "learning_rate": 2.2567803916738514e-06, + "loss": 0.4151, + "step": 20164 + }, + { + "epoch": 7.45, + "learning_rate": 2.256743441310506e-06, + "loss": 0.3897, + "step": 20165 + }, + { + "epoch": 7.45, + "learning_rate": 2.256706490947161e-06, + "loss": 0.3266, + "step": 20166 + }, + { + "epoch": 7.45, + "learning_rate": 2.2566695405838157e-06, + "loss": 0.4099, + "step": 20167 + }, + { + "epoch": 7.45, + "learning_rate": 2.256632590220471e-06, + "loss": 0.403, + "step": 20168 + }, + { + "epoch": 7.45, + "learning_rate": 2.2565956398571254e-06, + "loss": 0.4812, + "step": 20169 + }, + { + "epoch": 7.45, + "learning_rate": 2.25655868949378e-06, + "loss": 0.4484, + "step": 20170 + }, + { + "epoch": 7.45, + "learning_rate": 2.2565217391304347e-06, + "loss": 0.3909, + "step": 20171 + }, + { + "epoch": 7.45, + "learning_rate": 2.2564847887670898e-06, + "loss": 0.5653, + "step": 20172 + }, + { + "epoch": 7.45, + "learning_rate": 2.2564478384037444e-06, + "loss": 0.5089, + "step": 20173 + }, + { + "epoch": 7.45, + "learning_rate": 2.256410888040399e-06, + "loss": 0.4785, + "step": 20174 + }, + { + "epoch": 7.45, + "learning_rate": 2.256373937677054e-06, + "loss": 0.3792, + "step": 20175 + }, + { + "epoch": 7.45, + "learning_rate": 2.2563369873137087e-06, + "loss": 0.583, + "step": 20176 + }, + { + "epoch": 7.45, + "learning_rate": 2.2563000369503634e-06, + "loss": 0.3896, + "step": 20177 + }, + { + "epoch": 7.45, + "learning_rate": 2.256263086587018e-06, + "loss": 0.5163, + "step": 20178 + }, + { + "epoch": 7.45, + "learning_rate": 2.256226136223673e-06, + "loss": 0.4537, + "step": 20179 + }, + { + "epoch": 7.45, + "learning_rate": 2.2561891858603277e-06, + "loss": 0.5231, + "step": 20180 + }, + { + "epoch": 7.45, + "learning_rate": 2.2561522354969828e-06, + "loss": 0.4427, + "step": 20181 + }, + { + "epoch": 7.45, + "learning_rate": 2.256115285133637e-06, + "loss": 0.3567, + "step": 20182 + }, + { + "epoch": 7.45, + "learning_rate": 2.256078334770292e-06, + "loss": 0.4278, + "step": 20183 + }, + { + "epoch": 7.45, + "learning_rate": 2.2560413844069467e-06, + "loss": 0.4988, + "step": 20184 + }, + { + "epoch": 7.45, + "learning_rate": 2.2560044340436013e-06, + "loss": 0.4627, + "step": 20185 + }, + { + "epoch": 7.45, + "learning_rate": 2.2559674836802564e-06, + "loss": 0.3437, + "step": 20186 + }, + { + "epoch": 7.45, + "learning_rate": 2.255930533316911e-06, + "loss": 0.3562, + "step": 20187 + }, + { + "epoch": 7.45, + "learning_rate": 2.2558935829535656e-06, + "loss": 0.4065, + "step": 20188 + }, + { + "epoch": 7.46, + "learning_rate": 2.2558566325902203e-06, + "loss": 0.4026, + "step": 20189 + }, + { + "epoch": 7.46, + "learning_rate": 2.2558196822268753e-06, + "loss": 0.4927, + "step": 20190 + }, + { + "epoch": 7.46, + "learning_rate": 2.25578273186353e-06, + "loss": 0.4822, + "step": 20191 + }, + { + "epoch": 7.46, + "learning_rate": 2.255745781500185e-06, + "loss": 0.3736, + "step": 20192 + }, + { + "epoch": 7.46, + "learning_rate": 2.2557088311368397e-06, + "loss": 0.3654, + "step": 20193 + }, + { + "epoch": 7.46, + "learning_rate": 2.2556718807734943e-06, + "loss": 0.5978, + "step": 20194 + }, + { + "epoch": 7.46, + "learning_rate": 2.255634930410149e-06, + "loss": 0.4023, + "step": 20195 + }, + { + "epoch": 7.46, + "learning_rate": 2.2555979800468036e-06, + "loss": 0.4866, + "step": 20196 + }, + { + "epoch": 7.46, + "learning_rate": 2.2555610296834586e-06, + "loss": 0.5037, + "step": 20197 + }, + { + "epoch": 7.46, + "learning_rate": 2.2555240793201133e-06, + "loss": 0.4348, + "step": 20198 + }, + { + "epoch": 7.46, + "learning_rate": 2.2554871289567683e-06, + "loss": 0.3568, + "step": 20199 + }, + { + "epoch": 7.46, + "learning_rate": 2.255450178593423e-06, + "loss": 0.4343, + "step": 20200 + }, + { + "epoch": 7.46, + "learning_rate": 2.2554132282300776e-06, + "loss": 0.4748, + "step": 20201 + }, + { + "epoch": 7.46, + "learning_rate": 2.2553762778667322e-06, + "loss": 0.3599, + "step": 20202 + }, + { + "epoch": 7.46, + "learning_rate": 2.2553393275033873e-06, + "loss": 0.3904, + "step": 20203 + }, + { + "epoch": 7.46, + "learning_rate": 2.255302377140042e-06, + "loss": 0.3737, + "step": 20204 + }, + { + "epoch": 7.46, + "learning_rate": 2.2552654267766966e-06, + "loss": 0.4463, + "step": 20205 + }, + { + "epoch": 7.46, + "learning_rate": 2.2552284764133516e-06, + "loss": 0.238, + "step": 20206 + }, + { + "epoch": 7.46, + "learning_rate": 2.2551915260500063e-06, + "loss": 0.3891, + "step": 20207 + }, + { + "epoch": 7.46, + "learning_rate": 2.255154575686661e-06, + "loss": 0.4358, + "step": 20208 + }, + { + "epoch": 7.46, + "learning_rate": 2.2551176253233155e-06, + "loss": 0.3329, + "step": 20209 + }, + { + "epoch": 7.46, + "learning_rate": 2.2550806749599706e-06, + "loss": 0.4409, + "step": 20210 + }, + { + "epoch": 7.46, + "learning_rate": 2.2550437245966252e-06, + "loss": 0.3019, + "step": 20211 + }, + { + "epoch": 7.46, + "learning_rate": 2.2550067742332803e-06, + "loss": 0.4573, + "step": 20212 + }, + { + "epoch": 7.46, + "learning_rate": 2.254969823869935e-06, + "loss": 0.4007, + "step": 20213 + }, + { + "epoch": 7.46, + "learning_rate": 2.2549328735065896e-06, + "loss": 0.3705, + "step": 20214 + }, + { + "epoch": 7.46, + "learning_rate": 2.2548959231432442e-06, + "loss": 0.3804, + "step": 20215 + }, + { + "epoch": 7.47, + "learning_rate": 2.254858972779899e-06, + "loss": 0.3119, + "step": 20216 + }, + { + "epoch": 7.47, + "learning_rate": 2.254822022416554e-06, + "loss": 0.2739, + "step": 20217 + }, + { + "epoch": 7.47, + "learning_rate": 2.2547850720532085e-06, + "loss": 0.3554, + "step": 20218 + }, + { + "epoch": 7.47, + "learning_rate": 2.2547481216898636e-06, + "loss": 0.5442, + "step": 20219 + }, + { + "epoch": 7.47, + "learning_rate": 2.2547111713265182e-06, + "loss": 0.3269, + "step": 20220 + }, + { + "epoch": 7.47, + "learning_rate": 2.254674220963173e-06, + "loss": 0.4269, + "step": 20221 + }, + { + "epoch": 7.47, + "learning_rate": 2.2546372705998275e-06, + "loss": 0.4195, + "step": 20222 + }, + { + "epoch": 7.47, + "learning_rate": 2.2546003202364826e-06, + "loss": 0.3652, + "step": 20223 + }, + { + "epoch": 7.47, + "learning_rate": 2.254563369873137e-06, + "loss": 0.4458, + "step": 20224 + }, + { + "epoch": 7.47, + "learning_rate": 2.254526419509792e-06, + "loss": 0.3852, + "step": 20225 + }, + { + "epoch": 7.47, + "learning_rate": 2.254489469146447e-06, + "loss": 0.3619, + "step": 20226 + }, + { + "epoch": 7.47, + "learning_rate": 2.254452518783101e-06, + "loss": 0.4194, + "step": 20227 + }, + { + "epoch": 7.47, + "learning_rate": 2.254415568419756e-06, + "loss": 0.3684, + "step": 20228 + }, + { + "epoch": 7.47, + "learning_rate": 2.254378618056411e-06, + "loss": 0.3806, + "step": 20229 + }, + { + "epoch": 7.47, + "learning_rate": 2.254341667693066e-06, + "loss": 0.5093, + "step": 20230 + }, + { + "epoch": 7.47, + "learning_rate": 2.2543047173297205e-06, + "loss": 0.4987, + "step": 20231 + }, + { + "epoch": 7.47, + "learning_rate": 2.2542677669663756e-06, + "loss": 0.3265, + "step": 20232 + }, + { + "epoch": 7.47, + "learning_rate": 2.2542308166030298e-06, + "loss": 0.5442, + "step": 20233 + }, + { + "epoch": 7.47, + "learning_rate": 2.2541938662396844e-06, + "loss": 0.414, + "step": 20234 + }, + { + "epoch": 7.47, + "learning_rate": 2.2541569158763395e-06, + "loss": 0.5086, + "step": 20235 + }, + { + "epoch": 7.47, + "learning_rate": 2.254119965512994e-06, + "loss": 0.5946, + "step": 20236 + }, + { + "epoch": 7.47, + "learning_rate": 2.254083015149649e-06, + "loss": 0.4257, + "step": 20237 + }, + { + "epoch": 7.47, + "learning_rate": 2.254046064786304e-06, + "loss": 0.3951, + "step": 20238 + }, + { + "epoch": 7.47, + "learning_rate": 2.2540091144229585e-06, + "loss": 0.339, + "step": 20239 + }, + { + "epoch": 7.47, + "learning_rate": 2.253972164059613e-06, + "loss": 0.4411, + "step": 20240 + }, + { + "epoch": 7.47, + "learning_rate": 2.253935213696268e-06, + "loss": 0.4756, + "step": 20241 + }, + { + "epoch": 7.47, + "learning_rate": 2.2538982633329228e-06, + "loss": 0.3313, + "step": 20242 + }, + { + "epoch": 7.48, + "learning_rate": 2.2538613129695774e-06, + "loss": 0.4748, + "step": 20243 + }, + { + "epoch": 7.48, + "learning_rate": 2.2538243626062325e-06, + "loss": 0.4116, + "step": 20244 + }, + { + "epoch": 7.48, + "learning_rate": 2.253787412242887e-06, + "loss": 0.4277, + "step": 20245 + }, + { + "epoch": 7.48, + "learning_rate": 2.2537504618795418e-06, + "loss": 0.4728, + "step": 20246 + }, + { + "epoch": 7.48, + "learning_rate": 2.2537135115161964e-06, + "loss": 0.4951, + "step": 20247 + }, + { + "epoch": 7.48, + "learning_rate": 2.2536765611528514e-06, + "loss": 0.3854, + "step": 20248 + }, + { + "epoch": 7.48, + "learning_rate": 2.253639610789506e-06, + "loss": 0.3239, + "step": 20249 + }, + { + "epoch": 7.48, + "learning_rate": 2.253602660426161e-06, + "loss": 0.3207, + "step": 20250 + }, + { + "epoch": 7.48, + "learning_rate": 2.2535657100628158e-06, + "loss": 0.4316, + "step": 20251 + }, + { + "epoch": 7.48, + "learning_rate": 2.2535287596994704e-06, + "loss": 0.3923, + "step": 20252 + }, + { + "epoch": 7.48, + "learning_rate": 2.253491809336125e-06, + "loss": 0.3365, + "step": 20253 + }, + { + "epoch": 7.48, + "learning_rate": 2.2534548589727797e-06, + "loss": 0.3688, + "step": 20254 + }, + { + "epoch": 7.48, + "learning_rate": 2.2534179086094348e-06, + "loss": 0.4472, + "step": 20255 + }, + { + "epoch": 7.48, + "learning_rate": 2.2533809582460894e-06, + "loss": 0.4076, + "step": 20256 + }, + { + "epoch": 7.48, + "learning_rate": 2.2533440078827444e-06, + "loss": 0.3554, + "step": 20257 + }, + { + "epoch": 7.48, + "learning_rate": 2.253307057519399e-06, + "loss": 0.3921, + "step": 20258 + }, + { + "epoch": 7.48, + "learning_rate": 2.2532701071560537e-06, + "loss": 0.4316, + "step": 20259 + }, + { + "epoch": 7.48, + "learning_rate": 2.2532331567927084e-06, + "loss": 0.4521, + "step": 20260 + }, + { + "epoch": 7.48, + "learning_rate": 2.2531962064293634e-06, + "loss": 0.3613, + "step": 20261 + }, + { + "epoch": 7.48, + "learning_rate": 2.253159256066018e-06, + "loss": 0.3757, + "step": 20262 + }, + { + "epoch": 7.48, + "learning_rate": 2.2531223057026727e-06, + "loss": 0.3946, + "step": 20263 + }, + { + "epoch": 7.48, + "learning_rate": 2.2530853553393278e-06, + "loss": 0.4896, + "step": 20264 + }, + { + "epoch": 7.48, + "learning_rate": 2.2530484049759824e-06, + "loss": 0.321, + "step": 20265 + }, + { + "epoch": 7.48, + "learning_rate": 2.253011454612637e-06, + "loss": 0.4619, + "step": 20266 + }, + { + "epoch": 7.48, + "learning_rate": 2.2529745042492917e-06, + "loss": 0.4971, + "step": 20267 + }, + { + "epoch": 7.48, + "learning_rate": 2.2529375538859467e-06, + "loss": 0.3286, + "step": 20268 + }, + { + "epoch": 7.48, + "learning_rate": 2.2529006035226014e-06, + "loss": 0.4134, + "step": 20269 + }, + { + "epoch": 7.49, + "learning_rate": 2.2528636531592564e-06, + "loss": 0.4865, + "step": 20270 + }, + { + "epoch": 7.49, + "learning_rate": 2.252826702795911e-06, + "loss": 0.4189, + "step": 20271 + }, + { + "epoch": 7.49, + "learning_rate": 2.2527897524325657e-06, + "loss": 0.5608, + "step": 20272 + }, + { + "epoch": 7.49, + "learning_rate": 2.2527528020692203e-06, + "loss": 0.4206, + "step": 20273 + }, + { + "epoch": 7.49, + "learning_rate": 2.252715851705875e-06, + "loss": 0.3858, + "step": 20274 + }, + { + "epoch": 7.49, + "learning_rate": 2.25267890134253e-06, + "loss": 0.4243, + "step": 20275 + }, + { + "epoch": 7.49, + "learning_rate": 2.2526419509791847e-06, + "loss": 0.4293, + "step": 20276 + }, + { + "epoch": 7.49, + "learning_rate": 2.2526050006158397e-06, + "loss": 0.4241, + "step": 20277 + }, + { + "epoch": 7.49, + "learning_rate": 2.252568050252494e-06, + "loss": 0.4759, + "step": 20278 + }, + { + "epoch": 7.49, + "learning_rate": 2.252531099889149e-06, + "loss": 0.4671, + "step": 20279 + }, + { + "epoch": 7.49, + "learning_rate": 2.2524941495258036e-06, + "loss": 0.4163, + "step": 20280 + }, + { + "epoch": 7.49, + "learning_rate": 2.2524571991624587e-06, + "loss": 0.3913, + "step": 20281 + }, + { + "epoch": 7.49, + "learning_rate": 2.2524202487991133e-06, + "loss": 0.4181, + "step": 20282 + }, + { + "epoch": 7.49, + "learning_rate": 2.252383298435768e-06, + "loss": 0.3622, + "step": 20283 + }, + { + "epoch": 7.49, + "learning_rate": 2.2523463480724226e-06, + "loss": 0.4487, + "step": 20284 + }, + { + "epoch": 7.49, + "learning_rate": 2.2523093977090772e-06, + "loss": 0.3307, + "step": 20285 + }, + { + "epoch": 7.49, + "learning_rate": 2.2522724473457323e-06, + "loss": 0.4588, + "step": 20286 + }, + { + "epoch": 7.49, + "learning_rate": 2.252235496982387e-06, + "loss": 0.4127, + "step": 20287 + }, + { + "epoch": 7.49, + "learning_rate": 2.252198546619042e-06, + "loss": 0.3082, + "step": 20288 + }, + { + "epoch": 7.49, + "learning_rate": 2.2521615962556966e-06, + "loss": 0.4615, + "step": 20289 + }, + { + "epoch": 7.49, + "learning_rate": 2.2521246458923513e-06, + "loss": 0.4715, + "step": 20290 + }, + { + "epoch": 7.49, + "learning_rate": 2.252087695529006e-06, + "loss": 0.4887, + "step": 20291 + }, + { + "epoch": 7.49, + "learning_rate": 2.252050745165661e-06, + "loss": 0.5104, + "step": 20292 + }, + { + "epoch": 7.49, + "learning_rate": 2.2520137948023156e-06, + "loss": 0.4712, + "step": 20293 + }, + { + "epoch": 7.49, + "learning_rate": 2.2519768444389702e-06, + "loss": 0.4715, + "step": 20294 + }, + { + "epoch": 7.49, + "learning_rate": 2.2519398940756253e-06, + "loss": 0.3682, + "step": 20295 + }, + { + "epoch": 7.49, + "learning_rate": 2.25190294371228e-06, + "loss": 0.2713, + "step": 20296 + }, + { + "epoch": 7.5, + "learning_rate": 2.2518659933489346e-06, + "loss": 0.4552, + "step": 20297 + }, + { + "epoch": 7.5, + "learning_rate": 2.251829042985589e-06, + "loss": 0.427, + "step": 20298 + }, + { + "epoch": 7.5, + "learning_rate": 2.2517920926222443e-06, + "loss": 0.3687, + "step": 20299 + }, + { + "epoch": 7.5, + "learning_rate": 2.251755142258899e-06, + "loss": 0.4447, + "step": 20300 + }, + { + "epoch": 7.5, + "learning_rate": 2.251718191895554e-06, + "loss": 0.4396, + "step": 20301 + }, + { + "epoch": 7.5, + "learning_rate": 2.2516812415322086e-06, + "loss": 0.4276, + "step": 20302 + }, + { + "epoch": 7.5, + "learning_rate": 2.2516442911688632e-06, + "loss": 0.5058, + "step": 20303 + }, + { + "epoch": 7.5, + "learning_rate": 2.251607340805518e-06, + "loss": 0.3563, + "step": 20304 + }, + { + "epoch": 7.5, + "learning_rate": 2.2515703904421725e-06, + "loss": 0.356, + "step": 20305 + }, + { + "epoch": 7.5, + "learning_rate": 2.2515334400788276e-06, + "loss": 0.3922, + "step": 20306 + }, + { + "epoch": 7.5, + "learning_rate": 2.251496489715482e-06, + "loss": 0.3909, + "step": 20307 + }, + { + "epoch": 7.5, + "learning_rate": 2.2514595393521373e-06, + "loss": 0.3452, + "step": 20308 + }, + { + "epoch": 7.5, + "learning_rate": 2.251422588988792e-06, + "loss": 0.421, + "step": 20309 + }, + { + "epoch": 7.5, + "learning_rate": 2.2513856386254465e-06, + "loss": 0.3357, + "step": 20310 + }, + { + "epoch": 7.5, + "learning_rate": 2.251348688262101e-06, + "loss": 0.4642, + "step": 20311 + }, + { + "epoch": 7.5, + "learning_rate": 2.2513117378987562e-06, + "loss": 0.484, + "step": 20312 + }, + { + "epoch": 7.5, + "learning_rate": 2.251274787535411e-06, + "loss": 0.3545, + "step": 20313 + }, + { + "epoch": 7.5, + "learning_rate": 2.2512378371720655e-06, + "loss": 0.555, + "step": 20314 + }, + { + "epoch": 7.5, + "learning_rate": 2.2512008868087206e-06, + "loss": 0.4127, + "step": 20315 + }, + { + "epoch": 7.5, + "learning_rate": 2.251163936445375e-06, + "loss": 0.4168, + "step": 20316 + }, + { + "epoch": 7.5, + "learning_rate": 2.25112698608203e-06, + "loss": 0.3825, + "step": 20317 + }, + { + "epoch": 7.5, + "learning_rate": 2.2510900357186845e-06, + "loss": 0.3604, + "step": 20318 + }, + { + "epoch": 7.5, + "learning_rate": 2.2510530853553395e-06, + "loss": 0.4915, + "step": 20319 + }, + { + "epoch": 7.5, + "learning_rate": 2.251016134991994e-06, + "loss": 0.4469, + "step": 20320 + }, + { + "epoch": 7.5, + "learning_rate": 2.2509791846286492e-06, + "loss": 0.3406, + "step": 20321 + }, + { + "epoch": 7.5, + "learning_rate": 2.250942234265304e-06, + "loss": 0.42, + "step": 20322 + }, + { + "epoch": 7.5, + "learning_rate": 2.250905283901958e-06, + "loss": 0.4441, + "step": 20323 + }, + { + "epoch": 7.51, + "learning_rate": 2.250868333538613e-06, + "loss": 0.4836, + "step": 20324 + }, + { + "epoch": 7.51, + "learning_rate": 2.2508313831752678e-06, + "loss": 0.382, + "step": 20325 + }, + { + "epoch": 7.51, + "learning_rate": 2.250794432811923e-06, + "loss": 0.4812, + "step": 20326 + }, + { + "epoch": 7.51, + "learning_rate": 2.2507574824485775e-06, + "loss": 0.3668, + "step": 20327 + }, + { + "epoch": 7.51, + "learning_rate": 2.2507205320852325e-06, + "loss": 0.5153, + "step": 20328 + }, + { + "epoch": 7.51, + "learning_rate": 2.2506835817218867e-06, + "loss": 0.3547, + "step": 20329 + }, + { + "epoch": 7.51, + "learning_rate": 2.250646631358542e-06, + "loss": 0.4957, + "step": 20330 + }, + { + "epoch": 7.51, + "learning_rate": 2.2506096809951964e-06, + "loss": 0.4092, + "step": 20331 + }, + { + "epoch": 7.51, + "learning_rate": 2.2505727306318515e-06, + "loss": 0.4209, + "step": 20332 + }, + { + "epoch": 7.51, + "learning_rate": 2.250535780268506e-06, + "loss": 0.4874, + "step": 20333 + }, + { + "epoch": 7.51, + "learning_rate": 2.2504988299051608e-06, + "loss": 0.513, + "step": 20334 + }, + { + "epoch": 7.51, + "learning_rate": 2.2504618795418154e-06, + "loss": 0.3221, + "step": 20335 + }, + { + "epoch": 7.51, + "learning_rate": 2.25042492917847e-06, + "loss": 0.463, + "step": 20336 + }, + { + "epoch": 7.51, + "learning_rate": 2.250387978815125e-06, + "loss": 0.466, + "step": 20337 + }, + { + "epoch": 7.51, + "learning_rate": 2.2503510284517797e-06, + "loss": 0.3741, + "step": 20338 + }, + { + "epoch": 7.51, + "learning_rate": 2.250314078088435e-06, + "loss": 0.4616, + "step": 20339 + }, + { + "epoch": 7.51, + "learning_rate": 2.2502771277250894e-06, + "loss": 0.4256, + "step": 20340 + }, + { + "epoch": 7.51, + "learning_rate": 2.250240177361744e-06, + "loss": 0.3921, + "step": 20341 + }, + { + "epoch": 7.51, + "learning_rate": 2.2502032269983987e-06, + "loss": 0.4523, + "step": 20342 + }, + { + "epoch": 7.51, + "learning_rate": 2.2501662766350533e-06, + "loss": 0.4034, + "step": 20343 + }, + { + "epoch": 7.51, + "learning_rate": 2.2501293262717084e-06, + "loss": 0.3441, + "step": 20344 + }, + { + "epoch": 7.51, + "learning_rate": 2.250092375908363e-06, + "loss": 0.5295, + "step": 20345 + }, + { + "epoch": 7.51, + "learning_rate": 2.250055425545018e-06, + "loss": 0.3952, + "step": 20346 + }, + { + "epoch": 7.51, + "learning_rate": 2.2500184751816727e-06, + "loss": 0.3266, + "step": 20347 + }, + { + "epoch": 7.51, + "learning_rate": 2.2499815248183274e-06, + "loss": 0.4375, + "step": 20348 + }, + { + "epoch": 7.51, + "learning_rate": 2.249944574454982e-06, + "loss": 0.4242, + "step": 20349 + }, + { + "epoch": 7.51, + "learning_rate": 2.249907624091637e-06, + "loss": 0.4873, + "step": 20350 + }, + { + "epoch": 7.52, + "learning_rate": 2.2498706737282917e-06, + "loss": 0.4575, + "step": 20351 + }, + { + "epoch": 7.52, + "learning_rate": 2.2498337233649468e-06, + "loss": 0.5204, + "step": 20352 + }, + { + "epoch": 7.52, + "learning_rate": 2.2497967730016014e-06, + "loss": 0.3855, + "step": 20353 + }, + { + "epoch": 7.52, + "learning_rate": 2.249759822638256e-06, + "loss": 0.4241, + "step": 20354 + }, + { + "epoch": 7.52, + "learning_rate": 2.2497228722749107e-06, + "loss": 0.3618, + "step": 20355 + }, + { + "epoch": 7.52, + "learning_rate": 2.2496859219115653e-06, + "loss": 0.4107, + "step": 20356 + }, + { + "epoch": 7.52, + "learning_rate": 2.2496489715482204e-06, + "loss": 0.3881, + "step": 20357 + }, + { + "epoch": 7.52, + "learning_rate": 2.249612021184875e-06, + "loss": 0.3797, + "step": 20358 + }, + { + "epoch": 7.52, + "learning_rate": 2.24957507082153e-06, + "loss": 0.3797, + "step": 20359 + }, + { + "epoch": 7.52, + "learning_rate": 2.2495381204581847e-06, + "loss": 0.5183, + "step": 20360 + }, + { + "epoch": 7.52, + "learning_rate": 2.2495011700948393e-06, + "loss": 0.3535, + "step": 20361 + }, + { + "epoch": 7.52, + "learning_rate": 2.249464219731494e-06, + "loss": 0.2423, + "step": 20362 + }, + { + "epoch": 7.52, + "learning_rate": 2.2494272693681486e-06, + "loss": 0.4759, + "step": 20363 + }, + { + "epoch": 7.52, + "learning_rate": 2.2493903190048037e-06, + "loss": 0.3275, + "step": 20364 + }, + { + "epoch": 7.52, + "learning_rate": 2.2493533686414583e-06, + "loss": 0.4229, + "step": 20365 + }, + { + "epoch": 7.52, + "learning_rate": 2.2493164182781134e-06, + "loss": 0.375, + "step": 20366 + }, + { + "epoch": 7.52, + "learning_rate": 2.2492794679147676e-06, + "loss": 0.4574, + "step": 20367 + }, + { + "epoch": 7.52, + "learning_rate": 2.2492425175514226e-06, + "loss": 0.4836, + "step": 20368 + }, + { + "epoch": 7.52, + "learning_rate": 2.2492055671880773e-06, + "loss": 0.3922, + "step": 20369 + }, + { + "epoch": 7.52, + "learning_rate": 2.2491686168247323e-06, + "loss": 0.4801, + "step": 20370 + }, + { + "epoch": 7.52, + "learning_rate": 2.249131666461387e-06, + "loss": 0.3696, + "step": 20371 + }, + { + "epoch": 7.52, + "learning_rate": 2.249094716098042e-06, + "loss": 0.3465, + "step": 20372 + }, + { + "epoch": 7.52, + "learning_rate": 2.2490577657346963e-06, + "loss": 0.4562, + "step": 20373 + }, + { + "epoch": 7.52, + "learning_rate": 2.249020815371351e-06, + "loss": 0.4048, + "step": 20374 + }, + { + "epoch": 7.52, + "learning_rate": 2.248983865008006e-06, + "loss": 0.2721, + "step": 20375 + }, + { + "epoch": 7.52, + "learning_rate": 2.2489469146446606e-06, + "loss": 0.4358, + "step": 20376 + }, + { + "epoch": 7.52, + "learning_rate": 2.2489099642813156e-06, + "loss": 0.4256, + "step": 20377 + }, + { + "epoch": 7.53, + "learning_rate": 2.2488730139179703e-06, + "loss": 0.4599, + "step": 20378 + }, + { + "epoch": 7.53, + "learning_rate": 2.248836063554625e-06, + "loss": 0.4369, + "step": 20379 + }, + { + "epoch": 7.53, + "learning_rate": 2.2487991131912796e-06, + "loss": 0.4412, + "step": 20380 + }, + { + "epoch": 7.53, + "learning_rate": 2.2487621628279346e-06, + "loss": 0.4346, + "step": 20381 + }, + { + "epoch": 7.53, + "learning_rate": 2.2487252124645892e-06, + "loss": 0.3918, + "step": 20382 + }, + { + "epoch": 7.53, + "learning_rate": 2.248688262101244e-06, + "loss": 0.4191, + "step": 20383 + }, + { + "epoch": 7.53, + "learning_rate": 2.248651311737899e-06, + "loss": 0.3106, + "step": 20384 + }, + { + "epoch": 7.53, + "learning_rate": 2.2486143613745536e-06, + "loss": 0.4715, + "step": 20385 + }, + { + "epoch": 7.53, + "learning_rate": 2.2485774110112082e-06, + "loss": 0.3816, + "step": 20386 + }, + { + "epoch": 7.53, + "learning_rate": 2.248540460647863e-06, + "loss": 0.2967, + "step": 20387 + }, + { + "epoch": 7.53, + "learning_rate": 2.248503510284518e-06, + "loss": 0.4902, + "step": 20388 + }, + { + "epoch": 7.53, + "learning_rate": 2.2484665599211726e-06, + "loss": 0.462, + "step": 20389 + }, + { + "epoch": 7.53, + "learning_rate": 2.2484296095578276e-06, + "loss": 0.4622, + "step": 20390 + }, + { + "epoch": 7.53, + "learning_rate": 2.2483926591944822e-06, + "loss": 0.4173, + "step": 20391 + }, + { + "epoch": 7.53, + "learning_rate": 2.248355708831137e-06, + "loss": 0.4575, + "step": 20392 + }, + { + "epoch": 7.53, + "learning_rate": 2.2483187584677915e-06, + "loss": 0.4987, + "step": 20393 + }, + { + "epoch": 7.53, + "learning_rate": 2.248281808104446e-06, + "loss": 0.3451, + "step": 20394 + }, + { + "epoch": 7.53, + "learning_rate": 2.2482448577411012e-06, + "loss": 0.3216, + "step": 20395 + }, + { + "epoch": 7.53, + "learning_rate": 2.248207907377756e-06, + "loss": 0.4441, + "step": 20396 + }, + { + "epoch": 7.53, + "learning_rate": 2.248170957014411e-06, + "loss": 0.476, + "step": 20397 + }, + { + "epoch": 7.53, + "learning_rate": 2.2481340066510655e-06, + "loss": 0.5064, + "step": 20398 + }, + { + "epoch": 7.53, + "learning_rate": 2.24809705628772e-06, + "loss": 0.3424, + "step": 20399 + }, + { + "epoch": 7.53, + "learning_rate": 2.248060105924375e-06, + "loss": 0.4048, + "step": 20400 + }, + { + "epoch": 7.53, + "learning_rate": 2.24802315556103e-06, + "loss": 0.441, + "step": 20401 + }, + { + "epoch": 7.53, + "learning_rate": 2.2479862051976845e-06, + "loss": 0.3142, + "step": 20402 + }, + { + "epoch": 7.53, + "learning_rate": 2.247949254834339e-06, + "loss": 0.3387, + "step": 20403 + }, + { + "epoch": 7.53, + "learning_rate": 2.2479123044709942e-06, + "loss": 0.4902, + "step": 20404 + }, + { + "epoch": 7.54, + "learning_rate": 2.247875354107649e-06, + "loss": 0.3809, + "step": 20405 + }, + { + "epoch": 7.54, + "learning_rate": 2.2478384037443035e-06, + "loss": 0.4676, + "step": 20406 + }, + { + "epoch": 7.54, + "learning_rate": 2.247801453380958e-06, + "loss": 0.4222, + "step": 20407 + }, + { + "epoch": 7.54, + "learning_rate": 2.247764503017613e-06, + "loss": 0.5145, + "step": 20408 + }, + { + "epoch": 7.54, + "learning_rate": 2.247727552654268e-06, + "loss": 0.4294, + "step": 20409 + }, + { + "epoch": 7.54, + "learning_rate": 2.247690602290923e-06, + "loss": 0.3963, + "step": 20410 + }, + { + "epoch": 7.54, + "learning_rate": 2.2476536519275775e-06, + "loss": 0.4026, + "step": 20411 + }, + { + "epoch": 7.54, + "learning_rate": 2.247616701564232e-06, + "loss": 0.4127, + "step": 20412 + }, + { + "epoch": 7.54, + "learning_rate": 2.247579751200887e-06, + "loss": 0.374, + "step": 20413 + }, + { + "epoch": 7.54, + "learning_rate": 2.2475428008375414e-06, + "loss": 0.5005, + "step": 20414 + }, + { + "epoch": 7.54, + "learning_rate": 2.2475058504741965e-06, + "loss": 0.3133, + "step": 20415 + }, + { + "epoch": 7.54, + "learning_rate": 2.247468900110851e-06, + "loss": 0.3593, + "step": 20416 + }, + { + "epoch": 7.54, + "learning_rate": 2.247431949747506e-06, + "loss": 0.3804, + "step": 20417 + }, + { + "epoch": 7.54, + "learning_rate": 2.2473949993841604e-06, + "loss": 0.4056, + "step": 20418 + }, + { + "epoch": 7.54, + "learning_rate": 2.2473580490208155e-06, + "loss": 0.3774, + "step": 20419 + }, + { + "epoch": 7.54, + "learning_rate": 2.24732109865747e-06, + "loss": 0.3959, + "step": 20420 + }, + { + "epoch": 7.54, + "learning_rate": 2.247284148294125e-06, + "loss": 0.5251, + "step": 20421 + }, + { + "epoch": 7.54, + "learning_rate": 2.24724719793078e-06, + "loss": 0.4265, + "step": 20422 + }, + { + "epoch": 7.54, + "learning_rate": 2.2472102475674344e-06, + "loss": 0.3336, + "step": 20423 + }, + { + "epoch": 7.54, + "learning_rate": 2.247173297204089e-06, + "loss": 0.4834, + "step": 20424 + }, + { + "epoch": 7.54, + "learning_rate": 2.2471363468407437e-06, + "loss": 0.4422, + "step": 20425 + }, + { + "epoch": 7.54, + "learning_rate": 2.2470993964773988e-06, + "loss": 0.3779, + "step": 20426 + }, + { + "epoch": 7.54, + "learning_rate": 2.2470624461140534e-06, + "loss": 0.4716, + "step": 20427 + }, + { + "epoch": 7.54, + "learning_rate": 2.2470254957507085e-06, + "loss": 0.3326, + "step": 20428 + }, + { + "epoch": 7.54, + "learning_rate": 2.246988545387363e-06, + "loss": 0.3669, + "step": 20429 + }, + { + "epoch": 7.54, + "learning_rate": 2.2469515950240177e-06, + "loss": 0.3479, + "step": 20430 + }, + { + "epoch": 7.54, + "learning_rate": 2.2469146446606724e-06, + "loss": 0.4214, + "step": 20431 + }, + { + "epoch": 7.55, + "learning_rate": 2.2468776942973274e-06, + "loss": 0.5093, + "step": 20432 + }, + { + "epoch": 7.55, + "learning_rate": 2.246840743933982e-06, + "loss": 0.4832, + "step": 20433 + }, + { + "epoch": 7.55, + "learning_rate": 2.2468037935706367e-06, + "loss": 0.5717, + "step": 20434 + }, + { + "epoch": 7.55, + "learning_rate": 2.2467668432072918e-06, + "loss": 0.4282, + "step": 20435 + }, + { + "epoch": 7.55, + "learning_rate": 2.2467298928439464e-06, + "loss": 0.421, + "step": 20436 + }, + { + "epoch": 7.55, + "learning_rate": 2.246692942480601e-06, + "loss": 0.5672, + "step": 20437 + }, + { + "epoch": 7.55, + "learning_rate": 2.2466559921172557e-06, + "loss": 0.492, + "step": 20438 + }, + { + "epoch": 7.55, + "learning_rate": 2.2466190417539107e-06, + "loss": 0.4166, + "step": 20439 + }, + { + "epoch": 7.55, + "learning_rate": 2.2465820913905654e-06, + "loss": 0.4295, + "step": 20440 + }, + { + "epoch": 7.55, + "learning_rate": 2.2465451410272204e-06, + "loss": 0.2855, + "step": 20441 + }, + { + "epoch": 7.55, + "learning_rate": 2.246508190663875e-06, + "loss": 0.3481, + "step": 20442 + }, + { + "epoch": 7.55, + "learning_rate": 2.2464712403005297e-06, + "loss": 0.5995, + "step": 20443 + }, + { + "epoch": 7.55, + "learning_rate": 2.2464342899371843e-06, + "loss": 0.4599, + "step": 20444 + }, + { + "epoch": 7.55, + "learning_rate": 2.246397339573839e-06, + "loss": 0.3691, + "step": 20445 + }, + { + "epoch": 7.55, + "learning_rate": 2.246360389210494e-06, + "loss": 0.4448, + "step": 20446 + }, + { + "epoch": 7.55, + "learning_rate": 2.2463234388471487e-06, + "loss": 0.4861, + "step": 20447 + }, + { + "epoch": 7.55, + "learning_rate": 2.2462864884838037e-06, + "loss": 0.4518, + "step": 20448 + }, + { + "epoch": 7.55, + "learning_rate": 2.2462495381204584e-06, + "loss": 0.3949, + "step": 20449 + }, + { + "epoch": 7.55, + "learning_rate": 2.246212587757113e-06, + "loss": 0.5294, + "step": 20450 + }, + { + "epoch": 7.55, + "learning_rate": 2.2461756373937676e-06, + "loss": 0.4794, + "step": 20451 + }, + { + "epoch": 7.55, + "learning_rate": 2.2461386870304227e-06, + "loss": 0.4348, + "step": 20452 + }, + { + "epoch": 7.55, + "learning_rate": 2.2461017366670773e-06, + "loss": 0.3561, + "step": 20453 + }, + { + "epoch": 7.55, + "learning_rate": 2.246064786303732e-06, + "loss": 0.4905, + "step": 20454 + }, + { + "epoch": 7.55, + "learning_rate": 2.246027835940387e-06, + "loss": 0.4297, + "step": 20455 + }, + { + "epoch": 7.55, + "learning_rate": 2.2459908855770417e-06, + "loss": 0.4854, + "step": 20456 + }, + { + "epoch": 7.55, + "learning_rate": 2.2459539352136963e-06, + "loss": 0.3784, + "step": 20457 + }, + { + "epoch": 7.55, + "learning_rate": 2.245916984850351e-06, + "loss": 0.5459, + "step": 20458 + }, + { + "epoch": 7.56, + "learning_rate": 2.245880034487006e-06, + "loss": 0.4298, + "step": 20459 + }, + { + "epoch": 7.56, + "learning_rate": 2.2458430841236606e-06, + "loss": 0.3946, + "step": 20460 + }, + { + "epoch": 7.56, + "learning_rate": 2.2458061337603157e-06, + "loss": 0.4349, + "step": 20461 + }, + { + "epoch": 7.56, + "learning_rate": 2.2457691833969703e-06, + "loss": 0.3543, + "step": 20462 + }, + { + "epoch": 7.56, + "learning_rate": 2.2457322330336245e-06, + "loss": 0.391, + "step": 20463 + }, + { + "epoch": 7.56, + "learning_rate": 2.2456952826702796e-06, + "loss": 0.4816, + "step": 20464 + }, + { + "epoch": 7.56, + "learning_rate": 2.2456583323069342e-06, + "loss": 0.3292, + "step": 20465 + }, + { + "epoch": 7.56, + "learning_rate": 2.2456213819435893e-06, + "loss": 0.2752, + "step": 20466 + }, + { + "epoch": 7.56, + "learning_rate": 2.245584431580244e-06, + "loss": 0.4151, + "step": 20467 + }, + { + "epoch": 7.56, + "learning_rate": 2.245547481216899e-06, + "loss": 0.3598, + "step": 20468 + }, + { + "epoch": 7.56, + "learning_rate": 2.245510530853553e-06, + "loss": 0.399, + "step": 20469 + }, + { + "epoch": 7.56, + "learning_rate": 2.2454735804902083e-06, + "loss": 0.4637, + "step": 20470 + }, + { + "epoch": 7.56, + "learning_rate": 2.245436630126863e-06, + "loss": 0.4895, + "step": 20471 + }, + { + "epoch": 7.56, + "learning_rate": 2.2453996797635175e-06, + "loss": 0.4942, + "step": 20472 + }, + { + "epoch": 7.56, + "learning_rate": 2.2453627294001726e-06, + "loss": 0.3767, + "step": 20473 + }, + { + "epoch": 7.56, + "learning_rate": 2.2453257790368272e-06, + "loss": 0.4905, + "step": 20474 + }, + { + "epoch": 7.56, + "learning_rate": 2.245288828673482e-06, + "loss": 0.5124, + "step": 20475 + }, + { + "epoch": 7.56, + "learning_rate": 2.2452518783101365e-06, + "loss": 0.4326, + "step": 20476 + }, + { + "epoch": 7.56, + "learning_rate": 2.2452149279467916e-06, + "loss": 0.2732, + "step": 20477 + }, + { + "epoch": 7.56, + "learning_rate": 2.245177977583446e-06, + "loss": 0.4344, + "step": 20478 + }, + { + "epoch": 7.56, + "learning_rate": 2.2451410272201013e-06, + "loss": 0.3091, + "step": 20479 + }, + { + "epoch": 7.56, + "learning_rate": 2.245104076856756e-06, + "loss": 0.3592, + "step": 20480 + }, + { + "epoch": 7.56, + "learning_rate": 2.2450671264934105e-06, + "loss": 0.3359, + "step": 20481 + }, + { + "epoch": 7.56, + "learning_rate": 2.245030176130065e-06, + "loss": 0.4043, + "step": 20482 + }, + { + "epoch": 7.56, + "learning_rate": 2.24499322576672e-06, + "loss": 0.4301, + "step": 20483 + }, + { + "epoch": 7.56, + "learning_rate": 2.244956275403375e-06, + "loss": 0.4378, + "step": 20484 + }, + { + "epoch": 7.56, + "learning_rate": 2.2449193250400295e-06, + "loss": 0.4525, + "step": 20485 + }, + { + "epoch": 7.56, + "learning_rate": 2.2448823746766846e-06, + "loss": 0.5166, + "step": 20486 + }, + { + "epoch": 7.57, + "learning_rate": 2.244845424313339e-06, + "loss": 0.4301, + "step": 20487 + }, + { + "epoch": 7.57, + "learning_rate": 2.244808473949994e-06, + "loss": 0.4698, + "step": 20488 + }, + { + "epoch": 7.57, + "learning_rate": 2.2447715235866485e-06, + "loss": 0.4977, + "step": 20489 + }, + { + "epoch": 7.57, + "learning_rate": 2.2447345732233035e-06, + "loss": 0.4017, + "step": 20490 + }, + { + "epoch": 7.57, + "learning_rate": 2.244697622859958e-06, + "loss": 0.5052, + "step": 20491 + }, + { + "epoch": 7.57, + "learning_rate": 2.244660672496613e-06, + "loss": 0.319, + "step": 20492 + }, + { + "epoch": 7.57, + "learning_rate": 2.244623722133268e-06, + "loss": 0.4775, + "step": 20493 + }, + { + "epoch": 7.57, + "learning_rate": 2.2445867717699225e-06, + "loss": 0.3873, + "step": 20494 + }, + { + "epoch": 7.57, + "learning_rate": 2.244549821406577e-06, + "loss": 0.3367, + "step": 20495 + }, + { + "epoch": 7.57, + "learning_rate": 2.2445128710432318e-06, + "loss": 0.4012, + "step": 20496 + }, + { + "epoch": 7.57, + "learning_rate": 2.244475920679887e-06, + "loss": 0.4425, + "step": 20497 + }, + { + "epoch": 7.57, + "learning_rate": 2.2444389703165415e-06, + "loss": 0.4761, + "step": 20498 + }, + { + "epoch": 7.57, + "learning_rate": 2.2444020199531965e-06, + "loss": 0.3522, + "step": 20499 + }, + { + "epoch": 7.57, + "learning_rate": 2.244365069589851e-06, + "loss": 0.4303, + "step": 20500 + }, + { + "epoch": 7.57, + "learning_rate": 2.244328119226506e-06, + "loss": 0.4047, + "step": 20501 + }, + { + "epoch": 7.57, + "learning_rate": 2.2442911688631604e-06, + "loss": 0.316, + "step": 20502 + }, + { + "epoch": 7.57, + "learning_rate": 2.244254218499815e-06, + "loss": 0.3342, + "step": 20503 + }, + { + "epoch": 7.57, + "learning_rate": 2.24421726813647e-06, + "loss": 0.3423, + "step": 20504 + }, + { + "epoch": 7.57, + "learning_rate": 2.2441803177731248e-06, + "loss": 0.4333, + "step": 20505 + }, + { + "epoch": 7.57, + "learning_rate": 2.24414336740978e-06, + "loss": 0.3283, + "step": 20506 + }, + { + "epoch": 7.57, + "learning_rate": 2.2441064170464345e-06, + "loss": 0.2775, + "step": 20507 + }, + { + "epoch": 7.57, + "learning_rate": 2.244069466683089e-06, + "loss": 0.7189, + "step": 20508 + }, + { + "epoch": 7.57, + "learning_rate": 2.2440325163197437e-06, + "loss": 0.3574, + "step": 20509 + }, + { + "epoch": 7.57, + "learning_rate": 2.243995565956399e-06, + "loss": 0.4702, + "step": 20510 + }, + { + "epoch": 7.57, + "learning_rate": 2.2439586155930534e-06, + "loss": 0.447, + "step": 20511 + }, + { + "epoch": 7.57, + "learning_rate": 2.243921665229708e-06, + "loss": 0.3991, + "step": 20512 + }, + { + "epoch": 7.57, + "learning_rate": 2.243884714866363e-06, + "loss": 0.481, + "step": 20513 + }, + { + "epoch": 7.58, + "learning_rate": 2.2438477645030174e-06, + "loss": 0.4937, + "step": 20514 + }, + { + "epoch": 7.58, + "learning_rate": 2.2438108141396724e-06, + "loss": 0.4064, + "step": 20515 + }, + { + "epoch": 7.58, + "learning_rate": 2.243773863776327e-06, + "loss": 0.2811, + "step": 20516 + }, + { + "epoch": 7.58, + "learning_rate": 2.243736913412982e-06, + "loss": 0.3285, + "step": 20517 + }, + { + "epoch": 7.58, + "learning_rate": 2.2436999630496367e-06, + "loss": 0.4179, + "step": 20518 + }, + { + "epoch": 7.58, + "learning_rate": 2.2436630126862914e-06, + "loss": 0.4763, + "step": 20519 + }, + { + "epoch": 7.58, + "learning_rate": 2.243626062322946e-06, + "loss": 0.5171, + "step": 20520 + }, + { + "epoch": 7.58, + "learning_rate": 2.243589111959601e-06, + "loss": 0.4616, + "step": 20521 + }, + { + "epoch": 7.58, + "learning_rate": 2.2435521615962557e-06, + "loss": 0.371, + "step": 20522 + }, + { + "epoch": 7.58, + "learning_rate": 2.2435152112329104e-06, + "loss": 0.4, + "step": 20523 + }, + { + "epoch": 7.58, + "learning_rate": 2.2434782608695654e-06, + "loss": 0.4379, + "step": 20524 + }, + { + "epoch": 7.58, + "learning_rate": 2.24344131050622e-06, + "loss": 0.3, + "step": 20525 + }, + { + "epoch": 7.58, + "learning_rate": 2.2434043601428747e-06, + "loss": 0.4569, + "step": 20526 + }, + { + "epoch": 7.58, + "learning_rate": 2.2433674097795293e-06, + "loss": 0.4613, + "step": 20527 + }, + { + "epoch": 7.58, + "learning_rate": 2.2433304594161844e-06, + "loss": 0.3985, + "step": 20528 + }, + { + "epoch": 7.58, + "learning_rate": 2.243293509052839e-06, + "loss": 0.3922, + "step": 20529 + }, + { + "epoch": 7.58, + "learning_rate": 2.243256558689494e-06, + "loss": 0.4226, + "step": 20530 + }, + { + "epoch": 7.58, + "learning_rate": 2.2432196083261487e-06, + "loss": 0.3486, + "step": 20531 + }, + { + "epoch": 7.58, + "learning_rate": 2.2431826579628033e-06, + "loss": 0.4975, + "step": 20532 + }, + { + "epoch": 7.58, + "learning_rate": 2.243145707599458e-06, + "loss": 0.3706, + "step": 20533 + }, + { + "epoch": 7.58, + "learning_rate": 2.2431087572361126e-06, + "loss": 0.4048, + "step": 20534 + }, + { + "epoch": 7.58, + "learning_rate": 2.2430718068727677e-06, + "loss": 0.3735, + "step": 20535 + }, + { + "epoch": 7.58, + "learning_rate": 2.2430348565094223e-06, + "loss": 0.313, + "step": 20536 + }, + { + "epoch": 7.58, + "learning_rate": 2.2429979061460774e-06, + "loss": 0.3509, + "step": 20537 + }, + { + "epoch": 7.58, + "learning_rate": 2.242960955782732e-06, + "loss": 0.3697, + "step": 20538 + }, + { + "epoch": 7.58, + "learning_rate": 2.2429240054193867e-06, + "loss": 0.4525, + "step": 20539 + }, + { + "epoch": 7.58, + "learning_rate": 2.2428870550560413e-06, + "loss": 0.366, + "step": 20540 + }, + { + "epoch": 7.59, + "learning_rate": 2.2428501046926963e-06, + "loss": 0.4923, + "step": 20541 + }, + { + "epoch": 7.59, + "learning_rate": 2.242813154329351e-06, + "loss": 0.499, + "step": 20542 + }, + { + "epoch": 7.59, + "learning_rate": 2.2427762039660056e-06, + "loss": 0.4574, + "step": 20543 + }, + { + "epoch": 7.59, + "learning_rate": 2.2427392536026607e-06, + "loss": 0.4021, + "step": 20544 + }, + { + "epoch": 7.59, + "learning_rate": 2.2427023032393153e-06, + "loss": 0.4115, + "step": 20545 + }, + { + "epoch": 7.59, + "learning_rate": 2.24266535287597e-06, + "loss": 0.5039, + "step": 20546 + }, + { + "epoch": 7.59, + "learning_rate": 2.2426284025126246e-06, + "loss": 0.5043, + "step": 20547 + }, + { + "epoch": 7.59, + "learning_rate": 2.2425914521492796e-06, + "loss": 0.4007, + "step": 20548 + }, + { + "epoch": 7.59, + "learning_rate": 2.2425545017859343e-06, + "loss": 0.4796, + "step": 20549 + }, + { + "epoch": 7.59, + "learning_rate": 2.2425175514225893e-06, + "loss": 0.3745, + "step": 20550 + }, + { + "epoch": 7.59, + "learning_rate": 2.242480601059244e-06, + "loss": 0.2972, + "step": 20551 + }, + { + "epoch": 7.59, + "learning_rate": 2.242443650695898e-06, + "loss": 0.5042, + "step": 20552 + }, + { + "epoch": 7.59, + "learning_rate": 2.2424067003325533e-06, + "loss": 0.5255, + "step": 20553 + }, + { + "epoch": 7.59, + "learning_rate": 2.242369749969208e-06, + "loss": 0.3594, + "step": 20554 + }, + { + "epoch": 7.59, + "learning_rate": 2.242332799605863e-06, + "loss": 0.41, + "step": 20555 + }, + { + "epoch": 7.59, + "learning_rate": 2.2422958492425176e-06, + "loss": 0.4101, + "step": 20556 + }, + { + "epoch": 7.59, + "learning_rate": 2.2422588988791726e-06, + "loss": 0.5207, + "step": 20557 + }, + { + "epoch": 7.59, + "learning_rate": 2.242221948515827e-06, + "loss": 0.3842, + "step": 20558 + }, + { + "epoch": 7.59, + "learning_rate": 2.242184998152482e-06, + "loss": 0.5079, + "step": 20559 + }, + { + "epoch": 7.59, + "learning_rate": 2.2421480477891366e-06, + "loss": 0.4979, + "step": 20560 + }, + { + "epoch": 7.59, + "learning_rate": 2.2421110974257916e-06, + "loss": 0.4854, + "step": 20561 + }, + { + "epoch": 7.59, + "learning_rate": 2.2420741470624463e-06, + "loss": 0.4591, + "step": 20562 + }, + { + "epoch": 7.59, + "learning_rate": 2.242037196699101e-06, + "loss": 0.4658, + "step": 20563 + }, + { + "epoch": 7.59, + "learning_rate": 2.2420002463357555e-06, + "loss": 0.4026, + "step": 20564 + }, + { + "epoch": 7.59, + "learning_rate": 2.24196329597241e-06, + "loss": 0.4999, + "step": 20565 + }, + { + "epoch": 7.59, + "learning_rate": 2.2419263456090652e-06, + "loss": 0.3408, + "step": 20566 + }, + { + "epoch": 7.59, + "learning_rate": 2.24188939524572e-06, + "loss": 0.4133, + "step": 20567 + }, + { + "epoch": 7.6, + "learning_rate": 2.241852444882375e-06, + "loss": 0.5434, + "step": 20568 + }, + { + "epoch": 7.6, + "learning_rate": 2.2418154945190296e-06, + "loss": 0.4107, + "step": 20569 + }, + { + "epoch": 7.6, + "learning_rate": 2.241778544155684e-06, + "loss": 0.3255, + "step": 20570 + }, + { + "epoch": 7.6, + "learning_rate": 2.241741593792339e-06, + "loss": 0.3244, + "step": 20571 + }, + { + "epoch": 7.6, + "learning_rate": 2.2417046434289935e-06, + "loss": 0.3966, + "step": 20572 + }, + { + "epoch": 7.6, + "learning_rate": 2.2416676930656485e-06, + "loss": 0.5251, + "step": 20573 + }, + { + "epoch": 7.6, + "learning_rate": 2.241630742702303e-06, + "loss": 0.415, + "step": 20574 + }, + { + "epoch": 7.6, + "learning_rate": 2.2415937923389582e-06, + "loss": 0.4412, + "step": 20575 + }, + { + "epoch": 7.6, + "learning_rate": 2.241556841975613e-06, + "loss": 0.5656, + "step": 20576 + }, + { + "epoch": 7.6, + "learning_rate": 2.2415198916122675e-06, + "loss": 0.3879, + "step": 20577 + }, + { + "epoch": 7.6, + "learning_rate": 2.241482941248922e-06, + "loss": 0.4301, + "step": 20578 + }, + { + "epoch": 7.6, + "learning_rate": 2.241445990885577e-06, + "loss": 0.4595, + "step": 20579 + }, + { + "epoch": 7.6, + "learning_rate": 2.241409040522232e-06, + "loss": 0.4602, + "step": 20580 + }, + { + "epoch": 7.6, + "learning_rate": 2.241372090158887e-06, + "loss": 0.3577, + "step": 20581 + }, + { + "epoch": 7.6, + "learning_rate": 2.2413351397955415e-06, + "loss": 0.4871, + "step": 20582 + }, + { + "epoch": 7.6, + "learning_rate": 2.241298189432196e-06, + "loss": 0.5369, + "step": 20583 + }, + { + "epoch": 7.6, + "learning_rate": 2.241261239068851e-06, + "loss": 0.3352, + "step": 20584 + }, + { + "epoch": 7.6, + "learning_rate": 2.2412242887055054e-06, + "loss": 0.5486, + "step": 20585 + }, + { + "epoch": 7.6, + "learning_rate": 2.2411873383421605e-06, + "loss": 0.3821, + "step": 20586 + }, + { + "epoch": 7.6, + "learning_rate": 2.241150387978815e-06, + "loss": 0.4368, + "step": 20587 + }, + { + "epoch": 7.6, + "learning_rate": 2.24111343761547e-06, + "loss": 0.2626, + "step": 20588 + }, + { + "epoch": 7.6, + "learning_rate": 2.241076487252125e-06, + "loss": 0.4014, + "step": 20589 + }, + { + "epoch": 7.6, + "learning_rate": 2.2410395368887795e-06, + "loss": 0.4111, + "step": 20590 + }, + { + "epoch": 7.6, + "learning_rate": 2.241002586525434e-06, + "loss": 0.4112, + "step": 20591 + }, + { + "epoch": 7.6, + "learning_rate": 2.2409656361620887e-06, + "loss": 0.4713, + "step": 20592 + }, + { + "epoch": 7.6, + "learning_rate": 2.240928685798744e-06, + "loss": 0.2641, + "step": 20593 + }, + { + "epoch": 7.6, + "learning_rate": 2.2408917354353984e-06, + "loss": 0.4607, + "step": 20594 + }, + { + "epoch": 7.61, + "learning_rate": 2.2408547850720535e-06, + "loss": 0.4594, + "step": 20595 + }, + { + "epoch": 7.61, + "learning_rate": 2.240817834708708e-06, + "loss": 0.4304, + "step": 20596 + }, + { + "epoch": 7.61, + "learning_rate": 2.2407808843453628e-06, + "loss": 0.5547, + "step": 20597 + }, + { + "epoch": 7.61, + "learning_rate": 2.2407439339820174e-06, + "loss": 0.524, + "step": 20598 + }, + { + "epoch": 7.61, + "learning_rate": 2.2407069836186725e-06, + "loss": 0.3536, + "step": 20599 + }, + { + "epoch": 7.61, + "learning_rate": 2.240670033255327e-06, + "loss": 0.3361, + "step": 20600 + }, + { + "epoch": 7.61, + "learning_rate": 2.240633082891982e-06, + "loss": 0.4876, + "step": 20601 + }, + { + "epoch": 7.61, + "learning_rate": 2.240596132528637e-06, + "loss": 0.4633, + "step": 20602 + }, + { + "epoch": 7.61, + "learning_rate": 2.240559182165291e-06, + "loss": 0.3686, + "step": 20603 + }, + { + "epoch": 7.61, + "learning_rate": 2.240522231801946e-06, + "loss": 0.3827, + "step": 20604 + }, + { + "epoch": 7.61, + "learning_rate": 2.2404852814386007e-06, + "loss": 0.3135, + "step": 20605 + }, + { + "epoch": 7.61, + "learning_rate": 2.2404483310752558e-06, + "loss": 0.3612, + "step": 20606 + }, + { + "epoch": 7.61, + "learning_rate": 2.2404113807119104e-06, + "loss": 0.2848, + "step": 20607 + }, + { + "epoch": 7.61, + "learning_rate": 2.2403744303485655e-06, + "loss": 0.4178, + "step": 20608 + }, + { + "epoch": 7.61, + "learning_rate": 2.2403374799852197e-06, + "loss": 0.4994, + "step": 20609 + }, + { + "epoch": 7.61, + "learning_rate": 2.2403005296218747e-06, + "loss": 0.3656, + "step": 20610 + }, + { + "epoch": 7.61, + "learning_rate": 2.2402635792585294e-06, + "loss": 0.4996, + "step": 20611 + }, + { + "epoch": 7.61, + "learning_rate": 2.240226628895184e-06, + "loss": 0.3789, + "step": 20612 + }, + { + "epoch": 7.61, + "learning_rate": 2.240189678531839e-06, + "loss": 0.3749, + "step": 20613 + }, + { + "epoch": 7.61, + "learning_rate": 2.2401527281684937e-06, + "loss": 0.3607, + "step": 20614 + }, + { + "epoch": 7.61, + "learning_rate": 2.2401157778051483e-06, + "loss": 0.4623, + "step": 20615 + }, + { + "epoch": 7.61, + "learning_rate": 2.240078827441803e-06, + "loss": 0.3396, + "step": 20616 + }, + { + "epoch": 7.61, + "learning_rate": 2.240041877078458e-06, + "loss": 0.3495, + "step": 20617 + }, + { + "epoch": 7.61, + "learning_rate": 2.2400049267151127e-06, + "loss": 0.3239, + "step": 20618 + }, + { + "epoch": 7.61, + "learning_rate": 2.2399679763517677e-06, + "loss": 0.6087, + "step": 20619 + }, + { + "epoch": 7.61, + "learning_rate": 2.2399310259884224e-06, + "loss": 0.5207, + "step": 20620 + }, + { + "epoch": 7.61, + "learning_rate": 2.239894075625077e-06, + "loss": 0.3691, + "step": 20621 + }, + { + "epoch": 7.62, + "learning_rate": 2.2398571252617316e-06, + "loss": 0.3374, + "step": 20622 + }, + { + "epoch": 7.62, + "learning_rate": 2.2398201748983863e-06, + "loss": 0.414, + "step": 20623 + }, + { + "epoch": 7.62, + "learning_rate": 2.2397832245350413e-06, + "loss": 0.3972, + "step": 20624 + }, + { + "epoch": 7.62, + "learning_rate": 2.239746274171696e-06, + "loss": 0.4087, + "step": 20625 + }, + { + "epoch": 7.62, + "learning_rate": 2.239709323808351e-06, + "loss": 0.4319, + "step": 20626 + }, + { + "epoch": 7.62, + "learning_rate": 2.2396723734450057e-06, + "loss": 0.3781, + "step": 20627 + }, + { + "epoch": 7.62, + "learning_rate": 2.2396354230816603e-06, + "loss": 0.5, + "step": 20628 + }, + { + "epoch": 7.62, + "learning_rate": 2.239598472718315e-06, + "loss": 0.4397, + "step": 20629 + }, + { + "epoch": 7.62, + "learning_rate": 2.23956152235497e-06, + "loss": 0.5302, + "step": 20630 + }, + { + "epoch": 7.62, + "learning_rate": 2.2395245719916246e-06, + "loss": 0.4761, + "step": 20631 + }, + { + "epoch": 7.62, + "learning_rate": 2.2394876216282793e-06, + "loss": 0.4444, + "step": 20632 + }, + { + "epoch": 7.62, + "learning_rate": 2.2394506712649343e-06, + "loss": 0.3062, + "step": 20633 + }, + { + "epoch": 7.62, + "learning_rate": 2.239413720901589e-06, + "loss": 0.4318, + "step": 20634 + }, + { + "epoch": 7.62, + "learning_rate": 2.2393767705382436e-06, + "loss": 0.5031, + "step": 20635 + }, + { + "epoch": 7.62, + "learning_rate": 2.2393398201748982e-06, + "loss": 0.5073, + "step": 20636 + }, + { + "epoch": 7.62, + "learning_rate": 2.2393028698115533e-06, + "loss": 0.6031, + "step": 20637 + }, + { + "epoch": 7.62, + "learning_rate": 2.239265919448208e-06, + "loss": 0.4642, + "step": 20638 + }, + { + "epoch": 7.62, + "learning_rate": 2.239228969084863e-06, + "loss": 0.3828, + "step": 20639 + }, + { + "epoch": 7.62, + "learning_rate": 2.2391920187215176e-06, + "loss": 0.5151, + "step": 20640 + }, + { + "epoch": 7.62, + "learning_rate": 2.2391550683581723e-06, + "loss": 0.4337, + "step": 20641 + }, + { + "epoch": 7.62, + "learning_rate": 2.239118117994827e-06, + "loss": 0.3622, + "step": 20642 + }, + { + "epoch": 7.62, + "learning_rate": 2.2390811676314815e-06, + "loss": 0.4446, + "step": 20643 + }, + { + "epoch": 7.62, + "learning_rate": 2.2390442172681366e-06, + "loss": 0.4356, + "step": 20644 + }, + { + "epoch": 7.62, + "learning_rate": 2.2390072669047912e-06, + "loss": 0.4049, + "step": 20645 + }, + { + "epoch": 7.62, + "learning_rate": 2.2389703165414463e-06, + "loss": 0.473, + "step": 20646 + }, + { + "epoch": 7.62, + "learning_rate": 2.238933366178101e-06, + "loss": 0.3897, + "step": 20647 + }, + { + "epoch": 7.62, + "learning_rate": 2.2388964158147556e-06, + "loss": 0.4332, + "step": 20648 + }, + { + "epoch": 7.63, + "learning_rate": 2.23885946545141e-06, + "loss": 0.4445, + "step": 20649 + }, + { + "epoch": 7.63, + "learning_rate": 2.2388225150880653e-06, + "loss": 0.3204, + "step": 20650 + }, + { + "epoch": 7.63, + "learning_rate": 2.23878556472472e-06, + "loss": 0.4565, + "step": 20651 + }, + { + "epoch": 7.63, + "learning_rate": 2.2387486143613745e-06, + "loss": 0.3715, + "step": 20652 + }, + { + "epoch": 7.63, + "learning_rate": 2.2387116639980296e-06, + "loss": 0.46, + "step": 20653 + }, + { + "epoch": 7.63, + "learning_rate": 2.238674713634684e-06, + "loss": 0.556, + "step": 20654 + }, + { + "epoch": 7.63, + "learning_rate": 2.238637763271339e-06, + "loss": 0.4774, + "step": 20655 + }, + { + "epoch": 7.63, + "learning_rate": 2.2386008129079935e-06, + "loss": 0.4645, + "step": 20656 + }, + { + "epoch": 7.63, + "learning_rate": 2.2385638625446486e-06, + "loss": 0.3687, + "step": 20657 + }, + { + "epoch": 7.63, + "learning_rate": 2.238526912181303e-06, + "loss": 0.4823, + "step": 20658 + }, + { + "epoch": 7.63, + "learning_rate": 2.2384899618179583e-06, + "loss": 0.4292, + "step": 20659 + }, + { + "epoch": 7.63, + "learning_rate": 2.2384530114546125e-06, + "loss": 0.4461, + "step": 20660 + }, + { + "epoch": 7.63, + "learning_rate": 2.2384160610912675e-06, + "loss": 0.4714, + "step": 20661 + }, + { + "epoch": 7.63, + "learning_rate": 2.238379110727922e-06, + "loss": 0.3982, + "step": 20662 + }, + { + "epoch": 7.63, + "learning_rate": 2.238342160364577e-06, + "loss": 0.525, + "step": 20663 + }, + { + "epoch": 7.63, + "learning_rate": 2.238305210001232e-06, + "loss": 0.537, + "step": 20664 + }, + { + "epoch": 7.63, + "learning_rate": 2.2382682596378865e-06, + "loss": 0.5393, + "step": 20665 + }, + { + "epoch": 7.63, + "learning_rate": 2.238231309274541e-06, + "loss": 0.4702, + "step": 20666 + }, + { + "epoch": 7.63, + "learning_rate": 2.2381943589111958e-06, + "loss": 0.365, + "step": 20667 + }, + { + "epoch": 7.63, + "learning_rate": 2.238157408547851e-06, + "loss": 0.3667, + "step": 20668 + }, + { + "epoch": 7.63, + "learning_rate": 2.2381204581845055e-06, + "loss": 0.4592, + "step": 20669 + }, + { + "epoch": 7.63, + "learning_rate": 2.2380835078211605e-06, + "loss": 0.4555, + "step": 20670 + }, + { + "epoch": 7.63, + "learning_rate": 2.238046557457815e-06, + "loss": 0.3077, + "step": 20671 + }, + { + "epoch": 7.63, + "learning_rate": 2.23800960709447e-06, + "loss": 0.4021, + "step": 20672 + }, + { + "epoch": 7.63, + "learning_rate": 2.2379726567311245e-06, + "loss": 0.5099, + "step": 20673 + }, + { + "epoch": 7.63, + "learning_rate": 2.237935706367779e-06, + "loss": 0.5299, + "step": 20674 + }, + { + "epoch": 7.63, + "learning_rate": 2.237898756004434e-06, + "loss": 0.3125, + "step": 20675 + }, + { + "epoch": 7.64, + "learning_rate": 2.2378618056410888e-06, + "loss": 0.3501, + "step": 20676 + }, + { + "epoch": 7.64, + "learning_rate": 2.237824855277744e-06, + "loss": 0.4459, + "step": 20677 + }, + { + "epoch": 7.64, + "learning_rate": 2.2377879049143985e-06, + "loss": 0.3773, + "step": 20678 + }, + { + "epoch": 7.64, + "learning_rate": 2.237750954551053e-06, + "loss": 0.3409, + "step": 20679 + }, + { + "epoch": 7.64, + "learning_rate": 2.2377140041877078e-06, + "loss": 0.4433, + "step": 20680 + }, + { + "epoch": 7.64, + "learning_rate": 2.237677053824363e-06, + "loss": 0.3033, + "step": 20681 + }, + { + "epoch": 7.64, + "learning_rate": 2.2376401034610174e-06, + "loss": 0.4954, + "step": 20682 + }, + { + "epoch": 7.64, + "learning_rate": 2.237603153097672e-06, + "loss": 0.4122, + "step": 20683 + }, + { + "epoch": 7.64, + "learning_rate": 2.237566202734327e-06, + "loss": 0.4761, + "step": 20684 + }, + { + "epoch": 7.64, + "learning_rate": 2.2375292523709818e-06, + "loss": 0.2801, + "step": 20685 + }, + { + "epoch": 7.64, + "learning_rate": 2.2374923020076364e-06, + "loss": 0.48, + "step": 20686 + }, + { + "epoch": 7.64, + "learning_rate": 2.237455351644291e-06, + "loss": 0.272, + "step": 20687 + }, + { + "epoch": 7.64, + "learning_rate": 2.237418401280946e-06, + "loss": 0.4605, + "step": 20688 + }, + { + "epoch": 7.64, + "learning_rate": 2.2373814509176008e-06, + "loss": 0.4635, + "step": 20689 + }, + { + "epoch": 7.64, + "learning_rate": 2.237344500554256e-06, + "loss": 0.4029, + "step": 20690 + }, + { + "epoch": 7.64, + "learning_rate": 2.2373075501909104e-06, + "loss": 0.3856, + "step": 20691 + }, + { + "epoch": 7.64, + "learning_rate": 2.237270599827565e-06, + "loss": 0.3872, + "step": 20692 + }, + { + "epoch": 7.64, + "learning_rate": 2.2372336494642197e-06, + "loss": 0.409, + "step": 20693 + }, + { + "epoch": 7.64, + "learning_rate": 2.2371966991008744e-06, + "loss": 0.4963, + "step": 20694 + }, + { + "epoch": 7.64, + "learning_rate": 2.2371597487375294e-06, + "loss": 0.3667, + "step": 20695 + }, + { + "epoch": 7.64, + "learning_rate": 2.237122798374184e-06, + "loss": 0.5807, + "step": 20696 + }, + { + "epoch": 7.64, + "learning_rate": 2.237085848010839e-06, + "loss": 0.3582, + "step": 20697 + }, + { + "epoch": 7.64, + "learning_rate": 2.2370488976474937e-06, + "loss": 0.5083, + "step": 20698 + }, + { + "epoch": 7.64, + "learning_rate": 2.2370119472841484e-06, + "loss": 0.3952, + "step": 20699 + }, + { + "epoch": 7.64, + "learning_rate": 2.236974996920803e-06, + "loss": 0.4146, + "step": 20700 + }, + { + "epoch": 7.64, + "learning_rate": 2.2369380465574577e-06, + "loss": 0.3714, + "step": 20701 + }, + { + "epoch": 7.64, + "learning_rate": 2.2369010961941127e-06, + "loss": 0.439, + "step": 20702 + }, + { + "epoch": 7.65, + "learning_rate": 2.2368641458307674e-06, + "loss": 0.3756, + "step": 20703 + }, + { + "epoch": 7.65, + "learning_rate": 2.2368271954674224e-06, + "loss": 0.3741, + "step": 20704 + }, + { + "epoch": 7.65, + "learning_rate": 2.2367902451040766e-06, + "loss": 0.3214, + "step": 20705 + }, + { + "epoch": 7.65, + "learning_rate": 2.2367532947407317e-06, + "loss": 0.5185, + "step": 20706 + }, + { + "epoch": 7.65, + "learning_rate": 2.2367163443773863e-06, + "loss": 0.4243, + "step": 20707 + }, + { + "epoch": 7.65, + "learning_rate": 2.2366793940140414e-06, + "loss": 0.411, + "step": 20708 + }, + { + "epoch": 7.65, + "learning_rate": 2.236642443650696e-06, + "loss": 0.3368, + "step": 20709 + }, + { + "epoch": 7.65, + "learning_rate": 2.2366054932873507e-06, + "loss": 0.5254, + "step": 20710 + }, + { + "epoch": 7.65, + "learning_rate": 2.2365685429240053e-06, + "loss": 0.3626, + "step": 20711 + }, + { + "epoch": 7.65, + "learning_rate": 2.23653159256066e-06, + "loss": 0.5263, + "step": 20712 + }, + { + "epoch": 7.65, + "learning_rate": 2.236494642197315e-06, + "loss": 0.4045, + "step": 20713 + }, + { + "epoch": 7.65, + "learning_rate": 2.2364576918339696e-06, + "loss": 0.4614, + "step": 20714 + }, + { + "epoch": 7.65, + "learning_rate": 2.2364207414706247e-06, + "loss": 0.5249, + "step": 20715 + }, + { + "epoch": 7.65, + "learning_rate": 2.2363837911072793e-06, + "loss": 0.5545, + "step": 20716 + }, + { + "epoch": 7.65, + "learning_rate": 2.236346840743934e-06, + "loss": 0.4389, + "step": 20717 + }, + { + "epoch": 7.65, + "learning_rate": 2.2363098903805886e-06, + "loss": 0.3387, + "step": 20718 + }, + { + "epoch": 7.65, + "learning_rate": 2.2362729400172437e-06, + "loss": 0.3397, + "step": 20719 + }, + { + "epoch": 7.65, + "learning_rate": 2.2362359896538983e-06, + "loss": 0.4759, + "step": 20720 + }, + { + "epoch": 7.65, + "learning_rate": 2.236199039290553e-06, + "loss": 0.4101, + "step": 20721 + }, + { + "epoch": 7.65, + "learning_rate": 2.236162088927208e-06, + "loss": 0.4366, + "step": 20722 + }, + { + "epoch": 7.65, + "learning_rate": 2.2361251385638626e-06, + "loss": 0.5209, + "step": 20723 + }, + { + "epoch": 7.65, + "learning_rate": 2.2360881882005173e-06, + "loss": 0.3762, + "step": 20724 + }, + { + "epoch": 7.65, + "learning_rate": 2.236051237837172e-06, + "loss": 0.4191, + "step": 20725 + }, + { + "epoch": 7.65, + "learning_rate": 2.236014287473827e-06, + "loss": 0.3606, + "step": 20726 + }, + { + "epoch": 7.65, + "learning_rate": 2.2359773371104816e-06, + "loss": 0.4589, + "step": 20727 + }, + { + "epoch": 7.65, + "learning_rate": 2.2359403867471367e-06, + "loss": 0.4305, + "step": 20728 + }, + { + "epoch": 7.65, + "learning_rate": 2.2359034363837913e-06, + "loss": 0.3446, + "step": 20729 + }, + { + "epoch": 7.66, + "learning_rate": 2.235866486020446e-06, + "loss": 0.4446, + "step": 20730 + }, + { + "epoch": 7.66, + "learning_rate": 2.2358295356571006e-06, + "loss": 0.4584, + "step": 20731 + }, + { + "epoch": 7.66, + "learning_rate": 2.235792585293755e-06, + "loss": 0.4067, + "step": 20732 + }, + { + "epoch": 7.66, + "learning_rate": 2.2357556349304103e-06, + "loss": 0.2976, + "step": 20733 + }, + { + "epoch": 7.66, + "learning_rate": 2.235718684567065e-06, + "loss": 0.4326, + "step": 20734 + }, + { + "epoch": 7.66, + "learning_rate": 2.23568173420372e-06, + "loss": 0.4053, + "step": 20735 + }, + { + "epoch": 7.66, + "learning_rate": 2.2356447838403746e-06, + "loss": 0.4171, + "step": 20736 + }, + { + "epoch": 7.66, + "learning_rate": 2.2356078334770292e-06, + "loss": 0.4553, + "step": 20737 + }, + { + "epoch": 7.66, + "learning_rate": 2.235570883113684e-06, + "loss": 0.4661, + "step": 20738 + }, + { + "epoch": 7.66, + "learning_rate": 2.235533932750339e-06, + "loss": 0.4206, + "step": 20739 + }, + { + "epoch": 7.66, + "learning_rate": 2.2354969823869936e-06, + "loss": 0.4976, + "step": 20740 + }, + { + "epoch": 7.66, + "learning_rate": 2.235460032023648e-06, + "loss": 0.4327, + "step": 20741 + }, + { + "epoch": 7.66, + "learning_rate": 2.2354230816603033e-06, + "loss": 0.5076, + "step": 20742 + }, + { + "epoch": 7.66, + "learning_rate": 2.2353861312969575e-06, + "loss": 0.4439, + "step": 20743 + }, + { + "epoch": 7.66, + "learning_rate": 2.2353491809336125e-06, + "loss": 0.3824, + "step": 20744 + }, + { + "epoch": 7.66, + "learning_rate": 2.235312230570267e-06, + "loss": 0.4879, + "step": 20745 + }, + { + "epoch": 7.66, + "learning_rate": 2.2352752802069222e-06, + "loss": 0.3978, + "step": 20746 + }, + { + "epoch": 7.66, + "learning_rate": 2.235238329843577e-06, + "loss": 0.3275, + "step": 20747 + }, + { + "epoch": 7.66, + "learning_rate": 2.235201379480232e-06, + "loss": 0.5861, + "step": 20748 + }, + { + "epoch": 7.66, + "learning_rate": 2.235164429116886e-06, + "loss": 0.3304, + "step": 20749 + }, + { + "epoch": 7.66, + "learning_rate": 2.235127478753541e-06, + "loss": 0.5467, + "step": 20750 + }, + { + "epoch": 7.66, + "learning_rate": 2.235090528390196e-06, + "loss": 0.4195, + "step": 20751 + }, + { + "epoch": 7.66, + "learning_rate": 2.2350535780268505e-06, + "loss": 0.5085, + "step": 20752 + }, + { + "epoch": 7.66, + "learning_rate": 2.2350166276635055e-06, + "loss": 0.4603, + "step": 20753 + }, + { + "epoch": 7.66, + "learning_rate": 2.23497967730016e-06, + "loss": 0.4288, + "step": 20754 + }, + { + "epoch": 7.66, + "learning_rate": 2.234942726936815e-06, + "loss": 0.2747, + "step": 20755 + }, + { + "epoch": 7.66, + "learning_rate": 2.2349057765734694e-06, + "loss": 0.4465, + "step": 20756 + }, + { + "epoch": 7.67, + "learning_rate": 2.2348688262101245e-06, + "loss": 0.3948, + "step": 20757 + }, + { + "epoch": 7.67, + "learning_rate": 2.234831875846779e-06, + "loss": 0.3583, + "step": 20758 + }, + { + "epoch": 7.67, + "learning_rate": 2.234794925483434e-06, + "loss": 0.3138, + "step": 20759 + }, + { + "epoch": 7.67, + "learning_rate": 2.234757975120089e-06, + "loss": 0.4591, + "step": 20760 + }, + { + "epoch": 7.67, + "learning_rate": 2.2347210247567435e-06, + "loss": 0.3986, + "step": 20761 + }, + { + "epoch": 7.67, + "learning_rate": 2.234684074393398e-06, + "loss": 0.4681, + "step": 20762 + }, + { + "epoch": 7.67, + "learning_rate": 2.2346471240300527e-06, + "loss": 0.343, + "step": 20763 + }, + { + "epoch": 7.67, + "learning_rate": 2.234610173666708e-06, + "loss": 0.5462, + "step": 20764 + }, + { + "epoch": 7.67, + "learning_rate": 2.2345732233033624e-06, + "loss": 0.466, + "step": 20765 + }, + { + "epoch": 7.67, + "learning_rate": 2.2345362729400175e-06, + "loss": 0.4209, + "step": 20766 + }, + { + "epoch": 7.67, + "learning_rate": 2.234499322576672e-06, + "loss": 0.3552, + "step": 20767 + }, + { + "epoch": 7.67, + "learning_rate": 2.2344623722133268e-06, + "loss": 0.5809, + "step": 20768 + }, + { + "epoch": 7.67, + "learning_rate": 2.2344254218499814e-06, + "loss": 0.5545, + "step": 20769 + }, + { + "epoch": 7.67, + "learning_rate": 2.2343884714866365e-06, + "loss": 0.4957, + "step": 20770 + }, + { + "epoch": 7.67, + "learning_rate": 2.234351521123291e-06, + "loss": 0.3612, + "step": 20771 + }, + { + "epoch": 7.67, + "learning_rate": 2.2343145707599457e-06, + "loss": 0.4783, + "step": 20772 + }, + { + "epoch": 7.67, + "learning_rate": 2.234277620396601e-06, + "loss": 0.3998, + "step": 20773 + }, + { + "epoch": 7.67, + "learning_rate": 2.2342406700332554e-06, + "loss": 0.4987, + "step": 20774 + }, + { + "epoch": 7.67, + "learning_rate": 2.23420371966991e-06, + "loss": 0.394, + "step": 20775 + }, + { + "epoch": 7.67, + "learning_rate": 2.2341667693065647e-06, + "loss": 0.5426, + "step": 20776 + }, + { + "epoch": 7.67, + "learning_rate": 2.2341298189432198e-06, + "loss": 0.4335, + "step": 20777 + }, + { + "epoch": 7.67, + "learning_rate": 2.2340928685798744e-06, + "loss": 0.3477, + "step": 20778 + }, + { + "epoch": 7.67, + "learning_rate": 2.2340559182165295e-06, + "loss": 0.4693, + "step": 20779 + }, + { + "epoch": 7.67, + "learning_rate": 2.234018967853184e-06, + "loss": 0.4168, + "step": 20780 + }, + { + "epoch": 7.67, + "learning_rate": 2.2339820174898387e-06, + "loss": 0.42, + "step": 20781 + }, + { + "epoch": 7.67, + "learning_rate": 2.2339450671264934e-06, + "loss": 0.2708, + "step": 20782 + }, + { + "epoch": 7.67, + "learning_rate": 2.233908116763148e-06, + "loss": 0.3793, + "step": 20783 + }, + { + "epoch": 7.68, + "learning_rate": 2.233871166399803e-06, + "loss": 0.3133, + "step": 20784 + }, + { + "epoch": 7.68, + "learning_rate": 2.2338342160364577e-06, + "loss": 0.4994, + "step": 20785 + }, + { + "epoch": 7.68, + "learning_rate": 2.2337972656731128e-06, + "loss": 0.4251, + "step": 20786 + }, + { + "epoch": 7.68, + "learning_rate": 2.2337603153097674e-06, + "loss": 0.5924, + "step": 20787 + }, + { + "epoch": 7.68, + "learning_rate": 2.233723364946422e-06, + "loss": 0.3982, + "step": 20788 + }, + { + "epoch": 7.68, + "learning_rate": 2.2336864145830767e-06, + "loss": 0.5053, + "step": 20789 + }, + { + "epoch": 7.68, + "learning_rate": 2.2336494642197317e-06, + "loss": 0.4099, + "step": 20790 + }, + { + "epoch": 7.68, + "learning_rate": 2.2336125138563864e-06, + "loss": 0.4954, + "step": 20791 + }, + { + "epoch": 7.68, + "learning_rate": 2.233575563493041e-06, + "loss": 0.4889, + "step": 20792 + }, + { + "epoch": 7.68, + "learning_rate": 2.233538613129696e-06, + "loss": 0.4554, + "step": 20793 + }, + { + "epoch": 7.68, + "learning_rate": 2.2335016627663503e-06, + "loss": 0.3759, + "step": 20794 + }, + { + "epoch": 7.68, + "learning_rate": 2.2334647124030053e-06, + "loss": 0.4014, + "step": 20795 + }, + { + "epoch": 7.68, + "learning_rate": 2.23342776203966e-06, + "loss": 0.4229, + "step": 20796 + }, + { + "epoch": 7.68, + "learning_rate": 2.233390811676315e-06, + "loss": 0.4354, + "step": 20797 + }, + { + "epoch": 7.68, + "learning_rate": 2.2333538613129697e-06, + "loss": 0.4864, + "step": 20798 + }, + { + "epoch": 7.68, + "learning_rate": 2.2333169109496247e-06, + "loss": 0.2954, + "step": 20799 + }, + { + "epoch": 7.68, + "learning_rate": 2.233279960586279e-06, + "loss": 0.5006, + "step": 20800 + }, + { + "epoch": 7.68, + "learning_rate": 2.2332430102229336e-06, + "loss": 0.3186, + "step": 20801 + }, + { + "epoch": 7.68, + "learning_rate": 2.2332060598595886e-06, + "loss": 0.3677, + "step": 20802 + }, + { + "epoch": 7.68, + "learning_rate": 2.2331691094962433e-06, + "loss": 0.4205, + "step": 20803 + }, + { + "epoch": 7.68, + "learning_rate": 2.2331321591328983e-06, + "loss": 0.4014, + "step": 20804 + }, + { + "epoch": 7.68, + "learning_rate": 2.233095208769553e-06, + "loss": 0.3554, + "step": 20805 + }, + { + "epoch": 7.68, + "learning_rate": 2.2330582584062076e-06, + "loss": 0.3615, + "step": 20806 + }, + { + "epoch": 7.68, + "learning_rate": 2.2330213080428622e-06, + "loss": 0.405, + "step": 20807 + }, + { + "epoch": 7.68, + "learning_rate": 2.2329843576795173e-06, + "loss": 0.4787, + "step": 20808 + }, + { + "epoch": 7.68, + "learning_rate": 2.232947407316172e-06, + "loss": 0.4449, + "step": 20809 + }, + { + "epoch": 7.68, + "learning_rate": 2.232910456952827e-06, + "loss": 0.5792, + "step": 20810 + }, + { + "epoch": 7.69, + "learning_rate": 2.2328735065894816e-06, + "loss": 0.4162, + "step": 20811 + }, + { + "epoch": 7.69, + "learning_rate": 2.2328365562261363e-06, + "loss": 0.2581, + "step": 20812 + }, + { + "epoch": 7.69, + "learning_rate": 2.232799605862791e-06, + "loss": 0.4826, + "step": 20813 + }, + { + "epoch": 7.69, + "learning_rate": 2.2327626554994456e-06, + "loss": 0.3247, + "step": 20814 + }, + { + "epoch": 7.69, + "learning_rate": 2.2327257051361006e-06, + "loss": 0.4351, + "step": 20815 + }, + { + "epoch": 7.69, + "learning_rate": 2.2326887547727552e-06, + "loss": 0.3549, + "step": 20816 + }, + { + "epoch": 7.69, + "learning_rate": 2.2326518044094103e-06, + "loss": 0.3482, + "step": 20817 + }, + { + "epoch": 7.69, + "learning_rate": 2.232614854046065e-06, + "loss": 0.3818, + "step": 20818 + }, + { + "epoch": 7.69, + "learning_rate": 2.2325779036827196e-06, + "loss": 0.4364, + "step": 20819 + }, + { + "epoch": 7.69, + "learning_rate": 2.2325409533193742e-06, + "loss": 0.4742, + "step": 20820 + }, + { + "epoch": 7.69, + "learning_rate": 2.232504002956029e-06, + "loss": 0.4234, + "step": 20821 + }, + { + "epoch": 7.69, + "learning_rate": 2.232467052592684e-06, + "loss": 0.4693, + "step": 20822 + }, + { + "epoch": 7.69, + "learning_rate": 2.2324301022293386e-06, + "loss": 0.4816, + "step": 20823 + }, + { + "epoch": 7.69, + "learning_rate": 2.2323931518659936e-06, + "loss": 0.4165, + "step": 20824 + }, + { + "epoch": 7.69, + "learning_rate": 2.2323562015026482e-06, + "loss": 0.4081, + "step": 20825 + }, + { + "epoch": 7.69, + "learning_rate": 2.232319251139303e-06, + "loss": 0.4594, + "step": 20826 + }, + { + "epoch": 7.69, + "learning_rate": 2.2322823007759575e-06, + "loss": 0.3863, + "step": 20827 + }, + { + "epoch": 7.69, + "learning_rate": 2.2322453504126126e-06, + "loss": 0.3984, + "step": 20828 + }, + { + "epoch": 7.69, + "learning_rate": 2.2322084000492672e-06, + "loss": 0.3838, + "step": 20829 + }, + { + "epoch": 7.69, + "learning_rate": 2.2321714496859223e-06, + "loss": 0.4348, + "step": 20830 + }, + { + "epoch": 7.69, + "learning_rate": 2.232134499322577e-06, + "loss": 0.305, + "step": 20831 + }, + { + "epoch": 7.69, + "learning_rate": 2.2320975489592315e-06, + "loss": 0.4191, + "step": 20832 + }, + { + "epoch": 7.69, + "learning_rate": 2.232060598595886e-06, + "loss": 0.4679, + "step": 20833 + }, + { + "epoch": 7.69, + "learning_rate": 2.232023648232541e-06, + "loss": 0.504, + "step": 20834 + }, + { + "epoch": 7.69, + "learning_rate": 2.231986697869196e-06, + "loss": 0.3921, + "step": 20835 + }, + { + "epoch": 7.69, + "learning_rate": 2.2319497475058505e-06, + "loss": 0.4195, + "step": 20836 + }, + { + "epoch": 7.69, + "learning_rate": 2.2319127971425056e-06, + "loss": 0.4009, + "step": 20837 + }, + { + "epoch": 7.69, + "learning_rate": 2.2318758467791602e-06, + "loss": 0.3178, + "step": 20838 + }, + { + "epoch": 7.7, + "learning_rate": 2.231838896415815e-06, + "loss": 0.455, + "step": 20839 + }, + { + "epoch": 7.7, + "learning_rate": 2.2318019460524695e-06, + "loss": 0.4428, + "step": 20840 + }, + { + "epoch": 7.7, + "learning_rate": 2.231764995689124e-06, + "loss": 0.3591, + "step": 20841 + }, + { + "epoch": 7.7, + "learning_rate": 2.231728045325779e-06, + "loss": 0.4298, + "step": 20842 + }, + { + "epoch": 7.7, + "learning_rate": 2.231691094962434e-06, + "loss": 0.3989, + "step": 20843 + }, + { + "epoch": 7.7, + "learning_rate": 2.231654144599089e-06, + "loss": 0.5756, + "step": 20844 + }, + { + "epoch": 7.7, + "learning_rate": 2.231617194235743e-06, + "loss": 0.4866, + "step": 20845 + }, + { + "epoch": 7.7, + "learning_rate": 2.231580243872398e-06, + "loss": 0.4139, + "step": 20846 + }, + { + "epoch": 7.7, + "learning_rate": 2.231543293509053e-06, + "loss": 0.4123, + "step": 20847 + }, + { + "epoch": 7.7, + "learning_rate": 2.231506343145708e-06, + "loss": 0.3903, + "step": 20848 + }, + { + "epoch": 7.7, + "learning_rate": 2.2314693927823625e-06, + "loss": 0.3342, + "step": 20849 + }, + { + "epoch": 7.7, + "learning_rate": 2.2314324424190175e-06, + "loss": 0.5487, + "step": 20850 + }, + { + "epoch": 7.7, + "learning_rate": 2.2313954920556718e-06, + "loss": 0.3914, + "step": 20851 + }, + { + "epoch": 7.7, + "learning_rate": 2.2313585416923264e-06, + "loss": 0.333, + "step": 20852 + }, + { + "epoch": 7.7, + "learning_rate": 2.2313215913289815e-06, + "loss": 0.3627, + "step": 20853 + }, + { + "epoch": 7.7, + "learning_rate": 2.231284640965636e-06, + "loss": 0.3888, + "step": 20854 + }, + { + "epoch": 7.7, + "learning_rate": 2.231247690602291e-06, + "loss": 0.4854, + "step": 20855 + }, + { + "epoch": 7.7, + "learning_rate": 2.2312107402389458e-06, + "loss": 0.4506, + "step": 20856 + }, + { + "epoch": 7.7, + "learning_rate": 2.2311737898756004e-06, + "loss": 0.4173, + "step": 20857 + }, + { + "epoch": 7.7, + "learning_rate": 2.231136839512255e-06, + "loss": 0.5229, + "step": 20858 + }, + { + "epoch": 7.7, + "learning_rate": 2.23109988914891e-06, + "loss": 0.4828, + "step": 20859 + }, + { + "epoch": 7.7, + "learning_rate": 2.2310629387855648e-06, + "loss": 0.4972, + "step": 20860 + }, + { + "epoch": 7.7, + "learning_rate": 2.2310259884222194e-06, + "loss": 0.3845, + "step": 20861 + }, + { + "epoch": 7.7, + "learning_rate": 2.2309890380588745e-06, + "loss": 0.3742, + "step": 20862 + }, + { + "epoch": 7.7, + "learning_rate": 2.230952087695529e-06, + "loss": 0.4018, + "step": 20863 + }, + { + "epoch": 7.7, + "learning_rate": 2.2309151373321837e-06, + "loss": 0.393, + "step": 20864 + }, + { + "epoch": 7.7, + "learning_rate": 2.2308781869688384e-06, + "loss": 0.4017, + "step": 20865 + }, + { + "epoch": 7.71, + "learning_rate": 2.2308412366054934e-06, + "loss": 0.484, + "step": 20866 + }, + { + "epoch": 7.71, + "learning_rate": 2.230804286242148e-06, + "loss": 0.52, + "step": 20867 + }, + { + "epoch": 7.71, + "learning_rate": 2.230767335878803e-06, + "loss": 0.3659, + "step": 20868 + }, + { + "epoch": 7.71, + "learning_rate": 2.2307303855154578e-06, + "loss": 0.362, + "step": 20869 + }, + { + "epoch": 7.71, + "learning_rate": 2.2306934351521124e-06, + "loss": 0.5276, + "step": 20870 + }, + { + "epoch": 7.71, + "learning_rate": 2.230656484788767e-06, + "loss": 0.4447, + "step": 20871 + }, + { + "epoch": 7.71, + "learning_rate": 2.2306195344254217e-06, + "loss": 0.3568, + "step": 20872 + }, + { + "epoch": 7.71, + "learning_rate": 2.2305825840620767e-06, + "loss": 0.4493, + "step": 20873 + }, + { + "epoch": 7.71, + "learning_rate": 2.2305456336987314e-06, + "loss": 0.4616, + "step": 20874 + }, + { + "epoch": 7.71, + "learning_rate": 2.2305086833353864e-06, + "loss": 0.452, + "step": 20875 + }, + { + "epoch": 7.71, + "learning_rate": 2.230471732972041e-06, + "loss": 0.413, + "step": 20876 + }, + { + "epoch": 7.71, + "learning_rate": 2.2304347826086957e-06, + "loss": 0.4272, + "step": 20877 + }, + { + "epoch": 7.71, + "learning_rate": 2.2303978322453503e-06, + "loss": 0.4108, + "step": 20878 + }, + { + "epoch": 7.71, + "learning_rate": 2.2303608818820054e-06, + "loss": 0.3943, + "step": 20879 + }, + { + "epoch": 7.71, + "learning_rate": 2.23032393151866e-06, + "loss": 0.4817, + "step": 20880 + }, + { + "epoch": 7.71, + "learning_rate": 2.2302869811553147e-06, + "loss": 0.488, + "step": 20881 + }, + { + "epoch": 7.71, + "learning_rate": 2.2302500307919697e-06, + "loss": 0.4348, + "step": 20882 + }, + { + "epoch": 7.71, + "learning_rate": 2.2302130804286244e-06, + "loss": 0.3391, + "step": 20883 + }, + { + "epoch": 7.71, + "learning_rate": 2.230176130065279e-06, + "loss": 0.4569, + "step": 20884 + }, + { + "epoch": 7.71, + "learning_rate": 2.2301391797019336e-06, + "loss": 0.3812, + "step": 20885 + }, + { + "epoch": 7.71, + "learning_rate": 2.2301022293385887e-06, + "loss": 0.4929, + "step": 20886 + }, + { + "epoch": 7.71, + "learning_rate": 2.2300652789752433e-06, + "loss": 0.4303, + "step": 20887 + }, + { + "epoch": 7.71, + "learning_rate": 2.2300283286118984e-06, + "loss": 0.4707, + "step": 20888 + }, + { + "epoch": 7.71, + "learning_rate": 2.229991378248553e-06, + "loss": 0.3676, + "step": 20889 + }, + { + "epoch": 7.71, + "learning_rate": 2.2299544278852077e-06, + "loss": 0.4426, + "step": 20890 + }, + { + "epoch": 7.71, + "learning_rate": 2.2299174775218623e-06, + "loss": 0.4705, + "step": 20891 + }, + { + "epoch": 7.71, + "learning_rate": 2.229880527158517e-06, + "loss": 0.3529, + "step": 20892 + }, + { + "epoch": 7.72, + "learning_rate": 2.229843576795172e-06, + "loss": 0.3996, + "step": 20893 + }, + { + "epoch": 7.72, + "learning_rate": 2.2298066264318266e-06, + "loss": 0.3862, + "step": 20894 + }, + { + "epoch": 7.72, + "learning_rate": 2.2297696760684813e-06, + "loss": 0.3128, + "step": 20895 + }, + { + "epoch": 7.72, + "learning_rate": 2.229732725705136e-06, + "loss": 0.4752, + "step": 20896 + }, + { + "epoch": 7.72, + "learning_rate": 2.229695775341791e-06, + "loss": 0.418, + "step": 20897 + }, + { + "epoch": 7.72, + "learning_rate": 2.2296588249784456e-06, + "loss": 0.409, + "step": 20898 + }, + { + "epoch": 7.72, + "learning_rate": 2.2296218746151007e-06, + "loss": 0.4136, + "step": 20899 + }, + { + "epoch": 7.72, + "learning_rate": 2.2295849242517553e-06, + "loss": 0.3307, + "step": 20900 + }, + { + "epoch": 7.72, + "learning_rate": 2.22954797388841e-06, + "loss": 0.4189, + "step": 20901 + }, + { + "epoch": 7.72, + "learning_rate": 2.2295110235250646e-06, + "loss": 0.4627, + "step": 20902 + }, + { + "epoch": 7.72, + "learning_rate": 2.229474073161719e-06, + "loss": 0.3896, + "step": 20903 + }, + { + "epoch": 7.72, + "learning_rate": 2.2294371227983743e-06, + "loss": 0.4909, + "step": 20904 + }, + { + "epoch": 7.72, + "learning_rate": 2.229400172435029e-06, + "loss": 0.4153, + "step": 20905 + }, + { + "epoch": 7.72, + "learning_rate": 2.229363222071684e-06, + "loss": 0.5147, + "step": 20906 + }, + { + "epoch": 7.72, + "learning_rate": 2.2293262717083386e-06, + "loss": 0.5694, + "step": 20907 + }, + { + "epoch": 7.72, + "learning_rate": 2.2292893213449932e-06, + "loss": 0.3841, + "step": 20908 + }, + { + "epoch": 7.72, + "learning_rate": 2.229252370981648e-06, + "loss": 0.4686, + "step": 20909 + }, + { + "epoch": 7.72, + "learning_rate": 2.229215420618303e-06, + "loss": 0.3233, + "step": 20910 + }, + { + "epoch": 7.72, + "learning_rate": 2.2291784702549576e-06, + "loss": 0.3735, + "step": 20911 + }, + { + "epoch": 7.72, + "learning_rate": 2.229141519891612e-06, + "loss": 0.5074, + "step": 20912 + }, + { + "epoch": 7.72, + "learning_rate": 2.2291045695282673e-06, + "loss": 0.5236, + "step": 20913 + }, + { + "epoch": 7.72, + "learning_rate": 2.229067619164922e-06, + "loss": 0.3455, + "step": 20914 + }, + { + "epoch": 7.72, + "learning_rate": 2.2290306688015765e-06, + "loss": 0.3786, + "step": 20915 + }, + { + "epoch": 7.72, + "learning_rate": 2.228993718438231e-06, + "loss": 0.3987, + "step": 20916 + }, + { + "epoch": 7.72, + "learning_rate": 2.2289567680748862e-06, + "loss": 0.2726, + "step": 20917 + }, + { + "epoch": 7.72, + "learning_rate": 2.228919817711541e-06, + "loss": 0.4099, + "step": 20918 + }, + { + "epoch": 7.72, + "learning_rate": 2.228882867348196e-06, + "loss": 0.2951, + "step": 20919 + }, + { + "epoch": 7.73, + "learning_rate": 2.2288459169848506e-06, + "loss": 0.3539, + "step": 20920 + }, + { + "epoch": 7.73, + "learning_rate": 2.228808966621505e-06, + "loss": 0.4349, + "step": 20921 + }, + { + "epoch": 7.73, + "learning_rate": 2.22877201625816e-06, + "loss": 0.3816, + "step": 20922 + }, + { + "epoch": 7.73, + "learning_rate": 2.2287350658948145e-06, + "loss": 0.4127, + "step": 20923 + }, + { + "epoch": 7.73, + "learning_rate": 2.2286981155314695e-06, + "loss": 0.5388, + "step": 20924 + }, + { + "epoch": 7.73, + "learning_rate": 2.228661165168124e-06, + "loss": 0.3824, + "step": 20925 + }, + { + "epoch": 7.73, + "learning_rate": 2.2286242148047792e-06, + "loss": 0.4311, + "step": 20926 + }, + { + "epoch": 7.73, + "learning_rate": 2.228587264441434e-06, + "loss": 0.3064, + "step": 20927 + }, + { + "epoch": 7.73, + "learning_rate": 2.2285503140780885e-06, + "loss": 0.5231, + "step": 20928 + }, + { + "epoch": 7.73, + "learning_rate": 2.228513363714743e-06, + "loss": 0.3299, + "step": 20929 + }, + { + "epoch": 7.73, + "learning_rate": 2.2284764133513978e-06, + "loss": 0.6183, + "step": 20930 + }, + { + "epoch": 7.73, + "learning_rate": 2.228439462988053e-06, + "loss": 0.3775, + "step": 20931 + }, + { + "epoch": 7.73, + "learning_rate": 2.2284025126247075e-06, + "loss": 0.3915, + "step": 20932 + }, + { + "epoch": 7.73, + "learning_rate": 2.2283655622613625e-06, + "loss": 0.5945, + "step": 20933 + }, + { + "epoch": 7.73, + "learning_rate": 2.2283286118980167e-06, + "loss": 0.4112, + "step": 20934 + }, + { + "epoch": 7.73, + "learning_rate": 2.228291661534672e-06, + "loss": 0.4565, + "step": 20935 + }, + { + "epoch": 7.73, + "learning_rate": 2.2282547111713264e-06, + "loss": 0.5925, + "step": 20936 + }, + { + "epoch": 7.73, + "learning_rate": 2.2282177608079815e-06, + "loss": 0.5071, + "step": 20937 + }, + { + "epoch": 7.73, + "learning_rate": 2.228180810444636e-06, + "loss": 0.3758, + "step": 20938 + }, + { + "epoch": 7.73, + "learning_rate": 2.228143860081291e-06, + "loss": 0.4891, + "step": 20939 + }, + { + "epoch": 7.73, + "learning_rate": 2.2281069097179454e-06, + "loss": 0.3176, + "step": 20940 + }, + { + "epoch": 7.73, + "learning_rate": 2.2280699593546e-06, + "loss": 0.3723, + "step": 20941 + }, + { + "epoch": 7.73, + "learning_rate": 2.228033008991255e-06, + "loss": 0.4215, + "step": 20942 + }, + { + "epoch": 7.73, + "learning_rate": 2.2279960586279097e-06, + "loss": 0.455, + "step": 20943 + }, + { + "epoch": 7.73, + "learning_rate": 2.227959108264565e-06, + "loss": 0.4245, + "step": 20944 + }, + { + "epoch": 7.73, + "learning_rate": 2.2279221579012194e-06, + "loss": 0.4383, + "step": 20945 + }, + { + "epoch": 7.73, + "learning_rate": 2.227885207537874e-06, + "loss": 0.463, + "step": 20946 + }, + { + "epoch": 7.74, + "learning_rate": 2.2278482571745287e-06, + "loss": 0.3796, + "step": 20947 + }, + { + "epoch": 7.74, + "learning_rate": 2.2278113068111838e-06, + "loss": 0.3701, + "step": 20948 + }, + { + "epoch": 7.74, + "learning_rate": 2.2277743564478384e-06, + "loss": 0.3746, + "step": 20949 + }, + { + "epoch": 7.74, + "learning_rate": 2.227737406084493e-06, + "loss": 0.3441, + "step": 20950 + }, + { + "epoch": 7.74, + "learning_rate": 2.227700455721148e-06, + "loss": 0.409, + "step": 20951 + }, + { + "epoch": 7.74, + "learning_rate": 2.2276635053578027e-06, + "loss": 0.4688, + "step": 20952 + }, + { + "epoch": 7.74, + "learning_rate": 2.2276265549944574e-06, + "loss": 0.4097, + "step": 20953 + }, + { + "epoch": 7.74, + "learning_rate": 2.227589604631112e-06, + "loss": 0.3652, + "step": 20954 + }, + { + "epoch": 7.74, + "learning_rate": 2.227552654267767e-06, + "loss": 0.4104, + "step": 20955 + }, + { + "epoch": 7.74, + "learning_rate": 2.2275157039044217e-06, + "loss": 0.3399, + "step": 20956 + }, + { + "epoch": 7.74, + "learning_rate": 2.2274787535410768e-06, + "loss": 0.5312, + "step": 20957 + }, + { + "epoch": 7.74, + "learning_rate": 2.2274418031777314e-06, + "loss": 0.4566, + "step": 20958 + }, + { + "epoch": 7.74, + "learning_rate": 2.227404852814386e-06, + "loss": 0.3958, + "step": 20959 + }, + { + "epoch": 7.74, + "learning_rate": 2.2273679024510407e-06, + "loss": 0.4158, + "step": 20960 + }, + { + "epoch": 7.74, + "learning_rate": 2.2273309520876953e-06, + "loss": 0.4158, + "step": 20961 + }, + { + "epoch": 7.74, + "learning_rate": 2.2272940017243504e-06, + "loss": 0.4461, + "step": 20962 + }, + { + "epoch": 7.74, + "learning_rate": 2.227257051361005e-06, + "loss": 0.4465, + "step": 20963 + }, + { + "epoch": 7.74, + "learning_rate": 2.22722010099766e-06, + "loss": 0.4418, + "step": 20964 + }, + { + "epoch": 7.74, + "learning_rate": 2.2271831506343147e-06, + "loss": 0.4579, + "step": 20965 + }, + { + "epoch": 7.74, + "learning_rate": 2.2271462002709693e-06, + "loss": 0.4777, + "step": 20966 + }, + { + "epoch": 7.74, + "learning_rate": 2.227109249907624e-06, + "loss": 0.4459, + "step": 20967 + }, + { + "epoch": 7.74, + "learning_rate": 2.227072299544279e-06, + "loss": 0.4592, + "step": 20968 + }, + { + "epoch": 7.74, + "learning_rate": 2.2270353491809337e-06, + "loss": 0.3577, + "step": 20969 + }, + { + "epoch": 7.74, + "learning_rate": 2.2269983988175883e-06, + "loss": 0.3587, + "step": 20970 + }, + { + "epoch": 7.74, + "learning_rate": 2.2269614484542434e-06, + "loss": 0.3898, + "step": 20971 + }, + { + "epoch": 7.74, + "learning_rate": 2.226924498090898e-06, + "loss": 0.4352, + "step": 20972 + }, + { + "epoch": 7.74, + "learning_rate": 2.2268875477275527e-06, + "loss": 0.4852, + "step": 20973 + }, + { + "epoch": 7.75, + "learning_rate": 2.2268505973642073e-06, + "loss": 0.436, + "step": 20974 + }, + { + "epoch": 7.75, + "learning_rate": 2.2268136470008623e-06, + "loss": 0.3217, + "step": 20975 + }, + { + "epoch": 7.75, + "learning_rate": 2.226776696637517e-06, + "loss": 0.3432, + "step": 20976 + }, + { + "epoch": 7.75, + "learning_rate": 2.226739746274172e-06, + "loss": 0.4529, + "step": 20977 + }, + { + "epoch": 7.75, + "learning_rate": 2.2267027959108267e-06, + "loss": 0.4575, + "step": 20978 + }, + { + "epoch": 7.75, + "learning_rate": 2.2266658455474813e-06, + "loss": 0.5172, + "step": 20979 + }, + { + "epoch": 7.75, + "learning_rate": 2.226628895184136e-06, + "loss": 0.4282, + "step": 20980 + }, + { + "epoch": 7.75, + "learning_rate": 2.2265919448207906e-06, + "loss": 0.3768, + "step": 20981 + }, + { + "epoch": 7.75, + "learning_rate": 2.2265549944574456e-06, + "loss": 0.41, + "step": 20982 + }, + { + "epoch": 7.75, + "learning_rate": 2.2265180440941003e-06, + "loss": 0.3977, + "step": 20983 + }, + { + "epoch": 7.75, + "learning_rate": 2.2264810937307553e-06, + "loss": 0.3567, + "step": 20984 + }, + { + "epoch": 7.75, + "learning_rate": 2.2264441433674096e-06, + "loss": 0.3919, + "step": 20985 + }, + { + "epoch": 7.75, + "learning_rate": 2.2264071930040646e-06, + "loss": 0.3981, + "step": 20986 + }, + { + "epoch": 7.75, + "learning_rate": 2.2263702426407193e-06, + "loss": 0.4563, + "step": 20987 + }, + { + "epoch": 7.75, + "learning_rate": 2.2263332922773743e-06, + "loss": 0.5131, + "step": 20988 + }, + { + "epoch": 7.75, + "learning_rate": 2.226296341914029e-06, + "loss": 0.5078, + "step": 20989 + }, + { + "epoch": 7.75, + "learning_rate": 2.2262593915506836e-06, + "loss": 0.3563, + "step": 20990 + }, + { + "epoch": 7.75, + "learning_rate": 2.2262224411873382e-06, + "loss": 0.5037, + "step": 20991 + }, + { + "epoch": 7.75, + "learning_rate": 2.226185490823993e-06, + "loss": 0.3541, + "step": 20992 + }, + { + "epoch": 7.75, + "learning_rate": 2.226148540460648e-06, + "loss": 0.482, + "step": 20993 + }, + { + "epoch": 7.75, + "learning_rate": 2.2261115900973026e-06, + "loss": 0.4353, + "step": 20994 + }, + { + "epoch": 7.75, + "learning_rate": 2.2260746397339576e-06, + "loss": 0.3382, + "step": 20995 + }, + { + "epoch": 7.75, + "learning_rate": 2.2260376893706123e-06, + "loss": 0.4534, + "step": 20996 + }, + { + "epoch": 7.75, + "learning_rate": 2.226000739007267e-06, + "loss": 0.4611, + "step": 20997 + }, + { + "epoch": 7.75, + "learning_rate": 2.2259637886439215e-06, + "loss": 0.4415, + "step": 20998 + }, + { + "epoch": 7.75, + "learning_rate": 2.2259268382805766e-06, + "loss": 0.3463, + "step": 20999 + }, + { + "epoch": 7.75, + "learning_rate": 2.2258898879172312e-06, + "loss": 0.3944, + "step": 21000 + }, + { + "epoch": 7.76, + "learning_rate": 2.225852937553886e-06, + "loss": 0.341, + "step": 21001 + }, + { + "epoch": 7.76, + "learning_rate": 2.225815987190541e-06, + "loss": 0.4258, + "step": 21002 + }, + { + "epoch": 7.76, + "learning_rate": 2.2257790368271956e-06, + "loss": 0.4063, + "step": 21003 + }, + { + "epoch": 7.76, + "learning_rate": 2.22574208646385e-06, + "loss": 0.3694, + "step": 21004 + }, + { + "epoch": 7.76, + "learning_rate": 2.225705136100505e-06, + "loss": 0.5523, + "step": 21005 + }, + { + "epoch": 7.76, + "learning_rate": 2.22566818573716e-06, + "loss": 0.4332, + "step": 21006 + }, + { + "epoch": 7.76, + "learning_rate": 2.2256312353738145e-06, + "loss": 0.3933, + "step": 21007 + }, + { + "epoch": 7.76, + "learning_rate": 2.2255942850104696e-06, + "loss": 0.2918, + "step": 21008 + }, + { + "epoch": 7.76, + "learning_rate": 2.2255573346471242e-06, + "loss": 0.4227, + "step": 21009 + }, + { + "epoch": 7.76, + "learning_rate": 2.225520384283779e-06, + "loss": 0.4412, + "step": 21010 + }, + { + "epoch": 7.76, + "learning_rate": 2.2254834339204335e-06, + "loss": 0.3408, + "step": 21011 + }, + { + "epoch": 7.76, + "learning_rate": 2.225446483557088e-06, + "loss": 0.3736, + "step": 21012 + }, + { + "epoch": 7.76, + "learning_rate": 2.225409533193743e-06, + "loss": 0.4749, + "step": 21013 + }, + { + "epoch": 7.76, + "learning_rate": 2.225372582830398e-06, + "loss": 0.318, + "step": 21014 + }, + { + "epoch": 7.76, + "learning_rate": 2.225335632467053e-06, + "loss": 0.3138, + "step": 21015 + }, + { + "epoch": 7.76, + "learning_rate": 2.2252986821037075e-06, + "loss": 0.4829, + "step": 21016 + }, + { + "epoch": 7.76, + "learning_rate": 2.225261731740362e-06, + "loss": 0.4893, + "step": 21017 + }, + { + "epoch": 7.76, + "learning_rate": 2.225224781377017e-06, + "loss": 0.3959, + "step": 21018 + }, + { + "epoch": 7.76, + "learning_rate": 2.225187831013672e-06, + "loss": 0.3538, + "step": 21019 + }, + { + "epoch": 7.76, + "learning_rate": 2.2251508806503265e-06, + "loss": 0.2696, + "step": 21020 + }, + { + "epoch": 7.76, + "learning_rate": 2.225113930286981e-06, + "loss": 0.4245, + "step": 21021 + }, + { + "epoch": 7.76, + "learning_rate": 2.225076979923636e-06, + "loss": 0.4477, + "step": 21022 + }, + { + "epoch": 7.76, + "learning_rate": 2.225040029560291e-06, + "loss": 0.4566, + "step": 21023 + }, + { + "epoch": 7.76, + "learning_rate": 2.2250030791969455e-06, + "loss": 0.4989, + "step": 21024 + }, + { + "epoch": 7.76, + "learning_rate": 2.2249661288336e-06, + "loss": 0.4005, + "step": 21025 + }, + { + "epoch": 7.76, + "learning_rate": 2.224929178470255e-06, + "loss": 0.4373, + "step": 21026 + }, + { + "epoch": 7.76, + "learning_rate": 2.22489222810691e-06, + "loss": 0.36, + "step": 21027 + }, + { + "epoch": 7.77, + "learning_rate": 2.224855277743565e-06, + "loss": 0.3859, + "step": 21028 + }, + { + "epoch": 7.77, + "learning_rate": 2.2248183273802195e-06, + "loss": 0.2994, + "step": 21029 + }, + { + "epoch": 7.77, + "learning_rate": 2.2247813770168737e-06, + "loss": 0.4204, + "step": 21030 + }, + { + "epoch": 7.77, + "learning_rate": 2.2247444266535288e-06, + "loss": 0.4343, + "step": 21031 + }, + { + "epoch": 7.77, + "learning_rate": 2.2247074762901834e-06, + "loss": 0.4778, + "step": 21032 + }, + { + "epoch": 7.77, + "learning_rate": 2.2246705259268385e-06, + "loss": 0.3586, + "step": 21033 + }, + { + "epoch": 7.77, + "learning_rate": 2.224633575563493e-06, + "loss": 0.3958, + "step": 21034 + }, + { + "epoch": 7.77, + "learning_rate": 2.224596625200148e-06, + "loss": 0.4268, + "step": 21035 + }, + { + "epoch": 7.77, + "learning_rate": 2.2245596748368024e-06, + "loss": 0.4289, + "step": 21036 + }, + { + "epoch": 7.77, + "learning_rate": 2.2245227244734574e-06, + "loss": 0.5096, + "step": 21037 + }, + { + "epoch": 7.77, + "learning_rate": 2.224485774110112e-06, + "loss": 0.4486, + "step": 21038 + }, + { + "epoch": 7.77, + "learning_rate": 2.224448823746767e-06, + "loss": 0.4439, + "step": 21039 + }, + { + "epoch": 7.77, + "learning_rate": 2.2244118733834218e-06, + "loss": 0.4732, + "step": 21040 + }, + { + "epoch": 7.77, + "learning_rate": 2.2243749230200764e-06, + "loss": 0.5586, + "step": 21041 + }, + { + "epoch": 7.77, + "learning_rate": 2.224337972656731e-06, + "loss": 0.5259, + "step": 21042 + }, + { + "epoch": 7.77, + "learning_rate": 2.2243010222933857e-06, + "loss": 0.4336, + "step": 21043 + }, + { + "epoch": 7.77, + "learning_rate": 2.2242640719300407e-06, + "loss": 0.3617, + "step": 21044 + }, + { + "epoch": 7.77, + "learning_rate": 2.2242271215666954e-06, + "loss": 0.3167, + "step": 21045 + }, + { + "epoch": 7.77, + "learning_rate": 2.2241901712033504e-06, + "loss": 0.4082, + "step": 21046 + }, + { + "epoch": 7.77, + "learning_rate": 2.224153220840005e-06, + "loss": 0.4326, + "step": 21047 + }, + { + "epoch": 7.77, + "learning_rate": 2.2241162704766597e-06, + "loss": 0.4286, + "step": 21048 + }, + { + "epoch": 7.77, + "learning_rate": 2.2240793201133143e-06, + "loss": 0.4891, + "step": 21049 + }, + { + "epoch": 7.77, + "learning_rate": 2.224042369749969e-06, + "loss": 0.499, + "step": 21050 + }, + { + "epoch": 7.77, + "learning_rate": 2.224005419386624e-06, + "loss": 0.4196, + "step": 21051 + }, + { + "epoch": 7.77, + "learning_rate": 2.2239684690232787e-06, + "loss": 0.3893, + "step": 21052 + }, + { + "epoch": 7.77, + "learning_rate": 2.2239315186599337e-06, + "loss": 0.46, + "step": 21053 + }, + { + "epoch": 7.77, + "learning_rate": 2.2238945682965884e-06, + "loss": 0.3201, + "step": 21054 + }, + { + "epoch": 7.78, + "learning_rate": 2.223857617933243e-06, + "loss": 0.5083, + "step": 21055 + }, + { + "epoch": 7.78, + "learning_rate": 2.2238206675698976e-06, + "loss": 0.4148, + "step": 21056 + }, + { + "epoch": 7.78, + "learning_rate": 2.2237837172065527e-06, + "loss": 0.3874, + "step": 21057 + }, + { + "epoch": 7.78, + "learning_rate": 2.2237467668432073e-06, + "loss": 0.3487, + "step": 21058 + }, + { + "epoch": 7.78, + "learning_rate": 2.2237098164798624e-06, + "loss": 0.5498, + "step": 21059 + }, + { + "epoch": 7.78, + "learning_rate": 2.223672866116517e-06, + "loss": 0.4513, + "step": 21060 + }, + { + "epoch": 7.78, + "learning_rate": 2.2236359157531717e-06, + "loss": 0.3537, + "step": 21061 + }, + { + "epoch": 7.78, + "learning_rate": 2.2235989653898263e-06, + "loss": 0.508, + "step": 21062 + }, + { + "epoch": 7.78, + "learning_rate": 2.223562015026481e-06, + "loss": 0.4305, + "step": 21063 + }, + { + "epoch": 7.78, + "learning_rate": 2.223525064663136e-06, + "loss": 0.5227, + "step": 21064 + }, + { + "epoch": 7.78, + "learning_rate": 2.2234881142997906e-06, + "loss": 0.4481, + "step": 21065 + }, + { + "epoch": 7.78, + "learning_rate": 2.2234511639364457e-06, + "loss": 0.5211, + "step": 21066 + }, + { + "epoch": 7.78, + "learning_rate": 2.2234142135731003e-06, + "loss": 0.4197, + "step": 21067 + }, + { + "epoch": 7.78, + "learning_rate": 2.223377263209755e-06, + "loss": 0.442, + "step": 21068 + }, + { + "epoch": 7.78, + "learning_rate": 2.2233403128464096e-06, + "loss": 0.407, + "step": 21069 + }, + { + "epoch": 7.78, + "learning_rate": 2.2233033624830642e-06, + "loss": 0.5568, + "step": 21070 + }, + { + "epoch": 7.78, + "learning_rate": 2.2232664121197193e-06, + "loss": 0.4386, + "step": 21071 + }, + { + "epoch": 7.78, + "learning_rate": 2.223229461756374e-06, + "loss": 0.4359, + "step": 21072 + }, + { + "epoch": 7.78, + "learning_rate": 2.223192511393029e-06, + "loss": 0.4509, + "step": 21073 + }, + { + "epoch": 7.78, + "learning_rate": 2.2231555610296836e-06, + "loss": 0.4292, + "step": 21074 + }, + { + "epoch": 7.78, + "learning_rate": 2.2231186106663383e-06, + "loss": 0.4386, + "step": 21075 + }, + { + "epoch": 7.78, + "learning_rate": 2.223081660302993e-06, + "loss": 0.3197, + "step": 21076 + }, + { + "epoch": 7.78, + "learning_rate": 2.223044709939648e-06, + "loss": 0.4512, + "step": 21077 + }, + { + "epoch": 7.78, + "learning_rate": 2.2230077595763026e-06, + "loss": 0.3262, + "step": 21078 + }, + { + "epoch": 7.78, + "learning_rate": 2.2229708092129577e-06, + "loss": 0.4367, + "step": 21079 + }, + { + "epoch": 7.78, + "learning_rate": 2.222933858849612e-06, + "loss": 0.4728, + "step": 21080 + }, + { + "epoch": 7.78, + "learning_rate": 2.2228969084862665e-06, + "loss": 0.4954, + "step": 21081 + }, + { + "epoch": 7.79, + "learning_rate": 2.2228599581229216e-06, + "loss": 0.4124, + "step": 21082 + }, + { + "epoch": 7.79, + "learning_rate": 2.222823007759576e-06, + "loss": 0.3837, + "step": 21083 + }, + { + "epoch": 7.79, + "learning_rate": 2.2227860573962313e-06, + "loss": 0.4575, + "step": 21084 + }, + { + "epoch": 7.79, + "learning_rate": 2.222749107032886e-06, + "loss": 0.3695, + "step": 21085 + }, + { + "epoch": 7.79, + "learning_rate": 2.2227121566695405e-06, + "loss": 0.4781, + "step": 21086 + }, + { + "epoch": 7.79, + "learning_rate": 2.222675206306195e-06, + "loss": 0.4139, + "step": 21087 + }, + { + "epoch": 7.79, + "learning_rate": 2.2226382559428502e-06, + "loss": 0.5022, + "step": 21088 + }, + { + "epoch": 7.79, + "learning_rate": 2.222601305579505e-06, + "loss": 0.4268, + "step": 21089 + }, + { + "epoch": 7.79, + "learning_rate": 2.2225643552161595e-06, + "loss": 0.4794, + "step": 21090 + }, + { + "epoch": 7.79, + "learning_rate": 2.2225274048528146e-06, + "loss": 0.3334, + "step": 21091 + }, + { + "epoch": 7.79, + "learning_rate": 2.222490454489469e-06, + "loss": 0.3657, + "step": 21092 + }, + { + "epoch": 7.79, + "learning_rate": 2.222453504126124e-06, + "loss": 0.3709, + "step": 21093 + }, + { + "epoch": 7.79, + "learning_rate": 2.2224165537627785e-06, + "loss": 0.4586, + "step": 21094 + }, + { + "epoch": 7.79, + "learning_rate": 2.2223796033994335e-06, + "loss": 0.4759, + "step": 21095 + }, + { + "epoch": 7.79, + "learning_rate": 2.222342653036088e-06, + "loss": 0.4006, + "step": 21096 + }, + { + "epoch": 7.79, + "learning_rate": 2.2223057026727432e-06, + "loss": 0.4348, + "step": 21097 + }, + { + "epoch": 7.79, + "learning_rate": 2.222268752309398e-06, + "loss": 0.392, + "step": 21098 + }, + { + "epoch": 7.79, + "learning_rate": 2.2222318019460525e-06, + "loss": 0.409, + "step": 21099 + }, + { + "epoch": 7.79, + "learning_rate": 2.222194851582707e-06, + "loss": 0.5461, + "step": 21100 + }, + { + "epoch": 7.79, + "learning_rate": 2.2221579012193618e-06, + "loss": 0.4775, + "step": 21101 + }, + { + "epoch": 7.79, + "learning_rate": 2.222120950856017e-06, + "loss": 0.4202, + "step": 21102 + }, + { + "epoch": 7.79, + "learning_rate": 2.2220840004926715e-06, + "loss": 0.3207, + "step": 21103 + }, + { + "epoch": 7.79, + "learning_rate": 2.2220470501293265e-06, + "loss": 0.5627, + "step": 21104 + }, + { + "epoch": 7.79, + "learning_rate": 2.222010099765981e-06, + "loss": 0.3871, + "step": 21105 + }, + { + "epoch": 7.79, + "learning_rate": 2.221973149402636e-06, + "loss": 0.3967, + "step": 21106 + }, + { + "epoch": 7.79, + "learning_rate": 2.2219361990392904e-06, + "loss": 0.4384, + "step": 21107 + }, + { + "epoch": 7.79, + "learning_rate": 2.2218992486759455e-06, + "loss": 0.3553, + "step": 21108 + }, + { + "epoch": 7.8, + "learning_rate": 2.2218622983126e-06, + "loss": 0.4482, + "step": 21109 + }, + { + "epoch": 7.8, + "learning_rate": 2.2218253479492548e-06, + "loss": 0.266, + "step": 21110 + }, + { + "epoch": 7.8, + "learning_rate": 2.22178839758591e-06, + "loss": 0.4874, + "step": 21111 + }, + { + "epoch": 7.8, + "learning_rate": 2.2217514472225645e-06, + "loss": 0.4713, + "step": 21112 + }, + { + "epoch": 7.8, + "learning_rate": 2.221714496859219e-06, + "loss": 0.4745, + "step": 21113 + }, + { + "epoch": 7.8, + "learning_rate": 2.2216775464958738e-06, + "loss": 0.4767, + "step": 21114 + }, + { + "epoch": 7.8, + "learning_rate": 2.221640596132529e-06, + "loss": 0.4622, + "step": 21115 + }, + { + "epoch": 7.8, + "learning_rate": 2.2216036457691834e-06, + "loss": 0.3788, + "step": 21116 + }, + { + "epoch": 7.8, + "learning_rate": 2.2215666954058385e-06, + "loss": 0.4926, + "step": 21117 + }, + { + "epoch": 7.8, + "learning_rate": 2.221529745042493e-06, + "loss": 0.4557, + "step": 21118 + }, + { + "epoch": 7.8, + "learning_rate": 2.2214927946791478e-06, + "loss": 0.4874, + "step": 21119 + }, + { + "epoch": 7.8, + "learning_rate": 2.2214558443158024e-06, + "loss": 0.3448, + "step": 21120 + }, + { + "epoch": 7.8, + "learning_rate": 2.221418893952457e-06, + "loss": 0.421, + "step": 21121 + }, + { + "epoch": 7.8, + "learning_rate": 2.221381943589112e-06, + "loss": 0.2969, + "step": 21122 + }, + { + "epoch": 7.8, + "learning_rate": 2.2213449932257667e-06, + "loss": 0.4762, + "step": 21123 + }, + { + "epoch": 7.8, + "learning_rate": 2.221308042862422e-06, + "loss": 0.3804, + "step": 21124 + }, + { + "epoch": 7.8, + "learning_rate": 2.221271092499076e-06, + "loss": 0.3785, + "step": 21125 + }, + { + "epoch": 7.8, + "learning_rate": 2.221234142135731e-06, + "loss": 0.4137, + "step": 21126 + }, + { + "epoch": 7.8, + "learning_rate": 2.2211971917723857e-06, + "loss": 0.2938, + "step": 21127 + }, + { + "epoch": 7.8, + "learning_rate": 2.2211602414090408e-06, + "loss": 0.3337, + "step": 21128 + }, + { + "epoch": 7.8, + "learning_rate": 2.2211232910456954e-06, + "loss": 0.4588, + "step": 21129 + }, + { + "epoch": 7.8, + "learning_rate": 2.22108634068235e-06, + "loss": 0.4084, + "step": 21130 + }, + { + "epoch": 7.8, + "learning_rate": 2.2210493903190047e-06, + "loss": 0.2979, + "step": 21131 + }, + { + "epoch": 7.8, + "learning_rate": 2.2210124399556593e-06, + "loss": 0.4741, + "step": 21132 + }, + { + "epoch": 7.8, + "learning_rate": 2.2209754895923144e-06, + "loss": 0.4853, + "step": 21133 + }, + { + "epoch": 7.8, + "learning_rate": 2.220938539228969e-06, + "loss": 0.4589, + "step": 21134 + }, + { + "epoch": 7.8, + "learning_rate": 2.220901588865624e-06, + "loss": 0.3661, + "step": 21135 + }, + { + "epoch": 7.81, + "learning_rate": 2.2208646385022787e-06, + "loss": 0.4084, + "step": 21136 + }, + { + "epoch": 7.81, + "learning_rate": 2.2208276881389334e-06, + "loss": 0.3764, + "step": 21137 + }, + { + "epoch": 7.81, + "learning_rate": 2.220790737775588e-06, + "loss": 0.4227, + "step": 21138 + }, + { + "epoch": 7.81, + "learning_rate": 2.220753787412243e-06, + "loss": 0.544, + "step": 21139 + }, + { + "epoch": 7.81, + "learning_rate": 2.2207168370488977e-06, + "loss": 0.3713, + "step": 21140 + }, + { + "epoch": 7.81, + "learning_rate": 2.2206798866855523e-06, + "loss": 0.3501, + "step": 21141 + }, + { + "epoch": 7.81, + "learning_rate": 2.2206429363222074e-06, + "loss": 0.5336, + "step": 21142 + }, + { + "epoch": 7.81, + "learning_rate": 2.220605985958862e-06, + "loss": 0.3878, + "step": 21143 + }, + { + "epoch": 7.81, + "learning_rate": 2.2205690355955167e-06, + "loss": 0.4343, + "step": 21144 + }, + { + "epoch": 7.81, + "learning_rate": 2.2205320852321713e-06, + "loss": 0.581, + "step": 21145 + }, + { + "epoch": 7.81, + "learning_rate": 2.2204951348688264e-06, + "loss": 0.3303, + "step": 21146 + }, + { + "epoch": 7.81, + "learning_rate": 2.220458184505481e-06, + "loss": 0.3264, + "step": 21147 + }, + { + "epoch": 7.81, + "learning_rate": 2.220421234142136e-06, + "loss": 0.3572, + "step": 21148 + }, + { + "epoch": 7.81, + "learning_rate": 2.2203842837787907e-06, + "loss": 0.474, + "step": 21149 + }, + { + "epoch": 7.81, + "learning_rate": 2.2203473334154453e-06, + "loss": 0.3685, + "step": 21150 + }, + { + "epoch": 7.81, + "learning_rate": 2.2203103830521e-06, + "loss": 0.5342, + "step": 21151 + }, + { + "epoch": 7.81, + "learning_rate": 2.2202734326887546e-06, + "loss": 0.4835, + "step": 21152 + }, + { + "epoch": 7.81, + "learning_rate": 2.2202364823254097e-06, + "loss": 0.3973, + "step": 21153 + }, + { + "epoch": 7.81, + "learning_rate": 2.2201995319620643e-06, + "loss": 0.4651, + "step": 21154 + }, + { + "epoch": 7.81, + "learning_rate": 2.2201625815987194e-06, + "loss": 0.4232, + "step": 21155 + }, + { + "epoch": 7.81, + "learning_rate": 2.220125631235374e-06, + "loss": 0.3746, + "step": 21156 + }, + { + "epoch": 7.81, + "learning_rate": 2.2200886808720286e-06, + "loss": 0.3472, + "step": 21157 + }, + { + "epoch": 7.81, + "learning_rate": 2.2200517305086833e-06, + "loss": 0.3736, + "step": 21158 + }, + { + "epoch": 7.81, + "learning_rate": 2.2200147801453383e-06, + "loss": 0.4361, + "step": 21159 + }, + { + "epoch": 7.81, + "learning_rate": 2.219977829781993e-06, + "loss": 0.3839, + "step": 21160 + }, + { + "epoch": 7.81, + "learning_rate": 2.2199408794186476e-06, + "loss": 0.3882, + "step": 21161 + }, + { + "epoch": 7.81, + "learning_rate": 2.2199039290553027e-06, + "loss": 0.471, + "step": 21162 + }, + { + "epoch": 7.81, + "learning_rate": 2.2198669786919573e-06, + "loss": 0.4807, + "step": 21163 + }, + { + "epoch": 7.82, + "learning_rate": 2.219830028328612e-06, + "loss": 0.4441, + "step": 21164 + }, + { + "epoch": 7.82, + "learning_rate": 2.2197930779652666e-06, + "loss": 0.3825, + "step": 21165 + }, + { + "epoch": 7.82, + "learning_rate": 2.2197561276019216e-06, + "loss": 0.3443, + "step": 21166 + }, + { + "epoch": 7.82, + "learning_rate": 2.2197191772385763e-06, + "loss": 0.5209, + "step": 21167 + }, + { + "epoch": 7.82, + "learning_rate": 2.2196822268752313e-06, + "loss": 0.4723, + "step": 21168 + }, + { + "epoch": 7.82, + "learning_rate": 2.219645276511886e-06, + "loss": 0.4282, + "step": 21169 + }, + { + "epoch": 7.82, + "learning_rate": 2.21960832614854e-06, + "loss": 0.5521, + "step": 21170 + }, + { + "epoch": 7.82, + "learning_rate": 2.2195713757851952e-06, + "loss": 0.3537, + "step": 21171 + }, + { + "epoch": 7.82, + "learning_rate": 2.21953442542185e-06, + "loss": 0.4467, + "step": 21172 + }, + { + "epoch": 7.82, + "learning_rate": 2.219497475058505e-06, + "loss": 0.4505, + "step": 21173 + }, + { + "epoch": 7.82, + "learning_rate": 2.2194605246951596e-06, + "loss": 0.4909, + "step": 21174 + }, + { + "epoch": 7.82, + "learning_rate": 2.2194235743318146e-06, + "loss": 0.3545, + "step": 21175 + }, + { + "epoch": 7.82, + "learning_rate": 2.219386623968469e-06, + "loss": 0.4211, + "step": 21176 + }, + { + "epoch": 7.82, + "learning_rate": 2.219349673605124e-06, + "loss": 0.5762, + "step": 21177 + }, + { + "epoch": 7.82, + "learning_rate": 2.2193127232417785e-06, + "loss": 0.4415, + "step": 21178 + }, + { + "epoch": 7.82, + "learning_rate": 2.219275772878433e-06, + "loss": 0.3838, + "step": 21179 + }, + { + "epoch": 7.82, + "learning_rate": 2.2192388225150882e-06, + "loss": 0.5686, + "step": 21180 + }, + { + "epoch": 7.82, + "learning_rate": 2.219201872151743e-06, + "loss": 0.3624, + "step": 21181 + }, + { + "epoch": 7.82, + "learning_rate": 2.2191649217883975e-06, + "loss": 0.5047, + "step": 21182 + }, + { + "epoch": 7.82, + "learning_rate": 2.219127971425052e-06, + "loss": 0.4169, + "step": 21183 + }, + { + "epoch": 7.82, + "learning_rate": 2.219091021061707e-06, + "loss": 0.5813, + "step": 21184 + }, + { + "epoch": 7.82, + "learning_rate": 2.219054070698362e-06, + "loss": 0.3871, + "step": 21185 + }, + { + "epoch": 7.82, + "learning_rate": 2.219017120335017e-06, + "loss": 0.3615, + "step": 21186 + }, + { + "epoch": 7.82, + "learning_rate": 2.2189801699716715e-06, + "loss": 0.4953, + "step": 21187 + }, + { + "epoch": 7.82, + "learning_rate": 2.218943219608326e-06, + "loss": 0.3894, + "step": 21188 + }, + { + "epoch": 7.82, + "learning_rate": 2.218906269244981e-06, + "loss": 0.4124, + "step": 21189 + }, + { + "epoch": 7.82, + "learning_rate": 2.2188693188816354e-06, + "loss": 0.313, + "step": 21190 + }, + { + "epoch": 7.83, + "learning_rate": 2.2188323685182905e-06, + "loss": 0.4168, + "step": 21191 + }, + { + "epoch": 7.83, + "learning_rate": 2.218795418154945e-06, + "loss": 0.3231, + "step": 21192 + }, + { + "epoch": 7.83, + "learning_rate": 2.2187584677916e-06, + "loss": 0.4132, + "step": 21193 + }, + { + "epoch": 7.83, + "learning_rate": 2.218721517428255e-06, + "loss": 0.5807, + "step": 21194 + }, + { + "epoch": 7.83, + "learning_rate": 2.2186845670649095e-06, + "loss": 0.5266, + "step": 21195 + }, + { + "epoch": 7.83, + "learning_rate": 2.218647616701564e-06, + "loss": 0.3117, + "step": 21196 + }, + { + "epoch": 7.83, + "learning_rate": 2.218610666338219e-06, + "loss": 0.3527, + "step": 21197 + }, + { + "epoch": 7.83, + "learning_rate": 2.218573715974874e-06, + "loss": 0.3839, + "step": 21198 + }, + { + "epoch": 7.83, + "learning_rate": 2.2185367656115284e-06, + "loss": 0.3481, + "step": 21199 + }, + { + "epoch": 7.83, + "learning_rate": 2.2184998152481835e-06, + "loss": 0.4517, + "step": 21200 + }, + { + "epoch": 7.83, + "learning_rate": 2.218462864884838e-06, + "loss": 0.4139, + "step": 21201 + }, + { + "epoch": 7.83, + "learning_rate": 2.2184259145214928e-06, + "loss": 0.3221, + "step": 21202 + }, + { + "epoch": 7.83, + "learning_rate": 2.2183889641581474e-06, + "loss": 0.4777, + "step": 21203 + }, + { + "epoch": 7.83, + "learning_rate": 2.2183520137948025e-06, + "loss": 0.4555, + "step": 21204 + }, + { + "epoch": 7.83, + "learning_rate": 2.218315063431457e-06, + "loss": 0.4116, + "step": 21205 + }, + { + "epoch": 7.83, + "learning_rate": 2.218278113068112e-06, + "loss": 0.4693, + "step": 21206 + }, + { + "epoch": 7.83, + "learning_rate": 2.218241162704767e-06, + "loss": 0.4138, + "step": 21207 + }, + { + "epoch": 7.83, + "learning_rate": 2.2182042123414214e-06, + "loss": 0.3533, + "step": 21208 + }, + { + "epoch": 7.83, + "learning_rate": 2.218167261978076e-06, + "loss": 0.4279, + "step": 21209 + }, + { + "epoch": 7.83, + "learning_rate": 2.2181303116147307e-06, + "loss": 0.3277, + "step": 21210 + }, + { + "epoch": 7.83, + "learning_rate": 2.2180933612513858e-06, + "loss": 0.3363, + "step": 21211 + }, + { + "epoch": 7.83, + "learning_rate": 2.2180564108880404e-06, + "loss": 0.3266, + "step": 21212 + }, + { + "epoch": 7.83, + "learning_rate": 2.2180194605246955e-06, + "loss": 0.4464, + "step": 21213 + }, + { + "epoch": 7.83, + "learning_rate": 2.21798251016135e-06, + "loss": 0.4053, + "step": 21214 + }, + { + "epoch": 7.83, + "learning_rate": 2.2179455597980047e-06, + "loss": 0.3693, + "step": 21215 + }, + { + "epoch": 7.83, + "learning_rate": 2.2179086094346594e-06, + "loss": 0.3483, + "step": 21216 + }, + { + "epoch": 7.83, + "learning_rate": 2.2178716590713144e-06, + "loss": 0.2978, + "step": 21217 + }, + { + "epoch": 7.84, + "learning_rate": 2.217834708707969e-06, + "loss": 0.3847, + "step": 21218 + }, + { + "epoch": 7.84, + "learning_rate": 2.2177977583446237e-06, + "loss": 0.4084, + "step": 21219 + }, + { + "epoch": 7.84, + "learning_rate": 2.2177608079812788e-06, + "loss": 0.3729, + "step": 21220 + }, + { + "epoch": 7.84, + "learning_rate": 2.217723857617933e-06, + "loss": 0.4331, + "step": 21221 + }, + { + "epoch": 7.84, + "learning_rate": 2.217686907254588e-06, + "loss": 0.402, + "step": 21222 + }, + { + "epoch": 7.84, + "learning_rate": 2.2176499568912427e-06, + "loss": 0.3771, + "step": 21223 + }, + { + "epoch": 7.84, + "learning_rate": 2.2176130065278977e-06, + "loss": 0.545, + "step": 21224 + }, + { + "epoch": 7.84, + "learning_rate": 2.2175760561645524e-06, + "loss": 0.5203, + "step": 21225 + }, + { + "epoch": 7.84, + "learning_rate": 2.2175391058012074e-06, + "loss": 0.2986, + "step": 21226 + }, + { + "epoch": 7.84, + "learning_rate": 2.2175021554378616e-06, + "loss": 0.3079, + "step": 21227 + }, + { + "epoch": 7.84, + "learning_rate": 2.2174652050745167e-06, + "loss": 0.5092, + "step": 21228 + }, + { + "epoch": 7.84, + "learning_rate": 2.2174282547111713e-06, + "loss": 0.376, + "step": 21229 + }, + { + "epoch": 7.84, + "learning_rate": 2.217391304347826e-06, + "loss": 0.3433, + "step": 21230 + }, + { + "epoch": 7.84, + "learning_rate": 2.217354353984481e-06, + "loss": 0.4914, + "step": 21231 + }, + { + "epoch": 7.84, + "learning_rate": 2.2173174036211357e-06, + "loss": 0.4414, + "step": 21232 + }, + { + "epoch": 7.84, + "learning_rate": 2.2172804532577903e-06, + "loss": 0.3479, + "step": 21233 + }, + { + "epoch": 7.84, + "learning_rate": 2.217243502894445e-06, + "loss": 0.5147, + "step": 21234 + }, + { + "epoch": 7.84, + "learning_rate": 2.2172065525311e-06, + "loss": 0.472, + "step": 21235 + }, + { + "epoch": 7.84, + "learning_rate": 2.2171696021677546e-06, + "loss": 0.3018, + "step": 21236 + }, + { + "epoch": 7.84, + "learning_rate": 2.2171326518044097e-06, + "loss": 0.43, + "step": 21237 + }, + { + "epoch": 7.84, + "learning_rate": 2.2170957014410643e-06, + "loss": 0.3888, + "step": 21238 + }, + { + "epoch": 7.84, + "learning_rate": 2.217058751077719e-06, + "loss": 0.4886, + "step": 21239 + }, + { + "epoch": 7.84, + "learning_rate": 2.2170218007143736e-06, + "loss": 0.4115, + "step": 21240 + }, + { + "epoch": 7.84, + "learning_rate": 2.2169848503510282e-06, + "loss": 0.36, + "step": 21241 + }, + { + "epoch": 7.84, + "learning_rate": 2.2169478999876833e-06, + "loss": 0.3807, + "step": 21242 + }, + { + "epoch": 7.84, + "learning_rate": 2.216910949624338e-06, + "loss": 0.4066, + "step": 21243 + }, + { + "epoch": 7.84, + "learning_rate": 2.216873999260993e-06, + "loss": 0.4292, + "step": 21244 + }, + { + "epoch": 7.85, + "learning_rate": 2.2168370488976476e-06, + "loss": 0.395, + "step": 21245 + }, + { + "epoch": 7.85, + "learning_rate": 2.2168000985343023e-06, + "loss": 0.495, + "step": 21246 + }, + { + "epoch": 7.85, + "learning_rate": 2.216763148170957e-06, + "loss": 0.3264, + "step": 21247 + }, + { + "epoch": 7.85, + "learning_rate": 2.216726197807612e-06, + "loss": 0.3716, + "step": 21248 + }, + { + "epoch": 7.85, + "learning_rate": 2.2166892474442666e-06, + "loss": 0.3899, + "step": 21249 + }, + { + "epoch": 7.85, + "learning_rate": 2.2166522970809212e-06, + "loss": 0.4219, + "step": 21250 + }, + { + "epoch": 7.85, + "learning_rate": 2.2166153467175763e-06, + "loss": 0.4905, + "step": 21251 + }, + { + "epoch": 7.85, + "learning_rate": 2.216578396354231e-06, + "loss": 0.4144, + "step": 21252 + }, + { + "epoch": 7.85, + "learning_rate": 2.2165414459908856e-06, + "loss": 0.4268, + "step": 21253 + }, + { + "epoch": 7.85, + "learning_rate": 2.2165044956275402e-06, + "loss": 0.4119, + "step": 21254 + }, + { + "epoch": 7.85, + "learning_rate": 2.2164675452641953e-06, + "loss": 0.4121, + "step": 21255 + }, + { + "epoch": 7.85, + "learning_rate": 2.21643059490085e-06, + "loss": 0.429, + "step": 21256 + }, + { + "epoch": 7.85, + "learning_rate": 2.216393644537505e-06, + "loss": 0.4153, + "step": 21257 + }, + { + "epoch": 7.85, + "learning_rate": 2.2163566941741596e-06, + "loss": 0.4098, + "step": 21258 + }, + { + "epoch": 7.85, + "learning_rate": 2.2163197438108142e-06, + "loss": 0.5063, + "step": 21259 + }, + { + "epoch": 7.85, + "learning_rate": 2.216282793447469e-06, + "loss": 0.5146, + "step": 21260 + }, + { + "epoch": 7.85, + "learning_rate": 2.2162458430841235e-06, + "loss": 0.4563, + "step": 21261 + }, + { + "epoch": 7.85, + "learning_rate": 2.2162088927207786e-06, + "loss": 0.3871, + "step": 21262 + }, + { + "epoch": 7.85, + "learning_rate": 2.2161719423574332e-06, + "loss": 0.3546, + "step": 21263 + }, + { + "epoch": 7.85, + "learning_rate": 2.2161349919940883e-06, + "loss": 0.3379, + "step": 21264 + }, + { + "epoch": 7.85, + "learning_rate": 2.2160980416307425e-06, + "loss": 0.5452, + "step": 21265 + }, + { + "epoch": 7.85, + "learning_rate": 2.2160610912673975e-06, + "loss": 0.3629, + "step": 21266 + }, + { + "epoch": 7.85, + "learning_rate": 2.216024140904052e-06, + "loss": 0.3912, + "step": 21267 + }, + { + "epoch": 7.85, + "learning_rate": 2.2159871905407072e-06, + "loss": 0.412, + "step": 21268 + }, + { + "epoch": 7.85, + "learning_rate": 2.215950240177362e-06, + "loss": 0.3802, + "step": 21269 + }, + { + "epoch": 7.85, + "learning_rate": 2.2159132898140165e-06, + "loss": 0.416, + "step": 21270 + }, + { + "epoch": 7.85, + "learning_rate": 2.215876339450671e-06, + "loss": 0.423, + "step": 21271 + }, + { + "epoch": 7.86, + "learning_rate": 2.215839389087326e-06, + "loss": 0.4001, + "step": 21272 + }, + { + "epoch": 7.86, + "learning_rate": 2.215802438723981e-06, + "loss": 0.3903, + "step": 21273 + }, + { + "epoch": 7.86, + "learning_rate": 2.2157654883606355e-06, + "loss": 0.3401, + "step": 21274 + }, + { + "epoch": 7.86, + "learning_rate": 2.2157285379972905e-06, + "loss": 0.5074, + "step": 21275 + }, + { + "epoch": 7.86, + "learning_rate": 2.215691587633945e-06, + "loss": 0.372, + "step": 21276 + }, + { + "epoch": 7.86, + "learning_rate": 2.2156546372706e-06, + "loss": 0.4182, + "step": 21277 + }, + { + "epoch": 7.86, + "learning_rate": 2.2156176869072545e-06, + "loss": 0.4467, + "step": 21278 + }, + { + "epoch": 7.86, + "learning_rate": 2.215580736543909e-06, + "loss": 0.3455, + "step": 21279 + }, + { + "epoch": 7.86, + "learning_rate": 2.215543786180564e-06, + "loss": 0.443, + "step": 21280 + }, + { + "epoch": 7.86, + "learning_rate": 2.215506835817219e-06, + "loss": 0.3682, + "step": 21281 + }, + { + "epoch": 7.86, + "learning_rate": 2.215469885453874e-06, + "loss": 0.4814, + "step": 21282 + }, + { + "epoch": 7.86, + "learning_rate": 2.2154329350905285e-06, + "loss": 0.3332, + "step": 21283 + }, + { + "epoch": 7.86, + "learning_rate": 2.215395984727183e-06, + "loss": 0.4641, + "step": 21284 + }, + { + "epoch": 7.86, + "learning_rate": 2.2153590343638378e-06, + "loss": 0.515, + "step": 21285 + }, + { + "epoch": 7.86, + "learning_rate": 2.215322084000493e-06, + "loss": 0.5369, + "step": 21286 + }, + { + "epoch": 7.86, + "learning_rate": 2.2152851336371475e-06, + "loss": 0.3858, + "step": 21287 + }, + { + "epoch": 7.86, + "learning_rate": 2.2152481832738025e-06, + "loss": 0.4373, + "step": 21288 + }, + { + "epoch": 7.86, + "learning_rate": 2.215211232910457e-06, + "loss": 0.3955, + "step": 21289 + }, + { + "epoch": 7.86, + "learning_rate": 2.2151742825471118e-06, + "loss": 0.4281, + "step": 21290 + }, + { + "epoch": 7.86, + "learning_rate": 2.2151373321837664e-06, + "loss": 0.4021, + "step": 21291 + }, + { + "epoch": 7.86, + "learning_rate": 2.215100381820421e-06, + "loss": 0.3754, + "step": 21292 + }, + { + "epoch": 7.86, + "learning_rate": 2.215063431457076e-06, + "loss": 0.447, + "step": 21293 + }, + { + "epoch": 7.86, + "learning_rate": 2.2150264810937308e-06, + "loss": 0.372, + "step": 21294 + }, + { + "epoch": 7.86, + "learning_rate": 2.214989530730386e-06, + "loss": 0.3448, + "step": 21295 + }, + { + "epoch": 7.86, + "learning_rate": 2.2149525803670405e-06, + "loss": 0.4508, + "step": 21296 + }, + { + "epoch": 7.86, + "learning_rate": 2.214915630003695e-06, + "loss": 0.4179, + "step": 21297 + }, + { + "epoch": 7.86, + "learning_rate": 2.2148786796403497e-06, + "loss": 0.3528, + "step": 21298 + }, + { + "epoch": 7.87, + "learning_rate": 2.2148417292770044e-06, + "loss": 0.42, + "step": 21299 + }, + { + "epoch": 7.87, + "learning_rate": 2.2148047789136594e-06, + "loss": 0.4484, + "step": 21300 + }, + { + "epoch": 7.87, + "learning_rate": 2.214767828550314e-06, + "loss": 0.3761, + "step": 21301 + }, + { + "epoch": 7.87, + "learning_rate": 2.214730878186969e-06, + "loss": 0.4548, + "step": 21302 + }, + { + "epoch": 7.87, + "learning_rate": 2.2146939278236238e-06, + "loss": 0.3991, + "step": 21303 + }, + { + "epoch": 7.87, + "learning_rate": 2.2146569774602784e-06, + "loss": 0.4614, + "step": 21304 + }, + { + "epoch": 7.87, + "learning_rate": 2.214620027096933e-06, + "loss": 0.4515, + "step": 21305 + }, + { + "epoch": 7.87, + "learning_rate": 2.214583076733588e-06, + "loss": 0.4319, + "step": 21306 + }, + { + "epoch": 7.87, + "learning_rate": 2.2145461263702427e-06, + "loss": 0.4754, + "step": 21307 + }, + { + "epoch": 7.87, + "learning_rate": 2.2145091760068978e-06, + "loss": 0.2886, + "step": 21308 + }, + { + "epoch": 7.87, + "learning_rate": 2.2144722256435524e-06, + "loss": 0.3189, + "step": 21309 + }, + { + "epoch": 7.87, + "learning_rate": 2.2144352752802066e-06, + "loss": 0.5157, + "step": 21310 + }, + { + "epoch": 7.87, + "learning_rate": 2.2143983249168617e-06, + "loss": 0.6092, + "step": 21311 + }, + { + "epoch": 7.87, + "learning_rate": 2.2143613745535163e-06, + "loss": 0.3562, + "step": 21312 + }, + { + "epoch": 7.87, + "learning_rate": 2.2143244241901714e-06, + "loss": 0.2864, + "step": 21313 + }, + { + "epoch": 7.87, + "learning_rate": 2.214287473826826e-06, + "loss": 0.382, + "step": 21314 + }, + { + "epoch": 7.87, + "learning_rate": 2.214250523463481e-06, + "loss": 0.3287, + "step": 21315 + }, + { + "epoch": 7.87, + "learning_rate": 2.2142135731001353e-06, + "loss": 0.4046, + "step": 21316 + }, + { + "epoch": 7.87, + "learning_rate": 2.2141766227367904e-06, + "loss": 0.3626, + "step": 21317 + }, + { + "epoch": 7.87, + "learning_rate": 2.214139672373445e-06, + "loss": 0.4566, + "step": 21318 + }, + { + "epoch": 7.87, + "learning_rate": 2.2141027220100996e-06, + "loss": 0.3344, + "step": 21319 + }, + { + "epoch": 7.87, + "learning_rate": 2.2140657716467547e-06, + "loss": 0.4075, + "step": 21320 + }, + { + "epoch": 7.87, + "learning_rate": 2.2140288212834093e-06, + "loss": 0.3579, + "step": 21321 + }, + { + "epoch": 7.87, + "learning_rate": 2.213991870920064e-06, + "loss": 0.3831, + "step": 21322 + }, + { + "epoch": 7.87, + "learning_rate": 2.2139549205567186e-06, + "loss": 0.42, + "step": 21323 + }, + { + "epoch": 7.87, + "learning_rate": 2.2139179701933737e-06, + "loss": 0.3136, + "step": 21324 + }, + { + "epoch": 7.87, + "learning_rate": 2.2138810198300283e-06, + "loss": 0.47, + "step": 21325 + }, + { + "epoch": 7.88, + "learning_rate": 2.2138440694666834e-06, + "loss": 0.2539, + "step": 21326 + }, + { + "epoch": 7.88, + "learning_rate": 2.213807119103338e-06, + "loss": 0.4107, + "step": 21327 + }, + { + "epoch": 7.88, + "learning_rate": 2.2137701687399926e-06, + "loss": 0.3798, + "step": 21328 + }, + { + "epoch": 7.88, + "learning_rate": 2.2137332183766473e-06, + "loss": 0.3649, + "step": 21329 + }, + { + "epoch": 7.88, + "learning_rate": 2.213696268013302e-06, + "loss": 0.4138, + "step": 21330 + }, + { + "epoch": 7.88, + "learning_rate": 2.213659317649957e-06, + "loss": 0.3847, + "step": 21331 + }, + { + "epoch": 7.88, + "learning_rate": 2.2136223672866116e-06, + "loss": 0.4383, + "step": 21332 + }, + { + "epoch": 7.88, + "learning_rate": 2.2135854169232667e-06, + "loss": 0.41, + "step": 21333 + }, + { + "epoch": 7.88, + "learning_rate": 2.2135484665599213e-06, + "loss": 0.505, + "step": 21334 + }, + { + "epoch": 7.88, + "learning_rate": 2.213511516196576e-06, + "loss": 0.3615, + "step": 21335 + }, + { + "epoch": 7.88, + "learning_rate": 2.2134745658332306e-06, + "loss": 0.2947, + "step": 21336 + }, + { + "epoch": 7.88, + "learning_rate": 2.2134376154698856e-06, + "loss": 0.4664, + "step": 21337 + }, + { + "epoch": 7.88, + "learning_rate": 2.2134006651065403e-06, + "loss": 0.3398, + "step": 21338 + }, + { + "epoch": 7.88, + "learning_rate": 2.213363714743195e-06, + "loss": 0.4486, + "step": 21339 + }, + { + "epoch": 7.88, + "learning_rate": 2.21332676437985e-06, + "loss": 0.4461, + "step": 21340 + }, + { + "epoch": 7.88, + "learning_rate": 2.2132898140165046e-06, + "loss": 0.5019, + "step": 21341 + }, + { + "epoch": 7.88, + "learning_rate": 2.2132528636531592e-06, + "loss": 0.569, + "step": 21342 + }, + { + "epoch": 7.88, + "learning_rate": 2.213215913289814e-06, + "loss": 0.3827, + "step": 21343 + }, + { + "epoch": 7.88, + "learning_rate": 2.213178962926469e-06, + "loss": 0.2984, + "step": 21344 + }, + { + "epoch": 7.88, + "learning_rate": 2.2131420125631236e-06, + "loss": 0.4879, + "step": 21345 + }, + { + "epoch": 7.88, + "learning_rate": 2.2131050621997786e-06, + "loss": 0.4122, + "step": 21346 + }, + { + "epoch": 7.88, + "learning_rate": 2.2130681118364333e-06, + "loss": 0.5293, + "step": 21347 + }, + { + "epoch": 7.88, + "learning_rate": 2.213031161473088e-06, + "loss": 0.4849, + "step": 21348 + }, + { + "epoch": 7.88, + "learning_rate": 2.2129942111097425e-06, + "loss": 0.4086, + "step": 21349 + }, + { + "epoch": 7.88, + "learning_rate": 2.212957260746397e-06, + "loss": 0.5997, + "step": 21350 + }, + { + "epoch": 7.88, + "learning_rate": 2.2129203103830522e-06, + "loss": 0.4095, + "step": 21351 + }, + { + "epoch": 7.88, + "learning_rate": 2.212883360019707e-06, + "loss": 0.3968, + "step": 21352 + }, + { + "epoch": 7.89, + "learning_rate": 2.212846409656362e-06, + "loss": 0.4175, + "step": 21353 + }, + { + "epoch": 7.89, + "learning_rate": 2.2128094592930166e-06, + "loss": 0.2241, + "step": 21354 + }, + { + "epoch": 7.89, + "learning_rate": 2.212772508929671e-06, + "loss": 0.3638, + "step": 21355 + }, + { + "epoch": 7.89, + "learning_rate": 2.212735558566326e-06, + "loss": 0.3382, + "step": 21356 + }, + { + "epoch": 7.89, + "learning_rate": 2.212698608202981e-06, + "loss": 0.3761, + "step": 21357 + }, + { + "epoch": 7.89, + "learning_rate": 2.2126616578396355e-06, + "loss": 0.3943, + "step": 21358 + }, + { + "epoch": 7.89, + "learning_rate": 2.21262470747629e-06, + "loss": 0.4579, + "step": 21359 + }, + { + "epoch": 7.89, + "learning_rate": 2.2125877571129452e-06, + "loss": 0.2762, + "step": 21360 + }, + { + "epoch": 7.89, + "learning_rate": 2.2125508067495994e-06, + "loss": 0.4754, + "step": 21361 + }, + { + "epoch": 7.89, + "learning_rate": 2.2125138563862545e-06, + "loss": 0.436, + "step": 21362 + }, + { + "epoch": 7.89, + "learning_rate": 2.212476906022909e-06, + "loss": 0.2997, + "step": 21363 + }, + { + "epoch": 7.89, + "learning_rate": 2.212439955659564e-06, + "loss": 0.546, + "step": 21364 + }, + { + "epoch": 7.89, + "learning_rate": 2.212403005296219e-06, + "loss": 0.2519, + "step": 21365 + }, + { + "epoch": 7.89, + "learning_rate": 2.212366054932874e-06, + "loss": 0.3551, + "step": 21366 + }, + { + "epoch": 7.89, + "learning_rate": 2.212329104569528e-06, + "loss": 0.3688, + "step": 21367 + }, + { + "epoch": 7.89, + "learning_rate": 2.212292154206183e-06, + "loss": 0.3056, + "step": 21368 + }, + { + "epoch": 7.89, + "learning_rate": 2.212255203842838e-06, + "loss": 0.3342, + "step": 21369 + }, + { + "epoch": 7.89, + "learning_rate": 2.2122182534794924e-06, + "loss": 0.396, + "step": 21370 + }, + { + "epoch": 7.89, + "learning_rate": 2.2121813031161475e-06, + "loss": 0.538, + "step": 21371 + }, + { + "epoch": 7.89, + "learning_rate": 2.212144352752802e-06, + "loss": 0.2946, + "step": 21372 + }, + { + "epoch": 7.89, + "learning_rate": 2.2121074023894568e-06, + "loss": 0.4647, + "step": 21373 + }, + { + "epoch": 7.89, + "learning_rate": 2.2120704520261114e-06, + "loss": 0.4334, + "step": 21374 + }, + { + "epoch": 7.89, + "learning_rate": 2.2120335016627665e-06, + "loss": 0.3301, + "step": 21375 + }, + { + "epoch": 7.89, + "learning_rate": 2.211996551299421e-06, + "loss": 0.4059, + "step": 21376 + }, + { + "epoch": 7.89, + "learning_rate": 2.211959600936076e-06, + "loss": 0.3797, + "step": 21377 + }, + { + "epoch": 7.89, + "learning_rate": 2.211922650572731e-06, + "loss": 0.3298, + "step": 21378 + }, + { + "epoch": 7.89, + "learning_rate": 2.2118857002093854e-06, + "loss": 0.3466, + "step": 21379 + }, + { + "epoch": 7.9, + "learning_rate": 2.21184874984604e-06, + "loss": 0.3446, + "step": 21380 + }, + { + "epoch": 7.9, + "learning_rate": 2.2118117994826947e-06, + "loss": 0.4108, + "step": 21381 + }, + { + "epoch": 7.9, + "learning_rate": 2.2117748491193498e-06, + "loss": 0.3647, + "step": 21382 + }, + { + "epoch": 7.9, + "learning_rate": 2.2117378987560044e-06, + "loss": 0.4004, + "step": 21383 + }, + { + "epoch": 7.9, + "learning_rate": 2.2117009483926595e-06, + "loss": 0.4356, + "step": 21384 + }, + { + "epoch": 7.9, + "learning_rate": 2.211663998029314e-06, + "loss": 0.4766, + "step": 21385 + }, + { + "epoch": 7.9, + "learning_rate": 2.2116270476659687e-06, + "loss": 0.4547, + "step": 21386 + }, + { + "epoch": 7.9, + "learning_rate": 2.2115900973026234e-06, + "loss": 0.391, + "step": 21387 + }, + { + "epoch": 7.9, + "learning_rate": 2.2115531469392784e-06, + "loss": 0.3667, + "step": 21388 + }, + { + "epoch": 7.9, + "learning_rate": 2.211516196575933e-06, + "loss": 0.368, + "step": 21389 + }, + { + "epoch": 7.9, + "learning_rate": 2.2114792462125877e-06, + "loss": 0.3109, + "step": 21390 + }, + { + "epoch": 7.9, + "learning_rate": 2.2114422958492428e-06, + "loss": 0.6181, + "step": 21391 + }, + { + "epoch": 7.9, + "learning_rate": 2.2114053454858974e-06, + "loss": 0.4672, + "step": 21392 + }, + { + "epoch": 7.9, + "learning_rate": 2.211368395122552e-06, + "loss": 0.4069, + "step": 21393 + }, + { + "epoch": 7.9, + "learning_rate": 2.2113314447592067e-06, + "loss": 0.5047, + "step": 21394 + }, + { + "epoch": 7.9, + "learning_rate": 2.2112944943958617e-06, + "loss": 0.3105, + "step": 21395 + }, + { + "epoch": 7.9, + "learning_rate": 2.2112575440325164e-06, + "loss": 0.4625, + "step": 21396 + }, + { + "epoch": 7.9, + "learning_rate": 2.2112205936691714e-06, + "loss": 0.3893, + "step": 21397 + }, + { + "epoch": 7.9, + "learning_rate": 2.211183643305826e-06, + "loss": 0.4201, + "step": 21398 + }, + { + "epoch": 7.9, + "learning_rate": 2.2111466929424807e-06, + "loss": 0.4482, + "step": 21399 + }, + { + "epoch": 7.9, + "learning_rate": 2.2111097425791353e-06, + "loss": 0.3351, + "step": 21400 + }, + { + "epoch": 7.9, + "learning_rate": 2.21107279221579e-06, + "loss": 0.5012, + "step": 21401 + }, + { + "epoch": 7.9, + "learning_rate": 2.211035841852445e-06, + "loss": 0.3034, + "step": 21402 + }, + { + "epoch": 7.9, + "learning_rate": 2.2109988914890997e-06, + "loss": 0.4667, + "step": 21403 + }, + { + "epoch": 7.9, + "learning_rate": 2.2109619411257547e-06, + "loss": 0.4151, + "step": 21404 + }, + { + "epoch": 7.9, + "learning_rate": 2.2109249907624094e-06, + "loss": 0.3597, + "step": 21405 + }, + { + "epoch": 7.9, + "learning_rate": 2.210888040399064e-06, + "loss": 0.3724, + "step": 21406 + }, + { + "epoch": 7.91, + "learning_rate": 2.2108510900357186e-06, + "loss": 0.4268, + "step": 21407 + }, + { + "epoch": 7.91, + "learning_rate": 2.2108141396723733e-06, + "loss": 0.3416, + "step": 21408 + }, + { + "epoch": 7.91, + "learning_rate": 2.2107771893090283e-06, + "loss": 0.4012, + "step": 21409 + }, + { + "epoch": 7.91, + "learning_rate": 2.210740238945683e-06, + "loss": 0.4124, + "step": 21410 + }, + { + "epoch": 7.91, + "learning_rate": 2.210703288582338e-06, + "loss": 0.4802, + "step": 21411 + }, + { + "epoch": 7.91, + "learning_rate": 2.2106663382189923e-06, + "loss": 0.4572, + "step": 21412 + }, + { + "epoch": 7.91, + "learning_rate": 2.2106293878556473e-06, + "loss": 0.3038, + "step": 21413 + }, + { + "epoch": 7.91, + "learning_rate": 2.210592437492302e-06, + "loss": 0.5066, + "step": 21414 + }, + { + "epoch": 7.91, + "learning_rate": 2.210555487128957e-06, + "loss": 0.4603, + "step": 21415 + }, + { + "epoch": 7.91, + "learning_rate": 2.2105185367656116e-06, + "loss": 0.2962, + "step": 21416 + }, + { + "epoch": 7.91, + "learning_rate": 2.2104815864022667e-06, + "loss": 0.4278, + "step": 21417 + }, + { + "epoch": 7.91, + "learning_rate": 2.210444636038921e-06, + "loss": 0.4342, + "step": 21418 + }, + { + "epoch": 7.91, + "learning_rate": 2.2104076856755756e-06, + "loss": 0.3631, + "step": 21419 + }, + { + "epoch": 7.91, + "learning_rate": 2.2103707353122306e-06, + "loss": 0.3158, + "step": 21420 + }, + { + "epoch": 7.91, + "learning_rate": 2.2103337849488853e-06, + "loss": 0.3152, + "step": 21421 + }, + { + "epoch": 7.91, + "learning_rate": 2.2102968345855403e-06, + "loss": 0.5481, + "step": 21422 + }, + { + "epoch": 7.91, + "learning_rate": 2.210259884222195e-06, + "loss": 0.3341, + "step": 21423 + }, + { + "epoch": 7.91, + "learning_rate": 2.2102229338588496e-06, + "loss": 0.3728, + "step": 21424 + }, + { + "epoch": 7.91, + "learning_rate": 2.2101859834955042e-06, + "loss": 0.426, + "step": 21425 + }, + { + "epoch": 7.91, + "learning_rate": 2.2101490331321593e-06, + "loss": 0.4994, + "step": 21426 + }, + { + "epoch": 7.91, + "learning_rate": 2.210112082768814e-06, + "loss": 0.46, + "step": 21427 + }, + { + "epoch": 7.91, + "learning_rate": 2.2100751324054686e-06, + "loss": 0.5261, + "step": 21428 + }, + { + "epoch": 7.91, + "learning_rate": 2.2100381820421236e-06, + "loss": 0.4442, + "step": 21429 + }, + { + "epoch": 7.91, + "learning_rate": 2.2100012316787783e-06, + "loss": 0.5024, + "step": 21430 + }, + { + "epoch": 7.91, + "learning_rate": 2.209964281315433e-06, + "loss": 0.3974, + "step": 21431 + }, + { + "epoch": 7.91, + "learning_rate": 2.2099273309520875e-06, + "loss": 0.3926, + "step": 21432 + }, + { + "epoch": 7.91, + "learning_rate": 2.2098903805887426e-06, + "loss": 0.3819, + "step": 21433 + }, + { + "epoch": 7.92, + "learning_rate": 2.2098534302253972e-06, + "loss": 0.3475, + "step": 21434 + }, + { + "epoch": 7.92, + "learning_rate": 2.2098164798620523e-06, + "loss": 0.2846, + "step": 21435 + }, + { + "epoch": 7.92, + "learning_rate": 2.209779529498707e-06, + "loss": 0.3658, + "step": 21436 + }, + { + "epoch": 7.92, + "learning_rate": 2.2097425791353616e-06, + "loss": 0.3613, + "step": 21437 + }, + { + "epoch": 7.92, + "learning_rate": 2.209705628772016e-06, + "loss": 0.5131, + "step": 21438 + }, + { + "epoch": 7.92, + "learning_rate": 2.209668678408671e-06, + "loss": 0.3196, + "step": 21439 + }, + { + "epoch": 7.92, + "learning_rate": 2.209631728045326e-06, + "loss": 0.3438, + "step": 21440 + }, + { + "epoch": 7.92, + "learning_rate": 2.2095947776819805e-06, + "loss": 0.3428, + "step": 21441 + }, + { + "epoch": 7.92, + "learning_rate": 2.2095578273186356e-06, + "loss": 0.3064, + "step": 21442 + }, + { + "epoch": 7.92, + "learning_rate": 2.2095208769552902e-06, + "loss": 0.4213, + "step": 21443 + }, + { + "epoch": 7.92, + "learning_rate": 2.209483926591945e-06, + "loss": 0.4518, + "step": 21444 + }, + { + "epoch": 7.92, + "learning_rate": 2.2094469762285995e-06, + "loss": 0.5156, + "step": 21445 + }, + { + "epoch": 7.92, + "learning_rate": 2.2094100258652546e-06, + "loss": 0.3765, + "step": 21446 + }, + { + "epoch": 7.92, + "learning_rate": 2.209373075501909e-06, + "loss": 0.3562, + "step": 21447 + }, + { + "epoch": 7.92, + "learning_rate": 2.209336125138564e-06, + "loss": 0.3198, + "step": 21448 + }, + { + "epoch": 7.92, + "learning_rate": 2.209299174775219e-06, + "loss": 0.3475, + "step": 21449 + }, + { + "epoch": 7.92, + "learning_rate": 2.2092622244118735e-06, + "loss": 0.5155, + "step": 21450 + }, + { + "epoch": 7.92, + "learning_rate": 2.209225274048528e-06, + "loss": 0.4582, + "step": 21451 + }, + { + "epoch": 7.92, + "learning_rate": 2.209188323685183e-06, + "loss": 0.4713, + "step": 21452 + }, + { + "epoch": 7.92, + "learning_rate": 2.209151373321838e-06, + "loss": 0.3974, + "step": 21453 + }, + { + "epoch": 7.92, + "learning_rate": 2.2091144229584925e-06, + "loss": 0.4411, + "step": 21454 + }, + { + "epoch": 7.92, + "learning_rate": 2.2090774725951476e-06, + "loss": 0.3984, + "step": 21455 + }, + { + "epoch": 7.92, + "learning_rate": 2.2090405222318018e-06, + "loss": 0.3144, + "step": 21456 + }, + { + "epoch": 7.92, + "learning_rate": 2.209003571868457e-06, + "loss": 0.4236, + "step": 21457 + }, + { + "epoch": 7.92, + "learning_rate": 2.2089666215051115e-06, + "loss": 0.3437, + "step": 21458 + }, + { + "epoch": 7.92, + "learning_rate": 2.208929671141766e-06, + "loss": 0.4462, + "step": 21459 + }, + { + "epoch": 7.92, + "learning_rate": 2.208892720778421e-06, + "loss": 0.4165, + "step": 21460 + }, + { + "epoch": 7.93, + "learning_rate": 2.208855770415076e-06, + "loss": 0.5678, + "step": 21461 + }, + { + "epoch": 7.93, + "learning_rate": 2.2088188200517304e-06, + "loss": 0.4473, + "step": 21462 + }, + { + "epoch": 7.93, + "learning_rate": 2.208781869688385e-06, + "loss": 0.3583, + "step": 21463 + }, + { + "epoch": 7.93, + "learning_rate": 2.20874491932504e-06, + "loss": 0.3956, + "step": 21464 + }, + { + "epoch": 7.93, + "learning_rate": 2.2087079689616948e-06, + "loss": 0.4168, + "step": 21465 + }, + { + "epoch": 7.93, + "learning_rate": 2.20867101859835e-06, + "loss": 0.3117, + "step": 21466 + }, + { + "epoch": 7.93, + "learning_rate": 2.2086340682350045e-06, + "loss": 0.3006, + "step": 21467 + }, + { + "epoch": 7.93, + "learning_rate": 2.208597117871659e-06, + "loss": 0.4699, + "step": 21468 + }, + { + "epoch": 7.93, + "learning_rate": 2.2085601675083137e-06, + "loss": 0.3719, + "step": 21469 + }, + { + "epoch": 7.93, + "learning_rate": 2.2085232171449684e-06, + "loss": 0.3435, + "step": 21470 + }, + { + "epoch": 7.93, + "learning_rate": 2.2084862667816234e-06, + "loss": 0.4442, + "step": 21471 + }, + { + "epoch": 7.93, + "learning_rate": 2.208449316418278e-06, + "loss": 0.3925, + "step": 21472 + }, + { + "epoch": 7.93, + "learning_rate": 2.208412366054933e-06, + "loss": 0.4686, + "step": 21473 + }, + { + "epoch": 7.93, + "learning_rate": 2.2083754156915878e-06, + "loss": 0.7272, + "step": 21474 + }, + { + "epoch": 7.93, + "learning_rate": 2.2083384653282424e-06, + "loss": 0.364, + "step": 21475 + }, + { + "epoch": 7.93, + "learning_rate": 2.208301514964897e-06, + "loss": 0.3442, + "step": 21476 + }, + { + "epoch": 7.93, + "learning_rate": 2.208264564601552e-06, + "loss": 0.5106, + "step": 21477 + }, + { + "epoch": 7.93, + "learning_rate": 2.2082276142382067e-06, + "loss": 0.434, + "step": 21478 + }, + { + "epoch": 7.93, + "learning_rate": 2.2081906638748614e-06, + "loss": 0.4884, + "step": 21479 + }, + { + "epoch": 7.93, + "learning_rate": 2.2081537135115164e-06, + "loss": 0.4256, + "step": 21480 + }, + { + "epoch": 7.93, + "learning_rate": 2.208116763148171e-06, + "loss": 0.5873, + "step": 21481 + }, + { + "epoch": 7.93, + "learning_rate": 2.2080798127848257e-06, + "loss": 0.374, + "step": 21482 + }, + { + "epoch": 7.93, + "learning_rate": 2.2080428624214803e-06, + "loss": 0.4809, + "step": 21483 + }, + { + "epoch": 7.93, + "learning_rate": 2.2080059120581354e-06, + "loss": 0.4007, + "step": 21484 + }, + { + "epoch": 7.93, + "learning_rate": 2.20796896169479e-06, + "loss": 0.4665, + "step": 21485 + }, + { + "epoch": 7.93, + "learning_rate": 2.207932011331445e-06, + "loss": 0.4412, + "step": 21486 + }, + { + "epoch": 7.93, + "learning_rate": 2.2078950609680997e-06, + "loss": 0.5533, + "step": 21487 + }, + { + "epoch": 7.94, + "learning_rate": 2.2078581106047544e-06, + "loss": 0.4312, + "step": 21488 + }, + { + "epoch": 7.94, + "learning_rate": 2.207821160241409e-06, + "loss": 0.453, + "step": 21489 + }, + { + "epoch": 7.94, + "learning_rate": 2.2077842098780636e-06, + "loss": 0.4444, + "step": 21490 + }, + { + "epoch": 7.94, + "learning_rate": 2.2077472595147187e-06, + "loss": 0.2485, + "step": 21491 + }, + { + "epoch": 7.94, + "learning_rate": 2.2077103091513733e-06, + "loss": 0.4627, + "step": 21492 + }, + { + "epoch": 7.94, + "learning_rate": 2.2076733587880284e-06, + "loss": 0.3534, + "step": 21493 + }, + { + "epoch": 7.94, + "learning_rate": 2.207636408424683e-06, + "loss": 0.3061, + "step": 21494 + }, + { + "epoch": 7.94, + "learning_rate": 2.2075994580613377e-06, + "loss": 0.3381, + "step": 21495 + }, + { + "epoch": 7.94, + "learning_rate": 2.2075625076979923e-06, + "loss": 0.4649, + "step": 21496 + }, + { + "epoch": 7.94, + "learning_rate": 2.2075255573346474e-06, + "loss": 0.3951, + "step": 21497 + }, + { + "epoch": 7.94, + "learning_rate": 2.207488606971302e-06, + "loss": 0.322, + "step": 21498 + }, + { + "epoch": 7.94, + "learning_rate": 2.2074516566079566e-06, + "loss": 0.4106, + "step": 21499 + }, + { + "epoch": 7.94, + "learning_rate": 2.2074147062446117e-06, + "loss": 0.5279, + "step": 21500 + }, + { + "epoch": 7.94, + "learning_rate": 2.207377755881266e-06, + "loss": 0.2746, + "step": 21501 + }, + { + "epoch": 7.94, + "learning_rate": 2.207340805517921e-06, + "loss": 0.453, + "step": 21502 + }, + { + "epoch": 7.94, + "learning_rate": 2.2073038551545756e-06, + "loss": 0.4269, + "step": 21503 + }, + { + "epoch": 7.94, + "learning_rate": 2.2072669047912307e-06, + "loss": 0.4102, + "step": 21504 + }, + { + "epoch": 7.94, + "learning_rate": 2.2072299544278853e-06, + "loss": 0.4689, + "step": 21505 + }, + { + "epoch": 7.94, + "learning_rate": 2.2071930040645404e-06, + "loss": 0.4364, + "step": 21506 + }, + { + "epoch": 7.94, + "learning_rate": 2.2071560537011946e-06, + "loss": 0.4243, + "step": 21507 + }, + { + "epoch": 7.94, + "learning_rate": 2.207119103337849e-06, + "loss": 0.4521, + "step": 21508 + }, + { + "epoch": 7.94, + "learning_rate": 2.2070821529745043e-06, + "loss": 0.3714, + "step": 21509 + }, + { + "epoch": 7.94, + "learning_rate": 2.207045202611159e-06, + "loss": 0.3888, + "step": 21510 + }, + { + "epoch": 7.94, + "learning_rate": 2.207008252247814e-06, + "loss": 0.3504, + "step": 21511 + }, + { + "epoch": 7.94, + "learning_rate": 2.2069713018844686e-06, + "loss": 0.3163, + "step": 21512 + }, + { + "epoch": 7.94, + "learning_rate": 2.2069343515211232e-06, + "loss": 0.4304, + "step": 21513 + }, + { + "epoch": 7.94, + "learning_rate": 2.206897401157778e-06, + "loss": 0.6213, + "step": 21514 + }, + { + "epoch": 7.94, + "learning_rate": 2.206860450794433e-06, + "loss": 0.4537, + "step": 21515 + }, + { + "epoch": 7.95, + "learning_rate": 2.2068235004310876e-06, + "loss": 0.4804, + "step": 21516 + }, + { + "epoch": 7.95, + "learning_rate": 2.2067865500677426e-06, + "loss": 0.4645, + "step": 21517 + }, + { + "epoch": 7.95, + "learning_rate": 2.2067495997043973e-06, + "loss": 0.2914, + "step": 21518 + }, + { + "epoch": 7.95, + "learning_rate": 2.206712649341052e-06, + "loss": 0.3466, + "step": 21519 + }, + { + "epoch": 7.95, + "learning_rate": 2.2066756989777065e-06, + "loss": 0.441, + "step": 21520 + }, + { + "epoch": 7.95, + "learning_rate": 2.206638748614361e-06, + "loss": 0.3938, + "step": 21521 + }, + { + "epoch": 7.95, + "learning_rate": 2.2066017982510162e-06, + "loss": 0.3693, + "step": 21522 + }, + { + "epoch": 7.95, + "learning_rate": 2.206564847887671e-06, + "loss": 0.3964, + "step": 21523 + }, + { + "epoch": 7.95, + "learning_rate": 2.206527897524326e-06, + "loss": 0.3699, + "step": 21524 + }, + { + "epoch": 7.95, + "learning_rate": 2.2064909471609806e-06, + "loss": 0.3575, + "step": 21525 + }, + { + "epoch": 7.95, + "learning_rate": 2.206453996797635e-06, + "loss": 0.4251, + "step": 21526 + }, + { + "epoch": 7.95, + "learning_rate": 2.20641704643429e-06, + "loss": 0.3882, + "step": 21527 + }, + { + "epoch": 7.95, + "learning_rate": 2.2063800960709445e-06, + "loss": 0.4691, + "step": 21528 + }, + { + "epoch": 7.95, + "learning_rate": 2.2063431457075995e-06, + "loss": 0.5218, + "step": 21529 + }, + { + "epoch": 7.95, + "learning_rate": 2.206306195344254e-06, + "loss": 0.4176, + "step": 21530 + }, + { + "epoch": 7.95, + "learning_rate": 2.2062692449809092e-06, + "loss": 0.3415, + "step": 21531 + }, + { + "epoch": 7.95, + "learning_rate": 2.206232294617564e-06, + "loss": 0.4441, + "step": 21532 + }, + { + "epoch": 7.95, + "learning_rate": 2.2061953442542185e-06, + "loss": 0.5574, + "step": 21533 + }, + { + "epoch": 7.95, + "learning_rate": 2.206158393890873e-06, + "loss": 0.3228, + "step": 21534 + }, + { + "epoch": 7.95, + "learning_rate": 2.206121443527528e-06, + "loss": 0.3172, + "step": 21535 + }, + { + "epoch": 7.95, + "learning_rate": 2.206084493164183e-06, + "loss": 0.4467, + "step": 21536 + }, + { + "epoch": 7.95, + "learning_rate": 2.206047542800838e-06, + "loss": 0.3699, + "step": 21537 + }, + { + "epoch": 7.95, + "learning_rate": 2.2060105924374925e-06, + "loss": 0.3857, + "step": 21538 + }, + { + "epoch": 7.95, + "learning_rate": 2.205973642074147e-06, + "loss": 0.3481, + "step": 21539 + }, + { + "epoch": 7.95, + "learning_rate": 2.205936691710802e-06, + "loss": 0.4831, + "step": 21540 + }, + { + "epoch": 7.95, + "learning_rate": 2.2058997413474564e-06, + "loss": 0.3867, + "step": 21541 + }, + { + "epoch": 7.95, + "learning_rate": 2.2058627909841115e-06, + "loss": 0.5831, + "step": 21542 + }, + { + "epoch": 7.96, + "learning_rate": 2.205825840620766e-06, + "loss": 0.3769, + "step": 21543 + }, + { + "epoch": 7.96, + "learning_rate": 2.205788890257421e-06, + "loss": 0.3308, + "step": 21544 + }, + { + "epoch": 7.96, + "learning_rate": 2.205751939894076e-06, + "loss": 0.465, + "step": 21545 + }, + { + "epoch": 7.96, + "learning_rate": 2.2057149895307305e-06, + "loss": 0.4354, + "step": 21546 + }, + { + "epoch": 7.96, + "learning_rate": 2.205678039167385e-06, + "loss": 0.4003, + "step": 21547 + }, + { + "epoch": 7.96, + "learning_rate": 2.2056410888040398e-06, + "loss": 0.4154, + "step": 21548 + }, + { + "epoch": 7.96, + "learning_rate": 2.205604138440695e-06, + "loss": 0.4445, + "step": 21549 + }, + { + "epoch": 7.96, + "learning_rate": 2.2055671880773494e-06, + "loss": 0.3541, + "step": 21550 + }, + { + "epoch": 7.96, + "learning_rate": 2.2055302377140045e-06, + "loss": 0.3092, + "step": 21551 + }, + { + "epoch": 7.96, + "learning_rate": 2.2054932873506587e-06, + "loss": 0.4241, + "step": 21552 + }, + { + "epoch": 7.96, + "learning_rate": 2.2054563369873138e-06, + "loss": 0.3461, + "step": 21553 + }, + { + "epoch": 7.96, + "learning_rate": 2.2054193866239684e-06, + "loss": 0.4749, + "step": 21554 + }, + { + "epoch": 7.96, + "learning_rate": 2.2053824362606235e-06, + "loss": 0.4277, + "step": 21555 + }, + { + "epoch": 7.96, + "learning_rate": 2.205345485897278e-06, + "loss": 0.3445, + "step": 21556 + }, + { + "epoch": 7.96, + "learning_rate": 2.205308535533933e-06, + "loss": 0.2611, + "step": 21557 + }, + { + "epoch": 7.96, + "learning_rate": 2.2052715851705874e-06, + "loss": 0.3616, + "step": 21558 + }, + { + "epoch": 7.96, + "learning_rate": 2.205234634807242e-06, + "loss": 0.4915, + "step": 21559 + }, + { + "epoch": 7.96, + "learning_rate": 2.205197684443897e-06, + "loss": 0.3993, + "step": 21560 + }, + { + "epoch": 7.96, + "learning_rate": 2.2051607340805517e-06, + "loss": 0.5312, + "step": 21561 + }, + { + "epoch": 7.96, + "learning_rate": 2.2051237837172068e-06, + "loss": 0.5314, + "step": 21562 + }, + { + "epoch": 7.96, + "learning_rate": 2.2050868333538614e-06, + "loss": 0.5025, + "step": 21563 + }, + { + "epoch": 7.96, + "learning_rate": 2.205049882990516e-06, + "loss": 0.524, + "step": 21564 + }, + { + "epoch": 7.96, + "learning_rate": 2.2050129326271707e-06, + "loss": 0.4173, + "step": 21565 + }, + { + "epoch": 7.96, + "learning_rate": 2.2049759822638257e-06, + "loss": 0.4244, + "step": 21566 + }, + { + "epoch": 7.96, + "learning_rate": 2.2049390319004804e-06, + "loss": 0.4258, + "step": 21567 + }, + { + "epoch": 7.96, + "learning_rate": 2.204902081537135e-06, + "loss": 0.5148, + "step": 21568 + }, + { + "epoch": 7.96, + "learning_rate": 2.20486513117379e-06, + "loss": 0.3902, + "step": 21569 + }, + { + "epoch": 7.97, + "learning_rate": 2.2048281808104447e-06, + "loss": 0.3169, + "step": 21570 + }, + { + "epoch": 7.97, + "learning_rate": 2.2047912304470994e-06, + "loss": 0.3962, + "step": 21571 + }, + { + "epoch": 7.97, + "learning_rate": 2.204754280083754e-06, + "loss": 0.3989, + "step": 21572 + }, + { + "epoch": 7.97, + "learning_rate": 2.204717329720409e-06, + "loss": 0.4262, + "step": 21573 + }, + { + "epoch": 7.97, + "learning_rate": 2.2046803793570637e-06, + "loss": 0.3863, + "step": 21574 + }, + { + "epoch": 7.97, + "learning_rate": 2.2046434289937187e-06, + "loss": 0.4161, + "step": 21575 + }, + { + "epoch": 7.97, + "learning_rate": 2.2046064786303734e-06, + "loss": 0.4482, + "step": 21576 + }, + { + "epoch": 7.97, + "learning_rate": 2.204569528267028e-06, + "loss": 0.5034, + "step": 21577 + }, + { + "epoch": 7.97, + "learning_rate": 2.2045325779036827e-06, + "loss": 0.4175, + "step": 21578 + }, + { + "epoch": 7.97, + "learning_rate": 2.2044956275403373e-06, + "loss": 0.473, + "step": 21579 + }, + { + "epoch": 7.97, + "learning_rate": 2.2044586771769924e-06, + "loss": 0.4449, + "step": 21580 + }, + { + "epoch": 7.97, + "learning_rate": 2.204421726813647e-06, + "loss": 0.4332, + "step": 21581 + }, + { + "epoch": 7.97, + "learning_rate": 2.204384776450302e-06, + "loss": 0.2414, + "step": 21582 + }, + { + "epoch": 7.97, + "learning_rate": 2.2043478260869567e-06, + "loss": 0.3381, + "step": 21583 + }, + { + "epoch": 7.97, + "learning_rate": 2.2043108757236113e-06, + "loss": 0.42, + "step": 21584 + }, + { + "epoch": 7.97, + "learning_rate": 2.204273925360266e-06, + "loss": 0.3798, + "step": 21585 + }, + { + "epoch": 7.97, + "learning_rate": 2.204236974996921e-06, + "loss": 0.3478, + "step": 21586 + }, + { + "epoch": 7.97, + "learning_rate": 2.2042000246335757e-06, + "loss": 0.5312, + "step": 21587 + }, + { + "epoch": 7.97, + "learning_rate": 2.2041630742702303e-06, + "loss": 0.3759, + "step": 21588 + }, + { + "epoch": 7.97, + "learning_rate": 2.2041261239068854e-06, + "loss": 0.4246, + "step": 21589 + }, + { + "epoch": 7.97, + "learning_rate": 2.20408917354354e-06, + "loss": 0.3925, + "step": 21590 + }, + { + "epoch": 7.97, + "learning_rate": 2.2040522231801946e-06, + "loss": 0.3281, + "step": 21591 + }, + { + "epoch": 7.97, + "learning_rate": 2.2040152728168493e-06, + "loss": 0.3806, + "step": 21592 + }, + { + "epoch": 7.97, + "learning_rate": 2.2039783224535043e-06, + "loss": 0.3403, + "step": 21593 + }, + { + "epoch": 7.97, + "learning_rate": 2.203941372090159e-06, + "loss": 0.4335, + "step": 21594 + }, + { + "epoch": 7.97, + "learning_rate": 2.203904421726814e-06, + "loss": 0.4065, + "step": 21595 + }, + { + "epoch": 7.97, + "learning_rate": 2.2038674713634687e-06, + "loss": 0.41, + "step": 21596 + }, + { + "epoch": 7.98, + "learning_rate": 2.2038305210001233e-06, + "loss": 0.5067, + "step": 21597 + }, + { + "epoch": 7.98, + "learning_rate": 2.203793570636778e-06, + "loss": 0.3395, + "step": 21598 + }, + { + "epoch": 7.98, + "learning_rate": 2.2037566202734326e-06, + "loss": 0.578, + "step": 21599 + }, + { + "epoch": 7.98, + "learning_rate": 2.2037196699100876e-06, + "loss": 0.6164, + "step": 21600 + }, + { + "epoch": 7.98, + "learning_rate": 2.2036827195467423e-06, + "loss": 0.3755, + "step": 21601 + }, + { + "epoch": 7.98, + "learning_rate": 2.2036457691833973e-06, + "loss": 0.4242, + "step": 21602 + }, + { + "epoch": 7.98, + "learning_rate": 2.2036088188200515e-06, + "loss": 0.4503, + "step": 21603 + }, + { + "epoch": 7.98, + "learning_rate": 2.2035718684567066e-06, + "loss": 0.4095, + "step": 21604 + }, + { + "epoch": 7.98, + "learning_rate": 2.2035349180933612e-06, + "loss": 0.4522, + "step": 21605 + }, + { + "epoch": 7.98, + "learning_rate": 2.2034979677300163e-06, + "loss": 0.3188, + "step": 21606 + }, + { + "epoch": 7.98, + "learning_rate": 2.203461017366671e-06, + "loss": 0.4461, + "step": 21607 + }, + { + "epoch": 7.98, + "learning_rate": 2.2034240670033256e-06, + "loss": 0.3823, + "step": 21608 + }, + { + "epoch": 7.98, + "learning_rate": 2.20338711663998e-06, + "loss": 0.333, + "step": 21609 + }, + { + "epoch": 7.98, + "learning_rate": 2.203350166276635e-06, + "loss": 0.4433, + "step": 21610 + }, + { + "epoch": 7.98, + "learning_rate": 2.20331321591329e-06, + "loss": 0.3018, + "step": 21611 + }, + { + "epoch": 7.98, + "learning_rate": 2.2032762655499445e-06, + "loss": 0.3956, + "step": 21612 + }, + { + "epoch": 7.98, + "learning_rate": 2.2032393151865996e-06, + "loss": 0.4309, + "step": 21613 + }, + { + "epoch": 7.98, + "learning_rate": 2.2032023648232542e-06, + "loss": 0.3603, + "step": 21614 + }, + { + "epoch": 7.98, + "learning_rate": 2.203165414459909e-06, + "loss": 0.4445, + "step": 21615 + }, + { + "epoch": 7.98, + "learning_rate": 2.2031284640965635e-06, + "loss": 0.4294, + "step": 21616 + }, + { + "epoch": 7.98, + "learning_rate": 2.2030915137332186e-06, + "loss": 0.4839, + "step": 21617 + }, + { + "epoch": 7.98, + "learning_rate": 2.203054563369873e-06, + "loss": 0.3655, + "step": 21618 + }, + { + "epoch": 7.98, + "learning_rate": 2.203017613006528e-06, + "loss": 0.3874, + "step": 21619 + }, + { + "epoch": 7.98, + "learning_rate": 2.202980662643183e-06, + "loss": 0.3715, + "step": 21620 + }, + { + "epoch": 7.98, + "learning_rate": 2.2029437122798375e-06, + "loss": 0.3263, + "step": 21621 + }, + { + "epoch": 7.98, + "learning_rate": 2.202906761916492e-06, + "loss": 0.6193, + "step": 21622 + }, + { + "epoch": 7.98, + "learning_rate": 2.202869811553147e-06, + "loss": 0.2509, + "step": 21623 + }, + { + "epoch": 7.99, + "learning_rate": 2.202832861189802e-06, + "loss": 0.3341, + "step": 21624 + }, + { + "epoch": 7.99, + "learning_rate": 2.2027959108264565e-06, + "loss": 0.4352, + "step": 21625 + }, + { + "epoch": 7.99, + "learning_rate": 2.2027589604631116e-06, + "loss": 0.4762, + "step": 21626 + }, + { + "epoch": 7.99, + "learning_rate": 2.202722010099766e-06, + "loss": 0.4739, + "step": 21627 + }, + { + "epoch": 7.99, + "learning_rate": 2.202685059736421e-06, + "loss": 0.3443, + "step": 21628 + }, + { + "epoch": 7.99, + "learning_rate": 2.2026481093730755e-06, + "loss": 0.3907, + "step": 21629 + }, + { + "epoch": 7.99, + "learning_rate": 2.20261115900973e-06, + "loss": 0.4658, + "step": 21630 + }, + { + "epoch": 7.99, + "learning_rate": 2.202574208646385e-06, + "loss": 0.3767, + "step": 21631 + }, + { + "epoch": 7.99, + "learning_rate": 2.20253725828304e-06, + "loss": 0.4395, + "step": 21632 + }, + { + "epoch": 7.99, + "learning_rate": 2.202500307919695e-06, + "loss": 0.4282, + "step": 21633 + }, + { + "epoch": 7.99, + "learning_rate": 2.2024633575563495e-06, + "loss": 0.5693, + "step": 21634 + }, + { + "epoch": 7.99, + "learning_rate": 2.202426407193004e-06, + "loss": 0.4014, + "step": 21635 + }, + { + "epoch": 7.99, + "learning_rate": 2.2023894568296588e-06, + "loss": 0.576, + "step": 21636 + }, + { + "epoch": 7.99, + "learning_rate": 2.2023525064663134e-06, + "loss": 0.3548, + "step": 21637 + }, + { + "epoch": 7.99, + "learning_rate": 2.2023155561029685e-06, + "loss": 0.4024, + "step": 21638 + }, + { + "epoch": 7.99, + "learning_rate": 2.202278605739623e-06, + "loss": 0.4898, + "step": 21639 + }, + { + "epoch": 7.99, + "learning_rate": 2.202241655376278e-06, + "loss": 0.3556, + "step": 21640 + }, + { + "epoch": 7.99, + "learning_rate": 2.2022047050129324e-06, + "loss": 0.3999, + "step": 21641 + }, + { + "epoch": 7.99, + "learning_rate": 2.2021677546495874e-06, + "loss": 0.3064, + "step": 21642 + }, + { + "epoch": 7.99, + "learning_rate": 2.202130804286242e-06, + "loss": 0.576, + "step": 21643 + }, + { + "epoch": 7.99, + "learning_rate": 2.202093853922897e-06, + "loss": 0.296, + "step": 21644 + }, + { + "epoch": 7.99, + "learning_rate": 2.2020569035595518e-06, + "loss": 0.4389, + "step": 21645 + }, + { + "epoch": 7.99, + "learning_rate": 2.202019953196207e-06, + "loss": 0.355, + "step": 21646 + }, + { + "epoch": 7.99, + "learning_rate": 2.201983002832861e-06, + "loss": 0.5823, + "step": 21647 + }, + { + "epoch": 7.99, + "learning_rate": 2.2019460524695157e-06, + "loss": 0.5091, + "step": 21648 + }, + { + "epoch": 7.99, + "learning_rate": 2.2019091021061707e-06, + "loss": 0.3579, + "step": 21649 + }, + { + "epoch": 7.99, + "learning_rate": 2.2018721517428254e-06, + "loss": 0.4989, + "step": 21650 + }, + { + "epoch": 8.0, + "learning_rate": 2.2018352013794804e-06, + "loss": 0.449, + "step": 21651 + }, + { + "epoch": 8.0, + "learning_rate": 2.201798251016135e-06, + "loss": 0.3872, + "step": 21652 + }, + { + "epoch": 8.0, + "learning_rate": 2.2017613006527897e-06, + "loss": 0.3182, + "step": 21653 + }, + { + "epoch": 8.0, + "learning_rate": 2.2017243502894443e-06, + "loss": 0.3586, + "step": 21654 + }, + { + "epoch": 8.0, + "learning_rate": 2.2016873999260994e-06, + "loss": 0.3577, + "step": 21655 + }, + { + "epoch": 8.0, + "learning_rate": 2.201650449562754e-06, + "loss": 0.3338, + "step": 21656 + }, + { + "epoch": 8.0, + "learning_rate": 2.2016134991994087e-06, + "loss": 0.3612, + "step": 21657 + }, + { + "epoch": 8.0, + "learning_rate": 2.2015765488360637e-06, + "loss": 0.5382, + "step": 21658 + }, + { + "epoch": 8.0, + "learning_rate": 2.2015395984727184e-06, + "loss": 0.36, + "step": 21659 + }, + { + "epoch": 8.0, + "learning_rate": 2.201502648109373e-06, + "loss": 0.4332, + "step": 21660 + }, + { + "epoch": 8.0, + "learning_rate": 2.2014656977460276e-06, + "loss": 0.374, + "step": 21661 + }, + { + "epoch": 8.0, + "learning_rate": 2.2014287473826827e-06, + "loss": 0.3666, + "step": 21662 + }, + { + "epoch": 8.0, + "learning_rate": 2.2013917970193373e-06, + "loss": 0.4375, + "step": 21663 + }, + { + "epoch": 8.0, + "learning_rate": 2.2013548466559924e-06, + "loss": 1.2677, + "step": 21664 + }, + { + "epoch": 8.0, + "eval_accuracy": 0.7874128457311723, + "eval_loss": 0.4615348279476166, + "eval_runtime": 25.4125, + "eval_samples_per_second": 1704.436, + "eval_steps_per_second": 26.64, + "step": 21664 + }, + { + "epoch": 8.0, + "learning_rate": 2.201317896292647e-06, + "loss": 0.4466, + "step": 21665 + }, + { + "epoch": 8.0, + "learning_rate": 2.2012809459293017e-06, + "loss": 0.3973, + "step": 21666 + }, + { + "epoch": 8.0, + "learning_rate": 2.2012439955659563e-06, + "loss": 0.3937, + "step": 21667 + }, + { + "epoch": 8.0, + "learning_rate": 2.201207045202611e-06, + "loss": 0.4261, + "step": 21668 + }, + { + "epoch": 8.0, + "learning_rate": 2.201170094839266e-06, + "loss": 0.3261, + "step": 21669 + }, + { + "epoch": 8.0, + "learning_rate": 2.2011331444759206e-06, + "loss": 0.4438, + "step": 21670 + }, + { + "epoch": 8.0, + "learning_rate": 2.2010961941125757e-06, + "loss": 0.4746, + "step": 21671 + }, + { + "epoch": 8.0, + "learning_rate": 2.2010592437492303e-06, + "loss": 0.363, + "step": 21672 + }, + { + "epoch": 8.0, + "learning_rate": 2.201022293385885e-06, + "loss": 0.4252, + "step": 21673 + }, + { + "epoch": 8.0, + "learning_rate": 2.2009853430225396e-06, + "loss": 0.4264, + "step": 21674 + }, + { + "epoch": 8.0, + "learning_rate": 2.2009483926591947e-06, + "loss": 0.3555, + "step": 21675 + }, + { + "epoch": 8.0, + "learning_rate": 2.2009114422958493e-06, + "loss": 0.4833, + "step": 21676 + }, + { + "epoch": 8.0, + "learning_rate": 2.200874491932504e-06, + "loss": 0.4892, + "step": 21677 + }, + { + "epoch": 8.01, + "learning_rate": 2.200837541569159e-06, + "loss": 0.4143, + "step": 21678 + }, + { + "epoch": 8.01, + "learning_rate": 2.2008005912058136e-06, + "loss": 0.3175, + "step": 21679 + }, + { + "epoch": 8.01, + "learning_rate": 2.2007636408424683e-06, + "loss": 0.4144, + "step": 21680 + }, + { + "epoch": 8.01, + "learning_rate": 2.200726690479123e-06, + "loss": 0.3885, + "step": 21681 + }, + { + "epoch": 8.01, + "learning_rate": 2.200689740115778e-06, + "loss": 0.3452, + "step": 21682 + }, + { + "epoch": 8.01, + "learning_rate": 2.2006527897524326e-06, + "loss": 0.4524, + "step": 21683 + }, + { + "epoch": 8.01, + "learning_rate": 2.2006158393890877e-06, + "loss": 0.5055, + "step": 21684 + }, + { + "epoch": 8.01, + "learning_rate": 2.2005788890257423e-06, + "loss": 0.2804, + "step": 21685 + }, + { + "epoch": 8.01, + "learning_rate": 2.200541938662397e-06, + "loss": 0.394, + "step": 21686 + }, + { + "epoch": 8.01, + "learning_rate": 2.2005049882990516e-06, + "loss": 0.2648, + "step": 21687 + }, + { + "epoch": 8.01, + "learning_rate": 2.2004680379357062e-06, + "loss": 0.3964, + "step": 21688 + }, + { + "epoch": 8.01, + "learning_rate": 2.2004310875723613e-06, + "loss": 0.3337, + "step": 21689 + }, + { + "epoch": 8.01, + "learning_rate": 2.200394137209016e-06, + "loss": 0.4721, + "step": 21690 + }, + { + "epoch": 8.01, + "learning_rate": 2.200357186845671e-06, + "loss": 0.5362, + "step": 21691 + }, + { + "epoch": 8.01, + "learning_rate": 2.200320236482325e-06, + "loss": 0.4785, + "step": 21692 + }, + { + "epoch": 8.01, + "learning_rate": 2.2002832861189802e-06, + "loss": 0.339, + "step": 21693 + }, + { + "epoch": 8.01, + "learning_rate": 2.200246335755635e-06, + "loss": 0.4109, + "step": 21694 + }, + { + "epoch": 8.01, + "learning_rate": 2.20020938539229e-06, + "loss": 0.4165, + "step": 21695 + }, + { + "epoch": 8.01, + "learning_rate": 2.2001724350289446e-06, + "loss": 0.3582, + "step": 21696 + }, + { + "epoch": 8.01, + "learning_rate": 2.2001354846655992e-06, + "loss": 0.3942, + "step": 21697 + }, + { + "epoch": 8.01, + "learning_rate": 2.200098534302254e-06, + "loss": 0.3178, + "step": 21698 + }, + { + "epoch": 8.01, + "learning_rate": 2.2000615839389085e-06, + "loss": 0.4017, + "step": 21699 + }, + { + "epoch": 8.01, + "learning_rate": 2.2000246335755635e-06, + "loss": 0.4638, + "step": 21700 + }, + { + "epoch": 8.01, + "learning_rate": 2.199987683212218e-06, + "loss": 0.3926, + "step": 21701 + }, + { + "epoch": 8.01, + "learning_rate": 2.1999507328488732e-06, + "loss": 0.396, + "step": 21702 + }, + { + "epoch": 8.01, + "learning_rate": 2.199913782485528e-06, + "loss": 0.5017, + "step": 21703 + }, + { + "epoch": 8.01, + "learning_rate": 2.1998768321221825e-06, + "loss": 0.3477, + "step": 21704 + }, + { + "epoch": 8.02, + "learning_rate": 2.199839881758837e-06, + "loss": 0.3775, + "step": 21705 + }, + { + "epoch": 8.02, + "learning_rate": 2.1998029313954922e-06, + "loss": 0.4323, + "step": 21706 + }, + { + "epoch": 8.02, + "learning_rate": 2.199765981032147e-06, + "loss": 0.4346, + "step": 21707 + }, + { + "epoch": 8.02, + "learning_rate": 2.1997290306688015e-06, + "loss": 0.2282, + "step": 21708 + }, + { + "epoch": 8.02, + "learning_rate": 2.1996920803054565e-06, + "loss": 0.4684, + "step": 21709 + }, + { + "epoch": 8.02, + "learning_rate": 2.199655129942111e-06, + "loss": 0.3593, + "step": 21710 + }, + { + "epoch": 8.02, + "learning_rate": 2.199618179578766e-06, + "loss": 0.408, + "step": 21711 + }, + { + "epoch": 8.02, + "learning_rate": 2.1995812292154205e-06, + "loss": 0.4324, + "step": 21712 + }, + { + "epoch": 8.02, + "learning_rate": 2.1995442788520755e-06, + "loss": 0.4489, + "step": 21713 + }, + { + "epoch": 8.02, + "learning_rate": 2.19950732848873e-06, + "loss": 0.2737, + "step": 21714 + }, + { + "epoch": 8.02, + "learning_rate": 2.199470378125385e-06, + "loss": 0.3425, + "step": 21715 + }, + { + "epoch": 8.02, + "learning_rate": 2.19943342776204e-06, + "loss": 0.3056, + "step": 21716 + }, + { + "epoch": 8.02, + "learning_rate": 2.1993964773986945e-06, + "loss": 0.4103, + "step": 21717 + }, + { + "epoch": 8.02, + "learning_rate": 2.199359527035349e-06, + "loss": 0.4462, + "step": 21718 + }, + { + "epoch": 8.02, + "learning_rate": 2.1993225766720038e-06, + "loss": 0.4167, + "step": 21719 + }, + { + "epoch": 8.02, + "learning_rate": 2.199285626308659e-06, + "loss": 0.5248, + "step": 21720 + }, + { + "epoch": 8.02, + "learning_rate": 2.1992486759453135e-06, + "loss": 0.3619, + "step": 21721 + }, + { + "epoch": 8.02, + "learning_rate": 2.1992117255819685e-06, + "loss": 0.3455, + "step": 21722 + }, + { + "epoch": 8.02, + "learning_rate": 2.199174775218623e-06, + "loss": 0.3605, + "step": 21723 + }, + { + "epoch": 8.02, + "learning_rate": 2.1991378248552778e-06, + "loss": 0.3581, + "step": 21724 + }, + { + "epoch": 8.02, + "learning_rate": 2.1991008744919324e-06, + "loss": 0.4309, + "step": 21725 + }, + { + "epoch": 8.02, + "learning_rate": 2.1990639241285875e-06, + "loss": 0.4519, + "step": 21726 + }, + { + "epoch": 8.02, + "learning_rate": 2.199026973765242e-06, + "loss": 0.4399, + "step": 21727 + }, + { + "epoch": 8.02, + "learning_rate": 2.1989900234018968e-06, + "loss": 0.388, + "step": 21728 + }, + { + "epoch": 8.02, + "learning_rate": 2.198953073038552e-06, + "loss": 0.5034, + "step": 21729 + }, + { + "epoch": 8.02, + "learning_rate": 2.1989161226752065e-06, + "loss": 0.5411, + "step": 21730 + }, + { + "epoch": 8.02, + "learning_rate": 2.198879172311861e-06, + "loss": 0.2732, + "step": 21731 + }, + { + "epoch": 8.03, + "learning_rate": 2.1988422219485157e-06, + "loss": 0.3319, + "step": 21732 + }, + { + "epoch": 8.03, + "learning_rate": 2.1988052715851708e-06, + "loss": 0.2613, + "step": 21733 + }, + { + "epoch": 8.03, + "learning_rate": 2.1987683212218254e-06, + "loss": 0.4437, + "step": 21734 + }, + { + "epoch": 8.03, + "learning_rate": 2.1987313708584805e-06, + "loss": 0.3853, + "step": 21735 + }, + { + "epoch": 8.03, + "learning_rate": 2.198694420495135e-06, + "loss": 0.4152, + "step": 21736 + }, + { + "epoch": 8.03, + "learning_rate": 2.1986574701317893e-06, + "loss": 0.4126, + "step": 21737 + }, + { + "epoch": 8.03, + "learning_rate": 2.1986205197684444e-06, + "loss": 0.3378, + "step": 21738 + }, + { + "epoch": 8.03, + "learning_rate": 2.198583569405099e-06, + "loss": 0.344, + "step": 21739 + }, + { + "epoch": 8.03, + "learning_rate": 2.198546619041754e-06, + "loss": 0.4841, + "step": 21740 + }, + { + "epoch": 8.03, + "learning_rate": 2.1985096686784087e-06, + "loss": 0.4135, + "step": 21741 + }, + { + "epoch": 8.03, + "learning_rate": 2.1984727183150638e-06, + "loss": 0.2981, + "step": 21742 + }, + { + "epoch": 8.03, + "learning_rate": 2.198435767951718e-06, + "loss": 0.2989, + "step": 21743 + }, + { + "epoch": 8.03, + "learning_rate": 2.198398817588373e-06, + "loss": 0.3832, + "step": 21744 + }, + { + "epoch": 8.03, + "learning_rate": 2.1983618672250277e-06, + "loss": 0.4892, + "step": 21745 + }, + { + "epoch": 8.03, + "learning_rate": 2.1983249168616828e-06, + "loss": 0.5381, + "step": 21746 + }, + { + "epoch": 8.03, + "learning_rate": 2.1982879664983374e-06, + "loss": 0.3988, + "step": 21747 + }, + { + "epoch": 8.03, + "learning_rate": 2.198251016134992e-06, + "loss": 0.3178, + "step": 21748 + }, + { + "epoch": 8.03, + "learning_rate": 2.1982140657716467e-06, + "loss": 0.2761, + "step": 21749 + }, + { + "epoch": 8.03, + "learning_rate": 2.1981771154083013e-06, + "loss": 0.4039, + "step": 21750 + }, + { + "epoch": 8.03, + "learning_rate": 2.1981401650449564e-06, + "loss": 0.422, + "step": 21751 + }, + { + "epoch": 8.03, + "learning_rate": 2.198103214681611e-06, + "loss": 0.3687, + "step": 21752 + }, + { + "epoch": 8.03, + "learning_rate": 2.198066264318266e-06, + "loss": 0.4137, + "step": 21753 + }, + { + "epoch": 8.03, + "learning_rate": 2.1980293139549207e-06, + "loss": 0.3703, + "step": 21754 + }, + { + "epoch": 8.03, + "learning_rate": 2.1979923635915753e-06, + "loss": 0.3973, + "step": 21755 + }, + { + "epoch": 8.03, + "learning_rate": 2.19795541322823e-06, + "loss": 0.4075, + "step": 21756 + }, + { + "epoch": 8.03, + "learning_rate": 2.1979184628648846e-06, + "loss": 0.4007, + "step": 21757 + }, + { + "epoch": 8.03, + "learning_rate": 2.1978815125015397e-06, + "loss": 0.3276, + "step": 21758 + }, + { + "epoch": 8.04, + "learning_rate": 2.1978445621381943e-06, + "loss": 0.3236, + "step": 21759 + }, + { + "epoch": 8.04, + "learning_rate": 2.1978076117748494e-06, + "loss": 0.3129, + "step": 21760 + }, + { + "epoch": 8.04, + "learning_rate": 2.197770661411504e-06, + "loss": 0.4491, + "step": 21761 + }, + { + "epoch": 8.04, + "learning_rate": 2.1977337110481586e-06, + "loss": 0.4687, + "step": 21762 + }, + { + "epoch": 8.04, + "learning_rate": 2.1976967606848133e-06, + "loss": 0.3506, + "step": 21763 + }, + { + "epoch": 8.04, + "learning_rate": 2.1976598103214683e-06, + "loss": 0.2937, + "step": 21764 + }, + { + "epoch": 8.04, + "learning_rate": 2.197622859958123e-06, + "loss": 0.5701, + "step": 21765 + }, + { + "epoch": 8.04, + "learning_rate": 2.197585909594778e-06, + "loss": 0.4203, + "step": 21766 + }, + { + "epoch": 8.04, + "learning_rate": 2.1975489592314327e-06, + "loss": 0.4116, + "step": 21767 + }, + { + "epoch": 8.04, + "learning_rate": 2.1975120088680873e-06, + "loss": 0.3821, + "step": 21768 + }, + { + "epoch": 8.04, + "learning_rate": 2.197475058504742e-06, + "loss": 0.3651, + "step": 21769 + }, + { + "epoch": 8.04, + "learning_rate": 2.1974381081413966e-06, + "loss": 0.3437, + "step": 21770 + }, + { + "epoch": 8.04, + "learning_rate": 2.1974011577780516e-06, + "loss": 0.4742, + "step": 21771 + }, + { + "epoch": 8.04, + "learning_rate": 2.1973642074147063e-06, + "loss": 0.3178, + "step": 21772 + }, + { + "epoch": 8.04, + "learning_rate": 2.1973272570513613e-06, + "loss": 0.5208, + "step": 21773 + }, + { + "epoch": 8.04, + "learning_rate": 2.197290306688016e-06, + "loss": 0.3251, + "step": 21774 + }, + { + "epoch": 8.04, + "learning_rate": 2.1972533563246706e-06, + "loss": 0.3367, + "step": 21775 + }, + { + "epoch": 8.04, + "learning_rate": 2.1972164059613252e-06, + "loss": 0.3265, + "step": 21776 + }, + { + "epoch": 8.04, + "learning_rate": 2.19717945559798e-06, + "loss": 0.2995, + "step": 21777 + }, + { + "epoch": 8.04, + "learning_rate": 2.197142505234635e-06, + "loss": 0.3435, + "step": 21778 + }, + { + "epoch": 8.04, + "learning_rate": 2.1971055548712896e-06, + "loss": 0.5064, + "step": 21779 + }, + { + "epoch": 8.04, + "learning_rate": 2.1970686045079446e-06, + "loss": 0.3985, + "step": 21780 + }, + { + "epoch": 8.04, + "learning_rate": 2.1970316541445993e-06, + "loss": 0.5065, + "step": 21781 + }, + { + "epoch": 8.04, + "learning_rate": 2.196994703781254e-06, + "loss": 0.3402, + "step": 21782 + }, + { + "epoch": 8.04, + "learning_rate": 2.1969577534179085e-06, + "loss": 0.3942, + "step": 21783 + }, + { + "epoch": 8.04, + "learning_rate": 2.1969208030545636e-06, + "loss": 0.4688, + "step": 21784 + }, + { + "epoch": 8.04, + "learning_rate": 2.1968838526912182e-06, + "loss": 0.5149, + "step": 21785 + }, + { + "epoch": 8.05, + "learning_rate": 2.1968469023278733e-06, + "loss": 0.3052, + "step": 21786 + }, + { + "epoch": 8.05, + "learning_rate": 2.196809951964528e-06, + "loss": 0.3241, + "step": 21787 + }, + { + "epoch": 8.05, + "learning_rate": 2.196773001601182e-06, + "loss": 0.4081, + "step": 21788 + }, + { + "epoch": 8.05, + "learning_rate": 2.196736051237837e-06, + "loss": 0.3379, + "step": 21789 + }, + { + "epoch": 8.05, + "learning_rate": 2.196699100874492e-06, + "loss": 0.4747, + "step": 21790 + }, + { + "epoch": 8.05, + "learning_rate": 2.196662150511147e-06, + "loss": 0.4354, + "step": 21791 + }, + { + "epoch": 8.05, + "learning_rate": 2.1966252001478015e-06, + "loss": 0.4648, + "step": 21792 + }, + { + "epoch": 8.05, + "learning_rate": 2.196588249784456e-06, + "loss": 0.4017, + "step": 21793 + }, + { + "epoch": 8.05, + "learning_rate": 2.196551299421111e-06, + "loss": 0.2671, + "step": 21794 + }, + { + "epoch": 8.05, + "learning_rate": 2.196514349057766e-06, + "loss": 0.4202, + "step": 21795 + }, + { + "epoch": 8.05, + "learning_rate": 2.1964773986944205e-06, + "loss": 0.3549, + "step": 21796 + }, + { + "epoch": 8.05, + "learning_rate": 2.196440448331075e-06, + "loss": 0.3839, + "step": 21797 + }, + { + "epoch": 8.05, + "learning_rate": 2.19640349796773e-06, + "loss": 0.3894, + "step": 21798 + }, + { + "epoch": 8.05, + "learning_rate": 2.196366547604385e-06, + "loss": 0.45, + "step": 21799 + }, + { + "epoch": 8.05, + "learning_rate": 2.1963295972410395e-06, + "loss": 0.5125, + "step": 21800 + }, + { + "epoch": 8.05, + "learning_rate": 2.196292646877694e-06, + "loss": 0.3905, + "step": 21801 + }, + { + "epoch": 8.05, + "learning_rate": 2.196255696514349e-06, + "loss": 0.2973, + "step": 21802 + }, + { + "epoch": 8.05, + "learning_rate": 2.196218746151004e-06, + "loss": 0.3415, + "step": 21803 + }, + { + "epoch": 8.05, + "learning_rate": 2.196181795787659e-06, + "loss": 0.4055, + "step": 21804 + }, + { + "epoch": 8.05, + "learning_rate": 2.1961448454243135e-06, + "loss": 0.3196, + "step": 21805 + }, + { + "epoch": 8.05, + "learning_rate": 2.196107895060968e-06, + "loss": 0.3759, + "step": 21806 + }, + { + "epoch": 8.05, + "learning_rate": 2.1960709446976228e-06, + "loss": 0.5272, + "step": 21807 + }, + { + "epoch": 8.05, + "learning_rate": 2.1960339943342774e-06, + "loss": 0.3978, + "step": 21808 + }, + { + "epoch": 8.05, + "learning_rate": 2.1959970439709325e-06, + "loss": 0.3527, + "step": 21809 + }, + { + "epoch": 8.05, + "learning_rate": 2.195960093607587e-06, + "loss": 0.2562, + "step": 21810 + }, + { + "epoch": 8.05, + "learning_rate": 2.195923143244242e-06, + "loss": 0.3549, + "step": 21811 + }, + { + "epoch": 8.05, + "learning_rate": 2.195886192880897e-06, + "loss": 0.3266, + "step": 21812 + }, + { + "epoch": 8.06, + "learning_rate": 2.1958492425175514e-06, + "loss": 0.3988, + "step": 21813 + }, + { + "epoch": 8.06, + "learning_rate": 2.195812292154206e-06, + "loss": 0.3971, + "step": 21814 + }, + { + "epoch": 8.06, + "learning_rate": 2.195775341790861e-06, + "loss": 0.4187, + "step": 21815 + }, + { + "epoch": 8.06, + "learning_rate": 2.1957383914275158e-06, + "loss": 0.3242, + "step": 21816 + }, + { + "epoch": 8.06, + "learning_rate": 2.1957014410641704e-06, + "loss": 0.3494, + "step": 21817 + }, + { + "epoch": 8.06, + "learning_rate": 2.1956644907008255e-06, + "loss": 0.4242, + "step": 21818 + }, + { + "epoch": 8.06, + "learning_rate": 2.19562754033748e-06, + "loss": 0.3898, + "step": 21819 + }, + { + "epoch": 8.06, + "learning_rate": 2.1955905899741347e-06, + "loss": 0.3107, + "step": 21820 + }, + { + "epoch": 8.06, + "learning_rate": 2.1955536396107894e-06, + "loss": 0.403, + "step": 21821 + }, + { + "epoch": 8.06, + "learning_rate": 2.1955166892474444e-06, + "loss": 0.4208, + "step": 21822 + }, + { + "epoch": 8.06, + "learning_rate": 2.195479738884099e-06, + "loss": 0.3371, + "step": 21823 + }, + { + "epoch": 8.06, + "learning_rate": 2.195442788520754e-06, + "loss": 0.3514, + "step": 21824 + }, + { + "epoch": 8.06, + "learning_rate": 2.1954058381574088e-06, + "loss": 0.3539, + "step": 21825 + }, + { + "epoch": 8.06, + "learning_rate": 2.1953688877940634e-06, + "loss": 0.4064, + "step": 21826 + }, + { + "epoch": 8.06, + "learning_rate": 2.195331937430718e-06, + "loss": 0.4634, + "step": 21827 + }, + { + "epoch": 8.06, + "learning_rate": 2.1952949870673727e-06, + "loss": 0.2299, + "step": 21828 + }, + { + "epoch": 8.06, + "learning_rate": 2.1952580367040277e-06, + "loss": 0.4341, + "step": 21829 + }, + { + "epoch": 8.06, + "learning_rate": 2.1952210863406824e-06, + "loss": 0.4359, + "step": 21830 + }, + { + "epoch": 8.06, + "learning_rate": 2.1951841359773374e-06, + "loss": 0.4119, + "step": 21831 + }, + { + "epoch": 8.06, + "learning_rate": 2.1951471856139916e-06, + "loss": 0.4104, + "step": 21832 + }, + { + "epoch": 8.06, + "learning_rate": 2.1951102352506467e-06, + "loss": 0.5319, + "step": 21833 + }, + { + "epoch": 8.06, + "learning_rate": 2.1950732848873013e-06, + "loss": 0.3258, + "step": 21834 + }, + { + "epoch": 8.06, + "learning_rate": 2.1950363345239564e-06, + "loss": 0.3945, + "step": 21835 + }, + { + "epoch": 8.06, + "learning_rate": 2.194999384160611e-06, + "loss": 0.4129, + "step": 21836 + }, + { + "epoch": 8.06, + "learning_rate": 2.1949624337972657e-06, + "loss": 0.2527, + "step": 21837 + }, + { + "epoch": 8.06, + "learning_rate": 2.1949254834339203e-06, + "loss": 0.39, + "step": 21838 + }, + { + "epoch": 8.06, + "learning_rate": 2.194888533070575e-06, + "loss": 0.3078, + "step": 21839 + }, + { + "epoch": 8.06, + "learning_rate": 2.19485158270723e-06, + "loss": 0.3735, + "step": 21840 + }, + { + "epoch": 8.07, + "learning_rate": 2.1948146323438846e-06, + "loss": 0.4234, + "step": 21841 + }, + { + "epoch": 8.07, + "learning_rate": 2.1947776819805397e-06, + "loss": 0.4285, + "step": 21842 + }, + { + "epoch": 8.07, + "learning_rate": 2.1947407316171943e-06, + "loss": 0.2938, + "step": 21843 + }, + { + "epoch": 8.07, + "learning_rate": 2.194703781253849e-06, + "loss": 0.3686, + "step": 21844 + }, + { + "epoch": 8.07, + "learning_rate": 2.1946668308905036e-06, + "loss": 0.4558, + "step": 21845 + }, + { + "epoch": 8.07, + "learning_rate": 2.1946298805271587e-06, + "loss": 0.3579, + "step": 21846 + }, + { + "epoch": 8.07, + "learning_rate": 2.1945929301638133e-06, + "loss": 0.4649, + "step": 21847 + }, + { + "epoch": 8.07, + "learning_rate": 2.194555979800468e-06, + "loss": 0.6413, + "step": 21848 + }, + { + "epoch": 8.07, + "learning_rate": 2.194519029437123e-06, + "loss": 0.3163, + "step": 21849 + }, + { + "epoch": 8.07, + "learning_rate": 2.1944820790737776e-06, + "loss": 0.4615, + "step": 21850 + }, + { + "epoch": 8.07, + "learning_rate": 2.1944451287104323e-06, + "loss": 0.4153, + "step": 21851 + }, + { + "epoch": 8.07, + "learning_rate": 2.194408178347087e-06, + "loss": 0.3432, + "step": 21852 + }, + { + "epoch": 8.07, + "learning_rate": 2.194371227983742e-06, + "loss": 0.455, + "step": 21853 + }, + { + "epoch": 8.07, + "learning_rate": 2.1943342776203966e-06, + "loss": 0.3325, + "step": 21854 + }, + { + "epoch": 8.07, + "learning_rate": 2.1942973272570517e-06, + "loss": 0.3854, + "step": 21855 + }, + { + "epoch": 8.07, + "learning_rate": 2.1942603768937063e-06, + "loss": 0.3628, + "step": 21856 + }, + { + "epoch": 8.07, + "learning_rate": 2.194223426530361e-06, + "loss": 0.4543, + "step": 21857 + }, + { + "epoch": 8.07, + "learning_rate": 2.1941864761670156e-06, + "loss": 0.3983, + "step": 21858 + }, + { + "epoch": 8.07, + "learning_rate": 2.1941495258036702e-06, + "loss": 0.3129, + "step": 21859 + }, + { + "epoch": 8.07, + "learning_rate": 2.1941125754403253e-06, + "loss": 0.336, + "step": 21860 + }, + { + "epoch": 8.07, + "learning_rate": 2.19407562507698e-06, + "loss": 0.3219, + "step": 21861 + }, + { + "epoch": 8.07, + "learning_rate": 2.194038674713635e-06, + "loss": 0.352, + "step": 21862 + }, + { + "epoch": 8.07, + "learning_rate": 2.1940017243502896e-06, + "loss": 0.2981, + "step": 21863 + }, + { + "epoch": 8.07, + "learning_rate": 2.1939647739869443e-06, + "loss": 0.4197, + "step": 21864 + }, + { + "epoch": 8.07, + "learning_rate": 2.193927823623599e-06, + "loss": 0.4249, + "step": 21865 + }, + { + "epoch": 8.07, + "learning_rate": 2.1938908732602535e-06, + "loss": 0.4682, + "step": 21866 + }, + { + "epoch": 8.07, + "learning_rate": 2.1938539228969086e-06, + "loss": 0.4655, + "step": 21867 + }, + { + "epoch": 8.08, + "learning_rate": 2.1938169725335632e-06, + "loss": 0.3456, + "step": 21868 + }, + { + "epoch": 8.08, + "learning_rate": 2.1937800221702183e-06, + "loss": 0.3706, + "step": 21869 + }, + { + "epoch": 8.08, + "learning_rate": 2.193743071806873e-06, + "loss": 0.3063, + "step": 21870 + }, + { + "epoch": 8.08, + "learning_rate": 2.1937061214435276e-06, + "loss": 0.4405, + "step": 21871 + }, + { + "epoch": 8.08, + "learning_rate": 2.193669171080182e-06, + "loss": 0.4858, + "step": 21872 + }, + { + "epoch": 8.08, + "learning_rate": 2.1936322207168372e-06, + "loss": 0.3209, + "step": 21873 + }, + { + "epoch": 8.08, + "learning_rate": 2.193595270353492e-06, + "loss": 0.3965, + "step": 21874 + }, + { + "epoch": 8.08, + "learning_rate": 2.193558319990147e-06, + "loss": 0.4002, + "step": 21875 + }, + { + "epoch": 8.08, + "learning_rate": 2.1935213696268016e-06, + "loss": 0.4004, + "step": 21876 + }, + { + "epoch": 8.08, + "learning_rate": 2.193484419263456e-06, + "loss": 0.3486, + "step": 21877 + }, + { + "epoch": 8.08, + "learning_rate": 2.193447468900111e-06, + "loss": 0.4192, + "step": 21878 + }, + { + "epoch": 8.08, + "learning_rate": 2.1934105185367655e-06, + "loss": 0.3149, + "step": 21879 + }, + { + "epoch": 8.08, + "learning_rate": 2.1933735681734206e-06, + "loss": 0.3878, + "step": 21880 + }, + { + "epoch": 8.08, + "learning_rate": 2.193336617810075e-06, + "loss": 0.2833, + "step": 21881 + }, + { + "epoch": 8.08, + "learning_rate": 2.1932996674467302e-06, + "loss": 0.3197, + "step": 21882 + }, + { + "epoch": 8.08, + "learning_rate": 2.1932627170833845e-06, + "loss": 0.3939, + "step": 21883 + }, + { + "epoch": 8.08, + "learning_rate": 2.1932257667200395e-06, + "loss": 0.3767, + "step": 21884 + }, + { + "epoch": 8.08, + "learning_rate": 2.193188816356694e-06, + "loss": 0.343, + "step": 21885 + }, + { + "epoch": 8.08, + "learning_rate": 2.193151865993349e-06, + "loss": 0.3139, + "step": 21886 + }, + { + "epoch": 8.08, + "learning_rate": 2.193114915630004e-06, + "loss": 0.5416, + "step": 21887 + }, + { + "epoch": 8.08, + "learning_rate": 2.1930779652666585e-06, + "loss": 0.3521, + "step": 21888 + }, + { + "epoch": 8.08, + "learning_rate": 2.193041014903313e-06, + "loss": 0.3586, + "step": 21889 + }, + { + "epoch": 8.08, + "learning_rate": 2.1930040645399678e-06, + "loss": 0.3749, + "step": 21890 + }, + { + "epoch": 8.08, + "learning_rate": 2.192967114176623e-06, + "loss": 0.3397, + "step": 21891 + }, + { + "epoch": 8.08, + "learning_rate": 2.1929301638132775e-06, + "loss": 0.2889, + "step": 21892 + }, + { + "epoch": 8.08, + "learning_rate": 2.1928932134499325e-06, + "loss": 0.4692, + "step": 21893 + }, + { + "epoch": 8.08, + "learning_rate": 2.192856263086587e-06, + "loss": 0.4126, + "step": 21894 + }, + { + "epoch": 8.09, + "learning_rate": 2.192819312723242e-06, + "loss": 0.4319, + "step": 21895 + }, + { + "epoch": 8.09, + "learning_rate": 2.1927823623598964e-06, + "loss": 0.4317, + "step": 21896 + }, + { + "epoch": 8.09, + "learning_rate": 2.192745411996551e-06, + "loss": 0.3618, + "step": 21897 + }, + { + "epoch": 8.09, + "learning_rate": 2.192708461633206e-06, + "loss": 0.3383, + "step": 21898 + }, + { + "epoch": 8.09, + "learning_rate": 2.1926715112698608e-06, + "loss": 0.3974, + "step": 21899 + }, + { + "epoch": 8.09, + "learning_rate": 2.192634560906516e-06, + "loss": 0.3351, + "step": 21900 + }, + { + "epoch": 8.09, + "learning_rate": 2.1925976105431705e-06, + "loss": 0.4651, + "step": 21901 + }, + { + "epoch": 8.09, + "learning_rate": 2.192560660179825e-06, + "loss": 0.3496, + "step": 21902 + }, + { + "epoch": 8.09, + "learning_rate": 2.1925237098164797e-06, + "loss": 0.4125, + "step": 21903 + }, + { + "epoch": 8.09, + "learning_rate": 2.192486759453135e-06, + "loss": 0.3522, + "step": 21904 + }, + { + "epoch": 8.09, + "learning_rate": 2.1924498090897894e-06, + "loss": 0.5086, + "step": 21905 + }, + { + "epoch": 8.09, + "learning_rate": 2.192412858726444e-06, + "loss": 0.2966, + "step": 21906 + }, + { + "epoch": 8.09, + "learning_rate": 2.192375908363099e-06, + "loss": 0.4153, + "step": 21907 + }, + { + "epoch": 8.09, + "learning_rate": 2.1923389579997538e-06, + "loss": 0.3597, + "step": 21908 + }, + { + "epoch": 8.09, + "learning_rate": 2.1923020076364084e-06, + "loss": 0.4091, + "step": 21909 + }, + { + "epoch": 8.09, + "learning_rate": 2.192265057273063e-06, + "loss": 0.4003, + "step": 21910 + }, + { + "epoch": 8.09, + "learning_rate": 2.192228106909718e-06, + "loss": 0.3355, + "step": 21911 + }, + { + "epoch": 8.09, + "learning_rate": 2.1921911565463727e-06, + "loss": 0.4752, + "step": 21912 + }, + { + "epoch": 8.09, + "learning_rate": 2.192154206183028e-06, + "loss": 0.4449, + "step": 21913 + }, + { + "epoch": 8.09, + "learning_rate": 2.1921172558196824e-06, + "loss": 0.465, + "step": 21914 + }, + { + "epoch": 8.09, + "learning_rate": 2.192080305456337e-06, + "loss": 0.3822, + "step": 21915 + }, + { + "epoch": 8.09, + "learning_rate": 2.1920433550929917e-06, + "loss": 0.443, + "step": 21916 + }, + { + "epoch": 8.09, + "learning_rate": 2.1920064047296463e-06, + "loss": 0.4617, + "step": 21917 + }, + { + "epoch": 8.09, + "learning_rate": 2.1919694543663014e-06, + "loss": 0.3229, + "step": 21918 + }, + { + "epoch": 8.09, + "learning_rate": 2.191932504002956e-06, + "loss": 0.3496, + "step": 21919 + }, + { + "epoch": 8.09, + "learning_rate": 2.191895553639611e-06, + "loss": 0.3636, + "step": 21920 + }, + { + "epoch": 8.09, + "learning_rate": 2.1918586032762657e-06, + "loss": 0.4069, + "step": 21921 + }, + { + "epoch": 8.1, + "learning_rate": 2.1918216529129204e-06, + "loss": 0.2824, + "step": 21922 + }, + { + "epoch": 8.1, + "learning_rate": 2.191784702549575e-06, + "loss": 0.3853, + "step": 21923 + }, + { + "epoch": 8.1, + "learning_rate": 2.19174775218623e-06, + "loss": 0.3797, + "step": 21924 + }, + { + "epoch": 8.1, + "learning_rate": 2.1917108018228847e-06, + "loss": 0.2658, + "step": 21925 + }, + { + "epoch": 8.1, + "learning_rate": 2.1916738514595393e-06, + "loss": 0.2968, + "step": 21926 + }, + { + "epoch": 8.1, + "learning_rate": 2.1916369010961944e-06, + "loss": 0.3282, + "step": 21927 + }, + { + "epoch": 8.1, + "learning_rate": 2.1915999507328486e-06, + "loss": 0.4251, + "step": 21928 + }, + { + "epoch": 8.1, + "learning_rate": 2.1915630003695037e-06, + "loss": 0.4008, + "step": 21929 + }, + { + "epoch": 8.1, + "learning_rate": 2.1915260500061583e-06, + "loss": 0.4878, + "step": 21930 + }, + { + "epoch": 8.1, + "learning_rate": 2.1914890996428134e-06, + "loss": 0.3637, + "step": 21931 + }, + { + "epoch": 8.1, + "learning_rate": 2.191452149279468e-06, + "loss": 0.3741, + "step": 21932 + }, + { + "epoch": 8.1, + "learning_rate": 2.191415198916123e-06, + "loss": 0.4225, + "step": 21933 + }, + { + "epoch": 8.1, + "learning_rate": 2.1913782485527773e-06, + "loss": 0.3427, + "step": 21934 + }, + { + "epoch": 8.1, + "learning_rate": 2.1913412981894323e-06, + "loss": 0.5106, + "step": 21935 + }, + { + "epoch": 8.1, + "learning_rate": 2.191304347826087e-06, + "loss": 0.364, + "step": 21936 + }, + { + "epoch": 8.1, + "learning_rate": 2.1912673974627416e-06, + "loss": 0.5006, + "step": 21937 + }, + { + "epoch": 8.1, + "learning_rate": 2.1912304470993967e-06, + "loss": 0.3861, + "step": 21938 + }, + { + "epoch": 8.1, + "learning_rate": 2.1911934967360513e-06, + "loss": 0.3982, + "step": 21939 + }, + { + "epoch": 8.1, + "learning_rate": 2.191156546372706e-06, + "loss": 0.3597, + "step": 21940 + }, + { + "epoch": 8.1, + "learning_rate": 2.1911195960093606e-06, + "loss": 0.6616, + "step": 21941 + }, + { + "epoch": 8.1, + "learning_rate": 2.1910826456460156e-06, + "loss": 0.4174, + "step": 21942 + }, + { + "epoch": 8.1, + "learning_rate": 2.1910456952826703e-06, + "loss": 0.3695, + "step": 21943 + }, + { + "epoch": 8.1, + "learning_rate": 2.1910087449193253e-06, + "loss": 0.3302, + "step": 21944 + }, + { + "epoch": 8.1, + "learning_rate": 2.19097179455598e-06, + "loss": 0.4156, + "step": 21945 + }, + { + "epoch": 8.1, + "learning_rate": 2.1909348441926346e-06, + "loss": 0.3981, + "step": 21946 + }, + { + "epoch": 8.1, + "learning_rate": 2.1908978938292892e-06, + "loss": 0.3339, + "step": 21947 + }, + { + "epoch": 8.1, + "learning_rate": 2.190860943465944e-06, + "loss": 0.3006, + "step": 21948 + }, + { + "epoch": 8.11, + "learning_rate": 2.190823993102599e-06, + "loss": 0.3745, + "step": 21949 + }, + { + "epoch": 8.11, + "learning_rate": 2.1907870427392536e-06, + "loss": 0.359, + "step": 21950 + }, + { + "epoch": 8.11, + "learning_rate": 2.1907500923759086e-06, + "loss": 0.4019, + "step": 21951 + }, + { + "epoch": 8.11, + "learning_rate": 2.1907131420125633e-06, + "loss": 0.5115, + "step": 21952 + }, + { + "epoch": 8.11, + "learning_rate": 2.190676191649218e-06, + "loss": 0.3652, + "step": 21953 + }, + { + "epoch": 8.11, + "learning_rate": 2.1906392412858725e-06, + "loss": 0.4128, + "step": 21954 + }, + { + "epoch": 8.11, + "learning_rate": 2.1906022909225276e-06, + "loss": 0.4367, + "step": 21955 + }, + { + "epoch": 8.11, + "learning_rate": 2.1905653405591822e-06, + "loss": 0.4789, + "step": 21956 + }, + { + "epoch": 8.11, + "learning_rate": 2.190528390195837e-06, + "loss": 0.4665, + "step": 21957 + }, + { + "epoch": 8.11, + "learning_rate": 2.190491439832492e-06, + "loss": 0.4222, + "step": 21958 + }, + { + "epoch": 8.11, + "learning_rate": 2.1904544894691466e-06, + "loss": 0.4346, + "step": 21959 + }, + { + "epoch": 8.11, + "learning_rate": 2.190417539105801e-06, + "loss": 0.2496, + "step": 21960 + }, + { + "epoch": 8.11, + "learning_rate": 2.190380588742456e-06, + "loss": 0.4283, + "step": 21961 + }, + { + "epoch": 8.11, + "learning_rate": 2.190343638379111e-06, + "loss": 0.3476, + "step": 21962 + }, + { + "epoch": 8.11, + "learning_rate": 2.1903066880157655e-06, + "loss": 0.3353, + "step": 21963 + }, + { + "epoch": 8.11, + "learning_rate": 2.1902697376524206e-06, + "loss": 0.3554, + "step": 21964 + }, + { + "epoch": 8.11, + "learning_rate": 2.1902327872890752e-06, + "loss": 0.3179, + "step": 21965 + }, + { + "epoch": 8.11, + "learning_rate": 2.19019583692573e-06, + "loss": 0.4386, + "step": 21966 + }, + { + "epoch": 8.11, + "learning_rate": 2.1901588865623845e-06, + "loss": 0.5369, + "step": 21967 + }, + { + "epoch": 8.11, + "learning_rate": 2.190121936199039e-06, + "loss": 0.3152, + "step": 21968 + }, + { + "epoch": 8.11, + "learning_rate": 2.190084985835694e-06, + "loss": 0.3262, + "step": 21969 + }, + { + "epoch": 8.11, + "learning_rate": 2.190048035472349e-06, + "loss": 0.4083, + "step": 21970 + }, + { + "epoch": 8.11, + "learning_rate": 2.190011085109004e-06, + "loss": 0.245, + "step": 21971 + }, + { + "epoch": 8.11, + "learning_rate": 2.1899741347456585e-06, + "loss": 0.3244, + "step": 21972 + }, + { + "epoch": 8.11, + "learning_rate": 2.189937184382313e-06, + "loss": 0.3301, + "step": 21973 + }, + { + "epoch": 8.11, + "learning_rate": 2.189900234018968e-06, + "loss": 0.3744, + "step": 21974 + }, + { + "epoch": 8.11, + "learning_rate": 2.189863283655623e-06, + "loss": 0.3404, + "step": 21975 + }, + { + "epoch": 8.12, + "learning_rate": 2.1898263332922775e-06, + "loss": 0.2455, + "step": 21976 + }, + { + "epoch": 8.12, + "learning_rate": 2.189789382928932e-06, + "loss": 0.3893, + "step": 21977 + }, + { + "epoch": 8.12, + "learning_rate": 2.189752432565587e-06, + "loss": 0.4396, + "step": 21978 + }, + { + "epoch": 8.12, + "learning_rate": 2.1897154822022414e-06, + "loss": 0.3591, + "step": 21979 + }, + { + "epoch": 8.12, + "learning_rate": 2.1896785318388965e-06, + "loss": 0.4517, + "step": 21980 + }, + { + "epoch": 8.12, + "learning_rate": 2.189641581475551e-06, + "loss": 0.2389, + "step": 21981 + }, + { + "epoch": 8.12, + "learning_rate": 2.189604631112206e-06, + "loss": 0.4213, + "step": 21982 + }, + { + "epoch": 8.12, + "learning_rate": 2.189567680748861e-06, + "loss": 0.4339, + "step": 21983 + }, + { + "epoch": 8.12, + "learning_rate": 2.1895307303855154e-06, + "loss": 0.4772, + "step": 21984 + }, + { + "epoch": 8.12, + "learning_rate": 2.18949378002217e-06, + "loss": 0.3388, + "step": 21985 + }, + { + "epoch": 8.12, + "learning_rate": 2.1894568296588247e-06, + "loss": 0.347, + "step": 21986 + }, + { + "epoch": 8.12, + "learning_rate": 2.1894198792954798e-06, + "loss": 0.4803, + "step": 21987 + }, + { + "epoch": 8.12, + "learning_rate": 2.1893829289321344e-06, + "loss": 0.3333, + "step": 21988 + }, + { + "epoch": 8.12, + "learning_rate": 2.1893459785687895e-06, + "loss": 0.401, + "step": 21989 + }, + { + "epoch": 8.12, + "learning_rate": 2.189309028205444e-06, + "loss": 0.3473, + "step": 21990 + }, + { + "epoch": 8.12, + "learning_rate": 2.1892720778420987e-06, + "loss": 0.3507, + "step": 21991 + }, + { + "epoch": 8.12, + "learning_rate": 2.1892351274787534e-06, + "loss": 0.2941, + "step": 21992 + }, + { + "epoch": 8.12, + "learning_rate": 2.1891981771154084e-06, + "loss": 0.4565, + "step": 21993 + }, + { + "epoch": 8.12, + "learning_rate": 2.189161226752063e-06, + "loss": 0.2941, + "step": 21994 + }, + { + "epoch": 8.12, + "learning_rate": 2.189124276388718e-06, + "loss": 0.4376, + "step": 21995 + }, + { + "epoch": 8.12, + "learning_rate": 2.1890873260253728e-06, + "loss": 0.3127, + "step": 21996 + }, + { + "epoch": 8.12, + "learning_rate": 2.1890503756620274e-06, + "loss": 0.4405, + "step": 21997 + }, + { + "epoch": 8.12, + "learning_rate": 2.189013425298682e-06, + "loss": 0.2803, + "step": 21998 + }, + { + "epoch": 8.12, + "learning_rate": 2.1889764749353367e-06, + "loss": 0.4479, + "step": 21999 + }, + { + "epoch": 8.12, + "learning_rate": 2.1889395245719917e-06, + "loss": 0.3677, + "step": 22000 + }, + { + "epoch": 8.12, + "learning_rate": 2.1889025742086464e-06, + "loss": 0.3618, + "step": 22001 + }, + { + "epoch": 8.12, + "learning_rate": 2.1888656238453014e-06, + "loss": 0.2467, + "step": 22002 + }, + { + "epoch": 8.13, + "learning_rate": 2.188828673481956e-06, + "loss": 0.3499, + "step": 22003 + }, + { + "epoch": 8.13, + "learning_rate": 2.1887917231186107e-06, + "loss": 0.5746, + "step": 22004 + }, + { + "epoch": 8.13, + "learning_rate": 2.1887547727552654e-06, + "loss": 0.4723, + "step": 22005 + }, + { + "epoch": 8.13, + "learning_rate": 2.18871782239192e-06, + "loss": 0.3875, + "step": 22006 + }, + { + "epoch": 8.13, + "learning_rate": 2.188680872028575e-06, + "loss": 0.421, + "step": 22007 + }, + { + "epoch": 8.13, + "learning_rate": 2.1886439216652297e-06, + "loss": 0.6022, + "step": 22008 + }, + { + "epoch": 8.13, + "learning_rate": 2.1886069713018847e-06, + "loss": 0.3671, + "step": 22009 + }, + { + "epoch": 8.13, + "learning_rate": 2.1885700209385394e-06, + "loss": 0.5236, + "step": 22010 + }, + { + "epoch": 8.13, + "learning_rate": 2.188533070575194e-06, + "loss": 0.4291, + "step": 22011 + }, + { + "epoch": 8.13, + "learning_rate": 2.1884961202118487e-06, + "loss": 0.5471, + "step": 22012 + }, + { + "epoch": 8.13, + "learning_rate": 2.1884591698485037e-06, + "loss": 0.3985, + "step": 22013 + }, + { + "epoch": 8.13, + "learning_rate": 2.1884222194851584e-06, + "loss": 0.4082, + "step": 22014 + }, + { + "epoch": 8.13, + "learning_rate": 2.1883852691218134e-06, + "loss": 0.4532, + "step": 22015 + }, + { + "epoch": 8.13, + "learning_rate": 2.188348318758468e-06, + "loss": 0.4572, + "step": 22016 + }, + { + "epoch": 8.13, + "learning_rate": 2.1883113683951223e-06, + "loss": 0.4459, + "step": 22017 + }, + { + "epoch": 8.13, + "learning_rate": 2.1882744180317773e-06, + "loss": 0.5215, + "step": 22018 + }, + { + "epoch": 8.13, + "learning_rate": 2.188237467668432e-06, + "loss": 0.3218, + "step": 22019 + }, + { + "epoch": 8.13, + "learning_rate": 2.188200517305087e-06, + "loss": 0.4247, + "step": 22020 + }, + { + "epoch": 8.13, + "learning_rate": 2.1881635669417417e-06, + "loss": 0.3949, + "step": 22021 + }, + { + "epoch": 8.13, + "learning_rate": 2.1881266165783967e-06, + "loss": 0.3599, + "step": 22022 + }, + { + "epoch": 8.13, + "learning_rate": 2.188089666215051e-06, + "loss": 0.3813, + "step": 22023 + }, + { + "epoch": 8.13, + "learning_rate": 2.188052715851706e-06, + "loss": 0.4047, + "step": 22024 + }, + { + "epoch": 8.13, + "learning_rate": 2.1880157654883606e-06, + "loss": 0.3813, + "step": 22025 + }, + { + "epoch": 8.13, + "learning_rate": 2.1879788151250153e-06, + "loss": 0.4603, + "step": 22026 + }, + { + "epoch": 8.13, + "learning_rate": 2.1879418647616703e-06, + "loss": 0.4066, + "step": 22027 + }, + { + "epoch": 8.13, + "learning_rate": 2.187904914398325e-06, + "loss": 0.46, + "step": 22028 + }, + { + "epoch": 8.13, + "learning_rate": 2.1878679640349796e-06, + "loss": 0.283, + "step": 22029 + }, + { + "epoch": 8.14, + "learning_rate": 2.1878310136716342e-06, + "loss": 0.3755, + "step": 22030 + }, + { + "epoch": 8.14, + "learning_rate": 2.1877940633082893e-06, + "loss": 0.4925, + "step": 22031 + }, + { + "epoch": 8.14, + "learning_rate": 2.187757112944944e-06, + "loss": 0.498, + "step": 22032 + }, + { + "epoch": 8.14, + "learning_rate": 2.187720162581599e-06, + "loss": 0.3489, + "step": 22033 + }, + { + "epoch": 8.14, + "learning_rate": 2.1876832122182536e-06, + "loss": 0.2421, + "step": 22034 + }, + { + "epoch": 8.14, + "learning_rate": 2.1876462618549083e-06, + "loss": 0.301, + "step": 22035 + }, + { + "epoch": 8.14, + "learning_rate": 2.187609311491563e-06, + "loss": 0.319, + "step": 22036 + }, + { + "epoch": 8.14, + "learning_rate": 2.1875723611282175e-06, + "loss": 0.3926, + "step": 22037 + }, + { + "epoch": 8.14, + "learning_rate": 2.1875354107648726e-06, + "loss": 0.5324, + "step": 22038 + }, + { + "epoch": 8.14, + "learning_rate": 2.1874984604015272e-06, + "loss": 0.4558, + "step": 22039 + }, + { + "epoch": 8.14, + "learning_rate": 2.1874615100381823e-06, + "loss": 0.435, + "step": 22040 + }, + { + "epoch": 8.14, + "learning_rate": 2.187424559674837e-06, + "loss": 0.4176, + "step": 22041 + }, + { + "epoch": 8.14, + "learning_rate": 2.1873876093114916e-06, + "loss": 0.4528, + "step": 22042 + }, + { + "epoch": 8.14, + "learning_rate": 2.187350658948146e-06, + "loss": 0.2688, + "step": 22043 + }, + { + "epoch": 8.14, + "learning_rate": 2.1873137085848013e-06, + "loss": 0.475, + "step": 22044 + }, + { + "epoch": 8.14, + "learning_rate": 2.187276758221456e-06, + "loss": 0.2961, + "step": 22045 + }, + { + "epoch": 8.14, + "learning_rate": 2.1872398078581105e-06, + "loss": 0.3858, + "step": 22046 + }, + { + "epoch": 8.14, + "learning_rate": 2.1872028574947656e-06, + "loss": 0.298, + "step": 22047 + }, + { + "epoch": 8.14, + "learning_rate": 2.1871659071314202e-06, + "loss": 0.4743, + "step": 22048 + }, + { + "epoch": 8.14, + "learning_rate": 2.187128956768075e-06, + "loss": 0.3489, + "step": 22049 + }, + { + "epoch": 8.14, + "learning_rate": 2.1870920064047295e-06, + "loss": 0.3586, + "step": 22050 + }, + { + "epoch": 8.14, + "learning_rate": 2.1870550560413846e-06, + "loss": 0.4775, + "step": 22051 + }, + { + "epoch": 8.14, + "learning_rate": 2.187018105678039e-06, + "loss": 0.429, + "step": 22052 + }, + { + "epoch": 8.14, + "learning_rate": 2.1869811553146943e-06, + "loss": 0.4328, + "step": 22053 + }, + { + "epoch": 8.14, + "learning_rate": 2.186944204951349e-06, + "loss": 0.4791, + "step": 22054 + }, + { + "epoch": 8.14, + "learning_rate": 2.1869072545880035e-06, + "loss": 0.2313, + "step": 22055 + }, + { + "epoch": 8.14, + "learning_rate": 2.186870304224658e-06, + "loss": 0.33, + "step": 22056 + }, + { + "epoch": 8.15, + "learning_rate": 2.186833353861313e-06, + "loss": 0.4241, + "step": 22057 + }, + { + "epoch": 8.15, + "learning_rate": 2.186796403497968e-06, + "loss": 0.5717, + "step": 22058 + }, + { + "epoch": 8.15, + "learning_rate": 2.1867594531346225e-06, + "loss": 0.3801, + "step": 22059 + }, + { + "epoch": 8.15, + "learning_rate": 2.1867225027712776e-06, + "loss": 0.3302, + "step": 22060 + }, + { + "epoch": 8.15, + "learning_rate": 2.186685552407932e-06, + "loss": 0.2998, + "step": 22061 + }, + { + "epoch": 8.15, + "learning_rate": 2.186648602044587e-06, + "loss": 0.4015, + "step": 22062 + }, + { + "epoch": 8.15, + "learning_rate": 2.1866116516812415e-06, + "loss": 0.4735, + "step": 22063 + }, + { + "epoch": 8.15, + "learning_rate": 2.1865747013178965e-06, + "loss": 0.394, + "step": 22064 + }, + { + "epoch": 8.15, + "learning_rate": 2.186537750954551e-06, + "loss": 0.3253, + "step": 22065 + }, + { + "epoch": 8.15, + "learning_rate": 2.186500800591206e-06, + "loss": 0.3707, + "step": 22066 + }, + { + "epoch": 8.15, + "learning_rate": 2.186463850227861e-06, + "loss": 0.5369, + "step": 22067 + }, + { + "epoch": 8.15, + "learning_rate": 2.186426899864515e-06, + "loss": 0.4202, + "step": 22068 + }, + { + "epoch": 8.15, + "learning_rate": 2.18638994950117e-06, + "loss": 0.3531, + "step": 22069 + }, + { + "epoch": 8.15, + "learning_rate": 2.1863529991378248e-06, + "loss": 0.4999, + "step": 22070 + }, + { + "epoch": 8.15, + "learning_rate": 2.18631604877448e-06, + "loss": 0.216, + "step": 22071 + }, + { + "epoch": 8.15, + "learning_rate": 2.1862790984111345e-06, + "loss": 0.401, + "step": 22072 + }, + { + "epoch": 8.15, + "learning_rate": 2.1862421480477895e-06, + "loss": 0.4162, + "step": 22073 + }, + { + "epoch": 8.15, + "learning_rate": 2.1862051976844437e-06, + "loss": 0.449, + "step": 22074 + }, + { + "epoch": 8.15, + "learning_rate": 2.186168247321099e-06, + "loss": 0.4628, + "step": 22075 + }, + { + "epoch": 8.15, + "learning_rate": 2.1861312969577534e-06, + "loss": 0.3782, + "step": 22076 + }, + { + "epoch": 8.15, + "learning_rate": 2.186094346594408e-06, + "loss": 0.4591, + "step": 22077 + }, + { + "epoch": 8.15, + "learning_rate": 2.186057396231063e-06, + "loss": 0.403, + "step": 22078 + }, + { + "epoch": 8.15, + "learning_rate": 2.1860204458677178e-06, + "loss": 0.3198, + "step": 22079 + }, + { + "epoch": 8.15, + "learning_rate": 2.1859834955043724e-06, + "loss": 0.4019, + "step": 22080 + }, + { + "epoch": 8.15, + "learning_rate": 2.185946545141027e-06, + "loss": 0.259, + "step": 22081 + }, + { + "epoch": 8.15, + "learning_rate": 2.185909594777682e-06, + "loss": 0.3621, + "step": 22082 + }, + { + "epoch": 8.15, + "learning_rate": 2.1858726444143367e-06, + "loss": 0.3779, + "step": 22083 + }, + { + "epoch": 8.16, + "learning_rate": 2.185835694050992e-06, + "loss": 0.3681, + "step": 22084 + }, + { + "epoch": 8.16, + "learning_rate": 2.1857987436876464e-06, + "loss": 0.4274, + "step": 22085 + }, + { + "epoch": 8.16, + "learning_rate": 2.185761793324301e-06, + "loss": 0.2358, + "step": 22086 + }, + { + "epoch": 8.16, + "learning_rate": 2.1857248429609557e-06, + "loss": 0.4073, + "step": 22087 + }, + { + "epoch": 8.16, + "learning_rate": 2.1856878925976103e-06, + "loss": 0.3034, + "step": 22088 + }, + { + "epoch": 8.16, + "learning_rate": 2.1856509422342654e-06, + "loss": 0.4712, + "step": 22089 + }, + { + "epoch": 8.16, + "learning_rate": 2.18561399187092e-06, + "loss": 0.3633, + "step": 22090 + }, + { + "epoch": 8.16, + "learning_rate": 2.185577041507575e-06, + "loss": 0.3665, + "step": 22091 + }, + { + "epoch": 8.16, + "learning_rate": 2.1855400911442297e-06, + "loss": 0.4513, + "step": 22092 + }, + { + "epoch": 8.16, + "learning_rate": 2.1855031407808844e-06, + "loss": 0.4284, + "step": 22093 + }, + { + "epoch": 8.16, + "learning_rate": 2.185466190417539e-06, + "loss": 0.4326, + "step": 22094 + }, + { + "epoch": 8.16, + "learning_rate": 2.1854292400541936e-06, + "loss": 0.3971, + "step": 22095 + }, + { + "epoch": 8.16, + "learning_rate": 2.1853922896908487e-06, + "loss": 0.3428, + "step": 22096 + }, + { + "epoch": 8.16, + "learning_rate": 2.1853553393275033e-06, + "loss": 0.4041, + "step": 22097 + }, + { + "epoch": 8.16, + "learning_rate": 2.1853183889641584e-06, + "loss": 0.4293, + "step": 22098 + }, + { + "epoch": 8.16, + "learning_rate": 2.185281438600813e-06, + "loss": 0.4118, + "step": 22099 + }, + { + "epoch": 8.16, + "learning_rate": 2.1852444882374677e-06, + "loss": 0.5071, + "step": 22100 + }, + { + "epoch": 8.16, + "learning_rate": 2.1852075378741223e-06, + "loss": 0.3708, + "step": 22101 + }, + { + "epoch": 8.16, + "learning_rate": 2.1851705875107774e-06, + "loss": 0.4792, + "step": 22102 + }, + { + "epoch": 8.16, + "learning_rate": 2.185133637147432e-06, + "loss": 0.3781, + "step": 22103 + }, + { + "epoch": 8.16, + "learning_rate": 2.185096686784087e-06, + "loss": 0.2678, + "step": 22104 + }, + { + "epoch": 8.16, + "learning_rate": 2.1850597364207417e-06, + "loss": 0.3779, + "step": 22105 + }, + { + "epoch": 8.16, + "learning_rate": 2.1850227860573963e-06, + "loss": 0.2956, + "step": 22106 + }, + { + "epoch": 8.16, + "learning_rate": 2.184985835694051e-06, + "loss": 0.4211, + "step": 22107 + }, + { + "epoch": 8.16, + "learning_rate": 2.1849488853307056e-06, + "loss": 0.5225, + "step": 22108 + }, + { + "epoch": 8.16, + "learning_rate": 2.1849119349673607e-06, + "loss": 0.3609, + "step": 22109 + }, + { + "epoch": 8.16, + "learning_rate": 2.1848749846040153e-06, + "loss": 0.4545, + "step": 22110 + }, + { + "epoch": 8.17, + "learning_rate": 2.1848380342406704e-06, + "loss": 0.419, + "step": 22111 + }, + { + "epoch": 8.17, + "learning_rate": 2.184801083877325e-06, + "loss": 0.2446, + "step": 22112 + }, + { + "epoch": 8.17, + "learning_rate": 2.1847641335139796e-06, + "loss": 0.3102, + "step": 22113 + }, + { + "epoch": 8.17, + "learning_rate": 2.1847271831506343e-06, + "loss": 0.4168, + "step": 22114 + }, + { + "epoch": 8.17, + "learning_rate": 2.184690232787289e-06, + "loss": 0.3007, + "step": 22115 + }, + { + "epoch": 8.17, + "learning_rate": 2.184653282423944e-06, + "loss": 0.3961, + "step": 22116 + }, + { + "epoch": 8.17, + "learning_rate": 2.1846163320605986e-06, + "loss": 0.2604, + "step": 22117 + }, + { + "epoch": 8.17, + "learning_rate": 2.1845793816972537e-06, + "loss": 0.3932, + "step": 22118 + }, + { + "epoch": 8.17, + "learning_rate": 2.184542431333908e-06, + "loss": 0.4339, + "step": 22119 + }, + { + "epoch": 8.17, + "learning_rate": 2.184505480970563e-06, + "loss": 0.3892, + "step": 22120 + }, + { + "epoch": 8.17, + "learning_rate": 2.1844685306072176e-06, + "loss": 0.3622, + "step": 22121 + }, + { + "epoch": 8.17, + "learning_rate": 2.1844315802438726e-06, + "loss": 0.3118, + "step": 22122 + }, + { + "epoch": 8.17, + "learning_rate": 2.1843946298805273e-06, + "loss": 0.4048, + "step": 22123 + }, + { + "epoch": 8.17, + "learning_rate": 2.1843576795171823e-06, + "loss": 0.2971, + "step": 22124 + }, + { + "epoch": 8.17, + "learning_rate": 2.1843207291538365e-06, + "loss": 0.5699, + "step": 22125 + }, + { + "epoch": 8.17, + "learning_rate": 2.184283778790491e-06, + "loss": 0.3699, + "step": 22126 + }, + { + "epoch": 8.17, + "learning_rate": 2.1842468284271462e-06, + "loss": 0.4389, + "step": 22127 + }, + { + "epoch": 8.17, + "learning_rate": 2.184209878063801e-06, + "loss": 0.4146, + "step": 22128 + }, + { + "epoch": 8.17, + "learning_rate": 2.184172927700456e-06, + "loss": 0.2846, + "step": 22129 + }, + { + "epoch": 8.17, + "learning_rate": 2.1841359773371106e-06, + "loss": 0.4234, + "step": 22130 + }, + { + "epoch": 8.17, + "learning_rate": 2.1840990269737652e-06, + "loss": 0.4597, + "step": 22131 + }, + { + "epoch": 8.17, + "learning_rate": 2.18406207661042e-06, + "loss": 0.3105, + "step": 22132 + }, + { + "epoch": 8.17, + "learning_rate": 2.184025126247075e-06, + "loss": 0.4155, + "step": 22133 + }, + { + "epoch": 8.17, + "learning_rate": 2.1839881758837295e-06, + "loss": 0.3873, + "step": 22134 + }, + { + "epoch": 8.17, + "learning_rate": 2.183951225520384e-06, + "loss": 0.3518, + "step": 22135 + }, + { + "epoch": 8.17, + "learning_rate": 2.1839142751570392e-06, + "loss": 0.5177, + "step": 22136 + }, + { + "epoch": 8.17, + "learning_rate": 2.183877324793694e-06, + "loss": 0.4666, + "step": 22137 + }, + { + "epoch": 8.18, + "learning_rate": 2.1838403744303485e-06, + "loss": 0.4008, + "step": 22138 + }, + { + "epoch": 8.18, + "learning_rate": 2.183803424067003e-06, + "loss": 0.3464, + "step": 22139 + }, + { + "epoch": 8.18, + "learning_rate": 2.183766473703658e-06, + "loss": 0.3353, + "step": 22140 + }, + { + "epoch": 8.18, + "learning_rate": 2.183729523340313e-06, + "loss": 0.4266, + "step": 22141 + }, + { + "epoch": 8.18, + "learning_rate": 2.183692572976968e-06, + "loss": 0.3601, + "step": 22142 + }, + { + "epoch": 8.18, + "learning_rate": 2.1836556226136225e-06, + "loss": 0.349, + "step": 22143 + }, + { + "epoch": 8.18, + "learning_rate": 2.183618672250277e-06, + "loss": 0.4483, + "step": 22144 + }, + { + "epoch": 8.18, + "learning_rate": 2.183581721886932e-06, + "loss": 0.4615, + "step": 22145 + }, + { + "epoch": 8.18, + "learning_rate": 2.1835447715235865e-06, + "loss": 0.4421, + "step": 22146 + }, + { + "epoch": 8.18, + "learning_rate": 2.1835078211602415e-06, + "loss": 0.5176, + "step": 22147 + }, + { + "epoch": 8.18, + "learning_rate": 2.183470870796896e-06, + "loss": 0.4121, + "step": 22148 + }, + { + "epoch": 8.18, + "learning_rate": 2.183433920433551e-06, + "loss": 0.4241, + "step": 22149 + }, + { + "epoch": 8.18, + "learning_rate": 2.183396970070206e-06, + "loss": 0.3141, + "step": 22150 + }, + { + "epoch": 8.18, + "learning_rate": 2.1833600197068605e-06, + "loss": 0.4068, + "step": 22151 + }, + { + "epoch": 8.18, + "learning_rate": 2.183323069343515e-06, + "loss": 0.4511, + "step": 22152 + }, + { + "epoch": 8.18, + "learning_rate": 2.18328611898017e-06, + "loss": 0.4721, + "step": 22153 + }, + { + "epoch": 8.18, + "learning_rate": 2.183249168616825e-06, + "loss": 0.3527, + "step": 22154 + }, + { + "epoch": 8.18, + "learning_rate": 2.1832122182534795e-06, + "loss": 0.4686, + "step": 22155 + }, + { + "epoch": 8.18, + "learning_rate": 2.1831752678901345e-06, + "loss": 0.3909, + "step": 22156 + }, + { + "epoch": 8.18, + "learning_rate": 2.183138317526789e-06, + "loss": 0.2631, + "step": 22157 + }, + { + "epoch": 8.18, + "learning_rate": 2.1831013671634438e-06, + "loss": 0.4082, + "step": 22158 + }, + { + "epoch": 8.18, + "learning_rate": 2.1830644168000984e-06, + "loss": 0.3231, + "step": 22159 + }, + { + "epoch": 8.18, + "learning_rate": 2.1830274664367535e-06, + "loss": 0.4512, + "step": 22160 + }, + { + "epoch": 8.18, + "learning_rate": 2.182990516073408e-06, + "loss": 0.4032, + "step": 22161 + }, + { + "epoch": 8.18, + "learning_rate": 2.182953565710063e-06, + "loss": 0.3234, + "step": 22162 + }, + { + "epoch": 8.18, + "learning_rate": 2.182916615346718e-06, + "loss": 0.424, + "step": 22163 + }, + { + "epoch": 8.18, + "learning_rate": 2.1828796649833725e-06, + "loss": 0.5094, + "step": 22164 + }, + { + "epoch": 8.19, + "learning_rate": 2.182842714620027e-06, + "loss": 0.3559, + "step": 22165 + }, + { + "epoch": 8.19, + "learning_rate": 2.1828057642566817e-06, + "loss": 0.4578, + "step": 22166 + }, + { + "epoch": 8.19, + "learning_rate": 2.1827688138933368e-06, + "loss": 0.3912, + "step": 22167 + }, + { + "epoch": 8.19, + "learning_rate": 2.1827318635299914e-06, + "loss": 0.464, + "step": 22168 + }, + { + "epoch": 8.19, + "learning_rate": 2.182694913166646e-06, + "loss": 0.4729, + "step": 22169 + }, + { + "epoch": 8.19, + "learning_rate": 2.1826579628033007e-06, + "loss": 0.2732, + "step": 22170 + }, + { + "epoch": 8.19, + "learning_rate": 2.1826210124399558e-06, + "loss": 0.4999, + "step": 22171 + }, + { + "epoch": 8.19, + "learning_rate": 2.1825840620766104e-06, + "loss": 0.3706, + "step": 22172 + }, + { + "epoch": 8.19, + "learning_rate": 2.1825471117132654e-06, + "loss": 0.3825, + "step": 22173 + }, + { + "epoch": 8.19, + "learning_rate": 2.18251016134992e-06, + "loss": 0.4316, + "step": 22174 + }, + { + "epoch": 8.19, + "learning_rate": 2.1824732109865747e-06, + "loss": 0.4394, + "step": 22175 + }, + { + "epoch": 8.19, + "learning_rate": 2.1824362606232294e-06, + "loss": 0.4137, + "step": 22176 + }, + { + "epoch": 8.19, + "learning_rate": 2.182399310259884e-06, + "loss": 0.4357, + "step": 22177 + }, + { + "epoch": 8.19, + "learning_rate": 2.182362359896539e-06, + "loss": 0.3244, + "step": 22178 + }, + { + "epoch": 8.19, + "learning_rate": 2.1823254095331937e-06, + "loss": 0.4231, + "step": 22179 + }, + { + "epoch": 8.19, + "learning_rate": 2.1822884591698488e-06, + "loss": 0.2948, + "step": 22180 + }, + { + "epoch": 8.19, + "learning_rate": 2.1822515088065034e-06, + "loss": 0.4029, + "step": 22181 + }, + { + "epoch": 8.19, + "learning_rate": 2.182214558443158e-06, + "loss": 0.3436, + "step": 22182 + }, + { + "epoch": 8.19, + "learning_rate": 2.1821776080798127e-06, + "loss": 0.4728, + "step": 22183 + }, + { + "epoch": 8.19, + "learning_rate": 2.1821406577164677e-06, + "loss": 0.3355, + "step": 22184 + }, + { + "epoch": 8.19, + "learning_rate": 2.1821037073531224e-06, + "loss": 0.5406, + "step": 22185 + }, + { + "epoch": 8.19, + "learning_rate": 2.182066756989777e-06, + "loss": 0.3203, + "step": 22186 + }, + { + "epoch": 8.19, + "learning_rate": 2.182029806626432e-06, + "loss": 0.5238, + "step": 22187 + }, + { + "epoch": 8.19, + "learning_rate": 2.1819928562630867e-06, + "loss": 0.5182, + "step": 22188 + }, + { + "epoch": 8.19, + "learning_rate": 2.1819559058997413e-06, + "loss": 0.3744, + "step": 22189 + }, + { + "epoch": 8.19, + "learning_rate": 2.181918955536396e-06, + "loss": 0.4898, + "step": 22190 + }, + { + "epoch": 8.19, + "learning_rate": 2.181882005173051e-06, + "loss": 0.448, + "step": 22191 + }, + { + "epoch": 8.19, + "learning_rate": 2.1818450548097057e-06, + "loss": 0.4686, + "step": 22192 + }, + { + "epoch": 8.2, + "learning_rate": 2.1818081044463607e-06, + "loss": 0.2769, + "step": 22193 + }, + { + "epoch": 8.2, + "learning_rate": 2.1817711540830154e-06, + "loss": 0.619, + "step": 22194 + }, + { + "epoch": 8.2, + "learning_rate": 2.18173420371967e-06, + "loss": 0.3809, + "step": 22195 + }, + { + "epoch": 8.2, + "learning_rate": 2.1816972533563246e-06, + "loss": 0.4731, + "step": 22196 + }, + { + "epoch": 8.2, + "learning_rate": 2.1816603029929793e-06, + "loss": 0.3177, + "step": 22197 + }, + { + "epoch": 8.2, + "learning_rate": 2.1816233526296343e-06, + "loss": 0.2899, + "step": 22198 + }, + { + "epoch": 8.2, + "learning_rate": 2.181586402266289e-06, + "loss": 0.2915, + "step": 22199 + }, + { + "epoch": 8.2, + "learning_rate": 2.181549451902944e-06, + "loss": 0.4248, + "step": 22200 + }, + { + "epoch": 8.2, + "learning_rate": 2.1815125015395987e-06, + "loss": 0.4259, + "step": 22201 + }, + { + "epoch": 8.2, + "learning_rate": 2.1814755511762533e-06, + "loss": 0.4365, + "step": 22202 + }, + { + "epoch": 8.2, + "learning_rate": 2.181438600812908e-06, + "loss": 0.3723, + "step": 22203 + }, + { + "epoch": 8.2, + "learning_rate": 2.181401650449563e-06, + "loss": 0.3985, + "step": 22204 + }, + { + "epoch": 8.2, + "learning_rate": 2.1813647000862176e-06, + "loss": 0.3999, + "step": 22205 + }, + { + "epoch": 8.2, + "learning_rate": 2.1813277497228723e-06, + "loss": 0.3094, + "step": 22206 + }, + { + "epoch": 8.2, + "learning_rate": 2.1812907993595273e-06, + "loss": 0.2565, + "step": 22207 + }, + { + "epoch": 8.2, + "learning_rate": 2.1812538489961815e-06, + "loss": 0.4679, + "step": 22208 + }, + { + "epoch": 8.2, + "learning_rate": 2.1812168986328366e-06, + "loss": 0.3528, + "step": 22209 + }, + { + "epoch": 8.2, + "learning_rate": 2.1811799482694912e-06, + "loss": 0.4083, + "step": 22210 + }, + { + "epoch": 8.2, + "learning_rate": 2.1811429979061463e-06, + "loss": 0.4365, + "step": 22211 + }, + { + "epoch": 8.2, + "learning_rate": 2.181106047542801e-06, + "loss": 0.4617, + "step": 22212 + }, + { + "epoch": 8.2, + "learning_rate": 2.181069097179456e-06, + "loss": 0.3766, + "step": 22213 + }, + { + "epoch": 8.2, + "learning_rate": 2.18103214681611e-06, + "loss": 0.4526, + "step": 22214 + }, + { + "epoch": 8.2, + "learning_rate": 2.180995196452765e-06, + "loss": 0.5555, + "step": 22215 + }, + { + "epoch": 8.2, + "learning_rate": 2.18095824608942e-06, + "loss": 0.4061, + "step": 22216 + }, + { + "epoch": 8.2, + "learning_rate": 2.1809212957260745e-06, + "loss": 0.2972, + "step": 22217 + }, + { + "epoch": 8.2, + "learning_rate": 2.1808843453627296e-06, + "loss": 0.256, + "step": 22218 + }, + { + "epoch": 8.2, + "learning_rate": 2.1808473949993842e-06, + "loss": 0.4594, + "step": 22219 + }, + { + "epoch": 8.21, + "learning_rate": 2.180810444636039e-06, + "loss": 0.3168, + "step": 22220 + }, + { + "epoch": 8.21, + "learning_rate": 2.1807734942726935e-06, + "loss": 0.3878, + "step": 22221 + }, + { + "epoch": 8.21, + "learning_rate": 2.1807365439093486e-06, + "loss": 0.298, + "step": 22222 + }, + { + "epoch": 8.21, + "learning_rate": 2.180699593546003e-06, + "loss": 0.3233, + "step": 22223 + }, + { + "epoch": 8.21, + "learning_rate": 2.1806626431826583e-06, + "loss": 0.2851, + "step": 22224 + }, + { + "epoch": 8.21, + "learning_rate": 2.180625692819313e-06, + "loss": 0.4177, + "step": 22225 + }, + { + "epoch": 8.21, + "learning_rate": 2.1805887424559675e-06, + "loss": 0.4541, + "step": 22226 + }, + { + "epoch": 8.21, + "learning_rate": 2.180551792092622e-06, + "loss": 0.3957, + "step": 22227 + }, + { + "epoch": 8.21, + "learning_rate": 2.180514841729277e-06, + "loss": 0.3636, + "step": 22228 + }, + { + "epoch": 8.21, + "learning_rate": 2.180477891365932e-06, + "loss": 0.4115, + "step": 22229 + }, + { + "epoch": 8.21, + "learning_rate": 2.1804409410025865e-06, + "loss": 0.4526, + "step": 22230 + }, + { + "epoch": 8.21, + "learning_rate": 2.1804039906392416e-06, + "loss": 0.39, + "step": 22231 + }, + { + "epoch": 8.21, + "learning_rate": 2.180367040275896e-06, + "loss": 0.3598, + "step": 22232 + }, + { + "epoch": 8.21, + "learning_rate": 2.180330089912551e-06, + "loss": 0.3817, + "step": 22233 + }, + { + "epoch": 8.21, + "learning_rate": 2.1802931395492055e-06, + "loss": 0.3838, + "step": 22234 + }, + { + "epoch": 8.21, + "learning_rate": 2.18025618918586e-06, + "loss": 0.4667, + "step": 22235 + }, + { + "epoch": 8.21, + "learning_rate": 2.180219238822515e-06, + "loss": 0.4813, + "step": 22236 + }, + { + "epoch": 8.21, + "learning_rate": 2.18018228845917e-06, + "loss": 0.3194, + "step": 22237 + }, + { + "epoch": 8.21, + "learning_rate": 2.180145338095825e-06, + "loss": 0.3077, + "step": 22238 + }, + { + "epoch": 8.21, + "learning_rate": 2.1801083877324795e-06, + "loss": 0.3595, + "step": 22239 + }, + { + "epoch": 8.21, + "learning_rate": 2.180071437369134e-06, + "loss": 0.3317, + "step": 22240 + }, + { + "epoch": 8.21, + "learning_rate": 2.1800344870057888e-06, + "loss": 0.3305, + "step": 22241 + }, + { + "epoch": 8.21, + "learning_rate": 2.179997536642444e-06, + "loss": 0.3873, + "step": 22242 + }, + { + "epoch": 8.21, + "learning_rate": 2.1799605862790985e-06, + "loss": 0.4212, + "step": 22243 + }, + { + "epoch": 8.21, + "learning_rate": 2.1799236359157535e-06, + "loss": 0.4835, + "step": 22244 + }, + { + "epoch": 8.21, + "learning_rate": 2.179886685552408e-06, + "loss": 0.4121, + "step": 22245 + }, + { + "epoch": 8.21, + "learning_rate": 2.179849735189063e-06, + "loss": 0.3956, + "step": 22246 + }, + { + "epoch": 8.22, + "learning_rate": 2.1798127848257174e-06, + "loss": 0.5702, + "step": 22247 + }, + { + "epoch": 8.22, + "learning_rate": 2.179775834462372e-06, + "loss": 0.3877, + "step": 22248 + }, + { + "epoch": 8.22, + "learning_rate": 2.179738884099027e-06, + "loss": 0.3255, + "step": 22249 + }, + { + "epoch": 8.22, + "learning_rate": 2.1797019337356818e-06, + "loss": 0.4362, + "step": 22250 + }, + { + "epoch": 8.22, + "learning_rate": 2.179664983372337e-06, + "loss": 0.3976, + "step": 22251 + }, + { + "epoch": 8.22, + "learning_rate": 2.1796280330089915e-06, + "loss": 0.4744, + "step": 22252 + }, + { + "epoch": 8.22, + "learning_rate": 2.179591082645646e-06, + "loss": 0.4358, + "step": 22253 + }, + { + "epoch": 8.22, + "learning_rate": 2.1795541322823007e-06, + "loss": 0.3085, + "step": 22254 + }, + { + "epoch": 8.22, + "learning_rate": 2.1795171819189554e-06, + "loss": 0.4341, + "step": 22255 + }, + { + "epoch": 8.22, + "learning_rate": 2.1794802315556104e-06, + "loss": 0.4324, + "step": 22256 + }, + { + "epoch": 8.22, + "learning_rate": 2.179443281192265e-06, + "loss": 0.4277, + "step": 22257 + }, + { + "epoch": 8.22, + "learning_rate": 2.17940633082892e-06, + "loss": 0.2858, + "step": 22258 + }, + { + "epoch": 8.22, + "learning_rate": 2.1793693804655743e-06, + "loss": 0.3862, + "step": 22259 + }, + { + "epoch": 8.22, + "learning_rate": 2.1793324301022294e-06, + "loss": 0.4462, + "step": 22260 + }, + { + "epoch": 8.22, + "learning_rate": 2.179295479738884e-06, + "loss": 0.3157, + "step": 22261 + }, + { + "epoch": 8.22, + "learning_rate": 2.179258529375539e-06, + "loss": 0.2824, + "step": 22262 + }, + { + "epoch": 8.22, + "learning_rate": 2.1792215790121937e-06, + "loss": 0.3859, + "step": 22263 + }, + { + "epoch": 8.22, + "learning_rate": 2.179184628648849e-06, + "loss": 0.3761, + "step": 22264 + }, + { + "epoch": 8.22, + "learning_rate": 2.179147678285503e-06, + "loss": 0.3642, + "step": 22265 + }, + { + "epoch": 8.22, + "learning_rate": 2.1791107279221576e-06, + "loss": 0.4667, + "step": 22266 + }, + { + "epoch": 8.22, + "learning_rate": 2.1790737775588127e-06, + "loss": 0.4166, + "step": 22267 + }, + { + "epoch": 8.22, + "learning_rate": 2.1790368271954673e-06, + "loss": 0.3829, + "step": 22268 + }, + { + "epoch": 8.22, + "learning_rate": 2.1789998768321224e-06, + "loss": 0.3733, + "step": 22269 + }, + { + "epoch": 8.22, + "learning_rate": 2.178962926468777e-06, + "loss": 0.4507, + "step": 22270 + }, + { + "epoch": 8.22, + "learning_rate": 2.1789259761054317e-06, + "loss": 0.2334, + "step": 22271 + }, + { + "epoch": 8.22, + "learning_rate": 2.1788890257420863e-06, + "loss": 0.4455, + "step": 22272 + }, + { + "epoch": 8.22, + "learning_rate": 2.1788520753787414e-06, + "loss": 0.3712, + "step": 22273 + }, + { + "epoch": 8.23, + "learning_rate": 2.178815125015396e-06, + "loss": 0.4877, + "step": 22274 + }, + { + "epoch": 8.23, + "learning_rate": 2.1787781746520506e-06, + "loss": 0.3921, + "step": 22275 + }, + { + "epoch": 8.23, + "learning_rate": 2.1787412242887057e-06, + "loss": 0.3182, + "step": 22276 + }, + { + "epoch": 8.23, + "learning_rate": 2.1787042739253603e-06, + "loss": 0.4195, + "step": 22277 + }, + { + "epoch": 8.23, + "learning_rate": 2.178667323562015e-06, + "loss": 0.4392, + "step": 22278 + }, + { + "epoch": 8.23, + "learning_rate": 2.1786303731986696e-06, + "loss": 0.3848, + "step": 22279 + }, + { + "epoch": 8.23, + "learning_rate": 2.1785934228353247e-06, + "loss": 0.392, + "step": 22280 + }, + { + "epoch": 8.23, + "learning_rate": 2.1785564724719793e-06, + "loss": 0.4763, + "step": 22281 + }, + { + "epoch": 8.23, + "learning_rate": 2.1785195221086344e-06, + "loss": 0.2637, + "step": 22282 + }, + { + "epoch": 8.23, + "learning_rate": 2.178482571745289e-06, + "loss": 0.3725, + "step": 22283 + }, + { + "epoch": 8.23, + "learning_rate": 2.1784456213819436e-06, + "loss": 0.3375, + "step": 22284 + }, + { + "epoch": 8.23, + "learning_rate": 2.1784086710185983e-06, + "loss": 0.3024, + "step": 22285 + }, + { + "epoch": 8.23, + "learning_rate": 2.178371720655253e-06, + "loss": 0.2887, + "step": 22286 + }, + { + "epoch": 8.23, + "learning_rate": 2.178334770291908e-06, + "loss": 0.3643, + "step": 22287 + }, + { + "epoch": 8.23, + "learning_rate": 2.1782978199285626e-06, + "loss": 0.5235, + "step": 22288 + }, + { + "epoch": 8.23, + "learning_rate": 2.1782608695652177e-06, + "loss": 0.5096, + "step": 22289 + }, + { + "epoch": 8.23, + "learning_rate": 2.1782239192018723e-06, + "loss": 0.4203, + "step": 22290 + }, + { + "epoch": 8.23, + "learning_rate": 2.178186968838527e-06, + "loss": 0.2748, + "step": 22291 + }, + { + "epoch": 8.23, + "learning_rate": 2.1781500184751816e-06, + "loss": 0.5047, + "step": 22292 + }, + { + "epoch": 8.23, + "learning_rate": 2.1781130681118366e-06, + "loss": 0.3794, + "step": 22293 + }, + { + "epoch": 8.23, + "learning_rate": 2.1780761177484913e-06, + "loss": 0.419, + "step": 22294 + }, + { + "epoch": 8.23, + "learning_rate": 2.178039167385146e-06, + "loss": 0.371, + "step": 22295 + }, + { + "epoch": 8.23, + "learning_rate": 2.178002217021801e-06, + "loss": 0.4731, + "step": 22296 + }, + { + "epoch": 8.23, + "learning_rate": 2.1779652666584556e-06, + "loss": 0.4193, + "step": 22297 + }, + { + "epoch": 8.23, + "learning_rate": 2.1779283162951102e-06, + "loss": 0.4221, + "step": 22298 + }, + { + "epoch": 8.23, + "learning_rate": 2.177891365931765e-06, + "loss": 0.2714, + "step": 22299 + }, + { + "epoch": 8.23, + "learning_rate": 2.17785441556842e-06, + "loss": 0.3481, + "step": 22300 + }, + { + "epoch": 8.24, + "learning_rate": 2.1778174652050746e-06, + "loss": 0.4888, + "step": 22301 + }, + { + "epoch": 8.24, + "learning_rate": 2.1777805148417296e-06, + "loss": 0.4777, + "step": 22302 + }, + { + "epoch": 8.24, + "learning_rate": 2.1777435644783843e-06, + "loss": 0.4686, + "step": 22303 + }, + { + "epoch": 8.24, + "learning_rate": 2.177706614115039e-06, + "loss": 0.478, + "step": 22304 + }, + { + "epoch": 8.24, + "learning_rate": 2.1776696637516936e-06, + "loss": 0.3946, + "step": 22305 + }, + { + "epoch": 8.24, + "learning_rate": 2.177632713388348e-06, + "loss": 0.32, + "step": 22306 + }, + { + "epoch": 8.24, + "learning_rate": 2.1775957630250032e-06, + "loss": 0.5616, + "step": 22307 + }, + { + "epoch": 8.24, + "learning_rate": 2.177558812661658e-06, + "loss": 0.4757, + "step": 22308 + }, + { + "epoch": 8.24, + "learning_rate": 2.177521862298313e-06, + "loss": 0.3906, + "step": 22309 + }, + { + "epoch": 8.24, + "learning_rate": 2.177484911934967e-06, + "loss": 0.3225, + "step": 22310 + }, + { + "epoch": 8.24, + "learning_rate": 2.1774479615716222e-06, + "loss": 0.3614, + "step": 22311 + }, + { + "epoch": 8.24, + "learning_rate": 2.177411011208277e-06, + "loss": 0.4276, + "step": 22312 + }, + { + "epoch": 8.24, + "learning_rate": 2.177374060844932e-06, + "loss": 0.3312, + "step": 22313 + }, + { + "epoch": 8.24, + "learning_rate": 2.1773371104815866e-06, + "loss": 0.4642, + "step": 22314 + }, + { + "epoch": 8.24, + "learning_rate": 2.177300160118241e-06, + "loss": 0.4806, + "step": 22315 + }, + { + "epoch": 8.24, + "learning_rate": 2.177263209754896e-06, + "loss": 0.402, + "step": 22316 + }, + { + "epoch": 8.24, + "learning_rate": 2.1772262593915505e-06, + "loss": 0.477, + "step": 22317 + }, + { + "epoch": 8.24, + "learning_rate": 2.1771893090282055e-06, + "loss": 0.4273, + "step": 22318 + }, + { + "epoch": 8.24, + "learning_rate": 2.17715235866486e-06, + "loss": 0.3338, + "step": 22319 + }, + { + "epoch": 8.24, + "learning_rate": 2.1771154083015152e-06, + "loss": 0.3689, + "step": 22320 + }, + { + "epoch": 8.24, + "learning_rate": 2.17707845793817e-06, + "loss": 0.4187, + "step": 22321 + }, + { + "epoch": 8.24, + "learning_rate": 2.1770415075748245e-06, + "loss": 0.362, + "step": 22322 + }, + { + "epoch": 8.24, + "learning_rate": 2.177004557211479e-06, + "loss": 0.4529, + "step": 22323 + }, + { + "epoch": 8.24, + "learning_rate": 2.1769676068481338e-06, + "loss": 0.3782, + "step": 22324 + }, + { + "epoch": 8.24, + "learning_rate": 2.176930656484789e-06, + "loss": 0.4445, + "step": 22325 + }, + { + "epoch": 8.24, + "learning_rate": 2.1768937061214435e-06, + "loss": 0.4477, + "step": 22326 + }, + { + "epoch": 8.24, + "learning_rate": 2.1768567557580985e-06, + "loss": 0.4546, + "step": 22327 + }, + { + "epoch": 8.25, + "learning_rate": 2.176819805394753e-06, + "loss": 0.4019, + "step": 22328 + }, + { + "epoch": 8.25, + "learning_rate": 2.176782855031408e-06, + "loss": 0.3682, + "step": 22329 + }, + { + "epoch": 8.25, + "learning_rate": 2.1767459046680624e-06, + "loss": 0.4001, + "step": 22330 + }, + { + "epoch": 8.25, + "learning_rate": 2.1767089543047175e-06, + "loss": 0.5558, + "step": 22331 + }, + { + "epoch": 8.25, + "learning_rate": 2.176672003941372e-06, + "loss": 0.3045, + "step": 22332 + }, + { + "epoch": 8.25, + "learning_rate": 2.176635053578027e-06, + "loss": 0.4944, + "step": 22333 + }, + { + "epoch": 8.25, + "learning_rate": 2.176598103214682e-06, + "loss": 0.4836, + "step": 22334 + }, + { + "epoch": 8.25, + "learning_rate": 2.1765611528513365e-06, + "loss": 0.3887, + "step": 22335 + }, + { + "epoch": 8.25, + "learning_rate": 2.176524202487991e-06, + "loss": 0.3669, + "step": 22336 + }, + { + "epoch": 8.25, + "learning_rate": 2.1764872521246457e-06, + "loss": 0.5836, + "step": 22337 + }, + { + "epoch": 8.25, + "learning_rate": 2.176450301761301e-06, + "loss": 0.3635, + "step": 22338 + }, + { + "epoch": 8.25, + "learning_rate": 2.1764133513979554e-06, + "loss": 0.3771, + "step": 22339 + }, + { + "epoch": 8.25, + "learning_rate": 2.1763764010346105e-06, + "loss": 0.3173, + "step": 22340 + }, + { + "epoch": 8.25, + "learning_rate": 2.176339450671265e-06, + "loss": 0.4075, + "step": 22341 + }, + { + "epoch": 8.25, + "learning_rate": 2.1763025003079198e-06, + "loss": 0.2901, + "step": 22342 + }, + { + "epoch": 8.25, + "learning_rate": 2.1762655499445744e-06, + "loss": 0.3043, + "step": 22343 + }, + { + "epoch": 8.25, + "learning_rate": 2.176228599581229e-06, + "loss": 0.3181, + "step": 22344 + }, + { + "epoch": 8.25, + "learning_rate": 2.176191649217884e-06, + "loss": 0.3597, + "step": 22345 + }, + { + "epoch": 8.25, + "learning_rate": 2.1761546988545387e-06, + "loss": 0.3131, + "step": 22346 + }, + { + "epoch": 8.25, + "learning_rate": 2.1761177484911938e-06, + "loss": 0.4249, + "step": 22347 + }, + { + "epoch": 8.25, + "learning_rate": 2.1760807981278484e-06, + "loss": 0.3511, + "step": 22348 + }, + { + "epoch": 8.25, + "learning_rate": 2.176043847764503e-06, + "loss": 0.4459, + "step": 22349 + }, + { + "epoch": 8.25, + "learning_rate": 2.1760068974011577e-06, + "loss": 0.5901, + "step": 22350 + }, + { + "epoch": 8.25, + "learning_rate": 2.1759699470378128e-06, + "loss": 0.4003, + "step": 22351 + }, + { + "epoch": 8.25, + "learning_rate": 2.1759329966744674e-06, + "loss": 0.4283, + "step": 22352 + }, + { + "epoch": 8.25, + "learning_rate": 2.1758960463111225e-06, + "loss": 0.3328, + "step": 22353 + }, + { + "epoch": 8.25, + "learning_rate": 2.1758590959477767e-06, + "loss": 0.4403, + "step": 22354 + }, + { + "epoch": 8.26, + "learning_rate": 2.1758221455844313e-06, + "loss": 0.2413, + "step": 22355 + }, + { + "epoch": 8.26, + "learning_rate": 2.1757851952210864e-06, + "loss": 0.3166, + "step": 22356 + }, + { + "epoch": 8.26, + "learning_rate": 2.175748244857741e-06, + "loss": 0.3671, + "step": 22357 + }, + { + "epoch": 8.26, + "learning_rate": 2.175711294494396e-06, + "loss": 0.3217, + "step": 22358 + }, + { + "epoch": 8.26, + "learning_rate": 2.1756743441310507e-06, + "loss": 0.6898, + "step": 22359 + }, + { + "epoch": 8.26, + "learning_rate": 2.1756373937677053e-06, + "loss": 0.3847, + "step": 22360 + }, + { + "epoch": 8.26, + "learning_rate": 2.17560044340436e-06, + "loss": 0.4877, + "step": 22361 + }, + { + "epoch": 8.26, + "learning_rate": 2.175563493041015e-06, + "loss": 0.4322, + "step": 22362 + }, + { + "epoch": 8.26, + "learning_rate": 2.1755265426776697e-06, + "loss": 0.4946, + "step": 22363 + }, + { + "epoch": 8.26, + "learning_rate": 2.1754895923143243e-06, + "loss": 0.4553, + "step": 22364 + }, + { + "epoch": 8.26, + "learning_rate": 2.1754526419509794e-06, + "loss": 0.3462, + "step": 22365 + }, + { + "epoch": 8.26, + "learning_rate": 2.175415691587634e-06, + "loss": 0.367, + "step": 22366 + }, + { + "epoch": 8.26, + "learning_rate": 2.1753787412242886e-06, + "loss": 0.3874, + "step": 22367 + }, + { + "epoch": 8.26, + "learning_rate": 2.1753417908609433e-06, + "loss": 0.3038, + "step": 22368 + }, + { + "epoch": 8.26, + "learning_rate": 2.1753048404975983e-06, + "loss": 0.3535, + "step": 22369 + }, + { + "epoch": 8.26, + "learning_rate": 2.175267890134253e-06, + "loss": 0.5063, + "step": 22370 + }, + { + "epoch": 8.26, + "learning_rate": 2.175230939770908e-06, + "loss": 0.3781, + "step": 22371 + }, + { + "epoch": 8.26, + "learning_rate": 2.1751939894075627e-06, + "loss": 0.4014, + "step": 22372 + }, + { + "epoch": 8.26, + "learning_rate": 2.1751570390442173e-06, + "loss": 0.4641, + "step": 22373 + }, + { + "epoch": 8.26, + "learning_rate": 2.175120088680872e-06, + "loss": 0.345, + "step": 22374 + }, + { + "epoch": 8.26, + "learning_rate": 2.1750831383175266e-06, + "loss": 0.3756, + "step": 22375 + }, + { + "epoch": 8.26, + "learning_rate": 2.1750461879541816e-06, + "loss": 0.3389, + "step": 22376 + }, + { + "epoch": 8.26, + "learning_rate": 2.1750092375908363e-06, + "loss": 0.3358, + "step": 22377 + }, + { + "epoch": 8.26, + "learning_rate": 2.1749722872274913e-06, + "loss": 0.4287, + "step": 22378 + }, + { + "epoch": 8.26, + "learning_rate": 2.174935336864146e-06, + "loss": 0.3181, + "step": 22379 + }, + { + "epoch": 8.26, + "learning_rate": 2.1748983865008006e-06, + "loss": 0.3007, + "step": 22380 + }, + { + "epoch": 8.26, + "learning_rate": 2.1748614361374552e-06, + "loss": 0.5266, + "step": 22381 + }, + { + "epoch": 8.27, + "learning_rate": 2.1748244857741103e-06, + "loss": 0.3931, + "step": 22382 + }, + { + "epoch": 8.27, + "learning_rate": 2.174787535410765e-06, + "loss": 0.4184, + "step": 22383 + }, + { + "epoch": 8.27, + "learning_rate": 2.1747505850474196e-06, + "loss": 0.2435, + "step": 22384 + }, + { + "epoch": 8.27, + "learning_rate": 2.1747136346840746e-06, + "loss": 0.273, + "step": 22385 + }, + { + "epoch": 8.27, + "learning_rate": 2.1746766843207293e-06, + "loss": 0.3679, + "step": 22386 + }, + { + "epoch": 8.27, + "learning_rate": 2.174639733957384e-06, + "loss": 0.3315, + "step": 22387 + }, + { + "epoch": 8.27, + "learning_rate": 2.1746027835940385e-06, + "loss": 0.3977, + "step": 22388 + }, + { + "epoch": 8.27, + "learning_rate": 2.1745658332306936e-06, + "loss": 0.3347, + "step": 22389 + }, + { + "epoch": 8.27, + "learning_rate": 2.1745288828673482e-06, + "loss": 0.3263, + "step": 22390 + }, + { + "epoch": 8.27, + "learning_rate": 2.1744919325040033e-06, + "loss": 0.4252, + "step": 22391 + }, + { + "epoch": 8.27, + "learning_rate": 2.174454982140658e-06, + "loss": 0.4256, + "step": 22392 + }, + { + "epoch": 8.27, + "learning_rate": 2.1744180317773126e-06, + "loss": 0.4183, + "step": 22393 + }, + { + "epoch": 8.27, + "learning_rate": 2.174381081413967e-06, + "loss": 0.3564, + "step": 22394 + }, + { + "epoch": 8.27, + "learning_rate": 2.174344131050622e-06, + "loss": 0.4501, + "step": 22395 + }, + { + "epoch": 8.27, + "learning_rate": 2.174307180687277e-06, + "loss": 0.5305, + "step": 22396 + }, + { + "epoch": 8.27, + "learning_rate": 2.1742702303239315e-06, + "loss": 0.3675, + "step": 22397 + }, + { + "epoch": 8.27, + "learning_rate": 2.1742332799605866e-06, + "loss": 0.4018, + "step": 22398 + }, + { + "epoch": 8.27, + "learning_rate": 2.174196329597241e-06, + "loss": 0.4126, + "step": 22399 + }, + { + "epoch": 8.27, + "learning_rate": 2.174159379233896e-06, + "loss": 0.3735, + "step": 22400 + }, + { + "epoch": 8.27, + "learning_rate": 2.1741224288705505e-06, + "loss": 0.4334, + "step": 22401 + }, + { + "epoch": 8.27, + "learning_rate": 2.1740854785072056e-06, + "loss": 0.3783, + "step": 22402 + }, + { + "epoch": 8.27, + "learning_rate": 2.17404852814386e-06, + "loss": 0.4095, + "step": 22403 + }, + { + "epoch": 8.27, + "learning_rate": 2.174011577780515e-06, + "loss": 0.4771, + "step": 22404 + }, + { + "epoch": 8.27, + "learning_rate": 2.1739746274171695e-06, + "loss": 0.4695, + "step": 22405 + }, + { + "epoch": 8.27, + "learning_rate": 2.173937677053824e-06, + "loss": 0.609, + "step": 22406 + }, + { + "epoch": 8.27, + "learning_rate": 2.173900726690479e-06, + "loss": 0.4107, + "step": 22407 + }, + { + "epoch": 8.27, + "learning_rate": 2.173863776327134e-06, + "loss": 0.3937, + "step": 22408 + }, + { + "epoch": 8.28, + "learning_rate": 2.173826825963789e-06, + "loss": 0.3984, + "step": 22409 + }, + { + "epoch": 8.28, + "learning_rate": 2.1737898756004435e-06, + "loss": 0.4706, + "step": 22410 + }, + { + "epoch": 8.28, + "learning_rate": 2.173752925237098e-06, + "loss": 0.3982, + "step": 22411 + }, + { + "epoch": 8.28, + "learning_rate": 2.1737159748737528e-06, + "loss": 0.3348, + "step": 22412 + }, + { + "epoch": 8.28, + "learning_rate": 2.173679024510408e-06, + "loss": 0.4164, + "step": 22413 + }, + { + "epoch": 8.28, + "learning_rate": 2.1736420741470625e-06, + "loss": 0.4134, + "step": 22414 + }, + { + "epoch": 8.28, + "learning_rate": 2.173605123783717e-06, + "loss": 0.3882, + "step": 22415 + }, + { + "epoch": 8.28, + "learning_rate": 2.173568173420372e-06, + "loss": 0.323, + "step": 22416 + }, + { + "epoch": 8.28, + "learning_rate": 2.173531223057027e-06, + "loss": 0.3378, + "step": 22417 + }, + { + "epoch": 8.28, + "learning_rate": 2.1734942726936814e-06, + "loss": 0.4888, + "step": 22418 + }, + { + "epoch": 8.28, + "learning_rate": 2.173457322330336e-06, + "loss": 0.3003, + "step": 22419 + }, + { + "epoch": 8.28, + "learning_rate": 2.173420371966991e-06, + "loss": 0.3676, + "step": 22420 + }, + { + "epoch": 8.28, + "learning_rate": 2.1733834216036458e-06, + "loss": 0.4612, + "step": 22421 + }, + { + "epoch": 8.28, + "learning_rate": 2.173346471240301e-06, + "loss": 0.4743, + "step": 22422 + }, + { + "epoch": 8.28, + "learning_rate": 2.1733095208769555e-06, + "loss": 0.3225, + "step": 22423 + }, + { + "epoch": 8.28, + "learning_rate": 2.17327257051361e-06, + "loss": 0.4328, + "step": 22424 + }, + { + "epoch": 8.28, + "learning_rate": 2.1732356201502647e-06, + "loss": 0.4201, + "step": 22425 + }, + { + "epoch": 8.28, + "learning_rate": 2.1731986697869194e-06, + "loss": 0.3628, + "step": 22426 + }, + { + "epoch": 8.28, + "learning_rate": 2.1731617194235744e-06, + "loss": 0.4423, + "step": 22427 + }, + { + "epoch": 8.28, + "learning_rate": 2.173124769060229e-06, + "loss": 0.3843, + "step": 22428 + }, + { + "epoch": 8.28, + "learning_rate": 2.173087818696884e-06, + "loss": 0.3719, + "step": 22429 + }, + { + "epoch": 8.28, + "learning_rate": 2.1730508683335388e-06, + "loss": 0.3241, + "step": 22430 + }, + { + "epoch": 8.28, + "learning_rate": 2.1730139179701934e-06, + "loss": 0.2737, + "step": 22431 + }, + { + "epoch": 8.28, + "learning_rate": 2.172976967606848e-06, + "loss": 0.3475, + "step": 22432 + }, + { + "epoch": 8.28, + "learning_rate": 2.172940017243503e-06, + "loss": 0.4164, + "step": 22433 + }, + { + "epoch": 8.28, + "learning_rate": 2.1729030668801577e-06, + "loss": 0.3823, + "step": 22434 + }, + { + "epoch": 8.28, + "learning_rate": 2.1728661165168124e-06, + "loss": 0.3322, + "step": 22435 + }, + { + "epoch": 8.29, + "learning_rate": 2.1728291661534674e-06, + "loss": 0.4009, + "step": 22436 + }, + { + "epoch": 8.29, + "learning_rate": 2.172792215790122e-06, + "loss": 0.3775, + "step": 22437 + }, + { + "epoch": 8.29, + "learning_rate": 2.1727552654267767e-06, + "loss": 0.427, + "step": 22438 + }, + { + "epoch": 8.29, + "learning_rate": 2.1727183150634314e-06, + "loss": 0.4288, + "step": 22439 + }, + { + "epoch": 8.29, + "learning_rate": 2.1726813647000864e-06, + "loss": 0.2266, + "step": 22440 + }, + { + "epoch": 8.29, + "learning_rate": 2.172644414336741e-06, + "loss": 0.3626, + "step": 22441 + }, + { + "epoch": 8.29, + "learning_rate": 2.172607463973396e-06, + "loss": 0.4448, + "step": 22442 + }, + { + "epoch": 8.29, + "learning_rate": 2.1725705136100507e-06, + "loss": 0.3747, + "step": 22443 + }, + { + "epoch": 8.29, + "learning_rate": 2.172533563246705e-06, + "loss": 0.468, + "step": 22444 + }, + { + "epoch": 8.29, + "learning_rate": 2.17249661288336e-06, + "loss": 0.491, + "step": 22445 + }, + { + "epoch": 8.29, + "learning_rate": 2.1724596625200147e-06, + "loss": 0.3612, + "step": 22446 + }, + { + "epoch": 8.29, + "learning_rate": 2.1724227121566697e-06, + "loss": 0.3882, + "step": 22447 + }, + { + "epoch": 8.29, + "learning_rate": 2.1723857617933243e-06, + "loss": 0.3813, + "step": 22448 + }, + { + "epoch": 8.29, + "learning_rate": 2.1723488114299794e-06, + "loss": 0.2838, + "step": 22449 + }, + { + "epoch": 8.29, + "learning_rate": 2.1723118610666336e-06, + "loss": 0.4079, + "step": 22450 + }, + { + "epoch": 8.29, + "learning_rate": 2.1722749107032887e-06, + "loss": 0.4637, + "step": 22451 + }, + { + "epoch": 8.29, + "learning_rate": 2.1722379603399433e-06, + "loss": 0.386, + "step": 22452 + }, + { + "epoch": 8.29, + "learning_rate": 2.1722010099765984e-06, + "loss": 0.3356, + "step": 22453 + }, + { + "epoch": 8.29, + "learning_rate": 2.172164059613253e-06, + "loss": 0.4003, + "step": 22454 + }, + { + "epoch": 8.29, + "learning_rate": 2.1721271092499077e-06, + "loss": 0.357, + "step": 22455 + }, + { + "epoch": 8.29, + "learning_rate": 2.1720901588865623e-06, + "loss": 0.4909, + "step": 22456 + }, + { + "epoch": 8.29, + "learning_rate": 2.172053208523217e-06, + "loss": 0.3681, + "step": 22457 + }, + { + "epoch": 8.29, + "learning_rate": 2.172016258159872e-06, + "loss": 0.3536, + "step": 22458 + }, + { + "epoch": 8.29, + "learning_rate": 2.1719793077965266e-06, + "loss": 0.4022, + "step": 22459 + }, + { + "epoch": 8.29, + "learning_rate": 2.1719423574331817e-06, + "loss": 0.4705, + "step": 22460 + }, + { + "epoch": 8.29, + "learning_rate": 2.1719054070698363e-06, + "loss": 0.4847, + "step": 22461 + }, + { + "epoch": 8.29, + "learning_rate": 2.171868456706491e-06, + "loss": 0.4005, + "step": 22462 + }, + { + "epoch": 8.3, + "learning_rate": 2.1718315063431456e-06, + "loss": 0.3978, + "step": 22463 + }, + { + "epoch": 8.3, + "learning_rate": 2.1717945559798002e-06, + "loss": 0.299, + "step": 22464 + }, + { + "epoch": 8.3, + "learning_rate": 2.1717576056164553e-06, + "loss": 0.3779, + "step": 22465 + }, + { + "epoch": 8.3, + "learning_rate": 2.17172065525311e-06, + "loss": 0.4437, + "step": 22466 + }, + { + "epoch": 8.3, + "learning_rate": 2.171683704889765e-06, + "loss": 0.4061, + "step": 22467 + }, + { + "epoch": 8.3, + "learning_rate": 2.1716467545264196e-06, + "loss": 0.366, + "step": 22468 + }, + { + "epoch": 8.3, + "learning_rate": 2.1716098041630743e-06, + "loss": 0.3922, + "step": 22469 + }, + { + "epoch": 8.3, + "learning_rate": 2.171572853799729e-06, + "loss": 0.3989, + "step": 22470 + }, + { + "epoch": 8.3, + "learning_rate": 2.171535903436384e-06, + "loss": 0.4913, + "step": 22471 + }, + { + "epoch": 8.3, + "learning_rate": 2.1714989530730386e-06, + "loss": 0.3409, + "step": 22472 + }, + { + "epoch": 8.3, + "learning_rate": 2.1714620027096936e-06, + "loss": 0.3955, + "step": 22473 + }, + { + "epoch": 8.3, + "learning_rate": 2.1714250523463483e-06, + "loss": 0.4274, + "step": 22474 + }, + { + "epoch": 8.3, + "learning_rate": 2.171388101983003e-06, + "loss": 0.4204, + "step": 22475 + }, + { + "epoch": 8.3, + "learning_rate": 2.1713511516196576e-06, + "loss": 0.3332, + "step": 22476 + }, + { + "epoch": 8.3, + "learning_rate": 2.171314201256312e-06, + "loss": 0.4086, + "step": 22477 + }, + { + "epoch": 8.3, + "learning_rate": 2.1712772508929673e-06, + "loss": 0.3538, + "step": 22478 + }, + { + "epoch": 8.3, + "learning_rate": 2.171240300529622e-06, + "loss": 0.2826, + "step": 22479 + }, + { + "epoch": 8.3, + "learning_rate": 2.171203350166277e-06, + "loss": 0.321, + "step": 22480 + }, + { + "epoch": 8.3, + "learning_rate": 2.1711663998029316e-06, + "loss": 0.3665, + "step": 22481 + }, + { + "epoch": 8.3, + "learning_rate": 2.1711294494395862e-06, + "loss": 0.5413, + "step": 22482 + }, + { + "epoch": 8.3, + "learning_rate": 2.171092499076241e-06, + "loss": 0.3546, + "step": 22483 + }, + { + "epoch": 8.3, + "learning_rate": 2.1710555487128955e-06, + "loss": 0.4373, + "step": 22484 + }, + { + "epoch": 8.3, + "learning_rate": 2.1710185983495506e-06, + "loss": 0.4188, + "step": 22485 + }, + { + "epoch": 8.3, + "learning_rate": 2.170981647986205e-06, + "loss": 0.428, + "step": 22486 + }, + { + "epoch": 8.3, + "learning_rate": 2.1709446976228603e-06, + "loss": 0.2867, + "step": 22487 + }, + { + "epoch": 8.3, + "learning_rate": 2.170907747259515e-06, + "loss": 0.3526, + "step": 22488 + }, + { + "epoch": 8.3, + "learning_rate": 2.1708707968961695e-06, + "loss": 0.4102, + "step": 22489 + }, + { + "epoch": 8.31, + "learning_rate": 2.170833846532824e-06, + "loss": 0.281, + "step": 22490 + }, + { + "epoch": 8.31, + "learning_rate": 2.1707968961694792e-06, + "loss": 0.3043, + "step": 22491 + }, + { + "epoch": 8.31, + "learning_rate": 2.170759945806134e-06, + "loss": 0.3203, + "step": 22492 + }, + { + "epoch": 8.31, + "learning_rate": 2.170722995442789e-06, + "loss": 0.425, + "step": 22493 + }, + { + "epoch": 8.31, + "learning_rate": 2.1706860450794436e-06, + "loss": 0.3455, + "step": 22494 + }, + { + "epoch": 8.31, + "learning_rate": 2.1706490947160978e-06, + "loss": 0.4487, + "step": 22495 + }, + { + "epoch": 8.31, + "learning_rate": 2.170612144352753e-06, + "loss": 0.3838, + "step": 22496 + }, + { + "epoch": 8.31, + "learning_rate": 2.1705751939894075e-06, + "loss": 0.3824, + "step": 22497 + }, + { + "epoch": 8.31, + "learning_rate": 2.1705382436260625e-06, + "loss": 0.3557, + "step": 22498 + }, + { + "epoch": 8.31, + "learning_rate": 2.170501293262717e-06, + "loss": 0.499, + "step": 22499 + }, + { + "epoch": 8.31, + "learning_rate": 2.1704643428993722e-06, + "loss": 0.3383, + "step": 22500 + }, + { + "epoch": 8.31, + "learning_rate": 2.1704273925360264e-06, + "loss": 0.4016, + "step": 22501 + }, + { + "epoch": 8.31, + "learning_rate": 2.1703904421726815e-06, + "loss": 0.2256, + "step": 22502 + }, + { + "epoch": 8.31, + "learning_rate": 2.170353491809336e-06, + "loss": 0.4501, + "step": 22503 + }, + { + "epoch": 8.31, + "learning_rate": 2.1703165414459908e-06, + "loss": 0.4319, + "step": 22504 + }, + { + "epoch": 8.31, + "learning_rate": 2.170279591082646e-06, + "loss": 0.2564, + "step": 22505 + }, + { + "epoch": 8.31, + "learning_rate": 2.1702426407193005e-06, + "loss": 0.2871, + "step": 22506 + }, + { + "epoch": 8.31, + "learning_rate": 2.170205690355955e-06, + "loss": 0.4524, + "step": 22507 + }, + { + "epoch": 8.31, + "learning_rate": 2.1701687399926097e-06, + "loss": 0.4427, + "step": 22508 + }, + { + "epoch": 8.31, + "learning_rate": 2.170131789629265e-06, + "loss": 0.4196, + "step": 22509 + }, + { + "epoch": 8.31, + "learning_rate": 2.1700948392659194e-06, + "loss": 0.3326, + "step": 22510 + }, + { + "epoch": 8.31, + "learning_rate": 2.1700578889025745e-06, + "loss": 0.3544, + "step": 22511 + }, + { + "epoch": 8.31, + "learning_rate": 2.170020938539229e-06, + "loss": 0.4712, + "step": 22512 + }, + { + "epoch": 8.31, + "learning_rate": 2.1699839881758838e-06, + "loss": 0.3276, + "step": 22513 + }, + { + "epoch": 8.31, + "learning_rate": 2.1699470378125384e-06, + "loss": 0.336, + "step": 22514 + }, + { + "epoch": 8.31, + "learning_rate": 2.169910087449193e-06, + "loss": 0.2708, + "step": 22515 + }, + { + "epoch": 8.31, + "learning_rate": 2.169873137085848e-06, + "loss": 0.3093, + "step": 22516 + }, + { + "epoch": 8.31, + "learning_rate": 2.1698361867225027e-06, + "loss": 0.3429, + "step": 22517 + }, + { + "epoch": 8.32, + "learning_rate": 2.169799236359158e-06, + "loss": 0.4296, + "step": 22518 + }, + { + "epoch": 8.32, + "learning_rate": 2.1697622859958124e-06, + "loss": 0.4137, + "step": 22519 + }, + { + "epoch": 8.32, + "learning_rate": 2.169725335632467e-06, + "loss": 0.2795, + "step": 22520 + }, + { + "epoch": 8.32, + "learning_rate": 2.1696883852691217e-06, + "loss": 0.5226, + "step": 22521 + }, + { + "epoch": 8.32, + "learning_rate": 2.1696514349057768e-06, + "loss": 0.3919, + "step": 22522 + }, + { + "epoch": 8.32, + "learning_rate": 2.1696144845424314e-06, + "loss": 0.259, + "step": 22523 + }, + { + "epoch": 8.32, + "learning_rate": 2.169577534179086e-06, + "loss": 0.3829, + "step": 22524 + }, + { + "epoch": 8.32, + "learning_rate": 2.169540583815741e-06, + "loss": 0.372, + "step": 22525 + }, + { + "epoch": 8.32, + "learning_rate": 2.1695036334523957e-06, + "loss": 0.4771, + "step": 22526 + }, + { + "epoch": 8.32, + "learning_rate": 2.1694666830890504e-06, + "loss": 0.34, + "step": 22527 + }, + { + "epoch": 8.32, + "learning_rate": 2.169429732725705e-06, + "loss": 0.471, + "step": 22528 + }, + { + "epoch": 8.32, + "learning_rate": 2.16939278236236e-06, + "loss": 0.3068, + "step": 22529 + }, + { + "epoch": 8.32, + "learning_rate": 2.1693558319990147e-06, + "loss": 0.324, + "step": 22530 + }, + { + "epoch": 8.32, + "learning_rate": 2.1693188816356698e-06, + "loss": 0.4213, + "step": 22531 + }, + { + "epoch": 8.32, + "learning_rate": 2.1692819312723244e-06, + "loss": 0.3122, + "step": 22532 + }, + { + "epoch": 8.32, + "learning_rate": 2.169244980908979e-06, + "loss": 0.4203, + "step": 22533 + }, + { + "epoch": 8.32, + "learning_rate": 2.1692080305456337e-06, + "loss": 0.2927, + "step": 22534 + }, + { + "epoch": 8.32, + "learning_rate": 2.1691710801822883e-06, + "loss": 0.4168, + "step": 22535 + }, + { + "epoch": 8.32, + "learning_rate": 2.1691341298189434e-06, + "loss": 0.3538, + "step": 22536 + }, + { + "epoch": 8.32, + "learning_rate": 2.169097179455598e-06, + "loss": 0.5268, + "step": 22537 + }, + { + "epoch": 8.32, + "learning_rate": 2.169060229092253e-06, + "loss": 0.5118, + "step": 22538 + }, + { + "epoch": 8.32, + "learning_rate": 2.1690232787289077e-06, + "loss": 0.3682, + "step": 22539 + }, + { + "epoch": 8.32, + "learning_rate": 2.1689863283655623e-06, + "loss": 0.3839, + "step": 22540 + }, + { + "epoch": 8.32, + "learning_rate": 2.168949378002217e-06, + "loss": 0.4192, + "step": 22541 + }, + { + "epoch": 8.32, + "learning_rate": 2.168912427638872e-06, + "loss": 0.3223, + "step": 22542 + }, + { + "epoch": 8.32, + "learning_rate": 2.1688754772755267e-06, + "loss": 0.5026, + "step": 22543 + }, + { + "epoch": 8.32, + "learning_rate": 2.1688385269121813e-06, + "loss": 0.3546, + "step": 22544 + }, + { + "epoch": 8.33, + "learning_rate": 2.168801576548836e-06, + "loss": 0.4261, + "step": 22545 + }, + { + "epoch": 8.33, + "learning_rate": 2.1687646261854906e-06, + "loss": 0.3097, + "step": 22546 + }, + { + "epoch": 8.33, + "learning_rate": 2.1687276758221456e-06, + "loss": 0.3924, + "step": 22547 + }, + { + "epoch": 8.33, + "learning_rate": 2.1686907254588003e-06, + "loss": 0.4056, + "step": 22548 + }, + { + "epoch": 8.33, + "learning_rate": 2.1686537750954553e-06, + "loss": 0.3699, + "step": 22549 + }, + { + "epoch": 8.33, + "learning_rate": 2.16861682473211e-06, + "loss": 0.4238, + "step": 22550 + }, + { + "epoch": 8.33, + "learning_rate": 2.1685798743687646e-06, + "loss": 0.4114, + "step": 22551 + }, + { + "epoch": 8.33, + "learning_rate": 2.1685429240054192e-06, + "loss": 0.3679, + "step": 22552 + }, + { + "epoch": 8.33, + "learning_rate": 2.168505973642074e-06, + "loss": 0.4196, + "step": 22553 + }, + { + "epoch": 8.33, + "learning_rate": 2.168469023278729e-06, + "loss": 0.3167, + "step": 22554 + }, + { + "epoch": 8.33, + "learning_rate": 2.1684320729153836e-06, + "loss": 0.4562, + "step": 22555 + }, + { + "epoch": 8.33, + "learning_rate": 2.1683951225520386e-06, + "loss": 0.3972, + "step": 22556 + }, + { + "epoch": 8.33, + "learning_rate": 2.1683581721886933e-06, + "loss": 0.2951, + "step": 22557 + }, + { + "epoch": 8.33, + "learning_rate": 2.168321221825348e-06, + "loss": 0.4059, + "step": 22558 + }, + { + "epoch": 8.33, + "learning_rate": 2.1682842714620025e-06, + "loss": 0.4483, + "step": 22559 + }, + { + "epoch": 8.33, + "learning_rate": 2.1682473210986576e-06, + "loss": 0.4478, + "step": 22560 + }, + { + "epoch": 8.33, + "learning_rate": 2.1682103707353122e-06, + "loss": 0.2546, + "step": 22561 + }, + { + "epoch": 8.33, + "learning_rate": 2.1681734203719673e-06, + "loss": 0.4802, + "step": 22562 + }, + { + "epoch": 8.33, + "learning_rate": 2.168136470008622e-06, + "loss": 0.3474, + "step": 22563 + }, + { + "epoch": 8.33, + "learning_rate": 2.1680995196452766e-06, + "loss": 0.3188, + "step": 22564 + }, + { + "epoch": 8.33, + "learning_rate": 2.1680625692819312e-06, + "loss": 0.3754, + "step": 22565 + }, + { + "epoch": 8.33, + "learning_rate": 2.168025618918586e-06, + "loss": 0.4217, + "step": 22566 + }, + { + "epoch": 8.33, + "learning_rate": 2.167988668555241e-06, + "loss": 0.4144, + "step": 22567 + }, + { + "epoch": 8.33, + "learning_rate": 2.1679517181918955e-06, + "loss": 0.3343, + "step": 22568 + }, + { + "epoch": 8.33, + "learning_rate": 2.1679147678285506e-06, + "loss": 0.4631, + "step": 22569 + }, + { + "epoch": 8.33, + "learning_rate": 2.1678778174652052e-06, + "loss": 0.3294, + "step": 22570 + }, + { + "epoch": 8.33, + "learning_rate": 2.16784086710186e-06, + "loss": 0.4025, + "step": 22571 + }, + { + "epoch": 8.34, + "learning_rate": 2.1678039167385145e-06, + "loss": 0.385, + "step": 22572 + }, + { + "epoch": 8.34, + "learning_rate": 2.167766966375169e-06, + "loss": 0.4647, + "step": 22573 + }, + { + "epoch": 8.34, + "learning_rate": 2.167730016011824e-06, + "loss": 0.4348, + "step": 22574 + }, + { + "epoch": 8.34, + "learning_rate": 2.167693065648479e-06, + "loss": 0.412, + "step": 22575 + }, + { + "epoch": 8.34, + "learning_rate": 2.167656115285134e-06, + "loss": 0.3014, + "step": 22576 + }, + { + "epoch": 8.34, + "learning_rate": 2.1676191649217885e-06, + "loss": 0.3415, + "step": 22577 + }, + { + "epoch": 8.34, + "learning_rate": 2.167582214558443e-06, + "loss": 0.2901, + "step": 22578 + }, + { + "epoch": 8.34, + "learning_rate": 2.167545264195098e-06, + "loss": 0.4103, + "step": 22579 + }, + { + "epoch": 8.34, + "learning_rate": 2.167508313831753e-06, + "loss": 0.4407, + "step": 22580 + }, + { + "epoch": 8.34, + "learning_rate": 2.1674713634684075e-06, + "loss": 0.4058, + "step": 22581 + }, + { + "epoch": 8.34, + "learning_rate": 2.1674344131050626e-06, + "loss": 0.3353, + "step": 22582 + }, + { + "epoch": 8.34, + "learning_rate": 2.167397462741717e-06, + "loss": 0.4037, + "step": 22583 + }, + { + "epoch": 8.34, + "learning_rate": 2.1673605123783714e-06, + "loss": 0.4356, + "step": 22584 + }, + { + "epoch": 8.34, + "learning_rate": 2.1673235620150265e-06, + "loss": 0.3657, + "step": 22585 + }, + { + "epoch": 8.34, + "learning_rate": 2.167286611651681e-06, + "loss": 0.373, + "step": 22586 + }, + { + "epoch": 8.34, + "learning_rate": 2.167249661288336e-06, + "loss": 0.3393, + "step": 22587 + }, + { + "epoch": 8.34, + "learning_rate": 2.167212710924991e-06, + "loss": 0.4404, + "step": 22588 + }, + { + "epoch": 8.34, + "learning_rate": 2.167175760561646e-06, + "loss": 0.3978, + "step": 22589 + }, + { + "epoch": 8.34, + "learning_rate": 2.1671388101983e-06, + "loss": 0.4153, + "step": 22590 + }, + { + "epoch": 8.34, + "learning_rate": 2.167101859834955e-06, + "loss": 0.3471, + "step": 22591 + }, + { + "epoch": 8.34, + "learning_rate": 2.1670649094716098e-06, + "loss": 0.4069, + "step": 22592 + }, + { + "epoch": 8.34, + "learning_rate": 2.1670279591082644e-06, + "loss": 0.4217, + "step": 22593 + }, + { + "epoch": 8.34, + "learning_rate": 2.1669910087449195e-06, + "loss": 0.3693, + "step": 22594 + }, + { + "epoch": 8.34, + "learning_rate": 2.166954058381574e-06, + "loss": 0.5437, + "step": 22595 + }, + { + "epoch": 8.34, + "learning_rate": 2.1669171080182288e-06, + "loss": 0.4563, + "step": 22596 + }, + { + "epoch": 8.34, + "learning_rate": 2.1668801576548834e-06, + "loss": 0.3935, + "step": 22597 + }, + { + "epoch": 8.34, + "learning_rate": 2.1668432072915384e-06, + "loss": 0.3576, + "step": 22598 + }, + { + "epoch": 8.35, + "learning_rate": 2.166806256928193e-06, + "loss": 0.2938, + "step": 22599 + }, + { + "epoch": 8.35, + "learning_rate": 2.166769306564848e-06, + "loss": 0.4049, + "step": 22600 + }, + { + "epoch": 8.35, + "learning_rate": 2.1667323562015028e-06, + "loss": 0.2701, + "step": 22601 + }, + { + "epoch": 8.35, + "learning_rate": 2.1666954058381574e-06, + "loss": 0.3352, + "step": 22602 + }, + { + "epoch": 8.35, + "learning_rate": 2.166658455474812e-06, + "loss": 0.3339, + "step": 22603 + }, + { + "epoch": 8.35, + "learning_rate": 2.1666215051114667e-06, + "loss": 0.2535, + "step": 22604 + }, + { + "epoch": 8.35, + "learning_rate": 2.1665845547481218e-06, + "loss": 0.4775, + "step": 22605 + }, + { + "epoch": 8.35, + "learning_rate": 2.1665476043847764e-06, + "loss": 0.3813, + "step": 22606 + }, + { + "epoch": 8.35, + "learning_rate": 2.1665106540214314e-06, + "loss": 0.3665, + "step": 22607 + }, + { + "epoch": 8.35, + "learning_rate": 2.166473703658086e-06, + "loss": 0.3706, + "step": 22608 + }, + { + "epoch": 8.35, + "learning_rate": 2.1664367532947407e-06, + "loss": 0.2828, + "step": 22609 + }, + { + "epoch": 8.35, + "learning_rate": 2.1663998029313954e-06, + "loss": 0.3939, + "step": 22610 + }, + { + "epoch": 8.35, + "learning_rate": 2.1663628525680504e-06, + "loss": 0.3782, + "step": 22611 + }, + { + "epoch": 8.35, + "learning_rate": 2.166325902204705e-06, + "loss": 0.5716, + "step": 22612 + }, + { + "epoch": 8.35, + "learning_rate": 2.1662889518413597e-06, + "loss": 0.3647, + "step": 22613 + }, + { + "epoch": 8.35, + "learning_rate": 2.1662520014780147e-06, + "loss": 0.3941, + "step": 22614 + }, + { + "epoch": 8.35, + "learning_rate": 2.1662150511146694e-06, + "loss": 0.3372, + "step": 22615 + }, + { + "epoch": 8.35, + "learning_rate": 2.166178100751324e-06, + "loss": 0.2712, + "step": 22616 + }, + { + "epoch": 8.35, + "learning_rate": 2.1661411503879787e-06, + "loss": 0.3674, + "step": 22617 + }, + { + "epoch": 8.35, + "learning_rate": 2.1661042000246337e-06, + "loss": 0.2862, + "step": 22618 + }, + { + "epoch": 8.35, + "learning_rate": 2.1660672496612884e-06, + "loss": 0.3751, + "step": 22619 + }, + { + "epoch": 8.35, + "learning_rate": 2.1660302992979434e-06, + "loss": 0.4416, + "step": 22620 + }, + { + "epoch": 8.35, + "learning_rate": 2.165993348934598e-06, + "loss": 0.3905, + "step": 22621 + }, + { + "epoch": 8.35, + "learning_rate": 2.1659563985712527e-06, + "loss": 0.4277, + "step": 22622 + }, + { + "epoch": 8.35, + "learning_rate": 2.1659194482079073e-06, + "loss": 0.4284, + "step": 22623 + }, + { + "epoch": 8.35, + "learning_rate": 2.165882497844562e-06, + "loss": 0.3677, + "step": 22624 + }, + { + "epoch": 8.35, + "learning_rate": 2.165845547481217e-06, + "loss": 0.3677, + "step": 22625 + }, + { + "epoch": 8.36, + "learning_rate": 2.1658085971178717e-06, + "loss": 0.4652, + "step": 22626 + }, + { + "epoch": 8.36, + "learning_rate": 2.1657716467545267e-06, + "loss": 0.4061, + "step": 22627 + }, + { + "epoch": 8.36, + "learning_rate": 2.1657346963911814e-06, + "loss": 0.4608, + "step": 22628 + }, + { + "epoch": 8.36, + "learning_rate": 2.165697746027836e-06, + "loss": 0.3999, + "step": 22629 + }, + { + "epoch": 8.36, + "learning_rate": 2.1656607956644906e-06, + "loss": 0.3587, + "step": 22630 + }, + { + "epoch": 8.36, + "learning_rate": 2.1656238453011457e-06, + "loss": 0.2606, + "step": 22631 + }, + { + "epoch": 8.36, + "learning_rate": 2.1655868949378003e-06, + "loss": 0.4389, + "step": 22632 + }, + { + "epoch": 8.36, + "learning_rate": 2.165549944574455e-06, + "loss": 0.3254, + "step": 22633 + }, + { + "epoch": 8.36, + "learning_rate": 2.16551299421111e-06, + "loss": 0.3652, + "step": 22634 + }, + { + "epoch": 8.36, + "learning_rate": 2.1654760438477642e-06, + "loss": 0.3613, + "step": 22635 + }, + { + "epoch": 8.36, + "learning_rate": 2.1654390934844193e-06, + "loss": 0.4157, + "step": 22636 + }, + { + "epoch": 8.36, + "learning_rate": 2.165402143121074e-06, + "loss": 0.418, + "step": 22637 + }, + { + "epoch": 8.36, + "learning_rate": 2.165365192757729e-06, + "loss": 0.4385, + "step": 22638 + }, + { + "epoch": 8.36, + "learning_rate": 2.1653282423943836e-06, + "loss": 0.4391, + "step": 22639 + }, + { + "epoch": 8.36, + "learning_rate": 2.1652912920310387e-06, + "loss": 0.3305, + "step": 22640 + }, + { + "epoch": 8.36, + "learning_rate": 2.165254341667693e-06, + "loss": 0.2487, + "step": 22641 + }, + { + "epoch": 8.36, + "learning_rate": 2.165217391304348e-06, + "loss": 0.3489, + "step": 22642 + }, + { + "epoch": 8.36, + "learning_rate": 2.1651804409410026e-06, + "loss": 0.2757, + "step": 22643 + }, + { + "epoch": 8.36, + "learning_rate": 2.1651434905776572e-06, + "loss": 0.349, + "step": 22644 + }, + { + "epoch": 8.36, + "learning_rate": 2.1651065402143123e-06, + "loss": 0.327, + "step": 22645 + }, + { + "epoch": 8.36, + "learning_rate": 2.165069589850967e-06, + "loss": 0.501, + "step": 22646 + }, + { + "epoch": 8.36, + "learning_rate": 2.1650326394876216e-06, + "loss": 0.4998, + "step": 22647 + }, + { + "epoch": 8.36, + "learning_rate": 2.164995689124276e-06, + "loss": 0.3153, + "step": 22648 + }, + { + "epoch": 8.36, + "learning_rate": 2.1649587387609313e-06, + "loss": 0.3687, + "step": 22649 + }, + { + "epoch": 8.36, + "learning_rate": 2.164921788397586e-06, + "loss": 0.4907, + "step": 22650 + }, + { + "epoch": 8.36, + "learning_rate": 2.164884838034241e-06, + "loss": 0.3087, + "step": 22651 + }, + { + "epoch": 8.36, + "learning_rate": 2.1648478876708956e-06, + "loss": 0.4423, + "step": 22652 + }, + { + "epoch": 8.37, + "learning_rate": 2.1648109373075502e-06, + "loss": 0.3146, + "step": 22653 + }, + { + "epoch": 8.37, + "learning_rate": 2.164773986944205e-06, + "loss": 0.2411, + "step": 22654 + }, + { + "epoch": 8.37, + "learning_rate": 2.1647370365808595e-06, + "loss": 0.2914, + "step": 22655 + }, + { + "epoch": 8.37, + "learning_rate": 2.1647000862175146e-06, + "loss": 0.3442, + "step": 22656 + }, + { + "epoch": 8.37, + "learning_rate": 2.164663135854169e-06, + "loss": 0.3408, + "step": 22657 + }, + { + "epoch": 8.37, + "learning_rate": 2.1646261854908243e-06, + "loss": 0.3603, + "step": 22658 + }, + { + "epoch": 8.37, + "learning_rate": 2.164589235127479e-06, + "loss": 0.3098, + "step": 22659 + }, + { + "epoch": 8.37, + "learning_rate": 2.1645522847641335e-06, + "loss": 0.4126, + "step": 22660 + }, + { + "epoch": 8.37, + "learning_rate": 2.164515334400788e-06, + "loss": 0.3822, + "step": 22661 + }, + { + "epoch": 8.37, + "learning_rate": 2.1644783840374432e-06, + "loss": 0.3759, + "step": 22662 + }, + { + "epoch": 8.37, + "learning_rate": 2.164441433674098e-06, + "loss": 0.2445, + "step": 22663 + }, + { + "epoch": 8.37, + "learning_rate": 2.1644044833107525e-06, + "loss": 0.398, + "step": 22664 + }, + { + "epoch": 8.37, + "learning_rate": 2.1643675329474076e-06, + "loss": 0.3278, + "step": 22665 + }, + { + "epoch": 8.37, + "learning_rate": 2.164330582584062e-06, + "loss": 0.3208, + "step": 22666 + }, + { + "epoch": 8.37, + "learning_rate": 2.164293632220717e-06, + "loss": 0.2734, + "step": 22667 + }, + { + "epoch": 8.37, + "learning_rate": 2.1642566818573715e-06, + "loss": 0.3865, + "step": 22668 + }, + { + "epoch": 8.37, + "learning_rate": 2.1642197314940265e-06, + "loss": 0.3086, + "step": 22669 + }, + { + "epoch": 8.37, + "learning_rate": 2.164182781130681e-06, + "loss": 0.4408, + "step": 22670 + }, + { + "epoch": 8.37, + "learning_rate": 2.1641458307673362e-06, + "loss": 0.4658, + "step": 22671 + }, + { + "epoch": 8.37, + "learning_rate": 2.164108880403991e-06, + "loss": 0.4207, + "step": 22672 + }, + { + "epoch": 8.37, + "learning_rate": 2.1640719300406455e-06, + "loss": 0.4482, + "step": 22673 + }, + { + "epoch": 8.37, + "learning_rate": 2.1640349796773e-06, + "loss": 0.4132, + "step": 22674 + }, + { + "epoch": 8.37, + "learning_rate": 2.1639980293139548e-06, + "loss": 0.4563, + "step": 22675 + }, + { + "epoch": 8.37, + "learning_rate": 2.16396107895061e-06, + "loss": 0.3672, + "step": 22676 + }, + { + "epoch": 8.37, + "learning_rate": 2.1639241285872645e-06, + "loss": 0.4548, + "step": 22677 + }, + { + "epoch": 8.37, + "learning_rate": 2.1638871782239195e-06, + "loss": 0.4709, + "step": 22678 + }, + { + "epoch": 8.37, + "learning_rate": 2.163850227860574e-06, + "loss": 0.3961, + "step": 22679 + }, + { + "epoch": 8.38, + "learning_rate": 2.163813277497229e-06, + "loss": 0.6135, + "step": 22680 + }, + { + "epoch": 8.38, + "learning_rate": 2.1637763271338834e-06, + "loss": 0.4329, + "step": 22681 + }, + { + "epoch": 8.38, + "learning_rate": 2.1637393767705385e-06, + "loss": 0.3167, + "step": 22682 + }, + { + "epoch": 8.38, + "learning_rate": 2.163702426407193e-06, + "loss": 0.4702, + "step": 22683 + }, + { + "epoch": 8.38, + "learning_rate": 2.1636654760438478e-06, + "loss": 0.3727, + "step": 22684 + }, + { + "epoch": 8.38, + "learning_rate": 2.163628525680503e-06, + "loss": 0.4469, + "step": 22685 + }, + { + "epoch": 8.38, + "learning_rate": 2.163591575317157e-06, + "loss": 0.3845, + "step": 22686 + }, + { + "epoch": 8.38, + "learning_rate": 2.163554624953812e-06, + "loss": 0.4125, + "step": 22687 + }, + { + "epoch": 8.38, + "learning_rate": 2.1635176745904667e-06, + "loss": 0.2952, + "step": 22688 + }, + { + "epoch": 8.38, + "learning_rate": 2.163480724227122e-06, + "loss": 0.5035, + "step": 22689 + }, + { + "epoch": 8.38, + "learning_rate": 2.1634437738637764e-06, + "loss": 0.2918, + "step": 22690 + }, + { + "epoch": 8.38, + "learning_rate": 2.1634068235004315e-06, + "loss": 0.4328, + "step": 22691 + }, + { + "epoch": 8.38, + "learning_rate": 2.1633698731370857e-06, + "loss": 0.4066, + "step": 22692 + }, + { + "epoch": 8.38, + "learning_rate": 2.1633329227737403e-06, + "loss": 0.459, + "step": 22693 + }, + { + "epoch": 8.38, + "learning_rate": 2.1632959724103954e-06, + "loss": 0.3675, + "step": 22694 + }, + { + "epoch": 8.38, + "learning_rate": 2.16325902204705e-06, + "loss": 0.3766, + "step": 22695 + }, + { + "epoch": 8.38, + "learning_rate": 2.163222071683705e-06, + "loss": 0.5234, + "step": 22696 + }, + { + "epoch": 8.38, + "learning_rate": 2.1631851213203597e-06, + "loss": 0.3673, + "step": 22697 + }, + { + "epoch": 8.38, + "learning_rate": 2.1631481709570144e-06, + "loss": 0.4678, + "step": 22698 + }, + { + "epoch": 8.38, + "learning_rate": 2.163111220593669e-06, + "loss": 0.4369, + "step": 22699 + }, + { + "epoch": 8.38, + "learning_rate": 2.163074270230324e-06, + "loss": 0.3872, + "step": 22700 + }, + { + "epoch": 8.38, + "learning_rate": 2.1630373198669787e-06, + "loss": 0.418, + "step": 22701 + }, + { + "epoch": 8.38, + "learning_rate": 2.1630003695036338e-06, + "loss": 0.2635, + "step": 22702 + }, + { + "epoch": 8.38, + "learning_rate": 2.1629634191402884e-06, + "loss": 0.3421, + "step": 22703 + }, + { + "epoch": 8.38, + "learning_rate": 2.162926468776943e-06, + "loss": 0.4208, + "step": 22704 + }, + { + "epoch": 8.38, + "learning_rate": 2.1628895184135977e-06, + "loss": 0.3599, + "step": 22705 + }, + { + "epoch": 8.38, + "learning_rate": 2.1628525680502523e-06, + "loss": 0.3392, + "step": 22706 + }, + { + "epoch": 8.39, + "learning_rate": 2.1628156176869074e-06, + "loss": 0.3811, + "step": 22707 + }, + { + "epoch": 8.39, + "learning_rate": 2.162778667323562e-06, + "loss": 0.3931, + "step": 22708 + }, + { + "epoch": 8.39, + "learning_rate": 2.162741716960217e-06, + "loss": 0.4495, + "step": 22709 + }, + { + "epoch": 8.39, + "learning_rate": 2.1627047665968717e-06, + "loss": 0.4601, + "step": 22710 + }, + { + "epoch": 8.39, + "learning_rate": 2.1626678162335263e-06, + "loss": 0.4336, + "step": 22711 + }, + { + "epoch": 8.39, + "learning_rate": 2.162630865870181e-06, + "loss": 0.3768, + "step": 22712 + }, + { + "epoch": 8.39, + "learning_rate": 2.1625939155068356e-06, + "loss": 0.4317, + "step": 22713 + }, + { + "epoch": 8.39, + "learning_rate": 2.1625569651434907e-06, + "loss": 0.4426, + "step": 22714 + }, + { + "epoch": 8.39, + "learning_rate": 2.1625200147801453e-06, + "loss": 0.3896, + "step": 22715 + }, + { + "epoch": 8.39, + "learning_rate": 2.1624830644168004e-06, + "loss": 0.2928, + "step": 22716 + }, + { + "epoch": 8.39, + "learning_rate": 2.162446114053455e-06, + "loss": 0.3618, + "step": 22717 + }, + { + "epoch": 8.39, + "learning_rate": 2.1624091636901096e-06, + "loss": 0.4394, + "step": 22718 + }, + { + "epoch": 8.39, + "learning_rate": 2.1623722133267643e-06, + "loss": 0.4292, + "step": 22719 + }, + { + "epoch": 8.39, + "learning_rate": 2.1623352629634193e-06, + "loss": 0.4825, + "step": 22720 + }, + { + "epoch": 8.39, + "learning_rate": 2.162298312600074e-06, + "loss": 0.4657, + "step": 22721 + }, + { + "epoch": 8.39, + "learning_rate": 2.162261362236729e-06, + "loss": 0.3143, + "step": 22722 + }, + { + "epoch": 8.39, + "learning_rate": 2.1622244118733837e-06, + "loss": 0.4977, + "step": 22723 + }, + { + "epoch": 8.39, + "learning_rate": 2.1621874615100383e-06, + "loss": 0.3681, + "step": 22724 + }, + { + "epoch": 8.39, + "learning_rate": 2.162150511146693e-06, + "loss": 0.3623, + "step": 22725 + }, + { + "epoch": 8.39, + "learning_rate": 2.1621135607833476e-06, + "loss": 0.4741, + "step": 22726 + }, + { + "epoch": 8.39, + "learning_rate": 2.1620766104200026e-06, + "loss": 0.4437, + "step": 22727 + }, + { + "epoch": 8.39, + "learning_rate": 2.1620396600566573e-06, + "loss": 0.334, + "step": 22728 + }, + { + "epoch": 8.39, + "learning_rate": 2.1620027096933123e-06, + "loss": 0.4193, + "step": 22729 + }, + { + "epoch": 8.39, + "learning_rate": 2.1619657593299666e-06, + "loss": 0.4511, + "step": 22730 + }, + { + "epoch": 8.39, + "learning_rate": 2.1619288089666216e-06, + "loss": 0.3836, + "step": 22731 + }, + { + "epoch": 8.39, + "learning_rate": 2.1618918586032762e-06, + "loss": 0.4095, + "step": 22732 + }, + { + "epoch": 8.39, + "learning_rate": 2.161854908239931e-06, + "loss": 0.3966, + "step": 22733 + }, + { + "epoch": 8.4, + "learning_rate": 2.161817957876586e-06, + "loss": 0.3759, + "step": 22734 + }, + { + "epoch": 8.4, + "learning_rate": 2.1617810075132406e-06, + "loss": 0.3361, + "step": 22735 + }, + { + "epoch": 8.4, + "learning_rate": 2.1617440571498952e-06, + "loss": 0.4208, + "step": 22736 + }, + { + "epoch": 8.4, + "learning_rate": 2.16170710678655e-06, + "loss": 0.4336, + "step": 22737 + }, + { + "epoch": 8.4, + "learning_rate": 2.161670156423205e-06, + "loss": 0.4932, + "step": 22738 + }, + { + "epoch": 8.4, + "learning_rate": 2.1616332060598596e-06, + "loss": 0.276, + "step": 22739 + }, + { + "epoch": 8.4, + "learning_rate": 2.1615962556965146e-06, + "loss": 0.4049, + "step": 22740 + }, + { + "epoch": 8.4, + "learning_rate": 2.1615593053331692e-06, + "loss": 0.3992, + "step": 22741 + }, + { + "epoch": 8.4, + "learning_rate": 2.161522354969824e-06, + "loss": 0.3707, + "step": 22742 + }, + { + "epoch": 8.4, + "learning_rate": 2.1614854046064785e-06, + "loss": 0.3113, + "step": 22743 + }, + { + "epoch": 8.4, + "learning_rate": 2.161448454243133e-06, + "loss": 0.4176, + "step": 22744 + }, + { + "epoch": 8.4, + "learning_rate": 2.1614115038797882e-06, + "loss": 0.3594, + "step": 22745 + }, + { + "epoch": 8.4, + "learning_rate": 2.161374553516443e-06, + "loss": 0.3208, + "step": 22746 + }, + { + "epoch": 8.4, + "learning_rate": 2.161337603153098e-06, + "loss": 0.3176, + "step": 22747 + }, + { + "epoch": 8.4, + "learning_rate": 2.1613006527897525e-06, + "loss": 0.4235, + "step": 22748 + }, + { + "epoch": 8.4, + "learning_rate": 2.161263702426407e-06, + "loss": 0.4192, + "step": 22749 + }, + { + "epoch": 8.4, + "learning_rate": 2.161226752063062e-06, + "loss": 0.4139, + "step": 22750 + }, + { + "epoch": 8.4, + "learning_rate": 2.161189801699717e-06, + "loss": 0.3544, + "step": 22751 + }, + { + "epoch": 8.4, + "learning_rate": 2.1611528513363715e-06, + "loss": 0.4301, + "step": 22752 + }, + { + "epoch": 8.4, + "learning_rate": 2.161115900973026e-06, + "loss": 0.3996, + "step": 22753 + }, + { + "epoch": 8.4, + "learning_rate": 2.1610789506096812e-06, + "loss": 0.5627, + "step": 22754 + }, + { + "epoch": 8.4, + "learning_rate": 2.161042000246336e-06, + "loss": 0.304, + "step": 22755 + }, + { + "epoch": 8.4, + "learning_rate": 2.1610050498829905e-06, + "loss": 0.4847, + "step": 22756 + }, + { + "epoch": 8.4, + "learning_rate": 2.160968099519645e-06, + "loss": 0.4921, + "step": 22757 + }, + { + "epoch": 8.4, + "learning_rate": 2.1609311491563e-06, + "loss": 0.4732, + "step": 22758 + }, + { + "epoch": 8.4, + "learning_rate": 2.160894198792955e-06, + "loss": 0.3717, + "step": 22759 + }, + { + "epoch": 8.4, + "learning_rate": 2.16085724842961e-06, + "loss": 0.3852, + "step": 22760 + }, + { + "epoch": 8.41, + "learning_rate": 2.1608202980662645e-06, + "loss": 0.3937, + "step": 22761 + }, + { + "epoch": 8.41, + "learning_rate": 2.160783347702919e-06, + "loss": 0.361, + "step": 22762 + }, + { + "epoch": 8.41, + "learning_rate": 2.160746397339574e-06, + "loss": 0.516, + "step": 22763 + }, + { + "epoch": 8.41, + "learning_rate": 2.1607094469762284e-06, + "loss": 0.366, + "step": 22764 + }, + { + "epoch": 8.41, + "learning_rate": 2.1606724966128835e-06, + "loss": 0.4871, + "step": 22765 + }, + { + "epoch": 8.41, + "learning_rate": 2.160635546249538e-06, + "loss": 0.5191, + "step": 22766 + }, + { + "epoch": 8.41, + "learning_rate": 2.160598595886193e-06, + "loss": 0.388, + "step": 22767 + }, + { + "epoch": 8.41, + "learning_rate": 2.160561645522848e-06, + "loss": 0.3859, + "step": 22768 + }, + { + "epoch": 8.41, + "learning_rate": 2.1605246951595025e-06, + "loss": 0.408, + "step": 22769 + }, + { + "epoch": 8.41, + "learning_rate": 2.160487744796157e-06, + "loss": 0.3617, + "step": 22770 + }, + { + "epoch": 8.41, + "learning_rate": 2.160450794432812e-06, + "loss": 0.3277, + "step": 22771 + }, + { + "epoch": 8.41, + "learning_rate": 2.160413844069467e-06, + "loss": 0.3035, + "step": 22772 + }, + { + "epoch": 8.41, + "learning_rate": 2.1603768937061214e-06, + "loss": 0.3979, + "step": 22773 + }, + { + "epoch": 8.41, + "learning_rate": 2.1603399433427765e-06, + "loss": 0.345, + "step": 22774 + }, + { + "epoch": 8.41, + "learning_rate": 2.1603029929794307e-06, + "loss": 0.3774, + "step": 22775 + }, + { + "epoch": 8.41, + "learning_rate": 2.1602660426160858e-06, + "loss": 0.2284, + "step": 22776 + }, + { + "epoch": 8.41, + "learning_rate": 2.1602290922527404e-06, + "loss": 0.3518, + "step": 22777 + }, + { + "epoch": 8.41, + "learning_rate": 2.1601921418893955e-06, + "loss": 0.4312, + "step": 22778 + }, + { + "epoch": 8.41, + "learning_rate": 2.16015519152605e-06, + "loss": 0.378, + "step": 22779 + }, + { + "epoch": 8.41, + "learning_rate": 2.160118241162705e-06, + "loss": 0.5268, + "step": 22780 + }, + { + "epoch": 8.41, + "learning_rate": 2.1600812907993594e-06, + "loss": 0.3153, + "step": 22781 + }, + { + "epoch": 8.41, + "learning_rate": 2.1600443404360144e-06, + "loss": 0.3812, + "step": 22782 + }, + { + "epoch": 8.41, + "learning_rate": 2.160007390072669e-06, + "loss": 0.3669, + "step": 22783 + }, + { + "epoch": 8.41, + "learning_rate": 2.1599704397093237e-06, + "loss": 0.4239, + "step": 22784 + }, + { + "epoch": 8.41, + "learning_rate": 2.1599334893459788e-06, + "loss": 0.361, + "step": 22785 + }, + { + "epoch": 8.41, + "learning_rate": 2.1598965389826334e-06, + "loss": 0.447, + "step": 22786 + }, + { + "epoch": 8.41, + "learning_rate": 2.159859588619288e-06, + "loss": 0.3435, + "step": 22787 + }, + { + "epoch": 8.42, + "learning_rate": 2.1598226382559427e-06, + "loss": 0.5525, + "step": 22788 + }, + { + "epoch": 8.42, + "learning_rate": 2.1597856878925977e-06, + "loss": 0.4498, + "step": 22789 + }, + { + "epoch": 8.42, + "learning_rate": 2.1597487375292524e-06, + "loss": 0.3472, + "step": 22790 + }, + { + "epoch": 8.42, + "learning_rate": 2.1597117871659074e-06, + "loss": 0.349, + "step": 22791 + }, + { + "epoch": 8.42, + "learning_rate": 2.159674836802562e-06, + "loss": 0.4199, + "step": 22792 + }, + { + "epoch": 8.42, + "learning_rate": 2.1596378864392167e-06, + "loss": 0.3775, + "step": 22793 + }, + { + "epoch": 8.42, + "learning_rate": 2.1596009360758713e-06, + "loss": 0.5006, + "step": 22794 + }, + { + "epoch": 8.42, + "learning_rate": 2.159563985712526e-06, + "loss": 0.3774, + "step": 22795 + }, + { + "epoch": 8.42, + "learning_rate": 2.159527035349181e-06, + "loss": 0.5644, + "step": 22796 + }, + { + "epoch": 8.42, + "learning_rate": 2.1594900849858357e-06, + "loss": 0.3223, + "step": 22797 + }, + { + "epoch": 8.42, + "learning_rate": 2.1594531346224907e-06, + "loss": 0.3598, + "step": 22798 + }, + { + "epoch": 8.42, + "learning_rate": 2.1594161842591454e-06, + "loss": 0.5028, + "step": 22799 + }, + { + "epoch": 8.42, + "learning_rate": 2.1593792338958e-06, + "loss": 0.3883, + "step": 22800 + }, + { + "epoch": 8.42, + "learning_rate": 2.1593422835324546e-06, + "loss": 0.4184, + "step": 22801 + }, + { + "epoch": 8.42, + "learning_rate": 2.1593053331691093e-06, + "loss": 0.3931, + "step": 22802 + }, + { + "epoch": 8.42, + "learning_rate": 2.1592683828057643e-06, + "loss": 0.3556, + "step": 22803 + }, + { + "epoch": 8.42, + "learning_rate": 2.159231432442419e-06, + "loss": 0.303, + "step": 22804 + }, + { + "epoch": 8.42, + "learning_rate": 2.159194482079074e-06, + "loss": 0.4723, + "step": 22805 + }, + { + "epoch": 8.42, + "learning_rate": 2.1591575317157287e-06, + "loss": 0.4474, + "step": 22806 + }, + { + "epoch": 8.42, + "learning_rate": 2.1591205813523833e-06, + "loss": 0.395, + "step": 22807 + }, + { + "epoch": 8.42, + "learning_rate": 2.159083630989038e-06, + "loss": 0.4843, + "step": 22808 + }, + { + "epoch": 8.42, + "learning_rate": 2.159046680625693e-06, + "loss": 0.4249, + "step": 22809 + }, + { + "epoch": 8.42, + "learning_rate": 2.1590097302623476e-06, + "loss": 0.4414, + "step": 22810 + }, + { + "epoch": 8.42, + "learning_rate": 2.1589727798990027e-06, + "loss": 0.2429, + "step": 22811 + }, + { + "epoch": 8.42, + "learning_rate": 2.1589358295356573e-06, + "loss": 0.4202, + "step": 22812 + }, + { + "epoch": 8.42, + "learning_rate": 2.158898879172312e-06, + "loss": 0.5171, + "step": 22813 + }, + { + "epoch": 8.42, + "learning_rate": 2.1588619288089666e-06, + "loss": 0.2864, + "step": 22814 + }, + { + "epoch": 8.43, + "learning_rate": 2.1588249784456212e-06, + "loss": 0.4407, + "step": 22815 + }, + { + "epoch": 8.43, + "learning_rate": 2.1587880280822763e-06, + "loss": 0.3844, + "step": 22816 + }, + { + "epoch": 8.43, + "learning_rate": 2.158751077718931e-06, + "loss": 0.4401, + "step": 22817 + }, + { + "epoch": 8.43, + "learning_rate": 2.158714127355586e-06, + "loss": 0.3956, + "step": 22818 + }, + { + "epoch": 8.43, + "learning_rate": 2.1586771769922406e-06, + "loss": 0.5037, + "step": 22819 + }, + { + "epoch": 8.43, + "learning_rate": 2.1586402266288953e-06, + "loss": 0.3815, + "step": 22820 + }, + { + "epoch": 8.43, + "learning_rate": 2.15860327626555e-06, + "loss": 0.4678, + "step": 22821 + }, + { + "epoch": 8.43, + "learning_rate": 2.1585663259022045e-06, + "loss": 0.3489, + "step": 22822 + }, + { + "epoch": 8.43, + "learning_rate": 2.1585293755388596e-06, + "loss": 0.3566, + "step": 22823 + }, + { + "epoch": 8.43, + "learning_rate": 2.1584924251755142e-06, + "loss": 0.2915, + "step": 22824 + }, + { + "epoch": 8.43, + "learning_rate": 2.1584554748121693e-06, + "loss": 0.3006, + "step": 22825 + }, + { + "epoch": 8.43, + "learning_rate": 2.1584185244488235e-06, + "loss": 0.2958, + "step": 22826 + }, + { + "epoch": 8.43, + "learning_rate": 2.1583815740854786e-06, + "loss": 0.444, + "step": 22827 + }, + { + "epoch": 8.43, + "learning_rate": 2.158344623722133e-06, + "loss": 0.3853, + "step": 22828 + }, + { + "epoch": 8.43, + "learning_rate": 2.1583076733587883e-06, + "loss": 0.415, + "step": 22829 + }, + { + "epoch": 8.43, + "learning_rate": 2.158270722995443e-06, + "loss": 0.4067, + "step": 22830 + }, + { + "epoch": 8.43, + "learning_rate": 2.158233772632098e-06, + "loss": 0.2752, + "step": 22831 + }, + { + "epoch": 8.43, + "learning_rate": 2.158196822268752e-06, + "loss": 0.3416, + "step": 22832 + }, + { + "epoch": 8.43, + "learning_rate": 2.158159871905407e-06, + "loss": 0.2654, + "step": 22833 + }, + { + "epoch": 8.43, + "learning_rate": 2.158122921542062e-06, + "loss": 0.2692, + "step": 22834 + }, + { + "epoch": 8.43, + "learning_rate": 2.1580859711787165e-06, + "loss": 0.464, + "step": 22835 + }, + { + "epoch": 8.43, + "learning_rate": 2.1580490208153716e-06, + "loss": 0.4595, + "step": 22836 + }, + { + "epoch": 8.43, + "learning_rate": 2.158012070452026e-06, + "loss": 0.3916, + "step": 22837 + }, + { + "epoch": 8.43, + "learning_rate": 2.157975120088681e-06, + "loss": 0.517, + "step": 22838 + }, + { + "epoch": 8.43, + "learning_rate": 2.1579381697253355e-06, + "loss": 0.3875, + "step": 22839 + }, + { + "epoch": 8.43, + "learning_rate": 2.1579012193619905e-06, + "loss": 0.354, + "step": 22840 + }, + { + "epoch": 8.43, + "learning_rate": 2.157864268998645e-06, + "loss": 0.3378, + "step": 22841 + }, + { + "epoch": 8.44, + "learning_rate": 2.1578273186353e-06, + "loss": 0.3241, + "step": 22842 + }, + { + "epoch": 8.44, + "learning_rate": 2.157790368271955e-06, + "loss": 0.5414, + "step": 22843 + }, + { + "epoch": 8.44, + "learning_rate": 2.1577534179086095e-06, + "loss": 0.5093, + "step": 22844 + }, + { + "epoch": 8.44, + "learning_rate": 2.157716467545264e-06, + "loss": 0.4461, + "step": 22845 + }, + { + "epoch": 8.44, + "learning_rate": 2.1576795171819188e-06, + "loss": 0.3644, + "step": 22846 + }, + { + "epoch": 8.44, + "learning_rate": 2.157642566818574e-06, + "loss": 0.2853, + "step": 22847 + }, + { + "epoch": 8.44, + "learning_rate": 2.1576056164552285e-06, + "loss": 0.4742, + "step": 22848 + }, + { + "epoch": 8.44, + "learning_rate": 2.1575686660918835e-06, + "loss": 0.4281, + "step": 22849 + }, + { + "epoch": 8.44, + "learning_rate": 2.157531715728538e-06, + "loss": 0.3639, + "step": 22850 + }, + { + "epoch": 8.44, + "learning_rate": 2.157494765365193e-06, + "loss": 0.3006, + "step": 22851 + }, + { + "epoch": 8.44, + "learning_rate": 2.1574578150018474e-06, + "loss": 0.365, + "step": 22852 + }, + { + "epoch": 8.44, + "learning_rate": 2.157420864638502e-06, + "loss": 0.3511, + "step": 22853 + }, + { + "epoch": 8.44, + "learning_rate": 2.157383914275157e-06, + "loss": 0.4934, + "step": 22854 + }, + { + "epoch": 8.44, + "learning_rate": 2.1573469639118118e-06, + "loss": 0.4779, + "step": 22855 + }, + { + "epoch": 8.44, + "learning_rate": 2.157310013548467e-06, + "loss": 0.387, + "step": 22856 + }, + { + "epoch": 8.44, + "learning_rate": 2.1572730631851215e-06, + "loss": 0.3336, + "step": 22857 + }, + { + "epoch": 8.44, + "learning_rate": 2.157236112821776e-06, + "loss": 0.3936, + "step": 22858 + }, + { + "epoch": 8.44, + "learning_rate": 2.1571991624584307e-06, + "loss": 0.3718, + "step": 22859 + }, + { + "epoch": 8.44, + "learning_rate": 2.157162212095086e-06, + "loss": 0.3753, + "step": 22860 + }, + { + "epoch": 8.44, + "learning_rate": 2.1571252617317404e-06, + "loss": 0.3618, + "step": 22861 + }, + { + "epoch": 8.44, + "learning_rate": 2.157088311368395e-06, + "loss": 0.4857, + "step": 22862 + }, + { + "epoch": 8.44, + "learning_rate": 2.15705136100505e-06, + "loss": 0.3792, + "step": 22863 + }, + { + "epoch": 8.44, + "learning_rate": 2.1570144106417048e-06, + "loss": 0.3809, + "step": 22864 + }, + { + "epoch": 8.44, + "learning_rate": 2.1569774602783594e-06, + "loss": 0.3312, + "step": 22865 + }, + { + "epoch": 8.44, + "learning_rate": 2.156940509915014e-06, + "loss": 0.4377, + "step": 22866 + }, + { + "epoch": 8.44, + "learning_rate": 2.156903559551669e-06, + "loss": 0.3755, + "step": 22867 + }, + { + "epoch": 8.44, + "learning_rate": 2.1568666091883237e-06, + "loss": 0.2896, + "step": 22868 + }, + { + "epoch": 8.44, + "learning_rate": 2.156829658824979e-06, + "loss": 0.3478, + "step": 22869 + }, + { + "epoch": 8.45, + "learning_rate": 2.1567927084616334e-06, + "loss": 0.3938, + "step": 22870 + }, + { + "epoch": 8.45, + "learning_rate": 2.156755758098288e-06, + "loss": 0.4693, + "step": 22871 + }, + { + "epoch": 8.45, + "learning_rate": 2.1567188077349427e-06, + "loss": 0.3277, + "step": 22872 + }, + { + "epoch": 8.45, + "learning_rate": 2.1566818573715974e-06, + "loss": 0.3104, + "step": 22873 + }, + { + "epoch": 8.45, + "learning_rate": 2.1566449070082524e-06, + "loss": 0.4839, + "step": 22874 + }, + { + "epoch": 8.45, + "learning_rate": 2.156607956644907e-06, + "loss": 0.4609, + "step": 22875 + }, + { + "epoch": 8.45, + "learning_rate": 2.156571006281562e-06, + "loss": 0.331, + "step": 22876 + }, + { + "epoch": 8.45, + "learning_rate": 2.1565340559182163e-06, + "loss": 0.3337, + "step": 22877 + }, + { + "epoch": 8.45, + "learning_rate": 2.1564971055548714e-06, + "loss": 0.4973, + "step": 22878 + }, + { + "epoch": 8.45, + "learning_rate": 2.156460155191526e-06, + "loss": 0.3446, + "step": 22879 + }, + { + "epoch": 8.45, + "learning_rate": 2.156423204828181e-06, + "loss": 0.3543, + "step": 22880 + }, + { + "epoch": 8.45, + "learning_rate": 2.1563862544648357e-06, + "loss": 0.4098, + "step": 22881 + }, + { + "epoch": 8.45, + "learning_rate": 2.1563493041014903e-06, + "loss": 0.4227, + "step": 22882 + }, + { + "epoch": 8.45, + "learning_rate": 2.156312353738145e-06, + "loss": 0.46, + "step": 22883 + }, + { + "epoch": 8.45, + "learning_rate": 2.1562754033747996e-06, + "loss": 0.4607, + "step": 22884 + }, + { + "epoch": 8.45, + "learning_rate": 2.1562384530114547e-06, + "loss": 0.4199, + "step": 22885 + }, + { + "epoch": 8.45, + "learning_rate": 2.1562015026481093e-06, + "loss": 0.4087, + "step": 22886 + }, + { + "epoch": 8.45, + "learning_rate": 2.1561645522847644e-06, + "loss": 0.4429, + "step": 22887 + }, + { + "epoch": 8.45, + "learning_rate": 2.156127601921419e-06, + "loss": 0.3427, + "step": 22888 + }, + { + "epoch": 8.45, + "learning_rate": 2.1560906515580737e-06, + "loss": 0.297, + "step": 22889 + }, + { + "epoch": 8.45, + "learning_rate": 2.1560537011947283e-06, + "loss": 0.513, + "step": 22890 + }, + { + "epoch": 8.45, + "learning_rate": 2.1560167508313833e-06, + "loss": 0.3462, + "step": 22891 + }, + { + "epoch": 8.45, + "learning_rate": 2.155979800468038e-06, + "loss": 0.3456, + "step": 22892 + }, + { + "epoch": 8.45, + "learning_rate": 2.1559428501046926e-06, + "loss": 0.4452, + "step": 22893 + }, + { + "epoch": 8.45, + "learning_rate": 2.1559058997413477e-06, + "loss": 0.3881, + "step": 22894 + }, + { + "epoch": 8.45, + "learning_rate": 2.1558689493780023e-06, + "loss": 0.3249, + "step": 22895 + }, + { + "epoch": 8.45, + "learning_rate": 2.155831999014657e-06, + "loss": 0.2813, + "step": 22896 + }, + { + "epoch": 8.46, + "learning_rate": 2.1557950486513116e-06, + "loss": 0.3484, + "step": 22897 + }, + { + "epoch": 8.46, + "learning_rate": 2.1557580982879666e-06, + "loss": 0.3994, + "step": 22898 + }, + { + "epoch": 8.46, + "learning_rate": 2.1557211479246213e-06, + "loss": 0.478, + "step": 22899 + }, + { + "epoch": 8.46, + "learning_rate": 2.1556841975612763e-06, + "loss": 0.3398, + "step": 22900 + }, + { + "epoch": 8.46, + "learning_rate": 2.155647247197931e-06, + "loss": 0.5151, + "step": 22901 + }, + { + "epoch": 8.46, + "learning_rate": 2.1556102968345856e-06, + "loss": 0.5159, + "step": 22902 + }, + { + "epoch": 8.46, + "learning_rate": 2.1555733464712403e-06, + "loss": 0.4128, + "step": 22903 + }, + { + "epoch": 8.46, + "learning_rate": 2.155536396107895e-06, + "loss": 0.5012, + "step": 22904 + }, + { + "epoch": 8.46, + "learning_rate": 2.15549944574455e-06, + "loss": 0.3326, + "step": 22905 + }, + { + "epoch": 8.46, + "learning_rate": 2.1554624953812046e-06, + "loss": 0.4848, + "step": 22906 + }, + { + "epoch": 8.46, + "learning_rate": 2.1554255450178596e-06, + "loss": 0.4277, + "step": 22907 + }, + { + "epoch": 8.46, + "learning_rate": 2.1553885946545143e-06, + "loss": 0.3921, + "step": 22908 + }, + { + "epoch": 8.46, + "learning_rate": 2.155351644291169e-06, + "loss": 0.4153, + "step": 22909 + }, + { + "epoch": 8.46, + "learning_rate": 2.1553146939278236e-06, + "loss": 0.3626, + "step": 22910 + }, + { + "epoch": 8.46, + "learning_rate": 2.1552777435644786e-06, + "loss": 0.4865, + "step": 22911 + }, + { + "epoch": 8.46, + "learning_rate": 2.1552407932011333e-06, + "loss": 0.3678, + "step": 22912 + }, + { + "epoch": 8.46, + "learning_rate": 2.155203842837788e-06, + "loss": 0.3853, + "step": 22913 + }, + { + "epoch": 8.46, + "learning_rate": 2.155166892474443e-06, + "loss": 0.3624, + "step": 22914 + }, + { + "epoch": 8.46, + "learning_rate": 2.155129942111097e-06, + "loss": 0.3439, + "step": 22915 + }, + { + "epoch": 8.46, + "learning_rate": 2.1550929917477522e-06, + "loss": 0.4052, + "step": 22916 + }, + { + "epoch": 8.46, + "learning_rate": 2.155056041384407e-06, + "loss": 0.3957, + "step": 22917 + }, + { + "epoch": 8.46, + "learning_rate": 2.155019091021062e-06, + "loss": 0.5736, + "step": 22918 + }, + { + "epoch": 8.46, + "learning_rate": 2.1549821406577166e-06, + "loss": 0.3353, + "step": 22919 + }, + { + "epoch": 8.46, + "learning_rate": 2.1549451902943716e-06, + "loss": 0.4985, + "step": 22920 + }, + { + "epoch": 8.46, + "learning_rate": 2.154908239931026e-06, + "loss": 0.2656, + "step": 22921 + }, + { + "epoch": 8.46, + "learning_rate": 2.1548712895676805e-06, + "loss": 0.3851, + "step": 22922 + }, + { + "epoch": 8.46, + "learning_rate": 2.1548343392043355e-06, + "loss": 0.4069, + "step": 22923 + }, + { + "epoch": 8.47, + "learning_rate": 2.15479738884099e-06, + "loss": 0.3242, + "step": 22924 + }, + { + "epoch": 8.47, + "learning_rate": 2.1547604384776452e-06, + "loss": 0.4233, + "step": 22925 + }, + { + "epoch": 8.47, + "learning_rate": 2.1547234881143e-06, + "loss": 0.5741, + "step": 22926 + }, + { + "epoch": 8.47, + "learning_rate": 2.1546865377509545e-06, + "loss": 0.3853, + "step": 22927 + }, + { + "epoch": 8.47, + "learning_rate": 2.154649587387609e-06, + "loss": 0.3069, + "step": 22928 + }, + { + "epoch": 8.47, + "learning_rate": 2.154612637024264e-06, + "loss": 0.3849, + "step": 22929 + }, + { + "epoch": 8.47, + "learning_rate": 2.154575686660919e-06, + "loss": 0.3216, + "step": 22930 + }, + { + "epoch": 8.47, + "learning_rate": 2.154538736297574e-06, + "loss": 0.4927, + "step": 22931 + }, + { + "epoch": 8.47, + "learning_rate": 2.1545017859342285e-06, + "loss": 0.419, + "step": 22932 + }, + { + "epoch": 8.47, + "learning_rate": 2.154464835570883e-06, + "loss": 0.4195, + "step": 22933 + }, + { + "epoch": 8.47, + "learning_rate": 2.154427885207538e-06, + "loss": 0.3445, + "step": 22934 + }, + { + "epoch": 8.47, + "learning_rate": 2.1543909348441924e-06, + "loss": 0.4309, + "step": 22935 + }, + { + "epoch": 8.47, + "learning_rate": 2.1543539844808475e-06, + "loss": 0.3813, + "step": 22936 + }, + { + "epoch": 8.47, + "learning_rate": 2.154317034117502e-06, + "loss": 0.5631, + "step": 22937 + }, + { + "epoch": 8.47, + "learning_rate": 2.154280083754157e-06, + "loss": 0.3579, + "step": 22938 + }, + { + "epoch": 8.47, + "learning_rate": 2.154243133390812e-06, + "loss": 0.3208, + "step": 22939 + }, + { + "epoch": 8.47, + "learning_rate": 2.1542061830274665e-06, + "loss": 0.3439, + "step": 22940 + }, + { + "epoch": 8.47, + "learning_rate": 2.154169232664121e-06, + "loss": 0.3754, + "step": 22941 + }, + { + "epoch": 8.47, + "learning_rate": 2.1541322823007757e-06, + "loss": 0.4042, + "step": 22942 + }, + { + "epoch": 8.47, + "learning_rate": 2.154095331937431e-06, + "loss": 0.3952, + "step": 22943 + }, + { + "epoch": 8.47, + "learning_rate": 2.1540583815740854e-06, + "loss": 0.4127, + "step": 22944 + }, + { + "epoch": 8.47, + "learning_rate": 2.1540214312107405e-06, + "loss": 0.3626, + "step": 22945 + }, + { + "epoch": 8.47, + "learning_rate": 2.153984480847395e-06, + "loss": 0.2847, + "step": 22946 + }, + { + "epoch": 8.47, + "learning_rate": 2.1539475304840498e-06, + "loss": 0.3534, + "step": 22947 + }, + { + "epoch": 8.47, + "learning_rate": 2.1539105801207044e-06, + "loss": 0.3669, + "step": 22948 + }, + { + "epoch": 8.47, + "learning_rate": 2.1538736297573595e-06, + "loss": 0.3688, + "step": 22949 + }, + { + "epoch": 8.47, + "learning_rate": 2.153836679394014e-06, + "loss": 0.3592, + "step": 22950 + }, + { + "epoch": 8.48, + "learning_rate": 2.153799729030669e-06, + "loss": 0.4291, + "step": 22951 + }, + { + "epoch": 8.48, + "learning_rate": 2.153762778667324e-06, + "loss": 0.4637, + "step": 22952 + }, + { + "epoch": 8.48, + "learning_rate": 2.1537258283039784e-06, + "loss": 0.5153, + "step": 22953 + }, + { + "epoch": 8.48, + "learning_rate": 2.153688877940633e-06, + "loss": 0.4712, + "step": 22954 + }, + { + "epoch": 8.48, + "learning_rate": 2.1536519275772877e-06, + "loss": 0.3531, + "step": 22955 + }, + { + "epoch": 8.48, + "learning_rate": 2.1536149772139428e-06, + "loss": 0.4886, + "step": 22956 + }, + { + "epoch": 8.48, + "learning_rate": 2.1535780268505974e-06, + "loss": 0.3479, + "step": 22957 + }, + { + "epoch": 8.48, + "learning_rate": 2.1535410764872525e-06, + "loss": 0.4049, + "step": 22958 + }, + { + "epoch": 8.48, + "learning_rate": 2.153504126123907e-06, + "loss": 0.2618, + "step": 22959 + }, + { + "epoch": 8.48, + "learning_rate": 2.1534671757605617e-06, + "loss": 0.3976, + "step": 22960 + }, + { + "epoch": 8.48, + "learning_rate": 2.1534302253972164e-06, + "loss": 0.5085, + "step": 22961 + }, + { + "epoch": 8.48, + "learning_rate": 2.153393275033871e-06, + "loss": 0.3062, + "step": 22962 + }, + { + "epoch": 8.48, + "learning_rate": 2.153356324670526e-06, + "loss": 0.2689, + "step": 22963 + }, + { + "epoch": 8.48, + "learning_rate": 2.1533193743071807e-06, + "loss": 0.4265, + "step": 22964 + }, + { + "epoch": 8.48, + "learning_rate": 2.1532824239438358e-06, + "loss": 0.405, + "step": 22965 + }, + { + "epoch": 8.48, + "learning_rate": 2.15324547358049e-06, + "loss": 0.4033, + "step": 22966 + }, + { + "epoch": 8.48, + "learning_rate": 2.153208523217145e-06, + "loss": 0.3528, + "step": 22967 + }, + { + "epoch": 8.48, + "learning_rate": 2.1531715728537997e-06, + "loss": 0.458, + "step": 22968 + }, + { + "epoch": 8.48, + "learning_rate": 2.1531346224904547e-06, + "loss": 0.3701, + "step": 22969 + }, + { + "epoch": 8.48, + "learning_rate": 2.1530976721271094e-06, + "loss": 0.4002, + "step": 22970 + }, + { + "epoch": 8.48, + "learning_rate": 2.1530607217637644e-06, + "loss": 0.3131, + "step": 22971 + }, + { + "epoch": 8.48, + "learning_rate": 2.1530237714004186e-06, + "loss": 0.3903, + "step": 22972 + }, + { + "epoch": 8.48, + "learning_rate": 2.1529868210370733e-06, + "loss": 0.3799, + "step": 22973 + }, + { + "epoch": 8.48, + "learning_rate": 2.1529498706737283e-06, + "loss": 0.2667, + "step": 22974 + }, + { + "epoch": 8.48, + "learning_rate": 2.152912920310383e-06, + "loss": 0.3397, + "step": 22975 + }, + { + "epoch": 8.48, + "learning_rate": 2.152875969947038e-06, + "loss": 0.3646, + "step": 22976 + }, + { + "epoch": 8.48, + "learning_rate": 2.1528390195836927e-06, + "loss": 0.2896, + "step": 22977 + }, + { + "epoch": 8.49, + "learning_rate": 2.1528020692203473e-06, + "loss": 0.3044, + "step": 22978 + }, + { + "epoch": 8.49, + "learning_rate": 2.152765118857002e-06, + "loss": 0.492, + "step": 22979 + }, + { + "epoch": 8.49, + "learning_rate": 2.152728168493657e-06, + "loss": 0.2607, + "step": 22980 + }, + { + "epoch": 8.49, + "learning_rate": 2.1526912181303116e-06, + "loss": 0.4032, + "step": 22981 + }, + { + "epoch": 8.49, + "learning_rate": 2.1526542677669663e-06, + "loss": 0.4775, + "step": 22982 + }, + { + "epoch": 8.49, + "learning_rate": 2.1526173174036213e-06, + "loss": 0.3823, + "step": 22983 + }, + { + "epoch": 8.49, + "learning_rate": 2.152580367040276e-06, + "loss": 0.4051, + "step": 22984 + }, + { + "epoch": 8.49, + "learning_rate": 2.1525434166769306e-06, + "loss": 0.4082, + "step": 22985 + }, + { + "epoch": 8.49, + "learning_rate": 2.1525064663135852e-06, + "loss": 0.4654, + "step": 22986 + }, + { + "epoch": 8.49, + "learning_rate": 2.1524695159502403e-06, + "loss": 0.4395, + "step": 22987 + }, + { + "epoch": 8.49, + "learning_rate": 2.152432565586895e-06, + "loss": 0.3929, + "step": 22988 + }, + { + "epoch": 8.49, + "learning_rate": 2.15239561522355e-06, + "loss": 0.4731, + "step": 22989 + }, + { + "epoch": 8.49, + "learning_rate": 2.1523586648602046e-06, + "loss": 0.3719, + "step": 22990 + }, + { + "epoch": 8.49, + "learning_rate": 2.1523217144968593e-06, + "loss": 0.5083, + "step": 22991 + }, + { + "epoch": 8.49, + "learning_rate": 2.152284764133514e-06, + "loss": 0.3579, + "step": 22992 + }, + { + "epoch": 8.49, + "learning_rate": 2.1522478137701685e-06, + "loss": 0.6064, + "step": 22993 + }, + { + "epoch": 8.49, + "learning_rate": 2.1522108634068236e-06, + "loss": 0.5384, + "step": 22994 + }, + { + "epoch": 8.49, + "learning_rate": 2.1521739130434782e-06, + "loss": 0.3364, + "step": 22995 + }, + { + "epoch": 8.49, + "learning_rate": 2.1521369626801333e-06, + "loss": 0.3537, + "step": 22996 + }, + { + "epoch": 8.49, + "learning_rate": 2.152100012316788e-06, + "loss": 0.3486, + "step": 22997 + }, + { + "epoch": 8.49, + "learning_rate": 2.1520630619534426e-06, + "loss": 0.3536, + "step": 22998 + }, + { + "epoch": 8.49, + "learning_rate": 2.152026111590097e-06, + "loss": 0.4409, + "step": 22999 + }, + { + "epoch": 8.49, + "learning_rate": 2.1519891612267523e-06, + "loss": 0.4889, + "step": 23000 + }, + { + "epoch": 8.49, + "learning_rate": 2.151952210863407e-06, + "loss": 0.3308, + "step": 23001 + }, + { + "epoch": 8.49, + "learning_rate": 2.1519152605000615e-06, + "loss": 0.4443, + "step": 23002 + }, + { + "epoch": 8.49, + "learning_rate": 2.1518783101367166e-06, + "loss": 0.3794, + "step": 23003 + }, + { + "epoch": 8.49, + "learning_rate": 2.1518413597733712e-06, + "loss": 0.4438, + "step": 23004 + }, + { + "epoch": 8.5, + "learning_rate": 2.151804409410026e-06, + "loss": 0.2626, + "step": 23005 + }, + { + "epoch": 8.5, + "learning_rate": 2.1517674590466805e-06, + "loss": 0.3403, + "step": 23006 + }, + { + "epoch": 8.5, + "learning_rate": 2.1517305086833356e-06, + "loss": 0.3718, + "step": 23007 + }, + { + "epoch": 8.5, + "learning_rate": 2.15169355831999e-06, + "loss": 0.2794, + "step": 23008 + }, + { + "epoch": 8.5, + "learning_rate": 2.1516566079566453e-06, + "loss": 0.4954, + "step": 23009 + }, + { + "epoch": 8.5, + "learning_rate": 2.1516196575933e-06, + "loss": 0.4731, + "step": 23010 + }, + { + "epoch": 8.5, + "learning_rate": 2.1515827072299545e-06, + "loss": 0.368, + "step": 23011 + }, + { + "epoch": 8.5, + "learning_rate": 2.151545756866609e-06, + "loss": 0.4795, + "step": 23012 + }, + { + "epoch": 8.5, + "learning_rate": 2.151508806503264e-06, + "loss": 0.3322, + "step": 23013 + }, + { + "epoch": 8.5, + "learning_rate": 2.151471856139919e-06, + "loss": 0.4428, + "step": 23014 + }, + { + "epoch": 8.5, + "learning_rate": 2.1514349057765735e-06, + "loss": 0.4082, + "step": 23015 + }, + { + "epoch": 8.5, + "learning_rate": 2.1513979554132286e-06, + "loss": 0.3247, + "step": 23016 + }, + { + "epoch": 8.5, + "learning_rate": 2.1513610050498828e-06, + "loss": 0.3028, + "step": 23017 + }, + { + "epoch": 8.5, + "learning_rate": 2.151324054686538e-06, + "loss": 0.3424, + "step": 23018 + }, + { + "epoch": 8.5, + "learning_rate": 2.1512871043231925e-06, + "loss": 0.3421, + "step": 23019 + }, + { + "epoch": 8.5, + "learning_rate": 2.1512501539598475e-06, + "loss": 0.4589, + "step": 23020 + }, + { + "epoch": 8.5, + "learning_rate": 2.151213203596502e-06, + "loss": 0.4551, + "step": 23021 + }, + { + "epoch": 8.5, + "learning_rate": 2.151176253233157e-06, + "loss": 0.3314, + "step": 23022 + }, + { + "epoch": 8.5, + "learning_rate": 2.1511393028698115e-06, + "loss": 0.3553, + "step": 23023 + }, + { + "epoch": 8.5, + "learning_rate": 2.151102352506466e-06, + "loss": 0.3254, + "step": 23024 + }, + { + "epoch": 8.5, + "learning_rate": 2.151065402143121e-06, + "loss": 0.4602, + "step": 23025 + }, + { + "epoch": 8.5, + "learning_rate": 2.1510284517797758e-06, + "loss": 0.3336, + "step": 23026 + }, + { + "epoch": 8.5, + "learning_rate": 2.150991501416431e-06, + "loss": 0.4569, + "step": 23027 + }, + { + "epoch": 8.5, + "learning_rate": 2.1509545510530855e-06, + "loss": 0.3341, + "step": 23028 + }, + { + "epoch": 8.5, + "learning_rate": 2.15091760068974e-06, + "loss": 0.4879, + "step": 23029 + }, + { + "epoch": 8.5, + "learning_rate": 2.1508806503263948e-06, + "loss": 0.2911, + "step": 23030 + }, + { + "epoch": 8.5, + "learning_rate": 2.1508436999630494e-06, + "loss": 0.4486, + "step": 23031 + }, + { + "epoch": 8.51, + "learning_rate": 2.1508067495997044e-06, + "loss": 0.3938, + "step": 23032 + }, + { + "epoch": 8.51, + "learning_rate": 2.150769799236359e-06, + "loss": 0.4746, + "step": 23033 + }, + { + "epoch": 8.51, + "learning_rate": 2.150732848873014e-06, + "loss": 0.2262, + "step": 23034 + }, + { + "epoch": 8.51, + "learning_rate": 2.1506958985096688e-06, + "loss": 0.3409, + "step": 23035 + }, + { + "epoch": 8.51, + "learning_rate": 2.1506589481463234e-06, + "loss": 0.3906, + "step": 23036 + }, + { + "epoch": 8.51, + "learning_rate": 2.150621997782978e-06, + "loss": 0.4447, + "step": 23037 + }, + { + "epoch": 8.51, + "learning_rate": 2.150585047419633e-06, + "loss": 0.2885, + "step": 23038 + }, + { + "epoch": 8.51, + "learning_rate": 2.1505480970562878e-06, + "loss": 0.2991, + "step": 23039 + }, + { + "epoch": 8.51, + "learning_rate": 2.150511146692943e-06, + "loss": 0.3795, + "step": 23040 + }, + { + "epoch": 8.51, + "learning_rate": 2.1504741963295974e-06, + "loss": 0.2707, + "step": 23041 + }, + { + "epoch": 8.51, + "learning_rate": 2.150437245966252e-06, + "loss": 0.6028, + "step": 23042 + }, + { + "epoch": 8.51, + "learning_rate": 2.1504002956029067e-06, + "loss": 0.3397, + "step": 23043 + }, + { + "epoch": 8.51, + "learning_rate": 2.1503633452395614e-06, + "loss": 0.4329, + "step": 23044 + }, + { + "epoch": 8.51, + "learning_rate": 2.1503263948762164e-06, + "loss": 0.5084, + "step": 23045 + }, + { + "epoch": 8.51, + "learning_rate": 2.150289444512871e-06, + "loss": 0.4221, + "step": 23046 + }, + { + "epoch": 8.51, + "learning_rate": 2.150252494149526e-06, + "loss": 0.5586, + "step": 23047 + }, + { + "epoch": 8.51, + "learning_rate": 2.1502155437861807e-06, + "loss": 0.3406, + "step": 23048 + }, + { + "epoch": 8.51, + "learning_rate": 2.1501785934228354e-06, + "loss": 0.3428, + "step": 23049 + }, + { + "epoch": 8.51, + "learning_rate": 2.15014164305949e-06, + "loss": 0.3384, + "step": 23050 + }, + { + "epoch": 8.51, + "learning_rate": 2.1501046926961447e-06, + "loss": 0.3978, + "step": 23051 + }, + { + "epoch": 8.51, + "learning_rate": 2.1500677423327997e-06, + "loss": 0.4205, + "step": 23052 + }, + { + "epoch": 8.51, + "learning_rate": 2.1500307919694544e-06, + "loss": 0.3706, + "step": 23053 + }, + { + "epoch": 8.51, + "learning_rate": 2.1499938416061094e-06, + "loss": 0.5727, + "step": 23054 + }, + { + "epoch": 8.51, + "learning_rate": 2.149956891242764e-06, + "loss": 0.4429, + "step": 23055 + }, + { + "epoch": 8.51, + "learning_rate": 2.1499199408794187e-06, + "loss": 0.3666, + "step": 23056 + }, + { + "epoch": 8.51, + "learning_rate": 2.1498829905160733e-06, + "loss": 0.3013, + "step": 23057 + }, + { + "epoch": 8.51, + "learning_rate": 2.1498460401527284e-06, + "loss": 0.3336, + "step": 23058 + }, + { + "epoch": 8.52, + "learning_rate": 2.149809089789383e-06, + "loss": 0.3838, + "step": 23059 + }, + { + "epoch": 8.52, + "learning_rate": 2.149772139426038e-06, + "loss": 0.3072, + "step": 23060 + }, + { + "epoch": 8.52, + "learning_rate": 2.1497351890626927e-06, + "loss": 0.2973, + "step": 23061 + }, + { + "epoch": 8.52, + "learning_rate": 2.149698238699347e-06, + "loss": 0.453, + "step": 23062 + }, + { + "epoch": 8.52, + "learning_rate": 2.149661288336002e-06, + "loss": 0.3967, + "step": 23063 + }, + { + "epoch": 8.52, + "learning_rate": 2.1496243379726566e-06, + "loss": 0.3163, + "step": 23064 + }, + { + "epoch": 8.52, + "learning_rate": 2.1495873876093117e-06, + "loss": 0.2799, + "step": 23065 + }, + { + "epoch": 8.52, + "learning_rate": 2.1495504372459663e-06, + "loss": 0.4566, + "step": 23066 + }, + { + "epoch": 8.52, + "learning_rate": 2.149513486882621e-06, + "loss": 0.5178, + "step": 23067 + }, + { + "epoch": 8.52, + "learning_rate": 2.1494765365192756e-06, + "loss": 0.3344, + "step": 23068 + }, + { + "epoch": 8.52, + "learning_rate": 2.1494395861559307e-06, + "loss": 0.4272, + "step": 23069 + }, + { + "epoch": 8.52, + "learning_rate": 2.1494026357925853e-06, + "loss": 0.3792, + "step": 23070 + }, + { + "epoch": 8.52, + "learning_rate": 2.14936568542924e-06, + "loss": 0.3819, + "step": 23071 + }, + { + "epoch": 8.52, + "learning_rate": 2.149328735065895e-06, + "loss": 0.4325, + "step": 23072 + }, + { + "epoch": 8.52, + "learning_rate": 2.1492917847025496e-06, + "loss": 0.3877, + "step": 23073 + }, + { + "epoch": 8.52, + "learning_rate": 2.1492548343392043e-06, + "loss": 0.3394, + "step": 23074 + }, + { + "epoch": 8.52, + "learning_rate": 2.149217883975859e-06, + "loss": 0.2976, + "step": 23075 + }, + { + "epoch": 8.52, + "learning_rate": 2.149180933612514e-06, + "loss": 0.347, + "step": 23076 + }, + { + "epoch": 8.52, + "learning_rate": 2.1491439832491686e-06, + "loss": 0.42, + "step": 23077 + }, + { + "epoch": 8.52, + "learning_rate": 2.1491070328858237e-06, + "loss": 0.4384, + "step": 23078 + }, + { + "epoch": 8.52, + "learning_rate": 2.1490700825224783e-06, + "loss": 0.3556, + "step": 23079 + }, + { + "epoch": 8.52, + "learning_rate": 2.149033132159133e-06, + "loss": 0.3616, + "step": 23080 + }, + { + "epoch": 8.52, + "learning_rate": 2.1489961817957876e-06, + "loss": 0.3208, + "step": 23081 + }, + { + "epoch": 8.52, + "learning_rate": 2.148959231432442e-06, + "loss": 0.2588, + "step": 23082 + }, + { + "epoch": 8.52, + "learning_rate": 2.1489222810690973e-06, + "loss": 0.2837, + "step": 23083 + }, + { + "epoch": 8.52, + "learning_rate": 2.148885330705752e-06, + "loss": 0.3886, + "step": 23084 + }, + { + "epoch": 8.52, + "learning_rate": 2.148848380342407e-06, + "loss": 0.4128, + "step": 23085 + }, + { + "epoch": 8.53, + "learning_rate": 2.1488114299790616e-06, + "loss": 0.3346, + "step": 23086 + }, + { + "epoch": 8.53, + "learning_rate": 2.1487744796157162e-06, + "loss": 0.3231, + "step": 23087 + }, + { + "epoch": 8.53, + "learning_rate": 2.148737529252371e-06, + "loss": 0.3798, + "step": 23088 + }, + { + "epoch": 8.53, + "learning_rate": 2.148700578889026e-06, + "loss": 0.4392, + "step": 23089 + }, + { + "epoch": 8.53, + "learning_rate": 2.1486636285256806e-06, + "loss": 0.4631, + "step": 23090 + }, + { + "epoch": 8.53, + "learning_rate": 2.148626678162335e-06, + "loss": 0.4265, + "step": 23091 + }, + { + "epoch": 8.53, + "learning_rate": 2.1485897277989903e-06, + "loss": 0.3796, + "step": 23092 + }, + { + "epoch": 8.53, + "learning_rate": 2.148552777435645e-06, + "loss": 0.4568, + "step": 23093 + }, + { + "epoch": 8.53, + "learning_rate": 2.1485158270722995e-06, + "loss": 0.4249, + "step": 23094 + }, + { + "epoch": 8.53, + "learning_rate": 2.148478876708954e-06, + "loss": 0.4106, + "step": 23095 + }, + { + "epoch": 8.53, + "learning_rate": 2.1484419263456092e-06, + "loss": 0.4056, + "step": 23096 + }, + { + "epoch": 8.53, + "learning_rate": 2.148404975982264e-06, + "loss": 0.3349, + "step": 23097 + }, + { + "epoch": 8.53, + "learning_rate": 2.148368025618919e-06, + "loss": 0.409, + "step": 23098 + }, + { + "epoch": 8.53, + "learning_rate": 2.1483310752555736e-06, + "loss": 0.4056, + "step": 23099 + }, + { + "epoch": 8.53, + "learning_rate": 2.148294124892228e-06, + "loss": 0.4166, + "step": 23100 + }, + { + "epoch": 8.53, + "learning_rate": 2.148257174528883e-06, + "loss": 0.4166, + "step": 23101 + }, + { + "epoch": 8.53, + "learning_rate": 2.1482202241655375e-06, + "loss": 0.4546, + "step": 23102 + }, + { + "epoch": 8.53, + "learning_rate": 2.1481832738021925e-06, + "loss": 0.3485, + "step": 23103 + }, + { + "epoch": 8.53, + "learning_rate": 2.148146323438847e-06, + "loss": 0.3436, + "step": 23104 + }, + { + "epoch": 8.53, + "learning_rate": 2.1481093730755022e-06, + "loss": 0.3125, + "step": 23105 + }, + { + "epoch": 8.53, + "learning_rate": 2.1480724227121564e-06, + "loss": 0.4578, + "step": 23106 + }, + { + "epoch": 8.53, + "learning_rate": 2.1480354723488115e-06, + "loss": 0.3408, + "step": 23107 + }, + { + "epoch": 8.53, + "learning_rate": 2.147998521985466e-06, + "loss": 0.4741, + "step": 23108 + }, + { + "epoch": 8.53, + "learning_rate": 2.147961571622121e-06, + "loss": 0.3329, + "step": 23109 + }, + { + "epoch": 8.53, + "learning_rate": 2.147924621258776e-06, + "loss": 0.3303, + "step": 23110 + }, + { + "epoch": 8.53, + "learning_rate": 2.1478876708954305e-06, + "loss": 0.3774, + "step": 23111 + }, + { + "epoch": 8.53, + "learning_rate": 2.147850720532085e-06, + "loss": 0.357, + "step": 23112 + }, + { + "epoch": 8.54, + "learning_rate": 2.1478137701687397e-06, + "loss": 0.445, + "step": 23113 + }, + { + "epoch": 8.54, + "learning_rate": 2.147776819805395e-06, + "loss": 0.3704, + "step": 23114 + }, + { + "epoch": 8.54, + "learning_rate": 2.1477398694420494e-06, + "loss": 0.4132, + "step": 23115 + }, + { + "epoch": 8.54, + "learning_rate": 2.1477029190787045e-06, + "loss": 0.5339, + "step": 23116 + }, + { + "epoch": 8.54, + "learning_rate": 2.147665968715359e-06, + "loss": 0.4391, + "step": 23117 + }, + { + "epoch": 8.54, + "learning_rate": 2.1476290183520138e-06, + "loss": 0.4149, + "step": 23118 + }, + { + "epoch": 8.54, + "learning_rate": 2.1475920679886684e-06, + "loss": 0.4185, + "step": 23119 + }, + { + "epoch": 8.54, + "learning_rate": 2.1475551176253235e-06, + "loss": 0.4061, + "step": 23120 + }, + { + "epoch": 8.54, + "learning_rate": 2.147518167261978e-06, + "loss": 0.3488, + "step": 23121 + }, + { + "epoch": 8.54, + "learning_rate": 2.1474812168986327e-06, + "loss": 0.4624, + "step": 23122 + }, + { + "epoch": 8.54, + "learning_rate": 2.147444266535288e-06, + "loss": 0.3147, + "step": 23123 + }, + { + "epoch": 8.54, + "learning_rate": 2.1474073161719424e-06, + "loss": 0.4468, + "step": 23124 + }, + { + "epoch": 8.54, + "learning_rate": 2.147370365808597e-06, + "loss": 0.4028, + "step": 23125 + }, + { + "epoch": 8.54, + "learning_rate": 2.1473334154452517e-06, + "loss": 0.4538, + "step": 23126 + }, + { + "epoch": 8.54, + "learning_rate": 2.1472964650819068e-06, + "loss": 0.422, + "step": 23127 + }, + { + "epoch": 8.54, + "learning_rate": 2.1472595147185614e-06, + "loss": 0.4266, + "step": 23128 + }, + { + "epoch": 8.54, + "learning_rate": 2.1472225643552165e-06, + "loss": 0.4753, + "step": 23129 + }, + { + "epoch": 8.54, + "learning_rate": 2.147185613991871e-06, + "loss": 0.3265, + "step": 23130 + }, + { + "epoch": 8.54, + "learning_rate": 2.1471486636285257e-06, + "loss": 0.3705, + "step": 23131 + }, + { + "epoch": 8.54, + "learning_rate": 2.1471117132651804e-06, + "loss": 0.4124, + "step": 23132 + }, + { + "epoch": 8.54, + "learning_rate": 2.147074762901835e-06, + "loss": 0.3188, + "step": 23133 + }, + { + "epoch": 8.54, + "learning_rate": 2.14703781253849e-06, + "loss": 0.5141, + "step": 23134 + }, + { + "epoch": 8.54, + "learning_rate": 2.1470008621751447e-06, + "loss": 0.4129, + "step": 23135 + }, + { + "epoch": 8.54, + "learning_rate": 2.1469639118117998e-06, + "loss": 0.4013, + "step": 23136 + }, + { + "epoch": 8.54, + "learning_rate": 2.1469269614484544e-06, + "loss": 0.3475, + "step": 23137 + }, + { + "epoch": 8.54, + "learning_rate": 2.146890011085109e-06, + "loss": 0.4204, + "step": 23138 + }, + { + "epoch": 8.54, + "learning_rate": 2.1468530607217637e-06, + "loss": 0.4811, + "step": 23139 + }, + { + "epoch": 8.55, + "learning_rate": 2.1468161103584187e-06, + "loss": 0.3513, + "step": 23140 + }, + { + "epoch": 8.55, + "learning_rate": 2.1467791599950734e-06, + "loss": 0.3352, + "step": 23141 + }, + { + "epoch": 8.55, + "learning_rate": 2.146742209631728e-06, + "loss": 0.3092, + "step": 23142 + }, + { + "epoch": 8.55, + "learning_rate": 2.146705259268383e-06, + "loss": 0.3087, + "step": 23143 + }, + { + "epoch": 8.55, + "learning_rate": 2.1466683089050377e-06, + "loss": 0.3702, + "step": 23144 + }, + { + "epoch": 8.55, + "learning_rate": 2.1466313585416923e-06, + "loss": 0.5741, + "step": 23145 + }, + { + "epoch": 8.55, + "learning_rate": 2.146594408178347e-06, + "loss": 0.4018, + "step": 23146 + }, + { + "epoch": 8.55, + "learning_rate": 2.146557457815002e-06, + "loss": 0.5001, + "step": 23147 + }, + { + "epoch": 8.55, + "learning_rate": 2.1465205074516567e-06, + "loss": 0.4259, + "step": 23148 + }, + { + "epoch": 8.55, + "learning_rate": 2.1464835570883117e-06, + "loss": 0.3223, + "step": 23149 + }, + { + "epoch": 8.55, + "learning_rate": 2.1464466067249664e-06, + "loss": 0.2602, + "step": 23150 + }, + { + "epoch": 8.55, + "learning_rate": 2.1464096563616206e-06, + "loss": 0.4105, + "step": 23151 + }, + { + "epoch": 8.55, + "learning_rate": 2.1463727059982756e-06, + "loss": 0.3415, + "step": 23152 + }, + { + "epoch": 8.55, + "learning_rate": 2.1463357556349303e-06, + "loss": 0.3083, + "step": 23153 + }, + { + "epoch": 8.55, + "learning_rate": 2.1462988052715853e-06, + "loss": 0.4357, + "step": 23154 + }, + { + "epoch": 8.55, + "learning_rate": 2.14626185490824e-06, + "loss": 0.4566, + "step": 23155 + }, + { + "epoch": 8.55, + "learning_rate": 2.146224904544895e-06, + "loss": 0.4455, + "step": 23156 + }, + { + "epoch": 8.55, + "learning_rate": 2.1461879541815492e-06, + "loss": 0.3504, + "step": 23157 + }, + { + "epoch": 8.55, + "learning_rate": 2.1461510038182043e-06, + "loss": 0.3452, + "step": 23158 + }, + { + "epoch": 8.55, + "learning_rate": 2.146114053454859e-06, + "loss": 0.243, + "step": 23159 + }, + { + "epoch": 8.55, + "learning_rate": 2.146077103091514e-06, + "loss": 0.329, + "step": 23160 + }, + { + "epoch": 8.55, + "learning_rate": 2.1460401527281686e-06, + "loss": 0.4755, + "step": 23161 + }, + { + "epoch": 8.55, + "learning_rate": 2.1460032023648233e-06, + "loss": 0.3705, + "step": 23162 + }, + { + "epoch": 8.55, + "learning_rate": 2.145966252001478e-06, + "loss": 0.4826, + "step": 23163 + }, + { + "epoch": 8.55, + "learning_rate": 2.1459293016381326e-06, + "loss": 0.3962, + "step": 23164 + }, + { + "epoch": 8.55, + "learning_rate": 2.1458923512747876e-06, + "loss": 0.3263, + "step": 23165 + }, + { + "epoch": 8.55, + "learning_rate": 2.1458554009114422e-06, + "loss": 0.4495, + "step": 23166 + }, + { + "epoch": 8.56, + "learning_rate": 2.1458184505480973e-06, + "loss": 0.357, + "step": 23167 + }, + { + "epoch": 8.56, + "learning_rate": 2.145781500184752e-06, + "loss": 0.3663, + "step": 23168 + }, + { + "epoch": 8.56, + "learning_rate": 2.1457445498214066e-06, + "loss": 0.5263, + "step": 23169 + }, + { + "epoch": 8.56, + "learning_rate": 2.1457075994580612e-06, + "loss": 0.5577, + "step": 23170 + }, + { + "epoch": 8.56, + "learning_rate": 2.145670649094716e-06, + "loss": 0.3942, + "step": 23171 + }, + { + "epoch": 8.56, + "learning_rate": 2.145633698731371e-06, + "loss": 0.4071, + "step": 23172 + }, + { + "epoch": 8.56, + "learning_rate": 2.1455967483680255e-06, + "loss": 0.4337, + "step": 23173 + }, + { + "epoch": 8.56, + "learning_rate": 2.1455597980046806e-06, + "loss": 0.4017, + "step": 23174 + }, + { + "epoch": 8.56, + "learning_rate": 2.1455228476413352e-06, + "loss": 0.3105, + "step": 23175 + }, + { + "epoch": 8.56, + "learning_rate": 2.14548589727799e-06, + "loss": 0.3474, + "step": 23176 + }, + { + "epoch": 8.56, + "learning_rate": 2.1454489469146445e-06, + "loss": 0.5371, + "step": 23177 + }, + { + "epoch": 8.56, + "learning_rate": 2.1454119965512996e-06, + "loss": 0.3797, + "step": 23178 + }, + { + "epoch": 8.56, + "learning_rate": 2.1453750461879542e-06, + "loss": 0.3867, + "step": 23179 + }, + { + "epoch": 8.56, + "learning_rate": 2.1453380958246093e-06, + "loss": 0.4224, + "step": 23180 + }, + { + "epoch": 8.56, + "learning_rate": 2.145301145461264e-06, + "loss": 0.4293, + "step": 23181 + }, + { + "epoch": 8.56, + "learning_rate": 2.1452641950979185e-06, + "loss": 0.4346, + "step": 23182 + }, + { + "epoch": 8.56, + "learning_rate": 2.145227244734573e-06, + "loss": 0.3392, + "step": 23183 + }, + { + "epoch": 8.56, + "learning_rate": 2.145190294371228e-06, + "loss": 0.491, + "step": 23184 + }, + { + "epoch": 8.56, + "learning_rate": 2.145153344007883e-06, + "loss": 0.3839, + "step": 23185 + }, + { + "epoch": 8.56, + "learning_rate": 2.1451163936445375e-06, + "loss": 0.4685, + "step": 23186 + }, + { + "epoch": 8.56, + "learning_rate": 2.1450794432811926e-06, + "loss": 0.4507, + "step": 23187 + }, + { + "epoch": 8.56, + "learning_rate": 2.1450424929178472e-06, + "loss": 0.5717, + "step": 23188 + }, + { + "epoch": 8.56, + "learning_rate": 2.145005542554502e-06, + "loss": 0.4151, + "step": 23189 + }, + { + "epoch": 8.56, + "learning_rate": 2.1449685921911565e-06, + "loss": 0.2917, + "step": 23190 + }, + { + "epoch": 8.56, + "learning_rate": 2.144931641827811e-06, + "loss": 0.2748, + "step": 23191 + }, + { + "epoch": 8.56, + "learning_rate": 2.144894691464466e-06, + "loss": 0.312, + "step": 23192 + }, + { + "epoch": 8.56, + "learning_rate": 2.144857741101121e-06, + "loss": 0.4397, + "step": 23193 + }, + { + "epoch": 8.56, + "learning_rate": 2.144820790737776e-06, + "loss": 0.4431, + "step": 23194 + }, + { + "epoch": 8.57, + "learning_rate": 2.1447838403744305e-06, + "loss": 0.3805, + "step": 23195 + }, + { + "epoch": 8.57, + "learning_rate": 2.144746890011085e-06, + "loss": 0.3986, + "step": 23196 + }, + { + "epoch": 8.57, + "learning_rate": 2.14470993964774e-06, + "loss": 0.3719, + "step": 23197 + }, + { + "epoch": 8.57, + "learning_rate": 2.144672989284395e-06, + "loss": 0.3313, + "step": 23198 + }, + { + "epoch": 8.57, + "learning_rate": 2.1446360389210495e-06, + "loss": 0.3202, + "step": 23199 + }, + { + "epoch": 8.57, + "learning_rate": 2.1445990885577045e-06, + "loss": 0.2966, + "step": 23200 + }, + { + "epoch": 8.57, + "learning_rate": 2.144562138194359e-06, + "loss": 0.3185, + "step": 23201 + }, + { + "epoch": 8.57, + "learning_rate": 2.1445251878310134e-06, + "loss": 0.3624, + "step": 23202 + }, + { + "epoch": 8.57, + "learning_rate": 2.1444882374676685e-06, + "loss": 0.5097, + "step": 23203 + }, + { + "epoch": 8.57, + "learning_rate": 2.144451287104323e-06, + "loss": 0.3391, + "step": 23204 + }, + { + "epoch": 8.57, + "learning_rate": 2.144414336740978e-06, + "loss": 0.4254, + "step": 23205 + }, + { + "epoch": 8.57, + "learning_rate": 2.1443773863776328e-06, + "loss": 0.4539, + "step": 23206 + }, + { + "epoch": 8.57, + "learning_rate": 2.144340436014288e-06, + "loss": 0.457, + "step": 23207 + }, + { + "epoch": 8.57, + "learning_rate": 2.144303485650942e-06, + "loss": 0.3497, + "step": 23208 + }, + { + "epoch": 8.57, + "learning_rate": 2.144266535287597e-06, + "loss": 0.33, + "step": 23209 + }, + { + "epoch": 8.57, + "learning_rate": 2.1442295849242518e-06, + "loss": 0.3921, + "step": 23210 + }, + { + "epoch": 8.57, + "learning_rate": 2.1441926345609064e-06, + "loss": 0.3688, + "step": 23211 + }, + { + "epoch": 8.57, + "learning_rate": 2.1441556841975615e-06, + "loss": 0.3664, + "step": 23212 + }, + { + "epoch": 8.57, + "learning_rate": 2.144118733834216e-06, + "loss": 0.258, + "step": 23213 + }, + { + "epoch": 8.57, + "learning_rate": 2.1440817834708707e-06, + "loss": 0.4204, + "step": 23214 + }, + { + "epoch": 8.57, + "learning_rate": 2.1440448331075254e-06, + "loss": 0.3924, + "step": 23215 + }, + { + "epoch": 8.57, + "learning_rate": 2.1440078827441804e-06, + "loss": 0.4404, + "step": 23216 + }, + { + "epoch": 8.57, + "learning_rate": 2.143970932380835e-06, + "loss": 0.2908, + "step": 23217 + }, + { + "epoch": 8.57, + "learning_rate": 2.14393398201749e-06, + "loss": 0.4061, + "step": 23218 + }, + { + "epoch": 8.57, + "learning_rate": 2.1438970316541448e-06, + "loss": 0.3889, + "step": 23219 + }, + { + "epoch": 8.57, + "learning_rate": 2.1438600812907994e-06, + "loss": 0.4349, + "step": 23220 + }, + { + "epoch": 8.57, + "learning_rate": 2.143823130927454e-06, + "loss": 0.3851, + "step": 23221 + }, + { + "epoch": 8.58, + "learning_rate": 2.1437861805641087e-06, + "loss": 0.3219, + "step": 23222 + }, + { + "epoch": 8.58, + "learning_rate": 2.1437492302007637e-06, + "loss": 0.2244, + "step": 23223 + }, + { + "epoch": 8.58, + "learning_rate": 2.1437122798374184e-06, + "loss": 0.3708, + "step": 23224 + }, + { + "epoch": 8.58, + "learning_rate": 2.1436753294740734e-06, + "loss": 0.4606, + "step": 23225 + }, + { + "epoch": 8.58, + "learning_rate": 2.143638379110728e-06, + "loss": 0.411, + "step": 23226 + }, + { + "epoch": 8.58, + "learning_rate": 2.1436014287473827e-06, + "loss": 0.4423, + "step": 23227 + }, + { + "epoch": 8.58, + "learning_rate": 2.1435644783840373e-06, + "loss": 0.3897, + "step": 23228 + }, + { + "epoch": 8.58, + "learning_rate": 2.1435275280206924e-06, + "loss": 0.5602, + "step": 23229 + }, + { + "epoch": 8.58, + "learning_rate": 2.143490577657347e-06, + "loss": 0.3988, + "step": 23230 + }, + { + "epoch": 8.58, + "learning_rate": 2.1434536272940017e-06, + "loss": 0.3277, + "step": 23231 + }, + { + "epoch": 8.58, + "learning_rate": 2.1434166769306567e-06, + "loss": 0.4577, + "step": 23232 + }, + { + "epoch": 8.58, + "learning_rate": 2.1433797265673114e-06, + "loss": 0.35, + "step": 23233 + }, + { + "epoch": 8.58, + "learning_rate": 2.143342776203966e-06, + "loss": 0.4106, + "step": 23234 + }, + { + "epoch": 8.58, + "learning_rate": 2.1433058258406206e-06, + "loss": 0.3682, + "step": 23235 + }, + { + "epoch": 8.58, + "learning_rate": 2.1432688754772757e-06, + "loss": 0.4954, + "step": 23236 + }, + { + "epoch": 8.58, + "learning_rate": 2.1432319251139303e-06, + "loss": 0.4659, + "step": 23237 + }, + { + "epoch": 8.58, + "learning_rate": 2.1431949747505854e-06, + "loss": 0.3951, + "step": 23238 + }, + { + "epoch": 8.58, + "learning_rate": 2.14315802438724e-06, + "loss": 0.4328, + "step": 23239 + }, + { + "epoch": 8.58, + "learning_rate": 2.1431210740238947e-06, + "loss": 0.5279, + "step": 23240 + }, + { + "epoch": 8.58, + "learning_rate": 2.1430841236605493e-06, + "loss": 0.4417, + "step": 23241 + }, + { + "epoch": 8.58, + "learning_rate": 2.143047173297204e-06, + "loss": 0.3589, + "step": 23242 + }, + { + "epoch": 8.58, + "learning_rate": 2.143010222933859e-06, + "loss": 0.4403, + "step": 23243 + }, + { + "epoch": 8.58, + "learning_rate": 2.1429732725705136e-06, + "loss": 0.5039, + "step": 23244 + }, + { + "epoch": 8.58, + "learning_rate": 2.1429363222071687e-06, + "loss": 0.3481, + "step": 23245 + }, + { + "epoch": 8.58, + "learning_rate": 2.1428993718438233e-06, + "loss": 0.3918, + "step": 23246 + }, + { + "epoch": 8.58, + "learning_rate": 2.142862421480478e-06, + "loss": 0.3645, + "step": 23247 + }, + { + "epoch": 8.58, + "learning_rate": 2.1428254711171326e-06, + "loss": 0.4644, + "step": 23248 + }, + { + "epoch": 8.59, + "learning_rate": 2.1427885207537877e-06, + "loss": 0.3435, + "step": 23249 + }, + { + "epoch": 8.59, + "learning_rate": 2.1427515703904423e-06, + "loss": 0.2463, + "step": 23250 + }, + { + "epoch": 8.59, + "learning_rate": 2.142714620027097e-06, + "loss": 0.3847, + "step": 23251 + }, + { + "epoch": 8.59, + "learning_rate": 2.142677669663752e-06, + "loss": 0.4475, + "step": 23252 + }, + { + "epoch": 8.59, + "learning_rate": 2.142640719300406e-06, + "loss": 0.3279, + "step": 23253 + }, + { + "epoch": 8.59, + "learning_rate": 2.1426037689370613e-06, + "loss": 0.3936, + "step": 23254 + }, + { + "epoch": 8.59, + "learning_rate": 2.142566818573716e-06, + "loss": 0.3829, + "step": 23255 + }, + { + "epoch": 8.59, + "learning_rate": 2.142529868210371e-06, + "loss": 0.2642, + "step": 23256 + }, + { + "epoch": 8.59, + "learning_rate": 2.1424929178470256e-06, + "loss": 0.3642, + "step": 23257 + }, + { + "epoch": 8.59, + "learning_rate": 2.1424559674836802e-06, + "loss": 0.3364, + "step": 23258 + }, + { + "epoch": 8.59, + "learning_rate": 2.142419017120335e-06, + "loss": 0.3517, + "step": 23259 + }, + { + "epoch": 8.59, + "learning_rate": 2.1423820667569895e-06, + "loss": 0.2652, + "step": 23260 + }, + { + "epoch": 8.59, + "learning_rate": 2.1423451163936446e-06, + "loss": 0.2989, + "step": 23261 + }, + { + "epoch": 8.59, + "learning_rate": 2.142308166030299e-06, + "loss": 0.4075, + "step": 23262 + }, + { + "epoch": 8.59, + "learning_rate": 2.1422712156669543e-06, + "loss": 0.3972, + "step": 23263 + }, + { + "epoch": 8.59, + "learning_rate": 2.142234265303609e-06, + "loss": 0.3833, + "step": 23264 + }, + { + "epoch": 8.59, + "learning_rate": 2.1421973149402635e-06, + "loss": 0.4984, + "step": 23265 + }, + { + "epoch": 8.59, + "learning_rate": 2.142160364576918e-06, + "loss": 0.3707, + "step": 23266 + }, + { + "epoch": 8.59, + "learning_rate": 2.1421234142135732e-06, + "loss": 0.462, + "step": 23267 + }, + { + "epoch": 8.59, + "learning_rate": 2.142086463850228e-06, + "loss": 0.3765, + "step": 23268 + }, + { + "epoch": 8.59, + "learning_rate": 2.142049513486883e-06, + "loss": 0.3403, + "step": 23269 + }, + { + "epoch": 8.59, + "learning_rate": 2.1420125631235376e-06, + "loss": 0.365, + "step": 23270 + }, + { + "epoch": 8.59, + "learning_rate": 2.141975612760192e-06, + "loss": 0.4141, + "step": 23271 + }, + { + "epoch": 8.59, + "learning_rate": 2.141938662396847e-06, + "loss": 0.3657, + "step": 23272 + }, + { + "epoch": 8.59, + "learning_rate": 2.1419017120335015e-06, + "loss": 0.4267, + "step": 23273 + }, + { + "epoch": 8.59, + "learning_rate": 2.1418647616701565e-06, + "loss": 0.5076, + "step": 23274 + }, + { + "epoch": 8.59, + "learning_rate": 2.141827811306811e-06, + "loss": 0.3143, + "step": 23275 + }, + { + "epoch": 8.6, + "learning_rate": 2.1417908609434662e-06, + "loss": 0.3267, + "step": 23276 + }, + { + "epoch": 8.6, + "learning_rate": 2.141753910580121e-06, + "loss": 0.6033, + "step": 23277 + }, + { + "epoch": 8.6, + "learning_rate": 2.1417169602167755e-06, + "loss": 0.3195, + "step": 23278 + }, + { + "epoch": 8.6, + "learning_rate": 2.14168000985343e-06, + "loss": 0.3968, + "step": 23279 + }, + { + "epoch": 8.6, + "learning_rate": 2.1416430594900848e-06, + "loss": 0.3847, + "step": 23280 + }, + { + "epoch": 8.6, + "learning_rate": 2.14160610912674e-06, + "loss": 0.5228, + "step": 23281 + }, + { + "epoch": 8.6, + "learning_rate": 2.1415691587633945e-06, + "loss": 0.3202, + "step": 23282 + }, + { + "epoch": 8.6, + "learning_rate": 2.1415322084000495e-06, + "loss": 0.4035, + "step": 23283 + }, + { + "epoch": 8.6, + "learning_rate": 2.141495258036704e-06, + "loss": 0.404, + "step": 23284 + }, + { + "epoch": 8.6, + "learning_rate": 2.141458307673359e-06, + "loss": 0.3835, + "step": 23285 + }, + { + "epoch": 8.6, + "learning_rate": 2.1414213573100134e-06, + "loss": 0.3147, + "step": 23286 + }, + { + "epoch": 8.6, + "learning_rate": 2.1413844069466685e-06, + "loss": 0.4559, + "step": 23287 + }, + { + "epoch": 8.6, + "learning_rate": 2.141347456583323e-06, + "loss": 0.4816, + "step": 23288 + }, + { + "epoch": 8.6, + "learning_rate": 2.141310506219978e-06, + "loss": 0.3151, + "step": 23289 + }, + { + "epoch": 8.6, + "learning_rate": 2.141273555856633e-06, + "loss": 0.3373, + "step": 23290 + }, + { + "epoch": 8.6, + "learning_rate": 2.141236605493287e-06, + "loss": 0.4636, + "step": 23291 + }, + { + "epoch": 8.6, + "learning_rate": 2.141199655129942e-06, + "loss": 0.3679, + "step": 23292 + }, + { + "epoch": 8.6, + "learning_rate": 2.1411627047665967e-06, + "loss": 0.4354, + "step": 23293 + }, + { + "epoch": 8.6, + "learning_rate": 2.141125754403252e-06, + "loss": 0.4233, + "step": 23294 + }, + { + "epoch": 8.6, + "learning_rate": 2.1410888040399064e-06, + "loss": 0.3168, + "step": 23295 + }, + { + "epoch": 8.6, + "learning_rate": 2.1410518536765615e-06, + "loss": 0.4159, + "step": 23296 + }, + { + "epoch": 8.6, + "learning_rate": 2.1410149033132157e-06, + "loss": 0.3211, + "step": 23297 + }, + { + "epoch": 8.6, + "learning_rate": 2.1409779529498708e-06, + "loss": 0.2946, + "step": 23298 + }, + { + "epoch": 8.6, + "learning_rate": 2.1409410025865254e-06, + "loss": 0.3438, + "step": 23299 + }, + { + "epoch": 8.6, + "learning_rate": 2.14090405222318e-06, + "loss": 0.4051, + "step": 23300 + }, + { + "epoch": 8.6, + "learning_rate": 2.140867101859835e-06, + "loss": 0.3904, + "step": 23301 + }, + { + "epoch": 8.6, + "learning_rate": 2.1408301514964897e-06, + "loss": 0.4744, + "step": 23302 + }, + { + "epoch": 8.61, + "learning_rate": 2.1407932011331444e-06, + "loss": 0.3689, + "step": 23303 + }, + { + "epoch": 8.61, + "learning_rate": 2.140756250769799e-06, + "loss": 0.5345, + "step": 23304 + }, + { + "epoch": 8.61, + "learning_rate": 2.140719300406454e-06, + "loss": 0.249, + "step": 23305 + }, + { + "epoch": 8.61, + "learning_rate": 2.1406823500431087e-06, + "loss": 0.4067, + "step": 23306 + }, + { + "epoch": 8.61, + "learning_rate": 2.1406453996797638e-06, + "loss": 0.4626, + "step": 23307 + }, + { + "epoch": 8.61, + "learning_rate": 2.1406084493164184e-06, + "loss": 0.4568, + "step": 23308 + }, + { + "epoch": 8.61, + "learning_rate": 2.140571498953073e-06, + "loss": 0.2568, + "step": 23309 + }, + { + "epoch": 8.61, + "learning_rate": 2.1405345485897277e-06, + "loss": 0.3558, + "step": 23310 + }, + { + "epoch": 8.61, + "learning_rate": 2.1404975982263823e-06, + "loss": 0.4, + "step": 23311 + }, + { + "epoch": 8.61, + "learning_rate": 2.1404606478630374e-06, + "loss": 0.4539, + "step": 23312 + }, + { + "epoch": 8.61, + "learning_rate": 2.140423697499692e-06, + "loss": 0.3183, + "step": 23313 + }, + { + "epoch": 8.61, + "learning_rate": 2.140386747136347e-06, + "loss": 0.3887, + "step": 23314 + }, + { + "epoch": 8.61, + "learning_rate": 2.1403497967730017e-06, + "loss": 0.3966, + "step": 23315 + }, + { + "epoch": 8.61, + "learning_rate": 2.1403128464096563e-06, + "loss": 0.3746, + "step": 23316 + }, + { + "epoch": 8.61, + "learning_rate": 2.140275896046311e-06, + "loss": 0.4255, + "step": 23317 + }, + { + "epoch": 8.61, + "learning_rate": 2.140238945682966e-06, + "loss": 0.3139, + "step": 23318 + }, + { + "epoch": 8.61, + "learning_rate": 2.1402019953196207e-06, + "loss": 0.4579, + "step": 23319 + }, + { + "epoch": 8.61, + "learning_rate": 2.1401650449562753e-06, + "loss": 0.2363, + "step": 23320 + }, + { + "epoch": 8.61, + "learning_rate": 2.1401280945929304e-06, + "loss": 0.3236, + "step": 23321 + }, + { + "epoch": 8.61, + "learning_rate": 2.140091144229585e-06, + "loss": 0.3051, + "step": 23322 + }, + { + "epoch": 8.61, + "learning_rate": 2.1400541938662396e-06, + "loss": 0.2487, + "step": 23323 + }, + { + "epoch": 8.61, + "learning_rate": 2.1400172435028943e-06, + "loss": 0.3981, + "step": 23324 + }, + { + "epoch": 8.61, + "learning_rate": 2.1399802931395493e-06, + "loss": 0.4401, + "step": 23325 + }, + { + "epoch": 8.61, + "learning_rate": 2.139943342776204e-06, + "loss": 0.4607, + "step": 23326 + }, + { + "epoch": 8.61, + "learning_rate": 2.139906392412859e-06, + "loss": 0.3968, + "step": 23327 + }, + { + "epoch": 8.61, + "learning_rate": 2.1398694420495137e-06, + "loss": 0.3478, + "step": 23328 + }, + { + "epoch": 8.61, + "learning_rate": 2.1398324916861683e-06, + "loss": 0.398, + "step": 23329 + }, + { + "epoch": 8.62, + "learning_rate": 2.139795541322823e-06, + "loss": 0.3277, + "step": 23330 + }, + { + "epoch": 8.62, + "learning_rate": 2.1397585909594776e-06, + "loss": 0.4186, + "step": 23331 + }, + { + "epoch": 8.62, + "learning_rate": 2.1397216405961326e-06, + "loss": 0.4731, + "step": 23332 + }, + { + "epoch": 8.62, + "learning_rate": 2.1396846902327873e-06, + "loss": 0.3145, + "step": 23333 + }, + { + "epoch": 8.62, + "learning_rate": 2.1396477398694423e-06, + "loss": 0.4153, + "step": 23334 + }, + { + "epoch": 8.62, + "learning_rate": 2.139610789506097e-06, + "loss": 0.3683, + "step": 23335 + }, + { + "epoch": 8.62, + "learning_rate": 2.1395738391427516e-06, + "loss": 0.4572, + "step": 23336 + }, + { + "epoch": 8.62, + "learning_rate": 2.1395368887794063e-06, + "loss": 0.3823, + "step": 23337 + }, + { + "epoch": 8.62, + "learning_rate": 2.1394999384160613e-06, + "loss": 0.4803, + "step": 23338 + }, + { + "epoch": 8.62, + "learning_rate": 2.139462988052716e-06, + "loss": 0.4232, + "step": 23339 + }, + { + "epoch": 8.62, + "learning_rate": 2.1394260376893706e-06, + "loss": 0.3943, + "step": 23340 + }, + { + "epoch": 8.62, + "learning_rate": 2.1393890873260256e-06, + "loss": 0.4179, + "step": 23341 + }, + { + "epoch": 8.62, + "learning_rate": 2.13935213696268e-06, + "loss": 0.3191, + "step": 23342 + }, + { + "epoch": 8.62, + "learning_rate": 2.139315186599335e-06, + "loss": 0.29, + "step": 23343 + }, + { + "epoch": 8.62, + "learning_rate": 2.1392782362359896e-06, + "loss": 0.3815, + "step": 23344 + }, + { + "epoch": 8.62, + "learning_rate": 2.1392412858726446e-06, + "loss": 0.3283, + "step": 23345 + }, + { + "epoch": 8.62, + "learning_rate": 2.1392043355092993e-06, + "loss": 0.3432, + "step": 23346 + }, + { + "epoch": 8.62, + "learning_rate": 2.1391673851459543e-06, + "loss": 0.3603, + "step": 23347 + }, + { + "epoch": 8.62, + "learning_rate": 2.1391304347826085e-06, + "loss": 0.4039, + "step": 23348 + }, + { + "epoch": 8.62, + "learning_rate": 2.1390934844192636e-06, + "loss": 0.3767, + "step": 23349 + }, + { + "epoch": 8.62, + "learning_rate": 2.1390565340559182e-06, + "loss": 0.4611, + "step": 23350 + }, + { + "epoch": 8.62, + "learning_rate": 2.139019583692573e-06, + "loss": 0.3054, + "step": 23351 + }, + { + "epoch": 8.62, + "learning_rate": 2.138982633329228e-06, + "loss": 0.4096, + "step": 23352 + }, + { + "epoch": 8.62, + "learning_rate": 2.1389456829658826e-06, + "loss": 0.4022, + "step": 23353 + }, + { + "epoch": 8.62, + "learning_rate": 2.138908732602537e-06, + "loss": 0.2946, + "step": 23354 + }, + { + "epoch": 8.62, + "learning_rate": 2.138871782239192e-06, + "loss": 0.3724, + "step": 23355 + }, + { + "epoch": 8.62, + "learning_rate": 2.138834831875847e-06, + "loss": 0.3088, + "step": 23356 + }, + { + "epoch": 8.63, + "learning_rate": 2.1387978815125015e-06, + "loss": 0.3923, + "step": 23357 + }, + { + "epoch": 8.63, + "learning_rate": 2.1387609311491566e-06, + "loss": 0.4032, + "step": 23358 + }, + { + "epoch": 8.63, + "learning_rate": 2.1387239807858112e-06, + "loss": 0.373, + "step": 23359 + }, + { + "epoch": 8.63, + "learning_rate": 2.138687030422466e-06, + "loss": 0.4862, + "step": 23360 + }, + { + "epoch": 8.63, + "learning_rate": 2.1386500800591205e-06, + "loss": 0.5714, + "step": 23361 + }, + { + "epoch": 8.63, + "learning_rate": 2.138613129695775e-06, + "loss": 0.41, + "step": 23362 + }, + { + "epoch": 8.63, + "learning_rate": 2.13857617933243e-06, + "loss": 0.3603, + "step": 23363 + }, + { + "epoch": 8.63, + "learning_rate": 2.138539228969085e-06, + "loss": 0.3864, + "step": 23364 + }, + { + "epoch": 8.63, + "learning_rate": 2.13850227860574e-06, + "loss": 0.3667, + "step": 23365 + }, + { + "epoch": 8.63, + "learning_rate": 2.1384653282423945e-06, + "loss": 0.2544, + "step": 23366 + }, + { + "epoch": 8.63, + "learning_rate": 2.138428377879049e-06, + "loss": 0.4909, + "step": 23367 + }, + { + "epoch": 8.63, + "learning_rate": 2.138391427515704e-06, + "loss": 0.4252, + "step": 23368 + }, + { + "epoch": 8.63, + "learning_rate": 2.138354477152359e-06, + "loss": 0.3731, + "step": 23369 + }, + { + "epoch": 8.63, + "learning_rate": 2.1383175267890135e-06, + "loss": 0.4315, + "step": 23370 + }, + { + "epoch": 8.63, + "learning_rate": 2.138280576425668e-06, + "loss": 0.4419, + "step": 23371 + }, + { + "epoch": 8.63, + "learning_rate": 2.138243626062323e-06, + "loss": 0.4002, + "step": 23372 + }, + { + "epoch": 8.63, + "learning_rate": 2.138206675698978e-06, + "loss": 0.3901, + "step": 23373 + }, + { + "epoch": 8.63, + "learning_rate": 2.1381697253356325e-06, + "loss": 0.2798, + "step": 23374 + }, + { + "epoch": 8.63, + "learning_rate": 2.138132774972287e-06, + "loss": 0.3504, + "step": 23375 + }, + { + "epoch": 8.63, + "learning_rate": 2.138095824608942e-06, + "loss": 0.3055, + "step": 23376 + }, + { + "epoch": 8.63, + "learning_rate": 2.138058874245597e-06, + "loss": 0.3392, + "step": 23377 + }, + { + "epoch": 8.63, + "learning_rate": 2.138021923882252e-06, + "loss": 0.2948, + "step": 23378 + }, + { + "epoch": 8.63, + "learning_rate": 2.1379849735189065e-06, + "loss": 0.4341, + "step": 23379 + }, + { + "epoch": 8.63, + "learning_rate": 2.137948023155561e-06, + "loss": 0.3881, + "step": 23380 + }, + { + "epoch": 8.63, + "learning_rate": 2.1379110727922158e-06, + "loss": 0.315, + "step": 23381 + }, + { + "epoch": 8.63, + "learning_rate": 2.1378741224288704e-06, + "loss": 0.452, + "step": 23382 + }, + { + "epoch": 8.63, + "learning_rate": 2.1378371720655255e-06, + "loss": 0.3507, + "step": 23383 + }, + { + "epoch": 8.64, + "learning_rate": 2.13780022170218e-06, + "loss": 0.4319, + "step": 23384 + }, + { + "epoch": 8.64, + "learning_rate": 2.137763271338835e-06, + "loss": 0.4057, + "step": 23385 + }, + { + "epoch": 8.64, + "learning_rate": 2.13772632097549e-06, + "loss": 0.3094, + "step": 23386 + }, + { + "epoch": 8.64, + "learning_rate": 2.1376893706121444e-06, + "loss": 0.4458, + "step": 23387 + }, + { + "epoch": 8.64, + "learning_rate": 2.137652420248799e-06, + "loss": 0.4167, + "step": 23388 + }, + { + "epoch": 8.64, + "learning_rate": 2.137615469885454e-06, + "loss": 0.4424, + "step": 23389 + }, + { + "epoch": 8.64, + "learning_rate": 2.1375785195221088e-06, + "loss": 0.4553, + "step": 23390 + }, + { + "epoch": 8.64, + "learning_rate": 2.1375415691587634e-06, + "loss": 0.3655, + "step": 23391 + }, + { + "epoch": 8.64, + "learning_rate": 2.1375046187954185e-06, + "loss": 0.3036, + "step": 23392 + }, + { + "epoch": 8.64, + "learning_rate": 2.1374676684320727e-06, + "loss": 0.507, + "step": 23393 + }, + { + "epoch": 8.64, + "learning_rate": 2.1374307180687277e-06, + "loss": 0.337, + "step": 23394 + }, + { + "epoch": 8.64, + "learning_rate": 2.1373937677053824e-06, + "loss": 0.4648, + "step": 23395 + }, + { + "epoch": 8.64, + "learning_rate": 2.1373568173420374e-06, + "loss": 0.4302, + "step": 23396 + }, + { + "epoch": 8.64, + "learning_rate": 2.137319866978692e-06, + "loss": 0.3549, + "step": 23397 + }, + { + "epoch": 8.64, + "learning_rate": 2.137282916615347e-06, + "loss": 0.3772, + "step": 23398 + }, + { + "epoch": 8.64, + "learning_rate": 2.1372459662520013e-06, + "loss": 0.3649, + "step": 23399 + }, + { + "epoch": 8.64, + "learning_rate": 2.137209015888656e-06, + "loss": 0.441, + "step": 23400 + }, + { + "epoch": 8.64, + "learning_rate": 2.137172065525311e-06, + "loss": 0.4585, + "step": 23401 + }, + { + "epoch": 8.64, + "learning_rate": 2.1371351151619657e-06, + "loss": 0.5162, + "step": 23402 + }, + { + "epoch": 8.64, + "learning_rate": 2.1370981647986207e-06, + "loss": 0.29, + "step": 23403 + }, + { + "epoch": 8.64, + "learning_rate": 2.1370612144352754e-06, + "loss": 0.3821, + "step": 23404 + }, + { + "epoch": 8.64, + "learning_rate": 2.13702426407193e-06, + "loss": 0.3873, + "step": 23405 + }, + { + "epoch": 8.64, + "learning_rate": 2.1369873137085846e-06, + "loss": 0.461, + "step": 23406 + }, + { + "epoch": 8.64, + "learning_rate": 2.1369503633452397e-06, + "loss": 0.35, + "step": 23407 + }, + { + "epoch": 8.64, + "learning_rate": 2.1369134129818943e-06, + "loss": 0.3195, + "step": 23408 + }, + { + "epoch": 8.64, + "learning_rate": 2.1368764626185494e-06, + "loss": 0.3523, + "step": 23409 + }, + { + "epoch": 8.64, + "learning_rate": 2.136839512255204e-06, + "loss": 0.4825, + "step": 23410 + }, + { + "epoch": 8.65, + "learning_rate": 2.1368025618918587e-06, + "loss": 0.326, + "step": 23411 + }, + { + "epoch": 8.65, + "learning_rate": 2.1367656115285133e-06, + "loss": 0.432, + "step": 23412 + }, + { + "epoch": 8.65, + "learning_rate": 2.136728661165168e-06, + "loss": 0.4896, + "step": 23413 + }, + { + "epoch": 8.65, + "learning_rate": 2.136691710801823e-06, + "loss": 0.3262, + "step": 23414 + }, + { + "epoch": 8.65, + "learning_rate": 2.1366547604384776e-06, + "loss": 0.4166, + "step": 23415 + }, + { + "epoch": 8.65, + "learning_rate": 2.1366178100751327e-06, + "loss": 0.4184, + "step": 23416 + }, + { + "epoch": 8.65, + "learning_rate": 2.1365808597117873e-06, + "loss": 0.3802, + "step": 23417 + }, + { + "epoch": 8.65, + "learning_rate": 2.136543909348442e-06, + "loss": 0.4482, + "step": 23418 + }, + { + "epoch": 8.65, + "learning_rate": 2.1365069589850966e-06, + "loss": 0.4778, + "step": 23419 + }, + { + "epoch": 8.65, + "learning_rate": 2.1364700086217512e-06, + "loss": 0.3668, + "step": 23420 + }, + { + "epoch": 8.65, + "learning_rate": 2.1364330582584063e-06, + "loss": 0.3165, + "step": 23421 + }, + { + "epoch": 8.65, + "learning_rate": 2.136396107895061e-06, + "loss": 0.3696, + "step": 23422 + }, + { + "epoch": 8.65, + "learning_rate": 2.136359157531716e-06, + "loss": 0.3707, + "step": 23423 + }, + { + "epoch": 8.65, + "learning_rate": 2.1363222071683706e-06, + "loss": 0.5169, + "step": 23424 + }, + { + "epoch": 8.65, + "learning_rate": 2.1362852568050253e-06, + "loss": 0.3967, + "step": 23425 + }, + { + "epoch": 8.65, + "learning_rate": 2.13624830644168e-06, + "loss": 0.3313, + "step": 23426 + }, + { + "epoch": 8.65, + "learning_rate": 2.136211356078335e-06, + "loss": 0.3581, + "step": 23427 + }, + { + "epoch": 8.65, + "learning_rate": 2.1361744057149896e-06, + "loss": 0.35, + "step": 23428 + }, + { + "epoch": 8.65, + "learning_rate": 2.1361374553516447e-06, + "loss": 0.4447, + "step": 23429 + }, + { + "epoch": 8.65, + "learning_rate": 2.1361005049882993e-06, + "loss": 0.4442, + "step": 23430 + }, + { + "epoch": 8.65, + "learning_rate": 2.136063554624954e-06, + "loss": 0.3765, + "step": 23431 + }, + { + "epoch": 8.65, + "learning_rate": 2.1360266042616086e-06, + "loss": 0.3886, + "step": 23432 + }, + { + "epoch": 8.65, + "learning_rate": 2.135989653898263e-06, + "loss": 0.4293, + "step": 23433 + }, + { + "epoch": 8.65, + "learning_rate": 2.1359527035349183e-06, + "loss": 0.2823, + "step": 23434 + }, + { + "epoch": 8.65, + "learning_rate": 2.135915753171573e-06, + "loss": 0.352, + "step": 23435 + }, + { + "epoch": 8.65, + "learning_rate": 2.135878802808228e-06, + "loss": 0.4021, + "step": 23436 + }, + { + "epoch": 8.65, + "learning_rate": 2.1358418524448826e-06, + "loss": 0.3022, + "step": 23437 + }, + { + "epoch": 8.66, + "learning_rate": 2.1358049020815372e-06, + "loss": 0.4082, + "step": 23438 + }, + { + "epoch": 8.66, + "learning_rate": 2.135767951718192e-06, + "loss": 0.2464, + "step": 23439 + }, + { + "epoch": 8.66, + "learning_rate": 2.1357310013548465e-06, + "loss": 0.4712, + "step": 23440 + }, + { + "epoch": 8.66, + "learning_rate": 2.1356940509915016e-06, + "loss": 0.5105, + "step": 23441 + }, + { + "epoch": 8.66, + "learning_rate": 2.135657100628156e-06, + "loss": 0.422, + "step": 23442 + }, + { + "epoch": 8.66, + "learning_rate": 2.135620150264811e-06, + "loss": 0.3772, + "step": 23443 + }, + { + "epoch": 8.66, + "learning_rate": 2.1355831999014655e-06, + "loss": 0.3018, + "step": 23444 + }, + { + "epoch": 8.66, + "learning_rate": 2.1355462495381205e-06, + "loss": 0.4631, + "step": 23445 + }, + { + "epoch": 8.66, + "learning_rate": 2.135509299174775e-06, + "loss": 0.3413, + "step": 23446 + }, + { + "epoch": 8.66, + "learning_rate": 2.1354723488114302e-06, + "loss": 0.3546, + "step": 23447 + }, + { + "epoch": 8.66, + "learning_rate": 2.135435398448085e-06, + "loss": 0.4555, + "step": 23448 + }, + { + "epoch": 8.66, + "learning_rate": 2.1353984480847395e-06, + "loss": 0.3827, + "step": 23449 + }, + { + "epoch": 8.66, + "learning_rate": 2.135361497721394e-06, + "loss": 0.3479, + "step": 23450 + }, + { + "epoch": 8.66, + "learning_rate": 2.1353245473580488e-06, + "loss": 0.4771, + "step": 23451 + }, + { + "epoch": 8.66, + "learning_rate": 2.135287596994704e-06, + "loss": 0.4312, + "step": 23452 + }, + { + "epoch": 8.66, + "learning_rate": 2.1352506466313585e-06, + "loss": 0.3971, + "step": 23453 + }, + { + "epoch": 8.66, + "learning_rate": 2.1352136962680135e-06, + "loss": 0.4247, + "step": 23454 + }, + { + "epoch": 8.66, + "learning_rate": 2.135176745904668e-06, + "loss": 0.4417, + "step": 23455 + }, + { + "epoch": 8.66, + "learning_rate": 2.135139795541323e-06, + "loss": 0.3693, + "step": 23456 + }, + { + "epoch": 8.66, + "learning_rate": 2.1351028451779774e-06, + "loss": 0.4615, + "step": 23457 + }, + { + "epoch": 8.66, + "learning_rate": 2.1350658948146325e-06, + "loss": 0.3665, + "step": 23458 + }, + { + "epoch": 8.66, + "learning_rate": 2.135028944451287e-06, + "loss": 0.3764, + "step": 23459 + }, + { + "epoch": 8.66, + "learning_rate": 2.1349919940879418e-06, + "loss": 0.4698, + "step": 23460 + }, + { + "epoch": 8.66, + "learning_rate": 2.134955043724597e-06, + "loss": 0.4094, + "step": 23461 + }, + { + "epoch": 8.66, + "learning_rate": 2.1349180933612515e-06, + "loss": 0.2934, + "step": 23462 + }, + { + "epoch": 8.66, + "learning_rate": 2.134881142997906e-06, + "loss": 0.4564, + "step": 23463 + }, + { + "epoch": 8.66, + "learning_rate": 2.1348441926345608e-06, + "loss": 0.3849, + "step": 23464 + }, + { + "epoch": 8.67, + "learning_rate": 2.134807242271216e-06, + "loss": 0.2877, + "step": 23465 + }, + { + "epoch": 8.67, + "learning_rate": 2.1347702919078704e-06, + "loss": 0.371, + "step": 23466 + }, + { + "epoch": 8.67, + "learning_rate": 2.1347333415445255e-06, + "loss": 0.4152, + "step": 23467 + }, + { + "epoch": 8.67, + "learning_rate": 2.13469639118118e-06, + "loss": 0.3164, + "step": 23468 + }, + { + "epoch": 8.67, + "learning_rate": 2.1346594408178348e-06, + "loss": 0.4479, + "step": 23469 + }, + { + "epoch": 8.67, + "learning_rate": 2.1346224904544894e-06, + "loss": 0.4282, + "step": 23470 + }, + { + "epoch": 8.67, + "learning_rate": 2.134585540091144e-06, + "loss": 0.4492, + "step": 23471 + }, + { + "epoch": 8.67, + "learning_rate": 2.134548589727799e-06, + "loss": 0.376, + "step": 23472 + }, + { + "epoch": 8.67, + "learning_rate": 2.1345116393644537e-06, + "loss": 0.4196, + "step": 23473 + }, + { + "epoch": 8.67, + "learning_rate": 2.134474689001109e-06, + "loss": 0.4248, + "step": 23474 + }, + { + "epoch": 8.67, + "learning_rate": 2.1344377386377634e-06, + "loss": 0.4242, + "step": 23475 + }, + { + "epoch": 8.67, + "learning_rate": 2.134400788274418e-06, + "loss": 0.2604, + "step": 23476 + }, + { + "epoch": 8.67, + "learning_rate": 2.1343638379110727e-06, + "loss": 0.3214, + "step": 23477 + }, + { + "epoch": 8.67, + "learning_rate": 2.1343268875477278e-06, + "loss": 0.3914, + "step": 23478 + }, + { + "epoch": 8.67, + "learning_rate": 2.1342899371843824e-06, + "loss": 0.2705, + "step": 23479 + }, + { + "epoch": 8.67, + "learning_rate": 2.134252986821037e-06, + "loss": 0.4165, + "step": 23480 + }, + { + "epoch": 8.67, + "learning_rate": 2.134216036457692e-06, + "loss": 0.55, + "step": 23481 + }, + { + "epoch": 8.67, + "learning_rate": 2.1341790860943463e-06, + "loss": 0.3726, + "step": 23482 + }, + { + "epoch": 8.67, + "learning_rate": 2.1341421357310014e-06, + "loss": 0.3886, + "step": 23483 + }, + { + "epoch": 8.67, + "learning_rate": 2.134105185367656e-06, + "loss": 0.4143, + "step": 23484 + }, + { + "epoch": 8.67, + "learning_rate": 2.134068235004311e-06, + "loss": 0.3799, + "step": 23485 + }, + { + "epoch": 8.67, + "learning_rate": 2.1340312846409657e-06, + "loss": 0.4471, + "step": 23486 + }, + { + "epoch": 8.67, + "learning_rate": 2.1339943342776208e-06, + "loss": 0.4442, + "step": 23487 + }, + { + "epoch": 8.67, + "learning_rate": 2.133957383914275e-06, + "loss": 0.3863, + "step": 23488 + }, + { + "epoch": 8.67, + "learning_rate": 2.1339204335509296e-06, + "loss": 0.4937, + "step": 23489 + }, + { + "epoch": 8.67, + "learning_rate": 2.1338834831875847e-06, + "loss": 0.4514, + "step": 23490 + }, + { + "epoch": 8.67, + "learning_rate": 2.1338465328242393e-06, + "loss": 0.4037, + "step": 23491 + }, + { + "epoch": 8.68, + "learning_rate": 2.1338095824608944e-06, + "loss": 0.3455, + "step": 23492 + }, + { + "epoch": 8.68, + "learning_rate": 2.133772632097549e-06, + "loss": 0.4896, + "step": 23493 + }, + { + "epoch": 8.68, + "learning_rate": 2.1337356817342037e-06, + "loss": 0.4874, + "step": 23494 + }, + { + "epoch": 8.68, + "learning_rate": 2.1336987313708583e-06, + "loss": 0.3459, + "step": 23495 + }, + { + "epoch": 8.68, + "learning_rate": 2.1336617810075134e-06, + "loss": 0.4936, + "step": 23496 + }, + { + "epoch": 8.68, + "learning_rate": 2.133624830644168e-06, + "loss": 0.3914, + "step": 23497 + }, + { + "epoch": 8.68, + "learning_rate": 2.133587880280823e-06, + "loss": 0.3901, + "step": 23498 + }, + { + "epoch": 8.68, + "learning_rate": 2.1335509299174777e-06, + "loss": 0.391, + "step": 23499 + }, + { + "epoch": 8.68, + "learning_rate": 2.1335139795541323e-06, + "loss": 0.3727, + "step": 23500 + }, + { + "epoch": 8.68, + "learning_rate": 2.133477029190787e-06, + "loss": 0.4044, + "step": 23501 + }, + { + "epoch": 8.68, + "learning_rate": 2.1334400788274416e-06, + "loss": 0.4882, + "step": 23502 + }, + { + "epoch": 8.68, + "learning_rate": 2.1334031284640967e-06, + "loss": 0.2754, + "step": 23503 + }, + { + "epoch": 8.68, + "learning_rate": 2.1333661781007513e-06, + "loss": 0.3008, + "step": 23504 + }, + { + "epoch": 8.68, + "learning_rate": 2.1333292277374064e-06, + "loss": 0.3031, + "step": 23505 + }, + { + "epoch": 8.68, + "learning_rate": 2.133292277374061e-06, + "loss": 0.3221, + "step": 23506 + }, + { + "epoch": 8.68, + "learning_rate": 2.1332553270107156e-06, + "loss": 0.3317, + "step": 23507 + }, + { + "epoch": 8.68, + "learning_rate": 2.1332183766473703e-06, + "loss": 0.3082, + "step": 23508 + }, + { + "epoch": 8.68, + "learning_rate": 2.133181426284025e-06, + "loss": 0.4837, + "step": 23509 + }, + { + "epoch": 8.68, + "learning_rate": 2.13314447592068e-06, + "loss": 0.4968, + "step": 23510 + }, + { + "epoch": 8.68, + "learning_rate": 2.1331075255573346e-06, + "loss": 0.4212, + "step": 23511 + }, + { + "epoch": 8.68, + "learning_rate": 2.1330705751939897e-06, + "loss": 0.4048, + "step": 23512 + }, + { + "epoch": 8.68, + "learning_rate": 2.1330336248306443e-06, + "loss": 0.4272, + "step": 23513 + }, + { + "epoch": 8.68, + "learning_rate": 2.132996674467299e-06, + "loss": 0.3912, + "step": 23514 + }, + { + "epoch": 8.68, + "learning_rate": 2.1329597241039536e-06, + "loss": 0.3543, + "step": 23515 + }, + { + "epoch": 8.68, + "learning_rate": 2.1329227737406086e-06, + "loss": 0.4158, + "step": 23516 + }, + { + "epoch": 8.68, + "learning_rate": 2.1328858233772633e-06, + "loss": 0.2871, + "step": 23517 + }, + { + "epoch": 8.68, + "learning_rate": 2.1328488730139183e-06, + "loss": 0.3512, + "step": 23518 + }, + { + "epoch": 8.69, + "learning_rate": 2.132811922650573e-06, + "loss": 0.364, + "step": 23519 + }, + { + "epoch": 8.69, + "learning_rate": 2.1327749722872276e-06, + "loss": 0.3881, + "step": 23520 + }, + { + "epoch": 8.69, + "learning_rate": 2.1327380219238822e-06, + "loss": 0.3576, + "step": 23521 + }, + { + "epoch": 8.69, + "learning_rate": 2.132701071560537e-06, + "loss": 0.4457, + "step": 23522 + }, + { + "epoch": 8.69, + "learning_rate": 2.132664121197192e-06, + "loss": 0.5162, + "step": 23523 + }, + { + "epoch": 8.69, + "learning_rate": 2.1326271708338466e-06, + "loss": 0.3692, + "step": 23524 + }, + { + "epoch": 8.69, + "learning_rate": 2.1325902204705016e-06, + "loss": 0.3191, + "step": 23525 + }, + { + "epoch": 8.69, + "learning_rate": 2.1325532701071563e-06, + "loss": 0.4212, + "step": 23526 + }, + { + "epoch": 8.69, + "learning_rate": 2.132516319743811e-06, + "loss": 0.4684, + "step": 23527 + }, + { + "epoch": 8.69, + "learning_rate": 2.1324793693804655e-06, + "loss": 0.3303, + "step": 23528 + }, + { + "epoch": 8.69, + "learning_rate": 2.13244241901712e-06, + "loss": 0.3986, + "step": 23529 + }, + { + "epoch": 8.69, + "learning_rate": 2.1324054686537752e-06, + "loss": 0.3299, + "step": 23530 + }, + { + "epoch": 8.69, + "learning_rate": 2.13236851829043e-06, + "loss": 0.398, + "step": 23531 + }, + { + "epoch": 8.69, + "learning_rate": 2.132331567927085e-06, + "loss": 0.3527, + "step": 23532 + }, + { + "epoch": 8.69, + "learning_rate": 2.132294617563739e-06, + "loss": 0.4713, + "step": 23533 + }, + { + "epoch": 8.69, + "learning_rate": 2.132257667200394e-06, + "loss": 0.436, + "step": 23534 + }, + { + "epoch": 8.69, + "learning_rate": 2.132220716837049e-06, + "loss": 0.4767, + "step": 23535 + }, + { + "epoch": 8.69, + "learning_rate": 2.132183766473704e-06, + "loss": 0.4621, + "step": 23536 + }, + { + "epoch": 8.69, + "learning_rate": 2.1321468161103585e-06, + "loss": 0.3867, + "step": 23537 + }, + { + "epoch": 8.69, + "learning_rate": 2.1321098657470136e-06, + "loss": 0.4122, + "step": 23538 + }, + { + "epoch": 8.69, + "learning_rate": 2.132072915383668e-06, + "loss": 0.2895, + "step": 23539 + }, + { + "epoch": 8.69, + "learning_rate": 2.1320359650203224e-06, + "loss": 0.4192, + "step": 23540 + }, + { + "epoch": 8.69, + "learning_rate": 2.1319990146569775e-06, + "loss": 0.38, + "step": 23541 + }, + { + "epoch": 8.69, + "learning_rate": 2.131962064293632e-06, + "loss": 0.3499, + "step": 23542 + }, + { + "epoch": 8.69, + "learning_rate": 2.131925113930287e-06, + "loss": 0.4037, + "step": 23543 + }, + { + "epoch": 8.69, + "learning_rate": 2.131888163566942e-06, + "loss": 0.3262, + "step": 23544 + }, + { + "epoch": 8.69, + "learning_rate": 2.1318512132035965e-06, + "loss": 0.4634, + "step": 23545 + }, + { + "epoch": 8.69, + "learning_rate": 2.131814262840251e-06, + "loss": 0.2399, + "step": 23546 + }, + { + "epoch": 8.7, + "learning_rate": 2.131777312476906e-06, + "loss": 0.4105, + "step": 23547 + }, + { + "epoch": 8.7, + "learning_rate": 2.131740362113561e-06, + "loss": 0.4099, + "step": 23548 + }, + { + "epoch": 8.7, + "learning_rate": 2.1317034117502154e-06, + "loss": 0.2799, + "step": 23549 + }, + { + "epoch": 8.7, + "learning_rate": 2.1316664613868705e-06, + "loss": 0.2895, + "step": 23550 + }, + { + "epoch": 8.7, + "learning_rate": 2.131629511023525e-06, + "loss": 0.3941, + "step": 23551 + }, + { + "epoch": 8.7, + "learning_rate": 2.1315925606601798e-06, + "loss": 0.2426, + "step": 23552 + }, + { + "epoch": 8.7, + "learning_rate": 2.1315556102968344e-06, + "loss": 0.3798, + "step": 23553 + }, + { + "epoch": 8.7, + "learning_rate": 2.1315186599334895e-06, + "loss": 0.259, + "step": 23554 + }, + { + "epoch": 8.7, + "learning_rate": 2.131481709570144e-06, + "loss": 0.3217, + "step": 23555 + }, + { + "epoch": 8.7, + "learning_rate": 2.131444759206799e-06, + "loss": 0.4129, + "step": 23556 + }, + { + "epoch": 8.7, + "learning_rate": 2.131407808843454e-06, + "loss": 0.31, + "step": 23557 + }, + { + "epoch": 8.7, + "learning_rate": 2.1313708584801084e-06, + "loss": 0.4452, + "step": 23558 + }, + { + "epoch": 8.7, + "learning_rate": 2.131333908116763e-06, + "loss": 0.2217, + "step": 23559 + }, + { + "epoch": 8.7, + "learning_rate": 2.1312969577534177e-06, + "loss": 0.408, + "step": 23560 + }, + { + "epoch": 8.7, + "learning_rate": 2.1312600073900728e-06, + "loss": 0.4189, + "step": 23561 + }, + { + "epoch": 8.7, + "learning_rate": 2.1312230570267274e-06, + "loss": 0.4885, + "step": 23562 + }, + { + "epoch": 8.7, + "learning_rate": 2.1311861066633825e-06, + "loss": 0.4375, + "step": 23563 + }, + { + "epoch": 8.7, + "learning_rate": 2.131149156300037e-06, + "loss": 0.3919, + "step": 23564 + }, + { + "epoch": 8.7, + "learning_rate": 2.1311122059366917e-06, + "loss": 0.3262, + "step": 23565 + }, + { + "epoch": 8.7, + "learning_rate": 2.1310752555733464e-06, + "loss": 0.3531, + "step": 23566 + }, + { + "epoch": 8.7, + "learning_rate": 2.1310383052100014e-06, + "loss": 0.326, + "step": 23567 + }, + { + "epoch": 8.7, + "learning_rate": 2.131001354846656e-06, + "loss": 0.3299, + "step": 23568 + }, + { + "epoch": 8.7, + "learning_rate": 2.1309644044833107e-06, + "loss": 0.4488, + "step": 23569 + }, + { + "epoch": 8.7, + "learning_rate": 2.1309274541199658e-06, + "loss": 0.3156, + "step": 23570 + }, + { + "epoch": 8.7, + "learning_rate": 2.1308905037566204e-06, + "loss": 0.4453, + "step": 23571 + }, + { + "epoch": 8.7, + "learning_rate": 2.130853553393275e-06, + "loss": 0.3513, + "step": 23572 + }, + { + "epoch": 8.7, + "learning_rate": 2.1308166030299297e-06, + "loss": 0.3869, + "step": 23573 + }, + { + "epoch": 8.71, + "learning_rate": 2.1307796526665847e-06, + "loss": 0.3936, + "step": 23574 + }, + { + "epoch": 8.71, + "learning_rate": 2.1307427023032394e-06, + "loss": 0.3995, + "step": 23575 + }, + { + "epoch": 8.71, + "learning_rate": 2.1307057519398944e-06, + "loss": 0.3932, + "step": 23576 + }, + { + "epoch": 8.71, + "learning_rate": 2.130668801576549e-06, + "loss": 0.3798, + "step": 23577 + }, + { + "epoch": 8.71, + "learning_rate": 2.1306318512132037e-06, + "loss": 0.4726, + "step": 23578 + }, + { + "epoch": 8.71, + "learning_rate": 2.1305949008498583e-06, + "loss": 0.4091, + "step": 23579 + }, + { + "epoch": 8.71, + "learning_rate": 2.130557950486513e-06, + "loss": 0.4174, + "step": 23580 + }, + { + "epoch": 8.71, + "learning_rate": 2.130521000123168e-06, + "loss": 0.4744, + "step": 23581 + }, + { + "epoch": 8.71, + "learning_rate": 2.1304840497598227e-06, + "loss": 0.3592, + "step": 23582 + }, + { + "epoch": 8.71, + "learning_rate": 2.1304470993964777e-06, + "loss": 0.56, + "step": 23583 + }, + { + "epoch": 8.71, + "learning_rate": 2.130410149033132e-06, + "loss": 0.4858, + "step": 23584 + }, + { + "epoch": 8.71, + "learning_rate": 2.130373198669787e-06, + "loss": 0.3742, + "step": 23585 + }, + { + "epoch": 8.71, + "learning_rate": 2.1303362483064416e-06, + "loss": 0.3764, + "step": 23586 + }, + { + "epoch": 8.71, + "learning_rate": 2.1302992979430967e-06, + "loss": 0.3875, + "step": 23587 + }, + { + "epoch": 8.71, + "learning_rate": 2.1302623475797513e-06, + "loss": 0.3164, + "step": 23588 + }, + { + "epoch": 8.71, + "learning_rate": 2.130225397216406e-06, + "loss": 0.4105, + "step": 23589 + }, + { + "epoch": 8.71, + "learning_rate": 2.1301884468530606e-06, + "loss": 0.4692, + "step": 23590 + }, + { + "epoch": 8.71, + "learning_rate": 2.1301514964897152e-06, + "loss": 0.3492, + "step": 23591 + }, + { + "epoch": 8.71, + "learning_rate": 2.1301145461263703e-06, + "loss": 0.3588, + "step": 23592 + }, + { + "epoch": 8.71, + "learning_rate": 2.130077595763025e-06, + "loss": 0.4902, + "step": 23593 + }, + { + "epoch": 8.71, + "learning_rate": 2.13004064539968e-06, + "loss": 0.471, + "step": 23594 + }, + { + "epoch": 8.71, + "learning_rate": 2.1300036950363346e-06, + "loss": 0.4403, + "step": 23595 + }, + { + "epoch": 8.71, + "learning_rate": 2.1299667446729893e-06, + "loss": 0.3749, + "step": 23596 + }, + { + "epoch": 8.71, + "learning_rate": 2.129929794309644e-06, + "loss": 0.321, + "step": 23597 + }, + { + "epoch": 8.71, + "learning_rate": 2.129892843946299e-06, + "loss": 0.3891, + "step": 23598 + }, + { + "epoch": 8.71, + "learning_rate": 2.1298558935829536e-06, + "loss": 0.3185, + "step": 23599 + }, + { + "epoch": 8.71, + "learning_rate": 2.1298189432196082e-06, + "loss": 0.4649, + "step": 23600 + }, + { + "epoch": 8.72, + "learning_rate": 2.1297819928562633e-06, + "loss": 0.2988, + "step": 23601 + }, + { + "epoch": 8.72, + "learning_rate": 2.129745042492918e-06, + "loss": 0.4187, + "step": 23602 + }, + { + "epoch": 8.72, + "learning_rate": 2.1297080921295726e-06, + "loss": 0.3397, + "step": 23603 + }, + { + "epoch": 8.72, + "learning_rate": 2.1296711417662272e-06, + "loss": 0.4216, + "step": 23604 + }, + { + "epoch": 8.72, + "learning_rate": 2.1296341914028823e-06, + "loss": 0.3551, + "step": 23605 + }, + { + "epoch": 8.72, + "learning_rate": 2.129597241039537e-06, + "loss": 0.3118, + "step": 23606 + }, + { + "epoch": 8.72, + "learning_rate": 2.129560290676192e-06, + "loss": 0.3705, + "step": 23607 + }, + { + "epoch": 8.72, + "learning_rate": 2.1295233403128466e-06, + "loss": 0.3095, + "step": 23608 + }, + { + "epoch": 8.72, + "learning_rate": 2.1294863899495012e-06, + "loss": 0.5199, + "step": 23609 + }, + { + "epoch": 8.72, + "learning_rate": 2.129449439586156e-06, + "loss": 0.2715, + "step": 23610 + }, + { + "epoch": 8.72, + "learning_rate": 2.1294124892228105e-06, + "loss": 0.4626, + "step": 23611 + }, + { + "epoch": 8.72, + "learning_rate": 2.1293755388594656e-06, + "loss": 0.4625, + "step": 23612 + }, + { + "epoch": 8.72, + "learning_rate": 2.1293385884961202e-06, + "loss": 0.2453, + "step": 23613 + }, + { + "epoch": 8.72, + "learning_rate": 2.1293016381327753e-06, + "loss": 0.4953, + "step": 23614 + }, + { + "epoch": 8.72, + "learning_rate": 2.12926468776943e-06, + "loss": 0.3804, + "step": 23615 + }, + { + "epoch": 8.72, + "learning_rate": 2.1292277374060845e-06, + "loss": 0.4855, + "step": 23616 + }, + { + "epoch": 8.72, + "learning_rate": 2.129190787042739e-06, + "loss": 0.4106, + "step": 23617 + }, + { + "epoch": 8.72, + "learning_rate": 2.1291538366793942e-06, + "loss": 0.3358, + "step": 23618 + }, + { + "epoch": 8.72, + "learning_rate": 2.129116886316049e-06, + "loss": 0.341, + "step": 23619 + }, + { + "epoch": 8.72, + "learning_rate": 2.1290799359527035e-06, + "loss": 0.48, + "step": 23620 + }, + { + "epoch": 8.72, + "learning_rate": 2.1290429855893586e-06, + "loss": 0.4491, + "step": 23621 + }, + { + "epoch": 8.72, + "learning_rate": 2.1290060352260132e-06, + "loss": 0.3036, + "step": 23622 + }, + { + "epoch": 8.72, + "learning_rate": 2.128969084862668e-06, + "loss": 0.4203, + "step": 23623 + }, + { + "epoch": 8.72, + "learning_rate": 2.1289321344993225e-06, + "loss": 0.3653, + "step": 23624 + }, + { + "epoch": 8.72, + "learning_rate": 2.1288951841359775e-06, + "loss": 0.4056, + "step": 23625 + }, + { + "epoch": 8.72, + "learning_rate": 2.128858233772632e-06, + "loss": 0.4061, + "step": 23626 + }, + { + "epoch": 8.72, + "learning_rate": 2.1288212834092872e-06, + "loss": 0.489, + "step": 23627 + }, + { + "epoch": 8.73, + "learning_rate": 2.1287843330459415e-06, + "loss": 0.3453, + "step": 23628 + }, + { + "epoch": 8.73, + "learning_rate": 2.128747382682596e-06, + "loss": 0.3709, + "step": 23629 + }, + { + "epoch": 8.73, + "learning_rate": 2.128710432319251e-06, + "loss": 0.4154, + "step": 23630 + }, + { + "epoch": 8.73, + "learning_rate": 2.1286734819559058e-06, + "loss": 0.3781, + "step": 23631 + }, + { + "epoch": 8.73, + "learning_rate": 2.128636531592561e-06, + "loss": 0.3299, + "step": 23632 + }, + { + "epoch": 8.73, + "learning_rate": 2.1285995812292155e-06, + "loss": 0.3551, + "step": 23633 + }, + { + "epoch": 8.73, + "learning_rate": 2.12856263086587e-06, + "loss": 0.3772, + "step": 23634 + }, + { + "epoch": 8.73, + "learning_rate": 2.1285256805025248e-06, + "loss": 0.4653, + "step": 23635 + }, + { + "epoch": 8.73, + "learning_rate": 2.12848873013918e-06, + "loss": 0.3717, + "step": 23636 + }, + { + "epoch": 8.73, + "learning_rate": 2.1284517797758345e-06, + "loss": 0.5881, + "step": 23637 + }, + { + "epoch": 8.73, + "learning_rate": 2.1284148294124895e-06, + "loss": 0.4421, + "step": 23638 + }, + { + "epoch": 8.73, + "learning_rate": 2.128377879049144e-06, + "loss": 0.3533, + "step": 23639 + }, + { + "epoch": 8.73, + "learning_rate": 2.1283409286857988e-06, + "loss": 0.4285, + "step": 23640 + }, + { + "epoch": 8.73, + "learning_rate": 2.1283039783224534e-06, + "loss": 0.3662, + "step": 23641 + }, + { + "epoch": 8.73, + "learning_rate": 2.128267027959108e-06, + "loss": 0.368, + "step": 23642 + }, + { + "epoch": 8.73, + "learning_rate": 2.128230077595763e-06, + "loss": 0.3136, + "step": 23643 + }, + { + "epoch": 8.73, + "learning_rate": 2.1281931272324178e-06, + "loss": 0.4267, + "step": 23644 + }, + { + "epoch": 8.73, + "learning_rate": 2.128156176869073e-06, + "loss": 0.307, + "step": 23645 + }, + { + "epoch": 8.73, + "learning_rate": 2.1281192265057275e-06, + "loss": 0.3938, + "step": 23646 + }, + { + "epoch": 8.73, + "learning_rate": 2.128082276142382e-06, + "loss": 0.3349, + "step": 23647 + }, + { + "epoch": 8.73, + "learning_rate": 2.1280453257790367e-06, + "loss": 0.3677, + "step": 23648 + }, + { + "epoch": 8.73, + "learning_rate": 2.1280083754156914e-06, + "loss": 0.3499, + "step": 23649 + }, + { + "epoch": 8.73, + "learning_rate": 2.1279714250523464e-06, + "loss": 0.4115, + "step": 23650 + }, + { + "epoch": 8.73, + "learning_rate": 2.127934474689001e-06, + "loss": 0.4317, + "step": 23651 + }, + { + "epoch": 8.73, + "learning_rate": 2.127897524325656e-06, + "loss": 0.3612, + "step": 23652 + }, + { + "epoch": 8.73, + "learning_rate": 2.1278605739623108e-06, + "loss": 0.465, + "step": 23653 + }, + { + "epoch": 8.73, + "learning_rate": 2.1278236235989654e-06, + "loss": 0.4314, + "step": 23654 + }, + { + "epoch": 8.74, + "learning_rate": 2.12778667323562e-06, + "loss": 0.5752, + "step": 23655 + }, + { + "epoch": 8.74, + "learning_rate": 2.127749722872275e-06, + "loss": 0.2739, + "step": 23656 + }, + { + "epoch": 8.74, + "learning_rate": 2.1277127725089297e-06, + "loss": 0.4767, + "step": 23657 + }, + { + "epoch": 8.74, + "learning_rate": 2.1276758221455848e-06, + "loss": 0.4318, + "step": 23658 + }, + { + "epoch": 8.74, + "learning_rate": 2.1276388717822394e-06, + "loss": 0.3848, + "step": 23659 + }, + { + "epoch": 8.74, + "learning_rate": 2.127601921418894e-06, + "loss": 0.3589, + "step": 23660 + }, + { + "epoch": 8.74, + "learning_rate": 2.1275649710555487e-06, + "loss": 0.3914, + "step": 23661 + }, + { + "epoch": 8.74, + "learning_rate": 2.1275280206922033e-06, + "loss": 0.3481, + "step": 23662 + }, + { + "epoch": 8.74, + "learning_rate": 2.1274910703288584e-06, + "loss": 0.4142, + "step": 23663 + }, + { + "epoch": 8.74, + "learning_rate": 2.127454119965513e-06, + "loss": 0.2923, + "step": 23664 + }, + { + "epoch": 8.74, + "learning_rate": 2.127417169602168e-06, + "loss": 0.372, + "step": 23665 + }, + { + "epoch": 8.74, + "learning_rate": 2.1273802192388227e-06, + "loss": 0.4198, + "step": 23666 + }, + { + "epoch": 8.74, + "learning_rate": 2.1273432688754774e-06, + "loss": 0.3596, + "step": 23667 + }, + { + "epoch": 8.74, + "learning_rate": 2.127306318512132e-06, + "loss": 0.2666, + "step": 23668 + }, + { + "epoch": 8.74, + "learning_rate": 2.1272693681487866e-06, + "loss": 0.2573, + "step": 23669 + }, + { + "epoch": 8.74, + "learning_rate": 2.1272324177854417e-06, + "loss": 0.3351, + "step": 23670 + }, + { + "epoch": 8.74, + "learning_rate": 2.1271954674220963e-06, + "loss": 0.4097, + "step": 23671 + }, + { + "epoch": 8.74, + "learning_rate": 2.1271585170587514e-06, + "loss": 0.4331, + "step": 23672 + }, + { + "epoch": 8.74, + "learning_rate": 2.1271215666954056e-06, + "loss": 0.3997, + "step": 23673 + }, + { + "epoch": 8.74, + "learning_rate": 2.1270846163320607e-06, + "loss": 0.3887, + "step": 23674 + }, + { + "epoch": 8.74, + "learning_rate": 2.1270476659687153e-06, + "loss": 0.418, + "step": 23675 + }, + { + "epoch": 8.74, + "learning_rate": 2.1270107156053704e-06, + "loss": 0.3988, + "step": 23676 + }, + { + "epoch": 8.74, + "learning_rate": 2.126973765242025e-06, + "loss": 0.3299, + "step": 23677 + }, + { + "epoch": 8.74, + "learning_rate": 2.12693681487868e-06, + "loss": 0.3023, + "step": 23678 + }, + { + "epoch": 8.74, + "learning_rate": 2.1268998645153343e-06, + "loss": 0.3602, + "step": 23679 + }, + { + "epoch": 8.74, + "learning_rate": 2.126862914151989e-06, + "loss": 0.3478, + "step": 23680 + }, + { + "epoch": 8.74, + "learning_rate": 2.126825963788644e-06, + "loss": 0.3793, + "step": 23681 + }, + { + "epoch": 8.75, + "learning_rate": 2.1267890134252986e-06, + "loss": 0.4998, + "step": 23682 + }, + { + "epoch": 8.75, + "learning_rate": 2.1267520630619537e-06, + "loss": 0.5323, + "step": 23683 + }, + { + "epoch": 8.75, + "learning_rate": 2.1267151126986083e-06, + "loss": 0.4556, + "step": 23684 + }, + { + "epoch": 8.75, + "learning_rate": 2.126678162335263e-06, + "loss": 0.3451, + "step": 23685 + }, + { + "epoch": 8.75, + "learning_rate": 2.1266412119719176e-06, + "loss": 0.4307, + "step": 23686 + }, + { + "epoch": 8.75, + "learning_rate": 2.1266042616085726e-06, + "loss": 0.3244, + "step": 23687 + }, + { + "epoch": 8.75, + "learning_rate": 2.1265673112452273e-06, + "loss": 0.4472, + "step": 23688 + }, + { + "epoch": 8.75, + "learning_rate": 2.126530360881882e-06, + "loss": 0.3097, + "step": 23689 + }, + { + "epoch": 8.75, + "learning_rate": 2.126493410518537e-06, + "loss": 0.4304, + "step": 23690 + }, + { + "epoch": 8.75, + "learning_rate": 2.1264564601551916e-06, + "loss": 0.5396, + "step": 23691 + }, + { + "epoch": 8.75, + "learning_rate": 2.1264195097918462e-06, + "loss": 0.3637, + "step": 23692 + }, + { + "epoch": 8.75, + "learning_rate": 2.126382559428501e-06, + "loss": 0.3045, + "step": 23693 + }, + { + "epoch": 8.75, + "learning_rate": 2.126345609065156e-06, + "loss": 0.3589, + "step": 23694 + }, + { + "epoch": 8.75, + "learning_rate": 2.1263086587018106e-06, + "loss": 0.4041, + "step": 23695 + }, + { + "epoch": 8.75, + "learning_rate": 2.1262717083384656e-06, + "loss": 0.4357, + "step": 23696 + }, + { + "epoch": 8.75, + "learning_rate": 2.1262347579751203e-06, + "loss": 0.527, + "step": 23697 + }, + { + "epoch": 8.75, + "learning_rate": 2.126197807611775e-06, + "loss": 0.6205, + "step": 23698 + }, + { + "epoch": 8.75, + "learning_rate": 2.1261608572484295e-06, + "loss": 0.5759, + "step": 23699 + }, + { + "epoch": 8.75, + "learning_rate": 2.126123906885084e-06, + "loss": 0.3943, + "step": 23700 + }, + { + "epoch": 8.75, + "learning_rate": 2.1260869565217392e-06, + "loss": 0.388, + "step": 23701 + }, + { + "epoch": 8.75, + "learning_rate": 2.126050006158394e-06, + "loss": 0.4202, + "step": 23702 + }, + { + "epoch": 8.75, + "learning_rate": 2.126013055795049e-06, + "loss": 0.3695, + "step": 23703 + }, + { + "epoch": 8.75, + "learning_rate": 2.1259761054317036e-06, + "loss": 0.3835, + "step": 23704 + }, + { + "epoch": 8.75, + "learning_rate": 2.125939155068358e-06, + "loss": 0.3883, + "step": 23705 + }, + { + "epoch": 8.75, + "learning_rate": 2.125902204705013e-06, + "loss": 0.4289, + "step": 23706 + }, + { + "epoch": 8.75, + "learning_rate": 2.125865254341668e-06, + "loss": 0.333, + "step": 23707 + }, + { + "epoch": 8.75, + "learning_rate": 2.1258283039783225e-06, + "loss": 0.4483, + "step": 23708 + }, + { + "epoch": 8.76, + "learning_rate": 2.125791353614977e-06, + "loss": 0.317, + "step": 23709 + }, + { + "epoch": 8.76, + "learning_rate": 2.1257544032516322e-06, + "loss": 0.3049, + "step": 23710 + }, + { + "epoch": 8.76, + "learning_rate": 2.125717452888287e-06, + "loss": 0.3878, + "step": 23711 + }, + { + "epoch": 8.76, + "learning_rate": 2.1256805025249415e-06, + "loss": 0.2813, + "step": 23712 + }, + { + "epoch": 8.76, + "learning_rate": 2.125643552161596e-06, + "loss": 0.4783, + "step": 23713 + }, + { + "epoch": 8.76, + "learning_rate": 2.125606601798251e-06, + "loss": 0.3245, + "step": 23714 + }, + { + "epoch": 8.76, + "learning_rate": 2.125569651434906e-06, + "loss": 0.3828, + "step": 23715 + }, + { + "epoch": 8.76, + "learning_rate": 2.125532701071561e-06, + "loss": 0.3772, + "step": 23716 + }, + { + "epoch": 8.76, + "learning_rate": 2.1254957507082155e-06, + "loss": 0.4416, + "step": 23717 + }, + { + "epoch": 8.76, + "learning_rate": 2.1254588003448697e-06, + "loss": 0.4643, + "step": 23718 + }, + { + "epoch": 8.76, + "learning_rate": 2.125421849981525e-06, + "loss": 0.3574, + "step": 23719 + }, + { + "epoch": 8.76, + "learning_rate": 2.1253848996181794e-06, + "loss": 0.3216, + "step": 23720 + }, + { + "epoch": 8.76, + "learning_rate": 2.1253479492548345e-06, + "loss": 0.4136, + "step": 23721 + }, + { + "epoch": 8.76, + "learning_rate": 2.125310998891489e-06, + "loss": 0.3055, + "step": 23722 + }, + { + "epoch": 8.76, + "learning_rate": 2.125274048528144e-06, + "loss": 0.282, + "step": 23723 + }, + { + "epoch": 8.76, + "learning_rate": 2.1252370981647984e-06, + "loss": 0.4006, + "step": 23724 + }, + { + "epoch": 8.76, + "learning_rate": 2.1252001478014535e-06, + "loss": 0.5019, + "step": 23725 + }, + { + "epoch": 8.76, + "learning_rate": 2.125163197438108e-06, + "loss": 0.4606, + "step": 23726 + }, + { + "epoch": 8.76, + "learning_rate": 2.125126247074763e-06, + "loss": 0.2778, + "step": 23727 + }, + { + "epoch": 8.76, + "learning_rate": 2.125089296711418e-06, + "loss": 0.4548, + "step": 23728 + }, + { + "epoch": 8.76, + "learning_rate": 2.1250523463480724e-06, + "loss": 0.3779, + "step": 23729 + }, + { + "epoch": 8.76, + "learning_rate": 2.125015395984727e-06, + "loss": 0.2775, + "step": 23730 + }, + { + "epoch": 8.76, + "learning_rate": 2.1249784456213817e-06, + "loss": 0.3406, + "step": 23731 + }, + { + "epoch": 8.76, + "learning_rate": 2.1249414952580368e-06, + "loss": 0.4358, + "step": 23732 + }, + { + "epoch": 8.76, + "learning_rate": 2.1249045448946914e-06, + "loss": 0.4214, + "step": 23733 + }, + { + "epoch": 8.76, + "learning_rate": 2.1248675945313465e-06, + "loss": 0.4786, + "step": 23734 + }, + { + "epoch": 8.76, + "learning_rate": 2.124830644168001e-06, + "loss": 0.3844, + "step": 23735 + }, + { + "epoch": 8.77, + "learning_rate": 2.1247936938046557e-06, + "loss": 0.3152, + "step": 23736 + }, + { + "epoch": 8.77, + "learning_rate": 2.1247567434413104e-06, + "loss": 0.3781, + "step": 23737 + }, + { + "epoch": 8.77, + "learning_rate": 2.124719793077965e-06, + "loss": 0.3417, + "step": 23738 + }, + { + "epoch": 8.77, + "learning_rate": 2.12468284271462e-06, + "loss": 0.5371, + "step": 23739 + }, + { + "epoch": 8.77, + "learning_rate": 2.1246458923512747e-06, + "loss": 0.3674, + "step": 23740 + }, + { + "epoch": 8.77, + "learning_rate": 2.1246089419879298e-06, + "loss": 0.4008, + "step": 23741 + }, + { + "epoch": 8.77, + "learning_rate": 2.1245719916245844e-06, + "loss": 0.4131, + "step": 23742 + }, + { + "epoch": 8.77, + "learning_rate": 2.124535041261239e-06, + "loss": 0.3506, + "step": 23743 + }, + { + "epoch": 8.77, + "learning_rate": 2.1244980908978937e-06, + "loss": 0.4209, + "step": 23744 + }, + { + "epoch": 8.77, + "learning_rate": 2.1244611405345487e-06, + "loss": 0.3733, + "step": 23745 + }, + { + "epoch": 8.77, + "learning_rate": 2.1244241901712034e-06, + "loss": 0.2947, + "step": 23746 + }, + { + "epoch": 8.77, + "learning_rate": 2.1243872398078584e-06, + "loss": 0.3714, + "step": 23747 + }, + { + "epoch": 8.77, + "learning_rate": 2.124350289444513e-06, + "loss": 0.4111, + "step": 23748 + }, + { + "epoch": 8.77, + "learning_rate": 2.1243133390811677e-06, + "loss": 0.4332, + "step": 23749 + }, + { + "epoch": 8.77, + "learning_rate": 2.1242763887178223e-06, + "loss": 0.3816, + "step": 23750 + }, + { + "epoch": 8.77, + "learning_rate": 2.124239438354477e-06, + "loss": 0.3875, + "step": 23751 + }, + { + "epoch": 8.77, + "learning_rate": 2.124202487991132e-06, + "loss": 0.3146, + "step": 23752 + }, + { + "epoch": 8.77, + "learning_rate": 2.1241655376277867e-06, + "loss": 0.3391, + "step": 23753 + }, + { + "epoch": 8.77, + "learning_rate": 2.1241285872644417e-06, + "loss": 0.2757, + "step": 23754 + }, + { + "epoch": 8.77, + "learning_rate": 2.1240916369010964e-06, + "loss": 0.4198, + "step": 23755 + }, + { + "epoch": 8.77, + "learning_rate": 2.124054686537751e-06, + "loss": 0.478, + "step": 23756 + }, + { + "epoch": 8.77, + "learning_rate": 2.1240177361744056e-06, + "loss": 0.3647, + "step": 23757 + }, + { + "epoch": 8.77, + "learning_rate": 2.1239807858110603e-06, + "loss": 0.3696, + "step": 23758 + }, + { + "epoch": 8.77, + "learning_rate": 2.1239438354477153e-06, + "loss": 0.4089, + "step": 23759 + }, + { + "epoch": 8.77, + "learning_rate": 2.12390688508437e-06, + "loss": 0.367, + "step": 23760 + }, + { + "epoch": 8.77, + "learning_rate": 2.123869934721025e-06, + "loss": 0.308, + "step": 23761 + }, + { + "epoch": 8.77, + "learning_rate": 2.1238329843576797e-06, + "loss": 0.3244, + "step": 23762 + }, + { + "epoch": 8.78, + "learning_rate": 2.1237960339943343e-06, + "loss": 0.2723, + "step": 23763 + }, + { + "epoch": 8.78, + "learning_rate": 2.123759083630989e-06, + "loss": 0.5206, + "step": 23764 + }, + { + "epoch": 8.78, + "learning_rate": 2.123722133267644e-06, + "loss": 0.33, + "step": 23765 + }, + { + "epoch": 8.78, + "learning_rate": 2.1236851829042986e-06, + "loss": 0.3595, + "step": 23766 + }, + { + "epoch": 8.78, + "learning_rate": 2.1236482325409537e-06, + "loss": 0.4127, + "step": 23767 + }, + { + "epoch": 8.78, + "learning_rate": 2.1236112821776083e-06, + "loss": 0.3107, + "step": 23768 + }, + { + "epoch": 8.78, + "learning_rate": 2.1235743318142626e-06, + "loss": 0.3432, + "step": 23769 + }, + { + "epoch": 8.78, + "learning_rate": 2.1235373814509176e-06, + "loss": 0.3867, + "step": 23770 + }, + { + "epoch": 8.78, + "learning_rate": 2.1235004310875723e-06, + "loss": 0.3322, + "step": 23771 + }, + { + "epoch": 8.78, + "learning_rate": 2.1234634807242273e-06, + "loss": 0.4993, + "step": 23772 + }, + { + "epoch": 8.78, + "learning_rate": 2.123426530360882e-06, + "loss": 0.3607, + "step": 23773 + }, + { + "epoch": 8.78, + "learning_rate": 2.123389579997537e-06, + "loss": 0.4247, + "step": 23774 + }, + { + "epoch": 8.78, + "learning_rate": 2.1233526296341912e-06, + "loss": 0.3588, + "step": 23775 + }, + { + "epoch": 8.78, + "learning_rate": 2.1233156792708463e-06, + "loss": 0.441, + "step": 23776 + }, + { + "epoch": 8.78, + "learning_rate": 2.123278728907501e-06, + "loss": 0.3957, + "step": 23777 + }, + { + "epoch": 8.78, + "learning_rate": 2.1232417785441556e-06, + "loss": 0.5195, + "step": 23778 + }, + { + "epoch": 8.78, + "learning_rate": 2.1232048281808106e-06, + "loss": 0.3128, + "step": 23779 + }, + { + "epoch": 8.78, + "learning_rate": 2.1231678778174653e-06, + "loss": 0.3869, + "step": 23780 + }, + { + "epoch": 8.78, + "learning_rate": 2.12313092745412e-06, + "loss": 0.2941, + "step": 23781 + }, + { + "epoch": 8.78, + "learning_rate": 2.1230939770907745e-06, + "loss": 0.4594, + "step": 23782 + }, + { + "epoch": 8.78, + "learning_rate": 2.1230570267274296e-06, + "loss": 0.2436, + "step": 23783 + }, + { + "epoch": 8.78, + "learning_rate": 2.1230200763640842e-06, + "loss": 0.4505, + "step": 23784 + }, + { + "epoch": 8.78, + "learning_rate": 2.1229831260007393e-06, + "loss": 0.3783, + "step": 23785 + }, + { + "epoch": 8.78, + "learning_rate": 2.122946175637394e-06, + "loss": 0.4908, + "step": 23786 + }, + { + "epoch": 8.78, + "learning_rate": 2.1229092252740486e-06, + "loss": 0.4589, + "step": 23787 + }, + { + "epoch": 8.78, + "learning_rate": 2.122872274910703e-06, + "loss": 0.2603, + "step": 23788 + }, + { + "epoch": 8.78, + "learning_rate": 2.122835324547358e-06, + "loss": 0.4225, + "step": 23789 + }, + { + "epoch": 8.79, + "learning_rate": 2.122798374184013e-06, + "loss": 0.4888, + "step": 23790 + }, + { + "epoch": 8.79, + "learning_rate": 2.1227614238206675e-06, + "loss": 0.4024, + "step": 23791 + }, + { + "epoch": 8.79, + "learning_rate": 2.1227244734573226e-06, + "loss": 0.3631, + "step": 23792 + }, + { + "epoch": 8.79, + "learning_rate": 2.1226875230939772e-06, + "loss": 0.5593, + "step": 23793 + }, + { + "epoch": 8.79, + "learning_rate": 2.122650572730632e-06, + "loss": 0.3268, + "step": 23794 + }, + { + "epoch": 8.79, + "learning_rate": 2.1226136223672865e-06, + "loss": 0.3865, + "step": 23795 + }, + { + "epoch": 8.79, + "learning_rate": 2.1225766720039416e-06, + "loss": 0.4463, + "step": 23796 + }, + { + "epoch": 8.79, + "learning_rate": 2.122539721640596e-06, + "loss": 0.3917, + "step": 23797 + }, + { + "epoch": 8.79, + "learning_rate": 2.122502771277251e-06, + "loss": 0.2324, + "step": 23798 + }, + { + "epoch": 8.79, + "learning_rate": 2.122465820913906e-06, + "loss": 0.4723, + "step": 23799 + }, + { + "epoch": 8.79, + "learning_rate": 2.1224288705505605e-06, + "loss": 0.4424, + "step": 23800 + }, + { + "epoch": 8.79, + "learning_rate": 2.122391920187215e-06, + "loss": 0.2961, + "step": 23801 + }, + { + "epoch": 8.79, + "learning_rate": 2.12235496982387e-06, + "loss": 0.4191, + "step": 23802 + }, + { + "epoch": 8.79, + "learning_rate": 2.122318019460525e-06, + "loss": 0.3736, + "step": 23803 + }, + { + "epoch": 8.79, + "learning_rate": 2.1222810690971795e-06, + "loss": 0.4769, + "step": 23804 + }, + { + "epoch": 8.79, + "learning_rate": 2.1222441187338346e-06, + "loss": 0.3135, + "step": 23805 + }, + { + "epoch": 8.79, + "learning_rate": 2.122207168370489e-06, + "loss": 0.3142, + "step": 23806 + }, + { + "epoch": 8.79, + "learning_rate": 2.122170218007144e-06, + "loss": 0.3348, + "step": 23807 + }, + { + "epoch": 8.79, + "learning_rate": 2.1221332676437985e-06, + "loss": 0.3576, + "step": 23808 + }, + { + "epoch": 8.79, + "learning_rate": 2.122096317280453e-06, + "loss": 0.3607, + "step": 23809 + }, + { + "epoch": 8.79, + "learning_rate": 2.122059366917108e-06, + "loss": 0.3597, + "step": 23810 + }, + { + "epoch": 8.79, + "learning_rate": 2.122022416553763e-06, + "loss": 0.5268, + "step": 23811 + }, + { + "epoch": 8.79, + "learning_rate": 2.121985466190418e-06, + "loss": 0.4286, + "step": 23812 + }, + { + "epoch": 8.79, + "learning_rate": 2.1219485158270725e-06, + "loss": 0.3676, + "step": 23813 + }, + { + "epoch": 8.79, + "learning_rate": 2.121911565463727e-06, + "loss": 0.3786, + "step": 23814 + }, + { + "epoch": 8.79, + "learning_rate": 2.1218746151003818e-06, + "loss": 0.342, + "step": 23815 + }, + { + "epoch": 8.79, + "learning_rate": 2.121837664737037e-06, + "loss": 0.3373, + "step": 23816 + }, + { + "epoch": 8.8, + "learning_rate": 2.1218007143736915e-06, + "loss": 0.4522, + "step": 23817 + }, + { + "epoch": 8.8, + "learning_rate": 2.121763764010346e-06, + "loss": 0.3888, + "step": 23818 + }, + { + "epoch": 8.8, + "learning_rate": 2.1217268136470007e-06, + "loss": 0.3805, + "step": 23819 + }, + { + "epoch": 8.8, + "learning_rate": 2.1216898632836554e-06, + "loss": 0.3902, + "step": 23820 + }, + { + "epoch": 8.8, + "learning_rate": 2.1216529129203104e-06, + "loss": 0.4911, + "step": 23821 + }, + { + "epoch": 8.8, + "learning_rate": 2.121615962556965e-06, + "loss": 0.309, + "step": 23822 + }, + { + "epoch": 8.8, + "learning_rate": 2.12157901219362e-06, + "loss": 0.3907, + "step": 23823 + }, + { + "epoch": 8.8, + "learning_rate": 2.1215420618302748e-06, + "loss": 0.4589, + "step": 23824 + }, + { + "epoch": 8.8, + "learning_rate": 2.1215051114669294e-06, + "loss": 0.5463, + "step": 23825 + }, + { + "epoch": 8.8, + "learning_rate": 2.121468161103584e-06, + "loss": 0.2718, + "step": 23826 + }, + { + "epoch": 8.8, + "learning_rate": 2.121431210740239e-06, + "loss": 0.3606, + "step": 23827 + }, + { + "epoch": 8.8, + "learning_rate": 2.1213942603768937e-06, + "loss": 0.3275, + "step": 23828 + }, + { + "epoch": 8.8, + "learning_rate": 2.1213573100135484e-06, + "loss": 0.4696, + "step": 23829 + }, + { + "epoch": 8.8, + "learning_rate": 2.1213203596502034e-06, + "loss": 0.4724, + "step": 23830 + }, + { + "epoch": 8.8, + "learning_rate": 2.121283409286858e-06, + "loss": 0.4702, + "step": 23831 + }, + { + "epoch": 8.8, + "learning_rate": 2.1212464589235127e-06, + "loss": 0.4228, + "step": 23832 + }, + { + "epoch": 8.8, + "learning_rate": 2.1212095085601673e-06, + "loss": 0.2827, + "step": 23833 + }, + { + "epoch": 8.8, + "learning_rate": 2.1211725581968224e-06, + "loss": 0.3614, + "step": 23834 + }, + { + "epoch": 8.8, + "learning_rate": 2.121135607833477e-06, + "loss": 0.4688, + "step": 23835 + }, + { + "epoch": 8.8, + "learning_rate": 2.121098657470132e-06, + "loss": 0.3657, + "step": 23836 + }, + { + "epoch": 8.8, + "learning_rate": 2.1210617071067867e-06, + "loss": 0.4609, + "step": 23837 + }, + { + "epoch": 8.8, + "learning_rate": 2.1210247567434414e-06, + "loss": 0.3842, + "step": 23838 + }, + { + "epoch": 8.8, + "learning_rate": 2.120987806380096e-06, + "loss": 0.4148, + "step": 23839 + }, + { + "epoch": 8.8, + "learning_rate": 2.1209508560167506e-06, + "loss": 0.5143, + "step": 23840 + }, + { + "epoch": 8.8, + "learning_rate": 2.1209139056534057e-06, + "loss": 0.4251, + "step": 23841 + }, + { + "epoch": 8.8, + "learning_rate": 2.1208769552900603e-06, + "loss": 0.4198, + "step": 23842 + }, + { + "epoch": 8.8, + "learning_rate": 2.1208400049267154e-06, + "loss": 0.3559, + "step": 23843 + }, + { + "epoch": 8.81, + "learning_rate": 2.12080305456337e-06, + "loss": 0.3971, + "step": 23844 + }, + { + "epoch": 8.81, + "learning_rate": 2.1207661042000247e-06, + "loss": 0.4173, + "step": 23845 + }, + { + "epoch": 8.81, + "learning_rate": 2.1207291538366793e-06, + "loss": 0.3428, + "step": 23846 + }, + { + "epoch": 8.81, + "learning_rate": 2.1206922034733344e-06, + "loss": 0.3795, + "step": 23847 + }, + { + "epoch": 8.81, + "learning_rate": 2.120655253109989e-06, + "loss": 0.3834, + "step": 23848 + }, + { + "epoch": 8.81, + "learning_rate": 2.1206183027466436e-06, + "loss": 0.3682, + "step": 23849 + }, + { + "epoch": 8.81, + "learning_rate": 2.1205813523832987e-06, + "loss": 0.3758, + "step": 23850 + }, + { + "epoch": 8.81, + "learning_rate": 2.1205444020199533e-06, + "loss": 0.5002, + "step": 23851 + }, + { + "epoch": 8.81, + "learning_rate": 2.120507451656608e-06, + "loss": 0.4052, + "step": 23852 + }, + { + "epoch": 8.81, + "learning_rate": 2.1204705012932626e-06, + "loss": 0.5211, + "step": 23853 + }, + { + "epoch": 8.81, + "learning_rate": 2.1204335509299177e-06, + "loss": 0.3594, + "step": 23854 + }, + { + "epoch": 8.81, + "learning_rate": 2.1203966005665723e-06, + "loss": 0.5376, + "step": 23855 + }, + { + "epoch": 8.81, + "learning_rate": 2.1203596502032274e-06, + "loss": 0.4058, + "step": 23856 + }, + { + "epoch": 8.81, + "learning_rate": 2.120322699839882e-06, + "loss": 0.3723, + "step": 23857 + }, + { + "epoch": 8.81, + "learning_rate": 2.120285749476536e-06, + "loss": 0.38, + "step": 23858 + }, + { + "epoch": 8.81, + "learning_rate": 2.1202487991131913e-06, + "loss": 0.427, + "step": 23859 + }, + { + "epoch": 8.81, + "learning_rate": 2.120211848749846e-06, + "loss": 0.3328, + "step": 23860 + }, + { + "epoch": 8.81, + "learning_rate": 2.120174898386501e-06, + "loss": 0.3232, + "step": 23861 + }, + { + "epoch": 8.81, + "learning_rate": 2.1201379480231556e-06, + "loss": 0.5054, + "step": 23862 + }, + { + "epoch": 8.81, + "learning_rate": 2.1201009976598107e-06, + "loss": 0.4798, + "step": 23863 + }, + { + "epoch": 8.81, + "learning_rate": 2.120064047296465e-06, + "loss": 0.4471, + "step": 23864 + }, + { + "epoch": 8.81, + "learning_rate": 2.12002709693312e-06, + "loss": 0.5649, + "step": 23865 + }, + { + "epoch": 8.81, + "learning_rate": 2.1199901465697746e-06, + "loss": 0.3812, + "step": 23866 + }, + { + "epoch": 8.81, + "learning_rate": 2.1199531962064296e-06, + "loss": 0.4023, + "step": 23867 + }, + { + "epoch": 8.81, + "learning_rate": 2.1199162458430843e-06, + "loss": 0.4719, + "step": 23868 + }, + { + "epoch": 8.81, + "learning_rate": 2.119879295479739e-06, + "loss": 0.364, + "step": 23869 + }, + { + "epoch": 8.81, + "learning_rate": 2.1198423451163935e-06, + "loss": 0.3302, + "step": 23870 + }, + { + "epoch": 8.81, + "learning_rate": 2.119805394753048e-06, + "loss": 0.4458, + "step": 23871 + }, + { + "epoch": 8.82, + "learning_rate": 2.1197684443897032e-06, + "loss": 0.4139, + "step": 23872 + }, + { + "epoch": 8.82, + "learning_rate": 2.119731494026358e-06, + "loss": 0.4446, + "step": 23873 + }, + { + "epoch": 8.82, + "learning_rate": 2.119694543663013e-06, + "loss": 0.2906, + "step": 23874 + }, + { + "epoch": 8.82, + "learning_rate": 2.1196575932996676e-06, + "loss": 0.3525, + "step": 23875 + }, + { + "epoch": 8.82, + "learning_rate": 2.119620642936322e-06, + "loss": 0.3199, + "step": 23876 + }, + { + "epoch": 8.82, + "learning_rate": 2.119583692572977e-06, + "loss": 0.2906, + "step": 23877 + }, + { + "epoch": 8.82, + "learning_rate": 2.1195467422096315e-06, + "loss": 0.2562, + "step": 23878 + }, + { + "epoch": 8.82, + "learning_rate": 2.1195097918462865e-06, + "loss": 0.502, + "step": 23879 + }, + { + "epoch": 8.82, + "learning_rate": 2.119472841482941e-06, + "loss": 0.4567, + "step": 23880 + }, + { + "epoch": 8.82, + "learning_rate": 2.1194358911195962e-06, + "loss": 0.396, + "step": 23881 + }, + { + "epoch": 8.82, + "learning_rate": 2.119398940756251e-06, + "loss": 0.3522, + "step": 23882 + }, + { + "epoch": 8.82, + "learning_rate": 2.1193619903929055e-06, + "loss": 0.2404, + "step": 23883 + }, + { + "epoch": 8.82, + "learning_rate": 2.11932504002956e-06, + "loss": 0.4609, + "step": 23884 + }, + { + "epoch": 8.82, + "learning_rate": 2.119288089666215e-06, + "loss": 0.394, + "step": 23885 + }, + { + "epoch": 8.82, + "learning_rate": 2.11925113930287e-06, + "loss": 0.2862, + "step": 23886 + }, + { + "epoch": 8.82, + "learning_rate": 2.119214188939525e-06, + "loss": 0.3896, + "step": 23887 + }, + { + "epoch": 8.82, + "learning_rate": 2.1191772385761795e-06, + "loss": 0.4161, + "step": 23888 + }, + { + "epoch": 8.82, + "learning_rate": 2.119140288212834e-06, + "loss": 0.3468, + "step": 23889 + }, + { + "epoch": 8.82, + "learning_rate": 2.119103337849489e-06, + "loss": 0.367, + "step": 23890 + }, + { + "epoch": 8.82, + "learning_rate": 2.1190663874861434e-06, + "loss": 0.5151, + "step": 23891 + }, + { + "epoch": 8.82, + "learning_rate": 2.1190294371227985e-06, + "loss": 0.4356, + "step": 23892 + }, + { + "epoch": 8.82, + "learning_rate": 2.118992486759453e-06, + "loss": 0.453, + "step": 23893 + }, + { + "epoch": 8.82, + "learning_rate": 2.118955536396108e-06, + "loss": 0.2202, + "step": 23894 + }, + { + "epoch": 8.82, + "learning_rate": 2.118918586032763e-06, + "loss": 0.5259, + "step": 23895 + }, + { + "epoch": 8.82, + "learning_rate": 2.1188816356694175e-06, + "loss": 0.4277, + "step": 23896 + }, + { + "epoch": 8.82, + "learning_rate": 2.118844685306072e-06, + "loss": 0.3825, + "step": 23897 + }, + { + "epoch": 8.82, + "learning_rate": 2.1188077349427268e-06, + "loss": 0.3429, + "step": 23898 + }, + { + "epoch": 8.83, + "learning_rate": 2.118770784579382e-06, + "loss": 0.3922, + "step": 23899 + }, + { + "epoch": 8.83, + "learning_rate": 2.1187338342160364e-06, + "loss": 0.3784, + "step": 23900 + }, + { + "epoch": 8.83, + "learning_rate": 2.1186968838526915e-06, + "loss": 0.4361, + "step": 23901 + }, + { + "epoch": 8.83, + "learning_rate": 2.118659933489346e-06, + "loss": 0.3268, + "step": 23902 + }, + { + "epoch": 8.83, + "learning_rate": 2.1186229831260008e-06, + "loss": 0.3269, + "step": 23903 + }, + { + "epoch": 8.83, + "learning_rate": 2.1185860327626554e-06, + "loss": 0.3737, + "step": 23904 + }, + { + "epoch": 8.83, + "learning_rate": 2.1185490823993105e-06, + "loss": 0.3978, + "step": 23905 + }, + { + "epoch": 8.83, + "learning_rate": 2.118512132035965e-06, + "loss": 0.3437, + "step": 23906 + }, + { + "epoch": 8.83, + "learning_rate": 2.11847518167262e-06, + "loss": 0.2785, + "step": 23907 + }, + { + "epoch": 8.83, + "learning_rate": 2.118438231309275e-06, + "loss": 0.2887, + "step": 23908 + }, + { + "epoch": 8.83, + "learning_rate": 2.118401280945929e-06, + "loss": 0.3208, + "step": 23909 + }, + { + "epoch": 8.83, + "learning_rate": 2.118364330582584e-06, + "loss": 0.3404, + "step": 23910 + }, + { + "epoch": 8.83, + "learning_rate": 2.1183273802192387e-06, + "loss": 0.3342, + "step": 23911 + }, + { + "epoch": 8.83, + "learning_rate": 2.1182904298558938e-06, + "loss": 0.3938, + "step": 23912 + }, + { + "epoch": 8.83, + "learning_rate": 2.1182534794925484e-06, + "loss": 0.3123, + "step": 23913 + }, + { + "epoch": 8.83, + "learning_rate": 2.1182165291292035e-06, + "loss": 0.2826, + "step": 23914 + }, + { + "epoch": 8.83, + "learning_rate": 2.1181795787658577e-06, + "loss": 0.327, + "step": 23915 + }, + { + "epoch": 8.83, + "learning_rate": 2.1181426284025127e-06, + "loss": 0.5389, + "step": 23916 + }, + { + "epoch": 8.83, + "learning_rate": 2.1181056780391674e-06, + "loss": 0.2657, + "step": 23917 + }, + { + "epoch": 8.83, + "learning_rate": 2.118068727675822e-06, + "loss": 0.5411, + "step": 23918 + }, + { + "epoch": 8.83, + "learning_rate": 2.118031777312477e-06, + "loss": 0.4255, + "step": 23919 + }, + { + "epoch": 8.83, + "learning_rate": 2.1179948269491317e-06, + "loss": 0.4384, + "step": 23920 + }, + { + "epoch": 8.83, + "learning_rate": 2.1179578765857864e-06, + "loss": 0.3456, + "step": 23921 + }, + { + "epoch": 8.83, + "learning_rate": 2.117920926222441e-06, + "loss": 0.4997, + "step": 23922 + }, + { + "epoch": 8.83, + "learning_rate": 2.117883975859096e-06, + "loss": 0.3447, + "step": 23923 + }, + { + "epoch": 8.83, + "learning_rate": 2.1178470254957507e-06, + "loss": 0.3423, + "step": 23924 + }, + { + "epoch": 8.83, + "learning_rate": 2.1178100751324057e-06, + "loss": 0.4427, + "step": 23925 + }, + { + "epoch": 8.84, + "learning_rate": 2.1177731247690604e-06, + "loss": 0.3397, + "step": 23926 + }, + { + "epoch": 8.84, + "learning_rate": 2.117736174405715e-06, + "loss": 0.3403, + "step": 23927 + }, + { + "epoch": 8.84, + "learning_rate": 2.1176992240423697e-06, + "loss": 0.4003, + "step": 23928 + }, + { + "epoch": 8.84, + "learning_rate": 2.1176622736790243e-06, + "loss": 0.4755, + "step": 23929 + }, + { + "epoch": 8.84, + "learning_rate": 2.1176253233156794e-06, + "loss": 0.3831, + "step": 23930 + }, + { + "epoch": 8.84, + "learning_rate": 2.117588372952334e-06, + "loss": 0.3589, + "step": 23931 + }, + { + "epoch": 8.84, + "learning_rate": 2.117551422588989e-06, + "loss": 0.3924, + "step": 23932 + }, + { + "epoch": 8.84, + "learning_rate": 2.1175144722256437e-06, + "loss": 0.3007, + "step": 23933 + }, + { + "epoch": 8.84, + "learning_rate": 2.1174775218622983e-06, + "loss": 0.416, + "step": 23934 + }, + { + "epoch": 8.84, + "learning_rate": 2.117440571498953e-06, + "loss": 0.3473, + "step": 23935 + }, + { + "epoch": 8.84, + "learning_rate": 2.117403621135608e-06, + "loss": 0.4039, + "step": 23936 + }, + { + "epoch": 8.84, + "learning_rate": 2.1173666707722627e-06, + "loss": 0.3087, + "step": 23937 + }, + { + "epoch": 8.84, + "learning_rate": 2.1173297204089173e-06, + "loss": 0.3298, + "step": 23938 + }, + { + "epoch": 8.84, + "learning_rate": 2.1172927700455723e-06, + "loss": 0.4171, + "step": 23939 + }, + { + "epoch": 8.84, + "learning_rate": 2.117255819682227e-06, + "loss": 0.4105, + "step": 23940 + }, + { + "epoch": 8.84, + "learning_rate": 2.1172188693188816e-06, + "loss": 0.3439, + "step": 23941 + }, + { + "epoch": 8.84, + "learning_rate": 2.1171819189555363e-06, + "loss": 0.4138, + "step": 23942 + }, + { + "epoch": 8.84, + "learning_rate": 2.1171449685921913e-06, + "loss": 0.3784, + "step": 23943 + }, + { + "epoch": 8.84, + "learning_rate": 2.117108018228846e-06, + "loss": 0.443, + "step": 23944 + }, + { + "epoch": 8.84, + "learning_rate": 2.117071067865501e-06, + "loss": 0.4015, + "step": 23945 + }, + { + "epoch": 8.84, + "learning_rate": 2.1170341175021557e-06, + "loss": 0.347, + "step": 23946 + }, + { + "epoch": 8.84, + "learning_rate": 2.1169971671388103e-06, + "loss": 0.3664, + "step": 23947 + }, + { + "epoch": 8.84, + "learning_rate": 2.116960216775465e-06, + "loss": 0.2757, + "step": 23948 + }, + { + "epoch": 8.84, + "learning_rate": 2.1169232664121196e-06, + "loss": 0.3531, + "step": 23949 + }, + { + "epoch": 8.84, + "learning_rate": 2.1168863160487746e-06, + "loss": 0.3701, + "step": 23950 + }, + { + "epoch": 8.84, + "learning_rate": 2.1168493656854293e-06, + "loss": 0.3556, + "step": 23951 + }, + { + "epoch": 8.84, + "learning_rate": 2.1168124153220843e-06, + "loss": 0.3456, + "step": 23952 + }, + { + "epoch": 8.85, + "learning_rate": 2.116775464958739e-06, + "loss": 0.3437, + "step": 23953 + }, + { + "epoch": 8.85, + "learning_rate": 2.1167385145953936e-06, + "loss": 0.3869, + "step": 23954 + }, + { + "epoch": 8.85, + "learning_rate": 2.1167015642320482e-06, + "loss": 0.3322, + "step": 23955 + }, + { + "epoch": 8.85, + "learning_rate": 2.1166646138687033e-06, + "loss": 0.2722, + "step": 23956 + }, + { + "epoch": 8.85, + "learning_rate": 2.116627663505358e-06, + "loss": 0.4635, + "step": 23957 + }, + { + "epoch": 8.85, + "learning_rate": 2.1165907131420126e-06, + "loss": 0.5381, + "step": 23958 + }, + { + "epoch": 8.85, + "learning_rate": 2.1165537627786676e-06, + "loss": 0.2635, + "step": 23959 + }, + { + "epoch": 8.85, + "learning_rate": 2.116516812415322e-06, + "loss": 0.4072, + "step": 23960 + }, + { + "epoch": 8.85, + "learning_rate": 2.116479862051977e-06, + "loss": 0.3915, + "step": 23961 + }, + { + "epoch": 8.85, + "learning_rate": 2.1164429116886315e-06, + "loss": 0.3469, + "step": 23962 + }, + { + "epoch": 8.85, + "learning_rate": 2.1164059613252866e-06, + "loss": 0.3967, + "step": 23963 + }, + { + "epoch": 8.85, + "learning_rate": 2.1163690109619412e-06, + "loss": 0.4954, + "step": 23964 + }, + { + "epoch": 8.85, + "learning_rate": 2.1163320605985963e-06, + "loss": 0.411, + "step": 23965 + }, + { + "epoch": 8.85, + "learning_rate": 2.1162951102352505e-06, + "loss": 0.343, + "step": 23966 + }, + { + "epoch": 8.85, + "learning_rate": 2.116258159871905e-06, + "loss": 0.2943, + "step": 23967 + }, + { + "epoch": 8.85, + "learning_rate": 2.11622120950856e-06, + "loss": 0.4042, + "step": 23968 + }, + { + "epoch": 8.85, + "learning_rate": 2.116184259145215e-06, + "loss": 0.4486, + "step": 23969 + }, + { + "epoch": 8.85, + "learning_rate": 2.11614730878187e-06, + "loss": 0.4744, + "step": 23970 + }, + { + "epoch": 8.85, + "learning_rate": 2.1161103584185245e-06, + "loss": 0.2556, + "step": 23971 + }, + { + "epoch": 8.85, + "learning_rate": 2.116073408055179e-06, + "loss": 0.3369, + "step": 23972 + }, + { + "epoch": 8.85, + "learning_rate": 2.116036457691834e-06, + "loss": 0.5277, + "step": 23973 + }, + { + "epoch": 8.85, + "learning_rate": 2.115999507328489e-06, + "loss": 0.476, + "step": 23974 + }, + { + "epoch": 8.85, + "learning_rate": 2.1159625569651435e-06, + "loss": 0.5035, + "step": 23975 + }, + { + "epoch": 8.85, + "learning_rate": 2.1159256066017986e-06, + "loss": 0.302, + "step": 23976 + }, + { + "epoch": 8.85, + "learning_rate": 2.115888656238453e-06, + "loss": 0.43, + "step": 23977 + }, + { + "epoch": 8.85, + "learning_rate": 2.115851705875108e-06, + "loss": 0.2837, + "step": 23978 + }, + { + "epoch": 8.85, + "learning_rate": 2.1158147555117625e-06, + "loss": 0.3208, + "step": 23979 + }, + { + "epoch": 8.86, + "learning_rate": 2.115777805148417e-06, + "loss": 0.4455, + "step": 23980 + }, + { + "epoch": 8.86, + "learning_rate": 2.115740854785072e-06, + "loss": 0.3599, + "step": 23981 + }, + { + "epoch": 8.86, + "learning_rate": 2.115703904421727e-06, + "loss": 0.3455, + "step": 23982 + }, + { + "epoch": 8.86, + "learning_rate": 2.115666954058382e-06, + "loss": 0.3643, + "step": 23983 + }, + { + "epoch": 8.86, + "learning_rate": 2.1156300036950365e-06, + "loss": 0.4228, + "step": 23984 + }, + { + "epoch": 8.86, + "learning_rate": 2.115593053331691e-06, + "loss": 0.3812, + "step": 23985 + }, + { + "epoch": 8.86, + "learning_rate": 2.1155561029683458e-06, + "loss": 0.3556, + "step": 23986 + }, + { + "epoch": 8.86, + "learning_rate": 2.1155191526050004e-06, + "loss": 0.3574, + "step": 23987 + }, + { + "epoch": 8.86, + "learning_rate": 2.1154822022416555e-06, + "loss": 0.4514, + "step": 23988 + }, + { + "epoch": 8.86, + "learning_rate": 2.11544525187831e-06, + "loss": 0.3714, + "step": 23989 + }, + { + "epoch": 8.86, + "learning_rate": 2.115408301514965e-06, + "loss": 0.4893, + "step": 23990 + }, + { + "epoch": 8.86, + "learning_rate": 2.11537135115162e-06, + "loss": 0.4191, + "step": 23991 + }, + { + "epoch": 8.86, + "learning_rate": 2.1153344007882744e-06, + "loss": 0.4251, + "step": 23992 + }, + { + "epoch": 8.86, + "learning_rate": 2.115297450424929e-06, + "loss": 0.4383, + "step": 23993 + }, + { + "epoch": 8.86, + "learning_rate": 2.115260500061584e-06, + "loss": 0.4415, + "step": 23994 + }, + { + "epoch": 8.86, + "learning_rate": 2.1152235496982388e-06, + "loss": 0.2936, + "step": 23995 + }, + { + "epoch": 8.86, + "learning_rate": 2.115186599334894e-06, + "loss": 0.358, + "step": 23996 + }, + { + "epoch": 8.86, + "learning_rate": 2.1151496489715485e-06, + "loss": 0.485, + "step": 23997 + }, + { + "epoch": 8.86, + "learning_rate": 2.115112698608203e-06, + "loss": 0.4102, + "step": 23998 + }, + { + "epoch": 8.86, + "learning_rate": 2.1150757482448577e-06, + "loss": 0.3524, + "step": 23999 + }, + { + "epoch": 8.86, + "learning_rate": 2.1150387978815124e-06, + "loss": 0.3603, + "step": 24000 + }, + { + "epoch": 8.86, + "learning_rate": 2.1150018475181674e-06, + "loss": 0.3008, + "step": 24001 + }, + { + "epoch": 8.86, + "learning_rate": 2.114964897154822e-06, + "loss": 0.4446, + "step": 24002 + }, + { + "epoch": 8.86, + "learning_rate": 2.114927946791477e-06, + "loss": 0.4116, + "step": 24003 + }, + { + "epoch": 8.86, + "learning_rate": 2.1148909964281313e-06, + "loss": 0.4037, + "step": 24004 + }, + { + "epoch": 8.86, + "learning_rate": 2.1148540460647864e-06, + "loss": 0.3665, + "step": 24005 + }, + { + "epoch": 8.86, + "learning_rate": 2.114817095701441e-06, + "loss": 0.4972, + "step": 24006 + }, + { + "epoch": 8.87, + "learning_rate": 2.1147801453380957e-06, + "loss": 0.3148, + "step": 24007 + }, + { + "epoch": 8.87, + "learning_rate": 2.1147431949747507e-06, + "loss": 0.4106, + "step": 24008 + }, + { + "epoch": 8.87, + "learning_rate": 2.1147062446114054e-06, + "loss": 0.4053, + "step": 24009 + }, + { + "epoch": 8.87, + "learning_rate": 2.11466929424806e-06, + "loss": 0.4692, + "step": 24010 + }, + { + "epoch": 8.87, + "learning_rate": 2.1146323438847146e-06, + "loss": 0.4132, + "step": 24011 + }, + { + "epoch": 8.87, + "learning_rate": 2.1145953935213697e-06, + "loss": 0.4283, + "step": 24012 + }, + { + "epoch": 8.87, + "learning_rate": 2.1145584431580243e-06, + "loss": 0.5404, + "step": 24013 + }, + { + "epoch": 8.87, + "learning_rate": 2.1145214927946794e-06, + "loss": 0.4806, + "step": 24014 + }, + { + "epoch": 8.87, + "learning_rate": 2.114484542431334e-06, + "loss": 0.4587, + "step": 24015 + }, + { + "epoch": 8.87, + "learning_rate": 2.1144475920679887e-06, + "loss": 0.356, + "step": 24016 + }, + { + "epoch": 8.87, + "learning_rate": 2.1144106417046433e-06, + "loss": 0.4146, + "step": 24017 + }, + { + "epoch": 8.87, + "learning_rate": 2.114373691341298e-06, + "loss": 0.49, + "step": 24018 + }, + { + "epoch": 8.87, + "learning_rate": 2.114336740977953e-06, + "loss": 0.2659, + "step": 24019 + }, + { + "epoch": 8.87, + "learning_rate": 2.1142997906146076e-06, + "loss": 0.3153, + "step": 24020 + }, + { + "epoch": 8.87, + "learning_rate": 2.1142628402512627e-06, + "loss": 0.4209, + "step": 24021 + }, + { + "epoch": 8.87, + "learning_rate": 2.1142258898879173e-06, + "loss": 0.4875, + "step": 24022 + }, + { + "epoch": 8.87, + "learning_rate": 2.114188939524572e-06, + "loss": 0.4511, + "step": 24023 + }, + { + "epoch": 8.87, + "learning_rate": 2.1141519891612266e-06, + "loss": 0.4396, + "step": 24024 + }, + { + "epoch": 8.87, + "learning_rate": 2.1141150387978817e-06, + "loss": 0.4352, + "step": 24025 + }, + { + "epoch": 8.87, + "learning_rate": 2.1140780884345363e-06, + "loss": 0.4871, + "step": 24026 + }, + { + "epoch": 8.87, + "learning_rate": 2.114041138071191e-06, + "loss": 0.3251, + "step": 24027 + }, + { + "epoch": 8.87, + "learning_rate": 2.114004187707846e-06, + "loss": 0.3869, + "step": 24028 + }, + { + "epoch": 8.87, + "learning_rate": 2.1139672373445006e-06, + "loss": 0.3726, + "step": 24029 + }, + { + "epoch": 8.87, + "learning_rate": 2.1139302869811553e-06, + "loss": 0.4402, + "step": 24030 + }, + { + "epoch": 8.87, + "learning_rate": 2.11389333661781e-06, + "loss": 0.3096, + "step": 24031 + }, + { + "epoch": 8.87, + "learning_rate": 2.113856386254465e-06, + "loss": 0.4908, + "step": 24032 + }, + { + "epoch": 8.87, + "learning_rate": 2.1138194358911196e-06, + "loss": 0.3495, + "step": 24033 + }, + { + "epoch": 8.88, + "learning_rate": 2.1137824855277747e-06, + "loss": 0.3573, + "step": 24034 + }, + { + "epoch": 8.88, + "learning_rate": 2.1137455351644293e-06, + "loss": 0.3389, + "step": 24035 + }, + { + "epoch": 8.88, + "learning_rate": 2.113708584801084e-06, + "loss": 0.4377, + "step": 24036 + }, + { + "epoch": 8.88, + "learning_rate": 2.1136716344377386e-06, + "loss": 0.2918, + "step": 24037 + }, + { + "epoch": 8.88, + "learning_rate": 2.1136346840743932e-06, + "loss": 0.372, + "step": 24038 + }, + { + "epoch": 8.88, + "learning_rate": 2.1135977337110483e-06, + "loss": 0.3814, + "step": 24039 + }, + { + "epoch": 8.88, + "learning_rate": 2.113560783347703e-06, + "loss": 0.366, + "step": 24040 + }, + { + "epoch": 8.88, + "learning_rate": 2.113523832984358e-06, + "loss": 0.347, + "step": 24041 + }, + { + "epoch": 8.88, + "learning_rate": 2.1134868826210126e-06, + "loss": 0.3688, + "step": 24042 + }, + { + "epoch": 8.88, + "learning_rate": 2.1134499322576672e-06, + "loss": 0.5383, + "step": 24043 + }, + { + "epoch": 8.88, + "learning_rate": 2.113412981894322e-06, + "loss": 0.3959, + "step": 24044 + }, + { + "epoch": 8.88, + "learning_rate": 2.113376031530977e-06, + "loss": 0.4501, + "step": 24045 + }, + { + "epoch": 8.88, + "learning_rate": 2.1133390811676316e-06, + "loss": 0.3746, + "step": 24046 + }, + { + "epoch": 8.88, + "learning_rate": 2.1133021308042862e-06, + "loss": 0.3665, + "step": 24047 + }, + { + "epoch": 8.88, + "learning_rate": 2.1132651804409413e-06, + "loss": 0.3422, + "step": 24048 + }, + { + "epoch": 8.88, + "learning_rate": 2.1132282300775955e-06, + "loss": 0.2948, + "step": 24049 + }, + { + "epoch": 8.88, + "learning_rate": 2.1131912797142505e-06, + "loss": 0.2795, + "step": 24050 + }, + { + "epoch": 8.88, + "learning_rate": 2.113154329350905e-06, + "loss": 0.3787, + "step": 24051 + }, + { + "epoch": 8.88, + "learning_rate": 2.1131173789875602e-06, + "loss": 0.2847, + "step": 24052 + }, + { + "epoch": 8.88, + "learning_rate": 2.113080428624215e-06, + "loss": 0.3005, + "step": 24053 + }, + { + "epoch": 8.88, + "learning_rate": 2.11304347826087e-06, + "loss": 0.3997, + "step": 24054 + }, + { + "epoch": 8.88, + "learning_rate": 2.113006527897524e-06, + "loss": 0.3493, + "step": 24055 + }, + { + "epoch": 8.88, + "learning_rate": 2.1129695775341792e-06, + "loss": 0.3967, + "step": 24056 + }, + { + "epoch": 8.88, + "learning_rate": 2.112932627170834e-06, + "loss": 0.5168, + "step": 24057 + }, + { + "epoch": 8.88, + "learning_rate": 2.1128956768074885e-06, + "loss": 0.4617, + "step": 24058 + }, + { + "epoch": 8.88, + "learning_rate": 2.1128587264441435e-06, + "loss": 0.3073, + "step": 24059 + }, + { + "epoch": 8.88, + "learning_rate": 2.112821776080798e-06, + "loss": 0.282, + "step": 24060 + }, + { + "epoch": 8.89, + "learning_rate": 2.112784825717453e-06, + "loss": 0.5082, + "step": 24061 + }, + { + "epoch": 8.89, + "learning_rate": 2.1127478753541075e-06, + "loss": 0.4478, + "step": 24062 + }, + { + "epoch": 8.89, + "learning_rate": 2.1127109249907625e-06, + "loss": 0.3893, + "step": 24063 + }, + { + "epoch": 8.89, + "learning_rate": 2.112673974627417e-06, + "loss": 0.2932, + "step": 24064 + }, + { + "epoch": 8.89, + "learning_rate": 2.112637024264072e-06, + "loss": 0.3476, + "step": 24065 + }, + { + "epoch": 8.89, + "learning_rate": 2.112600073900727e-06, + "loss": 0.429, + "step": 24066 + }, + { + "epoch": 8.89, + "learning_rate": 2.1125631235373815e-06, + "loss": 0.3354, + "step": 24067 + }, + { + "epoch": 8.89, + "learning_rate": 2.112526173174036e-06, + "loss": 0.4118, + "step": 24068 + }, + { + "epoch": 8.89, + "learning_rate": 2.1124892228106908e-06, + "loss": 0.3009, + "step": 24069 + }, + { + "epoch": 8.89, + "learning_rate": 2.112452272447346e-06, + "loss": 0.3929, + "step": 24070 + }, + { + "epoch": 8.89, + "learning_rate": 2.1124153220840005e-06, + "loss": 0.5223, + "step": 24071 + }, + { + "epoch": 8.89, + "learning_rate": 2.1123783717206555e-06, + "loss": 0.4605, + "step": 24072 + }, + { + "epoch": 8.89, + "learning_rate": 2.11234142135731e-06, + "loss": 0.3172, + "step": 24073 + }, + { + "epoch": 8.89, + "learning_rate": 2.1123044709939648e-06, + "loss": 0.4196, + "step": 24074 + }, + { + "epoch": 8.89, + "learning_rate": 2.1122675206306194e-06, + "loss": 0.2985, + "step": 24075 + }, + { + "epoch": 8.89, + "learning_rate": 2.1122305702672745e-06, + "loss": 0.4061, + "step": 24076 + }, + { + "epoch": 8.89, + "learning_rate": 2.112193619903929e-06, + "loss": 0.3409, + "step": 24077 + }, + { + "epoch": 8.89, + "learning_rate": 2.1121566695405838e-06, + "loss": 0.4592, + "step": 24078 + }, + { + "epoch": 8.89, + "learning_rate": 2.112119719177239e-06, + "loss": 0.4485, + "step": 24079 + }, + { + "epoch": 8.89, + "learning_rate": 2.1120827688138935e-06, + "loss": 0.2955, + "step": 24080 + }, + { + "epoch": 8.89, + "learning_rate": 2.112045818450548e-06, + "loss": 0.4506, + "step": 24081 + }, + { + "epoch": 8.89, + "learning_rate": 2.1120088680872027e-06, + "loss": 0.3853, + "step": 24082 + }, + { + "epoch": 8.89, + "learning_rate": 2.1119719177238578e-06, + "loss": 0.4328, + "step": 24083 + }, + { + "epoch": 8.89, + "learning_rate": 2.1119349673605124e-06, + "loss": 0.3215, + "step": 24084 + }, + { + "epoch": 8.89, + "learning_rate": 2.1118980169971675e-06, + "loss": 0.2583, + "step": 24085 + }, + { + "epoch": 8.89, + "learning_rate": 2.111861066633822e-06, + "loss": 0.362, + "step": 24086 + }, + { + "epoch": 8.89, + "learning_rate": 2.1118241162704768e-06, + "loss": 0.3319, + "step": 24087 + }, + { + "epoch": 8.9, + "learning_rate": 2.1117871659071314e-06, + "loss": 0.3788, + "step": 24088 + }, + { + "epoch": 8.9, + "learning_rate": 2.111750215543786e-06, + "loss": 0.3061, + "step": 24089 + }, + { + "epoch": 8.9, + "learning_rate": 2.111713265180441e-06, + "loss": 0.4054, + "step": 24090 + }, + { + "epoch": 8.9, + "learning_rate": 2.1116763148170957e-06, + "loss": 0.2795, + "step": 24091 + }, + { + "epoch": 8.9, + "learning_rate": 2.1116393644537508e-06, + "loss": 0.3969, + "step": 24092 + }, + { + "epoch": 8.9, + "learning_rate": 2.1116024140904054e-06, + "loss": 0.4487, + "step": 24093 + }, + { + "epoch": 8.9, + "learning_rate": 2.11156546372706e-06, + "loss": 0.3838, + "step": 24094 + }, + { + "epoch": 8.9, + "learning_rate": 2.1115285133637147e-06, + "loss": 0.455, + "step": 24095 + }, + { + "epoch": 8.9, + "learning_rate": 2.1114915630003698e-06, + "loss": 0.313, + "step": 24096 + }, + { + "epoch": 8.9, + "learning_rate": 2.1114546126370244e-06, + "loss": 0.4175, + "step": 24097 + }, + { + "epoch": 8.9, + "learning_rate": 2.111417662273679e-06, + "loss": 0.309, + "step": 24098 + }, + { + "epoch": 8.9, + "learning_rate": 2.111380711910334e-06, + "loss": 0.3633, + "step": 24099 + }, + { + "epoch": 8.9, + "learning_rate": 2.1113437615469883e-06, + "loss": 0.3987, + "step": 24100 + }, + { + "epoch": 8.9, + "learning_rate": 2.1113068111836434e-06, + "loss": 0.492, + "step": 24101 + }, + { + "epoch": 8.9, + "learning_rate": 2.111269860820298e-06, + "loss": 0.5446, + "step": 24102 + }, + { + "epoch": 8.9, + "learning_rate": 2.111232910456953e-06, + "loss": 0.2688, + "step": 24103 + }, + { + "epoch": 8.9, + "learning_rate": 2.1111959600936077e-06, + "loss": 0.3514, + "step": 24104 + }, + { + "epoch": 8.9, + "learning_rate": 2.1111590097302628e-06, + "loss": 0.4181, + "step": 24105 + }, + { + "epoch": 8.9, + "learning_rate": 2.111122059366917e-06, + "loss": 0.3909, + "step": 24106 + }, + { + "epoch": 8.9, + "learning_rate": 2.1110851090035716e-06, + "loss": 0.3711, + "step": 24107 + }, + { + "epoch": 8.9, + "learning_rate": 2.1110481586402267e-06, + "loss": 0.2541, + "step": 24108 + }, + { + "epoch": 8.9, + "learning_rate": 2.1110112082768813e-06, + "loss": 0.5233, + "step": 24109 + }, + { + "epoch": 8.9, + "learning_rate": 2.1109742579135364e-06, + "loss": 0.4048, + "step": 24110 + }, + { + "epoch": 8.9, + "learning_rate": 2.110937307550191e-06, + "loss": 0.278, + "step": 24111 + }, + { + "epoch": 8.9, + "learning_rate": 2.1109003571868456e-06, + "loss": 0.3764, + "step": 24112 + }, + { + "epoch": 8.9, + "learning_rate": 2.1108634068235003e-06, + "loss": 0.4022, + "step": 24113 + }, + { + "epoch": 8.9, + "learning_rate": 2.1108264564601553e-06, + "loss": 0.4317, + "step": 24114 + }, + { + "epoch": 8.91, + "learning_rate": 2.11078950609681e-06, + "loss": 0.317, + "step": 24115 + }, + { + "epoch": 8.91, + "learning_rate": 2.110752555733465e-06, + "loss": 0.3168, + "step": 24116 + }, + { + "epoch": 8.91, + "learning_rate": 2.1107156053701197e-06, + "loss": 0.4531, + "step": 24117 + }, + { + "epoch": 8.91, + "learning_rate": 2.1106786550067743e-06, + "loss": 0.318, + "step": 24118 + }, + { + "epoch": 8.91, + "learning_rate": 2.110641704643429e-06, + "loss": 0.3809, + "step": 24119 + }, + { + "epoch": 8.91, + "learning_rate": 2.1106047542800836e-06, + "loss": 0.3525, + "step": 24120 + }, + { + "epoch": 8.91, + "learning_rate": 2.1105678039167386e-06, + "loss": 0.4394, + "step": 24121 + }, + { + "epoch": 8.91, + "learning_rate": 2.1105308535533933e-06, + "loss": 0.3376, + "step": 24122 + }, + { + "epoch": 8.91, + "learning_rate": 2.1104939031900483e-06, + "loss": 0.2631, + "step": 24123 + }, + { + "epoch": 8.91, + "learning_rate": 2.110456952826703e-06, + "loss": 0.2789, + "step": 24124 + }, + { + "epoch": 8.91, + "learning_rate": 2.1104200024633576e-06, + "loss": 0.4848, + "step": 24125 + }, + { + "epoch": 8.91, + "learning_rate": 2.1103830521000122e-06, + "loss": 0.3169, + "step": 24126 + }, + { + "epoch": 8.91, + "learning_rate": 2.110346101736667e-06, + "loss": 0.4413, + "step": 24127 + }, + { + "epoch": 8.91, + "learning_rate": 2.110309151373322e-06, + "loss": 0.3653, + "step": 24128 + }, + { + "epoch": 8.91, + "learning_rate": 2.1102722010099766e-06, + "loss": 0.321, + "step": 24129 + }, + { + "epoch": 8.91, + "learning_rate": 2.1102352506466316e-06, + "loss": 0.329, + "step": 24130 + }, + { + "epoch": 8.91, + "learning_rate": 2.1101983002832863e-06, + "loss": 0.3103, + "step": 24131 + }, + { + "epoch": 8.91, + "learning_rate": 2.110161349919941e-06, + "loss": 0.328, + "step": 24132 + }, + { + "epoch": 8.91, + "learning_rate": 2.1101243995565955e-06, + "loss": 0.2647, + "step": 24133 + }, + { + "epoch": 8.91, + "learning_rate": 2.1100874491932506e-06, + "loss": 0.3084, + "step": 24134 + }, + { + "epoch": 8.91, + "learning_rate": 2.1100504988299052e-06, + "loss": 0.317, + "step": 24135 + }, + { + "epoch": 8.91, + "learning_rate": 2.1100135484665603e-06, + "loss": 0.6351, + "step": 24136 + }, + { + "epoch": 8.91, + "learning_rate": 2.109976598103215e-06, + "loss": 0.2989, + "step": 24137 + }, + { + "epoch": 8.91, + "learning_rate": 2.1099396477398696e-06, + "loss": 0.3478, + "step": 24138 + }, + { + "epoch": 8.91, + "learning_rate": 2.109902697376524e-06, + "loss": 0.352, + "step": 24139 + }, + { + "epoch": 8.91, + "learning_rate": 2.109865747013179e-06, + "loss": 0.4007, + "step": 24140 + }, + { + "epoch": 8.91, + "learning_rate": 2.109828796649834e-06, + "loss": 0.3649, + "step": 24141 + }, + { + "epoch": 8.92, + "learning_rate": 2.1097918462864885e-06, + "loss": 0.3444, + "step": 24142 + }, + { + "epoch": 8.92, + "learning_rate": 2.1097548959231436e-06, + "loss": 0.3161, + "step": 24143 + }, + { + "epoch": 8.92, + "learning_rate": 2.1097179455597982e-06, + "loss": 0.4437, + "step": 24144 + }, + { + "epoch": 8.92, + "learning_rate": 2.109680995196453e-06, + "loss": 0.3886, + "step": 24145 + }, + { + "epoch": 8.92, + "learning_rate": 2.1096440448331075e-06, + "loss": 0.5253, + "step": 24146 + }, + { + "epoch": 8.92, + "learning_rate": 2.109607094469762e-06, + "loss": 0.4456, + "step": 24147 + }, + { + "epoch": 8.92, + "learning_rate": 2.109570144106417e-06, + "loss": 0.4926, + "step": 24148 + }, + { + "epoch": 8.92, + "learning_rate": 2.109533193743072e-06, + "loss": 0.3919, + "step": 24149 + }, + { + "epoch": 8.92, + "learning_rate": 2.109496243379727e-06, + "loss": 0.3913, + "step": 24150 + }, + { + "epoch": 8.92, + "learning_rate": 2.109459293016381e-06, + "loss": 0.2775, + "step": 24151 + }, + { + "epoch": 8.92, + "learning_rate": 2.109422342653036e-06, + "loss": 0.3063, + "step": 24152 + }, + { + "epoch": 8.92, + "learning_rate": 2.109385392289691e-06, + "loss": 0.3737, + "step": 24153 + }, + { + "epoch": 8.92, + "learning_rate": 2.109348441926346e-06, + "loss": 0.4071, + "step": 24154 + }, + { + "epoch": 8.92, + "learning_rate": 2.1093114915630005e-06, + "loss": 0.3722, + "step": 24155 + }, + { + "epoch": 8.92, + "learning_rate": 2.109274541199655e-06, + "loss": 0.4485, + "step": 24156 + }, + { + "epoch": 8.92, + "learning_rate": 2.1092375908363098e-06, + "loss": 0.3601, + "step": 24157 + }, + { + "epoch": 8.92, + "learning_rate": 2.1092006404729644e-06, + "loss": 0.5047, + "step": 24158 + }, + { + "epoch": 8.92, + "learning_rate": 2.1091636901096195e-06, + "loss": 0.3323, + "step": 24159 + }, + { + "epoch": 8.92, + "learning_rate": 2.109126739746274e-06, + "loss": 0.4605, + "step": 24160 + }, + { + "epoch": 8.92, + "learning_rate": 2.109089789382929e-06, + "loss": 0.3342, + "step": 24161 + }, + { + "epoch": 8.92, + "learning_rate": 2.109052839019584e-06, + "loss": 0.4036, + "step": 24162 + }, + { + "epoch": 8.92, + "learning_rate": 2.1090158886562384e-06, + "loss": 0.6092, + "step": 24163 + }, + { + "epoch": 8.92, + "learning_rate": 2.108978938292893e-06, + "loss": 0.284, + "step": 24164 + }, + { + "epoch": 8.92, + "learning_rate": 2.108941987929548e-06, + "loss": 0.4547, + "step": 24165 + }, + { + "epoch": 8.92, + "learning_rate": 2.1089050375662028e-06, + "loss": 0.4676, + "step": 24166 + }, + { + "epoch": 8.92, + "learning_rate": 2.1088680872028574e-06, + "loss": 0.3867, + "step": 24167 + }, + { + "epoch": 8.92, + "learning_rate": 2.1088311368395125e-06, + "loss": 0.3618, + "step": 24168 + }, + { + "epoch": 8.93, + "learning_rate": 2.108794186476167e-06, + "loss": 0.4072, + "step": 24169 + }, + { + "epoch": 8.93, + "learning_rate": 2.1087572361128217e-06, + "loss": 0.2779, + "step": 24170 + }, + { + "epoch": 8.93, + "learning_rate": 2.1087202857494764e-06, + "loss": 0.4198, + "step": 24171 + }, + { + "epoch": 8.93, + "learning_rate": 2.1086833353861314e-06, + "loss": 0.3051, + "step": 24172 + }, + { + "epoch": 8.93, + "learning_rate": 2.108646385022786e-06, + "loss": 0.3266, + "step": 24173 + }, + { + "epoch": 8.93, + "learning_rate": 2.108609434659441e-06, + "loss": 0.2791, + "step": 24174 + }, + { + "epoch": 8.93, + "learning_rate": 2.1085724842960958e-06, + "loss": 0.4957, + "step": 24175 + }, + { + "epoch": 8.93, + "learning_rate": 2.1085355339327504e-06, + "loss": 0.425, + "step": 24176 + }, + { + "epoch": 8.93, + "learning_rate": 2.108498583569405e-06, + "loss": 0.4549, + "step": 24177 + }, + { + "epoch": 8.93, + "learning_rate": 2.1084616332060597e-06, + "loss": 0.3965, + "step": 24178 + }, + { + "epoch": 8.93, + "learning_rate": 2.1084246828427147e-06, + "loss": 0.4344, + "step": 24179 + }, + { + "epoch": 8.93, + "learning_rate": 2.1083877324793694e-06, + "loss": 0.3964, + "step": 24180 + }, + { + "epoch": 8.93, + "learning_rate": 2.1083507821160244e-06, + "loss": 0.4342, + "step": 24181 + }, + { + "epoch": 8.93, + "learning_rate": 2.108313831752679e-06, + "loss": 0.4816, + "step": 24182 + }, + { + "epoch": 8.93, + "learning_rate": 2.1082768813893337e-06, + "loss": 0.4346, + "step": 24183 + }, + { + "epoch": 8.93, + "learning_rate": 2.1082399310259883e-06, + "loss": 0.3508, + "step": 24184 + }, + { + "epoch": 8.93, + "learning_rate": 2.1082029806626434e-06, + "loss": 0.3058, + "step": 24185 + }, + { + "epoch": 8.93, + "learning_rate": 2.108166030299298e-06, + "loss": 0.3904, + "step": 24186 + }, + { + "epoch": 8.93, + "learning_rate": 2.1081290799359527e-06, + "loss": 0.3146, + "step": 24187 + }, + { + "epoch": 8.93, + "learning_rate": 2.1080921295726077e-06, + "loss": 0.5185, + "step": 24188 + }, + { + "epoch": 8.93, + "learning_rate": 2.108055179209262e-06, + "loss": 0.4091, + "step": 24189 + }, + { + "epoch": 8.93, + "learning_rate": 2.108018228845917e-06, + "loss": 0.3714, + "step": 24190 + }, + { + "epoch": 8.93, + "learning_rate": 2.1079812784825716e-06, + "loss": 0.3425, + "step": 24191 + }, + { + "epoch": 8.93, + "learning_rate": 2.1079443281192267e-06, + "loss": 0.5514, + "step": 24192 + }, + { + "epoch": 8.93, + "learning_rate": 2.1079073777558813e-06, + "loss": 0.4337, + "step": 24193 + }, + { + "epoch": 8.93, + "learning_rate": 2.1078704273925364e-06, + "loss": 0.4574, + "step": 24194 + }, + { + "epoch": 8.93, + "learning_rate": 2.1078334770291906e-06, + "loss": 0.4957, + "step": 24195 + }, + { + "epoch": 8.94, + "learning_rate": 2.1077965266658453e-06, + "loss": 0.2701, + "step": 24196 + }, + { + "epoch": 8.94, + "learning_rate": 2.1077595763025003e-06, + "loss": 0.3515, + "step": 24197 + }, + { + "epoch": 8.94, + "learning_rate": 2.107722625939155e-06, + "loss": 0.3236, + "step": 24198 + }, + { + "epoch": 8.94, + "learning_rate": 2.10768567557581e-06, + "loss": 0.3593, + "step": 24199 + }, + { + "epoch": 8.94, + "learning_rate": 2.1076487252124646e-06, + "loss": 0.4003, + "step": 24200 + }, + { + "epoch": 8.94, + "learning_rate": 2.1076117748491193e-06, + "loss": 0.4842, + "step": 24201 + }, + { + "epoch": 8.94, + "learning_rate": 2.107574824485774e-06, + "loss": 0.2882, + "step": 24202 + }, + { + "epoch": 8.94, + "learning_rate": 2.107537874122429e-06, + "loss": 0.4436, + "step": 24203 + }, + { + "epoch": 8.94, + "learning_rate": 2.1075009237590836e-06, + "loss": 0.3199, + "step": 24204 + }, + { + "epoch": 8.94, + "learning_rate": 2.1074639733957387e-06, + "loss": 0.3306, + "step": 24205 + }, + { + "epoch": 8.94, + "learning_rate": 2.1074270230323933e-06, + "loss": 0.323, + "step": 24206 + }, + { + "epoch": 8.94, + "learning_rate": 2.107390072669048e-06, + "loss": 0.2923, + "step": 24207 + }, + { + "epoch": 8.94, + "learning_rate": 2.1073531223057026e-06, + "loss": 0.5109, + "step": 24208 + }, + { + "epoch": 8.94, + "learning_rate": 2.1073161719423572e-06, + "loss": 0.3491, + "step": 24209 + }, + { + "epoch": 8.94, + "learning_rate": 2.1072792215790123e-06, + "loss": 0.456, + "step": 24210 + }, + { + "epoch": 8.94, + "learning_rate": 2.107242271215667e-06, + "loss": 0.3387, + "step": 24211 + }, + { + "epoch": 8.94, + "learning_rate": 2.107205320852322e-06, + "loss": 0.3505, + "step": 24212 + }, + { + "epoch": 8.94, + "learning_rate": 2.1071683704889766e-06, + "loss": 0.3339, + "step": 24213 + }, + { + "epoch": 8.94, + "learning_rate": 2.1071314201256313e-06, + "loss": 0.3364, + "step": 24214 + }, + { + "epoch": 8.94, + "learning_rate": 2.107094469762286e-06, + "loss": 0.4759, + "step": 24215 + }, + { + "epoch": 8.94, + "learning_rate": 2.1070575193989405e-06, + "loss": 0.2544, + "step": 24216 + }, + { + "epoch": 8.94, + "learning_rate": 2.1070205690355956e-06, + "loss": 0.3481, + "step": 24217 + }, + { + "epoch": 8.94, + "learning_rate": 2.1069836186722502e-06, + "loss": 0.3339, + "step": 24218 + }, + { + "epoch": 8.94, + "learning_rate": 2.1069466683089053e-06, + "loss": 0.5795, + "step": 24219 + }, + { + "epoch": 8.94, + "learning_rate": 2.10690971794556e-06, + "loss": 0.2318, + "step": 24220 + }, + { + "epoch": 8.94, + "learning_rate": 2.1068727675822146e-06, + "loss": 0.3528, + "step": 24221 + }, + { + "epoch": 8.94, + "learning_rate": 2.106835817218869e-06, + "loss": 0.4377, + "step": 24222 + }, + { + "epoch": 8.94, + "learning_rate": 2.1067988668555242e-06, + "loss": 0.2839, + "step": 24223 + }, + { + "epoch": 8.95, + "learning_rate": 2.106761916492179e-06, + "loss": 0.4065, + "step": 24224 + }, + { + "epoch": 8.95, + "learning_rate": 2.106724966128834e-06, + "loss": 0.4477, + "step": 24225 + }, + { + "epoch": 8.95, + "learning_rate": 2.1066880157654886e-06, + "loss": 0.2888, + "step": 24226 + }, + { + "epoch": 8.95, + "learning_rate": 2.1066510654021432e-06, + "loss": 0.3882, + "step": 24227 + }, + { + "epoch": 8.95, + "learning_rate": 2.106614115038798e-06, + "loss": 0.4037, + "step": 24228 + }, + { + "epoch": 8.95, + "learning_rate": 2.1065771646754525e-06, + "loss": 0.3987, + "step": 24229 + }, + { + "epoch": 8.95, + "learning_rate": 2.1065402143121076e-06, + "loss": 0.377, + "step": 24230 + }, + { + "epoch": 8.95, + "learning_rate": 2.106503263948762e-06, + "loss": 0.4066, + "step": 24231 + }, + { + "epoch": 8.95, + "learning_rate": 2.1064663135854172e-06, + "loss": 0.381, + "step": 24232 + }, + { + "epoch": 8.95, + "learning_rate": 2.106429363222072e-06, + "loss": 0.3825, + "step": 24233 + }, + { + "epoch": 8.95, + "learning_rate": 2.1063924128587265e-06, + "loss": 0.2956, + "step": 24234 + }, + { + "epoch": 8.95, + "learning_rate": 2.106355462495381e-06, + "loss": 0.3651, + "step": 24235 + }, + { + "epoch": 8.95, + "learning_rate": 2.106318512132036e-06, + "loss": 0.2736, + "step": 24236 + }, + { + "epoch": 8.95, + "learning_rate": 2.106281561768691e-06, + "loss": 0.5323, + "step": 24237 + }, + { + "epoch": 8.95, + "learning_rate": 2.1062446114053455e-06, + "loss": 0.2992, + "step": 24238 + }, + { + "epoch": 8.95, + "learning_rate": 2.1062076610420005e-06, + "loss": 0.4276, + "step": 24239 + }, + { + "epoch": 8.95, + "learning_rate": 2.1061707106786548e-06, + "loss": 0.2846, + "step": 24240 + }, + { + "epoch": 8.95, + "learning_rate": 2.10613376031531e-06, + "loss": 0.2746, + "step": 24241 + }, + { + "epoch": 8.95, + "learning_rate": 2.1060968099519645e-06, + "loss": 0.3958, + "step": 24242 + }, + { + "epoch": 8.95, + "learning_rate": 2.1060598595886195e-06, + "loss": 0.4533, + "step": 24243 + }, + { + "epoch": 8.95, + "learning_rate": 2.106022909225274e-06, + "loss": 0.5427, + "step": 24244 + }, + { + "epoch": 8.95, + "learning_rate": 2.1059859588619292e-06, + "loss": 0.3748, + "step": 24245 + }, + { + "epoch": 8.95, + "learning_rate": 2.1059490084985834e-06, + "loss": 0.5734, + "step": 24246 + }, + { + "epoch": 8.95, + "learning_rate": 2.105912058135238e-06, + "loss": 0.2713, + "step": 24247 + }, + { + "epoch": 8.95, + "learning_rate": 2.105875107771893e-06, + "loss": 0.3557, + "step": 24248 + }, + { + "epoch": 8.95, + "learning_rate": 2.1058381574085478e-06, + "loss": 0.3857, + "step": 24249 + }, + { + "epoch": 8.95, + "learning_rate": 2.105801207045203e-06, + "loss": 0.4545, + "step": 24250 + }, + { + "epoch": 8.96, + "learning_rate": 2.1057642566818575e-06, + "loss": 0.4804, + "step": 24251 + }, + { + "epoch": 8.96, + "learning_rate": 2.105727306318512e-06, + "loss": 0.3524, + "step": 24252 + }, + { + "epoch": 8.96, + "learning_rate": 2.1056903559551667e-06, + "loss": 0.3056, + "step": 24253 + }, + { + "epoch": 8.96, + "learning_rate": 2.105653405591822e-06, + "loss": 0.2705, + "step": 24254 + }, + { + "epoch": 8.96, + "learning_rate": 2.1056164552284764e-06, + "loss": 0.2903, + "step": 24255 + }, + { + "epoch": 8.96, + "learning_rate": 2.105579504865131e-06, + "loss": 0.222, + "step": 24256 + }, + { + "epoch": 8.96, + "learning_rate": 2.105542554501786e-06, + "loss": 0.3499, + "step": 24257 + }, + { + "epoch": 8.96, + "learning_rate": 2.1055056041384408e-06, + "loss": 0.432, + "step": 24258 + }, + { + "epoch": 8.96, + "learning_rate": 2.1054686537750954e-06, + "loss": 0.3789, + "step": 24259 + }, + { + "epoch": 8.96, + "learning_rate": 2.10543170341175e-06, + "loss": 0.4666, + "step": 24260 + }, + { + "epoch": 8.96, + "learning_rate": 2.105394753048405e-06, + "loss": 0.3531, + "step": 24261 + }, + { + "epoch": 8.96, + "learning_rate": 2.1053578026850597e-06, + "loss": 0.4368, + "step": 24262 + }, + { + "epoch": 8.96, + "learning_rate": 2.105320852321715e-06, + "loss": 0.3535, + "step": 24263 + }, + { + "epoch": 8.96, + "learning_rate": 2.1052839019583694e-06, + "loss": 0.3349, + "step": 24264 + }, + { + "epoch": 8.96, + "learning_rate": 2.105246951595024e-06, + "loss": 0.4223, + "step": 24265 + }, + { + "epoch": 8.96, + "learning_rate": 2.1052100012316787e-06, + "loss": 0.4632, + "step": 24266 + }, + { + "epoch": 8.96, + "learning_rate": 2.1051730508683333e-06, + "loss": 0.4227, + "step": 24267 + }, + { + "epoch": 8.96, + "learning_rate": 2.1051361005049884e-06, + "loss": 0.4084, + "step": 24268 + }, + { + "epoch": 8.96, + "learning_rate": 2.105099150141643e-06, + "loss": 0.4822, + "step": 24269 + }, + { + "epoch": 8.96, + "learning_rate": 2.105062199778298e-06, + "loss": 0.3618, + "step": 24270 + }, + { + "epoch": 8.96, + "learning_rate": 2.1050252494149527e-06, + "loss": 0.3782, + "step": 24271 + }, + { + "epoch": 8.96, + "learning_rate": 2.1049882990516074e-06, + "loss": 0.4636, + "step": 24272 + }, + { + "epoch": 8.96, + "learning_rate": 2.104951348688262e-06, + "loss": 0.3889, + "step": 24273 + }, + { + "epoch": 8.96, + "learning_rate": 2.104914398324917e-06, + "loss": 0.3924, + "step": 24274 + }, + { + "epoch": 8.96, + "learning_rate": 2.1048774479615717e-06, + "loss": 0.4838, + "step": 24275 + }, + { + "epoch": 8.96, + "learning_rate": 2.1048404975982263e-06, + "loss": 0.4327, + "step": 24276 + }, + { + "epoch": 8.96, + "learning_rate": 2.1048035472348814e-06, + "loss": 0.3832, + "step": 24277 + }, + { + "epoch": 8.97, + "learning_rate": 2.104766596871536e-06, + "loss": 0.3956, + "step": 24278 + }, + { + "epoch": 8.97, + "learning_rate": 2.1047296465081907e-06, + "loss": 0.5663, + "step": 24279 + }, + { + "epoch": 8.97, + "learning_rate": 2.1046926961448453e-06, + "loss": 0.4003, + "step": 24280 + }, + { + "epoch": 8.97, + "learning_rate": 2.1046557457815004e-06, + "loss": 0.3294, + "step": 24281 + }, + { + "epoch": 8.97, + "learning_rate": 2.104618795418155e-06, + "loss": 0.303, + "step": 24282 + }, + { + "epoch": 8.97, + "learning_rate": 2.10458184505481e-06, + "loss": 0.3837, + "step": 24283 + }, + { + "epoch": 8.97, + "learning_rate": 2.1045448946914647e-06, + "loss": 0.3944, + "step": 24284 + }, + { + "epoch": 8.97, + "learning_rate": 2.1045079443281193e-06, + "loss": 0.5155, + "step": 24285 + }, + { + "epoch": 8.97, + "learning_rate": 2.104470993964774e-06, + "loss": 0.3312, + "step": 24286 + }, + { + "epoch": 8.97, + "learning_rate": 2.1044340436014286e-06, + "loss": 0.3658, + "step": 24287 + }, + { + "epoch": 8.97, + "learning_rate": 2.1043970932380837e-06, + "loss": 0.4649, + "step": 24288 + }, + { + "epoch": 8.97, + "learning_rate": 2.1043601428747383e-06, + "loss": 0.3445, + "step": 24289 + }, + { + "epoch": 8.97, + "learning_rate": 2.1043231925113934e-06, + "loss": 0.277, + "step": 24290 + }, + { + "epoch": 8.97, + "learning_rate": 2.1042862421480476e-06, + "loss": 0.4226, + "step": 24291 + }, + { + "epoch": 8.97, + "learning_rate": 2.1042492917847026e-06, + "loss": 0.3229, + "step": 24292 + }, + { + "epoch": 8.97, + "learning_rate": 2.1042123414213573e-06, + "loss": 0.3398, + "step": 24293 + }, + { + "epoch": 8.97, + "learning_rate": 2.1041753910580123e-06, + "loss": 0.4158, + "step": 24294 + }, + { + "epoch": 8.97, + "learning_rate": 2.104138440694667e-06, + "loss": 0.2426, + "step": 24295 + }, + { + "epoch": 8.97, + "learning_rate": 2.1041014903313216e-06, + "loss": 0.3404, + "step": 24296 + }, + { + "epoch": 8.97, + "learning_rate": 2.1040645399679762e-06, + "loss": 0.3867, + "step": 24297 + }, + { + "epoch": 8.97, + "learning_rate": 2.104027589604631e-06, + "loss": 0.4996, + "step": 24298 + }, + { + "epoch": 8.97, + "learning_rate": 2.103990639241286e-06, + "loss": 0.3607, + "step": 24299 + }, + { + "epoch": 8.97, + "learning_rate": 2.1039536888779406e-06, + "loss": 0.3583, + "step": 24300 + }, + { + "epoch": 8.97, + "learning_rate": 2.1039167385145956e-06, + "loss": 0.3481, + "step": 24301 + }, + { + "epoch": 8.97, + "learning_rate": 2.1038797881512503e-06, + "loss": 0.3424, + "step": 24302 + }, + { + "epoch": 8.97, + "learning_rate": 2.103842837787905e-06, + "loss": 0.4224, + "step": 24303 + }, + { + "epoch": 8.97, + "learning_rate": 2.1038058874245595e-06, + "loss": 0.4223, + "step": 24304 + }, + { + "epoch": 8.98, + "learning_rate": 2.1037689370612146e-06, + "loss": 0.5806, + "step": 24305 + }, + { + "epoch": 8.98, + "learning_rate": 2.1037319866978692e-06, + "loss": 0.2913, + "step": 24306 + }, + { + "epoch": 8.98, + "learning_rate": 2.103695036334524e-06, + "loss": 0.3, + "step": 24307 + }, + { + "epoch": 8.98, + "learning_rate": 2.103658085971179e-06, + "loss": 0.4936, + "step": 24308 + }, + { + "epoch": 8.98, + "learning_rate": 2.1036211356078336e-06, + "loss": 0.4382, + "step": 24309 + }, + { + "epoch": 8.98, + "learning_rate": 2.103584185244488e-06, + "loss": 0.2634, + "step": 24310 + }, + { + "epoch": 8.98, + "learning_rate": 2.103547234881143e-06, + "loss": 0.3344, + "step": 24311 + }, + { + "epoch": 8.98, + "learning_rate": 2.103510284517798e-06, + "loss": 0.4385, + "step": 24312 + }, + { + "epoch": 8.98, + "learning_rate": 2.1034733341544525e-06, + "loss": 0.448, + "step": 24313 + }, + { + "epoch": 8.98, + "learning_rate": 2.1034363837911076e-06, + "loss": 0.3175, + "step": 24314 + }, + { + "epoch": 8.98, + "learning_rate": 2.1033994334277622e-06, + "loss": 0.3403, + "step": 24315 + }, + { + "epoch": 8.98, + "learning_rate": 2.103362483064417e-06, + "loss": 0.3728, + "step": 24316 + }, + { + "epoch": 8.98, + "learning_rate": 2.1033255327010715e-06, + "loss": 0.4467, + "step": 24317 + }, + { + "epoch": 8.98, + "learning_rate": 2.103288582337726e-06, + "loss": 0.4059, + "step": 24318 + }, + { + "epoch": 8.98, + "learning_rate": 2.103251631974381e-06, + "loss": 0.4336, + "step": 24319 + }, + { + "epoch": 8.98, + "learning_rate": 2.103214681611036e-06, + "loss": 0.3616, + "step": 24320 + }, + { + "epoch": 8.98, + "learning_rate": 2.103177731247691e-06, + "loss": 0.3438, + "step": 24321 + }, + { + "epoch": 8.98, + "learning_rate": 2.1031407808843455e-06, + "loss": 0.5578, + "step": 24322 + }, + { + "epoch": 8.98, + "learning_rate": 2.103103830521e-06, + "loss": 0.3989, + "step": 24323 + }, + { + "epoch": 8.98, + "learning_rate": 2.103066880157655e-06, + "loss": 0.3331, + "step": 24324 + }, + { + "epoch": 8.98, + "learning_rate": 2.10302992979431e-06, + "loss": 0.5517, + "step": 24325 + }, + { + "epoch": 8.98, + "learning_rate": 2.1029929794309645e-06, + "loss": 0.532, + "step": 24326 + }, + { + "epoch": 8.98, + "learning_rate": 2.102956029067619e-06, + "loss": 0.2769, + "step": 24327 + }, + { + "epoch": 8.98, + "learning_rate": 2.102919078704274e-06, + "loss": 0.3108, + "step": 24328 + }, + { + "epoch": 8.98, + "learning_rate": 2.102882128340929e-06, + "loss": 0.4662, + "step": 24329 + }, + { + "epoch": 8.98, + "learning_rate": 2.1028451779775835e-06, + "loss": 0.4546, + "step": 24330 + }, + { + "epoch": 8.98, + "learning_rate": 2.102808227614238e-06, + "loss": 0.5456, + "step": 24331 + }, + { + "epoch": 8.99, + "learning_rate": 2.102771277250893e-06, + "loss": 0.3015, + "step": 24332 + }, + { + "epoch": 8.99, + "learning_rate": 2.102734326887548e-06, + "loss": 0.3852, + "step": 24333 + }, + { + "epoch": 8.99, + "learning_rate": 2.102697376524203e-06, + "loss": 0.3455, + "step": 24334 + }, + { + "epoch": 8.99, + "learning_rate": 2.1026604261608575e-06, + "loss": 0.4434, + "step": 24335 + }, + { + "epoch": 8.99, + "learning_rate": 2.1026234757975117e-06, + "loss": 0.2741, + "step": 24336 + }, + { + "epoch": 8.99, + "learning_rate": 2.1025865254341668e-06, + "loss": 0.4605, + "step": 24337 + }, + { + "epoch": 8.99, + "learning_rate": 2.1025495750708214e-06, + "loss": 0.2922, + "step": 24338 + }, + { + "epoch": 8.99, + "learning_rate": 2.1025126247074765e-06, + "loss": 0.33, + "step": 24339 + }, + { + "epoch": 8.99, + "learning_rate": 2.102475674344131e-06, + "loss": 0.3616, + "step": 24340 + }, + { + "epoch": 8.99, + "learning_rate": 2.102438723980786e-06, + "loss": 0.3955, + "step": 24341 + }, + { + "epoch": 8.99, + "learning_rate": 2.1024017736174404e-06, + "loss": 0.5373, + "step": 24342 + }, + { + "epoch": 8.99, + "learning_rate": 2.1023648232540954e-06, + "loss": 0.2479, + "step": 24343 + }, + { + "epoch": 8.99, + "learning_rate": 2.10232787289075e-06, + "loss": 0.3515, + "step": 24344 + }, + { + "epoch": 8.99, + "learning_rate": 2.102290922527405e-06, + "loss": 0.4535, + "step": 24345 + }, + { + "epoch": 8.99, + "learning_rate": 2.1022539721640598e-06, + "loss": 0.3518, + "step": 24346 + }, + { + "epoch": 8.99, + "learning_rate": 2.1022170218007144e-06, + "loss": 0.48, + "step": 24347 + }, + { + "epoch": 8.99, + "learning_rate": 2.102180071437369e-06, + "loss": 0.3004, + "step": 24348 + }, + { + "epoch": 8.99, + "learning_rate": 2.1021431210740237e-06, + "loss": 0.4102, + "step": 24349 + }, + { + "epoch": 8.99, + "learning_rate": 2.1021061707106787e-06, + "loss": 0.4121, + "step": 24350 + }, + { + "epoch": 8.99, + "learning_rate": 2.1020692203473334e-06, + "loss": 0.3455, + "step": 24351 + }, + { + "epoch": 8.99, + "learning_rate": 2.1020322699839884e-06, + "loss": 0.5112, + "step": 24352 + }, + { + "epoch": 8.99, + "learning_rate": 2.101995319620643e-06, + "loss": 0.3211, + "step": 24353 + }, + { + "epoch": 8.99, + "learning_rate": 2.1019583692572977e-06, + "loss": 0.3319, + "step": 24354 + }, + { + "epoch": 8.99, + "learning_rate": 2.1019214188939524e-06, + "loss": 0.293, + "step": 24355 + }, + { + "epoch": 8.99, + "learning_rate": 2.101884468530607e-06, + "loss": 0.4047, + "step": 24356 + }, + { + "epoch": 8.99, + "learning_rate": 2.101847518167262e-06, + "loss": 0.3661, + "step": 24357 + }, + { + "epoch": 8.99, + "learning_rate": 2.1018105678039167e-06, + "loss": 0.3611, + "step": 24358 + }, + { + "epoch": 9.0, + "learning_rate": 2.1017736174405717e-06, + "loss": 0.4187, + "step": 24359 + }, + { + "epoch": 9.0, + "learning_rate": 2.1017366670772264e-06, + "loss": 0.5498, + "step": 24360 + }, + { + "epoch": 9.0, + "learning_rate": 2.101699716713881e-06, + "loss": 0.4043, + "step": 24361 + }, + { + "epoch": 9.0, + "learning_rate": 2.1016627663505357e-06, + "loss": 0.4022, + "step": 24362 + }, + { + "epoch": 9.0, + "learning_rate": 2.1016258159871907e-06, + "loss": 0.3928, + "step": 24363 + }, + { + "epoch": 9.0, + "learning_rate": 2.1015888656238454e-06, + "loss": 0.3348, + "step": 24364 + }, + { + "epoch": 9.0, + "learning_rate": 2.1015519152605004e-06, + "loss": 0.3626, + "step": 24365 + }, + { + "epoch": 9.0, + "learning_rate": 2.101514964897155e-06, + "loss": 0.3605, + "step": 24366 + }, + { + "epoch": 9.0, + "learning_rate": 2.1014780145338097e-06, + "loss": 0.497, + "step": 24367 + }, + { + "epoch": 9.0, + "learning_rate": 2.1014410641704643e-06, + "loss": 0.4131, + "step": 24368 + }, + { + "epoch": 9.0, + "learning_rate": 2.101404113807119e-06, + "loss": 0.4181, + "step": 24369 + }, + { + "epoch": 9.0, + "learning_rate": 2.101367163443774e-06, + "loss": 0.3027, + "step": 24370 + }, + { + "epoch": 9.0, + "learning_rate": 2.1013302130804287e-06, + "loss": 0.5096, + "step": 24371 + }, + { + "epoch": 9.0, + "learning_rate": 2.1012932627170837e-06, + "loss": 0.1489, + "step": 24372 + }, + { + "epoch": 9.0, + "eval_accuracy": 0.8014960520847763, + "eval_loss": 0.44180554151535034, + "eval_runtime": 25.273, + "eval_samples_per_second": 1713.845, + "eval_steps_per_second": 26.787, + "step": 24372 + }, + { + "epoch": 9.0, + "learning_rate": 2.1012563123537383e-06, + "loss": 0.3841, + "step": 24373 + }, + { + "epoch": 9.0, + "learning_rate": 2.101219361990393e-06, + "loss": 0.5108, + "step": 24374 + }, + { + "epoch": 9.0, + "learning_rate": 2.1011824116270476e-06, + "loss": 0.3088, + "step": 24375 + }, + { + "epoch": 9.0, + "learning_rate": 2.1011454612637023e-06, + "loss": 0.2991, + "step": 24376 + }, + { + "epoch": 9.0, + "learning_rate": 2.1011085109003573e-06, + "loss": 0.5314, + "step": 24377 + }, + { + "epoch": 9.0, + "learning_rate": 2.101071560537012e-06, + "loss": 0.3145, + "step": 24378 + }, + { + "epoch": 9.0, + "learning_rate": 2.101034610173667e-06, + "loss": 0.3747, + "step": 24379 + }, + { + "epoch": 9.0, + "learning_rate": 2.1009976598103212e-06, + "loss": 0.3269, + "step": 24380 + }, + { + "epoch": 9.0, + "learning_rate": 2.1009607094469763e-06, + "loss": 0.4605, + "step": 24381 + }, + { + "epoch": 9.0, + "learning_rate": 2.100923759083631e-06, + "loss": 0.2659, + "step": 24382 + }, + { + "epoch": 9.0, + "learning_rate": 2.100886808720286e-06, + "loss": 0.3745, + "step": 24383 + }, + { + "epoch": 9.0, + "learning_rate": 2.1008498583569406e-06, + "loss": 0.3535, + "step": 24384 + }, + { + "epoch": 9.0, + "learning_rate": 2.1008129079935957e-06, + "loss": 0.2938, + "step": 24385 + }, + { + "epoch": 9.01, + "learning_rate": 2.10077595763025e-06, + "loss": 0.3846, + "step": 24386 + }, + { + "epoch": 9.01, + "learning_rate": 2.1007390072669045e-06, + "loss": 0.3259, + "step": 24387 + }, + { + "epoch": 9.01, + "learning_rate": 2.1007020569035596e-06, + "loss": 0.3392, + "step": 24388 + }, + { + "epoch": 9.01, + "learning_rate": 2.1006651065402142e-06, + "loss": 0.3729, + "step": 24389 + }, + { + "epoch": 9.01, + "learning_rate": 2.1006281561768693e-06, + "loss": 0.4682, + "step": 24390 + }, + { + "epoch": 9.01, + "learning_rate": 2.100591205813524e-06, + "loss": 0.4034, + "step": 24391 + }, + { + "epoch": 9.01, + "learning_rate": 2.1005542554501786e-06, + "loss": 0.4245, + "step": 24392 + }, + { + "epoch": 9.01, + "learning_rate": 2.100517305086833e-06, + "loss": 0.4345, + "step": 24393 + }, + { + "epoch": 9.01, + "learning_rate": 2.1004803547234883e-06, + "loss": 0.3494, + "step": 24394 + }, + { + "epoch": 9.01, + "learning_rate": 2.100443404360143e-06, + "loss": 0.4381, + "step": 24395 + }, + { + "epoch": 9.01, + "learning_rate": 2.1004064539967975e-06, + "loss": 0.3063, + "step": 24396 + }, + { + "epoch": 9.01, + "learning_rate": 2.1003695036334526e-06, + "loss": 0.4526, + "step": 24397 + }, + { + "epoch": 9.01, + "learning_rate": 2.1003325532701072e-06, + "loss": 0.345, + "step": 24398 + }, + { + "epoch": 9.01, + "learning_rate": 2.100295602906762e-06, + "loss": 0.2949, + "step": 24399 + }, + { + "epoch": 9.01, + "learning_rate": 2.1002586525434165e-06, + "loss": 0.3752, + "step": 24400 + }, + { + "epoch": 9.01, + "learning_rate": 2.1002217021800716e-06, + "loss": 0.2795, + "step": 24401 + }, + { + "epoch": 9.01, + "learning_rate": 2.100184751816726e-06, + "loss": 0.5025, + "step": 24402 + }, + { + "epoch": 9.01, + "learning_rate": 2.1001478014533813e-06, + "loss": 0.3997, + "step": 24403 + }, + { + "epoch": 9.01, + "learning_rate": 2.100110851090036e-06, + "loss": 0.258, + "step": 24404 + }, + { + "epoch": 9.01, + "learning_rate": 2.1000739007266905e-06, + "loss": 0.394, + "step": 24405 + }, + { + "epoch": 9.01, + "learning_rate": 2.100036950363345e-06, + "loss": 0.3676, + "step": 24406 + }, + { + "epoch": 9.01, + "learning_rate": 2.1e-06, + "loss": 0.3022, + "step": 24407 + }, + { + "epoch": 9.01, + "learning_rate": 2.099963049636655e-06, + "loss": 0.3321, + "step": 24408 + }, + { + "epoch": 9.01, + "learning_rate": 2.0999260992733095e-06, + "loss": 0.3177, + "step": 24409 + }, + { + "epoch": 9.01, + "learning_rate": 2.0998891489099646e-06, + "loss": 0.4817, + "step": 24410 + }, + { + "epoch": 9.01, + "learning_rate": 2.099852198546619e-06, + "loss": 0.4764, + "step": 24411 + }, + { + "epoch": 9.01, + "learning_rate": 2.099815248183274e-06, + "loss": 0.3154, + "step": 24412 + }, + { + "epoch": 9.02, + "learning_rate": 2.0997782978199285e-06, + "loss": 0.2975, + "step": 24413 + }, + { + "epoch": 9.02, + "learning_rate": 2.0997413474565835e-06, + "loss": 0.3189, + "step": 24414 + }, + { + "epoch": 9.02, + "learning_rate": 2.099704397093238e-06, + "loss": 0.3709, + "step": 24415 + }, + { + "epoch": 9.02, + "learning_rate": 2.099667446729893e-06, + "loss": 0.4151, + "step": 24416 + }, + { + "epoch": 9.02, + "learning_rate": 2.099630496366548e-06, + "loss": 0.4848, + "step": 24417 + }, + { + "epoch": 9.02, + "learning_rate": 2.0995935460032025e-06, + "loss": 0.3993, + "step": 24418 + }, + { + "epoch": 9.02, + "learning_rate": 2.099556595639857e-06, + "loss": 0.3078, + "step": 24419 + }, + { + "epoch": 9.02, + "learning_rate": 2.0995196452765118e-06, + "loss": 0.2505, + "step": 24420 + }, + { + "epoch": 9.02, + "learning_rate": 2.099482694913167e-06, + "loss": 0.3923, + "step": 24421 + }, + { + "epoch": 9.02, + "learning_rate": 2.0994457445498215e-06, + "loss": 0.4413, + "step": 24422 + }, + { + "epoch": 9.02, + "learning_rate": 2.0994087941864765e-06, + "loss": 0.3373, + "step": 24423 + }, + { + "epoch": 9.02, + "learning_rate": 2.099371843823131e-06, + "loss": 0.44, + "step": 24424 + }, + { + "epoch": 9.02, + "learning_rate": 2.0993348934597854e-06, + "loss": 0.3601, + "step": 24425 + }, + { + "epoch": 9.02, + "learning_rate": 2.0992979430964404e-06, + "loss": 0.4309, + "step": 24426 + }, + { + "epoch": 9.02, + "learning_rate": 2.099260992733095e-06, + "loss": 0.376, + "step": 24427 + }, + { + "epoch": 9.02, + "learning_rate": 2.09922404236975e-06, + "loss": 0.3722, + "step": 24428 + }, + { + "epoch": 9.02, + "learning_rate": 2.0991870920064048e-06, + "loss": 0.369, + "step": 24429 + }, + { + "epoch": 9.02, + "learning_rate": 2.09915014164306e-06, + "loss": 0.4422, + "step": 24430 + }, + { + "epoch": 9.02, + "learning_rate": 2.099113191279714e-06, + "loss": 0.4581, + "step": 24431 + }, + { + "epoch": 9.02, + "learning_rate": 2.099076240916369e-06, + "loss": 0.4566, + "step": 24432 + }, + { + "epoch": 9.02, + "learning_rate": 2.0990392905530237e-06, + "loss": 0.2117, + "step": 24433 + }, + { + "epoch": 9.02, + "learning_rate": 2.099002340189679e-06, + "loss": 0.4726, + "step": 24434 + }, + { + "epoch": 9.02, + "learning_rate": 2.0989653898263334e-06, + "loss": 0.3449, + "step": 24435 + }, + { + "epoch": 9.02, + "learning_rate": 2.098928439462988e-06, + "loss": 0.4077, + "step": 24436 + }, + { + "epoch": 9.02, + "learning_rate": 2.0988914890996427e-06, + "loss": 0.4062, + "step": 24437 + }, + { + "epoch": 9.02, + "learning_rate": 2.0988545387362973e-06, + "loss": 0.4412, + "step": 24438 + }, + { + "epoch": 9.02, + "learning_rate": 2.0988175883729524e-06, + "loss": 0.3642, + "step": 24439 + }, + { + "epoch": 9.03, + "learning_rate": 2.098780638009607e-06, + "loss": 0.3528, + "step": 24440 + }, + { + "epoch": 9.03, + "learning_rate": 2.098743687646262e-06, + "loss": 0.4238, + "step": 24441 + }, + { + "epoch": 9.03, + "learning_rate": 2.0987067372829167e-06, + "loss": 0.3897, + "step": 24442 + }, + { + "epoch": 9.03, + "learning_rate": 2.0986697869195714e-06, + "loss": 0.4196, + "step": 24443 + }, + { + "epoch": 9.03, + "learning_rate": 2.098632836556226e-06, + "loss": 0.2424, + "step": 24444 + }, + { + "epoch": 9.03, + "learning_rate": 2.0985958861928806e-06, + "loss": 0.3214, + "step": 24445 + }, + { + "epoch": 9.03, + "learning_rate": 2.0985589358295357e-06, + "loss": 0.3123, + "step": 24446 + }, + { + "epoch": 9.03, + "learning_rate": 2.0985219854661903e-06, + "loss": 0.5298, + "step": 24447 + }, + { + "epoch": 9.03, + "learning_rate": 2.0984850351028454e-06, + "loss": 0.4634, + "step": 24448 + }, + { + "epoch": 9.03, + "learning_rate": 2.0984480847395e-06, + "loss": 0.3447, + "step": 24449 + }, + { + "epoch": 9.03, + "learning_rate": 2.0984111343761547e-06, + "loss": 0.3426, + "step": 24450 + }, + { + "epoch": 9.03, + "learning_rate": 2.0983741840128093e-06, + "loss": 0.3496, + "step": 24451 + }, + { + "epoch": 9.03, + "learning_rate": 2.0983372336494644e-06, + "loss": 0.4256, + "step": 24452 + }, + { + "epoch": 9.03, + "learning_rate": 2.098300283286119e-06, + "loss": 0.2928, + "step": 24453 + }, + { + "epoch": 9.03, + "learning_rate": 2.098263332922774e-06, + "loss": 0.3901, + "step": 24454 + }, + { + "epoch": 9.03, + "learning_rate": 2.0982263825594287e-06, + "loss": 0.44, + "step": 24455 + }, + { + "epoch": 9.03, + "learning_rate": 2.0981894321960833e-06, + "loss": 0.3352, + "step": 24456 + }, + { + "epoch": 9.03, + "learning_rate": 2.098152481832738e-06, + "loss": 0.4358, + "step": 24457 + }, + { + "epoch": 9.03, + "learning_rate": 2.0981155314693926e-06, + "loss": 0.5113, + "step": 24458 + }, + { + "epoch": 9.03, + "learning_rate": 2.0980785811060477e-06, + "loss": 0.3, + "step": 24459 + }, + { + "epoch": 9.03, + "learning_rate": 2.0980416307427023e-06, + "loss": 0.3054, + "step": 24460 + }, + { + "epoch": 9.03, + "learning_rate": 2.0980046803793574e-06, + "loss": 0.2756, + "step": 24461 + }, + { + "epoch": 9.03, + "learning_rate": 2.097967730016012e-06, + "loss": 0.2608, + "step": 24462 + }, + { + "epoch": 9.03, + "learning_rate": 2.0979307796526666e-06, + "loss": 0.3186, + "step": 24463 + }, + { + "epoch": 9.03, + "learning_rate": 2.0978938292893213e-06, + "loss": 0.299, + "step": 24464 + }, + { + "epoch": 9.03, + "learning_rate": 2.097856878925976e-06, + "loss": 0.4428, + "step": 24465 + }, + { + "epoch": 9.03, + "learning_rate": 2.097819928562631e-06, + "loss": 0.3643, + "step": 24466 + }, + { + "epoch": 9.04, + "learning_rate": 2.0977829781992856e-06, + "loss": 0.3741, + "step": 24467 + }, + { + "epoch": 9.04, + "learning_rate": 2.0977460278359407e-06, + "loss": 0.3366, + "step": 24468 + }, + { + "epoch": 9.04, + "learning_rate": 2.0977090774725953e-06, + "loss": 0.3191, + "step": 24469 + }, + { + "epoch": 9.04, + "learning_rate": 2.09767212710925e-06, + "loss": 0.2665, + "step": 24470 + }, + { + "epoch": 9.04, + "learning_rate": 2.0976351767459046e-06, + "loss": 0.3286, + "step": 24471 + }, + { + "epoch": 9.04, + "learning_rate": 2.0975982263825596e-06, + "loss": 0.4165, + "step": 24472 + }, + { + "epoch": 9.04, + "learning_rate": 2.0975612760192143e-06, + "loss": 0.2322, + "step": 24473 + }, + { + "epoch": 9.04, + "learning_rate": 2.0975243256558693e-06, + "loss": 0.4444, + "step": 24474 + }, + { + "epoch": 9.04, + "learning_rate": 2.097487375292524e-06, + "loss": 0.3687, + "step": 24475 + }, + { + "epoch": 9.04, + "learning_rate": 2.097450424929178e-06, + "loss": 0.4342, + "step": 24476 + }, + { + "epoch": 9.04, + "learning_rate": 2.0974134745658332e-06, + "loss": 0.31, + "step": 24477 + }, + { + "epoch": 9.04, + "learning_rate": 2.097376524202488e-06, + "loss": 0.347, + "step": 24478 + }, + { + "epoch": 9.04, + "learning_rate": 2.097339573839143e-06, + "loss": 0.3362, + "step": 24479 + }, + { + "epoch": 9.04, + "learning_rate": 2.0973026234757976e-06, + "loss": 0.3635, + "step": 24480 + }, + { + "epoch": 9.04, + "learning_rate": 2.0972656731124526e-06, + "loss": 0.4592, + "step": 24481 + }, + { + "epoch": 9.04, + "learning_rate": 2.097228722749107e-06, + "loss": 0.2616, + "step": 24482 + }, + { + "epoch": 9.04, + "learning_rate": 2.097191772385762e-06, + "loss": 0.5363, + "step": 24483 + }, + { + "epoch": 9.04, + "learning_rate": 2.0971548220224165e-06, + "loss": 0.4189, + "step": 24484 + }, + { + "epoch": 9.04, + "learning_rate": 2.097117871659071e-06, + "loss": 0.5428, + "step": 24485 + }, + { + "epoch": 9.04, + "learning_rate": 2.0970809212957262e-06, + "loss": 0.4454, + "step": 24486 + }, + { + "epoch": 9.04, + "learning_rate": 2.097043970932381e-06, + "loss": 0.364, + "step": 24487 + }, + { + "epoch": 9.04, + "learning_rate": 2.0970070205690355e-06, + "loss": 0.2644, + "step": 24488 + }, + { + "epoch": 9.04, + "learning_rate": 2.09697007020569e-06, + "loss": 0.3036, + "step": 24489 + }, + { + "epoch": 9.04, + "learning_rate": 2.096933119842345e-06, + "loss": 0.2789, + "step": 24490 + }, + { + "epoch": 9.04, + "learning_rate": 2.096896169479e-06, + "loss": 0.3615, + "step": 24491 + }, + { + "epoch": 9.04, + "learning_rate": 2.096859219115655e-06, + "loss": 0.3724, + "step": 24492 + }, + { + "epoch": 9.04, + "learning_rate": 2.0968222687523095e-06, + "loss": 0.3436, + "step": 24493 + }, + { + "epoch": 9.05, + "learning_rate": 2.096785318388964e-06, + "loss": 0.3645, + "step": 24494 + }, + { + "epoch": 9.05, + "learning_rate": 2.096748368025619e-06, + "loss": 0.3783, + "step": 24495 + }, + { + "epoch": 9.05, + "learning_rate": 2.0967114176622735e-06, + "loss": 0.3887, + "step": 24496 + }, + { + "epoch": 9.05, + "learning_rate": 2.0966744672989285e-06, + "loss": 0.387, + "step": 24497 + }, + { + "epoch": 9.05, + "learning_rate": 2.096637516935583e-06, + "loss": 0.4146, + "step": 24498 + }, + { + "epoch": 9.05, + "learning_rate": 2.096600566572238e-06, + "loss": 0.3152, + "step": 24499 + }, + { + "epoch": 9.05, + "learning_rate": 2.096563616208893e-06, + "loss": 0.4549, + "step": 24500 + }, + { + "epoch": 9.05, + "learning_rate": 2.0965266658455475e-06, + "loss": 0.3552, + "step": 24501 + }, + { + "epoch": 9.05, + "learning_rate": 2.096489715482202e-06, + "loss": 0.404, + "step": 24502 + }, + { + "epoch": 9.05, + "learning_rate": 2.096452765118857e-06, + "loss": 0.4974, + "step": 24503 + }, + { + "epoch": 9.05, + "learning_rate": 2.096415814755512e-06, + "loss": 0.3976, + "step": 24504 + }, + { + "epoch": 9.05, + "learning_rate": 2.0963788643921665e-06, + "loss": 0.5276, + "step": 24505 + }, + { + "epoch": 9.05, + "learning_rate": 2.0963419140288215e-06, + "loss": 0.366, + "step": 24506 + }, + { + "epoch": 9.05, + "learning_rate": 2.096304963665476e-06, + "loss": 0.351, + "step": 24507 + }, + { + "epoch": 9.05, + "learning_rate": 2.0962680133021308e-06, + "loss": 0.2411, + "step": 24508 + }, + { + "epoch": 9.05, + "learning_rate": 2.0962310629387854e-06, + "loss": 0.3144, + "step": 24509 + }, + { + "epoch": 9.05, + "learning_rate": 2.0961941125754405e-06, + "loss": 0.3251, + "step": 24510 + }, + { + "epoch": 9.05, + "learning_rate": 2.096157162212095e-06, + "loss": 0.4039, + "step": 24511 + }, + { + "epoch": 9.05, + "learning_rate": 2.09612021184875e-06, + "loss": 0.3234, + "step": 24512 + }, + { + "epoch": 9.05, + "learning_rate": 2.096083261485405e-06, + "loss": 0.2979, + "step": 24513 + }, + { + "epoch": 9.05, + "learning_rate": 2.0960463111220595e-06, + "loss": 0.335, + "step": 24514 + }, + { + "epoch": 9.05, + "learning_rate": 2.096009360758714e-06, + "loss": 0.4089, + "step": 24515 + }, + { + "epoch": 9.05, + "learning_rate": 2.0959724103953687e-06, + "loss": 0.2619, + "step": 24516 + }, + { + "epoch": 9.05, + "learning_rate": 2.0959354600320238e-06, + "loss": 0.2994, + "step": 24517 + }, + { + "epoch": 9.05, + "learning_rate": 2.0958985096686784e-06, + "loss": 0.3554, + "step": 24518 + }, + { + "epoch": 9.05, + "learning_rate": 2.0958615593053335e-06, + "loss": 0.2431, + "step": 24519 + }, + { + "epoch": 9.05, + "learning_rate": 2.095824608941988e-06, + "loss": 0.3487, + "step": 24520 + }, + { + "epoch": 9.06, + "learning_rate": 2.0957876585786428e-06, + "loss": 0.3529, + "step": 24521 + }, + { + "epoch": 9.06, + "learning_rate": 2.0957507082152974e-06, + "loss": 0.4582, + "step": 24522 + }, + { + "epoch": 9.06, + "learning_rate": 2.0957137578519524e-06, + "loss": 0.4734, + "step": 24523 + }, + { + "epoch": 9.06, + "learning_rate": 2.095676807488607e-06, + "loss": 0.3581, + "step": 24524 + }, + { + "epoch": 9.06, + "learning_rate": 2.0956398571252617e-06, + "loss": 0.3535, + "step": 24525 + }, + { + "epoch": 9.06, + "learning_rate": 2.0956029067619168e-06, + "loss": 0.3284, + "step": 24526 + }, + { + "epoch": 9.06, + "learning_rate": 2.095565956398571e-06, + "loss": 0.3, + "step": 24527 + }, + { + "epoch": 9.06, + "learning_rate": 2.095529006035226e-06, + "loss": 0.3075, + "step": 24528 + }, + { + "epoch": 9.06, + "learning_rate": 2.0954920556718807e-06, + "loss": 0.336, + "step": 24529 + }, + { + "epoch": 9.06, + "learning_rate": 2.0954551053085358e-06, + "loss": 0.3616, + "step": 24530 + }, + { + "epoch": 9.06, + "learning_rate": 2.0954181549451904e-06, + "loss": 0.3369, + "step": 24531 + }, + { + "epoch": 9.06, + "learning_rate": 2.095381204581845e-06, + "loss": 0.3745, + "step": 24532 + }, + { + "epoch": 9.06, + "learning_rate": 2.0953442542184997e-06, + "loss": 0.4681, + "step": 24533 + }, + { + "epoch": 9.06, + "learning_rate": 2.0953073038551547e-06, + "loss": 0.2954, + "step": 24534 + }, + { + "epoch": 9.06, + "learning_rate": 2.0952703534918094e-06, + "loss": 0.3635, + "step": 24535 + }, + { + "epoch": 9.06, + "learning_rate": 2.095233403128464e-06, + "loss": 0.3804, + "step": 24536 + }, + { + "epoch": 9.06, + "learning_rate": 2.095196452765119e-06, + "loss": 0.2475, + "step": 24537 + }, + { + "epoch": 9.06, + "learning_rate": 2.0951595024017737e-06, + "loss": 0.3764, + "step": 24538 + }, + { + "epoch": 9.06, + "learning_rate": 2.0951225520384283e-06, + "loss": 0.3008, + "step": 24539 + }, + { + "epoch": 9.06, + "learning_rate": 2.095085601675083e-06, + "loss": 0.3801, + "step": 24540 + }, + { + "epoch": 9.06, + "learning_rate": 2.095048651311738e-06, + "loss": 0.4219, + "step": 24541 + }, + { + "epoch": 9.06, + "learning_rate": 2.0950117009483927e-06, + "loss": 0.2722, + "step": 24542 + }, + { + "epoch": 9.06, + "learning_rate": 2.0949747505850477e-06, + "loss": 0.3476, + "step": 24543 + }, + { + "epoch": 9.06, + "learning_rate": 2.0949378002217024e-06, + "loss": 0.2993, + "step": 24544 + }, + { + "epoch": 9.06, + "learning_rate": 2.094900849858357e-06, + "loss": 0.3501, + "step": 24545 + }, + { + "epoch": 9.06, + "learning_rate": 2.0948638994950116e-06, + "loss": 0.2773, + "step": 24546 + }, + { + "epoch": 9.06, + "learning_rate": 2.0948269491316663e-06, + "loss": 0.3849, + "step": 24547 + }, + { + "epoch": 9.06, + "learning_rate": 2.0947899987683213e-06, + "loss": 0.3469, + "step": 24548 + }, + { + "epoch": 9.07, + "learning_rate": 2.094753048404976e-06, + "loss": 0.4105, + "step": 24549 + }, + { + "epoch": 9.07, + "learning_rate": 2.094716098041631e-06, + "loss": 0.4228, + "step": 24550 + }, + { + "epoch": 9.07, + "learning_rate": 2.0946791476782857e-06, + "loss": 0.418, + "step": 24551 + }, + { + "epoch": 9.07, + "learning_rate": 2.0946421973149403e-06, + "loss": 0.4318, + "step": 24552 + }, + { + "epoch": 9.07, + "learning_rate": 2.094605246951595e-06, + "loss": 0.3979, + "step": 24553 + }, + { + "epoch": 9.07, + "learning_rate": 2.09456829658825e-06, + "loss": 0.331, + "step": 24554 + }, + { + "epoch": 9.07, + "learning_rate": 2.0945313462249046e-06, + "loss": 0.4408, + "step": 24555 + }, + { + "epoch": 9.07, + "learning_rate": 2.0944943958615593e-06, + "loss": 0.3749, + "step": 24556 + }, + { + "epoch": 9.07, + "learning_rate": 2.0944574454982143e-06, + "loss": 0.4315, + "step": 24557 + }, + { + "epoch": 9.07, + "learning_rate": 2.094420495134869e-06, + "loss": 0.3423, + "step": 24558 + }, + { + "epoch": 9.07, + "learning_rate": 2.0943835447715236e-06, + "loss": 0.3107, + "step": 24559 + }, + { + "epoch": 9.07, + "learning_rate": 2.0943465944081782e-06, + "loss": 0.2701, + "step": 24560 + }, + { + "epoch": 9.07, + "learning_rate": 2.0943096440448333e-06, + "loss": 0.4314, + "step": 24561 + }, + { + "epoch": 9.07, + "learning_rate": 2.094272693681488e-06, + "loss": 0.4343, + "step": 24562 + }, + { + "epoch": 9.07, + "learning_rate": 2.094235743318143e-06, + "loss": 0.4602, + "step": 24563 + }, + { + "epoch": 9.07, + "learning_rate": 2.0941987929547976e-06, + "loss": 0.2855, + "step": 24564 + }, + { + "epoch": 9.07, + "learning_rate": 2.094161842591452e-06, + "loss": 0.2999, + "step": 24565 + }, + { + "epoch": 9.07, + "learning_rate": 2.094124892228107e-06, + "loss": 0.3379, + "step": 24566 + }, + { + "epoch": 9.07, + "learning_rate": 2.0940879418647615e-06, + "loss": 0.4553, + "step": 24567 + }, + { + "epoch": 9.07, + "learning_rate": 2.0940509915014166e-06, + "loss": 0.3174, + "step": 24568 + }, + { + "epoch": 9.07, + "learning_rate": 2.0940140411380712e-06, + "loss": 0.3531, + "step": 24569 + }, + { + "epoch": 9.07, + "learning_rate": 2.0939770907747263e-06, + "loss": 0.33, + "step": 24570 + }, + { + "epoch": 9.07, + "learning_rate": 2.0939401404113805e-06, + "loss": 0.318, + "step": 24571 + }, + { + "epoch": 9.07, + "learning_rate": 2.0939031900480356e-06, + "loss": 0.3487, + "step": 24572 + }, + { + "epoch": 9.07, + "learning_rate": 2.09386623968469e-06, + "loss": 0.215, + "step": 24573 + }, + { + "epoch": 9.07, + "learning_rate": 2.0938292893213453e-06, + "loss": 0.4436, + "step": 24574 + }, + { + "epoch": 9.07, + "learning_rate": 2.093792338958e-06, + "loss": 0.3018, + "step": 24575 + }, + { + "epoch": 9.08, + "learning_rate": 2.0937553885946545e-06, + "loss": 0.3224, + "step": 24576 + }, + { + "epoch": 9.08, + "learning_rate": 2.093718438231309e-06, + "loss": 0.3152, + "step": 24577 + }, + { + "epoch": 9.08, + "learning_rate": 2.093681487867964e-06, + "loss": 0.3479, + "step": 24578 + }, + { + "epoch": 9.08, + "learning_rate": 2.093644537504619e-06, + "loss": 0.2396, + "step": 24579 + }, + { + "epoch": 9.08, + "learning_rate": 2.0936075871412735e-06, + "loss": 0.3746, + "step": 24580 + }, + { + "epoch": 9.08, + "learning_rate": 2.0935706367779286e-06, + "loss": 0.4313, + "step": 24581 + }, + { + "epoch": 9.08, + "learning_rate": 2.093533686414583e-06, + "loss": 0.5193, + "step": 24582 + }, + { + "epoch": 9.08, + "learning_rate": 2.093496736051238e-06, + "loss": 0.3634, + "step": 24583 + }, + { + "epoch": 9.08, + "learning_rate": 2.0934597856878925e-06, + "loss": 0.3723, + "step": 24584 + }, + { + "epoch": 9.08, + "learning_rate": 2.093422835324547e-06, + "loss": 0.3541, + "step": 24585 + }, + { + "epoch": 9.08, + "learning_rate": 2.093385884961202e-06, + "loss": 0.3833, + "step": 24586 + }, + { + "epoch": 9.08, + "learning_rate": 2.093348934597857e-06, + "loss": 0.3464, + "step": 24587 + }, + { + "epoch": 9.08, + "learning_rate": 2.093311984234512e-06, + "loss": 0.2644, + "step": 24588 + }, + { + "epoch": 9.08, + "learning_rate": 2.0932750338711665e-06, + "loss": 0.371, + "step": 24589 + }, + { + "epoch": 9.08, + "learning_rate": 2.093238083507821e-06, + "loss": 0.3502, + "step": 24590 + }, + { + "epoch": 9.08, + "learning_rate": 2.0932011331444758e-06, + "loss": 0.2417, + "step": 24591 + }, + { + "epoch": 9.08, + "learning_rate": 2.093164182781131e-06, + "loss": 0.3447, + "step": 24592 + }, + { + "epoch": 9.08, + "learning_rate": 2.0931272324177855e-06, + "loss": 0.4294, + "step": 24593 + }, + { + "epoch": 9.08, + "learning_rate": 2.0930902820544405e-06, + "loss": 0.4829, + "step": 24594 + }, + { + "epoch": 9.08, + "learning_rate": 2.093053331691095e-06, + "loss": 0.3632, + "step": 24595 + }, + { + "epoch": 9.08, + "learning_rate": 2.09301638132775e-06, + "loss": 0.4955, + "step": 24596 + }, + { + "epoch": 9.08, + "learning_rate": 2.0929794309644044e-06, + "loss": 0.3862, + "step": 24597 + }, + { + "epoch": 9.08, + "learning_rate": 2.092942480601059e-06, + "loss": 0.3423, + "step": 24598 + }, + { + "epoch": 9.08, + "learning_rate": 2.092905530237714e-06, + "loss": 0.2564, + "step": 24599 + }, + { + "epoch": 9.08, + "learning_rate": 2.0928685798743688e-06, + "loss": 0.3529, + "step": 24600 + }, + { + "epoch": 9.08, + "learning_rate": 2.092831629511024e-06, + "loss": 0.2789, + "step": 24601 + }, + { + "epoch": 9.08, + "learning_rate": 2.0927946791476785e-06, + "loss": 0.5208, + "step": 24602 + }, + { + "epoch": 9.09, + "learning_rate": 2.092757728784333e-06, + "loss": 0.2777, + "step": 24603 + }, + { + "epoch": 9.09, + "learning_rate": 2.0927207784209877e-06, + "loss": 0.3894, + "step": 24604 + }, + { + "epoch": 9.09, + "learning_rate": 2.0926838280576424e-06, + "loss": 0.3738, + "step": 24605 + }, + { + "epoch": 9.09, + "learning_rate": 2.0926468776942974e-06, + "loss": 0.3264, + "step": 24606 + }, + { + "epoch": 9.09, + "learning_rate": 2.092609927330952e-06, + "loss": 0.344, + "step": 24607 + }, + { + "epoch": 9.09, + "learning_rate": 2.092572976967607e-06, + "loss": 0.2719, + "step": 24608 + }, + { + "epoch": 9.09, + "learning_rate": 2.0925360266042618e-06, + "loss": 0.3077, + "step": 24609 + }, + { + "epoch": 9.09, + "learning_rate": 2.0924990762409164e-06, + "loss": 0.3108, + "step": 24610 + }, + { + "epoch": 9.09, + "learning_rate": 2.092462125877571e-06, + "loss": 0.4071, + "step": 24611 + }, + { + "epoch": 9.09, + "learning_rate": 2.092425175514226e-06, + "loss": 0.4433, + "step": 24612 + }, + { + "epoch": 9.09, + "learning_rate": 2.0923882251508807e-06, + "loss": 0.444, + "step": 24613 + }, + { + "epoch": 9.09, + "learning_rate": 2.092351274787536e-06, + "loss": 0.2708, + "step": 24614 + }, + { + "epoch": 9.09, + "learning_rate": 2.0923143244241904e-06, + "loss": 0.3175, + "step": 24615 + }, + { + "epoch": 9.09, + "learning_rate": 2.0922773740608446e-06, + "loss": 0.3664, + "step": 24616 + }, + { + "epoch": 9.09, + "learning_rate": 2.0922404236974997e-06, + "loss": 0.4004, + "step": 24617 + }, + { + "epoch": 9.09, + "learning_rate": 2.0922034733341543e-06, + "loss": 0.4324, + "step": 24618 + }, + { + "epoch": 9.09, + "learning_rate": 2.0921665229708094e-06, + "loss": 0.2913, + "step": 24619 + }, + { + "epoch": 9.09, + "learning_rate": 2.092129572607464e-06, + "loss": 0.5971, + "step": 24620 + }, + { + "epoch": 9.09, + "learning_rate": 2.092092622244119e-06, + "loss": 0.375, + "step": 24621 + }, + { + "epoch": 9.09, + "learning_rate": 2.0920556718807733e-06, + "loss": 0.3761, + "step": 24622 + }, + { + "epoch": 9.09, + "learning_rate": 2.0920187215174284e-06, + "loss": 0.3197, + "step": 24623 + }, + { + "epoch": 9.09, + "learning_rate": 2.091981771154083e-06, + "loss": 0.6034, + "step": 24624 + }, + { + "epoch": 9.09, + "learning_rate": 2.0919448207907376e-06, + "loss": 0.3665, + "step": 24625 + }, + { + "epoch": 9.09, + "learning_rate": 2.0919078704273927e-06, + "loss": 0.4195, + "step": 24626 + }, + { + "epoch": 9.09, + "learning_rate": 2.0918709200640473e-06, + "loss": 0.3331, + "step": 24627 + }, + { + "epoch": 9.09, + "learning_rate": 2.091833969700702e-06, + "loss": 0.3054, + "step": 24628 + }, + { + "epoch": 9.09, + "learning_rate": 2.0917970193373566e-06, + "loss": 0.3894, + "step": 24629 + }, + { + "epoch": 9.1, + "learning_rate": 2.0917600689740117e-06, + "loss": 0.357, + "step": 24630 + }, + { + "epoch": 9.1, + "learning_rate": 2.0917231186106663e-06, + "loss": 0.4451, + "step": 24631 + }, + { + "epoch": 9.1, + "learning_rate": 2.0916861682473214e-06, + "loss": 0.3408, + "step": 24632 + }, + { + "epoch": 9.1, + "learning_rate": 2.091649217883976e-06, + "loss": 0.3849, + "step": 24633 + }, + { + "epoch": 9.1, + "learning_rate": 2.0916122675206306e-06, + "loss": 0.383, + "step": 24634 + }, + { + "epoch": 9.1, + "learning_rate": 2.0915753171572853e-06, + "loss": 0.3712, + "step": 24635 + }, + { + "epoch": 9.1, + "learning_rate": 2.09153836679394e-06, + "loss": 0.3944, + "step": 24636 + }, + { + "epoch": 9.1, + "learning_rate": 2.091501416430595e-06, + "loss": 0.3302, + "step": 24637 + }, + { + "epoch": 9.1, + "learning_rate": 2.0914644660672496e-06, + "loss": 0.2459, + "step": 24638 + }, + { + "epoch": 9.1, + "learning_rate": 2.0914275157039047e-06, + "loss": 0.4495, + "step": 24639 + }, + { + "epoch": 9.1, + "learning_rate": 2.0913905653405593e-06, + "loss": 0.3024, + "step": 24640 + }, + { + "epoch": 9.1, + "learning_rate": 2.091353614977214e-06, + "loss": 0.5391, + "step": 24641 + }, + { + "epoch": 9.1, + "learning_rate": 2.0913166646138686e-06, + "loss": 0.4916, + "step": 24642 + }, + { + "epoch": 9.1, + "learning_rate": 2.0912797142505236e-06, + "loss": 0.3853, + "step": 24643 + }, + { + "epoch": 9.1, + "learning_rate": 2.0912427638871783e-06, + "loss": 0.4334, + "step": 24644 + }, + { + "epoch": 9.1, + "learning_rate": 2.091205813523833e-06, + "loss": 0.243, + "step": 24645 + }, + { + "epoch": 9.1, + "learning_rate": 2.091168863160488e-06, + "loss": 0.3282, + "step": 24646 + }, + { + "epoch": 9.1, + "learning_rate": 2.0911319127971426e-06, + "loss": 0.2686, + "step": 24647 + }, + { + "epoch": 9.1, + "learning_rate": 2.0910949624337972e-06, + "loss": 0.2687, + "step": 24648 + }, + { + "epoch": 9.1, + "learning_rate": 2.091058012070452e-06, + "loss": 0.3665, + "step": 24649 + }, + { + "epoch": 9.1, + "learning_rate": 2.091021061707107e-06, + "loss": 0.3971, + "step": 24650 + }, + { + "epoch": 9.1, + "learning_rate": 2.0909841113437616e-06, + "loss": 0.3029, + "step": 24651 + }, + { + "epoch": 9.1, + "learning_rate": 2.0909471609804166e-06, + "loss": 0.4299, + "step": 24652 + }, + { + "epoch": 9.1, + "learning_rate": 2.0909102106170713e-06, + "loss": 0.218, + "step": 24653 + }, + { + "epoch": 9.1, + "learning_rate": 2.090873260253726e-06, + "loss": 0.3471, + "step": 24654 + }, + { + "epoch": 9.1, + "learning_rate": 2.0908363098903806e-06, + "loss": 0.3007, + "step": 24655 + }, + { + "epoch": 9.1, + "learning_rate": 2.090799359527035e-06, + "loss": 0.4008, + "step": 24656 + }, + { + "epoch": 9.11, + "learning_rate": 2.0907624091636902e-06, + "loss": 0.4345, + "step": 24657 + }, + { + "epoch": 9.11, + "learning_rate": 2.090725458800345e-06, + "loss": 0.3719, + "step": 24658 + }, + { + "epoch": 9.11, + "learning_rate": 2.090688508437e-06, + "loss": 0.3543, + "step": 24659 + }, + { + "epoch": 9.11, + "learning_rate": 2.0906515580736546e-06, + "loss": 0.3122, + "step": 24660 + }, + { + "epoch": 9.11, + "learning_rate": 2.0906146077103092e-06, + "loss": 0.4014, + "step": 24661 + }, + { + "epoch": 9.11, + "learning_rate": 2.090577657346964e-06, + "loss": 0.3367, + "step": 24662 + }, + { + "epoch": 9.11, + "learning_rate": 2.090540706983619e-06, + "loss": 0.3379, + "step": 24663 + }, + { + "epoch": 9.11, + "learning_rate": 2.0905037566202735e-06, + "loss": 0.2514, + "step": 24664 + }, + { + "epoch": 9.11, + "learning_rate": 2.090466806256928e-06, + "loss": 0.3795, + "step": 24665 + }, + { + "epoch": 9.11, + "learning_rate": 2.0904298558935832e-06, + "loss": 0.3071, + "step": 24666 + }, + { + "epoch": 9.11, + "learning_rate": 2.0903929055302375e-06, + "loss": 0.4113, + "step": 24667 + }, + { + "epoch": 9.11, + "learning_rate": 2.0903559551668925e-06, + "loss": 0.4548, + "step": 24668 + }, + { + "epoch": 9.11, + "learning_rate": 2.090319004803547e-06, + "loss": 0.3375, + "step": 24669 + }, + { + "epoch": 9.11, + "learning_rate": 2.0902820544402022e-06, + "loss": 0.4011, + "step": 24670 + }, + { + "epoch": 9.11, + "learning_rate": 2.090245104076857e-06, + "loss": 0.2826, + "step": 24671 + }, + { + "epoch": 9.11, + "learning_rate": 2.090208153713512e-06, + "loss": 0.3268, + "step": 24672 + }, + { + "epoch": 9.11, + "learning_rate": 2.090171203350166e-06, + "loss": 0.3443, + "step": 24673 + }, + { + "epoch": 9.11, + "learning_rate": 2.0901342529868208e-06, + "loss": 0.3611, + "step": 24674 + }, + { + "epoch": 9.11, + "learning_rate": 2.090097302623476e-06, + "loss": 0.4133, + "step": 24675 + }, + { + "epoch": 9.11, + "learning_rate": 2.0900603522601305e-06, + "loss": 0.4216, + "step": 24676 + }, + { + "epoch": 9.11, + "learning_rate": 2.0900234018967855e-06, + "loss": 0.3511, + "step": 24677 + }, + { + "epoch": 9.11, + "learning_rate": 2.08998645153344e-06, + "loss": 0.4464, + "step": 24678 + }, + { + "epoch": 9.11, + "learning_rate": 2.089949501170095e-06, + "loss": 0.4111, + "step": 24679 + }, + { + "epoch": 9.11, + "learning_rate": 2.0899125508067494e-06, + "loss": 0.3265, + "step": 24680 + }, + { + "epoch": 9.11, + "learning_rate": 2.0898756004434045e-06, + "loss": 0.2286, + "step": 24681 + }, + { + "epoch": 9.11, + "learning_rate": 2.089838650080059e-06, + "loss": 0.3315, + "step": 24682 + }, + { + "epoch": 9.11, + "learning_rate": 2.089801699716714e-06, + "loss": 0.3145, + "step": 24683 + }, + { + "epoch": 9.12, + "learning_rate": 2.089764749353369e-06, + "loss": 0.4368, + "step": 24684 + }, + { + "epoch": 9.12, + "learning_rate": 2.0897277989900235e-06, + "loss": 0.5114, + "step": 24685 + }, + { + "epoch": 9.12, + "learning_rate": 2.089690848626678e-06, + "loss": 0.3963, + "step": 24686 + }, + { + "epoch": 9.12, + "learning_rate": 2.0896538982633327e-06, + "loss": 0.3636, + "step": 24687 + }, + { + "epoch": 9.12, + "learning_rate": 2.089616947899988e-06, + "loss": 0.2935, + "step": 24688 + }, + { + "epoch": 9.12, + "learning_rate": 2.0895799975366424e-06, + "loss": 0.4021, + "step": 24689 + }, + { + "epoch": 9.12, + "learning_rate": 2.0895430471732975e-06, + "loss": 0.4066, + "step": 24690 + }, + { + "epoch": 9.12, + "learning_rate": 2.089506096809952e-06, + "loss": 0.2715, + "step": 24691 + }, + { + "epoch": 9.12, + "learning_rate": 2.0894691464466068e-06, + "loss": 0.3958, + "step": 24692 + }, + { + "epoch": 9.12, + "learning_rate": 2.0894321960832614e-06, + "loss": 0.4232, + "step": 24693 + }, + { + "epoch": 9.12, + "learning_rate": 2.089395245719916e-06, + "loss": 0.3027, + "step": 24694 + }, + { + "epoch": 9.12, + "learning_rate": 2.089358295356571e-06, + "loss": 0.2965, + "step": 24695 + }, + { + "epoch": 9.12, + "learning_rate": 2.0893213449932257e-06, + "loss": 0.3071, + "step": 24696 + }, + { + "epoch": 9.12, + "learning_rate": 2.0892843946298808e-06, + "loss": 0.362, + "step": 24697 + }, + { + "epoch": 9.12, + "learning_rate": 2.0892474442665354e-06, + "loss": 0.3598, + "step": 24698 + }, + { + "epoch": 9.12, + "learning_rate": 2.08921049390319e-06, + "loss": 0.4441, + "step": 24699 + }, + { + "epoch": 9.12, + "learning_rate": 2.0891735435398447e-06, + "loss": 0.4865, + "step": 24700 + }, + { + "epoch": 9.12, + "learning_rate": 2.0891365931764998e-06, + "loss": 0.2994, + "step": 24701 + }, + { + "epoch": 9.12, + "learning_rate": 2.0890996428131544e-06, + "loss": 0.507, + "step": 24702 + }, + { + "epoch": 9.12, + "learning_rate": 2.0890626924498095e-06, + "loss": 0.4124, + "step": 24703 + }, + { + "epoch": 9.12, + "learning_rate": 2.089025742086464e-06, + "loss": 0.2477, + "step": 24704 + }, + { + "epoch": 9.12, + "learning_rate": 2.0889887917231187e-06, + "loss": 0.3524, + "step": 24705 + }, + { + "epoch": 9.12, + "learning_rate": 2.0889518413597734e-06, + "loss": 0.4554, + "step": 24706 + }, + { + "epoch": 9.12, + "learning_rate": 2.088914890996428e-06, + "loss": 0.2967, + "step": 24707 + }, + { + "epoch": 9.12, + "learning_rate": 2.088877940633083e-06, + "loss": 0.3559, + "step": 24708 + }, + { + "epoch": 9.12, + "learning_rate": 2.0888409902697377e-06, + "loss": 0.3867, + "step": 24709 + }, + { + "epoch": 9.12, + "learning_rate": 2.0888040399063928e-06, + "loss": 0.2843, + "step": 24710 + }, + { + "epoch": 9.13, + "learning_rate": 2.0887670895430474e-06, + "loss": 0.301, + "step": 24711 + }, + { + "epoch": 9.13, + "learning_rate": 2.088730139179702e-06, + "loss": 0.4695, + "step": 24712 + }, + { + "epoch": 9.13, + "learning_rate": 2.0886931888163567e-06, + "loss": 0.6385, + "step": 24713 + }, + { + "epoch": 9.13, + "learning_rate": 2.0886562384530113e-06, + "loss": 0.2741, + "step": 24714 + }, + { + "epoch": 9.13, + "learning_rate": 2.0886192880896664e-06, + "loss": 0.4054, + "step": 24715 + }, + { + "epoch": 9.13, + "learning_rate": 2.088582337726321e-06, + "loss": 0.2115, + "step": 24716 + }, + { + "epoch": 9.13, + "learning_rate": 2.0885453873629756e-06, + "loss": 0.2536, + "step": 24717 + }, + { + "epoch": 9.13, + "learning_rate": 2.0885084369996303e-06, + "loss": 0.3772, + "step": 24718 + }, + { + "epoch": 9.13, + "learning_rate": 2.0884714866362853e-06, + "loss": 0.366, + "step": 24719 + }, + { + "epoch": 9.13, + "learning_rate": 2.08843453627294e-06, + "loss": 0.273, + "step": 24720 + }, + { + "epoch": 9.13, + "learning_rate": 2.088397585909595e-06, + "loss": 0.3016, + "step": 24721 + }, + { + "epoch": 9.13, + "learning_rate": 2.0883606355462497e-06, + "loss": 0.2487, + "step": 24722 + }, + { + "epoch": 9.13, + "learning_rate": 2.0883236851829043e-06, + "loss": 0.3226, + "step": 24723 + }, + { + "epoch": 9.13, + "learning_rate": 2.088286734819559e-06, + "loss": 0.4165, + "step": 24724 + }, + { + "epoch": 9.13, + "learning_rate": 2.0882497844562136e-06, + "loss": 0.3193, + "step": 24725 + }, + { + "epoch": 9.13, + "learning_rate": 2.0882128340928686e-06, + "loss": 0.3831, + "step": 24726 + }, + { + "epoch": 9.13, + "learning_rate": 2.0881758837295233e-06, + "loss": 0.317, + "step": 24727 + }, + { + "epoch": 9.13, + "learning_rate": 2.0881389333661783e-06, + "loss": 0.2411, + "step": 24728 + }, + { + "epoch": 9.13, + "learning_rate": 2.088101983002833e-06, + "loss": 0.2574, + "step": 24729 + }, + { + "epoch": 9.13, + "learning_rate": 2.0880650326394876e-06, + "loss": 0.3743, + "step": 24730 + }, + { + "epoch": 9.13, + "learning_rate": 2.0880280822761422e-06, + "loss": 0.2432, + "step": 24731 + }, + { + "epoch": 9.13, + "learning_rate": 2.0879911319127973e-06, + "loss": 0.4225, + "step": 24732 + }, + { + "epoch": 9.13, + "learning_rate": 2.087954181549452e-06, + "loss": 0.5178, + "step": 24733 + }, + { + "epoch": 9.13, + "learning_rate": 2.0879172311861066e-06, + "loss": 0.3515, + "step": 24734 + }, + { + "epoch": 9.13, + "learning_rate": 2.0878802808227616e-06, + "loss": 0.3132, + "step": 24735 + }, + { + "epoch": 9.13, + "learning_rate": 2.0878433304594163e-06, + "loss": 0.4356, + "step": 24736 + }, + { + "epoch": 9.13, + "learning_rate": 2.087806380096071e-06, + "loss": 0.3362, + "step": 24737 + }, + { + "epoch": 9.14, + "learning_rate": 2.0877694297327255e-06, + "loss": 0.2999, + "step": 24738 + }, + { + "epoch": 9.14, + "learning_rate": 2.0877324793693806e-06, + "loss": 0.3547, + "step": 24739 + }, + { + "epoch": 9.14, + "learning_rate": 2.0876955290060352e-06, + "loss": 0.4774, + "step": 24740 + }, + { + "epoch": 9.14, + "learning_rate": 2.0876585786426903e-06, + "loss": 0.4327, + "step": 24741 + }, + { + "epoch": 9.14, + "learning_rate": 2.087621628279345e-06, + "loss": 0.4023, + "step": 24742 + }, + { + "epoch": 9.14, + "learning_rate": 2.0875846779159996e-06, + "loss": 0.3524, + "step": 24743 + }, + { + "epoch": 9.14, + "learning_rate": 2.087547727552654e-06, + "loss": 0.2647, + "step": 24744 + }, + { + "epoch": 9.14, + "learning_rate": 2.087510777189309e-06, + "loss": 0.3463, + "step": 24745 + }, + { + "epoch": 9.14, + "learning_rate": 2.087473826825964e-06, + "loss": 0.3181, + "step": 24746 + }, + { + "epoch": 9.14, + "learning_rate": 2.0874368764626185e-06, + "loss": 0.3126, + "step": 24747 + }, + { + "epoch": 9.14, + "learning_rate": 2.0873999260992736e-06, + "loss": 0.3689, + "step": 24748 + }, + { + "epoch": 9.14, + "learning_rate": 2.0873629757359282e-06, + "loss": 0.4612, + "step": 24749 + }, + { + "epoch": 9.14, + "learning_rate": 2.087326025372583e-06, + "loss": 0.3765, + "step": 24750 + }, + { + "epoch": 9.14, + "learning_rate": 2.0872890750092375e-06, + "loss": 0.4493, + "step": 24751 + }, + { + "epoch": 9.14, + "learning_rate": 2.0872521246458926e-06, + "loss": 0.2933, + "step": 24752 + }, + { + "epoch": 9.14, + "learning_rate": 2.087215174282547e-06, + "loss": 0.2817, + "step": 24753 + }, + { + "epoch": 9.14, + "learning_rate": 2.087178223919202e-06, + "loss": 0.3423, + "step": 24754 + }, + { + "epoch": 9.14, + "learning_rate": 2.087141273555857e-06, + "loss": 0.4087, + "step": 24755 + }, + { + "epoch": 9.14, + "learning_rate": 2.087104323192511e-06, + "loss": 0.3518, + "step": 24756 + }, + { + "epoch": 9.14, + "learning_rate": 2.087067372829166e-06, + "loss": 0.3681, + "step": 24757 + }, + { + "epoch": 9.14, + "learning_rate": 2.087030422465821e-06, + "loss": 0.2081, + "step": 24758 + }, + { + "epoch": 9.14, + "learning_rate": 2.086993472102476e-06, + "loss": 0.3557, + "step": 24759 + }, + { + "epoch": 9.14, + "learning_rate": 2.0869565217391305e-06, + "loss": 0.4349, + "step": 24760 + }, + { + "epoch": 9.14, + "learning_rate": 2.0869195713757856e-06, + "loss": 0.3084, + "step": 24761 + }, + { + "epoch": 9.14, + "learning_rate": 2.0868826210124398e-06, + "loss": 0.4175, + "step": 24762 + }, + { + "epoch": 9.14, + "learning_rate": 2.086845670649095e-06, + "loss": 0.356, + "step": 24763 + }, + { + "epoch": 9.14, + "learning_rate": 2.0868087202857495e-06, + "loss": 0.5315, + "step": 24764 + }, + { + "epoch": 9.15, + "learning_rate": 2.086771769922404e-06, + "loss": 0.3525, + "step": 24765 + }, + { + "epoch": 9.15, + "learning_rate": 2.086734819559059e-06, + "loss": 0.3165, + "step": 24766 + }, + { + "epoch": 9.15, + "learning_rate": 2.086697869195714e-06, + "loss": 0.4937, + "step": 24767 + }, + { + "epoch": 9.15, + "learning_rate": 2.0866609188323684e-06, + "loss": 0.2992, + "step": 24768 + }, + { + "epoch": 9.15, + "learning_rate": 2.086623968469023e-06, + "loss": 0.4296, + "step": 24769 + }, + { + "epoch": 9.15, + "learning_rate": 2.086587018105678e-06, + "loss": 0.4211, + "step": 24770 + }, + { + "epoch": 9.15, + "learning_rate": 2.0865500677423328e-06, + "loss": 0.3391, + "step": 24771 + }, + { + "epoch": 9.15, + "learning_rate": 2.086513117378988e-06, + "loss": 0.5049, + "step": 24772 + }, + { + "epoch": 9.15, + "learning_rate": 2.0864761670156425e-06, + "loss": 0.2875, + "step": 24773 + }, + { + "epoch": 9.15, + "learning_rate": 2.086439216652297e-06, + "loss": 0.1738, + "step": 24774 + }, + { + "epoch": 9.15, + "learning_rate": 2.0864022662889517e-06, + "loss": 0.3416, + "step": 24775 + }, + { + "epoch": 9.15, + "learning_rate": 2.0863653159256064e-06, + "loss": 0.3486, + "step": 24776 + }, + { + "epoch": 9.15, + "learning_rate": 2.0863283655622614e-06, + "loss": 0.2957, + "step": 24777 + }, + { + "epoch": 9.15, + "learning_rate": 2.086291415198916e-06, + "loss": 0.3605, + "step": 24778 + }, + { + "epoch": 9.15, + "learning_rate": 2.086254464835571e-06, + "loss": 0.3668, + "step": 24779 + }, + { + "epoch": 9.15, + "learning_rate": 2.0862175144722258e-06, + "loss": 0.5513, + "step": 24780 + }, + { + "epoch": 9.15, + "learning_rate": 2.0861805641088804e-06, + "loss": 0.4126, + "step": 24781 + }, + { + "epoch": 9.15, + "learning_rate": 2.086143613745535e-06, + "loss": 0.3694, + "step": 24782 + }, + { + "epoch": 9.15, + "learning_rate": 2.08610666338219e-06, + "loss": 0.5554, + "step": 24783 + }, + { + "epoch": 9.15, + "learning_rate": 2.0860697130188447e-06, + "loss": 0.353, + "step": 24784 + }, + { + "epoch": 9.15, + "learning_rate": 2.0860327626554994e-06, + "loss": 0.3635, + "step": 24785 + }, + { + "epoch": 9.15, + "learning_rate": 2.0859958122921544e-06, + "loss": 0.3521, + "step": 24786 + }, + { + "epoch": 9.15, + "learning_rate": 2.085958861928809e-06, + "loss": 0.3594, + "step": 24787 + }, + { + "epoch": 9.15, + "learning_rate": 2.0859219115654637e-06, + "loss": 0.4253, + "step": 24788 + }, + { + "epoch": 9.15, + "learning_rate": 2.0858849612021184e-06, + "loss": 0.27, + "step": 24789 + }, + { + "epoch": 9.15, + "learning_rate": 2.0858480108387734e-06, + "loss": 0.4172, + "step": 24790 + }, + { + "epoch": 9.15, + "learning_rate": 2.085811060475428e-06, + "loss": 0.4486, + "step": 24791 + }, + { + "epoch": 9.16, + "learning_rate": 2.085774110112083e-06, + "loss": 0.3676, + "step": 24792 + }, + { + "epoch": 9.16, + "learning_rate": 2.0857371597487377e-06, + "loss": 0.472, + "step": 24793 + }, + { + "epoch": 9.16, + "learning_rate": 2.0857002093853924e-06, + "loss": 0.4422, + "step": 24794 + }, + { + "epoch": 9.16, + "learning_rate": 2.085663259022047e-06, + "loss": 0.232, + "step": 24795 + }, + { + "epoch": 9.16, + "learning_rate": 2.0856263086587017e-06, + "loss": 0.3856, + "step": 24796 + }, + { + "epoch": 9.16, + "learning_rate": 2.0855893582953567e-06, + "loss": 0.3126, + "step": 24797 + }, + { + "epoch": 9.16, + "learning_rate": 2.0855524079320113e-06, + "loss": 0.3847, + "step": 24798 + }, + { + "epoch": 9.16, + "learning_rate": 2.0855154575686664e-06, + "loss": 0.2693, + "step": 24799 + }, + { + "epoch": 9.16, + "learning_rate": 2.085478507205321e-06, + "loss": 0.254, + "step": 24800 + }, + { + "epoch": 9.16, + "learning_rate": 2.0854415568419757e-06, + "loss": 0.4119, + "step": 24801 + }, + { + "epoch": 9.16, + "learning_rate": 2.0854046064786303e-06, + "loss": 0.3924, + "step": 24802 + }, + { + "epoch": 9.16, + "learning_rate": 2.0853676561152854e-06, + "loss": 0.3357, + "step": 24803 + }, + { + "epoch": 9.16, + "learning_rate": 2.08533070575194e-06, + "loss": 0.3801, + "step": 24804 + }, + { + "epoch": 9.16, + "learning_rate": 2.0852937553885947e-06, + "loss": 0.3028, + "step": 24805 + }, + { + "epoch": 9.16, + "learning_rate": 2.0852568050252497e-06, + "loss": 0.4207, + "step": 24806 + }, + { + "epoch": 9.16, + "learning_rate": 2.085219854661904e-06, + "loss": 0.3229, + "step": 24807 + }, + { + "epoch": 9.16, + "learning_rate": 2.085182904298559e-06, + "loss": 0.3698, + "step": 24808 + }, + { + "epoch": 9.16, + "learning_rate": 2.0851459539352136e-06, + "loss": 0.3732, + "step": 24809 + }, + { + "epoch": 9.16, + "learning_rate": 2.0851090035718687e-06, + "loss": 0.4005, + "step": 24810 + }, + { + "epoch": 9.16, + "learning_rate": 2.0850720532085233e-06, + "loss": 0.3453, + "step": 24811 + }, + { + "epoch": 9.16, + "learning_rate": 2.0850351028451784e-06, + "loss": 0.3959, + "step": 24812 + }, + { + "epoch": 9.16, + "learning_rate": 2.0849981524818326e-06, + "loss": 0.3479, + "step": 24813 + }, + { + "epoch": 9.16, + "learning_rate": 2.0849612021184872e-06, + "loss": 0.2558, + "step": 24814 + }, + { + "epoch": 9.16, + "learning_rate": 2.0849242517551423e-06, + "loss": 0.3591, + "step": 24815 + }, + { + "epoch": 9.16, + "learning_rate": 2.084887301391797e-06, + "loss": 0.3453, + "step": 24816 + }, + { + "epoch": 9.16, + "learning_rate": 2.084850351028452e-06, + "loss": 0.4551, + "step": 24817 + }, + { + "epoch": 9.16, + "learning_rate": 2.0848134006651066e-06, + "loss": 0.3387, + "step": 24818 + }, + { + "epoch": 9.17, + "learning_rate": 2.0847764503017613e-06, + "loss": 0.3817, + "step": 24819 + }, + { + "epoch": 9.17, + "learning_rate": 2.084739499938416e-06, + "loss": 0.316, + "step": 24820 + }, + { + "epoch": 9.17, + "learning_rate": 2.084702549575071e-06, + "loss": 0.2507, + "step": 24821 + }, + { + "epoch": 9.17, + "learning_rate": 2.0846655992117256e-06, + "loss": 0.3641, + "step": 24822 + }, + { + "epoch": 9.17, + "learning_rate": 2.0846286488483806e-06, + "loss": 0.2853, + "step": 24823 + }, + { + "epoch": 9.17, + "learning_rate": 2.0845916984850353e-06, + "loss": 0.4093, + "step": 24824 + }, + { + "epoch": 9.17, + "learning_rate": 2.08455474812169e-06, + "loss": 0.39, + "step": 24825 + }, + { + "epoch": 9.17, + "learning_rate": 2.0845177977583446e-06, + "loss": 0.4326, + "step": 24826 + }, + { + "epoch": 9.17, + "learning_rate": 2.084480847394999e-06, + "loss": 0.3844, + "step": 24827 + }, + { + "epoch": 9.17, + "learning_rate": 2.0844438970316543e-06, + "loss": 0.4906, + "step": 24828 + }, + { + "epoch": 9.17, + "learning_rate": 2.084406946668309e-06, + "loss": 0.3526, + "step": 24829 + }, + { + "epoch": 9.17, + "learning_rate": 2.084369996304964e-06, + "loss": 0.2374, + "step": 24830 + }, + { + "epoch": 9.17, + "learning_rate": 2.0843330459416186e-06, + "loss": 0.3089, + "step": 24831 + }, + { + "epoch": 9.17, + "learning_rate": 2.0842960955782732e-06, + "loss": 0.334, + "step": 24832 + }, + { + "epoch": 9.17, + "learning_rate": 2.084259145214928e-06, + "loss": 0.4479, + "step": 24833 + }, + { + "epoch": 9.17, + "learning_rate": 2.0842221948515825e-06, + "loss": 0.327, + "step": 24834 + }, + { + "epoch": 9.17, + "learning_rate": 2.0841852444882376e-06, + "loss": 0.3541, + "step": 24835 + }, + { + "epoch": 9.17, + "learning_rate": 2.084148294124892e-06, + "loss": 0.2937, + "step": 24836 + }, + { + "epoch": 9.17, + "learning_rate": 2.0841113437615473e-06, + "loss": 0.4103, + "step": 24837 + }, + { + "epoch": 9.17, + "learning_rate": 2.084074393398202e-06, + "loss": 0.3331, + "step": 24838 + }, + { + "epoch": 9.17, + "learning_rate": 2.0840374430348565e-06, + "loss": 0.3338, + "step": 24839 + }, + { + "epoch": 9.17, + "learning_rate": 2.084000492671511e-06, + "loss": 0.5039, + "step": 24840 + }, + { + "epoch": 9.17, + "learning_rate": 2.0839635423081662e-06, + "loss": 0.3361, + "step": 24841 + }, + { + "epoch": 9.17, + "learning_rate": 2.083926591944821e-06, + "loss": 0.4486, + "step": 24842 + }, + { + "epoch": 9.17, + "learning_rate": 2.083889641581476e-06, + "loss": 0.4054, + "step": 24843 + }, + { + "epoch": 9.17, + "learning_rate": 2.0838526912181306e-06, + "loss": 0.6386, + "step": 24844 + }, + { + "epoch": 9.17, + "learning_rate": 2.083815740854785e-06, + "loss": 0.4106, + "step": 24845 + }, + { + "epoch": 9.18, + "learning_rate": 2.08377879049144e-06, + "loss": 0.3403, + "step": 24846 + }, + { + "epoch": 9.18, + "learning_rate": 2.0837418401280945e-06, + "loss": 0.5705, + "step": 24847 + }, + { + "epoch": 9.18, + "learning_rate": 2.0837048897647495e-06, + "loss": 0.2686, + "step": 24848 + }, + { + "epoch": 9.18, + "learning_rate": 2.083667939401404e-06, + "loss": 0.3704, + "step": 24849 + }, + { + "epoch": 9.18, + "learning_rate": 2.0836309890380592e-06, + "loss": 0.3696, + "step": 24850 + }, + { + "epoch": 9.18, + "learning_rate": 2.083594038674714e-06, + "loss": 0.3458, + "step": 24851 + }, + { + "epoch": 9.18, + "learning_rate": 2.0835570883113685e-06, + "loss": 0.3335, + "step": 24852 + }, + { + "epoch": 9.18, + "learning_rate": 2.083520137948023e-06, + "loss": 0.2838, + "step": 24853 + }, + { + "epoch": 9.18, + "learning_rate": 2.0834831875846778e-06, + "loss": 0.5172, + "step": 24854 + }, + { + "epoch": 9.18, + "learning_rate": 2.083446237221333e-06, + "loss": 0.4011, + "step": 24855 + }, + { + "epoch": 9.18, + "learning_rate": 2.0834092868579875e-06, + "loss": 0.3252, + "step": 24856 + }, + { + "epoch": 9.18, + "learning_rate": 2.0833723364946425e-06, + "loss": 0.2357, + "step": 24857 + }, + { + "epoch": 9.18, + "learning_rate": 2.0833353861312967e-06, + "loss": 0.3033, + "step": 24858 + }, + { + "epoch": 9.18, + "learning_rate": 2.083298435767952e-06, + "loss": 0.2843, + "step": 24859 + }, + { + "epoch": 9.18, + "learning_rate": 2.0832614854046064e-06, + "loss": 0.3816, + "step": 24860 + }, + { + "epoch": 9.18, + "learning_rate": 2.0832245350412615e-06, + "loss": 0.3689, + "step": 24861 + }, + { + "epoch": 9.18, + "learning_rate": 2.083187584677916e-06, + "loss": 0.4134, + "step": 24862 + }, + { + "epoch": 9.18, + "learning_rate": 2.083150634314571e-06, + "loss": 0.2385, + "step": 24863 + }, + { + "epoch": 9.18, + "learning_rate": 2.0831136839512254e-06, + "loss": 0.2798, + "step": 24864 + }, + { + "epoch": 9.18, + "learning_rate": 2.08307673358788e-06, + "loss": 0.4058, + "step": 24865 + }, + { + "epoch": 9.18, + "learning_rate": 2.083039783224535e-06, + "loss": 0.3712, + "step": 24866 + }, + { + "epoch": 9.18, + "learning_rate": 2.0830028328611897e-06, + "loss": 0.3869, + "step": 24867 + }, + { + "epoch": 9.18, + "learning_rate": 2.082965882497845e-06, + "loss": 0.3402, + "step": 24868 + }, + { + "epoch": 9.18, + "learning_rate": 2.0829289321344994e-06, + "loss": 0.301, + "step": 24869 + }, + { + "epoch": 9.18, + "learning_rate": 2.082891981771154e-06, + "loss": 0.369, + "step": 24870 + }, + { + "epoch": 9.18, + "learning_rate": 2.0828550314078087e-06, + "loss": 0.3114, + "step": 24871 + }, + { + "epoch": 9.18, + "learning_rate": 2.0828180810444638e-06, + "loss": 0.3209, + "step": 24872 + }, + { + "epoch": 9.19, + "learning_rate": 2.0827811306811184e-06, + "loss": 0.3716, + "step": 24873 + }, + { + "epoch": 9.19, + "learning_rate": 2.082744180317773e-06, + "loss": 0.3308, + "step": 24874 + }, + { + "epoch": 9.19, + "learning_rate": 2.082707229954428e-06, + "loss": 0.3371, + "step": 24875 + }, + { + "epoch": 9.19, + "learning_rate": 2.0826702795910827e-06, + "loss": 0.4455, + "step": 24876 + }, + { + "epoch": 9.19, + "learning_rate": 2.0826333292277374e-06, + "loss": 0.3241, + "step": 24877 + }, + { + "epoch": 9.19, + "learning_rate": 2.082596378864392e-06, + "loss": 0.2196, + "step": 24878 + }, + { + "epoch": 9.19, + "learning_rate": 2.082559428501047e-06, + "loss": 0.3799, + "step": 24879 + }, + { + "epoch": 9.19, + "learning_rate": 2.0825224781377017e-06, + "loss": 0.3656, + "step": 24880 + }, + { + "epoch": 9.19, + "learning_rate": 2.0824855277743568e-06, + "loss": 0.3168, + "step": 24881 + }, + { + "epoch": 9.19, + "learning_rate": 2.0824485774110114e-06, + "loss": 0.3281, + "step": 24882 + }, + { + "epoch": 9.19, + "learning_rate": 2.082411627047666e-06, + "loss": 0.494, + "step": 24883 + }, + { + "epoch": 9.19, + "learning_rate": 2.0823746766843207e-06, + "loss": 0.4777, + "step": 24884 + }, + { + "epoch": 9.19, + "learning_rate": 2.0823377263209753e-06, + "loss": 0.455, + "step": 24885 + }, + { + "epoch": 9.19, + "learning_rate": 2.0823007759576304e-06, + "loss": 0.382, + "step": 24886 + }, + { + "epoch": 9.19, + "learning_rate": 2.082263825594285e-06, + "loss": 0.3321, + "step": 24887 + }, + { + "epoch": 9.19, + "learning_rate": 2.08222687523094e-06, + "loss": 0.3415, + "step": 24888 + }, + { + "epoch": 9.19, + "learning_rate": 2.0821899248675947e-06, + "loss": 0.5049, + "step": 24889 + }, + { + "epoch": 9.19, + "learning_rate": 2.0821529745042493e-06, + "loss": 0.2657, + "step": 24890 + }, + { + "epoch": 9.19, + "learning_rate": 2.082116024140904e-06, + "loss": 0.3298, + "step": 24891 + }, + { + "epoch": 9.19, + "learning_rate": 2.082079073777559e-06, + "loss": 0.4292, + "step": 24892 + }, + { + "epoch": 9.19, + "learning_rate": 2.0820421234142137e-06, + "loss": 0.4719, + "step": 24893 + }, + { + "epoch": 9.19, + "learning_rate": 2.0820051730508683e-06, + "loss": 0.3823, + "step": 24894 + }, + { + "epoch": 9.19, + "learning_rate": 2.0819682226875234e-06, + "loss": 0.3879, + "step": 24895 + }, + { + "epoch": 9.19, + "learning_rate": 2.081931272324178e-06, + "loss": 0.3404, + "step": 24896 + }, + { + "epoch": 9.19, + "learning_rate": 2.0818943219608326e-06, + "loss": 0.3515, + "step": 24897 + }, + { + "epoch": 9.19, + "learning_rate": 2.0818573715974873e-06, + "loss": 0.3533, + "step": 24898 + }, + { + "epoch": 9.19, + "learning_rate": 2.0818204212341423e-06, + "loss": 0.4873, + "step": 24899 + }, + { + "epoch": 9.19, + "learning_rate": 2.081783470870797e-06, + "loss": 0.367, + "step": 24900 + }, + { + "epoch": 9.2, + "learning_rate": 2.081746520507452e-06, + "loss": 0.4008, + "step": 24901 + }, + { + "epoch": 9.2, + "learning_rate": 2.0817095701441062e-06, + "loss": 0.4534, + "step": 24902 + }, + { + "epoch": 9.2, + "learning_rate": 2.081672619780761e-06, + "loss": 0.4255, + "step": 24903 + }, + { + "epoch": 9.2, + "learning_rate": 2.081635669417416e-06, + "loss": 0.3002, + "step": 24904 + }, + { + "epoch": 9.2, + "learning_rate": 2.0815987190540706e-06, + "loss": 0.31, + "step": 24905 + }, + { + "epoch": 9.2, + "learning_rate": 2.0815617686907256e-06, + "loss": 0.3813, + "step": 24906 + }, + { + "epoch": 9.2, + "learning_rate": 2.0815248183273803e-06, + "loss": 0.367, + "step": 24907 + }, + { + "epoch": 9.2, + "learning_rate": 2.081487867964035e-06, + "loss": 0.4024, + "step": 24908 + }, + { + "epoch": 9.2, + "learning_rate": 2.0814509176006895e-06, + "loss": 0.2922, + "step": 24909 + }, + { + "epoch": 9.2, + "learning_rate": 2.0814139672373446e-06, + "loss": 0.379, + "step": 24910 + }, + { + "epoch": 9.2, + "learning_rate": 2.0813770168739992e-06, + "loss": 0.2215, + "step": 24911 + }, + { + "epoch": 9.2, + "learning_rate": 2.0813400665106543e-06, + "loss": 0.3971, + "step": 24912 + }, + { + "epoch": 9.2, + "learning_rate": 2.081303116147309e-06, + "loss": 0.2509, + "step": 24913 + }, + { + "epoch": 9.2, + "learning_rate": 2.0812661657839636e-06, + "loss": 0.3179, + "step": 24914 + }, + { + "epoch": 9.2, + "learning_rate": 2.081229215420618e-06, + "loss": 0.2774, + "step": 24915 + }, + { + "epoch": 9.2, + "learning_rate": 2.081192265057273e-06, + "loss": 0.2379, + "step": 24916 + }, + { + "epoch": 9.2, + "learning_rate": 2.081155314693928e-06, + "loss": 0.3883, + "step": 24917 + }, + { + "epoch": 9.2, + "learning_rate": 2.0811183643305825e-06, + "loss": 0.3841, + "step": 24918 + }, + { + "epoch": 9.2, + "learning_rate": 2.0810814139672376e-06, + "loss": 0.311, + "step": 24919 + }, + { + "epoch": 9.2, + "learning_rate": 2.0810444636038922e-06, + "loss": 0.3945, + "step": 24920 + }, + { + "epoch": 9.2, + "learning_rate": 2.081007513240547e-06, + "loss": 0.2743, + "step": 24921 + }, + { + "epoch": 9.2, + "learning_rate": 2.0809705628772015e-06, + "loss": 0.454, + "step": 24922 + }, + { + "epoch": 9.2, + "learning_rate": 2.080933612513856e-06, + "loss": 0.3517, + "step": 24923 + }, + { + "epoch": 9.2, + "learning_rate": 2.080896662150511e-06, + "loss": 0.5266, + "step": 24924 + }, + { + "epoch": 9.2, + "learning_rate": 2.080859711787166e-06, + "loss": 0.3585, + "step": 24925 + }, + { + "epoch": 9.2, + "learning_rate": 2.080822761423821e-06, + "loss": 0.451, + "step": 24926 + }, + { + "epoch": 9.2, + "learning_rate": 2.0807858110604755e-06, + "loss": 0.2796, + "step": 24927 + }, + { + "epoch": 9.21, + "learning_rate": 2.08074886069713e-06, + "loss": 0.3497, + "step": 24928 + }, + { + "epoch": 9.21, + "learning_rate": 2.080711910333785e-06, + "loss": 0.503, + "step": 24929 + }, + { + "epoch": 9.21, + "learning_rate": 2.08067495997044e-06, + "loss": 0.3764, + "step": 24930 + }, + { + "epoch": 9.21, + "learning_rate": 2.0806380096070945e-06, + "loss": 0.355, + "step": 24931 + }, + { + "epoch": 9.21, + "learning_rate": 2.0806010592437496e-06, + "loss": 0.404, + "step": 24932 + }, + { + "epoch": 9.21, + "learning_rate": 2.080564108880404e-06, + "loss": 0.462, + "step": 24933 + }, + { + "epoch": 9.21, + "learning_rate": 2.080527158517059e-06, + "loss": 0.4228, + "step": 24934 + }, + { + "epoch": 9.21, + "learning_rate": 2.0804902081537135e-06, + "loss": 0.3517, + "step": 24935 + }, + { + "epoch": 9.21, + "learning_rate": 2.080453257790368e-06, + "loss": 0.3065, + "step": 24936 + }, + { + "epoch": 9.21, + "learning_rate": 2.080416307427023e-06, + "loss": 0.307, + "step": 24937 + }, + { + "epoch": 9.21, + "learning_rate": 2.080379357063678e-06, + "loss": 0.3375, + "step": 24938 + }, + { + "epoch": 9.21, + "learning_rate": 2.080342406700333e-06, + "loss": 0.3473, + "step": 24939 + }, + { + "epoch": 9.21, + "learning_rate": 2.0803054563369875e-06, + "loss": 0.3384, + "step": 24940 + }, + { + "epoch": 9.21, + "learning_rate": 2.080268505973642e-06, + "loss": 0.3844, + "step": 24941 + }, + { + "epoch": 9.21, + "learning_rate": 2.0802315556102968e-06, + "loss": 0.3852, + "step": 24942 + }, + { + "epoch": 9.21, + "learning_rate": 2.0801946052469514e-06, + "loss": 0.3083, + "step": 24943 + }, + { + "epoch": 9.21, + "learning_rate": 2.0801576548836065e-06, + "loss": 0.3665, + "step": 24944 + }, + { + "epoch": 9.21, + "learning_rate": 2.080120704520261e-06, + "loss": 0.4578, + "step": 24945 + }, + { + "epoch": 9.21, + "learning_rate": 2.080083754156916e-06, + "loss": 0.3207, + "step": 24946 + }, + { + "epoch": 9.21, + "learning_rate": 2.0800468037935704e-06, + "loss": 0.3608, + "step": 24947 + }, + { + "epoch": 9.21, + "learning_rate": 2.0800098534302254e-06, + "loss": 0.4091, + "step": 24948 + }, + { + "epoch": 9.21, + "learning_rate": 2.07997290306688e-06, + "loss": 0.3197, + "step": 24949 + }, + { + "epoch": 9.21, + "learning_rate": 2.079935952703535e-06, + "loss": 0.3143, + "step": 24950 + }, + { + "epoch": 9.21, + "learning_rate": 2.0798990023401898e-06, + "loss": 0.2498, + "step": 24951 + }, + { + "epoch": 9.21, + "learning_rate": 2.079862051976845e-06, + "loss": 0.2866, + "step": 24952 + }, + { + "epoch": 9.21, + "learning_rate": 2.079825101613499e-06, + "loss": 0.4641, + "step": 24953 + }, + { + "epoch": 9.21, + "learning_rate": 2.0797881512501537e-06, + "loss": 0.4507, + "step": 24954 + }, + { + "epoch": 9.22, + "learning_rate": 2.0797512008868088e-06, + "loss": 0.3681, + "step": 24955 + }, + { + "epoch": 9.22, + "learning_rate": 2.0797142505234634e-06, + "loss": 0.3882, + "step": 24956 + }, + { + "epoch": 9.22, + "learning_rate": 2.0796773001601184e-06, + "loss": 0.2929, + "step": 24957 + }, + { + "epoch": 9.22, + "learning_rate": 2.079640349796773e-06, + "loss": 0.392, + "step": 24958 + }, + { + "epoch": 9.22, + "learning_rate": 2.0796033994334277e-06, + "loss": 0.4712, + "step": 24959 + }, + { + "epoch": 9.22, + "learning_rate": 2.0795664490700824e-06, + "loss": 0.3034, + "step": 24960 + }, + { + "epoch": 9.22, + "learning_rate": 2.0795294987067374e-06, + "loss": 0.4218, + "step": 24961 + }, + { + "epoch": 9.22, + "learning_rate": 2.079492548343392e-06, + "loss": 0.2642, + "step": 24962 + }, + { + "epoch": 9.22, + "learning_rate": 2.0794555979800467e-06, + "loss": 0.3739, + "step": 24963 + }, + { + "epoch": 9.22, + "learning_rate": 2.0794186476167017e-06, + "loss": 0.3877, + "step": 24964 + }, + { + "epoch": 9.22, + "learning_rate": 2.0793816972533564e-06, + "loss": 0.323, + "step": 24965 + }, + { + "epoch": 9.22, + "learning_rate": 2.079344746890011e-06, + "loss": 0.2603, + "step": 24966 + }, + { + "epoch": 9.22, + "learning_rate": 2.0793077965266657e-06, + "loss": 0.3261, + "step": 24967 + }, + { + "epoch": 9.22, + "learning_rate": 2.0792708461633207e-06, + "loss": 0.2789, + "step": 24968 + }, + { + "epoch": 9.22, + "learning_rate": 2.0792338957999754e-06, + "loss": 0.428, + "step": 24969 + }, + { + "epoch": 9.22, + "learning_rate": 2.0791969454366304e-06, + "loss": 0.2437, + "step": 24970 + }, + { + "epoch": 9.22, + "learning_rate": 2.079159995073285e-06, + "loss": 0.3644, + "step": 24971 + }, + { + "epoch": 9.22, + "learning_rate": 2.0791230447099397e-06, + "loss": 0.3407, + "step": 24972 + }, + { + "epoch": 9.22, + "learning_rate": 2.0790860943465943e-06, + "loss": 0.3068, + "step": 24973 + }, + { + "epoch": 9.22, + "learning_rate": 2.079049143983249e-06, + "loss": 0.3515, + "step": 24974 + }, + { + "epoch": 9.22, + "learning_rate": 2.079012193619904e-06, + "loss": 0.4881, + "step": 24975 + }, + { + "epoch": 9.22, + "learning_rate": 2.0789752432565587e-06, + "loss": 0.3095, + "step": 24976 + }, + { + "epoch": 9.22, + "learning_rate": 2.0789382928932137e-06, + "loss": 0.3592, + "step": 24977 + }, + { + "epoch": 9.22, + "learning_rate": 2.0789013425298684e-06, + "loss": 0.4263, + "step": 24978 + }, + { + "epoch": 9.22, + "learning_rate": 2.078864392166523e-06, + "loss": 0.3616, + "step": 24979 + }, + { + "epoch": 9.22, + "learning_rate": 2.0788274418031776e-06, + "loss": 0.4394, + "step": 24980 + }, + { + "epoch": 9.22, + "learning_rate": 2.0787904914398327e-06, + "loss": 0.457, + "step": 24981 + }, + { + "epoch": 9.23, + "learning_rate": 2.0787535410764873e-06, + "loss": 0.2826, + "step": 24982 + }, + { + "epoch": 9.23, + "learning_rate": 2.078716590713142e-06, + "loss": 0.3601, + "step": 24983 + }, + { + "epoch": 9.23, + "learning_rate": 2.078679640349797e-06, + "loss": 0.4912, + "step": 24984 + }, + { + "epoch": 9.23, + "learning_rate": 2.0786426899864517e-06, + "loss": 0.461, + "step": 24985 + }, + { + "epoch": 9.23, + "learning_rate": 2.0786057396231063e-06, + "loss": 0.416, + "step": 24986 + }, + { + "epoch": 9.23, + "learning_rate": 2.078568789259761e-06, + "loss": 0.429, + "step": 24987 + }, + { + "epoch": 9.23, + "learning_rate": 2.078531838896416e-06, + "loss": 0.3247, + "step": 24988 + }, + { + "epoch": 9.23, + "learning_rate": 2.0784948885330706e-06, + "loss": 0.4421, + "step": 24989 + }, + { + "epoch": 9.23, + "learning_rate": 2.0784579381697257e-06, + "loss": 0.443, + "step": 24990 + }, + { + "epoch": 9.23, + "learning_rate": 2.0784209878063803e-06, + "loss": 0.3091, + "step": 24991 + }, + { + "epoch": 9.23, + "learning_rate": 2.078384037443035e-06, + "loss": 0.2929, + "step": 24992 + }, + { + "epoch": 9.23, + "learning_rate": 2.0783470870796896e-06, + "loss": 0.3235, + "step": 24993 + }, + { + "epoch": 9.23, + "learning_rate": 2.0783101367163442e-06, + "loss": 0.3997, + "step": 24994 + }, + { + "epoch": 9.23, + "learning_rate": 2.0782731863529993e-06, + "loss": 0.323, + "step": 24995 + }, + { + "epoch": 9.23, + "learning_rate": 2.078236235989654e-06, + "loss": 0.3352, + "step": 24996 + }, + { + "epoch": 9.23, + "learning_rate": 2.078199285626309e-06, + "loss": 0.2554, + "step": 24997 + }, + { + "epoch": 9.23, + "learning_rate": 2.078162335262963e-06, + "loss": 0.2522, + "step": 24998 + }, + { + "epoch": 9.23, + "learning_rate": 2.0781253848996183e-06, + "loss": 0.4477, + "step": 24999 + }, + { + "epoch": 9.23, + "learning_rate": 2.078088434536273e-06, + "loss": 0.5649, + "step": 25000 + }, + { + "epoch": 9.23, + "learning_rate": 2.078051484172928e-06, + "loss": 0.3117, + "step": 25001 + }, + { + "epoch": 9.23, + "learning_rate": 2.0780145338095826e-06, + "loss": 0.2968, + "step": 25002 + }, + { + "epoch": 9.23, + "learning_rate": 2.0779775834462372e-06, + "loss": 0.4777, + "step": 25003 + }, + { + "epoch": 9.23, + "learning_rate": 2.077940633082892e-06, + "loss": 0.2422, + "step": 25004 + }, + { + "epoch": 9.23, + "learning_rate": 2.0779036827195465e-06, + "loss": 0.4599, + "step": 25005 + }, + { + "epoch": 9.23, + "learning_rate": 2.0778667323562016e-06, + "loss": 0.3113, + "step": 25006 + }, + { + "epoch": 9.23, + "learning_rate": 2.077829781992856e-06, + "loss": 0.4666, + "step": 25007 + }, + { + "epoch": 9.23, + "learning_rate": 2.0777928316295113e-06, + "loss": 0.5044, + "step": 25008 + }, + { + "epoch": 9.24, + "learning_rate": 2.077755881266166e-06, + "loss": 0.3201, + "step": 25009 + }, + { + "epoch": 9.24, + "learning_rate": 2.0777189309028205e-06, + "loss": 0.3391, + "step": 25010 + }, + { + "epoch": 9.24, + "learning_rate": 2.077681980539475e-06, + "loss": 0.34, + "step": 25011 + }, + { + "epoch": 9.24, + "learning_rate": 2.0776450301761302e-06, + "loss": 0.3595, + "step": 25012 + }, + { + "epoch": 9.24, + "learning_rate": 2.077608079812785e-06, + "loss": 0.3599, + "step": 25013 + }, + { + "epoch": 9.24, + "learning_rate": 2.0775711294494395e-06, + "loss": 0.4145, + "step": 25014 + }, + { + "epoch": 9.24, + "learning_rate": 2.0775341790860946e-06, + "loss": 0.2992, + "step": 25015 + }, + { + "epoch": 9.24, + "learning_rate": 2.077497228722749e-06, + "loss": 0.3969, + "step": 25016 + }, + { + "epoch": 9.24, + "learning_rate": 2.077460278359404e-06, + "loss": 0.4701, + "step": 25017 + }, + { + "epoch": 9.24, + "learning_rate": 2.0774233279960585e-06, + "loss": 0.2935, + "step": 25018 + }, + { + "epoch": 9.24, + "learning_rate": 2.0773863776327135e-06, + "loss": 0.2047, + "step": 25019 + }, + { + "epoch": 9.24, + "learning_rate": 2.077349427269368e-06, + "loss": 0.4593, + "step": 25020 + }, + { + "epoch": 9.24, + "learning_rate": 2.0773124769060232e-06, + "loss": 0.3438, + "step": 25021 + }, + { + "epoch": 9.24, + "learning_rate": 2.077275526542678e-06, + "loss": 0.2363, + "step": 25022 + }, + { + "epoch": 9.24, + "learning_rate": 2.0772385761793325e-06, + "loss": 0.4911, + "step": 25023 + }, + { + "epoch": 9.24, + "learning_rate": 2.077201625815987e-06, + "loss": 0.5348, + "step": 25024 + }, + { + "epoch": 9.24, + "learning_rate": 2.0771646754526418e-06, + "loss": 0.3662, + "step": 25025 + }, + { + "epoch": 9.24, + "learning_rate": 2.077127725089297e-06, + "loss": 0.3807, + "step": 25026 + }, + { + "epoch": 9.24, + "learning_rate": 2.0770907747259515e-06, + "loss": 0.3688, + "step": 25027 + }, + { + "epoch": 9.24, + "learning_rate": 2.0770538243626065e-06, + "loss": 0.4517, + "step": 25028 + }, + { + "epoch": 9.24, + "learning_rate": 2.077016873999261e-06, + "loss": 0.3408, + "step": 25029 + }, + { + "epoch": 9.24, + "learning_rate": 2.076979923635916e-06, + "loss": 0.2878, + "step": 25030 + }, + { + "epoch": 9.24, + "learning_rate": 2.0769429732725704e-06, + "loss": 0.373, + "step": 25031 + }, + { + "epoch": 9.24, + "learning_rate": 2.0769060229092255e-06, + "loss": 0.4695, + "step": 25032 + }, + { + "epoch": 9.24, + "learning_rate": 2.07686907254588e-06, + "loss": 0.265, + "step": 25033 + }, + { + "epoch": 9.24, + "learning_rate": 2.0768321221825348e-06, + "loss": 0.3108, + "step": 25034 + }, + { + "epoch": 9.24, + "learning_rate": 2.07679517181919e-06, + "loss": 0.3263, + "step": 25035 + }, + { + "epoch": 9.25, + "learning_rate": 2.0767582214558445e-06, + "loss": 0.3692, + "step": 25036 + }, + { + "epoch": 9.25, + "learning_rate": 2.076721271092499e-06, + "loss": 0.316, + "step": 25037 + }, + { + "epoch": 9.25, + "learning_rate": 2.0766843207291537e-06, + "loss": 0.4815, + "step": 25038 + }, + { + "epoch": 9.25, + "learning_rate": 2.076647370365809e-06, + "loss": 0.3873, + "step": 25039 + }, + { + "epoch": 9.25, + "learning_rate": 2.0766104200024634e-06, + "loss": 0.3359, + "step": 25040 + }, + { + "epoch": 9.25, + "learning_rate": 2.0765734696391185e-06, + "loss": 0.4037, + "step": 25041 + }, + { + "epoch": 9.25, + "learning_rate": 2.076536519275773e-06, + "loss": 0.3817, + "step": 25042 + }, + { + "epoch": 9.25, + "learning_rate": 2.0764995689124273e-06, + "loss": 0.4969, + "step": 25043 + }, + { + "epoch": 9.25, + "learning_rate": 2.0764626185490824e-06, + "loss": 0.3404, + "step": 25044 + }, + { + "epoch": 9.25, + "learning_rate": 2.076425668185737e-06, + "loss": 0.3965, + "step": 25045 + }, + { + "epoch": 9.25, + "learning_rate": 2.076388717822392e-06, + "loss": 0.3229, + "step": 25046 + }, + { + "epoch": 9.25, + "learning_rate": 2.0763517674590467e-06, + "loss": 0.3082, + "step": 25047 + }, + { + "epoch": 9.25, + "learning_rate": 2.076314817095702e-06, + "loss": 0.4257, + "step": 25048 + }, + { + "epoch": 9.25, + "learning_rate": 2.076277866732356e-06, + "loss": 0.2282, + "step": 25049 + }, + { + "epoch": 9.25, + "learning_rate": 2.076240916369011e-06, + "loss": 0.4331, + "step": 25050 + }, + { + "epoch": 9.25, + "learning_rate": 2.0762039660056657e-06, + "loss": 0.4501, + "step": 25051 + }, + { + "epoch": 9.25, + "learning_rate": 2.0761670156423208e-06, + "loss": 0.3815, + "step": 25052 + }, + { + "epoch": 9.25, + "learning_rate": 2.0761300652789754e-06, + "loss": 0.3231, + "step": 25053 + }, + { + "epoch": 9.25, + "learning_rate": 2.07609311491563e-06, + "loss": 0.3896, + "step": 25054 + }, + { + "epoch": 9.25, + "learning_rate": 2.0760561645522847e-06, + "loss": 0.45, + "step": 25055 + }, + { + "epoch": 9.25, + "learning_rate": 2.0760192141889393e-06, + "loss": 0.367, + "step": 25056 + }, + { + "epoch": 9.25, + "learning_rate": 2.0759822638255944e-06, + "loss": 0.3256, + "step": 25057 + }, + { + "epoch": 9.25, + "learning_rate": 2.075945313462249e-06, + "loss": 0.3797, + "step": 25058 + }, + { + "epoch": 9.25, + "learning_rate": 2.075908363098904e-06, + "loss": 0.4462, + "step": 25059 + }, + { + "epoch": 9.25, + "learning_rate": 2.0758714127355587e-06, + "loss": 0.2721, + "step": 25060 + }, + { + "epoch": 9.25, + "learning_rate": 2.0758344623722133e-06, + "loss": 0.3507, + "step": 25061 + }, + { + "epoch": 9.25, + "learning_rate": 2.075797512008868e-06, + "loss": 0.3837, + "step": 25062 + }, + { + "epoch": 9.26, + "learning_rate": 2.0757605616455226e-06, + "loss": 0.2988, + "step": 25063 + }, + { + "epoch": 9.26, + "learning_rate": 2.0757236112821777e-06, + "loss": 0.3482, + "step": 25064 + }, + { + "epoch": 9.26, + "learning_rate": 2.0756866609188323e-06, + "loss": 0.3482, + "step": 25065 + }, + { + "epoch": 9.26, + "learning_rate": 2.0756497105554874e-06, + "loss": 0.3764, + "step": 25066 + }, + { + "epoch": 9.26, + "learning_rate": 2.075612760192142e-06, + "loss": 0.3628, + "step": 25067 + }, + { + "epoch": 9.26, + "learning_rate": 2.0755758098287966e-06, + "loss": 0.3645, + "step": 25068 + }, + { + "epoch": 9.26, + "learning_rate": 2.0755388594654513e-06, + "loss": 0.3542, + "step": 25069 + }, + { + "epoch": 9.26, + "learning_rate": 2.0755019091021063e-06, + "loss": 0.388, + "step": 25070 + }, + { + "epoch": 9.26, + "learning_rate": 2.075464958738761e-06, + "loss": 0.279, + "step": 25071 + }, + { + "epoch": 9.26, + "learning_rate": 2.075428008375416e-06, + "loss": 0.3732, + "step": 25072 + }, + { + "epoch": 9.26, + "learning_rate": 2.0753910580120707e-06, + "loss": 0.3713, + "step": 25073 + }, + { + "epoch": 9.26, + "learning_rate": 2.0753541076487253e-06, + "loss": 0.3797, + "step": 25074 + }, + { + "epoch": 9.26, + "learning_rate": 2.07531715728538e-06, + "loss": 0.5055, + "step": 25075 + }, + { + "epoch": 9.26, + "learning_rate": 2.0752802069220346e-06, + "loss": 0.3403, + "step": 25076 + }, + { + "epoch": 9.26, + "learning_rate": 2.0752432565586896e-06, + "loss": 0.2807, + "step": 25077 + }, + { + "epoch": 9.26, + "learning_rate": 2.0752063061953443e-06, + "loss": 0.2421, + "step": 25078 + }, + { + "epoch": 9.26, + "learning_rate": 2.0751693558319993e-06, + "loss": 0.3964, + "step": 25079 + }, + { + "epoch": 9.26, + "learning_rate": 2.075132405468654e-06, + "loss": 0.495, + "step": 25080 + }, + { + "epoch": 9.26, + "learning_rate": 2.0750954551053086e-06, + "loss": 0.3374, + "step": 25081 + }, + { + "epoch": 9.26, + "learning_rate": 2.0750585047419632e-06, + "loss": 0.4048, + "step": 25082 + }, + { + "epoch": 9.26, + "learning_rate": 2.075021554378618e-06, + "loss": 0.2941, + "step": 25083 + }, + { + "epoch": 9.26, + "learning_rate": 2.074984604015273e-06, + "loss": 0.3727, + "step": 25084 + }, + { + "epoch": 9.26, + "learning_rate": 2.0749476536519276e-06, + "loss": 0.2831, + "step": 25085 + }, + { + "epoch": 9.26, + "learning_rate": 2.0749107032885826e-06, + "loss": 0.345, + "step": 25086 + }, + { + "epoch": 9.26, + "learning_rate": 2.0748737529252373e-06, + "loss": 0.3984, + "step": 25087 + }, + { + "epoch": 9.26, + "learning_rate": 2.074836802561892e-06, + "loss": 0.1634, + "step": 25088 + }, + { + "epoch": 9.26, + "learning_rate": 2.0747998521985466e-06, + "loss": 0.3838, + "step": 25089 + }, + { + "epoch": 9.27, + "learning_rate": 2.0747629018352016e-06, + "loss": 0.3327, + "step": 25090 + }, + { + "epoch": 9.27, + "learning_rate": 2.0747259514718562e-06, + "loss": 0.4959, + "step": 25091 + }, + { + "epoch": 9.27, + "learning_rate": 2.0746890011085113e-06, + "loss": 0.3719, + "step": 25092 + }, + { + "epoch": 9.27, + "learning_rate": 2.0746520507451655e-06, + "loss": 0.3388, + "step": 25093 + }, + { + "epoch": 9.27, + "learning_rate": 2.07461510038182e-06, + "loss": 0.1923, + "step": 25094 + }, + { + "epoch": 9.27, + "learning_rate": 2.0745781500184752e-06, + "loss": 0.5634, + "step": 25095 + }, + { + "epoch": 9.27, + "learning_rate": 2.07454119965513e-06, + "loss": 0.4381, + "step": 25096 + }, + { + "epoch": 9.27, + "learning_rate": 2.074504249291785e-06, + "loss": 0.265, + "step": 25097 + }, + { + "epoch": 9.27, + "learning_rate": 2.0744672989284395e-06, + "loss": 0.3212, + "step": 25098 + }, + { + "epoch": 9.27, + "learning_rate": 2.074430348565094e-06, + "loss": 0.3623, + "step": 25099 + }, + { + "epoch": 9.27, + "learning_rate": 2.074393398201749e-06, + "loss": 0.3413, + "step": 25100 + }, + { + "epoch": 9.27, + "learning_rate": 2.074356447838404e-06, + "loss": 0.3808, + "step": 25101 + }, + { + "epoch": 9.27, + "learning_rate": 2.0743194974750585e-06, + "loss": 0.3137, + "step": 25102 + }, + { + "epoch": 9.27, + "learning_rate": 2.074282547111713e-06, + "loss": 0.3942, + "step": 25103 + }, + { + "epoch": 9.27, + "learning_rate": 2.0742455967483682e-06, + "loss": 0.3475, + "step": 25104 + }, + { + "epoch": 9.27, + "learning_rate": 2.074208646385023e-06, + "loss": 0.2866, + "step": 25105 + }, + { + "epoch": 9.27, + "learning_rate": 2.0741716960216775e-06, + "loss": 0.296, + "step": 25106 + }, + { + "epoch": 9.27, + "learning_rate": 2.074134745658332e-06, + "loss": 0.301, + "step": 25107 + }, + { + "epoch": 9.27, + "learning_rate": 2.074097795294987e-06, + "loss": 0.4382, + "step": 25108 + }, + { + "epoch": 9.27, + "learning_rate": 2.074060844931642e-06, + "loss": 0.2787, + "step": 25109 + }, + { + "epoch": 9.27, + "learning_rate": 2.074023894568297e-06, + "loss": 0.509, + "step": 25110 + }, + { + "epoch": 9.27, + "learning_rate": 2.0739869442049515e-06, + "loss": 0.2888, + "step": 25111 + }, + { + "epoch": 9.27, + "learning_rate": 2.073949993841606e-06, + "loss": 0.3176, + "step": 25112 + }, + { + "epoch": 9.27, + "learning_rate": 2.073913043478261e-06, + "loss": 0.2817, + "step": 25113 + }, + { + "epoch": 9.27, + "learning_rate": 2.0738760931149154e-06, + "loss": 0.4227, + "step": 25114 + }, + { + "epoch": 9.27, + "learning_rate": 2.0738391427515705e-06, + "loss": 0.2951, + "step": 25115 + }, + { + "epoch": 9.27, + "learning_rate": 2.073802192388225e-06, + "loss": 0.3377, + "step": 25116 + }, + { + "epoch": 9.28, + "learning_rate": 2.07376524202488e-06, + "loss": 0.383, + "step": 25117 + }, + { + "epoch": 9.28, + "learning_rate": 2.073728291661535e-06, + "loss": 0.4137, + "step": 25118 + }, + { + "epoch": 9.28, + "learning_rate": 2.0736913412981895e-06, + "loss": 0.3454, + "step": 25119 + }, + { + "epoch": 9.28, + "learning_rate": 2.073654390934844e-06, + "loss": 0.3773, + "step": 25120 + }, + { + "epoch": 9.28, + "learning_rate": 2.073617440571499e-06, + "loss": 0.2894, + "step": 25121 + }, + { + "epoch": 9.28, + "learning_rate": 2.0735804902081538e-06, + "loss": 0.3349, + "step": 25122 + }, + { + "epoch": 9.28, + "learning_rate": 2.0735435398448084e-06, + "loss": 0.3191, + "step": 25123 + }, + { + "epoch": 9.28, + "learning_rate": 2.0735065894814635e-06, + "loss": 0.2618, + "step": 25124 + }, + { + "epoch": 9.28, + "learning_rate": 2.073469639118118e-06, + "loss": 0.3262, + "step": 25125 + }, + { + "epoch": 9.28, + "learning_rate": 2.0734326887547728e-06, + "loss": 0.2188, + "step": 25126 + }, + { + "epoch": 9.28, + "learning_rate": 2.0733957383914274e-06, + "loss": 0.2939, + "step": 25127 + }, + { + "epoch": 9.28, + "learning_rate": 2.0733587880280825e-06, + "loss": 0.3444, + "step": 25128 + }, + { + "epoch": 9.28, + "learning_rate": 2.073321837664737e-06, + "loss": 0.3769, + "step": 25129 + }, + { + "epoch": 9.28, + "learning_rate": 2.073284887301392e-06, + "loss": 0.4177, + "step": 25130 + }, + { + "epoch": 9.28, + "learning_rate": 2.0732479369380468e-06, + "loss": 0.4142, + "step": 25131 + }, + { + "epoch": 9.28, + "learning_rate": 2.073210986574701e-06, + "loss": 0.2533, + "step": 25132 + }, + { + "epoch": 9.28, + "learning_rate": 2.073174036211356e-06, + "loss": 0.3344, + "step": 25133 + }, + { + "epoch": 9.28, + "learning_rate": 2.0731370858480107e-06, + "loss": 0.3214, + "step": 25134 + }, + { + "epoch": 9.28, + "learning_rate": 2.0731001354846658e-06, + "loss": 0.2439, + "step": 25135 + }, + { + "epoch": 9.28, + "learning_rate": 2.0730631851213204e-06, + "loss": 0.2963, + "step": 25136 + }, + { + "epoch": 9.28, + "learning_rate": 2.0730262347579755e-06, + "loss": 0.277, + "step": 25137 + }, + { + "epoch": 9.28, + "learning_rate": 2.0729892843946297e-06, + "loss": 0.4185, + "step": 25138 + }, + { + "epoch": 9.28, + "learning_rate": 2.0729523340312847e-06, + "loss": 0.3694, + "step": 25139 + }, + { + "epoch": 9.28, + "learning_rate": 2.0729153836679394e-06, + "loss": 0.3324, + "step": 25140 + }, + { + "epoch": 9.28, + "learning_rate": 2.0728784333045944e-06, + "loss": 0.3934, + "step": 25141 + }, + { + "epoch": 9.28, + "learning_rate": 2.072841482941249e-06, + "loss": 0.3258, + "step": 25142 + }, + { + "epoch": 9.28, + "learning_rate": 2.0728045325779037e-06, + "loss": 0.2652, + "step": 25143 + }, + { + "epoch": 9.29, + "learning_rate": 2.0727675822145583e-06, + "loss": 0.5163, + "step": 25144 + }, + { + "epoch": 9.29, + "learning_rate": 2.072730631851213e-06, + "loss": 0.448, + "step": 25145 + }, + { + "epoch": 9.29, + "learning_rate": 2.072693681487868e-06, + "loss": 0.4296, + "step": 25146 + }, + { + "epoch": 9.29, + "learning_rate": 2.0726567311245227e-06, + "loss": 0.3517, + "step": 25147 + }, + { + "epoch": 9.29, + "learning_rate": 2.0726197807611777e-06, + "loss": 0.4121, + "step": 25148 + }, + { + "epoch": 9.29, + "learning_rate": 2.0725828303978324e-06, + "loss": 0.4, + "step": 25149 + }, + { + "epoch": 9.29, + "learning_rate": 2.072545880034487e-06, + "loss": 0.3233, + "step": 25150 + }, + { + "epoch": 9.29, + "learning_rate": 2.0725089296711416e-06, + "loss": 0.3973, + "step": 25151 + }, + { + "epoch": 9.29, + "learning_rate": 2.0724719793077963e-06, + "loss": 0.4675, + "step": 25152 + }, + { + "epoch": 9.29, + "learning_rate": 2.0724350289444513e-06, + "loss": 0.4087, + "step": 25153 + }, + { + "epoch": 9.29, + "learning_rate": 2.072398078581106e-06, + "loss": 0.3598, + "step": 25154 + }, + { + "epoch": 9.29, + "learning_rate": 2.072361128217761e-06, + "loss": 0.2754, + "step": 25155 + }, + { + "epoch": 9.29, + "learning_rate": 2.0723241778544157e-06, + "loss": 0.4676, + "step": 25156 + }, + { + "epoch": 9.29, + "learning_rate": 2.0722872274910703e-06, + "loss": 0.3387, + "step": 25157 + }, + { + "epoch": 9.29, + "learning_rate": 2.072250277127725e-06, + "loss": 0.343, + "step": 25158 + }, + { + "epoch": 9.29, + "learning_rate": 2.07221332676438e-06, + "loss": 0.2776, + "step": 25159 + }, + { + "epoch": 9.29, + "learning_rate": 2.0721763764010346e-06, + "loss": 0.4554, + "step": 25160 + }, + { + "epoch": 9.29, + "learning_rate": 2.0721394260376897e-06, + "loss": 0.2623, + "step": 25161 + }, + { + "epoch": 9.29, + "learning_rate": 2.0721024756743443e-06, + "loss": 0.332, + "step": 25162 + }, + { + "epoch": 9.29, + "learning_rate": 2.072065525310999e-06, + "loss": 0.3676, + "step": 25163 + }, + { + "epoch": 9.29, + "learning_rate": 2.0720285749476536e-06, + "loss": 0.4109, + "step": 25164 + }, + { + "epoch": 9.29, + "learning_rate": 2.0719916245843082e-06, + "loss": 0.3324, + "step": 25165 + }, + { + "epoch": 9.29, + "learning_rate": 2.0719546742209633e-06, + "loss": 0.4176, + "step": 25166 + }, + { + "epoch": 9.29, + "learning_rate": 2.071917723857618e-06, + "loss": 0.3044, + "step": 25167 + }, + { + "epoch": 9.29, + "learning_rate": 2.071880773494273e-06, + "loss": 0.3658, + "step": 25168 + }, + { + "epoch": 9.29, + "learning_rate": 2.0718438231309276e-06, + "loss": 0.3401, + "step": 25169 + }, + { + "epoch": 9.29, + "learning_rate": 2.0718068727675823e-06, + "loss": 0.3392, + "step": 25170 + }, + { + "epoch": 9.3, + "learning_rate": 2.071769922404237e-06, + "loss": 0.3353, + "step": 25171 + }, + { + "epoch": 9.3, + "learning_rate": 2.0717329720408915e-06, + "loss": 0.3052, + "step": 25172 + }, + { + "epoch": 9.3, + "learning_rate": 2.0716960216775466e-06, + "loss": 0.3532, + "step": 25173 + }, + { + "epoch": 9.3, + "learning_rate": 2.0716590713142012e-06, + "loss": 0.3639, + "step": 25174 + }, + { + "epoch": 9.3, + "learning_rate": 2.0716221209508563e-06, + "loss": 0.2567, + "step": 25175 + }, + { + "epoch": 9.3, + "learning_rate": 2.071585170587511e-06, + "loss": 0.4403, + "step": 25176 + }, + { + "epoch": 9.3, + "learning_rate": 2.0715482202241656e-06, + "loss": 0.3483, + "step": 25177 + }, + { + "epoch": 9.3, + "learning_rate": 2.07151126986082e-06, + "loss": 0.3217, + "step": 25178 + }, + { + "epoch": 9.3, + "learning_rate": 2.0714743194974753e-06, + "loss": 0.3463, + "step": 25179 + }, + { + "epoch": 9.3, + "learning_rate": 2.07143736913413e-06, + "loss": 0.4295, + "step": 25180 + }, + { + "epoch": 9.3, + "learning_rate": 2.071400418770785e-06, + "loss": 0.3871, + "step": 25181 + }, + { + "epoch": 9.3, + "learning_rate": 2.0713634684074396e-06, + "loss": 0.3955, + "step": 25182 + }, + { + "epoch": 9.3, + "learning_rate": 2.071326518044094e-06, + "loss": 0.3492, + "step": 25183 + }, + { + "epoch": 9.3, + "learning_rate": 2.071289567680749e-06, + "loss": 0.428, + "step": 25184 + }, + { + "epoch": 9.3, + "learning_rate": 2.0712526173174035e-06, + "loss": 0.3426, + "step": 25185 + }, + { + "epoch": 9.3, + "learning_rate": 2.0712156669540586e-06, + "loss": 0.3408, + "step": 25186 + }, + { + "epoch": 9.3, + "learning_rate": 2.071178716590713e-06, + "loss": 0.3665, + "step": 25187 + }, + { + "epoch": 9.3, + "learning_rate": 2.0711417662273683e-06, + "loss": 0.4768, + "step": 25188 + }, + { + "epoch": 9.3, + "learning_rate": 2.0711048158640225e-06, + "loss": 0.4015, + "step": 25189 + }, + { + "epoch": 9.3, + "learning_rate": 2.0710678655006775e-06, + "loss": 0.3875, + "step": 25190 + }, + { + "epoch": 9.3, + "learning_rate": 2.071030915137332e-06, + "loss": 0.3997, + "step": 25191 + }, + { + "epoch": 9.3, + "learning_rate": 2.070993964773987e-06, + "loss": 0.3417, + "step": 25192 + }, + { + "epoch": 9.3, + "learning_rate": 2.070957014410642e-06, + "loss": 0.3321, + "step": 25193 + }, + { + "epoch": 9.3, + "learning_rate": 2.0709200640472965e-06, + "loss": 0.3469, + "step": 25194 + }, + { + "epoch": 9.3, + "learning_rate": 2.070883113683951e-06, + "loss": 0.3699, + "step": 25195 + }, + { + "epoch": 9.3, + "learning_rate": 2.0708461633206058e-06, + "loss": 0.397, + "step": 25196 + }, + { + "epoch": 9.3, + "learning_rate": 2.070809212957261e-06, + "loss": 0.2535, + "step": 25197 + }, + { + "epoch": 9.31, + "learning_rate": 2.0707722625939155e-06, + "loss": 0.2723, + "step": 25198 + }, + { + "epoch": 9.31, + "learning_rate": 2.0707353122305705e-06, + "loss": 0.3134, + "step": 25199 + }, + { + "epoch": 9.31, + "learning_rate": 2.070698361867225e-06, + "loss": 0.5023, + "step": 25200 + }, + { + "epoch": 9.31, + "learning_rate": 2.07066141150388e-06, + "loss": 0.3373, + "step": 25201 + }, + { + "epoch": 9.31, + "learning_rate": 2.0706244611405344e-06, + "loss": 0.4551, + "step": 25202 + }, + { + "epoch": 9.31, + "learning_rate": 2.070587510777189e-06, + "loss": 0.2838, + "step": 25203 + }, + { + "epoch": 9.31, + "learning_rate": 2.070550560413844e-06, + "loss": 0.3215, + "step": 25204 + }, + { + "epoch": 9.31, + "learning_rate": 2.0705136100504988e-06, + "loss": 0.3475, + "step": 25205 + }, + { + "epoch": 9.31, + "learning_rate": 2.070476659687154e-06, + "loss": 0.3662, + "step": 25206 + }, + { + "epoch": 9.31, + "learning_rate": 2.0704397093238085e-06, + "loss": 0.352, + "step": 25207 + }, + { + "epoch": 9.31, + "learning_rate": 2.070402758960463e-06, + "loss": 0.315, + "step": 25208 + }, + { + "epoch": 9.31, + "learning_rate": 2.0703658085971177e-06, + "loss": 0.5176, + "step": 25209 + }, + { + "epoch": 9.31, + "learning_rate": 2.070328858233773e-06, + "loss": 0.3408, + "step": 25210 + }, + { + "epoch": 9.31, + "learning_rate": 2.0702919078704274e-06, + "loss": 0.3291, + "step": 25211 + }, + { + "epoch": 9.31, + "learning_rate": 2.070254957507082e-06, + "loss": 0.42, + "step": 25212 + }, + { + "epoch": 9.31, + "learning_rate": 2.070218007143737e-06, + "loss": 0.2939, + "step": 25213 + }, + { + "epoch": 9.31, + "learning_rate": 2.0701810567803918e-06, + "loss": 0.4456, + "step": 25214 + }, + { + "epoch": 9.31, + "learning_rate": 2.0701441064170464e-06, + "loss": 0.259, + "step": 25215 + }, + { + "epoch": 9.31, + "learning_rate": 2.070107156053701e-06, + "loss": 0.3338, + "step": 25216 + }, + { + "epoch": 9.31, + "learning_rate": 2.070070205690356e-06, + "loss": 0.4478, + "step": 25217 + }, + { + "epoch": 9.31, + "learning_rate": 2.0700332553270107e-06, + "loss": 0.2887, + "step": 25218 + }, + { + "epoch": 9.31, + "learning_rate": 2.069996304963666e-06, + "loss": 0.4102, + "step": 25219 + }, + { + "epoch": 9.31, + "learning_rate": 2.0699593546003204e-06, + "loss": 0.4612, + "step": 25220 + }, + { + "epoch": 9.31, + "learning_rate": 2.069922404236975e-06, + "loss": 0.4173, + "step": 25221 + }, + { + "epoch": 9.31, + "learning_rate": 2.0698854538736297e-06, + "loss": 0.4167, + "step": 25222 + }, + { + "epoch": 9.31, + "learning_rate": 2.0698485035102843e-06, + "loss": 0.3159, + "step": 25223 + }, + { + "epoch": 9.31, + "learning_rate": 2.0698115531469394e-06, + "loss": 0.3277, + "step": 25224 + }, + { + "epoch": 9.31, + "learning_rate": 2.069774602783594e-06, + "loss": 0.3319, + "step": 25225 + }, + { + "epoch": 9.32, + "learning_rate": 2.069737652420249e-06, + "loss": 0.387, + "step": 25226 + }, + { + "epoch": 9.32, + "learning_rate": 2.0697007020569037e-06, + "loss": 0.3383, + "step": 25227 + }, + { + "epoch": 9.32, + "learning_rate": 2.0696637516935584e-06, + "loss": 0.4438, + "step": 25228 + }, + { + "epoch": 9.32, + "learning_rate": 2.069626801330213e-06, + "loss": 0.3927, + "step": 25229 + }, + { + "epoch": 9.32, + "learning_rate": 2.069589850966868e-06, + "loss": 0.234, + "step": 25230 + }, + { + "epoch": 9.32, + "learning_rate": 2.0695529006035227e-06, + "loss": 0.4554, + "step": 25231 + }, + { + "epoch": 9.32, + "learning_rate": 2.0695159502401773e-06, + "loss": 0.2036, + "step": 25232 + }, + { + "epoch": 9.32, + "learning_rate": 2.0694789998768324e-06, + "loss": 0.4091, + "step": 25233 + }, + { + "epoch": 9.32, + "learning_rate": 2.0694420495134866e-06, + "loss": 0.4327, + "step": 25234 + }, + { + "epoch": 9.32, + "learning_rate": 2.0694050991501417e-06, + "loss": 0.287, + "step": 25235 + }, + { + "epoch": 9.32, + "learning_rate": 2.0693681487867963e-06, + "loss": 0.4009, + "step": 25236 + }, + { + "epoch": 9.32, + "learning_rate": 2.0693311984234514e-06, + "loss": 0.3727, + "step": 25237 + }, + { + "epoch": 9.32, + "learning_rate": 2.069294248060106e-06, + "loss": 0.379, + "step": 25238 + }, + { + "epoch": 9.32, + "learning_rate": 2.069257297696761e-06, + "loss": 0.3974, + "step": 25239 + }, + { + "epoch": 9.32, + "learning_rate": 2.0692203473334153e-06, + "loss": 0.3693, + "step": 25240 + }, + { + "epoch": 9.32, + "learning_rate": 2.0691833969700703e-06, + "loss": 0.4354, + "step": 25241 + }, + { + "epoch": 9.32, + "learning_rate": 2.069146446606725e-06, + "loss": 0.4537, + "step": 25242 + }, + { + "epoch": 9.32, + "learning_rate": 2.0691094962433796e-06, + "loss": 0.334, + "step": 25243 + }, + { + "epoch": 9.32, + "learning_rate": 2.0690725458800347e-06, + "loss": 0.4918, + "step": 25244 + }, + { + "epoch": 9.32, + "learning_rate": 2.0690355955166893e-06, + "loss": 0.2604, + "step": 25245 + }, + { + "epoch": 9.32, + "learning_rate": 2.068998645153344e-06, + "loss": 0.3789, + "step": 25246 + }, + { + "epoch": 9.32, + "learning_rate": 2.0689616947899986e-06, + "loss": 0.2817, + "step": 25247 + }, + { + "epoch": 9.32, + "learning_rate": 2.0689247444266536e-06, + "loss": 0.2816, + "step": 25248 + }, + { + "epoch": 9.32, + "learning_rate": 2.0688877940633083e-06, + "loss": 0.3222, + "step": 25249 + }, + { + "epoch": 9.32, + "learning_rate": 2.0688508436999633e-06, + "loss": 0.2976, + "step": 25250 + }, + { + "epoch": 9.32, + "learning_rate": 2.068813893336618e-06, + "loss": 0.3261, + "step": 25251 + }, + { + "epoch": 9.32, + "learning_rate": 2.0687769429732726e-06, + "loss": 0.3629, + "step": 25252 + }, + { + "epoch": 9.33, + "learning_rate": 2.0687399926099273e-06, + "loss": 0.4082, + "step": 25253 + }, + { + "epoch": 9.33, + "learning_rate": 2.068703042246582e-06, + "loss": 0.3917, + "step": 25254 + }, + { + "epoch": 9.33, + "learning_rate": 2.068666091883237e-06, + "loss": 0.3639, + "step": 25255 + }, + { + "epoch": 9.33, + "learning_rate": 2.0686291415198916e-06, + "loss": 0.502, + "step": 25256 + }, + { + "epoch": 9.33, + "learning_rate": 2.0685921911565466e-06, + "loss": 0.3916, + "step": 25257 + }, + { + "epoch": 9.33, + "learning_rate": 2.0685552407932013e-06, + "loss": 0.2638, + "step": 25258 + }, + { + "epoch": 9.33, + "learning_rate": 2.068518290429856e-06, + "loss": 0.4013, + "step": 25259 + }, + { + "epoch": 9.33, + "learning_rate": 2.0684813400665106e-06, + "loss": 0.3664, + "step": 25260 + }, + { + "epoch": 9.33, + "learning_rate": 2.0684443897031656e-06, + "loss": 0.2964, + "step": 25261 + }, + { + "epoch": 9.33, + "learning_rate": 2.0684074393398203e-06, + "loss": 0.3261, + "step": 25262 + }, + { + "epoch": 9.33, + "learning_rate": 2.068370488976475e-06, + "loss": 0.3528, + "step": 25263 + }, + { + "epoch": 9.33, + "learning_rate": 2.06833353861313e-06, + "loss": 0.4235, + "step": 25264 + }, + { + "epoch": 9.33, + "learning_rate": 2.0682965882497846e-06, + "loss": 0.28, + "step": 25265 + }, + { + "epoch": 9.33, + "learning_rate": 2.0682596378864392e-06, + "loss": 0.35, + "step": 25266 + }, + { + "epoch": 9.33, + "learning_rate": 2.068222687523094e-06, + "loss": 0.3482, + "step": 25267 + }, + { + "epoch": 9.33, + "learning_rate": 2.068185737159749e-06, + "loss": 0.3683, + "step": 25268 + }, + { + "epoch": 9.33, + "learning_rate": 2.0681487867964036e-06, + "loss": 0.3031, + "step": 25269 + }, + { + "epoch": 9.33, + "learning_rate": 2.0681118364330586e-06, + "loss": 0.4321, + "step": 25270 + }, + { + "epoch": 9.33, + "learning_rate": 2.0680748860697133e-06, + "loss": 0.3268, + "step": 25271 + }, + { + "epoch": 9.33, + "learning_rate": 2.068037935706368e-06, + "loss": 0.3251, + "step": 25272 + }, + { + "epoch": 9.33, + "learning_rate": 2.0680009853430225e-06, + "loss": 0.4235, + "step": 25273 + }, + { + "epoch": 9.33, + "learning_rate": 2.067964034979677e-06, + "loss": 0.3349, + "step": 25274 + }, + { + "epoch": 9.33, + "learning_rate": 2.0679270846163322e-06, + "loss": 0.3699, + "step": 25275 + }, + { + "epoch": 9.33, + "learning_rate": 2.067890134252987e-06, + "loss": 0.4196, + "step": 25276 + }, + { + "epoch": 9.33, + "learning_rate": 2.067853183889642e-06, + "loss": 0.3779, + "step": 25277 + }, + { + "epoch": 9.33, + "learning_rate": 2.067816233526296e-06, + "loss": 0.3436, + "step": 25278 + }, + { + "epoch": 9.33, + "learning_rate": 2.067779283162951e-06, + "loss": 0.359, + "step": 25279 + }, + { + "epoch": 9.34, + "learning_rate": 2.067742332799606e-06, + "loss": 0.3124, + "step": 25280 + }, + { + "epoch": 9.34, + "learning_rate": 2.067705382436261e-06, + "loss": 0.3307, + "step": 25281 + }, + { + "epoch": 9.34, + "learning_rate": 2.0676684320729155e-06, + "loss": 0.3714, + "step": 25282 + }, + { + "epoch": 9.34, + "learning_rate": 2.06763148170957e-06, + "loss": 0.3589, + "step": 25283 + }, + { + "epoch": 9.34, + "learning_rate": 2.067594531346225e-06, + "loss": 0.408, + "step": 25284 + }, + { + "epoch": 9.34, + "learning_rate": 2.0675575809828794e-06, + "loss": 0.4164, + "step": 25285 + }, + { + "epoch": 9.34, + "learning_rate": 2.0675206306195345e-06, + "loss": 0.3881, + "step": 25286 + }, + { + "epoch": 9.34, + "learning_rate": 2.067483680256189e-06, + "loss": 0.3207, + "step": 25287 + }, + { + "epoch": 9.34, + "learning_rate": 2.067446729892844e-06, + "loss": 0.3257, + "step": 25288 + }, + { + "epoch": 9.34, + "learning_rate": 2.067409779529499e-06, + "loss": 0.2947, + "step": 25289 + }, + { + "epoch": 9.34, + "learning_rate": 2.0673728291661535e-06, + "loss": 0.3607, + "step": 25290 + }, + { + "epoch": 9.34, + "learning_rate": 2.067335878802808e-06, + "loss": 0.2581, + "step": 25291 + }, + { + "epoch": 9.34, + "learning_rate": 2.0672989284394627e-06, + "loss": 0.4568, + "step": 25292 + }, + { + "epoch": 9.34, + "learning_rate": 2.067261978076118e-06, + "loss": 0.3193, + "step": 25293 + }, + { + "epoch": 9.34, + "learning_rate": 2.0672250277127724e-06, + "loss": 0.3399, + "step": 25294 + }, + { + "epoch": 9.34, + "learning_rate": 2.0671880773494275e-06, + "loss": 0.3372, + "step": 25295 + }, + { + "epoch": 9.34, + "learning_rate": 2.067151126986082e-06, + "loss": 0.4256, + "step": 25296 + }, + { + "epoch": 9.34, + "learning_rate": 2.0671141766227368e-06, + "loss": 0.3162, + "step": 25297 + }, + { + "epoch": 9.34, + "learning_rate": 2.0670772262593914e-06, + "loss": 0.4863, + "step": 25298 + }, + { + "epoch": 9.34, + "learning_rate": 2.0670402758960465e-06, + "loss": 0.4013, + "step": 25299 + }, + { + "epoch": 9.34, + "learning_rate": 2.067003325532701e-06, + "loss": 0.4203, + "step": 25300 + }, + { + "epoch": 9.34, + "learning_rate": 2.066966375169356e-06, + "loss": 0.4032, + "step": 25301 + }, + { + "epoch": 9.34, + "learning_rate": 2.066929424806011e-06, + "loss": 0.3191, + "step": 25302 + }, + { + "epoch": 9.34, + "learning_rate": 2.0668924744426654e-06, + "loss": 0.4063, + "step": 25303 + }, + { + "epoch": 9.34, + "learning_rate": 2.06685552407932e-06, + "loss": 0.3866, + "step": 25304 + }, + { + "epoch": 9.34, + "learning_rate": 2.0668185737159747e-06, + "loss": 0.2969, + "step": 25305 + }, + { + "epoch": 9.34, + "learning_rate": 2.0667816233526298e-06, + "loss": 0.4044, + "step": 25306 + }, + { + "epoch": 9.35, + "learning_rate": 2.0667446729892844e-06, + "loss": 0.4269, + "step": 25307 + }, + { + "epoch": 9.35, + "learning_rate": 2.0667077226259395e-06, + "loss": 0.3127, + "step": 25308 + }, + { + "epoch": 9.35, + "learning_rate": 2.066670772262594e-06, + "loss": 0.3609, + "step": 25309 + }, + { + "epoch": 9.35, + "learning_rate": 2.0666338218992487e-06, + "loss": 0.4622, + "step": 25310 + }, + { + "epoch": 9.35, + "learning_rate": 2.0665968715359034e-06, + "loss": 0.3849, + "step": 25311 + }, + { + "epoch": 9.35, + "learning_rate": 2.066559921172558e-06, + "loss": 0.4552, + "step": 25312 + }, + { + "epoch": 9.35, + "learning_rate": 2.066522970809213e-06, + "loss": 0.4663, + "step": 25313 + }, + { + "epoch": 9.35, + "learning_rate": 2.0664860204458677e-06, + "loss": 0.4862, + "step": 25314 + }, + { + "epoch": 9.35, + "learning_rate": 2.0664490700825228e-06, + "loss": 0.2878, + "step": 25315 + }, + { + "epoch": 9.35, + "learning_rate": 2.0664121197191774e-06, + "loss": 0.3783, + "step": 25316 + }, + { + "epoch": 9.35, + "learning_rate": 2.066375169355832e-06, + "loss": 0.3817, + "step": 25317 + }, + { + "epoch": 9.35, + "learning_rate": 2.0663382189924867e-06, + "loss": 0.3137, + "step": 25318 + }, + { + "epoch": 9.35, + "learning_rate": 2.0663012686291417e-06, + "loss": 0.3728, + "step": 25319 + }, + { + "epoch": 9.35, + "learning_rate": 2.0662643182657964e-06, + "loss": 0.3511, + "step": 25320 + }, + { + "epoch": 9.35, + "learning_rate": 2.0662273679024514e-06, + "loss": 0.3341, + "step": 25321 + }, + { + "epoch": 9.35, + "learning_rate": 2.066190417539106e-06, + "loss": 0.3941, + "step": 25322 + }, + { + "epoch": 9.35, + "learning_rate": 2.0661534671757603e-06, + "loss": 0.361, + "step": 25323 + }, + { + "epoch": 9.35, + "learning_rate": 2.0661165168124153e-06, + "loss": 0.3409, + "step": 25324 + }, + { + "epoch": 9.35, + "learning_rate": 2.06607956644907e-06, + "loss": 0.2822, + "step": 25325 + }, + { + "epoch": 9.35, + "learning_rate": 2.066042616085725e-06, + "loss": 0.5131, + "step": 25326 + }, + { + "epoch": 9.35, + "learning_rate": 2.0660056657223797e-06, + "loss": 0.3636, + "step": 25327 + }, + { + "epoch": 9.35, + "learning_rate": 2.0659687153590347e-06, + "loss": 0.3308, + "step": 25328 + }, + { + "epoch": 9.35, + "learning_rate": 2.065931764995689e-06, + "loss": 0.273, + "step": 25329 + }, + { + "epoch": 9.35, + "learning_rate": 2.065894814632344e-06, + "loss": 0.4972, + "step": 25330 + }, + { + "epoch": 9.35, + "learning_rate": 2.0658578642689986e-06, + "loss": 0.353, + "step": 25331 + }, + { + "epoch": 9.35, + "learning_rate": 2.0658209139056533e-06, + "loss": 0.3315, + "step": 25332 + }, + { + "epoch": 9.35, + "learning_rate": 2.0657839635423083e-06, + "loss": 0.3212, + "step": 25333 + }, + { + "epoch": 9.36, + "learning_rate": 2.065747013178963e-06, + "loss": 0.3224, + "step": 25334 + }, + { + "epoch": 9.36, + "learning_rate": 2.0657100628156176e-06, + "loss": 0.3552, + "step": 25335 + }, + { + "epoch": 9.36, + "learning_rate": 2.0656731124522722e-06, + "loss": 0.4487, + "step": 25336 + }, + { + "epoch": 9.36, + "learning_rate": 2.0656361620889273e-06, + "loss": 0.2861, + "step": 25337 + }, + { + "epoch": 9.36, + "learning_rate": 2.065599211725582e-06, + "loss": 0.3377, + "step": 25338 + }, + { + "epoch": 9.36, + "learning_rate": 2.065562261362237e-06, + "loss": 0.3852, + "step": 25339 + }, + { + "epoch": 9.36, + "learning_rate": 2.0655253109988916e-06, + "loss": 0.3997, + "step": 25340 + }, + { + "epoch": 9.36, + "learning_rate": 2.0654883606355463e-06, + "loss": 0.4738, + "step": 25341 + }, + { + "epoch": 9.36, + "learning_rate": 2.065451410272201e-06, + "loss": 0.3146, + "step": 25342 + }, + { + "epoch": 9.36, + "learning_rate": 2.0654144599088555e-06, + "loss": 0.3941, + "step": 25343 + }, + { + "epoch": 9.36, + "learning_rate": 2.0653775095455106e-06, + "loss": 0.3474, + "step": 25344 + }, + { + "epoch": 9.36, + "learning_rate": 2.0653405591821652e-06, + "loss": 0.2412, + "step": 25345 + }, + { + "epoch": 9.36, + "learning_rate": 2.0653036088188203e-06, + "loss": 0.3631, + "step": 25346 + }, + { + "epoch": 9.36, + "learning_rate": 2.065266658455475e-06, + "loss": 0.5084, + "step": 25347 + }, + { + "epoch": 9.36, + "learning_rate": 2.0652297080921296e-06, + "loss": 0.3265, + "step": 25348 + }, + { + "epoch": 9.36, + "learning_rate": 2.065192757728784e-06, + "loss": 0.4065, + "step": 25349 + }, + { + "epoch": 9.36, + "learning_rate": 2.0651558073654393e-06, + "loss": 0.3525, + "step": 25350 + }, + { + "epoch": 9.36, + "learning_rate": 2.065118857002094e-06, + "loss": 0.367, + "step": 25351 + }, + { + "epoch": 9.36, + "learning_rate": 2.0650819066387485e-06, + "loss": 0.3101, + "step": 25352 + }, + { + "epoch": 9.36, + "learning_rate": 2.0650449562754036e-06, + "loss": 0.287, + "step": 25353 + }, + { + "epoch": 9.36, + "learning_rate": 2.0650080059120582e-06, + "loss": 0.3275, + "step": 25354 + }, + { + "epoch": 9.36, + "learning_rate": 2.064971055548713e-06, + "loss": 0.3569, + "step": 25355 + }, + { + "epoch": 9.36, + "learning_rate": 2.0649341051853675e-06, + "loss": 0.4022, + "step": 25356 + }, + { + "epoch": 9.36, + "learning_rate": 2.0648971548220226e-06, + "loss": 0.2357, + "step": 25357 + }, + { + "epoch": 9.36, + "learning_rate": 2.064860204458677e-06, + "loss": 0.2774, + "step": 25358 + }, + { + "epoch": 9.36, + "learning_rate": 2.0648232540953323e-06, + "loss": 0.4577, + "step": 25359 + }, + { + "epoch": 9.36, + "learning_rate": 2.064786303731987e-06, + "loss": 0.4333, + "step": 25360 + }, + { + "epoch": 9.37, + "learning_rate": 2.0647493533686415e-06, + "loss": 0.3221, + "step": 25361 + }, + { + "epoch": 9.37, + "learning_rate": 2.064712403005296e-06, + "loss": 0.3263, + "step": 25362 + }, + { + "epoch": 9.37, + "learning_rate": 2.064675452641951e-06, + "loss": 0.3146, + "step": 25363 + }, + { + "epoch": 9.37, + "learning_rate": 2.064638502278606e-06, + "loss": 0.4253, + "step": 25364 + }, + { + "epoch": 9.37, + "learning_rate": 2.0646015519152605e-06, + "loss": 0.271, + "step": 25365 + }, + { + "epoch": 9.37, + "learning_rate": 2.0645646015519156e-06, + "loss": 0.4504, + "step": 25366 + }, + { + "epoch": 9.37, + "learning_rate": 2.06452765118857e-06, + "loss": 0.3058, + "step": 25367 + }, + { + "epoch": 9.37, + "learning_rate": 2.064490700825225e-06, + "loss": 0.415, + "step": 25368 + }, + { + "epoch": 9.37, + "learning_rate": 2.0644537504618795e-06, + "loss": 0.325, + "step": 25369 + }, + { + "epoch": 9.37, + "learning_rate": 2.0644168000985345e-06, + "loss": 0.2573, + "step": 25370 + }, + { + "epoch": 9.37, + "learning_rate": 2.064379849735189e-06, + "loss": 0.2841, + "step": 25371 + }, + { + "epoch": 9.37, + "learning_rate": 2.064342899371844e-06, + "loss": 0.3887, + "step": 25372 + }, + { + "epoch": 9.37, + "learning_rate": 2.064305949008499e-06, + "loss": 0.3238, + "step": 25373 + }, + { + "epoch": 9.37, + "learning_rate": 2.064268998645153e-06, + "loss": 0.3798, + "step": 25374 + }, + { + "epoch": 9.37, + "learning_rate": 2.064232048281808e-06, + "loss": 0.3381, + "step": 25375 + }, + { + "epoch": 9.37, + "learning_rate": 2.0641950979184628e-06, + "loss": 0.2936, + "step": 25376 + }, + { + "epoch": 9.37, + "learning_rate": 2.064158147555118e-06, + "loss": 0.3362, + "step": 25377 + }, + { + "epoch": 9.37, + "learning_rate": 2.0641211971917725e-06, + "loss": 0.5018, + "step": 25378 + }, + { + "epoch": 9.37, + "learning_rate": 2.0640842468284275e-06, + "loss": 0.2923, + "step": 25379 + }, + { + "epoch": 9.37, + "learning_rate": 2.0640472964650818e-06, + "loss": 0.2793, + "step": 25380 + }, + { + "epoch": 9.37, + "learning_rate": 2.0640103461017364e-06, + "loss": 0.3568, + "step": 25381 + }, + { + "epoch": 9.37, + "learning_rate": 2.0639733957383914e-06, + "loss": 0.5055, + "step": 25382 + }, + { + "epoch": 9.37, + "learning_rate": 2.063936445375046e-06, + "loss": 0.4506, + "step": 25383 + }, + { + "epoch": 9.37, + "learning_rate": 2.063899495011701e-06, + "loss": 0.3866, + "step": 25384 + }, + { + "epoch": 9.37, + "learning_rate": 2.0638625446483558e-06, + "loss": 0.4212, + "step": 25385 + }, + { + "epoch": 9.37, + "learning_rate": 2.0638255942850104e-06, + "loss": 0.2833, + "step": 25386 + }, + { + "epoch": 9.37, + "learning_rate": 2.063788643921665e-06, + "loss": 0.4537, + "step": 25387 + }, + { + "epoch": 9.38, + "learning_rate": 2.06375169355832e-06, + "loss": 0.3249, + "step": 25388 + }, + { + "epoch": 9.38, + "learning_rate": 2.0637147431949748e-06, + "loss": 0.375, + "step": 25389 + }, + { + "epoch": 9.38, + "learning_rate": 2.06367779283163e-06, + "loss": 0.3622, + "step": 25390 + }, + { + "epoch": 9.38, + "learning_rate": 2.0636408424682844e-06, + "loss": 0.5395, + "step": 25391 + }, + { + "epoch": 9.38, + "learning_rate": 2.063603892104939e-06, + "loss": 0.3028, + "step": 25392 + }, + { + "epoch": 9.38, + "learning_rate": 2.0635669417415937e-06, + "loss": 0.4193, + "step": 25393 + }, + { + "epoch": 9.38, + "learning_rate": 2.0635299913782484e-06, + "loss": 0.2648, + "step": 25394 + }, + { + "epoch": 9.38, + "learning_rate": 2.0634930410149034e-06, + "loss": 0.3157, + "step": 25395 + }, + { + "epoch": 9.38, + "learning_rate": 2.063456090651558e-06, + "loss": 0.4759, + "step": 25396 + }, + { + "epoch": 9.38, + "learning_rate": 2.063419140288213e-06, + "loss": 0.3402, + "step": 25397 + }, + { + "epoch": 9.38, + "learning_rate": 2.0633821899248677e-06, + "loss": 0.2921, + "step": 25398 + }, + { + "epoch": 9.38, + "learning_rate": 2.0633452395615224e-06, + "loss": 0.2371, + "step": 25399 + }, + { + "epoch": 9.38, + "learning_rate": 2.063308289198177e-06, + "loss": 0.2885, + "step": 25400 + }, + { + "epoch": 9.38, + "learning_rate": 2.0632713388348317e-06, + "loss": 0.4773, + "step": 25401 + }, + { + "epoch": 9.38, + "learning_rate": 2.0632343884714867e-06, + "loss": 0.4373, + "step": 25402 + }, + { + "epoch": 9.38, + "learning_rate": 2.0631974381081414e-06, + "loss": 0.3805, + "step": 25403 + }, + { + "epoch": 9.38, + "learning_rate": 2.0631604877447964e-06, + "loss": 0.4367, + "step": 25404 + }, + { + "epoch": 9.38, + "learning_rate": 2.063123537381451e-06, + "loss": 0.438, + "step": 25405 + }, + { + "epoch": 9.38, + "learning_rate": 2.0630865870181057e-06, + "loss": 0.4678, + "step": 25406 + }, + { + "epoch": 9.38, + "learning_rate": 2.0630496366547603e-06, + "loss": 0.3067, + "step": 25407 + }, + { + "epoch": 9.38, + "learning_rate": 2.0630126862914154e-06, + "loss": 0.3298, + "step": 25408 + }, + { + "epoch": 9.38, + "learning_rate": 2.06297573592807e-06, + "loss": 0.5293, + "step": 25409 + }, + { + "epoch": 9.38, + "learning_rate": 2.062938785564725e-06, + "loss": 0.309, + "step": 25410 + }, + { + "epoch": 9.38, + "learning_rate": 2.0629018352013797e-06, + "loss": 0.3333, + "step": 25411 + }, + { + "epoch": 9.38, + "learning_rate": 2.0628648848380344e-06, + "loss": 0.4027, + "step": 25412 + }, + { + "epoch": 9.38, + "learning_rate": 2.062827934474689e-06, + "loss": 0.374, + "step": 25413 + }, + { + "epoch": 9.38, + "learning_rate": 2.0627909841113436e-06, + "loss": 0.3575, + "step": 25414 + }, + { + "epoch": 9.39, + "learning_rate": 2.0627540337479987e-06, + "loss": 0.4085, + "step": 25415 + }, + { + "epoch": 9.39, + "learning_rate": 2.0627170833846533e-06, + "loss": 0.3827, + "step": 25416 + }, + { + "epoch": 9.39, + "learning_rate": 2.0626801330213084e-06, + "loss": 0.3674, + "step": 25417 + }, + { + "epoch": 9.39, + "learning_rate": 2.062643182657963e-06, + "loss": 0.3322, + "step": 25418 + }, + { + "epoch": 9.39, + "learning_rate": 2.0626062322946177e-06, + "loss": 0.3067, + "step": 25419 + }, + { + "epoch": 9.39, + "learning_rate": 2.0625692819312723e-06, + "loss": 0.5057, + "step": 25420 + }, + { + "epoch": 9.39, + "learning_rate": 2.062532331567927e-06, + "loss": 0.4341, + "step": 25421 + }, + { + "epoch": 9.39, + "learning_rate": 2.062495381204582e-06, + "loss": 0.3347, + "step": 25422 + }, + { + "epoch": 9.39, + "learning_rate": 2.0624584308412366e-06, + "loss": 0.2592, + "step": 25423 + }, + { + "epoch": 9.39, + "learning_rate": 2.0624214804778917e-06, + "loss": 0.2412, + "step": 25424 + }, + { + "epoch": 9.39, + "learning_rate": 2.062384530114546e-06, + "loss": 0.2118, + "step": 25425 + }, + { + "epoch": 9.39, + "learning_rate": 2.062347579751201e-06, + "loss": 0.2165, + "step": 25426 + }, + { + "epoch": 9.39, + "learning_rate": 2.0623106293878556e-06, + "loss": 0.4081, + "step": 25427 + }, + { + "epoch": 9.39, + "learning_rate": 2.0622736790245107e-06, + "loss": 0.3188, + "step": 25428 + }, + { + "epoch": 9.39, + "learning_rate": 2.0622367286611653e-06, + "loss": 0.5342, + "step": 25429 + }, + { + "epoch": 9.39, + "learning_rate": 2.06219977829782e-06, + "loss": 0.3967, + "step": 25430 + }, + { + "epoch": 9.39, + "learning_rate": 2.0621628279344746e-06, + "loss": 0.3831, + "step": 25431 + }, + { + "epoch": 9.39, + "learning_rate": 2.062125877571129e-06, + "loss": 0.3059, + "step": 25432 + }, + { + "epoch": 9.39, + "learning_rate": 2.0620889272077843e-06, + "loss": 0.3835, + "step": 25433 + }, + { + "epoch": 9.39, + "learning_rate": 2.062051976844439e-06, + "loss": 0.2497, + "step": 25434 + }, + { + "epoch": 9.39, + "learning_rate": 2.062015026481094e-06, + "loss": 0.2627, + "step": 25435 + }, + { + "epoch": 9.39, + "learning_rate": 2.0619780761177486e-06, + "loss": 0.4448, + "step": 25436 + }, + { + "epoch": 9.39, + "learning_rate": 2.0619411257544032e-06, + "loss": 0.317, + "step": 25437 + }, + { + "epoch": 9.39, + "learning_rate": 2.061904175391058e-06, + "loss": 0.3216, + "step": 25438 + }, + { + "epoch": 9.39, + "learning_rate": 2.061867225027713e-06, + "loss": 0.3805, + "step": 25439 + }, + { + "epoch": 9.39, + "learning_rate": 2.0618302746643676e-06, + "loss": 0.2954, + "step": 25440 + }, + { + "epoch": 9.39, + "learning_rate": 2.061793324301022e-06, + "loss": 0.3901, + "step": 25441 + }, + { + "epoch": 9.4, + "learning_rate": 2.0617563739376773e-06, + "loss": 0.5152, + "step": 25442 + }, + { + "epoch": 9.4, + "learning_rate": 2.061719423574332e-06, + "loss": 0.379, + "step": 25443 + }, + { + "epoch": 9.4, + "learning_rate": 2.0616824732109865e-06, + "loss": 0.4469, + "step": 25444 + }, + { + "epoch": 9.4, + "learning_rate": 2.061645522847641e-06, + "loss": 0.4003, + "step": 25445 + }, + { + "epoch": 9.4, + "learning_rate": 2.0616085724842962e-06, + "loss": 0.466, + "step": 25446 + }, + { + "epoch": 9.4, + "learning_rate": 2.061571622120951e-06, + "loss": 0.3437, + "step": 25447 + }, + { + "epoch": 9.4, + "learning_rate": 2.061534671757606e-06, + "loss": 0.2736, + "step": 25448 + }, + { + "epoch": 9.4, + "learning_rate": 2.0614977213942606e-06, + "loss": 0.5728, + "step": 25449 + }, + { + "epoch": 9.4, + "learning_rate": 2.061460771030915e-06, + "loss": 0.4708, + "step": 25450 + }, + { + "epoch": 9.4, + "learning_rate": 2.06142382066757e-06, + "loss": 0.5663, + "step": 25451 + }, + { + "epoch": 9.4, + "learning_rate": 2.0613868703042245e-06, + "loss": 0.4707, + "step": 25452 + }, + { + "epoch": 9.4, + "learning_rate": 2.0613499199408795e-06, + "loss": 0.4116, + "step": 25453 + }, + { + "epoch": 9.4, + "learning_rate": 2.061312969577534e-06, + "loss": 0.4802, + "step": 25454 + }, + { + "epoch": 9.4, + "learning_rate": 2.0612760192141892e-06, + "loss": 0.2278, + "step": 25455 + }, + { + "epoch": 9.4, + "learning_rate": 2.061239068850844e-06, + "loss": 0.3235, + "step": 25456 + }, + { + "epoch": 9.4, + "learning_rate": 2.0612021184874985e-06, + "loss": 0.4988, + "step": 25457 + }, + { + "epoch": 9.4, + "learning_rate": 2.061165168124153e-06, + "loss": 0.3146, + "step": 25458 + }, + { + "epoch": 9.4, + "learning_rate": 2.061128217760808e-06, + "loss": 0.3755, + "step": 25459 + }, + { + "epoch": 9.4, + "learning_rate": 2.061091267397463e-06, + "loss": 0.3275, + "step": 25460 + }, + { + "epoch": 9.4, + "learning_rate": 2.0610543170341175e-06, + "loss": 0.2914, + "step": 25461 + }, + { + "epoch": 9.4, + "learning_rate": 2.0610173666707725e-06, + "loss": 0.314, + "step": 25462 + }, + { + "epoch": 9.4, + "learning_rate": 2.0609804163074267e-06, + "loss": 0.3914, + "step": 25463 + }, + { + "epoch": 9.4, + "learning_rate": 2.060943465944082e-06, + "loss": 0.3878, + "step": 25464 + }, + { + "epoch": 9.4, + "learning_rate": 2.0609065155807364e-06, + "loss": 0.3839, + "step": 25465 + }, + { + "epoch": 9.4, + "learning_rate": 2.0608695652173915e-06, + "loss": 0.449, + "step": 25466 + }, + { + "epoch": 9.4, + "learning_rate": 2.060832614854046e-06, + "loss": 0.2498, + "step": 25467 + }, + { + "epoch": 9.4, + "learning_rate": 2.060795664490701e-06, + "loss": 0.2711, + "step": 25468 + }, + { + "epoch": 9.41, + "learning_rate": 2.0607587141273554e-06, + "loss": 0.4985, + "step": 25469 + }, + { + "epoch": 9.41, + "learning_rate": 2.0607217637640105e-06, + "loss": 0.3349, + "step": 25470 + }, + { + "epoch": 9.41, + "learning_rate": 2.060684813400665e-06, + "loss": 0.4076, + "step": 25471 + }, + { + "epoch": 9.41, + "learning_rate": 2.0606478630373197e-06, + "loss": 0.4732, + "step": 25472 + }, + { + "epoch": 9.41, + "learning_rate": 2.060610912673975e-06, + "loss": 0.4908, + "step": 25473 + }, + { + "epoch": 9.41, + "learning_rate": 2.0605739623106294e-06, + "loss": 0.4122, + "step": 25474 + }, + { + "epoch": 9.41, + "learning_rate": 2.060537011947284e-06, + "loss": 0.3156, + "step": 25475 + }, + { + "epoch": 9.41, + "learning_rate": 2.0605000615839387e-06, + "loss": 0.3173, + "step": 25476 + }, + { + "epoch": 9.41, + "learning_rate": 2.0604631112205938e-06, + "loss": 0.4516, + "step": 25477 + }, + { + "epoch": 9.41, + "learning_rate": 2.0604261608572484e-06, + "loss": 0.4401, + "step": 25478 + }, + { + "epoch": 9.41, + "learning_rate": 2.0603892104939035e-06, + "loss": 0.4231, + "step": 25479 + }, + { + "epoch": 9.41, + "learning_rate": 2.060352260130558e-06, + "loss": 0.5611, + "step": 25480 + }, + { + "epoch": 9.41, + "learning_rate": 2.0603153097672127e-06, + "loss": 0.4074, + "step": 25481 + }, + { + "epoch": 9.41, + "learning_rate": 2.0602783594038674e-06, + "loss": 0.3424, + "step": 25482 + }, + { + "epoch": 9.41, + "learning_rate": 2.060241409040522e-06, + "loss": 0.3714, + "step": 25483 + }, + { + "epoch": 9.41, + "learning_rate": 2.060204458677177e-06, + "loss": 0.338, + "step": 25484 + }, + { + "epoch": 9.41, + "learning_rate": 2.0601675083138317e-06, + "loss": 0.486, + "step": 25485 + }, + { + "epoch": 9.41, + "learning_rate": 2.0601305579504868e-06, + "loss": 0.4526, + "step": 25486 + }, + { + "epoch": 9.41, + "learning_rate": 2.0600936075871414e-06, + "loss": 0.4621, + "step": 25487 + }, + { + "epoch": 9.41, + "learning_rate": 2.060056657223796e-06, + "loss": 0.2755, + "step": 25488 + }, + { + "epoch": 9.41, + "learning_rate": 2.0600197068604507e-06, + "loss": 0.3735, + "step": 25489 + }, + { + "epoch": 9.41, + "learning_rate": 2.0599827564971057e-06, + "loss": 0.3821, + "step": 25490 + }, + { + "epoch": 9.41, + "learning_rate": 2.0599458061337604e-06, + "loss": 0.3881, + "step": 25491 + }, + { + "epoch": 9.41, + "learning_rate": 2.059908855770415e-06, + "loss": 0.3877, + "step": 25492 + }, + { + "epoch": 9.41, + "learning_rate": 2.05987190540707e-06, + "loss": 0.3691, + "step": 25493 + }, + { + "epoch": 9.41, + "learning_rate": 2.0598349550437247e-06, + "loss": 0.3535, + "step": 25494 + }, + { + "epoch": 9.41, + "learning_rate": 2.0597980046803793e-06, + "loss": 0.3997, + "step": 25495 + }, + { + "epoch": 9.42, + "learning_rate": 2.059761054317034e-06, + "loss": 0.2967, + "step": 25496 + }, + { + "epoch": 9.42, + "learning_rate": 2.059724103953689e-06, + "loss": 0.3082, + "step": 25497 + }, + { + "epoch": 9.42, + "learning_rate": 2.0596871535903437e-06, + "loss": 0.2804, + "step": 25498 + }, + { + "epoch": 9.42, + "learning_rate": 2.0596502032269987e-06, + "loss": 0.359, + "step": 25499 + }, + { + "epoch": 9.42, + "learning_rate": 2.0596132528636534e-06, + "loss": 0.238, + "step": 25500 + }, + { + "epoch": 9.42, + "learning_rate": 2.059576302500308e-06, + "loss": 0.3062, + "step": 25501 + }, + { + "epoch": 9.42, + "learning_rate": 2.0595393521369626e-06, + "loss": 0.3159, + "step": 25502 + }, + { + "epoch": 9.42, + "learning_rate": 2.0595024017736173e-06, + "loss": 0.3468, + "step": 25503 + }, + { + "epoch": 9.42, + "learning_rate": 2.0594654514102723e-06, + "loss": 0.3866, + "step": 25504 + }, + { + "epoch": 9.42, + "learning_rate": 2.059428501046927e-06, + "loss": 0.3363, + "step": 25505 + }, + { + "epoch": 9.42, + "learning_rate": 2.059391550683582e-06, + "loss": 0.3886, + "step": 25506 + }, + { + "epoch": 9.42, + "learning_rate": 2.0593546003202367e-06, + "loss": 0.4291, + "step": 25507 + }, + { + "epoch": 9.42, + "learning_rate": 2.0593176499568913e-06, + "loss": 0.4905, + "step": 25508 + }, + { + "epoch": 9.42, + "learning_rate": 2.059280699593546e-06, + "loss": 0.4089, + "step": 25509 + }, + { + "epoch": 9.42, + "learning_rate": 2.059243749230201e-06, + "loss": 0.3793, + "step": 25510 + }, + { + "epoch": 9.42, + "learning_rate": 2.0592067988668556e-06, + "loss": 0.3365, + "step": 25511 + }, + { + "epoch": 9.42, + "learning_rate": 2.0591698485035103e-06, + "loss": 0.3679, + "step": 25512 + }, + { + "epoch": 9.42, + "learning_rate": 2.0591328981401653e-06, + "loss": 0.4609, + "step": 25513 + }, + { + "epoch": 9.42, + "learning_rate": 2.0590959477768196e-06, + "loss": 0.3754, + "step": 25514 + }, + { + "epoch": 9.42, + "learning_rate": 2.0590589974134746e-06, + "loss": 0.4208, + "step": 25515 + }, + { + "epoch": 9.42, + "learning_rate": 2.0590220470501292e-06, + "loss": 0.2929, + "step": 25516 + }, + { + "epoch": 9.42, + "learning_rate": 2.0589850966867843e-06, + "loss": 0.352, + "step": 25517 + }, + { + "epoch": 9.42, + "learning_rate": 2.058948146323439e-06, + "loss": 0.3406, + "step": 25518 + }, + { + "epoch": 9.42, + "learning_rate": 2.058911195960094e-06, + "loss": 0.4233, + "step": 25519 + }, + { + "epoch": 9.42, + "learning_rate": 2.0588742455967482e-06, + "loss": 0.3107, + "step": 25520 + }, + { + "epoch": 9.42, + "learning_rate": 2.058837295233403e-06, + "loss": 0.4122, + "step": 25521 + }, + { + "epoch": 9.42, + "learning_rate": 2.058800344870058e-06, + "loss": 0.3127, + "step": 25522 + }, + { + "epoch": 9.43, + "learning_rate": 2.0587633945067125e-06, + "loss": 0.2906, + "step": 25523 + }, + { + "epoch": 9.43, + "learning_rate": 2.0587264441433676e-06, + "loss": 0.2447, + "step": 25524 + }, + { + "epoch": 9.43, + "learning_rate": 2.0586894937800222e-06, + "loss": 0.3883, + "step": 25525 + }, + { + "epoch": 9.43, + "learning_rate": 2.058652543416677e-06, + "loss": 0.3502, + "step": 25526 + }, + { + "epoch": 9.43, + "learning_rate": 2.0586155930533315e-06, + "loss": 0.406, + "step": 25527 + }, + { + "epoch": 9.43, + "learning_rate": 2.0585786426899866e-06, + "loss": 0.3511, + "step": 25528 + }, + { + "epoch": 9.43, + "learning_rate": 2.0585416923266412e-06, + "loss": 0.3614, + "step": 25529 + }, + { + "epoch": 9.43, + "learning_rate": 2.0585047419632963e-06, + "loss": 0.3694, + "step": 25530 + }, + { + "epoch": 9.43, + "learning_rate": 2.058467791599951e-06, + "loss": 0.2909, + "step": 25531 + }, + { + "epoch": 9.43, + "learning_rate": 2.0584308412366055e-06, + "loss": 0.3658, + "step": 25532 + }, + { + "epoch": 9.43, + "learning_rate": 2.05839389087326e-06, + "loss": 0.49, + "step": 25533 + }, + { + "epoch": 9.43, + "learning_rate": 2.058356940509915e-06, + "loss": 0.2496, + "step": 25534 + }, + { + "epoch": 9.43, + "learning_rate": 2.05831999014657e-06, + "loss": 0.3764, + "step": 25535 + }, + { + "epoch": 9.43, + "learning_rate": 2.0582830397832245e-06, + "loss": 0.278, + "step": 25536 + }, + { + "epoch": 9.43, + "learning_rate": 2.0582460894198796e-06, + "loss": 0.3657, + "step": 25537 + }, + { + "epoch": 9.43, + "learning_rate": 2.0582091390565342e-06, + "loss": 0.3638, + "step": 25538 + }, + { + "epoch": 9.43, + "learning_rate": 2.058172188693189e-06, + "loss": 0.4886, + "step": 25539 + }, + { + "epoch": 9.43, + "learning_rate": 2.0581352383298435e-06, + "loss": 0.3798, + "step": 25540 + }, + { + "epoch": 9.43, + "learning_rate": 2.058098287966498e-06, + "loss": 0.4936, + "step": 25541 + }, + { + "epoch": 9.43, + "learning_rate": 2.058061337603153e-06, + "loss": 0.4186, + "step": 25542 + }, + { + "epoch": 9.43, + "learning_rate": 2.058024387239808e-06, + "loss": 0.3818, + "step": 25543 + }, + { + "epoch": 9.43, + "learning_rate": 2.057987436876463e-06, + "loss": 0.3586, + "step": 25544 + }, + { + "epoch": 9.43, + "learning_rate": 2.0579504865131175e-06, + "loss": 0.4678, + "step": 25545 + }, + { + "epoch": 9.43, + "learning_rate": 2.057913536149772e-06, + "loss": 0.439, + "step": 25546 + }, + { + "epoch": 9.43, + "learning_rate": 2.057876585786427e-06, + "loss": 0.2524, + "step": 25547 + }, + { + "epoch": 9.43, + "learning_rate": 2.057839635423082e-06, + "loss": 0.3559, + "step": 25548 + }, + { + "epoch": 9.43, + "learning_rate": 2.0578026850597365e-06, + "loss": 0.258, + "step": 25549 + }, + { + "epoch": 9.44, + "learning_rate": 2.0577657346963915e-06, + "loss": 0.3041, + "step": 25550 + }, + { + "epoch": 9.44, + "learning_rate": 2.057728784333046e-06, + "loss": 0.4818, + "step": 25551 + }, + { + "epoch": 9.44, + "learning_rate": 2.057691833969701e-06, + "loss": 0.3014, + "step": 25552 + }, + { + "epoch": 9.44, + "learning_rate": 2.0576548836063555e-06, + "loss": 0.3593, + "step": 25553 + }, + { + "epoch": 9.44, + "learning_rate": 2.05761793324301e-06, + "loss": 0.3952, + "step": 25554 + }, + { + "epoch": 9.44, + "learning_rate": 2.057580982879665e-06, + "loss": 0.3579, + "step": 25555 + }, + { + "epoch": 9.44, + "learning_rate": 2.0575440325163198e-06, + "loss": 0.3038, + "step": 25556 + }, + { + "epoch": 9.44, + "learning_rate": 2.057507082152975e-06, + "loss": 0.2836, + "step": 25557 + }, + { + "epoch": 9.44, + "learning_rate": 2.0574701317896295e-06, + "loss": 0.3633, + "step": 25558 + }, + { + "epoch": 9.44, + "learning_rate": 2.057433181426284e-06, + "loss": 0.3282, + "step": 25559 + }, + { + "epoch": 9.44, + "learning_rate": 2.0573962310629388e-06, + "loss": 0.4815, + "step": 25560 + }, + { + "epoch": 9.44, + "learning_rate": 2.0573592806995934e-06, + "loss": 0.3402, + "step": 25561 + }, + { + "epoch": 9.44, + "learning_rate": 2.0573223303362485e-06, + "loss": 0.3961, + "step": 25562 + }, + { + "epoch": 9.44, + "learning_rate": 2.057285379972903e-06, + "loss": 0.4204, + "step": 25563 + }, + { + "epoch": 9.44, + "learning_rate": 2.057248429609558e-06, + "loss": 0.4855, + "step": 25564 + }, + { + "epoch": 9.44, + "learning_rate": 2.0572114792462124e-06, + "loss": 0.3632, + "step": 25565 + }, + { + "epoch": 9.44, + "learning_rate": 2.0571745288828674e-06, + "loss": 0.3112, + "step": 25566 + }, + { + "epoch": 9.44, + "learning_rate": 2.057137578519522e-06, + "loss": 0.388, + "step": 25567 + }, + { + "epoch": 9.44, + "learning_rate": 2.057100628156177e-06, + "loss": 0.4, + "step": 25568 + }, + { + "epoch": 9.44, + "learning_rate": 2.0570636777928318e-06, + "loss": 0.2909, + "step": 25569 + }, + { + "epoch": 9.44, + "learning_rate": 2.0570267274294864e-06, + "loss": 0.3736, + "step": 25570 + }, + { + "epoch": 9.44, + "learning_rate": 2.056989777066141e-06, + "loss": 0.3404, + "step": 25571 + }, + { + "epoch": 9.44, + "learning_rate": 2.0569528267027957e-06, + "loss": 0.3759, + "step": 25572 + }, + { + "epoch": 9.44, + "learning_rate": 2.0569158763394507e-06, + "loss": 0.4928, + "step": 25573 + }, + { + "epoch": 9.44, + "learning_rate": 2.0568789259761054e-06, + "loss": 0.3733, + "step": 25574 + }, + { + "epoch": 9.44, + "learning_rate": 2.0568419756127604e-06, + "loss": 0.3559, + "step": 25575 + }, + { + "epoch": 9.44, + "learning_rate": 2.056805025249415e-06, + "loss": 0.3312, + "step": 25576 + }, + { + "epoch": 9.44, + "learning_rate": 2.0567680748860697e-06, + "loss": 0.443, + "step": 25577 + }, + { + "epoch": 9.45, + "learning_rate": 2.0567311245227243e-06, + "loss": 0.4043, + "step": 25578 + }, + { + "epoch": 9.45, + "learning_rate": 2.0566941741593794e-06, + "loss": 0.3463, + "step": 25579 + }, + { + "epoch": 9.45, + "learning_rate": 2.056657223796034e-06, + "loss": 0.3599, + "step": 25580 + }, + { + "epoch": 9.45, + "learning_rate": 2.0566202734326887e-06, + "loss": 0.5053, + "step": 25581 + }, + { + "epoch": 9.45, + "learning_rate": 2.0565833230693437e-06, + "loss": 0.4037, + "step": 25582 + }, + { + "epoch": 9.45, + "learning_rate": 2.0565463727059984e-06, + "loss": 0.336, + "step": 25583 + }, + { + "epoch": 9.45, + "learning_rate": 2.056509422342653e-06, + "loss": 0.4989, + "step": 25584 + }, + { + "epoch": 9.45, + "learning_rate": 2.0564724719793076e-06, + "loss": 0.3192, + "step": 25585 + }, + { + "epoch": 9.45, + "learning_rate": 2.0564355216159627e-06, + "loss": 0.3854, + "step": 25586 + }, + { + "epoch": 9.45, + "learning_rate": 2.0563985712526173e-06, + "loss": 0.3378, + "step": 25587 + }, + { + "epoch": 9.45, + "learning_rate": 2.0563616208892724e-06, + "loss": 0.4673, + "step": 25588 + }, + { + "epoch": 9.45, + "learning_rate": 2.056324670525927e-06, + "loss": 0.5254, + "step": 25589 + }, + { + "epoch": 9.45, + "learning_rate": 2.0562877201625817e-06, + "loss": 0.3292, + "step": 25590 + }, + { + "epoch": 9.45, + "learning_rate": 2.0562507697992363e-06, + "loss": 0.3002, + "step": 25591 + }, + { + "epoch": 9.45, + "learning_rate": 2.056213819435891e-06, + "loss": 0.2884, + "step": 25592 + }, + { + "epoch": 9.45, + "learning_rate": 2.056176869072546e-06, + "loss": 0.3433, + "step": 25593 + }, + { + "epoch": 9.45, + "learning_rate": 2.0561399187092006e-06, + "loss": 0.3615, + "step": 25594 + }, + { + "epoch": 9.45, + "learning_rate": 2.0561029683458557e-06, + "loss": 0.2914, + "step": 25595 + }, + { + "epoch": 9.45, + "learning_rate": 2.0560660179825103e-06, + "loss": 0.4242, + "step": 25596 + }, + { + "epoch": 9.45, + "learning_rate": 2.056029067619165e-06, + "loss": 0.4027, + "step": 25597 + }, + { + "epoch": 9.45, + "learning_rate": 2.0559921172558196e-06, + "loss": 0.3552, + "step": 25598 + }, + { + "epoch": 9.45, + "learning_rate": 2.0559551668924747e-06, + "loss": 0.4764, + "step": 25599 + }, + { + "epoch": 9.45, + "learning_rate": 2.0559182165291293e-06, + "loss": 0.4014, + "step": 25600 + }, + { + "epoch": 9.45, + "learning_rate": 2.055881266165784e-06, + "loss": 0.4504, + "step": 25601 + }, + { + "epoch": 9.45, + "learning_rate": 2.055844315802439e-06, + "loss": 0.2376, + "step": 25602 + }, + { + "epoch": 9.45, + "learning_rate": 2.0558073654390936e-06, + "loss": 0.3346, + "step": 25603 + }, + { + "epoch": 9.45, + "learning_rate": 2.0557704150757483e-06, + "loss": 0.3025, + "step": 25604 + }, + { + "epoch": 9.46, + "learning_rate": 2.055733464712403e-06, + "loss": 0.3855, + "step": 25605 + }, + { + "epoch": 9.46, + "learning_rate": 2.055696514349058e-06, + "loss": 0.3388, + "step": 25606 + }, + { + "epoch": 9.46, + "learning_rate": 2.0556595639857126e-06, + "loss": 0.3327, + "step": 25607 + }, + { + "epoch": 9.46, + "learning_rate": 2.0556226136223677e-06, + "loss": 0.3572, + "step": 25608 + }, + { + "epoch": 9.46, + "learning_rate": 2.0555856632590223e-06, + "loss": 0.3575, + "step": 25609 + }, + { + "epoch": 9.46, + "learning_rate": 2.0555487128956765e-06, + "loss": 0.3491, + "step": 25610 + }, + { + "epoch": 9.46, + "learning_rate": 2.0555117625323316e-06, + "loss": 0.3616, + "step": 25611 + }, + { + "epoch": 9.46, + "learning_rate": 2.055474812168986e-06, + "loss": 0.3737, + "step": 25612 + }, + { + "epoch": 9.46, + "learning_rate": 2.0554378618056413e-06, + "loss": 0.3422, + "step": 25613 + }, + { + "epoch": 9.46, + "learning_rate": 2.055400911442296e-06, + "loss": 0.5068, + "step": 25614 + }, + { + "epoch": 9.46, + "learning_rate": 2.055363961078951e-06, + "loss": 0.3695, + "step": 25615 + }, + { + "epoch": 9.46, + "learning_rate": 2.055327010715605e-06, + "loss": 0.2634, + "step": 25616 + }, + { + "epoch": 9.46, + "learning_rate": 2.0552900603522602e-06, + "loss": 0.4738, + "step": 25617 + }, + { + "epoch": 9.46, + "learning_rate": 2.055253109988915e-06, + "loss": 0.4923, + "step": 25618 + }, + { + "epoch": 9.46, + "learning_rate": 2.05521615962557e-06, + "loss": 0.434, + "step": 25619 + }, + { + "epoch": 9.46, + "learning_rate": 2.0551792092622246e-06, + "loss": 0.3105, + "step": 25620 + }, + { + "epoch": 9.46, + "learning_rate": 2.055142258898879e-06, + "loss": 0.4067, + "step": 25621 + }, + { + "epoch": 9.46, + "learning_rate": 2.055105308535534e-06, + "loss": 0.2987, + "step": 25622 + }, + { + "epoch": 9.46, + "learning_rate": 2.0550683581721885e-06, + "loss": 0.3255, + "step": 25623 + }, + { + "epoch": 9.46, + "learning_rate": 2.0550314078088435e-06, + "loss": 0.371, + "step": 25624 + }, + { + "epoch": 9.46, + "learning_rate": 2.054994457445498e-06, + "loss": 0.4639, + "step": 25625 + }, + { + "epoch": 9.46, + "learning_rate": 2.0549575070821532e-06, + "loss": 0.2661, + "step": 25626 + }, + { + "epoch": 9.46, + "learning_rate": 2.054920556718808e-06, + "loss": 0.4053, + "step": 25627 + }, + { + "epoch": 9.46, + "learning_rate": 2.0548836063554625e-06, + "loss": 0.3476, + "step": 25628 + }, + { + "epoch": 9.46, + "learning_rate": 2.054846655992117e-06, + "loss": 0.4083, + "step": 25629 + }, + { + "epoch": 9.46, + "learning_rate": 2.0548097056287718e-06, + "loss": 0.3604, + "step": 25630 + }, + { + "epoch": 9.46, + "learning_rate": 2.054772755265427e-06, + "loss": 0.4211, + "step": 25631 + }, + { + "epoch": 9.47, + "learning_rate": 2.0547358049020815e-06, + "loss": 0.3283, + "step": 25632 + }, + { + "epoch": 9.47, + "learning_rate": 2.0546988545387365e-06, + "loss": 0.3505, + "step": 25633 + }, + { + "epoch": 9.47, + "learning_rate": 2.054661904175391e-06, + "loss": 0.4051, + "step": 25634 + }, + { + "epoch": 9.47, + "learning_rate": 2.054624953812046e-06, + "loss": 0.4096, + "step": 25635 + }, + { + "epoch": 9.47, + "learning_rate": 2.0545880034487004e-06, + "loss": 0.3615, + "step": 25636 + }, + { + "epoch": 9.47, + "learning_rate": 2.0545510530853555e-06, + "loss": 0.426, + "step": 25637 + }, + { + "epoch": 9.47, + "learning_rate": 2.05451410272201e-06, + "loss": 0.5116, + "step": 25638 + }, + { + "epoch": 9.47, + "learning_rate": 2.054477152358665e-06, + "loss": 0.428, + "step": 25639 + }, + { + "epoch": 9.47, + "learning_rate": 2.05444020199532e-06, + "loss": 0.3799, + "step": 25640 + }, + { + "epoch": 9.47, + "learning_rate": 2.0544032516319745e-06, + "loss": 0.3735, + "step": 25641 + }, + { + "epoch": 9.47, + "learning_rate": 2.054366301268629e-06, + "loss": 0.3129, + "step": 25642 + }, + { + "epoch": 9.47, + "learning_rate": 2.0543293509052837e-06, + "loss": 0.4062, + "step": 25643 + }, + { + "epoch": 9.47, + "learning_rate": 2.054292400541939e-06, + "loss": 0.2675, + "step": 25644 + }, + { + "epoch": 9.47, + "learning_rate": 2.0542554501785934e-06, + "loss": 0.3214, + "step": 25645 + }, + { + "epoch": 9.47, + "learning_rate": 2.0542184998152485e-06, + "loss": 0.3553, + "step": 25646 + }, + { + "epoch": 9.47, + "learning_rate": 2.054181549451903e-06, + "loss": 0.3737, + "step": 25647 + }, + { + "epoch": 9.47, + "learning_rate": 2.0541445990885578e-06, + "loss": 0.3456, + "step": 25648 + }, + { + "epoch": 9.47, + "learning_rate": 2.0541076487252124e-06, + "loss": 0.2627, + "step": 25649 + }, + { + "epoch": 9.47, + "learning_rate": 2.054070698361867e-06, + "loss": 0.5153, + "step": 25650 + }, + { + "epoch": 9.47, + "learning_rate": 2.054033747998522e-06, + "loss": 0.3029, + "step": 25651 + }, + { + "epoch": 9.47, + "learning_rate": 2.0539967976351767e-06, + "loss": 0.4218, + "step": 25652 + }, + { + "epoch": 9.47, + "learning_rate": 2.053959847271832e-06, + "loss": 0.2744, + "step": 25653 + }, + { + "epoch": 9.47, + "learning_rate": 2.053922896908486e-06, + "loss": 0.4147, + "step": 25654 + }, + { + "epoch": 9.47, + "learning_rate": 2.053885946545141e-06, + "loss": 0.3856, + "step": 25655 + }, + { + "epoch": 9.47, + "learning_rate": 2.0538489961817957e-06, + "loss": 0.4372, + "step": 25656 + }, + { + "epoch": 9.47, + "learning_rate": 2.0538120458184508e-06, + "loss": 0.4177, + "step": 25657 + }, + { + "epoch": 9.47, + "learning_rate": 2.0537750954551054e-06, + "loss": 0.3735, + "step": 25658 + }, + { + "epoch": 9.48, + "learning_rate": 2.0537381450917605e-06, + "loss": 0.3397, + "step": 25659 + }, + { + "epoch": 9.48, + "learning_rate": 2.0537011947284147e-06, + "loss": 0.3032, + "step": 25660 + }, + { + "epoch": 9.48, + "learning_rate": 2.0536642443650693e-06, + "loss": 0.2745, + "step": 25661 + }, + { + "epoch": 9.48, + "learning_rate": 2.0536272940017244e-06, + "loss": 0.3953, + "step": 25662 + }, + { + "epoch": 9.48, + "learning_rate": 2.053590343638379e-06, + "loss": 0.4103, + "step": 25663 + }, + { + "epoch": 9.48, + "learning_rate": 2.053553393275034e-06, + "loss": 0.4354, + "step": 25664 + }, + { + "epoch": 9.48, + "learning_rate": 2.0535164429116887e-06, + "loss": 0.3656, + "step": 25665 + }, + { + "epoch": 9.48, + "learning_rate": 2.0534794925483433e-06, + "loss": 0.4373, + "step": 25666 + }, + { + "epoch": 9.48, + "learning_rate": 2.053442542184998e-06, + "loss": 0.4063, + "step": 25667 + }, + { + "epoch": 9.48, + "learning_rate": 2.053405591821653e-06, + "loss": 0.4216, + "step": 25668 + }, + { + "epoch": 9.48, + "learning_rate": 2.0533686414583077e-06, + "loss": 0.3607, + "step": 25669 + }, + { + "epoch": 9.48, + "learning_rate": 2.0533316910949623e-06, + "loss": 0.2915, + "step": 25670 + }, + { + "epoch": 9.48, + "learning_rate": 2.0532947407316174e-06, + "loss": 0.2835, + "step": 25671 + }, + { + "epoch": 9.48, + "learning_rate": 2.053257790368272e-06, + "loss": 0.3515, + "step": 25672 + }, + { + "epoch": 9.48, + "learning_rate": 2.0532208400049266e-06, + "loss": 0.3566, + "step": 25673 + }, + { + "epoch": 9.48, + "learning_rate": 2.0531838896415813e-06, + "loss": 0.3702, + "step": 25674 + }, + { + "epoch": 9.48, + "learning_rate": 2.0531469392782363e-06, + "loss": 0.4707, + "step": 25675 + }, + { + "epoch": 9.48, + "learning_rate": 2.053109988914891e-06, + "loss": 0.3729, + "step": 25676 + }, + { + "epoch": 9.48, + "learning_rate": 2.053073038551546e-06, + "loss": 0.2993, + "step": 25677 + }, + { + "epoch": 9.48, + "learning_rate": 2.0530360881882007e-06, + "loss": 0.3637, + "step": 25678 + }, + { + "epoch": 9.48, + "learning_rate": 2.0529991378248553e-06, + "loss": 0.389, + "step": 25679 + }, + { + "epoch": 9.48, + "learning_rate": 2.05296218746151e-06, + "loss": 0.3916, + "step": 25680 + }, + { + "epoch": 9.48, + "learning_rate": 2.0529252370981646e-06, + "loss": 0.2792, + "step": 25681 + }, + { + "epoch": 9.48, + "learning_rate": 2.0528882867348196e-06, + "loss": 0.377, + "step": 25682 + }, + { + "epoch": 9.48, + "learning_rate": 2.0528513363714743e-06, + "loss": 0.4132, + "step": 25683 + }, + { + "epoch": 9.48, + "learning_rate": 2.0528143860081293e-06, + "loss": 0.3677, + "step": 25684 + }, + { + "epoch": 9.48, + "learning_rate": 2.052777435644784e-06, + "loss": 0.3061, + "step": 25685 + }, + { + "epoch": 9.49, + "learning_rate": 2.0527404852814386e-06, + "loss": 0.3562, + "step": 25686 + }, + { + "epoch": 9.49, + "learning_rate": 2.0527035349180933e-06, + "loss": 0.4503, + "step": 25687 + }, + { + "epoch": 9.49, + "learning_rate": 2.0526665845547483e-06, + "loss": 0.3835, + "step": 25688 + }, + { + "epoch": 9.49, + "learning_rate": 2.052629634191403e-06, + "loss": 0.3296, + "step": 25689 + }, + { + "epoch": 9.49, + "learning_rate": 2.0525926838280576e-06, + "loss": 0.2859, + "step": 25690 + }, + { + "epoch": 9.49, + "learning_rate": 2.0525557334647126e-06, + "loss": 0.3349, + "step": 25691 + }, + { + "epoch": 9.49, + "learning_rate": 2.0525187831013673e-06, + "loss": 0.3351, + "step": 25692 + }, + { + "epoch": 9.49, + "learning_rate": 2.052481832738022e-06, + "loss": 0.5111, + "step": 25693 + }, + { + "epoch": 9.49, + "learning_rate": 2.0524448823746766e-06, + "loss": 0.4365, + "step": 25694 + }, + { + "epoch": 9.49, + "learning_rate": 2.0524079320113316e-06, + "loss": 0.3851, + "step": 25695 + }, + { + "epoch": 9.49, + "learning_rate": 2.0523709816479863e-06, + "loss": 0.3732, + "step": 25696 + }, + { + "epoch": 9.49, + "learning_rate": 2.0523340312846413e-06, + "loss": 0.4035, + "step": 25697 + }, + { + "epoch": 9.49, + "learning_rate": 2.052297080921296e-06, + "loss": 0.3507, + "step": 25698 + }, + { + "epoch": 9.49, + "learning_rate": 2.0522601305579506e-06, + "loss": 0.2668, + "step": 25699 + }, + { + "epoch": 9.49, + "learning_rate": 2.0522231801946052e-06, + "loss": 0.2956, + "step": 25700 + }, + { + "epoch": 9.49, + "learning_rate": 2.05218622983126e-06, + "loss": 0.3807, + "step": 25701 + }, + { + "epoch": 9.49, + "learning_rate": 2.052149279467915e-06, + "loss": 0.3837, + "step": 25702 + }, + { + "epoch": 9.49, + "learning_rate": 2.0521123291045696e-06, + "loss": 0.4472, + "step": 25703 + }, + { + "epoch": 9.49, + "learning_rate": 2.0520753787412246e-06, + "loss": 0.3919, + "step": 25704 + }, + { + "epoch": 9.49, + "learning_rate": 2.052038428377879e-06, + "loss": 0.4696, + "step": 25705 + }, + { + "epoch": 9.49, + "learning_rate": 2.052001478014534e-06, + "loss": 0.3081, + "step": 25706 + }, + { + "epoch": 9.49, + "learning_rate": 2.0519645276511885e-06, + "loss": 0.3167, + "step": 25707 + }, + { + "epoch": 9.49, + "learning_rate": 2.0519275772878436e-06, + "loss": 0.3781, + "step": 25708 + }, + { + "epoch": 9.49, + "learning_rate": 2.0518906269244982e-06, + "loss": 0.4721, + "step": 25709 + }, + { + "epoch": 9.49, + "learning_rate": 2.051853676561153e-06, + "loss": 0.2507, + "step": 25710 + }, + { + "epoch": 9.49, + "learning_rate": 2.0518167261978075e-06, + "loss": 0.3096, + "step": 25711 + }, + { + "epoch": 9.49, + "learning_rate": 2.051779775834462e-06, + "loss": 0.4388, + "step": 25712 + }, + { + "epoch": 9.5, + "learning_rate": 2.051742825471117e-06, + "loss": 0.3087, + "step": 25713 + }, + { + "epoch": 9.5, + "learning_rate": 2.051705875107772e-06, + "loss": 0.3526, + "step": 25714 + }, + { + "epoch": 9.5, + "learning_rate": 2.051668924744427e-06, + "loss": 0.28, + "step": 25715 + }, + { + "epoch": 9.5, + "learning_rate": 2.0516319743810815e-06, + "loss": 0.3225, + "step": 25716 + }, + { + "epoch": 9.5, + "learning_rate": 2.051595024017736e-06, + "loss": 0.4126, + "step": 25717 + }, + { + "epoch": 9.5, + "learning_rate": 2.051558073654391e-06, + "loss": 0.3118, + "step": 25718 + }, + { + "epoch": 9.5, + "learning_rate": 2.051521123291046e-06, + "loss": 0.4451, + "step": 25719 + }, + { + "epoch": 9.5, + "learning_rate": 2.0514841729277005e-06, + "loss": 0.4671, + "step": 25720 + }, + { + "epoch": 9.5, + "learning_rate": 2.051447222564355e-06, + "loss": 0.4421, + "step": 25721 + }, + { + "epoch": 9.5, + "learning_rate": 2.05141027220101e-06, + "loss": 0.3141, + "step": 25722 + }, + { + "epoch": 9.5, + "learning_rate": 2.051373321837665e-06, + "loss": 0.318, + "step": 25723 + }, + { + "epoch": 9.5, + "learning_rate": 2.0513363714743195e-06, + "loss": 0.4155, + "step": 25724 + }, + { + "epoch": 9.5, + "learning_rate": 2.051299421110974e-06, + "loss": 0.2709, + "step": 25725 + }, + { + "epoch": 9.5, + "learning_rate": 2.051262470747629e-06, + "loss": 0.4055, + "step": 25726 + }, + { + "epoch": 9.5, + "learning_rate": 2.051225520384284e-06, + "loss": 0.2229, + "step": 25727 + }, + { + "epoch": 9.5, + "learning_rate": 2.051188570020939e-06, + "loss": 0.476, + "step": 25728 + }, + { + "epoch": 9.5, + "learning_rate": 2.0511516196575935e-06, + "loss": 0.3982, + "step": 25729 + }, + { + "epoch": 9.5, + "learning_rate": 2.051114669294248e-06, + "loss": 0.3704, + "step": 25730 + }, + { + "epoch": 9.5, + "learning_rate": 2.0510777189309028e-06, + "loss": 0.3589, + "step": 25731 + }, + { + "epoch": 9.5, + "learning_rate": 2.0510407685675574e-06, + "loss": 0.3409, + "step": 25732 + }, + { + "epoch": 9.5, + "learning_rate": 2.0510038182042125e-06, + "loss": 0.4335, + "step": 25733 + }, + { + "epoch": 9.5, + "learning_rate": 2.050966867840867e-06, + "loss": 0.4285, + "step": 25734 + }, + { + "epoch": 9.5, + "learning_rate": 2.050929917477522e-06, + "loss": 0.3074, + "step": 25735 + }, + { + "epoch": 9.5, + "learning_rate": 2.050892967114177e-06, + "loss": 0.3358, + "step": 25736 + }, + { + "epoch": 9.5, + "learning_rate": 2.0508560167508314e-06, + "loss": 0.3298, + "step": 25737 + }, + { + "epoch": 9.5, + "learning_rate": 2.050819066387486e-06, + "loss": 0.3981, + "step": 25738 + }, + { + "epoch": 9.5, + "learning_rate": 2.050782116024141e-06, + "loss": 0.4064, + "step": 25739 + }, + { + "epoch": 9.51, + "learning_rate": 2.0507451656607958e-06, + "loss": 0.2835, + "step": 25740 + }, + { + "epoch": 9.51, + "learning_rate": 2.0507082152974504e-06, + "loss": 0.3351, + "step": 25741 + }, + { + "epoch": 9.51, + "learning_rate": 2.0506712649341055e-06, + "loss": 0.3572, + "step": 25742 + }, + { + "epoch": 9.51, + "learning_rate": 2.05063431457076e-06, + "loss": 0.3135, + "step": 25743 + }, + { + "epoch": 9.51, + "learning_rate": 2.0505973642074147e-06, + "loss": 0.2219, + "step": 25744 + }, + { + "epoch": 9.51, + "learning_rate": 2.0505604138440694e-06, + "loss": 0.3114, + "step": 25745 + }, + { + "epoch": 9.51, + "learning_rate": 2.0505234634807244e-06, + "loss": 0.421, + "step": 25746 + }, + { + "epoch": 9.51, + "learning_rate": 2.050486513117379e-06, + "loss": 0.4407, + "step": 25747 + }, + { + "epoch": 9.51, + "learning_rate": 2.050449562754034e-06, + "loss": 0.3569, + "step": 25748 + }, + { + "epoch": 9.51, + "learning_rate": 2.0504126123906888e-06, + "loss": 0.2894, + "step": 25749 + }, + { + "epoch": 9.51, + "learning_rate": 2.050375662027343e-06, + "loss": 0.2638, + "step": 25750 + }, + { + "epoch": 9.51, + "learning_rate": 2.050338711663998e-06, + "loss": 0.3381, + "step": 25751 + }, + { + "epoch": 9.51, + "learning_rate": 2.0503017613006527e-06, + "loss": 0.3216, + "step": 25752 + }, + { + "epoch": 9.51, + "learning_rate": 2.0502648109373077e-06, + "loss": 0.3074, + "step": 25753 + }, + { + "epoch": 9.51, + "learning_rate": 2.0502278605739624e-06, + "loss": 0.4011, + "step": 25754 + }, + { + "epoch": 9.51, + "learning_rate": 2.0501909102106174e-06, + "loss": 0.3754, + "step": 25755 + }, + { + "epoch": 9.51, + "learning_rate": 2.0501539598472716e-06, + "loss": 0.3373, + "step": 25756 + }, + { + "epoch": 9.51, + "learning_rate": 2.0501170094839267e-06, + "loss": 0.4363, + "step": 25757 + }, + { + "epoch": 9.51, + "learning_rate": 2.0500800591205813e-06, + "loss": 0.3448, + "step": 25758 + }, + { + "epoch": 9.51, + "learning_rate": 2.0500431087572364e-06, + "loss": 0.3346, + "step": 25759 + }, + { + "epoch": 9.51, + "learning_rate": 2.050006158393891e-06, + "loss": 0.4634, + "step": 25760 + }, + { + "epoch": 9.51, + "learning_rate": 2.0499692080305457e-06, + "loss": 0.3498, + "step": 25761 + }, + { + "epoch": 9.51, + "learning_rate": 2.0499322576672003e-06, + "loss": 0.2224, + "step": 25762 + }, + { + "epoch": 9.51, + "learning_rate": 2.049895307303855e-06, + "loss": 0.3829, + "step": 25763 + }, + { + "epoch": 9.51, + "learning_rate": 2.04985835694051e-06, + "loss": 0.4984, + "step": 25764 + }, + { + "epoch": 9.51, + "learning_rate": 2.0498214065771646e-06, + "loss": 0.4905, + "step": 25765 + }, + { + "epoch": 9.51, + "learning_rate": 2.0497844562138197e-06, + "loss": 0.3155, + "step": 25766 + }, + { + "epoch": 9.52, + "learning_rate": 2.0497475058504743e-06, + "loss": 0.4065, + "step": 25767 + }, + { + "epoch": 9.52, + "learning_rate": 2.049710555487129e-06, + "loss": 0.3172, + "step": 25768 + }, + { + "epoch": 9.52, + "learning_rate": 2.0496736051237836e-06, + "loss": 0.4008, + "step": 25769 + }, + { + "epoch": 9.52, + "learning_rate": 2.0496366547604382e-06, + "loss": 0.2858, + "step": 25770 + }, + { + "epoch": 9.52, + "learning_rate": 2.0495997043970933e-06, + "loss": 0.4358, + "step": 25771 + }, + { + "epoch": 9.52, + "learning_rate": 2.049562754033748e-06, + "loss": 0.3547, + "step": 25772 + }, + { + "epoch": 9.52, + "learning_rate": 2.049525803670403e-06, + "loss": 0.4931, + "step": 25773 + }, + { + "epoch": 9.52, + "learning_rate": 2.0494888533070576e-06, + "loss": 0.2584, + "step": 25774 + }, + { + "epoch": 9.52, + "learning_rate": 2.0494519029437123e-06, + "loss": 0.4287, + "step": 25775 + }, + { + "epoch": 9.52, + "learning_rate": 2.049414952580367e-06, + "loss": 0.3552, + "step": 25776 + }, + { + "epoch": 9.52, + "learning_rate": 2.049378002217022e-06, + "loss": 0.3691, + "step": 25777 + }, + { + "epoch": 9.52, + "learning_rate": 2.0493410518536766e-06, + "loss": 0.3921, + "step": 25778 + }, + { + "epoch": 9.52, + "learning_rate": 2.0493041014903317e-06, + "loss": 0.3373, + "step": 25779 + }, + { + "epoch": 9.52, + "learning_rate": 2.0492671511269863e-06, + "loss": 0.2813, + "step": 25780 + }, + { + "epoch": 9.52, + "learning_rate": 2.049230200763641e-06, + "loss": 0.4197, + "step": 25781 + }, + { + "epoch": 9.52, + "learning_rate": 2.0491932504002956e-06, + "loss": 0.2676, + "step": 25782 + }, + { + "epoch": 9.52, + "learning_rate": 2.04915630003695e-06, + "loss": 0.4833, + "step": 25783 + }, + { + "epoch": 9.52, + "learning_rate": 2.0491193496736053e-06, + "loss": 0.3571, + "step": 25784 + }, + { + "epoch": 9.52, + "learning_rate": 2.04908239931026e-06, + "loss": 0.2595, + "step": 25785 + }, + { + "epoch": 9.52, + "learning_rate": 2.049045448946915e-06, + "loss": 0.4253, + "step": 25786 + }, + { + "epoch": 9.52, + "learning_rate": 2.0490084985835696e-06, + "loss": 0.3009, + "step": 25787 + }, + { + "epoch": 9.52, + "learning_rate": 2.0489715482202242e-06, + "loss": 0.3095, + "step": 25788 + }, + { + "epoch": 9.52, + "learning_rate": 2.048934597856879e-06, + "loss": 0.3918, + "step": 25789 + }, + { + "epoch": 9.52, + "learning_rate": 2.0488976474935335e-06, + "loss": 0.3648, + "step": 25790 + }, + { + "epoch": 9.52, + "learning_rate": 2.0488606971301886e-06, + "loss": 0.3986, + "step": 25791 + }, + { + "epoch": 9.52, + "learning_rate": 2.048823746766843e-06, + "loss": 0.4672, + "step": 25792 + }, + { + "epoch": 9.52, + "learning_rate": 2.0487867964034983e-06, + "loss": 0.406, + "step": 25793 + }, + { + "epoch": 9.53, + "learning_rate": 2.048749846040153e-06, + "loss": 0.3715, + "step": 25794 + }, + { + "epoch": 9.53, + "learning_rate": 2.0487128956768075e-06, + "loss": 0.3715, + "step": 25795 + }, + { + "epoch": 9.53, + "learning_rate": 2.048675945313462e-06, + "loss": 0.2821, + "step": 25796 + }, + { + "epoch": 9.53, + "learning_rate": 2.0486389949501172e-06, + "loss": 0.4809, + "step": 25797 + }, + { + "epoch": 9.53, + "learning_rate": 2.048602044586772e-06, + "loss": 0.3498, + "step": 25798 + }, + { + "epoch": 9.53, + "learning_rate": 2.0485650942234265e-06, + "loss": 0.4218, + "step": 25799 + }, + { + "epoch": 9.53, + "learning_rate": 2.0485281438600816e-06, + "loss": 0.2822, + "step": 25800 + }, + { + "epoch": 9.53, + "learning_rate": 2.0484911934967358e-06, + "loss": 0.2195, + "step": 25801 + }, + { + "epoch": 9.53, + "learning_rate": 2.048454243133391e-06, + "loss": 0.3348, + "step": 25802 + }, + { + "epoch": 9.53, + "learning_rate": 2.0484172927700455e-06, + "loss": 0.4482, + "step": 25803 + }, + { + "epoch": 9.53, + "learning_rate": 2.0483803424067005e-06, + "loss": 0.274, + "step": 25804 + }, + { + "epoch": 9.53, + "learning_rate": 2.048343392043355e-06, + "loss": 0.4066, + "step": 25805 + }, + { + "epoch": 9.53, + "learning_rate": 2.04830644168001e-06, + "loss": 0.3675, + "step": 25806 + }, + { + "epoch": 9.53, + "learning_rate": 2.0482694913166644e-06, + "loss": 0.2965, + "step": 25807 + }, + { + "epoch": 9.53, + "learning_rate": 2.0482325409533195e-06, + "loss": 0.3329, + "step": 25808 + }, + { + "epoch": 9.53, + "learning_rate": 2.048195590589974e-06, + "loss": 0.5246, + "step": 25809 + }, + { + "epoch": 9.53, + "learning_rate": 2.0481586402266288e-06, + "loss": 0.2062, + "step": 25810 + }, + { + "epoch": 9.53, + "learning_rate": 2.048121689863284e-06, + "loss": 0.2992, + "step": 25811 + }, + { + "epoch": 9.53, + "learning_rate": 2.0480847394999385e-06, + "loss": 0.36, + "step": 25812 + }, + { + "epoch": 9.53, + "learning_rate": 2.048047789136593e-06, + "loss": 0.2756, + "step": 25813 + }, + { + "epoch": 9.53, + "learning_rate": 2.0480108387732478e-06, + "loss": 0.3278, + "step": 25814 + }, + { + "epoch": 9.53, + "learning_rate": 2.047973888409903e-06, + "loss": 0.3074, + "step": 25815 + }, + { + "epoch": 9.53, + "learning_rate": 2.0479369380465574e-06, + "loss": 0.3424, + "step": 25816 + }, + { + "epoch": 9.53, + "learning_rate": 2.0478999876832125e-06, + "loss": 0.3454, + "step": 25817 + }, + { + "epoch": 9.53, + "learning_rate": 2.047863037319867e-06, + "loss": 0.3788, + "step": 25818 + }, + { + "epoch": 9.53, + "learning_rate": 2.0478260869565218e-06, + "loss": 0.3832, + "step": 25819 + }, + { + "epoch": 9.53, + "learning_rate": 2.0477891365931764e-06, + "loss": 0.31, + "step": 25820 + }, + { + "epoch": 9.54, + "learning_rate": 2.047752186229831e-06, + "loss": 0.3861, + "step": 25821 + }, + { + "epoch": 9.54, + "learning_rate": 2.047715235866486e-06, + "loss": 0.3445, + "step": 25822 + }, + { + "epoch": 9.54, + "learning_rate": 2.0476782855031407e-06, + "loss": 0.325, + "step": 25823 + }, + { + "epoch": 9.54, + "learning_rate": 2.047641335139796e-06, + "loss": 0.3768, + "step": 25824 + }, + { + "epoch": 9.54, + "learning_rate": 2.0476043847764504e-06, + "loss": 0.3246, + "step": 25825 + }, + { + "epoch": 9.54, + "learning_rate": 2.047567434413105e-06, + "loss": 0.3436, + "step": 25826 + }, + { + "epoch": 9.54, + "learning_rate": 2.0475304840497597e-06, + "loss": 0.4943, + "step": 25827 + }, + { + "epoch": 9.54, + "learning_rate": 2.0474935336864148e-06, + "loss": 0.4051, + "step": 25828 + }, + { + "epoch": 9.54, + "learning_rate": 2.0474565833230694e-06, + "loss": 0.3338, + "step": 25829 + }, + { + "epoch": 9.54, + "learning_rate": 2.047419632959724e-06, + "loss": 0.3987, + "step": 25830 + }, + { + "epoch": 9.54, + "learning_rate": 2.047382682596379e-06, + "loss": 0.3341, + "step": 25831 + }, + { + "epoch": 9.54, + "learning_rate": 2.0473457322330337e-06, + "loss": 0.3346, + "step": 25832 + }, + { + "epoch": 9.54, + "learning_rate": 2.0473087818696884e-06, + "loss": 0.3656, + "step": 25833 + }, + { + "epoch": 9.54, + "learning_rate": 2.047271831506343e-06, + "loss": 0.2603, + "step": 25834 + }, + { + "epoch": 9.54, + "learning_rate": 2.047234881142998e-06, + "loss": 0.3655, + "step": 25835 + }, + { + "epoch": 9.54, + "learning_rate": 2.0471979307796527e-06, + "loss": 0.3617, + "step": 25836 + }, + { + "epoch": 9.54, + "learning_rate": 2.0471609804163078e-06, + "loss": 0.3075, + "step": 25837 + }, + { + "epoch": 9.54, + "learning_rate": 2.0471240300529624e-06, + "loss": 0.2715, + "step": 25838 + }, + { + "epoch": 9.54, + "learning_rate": 2.0470870796896166e-06, + "loss": 0.408, + "step": 25839 + }, + { + "epoch": 9.54, + "learning_rate": 2.0470501293262717e-06, + "loss": 0.4512, + "step": 25840 + }, + { + "epoch": 9.54, + "learning_rate": 2.0470131789629263e-06, + "loss": 0.3978, + "step": 25841 + }, + { + "epoch": 9.54, + "learning_rate": 2.0469762285995814e-06, + "loss": 0.2513, + "step": 25842 + }, + { + "epoch": 9.54, + "learning_rate": 2.046939278236236e-06, + "loss": 0.4479, + "step": 25843 + }, + { + "epoch": 9.54, + "learning_rate": 2.046902327872891e-06, + "loss": 0.2054, + "step": 25844 + }, + { + "epoch": 9.54, + "learning_rate": 2.0468653775095453e-06, + "loss": 0.2532, + "step": 25845 + }, + { + "epoch": 9.54, + "learning_rate": 2.0468284271462004e-06, + "loss": 0.3309, + "step": 25846 + }, + { + "epoch": 9.54, + "learning_rate": 2.046791476782855e-06, + "loss": 0.3166, + "step": 25847 + }, + { + "epoch": 9.55, + "learning_rate": 2.04675452641951e-06, + "loss": 0.3446, + "step": 25848 + }, + { + "epoch": 9.55, + "learning_rate": 2.0467175760561647e-06, + "loss": 0.3966, + "step": 25849 + }, + { + "epoch": 9.55, + "learning_rate": 2.0466806256928193e-06, + "loss": 0.3407, + "step": 25850 + }, + { + "epoch": 9.55, + "learning_rate": 2.046643675329474e-06, + "loss": 0.5509, + "step": 25851 + }, + { + "epoch": 9.55, + "learning_rate": 2.0466067249661286e-06, + "loss": 0.3146, + "step": 25852 + }, + { + "epoch": 9.55, + "learning_rate": 2.0465697746027837e-06, + "loss": 0.4025, + "step": 25853 + }, + { + "epoch": 9.55, + "learning_rate": 2.0465328242394383e-06, + "loss": 0.4248, + "step": 25854 + }, + { + "epoch": 9.55, + "learning_rate": 2.0464958738760934e-06, + "loss": 0.438, + "step": 25855 + }, + { + "epoch": 9.55, + "learning_rate": 2.046458923512748e-06, + "loss": 0.4935, + "step": 25856 + }, + { + "epoch": 9.55, + "learning_rate": 2.0464219731494026e-06, + "loss": 0.4139, + "step": 25857 + }, + { + "epoch": 9.55, + "learning_rate": 2.0463850227860573e-06, + "loss": 0.2665, + "step": 25858 + }, + { + "epoch": 9.55, + "learning_rate": 2.046348072422712e-06, + "loss": 0.441, + "step": 25859 + }, + { + "epoch": 9.55, + "learning_rate": 2.046311122059367e-06, + "loss": 0.3532, + "step": 25860 + }, + { + "epoch": 9.55, + "learning_rate": 2.0462741716960216e-06, + "loss": 0.4182, + "step": 25861 + }, + { + "epoch": 9.55, + "learning_rate": 2.0462372213326767e-06, + "loss": 0.3378, + "step": 25862 + }, + { + "epoch": 9.55, + "learning_rate": 2.0462002709693313e-06, + "loss": 0.2871, + "step": 25863 + }, + { + "epoch": 9.55, + "learning_rate": 2.046163320605986e-06, + "loss": 0.3736, + "step": 25864 + }, + { + "epoch": 9.55, + "learning_rate": 2.0461263702426406e-06, + "loss": 0.3189, + "step": 25865 + }, + { + "epoch": 9.55, + "learning_rate": 2.0460894198792956e-06, + "loss": 0.4386, + "step": 25866 + }, + { + "epoch": 9.55, + "learning_rate": 2.0460524695159503e-06, + "loss": 0.4117, + "step": 25867 + }, + { + "epoch": 9.55, + "learning_rate": 2.0460155191526053e-06, + "loss": 0.328, + "step": 25868 + }, + { + "epoch": 9.55, + "learning_rate": 2.04597856878926e-06, + "loss": 0.4018, + "step": 25869 + }, + { + "epoch": 9.55, + "learning_rate": 2.0459416184259146e-06, + "loss": 0.6104, + "step": 25870 + }, + { + "epoch": 9.55, + "learning_rate": 2.0459046680625692e-06, + "loss": 0.308, + "step": 25871 + }, + { + "epoch": 9.55, + "learning_rate": 2.045867717699224e-06, + "loss": 0.3955, + "step": 25872 + }, + { + "epoch": 9.55, + "learning_rate": 2.045830767335879e-06, + "loss": 0.33, + "step": 25873 + }, + { + "epoch": 9.55, + "learning_rate": 2.0457938169725336e-06, + "loss": 0.5018, + "step": 25874 + }, + { + "epoch": 9.56, + "learning_rate": 2.0457568666091886e-06, + "loss": 0.2344, + "step": 25875 + }, + { + "epoch": 9.56, + "learning_rate": 2.0457199162458433e-06, + "loss": 0.4392, + "step": 25876 + }, + { + "epoch": 9.56, + "learning_rate": 2.045682965882498e-06, + "loss": 0.2876, + "step": 25877 + }, + { + "epoch": 9.56, + "learning_rate": 2.0456460155191525e-06, + "loss": 0.4638, + "step": 25878 + }, + { + "epoch": 9.56, + "learning_rate": 2.045609065155807e-06, + "loss": 0.2677, + "step": 25879 + }, + { + "epoch": 9.56, + "learning_rate": 2.0455721147924622e-06, + "loss": 0.3668, + "step": 25880 + }, + { + "epoch": 9.56, + "learning_rate": 2.045535164429117e-06, + "loss": 0.3439, + "step": 25881 + }, + { + "epoch": 9.56, + "learning_rate": 2.045498214065772e-06, + "loss": 0.3688, + "step": 25882 + }, + { + "epoch": 9.56, + "learning_rate": 2.0454612637024266e-06, + "loss": 0.2917, + "step": 25883 + }, + { + "epoch": 9.56, + "learning_rate": 2.045424313339081e-06, + "loss": 0.4314, + "step": 25884 + }, + { + "epoch": 9.56, + "learning_rate": 2.045387362975736e-06, + "loss": 0.335, + "step": 25885 + }, + { + "epoch": 9.56, + "learning_rate": 2.045350412612391e-06, + "loss": 0.3841, + "step": 25886 + }, + { + "epoch": 9.56, + "learning_rate": 2.0453134622490455e-06, + "loss": 0.3359, + "step": 25887 + }, + { + "epoch": 9.56, + "learning_rate": 2.0452765118857006e-06, + "loss": 0.377, + "step": 25888 + }, + { + "epoch": 9.56, + "learning_rate": 2.0452395615223552e-06, + "loss": 0.2528, + "step": 25889 + }, + { + "epoch": 9.56, + "learning_rate": 2.0452026111590094e-06, + "loss": 0.2782, + "step": 25890 + }, + { + "epoch": 9.56, + "learning_rate": 2.0451656607956645e-06, + "loss": 0.2981, + "step": 25891 + }, + { + "epoch": 9.56, + "learning_rate": 2.045128710432319e-06, + "loss": 0.3067, + "step": 25892 + }, + { + "epoch": 9.56, + "learning_rate": 2.045091760068974e-06, + "loss": 0.3929, + "step": 25893 + }, + { + "epoch": 9.56, + "learning_rate": 2.045054809705629e-06, + "loss": 0.2679, + "step": 25894 + }, + { + "epoch": 9.56, + "learning_rate": 2.045017859342284e-06, + "loss": 0.4127, + "step": 25895 + }, + { + "epoch": 9.56, + "learning_rate": 2.044980908978938e-06, + "loss": 0.392, + "step": 25896 + }, + { + "epoch": 9.56, + "learning_rate": 2.044943958615593e-06, + "loss": 0.3243, + "step": 25897 + }, + { + "epoch": 9.56, + "learning_rate": 2.044907008252248e-06, + "loss": 0.4215, + "step": 25898 + }, + { + "epoch": 9.56, + "learning_rate": 2.0448700578889024e-06, + "loss": 0.4135, + "step": 25899 + }, + { + "epoch": 9.56, + "learning_rate": 2.0448331075255575e-06, + "loss": 0.3951, + "step": 25900 + }, + { + "epoch": 9.56, + "learning_rate": 2.044796157162212e-06, + "loss": 0.3783, + "step": 25901 + }, + { + "epoch": 9.56, + "learning_rate": 2.0447592067988668e-06, + "loss": 0.3294, + "step": 25902 + }, + { + "epoch": 9.57, + "learning_rate": 2.0447222564355214e-06, + "loss": 0.2689, + "step": 25903 + }, + { + "epoch": 9.57, + "learning_rate": 2.0446853060721765e-06, + "loss": 0.4025, + "step": 25904 + }, + { + "epoch": 9.57, + "learning_rate": 2.044648355708831e-06, + "loss": 0.34, + "step": 25905 + }, + { + "epoch": 9.57, + "learning_rate": 2.044611405345486e-06, + "loss": 0.4123, + "step": 25906 + }, + { + "epoch": 9.57, + "learning_rate": 2.044574454982141e-06, + "loss": 0.3355, + "step": 25907 + }, + { + "epoch": 9.57, + "learning_rate": 2.0445375046187954e-06, + "loss": 0.4268, + "step": 25908 + }, + { + "epoch": 9.57, + "learning_rate": 2.04450055425545e-06, + "loss": 0.4223, + "step": 25909 + }, + { + "epoch": 9.57, + "learning_rate": 2.0444636038921047e-06, + "loss": 0.5013, + "step": 25910 + }, + { + "epoch": 9.57, + "learning_rate": 2.0444266535287598e-06, + "loss": 0.3044, + "step": 25911 + }, + { + "epoch": 9.57, + "learning_rate": 2.0443897031654144e-06, + "loss": 0.5551, + "step": 25912 + }, + { + "epoch": 9.57, + "learning_rate": 2.0443527528020695e-06, + "loss": 0.3377, + "step": 25913 + }, + { + "epoch": 9.57, + "learning_rate": 2.044315802438724e-06, + "loss": 0.1867, + "step": 25914 + }, + { + "epoch": 9.57, + "learning_rate": 2.0442788520753787e-06, + "loss": 0.4405, + "step": 25915 + }, + { + "epoch": 9.57, + "learning_rate": 2.0442419017120334e-06, + "loss": 0.2369, + "step": 25916 + }, + { + "epoch": 9.57, + "learning_rate": 2.0442049513486884e-06, + "loss": 0.4604, + "step": 25917 + }, + { + "epoch": 9.57, + "learning_rate": 2.044168000985343e-06, + "loss": 0.3115, + "step": 25918 + }, + { + "epoch": 9.57, + "learning_rate": 2.0441310506219977e-06, + "loss": 0.3065, + "step": 25919 + }, + { + "epoch": 9.57, + "learning_rate": 2.0440941002586528e-06, + "loss": 0.3898, + "step": 25920 + }, + { + "epoch": 9.57, + "learning_rate": 2.0440571498953074e-06, + "loss": 0.2589, + "step": 25921 + }, + { + "epoch": 9.57, + "learning_rate": 2.044020199531962e-06, + "loss": 0.4095, + "step": 25922 + }, + { + "epoch": 9.57, + "learning_rate": 2.0439832491686167e-06, + "loss": 0.3495, + "step": 25923 + }, + { + "epoch": 9.57, + "learning_rate": 2.0439462988052717e-06, + "loss": 0.4268, + "step": 25924 + }, + { + "epoch": 9.57, + "learning_rate": 2.0439093484419264e-06, + "loss": 0.1991, + "step": 25925 + }, + { + "epoch": 9.57, + "learning_rate": 2.0438723980785814e-06, + "loss": 0.299, + "step": 25926 + }, + { + "epoch": 9.57, + "learning_rate": 2.043835447715236e-06, + "loss": 0.3622, + "step": 25927 + }, + { + "epoch": 9.57, + "learning_rate": 2.0437984973518907e-06, + "loss": 0.4841, + "step": 25928 + }, + { + "epoch": 9.57, + "learning_rate": 2.0437615469885453e-06, + "loss": 0.455, + "step": 25929 + }, + { + "epoch": 9.58, + "learning_rate": 2.0437245966252e-06, + "loss": 0.3719, + "step": 25930 + }, + { + "epoch": 9.58, + "learning_rate": 2.043687646261855e-06, + "loss": 0.5978, + "step": 25931 + }, + { + "epoch": 9.58, + "learning_rate": 2.0436506958985097e-06, + "loss": 0.3277, + "step": 25932 + }, + { + "epoch": 9.58, + "learning_rate": 2.0436137455351647e-06, + "loss": 0.3839, + "step": 25933 + }, + { + "epoch": 9.58, + "learning_rate": 2.0435767951718194e-06, + "loss": 0.3317, + "step": 25934 + }, + { + "epoch": 9.58, + "learning_rate": 2.043539844808474e-06, + "loss": 0.4194, + "step": 25935 + }, + { + "epoch": 9.58, + "learning_rate": 2.0435028944451286e-06, + "loss": 0.3637, + "step": 25936 + }, + { + "epoch": 9.58, + "learning_rate": 2.0434659440817837e-06, + "loss": 0.2548, + "step": 25937 + }, + { + "epoch": 9.58, + "learning_rate": 2.0434289937184383e-06, + "loss": 0.4681, + "step": 25938 + }, + { + "epoch": 9.58, + "learning_rate": 2.043392043355093e-06, + "loss": 0.3898, + "step": 25939 + }, + { + "epoch": 9.58, + "learning_rate": 2.043355092991748e-06, + "loss": 0.3398, + "step": 25940 + }, + { + "epoch": 9.58, + "learning_rate": 2.0433181426284022e-06, + "loss": 0.4216, + "step": 25941 + }, + { + "epoch": 9.58, + "learning_rate": 2.0432811922650573e-06, + "loss": 0.4677, + "step": 25942 + }, + { + "epoch": 9.58, + "learning_rate": 2.043244241901712e-06, + "loss": 0.3731, + "step": 25943 + }, + { + "epoch": 9.58, + "learning_rate": 2.043207291538367e-06, + "loss": 0.3733, + "step": 25944 + }, + { + "epoch": 9.58, + "learning_rate": 2.0431703411750216e-06, + "loss": 0.4015, + "step": 25945 + }, + { + "epoch": 9.58, + "learning_rate": 2.0431333908116767e-06, + "loss": 0.405, + "step": 25946 + }, + { + "epoch": 9.58, + "learning_rate": 2.043096440448331e-06, + "loss": 0.4547, + "step": 25947 + }, + { + "epoch": 9.58, + "learning_rate": 2.043059490084986e-06, + "loss": 0.3137, + "step": 25948 + }, + { + "epoch": 9.58, + "learning_rate": 2.0430225397216406e-06, + "loss": 0.4022, + "step": 25949 + }, + { + "epoch": 9.58, + "learning_rate": 2.0429855893582952e-06, + "loss": 0.4298, + "step": 25950 + }, + { + "epoch": 9.58, + "learning_rate": 2.0429486389949503e-06, + "loss": 0.3529, + "step": 25951 + }, + { + "epoch": 9.58, + "learning_rate": 2.042911688631605e-06, + "loss": 0.4619, + "step": 25952 + }, + { + "epoch": 9.58, + "learning_rate": 2.0428747382682596e-06, + "loss": 0.4063, + "step": 25953 + }, + { + "epoch": 9.58, + "learning_rate": 2.0428377879049142e-06, + "loss": 0.3462, + "step": 25954 + }, + { + "epoch": 9.58, + "learning_rate": 2.0428008375415693e-06, + "loss": 0.2466, + "step": 25955 + }, + { + "epoch": 9.58, + "learning_rate": 2.042763887178224e-06, + "loss": 0.4646, + "step": 25956 + }, + { + "epoch": 9.59, + "learning_rate": 2.042726936814879e-06, + "loss": 0.4807, + "step": 25957 + }, + { + "epoch": 9.59, + "learning_rate": 2.0426899864515336e-06, + "loss": 0.309, + "step": 25958 + }, + { + "epoch": 9.59, + "learning_rate": 2.0426530360881882e-06, + "loss": 0.4766, + "step": 25959 + }, + { + "epoch": 9.59, + "learning_rate": 2.042616085724843e-06, + "loss": 0.4504, + "step": 25960 + }, + { + "epoch": 9.59, + "learning_rate": 2.0425791353614975e-06, + "loss": 0.2939, + "step": 25961 + }, + { + "epoch": 9.59, + "learning_rate": 2.0425421849981526e-06, + "loss": 0.3395, + "step": 25962 + }, + { + "epoch": 9.59, + "learning_rate": 2.0425052346348072e-06, + "loss": 0.4304, + "step": 25963 + }, + { + "epoch": 9.59, + "learning_rate": 2.0424682842714623e-06, + "loss": 0.2932, + "step": 25964 + }, + { + "epoch": 9.59, + "learning_rate": 2.042431333908117e-06, + "loss": 0.23, + "step": 25965 + }, + { + "epoch": 9.59, + "learning_rate": 2.0423943835447715e-06, + "loss": 0.4147, + "step": 25966 + }, + { + "epoch": 9.59, + "learning_rate": 2.042357433181426e-06, + "loss": 0.3932, + "step": 25967 + }, + { + "epoch": 9.59, + "learning_rate": 2.0423204828180812e-06, + "loss": 0.4358, + "step": 25968 + }, + { + "epoch": 9.59, + "learning_rate": 2.042283532454736e-06, + "loss": 0.6134, + "step": 25969 + }, + { + "epoch": 9.59, + "learning_rate": 2.0422465820913905e-06, + "loss": 0.3504, + "step": 25970 + }, + { + "epoch": 9.59, + "learning_rate": 2.0422096317280456e-06, + "loss": 0.3844, + "step": 25971 + }, + { + "epoch": 9.59, + "learning_rate": 2.0421726813647002e-06, + "loss": 0.2934, + "step": 25972 + }, + { + "epoch": 9.59, + "learning_rate": 2.042135731001355e-06, + "loss": 0.3692, + "step": 25973 + }, + { + "epoch": 9.59, + "learning_rate": 2.0420987806380095e-06, + "loss": 0.414, + "step": 25974 + }, + { + "epoch": 9.59, + "learning_rate": 2.0420618302746645e-06, + "loss": 0.4059, + "step": 25975 + }, + { + "epoch": 9.59, + "learning_rate": 2.042024879911319e-06, + "loss": 0.3592, + "step": 25976 + }, + { + "epoch": 9.59, + "learning_rate": 2.0419879295479742e-06, + "loss": 0.5076, + "step": 25977 + }, + { + "epoch": 9.59, + "learning_rate": 2.041950979184629e-06, + "loss": 0.3932, + "step": 25978 + }, + { + "epoch": 9.59, + "learning_rate": 2.0419140288212835e-06, + "loss": 0.3863, + "step": 25979 + }, + { + "epoch": 9.59, + "learning_rate": 2.041877078457938e-06, + "loss": 0.5174, + "step": 25980 + }, + { + "epoch": 9.59, + "learning_rate": 2.0418401280945928e-06, + "loss": 0.4133, + "step": 25981 + }, + { + "epoch": 9.59, + "learning_rate": 2.041803177731248e-06, + "loss": 0.3851, + "step": 25982 + }, + { + "epoch": 9.59, + "learning_rate": 2.0417662273679025e-06, + "loss": 0.3916, + "step": 25983 + }, + { + "epoch": 9.6, + "learning_rate": 2.0417292770045575e-06, + "loss": 0.4115, + "step": 25984 + }, + { + "epoch": 9.6, + "learning_rate": 2.041692326641212e-06, + "loss": 0.2784, + "step": 25985 + }, + { + "epoch": 9.6, + "learning_rate": 2.041655376277867e-06, + "loss": 0.3342, + "step": 25986 + }, + { + "epoch": 9.6, + "learning_rate": 2.0416184259145215e-06, + "loss": 0.43, + "step": 25987 + }, + { + "epoch": 9.6, + "learning_rate": 2.0415814755511765e-06, + "loss": 0.4487, + "step": 25988 + }, + { + "epoch": 9.6, + "learning_rate": 2.041544525187831e-06, + "loss": 0.3489, + "step": 25989 + }, + { + "epoch": 9.6, + "learning_rate": 2.0415075748244858e-06, + "loss": 0.3946, + "step": 25990 + }, + { + "epoch": 9.6, + "learning_rate": 2.0414706244611404e-06, + "loss": 0.3725, + "step": 25991 + }, + { + "epoch": 9.6, + "learning_rate": 2.041433674097795e-06, + "loss": 0.392, + "step": 25992 + }, + { + "epoch": 9.6, + "learning_rate": 2.04139672373445e-06, + "loss": 0.4036, + "step": 25993 + }, + { + "epoch": 9.6, + "learning_rate": 2.0413597733711048e-06, + "loss": 0.3716, + "step": 25994 + }, + { + "epoch": 9.6, + "learning_rate": 2.04132282300776e-06, + "loss": 0.4402, + "step": 25995 + }, + { + "epoch": 9.6, + "learning_rate": 2.0412858726444145e-06, + "loss": 0.3554, + "step": 25996 + }, + { + "epoch": 9.6, + "learning_rate": 2.041248922281069e-06, + "loss": 0.467, + "step": 25997 + }, + { + "epoch": 9.6, + "learning_rate": 2.0412119719177237e-06, + "loss": 0.5115, + "step": 25998 + }, + { + "epoch": 9.6, + "learning_rate": 2.0411750215543784e-06, + "loss": 0.4375, + "step": 25999 + }, + { + "epoch": 9.6, + "learning_rate": 2.0411380711910334e-06, + "loss": 0.3991, + "step": 26000 + }, + { + "epoch": 9.6, + "learning_rate": 2.041101120827688e-06, + "loss": 0.3349, + "step": 26001 + }, + { + "epoch": 9.6, + "learning_rate": 2.041064170464343e-06, + "loss": 0.333, + "step": 26002 + }, + { + "epoch": 9.6, + "learning_rate": 2.0410272201009978e-06, + "loss": 0.3699, + "step": 26003 + }, + { + "epoch": 9.6, + "learning_rate": 2.0409902697376524e-06, + "loss": 0.3124, + "step": 26004 + }, + { + "epoch": 9.6, + "learning_rate": 2.040953319374307e-06, + "loss": 0.4884, + "step": 26005 + }, + { + "epoch": 9.6, + "learning_rate": 2.040916369010962e-06, + "loss": 0.276, + "step": 26006 + }, + { + "epoch": 9.6, + "learning_rate": 2.0408794186476167e-06, + "loss": 0.3349, + "step": 26007 + }, + { + "epoch": 9.6, + "learning_rate": 2.0408424682842718e-06, + "loss": 0.4824, + "step": 26008 + }, + { + "epoch": 9.6, + "learning_rate": 2.0408055179209264e-06, + "loss": 0.3091, + "step": 26009 + }, + { + "epoch": 9.6, + "learning_rate": 2.040768567557581e-06, + "loss": 0.3709, + "step": 26010 + }, + { + "epoch": 9.61, + "learning_rate": 2.0407316171942357e-06, + "loss": 0.4731, + "step": 26011 + }, + { + "epoch": 9.61, + "learning_rate": 2.0406946668308903e-06, + "loss": 0.4104, + "step": 26012 + }, + { + "epoch": 9.61, + "learning_rate": 2.0406577164675454e-06, + "loss": 0.3695, + "step": 26013 + }, + { + "epoch": 9.61, + "learning_rate": 2.0406207661042e-06, + "loss": 0.2317, + "step": 26014 + }, + { + "epoch": 9.61, + "learning_rate": 2.040583815740855e-06, + "loss": 0.3027, + "step": 26015 + }, + { + "epoch": 9.61, + "learning_rate": 2.0405468653775097e-06, + "loss": 0.2802, + "step": 26016 + }, + { + "epoch": 9.61, + "learning_rate": 2.0405099150141644e-06, + "loss": 0.5788, + "step": 26017 + }, + { + "epoch": 9.61, + "learning_rate": 2.040472964650819e-06, + "loss": 0.4148, + "step": 26018 + }, + { + "epoch": 9.61, + "learning_rate": 2.0404360142874736e-06, + "loss": 0.3759, + "step": 26019 + }, + { + "epoch": 9.61, + "learning_rate": 2.0403990639241287e-06, + "loss": 0.3739, + "step": 26020 + }, + { + "epoch": 9.61, + "learning_rate": 2.0403621135607833e-06, + "loss": 0.2328, + "step": 26021 + }, + { + "epoch": 9.61, + "learning_rate": 2.0403251631974384e-06, + "loss": 0.498, + "step": 26022 + }, + { + "epoch": 9.61, + "learning_rate": 2.040288212834093e-06, + "loss": 0.3413, + "step": 26023 + }, + { + "epoch": 9.61, + "learning_rate": 2.0402512624707477e-06, + "loss": 0.2801, + "step": 26024 + }, + { + "epoch": 9.61, + "learning_rate": 2.0402143121074023e-06, + "loss": 0.4406, + "step": 26025 + }, + { + "epoch": 9.61, + "learning_rate": 2.0401773617440574e-06, + "loss": 0.3732, + "step": 26026 + }, + { + "epoch": 9.61, + "learning_rate": 2.040140411380712e-06, + "loss": 0.3162, + "step": 26027 + }, + { + "epoch": 9.61, + "learning_rate": 2.040103461017367e-06, + "loss": 0.397, + "step": 26028 + }, + { + "epoch": 9.61, + "learning_rate": 2.0400665106540217e-06, + "loss": 0.336, + "step": 26029 + }, + { + "epoch": 9.61, + "learning_rate": 2.040029560290676e-06, + "loss": 0.3429, + "step": 26030 + }, + { + "epoch": 9.61, + "learning_rate": 2.039992609927331e-06, + "loss": 0.2816, + "step": 26031 + }, + { + "epoch": 9.61, + "learning_rate": 2.0399556595639856e-06, + "loss": 0.2987, + "step": 26032 + }, + { + "epoch": 9.61, + "learning_rate": 2.0399187092006407e-06, + "loss": 0.3614, + "step": 26033 + }, + { + "epoch": 9.61, + "learning_rate": 2.0398817588372953e-06, + "loss": 0.2989, + "step": 26034 + }, + { + "epoch": 9.61, + "learning_rate": 2.0398448084739504e-06, + "loss": 0.4279, + "step": 26035 + }, + { + "epoch": 9.61, + "learning_rate": 2.0398078581106046e-06, + "loss": 0.4502, + "step": 26036 + }, + { + "epoch": 9.61, + "learning_rate": 2.0397709077472596e-06, + "loss": 0.3816, + "step": 26037 + }, + { + "epoch": 9.62, + "learning_rate": 2.0397339573839143e-06, + "loss": 0.279, + "step": 26038 + }, + { + "epoch": 9.62, + "learning_rate": 2.039697007020569e-06, + "loss": 0.4105, + "step": 26039 + }, + { + "epoch": 9.62, + "learning_rate": 2.039660056657224e-06, + "loss": 0.4587, + "step": 26040 + }, + { + "epoch": 9.62, + "learning_rate": 2.0396231062938786e-06, + "loss": 0.4223, + "step": 26041 + }, + { + "epoch": 9.62, + "learning_rate": 2.0395861559305332e-06, + "loss": 0.3329, + "step": 26042 + }, + { + "epoch": 9.62, + "learning_rate": 2.039549205567188e-06, + "loss": 0.3814, + "step": 26043 + }, + { + "epoch": 9.62, + "learning_rate": 2.039512255203843e-06, + "loss": 0.2558, + "step": 26044 + }, + { + "epoch": 9.62, + "learning_rate": 2.0394753048404976e-06, + "loss": 0.4816, + "step": 26045 + }, + { + "epoch": 9.62, + "learning_rate": 2.0394383544771526e-06, + "loss": 0.3285, + "step": 26046 + }, + { + "epoch": 9.62, + "learning_rate": 2.0394014041138073e-06, + "loss": 0.3684, + "step": 26047 + }, + { + "epoch": 9.62, + "learning_rate": 2.039364453750462e-06, + "loss": 0.3286, + "step": 26048 + }, + { + "epoch": 9.62, + "learning_rate": 2.0393275033871165e-06, + "loss": 0.3215, + "step": 26049 + }, + { + "epoch": 9.62, + "learning_rate": 2.039290553023771e-06, + "loss": 0.3607, + "step": 26050 + }, + { + "epoch": 9.62, + "learning_rate": 2.0392536026604262e-06, + "loss": 0.3343, + "step": 26051 + }, + { + "epoch": 9.62, + "learning_rate": 2.039216652297081e-06, + "loss": 0.4572, + "step": 26052 + }, + { + "epoch": 9.62, + "learning_rate": 2.039179701933736e-06, + "loss": 0.3459, + "step": 26053 + }, + { + "epoch": 9.62, + "learning_rate": 2.0391427515703906e-06, + "loss": 0.3343, + "step": 26054 + }, + { + "epoch": 9.62, + "learning_rate": 2.039105801207045e-06, + "loss": 0.3088, + "step": 26055 + }, + { + "epoch": 9.62, + "learning_rate": 2.0390688508437e-06, + "loss": 0.4638, + "step": 26056 + }, + { + "epoch": 9.62, + "learning_rate": 2.039031900480355e-06, + "loss": 0.2918, + "step": 26057 + }, + { + "epoch": 9.62, + "learning_rate": 2.0389949501170095e-06, + "loss": 0.36, + "step": 26058 + }, + { + "epoch": 9.62, + "learning_rate": 2.038957999753664e-06, + "loss": 0.317, + "step": 26059 + }, + { + "epoch": 9.62, + "learning_rate": 2.0389210493903192e-06, + "loss": 0.4357, + "step": 26060 + }, + { + "epoch": 9.62, + "learning_rate": 2.038884099026974e-06, + "loss": 0.41, + "step": 26061 + }, + { + "epoch": 9.62, + "learning_rate": 2.0388471486636285e-06, + "loss": 0.3668, + "step": 26062 + }, + { + "epoch": 9.62, + "learning_rate": 2.038810198300283e-06, + "loss": 0.2777, + "step": 26063 + }, + { + "epoch": 9.62, + "learning_rate": 2.038773247936938e-06, + "loss": 0.3642, + "step": 26064 + }, + { + "epoch": 9.63, + "learning_rate": 2.038736297573593e-06, + "loss": 0.4051, + "step": 26065 + }, + { + "epoch": 9.63, + "learning_rate": 2.038699347210248e-06, + "loss": 0.4441, + "step": 26066 + }, + { + "epoch": 9.63, + "learning_rate": 2.0386623968469025e-06, + "loss": 0.4578, + "step": 26067 + }, + { + "epoch": 9.63, + "learning_rate": 2.038625446483557e-06, + "loss": 0.3931, + "step": 26068 + }, + { + "epoch": 9.63, + "learning_rate": 2.038588496120212e-06, + "loss": 0.3513, + "step": 26069 + }, + { + "epoch": 9.63, + "learning_rate": 2.0385515457568664e-06, + "loss": 0.2962, + "step": 26070 + }, + { + "epoch": 9.63, + "learning_rate": 2.0385145953935215e-06, + "loss": 0.3662, + "step": 26071 + }, + { + "epoch": 9.63, + "learning_rate": 2.038477645030176e-06, + "loss": 0.3022, + "step": 26072 + }, + { + "epoch": 9.63, + "learning_rate": 2.038440694666831e-06, + "loss": 0.3538, + "step": 26073 + }, + { + "epoch": 9.63, + "learning_rate": 2.038403744303486e-06, + "loss": 0.412, + "step": 26074 + }, + { + "epoch": 9.63, + "learning_rate": 2.0383667939401405e-06, + "loss": 0.3772, + "step": 26075 + }, + { + "epoch": 9.63, + "learning_rate": 2.038329843576795e-06, + "loss": 0.302, + "step": 26076 + }, + { + "epoch": 9.63, + "learning_rate": 2.03829289321345e-06, + "loss": 0.3852, + "step": 26077 + }, + { + "epoch": 9.63, + "learning_rate": 2.038255942850105e-06, + "loss": 0.3841, + "step": 26078 + }, + { + "epoch": 9.63, + "learning_rate": 2.0382189924867594e-06, + "loss": 0.4794, + "step": 26079 + }, + { + "epoch": 9.63, + "learning_rate": 2.0381820421234145e-06, + "loss": 0.3783, + "step": 26080 + }, + { + "epoch": 9.63, + "learning_rate": 2.0381450917600687e-06, + "loss": 0.372, + "step": 26081 + }, + { + "epoch": 9.63, + "learning_rate": 2.0381081413967238e-06, + "loss": 0.4137, + "step": 26082 + }, + { + "epoch": 9.63, + "learning_rate": 2.0380711910333784e-06, + "loss": 0.2246, + "step": 26083 + }, + { + "epoch": 9.63, + "learning_rate": 2.0380342406700335e-06, + "loss": 0.3617, + "step": 26084 + }, + { + "epoch": 9.63, + "learning_rate": 2.037997290306688e-06, + "loss": 0.4402, + "step": 26085 + }, + { + "epoch": 9.63, + "learning_rate": 2.037960339943343e-06, + "loss": 0.575, + "step": 26086 + }, + { + "epoch": 9.63, + "learning_rate": 2.0379233895799974e-06, + "loss": 0.2518, + "step": 26087 + }, + { + "epoch": 9.63, + "learning_rate": 2.037886439216652e-06, + "loss": 0.353, + "step": 26088 + }, + { + "epoch": 9.63, + "learning_rate": 2.037849488853307e-06, + "loss": 0.4907, + "step": 26089 + }, + { + "epoch": 9.63, + "learning_rate": 2.0378125384899617e-06, + "loss": 0.4862, + "step": 26090 + }, + { + "epoch": 9.63, + "learning_rate": 2.0377755881266168e-06, + "loss": 0.4399, + "step": 26091 + }, + { + "epoch": 9.64, + "learning_rate": 2.0377386377632714e-06, + "loss": 0.3238, + "step": 26092 + }, + { + "epoch": 9.64, + "learning_rate": 2.037701687399926e-06, + "loss": 0.3009, + "step": 26093 + }, + { + "epoch": 9.64, + "learning_rate": 2.0376647370365807e-06, + "loss": 0.347, + "step": 26094 + }, + { + "epoch": 9.64, + "learning_rate": 2.0376277866732357e-06, + "loss": 0.3595, + "step": 26095 + }, + { + "epoch": 9.64, + "learning_rate": 2.0375908363098904e-06, + "loss": 0.2907, + "step": 26096 + }, + { + "epoch": 9.64, + "learning_rate": 2.0375538859465454e-06, + "loss": 0.4219, + "step": 26097 + }, + { + "epoch": 9.64, + "learning_rate": 2.0375169355832e-06, + "loss": 0.3999, + "step": 26098 + }, + { + "epoch": 9.64, + "learning_rate": 2.0374799852198547e-06, + "loss": 0.355, + "step": 26099 + }, + { + "epoch": 9.64, + "learning_rate": 2.0374430348565093e-06, + "loss": 0.3474, + "step": 26100 + }, + { + "epoch": 9.64, + "learning_rate": 2.037406084493164e-06, + "loss": 0.3889, + "step": 26101 + }, + { + "epoch": 9.64, + "learning_rate": 2.037369134129819e-06, + "loss": 0.3875, + "step": 26102 + }, + { + "epoch": 9.64, + "learning_rate": 2.0373321837664737e-06, + "loss": 0.3387, + "step": 26103 + }, + { + "epoch": 9.64, + "learning_rate": 2.0372952334031287e-06, + "loss": 0.2749, + "step": 26104 + }, + { + "epoch": 9.64, + "learning_rate": 2.0372582830397834e-06, + "loss": 0.257, + "step": 26105 + }, + { + "epoch": 9.64, + "learning_rate": 2.037221332676438e-06, + "loss": 0.2977, + "step": 26106 + }, + { + "epoch": 9.64, + "learning_rate": 2.0371843823130926e-06, + "loss": 0.3578, + "step": 26107 + }, + { + "epoch": 9.64, + "learning_rate": 2.0371474319497473e-06, + "loss": 0.2963, + "step": 26108 + }, + { + "epoch": 9.64, + "learning_rate": 2.0371104815864023e-06, + "loss": 0.3003, + "step": 26109 + }, + { + "epoch": 9.64, + "learning_rate": 2.037073531223057e-06, + "loss": 0.387, + "step": 26110 + }, + { + "epoch": 9.64, + "learning_rate": 2.037036580859712e-06, + "loss": 0.4136, + "step": 26111 + }, + { + "epoch": 9.64, + "learning_rate": 2.0369996304963667e-06, + "loss": 0.3373, + "step": 26112 + }, + { + "epoch": 9.64, + "learning_rate": 2.0369626801330213e-06, + "loss": 0.2902, + "step": 26113 + }, + { + "epoch": 9.64, + "learning_rate": 2.036925729769676e-06, + "loss": 0.3284, + "step": 26114 + }, + { + "epoch": 9.64, + "learning_rate": 2.036888779406331e-06, + "loss": 0.3607, + "step": 26115 + }, + { + "epoch": 9.64, + "learning_rate": 2.0368518290429856e-06, + "loss": 0.3906, + "step": 26116 + }, + { + "epoch": 9.64, + "learning_rate": 2.0368148786796407e-06, + "loss": 0.4601, + "step": 26117 + }, + { + "epoch": 9.64, + "learning_rate": 2.0367779283162953e-06, + "loss": 0.4443, + "step": 26118 + }, + { + "epoch": 9.65, + "learning_rate": 2.03674097795295e-06, + "loss": 0.3644, + "step": 26119 + }, + { + "epoch": 9.65, + "learning_rate": 2.0367040275896046e-06, + "loss": 0.5356, + "step": 26120 + }, + { + "epoch": 9.65, + "learning_rate": 2.0366670772262593e-06, + "loss": 0.2847, + "step": 26121 + }, + { + "epoch": 9.65, + "learning_rate": 2.0366301268629143e-06, + "loss": 0.3295, + "step": 26122 + }, + { + "epoch": 9.65, + "learning_rate": 2.036593176499569e-06, + "loss": 0.2631, + "step": 26123 + }, + { + "epoch": 9.65, + "learning_rate": 2.036556226136224e-06, + "loss": 0.4022, + "step": 26124 + }, + { + "epoch": 9.65, + "learning_rate": 2.0365192757728786e-06, + "loss": 0.2113, + "step": 26125 + }, + { + "epoch": 9.65, + "learning_rate": 2.0364823254095333e-06, + "loss": 0.3235, + "step": 26126 + }, + { + "epoch": 9.65, + "learning_rate": 2.036445375046188e-06, + "loss": 0.4584, + "step": 26127 + }, + { + "epoch": 9.65, + "learning_rate": 2.0364084246828426e-06, + "loss": 0.2476, + "step": 26128 + }, + { + "epoch": 9.65, + "learning_rate": 2.0363714743194976e-06, + "loss": 0.3552, + "step": 26129 + }, + { + "epoch": 9.65, + "learning_rate": 2.0363345239561523e-06, + "loss": 0.2938, + "step": 26130 + }, + { + "epoch": 9.65, + "learning_rate": 2.0362975735928073e-06, + "loss": 0.3246, + "step": 26131 + }, + { + "epoch": 9.65, + "learning_rate": 2.0362606232294615e-06, + "loss": 0.3391, + "step": 26132 + }, + { + "epoch": 9.65, + "learning_rate": 2.0362236728661166e-06, + "loss": 0.3752, + "step": 26133 + }, + { + "epoch": 9.65, + "learning_rate": 2.0361867225027712e-06, + "loss": 0.3582, + "step": 26134 + }, + { + "epoch": 9.65, + "learning_rate": 2.0361497721394263e-06, + "loss": 0.3502, + "step": 26135 + }, + { + "epoch": 9.65, + "learning_rate": 2.036112821776081e-06, + "loss": 0.4591, + "step": 26136 + }, + { + "epoch": 9.65, + "learning_rate": 2.036075871412736e-06, + "loss": 0.4178, + "step": 26137 + }, + { + "epoch": 9.65, + "learning_rate": 2.03603892104939e-06, + "loss": 0.4885, + "step": 26138 + }, + { + "epoch": 9.65, + "learning_rate": 2.036001970686045e-06, + "loss": 0.2605, + "step": 26139 + }, + { + "epoch": 9.65, + "learning_rate": 2.0359650203227e-06, + "loss": 0.5096, + "step": 26140 + }, + { + "epoch": 9.65, + "learning_rate": 2.0359280699593545e-06, + "loss": 0.3323, + "step": 26141 + }, + { + "epoch": 9.65, + "learning_rate": 2.0358911195960096e-06, + "loss": 0.424, + "step": 26142 + }, + { + "epoch": 9.65, + "learning_rate": 2.0358541692326642e-06, + "loss": 0.4632, + "step": 26143 + }, + { + "epoch": 9.65, + "learning_rate": 2.035817218869319e-06, + "loss": 0.4785, + "step": 26144 + }, + { + "epoch": 9.65, + "learning_rate": 2.0357802685059735e-06, + "loss": 0.33, + "step": 26145 + }, + { + "epoch": 9.66, + "learning_rate": 2.0357433181426286e-06, + "loss": 0.3676, + "step": 26146 + }, + { + "epoch": 9.66, + "learning_rate": 2.035706367779283e-06, + "loss": 0.4212, + "step": 26147 + }, + { + "epoch": 9.66, + "learning_rate": 2.035669417415938e-06, + "loss": 0.3891, + "step": 26148 + }, + { + "epoch": 9.66, + "learning_rate": 2.035632467052593e-06, + "loss": 0.2534, + "step": 26149 + }, + { + "epoch": 9.66, + "learning_rate": 2.0355955166892475e-06, + "loss": 0.4961, + "step": 26150 + }, + { + "epoch": 9.66, + "learning_rate": 2.035558566325902e-06, + "loss": 0.4419, + "step": 26151 + }, + { + "epoch": 9.66, + "learning_rate": 2.035521615962557e-06, + "loss": 0.3814, + "step": 26152 + }, + { + "epoch": 9.66, + "learning_rate": 2.035484665599212e-06, + "loss": 0.6562, + "step": 26153 + }, + { + "epoch": 9.66, + "learning_rate": 2.0354477152358665e-06, + "loss": 0.3134, + "step": 26154 + }, + { + "epoch": 9.66, + "learning_rate": 2.0354107648725215e-06, + "loss": 0.3682, + "step": 26155 + }, + { + "epoch": 9.66, + "learning_rate": 2.035373814509176e-06, + "loss": 0.3493, + "step": 26156 + }, + { + "epoch": 9.66, + "learning_rate": 2.035336864145831e-06, + "loss": 0.3482, + "step": 26157 + }, + { + "epoch": 9.66, + "learning_rate": 2.0352999137824855e-06, + "loss": 0.4331, + "step": 26158 + }, + { + "epoch": 9.66, + "learning_rate": 2.03526296341914e-06, + "loss": 0.3047, + "step": 26159 + }, + { + "epoch": 9.66, + "learning_rate": 2.035226013055795e-06, + "loss": 0.3741, + "step": 26160 + }, + { + "epoch": 9.66, + "learning_rate": 2.03518906269245e-06, + "loss": 0.3116, + "step": 26161 + }, + { + "epoch": 9.66, + "learning_rate": 2.035152112329105e-06, + "loss": 0.3341, + "step": 26162 + }, + { + "epoch": 9.66, + "learning_rate": 2.0351151619657595e-06, + "loss": 0.4293, + "step": 26163 + }, + { + "epoch": 9.66, + "learning_rate": 2.035078211602414e-06, + "loss": 0.4243, + "step": 26164 + }, + { + "epoch": 9.66, + "learning_rate": 2.0350412612390688e-06, + "loss": 0.3757, + "step": 26165 + }, + { + "epoch": 9.66, + "learning_rate": 2.035004310875724e-06, + "loss": 0.2293, + "step": 26166 + }, + { + "epoch": 9.66, + "learning_rate": 2.0349673605123785e-06, + "loss": 0.4312, + "step": 26167 + }, + { + "epoch": 9.66, + "learning_rate": 2.034930410149033e-06, + "loss": 0.362, + "step": 26168 + }, + { + "epoch": 9.66, + "learning_rate": 2.034893459785688e-06, + "loss": 0.4556, + "step": 26169 + }, + { + "epoch": 9.66, + "learning_rate": 2.034856509422343e-06, + "loss": 0.382, + "step": 26170 + }, + { + "epoch": 9.66, + "learning_rate": 2.0348195590589974e-06, + "loss": 0.4822, + "step": 26171 + }, + { + "epoch": 9.66, + "learning_rate": 2.034782608695652e-06, + "loss": 0.3712, + "step": 26172 + }, + { + "epoch": 9.67, + "learning_rate": 2.034745658332307e-06, + "loss": 0.3153, + "step": 26173 + }, + { + "epoch": 9.67, + "learning_rate": 2.0347087079689618e-06, + "loss": 0.4863, + "step": 26174 + }, + { + "epoch": 9.67, + "learning_rate": 2.034671757605617e-06, + "loss": 0.3148, + "step": 26175 + }, + { + "epoch": 9.67, + "learning_rate": 2.034634807242271e-06, + "loss": 0.3786, + "step": 26176 + }, + { + "epoch": 9.67, + "learning_rate": 2.034597856878926e-06, + "loss": 0.3516, + "step": 26177 + }, + { + "epoch": 9.67, + "learning_rate": 2.0345609065155807e-06, + "loss": 0.3647, + "step": 26178 + }, + { + "epoch": 9.67, + "learning_rate": 2.0345239561522354e-06, + "loss": 0.438, + "step": 26179 + }, + { + "epoch": 9.67, + "learning_rate": 2.0344870057888904e-06, + "loss": 0.4748, + "step": 26180 + }, + { + "epoch": 9.67, + "learning_rate": 2.034450055425545e-06, + "loss": 0.3056, + "step": 26181 + }, + { + "epoch": 9.67, + "learning_rate": 2.0344131050621997e-06, + "loss": 0.1696, + "step": 26182 + }, + { + "epoch": 9.67, + "learning_rate": 2.0343761546988543e-06, + "loss": 0.2832, + "step": 26183 + }, + { + "epoch": 9.67, + "learning_rate": 2.0343392043355094e-06, + "loss": 0.3811, + "step": 26184 + }, + { + "epoch": 9.67, + "learning_rate": 2.034302253972164e-06, + "loss": 0.2295, + "step": 26185 + }, + { + "epoch": 9.67, + "learning_rate": 2.034265303608819e-06, + "loss": 0.4043, + "step": 26186 + }, + { + "epoch": 9.67, + "learning_rate": 2.0342283532454737e-06, + "loss": 0.365, + "step": 26187 + }, + { + "epoch": 9.67, + "learning_rate": 2.0341914028821284e-06, + "loss": 0.331, + "step": 26188 + }, + { + "epoch": 9.67, + "learning_rate": 2.034154452518783e-06, + "loss": 0.401, + "step": 26189 + }, + { + "epoch": 9.67, + "learning_rate": 2.0341175021554376e-06, + "loss": 0.4237, + "step": 26190 + }, + { + "epoch": 9.67, + "learning_rate": 2.0340805517920927e-06, + "loss": 0.3114, + "step": 26191 + }, + { + "epoch": 9.67, + "learning_rate": 2.0340436014287473e-06, + "loss": 0.2441, + "step": 26192 + }, + { + "epoch": 9.67, + "learning_rate": 2.0340066510654024e-06, + "loss": 0.4194, + "step": 26193 + }, + { + "epoch": 9.67, + "learning_rate": 2.033969700702057e-06, + "loss": 0.3544, + "step": 26194 + }, + { + "epoch": 9.67, + "learning_rate": 2.0339327503387117e-06, + "loss": 0.3406, + "step": 26195 + }, + { + "epoch": 9.67, + "learning_rate": 2.0338957999753663e-06, + "loss": 0.3173, + "step": 26196 + }, + { + "epoch": 9.67, + "learning_rate": 2.0338588496120214e-06, + "loss": 0.3425, + "step": 26197 + }, + { + "epoch": 9.67, + "learning_rate": 2.033821899248676e-06, + "loss": 0.3189, + "step": 26198 + }, + { + "epoch": 9.67, + "learning_rate": 2.0337849488853306e-06, + "loss": 0.3761, + "step": 26199 + }, + { + "epoch": 9.68, + "learning_rate": 2.0337479985219857e-06, + "loss": 0.4447, + "step": 26200 + }, + { + "epoch": 9.68, + "learning_rate": 2.0337110481586403e-06, + "loss": 0.2791, + "step": 26201 + }, + { + "epoch": 9.68, + "learning_rate": 2.033674097795295e-06, + "loss": 0.2531, + "step": 26202 + }, + { + "epoch": 9.68, + "learning_rate": 2.0336371474319496e-06, + "loss": 0.3287, + "step": 26203 + }, + { + "epoch": 9.68, + "learning_rate": 2.0336001970686047e-06, + "loss": 0.4808, + "step": 26204 + }, + { + "epoch": 9.68, + "learning_rate": 2.0335632467052593e-06, + "loss": 0.4012, + "step": 26205 + }, + { + "epoch": 9.68, + "learning_rate": 2.0335262963419144e-06, + "loss": 0.462, + "step": 26206 + }, + { + "epoch": 9.68, + "learning_rate": 2.033489345978569e-06, + "loss": 0.379, + "step": 26207 + }, + { + "epoch": 9.68, + "learning_rate": 2.0334523956152236e-06, + "loss": 0.3407, + "step": 26208 + }, + { + "epoch": 9.68, + "learning_rate": 2.0334154452518783e-06, + "loss": 0.381, + "step": 26209 + }, + { + "epoch": 9.68, + "learning_rate": 2.033378494888533e-06, + "loss": 0.4883, + "step": 26210 + }, + { + "epoch": 9.68, + "learning_rate": 2.033341544525188e-06, + "loss": 0.3086, + "step": 26211 + }, + { + "epoch": 9.68, + "learning_rate": 2.0333045941618426e-06, + "loss": 0.383, + "step": 26212 + }, + { + "epoch": 9.68, + "learning_rate": 2.0332676437984977e-06, + "loss": 0.4064, + "step": 26213 + }, + { + "epoch": 9.68, + "learning_rate": 2.0332306934351523e-06, + "loss": 0.2836, + "step": 26214 + }, + { + "epoch": 9.68, + "learning_rate": 2.033193743071807e-06, + "loss": 0.2404, + "step": 26215 + }, + { + "epoch": 9.68, + "learning_rate": 2.0331567927084616e-06, + "loss": 0.3279, + "step": 26216 + }, + { + "epoch": 9.68, + "learning_rate": 2.0331198423451166e-06, + "loss": 0.3715, + "step": 26217 + }, + { + "epoch": 9.68, + "learning_rate": 2.0330828919817713e-06, + "loss": 0.3506, + "step": 26218 + }, + { + "epoch": 9.68, + "learning_rate": 2.033045941618426e-06, + "loss": 0.406, + "step": 26219 + }, + { + "epoch": 9.68, + "learning_rate": 2.033008991255081e-06, + "loss": 0.4247, + "step": 26220 + }, + { + "epoch": 9.68, + "learning_rate": 2.032972040891735e-06, + "loss": 0.3045, + "step": 26221 + }, + { + "epoch": 9.68, + "learning_rate": 2.0329350905283902e-06, + "loss": 0.3386, + "step": 26222 + }, + { + "epoch": 9.68, + "learning_rate": 2.032898140165045e-06, + "loss": 0.4286, + "step": 26223 + }, + { + "epoch": 9.68, + "learning_rate": 2.0328611898017e-06, + "loss": 0.5059, + "step": 26224 + }, + { + "epoch": 9.68, + "learning_rate": 2.0328242394383546e-06, + "loss": 0.3184, + "step": 26225 + }, + { + "epoch": 9.68, + "learning_rate": 2.0327872890750096e-06, + "loss": 0.4352, + "step": 26226 + }, + { + "epoch": 9.69, + "learning_rate": 2.032750338711664e-06, + "loss": 0.3534, + "step": 26227 + }, + { + "epoch": 9.69, + "learning_rate": 2.0327133883483185e-06, + "loss": 0.4082, + "step": 26228 + }, + { + "epoch": 9.69, + "learning_rate": 2.0326764379849735e-06, + "loss": 0.417, + "step": 26229 + }, + { + "epoch": 9.69, + "learning_rate": 2.032639487621628e-06, + "loss": 0.3474, + "step": 26230 + }, + { + "epoch": 9.69, + "learning_rate": 2.0326025372582832e-06, + "loss": 0.4052, + "step": 26231 + }, + { + "epoch": 9.69, + "learning_rate": 2.032565586894938e-06, + "loss": 0.3562, + "step": 26232 + }, + { + "epoch": 9.69, + "learning_rate": 2.0325286365315925e-06, + "loss": 0.4198, + "step": 26233 + }, + { + "epoch": 9.69, + "learning_rate": 2.032491686168247e-06, + "loss": 0.3333, + "step": 26234 + }, + { + "epoch": 9.69, + "learning_rate": 2.032454735804902e-06, + "loss": 0.3597, + "step": 26235 + }, + { + "epoch": 9.69, + "learning_rate": 2.032417785441557e-06, + "loss": 0.457, + "step": 26236 + }, + { + "epoch": 9.69, + "learning_rate": 2.032380835078212e-06, + "loss": 0.3909, + "step": 26237 + }, + { + "epoch": 9.69, + "learning_rate": 2.0323438847148665e-06, + "loss": 0.3781, + "step": 26238 + }, + { + "epoch": 9.69, + "learning_rate": 2.032306934351521e-06, + "loss": 0.421, + "step": 26239 + }, + { + "epoch": 9.69, + "learning_rate": 2.032269983988176e-06, + "loss": 0.3568, + "step": 26240 + }, + { + "epoch": 9.69, + "learning_rate": 2.0322330336248304e-06, + "loss": 0.3425, + "step": 26241 + }, + { + "epoch": 9.69, + "learning_rate": 2.0321960832614855e-06, + "loss": 0.3885, + "step": 26242 + }, + { + "epoch": 9.69, + "learning_rate": 2.03215913289814e-06, + "loss": 0.3304, + "step": 26243 + }, + { + "epoch": 9.69, + "learning_rate": 2.032122182534795e-06, + "loss": 0.3769, + "step": 26244 + }, + { + "epoch": 9.69, + "learning_rate": 2.03208523217145e-06, + "loss": 0.2716, + "step": 26245 + }, + { + "epoch": 9.69, + "learning_rate": 2.0320482818081045e-06, + "loss": 0.3428, + "step": 26246 + }, + { + "epoch": 9.69, + "learning_rate": 2.032011331444759e-06, + "loss": 0.3044, + "step": 26247 + }, + { + "epoch": 9.69, + "learning_rate": 2.0319743810814137e-06, + "loss": 0.5483, + "step": 26248 + }, + { + "epoch": 9.69, + "learning_rate": 2.031937430718069e-06, + "loss": 0.3599, + "step": 26249 + }, + { + "epoch": 9.69, + "learning_rate": 2.0319004803547234e-06, + "loss": 0.243, + "step": 26250 + }, + { + "epoch": 9.69, + "learning_rate": 2.0318635299913785e-06, + "loss": 0.2812, + "step": 26251 + }, + { + "epoch": 9.69, + "learning_rate": 2.031826579628033e-06, + "loss": 0.2756, + "step": 26252 + }, + { + "epoch": 9.69, + "learning_rate": 2.0317896292646878e-06, + "loss": 0.3452, + "step": 26253 + }, + { + "epoch": 9.69, + "learning_rate": 2.0317526789013424e-06, + "loss": 0.3669, + "step": 26254 + }, + { + "epoch": 9.7, + "learning_rate": 2.0317157285379975e-06, + "loss": 0.3872, + "step": 26255 + }, + { + "epoch": 9.7, + "learning_rate": 2.031678778174652e-06, + "loss": 0.3758, + "step": 26256 + }, + { + "epoch": 9.7, + "learning_rate": 2.031641827811307e-06, + "loss": 0.2464, + "step": 26257 + }, + { + "epoch": 9.7, + "learning_rate": 2.031604877447962e-06, + "loss": 0.3172, + "step": 26258 + }, + { + "epoch": 9.7, + "learning_rate": 2.0315679270846164e-06, + "loss": 0.2887, + "step": 26259 + }, + { + "epoch": 9.7, + "learning_rate": 2.031530976721271e-06, + "loss": 0.3545, + "step": 26260 + }, + { + "epoch": 9.7, + "learning_rate": 2.0314940263579257e-06, + "loss": 0.4857, + "step": 26261 + }, + { + "epoch": 9.7, + "learning_rate": 2.0314570759945808e-06, + "loss": 0.4565, + "step": 26262 + }, + { + "epoch": 9.7, + "learning_rate": 2.0314201256312354e-06, + "loss": 0.3305, + "step": 26263 + }, + { + "epoch": 9.7, + "learning_rate": 2.0313831752678905e-06, + "loss": 0.2833, + "step": 26264 + }, + { + "epoch": 9.7, + "learning_rate": 2.031346224904545e-06, + "loss": 0.3981, + "step": 26265 + }, + { + "epoch": 9.7, + "learning_rate": 2.0313092745411997e-06, + "loss": 0.365, + "step": 26266 + }, + { + "epoch": 9.7, + "learning_rate": 2.0312723241778544e-06, + "loss": 0.4487, + "step": 26267 + }, + { + "epoch": 9.7, + "learning_rate": 2.031235373814509e-06, + "loss": 0.4203, + "step": 26268 + }, + { + "epoch": 9.7, + "learning_rate": 2.031198423451164e-06, + "loss": 0.3122, + "step": 26269 + }, + { + "epoch": 9.7, + "learning_rate": 2.0311614730878187e-06, + "loss": 0.3983, + "step": 26270 + }, + { + "epoch": 9.7, + "learning_rate": 2.0311245227244738e-06, + "loss": 0.3837, + "step": 26271 + }, + { + "epoch": 9.7, + "learning_rate": 2.031087572361128e-06, + "loss": 0.4348, + "step": 26272 + }, + { + "epoch": 9.7, + "learning_rate": 2.031050621997783e-06, + "loss": 0.372, + "step": 26273 + }, + { + "epoch": 9.7, + "learning_rate": 2.0310136716344377e-06, + "loss": 0.3661, + "step": 26274 + }, + { + "epoch": 9.7, + "learning_rate": 2.0309767212710927e-06, + "loss": 0.4269, + "step": 26275 + }, + { + "epoch": 9.7, + "learning_rate": 2.0309397709077474e-06, + "loss": 0.3581, + "step": 26276 + }, + { + "epoch": 9.7, + "learning_rate": 2.030902820544402e-06, + "loss": 0.3575, + "step": 26277 + }, + { + "epoch": 9.7, + "learning_rate": 2.0308658701810567e-06, + "loss": 0.5146, + "step": 26278 + }, + { + "epoch": 9.7, + "learning_rate": 2.0308289198177113e-06, + "loss": 0.4269, + "step": 26279 + }, + { + "epoch": 9.7, + "learning_rate": 2.0307919694543664e-06, + "loss": 0.4643, + "step": 26280 + }, + { + "epoch": 9.7, + "learning_rate": 2.030755019091021e-06, + "loss": 0.2692, + "step": 26281 + }, + { + "epoch": 9.71, + "learning_rate": 2.030718068727676e-06, + "loss": 0.2464, + "step": 26282 + }, + { + "epoch": 9.71, + "learning_rate": 2.0306811183643307e-06, + "loss": 0.3071, + "step": 26283 + }, + { + "epoch": 9.71, + "learning_rate": 2.0306441680009853e-06, + "loss": 0.3325, + "step": 26284 + }, + { + "epoch": 9.71, + "learning_rate": 2.03060721763764e-06, + "loss": 0.351, + "step": 26285 + }, + { + "epoch": 9.71, + "learning_rate": 2.030570267274295e-06, + "loss": 0.3991, + "step": 26286 + }, + { + "epoch": 9.71, + "learning_rate": 2.0305333169109497e-06, + "loss": 0.3375, + "step": 26287 + }, + { + "epoch": 9.71, + "learning_rate": 2.0304963665476043e-06, + "loss": 0.3165, + "step": 26288 + }, + { + "epoch": 9.71, + "learning_rate": 2.0304594161842593e-06, + "loss": 0.2132, + "step": 26289 + }, + { + "epoch": 9.71, + "learning_rate": 2.030422465820914e-06, + "loss": 0.4587, + "step": 26290 + }, + { + "epoch": 9.71, + "learning_rate": 2.0303855154575686e-06, + "loss": 0.4567, + "step": 26291 + }, + { + "epoch": 9.71, + "learning_rate": 2.0303485650942233e-06, + "loss": 0.3752, + "step": 26292 + }, + { + "epoch": 9.71, + "learning_rate": 2.0303116147308783e-06, + "loss": 0.4165, + "step": 26293 + }, + { + "epoch": 9.71, + "learning_rate": 2.030274664367533e-06, + "loss": 0.4216, + "step": 26294 + }, + { + "epoch": 9.71, + "learning_rate": 2.030237714004188e-06, + "loss": 0.3128, + "step": 26295 + }, + { + "epoch": 9.71, + "learning_rate": 2.0302007636408427e-06, + "loss": 0.4523, + "step": 26296 + }, + { + "epoch": 9.71, + "learning_rate": 2.0301638132774973e-06, + "loss": 0.3472, + "step": 26297 + }, + { + "epoch": 9.71, + "learning_rate": 2.030126862914152e-06, + "loss": 0.3106, + "step": 26298 + }, + { + "epoch": 9.71, + "learning_rate": 2.0300899125508066e-06, + "loss": 0.3819, + "step": 26299 + }, + { + "epoch": 9.71, + "learning_rate": 2.0300529621874616e-06, + "loss": 0.1998, + "step": 26300 + }, + { + "epoch": 9.71, + "learning_rate": 2.0300160118241163e-06, + "loss": 0.3136, + "step": 26301 + }, + { + "epoch": 9.71, + "learning_rate": 2.0299790614607713e-06, + "loss": 0.3575, + "step": 26302 + }, + { + "epoch": 9.71, + "learning_rate": 2.029942111097426e-06, + "loss": 0.4993, + "step": 26303 + }, + { + "epoch": 9.71, + "learning_rate": 2.0299051607340806e-06, + "loss": 0.5069, + "step": 26304 + }, + { + "epoch": 9.71, + "learning_rate": 2.0298682103707352e-06, + "loss": 0.2692, + "step": 26305 + }, + { + "epoch": 9.71, + "learning_rate": 2.0298312600073903e-06, + "loss": 0.3492, + "step": 26306 + }, + { + "epoch": 9.71, + "learning_rate": 2.029794309644045e-06, + "loss": 0.4172, + "step": 26307 + }, + { + "epoch": 9.71, + "learning_rate": 2.0297573592806996e-06, + "loss": 0.3384, + "step": 26308 + }, + { + "epoch": 9.72, + "learning_rate": 2.0297204089173546e-06, + "loss": 0.5611, + "step": 26309 + }, + { + "epoch": 9.72, + "learning_rate": 2.0296834585540093e-06, + "loss": 0.2237, + "step": 26310 + }, + { + "epoch": 9.72, + "learning_rate": 2.029646508190664e-06, + "loss": 0.3386, + "step": 26311 + }, + { + "epoch": 9.72, + "learning_rate": 2.0296095578273185e-06, + "loss": 0.4825, + "step": 26312 + }, + { + "epoch": 9.72, + "learning_rate": 2.0295726074639736e-06, + "loss": 0.3567, + "step": 26313 + }, + { + "epoch": 9.72, + "learning_rate": 2.0295356571006282e-06, + "loss": 0.3176, + "step": 26314 + }, + { + "epoch": 9.72, + "learning_rate": 2.0294987067372833e-06, + "loss": 0.2348, + "step": 26315 + }, + { + "epoch": 9.72, + "learning_rate": 2.029461756373938e-06, + "loss": 0.3294, + "step": 26316 + }, + { + "epoch": 9.72, + "learning_rate": 2.029424806010592e-06, + "loss": 0.3471, + "step": 26317 + }, + { + "epoch": 9.72, + "learning_rate": 2.029387855647247e-06, + "loss": 0.351, + "step": 26318 + }, + { + "epoch": 9.72, + "learning_rate": 2.029350905283902e-06, + "loss": 0.3109, + "step": 26319 + }, + { + "epoch": 9.72, + "learning_rate": 2.029313954920557e-06, + "loss": 0.2527, + "step": 26320 + }, + { + "epoch": 9.72, + "learning_rate": 2.0292770045572115e-06, + "loss": 0.2301, + "step": 26321 + }, + { + "epoch": 9.72, + "learning_rate": 2.0292400541938666e-06, + "loss": 0.3302, + "step": 26322 + }, + { + "epoch": 9.72, + "learning_rate": 2.029203103830521e-06, + "loss": 0.3767, + "step": 26323 + }, + { + "epoch": 9.72, + "learning_rate": 2.029166153467176e-06, + "loss": 0.4396, + "step": 26324 + }, + { + "epoch": 9.72, + "learning_rate": 2.0291292031038305e-06, + "loss": 0.4847, + "step": 26325 + }, + { + "epoch": 9.72, + "learning_rate": 2.0290922527404856e-06, + "loss": 0.3259, + "step": 26326 + }, + { + "epoch": 9.72, + "learning_rate": 2.02905530237714e-06, + "loss": 0.2605, + "step": 26327 + }, + { + "epoch": 9.72, + "learning_rate": 2.029018352013795e-06, + "loss": 0.527, + "step": 26328 + }, + { + "epoch": 9.72, + "learning_rate": 2.0289814016504495e-06, + "loss": 0.3415, + "step": 26329 + }, + { + "epoch": 9.72, + "learning_rate": 2.028944451287104e-06, + "loss": 0.313, + "step": 26330 + }, + { + "epoch": 9.72, + "learning_rate": 2.028907500923759e-06, + "loss": 0.3786, + "step": 26331 + }, + { + "epoch": 9.72, + "learning_rate": 2.028870550560414e-06, + "loss": 0.4541, + "step": 26332 + }, + { + "epoch": 9.72, + "learning_rate": 2.028833600197069e-06, + "loss": 0.4128, + "step": 26333 + }, + { + "epoch": 9.72, + "learning_rate": 2.0287966498337235e-06, + "loss": 0.401, + "step": 26334 + }, + { + "epoch": 9.72, + "learning_rate": 2.028759699470378e-06, + "loss": 0.2924, + "step": 26335 + }, + { + "epoch": 9.73, + "learning_rate": 2.0287227491070328e-06, + "loss": 0.3956, + "step": 26336 + }, + { + "epoch": 9.73, + "learning_rate": 2.0286857987436874e-06, + "loss": 0.4319, + "step": 26337 + }, + { + "epoch": 9.73, + "learning_rate": 2.0286488483803425e-06, + "loss": 0.3881, + "step": 26338 + }, + { + "epoch": 9.73, + "learning_rate": 2.028611898016997e-06, + "loss": 0.3205, + "step": 26339 + }, + { + "epoch": 9.73, + "learning_rate": 2.028574947653652e-06, + "loss": 0.2832, + "step": 26340 + }, + { + "epoch": 9.73, + "learning_rate": 2.028537997290307e-06, + "loss": 0.3681, + "step": 26341 + }, + { + "epoch": 9.73, + "learning_rate": 2.0285010469269614e-06, + "loss": 0.3126, + "step": 26342 + }, + { + "epoch": 9.73, + "learning_rate": 2.028464096563616e-06, + "loss": 0.256, + "step": 26343 + }, + { + "epoch": 9.73, + "learning_rate": 2.028427146200271e-06, + "loss": 0.3301, + "step": 26344 + }, + { + "epoch": 9.73, + "learning_rate": 2.0283901958369258e-06, + "loss": 0.3722, + "step": 26345 + }, + { + "epoch": 9.73, + "learning_rate": 2.028353245473581e-06, + "loss": 0.3819, + "step": 26346 + }, + { + "epoch": 9.73, + "learning_rate": 2.0283162951102355e-06, + "loss": 0.3581, + "step": 26347 + }, + { + "epoch": 9.73, + "learning_rate": 2.02827934474689e-06, + "loss": 0.3523, + "step": 26348 + }, + { + "epoch": 9.73, + "learning_rate": 2.0282423943835447e-06, + "loss": 0.4131, + "step": 26349 + }, + { + "epoch": 9.73, + "learning_rate": 2.0282054440201994e-06, + "loss": 0.2919, + "step": 26350 + }, + { + "epoch": 9.73, + "learning_rate": 2.0281684936568544e-06, + "loss": 0.4615, + "step": 26351 + }, + { + "epoch": 9.73, + "learning_rate": 2.028131543293509e-06, + "loss": 0.3177, + "step": 26352 + }, + { + "epoch": 9.73, + "learning_rate": 2.028094592930164e-06, + "loss": 0.3231, + "step": 26353 + }, + { + "epoch": 9.73, + "learning_rate": 2.0280576425668188e-06, + "loss": 0.3958, + "step": 26354 + }, + { + "epoch": 9.73, + "learning_rate": 2.0280206922034734e-06, + "loss": 0.3835, + "step": 26355 + }, + { + "epoch": 9.73, + "learning_rate": 2.027983741840128e-06, + "loss": 0.3034, + "step": 26356 + }, + { + "epoch": 9.73, + "learning_rate": 2.0279467914767827e-06, + "loss": 0.3552, + "step": 26357 + }, + { + "epoch": 9.73, + "learning_rate": 2.0279098411134377e-06, + "loss": 0.339, + "step": 26358 + }, + { + "epoch": 9.73, + "learning_rate": 2.0278728907500924e-06, + "loss": 0.3846, + "step": 26359 + }, + { + "epoch": 9.73, + "learning_rate": 2.0278359403867474e-06, + "loss": 0.3864, + "step": 26360 + }, + { + "epoch": 9.73, + "learning_rate": 2.027798990023402e-06, + "loss": 0.3286, + "step": 26361 + }, + { + "epoch": 9.73, + "learning_rate": 2.0277620396600567e-06, + "loss": 0.3925, + "step": 26362 + }, + { + "epoch": 9.74, + "learning_rate": 2.0277250892967113e-06, + "loss": 0.3532, + "step": 26363 + }, + { + "epoch": 9.74, + "learning_rate": 2.0276881389333664e-06, + "loss": 0.4093, + "step": 26364 + }, + { + "epoch": 9.74, + "learning_rate": 2.027651188570021e-06, + "loss": 0.3626, + "step": 26365 + }, + { + "epoch": 9.74, + "learning_rate": 2.027614238206676e-06, + "loss": 0.3706, + "step": 26366 + }, + { + "epoch": 9.74, + "learning_rate": 2.0275772878433303e-06, + "loss": 0.2954, + "step": 26367 + }, + { + "epoch": 9.74, + "learning_rate": 2.027540337479985e-06, + "loss": 0.4454, + "step": 26368 + }, + { + "epoch": 9.74, + "learning_rate": 2.02750338711664e-06, + "loss": 0.2952, + "step": 26369 + }, + { + "epoch": 9.74, + "learning_rate": 2.0274664367532946e-06, + "loss": 0.2856, + "step": 26370 + }, + { + "epoch": 9.74, + "learning_rate": 2.0274294863899497e-06, + "loss": 0.4334, + "step": 26371 + }, + { + "epoch": 9.74, + "learning_rate": 2.0273925360266043e-06, + "loss": 0.3701, + "step": 26372 + }, + { + "epoch": 9.74, + "learning_rate": 2.027355585663259e-06, + "loss": 0.361, + "step": 26373 + }, + { + "epoch": 9.74, + "learning_rate": 2.0273186352999136e-06, + "loss": 0.4155, + "step": 26374 + }, + { + "epoch": 9.74, + "learning_rate": 2.0272816849365687e-06, + "loss": 0.196, + "step": 26375 + }, + { + "epoch": 9.74, + "learning_rate": 2.0272447345732233e-06, + "loss": 0.3889, + "step": 26376 + }, + { + "epoch": 9.74, + "learning_rate": 2.027207784209878e-06, + "loss": 0.3888, + "step": 26377 + }, + { + "epoch": 9.74, + "learning_rate": 2.027170833846533e-06, + "loss": 0.5155, + "step": 26378 + }, + { + "epoch": 9.74, + "learning_rate": 2.0271338834831876e-06, + "loss": 0.4177, + "step": 26379 + }, + { + "epoch": 9.74, + "learning_rate": 2.0270969331198423e-06, + "loss": 0.3698, + "step": 26380 + }, + { + "epoch": 9.74, + "learning_rate": 2.027059982756497e-06, + "loss": 0.3148, + "step": 26381 + }, + { + "epoch": 9.74, + "learning_rate": 2.027023032393152e-06, + "loss": 0.3085, + "step": 26382 + }, + { + "epoch": 9.74, + "learning_rate": 2.0269860820298066e-06, + "loss": 0.3674, + "step": 26383 + }, + { + "epoch": 9.74, + "learning_rate": 2.0269491316664617e-06, + "loss": 0.3034, + "step": 26384 + }, + { + "epoch": 9.74, + "learning_rate": 2.0269121813031163e-06, + "loss": 0.4404, + "step": 26385 + }, + { + "epoch": 9.74, + "learning_rate": 2.026875230939771e-06, + "loss": 0.2863, + "step": 26386 + }, + { + "epoch": 9.74, + "learning_rate": 2.0268382805764256e-06, + "loss": 0.3268, + "step": 26387 + }, + { + "epoch": 9.74, + "learning_rate": 2.0268013302130802e-06, + "loss": 0.3798, + "step": 26388 + }, + { + "epoch": 9.74, + "learning_rate": 2.0267643798497353e-06, + "loss": 0.5536, + "step": 26389 + }, + { + "epoch": 9.75, + "learning_rate": 2.02672742948639e-06, + "loss": 0.3179, + "step": 26390 + }, + { + "epoch": 9.75, + "learning_rate": 2.026690479123045e-06, + "loss": 0.3532, + "step": 26391 + }, + { + "epoch": 9.75, + "learning_rate": 2.0266535287596996e-06, + "loss": 0.3081, + "step": 26392 + }, + { + "epoch": 9.75, + "learning_rate": 2.0266165783963542e-06, + "loss": 0.295, + "step": 26393 + }, + { + "epoch": 9.75, + "learning_rate": 2.026579628033009e-06, + "loss": 0.461, + "step": 26394 + }, + { + "epoch": 9.75, + "learning_rate": 2.026542677669664e-06, + "loss": 0.5078, + "step": 26395 + }, + { + "epoch": 9.75, + "learning_rate": 2.0265057273063186e-06, + "loss": 0.3074, + "step": 26396 + }, + { + "epoch": 9.75, + "learning_rate": 2.0264687769429732e-06, + "loss": 0.3328, + "step": 26397 + }, + { + "epoch": 9.75, + "learning_rate": 2.0264318265796283e-06, + "loss": 0.336, + "step": 26398 + }, + { + "epoch": 9.75, + "learning_rate": 2.026394876216283e-06, + "loss": 0.4384, + "step": 26399 + }, + { + "epoch": 9.75, + "learning_rate": 2.0263579258529375e-06, + "loss": 0.3808, + "step": 26400 + }, + { + "epoch": 9.75, + "learning_rate": 2.026320975489592e-06, + "loss": 0.358, + "step": 26401 + }, + { + "epoch": 9.75, + "learning_rate": 2.0262840251262472e-06, + "loss": 0.461, + "step": 26402 + }, + { + "epoch": 9.75, + "learning_rate": 2.026247074762902e-06, + "loss": 0.4672, + "step": 26403 + }, + { + "epoch": 9.75, + "learning_rate": 2.026210124399557e-06, + "loss": 0.4232, + "step": 26404 + }, + { + "epoch": 9.75, + "learning_rate": 2.0261731740362116e-06, + "loss": 0.2615, + "step": 26405 + }, + { + "epoch": 9.75, + "learning_rate": 2.026136223672866e-06, + "loss": 0.1907, + "step": 26406 + }, + { + "epoch": 9.75, + "learning_rate": 2.026099273309521e-06, + "loss": 0.5545, + "step": 26407 + }, + { + "epoch": 9.75, + "learning_rate": 2.0260623229461755e-06, + "loss": 0.3472, + "step": 26408 + }, + { + "epoch": 9.75, + "learning_rate": 2.0260253725828305e-06, + "loss": 0.3853, + "step": 26409 + }, + { + "epoch": 9.75, + "learning_rate": 2.025988422219485e-06, + "loss": 0.2432, + "step": 26410 + }, + { + "epoch": 9.75, + "learning_rate": 2.0259514718561402e-06, + "loss": 0.355, + "step": 26411 + }, + { + "epoch": 9.75, + "learning_rate": 2.0259145214927945e-06, + "loss": 0.2989, + "step": 26412 + }, + { + "epoch": 9.75, + "learning_rate": 2.0258775711294495e-06, + "loss": 0.472, + "step": 26413 + }, + { + "epoch": 9.75, + "learning_rate": 2.025840620766104e-06, + "loss": 0.3545, + "step": 26414 + }, + { + "epoch": 9.75, + "learning_rate": 2.025803670402759e-06, + "loss": 0.3305, + "step": 26415 + }, + { + "epoch": 9.75, + "learning_rate": 2.025766720039414e-06, + "loss": 0.3424, + "step": 26416 + }, + { + "epoch": 9.76, + "learning_rate": 2.0257297696760685e-06, + "loss": 0.3945, + "step": 26417 + }, + { + "epoch": 9.76, + "learning_rate": 2.025692819312723e-06, + "loss": 0.4061, + "step": 26418 + }, + { + "epoch": 9.76, + "learning_rate": 2.0256558689493778e-06, + "loss": 0.3103, + "step": 26419 + }, + { + "epoch": 9.76, + "learning_rate": 2.025618918586033e-06, + "loss": 0.4765, + "step": 26420 + }, + { + "epoch": 9.76, + "learning_rate": 2.0255819682226875e-06, + "loss": 0.3512, + "step": 26421 + }, + { + "epoch": 9.76, + "learning_rate": 2.0255450178593425e-06, + "loss": 0.4101, + "step": 26422 + }, + { + "epoch": 9.76, + "learning_rate": 2.025508067495997e-06, + "loss": 0.4388, + "step": 26423 + }, + { + "epoch": 9.76, + "learning_rate": 2.0254711171326518e-06, + "loss": 0.2996, + "step": 26424 + }, + { + "epoch": 9.76, + "learning_rate": 2.0254341667693064e-06, + "loss": 0.4859, + "step": 26425 + }, + { + "epoch": 9.76, + "learning_rate": 2.0253972164059615e-06, + "loss": 0.2296, + "step": 26426 + }, + { + "epoch": 9.76, + "learning_rate": 2.025360266042616e-06, + "loss": 0.2764, + "step": 26427 + }, + { + "epoch": 9.76, + "learning_rate": 2.0253233156792708e-06, + "loss": 0.3744, + "step": 26428 + }, + { + "epoch": 9.76, + "learning_rate": 2.025286365315926e-06, + "loss": 0.3404, + "step": 26429 + }, + { + "epoch": 9.76, + "learning_rate": 2.0252494149525805e-06, + "loss": 0.4146, + "step": 26430 + }, + { + "epoch": 9.76, + "learning_rate": 2.025212464589235e-06, + "loss": 0.2848, + "step": 26431 + }, + { + "epoch": 9.76, + "learning_rate": 2.0251755142258897e-06, + "loss": 0.5221, + "step": 26432 + }, + { + "epoch": 9.76, + "learning_rate": 2.0251385638625448e-06, + "loss": 0.3031, + "step": 26433 + }, + { + "epoch": 9.76, + "learning_rate": 2.0251016134991994e-06, + "loss": 0.4408, + "step": 26434 + }, + { + "epoch": 9.76, + "learning_rate": 2.0250646631358545e-06, + "loss": 0.3025, + "step": 26435 + }, + { + "epoch": 9.76, + "learning_rate": 2.025027712772509e-06, + "loss": 0.2984, + "step": 26436 + }, + { + "epoch": 9.76, + "learning_rate": 2.0249907624091638e-06, + "loss": 0.3817, + "step": 26437 + }, + { + "epoch": 9.76, + "learning_rate": 2.0249538120458184e-06, + "loss": 0.351, + "step": 26438 + }, + { + "epoch": 9.76, + "learning_rate": 2.024916861682473e-06, + "loss": 0.3968, + "step": 26439 + }, + { + "epoch": 9.76, + "learning_rate": 2.024879911319128e-06, + "loss": 0.358, + "step": 26440 + }, + { + "epoch": 9.76, + "learning_rate": 2.0248429609557827e-06, + "loss": 0.2759, + "step": 26441 + }, + { + "epoch": 9.76, + "learning_rate": 2.0248060105924378e-06, + "loss": 0.3375, + "step": 26442 + }, + { + "epoch": 9.76, + "learning_rate": 2.0247690602290924e-06, + "loss": 0.3418, + "step": 26443 + }, + { + "epoch": 9.77, + "learning_rate": 2.024732109865747e-06, + "loss": 0.3366, + "step": 26444 + }, + { + "epoch": 9.77, + "learning_rate": 2.0246951595024017e-06, + "loss": 0.2685, + "step": 26445 + }, + { + "epoch": 9.77, + "learning_rate": 2.0246582091390568e-06, + "loss": 0.277, + "step": 26446 + }, + { + "epoch": 9.77, + "learning_rate": 2.0246212587757114e-06, + "loss": 0.4133, + "step": 26447 + }, + { + "epoch": 9.77, + "learning_rate": 2.024584308412366e-06, + "loss": 0.3833, + "step": 26448 + }, + { + "epoch": 9.77, + "learning_rate": 2.024547358049021e-06, + "loss": 0.2798, + "step": 26449 + }, + { + "epoch": 9.77, + "learning_rate": 2.0245104076856757e-06, + "loss": 0.4744, + "step": 26450 + }, + { + "epoch": 9.77, + "learning_rate": 2.0244734573223304e-06, + "loss": 0.2318, + "step": 26451 + }, + { + "epoch": 9.77, + "learning_rate": 2.024436506958985e-06, + "loss": 0.3337, + "step": 26452 + }, + { + "epoch": 9.77, + "learning_rate": 2.02439955659564e-06, + "loss": 0.2875, + "step": 26453 + }, + { + "epoch": 9.77, + "learning_rate": 2.0243626062322947e-06, + "loss": 0.4405, + "step": 26454 + }, + { + "epoch": 9.77, + "learning_rate": 2.0243256558689497e-06, + "loss": 0.4042, + "step": 26455 + }, + { + "epoch": 9.77, + "learning_rate": 2.0242887055056044e-06, + "loss": 0.3558, + "step": 26456 + }, + { + "epoch": 9.77, + "learning_rate": 2.0242517551422586e-06, + "loss": 0.2487, + "step": 26457 + }, + { + "epoch": 9.77, + "learning_rate": 2.0242148047789137e-06, + "loss": 0.3184, + "step": 26458 + }, + { + "epoch": 9.77, + "learning_rate": 2.0241778544155683e-06, + "loss": 0.3865, + "step": 26459 + }, + { + "epoch": 9.77, + "learning_rate": 2.0241409040522234e-06, + "loss": 0.4202, + "step": 26460 + }, + { + "epoch": 9.77, + "learning_rate": 2.024103953688878e-06, + "loss": 0.3561, + "step": 26461 + }, + { + "epoch": 9.77, + "learning_rate": 2.024067003325533e-06, + "loss": 0.3793, + "step": 26462 + }, + { + "epoch": 9.77, + "learning_rate": 2.0240300529621873e-06, + "loss": 0.3292, + "step": 26463 + }, + { + "epoch": 9.77, + "learning_rate": 2.0239931025988423e-06, + "loss": 0.3201, + "step": 26464 + }, + { + "epoch": 9.77, + "learning_rate": 2.023956152235497e-06, + "loss": 0.2829, + "step": 26465 + }, + { + "epoch": 9.77, + "learning_rate": 2.023919201872152e-06, + "loss": 0.4766, + "step": 26466 + }, + { + "epoch": 9.77, + "learning_rate": 2.0238822515088067e-06, + "loss": 0.3063, + "step": 26467 + }, + { + "epoch": 9.77, + "learning_rate": 2.0238453011454613e-06, + "loss": 0.2654, + "step": 26468 + }, + { + "epoch": 9.77, + "learning_rate": 2.023808350782116e-06, + "loss": 0.2873, + "step": 26469 + }, + { + "epoch": 9.77, + "learning_rate": 2.0237714004187706e-06, + "loss": 0.3625, + "step": 26470 + }, + { + "epoch": 9.78, + "learning_rate": 2.0237344500554256e-06, + "loss": 0.3517, + "step": 26471 + }, + { + "epoch": 9.78, + "learning_rate": 2.0236974996920803e-06, + "loss": 0.3462, + "step": 26472 + }, + { + "epoch": 9.78, + "learning_rate": 2.0236605493287353e-06, + "loss": 0.2908, + "step": 26473 + }, + { + "epoch": 9.78, + "learning_rate": 2.02362359896539e-06, + "loss": 0.3678, + "step": 26474 + }, + { + "epoch": 9.78, + "learning_rate": 2.0235866486020446e-06, + "loss": 0.3726, + "step": 26475 + }, + { + "epoch": 9.78, + "learning_rate": 2.0235496982386992e-06, + "loss": 0.4039, + "step": 26476 + }, + { + "epoch": 9.78, + "learning_rate": 2.023512747875354e-06, + "loss": 0.2748, + "step": 26477 + }, + { + "epoch": 9.78, + "learning_rate": 2.023475797512009e-06, + "loss": 0.3489, + "step": 26478 + }, + { + "epoch": 9.78, + "learning_rate": 2.0234388471486636e-06, + "loss": 0.2447, + "step": 26479 + }, + { + "epoch": 9.78, + "learning_rate": 2.0234018967853186e-06, + "loss": 0.2131, + "step": 26480 + }, + { + "epoch": 9.78, + "learning_rate": 2.0233649464219733e-06, + "loss": 0.4667, + "step": 26481 + }, + { + "epoch": 9.78, + "learning_rate": 2.023327996058628e-06, + "loss": 0.2929, + "step": 26482 + }, + { + "epoch": 9.78, + "learning_rate": 2.0232910456952825e-06, + "loss": 0.349, + "step": 26483 + }, + { + "epoch": 9.78, + "learning_rate": 2.0232540953319376e-06, + "loss": 0.3344, + "step": 26484 + }, + { + "epoch": 9.78, + "learning_rate": 2.0232171449685922e-06, + "loss": 0.5866, + "step": 26485 + }, + { + "epoch": 9.78, + "learning_rate": 2.0231801946052473e-06, + "loss": 0.46, + "step": 26486 + }, + { + "epoch": 9.78, + "learning_rate": 2.023143244241902e-06, + "loss": 0.2779, + "step": 26487 + }, + { + "epoch": 9.78, + "learning_rate": 2.0231062938785566e-06, + "loss": 0.3773, + "step": 26488 + }, + { + "epoch": 9.78, + "learning_rate": 2.023069343515211e-06, + "loss": 0.2647, + "step": 26489 + }, + { + "epoch": 9.78, + "learning_rate": 2.023032393151866e-06, + "loss": 0.3581, + "step": 26490 + }, + { + "epoch": 9.78, + "learning_rate": 2.022995442788521e-06, + "loss": 0.3268, + "step": 26491 + }, + { + "epoch": 9.78, + "learning_rate": 2.0229584924251755e-06, + "loss": 0.3193, + "step": 26492 + }, + { + "epoch": 9.78, + "learning_rate": 2.0229215420618306e-06, + "loss": 0.4388, + "step": 26493 + }, + { + "epoch": 9.78, + "learning_rate": 2.0228845916984852e-06, + "loss": 0.3312, + "step": 26494 + }, + { + "epoch": 9.78, + "learning_rate": 2.02284764133514e-06, + "loss": 0.324, + "step": 26495 + }, + { + "epoch": 9.78, + "learning_rate": 2.0228106909717945e-06, + "loss": 0.4635, + "step": 26496 + }, + { + "epoch": 9.78, + "learning_rate": 2.022773740608449e-06, + "loss": 0.3156, + "step": 26497 + }, + { + "epoch": 9.79, + "learning_rate": 2.022736790245104e-06, + "loss": 0.3165, + "step": 26498 + }, + { + "epoch": 9.79, + "learning_rate": 2.022699839881759e-06, + "loss": 0.4466, + "step": 26499 + }, + { + "epoch": 9.79, + "learning_rate": 2.022662889518414e-06, + "loss": 0.3695, + "step": 26500 + }, + { + "epoch": 9.79, + "learning_rate": 2.0226259391550685e-06, + "loss": 0.3259, + "step": 26501 + }, + { + "epoch": 9.79, + "learning_rate": 2.022588988791723e-06, + "loss": 0.2589, + "step": 26502 + }, + { + "epoch": 9.79, + "learning_rate": 2.022552038428378e-06, + "loss": 0.3687, + "step": 26503 + }, + { + "epoch": 9.79, + "learning_rate": 2.022515088065033e-06, + "loss": 0.3327, + "step": 26504 + }, + { + "epoch": 9.79, + "learning_rate": 2.0224781377016875e-06, + "loss": 0.3435, + "step": 26505 + }, + { + "epoch": 9.79, + "learning_rate": 2.022441187338342e-06, + "loss": 0.3598, + "step": 26506 + }, + { + "epoch": 9.79, + "learning_rate": 2.022404236974997e-06, + "loss": 0.4067, + "step": 26507 + }, + { + "epoch": 9.79, + "learning_rate": 2.0223672866116514e-06, + "loss": 0.3135, + "step": 26508 + }, + { + "epoch": 9.79, + "learning_rate": 2.0223303362483065e-06, + "loss": 0.3751, + "step": 26509 + }, + { + "epoch": 9.79, + "learning_rate": 2.022293385884961e-06, + "loss": 0.3828, + "step": 26510 + }, + { + "epoch": 9.79, + "learning_rate": 2.022256435521616e-06, + "loss": 0.2861, + "step": 26511 + }, + { + "epoch": 9.79, + "learning_rate": 2.022219485158271e-06, + "loss": 0.3419, + "step": 26512 + }, + { + "epoch": 9.79, + "learning_rate": 2.022182534794926e-06, + "loss": 0.2938, + "step": 26513 + }, + { + "epoch": 9.79, + "learning_rate": 2.02214558443158e-06, + "loss": 0.4004, + "step": 26514 + }, + { + "epoch": 9.79, + "learning_rate": 2.022108634068235e-06, + "loss": 0.3519, + "step": 26515 + }, + { + "epoch": 9.79, + "learning_rate": 2.0220716837048898e-06, + "loss": 0.413, + "step": 26516 + }, + { + "epoch": 9.79, + "learning_rate": 2.0220347333415444e-06, + "loss": 0.2808, + "step": 26517 + }, + { + "epoch": 9.79, + "learning_rate": 2.0219977829781995e-06, + "loss": 0.3424, + "step": 26518 + }, + { + "epoch": 9.79, + "learning_rate": 2.021960832614854e-06, + "loss": 0.2963, + "step": 26519 + }, + { + "epoch": 9.79, + "learning_rate": 2.0219238822515087e-06, + "loss": 0.3368, + "step": 26520 + }, + { + "epoch": 9.79, + "learning_rate": 2.0218869318881634e-06, + "loss": 0.4681, + "step": 26521 + }, + { + "epoch": 9.79, + "learning_rate": 2.0218499815248184e-06, + "loss": 0.2564, + "step": 26522 + }, + { + "epoch": 9.79, + "learning_rate": 2.021813031161473e-06, + "loss": 0.3144, + "step": 26523 + }, + { + "epoch": 9.79, + "learning_rate": 2.021776080798128e-06, + "loss": 0.4316, + "step": 26524 + }, + { + "epoch": 9.8, + "learning_rate": 2.0217391304347828e-06, + "loss": 0.4893, + "step": 26525 + }, + { + "epoch": 9.8, + "learning_rate": 2.0217021800714374e-06, + "loss": 0.3249, + "step": 26526 + }, + { + "epoch": 9.8, + "learning_rate": 2.021665229708092e-06, + "loss": 0.3941, + "step": 26527 + }, + { + "epoch": 9.8, + "learning_rate": 2.0216282793447467e-06, + "loss": 0.3162, + "step": 26528 + }, + { + "epoch": 9.8, + "learning_rate": 2.0215913289814017e-06, + "loss": 0.425, + "step": 26529 + }, + { + "epoch": 9.8, + "learning_rate": 2.0215543786180564e-06, + "loss": 0.3601, + "step": 26530 + }, + { + "epoch": 9.8, + "learning_rate": 2.0215174282547114e-06, + "loss": 0.3672, + "step": 26531 + }, + { + "epoch": 9.8, + "learning_rate": 2.021480477891366e-06, + "loss": 0.34, + "step": 26532 + }, + { + "epoch": 9.8, + "learning_rate": 2.0214435275280207e-06, + "loss": 0.3455, + "step": 26533 + }, + { + "epoch": 9.8, + "learning_rate": 2.0214065771646753e-06, + "loss": 0.3942, + "step": 26534 + }, + { + "epoch": 9.8, + "learning_rate": 2.0213696268013304e-06, + "loss": 0.5674, + "step": 26535 + }, + { + "epoch": 9.8, + "learning_rate": 2.021332676437985e-06, + "loss": 0.4204, + "step": 26536 + }, + { + "epoch": 9.8, + "learning_rate": 2.0212957260746397e-06, + "loss": 0.3588, + "step": 26537 + }, + { + "epoch": 9.8, + "learning_rate": 2.0212587757112947e-06, + "loss": 0.4154, + "step": 26538 + }, + { + "epoch": 9.8, + "learning_rate": 2.0212218253479494e-06, + "loss": 0.4479, + "step": 26539 + }, + { + "epoch": 9.8, + "learning_rate": 2.021184874984604e-06, + "loss": 0.3988, + "step": 26540 + }, + { + "epoch": 9.8, + "learning_rate": 2.0211479246212586e-06, + "loss": 0.3997, + "step": 26541 + }, + { + "epoch": 9.8, + "learning_rate": 2.0211109742579137e-06, + "loss": 0.3017, + "step": 26542 + }, + { + "epoch": 9.8, + "learning_rate": 2.0210740238945683e-06, + "loss": 0.4279, + "step": 26543 + }, + { + "epoch": 9.8, + "learning_rate": 2.0210370735312234e-06, + "loss": 0.2844, + "step": 26544 + }, + { + "epoch": 9.8, + "learning_rate": 2.021000123167878e-06, + "loss": 0.2848, + "step": 26545 + }, + { + "epoch": 9.8, + "learning_rate": 2.0209631728045327e-06, + "loss": 0.3982, + "step": 26546 + }, + { + "epoch": 9.8, + "learning_rate": 2.0209262224411873e-06, + "loss": 0.2755, + "step": 26547 + }, + { + "epoch": 9.8, + "learning_rate": 2.020889272077842e-06, + "loss": 0.2803, + "step": 26548 + }, + { + "epoch": 9.8, + "learning_rate": 2.020852321714497e-06, + "loss": 0.3741, + "step": 26549 + }, + { + "epoch": 9.8, + "learning_rate": 2.0208153713511516e-06, + "loss": 0.4073, + "step": 26550 + }, + { + "epoch": 9.8, + "learning_rate": 2.0207784209878067e-06, + "loss": 0.4911, + "step": 26551 + }, + { + "epoch": 9.81, + "learning_rate": 2.020741470624461e-06, + "loss": 0.44, + "step": 26552 + }, + { + "epoch": 9.81, + "learning_rate": 2.020704520261116e-06, + "loss": 0.3262, + "step": 26553 + }, + { + "epoch": 9.81, + "learning_rate": 2.0206675698977706e-06, + "loss": 0.4955, + "step": 26554 + }, + { + "epoch": 9.81, + "learning_rate": 2.0206306195344257e-06, + "loss": 0.3586, + "step": 26555 + }, + { + "epoch": 9.81, + "learning_rate": 2.0205936691710803e-06, + "loss": 0.1865, + "step": 26556 + }, + { + "epoch": 9.81, + "learning_rate": 2.020556718807735e-06, + "loss": 0.2866, + "step": 26557 + }, + { + "epoch": 9.81, + "learning_rate": 2.0205197684443896e-06, + "loss": 0.2956, + "step": 26558 + }, + { + "epoch": 9.81, + "learning_rate": 2.0204828180810442e-06, + "loss": 0.423, + "step": 26559 + }, + { + "epoch": 9.81, + "learning_rate": 2.0204458677176993e-06, + "loss": 0.3574, + "step": 26560 + }, + { + "epoch": 9.81, + "learning_rate": 2.020408917354354e-06, + "loss": 0.2929, + "step": 26561 + }, + { + "epoch": 9.81, + "learning_rate": 2.020371966991009e-06, + "loss": 0.3837, + "step": 26562 + }, + { + "epoch": 9.81, + "learning_rate": 2.0203350166276636e-06, + "loss": 0.53, + "step": 26563 + }, + { + "epoch": 9.81, + "learning_rate": 2.0202980662643183e-06, + "loss": 0.276, + "step": 26564 + }, + { + "epoch": 9.81, + "learning_rate": 2.020261115900973e-06, + "loss": 0.3087, + "step": 26565 + }, + { + "epoch": 9.81, + "learning_rate": 2.0202241655376275e-06, + "loss": 0.3586, + "step": 26566 + }, + { + "epoch": 9.81, + "learning_rate": 2.0201872151742826e-06, + "loss": 0.5081, + "step": 26567 + }, + { + "epoch": 9.81, + "learning_rate": 2.0201502648109372e-06, + "loss": 0.4037, + "step": 26568 + }, + { + "epoch": 9.81, + "learning_rate": 2.0201133144475923e-06, + "loss": 0.3027, + "step": 26569 + }, + { + "epoch": 9.81, + "learning_rate": 2.020076364084247e-06, + "loss": 0.2944, + "step": 26570 + }, + { + "epoch": 9.81, + "learning_rate": 2.0200394137209016e-06, + "loss": 0.4495, + "step": 26571 + }, + { + "epoch": 9.81, + "learning_rate": 2.020002463357556e-06, + "loss": 0.4452, + "step": 26572 + }, + { + "epoch": 9.81, + "learning_rate": 2.0199655129942112e-06, + "loss": 0.2449, + "step": 26573 + }, + { + "epoch": 9.81, + "learning_rate": 2.019928562630866e-06, + "loss": 0.3782, + "step": 26574 + }, + { + "epoch": 9.81, + "learning_rate": 2.019891612267521e-06, + "loss": 0.2772, + "step": 26575 + }, + { + "epoch": 9.81, + "learning_rate": 2.0198546619041756e-06, + "loss": 0.4649, + "step": 26576 + }, + { + "epoch": 9.81, + "learning_rate": 2.0198177115408302e-06, + "loss": 0.3674, + "step": 26577 + }, + { + "epoch": 9.81, + "learning_rate": 2.019780761177485e-06, + "loss": 0.3953, + "step": 26578 + }, + { + "epoch": 9.81, + "learning_rate": 2.0197438108141395e-06, + "loss": 0.4883, + "step": 26579 + }, + { + "epoch": 9.82, + "learning_rate": 2.0197068604507946e-06, + "loss": 0.1924, + "step": 26580 + }, + { + "epoch": 9.82, + "learning_rate": 2.019669910087449e-06, + "loss": 0.2685, + "step": 26581 + }, + { + "epoch": 9.82, + "learning_rate": 2.0196329597241042e-06, + "loss": 0.3256, + "step": 26582 + }, + { + "epoch": 9.82, + "learning_rate": 2.019596009360759e-06, + "loss": 0.3036, + "step": 26583 + }, + { + "epoch": 9.82, + "learning_rate": 2.0195590589974135e-06, + "loss": 0.2917, + "step": 26584 + }, + { + "epoch": 9.82, + "learning_rate": 2.019522108634068e-06, + "loss": 0.3987, + "step": 26585 + }, + { + "epoch": 9.82, + "learning_rate": 2.019485158270723e-06, + "loss": 0.3264, + "step": 26586 + }, + { + "epoch": 9.82, + "learning_rate": 2.019448207907378e-06, + "loss": 0.4234, + "step": 26587 + }, + { + "epoch": 9.82, + "learning_rate": 2.0194112575440325e-06, + "loss": 0.3653, + "step": 26588 + }, + { + "epoch": 9.82, + "learning_rate": 2.0193743071806875e-06, + "loss": 0.3923, + "step": 26589 + }, + { + "epoch": 9.82, + "learning_rate": 2.019337356817342e-06, + "loss": 0.4778, + "step": 26590 + }, + { + "epoch": 9.82, + "learning_rate": 2.019300406453997e-06, + "loss": 0.4043, + "step": 26591 + }, + { + "epoch": 9.82, + "learning_rate": 2.0192634560906515e-06, + "loss": 0.3022, + "step": 26592 + }, + { + "epoch": 9.82, + "learning_rate": 2.0192265057273065e-06, + "loss": 0.3849, + "step": 26593 + }, + { + "epoch": 9.82, + "learning_rate": 2.019189555363961e-06, + "loss": 0.4945, + "step": 26594 + }, + { + "epoch": 9.82, + "learning_rate": 2.0191526050006162e-06, + "loss": 0.44, + "step": 26595 + }, + { + "epoch": 9.82, + "learning_rate": 2.019115654637271e-06, + "loss": 0.5176, + "step": 26596 + }, + { + "epoch": 9.82, + "learning_rate": 2.019078704273925e-06, + "loss": 0.2857, + "step": 26597 + }, + { + "epoch": 9.82, + "learning_rate": 2.01904175391058e-06, + "loss": 0.4113, + "step": 26598 + }, + { + "epoch": 9.82, + "learning_rate": 2.0190048035472348e-06, + "loss": 0.3881, + "step": 26599 + }, + { + "epoch": 9.82, + "learning_rate": 2.01896785318389e-06, + "loss": 0.4203, + "step": 26600 + }, + { + "epoch": 9.82, + "learning_rate": 2.0189309028205445e-06, + "loss": 0.3436, + "step": 26601 + }, + { + "epoch": 9.82, + "learning_rate": 2.0188939524571995e-06, + "loss": 0.2999, + "step": 26602 + }, + { + "epoch": 9.82, + "learning_rate": 2.0188570020938537e-06, + "loss": 0.3993, + "step": 26603 + }, + { + "epoch": 9.82, + "learning_rate": 2.018820051730509e-06, + "loss": 0.3479, + "step": 26604 + }, + { + "epoch": 9.82, + "learning_rate": 2.0187831013671634e-06, + "loss": 0.3532, + "step": 26605 + }, + { + "epoch": 9.82, + "learning_rate": 2.018746151003818e-06, + "loss": 0.3498, + "step": 26606 + }, + { + "epoch": 9.83, + "learning_rate": 2.018709200640473e-06, + "loss": 0.3546, + "step": 26607 + }, + { + "epoch": 9.83, + "learning_rate": 2.0186722502771278e-06, + "loss": 0.2814, + "step": 26608 + }, + { + "epoch": 9.83, + "learning_rate": 2.0186352999137824e-06, + "loss": 0.4171, + "step": 26609 + }, + { + "epoch": 9.83, + "learning_rate": 2.018598349550437e-06, + "loss": 0.3512, + "step": 26610 + }, + { + "epoch": 9.83, + "learning_rate": 2.018561399187092e-06, + "loss": 0.4315, + "step": 26611 + }, + { + "epoch": 9.83, + "learning_rate": 2.0185244488237467e-06, + "loss": 0.2863, + "step": 26612 + }, + { + "epoch": 9.83, + "learning_rate": 2.0184874984604018e-06, + "loss": 0.2479, + "step": 26613 + }, + { + "epoch": 9.83, + "learning_rate": 2.0184505480970564e-06, + "loss": 0.3071, + "step": 26614 + }, + { + "epoch": 9.83, + "learning_rate": 2.018413597733711e-06, + "loss": 0.421, + "step": 26615 + }, + { + "epoch": 9.83, + "learning_rate": 2.0183766473703657e-06, + "loss": 0.2566, + "step": 26616 + }, + { + "epoch": 9.83, + "learning_rate": 2.0183396970070203e-06, + "loss": 0.2559, + "step": 26617 + }, + { + "epoch": 9.83, + "learning_rate": 2.0183027466436754e-06, + "loss": 0.493, + "step": 26618 + }, + { + "epoch": 9.83, + "learning_rate": 2.01826579628033e-06, + "loss": 0.3198, + "step": 26619 + }, + { + "epoch": 9.83, + "learning_rate": 2.018228845916985e-06, + "loss": 0.3434, + "step": 26620 + }, + { + "epoch": 9.83, + "learning_rate": 2.0181918955536397e-06, + "loss": 0.4291, + "step": 26621 + }, + { + "epoch": 9.83, + "learning_rate": 2.0181549451902944e-06, + "loss": 0.3336, + "step": 26622 + }, + { + "epoch": 9.83, + "learning_rate": 2.018117994826949e-06, + "loss": 0.3912, + "step": 26623 + }, + { + "epoch": 9.83, + "learning_rate": 2.018081044463604e-06, + "loss": 0.3713, + "step": 26624 + }, + { + "epoch": 9.83, + "learning_rate": 2.0180440941002587e-06, + "loss": 0.4276, + "step": 26625 + }, + { + "epoch": 9.83, + "learning_rate": 2.0180071437369133e-06, + "loss": 0.3825, + "step": 26626 + }, + { + "epoch": 9.83, + "learning_rate": 2.0179701933735684e-06, + "loss": 0.3586, + "step": 26627 + }, + { + "epoch": 9.83, + "learning_rate": 2.017933243010223e-06, + "loss": 0.4143, + "step": 26628 + }, + { + "epoch": 9.83, + "learning_rate": 2.0178962926468777e-06, + "loss": 0.4345, + "step": 26629 + }, + { + "epoch": 9.83, + "learning_rate": 2.0178593422835323e-06, + "loss": 0.2994, + "step": 26630 + }, + { + "epoch": 9.83, + "learning_rate": 2.0178223919201874e-06, + "loss": 0.298, + "step": 26631 + }, + { + "epoch": 9.83, + "learning_rate": 2.017785441556842e-06, + "loss": 0.4647, + "step": 26632 + }, + { + "epoch": 9.83, + "learning_rate": 2.017748491193497e-06, + "loss": 0.3594, + "step": 26633 + }, + { + "epoch": 9.84, + "learning_rate": 2.0177115408301517e-06, + "loss": 0.3317, + "step": 26634 + }, + { + "epoch": 9.84, + "learning_rate": 2.0176745904668063e-06, + "loss": 0.371, + "step": 26635 + }, + { + "epoch": 9.84, + "learning_rate": 2.017637640103461e-06, + "loss": 0.2199, + "step": 26636 + }, + { + "epoch": 9.84, + "learning_rate": 2.0176006897401156e-06, + "loss": 0.3377, + "step": 26637 + }, + { + "epoch": 9.84, + "learning_rate": 2.0175637393767707e-06, + "loss": 0.4768, + "step": 26638 + }, + { + "epoch": 9.84, + "learning_rate": 2.0175267890134253e-06, + "loss": 0.4266, + "step": 26639 + }, + { + "epoch": 9.84, + "learning_rate": 2.0174898386500804e-06, + "loss": 0.3154, + "step": 26640 + }, + { + "epoch": 9.84, + "learning_rate": 2.017452888286735e-06, + "loss": 0.3072, + "step": 26641 + }, + { + "epoch": 9.84, + "learning_rate": 2.0174159379233896e-06, + "loss": 0.282, + "step": 26642 + }, + { + "epoch": 9.84, + "learning_rate": 2.0173789875600443e-06, + "loss": 0.4103, + "step": 26643 + }, + { + "epoch": 9.84, + "learning_rate": 2.0173420371966993e-06, + "loss": 0.3891, + "step": 26644 + }, + { + "epoch": 9.84, + "learning_rate": 2.017305086833354e-06, + "loss": 0.395, + "step": 26645 + }, + { + "epoch": 9.84, + "learning_rate": 2.0172681364700086e-06, + "loss": 0.4, + "step": 26646 + }, + { + "epoch": 9.84, + "learning_rate": 2.0172311861066637e-06, + "loss": 0.3496, + "step": 26647 + }, + { + "epoch": 9.84, + "learning_rate": 2.017194235743318e-06, + "loss": 0.4662, + "step": 26648 + }, + { + "epoch": 9.84, + "learning_rate": 2.017157285379973e-06, + "loss": 0.392, + "step": 26649 + }, + { + "epoch": 9.84, + "learning_rate": 2.0171203350166276e-06, + "loss": 0.3109, + "step": 26650 + }, + { + "epoch": 9.84, + "learning_rate": 2.0170833846532826e-06, + "loss": 0.3974, + "step": 26651 + }, + { + "epoch": 9.84, + "learning_rate": 2.0170464342899373e-06, + "loss": 0.4178, + "step": 26652 + }, + { + "epoch": 9.84, + "learning_rate": 2.0170094839265923e-06, + "loss": 0.3278, + "step": 26653 + }, + { + "epoch": 9.84, + "learning_rate": 2.0169725335632465e-06, + "loss": 0.2695, + "step": 26654 + }, + { + "epoch": 9.84, + "learning_rate": 2.0169355831999016e-06, + "loss": 0.3707, + "step": 26655 + }, + { + "epoch": 9.84, + "learning_rate": 2.0168986328365562e-06, + "loss": 0.4172, + "step": 26656 + }, + { + "epoch": 9.84, + "learning_rate": 2.016861682473211e-06, + "loss": 0.4259, + "step": 26657 + }, + { + "epoch": 9.84, + "learning_rate": 2.016824732109866e-06, + "loss": 0.3625, + "step": 26658 + }, + { + "epoch": 9.84, + "learning_rate": 2.0167877817465206e-06, + "loss": 0.3132, + "step": 26659 + }, + { + "epoch": 9.84, + "learning_rate": 2.016750831383175e-06, + "loss": 0.291, + "step": 26660 + }, + { + "epoch": 9.85, + "learning_rate": 2.01671388101983e-06, + "loss": 0.4319, + "step": 26661 + }, + { + "epoch": 9.85, + "learning_rate": 2.016676930656485e-06, + "loss": 0.3515, + "step": 26662 + }, + { + "epoch": 9.85, + "learning_rate": 2.0166399802931395e-06, + "loss": 0.28, + "step": 26663 + }, + { + "epoch": 9.85, + "learning_rate": 2.0166030299297946e-06, + "loss": 0.3916, + "step": 26664 + }, + { + "epoch": 9.85, + "learning_rate": 2.0165660795664492e-06, + "loss": 0.355, + "step": 26665 + }, + { + "epoch": 9.85, + "learning_rate": 2.016529129203104e-06, + "loss": 0.3365, + "step": 26666 + }, + { + "epoch": 9.85, + "learning_rate": 2.0164921788397585e-06, + "loss": 0.3783, + "step": 26667 + }, + { + "epoch": 9.85, + "learning_rate": 2.016455228476413e-06, + "loss": 0.2979, + "step": 26668 + }, + { + "epoch": 9.85, + "learning_rate": 2.016418278113068e-06, + "loss": 0.3491, + "step": 26669 + }, + { + "epoch": 9.85, + "learning_rate": 2.016381327749723e-06, + "loss": 0.1762, + "step": 26670 + }, + { + "epoch": 9.85, + "learning_rate": 2.016344377386378e-06, + "loss": 0.3214, + "step": 26671 + }, + { + "epoch": 9.85, + "learning_rate": 2.0163074270230325e-06, + "loss": 0.4536, + "step": 26672 + }, + { + "epoch": 9.85, + "learning_rate": 2.016270476659687e-06, + "loss": 0.5404, + "step": 26673 + }, + { + "epoch": 9.85, + "learning_rate": 2.016233526296342e-06, + "loss": 0.4103, + "step": 26674 + }, + { + "epoch": 9.85, + "learning_rate": 2.016196575932997e-06, + "loss": 0.5534, + "step": 26675 + }, + { + "epoch": 9.85, + "learning_rate": 2.0161596255696515e-06, + "loss": 0.4904, + "step": 26676 + }, + { + "epoch": 9.85, + "learning_rate": 2.016122675206306e-06, + "loss": 0.3633, + "step": 26677 + }, + { + "epoch": 9.85, + "learning_rate": 2.016085724842961e-06, + "loss": 0.2611, + "step": 26678 + }, + { + "epoch": 9.85, + "learning_rate": 2.016048774479616e-06, + "loss": 0.2581, + "step": 26679 + }, + { + "epoch": 9.85, + "learning_rate": 2.0160118241162705e-06, + "loss": 0.4349, + "step": 26680 + }, + { + "epoch": 9.85, + "learning_rate": 2.015974873752925e-06, + "loss": 0.4227, + "step": 26681 + }, + { + "epoch": 9.85, + "learning_rate": 2.01593792338958e-06, + "loss": 0.4543, + "step": 26682 + }, + { + "epoch": 9.85, + "learning_rate": 2.015900973026235e-06, + "loss": 0.34, + "step": 26683 + }, + { + "epoch": 9.85, + "learning_rate": 2.01586402266289e-06, + "loss": 0.3602, + "step": 26684 + }, + { + "epoch": 9.85, + "learning_rate": 2.0158270722995445e-06, + "loss": 0.2698, + "step": 26685 + }, + { + "epoch": 9.85, + "learning_rate": 2.015790121936199e-06, + "loss": 0.2833, + "step": 26686 + }, + { + "epoch": 9.85, + "learning_rate": 2.0157531715728538e-06, + "loss": 0.4182, + "step": 26687 + }, + { + "epoch": 9.86, + "learning_rate": 2.0157162212095084e-06, + "loss": 0.4105, + "step": 26688 + }, + { + "epoch": 9.86, + "learning_rate": 2.0156792708461635e-06, + "loss": 0.4079, + "step": 26689 + }, + { + "epoch": 9.86, + "learning_rate": 2.015642320482818e-06, + "loss": 0.3317, + "step": 26690 + }, + { + "epoch": 9.86, + "learning_rate": 2.015605370119473e-06, + "loss": 0.3939, + "step": 26691 + }, + { + "epoch": 9.86, + "learning_rate": 2.015568419756128e-06, + "loss": 0.355, + "step": 26692 + }, + { + "epoch": 9.86, + "learning_rate": 2.0155314693927824e-06, + "loss": 0.3551, + "step": 26693 + }, + { + "epoch": 9.86, + "learning_rate": 2.015494519029437e-06, + "loss": 0.3741, + "step": 26694 + }, + { + "epoch": 9.86, + "learning_rate": 2.015457568666092e-06, + "loss": 0.2766, + "step": 26695 + }, + { + "epoch": 9.86, + "learning_rate": 2.0154206183027468e-06, + "loss": 0.4657, + "step": 26696 + }, + { + "epoch": 9.86, + "learning_rate": 2.0153836679394014e-06, + "loss": 0.5361, + "step": 26697 + }, + { + "epoch": 9.86, + "learning_rate": 2.0153467175760565e-06, + "loss": 0.4152, + "step": 26698 + }, + { + "epoch": 9.86, + "learning_rate": 2.0153097672127107e-06, + "loss": 0.2644, + "step": 26699 + }, + { + "epoch": 9.86, + "learning_rate": 2.0152728168493657e-06, + "loss": 0.3057, + "step": 26700 + }, + { + "epoch": 9.86, + "learning_rate": 2.0152358664860204e-06, + "loss": 0.3876, + "step": 26701 + }, + { + "epoch": 9.86, + "learning_rate": 2.0151989161226754e-06, + "loss": 0.3181, + "step": 26702 + }, + { + "epoch": 9.86, + "learning_rate": 2.01516196575933e-06, + "loss": 0.3274, + "step": 26703 + }, + { + "epoch": 9.86, + "learning_rate": 2.0151250153959847e-06, + "loss": 0.5128, + "step": 26704 + }, + { + "epoch": 9.86, + "learning_rate": 2.0150880650326394e-06, + "loss": 0.2799, + "step": 26705 + }, + { + "epoch": 9.86, + "learning_rate": 2.015051114669294e-06, + "loss": 0.3165, + "step": 26706 + }, + { + "epoch": 9.86, + "learning_rate": 2.015014164305949e-06, + "loss": 0.5334, + "step": 26707 + }, + { + "epoch": 9.86, + "learning_rate": 2.0149772139426037e-06, + "loss": 0.3126, + "step": 26708 + }, + { + "epoch": 9.86, + "learning_rate": 2.0149402635792587e-06, + "loss": 0.3346, + "step": 26709 + }, + { + "epoch": 9.86, + "learning_rate": 2.0149033132159134e-06, + "loss": 0.3318, + "step": 26710 + }, + { + "epoch": 9.86, + "learning_rate": 2.014866362852568e-06, + "loss": 0.5884, + "step": 26711 + }, + { + "epoch": 9.86, + "learning_rate": 2.0148294124892227e-06, + "loss": 0.3777, + "step": 26712 + }, + { + "epoch": 9.86, + "learning_rate": 2.0147924621258777e-06, + "loss": 0.3572, + "step": 26713 + }, + { + "epoch": 9.86, + "learning_rate": 2.0147555117625323e-06, + "loss": 0.3306, + "step": 26714 + }, + { + "epoch": 9.87, + "learning_rate": 2.0147185613991874e-06, + "loss": 0.3228, + "step": 26715 + }, + { + "epoch": 9.87, + "learning_rate": 2.014681611035842e-06, + "loss": 0.3485, + "step": 26716 + }, + { + "epoch": 9.87, + "learning_rate": 2.0146446606724967e-06, + "loss": 0.4462, + "step": 26717 + }, + { + "epoch": 9.87, + "learning_rate": 2.0146077103091513e-06, + "loss": 0.3195, + "step": 26718 + }, + { + "epoch": 9.87, + "learning_rate": 2.014570759945806e-06, + "loss": 0.3503, + "step": 26719 + }, + { + "epoch": 9.87, + "learning_rate": 2.014533809582461e-06, + "loss": 0.3967, + "step": 26720 + }, + { + "epoch": 9.87, + "learning_rate": 2.0144968592191157e-06, + "loss": 0.3915, + "step": 26721 + }, + { + "epoch": 9.87, + "learning_rate": 2.0144599088557707e-06, + "loss": 0.3828, + "step": 26722 + }, + { + "epoch": 9.87, + "learning_rate": 2.0144229584924253e-06, + "loss": 0.4359, + "step": 26723 + }, + { + "epoch": 9.87, + "learning_rate": 2.01438600812908e-06, + "loss": 0.3824, + "step": 26724 + }, + { + "epoch": 9.87, + "learning_rate": 2.0143490577657346e-06, + "loss": 0.3779, + "step": 26725 + }, + { + "epoch": 9.87, + "learning_rate": 2.0143121074023893e-06, + "loss": 0.2511, + "step": 26726 + }, + { + "epoch": 9.87, + "learning_rate": 2.0142751570390443e-06, + "loss": 0.3749, + "step": 26727 + }, + { + "epoch": 9.87, + "learning_rate": 2.014238206675699e-06, + "loss": 0.4639, + "step": 26728 + }, + { + "epoch": 9.87, + "learning_rate": 2.014201256312354e-06, + "loss": 0.2861, + "step": 26729 + }, + { + "epoch": 9.87, + "learning_rate": 2.0141643059490087e-06, + "loss": 0.3183, + "step": 26730 + }, + { + "epoch": 9.87, + "learning_rate": 2.0141273555856633e-06, + "loss": 0.377, + "step": 26731 + }, + { + "epoch": 9.87, + "learning_rate": 2.014090405222318e-06, + "loss": 0.2966, + "step": 26732 + }, + { + "epoch": 9.87, + "learning_rate": 2.014053454858973e-06, + "loss": 0.2601, + "step": 26733 + }, + { + "epoch": 9.87, + "learning_rate": 2.0140165044956276e-06, + "loss": 0.31, + "step": 26734 + }, + { + "epoch": 9.87, + "learning_rate": 2.0139795541322823e-06, + "loss": 0.3282, + "step": 26735 + }, + { + "epoch": 9.87, + "learning_rate": 2.0139426037689373e-06, + "loss": 0.3293, + "step": 26736 + }, + { + "epoch": 9.87, + "learning_rate": 2.0139056534055915e-06, + "loss": 0.2643, + "step": 26737 + }, + { + "epoch": 9.87, + "learning_rate": 2.0138687030422466e-06, + "loss": 0.3925, + "step": 26738 + }, + { + "epoch": 9.87, + "learning_rate": 2.0138317526789012e-06, + "loss": 0.2732, + "step": 26739 + }, + { + "epoch": 9.87, + "learning_rate": 2.0137948023155563e-06, + "loss": 0.4016, + "step": 26740 + }, + { + "epoch": 9.87, + "learning_rate": 2.013757851952211e-06, + "loss": 0.4188, + "step": 26741 + }, + { + "epoch": 9.88, + "learning_rate": 2.013720901588866e-06, + "loss": 0.4302, + "step": 26742 + }, + { + "epoch": 9.88, + "learning_rate": 2.01368395122552e-06, + "loss": 0.378, + "step": 26743 + }, + { + "epoch": 9.88, + "learning_rate": 2.0136470008621753e-06, + "loss": 0.3777, + "step": 26744 + }, + { + "epoch": 9.88, + "learning_rate": 2.01361005049883e-06, + "loss": 0.3494, + "step": 26745 + }, + { + "epoch": 9.88, + "learning_rate": 2.0135731001354845e-06, + "loss": 0.363, + "step": 26746 + }, + { + "epoch": 9.88, + "learning_rate": 2.0135361497721396e-06, + "loss": 0.4323, + "step": 26747 + }, + { + "epoch": 9.88, + "learning_rate": 2.0134991994087942e-06, + "loss": 0.4133, + "step": 26748 + }, + { + "epoch": 9.88, + "learning_rate": 2.013462249045449e-06, + "loss": 0.366, + "step": 26749 + }, + { + "epoch": 9.88, + "learning_rate": 2.0134252986821035e-06, + "loss": 0.3088, + "step": 26750 + }, + { + "epoch": 9.88, + "learning_rate": 2.0133883483187586e-06, + "loss": 0.4083, + "step": 26751 + }, + { + "epoch": 9.88, + "learning_rate": 2.013351397955413e-06, + "loss": 0.3854, + "step": 26752 + }, + { + "epoch": 9.88, + "learning_rate": 2.0133144475920683e-06, + "loss": 0.3721, + "step": 26753 + }, + { + "epoch": 9.88, + "learning_rate": 2.013277497228723e-06, + "loss": 0.381, + "step": 26754 + }, + { + "epoch": 9.88, + "learning_rate": 2.0132405468653775e-06, + "loss": 0.2211, + "step": 26755 + }, + { + "epoch": 9.88, + "learning_rate": 2.013203596502032e-06, + "loss": 0.389, + "step": 26756 + }, + { + "epoch": 9.88, + "learning_rate": 2.013166646138687e-06, + "loss": 0.2973, + "step": 26757 + }, + { + "epoch": 9.88, + "learning_rate": 2.013129695775342e-06, + "loss": 0.297, + "step": 26758 + }, + { + "epoch": 9.88, + "learning_rate": 2.0130927454119965e-06, + "loss": 0.487, + "step": 26759 + }, + { + "epoch": 9.88, + "learning_rate": 2.0130557950486516e-06, + "loss": 0.4388, + "step": 26760 + }, + { + "epoch": 9.88, + "learning_rate": 2.013018844685306e-06, + "loss": 0.4317, + "step": 26761 + }, + { + "epoch": 9.88, + "learning_rate": 2.012981894321961e-06, + "loss": 0.3637, + "step": 26762 + }, + { + "epoch": 9.88, + "learning_rate": 2.0129449439586155e-06, + "loss": 0.4045, + "step": 26763 + }, + { + "epoch": 9.88, + "learning_rate": 2.0129079935952705e-06, + "loss": 0.295, + "step": 26764 + }, + { + "epoch": 9.88, + "learning_rate": 2.012871043231925e-06, + "loss": 0.275, + "step": 26765 + }, + { + "epoch": 9.88, + "learning_rate": 2.01283409286858e-06, + "loss": 0.3679, + "step": 26766 + }, + { + "epoch": 9.88, + "learning_rate": 2.012797142505235e-06, + "loss": 0.2992, + "step": 26767 + }, + { + "epoch": 9.88, + "learning_rate": 2.0127601921418895e-06, + "loss": 0.3386, + "step": 26768 + }, + { + "epoch": 9.89, + "learning_rate": 2.012723241778544e-06, + "loss": 0.3602, + "step": 26769 + }, + { + "epoch": 9.89, + "learning_rate": 2.0126862914151988e-06, + "loss": 0.4306, + "step": 26770 + }, + { + "epoch": 9.89, + "learning_rate": 2.012649341051854e-06, + "loss": 0.2962, + "step": 26771 + }, + { + "epoch": 9.89, + "learning_rate": 2.0126123906885085e-06, + "loss": 0.3584, + "step": 26772 + }, + { + "epoch": 9.89, + "learning_rate": 2.0125754403251635e-06, + "loss": 0.3017, + "step": 26773 + }, + { + "epoch": 9.89, + "learning_rate": 2.012538489961818e-06, + "loss": 0.2293, + "step": 26774 + }, + { + "epoch": 9.89, + "learning_rate": 2.012501539598473e-06, + "loss": 0.3787, + "step": 26775 + }, + { + "epoch": 9.89, + "learning_rate": 2.0124645892351274e-06, + "loss": 0.3734, + "step": 26776 + }, + { + "epoch": 9.89, + "learning_rate": 2.012427638871782e-06, + "loss": 0.3933, + "step": 26777 + }, + { + "epoch": 9.89, + "learning_rate": 2.012390688508437e-06, + "loss": 0.3618, + "step": 26778 + }, + { + "epoch": 9.89, + "learning_rate": 2.0123537381450918e-06, + "loss": 0.5557, + "step": 26779 + }, + { + "epoch": 9.89, + "learning_rate": 2.012316787781747e-06, + "loss": 0.3885, + "step": 26780 + }, + { + "epoch": 9.89, + "learning_rate": 2.0122798374184015e-06, + "loss": 0.3289, + "step": 26781 + }, + { + "epoch": 9.89, + "learning_rate": 2.012242887055056e-06, + "loss": 0.353, + "step": 26782 + }, + { + "epoch": 9.89, + "learning_rate": 2.0122059366917107e-06, + "loss": 0.4457, + "step": 26783 + }, + { + "epoch": 9.89, + "learning_rate": 2.012168986328366e-06, + "loss": 0.377, + "step": 26784 + }, + { + "epoch": 9.89, + "learning_rate": 2.0121320359650204e-06, + "loss": 0.2869, + "step": 26785 + }, + { + "epoch": 9.89, + "learning_rate": 2.012095085601675e-06, + "loss": 0.3155, + "step": 26786 + }, + { + "epoch": 9.89, + "learning_rate": 2.01205813523833e-06, + "loss": 0.3102, + "step": 26787 + }, + { + "epoch": 9.89, + "learning_rate": 2.0120211848749843e-06, + "loss": 0.4706, + "step": 26788 + }, + { + "epoch": 9.89, + "learning_rate": 2.0119842345116394e-06, + "loss": 0.3295, + "step": 26789 + }, + { + "epoch": 9.89, + "learning_rate": 2.011947284148294e-06, + "loss": 0.2425, + "step": 26790 + }, + { + "epoch": 9.89, + "learning_rate": 2.011910333784949e-06, + "loss": 0.2966, + "step": 26791 + }, + { + "epoch": 9.89, + "learning_rate": 2.0118733834216037e-06, + "loss": 0.3896, + "step": 26792 + }, + { + "epoch": 9.89, + "learning_rate": 2.011836433058259e-06, + "loss": 0.3152, + "step": 26793 + }, + { + "epoch": 9.89, + "learning_rate": 2.011799482694913e-06, + "loss": 0.292, + "step": 26794 + }, + { + "epoch": 9.89, + "learning_rate": 2.0117625323315676e-06, + "loss": 0.4805, + "step": 26795 + }, + { + "epoch": 9.9, + "learning_rate": 2.0117255819682227e-06, + "loss": 0.3243, + "step": 26796 + }, + { + "epoch": 9.9, + "learning_rate": 2.0116886316048773e-06, + "loss": 0.3747, + "step": 26797 + }, + { + "epoch": 9.9, + "learning_rate": 2.0116516812415324e-06, + "loss": 0.4095, + "step": 26798 + }, + { + "epoch": 9.9, + "learning_rate": 2.011614730878187e-06, + "loss": 0.4697, + "step": 26799 + }, + { + "epoch": 9.9, + "learning_rate": 2.0115777805148417e-06, + "loss": 0.3623, + "step": 26800 + }, + { + "epoch": 9.9, + "learning_rate": 2.0115408301514963e-06, + "loss": 0.4251, + "step": 26801 + }, + { + "epoch": 9.9, + "learning_rate": 2.0115038797881514e-06, + "loss": 0.4222, + "step": 26802 + }, + { + "epoch": 9.9, + "learning_rate": 2.011466929424806e-06, + "loss": 0.3487, + "step": 26803 + }, + { + "epoch": 9.9, + "learning_rate": 2.011429979061461e-06, + "loss": 0.3645, + "step": 26804 + }, + { + "epoch": 9.9, + "learning_rate": 2.0113930286981157e-06, + "loss": 0.2588, + "step": 26805 + }, + { + "epoch": 9.9, + "learning_rate": 2.0113560783347703e-06, + "loss": 0.2657, + "step": 26806 + }, + { + "epoch": 9.9, + "learning_rate": 2.011319127971425e-06, + "loss": 0.4345, + "step": 26807 + }, + { + "epoch": 9.9, + "learning_rate": 2.0112821776080796e-06, + "loss": 0.4402, + "step": 26808 + }, + { + "epoch": 9.9, + "learning_rate": 2.0112452272447347e-06, + "loss": 0.3666, + "step": 26809 + }, + { + "epoch": 9.9, + "learning_rate": 2.0112082768813893e-06, + "loss": 0.2876, + "step": 26810 + }, + { + "epoch": 9.9, + "learning_rate": 2.0111713265180444e-06, + "loss": 0.4501, + "step": 26811 + }, + { + "epoch": 9.9, + "learning_rate": 2.011134376154699e-06, + "loss": 0.1877, + "step": 26812 + }, + { + "epoch": 9.9, + "learning_rate": 2.0110974257913536e-06, + "loss": 0.3491, + "step": 26813 + }, + { + "epoch": 9.9, + "learning_rate": 2.0110604754280083e-06, + "loss": 0.3401, + "step": 26814 + }, + { + "epoch": 9.9, + "learning_rate": 2.011023525064663e-06, + "loss": 0.454, + "step": 26815 + }, + { + "epoch": 9.9, + "learning_rate": 2.010986574701318e-06, + "loss": 0.3937, + "step": 26816 + }, + { + "epoch": 9.9, + "learning_rate": 2.0109496243379726e-06, + "loss": 0.324, + "step": 26817 + }, + { + "epoch": 9.9, + "learning_rate": 2.0109126739746277e-06, + "loss": 0.3108, + "step": 26818 + }, + { + "epoch": 9.9, + "learning_rate": 2.0108757236112823e-06, + "loss": 0.3939, + "step": 26819 + }, + { + "epoch": 9.9, + "learning_rate": 2.010838773247937e-06, + "loss": 0.4031, + "step": 26820 + }, + { + "epoch": 9.9, + "learning_rate": 2.0108018228845916e-06, + "loss": 0.3412, + "step": 26821 + }, + { + "epoch": 9.9, + "learning_rate": 2.0107648725212466e-06, + "loss": 0.3708, + "step": 26822 + }, + { + "epoch": 9.91, + "learning_rate": 2.0107279221579013e-06, + "loss": 0.2841, + "step": 26823 + }, + { + "epoch": 9.91, + "learning_rate": 2.0106909717945563e-06, + "loss": 0.4568, + "step": 26824 + }, + { + "epoch": 9.91, + "learning_rate": 2.010654021431211e-06, + "loss": 0.5563, + "step": 26825 + }, + { + "epoch": 9.91, + "learning_rate": 2.0106170710678656e-06, + "loss": 0.3832, + "step": 26826 + }, + { + "epoch": 9.91, + "learning_rate": 2.0105801207045202e-06, + "loss": 0.4288, + "step": 26827 + }, + { + "epoch": 9.91, + "learning_rate": 2.010543170341175e-06, + "loss": 0.3794, + "step": 26828 + }, + { + "epoch": 9.91, + "learning_rate": 2.01050621997783e-06, + "loss": 0.2049, + "step": 26829 + }, + { + "epoch": 9.91, + "learning_rate": 2.0104692696144846e-06, + "loss": 0.3503, + "step": 26830 + }, + { + "epoch": 9.91, + "learning_rate": 2.0104323192511396e-06, + "loss": 0.3643, + "step": 26831 + }, + { + "epoch": 9.91, + "learning_rate": 2.0103953688877943e-06, + "loss": 0.3487, + "step": 26832 + }, + { + "epoch": 9.91, + "learning_rate": 2.010358418524449e-06, + "loss": 0.2891, + "step": 26833 + }, + { + "epoch": 9.91, + "learning_rate": 2.0103214681611035e-06, + "loss": 0.5285, + "step": 26834 + }, + { + "epoch": 9.91, + "learning_rate": 2.010284517797758e-06, + "loss": 0.3805, + "step": 26835 + }, + { + "epoch": 9.91, + "learning_rate": 2.0102475674344132e-06, + "loss": 0.4339, + "step": 26836 + }, + { + "epoch": 9.91, + "learning_rate": 2.010210617071068e-06, + "loss": 0.2675, + "step": 26837 + }, + { + "epoch": 9.91, + "learning_rate": 2.010173666707723e-06, + "loss": 0.2476, + "step": 26838 + }, + { + "epoch": 9.91, + "learning_rate": 2.010136716344377e-06, + "loss": 0.3184, + "step": 26839 + }, + { + "epoch": 9.91, + "learning_rate": 2.010099765981032e-06, + "loss": 0.295, + "step": 26840 + }, + { + "epoch": 9.91, + "learning_rate": 2.010062815617687e-06, + "loss": 0.32, + "step": 26841 + }, + { + "epoch": 9.91, + "learning_rate": 2.010025865254342e-06, + "loss": 0.5248, + "step": 26842 + }, + { + "epoch": 9.91, + "learning_rate": 2.0099889148909965e-06, + "loss": 0.5544, + "step": 26843 + }, + { + "epoch": 9.91, + "learning_rate": 2.0099519645276516e-06, + "loss": 0.3439, + "step": 26844 + }, + { + "epoch": 9.91, + "learning_rate": 2.009915014164306e-06, + "loss": 0.3088, + "step": 26845 + }, + { + "epoch": 9.91, + "learning_rate": 2.0098780638009605e-06, + "loss": 0.3135, + "step": 26846 + }, + { + "epoch": 9.91, + "learning_rate": 2.0098411134376155e-06, + "loss": 0.4919, + "step": 26847 + }, + { + "epoch": 9.91, + "learning_rate": 2.00980416307427e-06, + "loss": 0.401, + "step": 26848 + }, + { + "epoch": 9.91, + "learning_rate": 2.009767212710925e-06, + "loss": 0.3078, + "step": 26849 + }, + { + "epoch": 9.92, + "learning_rate": 2.00973026234758e-06, + "loss": 0.303, + "step": 26850 + }, + { + "epoch": 9.92, + "learning_rate": 2.0096933119842345e-06, + "loss": 0.4397, + "step": 26851 + }, + { + "epoch": 9.92, + "learning_rate": 2.009656361620889e-06, + "loss": 0.4529, + "step": 26852 + }, + { + "epoch": 9.92, + "learning_rate": 2.009619411257544e-06, + "loss": 0.5404, + "step": 26853 + }, + { + "epoch": 9.92, + "learning_rate": 2.009582460894199e-06, + "loss": 0.3574, + "step": 26854 + }, + { + "epoch": 9.92, + "learning_rate": 2.0095455105308535e-06, + "loss": 0.3426, + "step": 26855 + }, + { + "epoch": 9.92, + "learning_rate": 2.0095085601675085e-06, + "loss": 0.3, + "step": 26856 + }, + { + "epoch": 9.92, + "learning_rate": 2.009471609804163e-06, + "loss": 0.3053, + "step": 26857 + }, + { + "epoch": 9.92, + "learning_rate": 2.0094346594408178e-06, + "loss": 0.3256, + "step": 26858 + }, + { + "epoch": 9.92, + "learning_rate": 2.0093977090774724e-06, + "loss": 0.4205, + "step": 26859 + }, + { + "epoch": 9.92, + "learning_rate": 2.0093607587141275e-06, + "loss": 0.3305, + "step": 26860 + }, + { + "epoch": 9.92, + "learning_rate": 2.009323808350782e-06, + "loss": 0.3369, + "step": 26861 + }, + { + "epoch": 9.92, + "learning_rate": 2.009286857987437e-06, + "loss": 0.2684, + "step": 26862 + }, + { + "epoch": 9.92, + "learning_rate": 2.009249907624092e-06, + "loss": 0.3979, + "step": 26863 + }, + { + "epoch": 9.92, + "learning_rate": 2.0092129572607464e-06, + "loss": 0.275, + "step": 26864 + }, + { + "epoch": 9.92, + "learning_rate": 2.009176006897401e-06, + "loss": 0.3261, + "step": 26865 + }, + { + "epoch": 9.92, + "learning_rate": 2.0091390565340557e-06, + "loss": 0.3468, + "step": 26866 + }, + { + "epoch": 9.92, + "learning_rate": 2.0091021061707108e-06, + "loss": 0.4036, + "step": 26867 + }, + { + "epoch": 9.92, + "learning_rate": 2.0090651558073654e-06, + "loss": 0.3599, + "step": 26868 + }, + { + "epoch": 9.92, + "learning_rate": 2.0090282054440205e-06, + "loss": 0.4041, + "step": 26869 + }, + { + "epoch": 9.92, + "learning_rate": 2.008991255080675e-06, + "loss": 0.4384, + "step": 26870 + }, + { + "epoch": 9.92, + "learning_rate": 2.0089543047173298e-06, + "loss": 0.4427, + "step": 26871 + }, + { + "epoch": 9.92, + "learning_rate": 2.0089173543539844e-06, + "loss": 0.3878, + "step": 26872 + }, + { + "epoch": 9.92, + "learning_rate": 2.0088804039906394e-06, + "loss": 0.447, + "step": 26873 + }, + { + "epoch": 9.92, + "learning_rate": 2.008843453627294e-06, + "loss": 0.2915, + "step": 26874 + }, + { + "epoch": 9.92, + "learning_rate": 2.0088065032639487e-06, + "loss": 0.346, + "step": 26875 + }, + { + "epoch": 9.92, + "learning_rate": 2.0087695529006038e-06, + "loss": 0.2033, + "step": 26876 + }, + { + "epoch": 9.93, + "learning_rate": 2.0087326025372584e-06, + "loss": 0.2648, + "step": 26877 + }, + { + "epoch": 9.93, + "learning_rate": 2.008695652173913e-06, + "loss": 0.3985, + "step": 26878 + }, + { + "epoch": 9.93, + "learning_rate": 2.0086587018105677e-06, + "loss": 0.3424, + "step": 26879 + }, + { + "epoch": 9.93, + "learning_rate": 2.0086217514472228e-06, + "loss": 0.4281, + "step": 26880 + }, + { + "epoch": 9.93, + "learning_rate": 2.0085848010838774e-06, + "loss": 0.3574, + "step": 26881 + }, + { + "epoch": 9.93, + "learning_rate": 2.0085478507205324e-06, + "loss": 0.2852, + "step": 26882 + }, + { + "epoch": 9.93, + "learning_rate": 2.008510900357187e-06, + "loss": 0.3657, + "step": 26883 + }, + { + "epoch": 9.93, + "learning_rate": 2.0084739499938417e-06, + "loss": 0.2579, + "step": 26884 + }, + { + "epoch": 9.93, + "learning_rate": 2.0084369996304964e-06, + "loss": 0.3372, + "step": 26885 + }, + { + "epoch": 9.93, + "learning_rate": 2.008400049267151e-06, + "loss": 0.2645, + "step": 26886 + }, + { + "epoch": 9.93, + "learning_rate": 2.008363098903806e-06, + "loss": 0.2907, + "step": 26887 + }, + { + "epoch": 9.93, + "learning_rate": 2.0083261485404607e-06, + "loss": 0.422, + "step": 26888 + }, + { + "epoch": 9.93, + "learning_rate": 2.0082891981771157e-06, + "loss": 0.2557, + "step": 26889 + }, + { + "epoch": 9.93, + "learning_rate": 2.00825224781377e-06, + "loss": 0.4597, + "step": 26890 + }, + { + "epoch": 9.93, + "learning_rate": 2.008215297450425e-06, + "loss": 0.436, + "step": 26891 + }, + { + "epoch": 9.93, + "learning_rate": 2.0081783470870797e-06, + "loss": 0.3167, + "step": 26892 + }, + { + "epoch": 9.93, + "learning_rate": 2.0081413967237347e-06, + "loss": 0.3059, + "step": 26893 + }, + { + "epoch": 9.93, + "learning_rate": 2.0081044463603894e-06, + "loss": 0.2675, + "step": 26894 + }, + { + "epoch": 9.93, + "learning_rate": 2.008067495997044e-06, + "loss": 0.359, + "step": 26895 + }, + { + "epoch": 9.93, + "learning_rate": 2.0080305456336986e-06, + "loss": 0.3523, + "step": 26896 + }, + { + "epoch": 9.93, + "learning_rate": 2.0079935952703533e-06, + "loss": 0.3485, + "step": 26897 + }, + { + "epoch": 9.93, + "learning_rate": 2.0079566449070083e-06, + "loss": 0.337, + "step": 26898 + }, + { + "epoch": 9.93, + "learning_rate": 2.007919694543663e-06, + "loss": 0.405, + "step": 26899 + }, + { + "epoch": 9.93, + "learning_rate": 2.007882744180318e-06, + "loss": 0.339, + "step": 26900 + }, + { + "epoch": 9.93, + "learning_rate": 2.0078457938169727e-06, + "loss": 0.3174, + "step": 26901 + }, + { + "epoch": 9.93, + "learning_rate": 2.0078088434536273e-06, + "loss": 0.2678, + "step": 26902 + }, + { + "epoch": 9.93, + "learning_rate": 2.007771893090282e-06, + "loss": 0.3235, + "step": 26903 + }, + { + "epoch": 9.94, + "learning_rate": 2.007734942726937e-06, + "loss": 0.5924, + "step": 26904 + }, + { + "epoch": 9.94, + "learning_rate": 2.0076979923635916e-06, + "loss": 0.4936, + "step": 26905 + }, + { + "epoch": 9.94, + "learning_rate": 2.0076610420002463e-06, + "loss": 0.4334, + "step": 26906 + }, + { + "epoch": 9.94, + "learning_rate": 2.0076240916369013e-06, + "loss": 0.3833, + "step": 26907 + }, + { + "epoch": 9.94, + "learning_rate": 2.007587141273556e-06, + "loss": 0.4632, + "step": 26908 + }, + { + "epoch": 9.94, + "learning_rate": 2.0075501909102106e-06, + "loss": 0.3148, + "step": 26909 + }, + { + "epoch": 9.94, + "learning_rate": 2.0075132405468652e-06, + "loss": 0.4602, + "step": 26910 + }, + { + "epoch": 9.94, + "learning_rate": 2.0074762901835203e-06, + "loss": 0.4367, + "step": 26911 + }, + { + "epoch": 9.94, + "learning_rate": 2.007439339820175e-06, + "loss": 0.4319, + "step": 26912 + }, + { + "epoch": 9.94, + "learning_rate": 2.00740238945683e-06, + "loss": 0.3098, + "step": 26913 + }, + { + "epoch": 9.94, + "learning_rate": 2.0073654390934846e-06, + "loss": 0.4351, + "step": 26914 + }, + { + "epoch": 9.94, + "learning_rate": 2.0073284887301393e-06, + "loss": 0.2963, + "step": 26915 + }, + { + "epoch": 9.94, + "learning_rate": 2.007291538366794e-06, + "loss": 0.5131, + "step": 26916 + }, + { + "epoch": 9.94, + "learning_rate": 2.0072545880034485e-06, + "loss": 0.4401, + "step": 26917 + }, + { + "epoch": 9.94, + "learning_rate": 2.0072176376401036e-06, + "loss": 0.4322, + "step": 26918 + }, + { + "epoch": 9.94, + "learning_rate": 2.0071806872767582e-06, + "loss": 0.4438, + "step": 26919 + }, + { + "epoch": 9.94, + "learning_rate": 2.0071437369134133e-06, + "loss": 0.4273, + "step": 26920 + }, + { + "epoch": 9.94, + "learning_rate": 2.007106786550068e-06, + "loss": 0.3133, + "step": 26921 + }, + { + "epoch": 9.94, + "learning_rate": 2.0070698361867226e-06, + "loss": 0.2294, + "step": 26922 + }, + { + "epoch": 9.94, + "learning_rate": 2.007032885823377e-06, + "loss": 0.4269, + "step": 26923 + }, + { + "epoch": 9.94, + "learning_rate": 2.0069959354600323e-06, + "loss": 0.4212, + "step": 26924 + }, + { + "epoch": 9.94, + "learning_rate": 2.006958985096687e-06, + "loss": 0.4043, + "step": 26925 + }, + { + "epoch": 9.94, + "learning_rate": 2.0069220347333415e-06, + "loss": 0.3198, + "step": 26926 + }, + { + "epoch": 9.94, + "learning_rate": 2.0068850843699966e-06, + "loss": 0.4138, + "step": 26927 + }, + { + "epoch": 9.94, + "learning_rate": 2.006848134006651e-06, + "loss": 0.3928, + "step": 26928 + }, + { + "epoch": 9.94, + "learning_rate": 2.006811183643306e-06, + "loss": 0.381, + "step": 26929 + }, + { + "epoch": 9.94, + "learning_rate": 2.0067742332799605e-06, + "loss": 0.3793, + "step": 26930 + }, + { + "epoch": 9.94, + "learning_rate": 2.0067372829166156e-06, + "loss": 0.347, + "step": 26931 + }, + { + "epoch": 9.95, + "learning_rate": 2.00670033255327e-06, + "loss": 0.2723, + "step": 26932 + }, + { + "epoch": 9.95, + "learning_rate": 2.0066633821899253e-06, + "loss": 0.3995, + "step": 26933 + }, + { + "epoch": 9.95, + "learning_rate": 2.0066264318265795e-06, + "loss": 0.3255, + "step": 26934 + }, + { + "epoch": 9.95, + "learning_rate": 2.006589481463234e-06, + "loss": 0.2498, + "step": 26935 + }, + { + "epoch": 9.95, + "learning_rate": 2.006552531099889e-06, + "loss": 0.3737, + "step": 26936 + }, + { + "epoch": 9.95, + "learning_rate": 2.006515580736544e-06, + "loss": 0.3196, + "step": 26937 + }, + { + "epoch": 9.95, + "learning_rate": 2.006478630373199e-06, + "loss": 0.3829, + "step": 26938 + }, + { + "epoch": 9.95, + "learning_rate": 2.0064416800098535e-06, + "loss": 0.3436, + "step": 26939 + }, + { + "epoch": 9.95, + "learning_rate": 2.006404729646508e-06, + "loss": 0.3826, + "step": 26940 + }, + { + "epoch": 9.95, + "learning_rate": 2.0063677792831628e-06, + "loss": 0.4305, + "step": 26941 + }, + { + "epoch": 9.95, + "learning_rate": 2.006330828919818e-06, + "loss": 0.2967, + "step": 26942 + }, + { + "epoch": 9.95, + "learning_rate": 2.0062938785564725e-06, + "loss": 0.4365, + "step": 26943 + }, + { + "epoch": 9.95, + "learning_rate": 2.0062569281931275e-06, + "loss": 0.2531, + "step": 26944 + }, + { + "epoch": 9.95, + "learning_rate": 2.006219977829782e-06, + "loss": 0.304, + "step": 26945 + }, + { + "epoch": 9.95, + "learning_rate": 2.006183027466437e-06, + "loss": 0.4302, + "step": 26946 + }, + { + "epoch": 9.95, + "learning_rate": 2.0061460771030914e-06, + "loss": 0.4312, + "step": 26947 + }, + { + "epoch": 9.95, + "learning_rate": 2.006109126739746e-06, + "loss": 0.302, + "step": 26948 + }, + { + "epoch": 9.95, + "learning_rate": 2.006072176376401e-06, + "loss": 0.2988, + "step": 26949 + }, + { + "epoch": 9.95, + "learning_rate": 2.0060352260130558e-06, + "loss": 0.3945, + "step": 26950 + }, + { + "epoch": 9.95, + "learning_rate": 2.005998275649711e-06, + "loss": 0.4189, + "step": 26951 + }, + { + "epoch": 9.95, + "learning_rate": 2.0059613252863655e-06, + "loss": 0.3737, + "step": 26952 + }, + { + "epoch": 9.95, + "learning_rate": 2.00592437492302e-06, + "loss": 0.2955, + "step": 26953 + }, + { + "epoch": 9.95, + "learning_rate": 2.0058874245596747e-06, + "loss": 0.3277, + "step": 26954 + }, + { + "epoch": 9.95, + "learning_rate": 2.0058504741963294e-06, + "loss": 0.2837, + "step": 26955 + }, + { + "epoch": 9.95, + "learning_rate": 2.0058135238329844e-06, + "loss": 0.2837, + "step": 26956 + }, + { + "epoch": 9.95, + "learning_rate": 2.005776573469639e-06, + "loss": 0.2396, + "step": 26957 + }, + { + "epoch": 9.95, + "learning_rate": 2.005739623106294e-06, + "loss": 0.3608, + "step": 26958 + }, + { + "epoch": 9.96, + "learning_rate": 2.0057026727429488e-06, + "loss": 0.2858, + "step": 26959 + }, + { + "epoch": 9.96, + "learning_rate": 2.0056657223796034e-06, + "loss": 0.47, + "step": 26960 + }, + { + "epoch": 9.96, + "learning_rate": 2.005628772016258e-06, + "loss": 0.1928, + "step": 26961 + }, + { + "epoch": 9.96, + "learning_rate": 2.005591821652913e-06, + "loss": 0.2559, + "step": 26962 + }, + { + "epoch": 9.96, + "learning_rate": 2.0055548712895677e-06, + "loss": 0.354, + "step": 26963 + }, + { + "epoch": 9.96, + "learning_rate": 2.0055179209262224e-06, + "loss": 0.3275, + "step": 26964 + }, + { + "epoch": 9.96, + "learning_rate": 2.0054809705628774e-06, + "loss": 0.4341, + "step": 26965 + }, + { + "epoch": 9.96, + "learning_rate": 2.005444020199532e-06, + "loss": 0.4627, + "step": 26966 + }, + { + "epoch": 9.96, + "learning_rate": 2.0054070698361867e-06, + "loss": 0.3687, + "step": 26967 + }, + { + "epoch": 9.96, + "learning_rate": 2.0053701194728413e-06, + "loss": 0.3877, + "step": 26968 + }, + { + "epoch": 9.96, + "learning_rate": 2.0053331691094964e-06, + "loss": 0.4411, + "step": 26969 + }, + { + "epoch": 9.96, + "learning_rate": 2.005296218746151e-06, + "loss": 0.3023, + "step": 26970 + }, + { + "epoch": 9.96, + "learning_rate": 2.005259268382806e-06, + "loss": 0.3117, + "step": 26971 + }, + { + "epoch": 9.96, + "learning_rate": 2.0052223180194607e-06, + "loss": 0.3335, + "step": 26972 + }, + { + "epoch": 9.96, + "learning_rate": 2.0051853676561154e-06, + "loss": 0.3541, + "step": 26973 + }, + { + "epoch": 9.96, + "learning_rate": 2.00514841729277e-06, + "loss": 0.2759, + "step": 26974 + }, + { + "epoch": 9.96, + "learning_rate": 2.0051114669294246e-06, + "loss": 0.2745, + "step": 26975 + }, + { + "epoch": 9.96, + "learning_rate": 2.0050745165660797e-06, + "loss": 0.3822, + "step": 26976 + }, + { + "epoch": 9.96, + "learning_rate": 2.0050375662027343e-06, + "loss": 0.2892, + "step": 26977 + }, + { + "epoch": 9.96, + "learning_rate": 2.0050006158393894e-06, + "loss": 0.4227, + "step": 26978 + }, + { + "epoch": 9.96, + "learning_rate": 2.0049636654760436e-06, + "loss": 0.3541, + "step": 26979 + }, + { + "epoch": 9.96, + "learning_rate": 2.0049267151126987e-06, + "loss": 0.3831, + "step": 26980 + }, + { + "epoch": 9.96, + "learning_rate": 2.0048897647493533e-06, + "loss": 0.3616, + "step": 26981 + }, + { + "epoch": 9.96, + "learning_rate": 2.0048528143860084e-06, + "loss": 0.4475, + "step": 26982 + }, + { + "epoch": 9.96, + "learning_rate": 2.004815864022663e-06, + "loss": 0.3798, + "step": 26983 + }, + { + "epoch": 9.96, + "learning_rate": 2.0047789136593176e-06, + "loss": 0.2578, + "step": 26984 + }, + { + "epoch": 9.96, + "learning_rate": 2.0047419632959723e-06, + "loss": 0.4627, + "step": 26985 + }, + { + "epoch": 9.97, + "learning_rate": 2.004705012932627e-06, + "loss": 0.4883, + "step": 26986 + }, + { + "epoch": 9.97, + "learning_rate": 2.004668062569282e-06, + "loss": 0.3288, + "step": 26987 + }, + { + "epoch": 9.97, + "learning_rate": 2.0046311122059366e-06, + "loss": 0.3992, + "step": 26988 + }, + { + "epoch": 9.97, + "learning_rate": 2.0045941618425917e-06, + "loss": 0.4576, + "step": 26989 + }, + { + "epoch": 9.97, + "learning_rate": 2.0045572114792463e-06, + "loss": 0.4291, + "step": 26990 + }, + { + "epoch": 9.97, + "learning_rate": 2.004520261115901e-06, + "loss": 0.4909, + "step": 26991 + }, + { + "epoch": 9.97, + "learning_rate": 2.0044833107525556e-06, + "loss": 0.3116, + "step": 26992 + }, + { + "epoch": 9.97, + "learning_rate": 2.0044463603892106e-06, + "loss": 0.3189, + "step": 26993 + }, + { + "epoch": 9.97, + "learning_rate": 2.0044094100258653e-06, + "loss": 0.3517, + "step": 26994 + }, + { + "epoch": 9.97, + "learning_rate": 2.00437245966252e-06, + "loss": 0.3763, + "step": 26995 + }, + { + "epoch": 9.97, + "learning_rate": 2.004335509299175e-06, + "loss": 0.4702, + "step": 26996 + }, + { + "epoch": 9.97, + "learning_rate": 2.0042985589358296e-06, + "loss": 0.2941, + "step": 26997 + }, + { + "epoch": 9.97, + "learning_rate": 2.0042616085724842e-06, + "loss": 0.4817, + "step": 26998 + }, + { + "epoch": 9.97, + "learning_rate": 2.004224658209139e-06, + "loss": 0.319, + "step": 26999 + }, + { + "epoch": 9.97, + "learning_rate": 2.004187707845794e-06, + "loss": 0.3811, + "step": 27000 + }, + { + "epoch": 9.97, + "learning_rate": 2.0041507574824486e-06, + "loss": 0.2805, + "step": 27001 + }, + { + "epoch": 9.97, + "learning_rate": 2.0041138071191036e-06, + "loss": 0.3424, + "step": 27002 + }, + { + "epoch": 9.97, + "learning_rate": 2.0040768567557583e-06, + "loss": 0.2457, + "step": 27003 + }, + { + "epoch": 9.97, + "learning_rate": 2.004039906392413e-06, + "loss": 0.3348, + "step": 27004 + }, + { + "epoch": 9.97, + "learning_rate": 2.0040029560290676e-06, + "loss": 0.227, + "step": 27005 + }, + { + "epoch": 9.97, + "learning_rate": 2.003966005665722e-06, + "loss": 0.3912, + "step": 27006 + }, + { + "epoch": 9.97, + "learning_rate": 2.0039290553023772e-06, + "loss": 0.4237, + "step": 27007 + }, + { + "epoch": 9.97, + "learning_rate": 2.003892104939032e-06, + "loss": 0.4112, + "step": 27008 + }, + { + "epoch": 9.97, + "learning_rate": 2.003855154575687e-06, + "loss": 0.4621, + "step": 27009 + }, + { + "epoch": 9.97, + "learning_rate": 2.0038182042123416e-06, + "loss": 0.5047, + "step": 27010 + }, + { + "epoch": 9.97, + "learning_rate": 2.0037812538489962e-06, + "loss": 0.3252, + "step": 27011 + }, + { + "epoch": 9.97, + "learning_rate": 2.003744303485651e-06, + "loss": 0.3661, + "step": 27012 + }, + { + "epoch": 9.98, + "learning_rate": 2.003707353122306e-06, + "loss": 0.3926, + "step": 27013 + }, + { + "epoch": 9.98, + "learning_rate": 2.0036704027589605e-06, + "loss": 0.3871, + "step": 27014 + }, + { + "epoch": 9.98, + "learning_rate": 2.003633452395615e-06, + "loss": 0.2464, + "step": 27015 + }, + { + "epoch": 9.98, + "learning_rate": 2.0035965020322702e-06, + "loss": 0.3132, + "step": 27016 + }, + { + "epoch": 9.98, + "learning_rate": 2.003559551668925e-06, + "loss": 0.3854, + "step": 27017 + }, + { + "epoch": 9.98, + "learning_rate": 2.0035226013055795e-06, + "loss": 0.228, + "step": 27018 + }, + { + "epoch": 9.98, + "learning_rate": 2.003485650942234e-06, + "loss": 0.2097, + "step": 27019 + }, + { + "epoch": 9.98, + "learning_rate": 2.0034487005788892e-06, + "loss": 0.5216, + "step": 27020 + }, + { + "epoch": 9.98, + "learning_rate": 2.003411750215544e-06, + "loss": 0.4085, + "step": 27021 + }, + { + "epoch": 9.98, + "learning_rate": 2.003374799852199e-06, + "loss": 0.3422, + "step": 27022 + }, + { + "epoch": 9.98, + "learning_rate": 2.0033378494888535e-06, + "loss": 0.4495, + "step": 27023 + }, + { + "epoch": 9.98, + "learning_rate": 2.0033008991255078e-06, + "loss": 0.2817, + "step": 27024 + }, + { + "epoch": 9.98, + "learning_rate": 2.003263948762163e-06, + "loss": 0.3653, + "step": 27025 + }, + { + "epoch": 9.98, + "learning_rate": 2.0032269983988175e-06, + "loss": 0.3568, + "step": 27026 + }, + { + "epoch": 9.98, + "learning_rate": 2.0031900480354725e-06, + "loss": 0.4353, + "step": 27027 + }, + { + "epoch": 9.98, + "learning_rate": 2.003153097672127e-06, + "loss": 0.3516, + "step": 27028 + }, + { + "epoch": 9.98, + "learning_rate": 2.0031161473087822e-06, + "loss": 0.2796, + "step": 27029 + }, + { + "epoch": 9.98, + "learning_rate": 2.0030791969454364e-06, + "loss": 0.5126, + "step": 27030 + }, + { + "epoch": 9.98, + "learning_rate": 2.0030422465820915e-06, + "loss": 0.3482, + "step": 27031 + }, + { + "epoch": 9.98, + "learning_rate": 2.003005296218746e-06, + "loss": 0.496, + "step": 27032 + }, + { + "epoch": 9.98, + "learning_rate": 2.002968345855401e-06, + "loss": 0.3844, + "step": 27033 + }, + { + "epoch": 9.98, + "learning_rate": 2.002931395492056e-06, + "loss": 0.4759, + "step": 27034 + }, + { + "epoch": 9.98, + "learning_rate": 2.0028944451287105e-06, + "loss": 0.4536, + "step": 27035 + }, + { + "epoch": 9.98, + "learning_rate": 2.002857494765365e-06, + "loss": 0.3551, + "step": 27036 + }, + { + "epoch": 9.98, + "learning_rate": 2.0028205444020197e-06, + "loss": 0.2721, + "step": 27037 + }, + { + "epoch": 9.98, + "learning_rate": 2.002783594038675e-06, + "loss": 0.3006, + "step": 27038 + }, + { + "epoch": 9.98, + "learning_rate": 2.0027466436753294e-06, + "loss": 0.4048, + "step": 27039 + }, + { + "epoch": 9.99, + "learning_rate": 2.0027096933119845e-06, + "loss": 0.3578, + "step": 27040 + }, + { + "epoch": 9.99, + "learning_rate": 2.002672742948639e-06, + "loss": 0.3361, + "step": 27041 + }, + { + "epoch": 9.99, + "learning_rate": 2.0026357925852938e-06, + "loss": 0.4358, + "step": 27042 + }, + { + "epoch": 9.99, + "learning_rate": 2.0025988422219484e-06, + "loss": 0.5053, + "step": 27043 + }, + { + "epoch": 9.99, + "learning_rate": 2.002561891858603e-06, + "loss": 0.3073, + "step": 27044 + }, + { + "epoch": 9.99, + "learning_rate": 2.002524941495258e-06, + "loss": 0.2417, + "step": 27045 + }, + { + "epoch": 9.99, + "learning_rate": 2.0024879911319127e-06, + "loss": 0.3318, + "step": 27046 + }, + { + "epoch": 9.99, + "learning_rate": 2.0024510407685678e-06, + "loss": 0.4419, + "step": 27047 + }, + { + "epoch": 9.99, + "learning_rate": 2.0024140904052224e-06, + "loss": 0.326, + "step": 27048 + }, + { + "epoch": 9.99, + "learning_rate": 2.002377140041877e-06, + "loss": 0.3328, + "step": 27049 + }, + { + "epoch": 9.99, + "learning_rate": 2.0023401896785317e-06, + "loss": 0.2902, + "step": 27050 + }, + { + "epoch": 9.99, + "learning_rate": 2.0023032393151868e-06, + "loss": 0.2805, + "step": 27051 + }, + { + "epoch": 9.99, + "learning_rate": 2.0022662889518414e-06, + "loss": 0.421, + "step": 27052 + }, + { + "epoch": 9.99, + "learning_rate": 2.0022293385884965e-06, + "loss": 0.4038, + "step": 27053 + }, + { + "epoch": 9.99, + "learning_rate": 2.002192388225151e-06, + "loss": 0.3509, + "step": 27054 + }, + { + "epoch": 9.99, + "learning_rate": 2.0021554378618057e-06, + "loss": 0.2513, + "step": 27055 + }, + { + "epoch": 9.99, + "learning_rate": 2.0021184874984604e-06, + "loss": 0.4337, + "step": 27056 + }, + { + "epoch": 9.99, + "learning_rate": 2.002081537135115e-06, + "loss": 0.442, + "step": 27057 + }, + { + "epoch": 9.99, + "learning_rate": 2.00204458677177e-06, + "loss": 0.4215, + "step": 27058 + }, + { + "epoch": 9.99, + "learning_rate": 2.0020076364084247e-06, + "loss": 0.317, + "step": 27059 + }, + { + "epoch": 9.99, + "learning_rate": 2.0019706860450798e-06, + "loss": 0.3314, + "step": 27060 + }, + { + "epoch": 9.99, + "learning_rate": 2.0019337356817344e-06, + "loss": 0.3714, + "step": 27061 + }, + { + "epoch": 9.99, + "learning_rate": 2.001896785318389e-06, + "loss": 0.3687, + "step": 27062 + }, + { + "epoch": 9.99, + "learning_rate": 2.0018598349550437e-06, + "loss": 0.5317, + "step": 27063 + }, + { + "epoch": 9.99, + "learning_rate": 2.0018228845916983e-06, + "loss": 0.4004, + "step": 27064 + }, + { + "epoch": 9.99, + "learning_rate": 2.0017859342283534e-06, + "loss": 0.371, + "step": 27065 + }, + { + "epoch": 9.99, + "learning_rate": 2.001748983865008e-06, + "loss": 0.4175, + "step": 27066 + }, + { + "epoch": 10.0, + "learning_rate": 2.001712033501663e-06, + "loss": 0.3554, + "step": 27067 + }, + { + "epoch": 10.0, + "learning_rate": 2.0016750831383177e-06, + "loss": 0.4515, + "step": 27068 + }, + { + "epoch": 10.0, + "learning_rate": 2.0016381327749723e-06, + "loss": 0.3194, + "step": 27069 + }, + { + "epoch": 10.0, + "learning_rate": 2.001601182411627e-06, + "loss": 0.2661, + "step": 27070 + }, + { + "epoch": 10.0, + "learning_rate": 2.001564232048282e-06, + "loss": 0.279, + "step": 27071 + }, + { + "epoch": 10.0, + "learning_rate": 2.0015272816849367e-06, + "loss": 0.2765, + "step": 27072 + }, + { + "epoch": 10.0, + "learning_rate": 2.0014903313215917e-06, + "loss": 0.4837, + "step": 27073 + }, + { + "epoch": 10.0, + "learning_rate": 2.0014533809582464e-06, + "loss": 0.3658, + "step": 27074 + }, + { + "epoch": 10.0, + "learning_rate": 2.0014164305949006e-06, + "loss": 0.3162, + "step": 27075 + }, + { + "epoch": 10.0, + "learning_rate": 2.0013794802315556e-06, + "loss": 0.2787, + "step": 27076 + }, + { + "epoch": 10.0, + "learning_rate": 2.0013425298682103e-06, + "loss": 0.2702, + "step": 27077 + }, + { + "epoch": 10.0, + "learning_rate": 2.0013055795048653e-06, + "loss": 0.2853, + "step": 27078 + }, + { + "epoch": 10.0, + "learning_rate": 2.00126862914152e-06, + "loss": 0.2786, + "step": 27079 + }, + { + "epoch": 10.0, + "learning_rate": 2.0012316787781746e-06, + "loss": 0.4526, + "step": 27080 + }, + { + "epoch": 10.0, + "eval_accuracy": 0.8232673038740361, + "eval_loss": 0.4079156219959259, + "eval_runtime": 25.305, + "eval_samples_per_second": 1711.679, + "eval_steps_per_second": 26.754, + "step": 27080 + }, + { + "epoch": 10.0, + "learning_rate": 2.0011947284148292e-06, + "loss": 0.3981, + "step": 27081 + }, + { + "epoch": 10.0, + "learning_rate": 2.0011577780514843e-06, + "loss": 0.3484, + "step": 27082 + }, + { + "epoch": 10.0, + "learning_rate": 2.001120827688139e-06, + "loss": 0.3729, + "step": 27083 + }, + { + "epoch": 10.0, + "learning_rate": 2.0010838773247936e-06, + "loss": 0.3836, + "step": 27084 + }, + { + "epoch": 10.0, + "learning_rate": 2.0010469269614486e-06, + "loss": 0.3323, + "step": 27085 + }, + { + "epoch": 10.0, + "learning_rate": 2.0010099765981033e-06, + "loss": 0.513, + "step": 27086 + }, + { + "epoch": 10.0, + "learning_rate": 2.000973026234758e-06, + "loss": 0.3374, + "step": 27087 + }, + { + "epoch": 10.0, + "learning_rate": 2.0009360758714125e-06, + "loss": 0.3489, + "step": 27088 + }, + { + "epoch": 10.0, + "learning_rate": 2.0008991255080676e-06, + "loss": 0.4441, + "step": 27089 + }, + { + "epoch": 10.0, + "learning_rate": 2.0008621751447222e-06, + "loss": 0.232, + "step": 27090 + }, + { + "epoch": 10.0, + "learning_rate": 2.0008252247813773e-06, + "loss": 0.3108, + "step": 27091 + }, + { + "epoch": 10.0, + "learning_rate": 2.000788274418032e-06, + "loss": 0.4202, + "step": 27092 + }, + { + "epoch": 10.0, + "learning_rate": 2.0007513240546866e-06, + "loss": 0.3058, + "step": 27093 + }, + { + "epoch": 10.01, + "learning_rate": 2.000714373691341e-06, + "loss": 0.4144, + "step": 27094 + }, + { + "epoch": 10.01, + "learning_rate": 2.000677423327996e-06, + "loss": 0.2561, + "step": 27095 + }, + { + "epoch": 10.01, + "learning_rate": 2.000640472964651e-06, + "loss": 0.4672, + "step": 27096 + }, + { + "epoch": 10.01, + "learning_rate": 2.0006035226013055e-06, + "loss": 0.3166, + "step": 27097 + }, + { + "epoch": 10.01, + "learning_rate": 2.0005665722379606e-06, + "loss": 0.3056, + "step": 27098 + }, + { + "epoch": 10.01, + "learning_rate": 2.0005296218746152e-06, + "loss": 0.2414, + "step": 27099 + }, + { + "epoch": 10.01, + "learning_rate": 2.00049267151127e-06, + "loss": 0.3642, + "step": 27100 + }, + { + "epoch": 10.01, + "learning_rate": 2.0004557211479245e-06, + "loss": 0.2847, + "step": 27101 + }, + { + "epoch": 10.01, + "learning_rate": 2.0004187707845796e-06, + "loss": 0.3502, + "step": 27102 + }, + { + "epoch": 10.01, + "learning_rate": 2.000381820421234e-06, + "loss": 0.3361, + "step": 27103 + }, + { + "epoch": 10.01, + "learning_rate": 2.000344870057889e-06, + "loss": 0.3791, + "step": 27104 + }, + { + "epoch": 10.01, + "learning_rate": 2.000307919694544e-06, + "loss": 0.4996, + "step": 27105 + }, + { + "epoch": 10.01, + "learning_rate": 2.0002709693311985e-06, + "loss": 0.3286, + "step": 27106 + }, + { + "epoch": 10.01, + "learning_rate": 2.000234018967853e-06, + "loss": 0.2855, + "step": 27107 + }, + { + "epoch": 10.01, + "learning_rate": 2.000197068604508e-06, + "loss": 0.2515, + "step": 27108 + }, + { + "epoch": 10.01, + "learning_rate": 2.000160118241163e-06, + "loss": 0.324, + "step": 27109 + }, + { + "epoch": 10.01, + "learning_rate": 2.0001231678778175e-06, + "loss": 0.3094, + "step": 27110 + }, + { + "epoch": 10.01, + "learning_rate": 2.0000862175144726e-06, + "loss": 0.3587, + "step": 27111 + }, + { + "epoch": 10.01, + "learning_rate": 2.000049267151127e-06, + "loss": 0.271, + "step": 27112 + }, + { + "epoch": 10.01, + "learning_rate": 2.000012316787782e-06, + "loss": 0.3089, + "step": 27113 + }, + { + "epoch": 10.01, + "learning_rate": 1.9999753664244365e-06, + "loss": 0.4004, + "step": 27114 + }, + { + "epoch": 10.01, + "learning_rate": 1.999938416061091e-06, + "loss": 0.4233, + "step": 27115 + }, + { + "epoch": 10.01, + "learning_rate": 1.999901465697746e-06, + "loss": 0.2928, + "step": 27116 + }, + { + "epoch": 10.01, + "learning_rate": 1.999864515334401e-06, + "loss": 0.3645, + "step": 27117 + }, + { + "epoch": 10.01, + "learning_rate": 1.999827564971056e-06, + "loss": 0.3524, + "step": 27118 + }, + { + "epoch": 10.01, + "learning_rate": 1.99979061460771e-06, + "loss": 0.2544, + "step": 27119 + }, + { + "epoch": 10.01, + "learning_rate": 1.999753664244365e-06, + "loss": 0.241, + "step": 27120 + }, + { + "epoch": 10.02, + "learning_rate": 1.9997167138810198e-06, + "loss": 0.3841, + "step": 27121 + }, + { + "epoch": 10.02, + "learning_rate": 1.999679763517675e-06, + "loss": 0.3282, + "step": 27122 + }, + { + "epoch": 10.02, + "learning_rate": 1.9996428131543295e-06, + "loss": 0.4217, + "step": 27123 + }, + { + "epoch": 10.02, + "learning_rate": 1.999605862790984e-06, + "loss": 0.3753, + "step": 27124 + }, + { + "epoch": 10.02, + "learning_rate": 1.9995689124276387e-06, + "loss": 0.3587, + "step": 27125 + }, + { + "epoch": 10.02, + "learning_rate": 1.9995319620642934e-06, + "loss": 0.4006, + "step": 27126 + }, + { + "epoch": 10.02, + "learning_rate": 1.9994950117009484e-06, + "loss": 0.3364, + "step": 27127 + }, + { + "epoch": 10.02, + "learning_rate": 1.999458061337603e-06, + "loss": 0.3898, + "step": 27128 + }, + { + "epoch": 10.02, + "learning_rate": 1.999421110974258e-06, + "loss": 0.4469, + "step": 27129 + }, + { + "epoch": 10.02, + "learning_rate": 1.9993841606109128e-06, + "loss": 0.4592, + "step": 27130 + }, + { + "epoch": 10.02, + "learning_rate": 1.9993472102475674e-06, + "loss": 0.3973, + "step": 27131 + }, + { + "epoch": 10.02, + "learning_rate": 1.999310259884222e-06, + "loss": 0.3354, + "step": 27132 + }, + { + "epoch": 10.02, + "learning_rate": 1.999273309520877e-06, + "loss": 0.3018, + "step": 27133 + }, + { + "epoch": 10.02, + "learning_rate": 1.9992363591575317e-06, + "loss": 0.3315, + "step": 27134 + }, + { + "epoch": 10.02, + "learning_rate": 1.9991994087941864e-06, + "loss": 0.3412, + "step": 27135 + }, + { + "epoch": 10.02, + "learning_rate": 1.9991624584308414e-06, + "loss": 0.3363, + "step": 27136 + }, + { + "epoch": 10.02, + "learning_rate": 1.999125508067496e-06, + "loss": 0.3722, + "step": 27137 + }, + { + "epoch": 10.02, + "learning_rate": 1.9990885577041507e-06, + "loss": 0.1985, + "step": 27138 + }, + { + "epoch": 10.02, + "learning_rate": 1.9990516073408054e-06, + "loss": 0.3734, + "step": 27139 + }, + { + "epoch": 10.02, + "learning_rate": 1.9990146569774604e-06, + "loss": 0.3189, + "step": 27140 + }, + { + "epoch": 10.02, + "learning_rate": 1.998977706614115e-06, + "loss": 0.3591, + "step": 27141 + }, + { + "epoch": 10.02, + "learning_rate": 1.99894075625077e-06, + "loss": 0.4296, + "step": 27142 + }, + { + "epoch": 10.02, + "learning_rate": 1.9989038058874247e-06, + "loss": 0.2775, + "step": 27143 + }, + { + "epoch": 10.02, + "learning_rate": 1.9988668555240794e-06, + "loss": 0.3576, + "step": 27144 + }, + { + "epoch": 10.02, + "learning_rate": 1.998829905160734e-06, + "loss": 0.3407, + "step": 27145 + }, + { + "epoch": 10.02, + "learning_rate": 1.9987929547973887e-06, + "loss": 0.2752, + "step": 27146 + }, + { + "epoch": 10.02, + "learning_rate": 1.9987560044340437e-06, + "loss": 0.4244, + "step": 27147 + }, + { + "epoch": 10.03, + "learning_rate": 1.9987190540706983e-06, + "loss": 0.2794, + "step": 27148 + }, + { + "epoch": 10.03, + "learning_rate": 1.9986821037073534e-06, + "loss": 0.2087, + "step": 27149 + }, + { + "epoch": 10.03, + "learning_rate": 1.998645153344008e-06, + "loss": 0.3236, + "step": 27150 + }, + { + "epoch": 10.03, + "learning_rate": 1.9986082029806627e-06, + "loss": 0.2796, + "step": 27151 + }, + { + "epoch": 10.03, + "learning_rate": 1.9985712526173173e-06, + "loss": 0.3792, + "step": 27152 + }, + { + "epoch": 10.03, + "learning_rate": 1.9985343022539724e-06, + "loss": 0.6635, + "step": 27153 + }, + { + "epoch": 10.03, + "learning_rate": 1.998497351890627e-06, + "loss": 0.3705, + "step": 27154 + }, + { + "epoch": 10.03, + "learning_rate": 1.9984604015272817e-06, + "loss": 0.293, + "step": 27155 + }, + { + "epoch": 10.03, + "learning_rate": 1.9984234511639367e-06, + "loss": 0.3302, + "step": 27156 + }, + { + "epoch": 10.03, + "learning_rate": 1.9983865008005913e-06, + "loss": 0.3822, + "step": 27157 + }, + { + "epoch": 10.03, + "learning_rate": 1.998349550437246e-06, + "loss": 0.334, + "step": 27158 + }, + { + "epoch": 10.03, + "learning_rate": 1.9983126000739006e-06, + "loss": 0.3918, + "step": 27159 + }, + { + "epoch": 10.03, + "learning_rate": 1.9982756497105557e-06, + "loss": 0.3601, + "step": 27160 + }, + { + "epoch": 10.03, + "learning_rate": 1.9982386993472103e-06, + "loss": 0.3749, + "step": 27161 + }, + { + "epoch": 10.03, + "learning_rate": 1.9982017489838654e-06, + "loss": 0.3222, + "step": 27162 + }, + { + "epoch": 10.03, + "learning_rate": 1.99816479862052e-06, + "loss": 0.3238, + "step": 27163 + }, + { + "epoch": 10.03, + "learning_rate": 1.9981278482571742e-06, + "loss": 0.3177, + "step": 27164 + }, + { + "epoch": 10.03, + "learning_rate": 1.9980908978938293e-06, + "loss": 0.3266, + "step": 27165 + }, + { + "epoch": 10.03, + "learning_rate": 1.998053947530484e-06, + "loss": 0.3177, + "step": 27166 + }, + { + "epoch": 10.03, + "learning_rate": 1.998016997167139e-06, + "loss": 0.3087, + "step": 27167 + }, + { + "epoch": 10.03, + "learning_rate": 1.9979800468037936e-06, + "loss": 0.2983, + "step": 27168 + }, + { + "epoch": 10.03, + "learning_rate": 1.9979430964404487e-06, + "loss": 0.3485, + "step": 27169 + }, + { + "epoch": 10.03, + "learning_rate": 1.997906146077103e-06, + "loss": 0.5645, + "step": 27170 + }, + { + "epoch": 10.03, + "learning_rate": 1.997869195713758e-06, + "loss": 0.3352, + "step": 27171 + }, + { + "epoch": 10.03, + "learning_rate": 1.9978322453504126e-06, + "loss": 0.4088, + "step": 27172 + }, + { + "epoch": 10.03, + "learning_rate": 1.9977952949870676e-06, + "loss": 0.3845, + "step": 27173 + }, + { + "epoch": 10.03, + "learning_rate": 1.9977583446237223e-06, + "loss": 0.3336, + "step": 27174 + }, + { + "epoch": 10.04, + "learning_rate": 1.997721394260377e-06, + "loss": 0.3359, + "step": 27175 + }, + { + "epoch": 10.04, + "learning_rate": 1.9976844438970316e-06, + "loss": 0.2775, + "step": 27176 + }, + { + "epoch": 10.04, + "learning_rate": 1.997647493533686e-06, + "loss": 0.257, + "step": 27177 + }, + { + "epoch": 10.04, + "learning_rate": 1.9976105431703413e-06, + "loss": 0.3122, + "step": 27178 + }, + { + "epoch": 10.04, + "learning_rate": 1.997573592806996e-06, + "loss": 0.3717, + "step": 27179 + }, + { + "epoch": 10.04, + "learning_rate": 1.997536642443651e-06, + "loss": 0.3776, + "step": 27180 + }, + { + "epoch": 10.04, + "learning_rate": 1.9974996920803056e-06, + "loss": 0.5303, + "step": 27181 + }, + { + "epoch": 10.04, + "learning_rate": 1.9974627417169602e-06, + "loss": 0.4325, + "step": 27182 + }, + { + "epoch": 10.04, + "learning_rate": 1.997425791353615e-06, + "loss": 0.3916, + "step": 27183 + }, + { + "epoch": 10.04, + "learning_rate": 1.9973888409902695e-06, + "loss": 0.2673, + "step": 27184 + }, + { + "epoch": 10.04, + "learning_rate": 1.9973518906269246e-06, + "loss": 0.2579, + "step": 27185 + }, + { + "epoch": 10.04, + "learning_rate": 1.997314940263579e-06, + "loss": 0.2272, + "step": 27186 + }, + { + "epoch": 10.04, + "learning_rate": 1.9972779899002343e-06, + "loss": 0.3257, + "step": 27187 + }, + { + "epoch": 10.04, + "learning_rate": 1.997241039536889e-06, + "loss": 0.3603, + "step": 27188 + }, + { + "epoch": 10.04, + "learning_rate": 1.9972040891735435e-06, + "loss": 0.2244, + "step": 27189 + }, + { + "epoch": 10.04, + "learning_rate": 1.997167138810198e-06, + "loss": 0.3889, + "step": 27190 + }, + { + "epoch": 10.04, + "learning_rate": 1.9971301884468532e-06, + "loss": 0.3057, + "step": 27191 + }, + { + "epoch": 10.04, + "learning_rate": 1.997093238083508e-06, + "loss": 0.4536, + "step": 27192 + }, + { + "epoch": 10.04, + "learning_rate": 1.9970562877201625e-06, + "loss": 0.3052, + "step": 27193 + }, + { + "epoch": 10.04, + "learning_rate": 1.9970193373568176e-06, + "loss": 0.3628, + "step": 27194 + }, + { + "epoch": 10.04, + "learning_rate": 1.996982386993472e-06, + "loss": 0.4943, + "step": 27195 + }, + { + "epoch": 10.04, + "learning_rate": 1.996945436630127e-06, + "loss": 0.346, + "step": 27196 + }, + { + "epoch": 10.04, + "learning_rate": 1.9969084862667815e-06, + "loss": 0.2865, + "step": 27197 + }, + { + "epoch": 10.04, + "learning_rate": 1.9968715359034365e-06, + "loss": 0.3655, + "step": 27198 + }, + { + "epoch": 10.04, + "learning_rate": 1.996834585540091e-06, + "loss": 0.2553, + "step": 27199 + }, + { + "epoch": 10.04, + "learning_rate": 1.9967976351767462e-06, + "loss": 0.5993, + "step": 27200 + }, + { + "epoch": 10.04, + "learning_rate": 1.996760684813401e-06, + "loss": 0.457, + "step": 27201 + }, + { + "epoch": 10.05, + "learning_rate": 1.9967237344500555e-06, + "loss": 0.4683, + "step": 27202 + }, + { + "epoch": 10.05, + "learning_rate": 1.99668678408671e-06, + "loss": 0.4881, + "step": 27203 + }, + { + "epoch": 10.05, + "learning_rate": 1.9966498337233648e-06, + "loss": 0.4411, + "step": 27204 + }, + { + "epoch": 10.05, + "learning_rate": 1.99661288336002e-06, + "loss": 0.4727, + "step": 27205 + }, + { + "epoch": 10.05, + "learning_rate": 1.9965759329966745e-06, + "loss": 0.247, + "step": 27206 + }, + { + "epoch": 10.05, + "learning_rate": 1.9965389826333295e-06, + "loss": 0.3435, + "step": 27207 + }, + { + "epoch": 10.05, + "learning_rate": 1.996502032269984e-06, + "loss": 0.2744, + "step": 27208 + }, + { + "epoch": 10.05, + "learning_rate": 1.996465081906639e-06, + "loss": 0.2908, + "step": 27209 + }, + { + "epoch": 10.05, + "learning_rate": 1.9964281315432934e-06, + "loss": 0.3625, + "step": 27210 + }, + { + "epoch": 10.05, + "learning_rate": 1.9963911811799485e-06, + "loss": 0.297, + "step": 27211 + }, + { + "epoch": 10.05, + "learning_rate": 1.996354230816603e-06, + "loss": 0.322, + "step": 27212 + }, + { + "epoch": 10.05, + "learning_rate": 1.9963172804532578e-06, + "loss": 0.2506, + "step": 27213 + }, + { + "epoch": 10.05, + "learning_rate": 1.996280330089913e-06, + "loss": 0.2735, + "step": 27214 + }, + { + "epoch": 10.05, + "learning_rate": 1.996243379726567e-06, + "loss": 0.3517, + "step": 27215 + }, + { + "epoch": 10.05, + "learning_rate": 1.996206429363222e-06, + "loss": 0.3847, + "step": 27216 + }, + { + "epoch": 10.05, + "learning_rate": 1.9961694789998767e-06, + "loss": 0.3613, + "step": 27217 + }, + { + "epoch": 10.05, + "learning_rate": 1.996132528636532e-06, + "loss": 0.3667, + "step": 27218 + }, + { + "epoch": 10.05, + "learning_rate": 1.9960955782731864e-06, + "loss": 0.311, + "step": 27219 + }, + { + "epoch": 10.05, + "learning_rate": 1.9960586279098415e-06, + "loss": 0.2985, + "step": 27220 + }, + { + "epoch": 10.05, + "learning_rate": 1.9960216775464957e-06, + "loss": 0.375, + "step": 27221 + }, + { + "epoch": 10.05, + "learning_rate": 1.9959847271831508e-06, + "loss": 0.3259, + "step": 27222 + }, + { + "epoch": 10.05, + "learning_rate": 1.9959477768198054e-06, + "loss": 0.4043, + "step": 27223 + }, + { + "epoch": 10.05, + "learning_rate": 1.99591082645646e-06, + "loss": 0.342, + "step": 27224 + }, + { + "epoch": 10.05, + "learning_rate": 1.995873876093115e-06, + "loss": 0.3697, + "step": 27225 + }, + { + "epoch": 10.05, + "learning_rate": 1.9958369257297697e-06, + "loss": 0.3316, + "step": 27226 + }, + { + "epoch": 10.05, + "learning_rate": 1.9957999753664244e-06, + "loss": 0.3626, + "step": 27227 + }, + { + "epoch": 10.05, + "learning_rate": 1.995763025003079e-06, + "loss": 0.2806, + "step": 27228 + }, + { + "epoch": 10.06, + "learning_rate": 1.995726074639734e-06, + "loss": 0.399, + "step": 27229 + }, + { + "epoch": 10.06, + "learning_rate": 1.9956891242763887e-06, + "loss": 0.3337, + "step": 27230 + }, + { + "epoch": 10.06, + "learning_rate": 1.9956521739130438e-06, + "loss": 0.4269, + "step": 27231 + }, + { + "epoch": 10.06, + "learning_rate": 1.9956152235496984e-06, + "loss": 0.4995, + "step": 27232 + }, + { + "epoch": 10.06, + "learning_rate": 1.995578273186353e-06, + "loss": 0.2922, + "step": 27233 + }, + { + "epoch": 10.06, + "learning_rate": 1.9955413228230077e-06, + "loss": 0.4061, + "step": 27234 + }, + { + "epoch": 10.06, + "learning_rate": 1.9955043724596623e-06, + "loss": 0.3559, + "step": 27235 + }, + { + "epoch": 10.06, + "learning_rate": 1.9954674220963174e-06, + "loss": 0.3925, + "step": 27236 + }, + { + "epoch": 10.06, + "learning_rate": 1.995430471732972e-06, + "loss": 0.2766, + "step": 27237 + }, + { + "epoch": 10.06, + "learning_rate": 1.995393521369627e-06, + "loss": 0.2579, + "step": 27238 + }, + { + "epoch": 10.06, + "learning_rate": 1.9953565710062817e-06, + "loss": 0.3637, + "step": 27239 + }, + { + "epoch": 10.06, + "learning_rate": 1.9953196206429363e-06, + "loss": 0.419, + "step": 27240 + }, + { + "epoch": 10.06, + "learning_rate": 1.995282670279591e-06, + "loss": 0.2703, + "step": 27241 + }, + { + "epoch": 10.06, + "learning_rate": 1.995245719916246e-06, + "loss": 0.2876, + "step": 27242 + }, + { + "epoch": 10.06, + "learning_rate": 1.9952087695529007e-06, + "loss": 0.411, + "step": 27243 + }, + { + "epoch": 10.06, + "learning_rate": 1.9951718191895553e-06, + "loss": 0.2654, + "step": 27244 + }, + { + "epoch": 10.06, + "learning_rate": 1.9951348688262104e-06, + "loss": 0.2757, + "step": 27245 + }, + { + "epoch": 10.06, + "learning_rate": 1.995097918462865e-06, + "loss": 0.3084, + "step": 27246 + }, + { + "epoch": 10.06, + "learning_rate": 1.9950609680995196e-06, + "loss": 0.4519, + "step": 27247 + }, + { + "epoch": 10.06, + "learning_rate": 1.9950240177361743e-06, + "loss": 0.3547, + "step": 27248 + }, + { + "epoch": 10.06, + "learning_rate": 1.9949870673728293e-06, + "loss": 0.244, + "step": 27249 + }, + { + "epoch": 10.06, + "learning_rate": 1.994950117009484e-06, + "loss": 0.3864, + "step": 27250 + }, + { + "epoch": 10.06, + "learning_rate": 1.994913166646139e-06, + "loss": 0.3865, + "step": 27251 + }, + { + "epoch": 10.06, + "learning_rate": 1.9948762162827937e-06, + "loss": 0.1998, + "step": 27252 + }, + { + "epoch": 10.06, + "learning_rate": 1.9948392659194483e-06, + "loss": 0.2608, + "step": 27253 + }, + { + "epoch": 10.06, + "learning_rate": 1.994802315556103e-06, + "loss": 0.4236, + "step": 27254 + }, + { + "epoch": 10.06, + "learning_rate": 1.9947653651927576e-06, + "loss": 0.3035, + "step": 27255 + }, + { + "epoch": 10.06, + "learning_rate": 1.9947284148294126e-06, + "loss": 0.4127, + "step": 27256 + }, + { + "epoch": 10.07, + "learning_rate": 1.9946914644660673e-06, + "loss": 0.4075, + "step": 27257 + }, + { + "epoch": 10.07, + "learning_rate": 1.9946545141027223e-06, + "loss": 0.2987, + "step": 27258 + }, + { + "epoch": 10.07, + "learning_rate": 1.994617563739377e-06, + "loss": 0.5447, + "step": 27259 + }, + { + "epoch": 10.07, + "learning_rate": 1.9945806133760316e-06, + "loss": 0.2876, + "step": 27260 + }, + { + "epoch": 10.07, + "learning_rate": 1.9945436630126862e-06, + "loss": 0.3055, + "step": 27261 + }, + { + "epoch": 10.07, + "learning_rate": 1.9945067126493413e-06, + "loss": 0.2628, + "step": 27262 + }, + { + "epoch": 10.07, + "learning_rate": 1.994469762285996e-06, + "loss": 0.4399, + "step": 27263 + }, + { + "epoch": 10.07, + "learning_rate": 1.9944328119226506e-06, + "loss": 0.381, + "step": 27264 + }, + { + "epoch": 10.07, + "learning_rate": 1.994395861559305e-06, + "loss": 0.5053, + "step": 27265 + }, + { + "epoch": 10.07, + "learning_rate": 1.99435891119596e-06, + "loss": 0.4138, + "step": 27266 + }, + { + "epoch": 10.07, + "learning_rate": 1.994321960832615e-06, + "loss": 0.3743, + "step": 27267 + }, + { + "epoch": 10.07, + "learning_rate": 1.9942850104692695e-06, + "loss": 0.365, + "step": 27268 + }, + { + "epoch": 10.07, + "learning_rate": 1.9942480601059246e-06, + "loss": 0.2201, + "step": 27269 + }, + { + "epoch": 10.07, + "learning_rate": 1.9942111097425792e-06, + "loss": 0.3675, + "step": 27270 + }, + { + "epoch": 10.07, + "learning_rate": 1.994174159379234e-06, + "loss": 0.3886, + "step": 27271 + }, + { + "epoch": 10.07, + "learning_rate": 1.9941372090158885e-06, + "loss": 0.2156, + "step": 27272 + }, + { + "epoch": 10.07, + "learning_rate": 1.994100258652543e-06, + "loss": 0.3909, + "step": 27273 + }, + { + "epoch": 10.07, + "learning_rate": 1.994063308289198e-06, + "loss": 0.3349, + "step": 27274 + }, + { + "epoch": 10.07, + "learning_rate": 1.994026357925853e-06, + "loss": 0.3606, + "step": 27275 + }, + { + "epoch": 10.07, + "learning_rate": 1.993989407562508e-06, + "loss": 0.3378, + "step": 27276 + }, + { + "epoch": 10.07, + "learning_rate": 1.9939524571991625e-06, + "loss": 0.2524, + "step": 27277 + }, + { + "epoch": 10.07, + "learning_rate": 1.993915506835817e-06, + "loss": 0.3912, + "step": 27278 + }, + { + "epoch": 10.07, + "learning_rate": 1.993878556472472e-06, + "loss": 0.4073, + "step": 27279 + }, + { + "epoch": 10.07, + "learning_rate": 1.993841606109127e-06, + "loss": 0.4803, + "step": 27280 + }, + { + "epoch": 10.07, + "learning_rate": 1.9938046557457815e-06, + "loss": 0.409, + "step": 27281 + }, + { + "epoch": 10.07, + "learning_rate": 1.9937677053824366e-06, + "loss": 0.3164, + "step": 27282 + }, + { + "epoch": 10.07, + "learning_rate": 1.993730755019091e-06, + "loss": 0.243, + "step": 27283 + }, + { + "epoch": 10.08, + "learning_rate": 1.993693804655746e-06, + "loss": 0.378, + "step": 27284 + }, + { + "epoch": 10.08, + "learning_rate": 1.9936568542924005e-06, + "loss": 0.3986, + "step": 27285 + }, + { + "epoch": 10.08, + "learning_rate": 1.993619903929055e-06, + "loss": 0.4274, + "step": 27286 + }, + { + "epoch": 10.08, + "learning_rate": 1.99358295356571e-06, + "loss": 0.3115, + "step": 27287 + }, + { + "epoch": 10.08, + "learning_rate": 1.993546003202365e-06, + "loss": 0.2456, + "step": 27288 + }, + { + "epoch": 10.08, + "learning_rate": 1.99350905283902e-06, + "loss": 0.248, + "step": 27289 + }, + { + "epoch": 10.08, + "learning_rate": 1.9934721024756745e-06, + "loss": 0.3532, + "step": 27290 + }, + { + "epoch": 10.08, + "learning_rate": 1.993435152112329e-06, + "loss": 0.3474, + "step": 27291 + }, + { + "epoch": 10.08, + "learning_rate": 1.9933982017489838e-06, + "loss": 0.3474, + "step": 27292 + }, + { + "epoch": 10.08, + "learning_rate": 1.9933612513856384e-06, + "loss": 0.316, + "step": 27293 + }, + { + "epoch": 10.08, + "learning_rate": 1.9933243010222935e-06, + "loss": 0.3117, + "step": 27294 + }, + { + "epoch": 10.08, + "learning_rate": 1.993287350658948e-06, + "loss": 0.3761, + "step": 27295 + }, + { + "epoch": 10.08, + "learning_rate": 1.993250400295603e-06, + "loss": 0.3168, + "step": 27296 + }, + { + "epoch": 10.08, + "learning_rate": 1.993213449932258e-06, + "loss": 0.3349, + "step": 27297 + }, + { + "epoch": 10.08, + "learning_rate": 1.9931764995689124e-06, + "loss": 0.316, + "step": 27298 + }, + { + "epoch": 10.08, + "learning_rate": 1.993139549205567e-06, + "loss": 0.236, + "step": 27299 + }, + { + "epoch": 10.08, + "learning_rate": 1.993102598842222e-06, + "loss": 0.2645, + "step": 27300 + }, + { + "epoch": 10.08, + "learning_rate": 1.9930656484788768e-06, + "loss": 0.3851, + "step": 27301 + }, + { + "epoch": 10.08, + "learning_rate": 1.993028698115532e-06, + "loss": 0.3025, + "step": 27302 + }, + { + "epoch": 10.08, + "learning_rate": 1.9929917477521865e-06, + "loss": 0.2887, + "step": 27303 + }, + { + "epoch": 10.08, + "learning_rate": 1.9929547973888407e-06, + "loss": 0.3596, + "step": 27304 + }, + { + "epoch": 10.08, + "learning_rate": 1.9929178470254958e-06, + "loss": 0.3604, + "step": 27305 + }, + { + "epoch": 10.08, + "learning_rate": 1.9928808966621504e-06, + "loss": 0.3617, + "step": 27306 + }, + { + "epoch": 10.08, + "learning_rate": 1.9928439462988054e-06, + "loss": 0.2841, + "step": 27307 + }, + { + "epoch": 10.08, + "learning_rate": 1.99280699593546e-06, + "loss": 0.2983, + "step": 27308 + }, + { + "epoch": 10.08, + "learning_rate": 1.992770045572115e-06, + "loss": 0.3986, + "step": 27309 + }, + { + "epoch": 10.08, + "learning_rate": 1.9927330952087694e-06, + "loss": 0.3425, + "step": 27310 + }, + { + "epoch": 10.09, + "learning_rate": 1.9926961448454244e-06, + "loss": 0.3552, + "step": 27311 + }, + { + "epoch": 10.09, + "learning_rate": 1.992659194482079e-06, + "loss": 0.3595, + "step": 27312 + }, + { + "epoch": 10.09, + "learning_rate": 1.9926222441187337e-06, + "loss": 0.3623, + "step": 27313 + }, + { + "epoch": 10.09, + "learning_rate": 1.9925852937553887e-06, + "loss": 0.3029, + "step": 27314 + }, + { + "epoch": 10.09, + "learning_rate": 1.9925483433920434e-06, + "loss": 0.263, + "step": 27315 + }, + { + "epoch": 10.09, + "learning_rate": 1.992511393028698e-06, + "loss": 0.4032, + "step": 27316 + }, + { + "epoch": 10.09, + "learning_rate": 1.9924744426653527e-06, + "loss": 0.2631, + "step": 27317 + }, + { + "epoch": 10.09, + "learning_rate": 1.9924374923020077e-06, + "loss": 0.3612, + "step": 27318 + }, + { + "epoch": 10.09, + "learning_rate": 1.9924005419386624e-06, + "loss": 0.2515, + "step": 27319 + }, + { + "epoch": 10.09, + "learning_rate": 1.9923635915753174e-06, + "loss": 0.2585, + "step": 27320 + }, + { + "epoch": 10.09, + "learning_rate": 1.992326641211972e-06, + "loss": 0.337, + "step": 27321 + }, + { + "epoch": 10.09, + "learning_rate": 1.9922896908486267e-06, + "loss": 0.3681, + "step": 27322 + }, + { + "epoch": 10.09, + "learning_rate": 1.9922527404852813e-06, + "loss": 0.3616, + "step": 27323 + }, + { + "epoch": 10.09, + "learning_rate": 1.992215790121936e-06, + "loss": 0.3383, + "step": 27324 + }, + { + "epoch": 10.09, + "learning_rate": 1.992178839758591e-06, + "loss": 0.2876, + "step": 27325 + }, + { + "epoch": 10.09, + "learning_rate": 1.9921418893952457e-06, + "loss": 0.4096, + "step": 27326 + }, + { + "epoch": 10.09, + "learning_rate": 1.9921049390319007e-06, + "loss": 0.328, + "step": 27327 + }, + { + "epoch": 10.09, + "learning_rate": 1.9920679886685554e-06, + "loss": 0.376, + "step": 27328 + }, + { + "epoch": 10.09, + "learning_rate": 1.99203103830521e-06, + "loss": 0.3155, + "step": 27329 + }, + { + "epoch": 10.09, + "learning_rate": 1.9919940879418646e-06, + "loss": 0.2769, + "step": 27330 + }, + { + "epoch": 10.09, + "learning_rate": 1.9919571375785197e-06, + "loss": 0.2683, + "step": 27331 + }, + { + "epoch": 10.09, + "learning_rate": 1.9919201872151743e-06, + "loss": 0.4083, + "step": 27332 + }, + { + "epoch": 10.09, + "learning_rate": 1.991883236851829e-06, + "loss": 0.3379, + "step": 27333 + }, + { + "epoch": 10.09, + "learning_rate": 1.991846286488484e-06, + "loss": 0.2613, + "step": 27334 + }, + { + "epoch": 10.09, + "learning_rate": 1.9918093361251387e-06, + "loss": 0.3206, + "step": 27335 + }, + { + "epoch": 10.09, + "learning_rate": 1.9917723857617933e-06, + "loss": 0.3985, + "step": 27336 + }, + { + "epoch": 10.09, + "learning_rate": 1.991735435398448e-06, + "loss": 0.3812, + "step": 27337 + }, + { + "epoch": 10.1, + "learning_rate": 1.991698485035103e-06, + "loss": 0.3907, + "step": 27338 + }, + { + "epoch": 10.1, + "learning_rate": 1.9916615346717576e-06, + "loss": 0.3289, + "step": 27339 + }, + { + "epoch": 10.1, + "learning_rate": 1.9916245843084127e-06, + "loss": 0.348, + "step": 27340 + }, + { + "epoch": 10.1, + "learning_rate": 1.9915876339450673e-06, + "loss": 0.4088, + "step": 27341 + }, + { + "epoch": 10.1, + "learning_rate": 1.991550683581722e-06, + "loss": 0.2394, + "step": 27342 + }, + { + "epoch": 10.1, + "learning_rate": 1.9915137332183766e-06, + "loss": 0.2741, + "step": 27343 + }, + { + "epoch": 10.1, + "learning_rate": 1.9914767828550312e-06, + "loss": 0.2401, + "step": 27344 + }, + { + "epoch": 10.1, + "learning_rate": 1.9914398324916863e-06, + "loss": 0.3256, + "step": 27345 + }, + { + "epoch": 10.1, + "learning_rate": 1.991402882128341e-06, + "loss": 0.4985, + "step": 27346 + }, + { + "epoch": 10.1, + "learning_rate": 1.991365931764996e-06, + "loss": 0.3383, + "step": 27347 + }, + { + "epoch": 10.1, + "learning_rate": 1.9913289814016506e-06, + "loss": 0.2405, + "step": 27348 + }, + { + "epoch": 10.1, + "learning_rate": 1.9912920310383053e-06, + "loss": 0.3149, + "step": 27349 + }, + { + "epoch": 10.1, + "learning_rate": 1.99125508067496e-06, + "loss": 0.5218, + "step": 27350 + }, + { + "epoch": 10.1, + "learning_rate": 1.991218130311615e-06, + "loss": 0.4431, + "step": 27351 + }, + { + "epoch": 10.1, + "learning_rate": 1.9911811799482696e-06, + "loss": 0.2538, + "step": 27352 + }, + { + "epoch": 10.1, + "learning_rate": 1.9911442295849242e-06, + "loss": 0.2422, + "step": 27353 + }, + { + "epoch": 10.1, + "learning_rate": 1.9911072792215793e-06, + "loss": 0.4095, + "step": 27354 + }, + { + "epoch": 10.1, + "learning_rate": 1.9910703288582335e-06, + "loss": 0.2572, + "step": 27355 + }, + { + "epoch": 10.1, + "learning_rate": 1.9910333784948886e-06, + "loss": 0.4352, + "step": 27356 + }, + { + "epoch": 10.1, + "learning_rate": 1.990996428131543e-06, + "loss": 0.3396, + "step": 27357 + }, + { + "epoch": 10.1, + "learning_rate": 1.9909594777681983e-06, + "loss": 0.2751, + "step": 27358 + }, + { + "epoch": 10.1, + "learning_rate": 1.990922527404853e-06, + "loss": 0.2867, + "step": 27359 + }, + { + "epoch": 10.1, + "learning_rate": 1.990885577041508e-06, + "loss": 0.5044, + "step": 27360 + }, + { + "epoch": 10.1, + "learning_rate": 1.990848626678162e-06, + "loss": 0.3163, + "step": 27361 + }, + { + "epoch": 10.1, + "learning_rate": 1.9908116763148172e-06, + "loss": 0.2832, + "step": 27362 + }, + { + "epoch": 10.1, + "learning_rate": 1.990774725951472e-06, + "loss": 0.3196, + "step": 27363 + }, + { + "epoch": 10.1, + "learning_rate": 1.9907377755881265e-06, + "loss": 0.3102, + "step": 27364 + }, + { + "epoch": 10.11, + "learning_rate": 1.9907008252247816e-06, + "loss": 0.2688, + "step": 27365 + }, + { + "epoch": 10.11, + "learning_rate": 1.990663874861436e-06, + "loss": 0.2297, + "step": 27366 + }, + { + "epoch": 10.11, + "learning_rate": 1.990626924498091e-06, + "loss": 0.3777, + "step": 27367 + }, + { + "epoch": 10.11, + "learning_rate": 1.9905899741347455e-06, + "loss": 0.3648, + "step": 27368 + }, + { + "epoch": 10.11, + "learning_rate": 1.9905530237714005e-06, + "loss": 0.6549, + "step": 27369 + }, + { + "epoch": 10.11, + "learning_rate": 1.990516073408055e-06, + "loss": 0.2659, + "step": 27370 + }, + { + "epoch": 10.11, + "learning_rate": 1.9904791230447102e-06, + "loss": 0.4129, + "step": 27371 + }, + { + "epoch": 10.11, + "learning_rate": 1.990442172681365e-06, + "loss": 0.3991, + "step": 27372 + }, + { + "epoch": 10.11, + "learning_rate": 1.9904052223180195e-06, + "loss": 0.3534, + "step": 27373 + }, + { + "epoch": 10.11, + "learning_rate": 1.990368271954674e-06, + "loss": 0.4356, + "step": 27374 + }, + { + "epoch": 10.11, + "learning_rate": 1.9903313215913288e-06, + "loss": 0.345, + "step": 27375 + }, + { + "epoch": 10.11, + "learning_rate": 1.990294371227984e-06, + "loss": 0.1962, + "step": 27376 + }, + { + "epoch": 10.11, + "learning_rate": 1.9902574208646385e-06, + "loss": 0.3267, + "step": 27377 + }, + { + "epoch": 10.11, + "learning_rate": 1.9902204705012935e-06, + "loss": 0.4092, + "step": 27378 + }, + { + "epoch": 10.11, + "learning_rate": 1.990183520137948e-06, + "loss": 0.2597, + "step": 27379 + }, + { + "epoch": 10.11, + "learning_rate": 1.990146569774603e-06, + "loss": 0.3442, + "step": 27380 + }, + { + "epoch": 10.11, + "learning_rate": 1.9901096194112574e-06, + "loss": 0.2635, + "step": 27381 + }, + { + "epoch": 10.11, + "learning_rate": 1.9900726690479125e-06, + "loss": 0.3588, + "step": 27382 + }, + { + "epoch": 10.11, + "learning_rate": 1.990035718684567e-06, + "loss": 0.3145, + "step": 27383 + }, + { + "epoch": 10.11, + "learning_rate": 1.9899987683212218e-06, + "loss": 0.4458, + "step": 27384 + }, + { + "epoch": 10.11, + "learning_rate": 1.989961817957877e-06, + "loss": 0.3642, + "step": 27385 + }, + { + "epoch": 10.11, + "learning_rate": 1.9899248675945315e-06, + "loss": 0.3263, + "step": 27386 + }, + { + "epoch": 10.11, + "learning_rate": 1.989887917231186e-06, + "loss": 0.3407, + "step": 27387 + }, + { + "epoch": 10.11, + "learning_rate": 1.9898509668678407e-06, + "loss": 0.2085, + "step": 27388 + }, + { + "epoch": 10.11, + "learning_rate": 1.989814016504496e-06, + "loss": 0.2693, + "step": 27389 + }, + { + "epoch": 10.11, + "learning_rate": 1.9897770661411504e-06, + "loss": 0.2933, + "step": 27390 + }, + { + "epoch": 10.11, + "learning_rate": 1.9897401157778055e-06, + "loss": 0.3354, + "step": 27391 + }, + { + "epoch": 10.12, + "learning_rate": 1.98970316541446e-06, + "loss": 0.487, + "step": 27392 + }, + { + "epoch": 10.12, + "learning_rate": 1.9896662150511148e-06, + "loss": 0.3483, + "step": 27393 + }, + { + "epoch": 10.12, + "learning_rate": 1.9896292646877694e-06, + "loss": 0.2657, + "step": 27394 + }, + { + "epoch": 10.12, + "learning_rate": 1.989592314324424e-06, + "loss": 0.3491, + "step": 27395 + }, + { + "epoch": 10.12, + "learning_rate": 1.989555363961079e-06, + "loss": 0.3398, + "step": 27396 + }, + { + "epoch": 10.12, + "learning_rate": 1.9895184135977337e-06, + "loss": 0.4406, + "step": 27397 + }, + { + "epoch": 10.12, + "learning_rate": 1.989481463234389e-06, + "loss": 0.2714, + "step": 27398 + }, + { + "epoch": 10.12, + "learning_rate": 1.9894445128710434e-06, + "loss": 0.3331, + "step": 27399 + }, + { + "epoch": 10.12, + "learning_rate": 1.989407562507698e-06, + "loss": 0.388, + "step": 27400 + }, + { + "epoch": 10.12, + "learning_rate": 1.9893706121443527e-06, + "loss": 0.3041, + "step": 27401 + }, + { + "epoch": 10.12, + "learning_rate": 1.9893336617810078e-06, + "loss": 0.3817, + "step": 27402 + }, + { + "epoch": 10.12, + "learning_rate": 1.9892967114176624e-06, + "loss": 0.4028, + "step": 27403 + }, + { + "epoch": 10.12, + "learning_rate": 1.989259761054317e-06, + "loss": 0.249, + "step": 27404 + }, + { + "epoch": 10.12, + "learning_rate": 1.989222810690972e-06, + "loss": 0.2223, + "step": 27405 + }, + { + "epoch": 10.12, + "learning_rate": 1.9891858603276263e-06, + "loss": 0.2537, + "step": 27406 + }, + { + "epoch": 10.12, + "learning_rate": 1.9891489099642814e-06, + "loss": 0.4142, + "step": 27407 + }, + { + "epoch": 10.12, + "learning_rate": 1.989111959600936e-06, + "loss": 0.3736, + "step": 27408 + }, + { + "epoch": 10.12, + "learning_rate": 1.989075009237591e-06, + "loss": 0.296, + "step": 27409 + }, + { + "epoch": 10.12, + "learning_rate": 1.9890380588742457e-06, + "loss": 0.3974, + "step": 27410 + }, + { + "epoch": 10.12, + "learning_rate": 1.9890011085109008e-06, + "loss": 0.4581, + "step": 27411 + }, + { + "epoch": 10.12, + "learning_rate": 1.988964158147555e-06, + "loss": 0.3539, + "step": 27412 + }, + { + "epoch": 10.12, + "learning_rate": 1.9889272077842096e-06, + "loss": 0.3757, + "step": 27413 + }, + { + "epoch": 10.12, + "learning_rate": 1.9888902574208647e-06, + "loss": 0.3417, + "step": 27414 + }, + { + "epoch": 10.12, + "learning_rate": 1.9888533070575193e-06, + "loss": 0.3128, + "step": 27415 + }, + { + "epoch": 10.12, + "learning_rate": 1.9888163566941744e-06, + "loss": 0.3425, + "step": 27416 + }, + { + "epoch": 10.12, + "learning_rate": 1.988779406330829e-06, + "loss": 0.345, + "step": 27417 + }, + { + "epoch": 10.12, + "learning_rate": 1.9887424559674836e-06, + "loss": 0.3899, + "step": 27418 + }, + { + "epoch": 10.13, + "learning_rate": 1.9887055056041383e-06, + "loss": 0.4946, + "step": 27419 + }, + { + "epoch": 10.13, + "learning_rate": 1.9886685552407933e-06, + "loss": 0.2719, + "step": 27420 + }, + { + "epoch": 10.13, + "learning_rate": 1.988631604877448e-06, + "loss": 0.2724, + "step": 27421 + }, + { + "epoch": 10.13, + "learning_rate": 1.988594654514103e-06, + "loss": 0.3714, + "step": 27422 + }, + { + "epoch": 10.13, + "learning_rate": 1.9885577041507577e-06, + "loss": 0.3405, + "step": 27423 + }, + { + "epoch": 10.13, + "learning_rate": 1.9885207537874123e-06, + "loss": 0.284, + "step": 27424 + }, + { + "epoch": 10.13, + "learning_rate": 1.988483803424067e-06, + "loss": 0.3103, + "step": 27425 + }, + { + "epoch": 10.13, + "learning_rate": 1.9884468530607216e-06, + "loss": 0.2742, + "step": 27426 + }, + { + "epoch": 10.13, + "learning_rate": 1.9884099026973766e-06, + "loss": 0.4092, + "step": 27427 + }, + { + "epoch": 10.13, + "learning_rate": 1.9883729523340313e-06, + "loss": 0.3141, + "step": 27428 + }, + { + "epoch": 10.13, + "learning_rate": 1.9883360019706863e-06, + "loss": 0.2488, + "step": 27429 + }, + { + "epoch": 10.13, + "learning_rate": 1.988299051607341e-06, + "loss": 0.3109, + "step": 27430 + }, + { + "epoch": 10.13, + "learning_rate": 1.9882621012439956e-06, + "loss": 0.4084, + "step": 27431 + }, + { + "epoch": 10.13, + "learning_rate": 1.9882251508806502e-06, + "loss": 0.3158, + "step": 27432 + }, + { + "epoch": 10.13, + "learning_rate": 1.988188200517305e-06, + "loss": 0.3188, + "step": 27433 + }, + { + "epoch": 10.13, + "learning_rate": 1.98815125015396e-06, + "loss": 0.4491, + "step": 27434 + }, + { + "epoch": 10.13, + "learning_rate": 1.9881142997906146e-06, + "loss": 0.4039, + "step": 27435 + }, + { + "epoch": 10.13, + "learning_rate": 1.9880773494272696e-06, + "loss": 0.4226, + "step": 27436 + }, + { + "epoch": 10.13, + "learning_rate": 1.9880403990639243e-06, + "loss": 0.2473, + "step": 27437 + }, + { + "epoch": 10.13, + "learning_rate": 1.988003448700579e-06, + "loss": 0.2884, + "step": 27438 + }, + { + "epoch": 10.13, + "learning_rate": 1.9879664983372336e-06, + "loss": 0.4085, + "step": 27439 + }, + { + "epoch": 10.13, + "learning_rate": 1.9879295479738886e-06, + "loss": 0.412, + "step": 27440 + }, + { + "epoch": 10.13, + "learning_rate": 1.9878925976105432e-06, + "loss": 0.4444, + "step": 27441 + }, + { + "epoch": 10.13, + "learning_rate": 1.987855647247198e-06, + "loss": 0.4357, + "step": 27442 + }, + { + "epoch": 10.13, + "learning_rate": 1.987818696883853e-06, + "loss": 0.299, + "step": 27443 + }, + { + "epoch": 10.13, + "learning_rate": 1.9877817465205076e-06, + "loss": 0.3545, + "step": 27444 + }, + { + "epoch": 10.13, + "learning_rate": 1.9877447961571622e-06, + "loss": 0.2843, + "step": 27445 + }, + { + "epoch": 10.14, + "learning_rate": 1.987707845793817e-06, + "loss": 0.31, + "step": 27446 + }, + { + "epoch": 10.14, + "learning_rate": 1.987670895430472e-06, + "loss": 0.3714, + "step": 27447 + }, + { + "epoch": 10.14, + "learning_rate": 1.9876339450671265e-06, + "loss": 0.2154, + "step": 27448 + }, + { + "epoch": 10.14, + "learning_rate": 1.9875969947037816e-06, + "loss": 0.3343, + "step": 27449 + }, + { + "epoch": 10.14, + "learning_rate": 1.9875600443404362e-06, + "loss": 0.2794, + "step": 27450 + }, + { + "epoch": 10.14, + "learning_rate": 1.987523093977091e-06, + "loss": 0.396, + "step": 27451 + }, + { + "epoch": 10.14, + "learning_rate": 1.9874861436137455e-06, + "loss": 0.2481, + "step": 27452 + }, + { + "epoch": 10.14, + "learning_rate": 1.9874491932504e-06, + "loss": 0.2971, + "step": 27453 + }, + { + "epoch": 10.14, + "learning_rate": 1.9874122428870552e-06, + "loss": 0.3188, + "step": 27454 + }, + { + "epoch": 10.14, + "learning_rate": 1.98737529252371e-06, + "loss": 0.4496, + "step": 27455 + }, + { + "epoch": 10.14, + "learning_rate": 1.9873383421603645e-06, + "loss": 0.3408, + "step": 27456 + }, + { + "epoch": 10.14, + "learning_rate": 1.987301391797019e-06, + "loss": 0.4041, + "step": 27457 + }, + { + "epoch": 10.14, + "learning_rate": 1.987264441433674e-06, + "loss": 0.3838, + "step": 27458 + }, + { + "epoch": 10.14, + "learning_rate": 1.987227491070329e-06, + "loss": 0.445, + "step": 27459 + }, + { + "epoch": 10.14, + "learning_rate": 1.987190540706984e-06, + "loss": 0.3496, + "step": 27460 + }, + { + "epoch": 10.14, + "learning_rate": 1.9871535903436385e-06, + "loss": 0.4854, + "step": 27461 + }, + { + "epoch": 10.14, + "learning_rate": 1.987116639980293e-06, + "loss": 0.3912, + "step": 27462 + }, + { + "epoch": 10.14, + "learning_rate": 1.987079689616948e-06, + "loss": 0.3059, + "step": 27463 + }, + { + "epoch": 10.14, + "learning_rate": 1.9870427392536024e-06, + "loss": 0.3878, + "step": 27464 + }, + { + "epoch": 10.14, + "learning_rate": 1.9870057888902575e-06, + "loss": 0.3163, + "step": 27465 + }, + { + "epoch": 10.14, + "learning_rate": 1.986968838526912e-06, + "loss": 0.3534, + "step": 27466 + }, + { + "epoch": 10.14, + "learning_rate": 1.986931888163567e-06, + "loss": 0.281, + "step": 27467 + }, + { + "epoch": 10.14, + "learning_rate": 1.986894937800222e-06, + "loss": 0.2601, + "step": 27468 + }, + { + "epoch": 10.14, + "learning_rate": 1.9868579874368765e-06, + "loss": 0.2671, + "step": 27469 + }, + { + "epoch": 10.14, + "learning_rate": 1.986821037073531e-06, + "loss": 0.3644, + "step": 27470 + }, + { + "epoch": 10.14, + "learning_rate": 1.986784086710186e-06, + "loss": 0.3757, + "step": 27471 + }, + { + "epoch": 10.14, + "learning_rate": 1.9867471363468408e-06, + "loss": 0.4123, + "step": 27472 + }, + { + "epoch": 10.15, + "learning_rate": 1.9867101859834954e-06, + "loss": 0.3479, + "step": 27473 + }, + { + "epoch": 10.15, + "learning_rate": 1.9866732356201505e-06, + "loss": 0.3359, + "step": 27474 + }, + { + "epoch": 10.15, + "learning_rate": 1.986636285256805e-06, + "loss": 0.3339, + "step": 27475 + }, + { + "epoch": 10.15, + "learning_rate": 1.9865993348934598e-06, + "loss": 0.3734, + "step": 27476 + }, + { + "epoch": 10.15, + "learning_rate": 1.9865623845301144e-06, + "loss": 0.4498, + "step": 27477 + }, + { + "epoch": 10.15, + "learning_rate": 1.9865254341667695e-06, + "loss": 0.2952, + "step": 27478 + }, + { + "epoch": 10.15, + "learning_rate": 1.986488483803424e-06, + "loss": 0.2968, + "step": 27479 + }, + { + "epoch": 10.15, + "learning_rate": 1.986451533440079e-06, + "loss": 0.4905, + "step": 27480 + }, + { + "epoch": 10.15, + "learning_rate": 1.9864145830767338e-06, + "loss": 0.2456, + "step": 27481 + }, + { + "epoch": 10.15, + "learning_rate": 1.9863776327133884e-06, + "loss": 0.3842, + "step": 27482 + }, + { + "epoch": 10.15, + "learning_rate": 1.986340682350043e-06, + "loss": 0.4007, + "step": 27483 + }, + { + "epoch": 10.15, + "learning_rate": 1.9863037319866977e-06, + "loss": 0.3079, + "step": 27484 + }, + { + "epoch": 10.15, + "learning_rate": 1.9862667816233528e-06, + "loss": 0.332, + "step": 27485 + }, + { + "epoch": 10.15, + "learning_rate": 1.9862298312600074e-06, + "loss": 0.3281, + "step": 27486 + }, + { + "epoch": 10.15, + "learning_rate": 1.9861928808966625e-06, + "loss": 0.4294, + "step": 27487 + }, + { + "epoch": 10.15, + "learning_rate": 1.986155930533317e-06, + "loss": 0.3302, + "step": 27488 + }, + { + "epoch": 10.15, + "learning_rate": 1.9861189801699717e-06, + "loss": 0.3734, + "step": 27489 + }, + { + "epoch": 10.15, + "learning_rate": 1.9860820298066264e-06, + "loss": 0.3937, + "step": 27490 + }, + { + "epoch": 10.15, + "learning_rate": 1.9860450794432814e-06, + "loss": 0.4236, + "step": 27491 + }, + { + "epoch": 10.15, + "learning_rate": 1.986008129079936e-06, + "loss": 0.293, + "step": 27492 + }, + { + "epoch": 10.15, + "learning_rate": 1.9859711787165907e-06, + "loss": 0.3548, + "step": 27493 + }, + { + "epoch": 10.15, + "learning_rate": 1.9859342283532458e-06, + "loss": 0.3276, + "step": 27494 + }, + { + "epoch": 10.15, + "learning_rate": 1.9858972779899e-06, + "loss": 0.4481, + "step": 27495 + }, + { + "epoch": 10.15, + "learning_rate": 1.985860327626555e-06, + "loss": 0.3953, + "step": 27496 + }, + { + "epoch": 10.15, + "learning_rate": 1.9858233772632097e-06, + "loss": 0.2999, + "step": 27497 + }, + { + "epoch": 10.15, + "learning_rate": 1.9857864268998647e-06, + "loss": 0.2844, + "step": 27498 + }, + { + "epoch": 10.15, + "learning_rate": 1.9857494765365194e-06, + "loss": 0.4553, + "step": 27499 + }, + { + "epoch": 10.16, + "learning_rate": 1.9857125261731744e-06, + "loss": 0.2755, + "step": 27500 + }, + { + "epoch": 10.16, + "learning_rate": 1.9856755758098286e-06, + "loss": 0.2999, + "step": 27501 + }, + { + "epoch": 10.16, + "learning_rate": 1.9856386254464833e-06, + "loss": 0.329, + "step": 27502 + }, + { + "epoch": 10.16, + "learning_rate": 1.9856016750831383e-06, + "loss": 0.2295, + "step": 27503 + }, + { + "epoch": 10.16, + "learning_rate": 1.985564724719793e-06, + "loss": 0.309, + "step": 27504 + }, + { + "epoch": 10.16, + "learning_rate": 1.985527774356448e-06, + "loss": 0.415, + "step": 27505 + }, + { + "epoch": 10.16, + "learning_rate": 1.9854908239931027e-06, + "loss": 0.3858, + "step": 27506 + }, + { + "epoch": 10.16, + "learning_rate": 1.9854538736297573e-06, + "loss": 0.4684, + "step": 27507 + }, + { + "epoch": 10.16, + "learning_rate": 1.985416923266412e-06, + "loss": 0.2313, + "step": 27508 + }, + { + "epoch": 10.16, + "learning_rate": 1.985379972903067e-06, + "loss": 0.3414, + "step": 27509 + }, + { + "epoch": 10.16, + "learning_rate": 1.9853430225397216e-06, + "loss": 0.2982, + "step": 27510 + }, + { + "epoch": 10.16, + "learning_rate": 1.9853060721763767e-06, + "loss": 0.3979, + "step": 27511 + }, + { + "epoch": 10.16, + "learning_rate": 1.9852691218130313e-06, + "loss": 0.3862, + "step": 27512 + }, + { + "epoch": 10.16, + "learning_rate": 1.985232171449686e-06, + "loss": 0.3269, + "step": 27513 + }, + { + "epoch": 10.16, + "learning_rate": 1.9851952210863406e-06, + "loss": 0.347, + "step": 27514 + }, + { + "epoch": 10.16, + "learning_rate": 1.9851582707229952e-06, + "loss": 0.2666, + "step": 27515 + }, + { + "epoch": 10.16, + "learning_rate": 1.9851213203596503e-06, + "loss": 0.4727, + "step": 27516 + }, + { + "epoch": 10.16, + "learning_rate": 1.985084369996305e-06, + "loss": 0.4306, + "step": 27517 + }, + { + "epoch": 10.16, + "learning_rate": 1.98504741963296e-06, + "loss": 0.2644, + "step": 27518 + }, + { + "epoch": 10.16, + "learning_rate": 1.9850104692696146e-06, + "loss": 0.3139, + "step": 27519 + }, + { + "epoch": 10.16, + "learning_rate": 1.9849735189062693e-06, + "loss": 0.2949, + "step": 27520 + }, + { + "epoch": 10.16, + "learning_rate": 1.984936568542924e-06, + "loss": 0.4112, + "step": 27521 + }, + { + "epoch": 10.16, + "learning_rate": 1.9848996181795785e-06, + "loss": 0.4599, + "step": 27522 + }, + { + "epoch": 10.16, + "learning_rate": 1.9848626678162336e-06, + "loss": 0.259, + "step": 27523 + }, + { + "epoch": 10.16, + "learning_rate": 1.9848257174528882e-06, + "loss": 0.3102, + "step": 27524 + }, + { + "epoch": 10.16, + "learning_rate": 1.9847887670895433e-06, + "loss": 0.3114, + "step": 27525 + }, + { + "epoch": 10.16, + "learning_rate": 1.984751816726198e-06, + "loss": 0.2892, + "step": 27526 + }, + { + "epoch": 10.17, + "learning_rate": 1.9847148663628526e-06, + "loss": 0.3029, + "step": 27527 + }, + { + "epoch": 10.17, + "learning_rate": 1.984677915999507e-06, + "loss": 0.4406, + "step": 27528 + }, + { + "epoch": 10.17, + "learning_rate": 1.9846409656361623e-06, + "loss": 0.4291, + "step": 27529 + }, + { + "epoch": 10.17, + "learning_rate": 1.984604015272817e-06, + "loss": 0.2604, + "step": 27530 + }, + { + "epoch": 10.17, + "learning_rate": 1.984567064909472e-06, + "loss": 0.2957, + "step": 27531 + }, + { + "epoch": 10.17, + "learning_rate": 1.9845301145461266e-06, + "loss": 0.2902, + "step": 27532 + }, + { + "epoch": 10.17, + "learning_rate": 1.9844931641827812e-06, + "loss": 0.3733, + "step": 27533 + }, + { + "epoch": 10.17, + "learning_rate": 1.984456213819436e-06, + "loss": 0.3046, + "step": 27534 + }, + { + "epoch": 10.17, + "learning_rate": 1.9844192634560905e-06, + "loss": 0.361, + "step": 27535 + }, + { + "epoch": 10.17, + "learning_rate": 1.9843823130927456e-06, + "loss": 0.3756, + "step": 27536 + }, + { + "epoch": 10.17, + "learning_rate": 1.9843453627294e-06, + "loss": 0.4113, + "step": 27537 + }, + { + "epoch": 10.17, + "learning_rate": 1.9843084123660553e-06, + "loss": 0.2716, + "step": 27538 + }, + { + "epoch": 10.17, + "learning_rate": 1.98427146200271e-06, + "loss": 0.3067, + "step": 27539 + }, + { + "epoch": 10.17, + "learning_rate": 1.9842345116393645e-06, + "loss": 0.3643, + "step": 27540 + }, + { + "epoch": 10.17, + "learning_rate": 1.984197561276019e-06, + "loss": 0.3059, + "step": 27541 + }, + { + "epoch": 10.17, + "learning_rate": 1.984160610912674e-06, + "loss": 0.3653, + "step": 27542 + }, + { + "epoch": 10.17, + "learning_rate": 1.984123660549329e-06, + "loss": 0.1618, + "step": 27543 + }, + { + "epoch": 10.17, + "learning_rate": 1.9840867101859835e-06, + "loss": 0.3083, + "step": 27544 + }, + { + "epoch": 10.17, + "learning_rate": 1.9840497598226386e-06, + "loss": 0.3408, + "step": 27545 + }, + { + "epoch": 10.17, + "learning_rate": 1.9840128094592928e-06, + "loss": 0.3096, + "step": 27546 + }, + { + "epoch": 10.17, + "learning_rate": 1.983975859095948e-06, + "loss": 0.3044, + "step": 27547 + }, + { + "epoch": 10.17, + "learning_rate": 1.9839389087326025e-06, + "loss": 0.2986, + "step": 27548 + }, + { + "epoch": 10.17, + "learning_rate": 1.9839019583692575e-06, + "loss": 0.3584, + "step": 27549 + }, + { + "epoch": 10.17, + "learning_rate": 1.983865008005912e-06, + "loss": 0.305, + "step": 27550 + }, + { + "epoch": 10.17, + "learning_rate": 1.9838280576425672e-06, + "loss": 0.2721, + "step": 27551 + }, + { + "epoch": 10.17, + "learning_rate": 1.9837911072792214e-06, + "loss": 0.4404, + "step": 27552 + }, + { + "epoch": 10.17, + "learning_rate": 1.983754156915876e-06, + "loss": 0.2992, + "step": 27553 + }, + { + "epoch": 10.18, + "learning_rate": 1.983717206552531e-06, + "loss": 0.29, + "step": 27554 + }, + { + "epoch": 10.18, + "learning_rate": 1.9836802561891858e-06, + "loss": 0.2837, + "step": 27555 + }, + { + "epoch": 10.18, + "learning_rate": 1.983643305825841e-06, + "loss": 0.3495, + "step": 27556 + }, + { + "epoch": 10.18, + "learning_rate": 1.9836063554624955e-06, + "loss": 0.4073, + "step": 27557 + }, + { + "epoch": 10.18, + "learning_rate": 1.98356940509915e-06, + "loss": 0.2872, + "step": 27558 + }, + { + "epoch": 10.18, + "learning_rate": 1.9835324547358047e-06, + "loss": 0.2629, + "step": 27559 + }, + { + "epoch": 10.18, + "learning_rate": 1.98349550437246e-06, + "loss": 0.3746, + "step": 27560 + }, + { + "epoch": 10.18, + "learning_rate": 1.9834585540091144e-06, + "loss": 0.2928, + "step": 27561 + }, + { + "epoch": 10.18, + "learning_rate": 1.983421603645769e-06, + "loss": 0.2829, + "step": 27562 + }, + { + "epoch": 10.18, + "learning_rate": 1.983384653282424e-06, + "loss": 0.211, + "step": 27563 + }, + { + "epoch": 10.18, + "learning_rate": 1.9833477029190788e-06, + "loss": 0.3388, + "step": 27564 + }, + { + "epoch": 10.18, + "learning_rate": 1.9833107525557334e-06, + "loss": 0.361, + "step": 27565 + }, + { + "epoch": 10.18, + "learning_rate": 1.983273802192388e-06, + "loss": 0.2444, + "step": 27566 + }, + { + "epoch": 10.18, + "learning_rate": 1.983236851829043e-06, + "loss": 0.2357, + "step": 27567 + }, + { + "epoch": 10.18, + "learning_rate": 1.9831999014656977e-06, + "loss": 0.4101, + "step": 27568 + }, + { + "epoch": 10.18, + "learning_rate": 1.983162951102353e-06, + "loss": 0.3438, + "step": 27569 + }, + { + "epoch": 10.18, + "learning_rate": 1.9831260007390074e-06, + "loss": 0.3137, + "step": 27570 + }, + { + "epoch": 10.18, + "learning_rate": 1.983089050375662e-06, + "loss": 0.3275, + "step": 27571 + }, + { + "epoch": 10.18, + "learning_rate": 1.9830521000123167e-06, + "loss": 0.2601, + "step": 27572 + }, + { + "epoch": 10.18, + "learning_rate": 1.9830151496489713e-06, + "loss": 0.283, + "step": 27573 + }, + { + "epoch": 10.18, + "learning_rate": 1.9829781992856264e-06, + "loss": 0.2524, + "step": 27574 + }, + { + "epoch": 10.18, + "learning_rate": 1.982941248922281e-06, + "loss": 0.3836, + "step": 27575 + }, + { + "epoch": 10.18, + "learning_rate": 1.982904298558936e-06, + "loss": 0.2249, + "step": 27576 + }, + { + "epoch": 10.18, + "learning_rate": 1.9828673481955907e-06, + "loss": 0.4368, + "step": 27577 + }, + { + "epoch": 10.18, + "learning_rate": 1.9828303978322454e-06, + "loss": 0.4046, + "step": 27578 + }, + { + "epoch": 10.18, + "learning_rate": 1.9827934474689e-06, + "loss": 0.4228, + "step": 27579 + }, + { + "epoch": 10.18, + "learning_rate": 1.982756497105555e-06, + "loss": 0.3146, + "step": 27580 + }, + { + "epoch": 10.19, + "learning_rate": 1.9827195467422097e-06, + "loss": 0.3548, + "step": 27581 + }, + { + "epoch": 10.19, + "learning_rate": 1.9826825963788643e-06, + "loss": 0.4851, + "step": 27582 + }, + { + "epoch": 10.19, + "learning_rate": 1.9826456460155194e-06, + "loss": 0.4591, + "step": 27583 + }, + { + "epoch": 10.19, + "learning_rate": 1.982608695652174e-06, + "loss": 0.374, + "step": 27584 + }, + { + "epoch": 10.19, + "learning_rate": 1.9825717452888287e-06, + "loss": 0.4379, + "step": 27585 + }, + { + "epoch": 10.19, + "learning_rate": 1.9825347949254833e-06, + "loss": 0.2412, + "step": 27586 + }, + { + "epoch": 10.19, + "learning_rate": 1.9824978445621384e-06, + "loss": 0.3257, + "step": 27587 + }, + { + "epoch": 10.19, + "learning_rate": 1.982460894198793e-06, + "loss": 0.3986, + "step": 27588 + }, + { + "epoch": 10.19, + "learning_rate": 1.982423943835448e-06, + "loss": 0.3615, + "step": 27589 + }, + { + "epoch": 10.19, + "learning_rate": 1.9823869934721027e-06, + "loss": 0.2951, + "step": 27590 + }, + { + "epoch": 10.19, + "learning_rate": 1.9823500431087573e-06, + "loss": 0.2105, + "step": 27591 + }, + { + "epoch": 10.19, + "learning_rate": 1.982313092745412e-06, + "loss": 0.2791, + "step": 27592 + }, + { + "epoch": 10.19, + "learning_rate": 1.9822761423820666e-06, + "loss": 0.3521, + "step": 27593 + }, + { + "epoch": 10.19, + "learning_rate": 1.9822391920187217e-06, + "loss": 0.3435, + "step": 27594 + }, + { + "epoch": 10.19, + "learning_rate": 1.9822022416553763e-06, + "loss": 0.3521, + "step": 27595 + }, + { + "epoch": 10.19, + "learning_rate": 1.9821652912920314e-06, + "loss": 0.337, + "step": 27596 + }, + { + "epoch": 10.19, + "learning_rate": 1.9821283409286856e-06, + "loss": 0.4264, + "step": 27597 + }, + { + "epoch": 10.19, + "learning_rate": 1.9820913905653406e-06, + "loss": 0.2449, + "step": 27598 + }, + { + "epoch": 10.19, + "learning_rate": 1.9820544402019953e-06, + "loss": 0.3571, + "step": 27599 + }, + { + "epoch": 10.19, + "learning_rate": 1.9820174898386503e-06, + "loss": 0.2401, + "step": 27600 + }, + { + "epoch": 10.19, + "learning_rate": 1.981980539475305e-06, + "loss": 0.3767, + "step": 27601 + }, + { + "epoch": 10.19, + "learning_rate": 1.9819435891119596e-06, + "loss": 0.4249, + "step": 27602 + }, + { + "epoch": 10.19, + "learning_rate": 1.9819066387486143e-06, + "loss": 0.4234, + "step": 27603 + }, + { + "epoch": 10.19, + "learning_rate": 1.981869688385269e-06, + "loss": 0.3319, + "step": 27604 + }, + { + "epoch": 10.19, + "learning_rate": 1.981832738021924e-06, + "loss": 0.2844, + "step": 27605 + }, + { + "epoch": 10.19, + "learning_rate": 1.9817957876585786e-06, + "loss": 0.3308, + "step": 27606 + }, + { + "epoch": 10.19, + "learning_rate": 1.9817588372952336e-06, + "loss": 0.3539, + "step": 27607 + }, + { + "epoch": 10.19, + "learning_rate": 1.9817218869318883e-06, + "loss": 0.3606, + "step": 27608 + }, + { + "epoch": 10.2, + "learning_rate": 1.981684936568543e-06, + "loss": 0.256, + "step": 27609 + }, + { + "epoch": 10.2, + "learning_rate": 1.9816479862051976e-06, + "loss": 0.4295, + "step": 27610 + }, + { + "epoch": 10.2, + "learning_rate": 1.9816110358418526e-06, + "loss": 0.4574, + "step": 27611 + }, + { + "epoch": 10.2, + "learning_rate": 1.9815740854785073e-06, + "loss": 0.391, + "step": 27612 + }, + { + "epoch": 10.2, + "learning_rate": 1.981537135115162e-06, + "loss": 0.3864, + "step": 27613 + }, + { + "epoch": 10.2, + "learning_rate": 1.981500184751817e-06, + "loss": 0.2832, + "step": 27614 + }, + { + "epoch": 10.2, + "learning_rate": 1.9814632343884716e-06, + "loss": 0.3453, + "step": 27615 + }, + { + "epoch": 10.2, + "learning_rate": 1.9814262840251262e-06, + "loss": 0.3096, + "step": 27616 + }, + { + "epoch": 10.2, + "learning_rate": 1.981389333661781e-06, + "loss": 0.2919, + "step": 27617 + }, + { + "epoch": 10.2, + "learning_rate": 1.981352383298436e-06, + "loss": 0.3322, + "step": 27618 + }, + { + "epoch": 10.2, + "learning_rate": 1.9813154329350906e-06, + "loss": 0.327, + "step": 27619 + }, + { + "epoch": 10.2, + "learning_rate": 1.9812784825717456e-06, + "loss": 0.456, + "step": 27620 + }, + { + "epoch": 10.2, + "learning_rate": 1.9812415322084003e-06, + "loss": 0.3429, + "step": 27621 + }, + { + "epoch": 10.2, + "learning_rate": 1.981204581845055e-06, + "loss": 0.4387, + "step": 27622 + }, + { + "epoch": 10.2, + "learning_rate": 1.9811676314817095e-06, + "loss": 0.4493, + "step": 27623 + }, + { + "epoch": 10.2, + "learning_rate": 1.981130681118364e-06, + "loss": 0.3327, + "step": 27624 + }, + { + "epoch": 10.2, + "learning_rate": 1.9810937307550192e-06, + "loss": 0.3204, + "step": 27625 + }, + { + "epoch": 10.2, + "learning_rate": 1.981056780391674e-06, + "loss": 0.3656, + "step": 27626 + }, + { + "epoch": 10.2, + "learning_rate": 1.981019830028329e-06, + "loss": 0.3295, + "step": 27627 + }, + { + "epoch": 10.2, + "learning_rate": 1.9809828796649836e-06, + "loss": 0.3256, + "step": 27628 + }, + { + "epoch": 10.2, + "learning_rate": 1.980945929301638e-06, + "loss": 0.3167, + "step": 27629 + }, + { + "epoch": 10.2, + "learning_rate": 1.980908978938293e-06, + "loss": 0.3169, + "step": 27630 + }, + { + "epoch": 10.2, + "learning_rate": 1.980872028574948e-06, + "loss": 0.2739, + "step": 27631 + }, + { + "epoch": 10.2, + "learning_rate": 1.9808350782116025e-06, + "loss": 0.378, + "step": 27632 + }, + { + "epoch": 10.2, + "learning_rate": 1.980798127848257e-06, + "loss": 0.2321, + "step": 27633 + }, + { + "epoch": 10.2, + "learning_rate": 1.9807611774849122e-06, + "loss": 0.3114, + "step": 27634 + }, + { + "epoch": 10.2, + "learning_rate": 1.980724227121567e-06, + "loss": 0.3341, + "step": 27635 + }, + { + "epoch": 10.21, + "learning_rate": 1.9806872767582215e-06, + "loss": 0.3011, + "step": 27636 + }, + { + "epoch": 10.21, + "learning_rate": 1.980650326394876e-06, + "loss": 0.4156, + "step": 27637 + }, + { + "epoch": 10.21, + "learning_rate": 1.980613376031531e-06, + "loss": 0.3834, + "step": 27638 + }, + { + "epoch": 10.21, + "learning_rate": 1.980576425668186e-06, + "loss": 0.2879, + "step": 27639 + }, + { + "epoch": 10.21, + "learning_rate": 1.980539475304841e-06, + "loss": 0.2071, + "step": 27640 + }, + { + "epoch": 10.21, + "learning_rate": 1.980502524941495e-06, + "loss": 0.4112, + "step": 27641 + }, + { + "epoch": 10.21, + "learning_rate": 1.9804655745781497e-06, + "loss": 0.3839, + "step": 27642 + }, + { + "epoch": 10.21, + "learning_rate": 1.980428624214805e-06, + "loss": 0.3994, + "step": 27643 + }, + { + "epoch": 10.21, + "learning_rate": 1.9803916738514594e-06, + "loss": 0.3501, + "step": 27644 + }, + { + "epoch": 10.21, + "learning_rate": 1.9803547234881145e-06, + "loss": 0.4118, + "step": 27645 + }, + { + "epoch": 10.21, + "learning_rate": 1.980317773124769e-06, + "loss": 0.3151, + "step": 27646 + }, + { + "epoch": 10.21, + "learning_rate": 1.9802808227614238e-06, + "loss": 0.4065, + "step": 27647 + }, + { + "epoch": 10.21, + "learning_rate": 1.9802438723980784e-06, + "loss": 0.3836, + "step": 27648 + }, + { + "epoch": 10.21, + "learning_rate": 1.9802069220347335e-06, + "loss": 0.2372, + "step": 27649 + }, + { + "epoch": 10.21, + "learning_rate": 1.980169971671388e-06, + "loss": 0.3897, + "step": 27650 + }, + { + "epoch": 10.21, + "learning_rate": 1.980133021308043e-06, + "loss": 0.3165, + "step": 27651 + }, + { + "epoch": 10.21, + "learning_rate": 1.980096070944698e-06, + "loss": 0.3219, + "step": 27652 + }, + { + "epoch": 10.21, + "learning_rate": 1.9800591205813524e-06, + "loss": 0.265, + "step": 27653 + }, + { + "epoch": 10.21, + "learning_rate": 1.980022170218007e-06, + "loss": 0.3857, + "step": 27654 + }, + { + "epoch": 10.21, + "learning_rate": 1.9799852198546617e-06, + "loss": 0.3382, + "step": 27655 + }, + { + "epoch": 10.21, + "learning_rate": 1.9799482694913168e-06, + "loss": 0.4413, + "step": 27656 + }, + { + "epoch": 10.21, + "learning_rate": 1.9799113191279714e-06, + "loss": 0.2838, + "step": 27657 + }, + { + "epoch": 10.21, + "learning_rate": 1.9798743687646265e-06, + "loss": 0.297, + "step": 27658 + }, + { + "epoch": 10.21, + "learning_rate": 1.979837418401281e-06, + "loss": 0.436, + "step": 27659 + }, + { + "epoch": 10.21, + "learning_rate": 1.9798004680379357e-06, + "loss": 0.2157, + "step": 27660 + }, + { + "epoch": 10.21, + "learning_rate": 1.9797635176745904e-06, + "loss": 0.3083, + "step": 27661 + }, + { + "epoch": 10.21, + "learning_rate": 1.979726567311245e-06, + "loss": 0.3163, + "step": 27662 + }, + { + "epoch": 10.22, + "learning_rate": 1.9796896169479e-06, + "loss": 0.3538, + "step": 27663 + }, + { + "epoch": 10.22, + "learning_rate": 1.9796526665845547e-06, + "loss": 0.3294, + "step": 27664 + }, + { + "epoch": 10.22, + "learning_rate": 1.9796157162212098e-06, + "loss": 0.3446, + "step": 27665 + }, + { + "epoch": 10.22, + "learning_rate": 1.9795787658578644e-06, + "loss": 0.3625, + "step": 27666 + }, + { + "epoch": 10.22, + "learning_rate": 1.979541815494519e-06, + "loss": 0.3111, + "step": 27667 + }, + { + "epoch": 10.22, + "learning_rate": 1.9795048651311737e-06, + "loss": 0.2365, + "step": 27668 + }, + { + "epoch": 10.22, + "learning_rate": 1.9794679147678287e-06, + "loss": 0.2506, + "step": 27669 + }, + { + "epoch": 10.22, + "learning_rate": 1.9794309644044834e-06, + "loss": 0.2336, + "step": 27670 + }, + { + "epoch": 10.22, + "learning_rate": 1.979394014041138e-06, + "loss": 0.331, + "step": 27671 + }, + { + "epoch": 10.22, + "learning_rate": 1.979357063677793e-06, + "loss": 0.2223, + "step": 27672 + }, + { + "epoch": 10.22, + "learning_rate": 1.9793201133144477e-06, + "loss": 0.4883, + "step": 27673 + }, + { + "epoch": 10.22, + "learning_rate": 1.9792831629511023e-06, + "loss": 0.403, + "step": 27674 + }, + { + "epoch": 10.22, + "learning_rate": 1.979246212587757e-06, + "loss": 0.3329, + "step": 27675 + }, + { + "epoch": 10.22, + "learning_rate": 1.979209262224412e-06, + "loss": 0.3596, + "step": 27676 + }, + { + "epoch": 10.22, + "learning_rate": 1.9791723118610667e-06, + "loss": 0.425, + "step": 27677 + }, + { + "epoch": 10.22, + "learning_rate": 1.9791353614977217e-06, + "loss": 0.2934, + "step": 27678 + }, + { + "epoch": 10.22, + "learning_rate": 1.9790984111343764e-06, + "loss": 0.3063, + "step": 27679 + }, + { + "epoch": 10.22, + "learning_rate": 1.979061460771031e-06, + "loss": 0.3764, + "step": 27680 + }, + { + "epoch": 10.22, + "learning_rate": 1.9790245104076856e-06, + "loss": 0.4411, + "step": 27681 + }, + { + "epoch": 10.22, + "learning_rate": 1.9789875600443403e-06, + "loss": 0.3216, + "step": 27682 + }, + { + "epoch": 10.22, + "learning_rate": 1.9789506096809953e-06, + "loss": 0.4379, + "step": 27683 + }, + { + "epoch": 10.22, + "learning_rate": 1.97891365931765e-06, + "loss": 0.448, + "step": 27684 + }, + { + "epoch": 10.22, + "learning_rate": 1.978876708954305e-06, + "loss": 0.3075, + "step": 27685 + }, + { + "epoch": 10.22, + "learning_rate": 1.9788397585909592e-06, + "loss": 0.3396, + "step": 27686 + }, + { + "epoch": 10.22, + "learning_rate": 1.9788028082276143e-06, + "loss": 0.4051, + "step": 27687 + }, + { + "epoch": 10.22, + "learning_rate": 1.978765857864269e-06, + "loss": 0.2817, + "step": 27688 + }, + { + "epoch": 10.22, + "learning_rate": 1.978728907500924e-06, + "loss": 0.2217, + "step": 27689 + }, + { + "epoch": 10.23, + "learning_rate": 1.9786919571375786e-06, + "loss": 0.4098, + "step": 27690 + }, + { + "epoch": 10.23, + "learning_rate": 1.9786550067742333e-06, + "loss": 0.3328, + "step": 27691 + }, + { + "epoch": 10.23, + "learning_rate": 1.978618056410888e-06, + "loss": 0.35, + "step": 27692 + }, + { + "epoch": 10.23, + "learning_rate": 1.9785811060475425e-06, + "loss": 0.439, + "step": 27693 + }, + { + "epoch": 10.23, + "learning_rate": 1.9785441556841976e-06, + "loss": 0.5025, + "step": 27694 + }, + { + "epoch": 10.23, + "learning_rate": 1.9785072053208522e-06, + "loss": 0.3637, + "step": 27695 + }, + { + "epoch": 10.23, + "learning_rate": 1.9784702549575073e-06, + "loss": 0.4216, + "step": 27696 + }, + { + "epoch": 10.23, + "learning_rate": 1.978433304594162e-06, + "loss": 0.285, + "step": 27697 + }, + { + "epoch": 10.23, + "learning_rate": 1.9783963542308166e-06, + "loss": 0.5849, + "step": 27698 + }, + { + "epoch": 10.23, + "learning_rate": 1.978359403867471e-06, + "loss": 0.3397, + "step": 27699 + }, + { + "epoch": 10.23, + "learning_rate": 1.9783224535041263e-06, + "loss": 0.3556, + "step": 27700 + }, + { + "epoch": 10.23, + "learning_rate": 1.978285503140781e-06, + "loss": 0.3784, + "step": 27701 + }, + { + "epoch": 10.23, + "learning_rate": 1.9782485527774355e-06, + "loss": 0.2735, + "step": 27702 + }, + { + "epoch": 10.23, + "learning_rate": 1.9782116024140906e-06, + "loss": 0.323, + "step": 27703 + }, + { + "epoch": 10.23, + "learning_rate": 1.9781746520507452e-06, + "loss": 0.4456, + "step": 27704 + }, + { + "epoch": 10.23, + "learning_rate": 1.9781377016874e-06, + "loss": 0.323, + "step": 27705 + }, + { + "epoch": 10.23, + "learning_rate": 1.9781007513240545e-06, + "loss": 0.3758, + "step": 27706 + }, + { + "epoch": 10.23, + "learning_rate": 1.9780638009607096e-06, + "loss": 0.3591, + "step": 27707 + }, + { + "epoch": 10.23, + "learning_rate": 1.978026850597364e-06, + "loss": 0.3075, + "step": 27708 + }, + { + "epoch": 10.23, + "learning_rate": 1.9779899002340193e-06, + "loss": 0.4312, + "step": 27709 + }, + { + "epoch": 10.23, + "learning_rate": 1.977952949870674e-06, + "loss": 0.3732, + "step": 27710 + }, + { + "epoch": 10.23, + "learning_rate": 1.9779159995073285e-06, + "loss": 0.4161, + "step": 27711 + }, + { + "epoch": 10.23, + "learning_rate": 1.977879049143983e-06, + "loss": 0.4086, + "step": 27712 + }, + { + "epoch": 10.23, + "learning_rate": 1.977842098780638e-06, + "loss": 0.3052, + "step": 27713 + }, + { + "epoch": 10.23, + "learning_rate": 1.977805148417293e-06, + "loss": 0.3432, + "step": 27714 + }, + { + "epoch": 10.23, + "learning_rate": 1.9777681980539475e-06, + "loss": 0.3919, + "step": 27715 + }, + { + "epoch": 10.23, + "learning_rate": 1.9777312476906026e-06, + "loss": 0.3881, + "step": 27716 + }, + { + "epoch": 10.24, + "learning_rate": 1.977694297327257e-06, + "loss": 0.3594, + "step": 27717 + }, + { + "epoch": 10.24, + "learning_rate": 1.977657346963912e-06, + "loss": 0.2629, + "step": 27718 + }, + { + "epoch": 10.24, + "learning_rate": 1.9776203966005665e-06, + "loss": 0.3616, + "step": 27719 + }, + { + "epoch": 10.24, + "learning_rate": 1.9775834462372215e-06, + "loss": 0.3757, + "step": 27720 + }, + { + "epoch": 10.24, + "learning_rate": 1.977546495873876e-06, + "loss": 0.3374, + "step": 27721 + }, + { + "epoch": 10.24, + "learning_rate": 1.977509545510531e-06, + "loss": 0.2826, + "step": 27722 + }, + { + "epoch": 10.24, + "learning_rate": 1.977472595147186e-06, + "loss": 0.3584, + "step": 27723 + }, + { + "epoch": 10.24, + "learning_rate": 1.9774356447838405e-06, + "loss": 0.3363, + "step": 27724 + }, + { + "epoch": 10.24, + "learning_rate": 1.977398694420495e-06, + "loss": 0.2675, + "step": 27725 + }, + { + "epoch": 10.24, + "learning_rate": 1.9773617440571498e-06, + "loss": 0.3744, + "step": 27726 + }, + { + "epoch": 10.24, + "learning_rate": 1.977324793693805e-06, + "loss": 0.3217, + "step": 27727 + }, + { + "epoch": 10.24, + "learning_rate": 1.9772878433304595e-06, + "loss": 0.4283, + "step": 27728 + }, + { + "epoch": 10.24, + "learning_rate": 1.9772508929671145e-06, + "loss": 0.3496, + "step": 27729 + }, + { + "epoch": 10.24, + "learning_rate": 1.977213942603769e-06, + "loss": 0.3204, + "step": 27730 + }, + { + "epoch": 10.24, + "learning_rate": 1.9771769922404234e-06, + "loss": 0.3764, + "step": 27731 + }, + { + "epoch": 10.24, + "learning_rate": 1.9771400418770784e-06, + "loss": 0.4147, + "step": 27732 + }, + { + "epoch": 10.24, + "learning_rate": 1.977103091513733e-06, + "loss": 0.411, + "step": 27733 + }, + { + "epoch": 10.24, + "learning_rate": 1.977066141150388e-06, + "loss": 0.2721, + "step": 27734 + }, + { + "epoch": 10.24, + "learning_rate": 1.9770291907870428e-06, + "loss": 0.3456, + "step": 27735 + }, + { + "epoch": 10.24, + "learning_rate": 1.976992240423698e-06, + "loss": 0.3711, + "step": 27736 + }, + { + "epoch": 10.24, + "learning_rate": 1.976955290060352e-06, + "loss": 0.2378, + "step": 27737 + }, + { + "epoch": 10.24, + "learning_rate": 1.976918339697007e-06, + "loss": 0.3182, + "step": 27738 + }, + { + "epoch": 10.24, + "learning_rate": 1.9768813893336617e-06, + "loss": 0.3082, + "step": 27739 + }, + { + "epoch": 10.24, + "learning_rate": 1.976844438970317e-06, + "loss": 0.2829, + "step": 27740 + }, + { + "epoch": 10.24, + "learning_rate": 1.9768074886069714e-06, + "loss": 0.3429, + "step": 27741 + }, + { + "epoch": 10.24, + "learning_rate": 1.976770538243626e-06, + "loss": 0.3688, + "step": 27742 + }, + { + "epoch": 10.24, + "learning_rate": 1.9767335878802807e-06, + "loss": 0.2714, + "step": 27743 + }, + { + "epoch": 10.25, + "learning_rate": 1.9766966375169354e-06, + "loss": 0.3741, + "step": 27744 + }, + { + "epoch": 10.25, + "learning_rate": 1.9766596871535904e-06, + "loss": 0.2713, + "step": 27745 + }, + { + "epoch": 10.25, + "learning_rate": 1.976622736790245e-06, + "loss": 0.3913, + "step": 27746 + }, + { + "epoch": 10.25, + "learning_rate": 1.9765857864269e-06, + "loss": 0.369, + "step": 27747 + }, + { + "epoch": 10.25, + "learning_rate": 1.9765488360635547e-06, + "loss": 0.3121, + "step": 27748 + }, + { + "epoch": 10.25, + "learning_rate": 1.9765118857002094e-06, + "loss": 0.2978, + "step": 27749 + }, + { + "epoch": 10.25, + "learning_rate": 1.976474935336864e-06, + "loss": 0.276, + "step": 27750 + }, + { + "epoch": 10.25, + "learning_rate": 1.9764379849735187e-06, + "loss": 0.4215, + "step": 27751 + }, + { + "epoch": 10.25, + "learning_rate": 1.9764010346101737e-06, + "loss": 0.4294, + "step": 27752 + }, + { + "epoch": 10.25, + "learning_rate": 1.9763640842468284e-06, + "loss": 0.4966, + "step": 27753 + }, + { + "epoch": 10.25, + "learning_rate": 1.9763271338834834e-06, + "loss": 0.3292, + "step": 27754 + }, + { + "epoch": 10.25, + "learning_rate": 1.976290183520138e-06, + "loss": 0.2753, + "step": 27755 + }, + { + "epoch": 10.25, + "learning_rate": 1.9762532331567927e-06, + "loss": 0.3269, + "step": 27756 + }, + { + "epoch": 10.25, + "learning_rate": 1.9762162827934473e-06, + "loss": 0.4192, + "step": 27757 + }, + { + "epoch": 10.25, + "learning_rate": 1.9761793324301024e-06, + "loss": 0.4421, + "step": 27758 + }, + { + "epoch": 10.25, + "learning_rate": 1.976142382066757e-06, + "loss": 0.48, + "step": 27759 + }, + { + "epoch": 10.25, + "learning_rate": 1.976105431703412e-06, + "loss": 0.33, + "step": 27760 + }, + { + "epoch": 10.25, + "learning_rate": 1.9760684813400667e-06, + "loss": 0.4355, + "step": 27761 + }, + { + "epoch": 10.25, + "learning_rate": 1.9760315309767214e-06, + "loss": 0.4896, + "step": 27762 + }, + { + "epoch": 10.25, + "learning_rate": 1.975994580613376e-06, + "loss": 0.3673, + "step": 27763 + }, + { + "epoch": 10.25, + "learning_rate": 1.9759576302500306e-06, + "loss": 0.2724, + "step": 27764 + }, + { + "epoch": 10.25, + "learning_rate": 1.9759206798866857e-06, + "loss": 0.2711, + "step": 27765 + }, + { + "epoch": 10.25, + "learning_rate": 1.9758837295233403e-06, + "loss": 0.3507, + "step": 27766 + }, + { + "epoch": 10.25, + "learning_rate": 1.9758467791599954e-06, + "loss": 0.2284, + "step": 27767 + }, + { + "epoch": 10.25, + "learning_rate": 1.97580982879665e-06, + "loss": 0.3538, + "step": 27768 + }, + { + "epoch": 10.25, + "learning_rate": 1.9757728784333047e-06, + "loss": 0.3729, + "step": 27769 + }, + { + "epoch": 10.25, + "learning_rate": 1.9757359280699593e-06, + "loss": 0.2416, + "step": 27770 + }, + { + "epoch": 10.26, + "learning_rate": 1.975698977706614e-06, + "loss": 0.3205, + "step": 27771 + }, + { + "epoch": 10.26, + "learning_rate": 1.975662027343269e-06, + "loss": 0.3005, + "step": 27772 + }, + { + "epoch": 10.26, + "learning_rate": 1.9756250769799236e-06, + "loss": 0.3743, + "step": 27773 + }, + { + "epoch": 10.26, + "learning_rate": 1.9755881266165787e-06, + "loss": 0.3913, + "step": 27774 + }, + { + "epoch": 10.26, + "learning_rate": 1.9755511762532333e-06, + "loss": 0.3082, + "step": 27775 + }, + { + "epoch": 10.26, + "learning_rate": 1.975514225889888e-06, + "loss": 0.3625, + "step": 27776 + }, + { + "epoch": 10.26, + "learning_rate": 1.9754772755265426e-06, + "loss": 0.3515, + "step": 27777 + }, + { + "epoch": 10.26, + "learning_rate": 1.9754403251631977e-06, + "loss": 0.4003, + "step": 27778 + }, + { + "epoch": 10.26, + "learning_rate": 1.9754033747998523e-06, + "loss": 0.4478, + "step": 27779 + }, + { + "epoch": 10.26, + "learning_rate": 1.9753664244365073e-06, + "loss": 0.3026, + "step": 27780 + }, + { + "epoch": 10.26, + "learning_rate": 1.975329474073162e-06, + "loss": 0.346, + "step": 27781 + }, + { + "epoch": 10.26, + "learning_rate": 1.975292523709816e-06, + "loss": 0.2939, + "step": 27782 + }, + { + "epoch": 10.26, + "learning_rate": 1.9752555733464713e-06, + "loss": 0.3567, + "step": 27783 + }, + { + "epoch": 10.26, + "learning_rate": 1.975218622983126e-06, + "loss": 0.2528, + "step": 27784 + }, + { + "epoch": 10.26, + "learning_rate": 1.975181672619781e-06, + "loss": 0.3578, + "step": 27785 + }, + { + "epoch": 10.26, + "learning_rate": 1.9751447222564356e-06, + "loss": 0.3939, + "step": 27786 + }, + { + "epoch": 10.26, + "learning_rate": 1.9751077718930907e-06, + "loss": 0.3437, + "step": 27787 + }, + { + "epoch": 10.26, + "learning_rate": 1.975070821529745e-06, + "loss": 0.377, + "step": 27788 + }, + { + "epoch": 10.26, + "learning_rate": 1.9750338711664e-06, + "loss": 0.3047, + "step": 27789 + }, + { + "epoch": 10.26, + "learning_rate": 1.9749969208030546e-06, + "loss": 0.4142, + "step": 27790 + }, + { + "epoch": 10.26, + "learning_rate": 1.974959970439709e-06, + "loss": 0.2872, + "step": 27791 + }, + { + "epoch": 10.26, + "learning_rate": 1.9749230200763643e-06, + "loss": 0.3005, + "step": 27792 + }, + { + "epoch": 10.26, + "learning_rate": 1.974886069713019e-06, + "loss": 0.2977, + "step": 27793 + }, + { + "epoch": 10.26, + "learning_rate": 1.9748491193496735e-06, + "loss": 0.2236, + "step": 27794 + }, + { + "epoch": 10.26, + "learning_rate": 1.974812168986328e-06, + "loss": 0.3757, + "step": 27795 + }, + { + "epoch": 10.26, + "learning_rate": 1.9747752186229832e-06, + "loss": 0.3165, + "step": 27796 + }, + { + "epoch": 10.26, + "learning_rate": 1.974738268259638e-06, + "loss": 0.3024, + "step": 27797 + }, + { + "epoch": 10.27, + "learning_rate": 1.974701317896293e-06, + "loss": 0.3749, + "step": 27798 + }, + { + "epoch": 10.27, + "learning_rate": 1.9746643675329476e-06, + "loss": 0.2724, + "step": 27799 + }, + { + "epoch": 10.27, + "learning_rate": 1.974627417169602e-06, + "loss": 0.5793, + "step": 27800 + }, + { + "epoch": 10.27, + "learning_rate": 1.974590466806257e-06, + "loss": 0.4147, + "step": 27801 + }, + { + "epoch": 10.27, + "learning_rate": 1.9745535164429115e-06, + "loss": 0.4038, + "step": 27802 + }, + { + "epoch": 10.27, + "learning_rate": 1.9745165660795665e-06, + "loss": 0.3694, + "step": 27803 + }, + { + "epoch": 10.27, + "learning_rate": 1.974479615716221e-06, + "loss": 0.2797, + "step": 27804 + }, + { + "epoch": 10.27, + "learning_rate": 1.9744426653528762e-06, + "loss": 0.4308, + "step": 27805 + }, + { + "epoch": 10.27, + "learning_rate": 1.974405714989531e-06, + "loss": 0.3479, + "step": 27806 + }, + { + "epoch": 10.27, + "learning_rate": 1.9743687646261855e-06, + "loss": 0.364, + "step": 27807 + }, + { + "epoch": 10.27, + "learning_rate": 1.97433181426284e-06, + "loss": 0.3606, + "step": 27808 + }, + { + "epoch": 10.27, + "learning_rate": 1.974294863899495e-06, + "loss": 0.3175, + "step": 27809 + }, + { + "epoch": 10.27, + "learning_rate": 1.97425791353615e-06, + "loss": 0.3122, + "step": 27810 + }, + { + "epoch": 10.27, + "learning_rate": 1.9742209631728045e-06, + "loss": 0.3756, + "step": 27811 + }, + { + "epoch": 10.27, + "learning_rate": 1.9741840128094595e-06, + "loss": 0.3362, + "step": 27812 + }, + { + "epoch": 10.27, + "learning_rate": 1.974147062446114e-06, + "loss": 0.24, + "step": 27813 + }, + { + "epoch": 10.27, + "learning_rate": 1.974110112082769e-06, + "loss": 0.3301, + "step": 27814 + }, + { + "epoch": 10.27, + "learning_rate": 1.9740731617194234e-06, + "loss": 0.4403, + "step": 27815 + }, + { + "epoch": 10.27, + "learning_rate": 1.9740362113560785e-06, + "loss": 0.4072, + "step": 27816 + }, + { + "epoch": 10.27, + "learning_rate": 1.973999260992733e-06, + "loss": 0.226, + "step": 27817 + }, + { + "epoch": 10.27, + "learning_rate": 1.973962310629388e-06, + "loss": 0.3651, + "step": 27818 + }, + { + "epoch": 10.27, + "learning_rate": 1.973925360266043e-06, + "loss": 0.323, + "step": 27819 + }, + { + "epoch": 10.27, + "learning_rate": 1.9738884099026975e-06, + "loss": 0.2537, + "step": 27820 + }, + { + "epoch": 10.27, + "learning_rate": 1.973851459539352e-06, + "loss": 0.3156, + "step": 27821 + }, + { + "epoch": 10.27, + "learning_rate": 1.9738145091760067e-06, + "loss": 0.2289, + "step": 27822 + }, + { + "epoch": 10.27, + "learning_rate": 1.973777558812662e-06, + "loss": 0.5932, + "step": 27823 + }, + { + "epoch": 10.27, + "learning_rate": 1.9737406084493164e-06, + "loss": 0.3338, + "step": 27824 + }, + { + "epoch": 10.28, + "learning_rate": 1.9737036580859715e-06, + "loss": 0.341, + "step": 27825 + }, + { + "epoch": 10.28, + "learning_rate": 1.9736667077226257e-06, + "loss": 0.3724, + "step": 27826 + }, + { + "epoch": 10.28, + "learning_rate": 1.9736297573592808e-06, + "loss": 0.3086, + "step": 27827 + }, + { + "epoch": 10.28, + "learning_rate": 1.9735928069959354e-06, + "loss": 0.3453, + "step": 27828 + }, + { + "epoch": 10.28, + "learning_rate": 1.9735558566325905e-06, + "loss": 0.3206, + "step": 27829 + }, + { + "epoch": 10.28, + "learning_rate": 1.973518906269245e-06, + "loss": 0.2696, + "step": 27830 + }, + { + "epoch": 10.28, + "learning_rate": 1.9734819559058997e-06, + "loss": 0.4156, + "step": 27831 + }, + { + "epoch": 10.28, + "learning_rate": 1.9734450055425544e-06, + "loss": 0.4049, + "step": 27832 + }, + { + "epoch": 10.28, + "learning_rate": 1.973408055179209e-06, + "loss": 0.4412, + "step": 27833 + }, + { + "epoch": 10.28, + "learning_rate": 1.973371104815864e-06, + "loss": 0.3107, + "step": 27834 + }, + { + "epoch": 10.28, + "learning_rate": 1.9733341544525187e-06, + "loss": 0.3508, + "step": 27835 + }, + { + "epoch": 10.28, + "learning_rate": 1.9732972040891738e-06, + "loss": 0.3602, + "step": 27836 + }, + { + "epoch": 10.28, + "learning_rate": 1.9732602537258284e-06, + "loss": 0.3012, + "step": 27837 + }, + { + "epoch": 10.28, + "learning_rate": 1.973223303362483e-06, + "loss": 0.3366, + "step": 27838 + }, + { + "epoch": 10.28, + "learning_rate": 1.9731863529991377e-06, + "loss": 0.3203, + "step": 27839 + }, + { + "epoch": 10.28, + "learning_rate": 1.9731494026357927e-06, + "loss": 0.3419, + "step": 27840 + }, + { + "epoch": 10.28, + "learning_rate": 1.9731124522724474e-06, + "loss": 0.3177, + "step": 27841 + }, + { + "epoch": 10.28, + "learning_rate": 1.973075501909102e-06, + "loss": 0.3559, + "step": 27842 + }, + { + "epoch": 10.28, + "learning_rate": 1.973038551545757e-06, + "loss": 0.4402, + "step": 27843 + }, + { + "epoch": 10.28, + "learning_rate": 1.9730016011824117e-06, + "loss": 0.3779, + "step": 27844 + }, + { + "epoch": 10.28, + "learning_rate": 1.9729646508190663e-06, + "loss": 0.4701, + "step": 27845 + }, + { + "epoch": 10.28, + "learning_rate": 1.972927700455721e-06, + "loss": 0.2888, + "step": 27846 + }, + { + "epoch": 10.28, + "learning_rate": 1.972890750092376e-06, + "loss": 0.4149, + "step": 27847 + }, + { + "epoch": 10.28, + "learning_rate": 1.9728537997290307e-06, + "loss": 0.4032, + "step": 27848 + }, + { + "epoch": 10.28, + "learning_rate": 1.9728168493656857e-06, + "loss": 0.3202, + "step": 27849 + }, + { + "epoch": 10.28, + "learning_rate": 1.9727798990023404e-06, + "loss": 0.2931, + "step": 27850 + }, + { + "epoch": 10.28, + "learning_rate": 1.972742948638995e-06, + "loss": 0.2893, + "step": 27851 + }, + { + "epoch": 10.29, + "learning_rate": 1.9727059982756496e-06, + "loss": 0.3264, + "step": 27852 + }, + { + "epoch": 10.29, + "learning_rate": 1.9726690479123043e-06, + "loss": 0.3691, + "step": 27853 + }, + { + "epoch": 10.29, + "learning_rate": 1.9726320975489593e-06, + "loss": 0.3395, + "step": 27854 + }, + { + "epoch": 10.29, + "learning_rate": 1.972595147185614e-06, + "loss": 0.2323, + "step": 27855 + }, + { + "epoch": 10.29, + "learning_rate": 1.972558196822269e-06, + "loss": 0.3651, + "step": 27856 + }, + { + "epoch": 10.29, + "learning_rate": 1.9725212464589237e-06, + "loss": 0.3478, + "step": 27857 + }, + { + "epoch": 10.29, + "learning_rate": 1.9724842960955783e-06, + "loss": 0.3398, + "step": 27858 + }, + { + "epoch": 10.29, + "learning_rate": 1.972447345732233e-06, + "loss": 0.2893, + "step": 27859 + }, + { + "epoch": 10.29, + "learning_rate": 1.972410395368888e-06, + "loss": 0.3271, + "step": 27860 + }, + { + "epoch": 10.29, + "learning_rate": 1.9723734450055426e-06, + "loss": 0.4733, + "step": 27861 + }, + { + "epoch": 10.29, + "learning_rate": 1.9723364946421973e-06, + "loss": 0.3555, + "step": 27862 + }, + { + "epoch": 10.29, + "learning_rate": 1.9722995442788523e-06, + "loss": 0.3526, + "step": 27863 + }, + { + "epoch": 10.29, + "learning_rate": 1.972262593915507e-06, + "loss": 0.3318, + "step": 27864 + }, + { + "epoch": 10.29, + "learning_rate": 1.9722256435521616e-06, + "loss": 0.3228, + "step": 27865 + }, + { + "epoch": 10.29, + "learning_rate": 1.9721886931888162e-06, + "loss": 0.3613, + "step": 27866 + }, + { + "epoch": 10.29, + "learning_rate": 1.9721517428254713e-06, + "loss": 0.3257, + "step": 27867 + }, + { + "epoch": 10.29, + "learning_rate": 1.972114792462126e-06, + "loss": 0.3063, + "step": 27868 + }, + { + "epoch": 10.29, + "learning_rate": 1.972077842098781e-06, + "loss": 0.3657, + "step": 27869 + }, + { + "epoch": 10.29, + "learning_rate": 1.9720408917354356e-06, + "loss": 0.2773, + "step": 27870 + }, + { + "epoch": 10.29, + "learning_rate": 1.97200394137209e-06, + "loss": 0.3244, + "step": 27871 + }, + { + "epoch": 10.29, + "learning_rate": 1.971966991008745e-06, + "loss": 0.3499, + "step": 27872 + }, + { + "epoch": 10.29, + "learning_rate": 1.9719300406453995e-06, + "loss": 0.402, + "step": 27873 + }, + { + "epoch": 10.29, + "learning_rate": 1.9718930902820546e-06, + "loss": 0.2561, + "step": 27874 + }, + { + "epoch": 10.29, + "learning_rate": 1.9718561399187092e-06, + "loss": 0.3559, + "step": 27875 + }, + { + "epoch": 10.29, + "learning_rate": 1.9718191895553643e-06, + "loss": 0.3836, + "step": 27876 + }, + { + "epoch": 10.29, + "learning_rate": 1.9717822391920185e-06, + "loss": 0.3872, + "step": 27877 + }, + { + "epoch": 10.29, + "learning_rate": 1.9717452888286736e-06, + "loss": 0.2866, + "step": 27878 + }, + { + "epoch": 10.3, + "learning_rate": 1.9717083384653282e-06, + "loss": 0.3931, + "step": 27879 + }, + { + "epoch": 10.3, + "learning_rate": 1.9716713881019833e-06, + "loss": 0.363, + "step": 27880 + }, + { + "epoch": 10.3, + "learning_rate": 1.971634437738638e-06, + "loss": 0.3565, + "step": 27881 + }, + { + "epoch": 10.3, + "learning_rate": 1.9715974873752925e-06, + "loss": 0.244, + "step": 27882 + }, + { + "epoch": 10.3, + "learning_rate": 1.971560537011947e-06, + "loss": 0.3945, + "step": 27883 + }, + { + "epoch": 10.3, + "learning_rate": 1.971523586648602e-06, + "loss": 0.3845, + "step": 27884 + }, + { + "epoch": 10.3, + "learning_rate": 1.971486636285257e-06, + "loss": 0.2421, + "step": 27885 + }, + { + "epoch": 10.3, + "learning_rate": 1.9714496859219115e-06, + "loss": 0.4507, + "step": 27886 + }, + { + "epoch": 10.3, + "learning_rate": 1.9714127355585666e-06, + "loss": 0.2728, + "step": 27887 + }, + { + "epoch": 10.3, + "learning_rate": 1.9713757851952212e-06, + "loss": 0.2409, + "step": 27888 + }, + { + "epoch": 10.3, + "learning_rate": 1.971338834831876e-06, + "loss": 0.2679, + "step": 27889 + }, + { + "epoch": 10.3, + "learning_rate": 1.9713018844685305e-06, + "loss": 0.2944, + "step": 27890 + }, + { + "epoch": 10.3, + "learning_rate": 1.971264934105185e-06, + "loss": 0.243, + "step": 27891 + }, + { + "epoch": 10.3, + "learning_rate": 1.97122798374184e-06, + "loss": 0.4515, + "step": 27892 + }, + { + "epoch": 10.3, + "learning_rate": 1.971191033378495e-06, + "loss": 0.339, + "step": 27893 + }, + { + "epoch": 10.3, + "learning_rate": 1.97115408301515e-06, + "loss": 0.3267, + "step": 27894 + }, + { + "epoch": 10.3, + "learning_rate": 1.9711171326518045e-06, + "loss": 0.4028, + "step": 27895 + }, + { + "epoch": 10.3, + "learning_rate": 1.971080182288459e-06, + "loss": 0.2983, + "step": 27896 + }, + { + "epoch": 10.3, + "learning_rate": 1.9710432319251138e-06, + "loss": 0.3456, + "step": 27897 + }, + { + "epoch": 10.3, + "learning_rate": 1.971006281561769e-06, + "loss": 0.2575, + "step": 27898 + }, + { + "epoch": 10.3, + "learning_rate": 1.9709693311984235e-06, + "loss": 0.3749, + "step": 27899 + }, + { + "epoch": 10.3, + "learning_rate": 1.970932380835078e-06, + "loss": 0.3795, + "step": 27900 + }, + { + "epoch": 10.3, + "learning_rate": 1.970895430471733e-06, + "loss": 0.3011, + "step": 27901 + }, + { + "epoch": 10.3, + "learning_rate": 1.970858480108388e-06, + "loss": 0.4331, + "step": 27902 + }, + { + "epoch": 10.3, + "learning_rate": 1.9708215297450425e-06, + "loss": 0.1999, + "step": 27903 + }, + { + "epoch": 10.3, + "learning_rate": 1.970784579381697e-06, + "loss": 0.2883, + "step": 27904 + }, + { + "epoch": 10.3, + "learning_rate": 1.970747629018352e-06, + "loss": 0.2414, + "step": 27905 + }, + { + "epoch": 10.31, + "learning_rate": 1.9707106786550068e-06, + "loss": 0.2473, + "step": 27906 + }, + { + "epoch": 10.31, + "learning_rate": 1.970673728291662e-06, + "loss": 0.3821, + "step": 27907 + }, + { + "epoch": 10.31, + "learning_rate": 1.9706367779283165e-06, + "loss": 0.2373, + "step": 27908 + }, + { + "epoch": 10.31, + "learning_rate": 1.970599827564971e-06, + "loss": 0.4328, + "step": 27909 + }, + { + "epoch": 10.31, + "learning_rate": 1.9705628772016258e-06, + "loss": 0.3853, + "step": 27910 + }, + { + "epoch": 10.31, + "learning_rate": 1.9705259268382804e-06, + "loss": 0.2084, + "step": 27911 + }, + { + "epoch": 10.31, + "learning_rate": 1.9704889764749355e-06, + "loss": 0.3465, + "step": 27912 + }, + { + "epoch": 10.31, + "learning_rate": 1.97045202611159e-06, + "loss": 0.3385, + "step": 27913 + }, + { + "epoch": 10.31, + "learning_rate": 1.970415075748245e-06, + "loss": 0.3904, + "step": 27914 + }, + { + "epoch": 10.31, + "learning_rate": 1.9703781253848998e-06, + "loss": 0.3926, + "step": 27915 + }, + { + "epoch": 10.31, + "learning_rate": 1.9703411750215544e-06, + "loss": 0.2929, + "step": 27916 + }, + { + "epoch": 10.31, + "learning_rate": 1.970304224658209e-06, + "loss": 0.3132, + "step": 27917 + }, + { + "epoch": 10.31, + "learning_rate": 1.970267274294864e-06, + "loss": 0.2278, + "step": 27918 + }, + { + "epoch": 10.31, + "learning_rate": 1.9702303239315188e-06, + "loss": 0.4707, + "step": 27919 + }, + { + "epoch": 10.31, + "learning_rate": 1.9701933735681734e-06, + "loss": 0.4031, + "step": 27920 + }, + { + "epoch": 10.31, + "learning_rate": 1.9701564232048285e-06, + "loss": 0.4325, + "step": 27921 + }, + { + "epoch": 10.31, + "learning_rate": 1.9701194728414827e-06, + "loss": 0.3975, + "step": 27922 + }, + { + "epoch": 10.31, + "learning_rate": 1.9700825224781377e-06, + "loss": 0.4216, + "step": 27923 + }, + { + "epoch": 10.31, + "learning_rate": 1.9700455721147924e-06, + "loss": 0.3961, + "step": 27924 + }, + { + "epoch": 10.31, + "learning_rate": 1.9700086217514474e-06, + "loss": 0.3797, + "step": 27925 + }, + { + "epoch": 10.31, + "learning_rate": 1.969971671388102e-06, + "loss": 0.2474, + "step": 27926 + }, + { + "epoch": 10.31, + "learning_rate": 1.969934721024757e-06, + "loss": 0.2848, + "step": 27927 + }, + { + "epoch": 10.31, + "learning_rate": 1.9698977706614113e-06, + "loss": 0.4354, + "step": 27928 + }, + { + "epoch": 10.31, + "learning_rate": 1.9698608202980664e-06, + "loss": 0.3424, + "step": 27929 + }, + { + "epoch": 10.31, + "learning_rate": 1.969823869934721e-06, + "loss": 0.2698, + "step": 27930 + }, + { + "epoch": 10.31, + "learning_rate": 1.9697869195713757e-06, + "loss": 0.3605, + "step": 27931 + }, + { + "epoch": 10.31, + "learning_rate": 1.9697499692080307e-06, + "loss": 0.2328, + "step": 27932 + }, + { + "epoch": 10.31, + "learning_rate": 1.9697130188446854e-06, + "loss": 0.3471, + "step": 27933 + }, + { + "epoch": 10.32, + "learning_rate": 1.96967606848134e-06, + "loss": 0.2418, + "step": 27934 + }, + { + "epoch": 10.32, + "learning_rate": 1.9696391181179946e-06, + "loss": 0.346, + "step": 27935 + }, + { + "epoch": 10.32, + "learning_rate": 1.9696021677546497e-06, + "loss": 0.3042, + "step": 27936 + }, + { + "epoch": 10.32, + "learning_rate": 1.9695652173913043e-06, + "loss": 0.3684, + "step": 27937 + }, + { + "epoch": 10.32, + "learning_rate": 1.9695282670279594e-06, + "loss": 0.3104, + "step": 27938 + }, + { + "epoch": 10.32, + "learning_rate": 1.969491316664614e-06, + "loss": 0.2869, + "step": 27939 + }, + { + "epoch": 10.32, + "learning_rate": 1.9694543663012687e-06, + "loss": 0.3795, + "step": 27940 + }, + { + "epoch": 10.32, + "learning_rate": 1.9694174159379233e-06, + "loss": 0.2813, + "step": 27941 + }, + { + "epoch": 10.32, + "learning_rate": 1.969380465574578e-06, + "loss": 0.4136, + "step": 27942 + }, + { + "epoch": 10.32, + "learning_rate": 1.969343515211233e-06, + "loss": 0.3564, + "step": 27943 + }, + { + "epoch": 10.32, + "learning_rate": 1.9693065648478876e-06, + "loss": 0.3664, + "step": 27944 + }, + { + "epoch": 10.32, + "learning_rate": 1.9692696144845427e-06, + "loss": 0.3079, + "step": 27945 + }, + { + "epoch": 10.32, + "learning_rate": 1.9692326641211973e-06, + "loss": 0.3341, + "step": 27946 + }, + { + "epoch": 10.32, + "learning_rate": 1.969195713757852e-06, + "loss": 0.3404, + "step": 27947 + }, + { + "epoch": 10.32, + "learning_rate": 1.9691587633945066e-06, + "loss": 0.288, + "step": 27948 + }, + { + "epoch": 10.32, + "learning_rate": 1.9691218130311617e-06, + "loss": 0.3595, + "step": 27949 + }, + { + "epoch": 10.32, + "learning_rate": 1.9690848626678163e-06, + "loss": 0.2953, + "step": 27950 + }, + { + "epoch": 10.32, + "learning_rate": 1.969047912304471e-06, + "loss": 0.3507, + "step": 27951 + }, + { + "epoch": 10.32, + "learning_rate": 1.969010961941126e-06, + "loss": 0.2515, + "step": 27952 + }, + { + "epoch": 10.32, + "learning_rate": 1.9689740115777806e-06, + "loss": 0.4265, + "step": 27953 + }, + { + "epoch": 10.32, + "learning_rate": 1.9689370612144353e-06, + "loss": 0.2399, + "step": 27954 + }, + { + "epoch": 10.32, + "learning_rate": 1.96890011085109e-06, + "loss": 0.2928, + "step": 27955 + }, + { + "epoch": 10.32, + "learning_rate": 1.968863160487745e-06, + "loss": 0.4147, + "step": 27956 + }, + { + "epoch": 10.32, + "learning_rate": 1.9688262101243996e-06, + "loss": 0.3472, + "step": 27957 + }, + { + "epoch": 10.32, + "learning_rate": 1.9687892597610547e-06, + "loss": 0.4193, + "step": 27958 + }, + { + "epoch": 10.32, + "learning_rate": 1.9687523093977093e-06, + "loss": 0.2905, + "step": 27959 + }, + { + "epoch": 10.32, + "learning_rate": 1.968715359034364e-06, + "loss": 0.4793, + "step": 27960 + }, + { + "epoch": 10.33, + "learning_rate": 1.9686784086710186e-06, + "loss": 0.3502, + "step": 27961 + }, + { + "epoch": 10.33, + "learning_rate": 1.968641458307673e-06, + "loss": 0.3578, + "step": 27962 + }, + { + "epoch": 10.33, + "learning_rate": 1.9686045079443283e-06, + "loss": 0.2353, + "step": 27963 + }, + { + "epoch": 10.33, + "learning_rate": 1.968567557580983e-06, + "loss": 0.2354, + "step": 27964 + }, + { + "epoch": 10.33, + "learning_rate": 1.968530607217638e-06, + "loss": 0.3352, + "step": 27965 + }, + { + "epoch": 10.33, + "learning_rate": 1.9684936568542926e-06, + "loss": 0.4165, + "step": 27966 + }, + { + "epoch": 10.33, + "learning_rate": 1.9684567064909472e-06, + "loss": 0.3527, + "step": 27967 + }, + { + "epoch": 10.33, + "learning_rate": 1.968419756127602e-06, + "loss": 0.2731, + "step": 27968 + }, + { + "epoch": 10.33, + "learning_rate": 1.968382805764257e-06, + "loss": 0.3338, + "step": 27969 + }, + { + "epoch": 10.33, + "learning_rate": 1.9683458554009116e-06, + "loss": 0.451, + "step": 27970 + }, + { + "epoch": 10.33, + "learning_rate": 1.968308905037566e-06, + "loss": 0.4583, + "step": 27971 + }, + { + "epoch": 10.33, + "learning_rate": 1.9682719546742213e-06, + "loss": 0.3655, + "step": 27972 + }, + { + "epoch": 10.33, + "learning_rate": 1.9682350043108755e-06, + "loss": 0.2111, + "step": 27973 + }, + { + "epoch": 10.33, + "learning_rate": 1.9681980539475305e-06, + "loss": 0.2399, + "step": 27974 + }, + { + "epoch": 10.33, + "learning_rate": 1.968161103584185e-06, + "loss": 0.4027, + "step": 27975 + }, + { + "epoch": 10.33, + "learning_rate": 1.9681241532208402e-06, + "loss": 0.3348, + "step": 27976 + }, + { + "epoch": 10.33, + "learning_rate": 1.968087202857495e-06, + "loss": 0.2917, + "step": 27977 + }, + { + "epoch": 10.33, + "learning_rate": 1.9680502524941495e-06, + "loss": 0.3067, + "step": 27978 + }, + { + "epoch": 10.33, + "learning_rate": 1.968013302130804e-06, + "loss": 0.3414, + "step": 27979 + }, + { + "epoch": 10.33, + "learning_rate": 1.9679763517674588e-06, + "loss": 0.4542, + "step": 27980 + }, + { + "epoch": 10.33, + "learning_rate": 1.967939401404114e-06, + "loss": 0.3354, + "step": 27981 + }, + { + "epoch": 10.33, + "learning_rate": 1.9679024510407685e-06, + "loss": 0.2918, + "step": 27982 + }, + { + "epoch": 10.33, + "learning_rate": 1.9678655006774235e-06, + "loss": 0.2385, + "step": 27983 + }, + { + "epoch": 10.33, + "learning_rate": 1.967828550314078e-06, + "loss": 0.3326, + "step": 27984 + }, + { + "epoch": 10.33, + "learning_rate": 1.967791599950733e-06, + "loss": 0.3281, + "step": 27985 + }, + { + "epoch": 10.33, + "learning_rate": 1.9677546495873874e-06, + "loss": 0.4526, + "step": 27986 + }, + { + "epoch": 10.33, + "learning_rate": 1.9677176992240425e-06, + "loss": 0.4279, + "step": 27987 + }, + { + "epoch": 10.34, + "learning_rate": 1.967680748860697e-06, + "loss": 0.3519, + "step": 27988 + }, + { + "epoch": 10.34, + "learning_rate": 1.967643798497352e-06, + "loss": 0.3344, + "step": 27989 + }, + { + "epoch": 10.34, + "learning_rate": 1.967606848134007e-06, + "loss": 0.3852, + "step": 27990 + }, + { + "epoch": 10.34, + "learning_rate": 1.9675698977706615e-06, + "loss": 0.2404, + "step": 27991 + }, + { + "epoch": 10.34, + "learning_rate": 1.967532947407316e-06, + "loss": 0.3315, + "step": 27992 + }, + { + "epoch": 10.34, + "learning_rate": 1.9674959970439707e-06, + "loss": 0.5744, + "step": 27993 + }, + { + "epoch": 10.34, + "learning_rate": 1.967459046680626e-06, + "loss": 0.291, + "step": 27994 + }, + { + "epoch": 10.34, + "learning_rate": 1.9674220963172804e-06, + "loss": 0.3962, + "step": 27995 + }, + { + "epoch": 10.34, + "learning_rate": 1.9673851459539355e-06, + "loss": 0.2992, + "step": 27996 + }, + { + "epoch": 10.34, + "learning_rate": 1.96734819559059e-06, + "loss": 0.3629, + "step": 27997 + }, + { + "epoch": 10.34, + "learning_rate": 1.9673112452272448e-06, + "loss": 0.2926, + "step": 27998 + }, + { + "epoch": 10.34, + "learning_rate": 1.9672742948638994e-06, + "loss": 0.3661, + "step": 27999 + }, + { + "epoch": 10.34, + "learning_rate": 1.967237344500554e-06, + "loss": 0.315, + "step": 28000 + }, + { + "epoch": 10.34, + "learning_rate": 1.967200394137209e-06, + "loss": 0.2911, + "step": 28001 + }, + { + "epoch": 10.34, + "learning_rate": 1.9671634437738637e-06, + "loss": 0.3202, + "step": 28002 + }, + { + "epoch": 10.34, + "learning_rate": 1.967126493410519e-06, + "loss": 0.3092, + "step": 28003 + }, + { + "epoch": 10.34, + "learning_rate": 1.9670895430471734e-06, + "loss": 0.2961, + "step": 28004 + }, + { + "epoch": 10.34, + "learning_rate": 1.967052592683828e-06, + "loss": 0.2524, + "step": 28005 + }, + { + "epoch": 10.34, + "learning_rate": 1.9670156423204827e-06, + "loss": 0.3848, + "step": 28006 + }, + { + "epoch": 10.34, + "learning_rate": 1.9669786919571378e-06, + "loss": 0.4167, + "step": 28007 + }, + { + "epoch": 10.34, + "learning_rate": 1.9669417415937924e-06, + "loss": 0.2556, + "step": 28008 + }, + { + "epoch": 10.34, + "learning_rate": 1.9669047912304475e-06, + "loss": 0.2281, + "step": 28009 + }, + { + "epoch": 10.34, + "learning_rate": 1.966867840867102e-06, + "loss": 0.3007, + "step": 28010 + }, + { + "epoch": 10.34, + "learning_rate": 1.9668308905037563e-06, + "loss": 0.2089, + "step": 28011 + }, + { + "epoch": 10.34, + "learning_rate": 1.9667939401404114e-06, + "loss": 0.306, + "step": 28012 + }, + { + "epoch": 10.34, + "learning_rate": 1.966756989777066e-06, + "loss": 0.3628, + "step": 28013 + }, + { + "epoch": 10.34, + "learning_rate": 1.966720039413721e-06, + "loss": 0.436, + "step": 28014 + }, + { + "epoch": 10.35, + "learning_rate": 1.9666830890503757e-06, + "loss": 0.2465, + "step": 28015 + }, + { + "epoch": 10.35, + "learning_rate": 1.9666461386870308e-06, + "loss": 0.2621, + "step": 28016 + }, + { + "epoch": 10.35, + "learning_rate": 1.966609188323685e-06, + "loss": 0.3977, + "step": 28017 + }, + { + "epoch": 10.35, + "learning_rate": 1.96657223796034e-06, + "loss": 0.3294, + "step": 28018 + }, + { + "epoch": 10.35, + "learning_rate": 1.9665352875969947e-06, + "loss": 0.3187, + "step": 28019 + }, + { + "epoch": 10.35, + "learning_rate": 1.9664983372336493e-06, + "loss": 0.3094, + "step": 28020 + }, + { + "epoch": 10.35, + "learning_rate": 1.9664613868703044e-06, + "loss": 0.3503, + "step": 28021 + }, + { + "epoch": 10.35, + "learning_rate": 1.966424436506959e-06, + "loss": 0.2666, + "step": 28022 + }, + { + "epoch": 10.35, + "learning_rate": 1.9663874861436136e-06, + "loss": 0.3269, + "step": 28023 + }, + { + "epoch": 10.35, + "learning_rate": 1.9663505357802683e-06, + "loss": 0.2484, + "step": 28024 + }, + { + "epoch": 10.35, + "learning_rate": 1.9663135854169233e-06, + "loss": 0.3476, + "step": 28025 + }, + { + "epoch": 10.35, + "learning_rate": 1.966276635053578e-06, + "loss": 0.3407, + "step": 28026 + }, + { + "epoch": 10.35, + "learning_rate": 1.966239684690233e-06, + "loss": 0.3885, + "step": 28027 + }, + { + "epoch": 10.35, + "learning_rate": 1.9662027343268877e-06, + "loss": 0.2792, + "step": 28028 + }, + { + "epoch": 10.35, + "learning_rate": 1.9661657839635423e-06, + "loss": 0.3778, + "step": 28029 + }, + { + "epoch": 10.35, + "learning_rate": 1.966128833600197e-06, + "loss": 0.3152, + "step": 28030 + }, + { + "epoch": 10.35, + "learning_rate": 1.9660918832368516e-06, + "loss": 0.392, + "step": 28031 + }, + { + "epoch": 10.35, + "learning_rate": 1.9660549328735066e-06, + "loss": 0.3033, + "step": 28032 + }, + { + "epoch": 10.35, + "learning_rate": 1.9660179825101613e-06, + "loss": 0.3377, + "step": 28033 + }, + { + "epoch": 10.35, + "learning_rate": 1.9659810321468163e-06, + "loss": 0.2443, + "step": 28034 + }, + { + "epoch": 10.35, + "learning_rate": 1.965944081783471e-06, + "loss": 0.3218, + "step": 28035 + }, + { + "epoch": 10.35, + "learning_rate": 1.9659071314201256e-06, + "loss": 0.3766, + "step": 28036 + }, + { + "epoch": 10.35, + "learning_rate": 1.9658701810567803e-06, + "loss": 0.344, + "step": 28037 + }, + { + "epoch": 10.35, + "learning_rate": 1.9658332306934353e-06, + "loss": 0.3173, + "step": 28038 + }, + { + "epoch": 10.35, + "learning_rate": 1.96579628033009e-06, + "loss": 0.5092, + "step": 28039 + }, + { + "epoch": 10.35, + "learning_rate": 1.9657593299667446e-06, + "loss": 0.3041, + "step": 28040 + }, + { + "epoch": 10.35, + "learning_rate": 1.9657223796033996e-06, + "loss": 0.4097, + "step": 28041 + }, + { + "epoch": 10.36, + "learning_rate": 1.9656854292400543e-06, + "loss": 0.2922, + "step": 28042 + }, + { + "epoch": 10.36, + "learning_rate": 1.965648478876709e-06, + "loss": 0.2853, + "step": 28043 + }, + { + "epoch": 10.36, + "learning_rate": 1.9656115285133636e-06, + "loss": 0.2069, + "step": 28044 + }, + { + "epoch": 10.36, + "learning_rate": 1.9655745781500186e-06, + "loss": 0.3214, + "step": 28045 + }, + { + "epoch": 10.36, + "learning_rate": 1.9655376277866733e-06, + "loss": 0.3991, + "step": 28046 + }, + { + "epoch": 10.36, + "learning_rate": 1.9655006774233283e-06, + "loss": 0.4953, + "step": 28047 + }, + { + "epoch": 10.36, + "learning_rate": 1.965463727059983e-06, + "loss": 0.2862, + "step": 28048 + }, + { + "epoch": 10.36, + "learning_rate": 1.9654267766966376e-06, + "loss": 0.4547, + "step": 28049 + }, + { + "epoch": 10.36, + "learning_rate": 1.9653898263332922e-06, + "loss": 0.3357, + "step": 28050 + }, + { + "epoch": 10.36, + "learning_rate": 1.965352875969947e-06, + "loss": 0.3499, + "step": 28051 + }, + { + "epoch": 10.36, + "learning_rate": 1.965315925606602e-06, + "loss": 0.3316, + "step": 28052 + }, + { + "epoch": 10.36, + "learning_rate": 1.9652789752432566e-06, + "loss": 0.2444, + "step": 28053 + }, + { + "epoch": 10.36, + "learning_rate": 1.9652420248799116e-06, + "loss": 0.3842, + "step": 28054 + }, + { + "epoch": 10.36, + "learning_rate": 1.9652050745165663e-06, + "loss": 0.2754, + "step": 28055 + }, + { + "epoch": 10.36, + "learning_rate": 1.965168124153221e-06, + "loss": 0.3799, + "step": 28056 + }, + { + "epoch": 10.36, + "learning_rate": 1.9651311737898755e-06, + "loss": 0.3126, + "step": 28057 + }, + { + "epoch": 10.36, + "learning_rate": 1.9650942234265306e-06, + "loss": 0.3902, + "step": 28058 + }, + { + "epoch": 10.36, + "learning_rate": 1.9650572730631852e-06, + "loss": 0.4275, + "step": 28059 + }, + { + "epoch": 10.36, + "learning_rate": 1.96502032269984e-06, + "loss": 0.3662, + "step": 28060 + }, + { + "epoch": 10.36, + "learning_rate": 1.964983372336495e-06, + "loss": 0.3028, + "step": 28061 + }, + { + "epoch": 10.36, + "learning_rate": 1.964946421973149e-06, + "loss": 0.3955, + "step": 28062 + }, + { + "epoch": 10.36, + "learning_rate": 1.964909471609804e-06, + "loss": 0.3794, + "step": 28063 + }, + { + "epoch": 10.36, + "learning_rate": 1.964872521246459e-06, + "loss": 0.3387, + "step": 28064 + }, + { + "epoch": 10.36, + "learning_rate": 1.964835570883114e-06, + "loss": 0.4194, + "step": 28065 + }, + { + "epoch": 10.36, + "learning_rate": 1.9647986205197685e-06, + "loss": 0.3451, + "step": 28066 + }, + { + "epoch": 10.36, + "learning_rate": 1.9647616701564236e-06, + "loss": 0.4308, + "step": 28067 + }, + { + "epoch": 10.36, + "learning_rate": 1.964724719793078e-06, + "loss": 0.2594, + "step": 28068 + }, + { + "epoch": 10.37, + "learning_rate": 1.964687769429733e-06, + "loss": 0.3319, + "step": 28069 + }, + { + "epoch": 10.37, + "learning_rate": 1.9646508190663875e-06, + "loss": 0.6521, + "step": 28070 + }, + { + "epoch": 10.37, + "learning_rate": 1.964613868703042e-06, + "loss": 0.3947, + "step": 28071 + }, + { + "epoch": 10.37, + "learning_rate": 1.964576918339697e-06, + "loss": 0.4737, + "step": 28072 + }, + { + "epoch": 10.37, + "learning_rate": 1.964539967976352e-06, + "loss": 0.3614, + "step": 28073 + }, + { + "epoch": 10.37, + "learning_rate": 1.9645030176130065e-06, + "loss": 0.2945, + "step": 28074 + }, + { + "epoch": 10.37, + "learning_rate": 1.964466067249661e-06, + "loss": 0.4151, + "step": 28075 + }, + { + "epoch": 10.37, + "learning_rate": 1.964429116886316e-06, + "loss": 0.2957, + "step": 28076 + }, + { + "epoch": 10.37, + "learning_rate": 1.964392166522971e-06, + "loss": 0.3777, + "step": 28077 + }, + { + "epoch": 10.37, + "learning_rate": 1.964355216159626e-06, + "loss": 0.3786, + "step": 28078 + }, + { + "epoch": 10.37, + "learning_rate": 1.9643182657962805e-06, + "loss": 0.3137, + "step": 28079 + }, + { + "epoch": 10.37, + "learning_rate": 1.964281315432935e-06, + "loss": 0.369, + "step": 28080 + }, + { + "epoch": 10.37, + "learning_rate": 1.9642443650695898e-06, + "loss": 0.1917, + "step": 28081 + }, + { + "epoch": 10.37, + "learning_rate": 1.9642074147062444e-06, + "loss": 0.3196, + "step": 28082 + }, + { + "epoch": 10.37, + "learning_rate": 1.9641704643428995e-06, + "loss": 0.2743, + "step": 28083 + }, + { + "epoch": 10.37, + "learning_rate": 1.964133513979554e-06, + "loss": 0.415, + "step": 28084 + }, + { + "epoch": 10.37, + "learning_rate": 1.964096563616209e-06, + "loss": 0.2716, + "step": 28085 + }, + { + "epoch": 10.37, + "learning_rate": 1.964059613252864e-06, + "loss": 0.4396, + "step": 28086 + }, + { + "epoch": 10.37, + "learning_rate": 1.9640226628895184e-06, + "loss": 0.4321, + "step": 28087 + }, + { + "epoch": 10.37, + "learning_rate": 1.963985712526173e-06, + "loss": 0.323, + "step": 28088 + }, + { + "epoch": 10.37, + "learning_rate": 1.963948762162828e-06, + "loss": 0.2961, + "step": 28089 + }, + { + "epoch": 10.37, + "learning_rate": 1.9639118117994828e-06, + "loss": 0.289, + "step": 28090 + }, + { + "epoch": 10.37, + "learning_rate": 1.9638748614361374e-06, + "loss": 0.2841, + "step": 28091 + }, + { + "epoch": 10.37, + "learning_rate": 1.9638379110727925e-06, + "loss": 0.3369, + "step": 28092 + }, + { + "epoch": 10.37, + "learning_rate": 1.963800960709447e-06, + "loss": 0.3431, + "step": 28093 + }, + { + "epoch": 10.37, + "learning_rate": 1.9637640103461017e-06, + "loss": 0.3788, + "step": 28094 + }, + { + "epoch": 10.37, + "learning_rate": 1.9637270599827564e-06, + "loss": 0.3238, + "step": 28095 + }, + { + "epoch": 10.38, + "learning_rate": 1.9636901096194114e-06, + "loss": 0.2562, + "step": 28096 + }, + { + "epoch": 10.38, + "learning_rate": 1.963653159256066e-06, + "loss": 0.4245, + "step": 28097 + }, + { + "epoch": 10.38, + "learning_rate": 1.963616208892721e-06, + "loss": 0.2469, + "step": 28098 + }, + { + "epoch": 10.38, + "learning_rate": 1.9635792585293758e-06, + "loss": 0.3323, + "step": 28099 + }, + { + "epoch": 10.38, + "learning_rate": 1.9635423081660304e-06, + "loss": 0.3502, + "step": 28100 + }, + { + "epoch": 10.38, + "learning_rate": 1.963505357802685e-06, + "loss": 0.3641, + "step": 28101 + }, + { + "epoch": 10.38, + "learning_rate": 1.9634684074393397e-06, + "loss": 0.3571, + "step": 28102 + }, + { + "epoch": 10.38, + "learning_rate": 1.9634314570759947e-06, + "loss": 0.3768, + "step": 28103 + }, + { + "epoch": 10.38, + "learning_rate": 1.9633945067126494e-06, + "loss": 0.4382, + "step": 28104 + }, + { + "epoch": 10.38, + "learning_rate": 1.9633575563493044e-06, + "loss": 0.4403, + "step": 28105 + }, + { + "epoch": 10.38, + "learning_rate": 1.963320605985959e-06, + "loss": 0.3541, + "step": 28106 + }, + { + "epoch": 10.38, + "learning_rate": 1.9632836556226137e-06, + "loss": 0.2967, + "step": 28107 + }, + { + "epoch": 10.38, + "learning_rate": 1.9632467052592683e-06, + "loss": 0.4816, + "step": 28108 + }, + { + "epoch": 10.38, + "learning_rate": 1.9632097548959234e-06, + "loss": 0.288, + "step": 28109 + }, + { + "epoch": 10.38, + "learning_rate": 1.963172804532578e-06, + "loss": 0.3746, + "step": 28110 + }, + { + "epoch": 10.38, + "learning_rate": 1.9631358541692327e-06, + "loss": 0.1833, + "step": 28111 + }, + { + "epoch": 10.38, + "learning_rate": 1.9630989038058877e-06, + "loss": 0.387, + "step": 28112 + }, + { + "epoch": 10.38, + "learning_rate": 1.963061953442542e-06, + "loss": 0.2781, + "step": 28113 + }, + { + "epoch": 10.38, + "learning_rate": 1.963025003079197e-06, + "loss": 0.3375, + "step": 28114 + }, + { + "epoch": 10.38, + "learning_rate": 1.9629880527158516e-06, + "loss": 0.2642, + "step": 28115 + }, + { + "epoch": 10.38, + "learning_rate": 1.9629511023525067e-06, + "loss": 0.5497, + "step": 28116 + }, + { + "epoch": 10.38, + "learning_rate": 1.9629141519891613e-06, + "loss": 0.2761, + "step": 28117 + }, + { + "epoch": 10.38, + "learning_rate": 1.9628772016258164e-06, + "loss": 0.265, + "step": 28118 + }, + { + "epoch": 10.38, + "learning_rate": 1.9628402512624706e-06, + "loss": 0.2737, + "step": 28119 + }, + { + "epoch": 10.38, + "learning_rate": 1.9628033008991252e-06, + "loss": 0.3763, + "step": 28120 + }, + { + "epoch": 10.38, + "learning_rate": 1.9627663505357803e-06, + "loss": 0.3596, + "step": 28121 + }, + { + "epoch": 10.38, + "learning_rate": 1.962729400172435e-06, + "loss": 0.3456, + "step": 28122 + }, + { + "epoch": 10.39, + "learning_rate": 1.96269244980909e-06, + "loss": 0.4643, + "step": 28123 + }, + { + "epoch": 10.39, + "learning_rate": 1.9626554994457446e-06, + "loss": 0.4168, + "step": 28124 + }, + { + "epoch": 10.39, + "learning_rate": 1.9626185490823993e-06, + "loss": 0.4396, + "step": 28125 + }, + { + "epoch": 10.39, + "learning_rate": 1.962581598719054e-06, + "loss": 0.3185, + "step": 28126 + }, + { + "epoch": 10.39, + "learning_rate": 1.962544648355709e-06, + "loss": 0.2475, + "step": 28127 + }, + { + "epoch": 10.39, + "learning_rate": 1.9625076979923636e-06, + "loss": 0.3234, + "step": 28128 + }, + { + "epoch": 10.39, + "learning_rate": 1.9624707476290182e-06, + "loss": 0.3031, + "step": 28129 + }, + { + "epoch": 10.39, + "learning_rate": 1.9624337972656733e-06, + "loss": 0.3678, + "step": 28130 + }, + { + "epoch": 10.39, + "learning_rate": 1.962396846902328e-06, + "loss": 0.3819, + "step": 28131 + }, + { + "epoch": 10.39, + "learning_rate": 1.9623598965389826e-06, + "loss": 0.3498, + "step": 28132 + }, + { + "epoch": 10.39, + "learning_rate": 1.962322946175637e-06, + "loss": 0.3513, + "step": 28133 + }, + { + "epoch": 10.39, + "learning_rate": 1.9622859958122923e-06, + "loss": 0.3439, + "step": 28134 + }, + { + "epoch": 10.39, + "learning_rate": 1.962249045448947e-06, + "loss": 0.4276, + "step": 28135 + }, + { + "epoch": 10.39, + "learning_rate": 1.962212095085602e-06, + "loss": 0.5448, + "step": 28136 + }, + { + "epoch": 10.39, + "learning_rate": 1.9621751447222566e-06, + "loss": 0.4533, + "step": 28137 + }, + { + "epoch": 10.39, + "learning_rate": 1.9621381943589112e-06, + "loss": 0.4803, + "step": 28138 + }, + { + "epoch": 10.39, + "learning_rate": 1.962101243995566e-06, + "loss": 0.2991, + "step": 28139 + }, + { + "epoch": 10.39, + "learning_rate": 1.9620642936322205e-06, + "loss": 0.3935, + "step": 28140 + }, + { + "epoch": 10.39, + "learning_rate": 1.9620273432688756e-06, + "loss": 0.2518, + "step": 28141 + }, + { + "epoch": 10.39, + "learning_rate": 1.96199039290553e-06, + "loss": 0.3149, + "step": 28142 + }, + { + "epoch": 10.39, + "learning_rate": 1.9619534425421853e-06, + "loss": 0.3396, + "step": 28143 + }, + { + "epoch": 10.39, + "learning_rate": 1.96191649217884e-06, + "loss": 0.4534, + "step": 28144 + }, + { + "epoch": 10.39, + "learning_rate": 1.9618795418154945e-06, + "loss": 0.3374, + "step": 28145 + }, + { + "epoch": 10.39, + "learning_rate": 1.961842591452149e-06, + "loss": 0.3532, + "step": 28146 + }, + { + "epoch": 10.39, + "learning_rate": 1.9618056410888042e-06, + "loss": 0.3466, + "step": 28147 + }, + { + "epoch": 10.39, + "learning_rate": 1.961768690725459e-06, + "loss": 0.2904, + "step": 28148 + }, + { + "epoch": 10.39, + "learning_rate": 1.9617317403621135e-06, + "loss": 0.3016, + "step": 28149 + }, + { + "epoch": 10.4, + "learning_rate": 1.9616947899987686e-06, + "loss": 0.3924, + "step": 28150 + }, + { + "epoch": 10.4, + "learning_rate": 1.961657839635423e-06, + "loss": 0.4242, + "step": 28151 + }, + { + "epoch": 10.4, + "learning_rate": 1.961620889272078e-06, + "loss": 0.3322, + "step": 28152 + }, + { + "epoch": 10.4, + "learning_rate": 1.9615839389087325e-06, + "loss": 0.205, + "step": 28153 + }, + { + "epoch": 10.4, + "learning_rate": 1.9615469885453875e-06, + "loss": 0.3697, + "step": 28154 + }, + { + "epoch": 10.4, + "learning_rate": 1.961510038182042e-06, + "loss": 0.3696, + "step": 28155 + }, + { + "epoch": 10.4, + "learning_rate": 1.9614730878186972e-06, + "loss": 0.5044, + "step": 28156 + }, + { + "epoch": 10.4, + "learning_rate": 1.961436137455352e-06, + "loss": 0.4671, + "step": 28157 + }, + { + "epoch": 10.4, + "learning_rate": 1.9613991870920065e-06, + "loss": 0.4173, + "step": 28158 + }, + { + "epoch": 10.4, + "learning_rate": 1.961362236728661e-06, + "loss": 0.346, + "step": 28159 + }, + { + "epoch": 10.4, + "learning_rate": 1.9613252863653158e-06, + "loss": 0.3957, + "step": 28160 + }, + { + "epoch": 10.4, + "learning_rate": 1.961288336001971e-06, + "loss": 0.2612, + "step": 28161 + }, + { + "epoch": 10.4, + "learning_rate": 1.9612513856386255e-06, + "loss": 0.3476, + "step": 28162 + }, + { + "epoch": 10.4, + "learning_rate": 1.9612144352752805e-06, + "loss": 0.3153, + "step": 28163 + }, + { + "epoch": 10.4, + "learning_rate": 1.9611774849119348e-06, + "loss": 0.3397, + "step": 28164 + }, + { + "epoch": 10.4, + "learning_rate": 1.96114053454859e-06, + "loss": 0.3069, + "step": 28165 + }, + { + "epoch": 10.4, + "learning_rate": 1.9611035841852444e-06, + "loss": 0.3873, + "step": 28166 + }, + { + "epoch": 10.4, + "learning_rate": 1.9610666338218995e-06, + "loss": 0.3242, + "step": 28167 + }, + { + "epoch": 10.4, + "learning_rate": 1.961029683458554e-06, + "loss": 0.4074, + "step": 28168 + }, + { + "epoch": 10.4, + "learning_rate": 1.9609927330952088e-06, + "loss": 0.3751, + "step": 28169 + }, + { + "epoch": 10.4, + "learning_rate": 1.9609557827318634e-06, + "loss": 0.2841, + "step": 28170 + }, + { + "epoch": 10.4, + "learning_rate": 1.960918832368518e-06, + "loss": 0.2624, + "step": 28171 + }, + { + "epoch": 10.4, + "learning_rate": 1.960881882005173e-06, + "loss": 0.2971, + "step": 28172 + }, + { + "epoch": 10.4, + "learning_rate": 1.9608449316418277e-06, + "loss": 0.2577, + "step": 28173 + }, + { + "epoch": 10.4, + "learning_rate": 1.960807981278483e-06, + "loss": 0.3645, + "step": 28174 + }, + { + "epoch": 10.4, + "learning_rate": 1.9607710309151374e-06, + "loss": 0.527, + "step": 28175 + }, + { + "epoch": 10.4, + "learning_rate": 1.960734080551792e-06, + "loss": 0.3055, + "step": 28176 + }, + { + "epoch": 10.41, + "learning_rate": 1.9606971301884467e-06, + "loss": 0.3473, + "step": 28177 + }, + { + "epoch": 10.41, + "learning_rate": 1.9606601798251018e-06, + "loss": 0.3655, + "step": 28178 + }, + { + "epoch": 10.41, + "learning_rate": 1.9606232294617564e-06, + "loss": 0.3454, + "step": 28179 + }, + { + "epoch": 10.41, + "learning_rate": 1.960586279098411e-06, + "loss": 0.327, + "step": 28180 + }, + { + "epoch": 10.41, + "learning_rate": 1.960549328735066e-06, + "loss": 0.2455, + "step": 28181 + }, + { + "epoch": 10.41, + "learning_rate": 1.9605123783717207e-06, + "loss": 0.2819, + "step": 28182 + }, + { + "epoch": 10.41, + "learning_rate": 1.9604754280083754e-06, + "loss": 0.2861, + "step": 28183 + }, + { + "epoch": 10.41, + "learning_rate": 1.96043847764503e-06, + "loss": 0.3139, + "step": 28184 + }, + { + "epoch": 10.41, + "learning_rate": 1.960401527281685e-06, + "loss": 0.3035, + "step": 28185 + }, + { + "epoch": 10.41, + "learning_rate": 1.9603645769183397e-06, + "loss": 0.2864, + "step": 28186 + }, + { + "epoch": 10.41, + "learning_rate": 1.9603276265549948e-06, + "loss": 0.3186, + "step": 28187 + }, + { + "epoch": 10.41, + "learning_rate": 1.9602906761916494e-06, + "loss": 0.3705, + "step": 28188 + }, + { + "epoch": 10.41, + "learning_rate": 1.960253725828304e-06, + "loss": 0.3213, + "step": 28189 + }, + { + "epoch": 10.41, + "learning_rate": 1.9602167754649587e-06, + "loss": 0.4215, + "step": 28190 + }, + { + "epoch": 10.41, + "learning_rate": 1.9601798251016133e-06, + "loss": 0.3742, + "step": 28191 + }, + { + "epoch": 10.41, + "learning_rate": 1.9601428747382684e-06, + "loss": 0.4528, + "step": 28192 + }, + { + "epoch": 10.41, + "learning_rate": 1.960105924374923e-06, + "loss": 0.3362, + "step": 28193 + }, + { + "epoch": 10.41, + "learning_rate": 1.960068974011578e-06, + "loss": 0.3146, + "step": 28194 + }, + { + "epoch": 10.41, + "learning_rate": 1.9600320236482327e-06, + "loss": 0.3924, + "step": 28195 + }, + { + "epoch": 10.41, + "learning_rate": 1.9599950732848874e-06, + "loss": 0.3279, + "step": 28196 + }, + { + "epoch": 10.41, + "learning_rate": 1.959958122921542e-06, + "loss": 0.242, + "step": 28197 + }, + { + "epoch": 10.41, + "learning_rate": 1.959921172558197e-06, + "loss": 0.3356, + "step": 28198 + }, + { + "epoch": 10.41, + "learning_rate": 1.9598842221948517e-06, + "loss": 0.3627, + "step": 28199 + }, + { + "epoch": 10.41, + "learning_rate": 1.9598472718315063e-06, + "loss": 0.4312, + "step": 28200 + }, + { + "epoch": 10.41, + "learning_rate": 1.9598103214681614e-06, + "loss": 0.3005, + "step": 28201 + }, + { + "epoch": 10.41, + "learning_rate": 1.9597733711048156e-06, + "loss": 0.2117, + "step": 28202 + }, + { + "epoch": 10.41, + "learning_rate": 1.9597364207414707e-06, + "loss": 0.3347, + "step": 28203 + }, + { + "epoch": 10.42, + "learning_rate": 1.9596994703781253e-06, + "loss": 0.3414, + "step": 28204 + }, + { + "epoch": 10.42, + "learning_rate": 1.9596625200147803e-06, + "loss": 0.2986, + "step": 28205 + }, + { + "epoch": 10.42, + "learning_rate": 1.959625569651435e-06, + "loss": 0.3108, + "step": 28206 + }, + { + "epoch": 10.42, + "learning_rate": 1.95958861928809e-06, + "loss": 0.3526, + "step": 28207 + }, + { + "epoch": 10.42, + "learning_rate": 1.9595516689247443e-06, + "loss": 0.36, + "step": 28208 + }, + { + "epoch": 10.42, + "learning_rate": 1.959514718561399e-06, + "loss": 0.4175, + "step": 28209 + }, + { + "epoch": 10.42, + "learning_rate": 1.959477768198054e-06, + "loss": 0.3329, + "step": 28210 + }, + { + "epoch": 10.42, + "learning_rate": 1.9594408178347086e-06, + "loss": 0.3368, + "step": 28211 + }, + { + "epoch": 10.42, + "learning_rate": 1.9594038674713637e-06, + "loss": 0.4281, + "step": 28212 + }, + { + "epoch": 10.42, + "learning_rate": 1.9593669171080183e-06, + "loss": 0.3625, + "step": 28213 + }, + { + "epoch": 10.42, + "learning_rate": 1.959329966744673e-06, + "loss": 0.2722, + "step": 28214 + }, + { + "epoch": 10.42, + "learning_rate": 1.9592930163813276e-06, + "loss": 0.3195, + "step": 28215 + }, + { + "epoch": 10.42, + "learning_rate": 1.9592560660179826e-06, + "loss": 0.3954, + "step": 28216 + }, + { + "epoch": 10.42, + "learning_rate": 1.9592191156546373e-06, + "loss": 0.2264, + "step": 28217 + }, + { + "epoch": 10.42, + "learning_rate": 1.9591821652912923e-06, + "loss": 0.2884, + "step": 28218 + }, + { + "epoch": 10.42, + "learning_rate": 1.959145214927947e-06, + "loss": 0.2666, + "step": 28219 + }, + { + "epoch": 10.42, + "learning_rate": 1.9591082645646016e-06, + "loss": 0.2964, + "step": 28220 + }, + { + "epoch": 10.42, + "learning_rate": 1.9590713142012562e-06, + "loss": 0.338, + "step": 28221 + }, + { + "epoch": 10.42, + "learning_rate": 1.959034363837911e-06, + "loss": 0.4825, + "step": 28222 + }, + { + "epoch": 10.42, + "learning_rate": 1.958997413474566e-06, + "loss": 0.258, + "step": 28223 + }, + { + "epoch": 10.42, + "learning_rate": 1.9589604631112206e-06, + "loss": 0.4789, + "step": 28224 + }, + { + "epoch": 10.42, + "learning_rate": 1.9589235127478756e-06, + "loss": 0.4249, + "step": 28225 + }, + { + "epoch": 10.42, + "learning_rate": 1.9588865623845303e-06, + "loss": 0.3444, + "step": 28226 + }, + { + "epoch": 10.42, + "learning_rate": 1.958849612021185e-06, + "loss": 0.3347, + "step": 28227 + }, + { + "epoch": 10.42, + "learning_rate": 1.9588126616578395e-06, + "loss": 0.4507, + "step": 28228 + }, + { + "epoch": 10.42, + "learning_rate": 1.958775711294494e-06, + "loss": 0.5733, + "step": 28229 + }, + { + "epoch": 10.42, + "learning_rate": 1.9587387609311492e-06, + "loss": 0.4126, + "step": 28230 + }, + { + "epoch": 10.43, + "learning_rate": 1.958701810567804e-06, + "loss": 0.408, + "step": 28231 + }, + { + "epoch": 10.43, + "learning_rate": 1.958664860204459e-06, + "loss": 0.3349, + "step": 28232 + }, + { + "epoch": 10.43, + "learning_rate": 1.9586279098411136e-06, + "loss": 0.3698, + "step": 28233 + }, + { + "epoch": 10.43, + "learning_rate": 1.958590959477768e-06, + "loss": 0.2823, + "step": 28234 + }, + { + "epoch": 10.43, + "learning_rate": 1.958554009114423e-06, + "loss": 0.3702, + "step": 28235 + }, + { + "epoch": 10.43, + "learning_rate": 1.958517058751078e-06, + "loss": 0.3768, + "step": 28236 + }, + { + "epoch": 10.43, + "learning_rate": 1.9584801083877325e-06, + "loss": 0.2952, + "step": 28237 + }, + { + "epoch": 10.43, + "learning_rate": 1.9584431580243876e-06, + "loss": 0.2529, + "step": 28238 + }, + { + "epoch": 10.43, + "learning_rate": 1.9584062076610422e-06, + "loss": 0.1746, + "step": 28239 + }, + { + "epoch": 10.43, + "learning_rate": 1.958369257297697e-06, + "loss": 0.3196, + "step": 28240 + }, + { + "epoch": 10.43, + "learning_rate": 1.9583323069343515e-06, + "loss": 0.3017, + "step": 28241 + }, + { + "epoch": 10.43, + "learning_rate": 1.958295356571006e-06, + "loss": 0.4276, + "step": 28242 + }, + { + "epoch": 10.43, + "learning_rate": 1.958258406207661e-06, + "loss": 0.2489, + "step": 28243 + }, + { + "epoch": 10.43, + "learning_rate": 1.958221455844316e-06, + "loss": 0.2964, + "step": 28244 + }, + { + "epoch": 10.43, + "learning_rate": 1.958184505480971e-06, + "loss": 0.4933, + "step": 28245 + }, + { + "epoch": 10.43, + "learning_rate": 1.9581475551176255e-06, + "loss": 0.3717, + "step": 28246 + }, + { + "epoch": 10.43, + "learning_rate": 1.95811060475428e-06, + "loss": 0.499, + "step": 28247 + }, + { + "epoch": 10.43, + "learning_rate": 1.958073654390935e-06, + "loss": 0.3894, + "step": 28248 + }, + { + "epoch": 10.43, + "learning_rate": 1.9580367040275894e-06, + "loss": 0.3189, + "step": 28249 + }, + { + "epoch": 10.43, + "learning_rate": 1.9579997536642445e-06, + "loss": 0.4924, + "step": 28250 + }, + { + "epoch": 10.43, + "learning_rate": 1.957962803300899e-06, + "loss": 0.3146, + "step": 28251 + }, + { + "epoch": 10.43, + "learning_rate": 1.957925852937554e-06, + "loss": 0.2996, + "step": 28252 + }, + { + "epoch": 10.43, + "learning_rate": 1.9578889025742084e-06, + "loss": 0.2475, + "step": 28253 + }, + { + "epoch": 10.43, + "learning_rate": 1.9578519522108635e-06, + "loss": 0.2153, + "step": 28254 + }, + { + "epoch": 10.43, + "learning_rate": 1.957815001847518e-06, + "loss": 0.381, + "step": 28255 + }, + { + "epoch": 10.43, + "learning_rate": 1.957778051484173e-06, + "loss": 0.2993, + "step": 28256 + }, + { + "epoch": 10.43, + "learning_rate": 1.957741101120828e-06, + "loss": 0.339, + "step": 28257 + }, + { + "epoch": 10.44, + "learning_rate": 1.957704150757483e-06, + "loss": 0.3547, + "step": 28258 + }, + { + "epoch": 10.44, + "learning_rate": 1.957667200394137e-06, + "loss": 0.4188, + "step": 28259 + }, + { + "epoch": 10.44, + "learning_rate": 1.9576302500307917e-06, + "loss": 0.3058, + "step": 28260 + }, + { + "epoch": 10.44, + "learning_rate": 1.9575932996674468e-06, + "loss": 0.2835, + "step": 28261 + }, + { + "epoch": 10.44, + "learning_rate": 1.9575563493041014e-06, + "loss": 0.3291, + "step": 28262 + }, + { + "epoch": 10.44, + "learning_rate": 1.9575193989407565e-06, + "loss": 0.3094, + "step": 28263 + }, + { + "epoch": 10.44, + "learning_rate": 1.957482448577411e-06, + "loss": 0.3168, + "step": 28264 + }, + { + "epoch": 10.44, + "learning_rate": 1.9574454982140657e-06, + "loss": 0.2561, + "step": 28265 + }, + { + "epoch": 10.44, + "learning_rate": 1.9574085478507204e-06, + "loss": 0.283, + "step": 28266 + }, + { + "epoch": 10.44, + "learning_rate": 1.9573715974873754e-06, + "loss": 0.3215, + "step": 28267 + }, + { + "epoch": 10.44, + "learning_rate": 1.95733464712403e-06, + "loss": 0.324, + "step": 28268 + }, + { + "epoch": 10.44, + "learning_rate": 1.9572976967606847e-06, + "loss": 0.2941, + "step": 28269 + }, + { + "epoch": 10.44, + "learning_rate": 1.9572607463973398e-06, + "loss": 0.3411, + "step": 28270 + }, + { + "epoch": 10.44, + "learning_rate": 1.9572237960339944e-06, + "loss": 0.3342, + "step": 28271 + }, + { + "epoch": 10.44, + "learning_rate": 1.957186845670649e-06, + "loss": 0.3685, + "step": 28272 + }, + { + "epoch": 10.44, + "learning_rate": 1.9571498953073037e-06, + "loss": 0.2816, + "step": 28273 + }, + { + "epoch": 10.44, + "learning_rate": 1.9571129449439587e-06, + "loss": 0.2942, + "step": 28274 + }, + { + "epoch": 10.44, + "learning_rate": 1.9570759945806134e-06, + "loss": 0.2895, + "step": 28275 + }, + { + "epoch": 10.44, + "learning_rate": 1.9570390442172684e-06, + "loss": 0.2587, + "step": 28276 + }, + { + "epoch": 10.44, + "learning_rate": 1.957002093853923e-06, + "loss": 0.3556, + "step": 28277 + }, + { + "epoch": 10.44, + "learning_rate": 1.9569651434905777e-06, + "loss": 0.364, + "step": 28278 + }, + { + "epoch": 10.44, + "learning_rate": 1.9569281931272323e-06, + "loss": 0.3705, + "step": 28279 + }, + { + "epoch": 10.44, + "learning_rate": 1.956891242763887e-06, + "loss": 0.3628, + "step": 28280 + }, + { + "epoch": 10.44, + "learning_rate": 1.956854292400542e-06, + "loss": 0.3429, + "step": 28281 + }, + { + "epoch": 10.44, + "learning_rate": 1.9568173420371967e-06, + "loss": 0.336, + "step": 28282 + }, + { + "epoch": 10.44, + "learning_rate": 1.9567803916738517e-06, + "loss": 0.2998, + "step": 28283 + }, + { + "epoch": 10.44, + "learning_rate": 1.9567434413105064e-06, + "loss": 0.3124, + "step": 28284 + }, + { + "epoch": 10.44, + "learning_rate": 1.956706490947161e-06, + "loss": 0.3394, + "step": 28285 + }, + { + "epoch": 10.45, + "learning_rate": 1.9566695405838156e-06, + "loss": 0.315, + "step": 28286 + }, + { + "epoch": 10.45, + "learning_rate": 1.9566325902204707e-06, + "loss": 0.3357, + "step": 28287 + }, + { + "epoch": 10.45, + "learning_rate": 1.9565956398571253e-06, + "loss": 0.3211, + "step": 28288 + }, + { + "epoch": 10.45, + "learning_rate": 1.95655868949378e-06, + "loss": 0.3697, + "step": 28289 + }, + { + "epoch": 10.45, + "learning_rate": 1.956521739130435e-06, + "loss": 0.3859, + "step": 28290 + }, + { + "epoch": 10.45, + "learning_rate": 1.9564847887670897e-06, + "loss": 0.4553, + "step": 28291 + }, + { + "epoch": 10.45, + "learning_rate": 1.9564478384037443e-06, + "loss": 0.1937, + "step": 28292 + }, + { + "epoch": 10.45, + "learning_rate": 1.956410888040399e-06, + "loss": 0.2778, + "step": 28293 + }, + { + "epoch": 10.45, + "learning_rate": 1.956373937677054e-06, + "loss": 0.3198, + "step": 28294 + }, + { + "epoch": 10.45, + "learning_rate": 1.9563369873137086e-06, + "loss": 0.2009, + "step": 28295 + }, + { + "epoch": 10.45, + "learning_rate": 1.9563000369503637e-06, + "loss": 0.3896, + "step": 28296 + }, + { + "epoch": 10.45, + "learning_rate": 1.9562630865870183e-06, + "loss": 0.3987, + "step": 28297 + }, + { + "epoch": 10.45, + "learning_rate": 1.956226136223673e-06, + "loss": 0.4159, + "step": 28298 + }, + { + "epoch": 10.45, + "learning_rate": 1.9561891858603276e-06, + "loss": 0.2891, + "step": 28299 + }, + { + "epoch": 10.45, + "learning_rate": 1.9561522354969822e-06, + "loss": 0.3566, + "step": 28300 + }, + { + "epoch": 10.45, + "learning_rate": 1.9561152851336373e-06, + "loss": 0.2198, + "step": 28301 + }, + { + "epoch": 10.45, + "learning_rate": 1.956078334770292e-06, + "loss": 0.3912, + "step": 28302 + }, + { + "epoch": 10.45, + "learning_rate": 1.956041384406947e-06, + "loss": 0.2893, + "step": 28303 + }, + { + "epoch": 10.45, + "learning_rate": 1.9560044340436012e-06, + "loss": 0.2809, + "step": 28304 + }, + { + "epoch": 10.45, + "learning_rate": 1.9559674836802563e-06, + "loss": 0.3832, + "step": 28305 + }, + { + "epoch": 10.45, + "learning_rate": 1.955930533316911e-06, + "loss": 0.2657, + "step": 28306 + }, + { + "epoch": 10.45, + "learning_rate": 1.955893582953566e-06, + "loss": 0.2586, + "step": 28307 + }, + { + "epoch": 10.45, + "learning_rate": 1.9558566325902206e-06, + "loss": 0.2927, + "step": 28308 + }, + { + "epoch": 10.45, + "learning_rate": 1.9558196822268752e-06, + "loss": 0.3742, + "step": 28309 + }, + { + "epoch": 10.45, + "learning_rate": 1.95578273186353e-06, + "loss": 0.383, + "step": 28310 + }, + { + "epoch": 10.45, + "learning_rate": 1.9557457815001845e-06, + "loss": 0.4043, + "step": 28311 + }, + { + "epoch": 10.45, + "learning_rate": 1.9557088311368396e-06, + "loss": 0.2736, + "step": 28312 + }, + { + "epoch": 10.46, + "learning_rate": 1.9556718807734942e-06, + "loss": 0.3023, + "step": 28313 + }, + { + "epoch": 10.46, + "learning_rate": 1.9556349304101493e-06, + "loss": 0.3402, + "step": 28314 + }, + { + "epoch": 10.46, + "learning_rate": 1.955597980046804e-06, + "loss": 0.2608, + "step": 28315 + }, + { + "epoch": 10.46, + "learning_rate": 1.9555610296834585e-06, + "loss": 0.4175, + "step": 28316 + }, + { + "epoch": 10.46, + "learning_rate": 1.955524079320113e-06, + "loss": 0.3563, + "step": 28317 + }, + { + "epoch": 10.46, + "learning_rate": 1.9554871289567682e-06, + "loss": 0.3045, + "step": 28318 + }, + { + "epoch": 10.46, + "learning_rate": 1.955450178593423e-06, + "loss": 0.4173, + "step": 28319 + }, + { + "epoch": 10.46, + "learning_rate": 1.9554132282300775e-06, + "loss": 0.4475, + "step": 28320 + }, + { + "epoch": 10.46, + "learning_rate": 1.9553762778667326e-06, + "loss": 0.3835, + "step": 28321 + }, + { + "epoch": 10.46, + "learning_rate": 1.9553393275033872e-06, + "loss": 0.3244, + "step": 28322 + }, + { + "epoch": 10.46, + "learning_rate": 1.955302377140042e-06, + "loss": 0.4008, + "step": 28323 + }, + { + "epoch": 10.46, + "learning_rate": 1.9552654267766965e-06, + "loss": 0.3123, + "step": 28324 + }, + { + "epoch": 10.46, + "learning_rate": 1.9552284764133515e-06, + "loss": 0.2704, + "step": 28325 + }, + { + "epoch": 10.46, + "learning_rate": 1.955191526050006e-06, + "loss": 0.3609, + "step": 28326 + }, + { + "epoch": 10.46, + "learning_rate": 1.9551545756866612e-06, + "loss": 0.3348, + "step": 28327 + }, + { + "epoch": 10.46, + "learning_rate": 1.955117625323316e-06, + "loss": 0.36, + "step": 28328 + }, + { + "epoch": 10.46, + "learning_rate": 1.9550806749599705e-06, + "loss": 0.483, + "step": 28329 + }, + { + "epoch": 10.46, + "learning_rate": 1.955043724596625e-06, + "loss": 0.2886, + "step": 28330 + }, + { + "epoch": 10.46, + "learning_rate": 1.9550067742332798e-06, + "loss": 0.3199, + "step": 28331 + }, + { + "epoch": 10.46, + "learning_rate": 1.954969823869935e-06, + "loss": 0.4676, + "step": 28332 + }, + { + "epoch": 10.46, + "learning_rate": 1.9549328735065895e-06, + "loss": 0.3113, + "step": 28333 + }, + { + "epoch": 10.46, + "learning_rate": 1.9548959231432445e-06, + "loss": 0.3905, + "step": 28334 + }, + { + "epoch": 10.46, + "learning_rate": 1.954858972779899e-06, + "loss": 0.4547, + "step": 28335 + }, + { + "epoch": 10.46, + "learning_rate": 1.954822022416554e-06, + "loss": 0.3836, + "step": 28336 + }, + { + "epoch": 10.46, + "learning_rate": 1.9547850720532085e-06, + "loss": 0.2996, + "step": 28337 + }, + { + "epoch": 10.46, + "learning_rate": 1.9547481216898635e-06, + "loss": 0.2528, + "step": 28338 + }, + { + "epoch": 10.46, + "learning_rate": 1.954711171326518e-06, + "loss": 0.4372, + "step": 28339 + }, + { + "epoch": 10.47, + "learning_rate": 1.9546742209631728e-06, + "loss": 0.2963, + "step": 28340 + }, + { + "epoch": 10.47, + "learning_rate": 1.954637270599828e-06, + "loss": 0.2917, + "step": 28341 + }, + { + "epoch": 10.47, + "learning_rate": 1.9546003202364825e-06, + "loss": 0.2598, + "step": 28342 + }, + { + "epoch": 10.47, + "learning_rate": 1.954563369873137e-06, + "loss": 0.423, + "step": 28343 + }, + { + "epoch": 10.47, + "learning_rate": 1.9545264195097918e-06, + "loss": 0.3475, + "step": 28344 + }, + { + "epoch": 10.47, + "learning_rate": 1.954489469146447e-06, + "loss": 0.3996, + "step": 28345 + }, + { + "epoch": 10.47, + "learning_rate": 1.9544525187831015e-06, + "loss": 0.3635, + "step": 28346 + }, + { + "epoch": 10.47, + "learning_rate": 1.9544155684197565e-06, + "loss": 0.2224, + "step": 28347 + }, + { + "epoch": 10.47, + "learning_rate": 1.954378618056411e-06, + "loss": 0.3597, + "step": 28348 + }, + { + "epoch": 10.47, + "learning_rate": 1.9543416676930654e-06, + "loss": 0.2895, + "step": 28349 + }, + { + "epoch": 10.47, + "learning_rate": 1.9543047173297204e-06, + "loss": 0.489, + "step": 28350 + }, + { + "epoch": 10.47, + "learning_rate": 1.954267766966375e-06, + "loss": 0.2954, + "step": 28351 + }, + { + "epoch": 10.47, + "learning_rate": 1.95423081660303e-06, + "loss": 0.3655, + "step": 28352 + }, + { + "epoch": 10.47, + "learning_rate": 1.9541938662396848e-06, + "loss": 0.276, + "step": 28353 + }, + { + "epoch": 10.47, + "learning_rate": 1.9541569158763394e-06, + "loss": 0.4791, + "step": 28354 + }, + { + "epoch": 10.47, + "learning_rate": 1.954119965512994e-06, + "loss": 0.4331, + "step": 28355 + }, + { + "epoch": 10.47, + "learning_rate": 1.954083015149649e-06, + "loss": 0.2959, + "step": 28356 + }, + { + "epoch": 10.47, + "learning_rate": 1.9540460647863037e-06, + "loss": 0.3293, + "step": 28357 + }, + { + "epoch": 10.47, + "learning_rate": 1.9540091144229584e-06, + "loss": 0.399, + "step": 28358 + }, + { + "epoch": 10.47, + "learning_rate": 1.9539721640596134e-06, + "loss": 0.3837, + "step": 28359 + }, + { + "epoch": 10.47, + "learning_rate": 1.953935213696268e-06, + "loss": 0.3544, + "step": 28360 + }, + { + "epoch": 10.47, + "learning_rate": 1.9538982633329227e-06, + "loss": 0.3516, + "step": 28361 + }, + { + "epoch": 10.47, + "learning_rate": 1.9538613129695773e-06, + "loss": 0.3425, + "step": 28362 + }, + { + "epoch": 10.47, + "learning_rate": 1.9538243626062324e-06, + "loss": 0.3088, + "step": 28363 + }, + { + "epoch": 10.47, + "learning_rate": 1.953787412242887e-06, + "loss": 0.3225, + "step": 28364 + }, + { + "epoch": 10.47, + "learning_rate": 1.953750461879542e-06, + "loss": 0.2787, + "step": 28365 + }, + { + "epoch": 10.47, + "learning_rate": 1.9537135115161967e-06, + "loss": 0.3589, + "step": 28366 + }, + { + "epoch": 10.48, + "learning_rate": 1.9536765611528514e-06, + "loss": 0.518, + "step": 28367 + }, + { + "epoch": 10.48, + "learning_rate": 1.953639610789506e-06, + "loss": 0.312, + "step": 28368 + }, + { + "epoch": 10.48, + "learning_rate": 1.9536026604261606e-06, + "loss": 0.3423, + "step": 28369 + }, + { + "epoch": 10.48, + "learning_rate": 1.9535657100628157e-06, + "loss": 0.2655, + "step": 28370 + }, + { + "epoch": 10.48, + "learning_rate": 1.9535287596994703e-06, + "loss": 0.3291, + "step": 28371 + }, + { + "epoch": 10.48, + "learning_rate": 1.9534918093361254e-06, + "loss": 0.2974, + "step": 28372 + }, + { + "epoch": 10.48, + "learning_rate": 1.95345485897278e-06, + "loss": 0.4694, + "step": 28373 + }, + { + "epoch": 10.48, + "learning_rate": 1.9534179086094347e-06, + "loss": 0.4403, + "step": 28374 + }, + { + "epoch": 10.48, + "learning_rate": 1.9533809582460893e-06, + "loss": 0.3325, + "step": 28375 + }, + { + "epoch": 10.48, + "learning_rate": 1.9533440078827444e-06, + "loss": 0.2612, + "step": 28376 + }, + { + "epoch": 10.48, + "learning_rate": 1.953307057519399e-06, + "loss": 0.3342, + "step": 28377 + }, + { + "epoch": 10.48, + "learning_rate": 1.9532701071560536e-06, + "loss": 0.4164, + "step": 28378 + }, + { + "epoch": 10.48, + "learning_rate": 1.9532331567927087e-06, + "loss": 0.241, + "step": 28379 + }, + { + "epoch": 10.48, + "learning_rate": 1.9531962064293633e-06, + "loss": 0.2095, + "step": 28380 + }, + { + "epoch": 10.48, + "learning_rate": 1.953159256066018e-06, + "loss": 0.3621, + "step": 28381 + }, + { + "epoch": 10.48, + "learning_rate": 1.9531223057026726e-06, + "loss": 0.3368, + "step": 28382 + }, + { + "epoch": 10.48, + "learning_rate": 1.9530853553393277e-06, + "loss": 0.4512, + "step": 28383 + }, + { + "epoch": 10.48, + "learning_rate": 1.9530484049759823e-06, + "loss": 0.3449, + "step": 28384 + }, + { + "epoch": 10.48, + "learning_rate": 1.9530114546126374e-06, + "loss": 0.2765, + "step": 28385 + }, + { + "epoch": 10.48, + "learning_rate": 1.952974504249292e-06, + "loss": 0.4125, + "step": 28386 + }, + { + "epoch": 10.48, + "learning_rate": 1.9529375538859466e-06, + "loss": 0.2916, + "step": 28387 + }, + { + "epoch": 10.48, + "learning_rate": 1.9529006035226013e-06, + "loss": 0.2552, + "step": 28388 + }, + { + "epoch": 10.48, + "learning_rate": 1.952863653159256e-06, + "loss": 0.3892, + "step": 28389 + }, + { + "epoch": 10.48, + "learning_rate": 1.952826702795911e-06, + "loss": 0.2847, + "step": 28390 + }, + { + "epoch": 10.48, + "learning_rate": 1.9527897524325656e-06, + "loss": 0.3464, + "step": 28391 + }, + { + "epoch": 10.48, + "learning_rate": 1.9527528020692207e-06, + "loss": 0.2608, + "step": 28392 + }, + { + "epoch": 10.48, + "learning_rate": 1.952715851705875e-06, + "loss": 0.2366, + "step": 28393 + }, + { + "epoch": 10.49, + "learning_rate": 1.95267890134253e-06, + "loss": 0.4523, + "step": 28394 + }, + { + "epoch": 10.49, + "learning_rate": 1.9526419509791846e-06, + "loss": 0.3988, + "step": 28395 + }, + { + "epoch": 10.49, + "learning_rate": 1.9526050006158396e-06, + "loss": 0.4324, + "step": 28396 + }, + { + "epoch": 10.49, + "learning_rate": 1.9525680502524943e-06, + "loss": 0.3975, + "step": 28397 + }, + { + "epoch": 10.49, + "learning_rate": 1.952531099889149e-06, + "loss": 0.2548, + "step": 28398 + }, + { + "epoch": 10.49, + "learning_rate": 1.9524941495258035e-06, + "loss": 0.287, + "step": 28399 + }, + { + "epoch": 10.49, + "learning_rate": 1.952457199162458e-06, + "loss": 0.4843, + "step": 28400 + }, + { + "epoch": 10.49, + "learning_rate": 1.9524202487991132e-06, + "loss": 0.5125, + "step": 28401 + }, + { + "epoch": 10.49, + "learning_rate": 1.952383298435768e-06, + "loss": 0.3421, + "step": 28402 + }, + { + "epoch": 10.49, + "learning_rate": 1.952346348072423e-06, + "loss": 0.2697, + "step": 28403 + }, + { + "epoch": 10.49, + "learning_rate": 1.9523093977090776e-06, + "loss": 0.2314, + "step": 28404 + }, + { + "epoch": 10.49, + "learning_rate": 1.952272447345732e-06, + "loss": 0.2776, + "step": 28405 + }, + { + "epoch": 10.49, + "learning_rate": 1.952235496982387e-06, + "loss": 0.3361, + "step": 28406 + }, + { + "epoch": 10.49, + "learning_rate": 1.952198546619042e-06, + "loss": 0.3323, + "step": 28407 + }, + { + "epoch": 10.49, + "learning_rate": 1.9521615962556965e-06, + "loss": 0.2477, + "step": 28408 + }, + { + "epoch": 10.49, + "learning_rate": 1.952124645892351e-06, + "loss": 0.2288, + "step": 28409 + }, + { + "epoch": 10.49, + "learning_rate": 1.9520876955290062e-06, + "loss": 0.2786, + "step": 28410 + }, + { + "epoch": 10.49, + "learning_rate": 1.952050745165661e-06, + "loss": 0.3104, + "step": 28411 + }, + { + "epoch": 10.49, + "learning_rate": 1.9520137948023155e-06, + "loss": 0.3482, + "step": 28412 + }, + { + "epoch": 10.49, + "learning_rate": 1.95197684443897e-06, + "loss": 0.2961, + "step": 28413 + }, + { + "epoch": 10.49, + "learning_rate": 1.951939894075625e-06, + "loss": 0.2626, + "step": 28414 + }, + { + "epoch": 10.49, + "learning_rate": 1.95190294371228e-06, + "loss": 0.4958, + "step": 28415 + }, + { + "epoch": 10.49, + "learning_rate": 1.951865993348935e-06, + "loss": 0.5138, + "step": 28416 + }, + { + "epoch": 10.49, + "learning_rate": 1.9518290429855895e-06, + "loss": 0.3357, + "step": 28417 + }, + { + "epoch": 10.49, + "learning_rate": 1.951792092622244e-06, + "loss": 0.3979, + "step": 28418 + }, + { + "epoch": 10.49, + "learning_rate": 1.951755142258899e-06, + "loss": 0.2796, + "step": 28419 + }, + { + "epoch": 10.49, + "learning_rate": 1.9517181918955534e-06, + "loss": 0.3081, + "step": 28420 + }, + { + "epoch": 10.5, + "learning_rate": 1.9516812415322085e-06, + "loss": 0.3152, + "step": 28421 + }, + { + "epoch": 10.5, + "learning_rate": 1.951644291168863e-06, + "loss": 0.3531, + "step": 28422 + }, + { + "epoch": 10.5, + "learning_rate": 1.951607340805518e-06, + "loss": 0.2704, + "step": 28423 + }, + { + "epoch": 10.5, + "learning_rate": 1.951570390442173e-06, + "loss": 0.2893, + "step": 28424 + }, + { + "epoch": 10.5, + "learning_rate": 1.9515334400788275e-06, + "loss": 0.3507, + "step": 28425 + }, + { + "epoch": 10.5, + "learning_rate": 1.951496489715482e-06, + "loss": 0.3651, + "step": 28426 + }, + { + "epoch": 10.5, + "learning_rate": 1.951459539352137e-06, + "loss": 0.4197, + "step": 28427 + }, + { + "epoch": 10.5, + "learning_rate": 1.951422588988792e-06, + "loss": 0.4085, + "step": 28428 + }, + { + "epoch": 10.5, + "learning_rate": 1.9513856386254464e-06, + "loss": 0.3021, + "step": 28429 + }, + { + "epoch": 10.5, + "learning_rate": 1.9513486882621015e-06, + "loss": 0.3436, + "step": 28430 + }, + { + "epoch": 10.5, + "learning_rate": 1.951311737898756e-06, + "loss": 0.3612, + "step": 28431 + }, + { + "epoch": 10.5, + "learning_rate": 1.9512747875354108e-06, + "loss": 0.3744, + "step": 28432 + }, + { + "epoch": 10.5, + "learning_rate": 1.9512378371720654e-06, + "loss": 0.3484, + "step": 28433 + }, + { + "epoch": 10.5, + "learning_rate": 1.9512008868087205e-06, + "loss": 0.3337, + "step": 28434 + }, + { + "epoch": 10.5, + "learning_rate": 1.951163936445375e-06, + "loss": 0.2563, + "step": 28435 + }, + { + "epoch": 10.5, + "learning_rate": 1.95112698608203e-06, + "loss": 0.4356, + "step": 28436 + }, + { + "epoch": 10.5, + "learning_rate": 1.951090035718685e-06, + "loss": 0.2988, + "step": 28437 + }, + { + "epoch": 10.5, + "learning_rate": 1.951053085355339e-06, + "loss": 0.3482, + "step": 28438 + }, + { + "epoch": 10.5, + "learning_rate": 1.951016134991994e-06, + "loss": 0.4998, + "step": 28439 + }, + { + "epoch": 10.5, + "learning_rate": 1.9509791846286487e-06, + "loss": 0.534, + "step": 28440 + }, + { + "epoch": 10.5, + "learning_rate": 1.9509422342653038e-06, + "loss": 0.2956, + "step": 28441 + }, + { + "epoch": 10.5, + "learning_rate": 1.9509052839019584e-06, + "loss": 0.3638, + "step": 28442 + }, + { + "epoch": 10.5, + "learning_rate": 1.9508683335386135e-06, + "loss": 0.2175, + "step": 28443 + }, + { + "epoch": 10.5, + "learning_rate": 1.9508313831752677e-06, + "loss": 0.3523, + "step": 28444 + }, + { + "epoch": 10.5, + "learning_rate": 1.9507944328119227e-06, + "loss": 0.4041, + "step": 28445 + }, + { + "epoch": 10.5, + "learning_rate": 1.9507574824485774e-06, + "loss": 0.3058, + "step": 28446 + }, + { + "epoch": 10.5, + "learning_rate": 1.9507205320852324e-06, + "loss": 0.4301, + "step": 28447 + }, + { + "epoch": 10.51, + "learning_rate": 1.950683581721887e-06, + "loss": 0.3205, + "step": 28448 + }, + { + "epoch": 10.51, + "learning_rate": 1.9506466313585417e-06, + "loss": 0.2418, + "step": 28449 + }, + { + "epoch": 10.51, + "learning_rate": 1.9506096809951963e-06, + "loss": 0.3041, + "step": 28450 + }, + { + "epoch": 10.51, + "learning_rate": 1.950572730631851e-06, + "loss": 0.4395, + "step": 28451 + }, + { + "epoch": 10.51, + "learning_rate": 1.950535780268506e-06, + "loss": 0.3615, + "step": 28452 + }, + { + "epoch": 10.51, + "learning_rate": 1.9504988299051607e-06, + "loss": 0.3657, + "step": 28453 + }, + { + "epoch": 10.51, + "learning_rate": 1.9504618795418157e-06, + "loss": 0.3414, + "step": 28454 + }, + { + "epoch": 10.51, + "learning_rate": 1.9504249291784704e-06, + "loss": 0.5168, + "step": 28455 + }, + { + "epoch": 10.51, + "learning_rate": 1.950387978815125e-06, + "loss": 0.37, + "step": 28456 + }, + { + "epoch": 10.51, + "learning_rate": 1.9503510284517796e-06, + "loss": 0.2803, + "step": 28457 + }, + { + "epoch": 10.51, + "learning_rate": 1.9503140780884343e-06, + "loss": 0.3657, + "step": 28458 + }, + { + "epoch": 10.51, + "learning_rate": 1.9502771277250893e-06, + "loss": 0.348, + "step": 28459 + }, + { + "epoch": 10.51, + "learning_rate": 1.950240177361744e-06, + "loss": 0.3571, + "step": 28460 + }, + { + "epoch": 10.51, + "learning_rate": 1.950203226998399e-06, + "loss": 0.2635, + "step": 28461 + }, + { + "epoch": 10.51, + "learning_rate": 1.9501662766350537e-06, + "loss": 0.2196, + "step": 28462 + }, + { + "epoch": 10.51, + "learning_rate": 1.9501293262717083e-06, + "loss": 0.3751, + "step": 28463 + }, + { + "epoch": 10.51, + "learning_rate": 1.950092375908363e-06, + "loss": 0.3419, + "step": 28464 + }, + { + "epoch": 10.51, + "learning_rate": 1.950055425545018e-06, + "loss": 0.3618, + "step": 28465 + }, + { + "epoch": 10.51, + "learning_rate": 1.9500184751816726e-06, + "loss": 0.2865, + "step": 28466 + }, + { + "epoch": 10.51, + "learning_rate": 1.9499815248183277e-06, + "loss": 0.315, + "step": 28467 + }, + { + "epoch": 10.51, + "learning_rate": 1.9499445744549823e-06, + "loss": 0.4051, + "step": 28468 + }, + { + "epoch": 10.51, + "learning_rate": 1.949907624091637e-06, + "loss": 0.3548, + "step": 28469 + }, + { + "epoch": 10.51, + "learning_rate": 1.9498706737282916e-06, + "loss": 0.3224, + "step": 28470 + }, + { + "epoch": 10.51, + "learning_rate": 1.9498337233649463e-06, + "loss": 0.4605, + "step": 28471 + }, + { + "epoch": 10.51, + "learning_rate": 1.9497967730016013e-06, + "loss": 0.6088, + "step": 28472 + }, + { + "epoch": 10.51, + "learning_rate": 1.949759822638256e-06, + "loss": 0.378, + "step": 28473 + }, + { + "epoch": 10.51, + "learning_rate": 1.949722872274911e-06, + "loss": 0.2154, + "step": 28474 + }, + { + "epoch": 10.52, + "learning_rate": 1.9496859219115656e-06, + "loss": 0.2629, + "step": 28475 + }, + { + "epoch": 10.52, + "learning_rate": 1.9496489715482203e-06, + "loss": 0.3223, + "step": 28476 + }, + { + "epoch": 10.52, + "learning_rate": 1.949612021184875e-06, + "loss": 0.3195, + "step": 28477 + }, + { + "epoch": 10.52, + "learning_rate": 1.9495750708215296e-06, + "loss": 0.267, + "step": 28478 + }, + { + "epoch": 10.52, + "learning_rate": 1.9495381204581846e-06, + "loss": 0.2593, + "step": 28479 + }, + { + "epoch": 10.52, + "learning_rate": 1.9495011700948393e-06, + "loss": 0.2648, + "step": 28480 + }, + { + "epoch": 10.52, + "learning_rate": 1.9494642197314943e-06, + "loss": 0.2864, + "step": 28481 + }, + { + "epoch": 10.52, + "learning_rate": 1.949427269368149e-06, + "loss": 0.3576, + "step": 28482 + }, + { + "epoch": 10.52, + "learning_rate": 1.9493903190048036e-06, + "loss": 0.333, + "step": 28483 + }, + { + "epoch": 10.52, + "learning_rate": 1.9493533686414582e-06, + "loss": 0.358, + "step": 28484 + }, + { + "epoch": 10.52, + "learning_rate": 1.9493164182781133e-06, + "loss": 0.2501, + "step": 28485 + }, + { + "epoch": 10.52, + "learning_rate": 1.949279467914768e-06, + "loss": 0.3035, + "step": 28486 + }, + { + "epoch": 10.52, + "learning_rate": 1.949242517551423e-06, + "loss": 0.2268, + "step": 28487 + }, + { + "epoch": 10.52, + "learning_rate": 1.9492055671880776e-06, + "loss": 0.3047, + "step": 28488 + }, + { + "epoch": 10.52, + "learning_rate": 1.949168616824732e-06, + "loss": 0.2308, + "step": 28489 + }, + { + "epoch": 10.52, + "learning_rate": 1.949131666461387e-06, + "loss": 0.4634, + "step": 28490 + }, + { + "epoch": 10.52, + "learning_rate": 1.9490947160980415e-06, + "loss": 0.3579, + "step": 28491 + }, + { + "epoch": 10.52, + "learning_rate": 1.9490577657346966e-06, + "loss": 0.2975, + "step": 28492 + }, + { + "epoch": 10.52, + "learning_rate": 1.9490208153713512e-06, + "loss": 0.373, + "step": 28493 + }, + { + "epoch": 10.52, + "learning_rate": 1.9489838650080063e-06, + "loss": 0.3965, + "step": 28494 + }, + { + "epoch": 10.52, + "learning_rate": 1.9489469146446605e-06, + "loss": 0.3643, + "step": 28495 + }, + { + "epoch": 10.52, + "learning_rate": 1.9489099642813156e-06, + "loss": 0.2648, + "step": 28496 + }, + { + "epoch": 10.52, + "learning_rate": 1.94887301391797e-06, + "loss": 0.3534, + "step": 28497 + }, + { + "epoch": 10.52, + "learning_rate": 1.948836063554625e-06, + "loss": 0.3583, + "step": 28498 + }, + { + "epoch": 10.52, + "learning_rate": 1.94879911319128e-06, + "loss": 0.2903, + "step": 28499 + }, + { + "epoch": 10.52, + "learning_rate": 1.9487621628279345e-06, + "loss": 0.3269, + "step": 28500 + }, + { + "epoch": 10.52, + "learning_rate": 1.948725212464589e-06, + "loss": 0.3731, + "step": 28501 + }, + { + "epoch": 10.53, + "learning_rate": 1.948688262101244e-06, + "loss": 0.408, + "step": 28502 + }, + { + "epoch": 10.53, + "learning_rate": 1.948651311737899e-06, + "loss": 0.277, + "step": 28503 + }, + { + "epoch": 10.53, + "learning_rate": 1.9486143613745535e-06, + "loss": 0.3433, + "step": 28504 + }, + { + "epoch": 10.53, + "learning_rate": 1.9485774110112085e-06, + "loss": 0.306, + "step": 28505 + }, + { + "epoch": 10.53, + "learning_rate": 1.948540460647863e-06, + "loss": 0.3185, + "step": 28506 + }, + { + "epoch": 10.53, + "learning_rate": 1.948503510284518e-06, + "loss": 0.2357, + "step": 28507 + }, + { + "epoch": 10.53, + "learning_rate": 1.9484665599211725e-06, + "loss": 0.412, + "step": 28508 + }, + { + "epoch": 10.53, + "learning_rate": 1.948429609557827e-06, + "loss": 0.3332, + "step": 28509 + }, + { + "epoch": 10.53, + "learning_rate": 1.948392659194482e-06, + "loss": 0.3963, + "step": 28510 + }, + { + "epoch": 10.53, + "learning_rate": 1.948355708831137e-06, + "loss": 0.3674, + "step": 28511 + }, + { + "epoch": 10.53, + "learning_rate": 1.948318758467792e-06, + "loss": 0.258, + "step": 28512 + }, + { + "epoch": 10.53, + "learning_rate": 1.9482818081044465e-06, + "loss": 0.3447, + "step": 28513 + }, + { + "epoch": 10.53, + "learning_rate": 1.948244857741101e-06, + "loss": 0.2816, + "step": 28514 + }, + { + "epoch": 10.53, + "learning_rate": 1.9482079073777558e-06, + "loss": 0.2857, + "step": 28515 + }, + { + "epoch": 10.53, + "learning_rate": 1.948170957014411e-06, + "loss": 0.2777, + "step": 28516 + }, + { + "epoch": 10.53, + "learning_rate": 1.9481340066510655e-06, + "loss": 0.4039, + "step": 28517 + }, + { + "epoch": 10.53, + "learning_rate": 1.94809705628772e-06, + "loss": 0.2413, + "step": 28518 + }, + { + "epoch": 10.53, + "learning_rate": 1.948060105924375e-06, + "loss": 0.3544, + "step": 28519 + }, + { + "epoch": 10.53, + "learning_rate": 1.94802315556103e-06, + "loss": 0.3375, + "step": 28520 + }, + { + "epoch": 10.53, + "learning_rate": 1.9479862051976844e-06, + "loss": 0.3312, + "step": 28521 + }, + { + "epoch": 10.53, + "learning_rate": 1.947949254834339e-06, + "loss": 0.3213, + "step": 28522 + }, + { + "epoch": 10.53, + "learning_rate": 1.947912304470994e-06, + "loss": 0.3162, + "step": 28523 + }, + { + "epoch": 10.53, + "learning_rate": 1.9478753541076488e-06, + "loss": 0.321, + "step": 28524 + }, + { + "epoch": 10.53, + "learning_rate": 1.947838403744304e-06, + "loss": 0.2641, + "step": 28525 + }, + { + "epoch": 10.53, + "learning_rate": 1.9478014533809585e-06, + "loss": 0.3654, + "step": 28526 + }, + { + "epoch": 10.53, + "learning_rate": 1.947764503017613e-06, + "loss": 0.3143, + "step": 28527 + }, + { + "epoch": 10.53, + "learning_rate": 1.9477275526542677e-06, + "loss": 0.3101, + "step": 28528 + }, + { + "epoch": 10.54, + "learning_rate": 1.9476906022909224e-06, + "loss": 0.3027, + "step": 28529 + }, + { + "epoch": 10.54, + "learning_rate": 1.9476536519275774e-06, + "loss": 0.4907, + "step": 28530 + }, + { + "epoch": 10.54, + "learning_rate": 1.947616701564232e-06, + "loss": 0.2726, + "step": 28531 + }, + { + "epoch": 10.54, + "learning_rate": 1.947579751200887e-06, + "loss": 0.341, + "step": 28532 + }, + { + "epoch": 10.54, + "learning_rate": 1.9475428008375418e-06, + "loss": 0.3351, + "step": 28533 + }, + { + "epoch": 10.54, + "learning_rate": 1.9475058504741964e-06, + "loss": 0.252, + "step": 28534 + }, + { + "epoch": 10.54, + "learning_rate": 1.947468900110851e-06, + "loss": 0.3767, + "step": 28535 + }, + { + "epoch": 10.54, + "learning_rate": 1.947431949747506e-06, + "loss": 0.273, + "step": 28536 + }, + { + "epoch": 10.54, + "learning_rate": 1.9473949993841607e-06, + "loss": 0.3258, + "step": 28537 + }, + { + "epoch": 10.54, + "learning_rate": 1.9473580490208154e-06, + "loss": 0.4334, + "step": 28538 + }, + { + "epoch": 10.54, + "learning_rate": 1.94732109865747e-06, + "loss": 0.3522, + "step": 28539 + }, + { + "epoch": 10.54, + "learning_rate": 1.9472841482941246e-06, + "loss": 0.2954, + "step": 28540 + }, + { + "epoch": 10.54, + "learning_rate": 1.9472471979307797e-06, + "loss": 0.3998, + "step": 28541 + }, + { + "epoch": 10.54, + "learning_rate": 1.9472102475674343e-06, + "loss": 0.4074, + "step": 28542 + }, + { + "epoch": 10.54, + "learning_rate": 1.9471732972040894e-06, + "loss": 0.3801, + "step": 28543 + }, + { + "epoch": 10.54, + "learning_rate": 1.947136346840744e-06, + "loss": 0.3827, + "step": 28544 + }, + { + "epoch": 10.54, + "learning_rate": 1.9470993964773987e-06, + "loss": 0.3309, + "step": 28545 + }, + { + "epoch": 10.54, + "learning_rate": 1.9470624461140533e-06, + "loss": 0.4018, + "step": 28546 + }, + { + "epoch": 10.54, + "learning_rate": 1.9470254957507084e-06, + "loss": 0.3876, + "step": 28547 + }, + { + "epoch": 10.54, + "learning_rate": 1.946988545387363e-06, + "loss": 0.3291, + "step": 28548 + }, + { + "epoch": 10.54, + "learning_rate": 1.9469515950240176e-06, + "loss": 0.3144, + "step": 28549 + }, + { + "epoch": 10.54, + "learning_rate": 1.9469146446606727e-06, + "loss": 0.4005, + "step": 28550 + }, + { + "epoch": 10.54, + "learning_rate": 1.9468776942973273e-06, + "loss": 0.3735, + "step": 28551 + }, + { + "epoch": 10.54, + "learning_rate": 1.946840743933982e-06, + "loss": 0.4479, + "step": 28552 + }, + { + "epoch": 10.54, + "learning_rate": 1.9468037935706366e-06, + "loss": 0.5037, + "step": 28553 + }, + { + "epoch": 10.54, + "learning_rate": 1.9467668432072917e-06, + "loss": 0.3426, + "step": 28554 + }, + { + "epoch": 10.54, + "learning_rate": 1.9467298928439463e-06, + "loss": 0.2985, + "step": 28555 + }, + { + "epoch": 10.55, + "learning_rate": 1.9466929424806014e-06, + "loss": 0.2971, + "step": 28556 + }, + { + "epoch": 10.55, + "learning_rate": 1.946655992117256e-06, + "loss": 0.3825, + "step": 28557 + }, + { + "epoch": 10.55, + "learning_rate": 1.9466190417539106e-06, + "loss": 0.3919, + "step": 28558 + }, + { + "epoch": 10.55, + "learning_rate": 1.9465820913905653e-06, + "loss": 0.4557, + "step": 28559 + }, + { + "epoch": 10.55, + "learning_rate": 1.94654514102722e-06, + "loss": 0.3327, + "step": 28560 + }, + { + "epoch": 10.55, + "learning_rate": 1.946508190663875e-06, + "loss": 0.2792, + "step": 28561 + }, + { + "epoch": 10.55, + "learning_rate": 1.9464712403005296e-06, + "loss": 0.4263, + "step": 28562 + }, + { + "epoch": 10.55, + "learning_rate": 1.9464342899371847e-06, + "loss": 0.4432, + "step": 28563 + }, + { + "epoch": 10.55, + "learning_rate": 1.9463973395738393e-06, + "loss": 0.3872, + "step": 28564 + }, + { + "epoch": 10.55, + "learning_rate": 1.946360389210494e-06, + "loss": 0.3601, + "step": 28565 + }, + { + "epoch": 10.55, + "learning_rate": 1.9463234388471486e-06, + "loss": 0.3866, + "step": 28566 + }, + { + "epoch": 10.55, + "learning_rate": 1.9462864884838036e-06, + "loss": 0.2483, + "step": 28567 + }, + { + "epoch": 10.55, + "learning_rate": 1.9462495381204583e-06, + "loss": 0.2547, + "step": 28568 + }, + { + "epoch": 10.55, + "learning_rate": 1.946212587757113e-06, + "loss": 0.5393, + "step": 28569 + }, + { + "epoch": 10.55, + "learning_rate": 1.946175637393768e-06, + "loss": 0.3456, + "step": 28570 + }, + { + "epoch": 10.55, + "learning_rate": 1.9461386870304226e-06, + "loss": 0.3357, + "step": 28571 + }, + { + "epoch": 10.55, + "learning_rate": 1.9461017366670772e-06, + "loss": 0.2822, + "step": 28572 + }, + { + "epoch": 10.55, + "learning_rate": 1.946064786303732e-06, + "loss": 0.3518, + "step": 28573 + }, + { + "epoch": 10.55, + "learning_rate": 1.946027835940387e-06, + "loss": 0.2696, + "step": 28574 + }, + { + "epoch": 10.55, + "learning_rate": 1.9459908855770416e-06, + "loss": 0.2717, + "step": 28575 + }, + { + "epoch": 10.55, + "learning_rate": 1.9459539352136966e-06, + "loss": 0.4405, + "step": 28576 + }, + { + "epoch": 10.55, + "learning_rate": 1.9459169848503513e-06, + "loss": 0.4613, + "step": 28577 + }, + { + "epoch": 10.55, + "learning_rate": 1.9458800344870055e-06, + "loss": 0.2927, + "step": 28578 + }, + { + "epoch": 10.55, + "learning_rate": 1.9458430841236605e-06, + "loss": 0.42, + "step": 28579 + }, + { + "epoch": 10.55, + "learning_rate": 1.945806133760315e-06, + "loss": 0.255, + "step": 28580 + }, + { + "epoch": 10.55, + "learning_rate": 1.9457691833969702e-06, + "loss": 0.2641, + "step": 28581 + }, + { + "epoch": 10.55, + "learning_rate": 1.945732233033625e-06, + "loss": 0.3855, + "step": 28582 + }, + { + "epoch": 10.56, + "learning_rate": 1.94569528267028e-06, + "loss": 0.1995, + "step": 28583 + }, + { + "epoch": 10.56, + "learning_rate": 1.945658332306934e-06, + "loss": 0.1855, + "step": 28584 + }, + { + "epoch": 10.56, + "learning_rate": 1.945621381943589e-06, + "loss": 0.3414, + "step": 28585 + }, + { + "epoch": 10.56, + "learning_rate": 1.945584431580244e-06, + "loss": 0.2399, + "step": 28586 + }, + { + "epoch": 10.56, + "learning_rate": 1.9455474812168985e-06, + "loss": 0.3046, + "step": 28587 + }, + { + "epoch": 10.56, + "learning_rate": 1.9455105308535535e-06, + "loss": 0.3761, + "step": 28588 + }, + { + "epoch": 10.56, + "learning_rate": 1.945473580490208e-06, + "loss": 0.4089, + "step": 28589 + }, + { + "epoch": 10.56, + "learning_rate": 1.945436630126863e-06, + "loss": 0.3395, + "step": 28590 + }, + { + "epoch": 10.56, + "learning_rate": 1.9453996797635174e-06, + "loss": 0.3754, + "step": 28591 + }, + { + "epoch": 10.56, + "learning_rate": 1.9453627294001725e-06, + "loss": 0.3541, + "step": 28592 + }, + { + "epoch": 10.56, + "learning_rate": 1.945325779036827e-06, + "loss": 0.3578, + "step": 28593 + }, + { + "epoch": 10.56, + "learning_rate": 1.945288828673482e-06, + "loss": 0.4696, + "step": 28594 + }, + { + "epoch": 10.56, + "learning_rate": 1.945251878310137e-06, + "loss": 0.5093, + "step": 28595 + }, + { + "epoch": 10.56, + "learning_rate": 1.9452149279467915e-06, + "loss": 0.3053, + "step": 28596 + }, + { + "epoch": 10.56, + "learning_rate": 1.945177977583446e-06, + "loss": 0.3023, + "step": 28597 + }, + { + "epoch": 10.56, + "learning_rate": 1.9451410272201007e-06, + "loss": 0.3174, + "step": 28598 + }, + { + "epoch": 10.56, + "learning_rate": 1.945104076856756e-06, + "loss": 0.2546, + "step": 28599 + }, + { + "epoch": 10.56, + "learning_rate": 1.9450671264934104e-06, + "loss": 0.4115, + "step": 28600 + }, + { + "epoch": 10.56, + "learning_rate": 1.9450301761300655e-06, + "loss": 0.3545, + "step": 28601 + }, + { + "epoch": 10.56, + "learning_rate": 1.94499322576672e-06, + "loss": 0.4065, + "step": 28602 + }, + { + "epoch": 10.56, + "learning_rate": 1.9449562754033748e-06, + "loss": 0.2238, + "step": 28603 + }, + { + "epoch": 10.56, + "learning_rate": 1.9449193250400294e-06, + "loss": 0.429, + "step": 28604 + }, + { + "epoch": 10.56, + "learning_rate": 1.9448823746766845e-06, + "loss": 0.2889, + "step": 28605 + }, + { + "epoch": 10.56, + "learning_rate": 1.944845424313339e-06, + "loss": 0.3136, + "step": 28606 + }, + { + "epoch": 10.56, + "learning_rate": 1.9448084739499937e-06, + "loss": 0.3448, + "step": 28607 + }, + { + "epoch": 10.56, + "learning_rate": 1.944771523586649e-06, + "loss": 0.3061, + "step": 28608 + }, + { + "epoch": 10.56, + "learning_rate": 1.9447345732233034e-06, + "loss": 0.2464, + "step": 28609 + }, + { + "epoch": 10.56, + "learning_rate": 1.944697622859958e-06, + "loss": 0.4012, + "step": 28610 + }, + { + "epoch": 10.57, + "learning_rate": 1.9446606724966127e-06, + "loss": 0.2092, + "step": 28611 + }, + { + "epoch": 10.57, + "learning_rate": 1.9446237221332678e-06, + "loss": 0.5592, + "step": 28612 + }, + { + "epoch": 10.57, + "learning_rate": 1.9445867717699224e-06, + "loss": 0.3483, + "step": 28613 + }, + { + "epoch": 10.57, + "learning_rate": 1.9445498214065775e-06, + "loss": 0.364, + "step": 28614 + }, + { + "epoch": 10.57, + "learning_rate": 1.944512871043232e-06, + "loss": 0.4314, + "step": 28615 + }, + { + "epoch": 10.57, + "learning_rate": 1.9444759206798867e-06, + "loss": 0.4769, + "step": 28616 + }, + { + "epoch": 10.57, + "learning_rate": 1.9444389703165414e-06, + "loss": 0.3885, + "step": 28617 + }, + { + "epoch": 10.57, + "learning_rate": 1.944402019953196e-06, + "loss": 0.2776, + "step": 28618 + }, + { + "epoch": 10.57, + "learning_rate": 1.944365069589851e-06, + "loss": 0.2651, + "step": 28619 + }, + { + "epoch": 10.57, + "learning_rate": 1.9443281192265057e-06, + "loss": 0.3747, + "step": 28620 + }, + { + "epoch": 10.57, + "learning_rate": 1.9442911688631608e-06, + "loss": 0.3093, + "step": 28621 + }, + { + "epoch": 10.57, + "learning_rate": 1.9442542184998154e-06, + "loss": 0.4377, + "step": 28622 + }, + { + "epoch": 10.57, + "learning_rate": 1.94421726813647e-06, + "loss": 0.2599, + "step": 28623 + }, + { + "epoch": 10.57, + "learning_rate": 1.9441803177731247e-06, + "loss": 0.4188, + "step": 28624 + }, + { + "epoch": 10.57, + "learning_rate": 1.9441433674097797e-06, + "loss": 0.2173, + "step": 28625 + }, + { + "epoch": 10.57, + "learning_rate": 1.9441064170464344e-06, + "loss": 0.2401, + "step": 28626 + }, + { + "epoch": 10.57, + "learning_rate": 1.944069466683089e-06, + "loss": 0.3059, + "step": 28627 + }, + { + "epoch": 10.57, + "learning_rate": 1.944032516319744e-06, + "loss": 0.3509, + "step": 28628 + }, + { + "epoch": 10.57, + "learning_rate": 1.9439955659563983e-06, + "loss": 0.3154, + "step": 28629 + }, + { + "epoch": 10.57, + "learning_rate": 1.9439586155930534e-06, + "loss": 0.2835, + "step": 28630 + }, + { + "epoch": 10.57, + "learning_rate": 1.943921665229708e-06, + "loss": 0.2953, + "step": 28631 + }, + { + "epoch": 10.57, + "learning_rate": 1.943884714866363e-06, + "loss": 0.2848, + "step": 28632 + }, + { + "epoch": 10.57, + "learning_rate": 1.9438477645030177e-06, + "loss": 0.4164, + "step": 28633 + }, + { + "epoch": 10.57, + "learning_rate": 1.9438108141396727e-06, + "loss": 0.3363, + "step": 28634 + }, + { + "epoch": 10.57, + "learning_rate": 1.943773863776327e-06, + "loss": 0.2956, + "step": 28635 + }, + { + "epoch": 10.57, + "learning_rate": 1.943736913412982e-06, + "loss": 0.2091, + "step": 28636 + }, + { + "epoch": 10.57, + "learning_rate": 1.9436999630496367e-06, + "loss": 0.3622, + "step": 28637 + }, + { + "epoch": 10.58, + "learning_rate": 1.9436630126862913e-06, + "loss": 0.3469, + "step": 28638 + }, + { + "epoch": 10.58, + "learning_rate": 1.9436260623229463e-06, + "loss": 0.2501, + "step": 28639 + }, + { + "epoch": 10.58, + "learning_rate": 1.943589111959601e-06, + "loss": 0.3689, + "step": 28640 + }, + { + "epoch": 10.58, + "learning_rate": 1.9435521615962556e-06, + "loss": 0.2502, + "step": 28641 + }, + { + "epoch": 10.58, + "learning_rate": 1.9435152112329103e-06, + "loss": 0.3939, + "step": 28642 + }, + { + "epoch": 10.58, + "learning_rate": 1.9434782608695653e-06, + "loss": 0.3206, + "step": 28643 + }, + { + "epoch": 10.58, + "learning_rate": 1.94344131050622e-06, + "loss": 0.3567, + "step": 28644 + }, + { + "epoch": 10.58, + "learning_rate": 1.943404360142875e-06, + "loss": 0.3196, + "step": 28645 + }, + { + "epoch": 10.58, + "learning_rate": 1.9433674097795297e-06, + "loss": 0.435, + "step": 28646 + }, + { + "epoch": 10.58, + "learning_rate": 1.9433304594161843e-06, + "loss": 0.3287, + "step": 28647 + }, + { + "epoch": 10.58, + "learning_rate": 1.943293509052839e-06, + "loss": 0.3384, + "step": 28648 + }, + { + "epoch": 10.58, + "learning_rate": 1.9432565586894936e-06, + "loss": 0.278, + "step": 28649 + }, + { + "epoch": 10.58, + "learning_rate": 1.9432196083261486e-06, + "loss": 0.2467, + "step": 28650 + }, + { + "epoch": 10.58, + "learning_rate": 1.9431826579628033e-06, + "loss": 0.5492, + "step": 28651 + }, + { + "epoch": 10.58, + "learning_rate": 1.9431457075994583e-06, + "loss": 0.2376, + "step": 28652 + }, + { + "epoch": 10.58, + "learning_rate": 1.943108757236113e-06, + "loss": 0.2732, + "step": 28653 + }, + { + "epoch": 10.58, + "learning_rate": 1.9430718068727676e-06, + "loss": 0.3072, + "step": 28654 + }, + { + "epoch": 10.58, + "learning_rate": 1.9430348565094222e-06, + "loss": 0.2204, + "step": 28655 + }, + { + "epoch": 10.58, + "learning_rate": 1.9429979061460773e-06, + "loss": 0.3649, + "step": 28656 + }, + { + "epoch": 10.58, + "learning_rate": 1.942960955782732e-06, + "loss": 0.3309, + "step": 28657 + }, + { + "epoch": 10.58, + "learning_rate": 1.9429240054193866e-06, + "loss": 0.3851, + "step": 28658 + }, + { + "epoch": 10.58, + "learning_rate": 1.9428870550560416e-06, + "loss": 0.4294, + "step": 28659 + }, + { + "epoch": 10.58, + "learning_rate": 1.9428501046926963e-06, + "loss": 0.2569, + "step": 28660 + }, + { + "epoch": 10.58, + "learning_rate": 1.942813154329351e-06, + "loss": 0.3229, + "step": 28661 + }, + { + "epoch": 10.58, + "learning_rate": 1.9427762039660055e-06, + "loss": 0.3268, + "step": 28662 + }, + { + "epoch": 10.58, + "learning_rate": 1.9427392536026606e-06, + "loss": 0.295, + "step": 28663 + }, + { + "epoch": 10.58, + "learning_rate": 1.9427023032393152e-06, + "loss": 0.2433, + "step": 28664 + }, + { + "epoch": 10.59, + "learning_rate": 1.9426653528759703e-06, + "loss": 0.4248, + "step": 28665 + }, + { + "epoch": 10.59, + "learning_rate": 1.942628402512625e-06, + "loss": 0.3802, + "step": 28666 + }, + { + "epoch": 10.59, + "learning_rate": 1.9425914521492796e-06, + "loss": 0.3215, + "step": 28667 + }, + { + "epoch": 10.59, + "learning_rate": 1.942554501785934e-06, + "loss": 0.2731, + "step": 28668 + }, + { + "epoch": 10.59, + "learning_rate": 1.942517551422589e-06, + "loss": 0.2835, + "step": 28669 + }, + { + "epoch": 10.59, + "learning_rate": 1.942480601059244e-06, + "loss": 0.2944, + "step": 28670 + }, + { + "epoch": 10.59, + "learning_rate": 1.9424436506958985e-06, + "loss": 0.2687, + "step": 28671 + }, + { + "epoch": 10.59, + "learning_rate": 1.9424067003325536e-06, + "loss": 0.3075, + "step": 28672 + }, + { + "epoch": 10.59, + "learning_rate": 1.9423697499692082e-06, + "loss": 0.3519, + "step": 28673 + }, + { + "epoch": 10.59, + "learning_rate": 1.942332799605863e-06, + "loss": 0.3199, + "step": 28674 + }, + { + "epoch": 10.59, + "learning_rate": 1.9422958492425175e-06, + "loss": 0.2449, + "step": 28675 + }, + { + "epoch": 10.59, + "learning_rate": 1.9422588988791726e-06, + "loss": 0.4136, + "step": 28676 + }, + { + "epoch": 10.59, + "learning_rate": 1.942221948515827e-06, + "loss": 0.4801, + "step": 28677 + }, + { + "epoch": 10.59, + "learning_rate": 1.942184998152482e-06, + "loss": 0.4082, + "step": 28678 + }, + { + "epoch": 10.59, + "learning_rate": 1.942148047789137e-06, + "loss": 0.4803, + "step": 28679 + }, + { + "epoch": 10.59, + "learning_rate": 1.942111097425791e-06, + "loss": 0.4753, + "step": 28680 + }, + { + "epoch": 10.59, + "learning_rate": 1.942074147062446e-06, + "loss": 0.3858, + "step": 28681 + }, + { + "epoch": 10.59, + "learning_rate": 1.942037196699101e-06, + "loss": 0.4029, + "step": 28682 + }, + { + "epoch": 10.59, + "learning_rate": 1.942000246335756e-06, + "loss": 0.4127, + "step": 28683 + }, + { + "epoch": 10.59, + "learning_rate": 1.9419632959724105e-06, + "loss": 0.2941, + "step": 28684 + }, + { + "epoch": 10.59, + "learning_rate": 1.9419263456090656e-06, + "loss": 0.428, + "step": 28685 + }, + { + "epoch": 10.59, + "learning_rate": 1.9418893952457198e-06, + "loss": 0.1811, + "step": 28686 + }, + { + "epoch": 10.59, + "learning_rate": 1.9418524448823744e-06, + "loss": 0.3494, + "step": 28687 + }, + { + "epoch": 10.59, + "learning_rate": 1.9418154945190295e-06, + "loss": 0.3288, + "step": 28688 + }, + { + "epoch": 10.59, + "learning_rate": 1.941778544155684e-06, + "loss": 0.2162, + "step": 28689 + }, + { + "epoch": 10.59, + "learning_rate": 1.941741593792339e-06, + "loss": 0.3483, + "step": 28690 + }, + { + "epoch": 10.59, + "learning_rate": 1.941704643428994e-06, + "loss": 0.3508, + "step": 28691 + }, + { + "epoch": 10.6, + "learning_rate": 1.9416676930656484e-06, + "loss": 0.4598, + "step": 28692 + }, + { + "epoch": 10.6, + "learning_rate": 1.941630742702303e-06, + "loss": 0.1846, + "step": 28693 + }, + { + "epoch": 10.6, + "learning_rate": 1.941593792338958e-06, + "loss": 0.2731, + "step": 28694 + }, + { + "epoch": 10.6, + "learning_rate": 1.9415568419756128e-06, + "loss": 0.2531, + "step": 28695 + }, + { + "epoch": 10.6, + "learning_rate": 1.941519891612268e-06, + "loss": 0.304, + "step": 28696 + }, + { + "epoch": 10.6, + "learning_rate": 1.9414829412489225e-06, + "loss": 0.3048, + "step": 28697 + }, + { + "epoch": 10.6, + "learning_rate": 1.941445990885577e-06, + "loss": 0.2897, + "step": 28698 + }, + { + "epoch": 10.6, + "learning_rate": 1.9414090405222317e-06, + "loss": 0.4012, + "step": 28699 + }, + { + "epoch": 10.6, + "learning_rate": 1.9413720901588864e-06, + "loss": 0.3192, + "step": 28700 + }, + { + "epoch": 10.6, + "learning_rate": 1.9413351397955414e-06, + "loss": 0.3921, + "step": 28701 + }, + { + "epoch": 10.6, + "learning_rate": 1.941298189432196e-06, + "loss": 0.2913, + "step": 28702 + }, + { + "epoch": 10.6, + "learning_rate": 1.941261239068851e-06, + "loss": 0.3576, + "step": 28703 + }, + { + "epoch": 10.6, + "learning_rate": 1.9412242887055058e-06, + "loss": 0.269, + "step": 28704 + }, + { + "epoch": 10.6, + "learning_rate": 1.9411873383421604e-06, + "loss": 0.506, + "step": 28705 + }, + { + "epoch": 10.6, + "learning_rate": 1.941150387978815e-06, + "loss": 0.2752, + "step": 28706 + }, + { + "epoch": 10.6, + "learning_rate": 1.9411134376154697e-06, + "loss": 0.3998, + "step": 28707 + }, + { + "epoch": 10.6, + "learning_rate": 1.9410764872521247e-06, + "loss": 0.3275, + "step": 28708 + }, + { + "epoch": 10.6, + "learning_rate": 1.9410395368887794e-06, + "loss": 0.5657, + "step": 28709 + }, + { + "epoch": 10.6, + "learning_rate": 1.9410025865254344e-06, + "loss": 0.1992, + "step": 28710 + }, + { + "epoch": 10.6, + "learning_rate": 1.940965636162089e-06, + "loss": 0.4436, + "step": 28711 + }, + { + "epoch": 10.6, + "learning_rate": 1.9409286857987437e-06, + "loss": 0.5225, + "step": 28712 + }, + { + "epoch": 10.6, + "learning_rate": 1.9408917354353983e-06, + "loss": 0.3617, + "step": 28713 + }, + { + "epoch": 10.6, + "learning_rate": 1.9408547850720534e-06, + "loss": 0.4039, + "step": 28714 + }, + { + "epoch": 10.6, + "learning_rate": 1.940817834708708e-06, + "loss": 0.3676, + "step": 28715 + }, + { + "epoch": 10.6, + "learning_rate": 1.940780884345363e-06, + "loss": 0.3222, + "step": 28716 + }, + { + "epoch": 10.6, + "learning_rate": 1.9407439339820177e-06, + "loss": 0.3249, + "step": 28717 + }, + { + "epoch": 10.6, + "learning_rate": 1.9407069836186724e-06, + "loss": 0.4277, + "step": 28718 + }, + { + "epoch": 10.61, + "learning_rate": 1.940670033255327e-06, + "loss": 0.3579, + "step": 28719 + }, + { + "epoch": 10.61, + "learning_rate": 1.9406330828919816e-06, + "loss": 0.3369, + "step": 28720 + }, + { + "epoch": 10.61, + "learning_rate": 1.9405961325286367e-06, + "loss": 0.2493, + "step": 28721 + }, + { + "epoch": 10.61, + "learning_rate": 1.9405591821652913e-06, + "loss": 0.3848, + "step": 28722 + }, + { + "epoch": 10.61, + "learning_rate": 1.9405222318019464e-06, + "loss": 0.2855, + "step": 28723 + }, + { + "epoch": 10.61, + "learning_rate": 1.940485281438601e-06, + "loss": 0.22, + "step": 28724 + }, + { + "epoch": 10.61, + "learning_rate": 1.9404483310752557e-06, + "loss": 0.2514, + "step": 28725 + }, + { + "epoch": 10.61, + "learning_rate": 1.9404113807119103e-06, + "loss": 0.235, + "step": 28726 + }, + { + "epoch": 10.61, + "learning_rate": 1.940374430348565e-06, + "loss": 0.3112, + "step": 28727 + }, + { + "epoch": 10.61, + "learning_rate": 1.94033747998522e-06, + "loss": 0.3394, + "step": 28728 + }, + { + "epoch": 10.61, + "learning_rate": 1.9403005296218746e-06, + "loss": 0.3072, + "step": 28729 + }, + { + "epoch": 10.61, + "learning_rate": 1.9402635792585293e-06, + "loss": 0.2476, + "step": 28730 + }, + { + "epoch": 10.61, + "learning_rate": 1.940226628895184e-06, + "loss": 0.3784, + "step": 28731 + }, + { + "epoch": 10.61, + "learning_rate": 1.940189678531839e-06, + "loss": 0.4734, + "step": 28732 + }, + { + "epoch": 10.61, + "learning_rate": 1.9401527281684936e-06, + "loss": 0.2843, + "step": 28733 + }, + { + "epoch": 10.61, + "learning_rate": 1.9401157778051487e-06, + "loss": 0.257, + "step": 28734 + }, + { + "epoch": 10.61, + "learning_rate": 1.9400788274418033e-06, + "loss": 0.3211, + "step": 28735 + }, + { + "epoch": 10.61, + "learning_rate": 1.940041877078458e-06, + "loss": 0.2373, + "step": 28736 + }, + { + "epoch": 10.61, + "learning_rate": 1.9400049267151126e-06, + "loss": 0.2755, + "step": 28737 + }, + { + "epoch": 10.61, + "learning_rate": 1.9399679763517672e-06, + "loss": 0.3916, + "step": 28738 + }, + { + "epoch": 10.61, + "learning_rate": 1.9399310259884223e-06, + "loss": 0.4198, + "step": 28739 + }, + { + "epoch": 10.61, + "learning_rate": 1.939894075625077e-06, + "loss": 0.2058, + "step": 28740 + }, + { + "epoch": 10.61, + "learning_rate": 1.939857125261732e-06, + "loss": 0.377, + "step": 28741 + }, + { + "epoch": 10.61, + "learning_rate": 1.9398201748983866e-06, + "loss": 0.1821, + "step": 28742 + }, + { + "epoch": 10.61, + "learning_rate": 1.9397832245350412e-06, + "loss": 0.4385, + "step": 28743 + }, + { + "epoch": 10.61, + "learning_rate": 1.939746274171696e-06, + "loss": 0.388, + "step": 28744 + }, + { + "epoch": 10.61, + "learning_rate": 1.939709323808351e-06, + "loss": 0.3912, + "step": 28745 + }, + { + "epoch": 10.62, + "learning_rate": 1.9396723734450056e-06, + "loss": 0.3309, + "step": 28746 + }, + { + "epoch": 10.62, + "learning_rate": 1.9396354230816602e-06, + "loss": 0.4611, + "step": 28747 + }, + { + "epoch": 10.62, + "learning_rate": 1.9395984727183153e-06, + "loss": 0.3257, + "step": 28748 + }, + { + "epoch": 10.62, + "learning_rate": 1.93956152235497e-06, + "loss": 0.3821, + "step": 28749 + }, + { + "epoch": 10.62, + "learning_rate": 1.9395245719916245e-06, + "loss": 0.2771, + "step": 28750 + }, + { + "epoch": 10.62, + "learning_rate": 1.939487621628279e-06, + "loss": 0.3801, + "step": 28751 + }, + { + "epoch": 10.62, + "learning_rate": 1.9394506712649342e-06, + "loss": 0.4132, + "step": 28752 + }, + { + "epoch": 10.62, + "learning_rate": 1.939413720901589e-06, + "loss": 0.3369, + "step": 28753 + }, + { + "epoch": 10.62, + "learning_rate": 1.939376770538244e-06, + "loss": 0.3532, + "step": 28754 + }, + { + "epoch": 10.62, + "learning_rate": 1.9393398201748986e-06, + "loss": 0.3167, + "step": 28755 + }, + { + "epoch": 10.62, + "learning_rate": 1.939302869811553e-06, + "loss": 0.4756, + "step": 28756 + }, + { + "epoch": 10.62, + "learning_rate": 1.939265919448208e-06, + "loss": 0.3705, + "step": 28757 + }, + { + "epoch": 10.62, + "learning_rate": 1.9392289690848625e-06, + "loss": 0.3946, + "step": 28758 + }, + { + "epoch": 10.62, + "learning_rate": 1.9391920187215175e-06, + "loss": 0.322, + "step": 28759 + }, + { + "epoch": 10.62, + "learning_rate": 1.939155068358172e-06, + "loss": 0.3031, + "step": 28760 + }, + { + "epoch": 10.62, + "learning_rate": 1.9391181179948272e-06, + "loss": 0.3481, + "step": 28761 + }, + { + "epoch": 10.62, + "learning_rate": 1.939081167631482e-06, + "loss": 0.23, + "step": 28762 + }, + { + "epoch": 10.62, + "learning_rate": 1.9390442172681365e-06, + "loss": 0.3118, + "step": 28763 + }, + { + "epoch": 10.62, + "learning_rate": 1.939007266904791e-06, + "loss": 0.4083, + "step": 28764 + }, + { + "epoch": 10.62, + "learning_rate": 1.938970316541446e-06, + "loss": 0.4631, + "step": 28765 + }, + { + "epoch": 10.62, + "learning_rate": 1.938933366178101e-06, + "loss": 0.3289, + "step": 28766 + }, + { + "epoch": 10.62, + "learning_rate": 1.9388964158147555e-06, + "loss": 0.3804, + "step": 28767 + }, + { + "epoch": 10.62, + "learning_rate": 1.9388594654514105e-06, + "loss": 0.3892, + "step": 28768 + }, + { + "epoch": 10.62, + "learning_rate": 1.9388225150880648e-06, + "loss": 0.2908, + "step": 28769 + }, + { + "epoch": 10.62, + "learning_rate": 1.93878556472472e-06, + "loss": 0.3985, + "step": 28770 + }, + { + "epoch": 10.62, + "learning_rate": 1.9387486143613745e-06, + "loss": 0.3758, + "step": 28771 + }, + { + "epoch": 10.62, + "learning_rate": 1.9387116639980295e-06, + "loss": 0.2955, + "step": 28772 + }, + { + "epoch": 10.63, + "learning_rate": 1.938674713634684e-06, + "loss": 0.3487, + "step": 28773 + }, + { + "epoch": 10.63, + "learning_rate": 1.938637763271339e-06, + "loss": 0.2766, + "step": 28774 + }, + { + "epoch": 10.63, + "learning_rate": 1.9386008129079934e-06, + "loss": 0.3278, + "step": 28775 + }, + { + "epoch": 10.63, + "learning_rate": 1.9385638625446485e-06, + "loss": 0.2535, + "step": 28776 + }, + { + "epoch": 10.63, + "learning_rate": 1.938526912181303e-06, + "loss": 0.3902, + "step": 28777 + }, + { + "epoch": 10.63, + "learning_rate": 1.9384899618179578e-06, + "loss": 0.3231, + "step": 28778 + }, + { + "epoch": 10.63, + "learning_rate": 1.938453011454613e-06, + "loss": 0.3312, + "step": 28779 + }, + { + "epoch": 10.63, + "learning_rate": 1.9384160610912675e-06, + "loss": 0.3903, + "step": 28780 + }, + { + "epoch": 10.63, + "learning_rate": 1.938379110727922e-06, + "loss": 0.2406, + "step": 28781 + }, + { + "epoch": 10.63, + "learning_rate": 1.9383421603645767e-06, + "loss": 0.352, + "step": 28782 + }, + { + "epoch": 10.63, + "learning_rate": 1.9383052100012318e-06, + "loss": 0.2708, + "step": 28783 + }, + { + "epoch": 10.63, + "learning_rate": 1.9382682596378864e-06, + "loss": 0.2956, + "step": 28784 + }, + { + "epoch": 10.63, + "learning_rate": 1.9382313092745415e-06, + "loss": 0.4276, + "step": 28785 + }, + { + "epoch": 10.63, + "learning_rate": 1.938194358911196e-06, + "loss": 0.3469, + "step": 28786 + }, + { + "epoch": 10.63, + "learning_rate": 1.9381574085478508e-06, + "loss": 0.3479, + "step": 28787 + }, + { + "epoch": 10.63, + "learning_rate": 1.9381204581845054e-06, + "loss": 0.2284, + "step": 28788 + }, + { + "epoch": 10.63, + "learning_rate": 1.93808350782116e-06, + "loss": 0.3326, + "step": 28789 + }, + { + "epoch": 10.63, + "learning_rate": 1.938046557457815e-06, + "loss": 0.3545, + "step": 28790 + }, + { + "epoch": 10.63, + "learning_rate": 1.9380096070944697e-06, + "loss": 0.2649, + "step": 28791 + }, + { + "epoch": 10.63, + "learning_rate": 1.9379726567311248e-06, + "loss": 0.2638, + "step": 28792 + }, + { + "epoch": 10.63, + "learning_rate": 1.9379357063677794e-06, + "loss": 0.301, + "step": 28793 + }, + { + "epoch": 10.63, + "learning_rate": 1.937898756004434e-06, + "loss": 0.46, + "step": 28794 + }, + { + "epoch": 10.63, + "learning_rate": 1.9378618056410887e-06, + "loss": 0.3422, + "step": 28795 + }, + { + "epoch": 10.63, + "learning_rate": 1.9378248552777438e-06, + "loss": 0.3223, + "step": 28796 + }, + { + "epoch": 10.63, + "learning_rate": 1.9377879049143984e-06, + "loss": 0.2947, + "step": 28797 + }, + { + "epoch": 10.63, + "learning_rate": 1.937750954551053e-06, + "loss": 0.3398, + "step": 28798 + }, + { + "epoch": 10.63, + "learning_rate": 1.937714004187708e-06, + "loss": 0.2672, + "step": 28799 + }, + { + "epoch": 10.64, + "learning_rate": 1.9376770538243627e-06, + "loss": 0.5269, + "step": 28800 + }, + { + "epoch": 10.64, + "learning_rate": 1.9376401034610174e-06, + "loss": 0.3428, + "step": 28801 + }, + { + "epoch": 10.64, + "learning_rate": 1.937603153097672e-06, + "loss": 0.3499, + "step": 28802 + }, + { + "epoch": 10.64, + "learning_rate": 1.937566202734327e-06, + "loss": 0.3588, + "step": 28803 + }, + { + "epoch": 10.64, + "learning_rate": 1.9375292523709817e-06, + "loss": 0.2253, + "step": 28804 + }, + { + "epoch": 10.64, + "learning_rate": 1.9374923020076367e-06, + "loss": 0.3524, + "step": 28805 + }, + { + "epoch": 10.64, + "learning_rate": 1.9374553516442914e-06, + "loss": 0.3842, + "step": 28806 + }, + { + "epoch": 10.64, + "learning_rate": 1.937418401280946e-06, + "loss": 0.4047, + "step": 28807 + }, + { + "epoch": 10.64, + "learning_rate": 1.9373814509176007e-06, + "loss": 0.2923, + "step": 28808 + }, + { + "epoch": 10.64, + "learning_rate": 1.9373445005542553e-06, + "loss": 0.2834, + "step": 28809 + }, + { + "epoch": 10.64, + "learning_rate": 1.9373075501909104e-06, + "loss": 0.2409, + "step": 28810 + }, + { + "epoch": 10.64, + "learning_rate": 1.937270599827565e-06, + "loss": 0.2217, + "step": 28811 + }, + { + "epoch": 10.64, + "learning_rate": 1.93723364946422e-06, + "loss": 0.4067, + "step": 28812 + }, + { + "epoch": 10.64, + "learning_rate": 1.9371966991008747e-06, + "loss": 0.3002, + "step": 28813 + }, + { + "epoch": 10.64, + "learning_rate": 1.9371597487375293e-06, + "loss": 0.3425, + "step": 28814 + }, + { + "epoch": 10.64, + "learning_rate": 1.937122798374184e-06, + "loss": 0.2793, + "step": 28815 + }, + { + "epoch": 10.64, + "learning_rate": 1.9370858480108386e-06, + "loss": 0.4014, + "step": 28816 + }, + { + "epoch": 10.64, + "learning_rate": 1.9370488976474937e-06, + "loss": 0.3575, + "step": 28817 + }, + { + "epoch": 10.64, + "learning_rate": 1.9370119472841483e-06, + "loss": 0.3586, + "step": 28818 + }, + { + "epoch": 10.64, + "learning_rate": 1.9369749969208034e-06, + "loss": 0.4038, + "step": 28819 + }, + { + "epoch": 10.64, + "learning_rate": 1.9369380465574576e-06, + "loss": 0.2793, + "step": 28820 + }, + { + "epoch": 10.64, + "learning_rate": 1.9369010961941126e-06, + "loss": 0.4335, + "step": 28821 + }, + { + "epoch": 10.64, + "learning_rate": 1.9368641458307673e-06, + "loss": 0.2939, + "step": 28822 + }, + { + "epoch": 10.64, + "learning_rate": 1.9368271954674223e-06, + "loss": 0.33, + "step": 28823 + }, + { + "epoch": 10.64, + "learning_rate": 1.936790245104077e-06, + "loss": 0.518, + "step": 28824 + }, + { + "epoch": 10.64, + "learning_rate": 1.936753294740732e-06, + "loss": 0.5327, + "step": 28825 + }, + { + "epoch": 10.64, + "learning_rate": 1.9367163443773862e-06, + "loss": 0.2851, + "step": 28826 + }, + { + "epoch": 10.65, + "learning_rate": 1.936679394014041e-06, + "loss": 0.3915, + "step": 28827 + }, + { + "epoch": 10.65, + "learning_rate": 1.936642443650696e-06, + "loss": 0.217, + "step": 28828 + }, + { + "epoch": 10.65, + "learning_rate": 1.9366054932873506e-06, + "loss": 0.3121, + "step": 28829 + }, + { + "epoch": 10.65, + "learning_rate": 1.9365685429240056e-06, + "loss": 0.3672, + "step": 28830 + }, + { + "epoch": 10.65, + "learning_rate": 1.9365315925606603e-06, + "loss": 0.2441, + "step": 28831 + }, + { + "epoch": 10.65, + "learning_rate": 1.936494642197315e-06, + "loss": 0.4866, + "step": 28832 + }, + { + "epoch": 10.65, + "learning_rate": 1.9364576918339695e-06, + "loss": 0.2898, + "step": 28833 + }, + { + "epoch": 10.65, + "learning_rate": 1.9364207414706246e-06, + "loss": 0.3658, + "step": 28834 + }, + { + "epoch": 10.65, + "learning_rate": 1.9363837911072792e-06, + "loss": 0.3763, + "step": 28835 + }, + { + "epoch": 10.65, + "learning_rate": 1.936346840743934e-06, + "loss": 0.2964, + "step": 28836 + }, + { + "epoch": 10.65, + "learning_rate": 1.936309890380589e-06, + "loss": 0.3534, + "step": 28837 + }, + { + "epoch": 10.65, + "learning_rate": 1.9362729400172436e-06, + "loss": 0.2537, + "step": 28838 + }, + { + "epoch": 10.65, + "learning_rate": 1.936235989653898e-06, + "loss": 0.3816, + "step": 28839 + }, + { + "epoch": 10.65, + "learning_rate": 1.936199039290553e-06, + "loss": 0.4208, + "step": 28840 + }, + { + "epoch": 10.65, + "learning_rate": 1.936162088927208e-06, + "loss": 0.3157, + "step": 28841 + }, + { + "epoch": 10.65, + "learning_rate": 1.9361251385638625e-06, + "loss": 0.2832, + "step": 28842 + }, + { + "epoch": 10.65, + "learning_rate": 1.9360881882005176e-06, + "loss": 0.3018, + "step": 28843 + }, + { + "epoch": 10.65, + "learning_rate": 1.9360512378371722e-06, + "loss": 0.2989, + "step": 28844 + }, + { + "epoch": 10.65, + "learning_rate": 1.936014287473827e-06, + "loss": 0.4858, + "step": 28845 + }, + { + "epoch": 10.65, + "learning_rate": 1.9359773371104815e-06, + "loss": 0.3242, + "step": 28846 + }, + { + "epoch": 10.65, + "learning_rate": 1.935940386747136e-06, + "loss": 0.3243, + "step": 28847 + }, + { + "epoch": 10.65, + "learning_rate": 1.935903436383791e-06, + "loss": 0.4218, + "step": 28848 + }, + { + "epoch": 10.65, + "learning_rate": 1.935866486020446e-06, + "loss": 0.3955, + "step": 28849 + }, + { + "epoch": 10.65, + "learning_rate": 1.935829535657101e-06, + "loss": 0.2731, + "step": 28850 + }, + { + "epoch": 10.65, + "learning_rate": 1.9357925852937555e-06, + "loss": 0.2548, + "step": 28851 + }, + { + "epoch": 10.65, + "learning_rate": 1.93575563493041e-06, + "loss": 0.447, + "step": 28852 + }, + { + "epoch": 10.65, + "learning_rate": 1.935718684567065e-06, + "loss": 0.3985, + "step": 28853 + }, + { + "epoch": 10.66, + "learning_rate": 1.93568173420372e-06, + "loss": 0.3608, + "step": 28854 + }, + { + "epoch": 10.66, + "learning_rate": 1.9356447838403745e-06, + "loss": 0.4516, + "step": 28855 + }, + { + "epoch": 10.66, + "learning_rate": 1.935607833477029e-06, + "loss": 0.4318, + "step": 28856 + }, + { + "epoch": 10.66, + "learning_rate": 1.935570883113684e-06, + "loss": 0.3013, + "step": 28857 + }, + { + "epoch": 10.66, + "learning_rate": 1.935533932750339e-06, + "loss": 0.2963, + "step": 28858 + }, + { + "epoch": 10.66, + "learning_rate": 1.9354969823869935e-06, + "loss": 0.3874, + "step": 28859 + }, + { + "epoch": 10.66, + "learning_rate": 1.935460032023648e-06, + "loss": 0.3165, + "step": 28860 + }, + { + "epoch": 10.66, + "learning_rate": 1.935423081660303e-06, + "loss": 0.2946, + "step": 28861 + }, + { + "epoch": 10.66, + "learning_rate": 1.935386131296958e-06, + "loss": 0.3174, + "step": 28862 + }, + { + "epoch": 10.66, + "learning_rate": 1.935349180933613e-06, + "loss": 0.3858, + "step": 28863 + }, + { + "epoch": 10.66, + "learning_rate": 1.9353122305702675e-06, + "loss": 0.4319, + "step": 28864 + }, + { + "epoch": 10.66, + "learning_rate": 1.935275280206922e-06, + "loss": 0.1984, + "step": 28865 + }, + { + "epoch": 10.66, + "learning_rate": 1.9352383298435768e-06, + "loss": 0.3152, + "step": 28866 + }, + { + "epoch": 10.66, + "learning_rate": 1.9352013794802314e-06, + "loss": 0.2939, + "step": 28867 + }, + { + "epoch": 10.66, + "learning_rate": 1.9351644291168865e-06, + "loss": 0.375, + "step": 28868 + }, + { + "epoch": 10.66, + "learning_rate": 1.935127478753541e-06, + "loss": 0.3831, + "step": 28869 + }, + { + "epoch": 10.66, + "learning_rate": 1.935090528390196e-06, + "loss": 0.3562, + "step": 28870 + }, + { + "epoch": 10.66, + "learning_rate": 1.9350535780268504e-06, + "loss": 0.3692, + "step": 28871 + }, + { + "epoch": 10.66, + "learning_rate": 1.9350166276635054e-06, + "loss": 0.2894, + "step": 28872 + }, + { + "epoch": 10.66, + "learning_rate": 1.93497967730016e-06, + "loss": 0.3009, + "step": 28873 + }, + { + "epoch": 10.66, + "learning_rate": 1.934942726936815e-06, + "loss": 0.3352, + "step": 28874 + }, + { + "epoch": 10.66, + "learning_rate": 1.9349057765734698e-06, + "loss": 0.2748, + "step": 28875 + }, + { + "epoch": 10.66, + "learning_rate": 1.9348688262101244e-06, + "loss": 0.3732, + "step": 28876 + }, + { + "epoch": 10.66, + "learning_rate": 1.934831875846779e-06, + "loss": 0.2137, + "step": 28877 + }, + { + "epoch": 10.66, + "learning_rate": 1.9347949254834337e-06, + "loss": 0.2973, + "step": 28878 + }, + { + "epoch": 10.66, + "learning_rate": 1.9347579751200887e-06, + "loss": 0.3555, + "step": 28879 + }, + { + "epoch": 10.66, + "learning_rate": 1.9347210247567434e-06, + "loss": 0.235, + "step": 28880 + }, + { + "epoch": 10.67, + "learning_rate": 1.9346840743933984e-06, + "loss": 0.2789, + "step": 28881 + }, + { + "epoch": 10.67, + "learning_rate": 1.934647124030053e-06, + "loss": 0.4479, + "step": 28882 + }, + { + "epoch": 10.67, + "learning_rate": 1.9346101736667077e-06, + "loss": 0.3143, + "step": 28883 + }, + { + "epoch": 10.67, + "learning_rate": 1.9345732233033623e-06, + "loss": 0.3257, + "step": 28884 + }, + { + "epoch": 10.67, + "learning_rate": 1.9345362729400174e-06, + "loss": 0.3325, + "step": 28885 + }, + { + "epoch": 10.67, + "learning_rate": 1.934499322576672e-06, + "loss": 0.2674, + "step": 28886 + }, + { + "epoch": 10.67, + "learning_rate": 1.9344623722133267e-06, + "loss": 0.4319, + "step": 28887 + }, + { + "epoch": 10.67, + "learning_rate": 1.9344254218499817e-06, + "loss": 0.4199, + "step": 28888 + }, + { + "epoch": 10.67, + "learning_rate": 1.9343884714866364e-06, + "loss": 0.3534, + "step": 28889 + }, + { + "epoch": 10.67, + "learning_rate": 1.934351521123291e-06, + "loss": 0.3741, + "step": 28890 + }, + { + "epoch": 10.67, + "learning_rate": 1.9343145707599456e-06, + "loss": 0.3495, + "step": 28891 + }, + { + "epoch": 10.67, + "learning_rate": 1.9342776203966007e-06, + "loss": 0.4076, + "step": 28892 + }, + { + "epoch": 10.67, + "learning_rate": 1.9342406700332553e-06, + "loss": 0.4278, + "step": 28893 + }, + { + "epoch": 10.67, + "learning_rate": 1.9342037196699104e-06, + "loss": 0.3571, + "step": 28894 + }, + { + "epoch": 10.67, + "learning_rate": 1.934166769306565e-06, + "loss": 0.3972, + "step": 28895 + }, + { + "epoch": 10.67, + "learning_rate": 1.9341298189432197e-06, + "loss": 0.2323, + "step": 28896 + }, + { + "epoch": 10.67, + "learning_rate": 1.9340928685798743e-06, + "loss": 0.3022, + "step": 28897 + }, + { + "epoch": 10.67, + "learning_rate": 1.934055918216529e-06, + "loss": 0.3172, + "step": 28898 + }, + { + "epoch": 10.67, + "learning_rate": 1.934018967853184e-06, + "loss": 0.3576, + "step": 28899 + }, + { + "epoch": 10.67, + "learning_rate": 1.9339820174898386e-06, + "loss": 0.2244, + "step": 28900 + }, + { + "epoch": 10.67, + "learning_rate": 1.9339450671264937e-06, + "loss": 0.4092, + "step": 28901 + }, + { + "epoch": 10.67, + "learning_rate": 1.9339081167631483e-06, + "loss": 0.288, + "step": 28902 + }, + { + "epoch": 10.67, + "learning_rate": 1.933871166399803e-06, + "loss": 0.4656, + "step": 28903 + }, + { + "epoch": 10.67, + "learning_rate": 1.9338342160364576e-06, + "loss": 0.3914, + "step": 28904 + }, + { + "epoch": 10.67, + "learning_rate": 1.9337972656731127e-06, + "loss": 0.3013, + "step": 28905 + }, + { + "epoch": 10.67, + "learning_rate": 1.9337603153097673e-06, + "loss": 0.3396, + "step": 28906 + }, + { + "epoch": 10.67, + "learning_rate": 1.933723364946422e-06, + "loss": 0.4136, + "step": 28907 + }, + { + "epoch": 10.68, + "learning_rate": 1.933686414583077e-06, + "loss": 0.3784, + "step": 28908 + }, + { + "epoch": 10.68, + "learning_rate": 1.9336494642197316e-06, + "loss": 0.4307, + "step": 28909 + }, + { + "epoch": 10.68, + "learning_rate": 1.9336125138563863e-06, + "loss": 0.2961, + "step": 28910 + }, + { + "epoch": 10.68, + "learning_rate": 1.933575563493041e-06, + "loss": 0.6096, + "step": 28911 + }, + { + "epoch": 10.68, + "learning_rate": 1.933538613129696e-06, + "loss": 0.328, + "step": 28912 + }, + { + "epoch": 10.68, + "learning_rate": 1.9335016627663506e-06, + "loss": 0.2528, + "step": 28913 + }, + { + "epoch": 10.68, + "learning_rate": 1.9334647124030057e-06, + "loss": 0.3374, + "step": 28914 + }, + { + "epoch": 10.68, + "learning_rate": 1.93342776203966e-06, + "loss": 0.4185, + "step": 28915 + }, + { + "epoch": 10.68, + "learning_rate": 1.9333908116763145e-06, + "loss": 0.3003, + "step": 28916 + }, + { + "epoch": 10.68, + "learning_rate": 1.9333538613129696e-06, + "loss": 0.3462, + "step": 28917 + }, + { + "epoch": 10.68, + "learning_rate": 1.9333169109496242e-06, + "loss": 0.3857, + "step": 28918 + }, + { + "epoch": 10.68, + "learning_rate": 1.9332799605862793e-06, + "loss": 0.2797, + "step": 28919 + }, + { + "epoch": 10.68, + "learning_rate": 1.933243010222934e-06, + "loss": 0.3047, + "step": 28920 + }, + { + "epoch": 10.68, + "learning_rate": 1.9332060598595886e-06, + "loss": 0.2738, + "step": 28921 + }, + { + "epoch": 10.68, + "learning_rate": 1.933169109496243e-06, + "loss": 0.3413, + "step": 28922 + }, + { + "epoch": 10.68, + "learning_rate": 1.9331321591328982e-06, + "loss": 0.4541, + "step": 28923 + }, + { + "epoch": 10.68, + "learning_rate": 1.933095208769553e-06, + "loss": 0.3991, + "step": 28924 + }, + { + "epoch": 10.68, + "learning_rate": 1.933058258406208e-06, + "loss": 0.4006, + "step": 28925 + }, + { + "epoch": 10.68, + "learning_rate": 1.9330213080428626e-06, + "loss": 0.3083, + "step": 28926 + }, + { + "epoch": 10.68, + "learning_rate": 1.9329843576795172e-06, + "loss": 0.3918, + "step": 28927 + }, + { + "epoch": 10.68, + "learning_rate": 1.932947407316172e-06, + "loss": 0.2899, + "step": 28928 + }, + { + "epoch": 10.68, + "learning_rate": 1.9329104569528265e-06, + "loss": 0.3681, + "step": 28929 + }, + { + "epoch": 10.68, + "learning_rate": 1.9328735065894816e-06, + "loss": 0.3624, + "step": 28930 + }, + { + "epoch": 10.68, + "learning_rate": 1.932836556226136e-06, + "loss": 0.3248, + "step": 28931 + }, + { + "epoch": 10.68, + "learning_rate": 1.9327996058627912e-06, + "loss": 0.3015, + "step": 28932 + }, + { + "epoch": 10.68, + "learning_rate": 1.932762655499446e-06, + "loss": 0.3359, + "step": 28933 + }, + { + "epoch": 10.68, + "learning_rate": 1.9327257051361005e-06, + "loss": 0.2531, + "step": 28934 + }, + { + "epoch": 10.69, + "learning_rate": 1.932688754772755e-06, + "loss": 0.3165, + "step": 28935 + }, + { + "epoch": 10.69, + "learning_rate": 1.93265180440941e-06, + "loss": 0.3293, + "step": 28936 + }, + { + "epoch": 10.69, + "learning_rate": 1.932614854046065e-06, + "loss": 0.3305, + "step": 28937 + }, + { + "epoch": 10.69, + "learning_rate": 1.9325779036827195e-06, + "loss": 0.375, + "step": 28938 + }, + { + "epoch": 10.69, + "learning_rate": 1.9325409533193745e-06, + "loss": 0.3538, + "step": 28939 + }, + { + "epoch": 10.69, + "learning_rate": 1.932504002956029e-06, + "loss": 0.2473, + "step": 28940 + }, + { + "epoch": 10.69, + "learning_rate": 1.932467052592684e-06, + "loss": 0.3593, + "step": 28941 + }, + { + "epoch": 10.69, + "learning_rate": 1.9324301022293385e-06, + "loss": 0.4694, + "step": 28942 + }, + { + "epoch": 10.69, + "learning_rate": 1.9323931518659935e-06, + "loss": 0.3846, + "step": 28943 + }, + { + "epoch": 10.69, + "learning_rate": 1.932356201502648e-06, + "loss": 0.213, + "step": 28944 + }, + { + "epoch": 10.69, + "learning_rate": 1.9323192511393032e-06, + "loss": 0.4067, + "step": 28945 + }, + { + "epoch": 10.69, + "learning_rate": 1.932282300775958e-06, + "loss": 0.487, + "step": 28946 + }, + { + "epoch": 10.69, + "learning_rate": 1.9322453504126125e-06, + "loss": 0.3369, + "step": 28947 + }, + { + "epoch": 10.69, + "learning_rate": 1.932208400049267e-06, + "loss": 0.451, + "step": 28948 + }, + { + "epoch": 10.69, + "learning_rate": 1.9321714496859218e-06, + "loss": 0.356, + "step": 28949 + }, + { + "epoch": 10.69, + "learning_rate": 1.932134499322577e-06, + "loss": 0.3386, + "step": 28950 + }, + { + "epoch": 10.69, + "learning_rate": 1.9320975489592315e-06, + "loss": 0.3435, + "step": 28951 + }, + { + "epoch": 10.69, + "learning_rate": 1.9320605985958865e-06, + "loss": 0.3708, + "step": 28952 + }, + { + "epoch": 10.69, + "learning_rate": 1.932023648232541e-06, + "loss": 0.3637, + "step": 28953 + }, + { + "epoch": 10.69, + "learning_rate": 1.931986697869196e-06, + "loss": 0.2282, + "step": 28954 + }, + { + "epoch": 10.69, + "learning_rate": 1.9319497475058504e-06, + "loss": 0.317, + "step": 28955 + }, + { + "epoch": 10.69, + "learning_rate": 1.931912797142505e-06, + "loss": 0.3955, + "step": 28956 + }, + { + "epoch": 10.69, + "learning_rate": 1.93187584677916e-06, + "loss": 0.4387, + "step": 28957 + }, + { + "epoch": 10.69, + "learning_rate": 1.9318388964158148e-06, + "loss": 0.4131, + "step": 28958 + }, + { + "epoch": 10.69, + "learning_rate": 1.93180194605247e-06, + "loss": 0.2745, + "step": 28959 + }, + { + "epoch": 10.69, + "learning_rate": 1.931764995689124e-06, + "loss": 0.3144, + "step": 28960 + }, + { + "epoch": 10.69, + "learning_rate": 1.931728045325779e-06, + "loss": 0.2718, + "step": 28961 + }, + { + "epoch": 10.69, + "learning_rate": 1.9316910949624337e-06, + "loss": 0.4274, + "step": 28962 + }, + { + "epoch": 10.7, + "learning_rate": 1.9316541445990888e-06, + "loss": 0.3184, + "step": 28963 + }, + { + "epoch": 10.7, + "learning_rate": 1.9316171942357434e-06, + "loss": 0.4002, + "step": 28964 + }, + { + "epoch": 10.7, + "learning_rate": 1.9315802438723985e-06, + "loss": 0.3298, + "step": 28965 + }, + { + "epoch": 10.7, + "learning_rate": 1.9315432935090527e-06, + "loss": 0.3526, + "step": 28966 + }, + { + "epoch": 10.7, + "learning_rate": 1.9315063431457073e-06, + "loss": 0.21, + "step": 28967 + }, + { + "epoch": 10.7, + "learning_rate": 1.9314693927823624e-06, + "loss": 0.3857, + "step": 28968 + }, + { + "epoch": 10.7, + "learning_rate": 1.931432442419017e-06, + "loss": 0.396, + "step": 28969 + }, + { + "epoch": 10.7, + "learning_rate": 1.931395492055672e-06, + "loss": 0.2936, + "step": 28970 + }, + { + "epoch": 10.7, + "learning_rate": 1.9313585416923267e-06, + "loss": 0.2293, + "step": 28971 + }, + { + "epoch": 10.7, + "learning_rate": 1.9313215913289814e-06, + "loss": 0.4349, + "step": 28972 + }, + { + "epoch": 10.7, + "learning_rate": 1.931284640965636e-06, + "loss": 0.2007, + "step": 28973 + }, + { + "epoch": 10.7, + "learning_rate": 1.931247690602291e-06, + "loss": 0.3702, + "step": 28974 + }, + { + "epoch": 10.7, + "learning_rate": 1.9312107402389457e-06, + "loss": 0.4307, + "step": 28975 + }, + { + "epoch": 10.7, + "learning_rate": 1.9311737898756003e-06, + "loss": 0.3237, + "step": 28976 + }, + { + "epoch": 10.7, + "learning_rate": 1.9311368395122554e-06, + "loss": 0.3098, + "step": 28977 + }, + { + "epoch": 10.7, + "learning_rate": 1.93109988914891e-06, + "loss": 0.4245, + "step": 28978 + }, + { + "epoch": 10.7, + "learning_rate": 1.9310629387855647e-06, + "loss": 0.3964, + "step": 28979 + }, + { + "epoch": 10.7, + "learning_rate": 1.9310259884222193e-06, + "loss": 0.3195, + "step": 28980 + }, + { + "epoch": 10.7, + "learning_rate": 1.9309890380588744e-06, + "loss": 0.2764, + "step": 28981 + }, + { + "epoch": 10.7, + "learning_rate": 1.930952087695529e-06, + "loss": 0.2738, + "step": 28982 + }, + { + "epoch": 10.7, + "learning_rate": 1.930915137332184e-06, + "loss": 0.3418, + "step": 28983 + }, + { + "epoch": 10.7, + "learning_rate": 1.9308781869688387e-06, + "loss": 0.3866, + "step": 28984 + }, + { + "epoch": 10.7, + "learning_rate": 1.9308412366054933e-06, + "loss": 0.3319, + "step": 28985 + }, + { + "epoch": 10.7, + "learning_rate": 1.930804286242148e-06, + "loss": 0.4615, + "step": 28986 + }, + { + "epoch": 10.7, + "learning_rate": 1.9307673358788026e-06, + "loss": 0.3335, + "step": 28987 + }, + { + "epoch": 10.7, + "learning_rate": 1.9307303855154577e-06, + "loss": 0.2271, + "step": 28988 + }, + { + "epoch": 10.7, + "learning_rate": 1.9306934351521123e-06, + "loss": 0.245, + "step": 28989 + }, + { + "epoch": 10.71, + "learning_rate": 1.9306564847887674e-06, + "loss": 0.2864, + "step": 28990 + }, + { + "epoch": 10.71, + "learning_rate": 1.930619534425422e-06, + "loss": 0.2983, + "step": 28991 + }, + { + "epoch": 10.71, + "learning_rate": 1.9305825840620766e-06, + "loss": 0.3403, + "step": 28992 + }, + { + "epoch": 10.71, + "learning_rate": 1.9305456336987313e-06, + "loss": 0.496, + "step": 28993 + }, + { + "epoch": 10.71, + "learning_rate": 1.9305086833353863e-06, + "loss": 0.247, + "step": 28994 + }, + { + "epoch": 10.71, + "learning_rate": 1.930471732972041e-06, + "loss": 0.388, + "step": 28995 + }, + { + "epoch": 10.71, + "learning_rate": 1.9304347826086956e-06, + "loss": 0.3401, + "step": 28996 + }, + { + "epoch": 10.71, + "learning_rate": 1.9303978322453507e-06, + "loss": 0.3744, + "step": 28997 + }, + { + "epoch": 10.71, + "learning_rate": 1.9303608818820053e-06, + "loss": 0.3791, + "step": 28998 + }, + { + "epoch": 10.71, + "learning_rate": 1.93032393151866e-06, + "loss": 0.4593, + "step": 28999 + }, + { + "epoch": 10.71, + "learning_rate": 1.9302869811553146e-06, + "loss": 0.3201, + "step": 29000 + }, + { + "epoch": 10.71, + "learning_rate": 1.9302500307919696e-06, + "loss": 0.3855, + "step": 29001 + }, + { + "epoch": 10.71, + "learning_rate": 1.9302130804286243e-06, + "loss": 0.3864, + "step": 29002 + }, + { + "epoch": 10.71, + "learning_rate": 1.9301761300652793e-06, + "loss": 0.3803, + "step": 29003 + }, + { + "epoch": 10.71, + "learning_rate": 1.930139179701934e-06, + "loss": 0.4234, + "step": 29004 + }, + { + "epoch": 10.71, + "learning_rate": 1.9301022293385886e-06, + "loss": 0.2983, + "step": 29005 + }, + { + "epoch": 10.71, + "learning_rate": 1.9300652789752432e-06, + "loss": 0.3179, + "step": 29006 + }, + { + "epoch": 10.71, + "learning_rate": 1.930028328611898e-06, + "loss": 0.3669, + "step": 29007 + }, + { + "epoch": 10.71, + "learning_rate": 1.929991378248553e-06, + "loss": 0.3576, + "step": 29008 + }, + { + "epoch": 10.71, + "learning_rate": 1.9299544278852076e-06, + "loss": 0.2671, + "step": 29009 + }, + { + "epoch": 10.71, + "learning_rate": 1.9299174775218626e-06, + "loss": 0.3988, + "step": 29010 + }, + { + "epoch": 10.71, + "learning_rate": 1.929880527158517e-06, + "loss": 0.2371, + "step": 29011 + }, + { + "epoch": 10.71, + "learning_rate": 1.929843576795172e-06, + "loss": 0.2883, + "step": 29012 + }, + { + "epoch": 10.71, + "learning_rate": 1.9298066264318265e-06, + "loss": 0.5476, + "step": 29013 + }, + { + "epoch": 10.71, + "learning_rate": 1.9297696760684816e-06, + "loss": 0.3188, + "step": 29014 + }, + { + "epoch": 10.71, + "learning_rate": 1.9297327257051362e-06, + "loss": 0.3703, + "step": 29015 + }, + { + "epoch": 10.71, + "learning_rate": 1.929695775341791e-06, + "loss": 0.4047, + "step": 29016 + }, + { + "epoch": 10.72, + "learning_rate": 1.9296588249784455e-06, + "loss": 0.2675, + "step": 29017 + }, + { + "epoch": 10.72, + "learning_rate": 1.9296218746151e-06, + "loss": 0.364, + "step": 29018 + }, + { + "epoch": 10.72, + "learning_rate": 1.929584924251755e-06, + "loss": 0.2441, + "step": 29019 + }, + { + "epoch": 10.72, + "learning_rate": 1.92954797388841e-06, + "loss": 0.3933, + "step": 29020 + }, + { + "epoch": 10.72, + "learning_rate": 1.929511023525065e-06, + "loss": 0.3189, + "step": 29021 + }, + { + "epoch": 10.72, + "learning_rate": 1.9294740731617195e-06, + "loss": 0.4041, + "step": 29022 + }, + { + "epoch": 10.72, + "learning_rate": 1.929437122798374e-06, + "loss": 0.4117, + "step": 29023 + }, + { + "epoch": 10.72, + "learning_rate": 1.929400172435029e-06, + "loss": 0.3261, + "step": 29024 + }, + { + "epoch": 10.72, + "learning_rate": 1.929363222071684e-06, + "loss": 0.281, + "step": 29025 + }, + { + "epoch": 10.72, + "learning_rate": 1.9293262717083385e-06, + "loss": 0.3793, + "step": 29026 + }, + { + "epoch": 10.72, + "learning_rate": 1.929289321344993e-06, + "loss": 0.3026, + "step": 29027 + }, + { + "epoch": 10.72, + "learning_rate": 1.929252370981648e-06, + "loss": 0.3837, + "step": 29028 + }, + { + "epoch": 10.72, + "learning_rate": 1.929215420618303e-06, + "loss": 0.4243, + "step": 29029 + }, + { + "epoch": 10.72, + "learning_rate": 1.9291784702549575e-06, + "loss": 0.364, + "step": 29030 + }, + { + "epoch": 10.72, + "learning_rate": 1.929141519891612e-06, + "loss": 0.382, + "step": 29031 + }, + { + "epoch": 10.72, + "learning_rate": 1.929104569528267e-06, + "loss": 0.3239, + "step": 29032 + }, + { + "epoch": 10.72, + "learning_rate": 1.929067619164922e-06, + "loss": 0.2567, + "step": 29033 + }, + { + "epoch": 10.72, + "learning_rate": 1.929030668801577e-06, + "loss": 0.2862, + "step": 29034 + }, + { + "epoch": 10.72, + "learning_rate": 1.9289937184382315e-06, + "loss": 0.3476, + "step": 29035 + }, + { + "epoch": 10.72, + "learning_rate": 1.928956768074886e-06, + "loss": 0.2855, + "step": 29036 + }, + { + "epoch": 10.72, + "learning_rate": 1.9289198177115408e-06, + "loss": 0.2353, + "step": 29037 + }, + { + "epoch": 10.72, + "learning_rate": 1.9288828673481954e-06, + "loss": 0.3609, + "step": 29038 + }, + { + "epoch": 10.72, + "learning_rate": 1.9288459169848505e-06, + "loss": 0.3606, + "step": 29039 + }, + { + "epoch": 10.72, + "learning_rate": 1.928808966621505e-06, + "loss": 0.3343, + "step": 29040 + }, + { + "epoch": 10.72, + "learning_rate": 1.92877201625816e-06, + "loss": 0.3888, + "step": 29041 + }, + { + "epoch": 10.72, + "learning_rate": 1.928735065894815e-06, + "loss": 0.3374, + "step": 29042 + }, + { + "epoch": 10.72, + "learning_rate": 1.9286981155314694e-06, + "loss": 0.317, + "step": 29043 + }, + { + "epoch": 10.73, + "learning_rate": 1.928661165168124e-06, + "loss": 0.282, + "step": 29044 + }, + { + "epoch": 10.73, + "learning_rate": 1.928624214804779e-06, + "loss": 0.3001, + "step": 29045 + }, + { + "epoch": 10.73, + "learning_rate": 1.9285872644414338e-06, + "loss": 0.2562, + "step": 29046 + }, + { + "epoch": 10.73, + "learning_rate": 1.9285503140780884e-06, + "loss": 0.2541, + "step": 29047 + }, + { + "epoch": 10.73, + "learning_rate": 1.9285133637147435e-06, + "loss": 0.2839, + "step": 29048 + }, + { + "epoch": 10.73, + "learning_rate": 1.928476413351398e-06, + "loss": 0.2787, + "step": 29049 + }, + { + "epoch": 10.73, + "learning_rate": 1.9284394629880527e-06, + "loss": 0.3039, + "step": 29050 + }, + { + "epoch": 10.73, + "learning_rate": 1.9284025126247074e-06, + "loss": 0.3011, + "step": 29051 + }, + { + "epoch": 10.73, + "learning_rate": 1.9283655622613624e-06, + "loss": 0.4929, + "step": 29052 + }, + { + "epoch": 10.73, + "learning_rate": 1.928328611898017e-06, + "loss": 0.3768, + "step": 29053 + }, + { + "epoch": 10.73, + "learning_rate": 1.928291661534672e-06, + "loss": 0.2482, + "step": 29054 + }, + { + "epoch": 10.73, + "learning_rate": 1.9282547111713268e-06, + "loss": 0.4129, + "step": 29055 + }, + { + "epoch": 10.73, + "learning_rate": 1.928217760807981e-06, + "loss": 0.3967, + "step": 29056 + }, + { + "epoch": 10.73, + "learning_rate": 1.928180810444636e-06, + "loss": 0.3614, + "step": 29057 + }, + { + "epoch": 10.73, + "learning_rate": 1.9281438600812907e-06, + "loss": 0.3621, + "step": 29058 + }, + { + "epoch": 10.73, + "learning_rate": 1.9281069097179457e-06, + "loss": 0.3433, + "step": 29059 + }, + { + "epoch": 10.73, + "learning_rate": 1.9280699593546004e-06, + "loss": 0.3404, + "step": 29060 + }, + { + "epoch": 10.73, + "learning_rate": 1.9280330089912554e-06, + "loss": 0.4527, + "step": 29061 + }, + { + "epoch": 10.73, + "learning_rate": 1.9279960586279097e-06, + "loss": 0.3023, + "step": 29062 + }, + { + "epoch": 10.73, + "learning_rate": 1.9279591082645647e-06, + "loss": 0.325, + "step": 29063 + }, + { + "epoch": 10.73, + "learning_rate": 1.9279221579012193e-06, + "loss": 0.4163, + "step": 29064 + }, + { + "epoch": 10.73, + "learning_rate": 1.927885207537874e-06, + "loss": 0.3951, + "step": 29065 + }, + { + "epoch": 10.73, + "learning_rate": 1.927848257174529e-06, + "loss": 0.3589, + "step": 29066 + }, + { + "epoch": 10.73, + "learning_rate": 1.9278113068111837e-06, + "loss": 0.419, + "step": 29067 + }, + { + "epoch": 10.73, + "learning_rate": 1.9277743564478383e-06, + "loss": 0.1648, + "step": 29068 + }, + { + "epoch": 10.73, + "learning_rate": 1.927737406084493e-06, + "loss": 0.3169, + "step": 29069 + }, + { + "epoch": 10.73, + "learning_rate": 1.927700455721148e-06, + "loss": 0.3486, + "step": 29070 + }, + { + "epoch": 10.74, + "learning_rate": 1.9276635053578027e-06, + "loss": 0.3184, + "step": 29071 + }, + { + "epoch": 10.74, + "learning_rate": 1.9276265549944577e-06, + "loss": 0.3221, + "step": 29072 + }, + { + "epoch": 10.74, + "learning_rate": 1.9275896046311123e-06, + "loss": 0.4334, + "step": 29073 + }, + { + "epoch": 10.74, + "learning_rate": 1.927552654267767e-06, + "loss": 0.3641, + "step": 29074 + }, + { + "epoch": 10.74, + "learning_rate": 1.9275157039044216e-06, + "loss": 0.3612, + "step": 29075 + }, + { + "epoch": 10.74, + "learning_rate": 1.9274787535410763e-06, + "loss": 0.3443, + "step": 29076 + }, + { + "epoch": 10.74, + "learning_rate": 1.9274418031777313e-06, + "loss": 0.3991, + "step": 29077 + }, + { + "epoch": 10.74, + "learning_rate": 1.927404852814386e-06, + "loss": 0.3133, + "step": 29078 + }, + { + "epoch": 10.74, + "learning_rate": 1.927367902451041e-06, + "loss": 0.3702, + "step": 29079 + }, + { + "epoch": 10.74, + "learning_rate": 1.9273309520876957e-06, + "loss": 0.4464, + "step": 29080 + }, + { + "epoch": 10.74, + "learning_rate": 1.9272940017243503e-06, + "loss": 0.3171, + "step": 29081 + }, + { + "epoch": 10.74, + "learning_rate": 1.927257051361005e-06, + "loss": 0.2605, + "step": 29082 + }, + { + "epoch": 10.74, + "learning_rate": 1.92722010099766e-06, + "loss": 0.3237, + "step": 29083 + }, + { + "epoch": 10.74, + "learning_rate": 1.9271831506343146e-06, + "loss": 0.2198, + "step": 29084 + }, + { + "epoch": 10.74, + "learning_rate": 1.9271462002709693e-06, + "loss": 0.3277, + "step": 29085 + }, + { + "epoch": 10.74, + "learning_rate": 1.9271092499076243e-06, + "loss": 0.3902, + "step": 29086 + }, + { + "epoch": 10.74, + "learning_rate": 1.927072299544279e-06, + "loss": 0.4498, + "step": 29087 + }, + { + "epoch": 10.74, + "learning_rate": 1.9270353491809336e-06, + "loss": 0.3633, + "step": 29088 + }, + { + "epoch": 10.74, + "learning_rate": 1.9269983988175882e-06, + "loss": 0.2973, + "step": 29089 + }, + { + "epoch": 10.74, + "learning_rate": 1.9269614484542433e-06, + "loss": 0.3299, + "step": 29090 + }, + { + "epoch": 10.74, + "learning_rate": 1.926924498090898e-06, + "loss": 0.3087, + "step": 29091 + }, + { + "epoch": 10.74, + "learning_rate": 1.926887547727553e-06, + "loss": 0.4014, + "step": 29092 + }, + { + "epoch": 10.74, + "learning_rate": 1.9268505973642076e-06, + "loss": 0.2905, + "step": 29093 + }, + { + "epoch": 10.74, + "learning_rate": 1.9268136470008623e-06, + "loss": 0.4725, + "step": 29094 + }, + { + "epoch": 10.74, + "learning_rate": 1.926776696637517e-06, + "loss": 0.3544, + "step": 29095 + }, + { + "epoch": 10.74, + "learning_rate": 1.9267397462741715e-06, + "loss": 0.3575, + "step": 29096 + }, + { + "epoch": 10.74, + "learning_rate": 1.9267027959108266e-06, + "loss": 0.3636, + "step": 29097 + }, + { + "epoch": 10.75, + "learning_rate": 1.9266658455474812e-06, + "loss": 0.2832, + "step": 29098 + }, + { + "epoch": 10.75, + "learning_rate": 1.9266288951841363e-06, + "loss": 0.3993, + "step": 29099 + }, + { + "epoch": 10.75, + "learning_rate": 1.9265919448207905e-06, + "loss": 0.3086, + "step": 29100 + }, + { + "epoch": 10.75, + "learning_rate": 1.9265549944574456e-06, + "loss": 0.3192, + "step": 29101 + }, + { + "epoch": 10.75, + "learning_rate": 1.9265180440941e-06, + "loss": 0.2911, + "step": 29102 + }, + { + "epoch": 10.75, + "learning_rate": 1.9264810937307553e-06, + "loss": 0.3099, + "step": 29103 + }, + { + "epoch": 10.75, + "learning_rate": 1.92644414336741e-06, + "loss": 0.278, + "step": 29104 + }, + { + "epoch": 10.75, + "learning_rate": 1.9264071930040645e-06, + "loss": 0.2669, + "step": 29105 + }, + { + "epoch": 10.75, + "learning_rate": 1.926370242640719e-06, + "loss": 0.4169, + "step": 29106 + }, + { + "epoch": 10.75, + "learning_rate": 1.926333292277374e-06, + "loss": 0.2724, + "step": 29107 + }, + { + "epoch": 10.75, + "learning_rate": 1.926296341914029e-06, + "loss": 0.3707, + "step": 29108 + }, + { + "epoch": 10.75, + "learning_rate": 1.9262593915506835e-06, + "loss": 0.3384, + "step": 29109 + }, + { + "epoch": 10.75, + "learning_rate": 1.9262224411873386e-06, + "loss": 0.3129, + "step": 29110 + }, + { + "epoch": 10.75, + "learning_rate": 1.926185490823993e-06, + "loss": 0.4231, + "step": 29111 + }, + { + "epoch": 10.75, + "learning_rate": 1.926148540460648e-06, + "loss": 0.281, + "step": 29112 + }, + { + "epoch": 10.75, + "learning_rate": 1.9261115900973025e-06, + "loss": 0.3435, + "step": 29113 + }, + { + "epoch": 10.75, + "learning_rate": 1.9260746397339575e-06, + "loss": 0.4696, + "step": 29114 + }, + { + "epoch": 10.75, + "learning_rate": 1.926037689370612e-06, + "loss": 0.3076, + "step": 29115 + }, + { + "epoch": 10.75, + "learning_rate": 1.926000739007267e-06, + "loss": 0.336, + "step": 29116 + }, + { + "epoch": 10.75, + "learning_rate": 1.925963788643922e-06, + "loss": 0.3356, + "step": 29117 + }, + { + "epoch": 10.75, + "learning_rate": 1.9259268382805765e-06, + "loss": 0.3572, + "step": 29118 + }, + { + "epoch": 10.75, + "learning_rate": 1.925889887917231e-06, + "loss": 0.446, + "step": 29119 + }, + { + "epoch": 10.75, + "learning_rate": 1.9258529375538858e-06, + "loss": 0.2777, + "step": 29120 + }, + { + "epoch": 10.75, + "learning_rate": 1.925815987190541e-06, + "loss": 0.3833, + "step": 29121 + }, + { + "epoch": 10.75, + "learning_rate": 1.9257790368271955e-06, + "loss": 0.3702, + "step": 29122 + }, + { + "epoch": 10.75, + "learning_rate": 1.9257420864638505e-06, + "loss": 0.2292, + "step": 29123 + }, + { + "epoch": 10.75, + "learning_rate": 1.925705136100505e-06, + "loss": 0.5009, + "step": 29124 + }, + { + "epoch": 10.76, + "learning_rate": 1.92566818573716e-06, + "loss": 0.5436, + "step": 29125 + }, + { + "epoch": 10.76, + "learning_rate": 1.9256312353738144e-06, + "loss": 0.4155, + "step": 29126 + }, + { + "epoch": 10.76, + "learning_rate": 1.925594285010469e-06, + "loss": 0.3629, + "step": 29127 + }, + { + "epoch": 10.76, + "learning_rate": 1.925557334647124e-06, + "loss": 0.2913, + "step": 29128 + }, + { + "epoch": 10.76, + "learning_rate": 1.9255203842837788e-06, + "loss": 0.303, + "step": 29129 + }, + { + "epoch": 10.76, + "learning_rate": 1.925483433920434e-06, + "loss": 0.3888, + "step": 29130 + }, + { + "epoch": 10.76, + "learning_rate": 1.9254464835570885e-06, + "loss": 0.3427, + "step": 29131 + }, + { + "epoch": 10.76, + "learning_rate": 1.925409533193743e-06, + "loss": 0.4857, + "step": 29132 + }, + { + "epoch": 10.76, + "learning_rate": 1.9253725828303977e-06, + "loss": 0.435, + "step": 29133 + }, + { + "epoch": 10.76, + "learning_rate": 1.925335632467053e-06, + "loss": 0.3627, + "step": 29134 + }, + { + "epoch": 10.76, + "learning_rate": 1.9252986821037074e-06, + "loss": 0.2708, + "step": 29135 + }, + { + "epoch": 10.76, + "learning_rate": 1.925261731740362e-06, + "loss": 0.3904, + "step": 29136 + }, + { + "epoch": 10.76, + "learning_rate": 1.925224781377017e-06, + "loss": 0.4017, + "step": 29137 + }, + { + "epoch": 10.76, + "learning_rate": 1.9251878310136718e-06, + "loss": 0.3713, + "step": 29138 + }, + { + "epoch": 10.76, + "learning_rate": 1.9251508806503264e-06, + "loss": 0.3635, + "step": 29139 + }, + { + "epoch": 10.76, + "learning_rate": 1.925113930286981e-06, + "loss": 0.2555, + "step": 29140 + }, + { + "epoch": 10.76, + "learning_rate": 1.925076979923636e-06, + "loss": 0.2697, + "step": 29141 + }, + { + "epoch": 10.76, + "learning_rate": 1.9250400295602907e-06, + "loss": 0.3475, + "step": 29142 + }, + { + "epoch": 10.76, + "learning_rate": 1.925003079196946e-06, + "loss": 0.4187, + "step": 29143 + }, + { + "epoch": 10.76, + "learning_rate": 1.9249661288336004e-06, + "loss": 0.2828, + "step": 29144 + }, + { + "epoch": 10.76, + "learning_rate": 1.9249291784702546e-06, + "loss": 0.381, + "step": 29145 + }, + { + "epoch": 10.76, + "learning_rate": 1.9248922281069097e-06, + "loss": 0.2783, + "step": 29146 + }, + { + "epoch": 10.76, + "learning_rate": 1.9248552777435643e-06, + "loss": 0.4728, + "step": 29147 + }, + { + "epoch": 10.76, + "learning_rate": 1.9248183273802194e-06, + "loss": 0.3597, + "step": 29148 + }, + { + "epoch": 10.76, + "learning_rate": 1.924781377016874e-06, + "loss": 0.2738, + "step": 29149 + }, + { + "epoch": 10.76, + "learning_rate": 1.924744426653529e-06, + "loss": 0.3339, + "step": 29150 + }, + { + "epoch": 10.76, + "learning_rate": 1.9247074762901833e-06, + "loss": 0.3064, + "step": 29151 + }, + { + "epoch": 10.77, + "learning_rate": 1.9246705259268384e-06, + "loss": 0.412, + "step": 29152 + }, + { + "epoch": 10.77, + "learning_rate": 1.924633575563493e-06, + "loss": 0.2812, + "step": 29153 + }, + { + "epoch": 10.77, + "learning_rate": 1.924596625200148e-06, + "loss": 0.3282, + "step": 29154 + }, + { + "epoch": 10.77, + "learning_rate": 1.9245596748368027e-06, + "loss": 0.248, + "step": 29155 + }, + { + "epoch": 10.77, + "learning_rate": 1.9245227244734573e-06, + "loss": 0.2901, + "step": 29156 + }, + { + "epoch": 10.77, + "learning_rate": 1.924485774110112e-06, + "loss": 0.3262, + "step": 29157 + }, + { + "epoch": 10.77, + "learning_rate": 1.9244488237467666e-06, + "loss": 0.3245, + "step": 29158 + }, + { + "epoch": 10.77, + "learning_rate": 1.9244118733834217e-06, + "loss": 0.2996, + "step": 29159 + }, + { + "epoch": 10.77, + "learning_rate": 1.9243749230200763e-06, + "loss": 0.2857, + "step": 29160 + }, + { + "epoch": 10.77, + "learning_rate": 1.9243379726567314e-06, + "loss": 0.2702, + "step": 29161 + }, + { + "epoch": 10.77, + "learning_rate": 1.924301022293386e-06, + "loss": 0.4441, + "step": 29162 + }, + { + "epoch": 10.77, + "learning_rate": 1.9242640719300406e-06, + "loss": 0.4542, + "step": 29163 + }, + { + "epoch": 10.77, + "learning_rate": 1.9242271215666953e-06, + "loss": 0.3039, + "step": 29164 + }, + { + "epoch": 10.77, + "learning_rate": 1.92419017120335e-06, + "loss": 0.329, + "step": 29165 + }, + { + "epoch": 10.77, + "learning_rate": 1.924153220840005e-06, + "loss": 0.3471, + "step": 29166 + }, + { + "epoch": 10.77, + "learning_rate": 1.9241162704766596e-06, + "loss": 0.4761, + "step": 29167 + }, + { + "epoch": 10.77, + "learning_rate": 1.9240793201133147e-06, + "loss": 0.3273, + "step": 29168 + }, + { + "epoch": 10.77, + "learning_rate": 1.9240423697499693e-06, + "loss": 0.4964, + "step": 29169 + }, + { + "epoch": 10.77, + "learning_rate": 1.924005419386624e-06, + "loss": 0.421, + "step": 29170 + }, + { + "epoch": 10.77, + "learning_rate": 1.9239684690232786e-06, + "loss": 0.2185, + "step": 29171 + }, + { + "epoch": 10.77, + "learning_rate": 1.9239315186599336e-06, + "loss": 0.4647, + "step": 29172 + }, + { + "epoch": 10.77, + "learning_rate": 1.9238945682965883e-06, + "loss": 0.1814, + "step": 29173 + }, + { + "epoch": 10.77, + "learning_rate": 1.9238576179332433e-06, + "loss": 0.273, + "step": 29174 + }, + { + "epoch": 10.77, + "learning_rate": 1.923820667569898e-06, + "loss": 0.3323, + "step": 29175 + }, + { + "epoch": 10.77, + "learning_rate": 1.9237837172065526e-06, + "loss": 0.2821, + "step": 29176 + }, + { + "epoch": 10.77, + "learning_rate": 1.9237467668432072e-06, + "loss": 0.3549, + "step": 29177 + }, + { + "epoch": 10.77, + "learning_rate": 1.923709816479862e-06, + "loss": 0.3098, + "step": 29178 + }, + { + "epoch": 10.78, + "learning_rate": 1.923672866116517e-06, + "loss": 0.3896, + "step": 29179 + }, + { + "epoch": 10.78, + "learning_rate": 1.9236359157531716e-06, + "loss": 0.2907, + "step": 29180 + }, + { + "epoch": 10.78, + "learning_rate": 1.9235989653898266e-06, + "loss": 0.3533, + "step": 29181 + }, + { + "epoch": 10.78, + "learning_rate": 1.9235620150264813e-06, + "loss": 0.2155, + "step": 29182 + }, + { + "epoch": 10.78, + "learning_rate": 1.923525064663136e-06, + "loss": 0.2998, + "step": 29183 + }, + { + "epoch": 10.78, + "learning_rate": 1.9234881142997905e-06, + "loss": 0.3083, + "step": 29184 + }, + { + "epoch": 10.78, + "learning_rate": 1.923451163936445e-06, + "loss": 0.4107, + "step": 29185 + }, + { + "epoch": 10.78, + "learning_rate": 1.9234142135731002e-06, + "loss": 0.3522, + "step": 29186 + }, + { + "epoch": 10.78, + "learning_rate": 1.923377263209755e-06, + "loss": 0.5633, + "step": 29187 + }, + { + "epoch": 10.78, + "learning_rate": 1.92334031284641e-06, + "loss": 0.3319, + "step": 29188 + }, + { + "epoch": 10.78, + "learning_rate": 1.9233033624830646e-06, + "loss": 0.2591, + "step": 29189 + }, + { + "epoch": 10.78, + "learning_rate": 1.923266412119719e-06, + "loss": 0.3927, + "step": 29190 + }, + { + "epoch": 10.78, + "learning_rate": 1.923229461756374e-06, + "loss": 0.4079, + "step": 29191 + }, + { + "epoch": 10.78, + "learning_rate": 1.923192511393029e-06, + "loss": 0.3438, + "step": 29192 + }, + { + "epoch": 10.78, + "learning_rate": 1.9231555610296835e-06, + "loss": 0.3024, + "step": 29193 + }, + { + "epoch": 10.78, + "learning_rate": 1.9231186106663386e-06, + "loss": 0.2584, + "step": 29194 + }, + { + "epoch": 10.78, + "learning_rate": 1.9230816603029932e-06, + "loss": 0.3394, + "step": 29195 + }, + { + "epoch": 10.78, + "learning_rate": 1.9230447099396475e-06, + "loss": 0.4999, + "step": 29196 + }, + { + "epoch": 10.78, + "learning_rate": 1.9230077595763025e-06, + "loss": 0.475, + "step": 29197 + }, + { + "epoch": 10.78, + "learning_rate": 1.922970809212957e-06, + "loss": 0.4357, + "step": 29198 + }, + { + "epoch": 10.78, + "learning_rate": 1.922933858849612e-06, + "loss": 0.2703, + "step": 29199 + }, + { + "epoch": 10.78, + "learning_rate": 1.922896908486267e-06, + "loss": 0.2067, + "step": 29200 + }, + { + "epoch": 10.78, + "learning_rate": 1.922859958122922e-06, + "loss": 0.3161, + "step": 29201 + }, + { + "epoch": 10.78, + "learning_rate": 1.922823007759576e-06, + "loss": 0.3727, + "step": 29202 + }, + { + "epoch": 10.78, + "learning_rate": 1.922786057396231e-06, + "loss": 0.3647, + "step": 29203 + }, + { + "epoch": 10.78, + "learning_rate": 1.922749107032886e-06, + "loss": 0.3853, + "step": 29204 + }, + { + "epoch": 10.78, + "learning_rate": 1.9227121566695405e-06, + "loss": 0.4128, + "step": 29205 + }, + { + "epoch": 10.79, + "learning_rate": 1.9226752063061955e-06, + "loss": 0.3231, + "step": 29206 + }, + { + "epoch": 10.79, + "learning_rate": 1.92263825594285e-06, + "loss": 0.4831, + "step": 29207 + }, + { + "epoch": 10.79, + "learning_rate": 1.9226013055795048e-06, + "loss": 0.3147, + "step": 29208 + }, + { + "epoch": 10.79, + "learning_rate": 1.9225643552161594e-06, + "loss": 0.3054, + "step": 29209 + }, + { + "epoch": 10.79, + "learning_rate": 1.9225274048528145e-06, + "loss": 0.3022, + "step": 29210 + }, + { + "epoch": 10.79, + "learning_rate": 1.922490454489469e-06, + "loss": 0.2991, + "step": 29211 + }, + { + "epoch": 10.79, + "learning_rate": 1.922453504126124e-06, + "loss": 0.4055, + "step": 29212 + }, + { + "epoch": 10.79, + "learning_rate": 1.922416553762779e-06, + "loss": 0.2829, + "step": 29213 + }, + { + "epoch": 10.79, + "learning_rate": 1.9223796033994334e-06, + "loss": 0.2038, + "step": 29214 + }, + { + "epoch": 10.79, + "learning_rate": 1.922342653036088e-06, + "loss": 0.412, + "step": 29215 + }, + { + "epoch": 10.79, + "learning_rate": 1.9223057026727427e-06, + "loss": 0.3858, + "step": 29216 + }, + { + "epoch": 10.79, + "learning_rate": 1.9222687523093978e-06, + "loss": 0.2627, + "step": 29217 + }, + { + "epoch": 10.79, + "learning_rate": 1.9222318019460524e-06, + "loss": 0.4364, + "step": 29218 + }, + { + "epoch": 10.79, + "learning_rate": 1.9221948515827075e-06, + "loss": 0.2403, + "step": 29219 + }, + { + "epoch": 10.79, + "learning_rate": 1.922157901219362e-06, + "loss": 0.4064, + "step": 29220 + }, + { + "epoch": 10.79, + "learning_rate": 1.9221209508560168e-06, + "loss": 0.4514, + "step": 29221 + }, + { + "epoch": 10.79, + "learning_rate": 1.9220840004926714e-06, + "loss": 0.3666, + "step": 29222 + }, + { + "epoch": 10.79, + "learning_rate": 1.9220470501293264e-06, + "loss": 0.3468, + "step": 29223 + }, + { + "epoch": 10.79, + "learning_rate": 1.922010099765981e-06, + "loss": 0.3975, + "step": 29224 + }, + { + "epoch": 10.79, + "learning_rate": 1.9219731494026357e-06, + "loss": 0.4046, + "step": 29225 + }, + { + "epoch": 10.79, + "learning_rate": 1.9219361990392908e-06, + "loss": 0.3904, + "step": 29226 + }, + { + "epoch": 10.79, + "learning_rate": 1.9218992486759454e-06, + "loss": 0.3807, + "step": 29227 + }, + { + "epoch": 10.79, + "learning_rate": 1.9218622983126e-06, + "loss": 0.3266, + "step": 29228 + }, + { + "epoch": 10.79, + "learning_rate": 1.9218253479492547e-06, + "loss": 0.2994, + "step": 29229 + }, + { + "epoch": 10.79, + "learning_rate": 1.9217883975859097e-06, + "loss": 0.1783, + "step": 29230 + }, + { + "epoch": 10.79, + "learning_rate": 1.9217514472225644e-06, + "loss": 0.3727, + "step": 29231 + }, + { + "epoch": 10.79, + "learning_rate": 1.9217144968592194e-06, + "loss": 0.3833, + "step": 29232 + }, + { + "epoch": 10.8, + "learning_rate": 1.921677546495874e-06, + "loss": 0.43, + "step": 29233 + }, + { + "epoch": 10.8, + "learning_rate": 1.9216405961325287e-06, + "loss": 0.3986, + "step": 29234 + }, + { + "epoch": 10.8, + "learning_rate": 1.9216036457691834e-06, + "loss": 0.2803, + "step": 29235 + }, + { + "epoch": 10.8, + "learning_rate": 1.921566695405838e-06, + "loss": 0.2583, + "step": 29236 + }, + { + "epoch": 10.8, + "learning_rate": 1.921529745042493e-06, + "loss": 0.3429, + "step": 29237 + }, + { + "epoch": 10.8, + "learning_rate": 1.9214927946791477e-06, + "loss": 0.3321, + "step": 29238 + }, + { + "epoch": 10.8, + "learning_rate": 1.9214558443158027e-06, + "loss": 0.4002, + "step": 29239 + }, + { + "epoch": 10.8, + "learning_rate": 1.9214188939524574e-06, + "loss": 0.3618, + "step": 29240 + }, + { + "epoch": 10.8, + "learning_rate": 1.921381943589112e-06, + "loss": 0.2203, + "step": 29241 + }, + { + "epoch": 10.8, + "learning_rate": 1.9213449932257667e-06, + "loss": 0.3673, + "step": 29242 + }, + { + "epoch": 10.8, + "learning_rate": 1.9213080428624217e-06, + "loss": 0.331, + "step": 29243 + }, + { + "epoch": 10.8, + "learning_rate": 1.9212710924990764e-06, + "loss": 0.2606, + "step": 29244 + }, + { + "epoch": 10.8, + "learning_rate": 1.921234142135731e-06, + "loss": 0.2627, + "step": 29245 + }, + { + "epoch": 10.8, + "learning_rate": 1.921197191772386e-06, + "loss": 0.3249, + "step": 29246 + }, + { + "epoch": 10.8, + "learning_rate": 1.9211602414090403e-06, + "loss": 0.2529, + "step": 29247 + }, + { + "epoch": 10.8, + "learning_rate": 1.9211232910456953e-06, + "loss": 0.3858, + "step": 29248 + }, + { + "epoch": 10.8, + "learning_rate": 1.92108634068235e-06, + "loss": 0.3053, + "step": 29249 + }, + { + "epoch": 10.8, + "learning_rate": 1.921049390319005e-06, + "loss": 0.4113, + "step": 29250 + }, + { + "epoch": 10.8, + "learning_rate": 1.9210124399556597e-06, + "loss": 0.3288, + "step": 29251 + }, + { + "epoch": 10.8, + "learning_rate": 1.9209754895923147e-06, + "loss": 0.3534, + "step": 29252 + }, + { + "epoch": 10.8, + "learning_rate": 1.920938539228969e-06, + "loss": 0.4597, + "step": 29253 + }, + { + "epoch": 10.8, + "learning_rate": 1.920901588865624e-06, + "loss": 0.4049, + "step": 29254 + }, + { + "epoch": 10.8, + "learning_rate": 1.9208646385022786e-06, + "loss": 0.4232, + "step": 29255 + }, + { + "epoch": 10.8, + "learning_rate": 1.9208276881389333e-06, + "loss": 0.3215, + "step": 29256 + }, + { + "epoch": 10.8, + "learning_rate": 1.9207907377755883e-06, + "loss": 0.3226, + "step": 29257 + }, + { + "epoch": 10.8, + "learning_rate": 1.920753787412243e-06, + "loss": 0.2028, + "step": 29258 + }, + { + "epoch": 10.8, + "learning_rate": 1.9207168370488976e-06, + "loss": 0.3702, + "step": 29259 + }, + { + "epoch": 10.81, + "learning_rate": 1.9206798866855522e-06, + "loss": 0.3397, + "step": 29260 + }, + { + "epoch": 10.81, + "learning_rate": 1.9206429363222073e-06, + "loss": 0.3264, + "step": 29261 + }, + { + "epoch": 10.81, + "learning_rate": 1.920605985958862e-06, + "loss": 0.4216, + "step": 29262 + }, + { + "epoch": 10.81, + "learning_rate": 1.920569035595517e-06, + "loss": 0.2536, + "step": 29263 + }, + { + "epoch": 10.81, + "learning_rate": 1.9205320852321716e-06, + "loss": 0.3803, + "step": 29264 + }, + { + "epoch": 10.81, + "learning_rate": 1.9204951348688263e-06, + "loss": 0.3309, + "step": 29265 + }, + { + "epoch": 10.81, + "learning_rate": 1.920458184505481e-06, + "loss": 0.3508, + "step": 29266 + }, + { + "epoch": 10.81, + "learning_rate": 1.9204212341421355e-06, + "loss": 0.2365, + "step": 29267 + }, + { + "epoch": 10.81, + "learning_rate": 1.9203842837787906e-06, + "loss": 0.2986, + "step": 29268 + }, + { + "epoch": 10.81, + "learning_rate": 1.9203473334154452e-06, + "loss": 0.3827, + "step": 29269 + }, + { + "epoch": 10.81, + "learning_rate": 1.9203103830521003e-06, + "loss": 0.2644, + "step": 29270 + }, + { + "epoch": 10.81, + "learning_rate": 1.920273432688755e-06, + "loss": 0.2186, + "step": 29271 + }, + { + "epoch": 10.81, + "learning_rate": 1.9202364823254096e-06, + "loss": 0.3253, + "step": 29272 + }, + { + "epoch": 10.81, + "learning_rate": 1.920199531962064e-06, + "loss": 0.4047, + "step": 29273 + }, + { + "epoch": 10.81, + "learning_rate": 1.9201625815987193e-06, + "loss": 0.3802, + "step": 29274 + }, + { + "epoch": 10.81, + "learning_rate": 1.920125631235374e-06, + "loss": 0.4142, + "step": 29275 + }, + { + "epoch": 10.81, + "learning_rate": 1.9200886808720285e-06, + "loss": 0.2531, + "step": 29276 + }, + { + "epoch": 10.81, + "learning_rate": 1.9200517305086836e-06, + "loss": 0.2744, + "step": 29277 + }, + { + "epoch": 10.81, + "learning_rate": 1.9200147801453382e-06, + "loss": 0.3284, + "step": 29278 + }, + { + "epoch": 10.81, + "learning_rate": 1.919977829781993e-06, + "loss": 0.4014, + "step": 29279 + }, + { + "epoch": 10.81, + "learning_rate": 1.9199408794186475e-06, + "loss": 0.3607, + "step": 29280 + }, + { + "epoch": 10.81, + "learning_rate": 1.9199039290553026e-06, + "loss": 0.3174, + "step": 29281 + }, + { + "epoch": 10.81, + "learning_rate": 1.919866978691957e-06, + "loss": 0.3165, + "step": 29282 + }, + { + "epoch": 10.81, + "learning_rate": 1.9198300283286123e-06, + "loss": 0.3009, + "step": 29283 + }, + { + "epoch": 10.81, + "learning_rate": 1.919793077965267e-06, + "loss": 0.4698, + "step": 29284 + }, + { + "epoch": 10.81, + "learning_rate": 1.919756127601921e-06, + "loss": 0.2419, + "step": 29285 + }, + { + "epoch": 10.81, + "learning_rate": 1.919719177238576e-06, + "loss": 0.2879, + "step": 29286 + }, + { + "epoch": 10.81, + "learning_rate": 1.919682226875231e-06, + "loss": 0.2502, + "step": 29287 + }, + { + "epoch": 10.82, + "learning_rate": 1.919645276511886e-06, + "loss": 0.3269, + "step": 29288 + }, + { + "epoch": 10.82, + "learning_rate": 1.9196083261485405e-06, + "loss": 0.3557, + "step": 29289 + }, + { + "epoch": 10.82, + "learning_rate": 1.9195713757851956e-06, + "loss": 0.3269, + "step": 29290 + }, + { + "epoch": 10.82, + "learning_rate": 1.9195344254218498e-06, + "loss": 0.3315, + "step": 29291 + }, + { + "epoch": 10.82, + "learning_rate": 1.919497475058505e-06, + "loss": 0.4527, + "step": 29292 + }, + { + "epoch": 10.82, + "learning_rate": 1.9194605246951595e-06, + "loss": 0.2711, + "step": 29293 + }, + { + "epoch": 10.82, + "learning_rate": 1.919423574331814e-06, + "loss": 0.399, + "step": 29294 + }, + { + "epoch": 10.82, + "learning_rate": 1.919386623968469e-06, + "loss": 0.354, + "step": 29295 + }, + { + "epoch": 10.82, + "learning_rate": 1.919349673605124e-06, + "loss": 0.3073, + "step": 29296 + }, + { + "epoch": 10.82, + "learning_rate": 1.9193127232417784e-06, + "loss": 0.2749, + "step": 29297 + }, + { + "epoch": 10.82, + "learning_rate": 1.919275772878433e-06, + "loss": 0.2529, + "step": 29298 + }, + { + "epoch": 10.82, + "learning_rate": 1.919238822515088e-06, + "loss": 0.4272, + "step": 29299 + }, + { + "epoch": 10.82, + "learning_rate": 1.9192018721517428e-06, + "loss": 0.3785, + "step": 29300 + }, + { + "epoch": 10.82, + "learning_rate": 1.919164921788398e-06, + "loss": 0.3712, + "step": 29301 + }, + { + "epoch": 10.82, + "learning_rate": 1.9191279714250525e-06, + "loss": 0.3284, + "step": 29302 + }, + { + "epoch": 10.82, + "learning_rate": 1.919091021061707e-06, + "loss": 0.3713, + "step": 29303 + }, + { + "epoch": 10.82, + "learning_rate": 1.9190540706983617e-06, + "loss": 0.2657, + "step": 29304 + }, + { + "epoch": 10.82, + "learning_rate": 1.9190171203350164e-06, + "loss": 0.2924, + "step": 29305 + }, + { + "epoch": 10.82, + "learning_rate": 1.9189801699716714e-06, + "loss": 0.4471, + "step": 29306 + }, + { + "epoch": 10.82, + "learning_rate": 1.918943219608326e-06, + "loss": 0.332, + "step": 29307 + }, + { + "epoch": 10.82, + "learning_rate": 1.918906269244981e-06, + "loss": 0.276, + "step": 29308 + }, + { + "epoch": 10.82, + "learning_rate": 1.9188693188816358e-06, + "loss": 0.3379, + "step": 29309 + }, + { + "epoch": 10.82, + "learning_rate": 1.9188323685182904e-06, + "loss": 0.3239, + "step": 29310 + }, + { + "epoch": 10.82, + "learning_rate": 1.918795418154945e-06, + "loss": 0.4446, + "step": 29311 + }, + { + "epoch": 10.82, + "learning_rate": 1.9187584677916e-06, + "loss": 0.3109, + "step": 29312 + }, + { + "epoch": 10.82, + "learning_rate": 1.9187215174282547e-06, + "loss": 0.3921, + "step": 29313 + }, + { + "epoch": 10.82, + "learning_rate": 1.9186845670649094e-06, + "loss": 0.4516, + "step": 29314 + }, + { + "epoch": 10.83, + "learning_rate": 1.9186476167015644e-06, + "loss": 0.2968, + "step": 29315 + }, + { + "epoch": 10.83, + "learning_rate": 1.918610666338219e-06, + "loss": 0.2113, + "step": 29316 + }, + { + "epoch": 10.83, + "learning_rate": 1.9185737159748737e-06, + "loss": 0.3366, + "step": 29317 + }, + { + "epoch": 10.83, + "learning_rate": 1.9185367656115283e-06, + "loss": 0.2608, + "step": 29318 + }, + { + "epoch": 10.83, + "learning_rate": 1.9184998152481834e-06, + "loss": 0.2718, + "step": 29319 + }, + { + "epoch": 10.83, + "learning_rate": 1.918462864884838e-06, + "loss": 0.4242, + "step": 29320 + }, + { + "epoch": 10.83, + "learning_rate": 1.918425914521493e-06, + "loss": 0.3964, + "step": 29321 + }, + { + "epoch": 10.83, + "learning_rate": 1.9183889641581477e-06, + "loss": 0.3189, + "step": 29322 + }, + { + "epoch": 10.83, + "learning_rate": 1.9183520137948024e-06, + "loss": 0.2921, + "step": 29323 + }, + { + "epoch": 10.83, + "learning_rate": 1.918315063431457e-06, + "loss": 0.4014, + "step": 29324 + }, + { + "epoch": 10.83, + "learning_rate": 1.9182781130681116e-06, + "loss": 0.3262, + "step": 29325 + }, + { + "epoch": 10.83, + "learning_rate": 1.9182411627047667e-06, + "loss": 0.4191, + "step": 29326 + }, + { + "epoch": 10.83, + "learning_rate": 1.9182042123414213e-06, + "loss": 0.3027, + "step": 29327 + }, + { + "epoch": 10.83, + "learning_rate": 1.9181672619780764e-06, + "loss": 0.2382, + "step": 29328 + }, + { + "epoch": 10.83, + "learning_rate": 1.918130311614731e-06, + "loss": 0.3189, + "step": 29329 + }, + { + "epoch": 10.83, + "learning_rate": 1.9180933612513857e-06, + "loss": 0.3287, + "step": 29330 + }, + { + "epoch": 10.83, + "learning_rate": 1.9180564108880403e-06, + "loss": 0.2775, + "step": 29331 + }, + { + "epoch": 10.83, + "learning_rate": 1.9180194605246954e-06, + "loss": 0.317, + "step": 29332 + }, + { + "epoch": 10.83, + "learning_rate": 1.91798251016135e-06, + "loss": 0.3632, + "step": 29333 + }, + { + "epoch": 10.83, + "learning_rate": 1.9179455597980046e-06, + "loss": 0.3296, + "step": 29334 + }, + { + "epoch": 10.83, + "learning_rate": 1.9179086094346597e-06, + "loss": 0.3773, + "step": 29335 + }, + { + "epoch": 10.83, + "learning_rate": 1.917871659071314e-06, + "loss": 0.3824, + "step": 29336 + }, + { + "epoch": 10.83, + "learning_rate": 1.917834708707969e-06, + "loss": 0.3251, + "step": 29337 + }, + { + "epoch": 10.83, + "learning_rate": 1.9177977583446236e-06, + "loss": 0.2661, + "step": 29338 + }, + { + "epoch": 10.83, + "learning_rate": 1.9177608079812787e-06, + "loss": 0.355, + "step": 29339 + }, + { + "epoch": 10.83, + "learning_rate": 1.9177238576179333e-06, + "loss": 0.3331, + "step": 29340 + }, + { + "epoch": 10.83, + "learning_rate": 1.9176869072545884e-06, + "loss": 0.4275, + "step": 29341 + }, + { + "epoch": 10.84, + "learning_rate": 1.9176499568912426e-06, + "loss": 0.2256, + "step": 29342 + }, + { + "epoch": 10.84, + "learning_rate": 1.9176130065278976e-06, + "loss": 0.3657, + "step": 29343 + }, + { + "epoch": 10.84, + "learning_rate": 1.9175760561645523e-06, + "loss": 0.3774, + "step": 29344 + }, + { + "epoch": 10.84, + "learning_rate": 1.917539105801207e-06, + "loss": 0.5093, + "step": 29345 + }, + { + "epoch": 10.84, + "learning_rate": 1.917502155437862e-06, + "loss": 0.3593, + "step": 29346 + }, + { + "epoch": 10.84, + "learning_rate": 1.9174652050745166e-06, + "loss": 0.3539, + "step": 29347 + }, + { + "epoch": 10.84, + "learning_rate": 1.9174282547111712e-06, + "loss": 0.4181, + "step": 29348 + }, + { + "epoch": 10.84, + "learning_rate": 1.917391304347826e-06, + "loss": 0.2399, + "step": 29349 + }, + { + "epoch": 10.84, + "learning_rate": 1.917354353984481e-06, + "loss": 0.3409, + "step": 29350 + }, + { + "epoch": 10.84, + "learning_rate": 1.9173174036211356e-06, + "loss": 0.2909, + "step": 29351 + }, + { + "epoch": 10.84, + "learning_rate": 1.9172804532577906e-06, + "loss": 0.2737, + "step": 29352 + }, + { + "epoch": 10.84, + "learning_rate": 1.9172435028944453e-06, + "loss": 0.288, + "step": 29353 + }, + { + "epoch": 10.84, + "learning_rate": 1.9172065525311e-06, + "loss": 0.3247, + "step": 29354 + }, + { + "epoch": 10.84, + "learning_rate": 1.9171696021677546e-06, + "loss": 0.3901, + "step": 29355 + }, + { + "epoch": 10.84, + "learning_rate": 1.917132651804409e-06, + "loss": 0.2572, + "step": 29356 + }, + { + "epoch": 10.84, + "learning_rate": 1.9170957014410642e-06, + "loss": 0.3298, + "step": 29357 + }, + { + "epoch": 10.84, + "learning_rate": 1.917058751077719e-06, + "loss": 0.2764, + "step": 29358 + }, + { + "epoch": 10.84, + "learning_rate": 1.917021800714374e-06, + "loss": 0.3988, + "step": 29359 + }, + { + "epoch": 10.84, + "learning_rate": 1.9169848503510286e-06, + "loss": 0.2714, + "step": 29360 + }, + { + "epoch": 10.84, + "learning_rate": 1.9169478999876832e-06, + "loss": 0.363, + "step": 29361 + }, + { + "epoch": 10.84, + "learning_rate": 1.916910949624338e-06, + "loss": 0.3605, + "step": 29362 + }, + { + "epoch": 10.84, + "learning_rate": 1.916873999260993e-06, + "loss": 0.3848, + "step": 29363 + }, + { + "epoch": 10.84, + "learning_rate": 1.9168370488976475e-06, + "loss": 0.3748, + "step": 29364 + }, + { + "epoch": 10.84, + "learning_rate": 1.916800098534302e-06, + "loss": 0.3819, + "step": 29365 + }, + { + "epoch": 10.84, + "learning_rate": 1.9167631481709572e-06, + "loss": 0.3298, + "step": 29366 + }, + { + "epoch": 10.84, + "learning_rate": 1.916726197807612e-06, + "loss": 0.3541, + "step": 29367 + }, + { + "epoch": 10.84, + "learning_rate": 1.9166892474442665e-06, + "loss": 0.3292, + "step": 29368 + }, + { + "epoch": 10.85, + "learning_rate": 1.916652297080921e-06, + "loss": 0.2245, + "step": 29369 + }, + { + "epoch": 10.85, + "learning_rate": 1.9166153467175762e-06, + "loss": 0.4774, + "step": 29370 + }, + { + "epoch": 10.85, + "learning_rate": 1.916578396354231e-06, + "loss": 0.2762, + "step": 29371 + }, + { + "epoch": 10.85, + "learning_rate": 1.916541445990886e-06, + "loss": 0.278, + "step": 29372 + }, + { + "epoch": 10.85, + "learning_rate": 1.9165044956275405e-06, + "loss": 0.2643, + "step": 29373 + }, + { + "epoch": 10.85, + "learning_rate": 1.916467545264195e-06, + "loss": 0.3934, + "step": 29374 + }, + { + "epoch": 10.85, + "learning_rate": 1.91643059490085e-06, + "loss": 0.2738, + "step": 29375 + }, + { + "epoch": 10.85, + "learning_rate": 1.9163936445375045e-06, + "loss": 0.3595, + "step": 29376 + }, + { + "epoch": 10.85, + "learning_rate": 1.9163566941741595e-06, + "loss": 0.2176, + "step": 29377 + }, + { + "epoch": 10.85, + "learning_rate": 1.916319743810814e-06, + "loss": 0.3573, + "step": 29378 + }, + { + "epoch": 10.85, + "learning_rate": 1.9162827934474692e-06, + "loss": 0.4108, + "step": 29379 + }, + { + "epoch": 10.85, + "learning_rate": 1.916245843084124e-06, + "loss": 0.2884, + "step": 29380 + }, + { + "epoch": 10.85, + "learning_rate": 1.9162088927207785e-06, + "loss": 0.4131, + "step": 29381 + }, + { + "epoch": 10.85, + "learning_rate": 1.916171942357433e-06, + "loss": 0.2516, + "step": 29382 + }, + { + "epoch": 10.85, + "learning_rate": 1.916134991994088e-06, + "loss": 0.2836, + "step": 29383 + }, + { + "epoch": 10.85, + "learning_rate": 1.916098041630743e-06, + "loss": 0.5202, + "step": 29384 + }, + { + "epoch": 10.85, + "learning_rate": 1.9160610912673975e-06, + "loss": 0.2882, + "step": 29385 + }, + { + "epoch": 10.85, + "learning_rate": 1.9160241409040525e-06, + "loss": 0.3363, + "step": 29386 + }, + { + "epoch": 10.85, + "learning_rate": 1.9159871905407067e-06, + "loss": 0.3327, + "step": 29387 + }, + { + "epoch": 10.85, + "learning_rate": 1.9159502401773618e-06, + "loss": 0.3994, + "step": 29388 + }, + { + "epoch": 10.85, + "learning_rate": 1.9159132898140164e-06, + "loss": 0.3543, + "step": 29389 + }, + { + "epoch": 10.85, + "learning_rate": 1.9158763394506715e-06, + "loss": 0.4052, + "step": 29390 + }, + { + "epoch": 10.85, + "learning_rate": 1.915839389087326e-06, + "loss": 0.2231, + "step": 29391 + }, + { + "epoch": 10.85, + "learning_rate": 1.915802438723981e-06, + "loss": 0.355, + "step": 29392 + }, + { + "epoch": 10.85, + "learning_rate": 1.9157654883606354e-06, + "loss": 0.2138, + "step": 29393 + }, + { + "epoch": 10.85, + "learning_rate": 1.91572853799729e-06, + "loss": 0.3556, + "step": 29394 + }, + { + "epoch": 10.85, + "learning_rate": 1.915691587633945e-06, + "loss": 0.3193, + "step": 29395 + }, + { + "epoch": 10.86, + "learning_rate": 1.9156546372705997e-06, + "loss": 0.2865, + "step": 29396 + }, + { + "epoch": 10.86, + "learning_rate": 1.9156176869072548e-06, + "loss": 0.3853, + "step": 29397 + }, + { + "epoch": 10.86, + "learning_rate": 1.9155807365439094e-06, + "loss": 0.2888, + "step": 29398 + }, + { + "epoch": 10.86, + "learning_rate": 1.915543786180564e-06, + "loss": 0.2644, + "step": 29399 + }, + { + "epoch": 10.86, + "learning_rate": 1.9155068358172187e-06, + "loss": 0.4293, + "step": 29400 + }, + { + "epoch": 10.86, + "learning_rate": 1.9154698854538738e-06, + "loss": 0.3712, + "step": 29401 + }, + { + "epoch": 10.86, + "learning_rate": 1.9154329350905284e-06, + "loss": 0.3455, + "step": 29402 + }, + { + "epoch": 10.86, + "learning_rate": 1.9153959847271835e-06, + "loss": 0.2166, + "step": 29403 + }, + { + "epoch": 10.86, + "learning_rate": 1.915359034363838e-06, + "loss": 0.3569, + "step": 29404 + }, + { + "epoch": 10.86, + "learning_rate": 1.9153220840004927e-06, + "loss": 0.3518, + "step": 29405 + }, + { + "epoch": 10.86, + "learning_rate": 1.9152851336371474e-06, + "loss": 0.2855, + "step": 29406 + }, + { + "epoch": 10.86, + "learning_rate": 1.915248183273802e-06, + "loss": 0.3042, + "step": 29407 + }, + { + "epoch": 10.86, + "learning_rate": 1.915211232910457e-06, + "loss": 0.3219, + "step": 29408 + }, + { + "epoch": 10.86, + "learning_rate": 1.9151742825471117e-06, + "loss": 0.4093, + "step": 29409 + }, + { + "epoch": 10.86, + "learning_rate": 1.9151373321837668e-06, + "loss": 0.424, + "step": 29410 + }, + { + "epoch": 10.86, + "learning_rate": 1.9151003818204214e-06, + "loss": 0.4067, + "step": 29411 + }, + { + "epoch": 10.86, + "learning_rate": 1.915063431457076e-06, + "loss": 0.328, + "step": 29412 + }, + { + "epoch": 10.86, + "learning_rate": 1.9150264810937307e-06, + "loss": 0.4267, + "step": 29413 + }, + { + "epoch": 10.86, + "learning_rate": 1.9149895307303853e-06, + "loss": 0.4085, + "step": 29414 + }, + { + "epoch": 10.86, + "learning_rate": 1.9149525803670404e-06, + "loss": 0.3833, + "step": 29415 + }, + { + "epoch": 10.86, + "learning_rate": 1.914915630003695e-06, + "loss": 0.2737, + "step": 29416 + }, + { + "epoch": 10.86, + "learning_rate": 1.91487867964035e-06, + "loss": 0.1982, + "step": 29417 + }, + { + "epoch": 10.86, + "learning_rate": 1.9148417292770047e-06, + "loss": 0.3884, + "step": 29418 + }, + { + "epoch": 10.86, + "learning_rate": 1.9148047789136593e-06, + "loss": 0.4588, + "step": 29419 + }, + { + "epoch": 10.86, + "learning_rate": 1.914767828550314e-06, + "loss": 0.3714, + "step": 29420 + }, + { + "epoch": 10.86, + "learning_rate": 1.914730878186969e-06, + "loss": 0.3961, + "step": 29421 + }, + { + "epoch": 10.86, + "learning_rate": 1.9146939278236237e-06, + "loss": 0.4855, + "step": 29422 + }, + { + "epoch": 10.87, + "learning_rate": 1.9146569774602787e-06, + "loss": 0.3603, + "step": 29423 + }, + { + "epoch": 10.87, + "learning_rate": 1.9146200270969334e-06, + "loss": 0.2299, + "step": 29424 + }, + { + "epoch": 10.87, + "learning_rate": 1.914583076733588e-06, + "loss": 0.4114, + "step": 29425 + }, + { + "epoch": 10.87, + "learning_rate": 1.9145461263702426e-06, + "loss": 0.3437, + "step": 29426 + }, + { + "epoch": 10.87, + "learning_rate": 1.9145091760068973e-06, + "loss": 0.3279, + "step": 29427 + }, + { + "epoch": 10.87, + "learning_rate": 1.9144722256435523e-06, + "loss": 0.267, + "step": 29428 + }, + { + "epoch": 10.87, + "learning_rate": 1.914435275280207e-06, + "loss": 0.4036, + "step": 29429 + }, + { + "epoch": 10.87, + "learning_rate": 1.914398324916862e-06, + "loss": 0.3622, + "step": 29430 + }, + { + "epoch": 10.87, + "learning_rate": 1.9143613745535167e-06, + "loss": 0.3499, + "step": 29431 + }, + { + "epoch": 10.87, + "learning_rate": 1.9143244241901713e-06, + "loss": 0.3555, + "step": 29432 + }, + { + "epoch": 10.87, + "learning_rate": 1.914287473826826e-06, + "loss": 0.3511, + "step": 29433 + }, + { + "epoch": 10.87, + "learning_rate": 1.9142505234634806e-06, + "loss": 0.4552, + "step": 29434 + }, + { + "epoch": 10.87, + "learning_rate": 1.9142135731001356e-06, + "loss": 0.3349, + "step": 29435 + }, + { + "epoch": 10.87, + "learning_rate": 1.9141766227367903e-06, + "loss": 0.3629, + "step": 29436 + }, + { + "epoch": 10.87, + "learning_rate": 1.9141396723734453e-06, + "loss": 0.2837, + "step": 29437 + }, + { + "epoch": 10.87, + "learning_rate": 1.9141027220100995e-06, + "loss": 0.2803, + "step": 29438 + }, + { + "epoch": 10.87, + "learning_rate": 1.9140657716467546e-06, + "loss": 0.2684, + "step": 29439 + }, + { + "epoch": 10.87, + "learning_rate": 1.9140288212834092e-06, + "loss": 0.3064, + "step": 29440 + }, + { + "epoch": 10.87, + "learning_rate": 1.9139918709200643e-06, + "loss": 0.3401, + "step": 29441 + }, + { + "epoch": 10.87, + "learning_rate": 1.913954920556719e-06, + "loss": 0.3228, + "step": 29442 + }, + { + "epoch": 10.87, + "learning_rate": 1.9139179701933736e-06, + "loss": 0.5544, + "step": 29443 + }, + { + "epoch": 10.87, + "learning_rate": 1.913881019830028e-06, + "loss": 0.2584, + "step": 29444 + }, + { + "epoch": 10.87, + "learning_rate": 1.913844069466683e-06, + "loss": 0.3529, + "step": 29445 + }, + { + "epoch": 10.87, + "learning_rate": 1.913807119103338e-06, + "loss": 0.3056, + "step": 29446 + }, + { + "epoch": 10.87, + "learning_rate": 1.9137701687399925e-06, + "loss": 0.3409, + "step": 29447 + }, + { + "epoch": 10.87, + "learning_rate": 1.9137332183766476e-06, + "loss": 0.3364, + "step": 29448 + }, + { + "epoch": 10.87, + "learning_rate": 1.9136962680133022e-06, + "loss": 0.3489, + "step": 29449 + }, + { + "epoch": 10.88, + "learning_rate": 1.913659317649957e-06, + "loss": 0.2666, + "step": 29450 + }, + { + "epoch": 10.88, + "learning_rate": 1.9136223672866115e-06, + "loss": 0.2877, + "step": 29451 + }, + { + "epoch": 10.88, + "learning_rate": 1.9135854169232666e-06, + "loss": 0.3877, + "step": 29452 + }, + { + "epoch": 10.88, + "learning_rate": 1.913548466559921e-06, + "loss": 0.2785, + "step": 29453 + }, + { + "epoch": 10.88, + "learning_rate": 1.913511516196576e-06, + "loss": 0.2741, + "step": 29454 + }, + { + "epoch": 10.88, + "learning_rate": 1.913474565833231e-06, + "loss": 0.286, + "step": 29455 + }, + { + "epoch": 10.88, + "learning_rate": 1.9134376154698855e-06, + "loss": 0.2919, + "step": 29456 + }, + { + "epoch": 10.88, + "learning_rate": 1.91340066510654e-06, + "loss": 0.3971, + "step": 29457 + }, + { + "epoch": 10.88, + "learning_rate": 1.913363714743195e-06, + "loss": 0.2799, + "step": 29458 + }, + { + "epoch": 10.88, + "learning_rate": 1.91332676437985e-06, + "loss": 0.3075, + "step": 29459 + }, + { + "epoch": 10.88, + "learning_rate": 1.9132898140165045e-06, + "loss": 0.312, + "step": 29460 + }, + { + "epoch": 10.88, + "learning_rate": 1.9132528636531596e-06, + "loss": 0.4076, + "step": 29461 + }, + { + "epoch": 10.88, + "learning_rate": 1.913215913289814e-06, + "loss": 0.2828, + "step": 29462 + }, + { + "epoch": 10.88, + "learning_rate": 1.913178962926469e-06, + "loss": 0.4567, + "step": 29463 + }, + { + "epoch": 10.88, + "learning_rate": 1.9131420125631235e-06, + "loss": 0.3502, + "step": 29464 + }, + { + "epoch": 10.88, + "learning_rate": 1.913105062199778e-06, + "loss": 0.3966, + "step": 29465 + }, + { + "epoch": 10.88, + "learning_rate": 1.913068111836433e-06, + "loss": 0.3414, + "step": 29466 + }, + { + "epoch": 10.88, + "learning_rate": 1.913031161473088e-06, + "loss": 0.2519, + "step": 29467 + }, + { + "epoch": 10.88, + "learning_rate": 1.912994211109743e-06, + "loss": 0.3853, + "step": 29468 + }, + { + "epoch": 10.88, + "learning_rate": 1.9129572607463975e-06, + "loss": 0.2605, + "step": 29469 + }, + { + "epoch": 10.88, + "learning_rate": 1.912920310383052e-06, + "loss": 0.2326, + "step": 29470 + }, + { + "epoch": 10.88, + "learning_rate": 1.9128833600197068e-06, + "loss": 0.3598, + "step": 29471 + }, + { + "epoch": 10.88, + "learning_rate": 1.912846409656362e-06, + "loss": 0.3383, + "step": 29472 + }, + { + "epoch": 10.88, + "learning_rate": 1.9128094592930165e-06, + "loss": 0.2759, + "step": 29473 + }, + { + "epoch": 10.88, + "learning_rate": 1.912772508929671e-06, + "loss": 0.3672, + "step": 29474 + }, + { + "epoch": 10.88, + "learning_rate": 1.912735558566326e-06, + "loss": 0.2872, + "step": 29475 + }, + { + "epoch": 10.88, + "learning_rate": 1.9126986082029804e-06, + "loss": 0.2326, + "step": 29476 + }, + { + "epoch": 10.89, + "learning_rate": 1.9126616578396354e-06, + "loss": 0.3445, + "step": 29477 + }, + { + "epoch": 10.89, + "learning_rate": 1.91262470747629e-06, + "loss": 0.2994, + "step": 29478 + }, + { + "epoch": 10.89, + "learning_rate": 1.912587757112945e-06, + "loss": 0.4047, + "step": 29479 + }, + { + "epoch": 10.89, + "learning_rate": 1.9125508067495998e-06, + "loss": 0.2465, + "step": 29480 + }, + { + "epoch": 10.89, + "learning_rate": 1.912513856386255e-06, + "loss": 0.4307, + "step": 29481 + }, + { + "epoch": 10.89, + "learning_rate": 1.912476906022909e-06, + "loss": 0.3335, + "step": 29482 + }, + { + "epoch": 10.89, + "learning_rate": 1.912439955659564e-06, + "loss": 0.2941, + "step": 29483 + }, + { + "epoch": 10.89, + "learning_rate": 1.9124030052962187e-06, + "loss": 0.4022, + "step": 29484 + }, + { + "epoch": 10.89, + "learning_rate": 1.9123660549328734e-06, + "loss": 0.3592, + "step": 29485 + }, + { + "epoch": 10.89, + "learning_rate": 1.9123291045695284e-06, + "loss": 0.2774, + "step": 29486 + }, + { + "epoch": 10.89, + "learning_rate": 1.912292154206183e-06, + "loss": 0.2868, + "step": 29487 + }, + { + "epoch": 10.89, + "learning_rate": 1.9122552038428377e-06, + "loss": 0.3442, + "step": 29488 + }, + { + "epoch": 10.89, + "learning_rate": 1.9122182534794924e-06, + "loss": 0.291, + "step": 29489 + }, + { + "epoch": 10.89, + "learning_rate": 1.9121813031161474e-06, + "loss": 0.4811, + "step": 29490 + }, + { + "epoch": 10.89, + "learning_rate": 1.912144352752802e-06, + "loss": 0.3755, + "step": 29491 + }, + { + "epoch": 10.89, + "learning_rate": 1.912107402389457e-06, + "loss": 0.2944, + "step": 29492 + }, + { + "epoch": 10.89, + "learning_rate": 1.9120704520261117e-06, + "loss": 0.435, + "step": 29493 + }, + { + "epoch": 10.89, + "learning_rate": 1.9120335016627664e-06, + "loss": 0.3008, + "step": 29494 + }, + { + "epoch": 10.89, + "learning_rate": 1.911996551299421e-06, + "loss": 0.6065, + "step": 29495 + }, + { + "epoch": 10.89, + "learning_rate": 1.9119596009360757e-06, + "loss": 0.4036, + "step": 29496 + }, + { + "epoch": 10.89, + "learning_rate": 1.9119226505727307e-06, + "loss": 0.2557, + "step": 29497 + }, + { + "epoch": 10.89, + "learning_rate": 1.9118857002093853e-06, + "loss": 0.3676, + "step": 29498 + }, + { + "epoch": 10.89, + "learning_rate": 1.9118487498460404e-06, + "loss": 0.3287, + "step": 29499 + }, + { + "epoch": 10.89, + "learning_rate": 1.911811799482695e-06, + "loss": 0.3352, + "step": 29500 + }, + { + "epoch": 10.89, + "learning_rate": 1.9117748491193497e-06, + "loss": 0.234, + "step": 29501 + }, + { + "epoch": 10.89, + "learning_rate": 1.9117378987560043e-06, + "loss": 0.3598, + "step": 29502 + }, + { + "epoch": 10.89, + "learning_rate": 1.9117009483926594e-06, + "loss": 0.4126, + "step": 29503 + }, + { + "epoch": 10.9, + "learning_rate": 1.911663998029314e-06, + "loss": 0.2132, + "step": 29504 + }, + { + "epoch": 10.9, + "learning_rate": 1.9116270476659687e-06, + "loss": 0.2836, + "step": 29505 + }, + { + "epoch": 10.9, + "learning_rate": 1.9115900973026237e-06, + "loss": 0.2507, + "step": 29506 + }, + { + "epoch": 10.9, + "learning_rate": 1.9115531469392783e-06, + "loss": 0.5057, + "step": 29507 + }, + { + "epoch": 10.9, + "learning_rate": 1.911516196575933e-06, + "loss": 0.2077, + "step": 29508 + }, + { + "epoch": 10.9, + "learning_rate": 1.9114792462125876e-06, + "loss": 0.2379, + "step": 29509 + }, + { + "epoch": 10.9, + "learning_rate": 1.9114422958492427e-06, + "loss": 0.2123, + "step": 29510 + }, + { + "epoch": 10.9, + "learning_rate": 1.9114053454858973e-06, + "loss": 0.3392, + "step": 29511 + }, + { + "epoch": 10.9, + "learning_rate": 1.9113683951225524e-06, + "loss": 0.293, + "step": 29512 + }, + { + "epoch": 10.9, + "learning_rate": 1.911331444759207e-06, + "loss": 0.4528, + "step": 29513 + }, + { + "epoch": 10.9, + "learning_rate": 1.9112944943958616e-06, + "loss": 0.3041, + "step": 29514 + }, + { + "epoch": 10.9, + "learning_rate": 1.9112575440325163e-06, + "loss": 0.4546, + "step": 29515 + }, + { + "epoch": 10.9, + "learning_rate": 1.911220593669171e-06, + "loss": 0.3409, + "step": 29516 + }, + { + "epoch": 10.9, + "learning_rate": 1.911183643305826e-06, + "loss": 0.3651, + "step": 29517 + }, + { + "epoch": 10.9, + "learning_rate": 1.9111466929424806e-06, + "loss": 0.4363, + "step": 29518 + }, + { + "epoch": 10.9, + "learning_rate": 1.9111097425791357e-06, + "loss": 0.2514, + "step": 29519 + }, + { + "epoch": 10.9, + "learning_rate": 1.9110727922157903e-06, + "loss": 0.3235, + "step": 29520 + }, + { + "epoch": 10.9, + "learning_rate": 1.911035841852445e-06, + "loss": 0.3098, + "step": 29521 + }, + { + "epoch": 10.9, + "learning_rate": 1.9109988914890996e-06, + "loss": 0.2925, + "step": 29522 + }, + { + "epoch": 10.9, + "learning_rate": 1.9109619411257542e-06, + "loss": 0.2712, + "step": 29523 + }, + { + "epoch": 10.9, + "learning_rate": 1.9109249907624093e-06, + "loss": 0.3974, + "step": 29524 + }, + { + "epoch": 10.9, + "learning_rate": 1.910888040399064e-06, + "loss": 0.4034, + "step": 29525 + }, + { + "epoch": 10.9, + "learning_rate": 1.910851090035719e-06, + "loss": 0.2626, + "step": 29526 + }, + { + "epoch": 10.9, + "learning_rate": 1.910814139672373e-06, + "loss": 0.2252, + "step": 29527 + }, + { + "epoch": 10.9, + "learning_rate": 1.9107771893090283e-06, + "loss": 0.3444, + "step": 29528 + }, + { + "epoch": 10.9, + "learning_rate": 1.910740238945683e-06, + "loss": 0.2506, + "step": 29529 + }, + { + "epoch": 10.9, + "learning_rate": 1.910703288582338e-06, + "loss": 0.3592, + "step": 29530 + }, + { + "epoch": 10.91, + "learning_rate": 1.9106663382189926e-06, + "loss": 0.4161, + "step": 29531 + }, + { + "epoch": 10.91, + "learning_rate": 1.9106293878556476e-06, + "loss": 0.4087, + "step": 29532 + }, + { + "epoch": 10.91, + "learning_rate": 1.910592437492302e-06, + "loss": 0.2799, + "step": 29533 + }, + { + "epoch": 10.91, + "learning_rate": 1.9105554871289565e-06, + "loss": 0.2445, + "step": 29534 + }, + { + "epoch": 10.91, + "learning_rate": 1.9105185367656116e-06, + "loss": 0.3839, + "step": 29535 + }, + { + "epoch": 10.91, + "learning_rate": 1.910481586402266e-06, + "loss": 0.3909, + "step": 29536 + }, + { + "epoch": 10.91, + "learning_rate": 1.9104446360389213e-06, + "loss": 0.3011, + "step": 29537 + }, + { + "epoch": 10.91, + "learning_rate": 1.910407685675576e-06, + "loss": 0.3289, + "step": 29538 + }, + { + "epoch": 10.91, + "learning_rate": 1.9103707353122305e-06, + "loss": 0.4076, + "step": 29539 + }, + { + "epoch": 10.91, + "learning_rate": 1.910333784948885e-06, + "loss": 0.2776, + "step": 29540 + }, + { + "epoch": 10.91, + "learning_rate": 1.9102968345855402e-06, + "loss": 0.3631, + "step": 29541 + }, + { + "epoch": 10.91, + "learning_rate": 1.910259884222195e-06, + "loss": 0.2581, + "step": 29542 + }, + { + "epoch": 10.91, + "learning_rate": 1.9102229338588495e-06, + "loss": 0.3682, + "step": 29543 + }, + { + "epoch": 10.91, + "learning_rate": 1.9101859834955046e-06, + "loss": 0.3063, + "step": 29544 + }, + { + "epoch": 10.91, + "learning_rate": 1.910149033132159e-06, + "loss": 0.3365, + "step": 29545 + }, + { + "epoch": 10.91, + "learning_rate": 1.910112082768814e-06, + "loss": 0.2523, + "step": 29546 + }, + { + "epoch": 10.91, + "learning_rate": 1.9100751324054685e-06, + "loss": 0.3328, + "step": 29547 + }, + { + "epoch": 10.91, + "learning_rate": 1.9100381820421235e-06, + "loss": 0.4273, + "step": 29548 + }, + { + "epoch": 10.91, + "learning_rate": 1.910001231678778e-06, + "loss": 0.4186, + "step": 29549 + }, + { + "epoch": 10.91, + "learning_rate": 1.9099642813154332e-06, + "loss": 0.4452, + "step": 29550 + }, + { + "epoch": 10.91, + "learning_rate": 1.909927330952088e-06, + "loss": 0.2485, + "step": 29551 + }, + { + "epoch": 10.91, + "learning_rate": 1.9098903805887425e-06, + "loss": 0.2632, + "step": 29552 + }, + { + "epoch": 10.91, + "learning_rate": 1.909853430225397e-06, + "loss": 0.3481, + "step": 29553 + }, + { + "epoch": 10.91, + "learning_rate": 1.9098164798620518e-06, + "loss": 0.3515, + "step": 29554 + }, + { + "epoch": 10.91, + "learning_rate": 1.909779529498707e-06, + "loss": 0.3807, + "step": 29555 + }, + { + "epoch": 10.91, + "learning_rate": 1.9097425791353615e-06, + "loss": 0.2488, + "step": 29556 + }, + { + "epoch": 10.91, + "learning_rate": 1.9097056287720165e-06, + "loss": 0.3555, + "step": 29557 + }, + { + "epoch": 10.92, + "learning_rate": 1.909668678408671e-06, + "loss": 0.4487, + "step": 29558 + }, + { + "epoch": 10.92, + "learning_rate": 1.909631728045326e-06, + "loss": 0.3324, + "step": 29559 + }, + { + "epoch": 10.92, + "learning_rate": 1.9095947776819804e-06, + "loss": 0.2676, + "step": 29560 + }, + { + "epoch": 10.92, + "learning_rate": 1.9095578273186355e-06, + "loss": 0.2328, + "step": 29561 + }, + { + "epoch": 10.92, + "learning_rate": 1.90952087695529e-06, + "loss": 0.2801, + "step": 29562 + }, + { + "epoch": 10.92, + "learning_rate": 1.9094839265919448e-06, + "loss": 0.2578, + "step": 29563 + }, + { + "epoch": 10.92, + "learning_rate": 1.9094469762286e-06, + "loss": 0.4734, + "step": 29564 + }, + { + "epoch": 10.92, + "learning_rate": 1.9094100258652545e-06, + "loss": 0.1888, + "step": 29565 + }, + { + "epoch": 10.92, + "learning_rate": 1.909373075501909e-06, + "loss": 0.3743, + "step": 29566 + }, + { + "epoch": 10.92, + "learning_rate": 1.9093361251385637e-06, + "loss": 0.2465, + "step": 29567 + }, + { + "epoch": 10.92, + "learning_rate": 1.909299174775219e-06, + "loss": 0.2594, + "step": 29568 + }, + { + "epoch": 10.92, + "learning_rate": 1.9092622244118734e-06, + "loss": 0.4931, + "step": 29569 + }, + { + "epoch": 10.92, + "learning_rate": 1.9092252740485285e-06, + "loss": 0.3504, + "step": 29570 + }, + { + "epoch": 10.92, + "learning_rate": 1.909188323685183e-06, + "loss": 0.4704, + "step": 29571 + }, + { + "epoch": 10.92, + "learning_rate": 1.9091513733218378e-06, + "loss": 0.2967, + "step": 29572 + }, + { + "epoch": 10.92, + "learning_rate": 1.9091144229584924e-06, + "loss": 0.2849, + "step": 29573 + }, + { + "epoch": 10.92, + "learning_rate": 1.909077472595147e-06, + "loss": 0.3533, + "step": 29574 + }, + { + "epoch": 10.92, + "learning_rate": 1.909040522231802e-06, + "loss": 0.3473, + "step": 29575 + }, + { + "epoch": 10.92, + "learning_rate": 1.9090035718684567e-06, + "loss": 0.3924, + "step": 29576 + }, + { + "epoch": 10.92, + "learning_rate": 1.908966621505112e-06, + "loss": 0.3256, + "step": 29577 + }, + { + "epoch": 10.92, + "learning_rate": 1.908929671141766e-06, + "loss": 0.3824, + "step": 29578 + }, + { + "epoch": 10.92, + "learning_rate": 1.908892720778421e-06, + "loss": 0.3563, + "step": 29579 + }, + { + "epoch": 10.92, + "learning_rate": 1.9088557704150757e-06, + "loss": 0.2873, + "step": 29580 + }, + { + "epoch": 10.92, + "learning_rate": 1.9088188200517308e-06, + "loss": 0.2965, + "step": 29581 + }, + { + "epoch": 10.92, + "learning_rate": 1.9087818696883854e-06, + "loss": 0.2441, + "step": 29582 + }, + { + "epoch": 10.92, + "learning_rate": 1.90874491932504e-06, + "loss": 0.3356, + "step": 29583 + }, + { + "epoch": 10.92, + "learning_rate": 1.9087079689616947e-06, + "loss": 0.2788, + "step": 29584 + }, + { + "epoch": 10.93, + "learning_rate": 1.9086710185983493e-06, + "loss": 0.2986, + "step": 29585 + }, + { + "epoch": 10.93, + "learning_rate": 1.9086340682350044e-06, + "loss": 0.2131, + "step": 29586 + }, + { + "epoch": 10.93, + "learning_rate": 1.908597117871659e-06, + "loss": 0.3448, + "step": 29587 + }, + { + "epoch": 10.93, + "learning_rate": 1.908560167508314e-06, + "loss": 0.4133, + "step": 29588 + }, + { + "epoch": 10.93, + "learning_rate": 1.9085232171449687e-06, + "loss": 0.3146, + "step": 29589 + }, + { + "epoch": 10.93, + "learning_rate": 1.9084862667816233e-06, + "loss": 0.3827, + "step": 29590 + }, + { + "epoch": 10.93, + "learning_rate": 1.908449316418278e-06, + "loss": 0.3393, + "step": 29591 + }, + { + "epoch": 10.93, + "learning_rate": 1.908412366054933e-06, + "loss": 0.4456, + "step": 29592 + }, + { + "epoch": 10.93, + "learning_rate": 1.9083754156915877e-06, + "loss": 0.344, + "step": 29593 + }, + { + "epoch": 10.93, + "learning_rate": 1.9083384653282423e-06, + "loss": 0.3196, + "step": 29594 + }, + { + "epoch": 10.93, + "learning_rate": 1.9083015149648974e-06, + "loss": 0.4005, + "step": 29595 + }, + { + "epoch": 10.93, + "learning_rate": 1.908264564601552e-06, + "loss": 0.2499, + "step": 29596 + }, + { + "epoch": 10.93, + "learning_rate": 1.9082276142382066e-06, + "loss": 0.4219, + "step": 29597 + }, + { + "epoch": 10.93, + "learning_rate": 1.9081906638748613e-06, + "loss": 0.308, + "step": 29598 + }, + { + "epoch": 10.93, + "learning_rate": 1.9081537135115163e-06, + "loss": 0.2976, + "step": 29599 + }, + { + "epoch": 10.93, + "learning_rate": 1.908116763148171e-06, + "loss": 0.3667, + "step": 29600 + }, + { + "epoch": 10.93, + "learning_rate": 1.908079812784826e-06, + "loss": 0.2883, + "step": 29601 + }, + { + "epoch": 10.93, + "learning_rate": 1.9080428624214807e-06, + "loss": 0.3787, + "step": 29602 + }, + { + "epoch": 10.93, + "learning_rate": 1.9080059120581353e-06, + "loss": 0.313, + "step": 29603 + }, + { + "epoch": 10.93, + "learning_rate": 1.90796896169479e-06, + "loss": 0.5114, + "step": 29604 + }, + { + "epoch": 10.93, + "learning_rate": 1.9079320113314446e-06, + "loss": 0.3441, + "step": 29605 + }, + { + "epoch": 10.93, + "learning_rate": 1.9078950609680996e-06, + "loss": 0.4197, + "step": 29606 + }, + { + "epoch": 10.93, + "learning_rate": 1.9078581106047543e-06, + "loss": 0.3238, + "step": 29607 + }, + { + "epoch": 10.93, + "learning_rate": 1.9078211602414093e-06, + "loss": 0.3039, + "step": 29608 + }, + { + "epoch": 10.93, + "learning_rate": 1.907784209878064e-06, + "loss": 0.2788, + "step": 29609 + }, + { + "epoch": 10.93, + "learning_rate": 1.9077472595147186e-06, + "loss": 0.2808, + "step": 29610 + }, + { + "epoch": 10.93, + "learning_rate": 1.9077103091513732e-06, + "loss": 0.3789, + "step": 29611 + }, + { + "epoch": 10.94, + "learning_rate": 1.9076733587880283e-06, + "loss": 0.4743, + "step": 29612 + }, + { + "epoch": 10.94, + "learning_rate": 1.907636408424683e-06, + "loss": 0.3008, + "step": 29613 + }, + { + "epoch": 10.94, + "learning_rate": 1.9075994580613376e-06, + "loss": 0.3744, + "step": 29614 + }, + { + "epoch": 10.94, + "learning_rate": 1.9075625076979926e-06, + "loss": 0.3016, + "step": 29615 + }, + { + "epoch": 10.94, + "learning_rate": 1.9075255573346473e-06, + "loss": 0.3649, + "step": 29616 + }, + { + "epoch": 10.94, + "learning_rate": 1.907488606971302e-06, + "loss": 0.3363, + "step": 29617 + }, + { + "epoch": 10.94, + "learning_rate": 1.9074516566079565e-06, + "loss": 0.1992, + "step": 29618 + }, + { + "epoch": 10.94, + "learning_rate": 1.9074147062446116e-06, + "loss": 0.2365, + "step": 29619 + }, + { + "epoch": 10.94, + "learning_rate": 1.9073777558812662e-06, + "loss": 0.3985, + "step": 29620 + }, + { + "epoch": 10.94, + "learning_rate": 1.907340805517921e-06, + "loss": 0.3499, + "step": 29621 + }, + { + "epoch": 10.94, + "learning_rate": 1.9073038551545757e-06, + "loss": 0.2578, + "step": 29622 + }, + { + "epoch": 10.94, + "learning_rate": 1.9072669047912304e-06, + "loss": 0.3559, + "step": 29623 + }, + { + "epoch": 10.94, + "learning_rate": 1.9072299544278852e-06, + "loss": 0.4248, + "step": 29624 + }, + { + "epoch": 10.94, + "learning_rate": 1.9071930040645398e-06, + "loss": 0.2896, + "step": 29625 + }, + { + "epoch": 10.94, + "learning_rate": 1.907156053701195e-06, + "loss": 0.3769, + "step": 29626 + }, + { + "epoch": 10.94, + "learning_rate": 1.9071191033378495e-06, + "loss": 0.3082, + "step": 29627 + }, + { + "epoch": 10.94, + "learning_rate": 1.9070821529745044e-06, + "loss": 0.4812, + "step": 29628 + }, + { + "epoch": 10.94, + "learning_rate": 1.907045202611159e-06, + "loss": 0.3013, + "step": 29629 + }, + { + "epoch": 10.94, + "learning_rate": 1.9070082522478139e-06, + "loss": 0.3314, + "step": 29630 + }, + { + "epoch": 10.94, + "learning_rate": 1.9069713018844685e-06, + "loss": 0.4189, + "step": 29631 + }, + { + "epoch": 10.94, + "learning_rate": 1.9069343515211236e-06, + "loss": 0.3064, + "step": 29632 + }, + { + "epoch": 10.94, + "learning_rate": 1.9068974011577782e-06, + "loss": 0.369, + "step": 29633 + }, + { + "epoch": 10.94, + "learning_rate": 1.9068604507944326e-06, + "loss": 0.2783, + "step": 29634 + }, + { + "epoch": 10.94, + "learning_rate": 1.9068235004310877e-06, + "loss": 0.3445, + "step": 29635 + }, + { + "epoch": 10.94, + "learning_rate": 1.9067865500677423e-06, + "loss": 0.3428, + "step": 29636 + }, + { + "epoch": 10.94, + "learning_rate": 1.9067495997043972e-06, + "loss": 0.2707, + "step": 29637 + }, + { + "epoch": 10.94, + "learning_rate": 1.9067126493410518e-06, + "loss": 0.4526, + "step": 29638 + }, + { + "epoch": 10.94, + "learning_rate": 1.9066756989777069e-06, + "loss": 0.2707, + "step": 29639 + }, + { + "epoch": 10.95, + "learning_rate": 1.9066387486143613e-06, + "loss": 0.3645, + "step": 29640 + }, + { + "epoch": 10.95, + "learning_rate": 1.9066017982510164e-06, + "loss": 0.3995, + "step": 29641 + }, + { + "epoch": 10.95, + "learning_rate": 1.906564847887671e-06, + "loss": 0.2965, + "step": 29642 + }, + { + "epoch": 10.95, + "learning_rate": 1.9065278975243256e-06, + "loss": 0.2578, + "step": 29643 + }, + { + "epoch": 10.95, + "learning_rate": 1.9064909471609805e-06, + "loss": 0.388, + "step": 29644 + }, + { + "epoch": 10.95, + "learning_rate": 1.9064539967976351e-06, + "loss": 0.2804, + "step": 29645 + }, + { + "epoch": 10.95, + "learning_rate": 1.90641704643429e-06, + "loss": 0.4259, + "step": 29646 + }, + { + "epoch": 10.95, + "learning_rate": 1.9063800960709446e-06, + "loss": 0.3515, + "step": 29647 + }, + { + "epoch": 10.95, + "learning_rate": 1.9063431457075997e-06, + "loss": 0.269, + "step": 29648 + }, + { + "epoch": 10.95, + "learning_rate": 1.9063061953442543e-06, + "loss": 0.3507, + "step": 29649 + }, + { + "epoch": 10.95, + "learning_rate": 1.9062692449809091e-06, + "loss": 0.2523, + "step": 29650 + }, + { + "epoch": 10.95, + "learning_rate": 1.9062322946175638e-06, + "loss": 0.405, + "step": 29651 + }, + { + "epoch": 10.95, + "learning_rate": 1.9061953442542186e-06, + "loss": 0.4829, + "step": 29652 + }, + { + "epoch": 10.95, + "learning_rate": 1.9061583938908733e-06, + "loss": 0.4366, + "step": 29653 + }, + { + "epoch": 10.95, + "learning_rate": 1.906121443527528e-06, + "loss": 0.4091, + "step": 29654 + }, + { + "epoch": 10.95, + "learning_rate": 1.906084493164183e-06, + "loss": 0.2365, + "step": 29655 + }, + { + "epoch": 10.95, + "learning_rate": 1.9060475428008374e-06, + "loss": 0.184, + "step": 29656 + }, + { + "epoch": 10.95, + "learning_rate": 1.9060105924374924e-06, + "loss": 0.2495, + "step": 29657 + }, + { + "epoch": 10.95, + "learning_rate": 1.905973642074147e-06, + "loss": 0.3634, + "step": 29658 + }, + { + "epoch": 10.95, + "learning_rate": 1.905936691710802e-06, + "loss": 0.3372, + "step": 29659 + }, + { + "epoch": 10.95, + "learning_rate": 1.9058997413474566e-06, + "loss": 0.3923, + "step": 29660 + }, + { + "epoch": 10.95, + "learning_rate": 1.9058627909841116e-06, + "loss": 0.2895, + "step": 29661 + }, + { + "epoch": 10.95, + "learning_rate": 1.905825840620766e-06, + "loss": 0.2957, + "step": 29662 + }, + { + "epoch": 10.95, + "learning_rate": 1.9057888902574207e-06, + "loss": 0.3428, + "step": 29663 + }, + { + "epoch": 10.95, + "learning_rate": 1.9057519398940757e-06, + "loss": 0.3125, + "step": 29664 + }, + { + "epoch": 10.95, + "learning_rate": 1.9057149895307304e-06, + "loss": 0.3712, + "step": 29665 + }, + { + "epoch": 10.95, + "learning_rate": 1.9056780391673852e-06, + "loss": 0.3942, + "step": 29666 + }, + { + "epoch": 10.96, + "learning_rate": 1.9056410888040399e-06, + "loss": 0.2085, + "step": 29667 + }, + { + "epoch": 10.96, + "learning_rate": 1.9056041384406947e-06, + "loss": 0.3982, + "step": 29668 + }, + { + "epoch": 10.96, + "learning_rate": 1.9055671880773494e-06, + "loss": 0.3301, + "step": 29669 + }, + { + "epoch": 10.96, + "learning_rate": 1.9055302377140044e-06, + "loss": 0.3468, + "step": 29670 + }, + { + "epoch": 10.96, + "learning_rate": 1.905493287350659e-06, + "loss": 0.3404, + "step": 29671 + }, + { + "epoch": 10.96, + "learning_rate": 1.905456336987314e-06, + "loss": 0.3685, + "step": 29672 + }, + { + "epoch": 10.96, + "learning_rate": 1.9054193866239685e-06, + "loss": 0.4013, + "step": 29673 + }, + { + "epoch": 10.96, + "learning_rate": 1.9053824362606232e-06, + "loss": 0.2632, + "step": 29674 + }, + { + "epoch": 10.96, + "learning_rate": 1.905345485897278e-06, + "loss": 0.372, + "step": 29675 + }, + { + "epoch": 10.96, + "learning_rate": 1.9053085355339327e-06, + "loss": 0.2504, + "step": 29676 + }, + { + "epoch": 10.96, + "learning_rate": 1.9052715851705877e-06, + "loss": 0.3926, + "step": 29677 + }, + { + "epoch": 10.96, + "learning_rate": 1.9052346348072424e-06, + "loss": 0.399, + "step": 29678 + }, + { + "epoch": 10.96, + "learning_rate": 1.9051976844438972e-06, + "loss": 0.3174, + "step": 29679 + }, + { + "epoch": 10.96, + "learning_rate": 1.9051607340805518e-06, + "loss": 0.4899, + "step": 29680 + }, + { + "epoch": 10.96, + "learning_rate": 1.9051237837172067e-06, + "loss": 0.3509, + "step": 29681 + }, + { + "epoch": 10.96, + "learning_rate": 1.9050868333538613e-06, + "loss": 0.2997, + "step": 29682 + }, + { + "epoch": 10.96, + "learning_rate": 1.905049882990516e-06, + "loss": 0.3229, + "step": 29683 + }, + { + "epoch": 10.96, + "learning_rate": 1.9050129326271708e-06, + "loss": 0.4286, + "step": 29684 + }, + { + "epoch": 10.96, + "learning_rate": 1.9049759822638254e-06, + "loss": 0.3691, + "step": 29685 + }, + { + "epoch": 10.96, + "learning_rate": 1.9049390319004805e-06, + "loss": 0.2426, + "step": 29686 + }, + { + "epoch": 10.96, + "learning_rate": 1.9049020815371351e-06, + "loss": 0.2704, + "step": 29687 + }, + { + "epoch": 10.96, + "learning_rate": 1.90486513117379e-06, + "loss": 0.3529, + "step": 29688 + }, + { + "epoch": 10.96, + "learning_rate": 1.9048281808104446e-06, + "loss": 0.4045, + "step": 29689 + }, + { + "epoch": 10.96, + "learning_rate": 1.9047912304470995e-06, + "loss": 0.3553, + "step": 29690 + }, + { + "epoch": 10.96, + "learning_rate": 1.9047542800837541e-06, + "loss": 0.2185, + "step": 29691 + }, + { + "epoch": 10.96, + "learning_rate": 1.9047173297204092e-06, + "loss": 0.2088, + "step": 29692 + }, + { + "epoch": 10.96, + "learning_rate": 1.9046803793570638e-06, + "loss": 0.2509, + "step": 29693 + }, + { + "epoch": 10.97, + "learning_rate": 1.9046434289937184e-06, + "loss": 0.2019, + "step": 29694 + }, + { + "epoch": 10.97, + "learning_rate": 1.9046064786303733e-06, + "loss": 0.4267, + "step": 29695 + }, + { + "epoch": 10.97, + "learning_rate": 1.904569528267028e-06, + "loss": 0.2853, + "step": 29696 + }, + { + "epoch": 10.97, + "learning_rate": 1.9045325779036828e-06, + "loss": 0.2529, + "step": 29697 + }, + { + "epoch": 10.97, + "learning_rate": 1.9044956275403374e-06, + "loss": 0.3134, + "step": 29698 + }, + { + "epoch": 10.97, + "learning_rate": 1.9044586771769925e-06, + "loss": 0.3501, + "step": 29699 + }, + { + "epoch": 10.97, + "learning_rate": 1.9044217268136471e-06, + "loss": 0.5052, + "step": 29700 + }, + { + "epoch": 10.97, + "learning_rate": 1.904384776450302e-06, + "loss": 0.2616, + "step": 29701 + }, + { + "epoch": 10.97, + "learning_rate": 1.9043478260869566e-06, + "loss": 0.3082, + "step": 29702 + }, + { + "epoch": 10.97, + "learning_rate": 1.9043108757236112e-06, + "loss": 0.3393, + "step": 29703 + }, + { + "epoch": 10.97, + "learning_rate": 1.904273925360266e-06, + "loss": 0.3011, + "step": 29704 + }, + { + "epoch": 10.97, + "learning_rate": 1.9042369749969207e-06, + "loss": 0.2804, + "step": 29705 + }, + { + "epoch": 10.97, + "learning_rate": 1.9042000246335758e-06, + "loss": 0.2357, + "step": 29706 + }, + { + "epoch": 10.97, + "learning_rate": 1.9041630742702302e-06, + "loss": 0.3215, + "step": 29707 + }, + { + "epoch": 10.97, + "learning_rate": 1.9041261239068853e-06, + "loss": 0.4287, + "step": 29708 + }, + { + "epoch": 10.97, + "learning_rate": 1.90408917354354e-06, + "loss": 0.3065, + "step": 29709 + }, + { + "epoch": 10.97, + "learning_rate": 1.9040522231801947e-06, + "loss": 0.263, + "step": 29710 + }, + { + "epoch": 10.97, + "learning_rate": 1.9040152728168494e-06, + "loss": 0.3168, + "step": 29711 + }, + { + "epoch": 10.97, + "learning_rate": 1.9039783224535044e-06, + "loss": 0.3079, + "step": 29712 + }, + { + "epoch": 10.97, + "learning_rate": 1.9039413720901589e-06, + "loss": 0.3453, + "step": 29713 + }, + { + "epoch": 10.97, + "learning_rate": 1.9039044217268135e-06, + "loss": 0.3522, + "step": 29714 + }, + { + "epoch": 10.97, + "learning_rate": 1.9038674713634686e-06, + "loss": 0.3566, + "step": 29715 + }, + { + "epoch": 10.97, + "learning_rate": 1.9038305210001232e-06, + "loss": 0.2893, + "step": 29716 + }, + { + "epoch": 10.97, + "learning_rate": 1.903793570636778e-06, + "loss": 0.2774, + "step": 29717 + }, + { + "epoch": 10.97, + "learning_rate": 1.9037566202734327e-06, + "loss": 0.3643, + "step": 29718 + }, + { + "epoch": 10.97, + "learning_rate": 1.9037196699100875e-06, + "loss": 0.3249, + "step": 29719 + }, + { + "epoch": 10.97, + "learning_rate": 1.9036827195467422e-06, + "loss": 0.4551, + "step": 29720 + }, + { + "epoch": 10.98, + "learning_rate": 1.9036457691833972e-06, + "loss": 0.3968, + "step": 29721 + }, + { + "epoch": 10.98, + "learning_rate": 1.9036088188200519e-06, + "loss": 0.2898, + "step": 29722 + }, + { + "epoch": 10.98, + "learning_rate": 1.9035718684567063e-06, + "loss": 0.3105, + "step": 29723 + }, + { + "epoch": 10.98, + "learning_rate": 1.9035349180933613e-06, + "loss": 0.4868, + "step": 29724 + }, + { + "epoch": 10.98, + "learning_rate": 1.903497967730016e-06, + "loss": 0.335, + "step": 29725 + }, + { + "epoch": 10.98, + "learning_rate": 1.9034610173666708e-06, + "loss": 0.2725, + "step": 29726 + }, + { + "epoch": 10.98, + "learning_rate": 1.9034240670033255e-06, + "loss": 0.3989, + "step": 29727 + }, + { + "epoch": 10.98, + "learning_rate": 1.9033871166399805e-06, + "loss": 0.2224, + "step": 29728 + }, + { + "epoch": 10.98, + "learning_rate": 1.903350166276635e-06, + "loss": 0.4154, + "step": 29729 + }, + { + "epoch": 10.98, + "learning_rate": 1.90331321591329e-06, + "loss": 0.2254, + "step": 29730 + }, + { + "epoch": 10.98, + "learning_rate": 1.9032762655499446e-06, + "loss": 0.3244, + "step": 29731 + }, + { + "epoch": 10.98, + "learning_rate": 1.9032393151865995e-06, + "loss": 0.377, + "step": 29732 + }, + { + "epoch": 10.98, + "learning_rate": 1.9032023648232541e-06, + "loss": 0.337, + "step": 29733 + }, + { + "epoch": 10.98, + "learning_rate": 1.9031654144599088e-06, + "loss": 0.3003, + "step": 29734 + }, + { + "epoch": 10.98, + "learning_rate": 1.9031284640965636e-06, + "loss": 0.2316, + "step": 29735 + }, + { + "epoch": 10.98, + "learning_rate": 1.9030915137332183e-06, + "loss": 0.3703, + "step": 29736 + }, + { + "epoch": 10.98, + "learning_rate": 1.9030545633698733e-06, + "loss": 0.2572, + "step": 29737 + }, + { + "epoch": 10.98, + "learning_rate": 1.903017613006528e-06, + "loss": 0.3048, + "step": 29738 + }, + { + "epoch": 10.98, + "learning_rate": 1.9029806626431828e-06, + "loss": 0.3798, + "step": 29739 + }, + { + "epoch": 10.98, + "learning_rate": 1.9029437122798374e-06, + "loss": 0.2445, + "step": 29740 + }, + { + "epoch": 10.98, + "learning_rate": 1.9029067619164923e-06, + "loss": 0.3575, + "step": 29741 + }, + { + "epoch": 10.98, + "learning_rate": 1.902869811553147e-06, + "loss": 0.3435, + "step": 29742 + }, + { + "epoch": 10.98, + "learning_rate": 1.9028328611898016e-06, + "loss": 0.4847, + "step": 29743 + }, + { + "epoch": 10.98, + "learning_rate": 1.9027959108264566e-06, + "loss": 0.2535, + "step": 29744 + }, + { + "epoch": 10.98, + "learning_rate": 1.9027589604631113e-06, + "loss": 0.3754, + "step": 29745 + }, + { + "epoch": 10.98, + "learning_rate": 1.902722010099766e-06, + "loss": 0.2691, + "step": 29746 + }, + { + "epoch": 10.98, + "learning_rate": 1.9026850597364207e-06, + "loss": 0.2944, + "step": 29747 + }, + { + "epoch": 10.99, + "learning_rate": 1.9026481093730756e-06, + "loss": 0.2398, + "step": 29748 + }, + { + "epoch": 10.99, + "learning_rate": 1.9026111590097302e-06, + "loss": 0.3173, + "step": 29749 + }, + { + "epoch": 10.99, + "learning_rate": 1.9025742086463853e-06, + "loss": 0.3059, + "step": 29750 + }, + { + "epoch": 10.99, + "learning_rate": 1.90253725828304e-06, + "loss": 0.3941, + "step": 29751 + }, + { + "epoch": 10.99, + "learning_rate": 1.9025003079196943e-06, + "loss": 0.3448, + "step": 29752 + }, + { + "epoch": 10.99, + "learning_rate": 1.9024633575563494e-06, + "loss": 0.2791, + "step": 29753 + }, + { + "epoch": 10.99, + "learning_rate": 1.902426407193004e-06, + "loss": 0.2309, + "step": 29754 + }, + { + "epoch": 10.99, + "learning_rate": 1.9023894568296589e-06, + "loss": 0.4063, + "step": 29755 + }, + { + "epoch": 10.99, + "learning_rate": 1.9023525064663135e-06, + "loss": 0.3919, + "step": 29756 + }, + { + "epoch": 10.99, + "learning_rate": 1.9023155561029686e-06, + "loss": 0.336, + "step": 29757 + }, + { + "epoch": 10.99, + "learning_rate": 1.902278605739623e-06, + "loss": 0.3284, + "step": 29758 + }, + { + "epoch": 10.99, + "learning_rate": 1.902241655376278e-06, + "loss": 0.3489, + "step": 29759 + }, + { + "epoch": 10.99, + "learning_rate": 1.9022047050129327e-06, + "loss": 0.3735, + "step": 29760 + }, + { + "epoch": 10.99, + "learning_rate": 1.9021677546495876e-06, + "loss": 0.361, + "step": 29761 + }, + { + "epoch": 10.99, + "learning_rate": 1.9021308042862422e-06, + "loss": 0.28, + "step": 29762 + }, + { + "epoch": 10.99, + "learning_rate": 1.9020938539228968e-06, + "loss": 0.2495, + "step": 29763 + }, + { + "epoch": 10.99, + "learning_rate": 1.9020569035595517e-06, + "loss": 0.2564, + "step": 29764 + }, + { + "epoch": 10.99, + "learning_rate": 1.9020199531962063e-06, + "loss": 0.2209, + "step": 29765 + }, + { + "epoch": 10.99, + "learning_rate": 1.9019830028328614e-06, + "loss": 0.2792, + "step": 29766 + }, + { + "epoch": 10.99, + "learning_rate": 1.901946052469516e-06, + "loss": 0.3716, + "step": 29767 + }, + { + "epoch": 10.99, + "learning_rate": 1.9019091021061709e-06, + "loss": 0.4787, + "step": 29768 + }, + { + "epoch": 10.99, + "learning_rate": 1.9018721517428255e-06, + "loss": 0.3005, + "step": 29769 + }, + { + "epoch": 10.99, + "learning_rate": 1.9018352013794803e-06, + "loss": 0.3221, + "step": 29770 + }, + { + "epoch": 10.99, + "learning_rate": 1.901798251016135e-06, + "loss": 0.4882, + "step": 29771 + }, + { + "epoch": 10.99, + "learning_rate": 1.9017613006527896e-06, + "loss": 0.4749, + "step": 29772 + }, + { + "epoch": 10.99, + "learning_rate": 1.9017243502894447e-06, + "loss": 0.303, + "step": 29773 + }, + { + "epoch": 10.99, + "learning_rate": 1.901687399926099e-06, + "loss": 0.25, + "step": 29774 + }, + { + "epoch": 11.0, + "learning_rate": 1.9016504495627542e-06, + "loss": 0.3339, + "step": 29775 + }, + { + "epoch": 11.0, + "learning_rate": 1.9016134991994088e-06, + "loss": 0.326, + "step": 29776 + }, + { + "epoch": 11.0, + "learning_rate": 1.9015765488360636e-06, + "loss": 0.3785, + "step": 29777 + }, + { + "epoch": 11.0, + "learning_rate": 1.9015395984727183e-06, + "loss": 0.257, + "step": 29778 + }, + { + "epoch": 11.0, + "learning_rate": 1.9015026481093733e-06, + "loss": 0.316, + "step": 29779 + }, + { + "epoch": 11.0, + "learning_rate": 1.9014656977460278e-06, + "loss": 0.3957, + "step": 29780 + }, + { + "epoch": 11.0, + "learning_rate": 1.9014287473826828e-06, + "loss": 0.3855, + "step": 29781 + }, + { + "epoch": 11.0, + "learning_rate": 1.9013917970193375e-06, + "loss": 0.1748, + "step": 29782 + }, + { + "epoch": 11.0, + "learning_rate": 1.901354846655992e-06, + "loss": 0.3767, + "step": 29783 + }, + { + "epoch": 11.0, + "learning_rate": 1.901317896292647e-06, + "loss": 0.4659, + "step": 29784 + }, + { + "epoch": 11.0, + "learning_rate": 1.9012809459293016e-06, + "loss": 0.243, + "step": 29785 + }, + { + "epoch": 11.0, + "learning_rate": 1.9012439955659564e-06, + "loss": 0.3475, + "step": 29786 + }, + { + "epoch": 11.0, + "learning_rate": 1.901207045202611e-06, + "loss": 0.4331, + "step": 29787 + }, + { + "epoch": 11.0, + "learning_rate": 1.9011700948392661e-06, + "loss": 0.3377, + "step": 29788 + }, + { + "epoch": 11.0, + "eval_accuracy": 0.8253220667682505, + "eval_loss": 0.4101983308792114, + "eval_runtime": 25.1932, + "eval_samples_per_second": 1719.272, + "eval_steps_per_second": 26.872, + "step": 29788 + }, + { + "epoch": 11.0, + "learning_rate": 1.9011331444759208e-06, + "loss": 0.2996, + "step": 29789 + }, + { + "epoch": 11.0, + "learning_rate": 1.9010961941125756e-06, + "loss": 0.3893, + "step": 29790 + }, + { + "epoch": 11.0, + "learning_rate": 1.9010592437492302e-06, + "loss": 0.402, + "step": 29791 + }, + { + "epoch": 11.0, + "learning_rate": 1.9010222933858849e-06, + "loss": 0.2859, + "step": 29792 + }, + { + "epoch": 11.0, + "learning_rate": 1.9009853430225397e-06, + "loss": 0.2725, + "step": 29793 + }, + { + "epoch": 11.0, + "learning_rate": 1.9009483926591944e-06, + "loss": 0.2556, + "step": 29794 + }, + { + "epoch": 11.0, + "learning_rate": 1.9009114422958494e-06, + "loss": 0.3639, + "step": 29795 + }, + { + "epoch": 11.0, + "learning_rate": 1.9008744919325039e-06, + "loss": 0.2996, + "step": 29796 + }, + { + "epoch": 11.0, + "learning_rate": 1.900837541569159e-06, + "loss": 0.2949, + "step": 29797 + }, + { + "epoch": 11.0, + "learning_rate": 1.9008005912058135e-06, + "loss": 0.3592, + "step": 29798 + }, + { + "epoch": 11.0, + "learning_rate": 1.9007636408424684e-06, + "loss": 0.2361, + "step": 29799 + }, + { + "epoch": 11.0, + "learning_rate": 1.900726690479123e-06, + "loss": 0.2379, + "step": 29800 + }, + { + "epoch": 11.0, + "learning_rate": 1.900689740115778e-06, + "loss": 0.2715, + "step": 29801 + }, + { + "epoch": 11.01, + "learning_rate": 1.9006527897524325e-06, + "loss": 0.2033, + "step": 29802 + }, + { + "epoch": 11.01, + "learning_rate": 1.9006158393890872e-06, + "loss": 0.2689, + "step": 29803 + }, + { + "epoch": 11.01, + "learning_rate": 1.9005788890257422e-06, + "loss": 0.4168, + "step": 29804 + }, + { + "epoch": 11.01, + "learning_rate": 1.9005419386623969e-06, + "loss": 0.2269, + "step": 29805 + }, + { + "epoch": 11.01, + "learning_rate": 1.9005049882990517e-06, + "loss": 0.3122, + "step": 29806 + }, + { + "epoch": 11.01, + "learning_rate": 1.9004680379357063e-06, + "loss": 0.3243, + "step": 29807 + }, + { + "epoch": 11.01, + "learning_rate": 1.9004310875723612e-06, + "loss": 0.3301, + "step": 29808 + }, + { + "epoch": 11.01, + "learning_rate": 1.9003941372090158e-06, + "loss": 0.242, + "step": 29809 + }, + { + "epoch": 11.01, + "learning_rate": 1.9003571868456709e-06, + "loss": 0.3691, + "step": 29810 + }, + { + "epoch": 11.01, + "learning_rate": 1.9003202364823255e-06, + "loss": 0.4387, + "step": 29811 + }, + { + "epoch": 11.01, + "learning_rate": 1.9002832861189802e-06, + "loss": 0.4817, + "step": 29812 + }, + { + "epoch": 11.01, + "learning_rate": 1.900246335755635e-06, + "loss": 0.4603, + "step": 29813 + }, + { + "epoch": 11.01, + "learning_rate": 1.9002093853922896e-06, + "loss": 0.2615, + "step": 29814 + }, + { + "epoch": 11.01, + "learning_rate": 1.9001724350289445e-06, + "loss": 0.2661, + "step": 29815 + }, + { + "epoch": 11.01, + "learning_rate": 1.9001354846655991e-06, + "loss": 0.3252, + "step": 29816 + }, + { + "epoch": 11.01, + "learning_rate": 1.9000985343022542e-06, + "loss": 0.4454, + "step": 29817 + }, + { + "epoch": 11.01, + "learning_rate": 1.9000615839389088e-06, + "loss": 0.2762, + "step": 29818 + }, + { + "epoch": 11.01, + "learning_rate": 1.9000246335755637e-06, + "loss": 0.2915, + "step": 29819 + }, + { + "epoch": 11.01, + "learning_rate": 1.8999876832122183e-06, + "loss": 0.4684, + "step": 29820 + }, + { + "epoch": 11.01, + "learning_rate": 1.8999507328488732e-06, + "loss": 0.1257, + "step": 29821 + }, + { + "epoch": 11.01, + "learning_rate": 1.8999137824855278e-06, + "loss": 0.2269, + "step": 29822 + }, + { + "epoch": 11.01, + "learning_rate": 1.8998768321221824e-06, + "loss": 0.3377, + "step": 29823 + }, + { + "epoch": 11.01, + "learning_rate": 1.8998398817588375e-06, + "loss": 0.328, + "step": 29824 + }, + { + "epoch": 11.01, + "learning_rate": 1.899802931395492e-06, + "loss": 0.1826, + "step": 29825 + }, + { + "epoch": 11.01, + "learning_rate": 1.899765981032147e-06, + "loss": 0.2819, + "step": 29826 + }, + { + "epoch": 11.01, + "learning_rate": 1.8997290306688016e-06, + "loss": 0.312, + "step": 29827 + }, + { + "epoch": 11.01, + "learning_rate": 1.8996920803054565e-06, + "loss": 0.4702, + "step": 29828 + }, + { + "epoch": 11.02, + "learning_rate": 1.899655129942111e-06, + "loss": 0.2234, + "step": 29829 + }, + { + "epoch": 11.02, + "learning_rate": 1.8996181795787661e-06, + "loss": 0.374, + "step": 29830 + }, + { + "epoch": 11.02, + "learning_rate": 1.8995812292154206e-06, + "loss": 0.4022, + "step": 29831 + }, + { + "epoch": 11.02, + "learning_rate": 1.8995442788520752e-06, + "loss": 0.2997, + "step": 29832 + }, + { + "epoch": 11.02, + "learning_rate": 1.8995073284887303e-06, + "loss": 0.2925, + "step": 29833 + }, + { + "epoch": 11.02, + "learning_rate": 1.899470378125385e-06, + "loss": 0.4008, + "step": 29834 + }, + { + "epoch": 11.02, + "learning_rate": 1.8994334277620398e-06, + "loss": 0.2487, + "step": 29835 + }, + { + "epoch": 11.02, + "learning_rate": 1.8993964773986944e-06, + "loss": 0.3097, + "step": 29836 + }, + { + "epoch": 11.02, + "learning_rate": 1.8993595270353492e-06, + "loss": 0.385, + "step": 29837 + }, + { + "epoch": 11.02, + "learning_rate": 1.8993225766720039e-06, + "loss": 0.4102, + "step": 29838 + }, + { + "epoch": 11.02, + "learning_rate": 1.899285626308659e-06, + "loss": 0.329, + "step": 29839 + }, + { + "epoch": 11.02, + "learning_rate": 1.8992486759453136e-06, + "loss": 0.2565, + "step": 29840 + }, + { + "epoch": 11.02, + "learning_rate": 1.8992117255819684e-06, + "loss": 0.4209, + "step": 29841 + }, + { + "epoch": 11.02, + "learning_rate": 1.899174775218623e-06, + "loss": 0.2801, + "step": 29842 + }, + { + "epoch": 11.02, + "learning_rate": 1.8991378248552777e-06, + "loss": 0.2302, + "step": 29843 + }, + { + "epoch": 11.02, + "learning_rate": 1.8991008744919325e-06, + "loss": 0.225, + "step": 29844 + }, + { + "epoch": 11.02, + "learning_rate": 1.8990639241285872e-06, + "loss": 0.455, + "step": 29845 + }, + { + "epoch": 11.02, + "learning_rate": 1.8990269737652422e-06, + "loss": 0.3234, + "step": 29846 + }, + { + "epoch": 11.02, + "learning_rate": 1.8989900234018967e-06, + "loss": 0.4315, + "step": 29847 + }, + { + "epoch": 11.02, + "learning_rate": 1.8989530730385517e-06, + "loss": 0.3229, + "step": 29848 + }, + { + "epoch": 11.02, + "learning_rate": 1.8989161226752064e-06, + "loss": 0.262, + "step": 29849 + }, + { + "epoch": 11.02, + "learning_rate": 1.8988791723118612e-06, + "loss": 0.3727, + "step": 29850 + }, + { + "epoch": 11.02, + "learning_rate": 1.8988422219485158e-06, + "loss": 0.4396, + "step": 29851 + }, + { + "epoch": 11.02, + "learning_rate": 1.8988052715851705e-06, + "loss": 0.1556, + "step": 29852 + }, + { + "epoch": 11.02, + "learning_rate": 1.8987683212218253e-06, + "loss": 0.2959, + "step": 29853 + }, + { + "epoch": 11.02, + "learning_rate": 1.89873137085848e-06, + "loss": 0.2614, + "step": 29854 + }, + { + "epoch": 11.02, + "learning_rate": 1.898694420495135e-06, + "loss": 0.3101, + "step": 29855 + }, + { + "epoch": 11.03, + "learning_rate": 1.8986574701317897e-06, + "loss": 0.2234, + "step": 29856 + }, + { + "epoch": 11.03, + "learning_rate": 1.8986205197684445e-06, + "loss": 0.3343, + "step": 29857 + }, + { + "epoch": 11.03, + "learning_rate": 1.8985835694050991e-06, + "loss": 0.3833, + "step": 29858 + }, + { + "epoch": 11.03, + "learning_rate": 1.898546619041754e-06, + "loss": 0.3845, + "step": 29859 + }, + { + "epoch": 11.03, + "learning_rate": 1.8985096686784086e-06, + "loss": 0.3572, + "step": 29860 + }, + { + "epoch": 11.03, + "learning_rate": 1.8984727183150637e-06, + "loss": 0.2883, + "step": 29861 + }, + { + "epoch": 11.03, + "learning_rate": 1.8984357679517183e-06, + "loss": 0.3943, + "step": 29862 + }, + { + "epoch": 11.03, + "learning_rate": 1.898398817588373e-06, + "loss": 0.3472, + "step": 29863 + }, + { + "epoch": 11.03, + "learning_rate": 1.8983618672250278e-06, + "loss": 0.3679, + "step": 29864 + }, + { + "epoch": 11.03, + "learning_rate": 1.8983249168616824e-06, + "loss": 0.4283, + "step": 29865 + }, + { + "epoch": 11.03, + "learning_rate": 1.8982879664983373e-06, + "loss": 0.306, + "step": 29866 + }, + { + "epoch": 11.03, + "learning_rate": 1.898251016134992e-06, + "loss": 0.2816, + "step": 29867 + }, + { + "epoch": 11.03, + "learning_rate": 1.898214065771647e-06, + "loss": 0.3618, + "step": 29868 + }, + { + "epoch": 11.03, + "learning_rate": 1.8981771154083014e-06, + "loss": 0.2885, + "step": 29869 + }, + { + "epoch": 11.03, + "learning_rate": 1.8981401650449565e-06, + "loss": 0.2715, + "step": 29870 + }, + { + "epoch": 11.03, + "learning_rate": 1.8981032146816111e-06, + "loss": 0.2634, + "step": 29871 + }, + { + "epoch": 11.03, + "learning_rate": 1.8980662643182657e-06, + "loss": 0.3478, + "step": 29872 + }, + { + "epoch": 11.03, + "learning_rate": 1.8980293139549206e-06, + "loss": 0.3222, + "step": 29873 + }, + { + "epoch": 11.03, + "learning_rate": 1.8979923635915752e-06, + "loss": 0.2335, + "step": 29874 + }, + { + "epoch": 11.03, + "learning_rate": 1.89795541322823e-06, + "loss": 0.2223, + "step": 29875 + }, + { + "epoch": 11.03, + "learning_rate": 1.8979184628648847e-06, + "loss": 0.3192, + "step": 29876 + }, + { + "epoch": 11.03, + "learning_rate": 1.8978815125015398e-06, + "loss": 0.2921, + "step": 29877 + }, + { + "epoch": 11.03, + "learning_rate": 1.8978445621381944e-06, + "loss": 0.3036, + "step": 29878 + }, + { + "epoch": 11.03, + "learning_rate": 1.8978076117748493e-06, + "loss": 0.1615, + "step": 29879 + }, + { + "epoch": 11.03, + "learning_rate": 1.897770661411504e-06, + "loss": 0.221, + "step": 29880 + }, + { + "epoch": 11.03, + "learning_rate": 1.8977337110481587e-06, + "loss": 0.2837, + "step": 29881 + }, + { + "epoch": 11.03, + "learning_rate": 1.8976967606848134e-06, + "loss": 0.4563, + "step": 29882 + }, + { + "epoch": 11.04, + "learning_rate": 1.897659810321468e-06, + "loss": 0.3611, + "step": 29883 + }, + { + "epoch": 11.04, + "learning_rate": 1.897622859958123e-06, + "loss": 0.3102, + "step": 29884 + }, + { + "epoch": 11.04, + "learning_rate": 1.8975859095947777e-06, + "loss": 0.2926, + "step": 29885 + }, + { + "epoch": 11.04, + "learning_rate": 1.8975489592314326e-06, + "loss": 0.2762, + "step": 29886 + }, + { + "epoch": 11.04, + "learning_rate": 1.8975120088680872e-06, + "loss": 0.2493, + "step": 29887 + }, + { + "epoch": 11.04, + "learning_rate": 1.897475058504742e-06, + "loss": 0.3847, + "step": 29888 + }, + { + "epoch": 11.04, + "learning_rate": 1.8974381081413967e-06, + "loss": 0.1862, + "step": 29889 + }, + { + "epoch": 11.04, + "learning_rate": 1.8974011577780517e-06, + "loss": 0.3263, + "step": 29890 + }, + { + "epoch": 11.04, + "learning_rate": 1.8973642074147064e-06, + "loss": 0.375, + "step": 29891 + }, + { + "epoch": 11.04, + "learning_rate": 1.8973272570513608e-06, + "loss": 0.3049, + "step": 29892 + }, + { + "epoch": 11.04, + "learning_rate": 1.8972903066880159e-06, + "loss": 0.2594, + "step": 29893 + }, + { + "epoch": 11.04, + "learning_rate": 1.8972533563246705e-06, + "loss": 0.3845, + "step": 29894 + }, + { + "epoch": 11.04, + "learning_rate": 1.8972164059613254e-06, + "loss": 0.3127, + "step": 29895 + }, + { + "epoch": 11.04, + "learning_rate": 1.89717945559798e-06, + "loss": 0.2539, + "step": 29896 + }, + { + "epoch": 11.04, + "learning_rate": 1.897142505234635e-06, + "loss": 0.4399, + "step": 29897 + }, + { + "epoch": 11.04, + "learning_rate": 1.8971055548712895e-06, + "loss": 0.2849, + "step": 29898 + }, + { + "epoch": 11.04, + "learning_rate": 1.8970686045079445e-06, + "loss": 0.2077, + "step": 29899 + }, + { + "epoch": 11.04, + "learning_rate": 1.8970316541445992e-06, + "loss": 0.3573, + "step": 29900 + }, + { + "epoch": 11.04, + "learning_rate": 1.896994703781254e-06, + "loss": 0.4848, + "step": 29901 + }, + { + "epoch": 11.04, + "learning_rate": 1.8969577534179087e-06, + "loss": 0.4189, + "step": 29902 + }, + { + "epoch": 11.04, + "learning_rate": 1.8969208030545633e-06, + "loss": 0.2172, + "step": 29903 + }, + { + "epoch": 11.04, + "learning_rate": 1.8968838526912181e-06, + "loss": 0.3009, + "step": 29904 + }, + { + "epoch": 11.04, + "learning_rate": 1.8968469023278728e-06, + "loss": 0.4085, + "step": 29905 + }, + { + "epoch": 11.04, + "learning_rate": 1.8968099519645278e-06, + "loss": 0.2501, + "step": 29906 + }, + { + "epoch": 11.04, + "learning_rate": 1.8967730016011825e-06, + "loss": 0.2812, + "step": 29907 + }, + { + "epoch": 11.04, + "learning_rate": 1.8967360512378373e-06, + "loss": 0.3662, + "step": 29908 + }, + { + "epoch": 11.04, + "learning_rate": 1.896699100874492e-06, + "loss": 0.3442, + "step": 29909 + }, + { + "epoch": 11.05, + "learning_rate": 1.8966621505111468e-06, + "loss": 0.3629, + "step": 29910 + }, + { + "epoch": 11.05, + "learning_rate": 1.8966252001478014e-06, + "loss": 0.2777, + "step": 29911 + }, + { + "epoch": 11.05, + "learning_rate": 1.896588249784456e-06, + "loss": 0.4097, + "step": 29912 + }, + { + "epoch": 11.05, + "learning_rate": 1.8965512994211111e-06, + "loss": 0.396, + "step": 29913 + }, + { + "epoch": 11.05, + "learning_rate": 1.8965143490577656e-06, + "loss": 0.252, + "step": 29914 + }, + { + "epoch": 11.05, + "learning_rate": 1.8964773986944206e-06, + "loss": 0.2959, + "step": 29915 + }, + { + "epoch": 11.05, + "learning_rate": 1.8964404483310753e-06, + "loss": 0.3486, + "step": 29916 + }, + { + "epoch": 11.05, + "learning_rate": 1.8964034979677301e-06, + "loss": 0.2636, + "step": 29917 + }, + { + "epoch": 11.05, + "learning_rate": 1.8963665476043847e-06, + "loss": 0.274, + "step": 29918 + }, + { + "epoch": 11.05, + "learning_rate": 1.8963295972410398e-06, + "loss": 0.297, + "step": 29919 + }, + { + "epoch": 11.05, + "learning_rate": 1.8962926468776942e-06, + "loss": 0.2424, + "step": 29920 + }, + { + "epoch": 11.05, + "learning_rate": 1.8962556965143493e-06, + "loss": 0.2854, + "step": 29921 + }, + { + "epoch": 11.05, + "learning_rate": 1.896218746151004e-06, + "loss": 0.3534, + "step": 29922 + }, + { + "epoch": 11.05, + "learning_rate": 1.8961817957876586e-06, + "loss": 0.2396, + "step": 29923 + }, + { + "epoch": 11.05, + "learning_rate": 1.8961448454243134e-06, + "loss": 0.3786, + "step": 29924 + }, + { + "epoch": 11.05, + "learning_rate": 1.896107895060968e-06, + "loss": 0.3059, + "step": 29925 + }, + { + "epoch": 11.05, + "learning_rate": 1.8960709446976229e-06, + "loss": 0.222, + "step": 29926 + }, + { + "epoch": 11.05, + "learning_rate": 1.8960339943342775e-06, + "loss": 0.3448, + "step": 29927 + }, + { + "epoch": 11.05, + "learning_rate": 1.8959970439709326e-06, + "loss": 0.2241, + "step": 29928 + }, + { + "epoch": 11.05, + "learning_rate": 1.8959600936075872e-06, + "loss": 0.375, + "step": 29929 + }, + { + "epoch": 11.05, + "learning_rate": 1.895923143244242e-06, + "loss": 0.3084, + "step": 29930 + }, + { + "epoch": 11.05, + "learning_rate": 1.8958861928808967e-06, + "loss": 0.2893, + "step": 29931 + }, + { + "epoch": 11.05, + "learning_rate": 1.8958492425175513e-06, + "loss": 0.3216, + "step": 29932 + }, + { + "epoch": 11.05, + "learning_rate": 1.8958122921542062e-06, + "loss": 0.3045, + "step": 29933 + }, + { + "epoch": 11.05, + "learning_rate": 1.8957753417908608e-06, + "loss": 0.2674, + "step": 29934 + }, + { + "epoch": 11.05, + "learning_rate": 1.8957383914275159e-06, + "loss": 0.3309, + "step": 29935 + }, + { + "epoch": 11.05, + "learning_rate": 1.8957014410641705e-06, + "loss": 0.2438, + "step": 29936 + }, + { + "epoch": 11.06, + "learning_rate": 1.8956644907008254e-06, + "loss": 0.386, + "step": 29937 + }, + { + "epoch": 11.06, + "learning_rate": 1.89562754033748e-06, + "loss": 0.3162, + "step": 29938 + }, + { + "epoch": 11.06, + "learning_rate": 1.8955905899741349e-06, + "loss": 0.3186, + "step": 29939 + }, + { + "epoch": 11.06, + "learning_rate": 1.8955536396107895e-06, + "loss": 0.3021, + "step": 29940 + }, + { + "epoch": 11.06, + "learning_rate": 1.8955166892474446e-06, + "loss": 0.3356, + "step": 29941 + }, + { + "epoch": 11.06, + "learning_rate": 1.8954797388840992e-06, + "loss": 0.23, + "step": 29942 + }, + { + "epoch": 11.06, + "learning_rate": 1.8954427885207536e-06, + "loss": 0.4234, + "step": 29943 + }, + { + "epoch": 11.06, + "learning_rate": 1.8954058381574087e-06, + "loss": 0.2477, + "step": 29944 + }, + { + "epoch": 11.06, + "learning_rate": 1.8953688877940633e-06, + "loss": 0.3131, + "step": 29945 + }, + { + "epoch": 11.06, + "learning_rate": 1.8953319374307182e-06, + "loss": 0.3191, + "step": 29946 + }, + { + "epoch": 11.06, + "learning_rate": 1.8952949870673728e-06, + "loss": 0.3715, + "step": 29947 + }, + { + "epoch": 11.06, + "learning_rate": 1.8952580367040276e-06, + "loss": 0.4036, + "step": 29948 + }, + { + "epoch": 11.06, + "learning_rate": 1.8952210863406823e-06, + "loss": 0.3029, + "step": 29949 + }, + { + "epoch": 11.06, + "learning_rate": 1.8951841359773373e-06, + "loss": 0.3613, + "step": 29950 + }, + { + "epoch": 11.06, + "learning_rate": 1.895147185613992e-06, + "loss": 0.4854, + "step": 29951 + }, + { + "epoch": 11.06, + "learning_rate": 1.8951102352506466e-06, + "loss": 0.3461, + "step": 29952 + }, + { + "epoch": 11.06, + "learning_rate": 1.8950732848873015e-06, + "loss": 0.3682, + "step": 29953 + }, + { + "epoch": 11.06, + "learning_rate": 1.895036334523956e-06, + "loss": 0.3276, + "step": 29954 + }, + { + "epoch": 11.06, + "learning_rate": 1.894999384160611e-06, + "loss": 0.2978, + "step": 29955 + }, + { + "epoch": 11.06, + "learning_rate": 1.8949624337972656e-06, + "loss": 0.362, + "step": 29956 + }, + { + "epoch": 11.06, + "learning_rate": 1.8949254834339206e-06, + "loss": 0.2985, + "step": 29957 + }, + { + "epoch": 11.06, + "learning_rate": 1.8948885330705753e-06, + "loss": 0.247, + "step": 29958 + }, + { + "epoch": 11.06, + "learning_rate": 1.8948515827072301e-06, + "loss": 0.3527, + "step": 29959 + }, + { + "epoch": 11.06, + "learning_rate": 1.8948146323438848e-06, + "loss": 0.4664, + "step": 29960 + }, + { + "epoch": 11.06, + "learning_rate": 1.8947776819805396e-06, + "loss": 0.3292, + "step": 29961 + }, + { + "epoch": 11.06, + "learning_rate": 1.8947407316171943e-06, + "loss": 0.4308, + "step": 29962 + }, + { + "epoch": 11.06, + "learning_rate": 1.8947037812538489e-06, + "loss": 0.4179, + "step": 29963 + }, + { + "epoch": 11.06, + "learning_rate": 1.894666830890504e-06, + "loss": 0.2969, + "step": 29964 + }, + { + "epoch": 11.07, + "learning_rate": 1.8946298805271584e-06, + "loss": 0.2605, + "step": 29965 + }, + { + "epoch": 11.07, + "learning_rate": 1.8945929301638134e-06, + "loss": 0.362, + "step": 29966 + }, + { + "epoch": 11.07, + "learning_rate": 1.894555979800468e-06, + "loss": 0.444, + "step": 29967 + }, + { + "epoch": 11.07, + "learning_rate": 1.894519029437123e-06, + "loss": 0.2832, + "step": 29968 + }, + { + "epoch": 11.07, + "learning_rate": 1.8944820790737776e-06, + "loss": 0.3464, + "step": 29969 + }, + { + "epoch": 11.07, + "learning_rate": 1.8944451287104326e-06, + "loss": 0.3071, + "step": 29970 + }, + { + "epoch": 11.07, + "learning_rate": 1.894408178347087e-06, + "loss": 0.3063, + "step": 29971 + }, + { + "epoch": 11.07, + "learning_rate": 1.8943712279837417e-06, + "loss": 0.3778, + "step": 29972 + }, + { + "epoch": 11.07, + "learning_rate": 1.8943342776203967e-06, + "loss": 0.2606, + "step": 29973 + }, + { + "epoch": 11.07, + "learning_rate": 1.8942973272570514e-06, + "loss": 0.3521, + "step": 29974 + }, + { + "epoch": 11.07, + "learning_rate": 1.8942603768937062e-06, + "loss": 0.3582, + "step": 29975 + }, + { + "epoch": 11.07, + "learning_rate": 1.8942234265303609e-06, + "loss": 0.2517, + "step": 29976 + }, + { + "epoch": 11.07, + "learning_rate": 1.8941864761670157e-06, + "loss": 0.2762, + "step": 29977 + }, + { + "epoch": 11.07, + "learning_rate": 1.8941495258036703e-06, + "loss": 0.204, + "step": 29978 + }, + { + "epoch": 11.07, + "learning_rate": 1.8941125754403254e-06, + "loss": 0.344, + "step": 29979 + }, + { + "epoch": 11.07, + "learning_rate": 1.89407562507698e-06, + "loss": 0.2908, + "step": 29980 + }, + { + "epoch": 11.07, + "learning_rate": 1.8940386747136345e-06, + "loss": 0.3215, + "step": 29981 + }, + { + "epoch": 11.07, + "learning_rate": 1.8940017243502895e-06, + "loss": 0.2587, + "step": 29982 + }, + { + "epoch": 11.07, + "learning_rate": 1.8939647739869442e-06, + "loss": 0.3647, + "step": 29983 + }, + { + "epoch": 11.07, + "learning_rate": 1.893927823623599e-06, + "loss": 0.3193, + "step": 29984 + }, + { + "epoch": 11.07, + "learning_rate": 1.8938908732602536e-06, + "loss": 0.3543, + "step": 29985 + }, + { + "epoch": 11.07, + "learning_rate": 1.8938539228969087e-06, + "loss": 0.3379, + "step": 29986 + }, + { + "epoch": 11.07, + "learning_rate": 1.8938169725335631e-06, + "loss": 0.3093, + "step": 29987 + }, + { + "epoch": 11.07, + "learning_rate": 1.8937800221702182e-06, + "loss": 0.2207, + "step": 29988 + }, + { + "epoch": 11.07, + "learning_rate": 1.8937430718068728e-06, + "loss": 0.3788, + "step": 29989 + }, + { + "epoch": 11.07, + "learning_rate": 1.8937061214435277e-06, + "loss": 0.2577, + "step": 29990 + }, + { + "epoch": 11.07, + "learning_rate": 1.8936691710801823e-06, + "loss": 0.2921, + "step": 29991 + }, + { + "epoch": 11.08, + "learning_rate": 1.893632220716837e-06, + "loss": 0.3457, + "step": 29992 + }, + { + "epoch": 11.08, + "learning_rate": 1.8935952703534918e-06, + "loss": 0.25, + "step": 29993 + }, + { + "epoch": 11.08, + "learning_rate": 1.8935583199901464e-06, + "loss": 0.298, + "step": 29994 + }, + { + "epoch": 11.08, + "learning_rate": 1.8935213696268015e-06, + "loss": 0.2396, + "step": 29995 + }, + { + "epoch": 11.08, + "learning_rate": 1.8934844192634561e-06, + "loss": 0.2818, + "step": 29996 + }, + { + "epoch": 11.08, + "learning_rate": 1.893447468900111e-06, + "loss": 0.3296, + "step": 29997 + }, + { + "epoch": 11.08, + "learning_rate": 1.8934105185367656e-06, + "loss": 0.3806, + "step": 29998 + }, + { + "epoch": 11.08, + "learning_rate": 1.8933735681734205e-06, + "loss": 0.3032, + "step": 29999 + }, + { + "epoch": 11.08, + "learning_rate": 1.893336617810075e-06, + "loss": 0.3243, + "step": 30000 + }, + { + "epoch": 11.08, + "learning_rate": 1.8932996674467297e-06, + "loss": 0.2154, + "step": 30001 + }, + { + "epoch": 11.08, + "learning_rate": 1.8932627170833848e-06, + "loss": 0.3896, + "step": 30002 + }, + { + "epoch": 11.08, + "learning_rate": 1.8932257667200394e-06, + "loss": 0.259, + "step": 30003 + }, + { + "epoch": 11.08, + "learning_rate": 1.8931888163566943e-06, + "loss": 0.3605, + "step": 30004 + }, + { + "epoch": 11.08, + "learning_rate": 1.893151865993349e-06, + "loss": 0.3901, + "step": 30005 + }, + { + "epoch": 11.08, + "learning_rate": 1.8931149156300038e-06, + "loss": 0.3216, + "step": 30006 + }, + { + "epoch": 11.08, + "learning_rate": 1.8930779652666584e-06, + "loss": 0.3319, + "step": 30007 + }, + { + "epoch": 11.08, + "learning_rate": 1.8930410149033135e-06, + "loss": 0.3506, + "step": 30008 + }, + { + "epoch": 11.08, + "learning_rate": 1.893004064539968e-06, + "loss": 0.248, + "step": 30009 + }, + { + "epoch": 11.08, + "learning_rate": 1.892967114176623e-06, + "loss": 0.2917, + "step": 30010 + }, + { + "epoch": 11.08, + "learning_rate": 1.8929301638132776e-06, + "loss": 0.3708, + "step": 30011 + }, + { + "epoch": 11.08, + "learning_rate": 1.8928932134499322e-06, + "loss": 0.374, + "step": 30012 + }, + { + "epoch": 11.08, + "learning_rate": 1.892856263086587e-06, + "loss": 0.3816, + "step": 30013 + }, + { + "epoch": 11.08, + "learning_rate": 1.8928193127232417e-06, + "loss": 0.3962, + "step": 30014 + }, + { + "epoch": 11.08, + "learning_rate": 1.8927823623598968e-06, + "loss": 0.2022, + "step": 30015 + }, + { + "epoch": 11.08, + "learning_rate": 1.8927454119965512e-06, + "loss": 0.2949, + "step": 30016 + }, + { + "epoch": 11.08, + "learning_rate": 1.8927084616332062e-06, + "loss": 0.2686, + "step": 30017 + }, + { + "epoch": 11.08, + "learning_rate": 1.8926715112698609e-06, + "loss": 0.2733, + "step": 30018 + }, + { + "epoch": 11.09, + "learning_rate": 1.8926345609065157e-06, + "loss": 0.4018, + "step": 30019 + }, + { + "epoch": 11.09, + "learning_rate": 1.8925976105431704e-06, + "loss": 0.3538, + "step": 30020 + }, + { + "epoch": 11.09, + "learning_rate": 1.892560660179825e-06, + "loss": 0.2441, + "step": 30021 + }, + { + "epoch": 11.09, + "learning_rate": 1.8925237098164798e-06, + "loss": 0.3308, + "step": 30022 + }, + { + "epoch": 11.09, + "learning_rate": 1.8924867594531345e-06, + "loss": 0.3939, + "step": 30023 + }, + { + "epoch": 11.09, + "learning_rate": 1.8924498090897895e-06, + "loss": 0.325, + "step": 30024 + }, + { + "epoch": 11.09, + "learning_rate": 1.8924128587264442e-06, + "loss": 0.2115, + "step": 30025 + }, + { + "epoch": 11.09, + "learning_rate": 1.892375908363099e-06, + "loss": 0.3355, + "step": 30026 + }, + { + "epoch": 11.09, + "learning_rate": 1.8923389579997537e-06, + "loss": 0.3363, + "step": 30027 + }, + { + "epoch": 11.09, + "learning_rate": 1.8923020076364085e-06, + "loss": 0.3949, + "step": 30028 + }, + { + "epoch": 11.09, + "learning_rate": 1.8922650572730632e-06, + "loss": 0.3229, + "step": 30029 + }, + { + "epoch": 11.09, + "learning_rate": 1.8922281069097182e-06, + "loss": 0.297, + "step": 30030 + }, + { + "epoch": 11.09, + "learning_rate": 1.8921911565463728e-06, + "loss": 0.2414, + "step": 30031 + }, + { + "epoch": 11.09, + "learning_rate": 1.8921542061830273e-06, + "loss": 0.3159, + "step": 30032 + }, + { + "epoch": 11.09, + "learning_rate": 1.8921172558196823e-06, + "loss": 0.253, + "step": 30033 + }, + { + "epoch": 11.09, + "learning_rate": 1.892080305456337e-06, + "loss": 0.4022, + "step": 30034 + }, + { + "epoch": 11.09, + "learning_rate": 1.8920433550929918e-06, + "loss": 0.2977, + "step": 30035 + }, + { + "epoch": 11.09, + "learning_rate": 1.8920064047296465e-06, + "loss": 0.3102, + "step": 30036 + }, + { + "epoch": 11.09, + "learning_rate": 1.8919694543663015e-06, + "loss": 0.2476, + "step": 30037 + }, + { + "epoch": 11.09, + "learning_rate": 1.891932504002956e-06, + "loss": 0.2848, + "step": 30038 + }, + { + "epoch": 11.09, + "learning_rate": 1.891895553639611e-06, + "loss": 0.3195, + "step": 30039 + }, + { + "epoch": 11.09, + "learning_rate": 1.8918586032762656e-06, + "loss": 0.2304, + "step": 30040 + }, + { + "epoch": 11.09, + "learning_rate": 1.8918216529129203e-06, + "loss": 0.3023, + "step": 30041 + }, + { + "epoch": 11.09, + "learning_rate": 1.8917847025495751e-06, + "loss": 0.3917, + "step": 30042 + }, + { + "epoch": 11.09, + "learning_rate": 1.8917477521862298e-06, + "loss": 0.34, + "step": 30043 + }, + { + "epoch": 11.09, + "learning_rate": 1.8917108018228846e-06, + "loss": 0.4099, + "step": 30044 + }, + { + "epoch": 11.09, + "learning_rate": 1.8916738514595392e-06, + "loss": 0.258, + "step": 30045 + }, + { + "epoch": 11.1, + "learning_rate": 1.8916369010961943e-06, + "loss": 0.2763, + "step": 30046 + }, + { + "epoch": 11.1, + "learning_rate": 1.891599950732849e-06, + "loss": 0.2501, + "step": 30047 + }, + { + "epoch": 11.1, + "learning_rate": 1.8915630003695038e-06, + "loss": 0.2372, + "step": 30048 + }, + { + "epoch": 11.1, + "learning_rate": 1.8915260500061584e-06, + "loss": 0.2732, + "step": 30049 + }, + { + "epoch": 11.1, + "learning_rate": 1.8914890996428133e-06, + "loss": 0.3525, + "step": 30050 + }, + { + "epoch": 11.1, + "learning_rate": 1.891452149279468e-06, + "loss": 0.2132, + "step": 30051 + }, + { + "epoch": 11.1, + "learning_rate": 1.8914151989161225e-06, + "loss": 0.2825, + "step": 30052 + }, + { + "epoch": 11.1, + "learning_rate": 1.8913782485527776e-06, + "loss": 0.4989, + "step": 30053 + }, + { + "epoch": 11.1, + "learning_rate": 1.891341298189432e-06, + "loss": 0.2562, + "step": 30054 + }, + { + "epoch": 11.1, + "learning_rate": 1.891304347826087e-06, + "loss": 0.26, + "step": 30055 + }, + { + "epoch": 11.1, + "learning_rate": 1.8912673974627417e-06, + "loss": 0.378, + "step": 30056 + }, + { + "epoch": 11.1, + "learning_rate": 1.8912304470993966e-06, + "loss": 0.4716, + "step": 30057 + }, + { + "epoch": 11.1, + "learning_rate": 1.8911934967360512e-06, + "loss": 0.307, + "step": 30058 + }, + { + "epoch": 11.1, + "learning_rate": 1.8911565463727063e-06, + "loss": 0.2892, + "step": 30059 + }, + { + "epoch": 11.1, + "learning_rate": 1.8911195960093607e-06, + "loss": 0.2854, + "step": 30060 + }, + { + "epoch": 11.1, + "learning_rate": 1.8910826456460153e-06, + "loss": 0.2878, + "step": 30061 + }, + { + "epoch": 11.1, + "learning_rate": 1.8910456952826704e-06, + "loss": 0.4692, + "step": 30062 + }, + { + "epoch": 11.1, + "learning_rate": 1.891008744919325e-06, + "loss": 0.4575, + "step": 30063 + }, + { + "epoch": 11.1, + "learning_rate": 1.8909717945559799e-06, + "loss": 0.3169, + "step": 30064 + }, + { + "epoch": 11.1, + "learning_rate": 1.8909348441926345e-06, + "loss": 0.2866, + "step": 30065 + }, + { + "epoch": 11.1, + "learning_rate": 1.8908978938292894e-06, + "loss": 0.2803, + "step": 30066 + }, + { + "epoch": 11.1, + "learning_rate": 1.890860943465944e-06, + "loss": 0.2204, + "step": 30067 + }, + { + "epoch": 11.1, + "learning_rate": 1.890823993102599e-06, + "loss": 0.2559, + "step": 30068 + }, + { + "epoch": 11.1, + "learning_rate": 1.8907870427392537e-06, + "loss": 0.3971, + "step": 30069 + }, + { + "epoch": 11.1, + "learning_rate": 1.8907500923759085e-06, + "loss": 0.3237, + "step": 30070 + }, + { + "epoch": 11.1, + "learning_rate": 1.8907131420125632e-06, + "loss": 0.3726, + "step": 30071 + }, + { + "epoch": 11.1, + "learning_rate": 1.8906761916492178e-06, + "loss": 0.2598, + "step": 30072 + }, + { + "epoch": 11.11, + "learning_rate": 1.8906392412858727e-06, + "loss": 0.3657, + "step": 30073 + }, + { + "epoch": 11.11, + "learning_rate": 1.8906022909225273e-06, + "loss": 0.2749, + "step": 30074 + }, + { + "epoch": 11.11, + "learning_rate": 1.8905653405591824e-06, + "loss": 0.3136, + "step": 30075 + }, + { + "epoch": 11.11, + "learning_rate": 1.890528390195837e-06, + "loss": 0.4675, + "step": 30076 + }, + { + "epoch": 11.11, + "learning_rate": 1.8904914398324918e-06, + "loss": 0.4282, + "step": 30077 + }, + { + "epoch": 11.11, + "learning_rate": 1.8904544894691465e-06, + "loss": 0.3759, + "step": 30078 + }, + { + "epoch": 11.11, + "learning_rate": 1.8904175391058013e-06, + "loss": 0.354, + "step": 30079 + }, + { + "epoch": 11.11, + "learning_rate": 1.890380588742456e-06, + "loss": 0.3298, + "step": 30080 + }, + { + "epoch": 11.11, + "learning_rate": 1.8903436383791106e-06, + "loss": 0.2311, + "step": 30081 + }, + { + "epoch": 11.11, + "learning_rate": 1.8903066880157657e-06, + "loss": 0.3878, + "step": 30082 + }, + { + "epoch": 11.11, + "learning_rate": 1.89026973765242e-06, + "loss": 0.3099, + "step": 30083 + }, + { + "epoch": 11.11, + "learning_rate": 1.8902327872890751e-06, + "loss": 0.2582, + "step": 30084 + }, + { + "epoch": 11.11, + "learning_rate": 1.8901958369257298e-06, + "loss": 0.2808, + "step": 30085 + }, + { + "epoch": 11.11, + "learning_rate": 1.8901588865623846e-06, + "loss": 0.4555, + "step": 30086 + }, + { + "epoch": 11.11, + "learning_rate": 1.8901219361990393e-06, + "loss": 0.299, + "step": 30087 + }, + { + "epoch": 11.11, + "learning_rate": 1.8900849858356943e-06, + "loss": 0.3624, + "step": 30088 + }, + { + "epoch": 11.11, + "learning_rate": 1.8900480354723487e-06, + "loss": 0.2981, + "step": 30089 + }, + { + "epoch": 11.11, + "learning_rate": 1.8900110851090038e-06, + "loss": 0.3428, + "step": 30090 + }, + { + "epoch": 11.11, + "learning_rate": 1.8899741347456584e-06, + "loss": 0.2931, + "step": 30091 + }, + { + "epoch": 11.11, + "learning_rate": 1.889937184382313e-06, + "loss": 0.3199, + "step": 30092 + }, + { + "epoch": 11.11, + "learning_rate": 1.889900234018968e-06, + "loss": 0.3974, + "step": 30093 + }, + { + "epoch": 11.11, + "learning_rate": 1.8898632836556226e-06, + "loss": 0.4808, + "step": 30094 + }, + { + "epoch": 11.11, + "learning_rate": 1.8898263332922774e-06, + "loss": 0.3258, + "step": 30095 + }, + { + "epoch": 11.11, + "learning_rate": 1.889789382928932e-06, + "loss": 0.3926, + "step": 30096 + }, + { + "epoch": 11.11, + "learning_rate": 1.8897524325655871e-06, + "loss": 0.3527, + "step": 30097 + }, + { + "epoch": 11.11, + "learning_rate": 1.8897154822022417e-06, + "loss": 0.254, + "step": 30098 + }, + { + "epoch": 11.11, + "learning_rate": 1.8896785318388966e-06, + "loss": 0.3889, + "step": 30099 + }, + { + "epoch": 11.12, + "learning_rate": 1.8896415814755512e-06, + "loss": 0.2531, + "step": 30100 + }, + { + "epoch": 11.12, + "learning_rate": 1.8896046311122059e-06, + "loss": 0.454, + "step": 30101 + }, + { + "epoch": 11.12, + "learning_rate": 1.8895676807488607e-06, + "loss": 0.1978, + "step": 30102 + }, + { + "epoch": 11.12, + "learning_rate": 1.8895307303855154e-06, + "loss": 0.4382, + "step": 30103 + }, + { + "epoch": 11.12, + "learning_rate": 1.8894937800221704e-06, + "loss": 0.3665, + "step": 30104 + }, + { + "epoch": 11.12, + "learning_rate": 1.8894568296588248e-06, + "loss": 0.4325, + "step": 30105 + }, + { + "epoch": 11.12, + "learning_rate": 1.88941987929548e-06, + "loss": 0.4634, + "step": 30106 + }, + { + "epoch": 11.12, + "learning_rate": 1.8893829289321345e-06, + "loss": 0.3864, + "step": 30107 + }, + { + "epoch": 11.12, + "learning_rate": 1.8893459785687894e-06, + "loss": 0.3421, + "step": 30108 + }, + { + "epoch": 11.12, + "learning_rate": 1.889309028205444e-06, + "loss": 0.3791, + "step": 30109 + }, + { + "epoch": 11.12, + "learning_rate": 1.889272077842099e-06, + "loss": 0.2135, + "step": 30110 + }, + { + "epoch": 11.12, + "learning_rate": 1.8892351274787535e-06, + "loss": 0.2389, + "step": 30111 + }, + { + "epoch": 11.12, + "learning_rate": 1.8891981771154081e-06, + "loss": 0.3082, + "step": 30112 + }, + { + "epoch": 11.12, + "learning_rate": 1.8891612267520632e-06, + "loss": 0.184, + "step": 30113 + }, + { + "epoch": 11.12, + "learning_rate": 1.8891242763887178e-06, + "loss": 0.2898, + "step": 30114 + }, + { + "epoch": 11.12, + "learning_rate": 1.8890873260253727e-06, + "loss": 0.338, + "step": 30115 + }, + { + "epoch": 11.12, + "learning_rate": 1.8890503756620273e-06, + "loss": 0.2816, + "step": 30116 + }, + { + "epoch": 11.12, + "learning_rate": 1.8890134252986822e-06, + "loss": 0.2914, + "step": 30117 + }, + { + "epoch": 11.12, + "learning_rate": 1.8889764749353368e-06, + "loss": 0.349, + "step": 30118 + }, + { + "epoch": 11.12, + "learning_rate": 1.8889395245719919e-06, + "loss": 0.3704, + "step": 30119 + }, + { + "epoch": 11.12, + "learning_rate": 1.8889025742086465e-06, + "loss": 0.3481, + "step": 30120 + }, + { + "epoch": 11.12, + "learning_rate": 1.8888656238453011e-06, + "loss": 0.3114, + "step": 30121 + }, + { + "epoch": 11.12, + "learning_rate": 1.888828673481956e-06, + "loss": 0.2711, + "step": 30122 + }, + { + "epoch": 11.12, + "learning_rate": 1.8887917231186106e-06, + "loss": 0.3304, + "step": 30123 + }, + { + "epoch": 11.12, + "learning_rate": 1.8887547727552655e-06, + "loss": 0.2806, + "step": 30124 + }, + { + "epoch": 11.12, + "learning_rate": 1.8887178223919201e-06, + "loss": 0.2611, + "step": 30125 + }, + { + "epoch": 11.12, + "learning_rate": 1.8886808720285752e-06, + "loss": 0.3003, + "step": 30126 + }, + { + "epoch": 11.13, + "learning_rate": 1.8886439216652298e-06, + "loss": 0.2664, + "step": 30127 + }, + { + "epoch": 11.13, + "learning_rate": 1.8886069713018847e-06, + "loss": 0.3029, + "step": 30128 + }, + { + "epoch": 11.13, + "learning_rate": 1.8885700209385393e-06, + "loss": 0.3232, + "step": 30129 + }, + { + "epoch": 11.13, + "learning_rate": 1.8885330705751941e-06, + "loss": 0.3934, + "step": 30130 + }, + { + "epoch": 11.13, + "learning_rate": 1.8884961202118488e-06, + "loss": 0.2495, + "step": 30131 + }, + { + "epoch": 11.13, + "learning_rate": 1.8884591698485034e-06, + "loss": 0.4274, + "step": 30132 + }, + { + "epoch": 11.13, + "learning_rate": 1.8884222194851583e-06, + "loss": 0.2463, + "step": 30133 + }, + { + "epoch": 11.13, + "learning_rate": 1.888385269121813e-06, + "loss": 0.1899, + "step": 30134 + }, + { + "epoch": 11.13, + "learning_rate": 1.888348318758468e-06, + "loss": 0.2784, + "step": 30135 + }, + { + "epoch": 11.13, + "learning_rate": 1.8883113683951226e-06, + "loss": 0.3737, + "step": 30136 + }, + { + "epoch": 11.13, + "learning_rate": 1.8882744180317774e-06, + "loss": 0.2015, + "step": 30137 + }, + { + "epoch": 11.13, + "learning_rate": 1.888237467668432e-06, + "loss": 0.3257, + "step": 30138 + }, + { + "epoch": 11.13, + "learning_rate": 1.888200517305087e-06, + "loss": 0.3287, + "step": 30139 + }, + { + "epoch": 11.13, + "learning_rate": 1.8881635669417416e-06, + "loss": 0.2802, + "step": 30140 + }, + { + "epoch": 11.13, + "learning_rate": 1.8881266165783962e-06, + "loss": 0.378, + "step": 30141 + }, + { + "epoch": 11.13, + "learning_rate": 1.8880896662150513e-06, + "loss": 0.4436, + "step": 30142 + }, + { + "epoch": 11.13, + "learning_rate": 1.8880527158517059e-06, + "loss": 0.2256, + "step": 30143 + }, + { + "epoch": 11.13, + "learning_rate": 1.8880157654883607e-06, + "loss": 0.4989, + "step": 30144 + }, + { + "epoch": 11.13, + "learning_rate": 1.8879788151250154e-06, + "loss": 0.4341, + "step": 30145 + }, + { + "epoch": 11.13, + "learning_rate": 1.8879418647616702e-06, + "loss": 0.1946, + "step": 30146 + }, + { + "epoch": 11.13, + "learning_rate": 1.8879049143983249e-06, + "loss": 0.2312, + "step": 30147 + }, + { + "epoch": 11.13, + "learning_rate": 1.88786796403498e-06, + "loss": 0.2877, + "step": 30148 + }, + { + "epoch": 11.13, + "learning_rate": 1.8878310136716346e-06, + "loss": 0.2658, + "step": 30149 + }, + { + "epoch": 11.13, + "learning_rate": 1.8877940633082894e-06, + "loss": 0.3504, + "step": 30150 + }, + { + "epoch": 11.13, + "learning_rate": 1.887757112944944e-06, + "loss": 0.4428, + "step": 30151 + }, + { + "epoch": 11.13, + "learning_rate": 1.8877201625815987e-06, + "loss": 0.3284, + "step": 30152 + }, + { + "epoch": 11.13, + "learning_rate": 1.8876832122182535e-06, + "loss": 0.2611, + "step": 30153 + }, + { + "epoch": 11.14, + "learning_rate": 1.8876462618549082e-06, + "loss": 0.2325, + "step": 30154 + }, + { + "epoch": 11.14, + "learning_rate": 1.8876093114915632e-06, + "loss": 0.1879, + "step": 30155 + }, + { + "epoch": 11.14, + "learning_rate": 1.8875723611282176e-06, + "loss": 0.1793, + "step": 30156 + }, + { + "epoch": 11.14, + "learning_rate": 1.8875354107648727e-06, + "loss": 0.3085, + "step": 30157 + }, + { + "epoch": 11.14, + "learning_rate": 1.8874984604015273e-06, + "loss": 0.4596, + "step": 30158 + }, + { + "epoch": 11.14, + "learning_rate": 1.8874615100381822e-06, + "loss": 0.1989, + "step": 30159 + }, + { + "epoch": 11.14, + "learning_rate": 1.8874245596748368e-06, + "loss": 0.3317, + "step": 30160 + }, + { + "epoch": 11.14, + "learning_rate": 1.8873876093114915e-06, + "loss": 0.2126, + "step": 30161 + }, + { + "epoch": 11.14, + "learning_rate": 1.8873506589481463e-06, + "loss": 0.3437, + "step": 30162 + }, + { + "epoch": 11.14, + "learning_rate": 1.887313708584801e-06, + "loss": 0.3294, + "step": 30163 + }, + { + "epoch": 11.14, + "learning_rate": 1.887276758221456e-06, + "loss": 0.2917, + "step": 30164 + }, + { + "epoch": 11.14, + "learning_rate": 1.8872398078581106e-06, + "loss": 0.2502, + "step": 30165 + }, + { + "epoch": 11.14, + "learning_rate": 1.8872028574947655e-06, + "loss": 0.3572, + "step": 30166 + }, + { + "epoch": 11.14, + "learning_rate": 1.8871659071314201e-06, + "loss": 0.2641, + "step": 30167 + }, + { + "epoch": 11.14, + "learning_rate": 1.887128956768075e-06, + "loss": 0.3514, + "step": 30168 + }, + { + "epoch": 11.14, + "learning_rate": 1.8870920064047296e-06, + "loss": 0.3042, + "step": 30169 + }, + { + "epoch": 11.14, + "learning_rate": 1.8870550560413847e-06, + "loss": 0.2302, + "step": 30170 + }, + { + "epoch": 11.14, + "learning_rate": 1.8870181056780393e-06, + "loss": 0.2717, + "step": 30171 + }, + { + "epoch": 11.14, + "learning_rate": 1.8869811553146937e-06, + "loss": 0.2085, + "step": 30172 + }, + { + "epoch": 11.14, + "learning_rate": 1.8869442049513488e-06, + "loss": 0.3117, + "step": 30173 + }, + { + "epoch": 11.14, + "learning_rate": 1.8869072545880034e-06, + "loss": 0.3732, + "step": 30174 + }, + { + "epoch": 11.14, + "learning_rate": 1.8868703042246583e-06, + "loss": 0.3473, + "step": 30175 + }, + { + "epoch": 11.14, + "learning_rate": 1.886833353861313e-06, + "loss": 0.3074, + "step": 30176 + }, + { + "epoch": 11.14, + "learning_rate": 1.886796403497968e-06, + "loss": 0.2868, + "step": 30177 + }, + { + "epoch": 11.14, + "learning_rate": 1.8867594531346224e-06, + "loss": 0.2673, + "step": 30178 + }, + { + "epoch": 11.14, + "learning_rate": 1.8867225027712775e-06, + "loss": 0.2997, + "step": 30179 + }, + { + "epoch": 11.14, + "learning_rate": 1.886685552407932e-06, + "loss": 0.3526, + "step": 30180 + }, + { + "epoch": 11.15, + "learning_rate": 1.8866486020445867e-06, + "loss": 0.2002, + "step": 30181 + }, + { + "epoch": 11.15, + "learning_rate": 1.8866116516812416e-06, + "loss": 0.2899, + "step": 30182 + }, + { + "epoch": 11.15, + "learning_rate": 1.8865747013178962e-06, + "loss": 0.2884, + "step": 30183 + }, + { + "epoch": 11.15, + "learning_rate": 1.886537750954551e-06, + "loss": 0.2932, + "step": 30184 + }, + { + "epoch": 11.15, + "learning_rate": 1.8865008005912057e-06, + "loss": 0.3362, + "step": 30185 + }, + { + "epoch": 11.15, + "learning_rate": 1.8864638502278608e-06, + "loss": 0.2325, + "step": 30186 + }, + { + "epoch": 11.15, + "learning_rate": 1.8864268998645154e-06, + "loss": 0.237, + "step": 30187 + }, + { + "epoch": 11.15, + "learning_rate": 1.8863899495011703e-06, + "loss": 0.4014, + "step": 30188 + }, + { + "epoch": 11.15, + "learning_rate": 1.8863529991378249e-06, + "loss": 0.2501, + "step": 30189 + }, + { + "epoch": 11.15, + "learning_rate": 1.8863160487744797e-06, + "loss": 0.2606, + "step": 30190 + }, + { + "epoch": 11.15, + "learning_rate": 1.8862790984111344e-06, + "loss": 0.2794, + "step": 30191 + }, + { + "epoch": 11.15, + "learning_rate": 1.886242148047789e-06, + "loss": 0.3384, + "step": 30192 + }, + { + "epoch": 11.15, + "learning_rate": 1.886205197684444e-06, + "loss": 0.4097, + "step": 30193 + }, + { + "epoch": 11.15, + "learning_rate": 1.8861682473210987e-06, + "loss": 0.2442, + "step": 30194 + }, + { + "epoch": 11.15, + "learning_rate": 1.8861312969577536e-06, + "loss": 0.2954, + "step": 30195 + }, + { + "epoch": 11.15, + "learning_rate": 1.8860943465944082e-06, + "loss": 0.2348, + "step": 30196 + }, + { + "epoch": 11.15, + "learning_rate": 1.886057396231063e-06, + "loss": 0.3087, + "step": 30197 + }, + { + "epoch": 11.15, + "learning_rate": 1.8860204458677177e-06, + "loss": 0.3506, + "step": 30198 + }, + { + "epoch": 11.15, + "learning_rate": 1.8859834955043727e-06, + "loss": 0.475, + "step": 30199 + }, + { + "epoch": 11.15, + "learning_rate": 1.8859465451410274e-06, + "loss": 0.3583, + "step": 30200 + }, + { + "epoch": 11.15, + "learning_rate": 1.8859095947776818e-06, + "loss": 0.4025, + "step": 30201 + }, + { + "epoch": 11.15, + "learning_rate": 1.8858726444143369e-06, + "loss": 0.4138, + "step": 30202 + }, + { + "epoch": 11.15, + "learning_rate": 1.8858356940509915e-06, + "loss": 0.2528, + "step": 30203 + }, + { + "epoch": 11.15, + "learning_rate": 1.8857987436876463e-06, + "loss": 0.3644, + "step": 30204 + }, + { + "epoch": 11.15, + "learning_rate": 1.885761793324301e-06, + "loss": 0.2936, + "step": 30205 + }, + { + "epoch": 11.15, + "learning_rate": 1.885724842960956e-06, + "loss": 0.354, + "step": 30206 + }, + { + "epoch": 11.15, + "learning_rate": 1.8856878925976105e-06, + "loss": 0.391, + "step": 30207 + }, + { + "epoch": 11.16, + "learning_rate": 1.8856509422342655e-06, + "loss": 0.2744, + "step": 30208 + }, + { + "epoch": 11.16, + "learning_rate": 1.8856139918709202e-06, + "loss": 0.3984, + "step": 30209 + }, + { + "epoch": 11.16, + "learning_rate": 1.8855770415075748e-06, + "loss": 0.3184, + "step": 30210 + }, + { + "epoch": 11.16, + "learning_rate": 1.8855400911442296e-06, + "loss": 0.2805, + "step": 30211 + }, + { + "epoch": 11.16, + "learning_rate": 1.8855031407808843e-06, + "loss": 0.4342, + "step": 30212 + }, + { + "epoch": 11.16, + "learning_rate": 1.8854661904175391e-06, + "loss": 0.2929, + "step": 30213 + }, + { + "epoch": 11.16, + "learning_rate": 1.8854292400541938e-06, + "loss": 0.3553, + "step": 30214 + }, + { + "epoch": 11.16, + "learning_rate": 1.8853922896908488e-06, + "loss": 0.3761, + "step": 30215 + }, + { + "epoch": 11.16, + "learning_rate": 1.8853553393275035e-06, + "loss": 0.3302, + "step": 30216 + }, + { + "epoch": 11.16, + "learning_rate": 1.8853183889641583e-06, + "loss": 0.3423, + "step": 30217 + }, + { + "epoch": 11.16, + "learning_rate": 1.885281438600813e-06, + "loss": 0.237, + "step": 30218 + }, + { + "epoch": 11.16, + "learning_rate": 1.8852444882374678e-06, + "loss": 0.3352, + "step": 30219 + }, + { + "epoch": 11.16, + "learning_rate": 1.8852075378741224e-06, + "loss": 0.3135, + "step": 30220 + }, + { + "epoch": 11.16, + "learning_rate": 1.885170587510777e-06, + "loss": 0.3844, + "step": 30221 + }, + { + "epoch": 11.16, + "learning_rate": 1.8851336371474321e-06, + "loss": 0.3541, + "step": 30222 + }, + { + "epoch": 11.16, + "learning_rate": 1.8850966867840865e-06, + "loss": 0.229, + "step": 30223 + }, + { + "epoch": 11.16, + "learning_rate": 1.8850597364207416e-06, + "loss": 0.2982, + "step": 30224 + }, + { + "epoch": 11.16, + "learning_rate": 1.8850227860573962e-06, + "loss": 0.2344, + "step": 30225 + }, + { + "epoch": 11.16, + "learning_rate": 1.884985835694051e-06, + "loss": 0.1523, + "step": 30226 + }, + { + "epoch": 11.16, + "learning_rate": 1.8849488853307057e-06, + "loss": 0.3189, + "step": 30227 + }, + { + "epoch": 11.16, + "learning_rate": 1.8849119349673608e-06, + "loss": 0.2449, + "step": 30228 + }, + { + "epoch": 11.16, + "learning_rate": 1.8848749846040152e-06, + "loss": 0.3452, + "step": 30229 + }, + { + "epoch": 11.16, + "learning_rate": 1.8848380342406699e-06, + "loss": 0.2147, + "step": 30230 + }, + { + "epoch": 11.16, + "learning_rate": 1.884801083877325e-06, + "loss": 0.2829, + "step": 30231 + }, + { + "epoch": 11.16, + "learning_rate": 1.8847641335139795e-06, + "loss": 0.2151, + "step": 30232 + }, + { + "epoch": 11.16, + "learning_rate": 1.8847271831506344e-06, + "loss": 0.3415, + "step": 30233 + }, + { + "epoch": 11.16, + "learning_rate": 1.884690232787289e-06, + "loss": 0.2133, + "step": 30234 + }, + { + "epoch": 11.17, + "learning_rate": 1.8846532824239439e-06, + "loss": 0.2237, + "step": 30235 + }, + { + "epoch": 11.17, + "learning_rate": 1.8846163320605985e-06, + "loss": 0.3933, + "step": 30236 + }, + { + "epoch": 11.17, + "learning_rate": 1.8845793816972536e-06, + "loss": 0.3882, + "step": 30237 + }, + { + "epoch": 11.17, + "learning_rate": 1.8845424313339082e-06, + "loss": 0.405, + "step": 30238 + }, + { + "epoch": 11.17, + "learning_rate": 1.884505480970563e-06, + "loss": 0.2489, + "step": 30239 + }, + { + "epoch": 11.17, + "learning_rate": 1.8844685306072177e-06, + "loss": 0.3573, + "step": 30240 + }, + { + "epoch": 11.17, + "learning_rate": 1.8844315802438723e-06, + "loss": 0.3442, + "step": 30241 + }, + { + "epoch": 11.17, + "learning_rate": 1.8843946298805272e-06, + "loss": 0.2538, + "step": 30242 + }, + { + "epoch": 11.17, + "learning_rate": 1.8843576795171818e-06, + "loss": 0.2331, + "step": 30243 + }, + { + "epoch": 11.17, + "learning_rate": 1.8843207291538369e-06, + "loss": 0.2082, + "step": 30244 + }, + { + "epoch": 11.17, + "learning_rate": 1.8842837787904913e-06, + "loss": 0.2371, + "step": 30245 + }, + { + "epoch": 11.17, + "learning_rate": 1.8842468284271464e-06, + "loss": 0.3047, + "step": 30246 + }, + { + "epoch": 11.17, + "learning_rate": 1.884209878063801e-06, + "loss": 0.2855, + "step": 30247 + }, + { + "epoch": 11.17, + "learning_rate": 1.8841729277004558e-06, + "loss": 0.2327, + "step": 30248 + }, + { + "epoch": 11.17, + "learning_rate": 1.8841359773371105e-06, + "loss": 0.3081, + "step": 30249 + }, + { + "epoch": 11.17, + "learning_rate": 1.8840990269737651e-06, + "loss": 0.4825, + "step": 30250 + }, + { + "epoch": 11.17, + "learning_rate": 1.88406207661042e-06, + "loss": 0.2649, + "step": 30251 + }, + { + "epoch": 11.17, + "learning_rate": 1.8840251262470746e-06, + "loss": 0.2981, + "step": 30252 + }, + { + "epoch": 11.17, + "learning_rate": 1.8839881758837297e-06, + "loss": 0.2517, + "step": 30253 + }, + { + "epoch": 11.17, + "learning_rate": 1.8839512255203843e-06, + "loss": 0.306, + "step": 30254 + }, + { + "epoch": 11.17, + "learning_rate": 1.8839142751570391e-06, + "loss": 0.244, + "step": 30255 + }, + { + "epoch": 11.17, + "learning_rate": 1.8838773247936938e-06, + "loss": 0.2405, + "step": 30256 + }, + { + "epoch": 11.17, + "learning_rate": 1.8838403744303486e-06, + "loss": 0.2317, + "step": 30257 + }, + { + "epoch": 11.17, + "learning_rate": 1.8838034240670033e-06, + "loss": 0.3507, + "step": 30258 + }, + { + "epoch": 11.17, + "learning_rate": 1.8837664737036583e-06, + "loss": 0.2035, + "step": 30259 + }, + { + "epoch": 11.17, + "learning_rate": 1.883729523340313e-06, + "loss": 0.3567, + "step": 30260 + }, + { + "epoch": 11.17, + "learning_rate": 1.8836925729769676e-06, + "loss": 0.3158, + "step": 30261 + }, + { + "epoch": 11.18, + "learning_rate": 1.8836556226136225e-06, + "loss": 0.4746, + "step": 30262 + }, + { + "epoch": 11.18, + "learning_rate": 1.883618672250277e-06, + "loss": 0.4867, + "step": 30263 + }, + { + "epoch": 11.18, + "learning_rate": 1.883581721886932e-06, + "loss": 0.2814, + "step": 30264 + }, + { + "epoch": 11.18, + "learning_rate": 1.8835447715235866e-06, + "loss": 0.4377, + "step": 30265 + }, + { + "epoch": 11.18, + "learning_rate": 1.8835078211602416e-06, + "loss": 0.2593, + "step": 30266 + }, + { + "epoch": 11.18, + "learning_rate": 1.8834708707968963e-06, + "loss": 0.3388, + "step": 30267 + }, + { + "epoch": 11.18, + "learning_rate": 1.8834339204335511e-06, + "loss": 0.312, + "step": 30268 + }, + { + "epoch": 11.18, + "learning_rate": 1.8833969700702058e-06, + "loss": 0.2862, + "step": 30269 + }, + { + "epoch": 11.18, + "learning_rate": 1.8833600197068604e-06, + "loss": 0.2948, + "step": 30270 + }, + { + "epoch": 11.18, + "learning_rate": 1.8833230693435152e-06, + "loss": 0.3479, + "step": 30271 + }, + { + "epoch": 11.18, + "learning_rate": 1.8832861189801699e-06, + "loss": 0.2056, + "step": 30272 + }, + { + "epoch": 11.18, + "learning_rate": 1.883249168616825e-06, + "loss": 0.2769, + "step": 30273 + }, + { + "epoch": 11.18, + "learning_rate": 1.8832122182534794e-06, + "loss": 0.2302, + "step": 30274 + }, + { + "epoch": 11.18, + "learning_rate": 1.8831752678901344e-06, + "loss": 0.2879, + "step": 30275 + }, + { + "epoch": 11.18, + "learning_rate": 1.883138317526789e-06, + "loss": 0.1974, + "step": 30276 + }, + { + "epoch": 11.18, + "learning_rate": 1.883101367163444e-06, + "loss": 0.2519, + "step": 30277 + }, + { + "epoch": 11.18, + "learning_rate": 1.8830644168000985e-06, + "loss": 0.3584, + "step": 30278 + }, + { + "epoch": 11.18, + "learning_rate": 1.8830274664367536e-06, + "loss": 0.2603, + "step": 30279 + }, + { + "epoch": 11.18, + "learning_rate": 1.882990516073408e-06, + "loss": 0.3387, + "step": 30280 + }, + { + "epoch": 11.18, + "learning_rate": 1.8829535657100627e-06, + "loss": 0.303, + "step": 30281 + }, + { + "epoch": 11.18, + "learning_rate": 1.8829166153467177e-06, + "loss": 0.2759, + "step": 30282 + }, + { + "epoch": 11.18, + "learning_rate": 1.8828796649833724e-06, + "loss": 0.4109, + "step": 30283 + }, + { + "epoch": 11.18, + "learning_rate": 1.8828427146200272e-06, + "loss": 0.3847, + "step": 30284 + }, + { + "epoch": 11.18, + "learning_rate": 1.8828057642566818e-06, + "loss": 0.2564, + "step": 30285 + }, + { + "epoch": 11.18, + "learning_rate": 1.8827688138933367e-06, + "loss": 0.2776, + "step": 30286 + }, + { + "epoch": 11.18, + "learning_rate": 1.8827318635299913e-06, + "loss": 0.4127, + "step": 30287 + }, + { + "epoch": 11.18, + "learning_rate": 1.8826949131666464e-06, + "loss": 0.3631, + "step": 30288 + }, + { + "epoch": 11.19, + "learning_rate": 1.882657962803301e-06, + "loss": 0.3377, + "step": 30289 + }, + { + "epoch": 11.19, + "learning_rate": 1.8826210124399554e-06, + "loss": 0.2936, + "step": 30290 + }, + { + "epoch": 11.19, + "learning_rate": 1.8825840620766105e-06, + "loss": 0.1385, + "step": 30291 + }, + { + "epoch": 11.19, + "learning_rate": 1.8825471117132651e-06, + "loss": 0.4489, + "step": 30292 + }, + { + "epoch": 11.19, + "learning_rate": 1.88251016134992e-06, + "loss": 0.2537, + "step": 30293 + }, + { + "epoch": 11.19, + "learning_rate": 1.8824732109865746e-06, + "loss": 0.3284, + "step": 30294 + }, + { + "epoch": 11.19, + "learning_rate": 1.8824362606232297e-06, + "loss": 0.453, + "step": 30295 + }, + { + "epoch": 11.19, + "learning_rate": 1.8823993102598841e-06, + "loss": 0.4309, + "step": 30296 + }, + { + "epoch": 11.19, + "learning_rate": 1.8823623598965392e-06, + "loss": 0.3355, + "step": 30297 + }, + { + "epoch": 11.19, + "learning_rate": 1.8823254095331938e-06, + "loss": 0.3138, + "step": 30298 + }, + { + "epoch": 11.19, + "learning_rate": 1.8822884591698487e-06, + "loss": 0.2439, + "step": 30299 + }, + { + "epoch": 11.19, + "learning_rate": 1.8822515088065033e-06, + "loss": 0.2769, + "step": 30300 + }, + { + "epoch": 11.19, + "learning_rate": 1.882214558443158e-06, + "loss": 0.3825, + "step": 30301 + }, + { + "epoch": 11.19, + "learning_rate": 1.8821776080798128e-06, + "loss": 0.2408, + "step": 30302 + }, + { + "epoch": 11.19, + "learning_rate": 1.8821406577164674e-06, + "loss": 0.2866, + "step": 30303 + }, + { + "epoch": 11.19, + "learning_rate": 1.8821037073531225e-06, + "loss": 0.4562, + "step": 30304 + }, + { + "epoch": 11.19, + "learning_rate": 1.8820667569897771e-06, + "loss": 0.3418, + "step": 30305 + }, + { + "epoch": 11.19, + "learning_rate": 1.882029806626432e-06, + "loss": 0.2777, + "step": 30306 + }, + { + "epoch": 11.19, + "learning_rate": 1.8819928562630866e-06, + "loss": 0.2779, + "step": 30307 + }, + { + "epoch": 11.19, + "learning_rate": 1.8819559058997414e-06, + "loss": 0.2765, + "step": 30308 + }, + { + "epoch": 11.19, + "learning_rate": 1.881918955536396e-06, + "loss": 0.5342, + "step": 30309 + }, + { + "epoch": 11.19, + "learning_rate": 1.8818820051730507e-06, + "loss": 0.1819, + "step": 30310 + }, + { + "epoch": 11.19, + "learning_rate": 1.8818450548097058e-06, + "loss": 0.3975, + "step": 30311 + }, + { + "epoch": 11.19, + "learning_rate": 1.8818081044463604e-06, + "loss": 0.2934, + "step": 30312 + }, + { + "epoch": 11.19, + "learning_rate": 1.8817711540830153e-06, + "loss": 0.1998, + "step": 30313 + }, + { + "epoch": 11.19, + "learning_rate": 1.88173420371967e-06, + "loss": 0.3081, + "step": 30314 + }, + { + "epoch": 11.19, + "learning_rate": 1.8816972533563247e-06, + "loss": 0.3376, + "step": 30315 + }, + { + "epoch": 11.19, + "learning_rate": 1.8816603029929794e-06, + "loss": 0.357, + "step": 30316 + }, + { + "epoch": 11.2, + "learning_rate": 1.8816233526296344e-06, + "loss": 0.2765, + "step": 30317 + }, + { + "epoch": 11.2, + "learning_rate": 1.8815864022662889e-06, + "loss": 0.3818, + "step": 30318 + }, + { + "epoch": 11.2, + "learning_rate": 1.881549451902944e-06, + "loss": 0.4226, + "step": 30319 + }, + { + "epoch": 11.2, + "learning_rate": 1.8815125015395986e-06, + "loss": 0.2744, + "step": 30320 + }, + { + "epoch": 11.2, + "learning_rate": 1.8814755511762532e-06, + "loss": 0.4106, + "step": 30321 + }, + { + "epoch": 11.2, + "learning_rate": 1.881438600812908e-06, + "loss": 0.2905, + "step": 30322 + }, + { + "epoch": 11.2, + "learning_rate": 1.8814016504495627e-06, + "loss": 0.4387, + "step": 30323 + }, + { + "epoch": 11.2, + "learning_rate": 1.8813647000862175e-06, + "loss": 0.3192, + "step": 30324 + }, + { + "epoch": 11.2, + "learning_rate": 1.8813277497228722e-06, + "loss": 0.3513, + "step": 30325 + }, + { + "epoch": 11.2, + "learning_rate": 1.8812907993595272e-06, + "loss": 0.2969, + "step": 30326 + }, + { + "epoch": 11.2, + "learning_rate": 1.8812538489961819e-06, + "loss": 0.3475, + "step": 30327 + }, + { + "epoch": 11.2, + "learning_rate": 1.8812168986328367e-06, + "loss": 0.2958, + "step": 30328 + }, + { + "epoch": 11.2, + "learning_rate": 1.8811799482694914e-06, + "loss": 0.2739, + "step": 30329 + }, + { + "epoch": 11.2, + "learning_rate": 1.881142997906146e-06, + "loss": 0.3622, + "step": 30330 + }, + { + "epoch": 11.2, + "learning_rate": 1.8811060475428008e-06, + "loss": 0.3481, + "step": 30331 + }, + { + "epoch": 11.2, + "learning_rate": 1.8810690971794555e-06, + "loss": 0.2887, + "step": 30332 + }, + { + "epoch": 11.2, + "learning_rate": 1.8810321468161105e-06, + "loss": 0.344, + "step": 30333 + }, + { + "epoch": 11.2, + "learning_rate": 1.8809951964527652e-06, + "loss": 0.2116, + "step": 30334 + }, + { + "epoch": 11.2, + "learning_rate": 1.88095824608942e-06, + "loss": 0.3788, + "step": 30335 + }, + { + "epoch": 11.2, + "learning_rate": 1.8809212957260747e-06, + "loss": 0.3126, + "step": 30336 + }, + { + "epoch": 11.2, + "learning_rate": 1.8808843453627295e-06, + "loss": 0.2578, + "step": 30337 + }, + { + "epoch": 11.2, + "learning_rate": 1.8808473949993841e-06, + "loss": 0.1902, + "step": 30338 + }, + { + "epoch": 11.2, + "learning_rate": 1.8808104446360392e-06, + "loss": 0.2938, + "step": 30339 + }, + { + "epoch": 11.2, + "learning_rate": 1.8807734942726938e-06, + "loss": 0.4348, + "step": 30340 + }, + { + "epoch": 11.2, + "learning_rate": 1.8807365439093483e-06, + "loss": 0.4904, + "step": 30341 + }, + { + "epoch": 11.2, + "learning_rate": 1.8806995935460033e-06, + "loss": 0.2446, + "step": 30342 + }, + { + "epoch": 11.2, + "learning_rate": 1.880662643182658e-06, + "loss": 0.3426, + "step": 30343 + }, + { + "epoch": 11.21, + "learning_rate": 1.8806256928193128e-06, + "loss": 0.306, + "step": 30344 + }, + { + "epoch": 11.21, + "learning_rate": 1.8805887424559674e-06, + "loss": 0.2685, + "step": 30345 + }, + { + "epoch": 11.21, + "learning_rate": 1.8805517920926225e-06, + "loss": 0.242, + "step": 30346 + }, + { + "epoch": 11.21, + "learning_rate": 1.880514841729277e-06, + "loss": 0.3287, + "step": 30347 + }, + { + "epoch": 11.21, + "learning_rate": 1.880477891365932e-06, + "loss": 0.3617, + "step": 30348 + }, + { + "epoch": 11.21, + "learning_rate": 1.8804409410025866e-06, + "loss": 0.2359, + "step": 30349 + }, + { + "epoch": 11.21, + "learning_rate": 1.8804039906392413e-06, + "loss": 0.2824, + "step": 30350 + }, + { + "epoch": 11.21, + "learning_rate": 1.8803670402758961e-06, + "loss": 0.2796, + "step": 30351 + }, + { + "epoch": 11.21, + "learning_rate": 1.8803300899125507e-06, + "loss": 0.2331, + "step": 30352 + }, + { + "epoch": 11.21, + "learning_rate": 1.8802931395492056e-06, + "loss": 0.1594, + "step": 30353 + }, + { + "epoch": 11.21, + "learning_rate": 1.8802561891858602e-06, + "loss": 0.4013, + "step": 30354 + }, + { + "epoch": 11.21, + "learning_rate": 1.8802192388225153e-06, + "loss": 0.2141, + "step": 30355 + }, + { + "epoch": 11.21, + "learning_rate": 1.88018228845917e-06, + "loss": 0.1923, + "step": 30356 + }, + { + "epoch": 11.21, + "learning_rate": 1.8801453380958248e-06, + "loss": 0.4086, + "step": 30357 + }, + { + "epoch": 11.21, + "learning_rate": 1.8801083877324794e-06, + "loss": 0.267, + "step": 30358 + }, + { + "epoch": 11.21, + "learning_rate": 1.8800714373691343e-06, + "loss": 0.4654, + "step": 30359 + }, + { + "epoch": 11.21, + "learning_rate": 1.8800344870057889e-06, + "loss": 0.2252, + "step": 30360 + }, + { + "epoch": 11.21, + "learning_rate": 1.8799975366424435e-06, + "loss": 0.2975, + "step": 30361 + }, + { + "epoch": 11.21, + "learning_rate": 1.8799605862790986e-06, + "loss": 0.2011, + "step": 30362 + }, + { + "epoch": 11.21, + "learning_rate": 1.879923635915753e-06, + "loss": 0.3212, + "step": 30363 + }, + { + "epoch": 11.21, + "learning_rate": 1.879886685552408e-06, + "loss": 0.3504, + "step": 30364 + }, + { + "epoch": 11.21, + "learning_rate": 1.8798497351890627e-06, + "loss": 0.3558, + "step": 30365 + }, + { + "epoch": 11.21, + "learning_rate": 1.8798127848257176e-06, + "loss": 0.3122, + "step": 30366 + }, + { + "epoch": 11.21, + "learning_rate": 1.8797758344623722e-06, + "loss": 0.365, + "step": 30367 + }, + { + "epoch": 11.21, + "learning_rate": 1.8797388840990273e-06, + "loss": 0.33, + "step": 30368 + }, + { + "epoch": 11.21, + "learning_rate": 1.8797019337356817e-06, + "loss": 0.3117, + "step": 30369 + }, + { + "epoch": 11.21, + "learning_rate": 1.8796649833723363e-06, + "loss": 0.4331, + "step": 30370 + }, + { + "epoch": 11.22, + "learning_rate": 1.8796280330089914e-06, + "loss": 0.3374, + "step": 30371 + }, + { + "epoch": 11.22, + "learning_rate": 1.879591082645646e-06, + "loss": 0.4802, + "step": 30372 + }, + { + "epoch": 11.22, + "learning_rate": 1.8795541322823009e-06, + "loss": 0.2431, + "step": 30373 + }, + { + "epoch": 11.22, + "learning_rate": 1.8795171819189555e-06, + "loss": 0.2371, + "step": 30374 + }, + { + "epoch": 11.22, + "learning_rate": 1.8794802315556103e-06, + "loss": 0.3236, + "step": 30375 + }, + { + "epoch": 11.22, + "learning_rate": 1.879443281192265e-06, + "loss": 0.2029, + "step": 30376 + }, + { + "epoch": 11.22, + "learning_rate": 1.87940633082892e-06, + "loss": 0.3601, + "step": 30377 + }, + { + "epoch": 11.22, + "learning_rate": 1.8793693804655747e-06, + "loss": 0.3588, + "step": 30378 + }, + { + "epoch": 11.22, + "learning_rate": 1.8793324301022295e-06, + "loss": 0.2493, + "step": 30379 + }, + { + "epoch": 11.22, + "learning_rate": 1.8792954797388842e-06, + "loss": 0.3593, + "step": 30380 + }, + { + "epoch": 11.22, + "learning_rate": 1.8792585293755388e-06, + "loss": 0.3173, + "step": 30381 + }, + { + "epoch": 11.22, + "learning_rate": 1.8792215790121936e-06, + "loss": 0.3955, + "step": 30382 + }, + { + "epoch": 11.22, + "learning_rate": 1.8791846286488483e-06, + "loss": 0.3185, + "step": 30383 + }, + { + "epoch": 11.22, + "learning_rate": 1.8791476782855033e-06, + "loss": 0.2668, + "step": 30384 + }, + { + "epoch": 11.22, + "learning_rate": 1.879110727922158e-06, + "loss": 0.3634, + "step": 30385 + }, + { + "epoch": 11.22, + "learning_rate": 1.8790737775588128e-06, + "loss": 0.2713, + "step": 30386 + }, + { + "epoch": 11.22, + "learning_rate": 1.8790368271954675e-06, + "loss": 0.4416, + "step": 30387 + }, + { + "epoch": 11.22, + "learning_rate": 1.8789998768321223e-06, + "loss": 0.2218, + "step": 30388 + }, + { + "epoch": 11.22, + "learning_rate": 1.878962926468777e-06, + "loss": 0.3078, + "step": 30389 + }, + { + "epoch": 11.22, + "learning_rate": 1.8789259761054316e-06, + "loss": 0.5857, + "step": 30390 + }, + { + "epoch": 11.22, + "learning_rate": 1.8788890257420866e-06, + "loss": 0.4204, + "step": 30391 + }, + { + "epoch": 11.22, + "learning_rate": 1.878852075378741e-06, + "loss": 0.2874, + "step": 30392 + }, + { + "epoch": 11.22, + "learning_rate": 1.8788151250153961e-06, + "loss": 0.3288, + "step": 30393 + }, + { + "epoch": 11.22, + "learning_rate": 1.8787781746520508e-06, + "loss": 0.2691, + "step": 30394 + }, + { + "epoch": 11.22, + "learning_rate": 1.8787412242887056e-06, + "loss": 0.2239, + "step": 30395 + }, + { + "epoch": 11.22, + "learning_rate": 1.8787042739253603e-06, + "loss": 0.3221, + "step": 30396 + }, + { + "epoch": 11.22, + "learning_rate": 1.878667323562015e-06, + "loss": 0.297, + "step": 30397 + }, + { + "epoch": 11.23, + "learning_rate": 1.8786303731986697e-06, + "loss": 0.4108, + "step": 30398 + }, + { + "epoch": 11.23, + "learning_rate": 1.8785934228353248e-06, + "loss": 0.4575, + "step": 30399 + }, + { + "epoch": 11.23, + "learning_rate": 1.8785564724719794e-06, + "loss": 0.3631, + "step": 30400 + }, + { + "epoch": 11.23, + "learning_rate": 1.878519522108634e-06, + "loss": 0.3475, + "step": 30401 + }, + { + "epoch": 11.23, + "learning_rate": 1.878482571745289e-06, + "loss": 0.338, + "step": 30402 + }, + { + "epoch": 11.23, + "learning_rate": 1.8784456213819436e-06, + "loss": 0.3921, + "step": 30403 + }, + { + "epoch": 11.23, + "learning_rate": 1.8784086710185984e-06, + "loss": 0.356, + "step": 30404 + }, + { + "epoch": 11.23, + "learning_rate": 1.878371720655253e-06, + "loss": 0.4243, + "step": 30405 + }, + { + "epoch": 11.23, + "learning_rate": 1.878334770291908e-06, + "loss": 0.2795, + "step": 30406 + }, + { + "epoch": 11.23, + "learning_rate": 1.8782978199285627e-06, + "loss": 0.2806, + "step": 30407 + }, + { + "epoch": 11.23, + "learning_rate": 1.8782608695652176e-06, + "loss": 0.2694, + "step": 30408 + }, + { + "epoch": 11.23, + "learning_rate": 1.8782239192018722e-06, + "loss": 0.3536, + "step": 30409 + }, + { + "epoch": 11.23, + "learning_rate": 1.8781869688385269e-06, + "loss": 0.2006, + "step": 30410 + }, + { + "epoch": 11.23, + "learning_rate": 1.8781500184751817e-06, + "loss": 0.3544, + "step": 30411 + }, + { + "epoch": 11.23, + "learning_rate": 1.8781130681118363e-06, + "loss": 0.2569, + "step": 30412 + }, + { + "epoch": 11.23, + "learning_rate": 1.8780761177484914e-06, + "loss": 0.3932, + "step": 30413 + }, + { + "epoch": 11.23, + "learning_rate": 1.8780391673851458e-06, + "loss": 0.2474, + "step": 30414 + }, + { + "epoch": 11.23, + "learning_rate": 1.8780022170218009e-06, + "loss": 0.4086, + "step": 30415 + }, + { + "epoch": 11.23, + "learning_rate": 1.8779652666584555e-06, + "loss": 0.2705, + "step": 30416 + }, + { + "epoch": 11.23, + "learning_rate": 1.8779283162951104e-06, + "loss": 0.2909, + "step": 30417 + }, + { + "epoch": 11.23, + "learning_rate": 1.877891365931765e-06, + "loss": 0.341, + "step": 30418 + }, + { + "epoch": 11.23, + "learning_rate": 1.87785441556842e-06, + "loss": 0.2737, + "step": 30419 + }, + { + "epoch": 11.23, + "learning_rate": 1.8778174652050745e-06, + "loss": 0.2945, + "step": 30420 + }, + { + "epoch": 11.23, + "learning_rate": 1.8777805148417291e-06, + "loss": 0.3396, + "step": 30421 + }, + { + "epoch": 11.23, + "learning_rate": 1.8777435644783842e-06, + "loss": 0.3478, + "step": 30422 + }, + { + "epoch": 11.23, + "learning_rate": 1.8777066141150388e-06, + "loss": 0.4246, + "step": 30423 + }, + { + "epoch": 11.23, + "learning_rate": 1.8776696637516937e-06, + "loss": 0.3227, + "step": 30424 + }, + { + "epoch": 11.24, + "learning_rate": 1.8776327133883483e-06, + "loss": 0.4342, + "step": 30425 + }, + { + "epoch": 11.24, + "learning_rate": 1.8775957630250032e-06, + "loss": 0.2685, + "step": 30426 + }, + { + "epoch": 11.24, + "learning_rate": 1.8775588126616578e-06, + "loss": 0.3789, + "step": 30427 + }, + { + "epoch": 11.24, + "learning_rate": 1.8775218622983129e-06, + "loss": 0.2793, + "step": 30428 + }, + { + "epoch": 11.24, + "learning_rate": 1.8774849119349675e-06, + "loss": 0.4149, + "step": 30429 + }, + { + "epoch": 11.24, + "learning_rate": 1.877447961571622e-06, + "loss": 0.3119, + "step": 30430 + }, + { + "epoch": 11.24, + "learning_rate": 1.877411011208277e-06, + "loss": 0.3631, + "step": 30431 + }, + { + "epoch": 11.24, + "learning_rate": 1.8773740608449316e-06, + "loss": 0.3834, + "step": 30432 + }, + { + "epoch": 11.24, + "learning_rate": 1.8773371104815865e-06, + "loss": 0.2255, + "step": 30433 + }, + { + "epoch": 11.24, + "learning_rate": 1.877300160118241e-06, + "loss": 0.3361, + "step": 30434 + }, + { + "epoch": 11.24, + "learning_rate": 1.8772632097548962e-06, + "loss": 0.3605, + "step": 30435 + }, + { + "epoch": 11.24, + "learning_rate": 1.8772262593915506e-06, + "loss": 0.3569, + "step": 30436 + }, + { + "epoch": 11.24, + "learning_rate": 1.8771893090282056e-06, + "loss": 0.4315, + "step": 30437 + }, + { + "epoch": 11.24, + "learning_rate": 1.8771523586648603e-06, + "loss": 0.3711, + "step": 30438 + }, + { + "epoch": 11.24, + "learning_rate": 1.877115408301515e-06, + "loss": 0.2993, + "step": 30439 + }, + { + "epoch": 11.24, + "learning_rate": 1.8770784579381698e-06, + "loss": 0.3137, + "step": 30440 + }, + { + "epoch": 11.24, + "learning_rate": 1.8770415075748244e-06, + "loss": 0.3002, + "step": 30441 + }, + { + "epoch": 11.24, + "learning_rate": 1.8770045572114792e-06, + "loss": 0.2731, + "step": 30442 + }, + { + "epoch": 11.24, + "learning_rate": 1.8769676068481339e-06, + "loss": 0.2528, + "step": 30443 + }, + { + "epoch": 11.24, + "learning_rate": 1.876930656484789e-06, + "loss": 0.2873, + "step": 30444 + }, + { + "epoch": 11.24, + "learning_rate": 1.8768937061214436e-06, + "loss": 0.6365, + "step": 30445 + }, + { + "epoch": 11.24, + "learning_rate": 1.8768567557580984e-06, + "loss": 0.3633, + "step": 30446 + }, + { + "epoch": 11.24, + "learning_rate": 1.876819805394753e-06, + "loss": 0.2646, + "step": 30447 + }, + { + "epoch": 11.24, + "learning_rate": 1.876782855031408e-06, + "loss": 0.4171, + "step": 30448 + }, + { + "epoch": 11.24, + "learning_rate": 1.8767459046680625e-06, + "loss": 0.2284, + "step": 30449 + }, + { + "epoch": 11.24, + "learning_rate": 1.8767089543047172e-06, + "loss": 0.2147, + "step": 30450 + }, + { + "epoch": 11.24, + "learning_rate": 1.8766720039413722e-06, + "loss": 0.2605, + "step": 30451 + }, + { + "epoch": 11.25, + "learning_rate": 1.8766350535780269e-06, + "loss": 0.3759, + "step": 30452 + }, + { + "epoch": 11.25, + "learning_rate": 1.8765981032146817e-06, + "loss": 0.3952, + "step": 30453 + }, + { + "epoch": 11.25, + "learning_rate": 1.8765611528513364e-06, + "loss": 0.4054, + "step": 30454 + }, + { + "epoch": 11.25, + "learning_rate": 1.8765242024879912e-06, + "loss": 0.2848, + "step": 30455 + }, + { + "epoch": 11.25, + "learning_rate": 1.8764872521246458e-06, + "loss": 0.347, + "step": 30456 + }, + { + "epoch": 11.25, + "learning_rate": 1.876450301761301e-06, + "loss": 0.3119, + "step": 30457 + }, + { + "epoch": 11.25, + "learning_rate": 1.8764133513979555e-06, + "loss": 0.3562, + "step": 30458 + }, + { + "epoch": 11.25, + "learning_rate": 1.87637640103461e-06, + "loss": 0.3302, + "step": 30459 + }, + { + "epoch": 11.25, + "learning_rate": 1.876339450671265e-06, + "loss": 0.2523, + "step": 30460 + }, + { + "epoch": 11.25, + "learning_rate": 1.8763025003079197e-06, + "loss": 0.3719, + "step": 30461 + }, + { + "epoch": 11.25, + "learning_rate": 1.8762655499445745e-06, + "loss": 0.2914, + "step": 30462 + }, + { + "epoch": 11.25, + "learning_rate": 1.8762285995812292e-06, + "loss": 0.3224, + "step": 30463 + }, + { + "epoch": 11.25, + "learning_rate": 1.8761916492178842e-06, + "loss": 0.3206, + "step": 30464 + }, + { + "epoch": 11.25, + "learning_rate": 1.8761546988545386e-06, + "loss": 0.332, + "step": 30465 + }, + { + "epoch": 11.25, + "learning_rate": 1.8761177484911937e-06, + "loss": 0.3005, + "step": 30466 + }, + { + "epoch": 11.25, + "learning_rate": 1.8760807981278483e-06, + "loss": 0.2798, + "step": 30467 + }, + { + "epoch": 11.25, + "learning_rate": 1.8760438477645032e-06, + "loss": 0.3854, + "step": 30468 + }, + { + "epoch": 11.25, + "learning_rate": 1.8760068974011578e-06, + "loss": 0.2245, + "step": 30469 + }, + { + "epoch": 11.25, + "learning_rate": 1.8759699470378125e-06, + "loss": 0.3377, + "step": 30470 + }, + { + "epoch": 11.25, + "learning_rate": 1.8759329966744673e-06, + "loss": 0.2016, + "step": 30471 + }, + { + "epoch": 11.25, + "learning_rate": 1.875896046311122e-06, + "loss": 0.3769, + "step": 30472 + }, + { + "epoch": 11.25, + "learning_rate": 1.875859095947777e-06, + "loss": 0.2266, + "step": 30473 + }, + { + "epoch": 11.25, + "learning_rate": 1.8758221455844316e-06, + "loss": 0.2692, + "step": 30474 + }, + { + "epoch": 11.25, + "learning_rate": 1.8757851952210865e-06, + "loss": 0.1865, + "step": 30475 + }, + { + "epoch": 11.25, + "learning_rate": 1.8757482448577411e-06, + "loss": 0.3484, + "step": 30476 + }, + { + "epoch": 11.25, + "learning_rate": 1.875711294494396e-06, + "loss": 0.1969, + "step": 30477 + }, + { + "epoch": 11.25, + "learning_rate": 1.8756743441310506e-06, + "loss": 0.2868, + "step": 30478 + }, + { + "epoch": 11.26, + "learning_rate": 1.8756373937677052e-06, + "loss": 0.2624, + "step": 30479 + }, + { + "epoch": 11.26, + "learning_rate": 1.8756004434043603e-06, + "loss": 0.2736, + "step": 30480 + }, + { + "epoch": 11.26, + "learning_rate": 1.8755634930410147e-06, + "loss": 0.3269, + "step": 30481 + }, + { + "epoch": 11.26, + "learning_rate": 1.8755265426776698e-06, + "loss": 0.3473, + "step": 30482 + }, + { + "epoch": 11.26, + "learning_rate": 1.8754895923143244e-06, + "loss": 0.3367, + "step": 30483 + }, + { + "epoch": 11.26, + "learning_rate": 1.8754526419509793e-06, + "loss": 0.3387, + "step": 30484 + }, + { + "epoch": 11.26, + "learning_rate": 1.875415691587634e-06, + "loss": 0.2833, + "step": 30485 + }, + { + "epoch": 11.26, + "learning_rate": 1.875378741224289e-06, + "loss": 0.2822, + "step": 30486 + }, + { + "epoch": 11.26, + "learning_rate": 1.8753417908609434e-06, + "loss": 0.2978, + "step": 30487 + }, + { + "epoch": 11.26, + "learning_rate": 1.8753048404975984e-06, + "loss": 0.308, + "step": 30488 + }, + { + "epoch": 11.26, + "learning_rate": 1.875267890134253e-06, + "loss": 0.2467, + "step": 30489 + }, + { + "epoch": 11.26, + "learning_rate": 1.8752309397709077e-06, + "loss": 0.3166, + "step": 30490 + }, + { + "epoch": 11.26, + "learning_rate": 1.8751939894075626e-06, + "loss": 0.2916, + "step": 30491 + }, + { + "epoch": 11.26, + "learning_rate": 1.8751570390442172e-06, + "loss": 0.2795, + "step": 30492 + }, + { + "epoch": 11.26, + "learning_rate": 1.875120088680872e-06, + "loss": 0.3639, + "step": 30493 + }, + { + "epoch": 11.26, + "learning_rate": 1.8750831383175267e-06, + "loss": 0.3449, + "step": 30494 + }, + { + "epoch": 11.26, + "learning_rate": 1.8750461879541818e-06, + "loss": 0.2261, + "step": 30495 + }, + { + "epoch": 11.26, + "learning_rate": 1.8750092375908364e-06, + "loss": 0.313, + "step": 30496 + }, + { + "epoch": 11.26, + "learning_rate": 1.8749722872274912e-06, + "loss": 0.4241, + "step": 30497 + }, + { + "epoch": 11.26, + "learning_rate": 1.8749353368641459e-06, + "loss": 0.2518, + "step": 30498 + }, + { + "epoch": 11.26, + "learning_rate": 1.8748983865008005e-06, + "loss": 0.3358, + "step": 30499 + }, + { + "epoch": 11.26, + "learning_rate": 1.8748614361374554e-06, + "loss": 0.39, + "step": 30500 + }, + { + "epoch": 11.26, + "learning_rate": 1.87482448577411e-06, + "loss": 0.346, + "step": 30501 + }, + { + "epoch": 11.26, + "learning_rate": 1.874787535410765e-06, + "loss": 0.3011, + "step": 30502 + }, + { + "epoch": 11.26, + "learning_rate": 1.8747505850474197e-06, + "loss": 0.3139, + "step": 30503 + }, + { + "epoch": 11.26, + "learning_rate": 1.8747136346840745e-06, + "loss": 0.2182, + "step": 30504 + }, + { + "epoch": 11.26, + "learning_rate": 1.8746766843207292e-06, + "loss": 0.3472, + "step": 30505 + }, + { + "epoch": 11.27, + "learning_rate": 1.874639733957384e-06, + "loss": 0.3513, + "step": 30506 + }, + { + "epoch": 11.27, + "learning_rate": 1.8746027835940387e-06, + "loss": 0.4277, + "step": 30507 + }, + { + "epoch": 11.27, + "learning_rate": 1.8745658332306937e-06, + "loss": 0.2748, + "step": 30508 + }, + { + "epoch": 11.27, + "learning_rate": 1.8745288828673481e-06, + "loss": 0.3114, + "step": 30509 + }, + { + "epoch": 11.27, + "learning_rate": 1.8744919325040028e-06, + "loss": 0.3447, + "step": 30510 + }, + { + "epoch": 11.27, + "learning_rate": 1.8744549821406578e-06, + "loss": 0.3671, + "step": 30511 + }, + { + "epoch": 11.27, + "learning_rate": 1.8744180317773125e-06, + "loss": 0.4016, + "step": 30512 + }, + { + "epoch": 11.27, + "learning_rate": 1.8743810814139673e-06, + "loss": 0.263, + "step": 30513 + }, + { + "epoch": 11.27, + "learning_rate": 1.874344131050622e-06, + "loss": 0.3752, + "step": 30514 + }, + { + "epoch": 11.27, + "learning_rate": 1.8743071806872768e-06, + "loss": 0.2908, + "step": 30515 + }, + { + "epoch": 11.27, + "learning_rate": 1.8742702303239314e-06, + "loss": 0.3258, + "step": 30516 + }, + { + "epoch": 11.27, + "learning_rate": 1.8742332799605865e-06, + "loss": 0.5228, + "step": 30517 + }, + { + "epoch": 11.27, + "learning_rate": 1.8741963295972411e-06, + "loss": 0.3071, + "step": 30518 + }, + { + "epoch": 11.27, + "learning_rate": 1.8741593792338958e-06, + "loss": 0.367, + "step": 30519 + }, + { + "epoch": 11.27, + "learning_rate": 1.8741224288705506e-06, + "loss": 0.2606, + "step": 30520 + }, + { + "epoch": 11.27, + "learning_rate": 1.8740854785072053e-06, + "loss": 0.3348, + "step": 30521 + }, + { + "epoch": 11.27, + "learning_rate": 1.8740485281438601e-06, + "loss": 0.359, + "step": 30522 + }, + { + "epoch": 11.27, + "learning_rate": 1.8740115777805147e-06, + "loss": 0.3863, + "step": 30523 + }, + { + "epoch": 11.27, + "learning_rate": 1.8739746274171698e-06, + "loss": 0.2658, + "step": 30524 + }, + { + "epoch": 11.27, + "learning_rate": 1.8739376770538244e-06, + "loss": 0.4337, + "step": 30525 + }, + { + "epoch": 11.27, + "learning_rate": 1.8739007266904793e-06, + "loss": 0.2603, + "step": 30526 + }, + { + "epoch": 11.27, + "learning_rate": 1.873863776327134e-06, + "loss": 0.3701, + "step": 30527 + }, + { + "epoch": 11.27, + "learning_rate": 1.8738268259637888e-06, + "loss": 0.3023, + "step": 30528 + }, + { + "epoch": 11.27, + "learning_rate": 1.8737898756004434e-06, + "loss": 0.3113, + "step": 30529 + }, + { + "epoch": 11.27, + "learning_rate": 1.873752925237098e-06, + "loss": 0.3529, + "step": 30530 + }, + { + "epoch": 11.27, + "learning_rate": 1.8737159748737531e-06, + "loss": 0.2595, + "step": 30531 + }, + { + "epoch": 11.27, + "learning_rate": 1.8736790245104075e-06, + "loss": 0.2932, + "step": 30532 + }, + { + "epoch": 11.28, + "learning_rate": 1.8736420741470626e-06, + "loss": 0.3918, + "step": 30533 + }, + { + "epoch": 11.28, + "learning_rate": 1.8736051237837172e-06, + "loss": 0.2878, + "step": 30534 + }, + { + "epoch": 11.28, + "learning_rate": 1.873568173420372e-06, + "loss": 0.2292, + "step": 30535 + }, + { + "epoch": 11.28, + "learning_rate": 1.8735312230570267e-06, + "loss": 0.3194, + "step": 30536 + }, + { + "epoch": 11.28, + "learning_rate": 1.8734942726936818e-06, + "loss": 0.2295, + "step": 30537 + }, + { + "epoch": 11.28, + "learning_rate": 1.8734573223303362e-06, + "loss": 0.4704, + "step": 30538 + }, + { + "epoch": 11.28, + "learning_rate": 1.8734203719669908e-06, + "loss": 0.3549, + "step": 30539 + }, + { + "epoch": 11.28, + "learning_rate": 1.873383421603646e-06, + "loss": 0.411, + "step": 30540 + }, + { + "epoch": 11.28, + "learning_rate": 1.8733464712403005e-06, + "loss": 0.2116, + "step": 30541 + }, + { + "epoch": 11.28, + "learning_rate": 1.8733095208769554e-06, + "loss": 0.2705, + "step": 30542 + }, + { + "epoch": 11.28, + "learning_rate": 1.87327257051361e-06, + "loss": 0.2671, + "step": 30543 + }, + { + "epoch": 11.28, + "learning_rate": 1.8732356201502649e-06, + "loss": 0.2624, + "step": 30544 + }, + { + "epoch": 11.28, + "learning_rate": 1.8731986697869195e-06, + "loss": 0.2463, + "step": 30545 + }, + { + "epoch": 11.28, + "learning_rate": 1.8731617194235746e-06, + "loss": 0.3159, + "step": 30546 + }, + { + "epoch": 11.28, + "learning_rate": 1.8731247690602292e-06, + "loss": 0.276, + "step": 30547 + }, + { + "epoch": 11.28, + "learning_rate": 1.873087818696884e-06, + "loss": 0.3582, + "step": 30548 + }, + { + "epoch": 11.28, + "learning_rate": 1.8730508683335387e-06, + "loss": 0.3759, + "step": 30549 + }, + { + "epoch": 11.28, + "learning_rate": 1.8730139179701933e-06, + "loss": 0.2164, + "step": 30550 + }, + { + "epoch": 11.28, + "learning_rate": 1.8729769676068482e-06, + "loss": 0.3141, + "step": 30551 + }, + { + "epoch": 11.28, + "learning_rate": 1.8729400172435028e-06, + "loss": 0.3308, + "step": 30552 + }, + { + "epoch": 11.28, + "learning_rate": 1.8729030668801579e-06, + "loss": 0.3297, + "step": 30553 + }, + { + "epoch": 11.28, + "learning_rate": 1.8728661165168123e-06, + "loss": 0.2148, + "step": 30554 + }, + { + "epoch": 11.28, + "learning_rate": 1.8728291661534673e-06, + "loss": 0.324, + "step": 30555 + }, + { + "epoch": 11.28, + "learning_rate": 1.872792215790122e-06, + "loss": 0.3032, + "step": 30556 + }, + { + "epoch": 11.28, + "learning_rate": 1.8727552654267768e-06, + "loss": 0.2664, + "step": 30557 + }, + { + "epoch": 11.28, + "learning_rate": 1.8727183150634315e-06, + "loss": 0.2688, + "step": 30558 + }, + { + "epoch": 11.28, + "learning_rate": 1.8726813647000861e-06, + "loss": 0.4893, + "step": 30559 + }, + { + "epoch": 11.29, + "learning_rate": 1.872644414336741e-06, + "loss": 0.3052, + "step": 30560 + }, + { + "epoch": 11.29, + "learning_rate": 1.8726074639733956e-06, + "loss": 0.4311, + "step": 30561 + }, + { + "epoch": 11.29, + "learning_rate": 1.8725705136100507e-06, + "loss": 0.3912, + "step": 30562 + }, + { + "epoch": 11.29, + "learning_rate": 1.8725335632467053e-06, + "loss": 0.2374, + "step": 30563 + }, + { + "epoch": 11.29, + "learning_rate": 1.8724966128833601e-06, + "loss": 0.3664, + "step": 30564 + }, + { + "epoch": 11.29, + "learning_rate": 1.8724596625200148e-06, + "loss": 0.3267, + "step": 30565 + }, + { + "epoch": 11.29, + "learning_rate": 1.8724227121566696e-06, + "loss": 0.3014, + "step": 30566 + }, + { + "epoch": 11.29, + "learning_rate": 1.8723857617933243e-06, + "loss": 0.4912, + "step": 30567 + }, + { + "epoch": 11.29, + "learning_rate": 1.8723488114299793e-06, + "loss": 0.3606, + "step": 30568 + }, + { + "epoch": 11.29, + "learning_rate": 1.872311861066634e-06, + "loss": 0.451, + "step": 30569 + }, + { + "epoch": 11.29, + "learning_rate": 1.8722749107032886e-06, + "loss": 0.2659, + "step": 30570 + }, + { + "epoch": 11.29, + "learning_rate": 1.8722379603399434e-06, + "loss": 0.2466, + "step": 30571 + }, + { + "epoch": 11.29, + "learning_rate": 1.872201009976598e-06, + "loss": 0.4133, + "step": 30572 + }, + { + "epoch": 11.29, + "learning_rate": 1.872164059613253e-06, + "loss": 0.367, + "step": 30573 + }, + { + "epoch": 11.29, + "learning_rate": 1.8721271092499076e-06, + "loss": 0.3865, + "step": 30574 + }, + { + "epoch": 11.29, + "learning_rate": 1.8720901588865626e-06, + "loss": 0.3427, + "step": 30575 + }, + { + "epoch": 11.29, + "learning_rate": 1.8720532085232173e-06, + "loss": 0.3335, + "step": 30576 + }, + { + "epoch": 11.29, + "learning_rate": 1.872016258159872e-06, + "loss": 0.3296, + "step": 30577 + }, + { + "epoch": 11.29, + "learning_rate": 1.8719793077965267e-06, + "loss": 0.311, + "step": 30578 + }, + { + "epoch": 11.29, + "learning_rate": 1.8719423574331814e-06, + "loss": 0.382, + "step": 30579 + }, + { + "epoch": 11.29, + "learning_rate": 1.8719054070698362e-06, + "loss": 0.3532, + "step": 30580 + }, + { + "epoch": 11.29, + "learning_rate": 1.8718684567064909e-06, + "loss": 0.2244, + "step": 30581 + }, + { + "epoch": 11.29, + "learning_rate": 1.8718315063431457e-06, + "loss": 0.2075, + "step": 30582 + }, + { + "epoch": 11.29, + "learning_rate": 1.8717945559798003e-06, + "loss": 0.3756, + "step": 30583 + }, + { + "epoch": 11.29, + "learning_rate": 1.8717576056164554e-06, + "loss": 0.2383, + "step": 30584 + }, + { + "epoch": 11.29, + "learning_rate": 1.87172065525311e-06, + "loss": 0.2737, + "step": 30585 + }, + { + "epoch": 11.29, + "learning_rate": 1.8716837048897649e-06, + "loss": 0.2263, + "step": 30586 + }, + { + "epoch": 11.3, + "learning_rate": 1.8716467545264195e-06, + "loss": 0.3978, + "step": 30587 + }, + { + "epoch": 11.3, + "learning_rate": 1.8716098041630744e-06, + "loss": 0.46, + "step": 30588 + }, + { + "epoch": 11.3, + "learning_rate": 1.871572853799729e-06, + "loss": 0.4397, + "step": 30589 + }, + { + "epoch": 11.3, + "learning_rate": 1.8715359034363836e-06, + "loss": 0.3703, + "step": 30590 + }, + { + "epoch": 11.3, + "learning_rate": 1.8714989530730387e-06, + "loss": 0.2883, + "step": 30591 + }, + { + "epoch": 11.3, + "learning_rate": 1.8714620027096933e-06, + "loss": 0.2925, + "step": 30592 + }, + { + "epoch": 11.3, + "learning_rate": 1.8714250523463482e-06, + "loss": 0.2749, + "step": 30593 + }, + { + "epoch": 11.3, + "learning_rate": 1.8713881019830028e-06, + "loss": 0.2542, + "step": 30594 + }, + { + "epoch": 11.3, + "learning_rate": 1.8713511516196577e-06, + "loss": 0.3835, + "step": 30595 + }, + { + "epoch": 11.3, + "learning_rate": 1.8713142012563123e-06, + "loss": 0.2959, + "step": 30596 + }, + { + "epoch": 11.3, + "learning_rate": 1.8712772508929674e-06, + "loss": 0.3142, + "step": 30597 + }, + { + "epoch": 11.3, + "learning_rate": 1.871240300529622e-06, + "loss": 0.3921, + "step": 30598 + }, + { + "epoch": 11.3, + "learning_rate": 1.8712033501662764e-06, + "loss": 0.4371, + "step": 30599 + }, + { + "epoch": 11.3, + "learning_rate": 1.8711663998029315e-06, + "loss": 0.3946, + "step": 30600 + }, + { + "epoch": 11.3, + "learning_rate": 1.8711294494395861e-06, + "loss": 0.3161, + "step": 30601 + }, + { + "epoch": 11.3, + "learning_rate": 1.871092499076241e-06, + "loss": 0.5026, + "step": 30602 + }, + { + "epoch": 11.3, + "learning_rate": 1.8710555487128956e-06, + "loss": 0.3905, + "step": 30603 + }, + { + "epoch": 11.3, + "learning_rate": 1.8710185983495507e-06, + "loss": 0.2889, + "step": 30604 + }, + { + "epoch": 11.3, + "learning_rate": 1.870981647986205e-06, + "loss": 0.4938, + "step": 30605 + }, + { + "epoch": 11.3, + "learning_rate": 1.8709446976228602e-06, + "loss": 0.2927, + "step": 30606 + }, + { + "epoch": 11.3, + "learning_rate": 1.8709077472595148e-06, + "loss": 0.2415, + "step": 30607 + }, + { + "epoch": 11.3, + "learning_rate": 1.8708707968961696e-06, + "loss": 0.2791, + "step": 30608 + }, + { + "epoch": 11.3, + "learning_rate": 1.8708338465328243e-06, + "loss": 0.27, + "step": 30609 + }, + { + "epoch": 11.3, + "learning_rate": 1.870796896169479e-06, + "loss": 0.2802, + "step": 30610 + }, + { + "epoch": 11.3, + "learning_rate": 1.8707599458061338e-06, + "loss": 0.3138, + "step": 30611 + }, + { + "epoch": 11.3, + "learning_rate": 1.8707229954427884e-06, + "loss": 0.3012, + "step": 30612 + }, + { + "epoch": 11.3, + "learning_rate": 1.8706860450794435e-06, + "loss": 0.4663, + "step": 30613 + }, + { + "epoch": 11.31, + "learning_rate": 1.870649094716098e-06, + "loss": 0.3296, + "step": 30614 + }, + { + "epoch": 11.31, + "learning_rate": 1.870612144352753e-06, + "loss": 0.2418, + "step": 30615 + }, + { + "epoch": 11.31, + "learning_rate": 1.8705751939894076e-06, + "loss": 0.4079, + "step": 30616 + }, + { + "epoch": 11.31, + "learning_rate": 1.8705382436260624e-06, + "loss": 0.3292, + "step": 30617 + }, + { + "epoch": 11.31, + "learning_rate": 1.870501293262717e-06, + "loss": 0.4398, + "step": 30618 + }, + { + "epoch": 11.31, + "learning_rate": 1.8704643428993717e-06, + "loss": 0.3134, + "step": 30619 + }, + { + "epoch": 11.31, + "learning_rate": 1.8704273925360268e-06, + "loss": 0.3057, + "step": 30620 + }, + { + "epoch": 11.31, + "learning_rate": 1.8703904421726812e-06, + "loss": 0.3531, + "step": 30621 + }, + { + "epoch": 11.31, + "learning_rate": 1.8703534918093362e-06, + "loss": 0.3834, + "step": 30622 + }, + { + "epoch": 11.31, + "learning_rate": 1.8703165414459909e-06, + "loss": 0.4608, + "step": 30623 + }, + { + "epoch": 11.31, + "learning_rate": 1.8702795910826457e-06, + "loss": 0.3833, + "step": 30624 + }, + { + "epoch": 11.31, + "learning_rate": 1.8702426407193004e-06, + "loss": 0.3891, + "step": 30625 + }, + { + "epoch": 11.31, + "learning_rate": 1.8702056903559554e-06, + "loss": 0.308, + "step": 30626 + }, + { + "epoch": 11.31, + "learning_rate": 1.8701687399926099e-06, + "loss": 0.4139, + "step": 30627 + }, + { + "epoch": 11.31, + "learning_rate": 1.870131789629265e-06, + "loss": 0.2328, + "step": 30628 + }, + { + "epoch": 11.31, + "learning_rate": 1.8700948392659196e-06, + "loss": 0.3455, + "step": 30629 + }, + { + "epoch": 11.31, + "learning_rate": 1.8700578889025742e-06, + "loss": 0.2441, + "step": 30630 + }, + { + "epoch": 11.31, + "learning_rate": 1.870020938539229e-06, + "loss": 0.3316, + "step": 30631 + }, + { + "epoch": 11.31, + "learning_rate": 1.8699839881758837e-06, + "loss": 0.3505, + "step": 30632 + }, + { + "epoch": 11.31, + "learning_rate": 1.8699470378125385e-06, + "loss": 0.4215, + "step": 30633 + }, + { + "epoch": 11.31, + "learning_rate": 1.8699100874491932e-06, + "loss": 0.3036, + "step": 30634 + }, + { + "epoch": 11.31, + "learning_rate": 1.8698731370858482e-06, + "loss": 0.4494, + "step": 30635 + }, + { + "epoch": 11.31, + "learning_rate": 1.8698361867225029e-06, + "loss": 0.401, + "step": 30636 + }, + { + "epoch": 11.31, + "learning_rate": 1.8697992363591577e-06, + "loss": 0.2678, + "step": 30637 + }, + { + "epoch": 11.31, + "learning_rate": 1.8697622859958123e-06, + "loss": 0.2837, + "step": 30638 + }, + { + "epoch": 11.31, + "learning_rate": 1.869725335632467e-06, + "loss": 0.3265, + "step": 30639 + }, + { + "epoch": 11.31, + "learning_rate": 1.8696883852691218e-06, + "loss": 0.3102, + "step": 30640 + }, + { + "epoch": 11.31, + "learning_rate": 1.8696514349057765e-06, + "loss": 0.2912, + "step": 30641 + }, + { + "epoch": 11.32, + "learning_rate": 1.8696144845424315e-06, + "loss": 0.3551, + "step": 30642 + }, + { + "epoch": 11.32, + "learning_rate": 1.8695775341790862e-06, + "loss": 0.3605, + "step": 30643 + }, + { + "epoch": 11.32, + "learning_rate": 1.869540583815741e-06, + "loss": 0.2985, + "step": 30644 + }, + { + "epoch": 11.32, + "learning_rate": 1.8695036334523956e-06, + "loss": 0.2849, + "step": 30645 + }, + { + "epoch": 11.32, + "learning_rate": 1.8694666830890505e-06, + "loss": 0.2353, + "step": 30646 + }, + { + "epoch": 11.32, + "learning_rate": 1.8694297327257051e-06, + "loss": 0.1922, + "step": 30647 + }, + { + "epoch": 11.32, + "learning_rate": 1.8693927823623602e-06, + "loss": 0.3683, + "step": 30648 + }, + { + "epoch": 11.32, + "learning_rate": 1.8693558319990148e-06, + "loss": 0.3167, + "step": 30649 + }, + { + "epoch": 11.32, + "learning_rate": 1.8693188816356692e-06, + "loss": 0.3353, + "step": 30650 + }, + { + "epoch": 11.32, + "learning_rate": 1.8692819312723243e-06, + "loss": 0.3204, + "step": 30651 + }, + { + "epoch": 11.32, + "learning_rate": 1.869244980908979e-06, + "loss": 0.4346, + "step": 30652 + }, + { + "epoch": 11.32, + "learning_rate": 1.8692080305456338e-06, + "loss": 0.4436, + "step": 30653 + }, + { + "epoch": 11.32, + "learning_rate": 1.8691710801822884e-06, + "loss": 0.2285, + "step": 30654 + }, + { + "epoch": 11.32, + "learning_rate": 1.8691341298189435e-06, + "loss": 0.3327, + "step": 30655 + }, + { + "epoch": 11.32, + "learning_rate": 1.869097179455598e-06, + "loss": 0.3065, + "step": 30656 + }, + { + "epoch": 11.32, + "learning_rate": 1.869060229092253e-06, + "loss": 0.4752, + "step": 30657 + }, + { + "epoch": 11.32, + "learning_rate": 1.8690232787289076e-06, + "loss": 0.3437, + "step": 30658 + }, + { + "epoch": 11.32, + "learning_rate": 1.8689863283655622e-06, + "loss": 0.4105, + "step": 30659 + }, + { + "epoch": 11.32, + "learning_rate": 1.868949378002217e-06, + "loss": 0.3184, + "step": 30660 + }, + { + "epoch": 11.32, + "learning_rate": 1.8689124276388717e-06, + "loss": 0.2909, + "step": 30661 + }, + { + "epoch": 11.32, + "learning_rate": 1.8688754772755266e-06, + "loss": 0.29, + "step": 30662 + }, + { + "epoch": 11.32, + "learning_rate": 1.8688385269121812e-06, + "loss": 0.3637, + "step": 30663 + }, + { + "epoch": 11.32, + "learning_rate": 1.8688015765488363e-06, + "loss": 0.4277, + "step": 30664 + }, + { + "epoch": 11.32, + "learning_rate": 1.868764626185491e-06, + "loss": 0.4541, + "step": 30665 + }, + { + "epoch": 11.32, + "learning_rate": 1.8687276758221458e-06, + "loss": 0.4907, + "step": 30666 + }, + { + "epoch": 11.32, + "learning_rate": 1.8686907254588004e-06, + "loss": 0.2443, + "step": 30667 + }, + { + "epoch": 11.32, + "learning_rate": 1.8686537750954552e-06, + "loss": 0.3254, + "step": 30668 + }, + { + "epoch": 11.33, + "learning_rate": 1.8686168247321099e-06, + "loss": 0.3708, + "step": 30669 + }, + { + "epoch": 11.33, + "learning_rate": 1.8685798743687645e-06, + "loss": 0.3156, + "step": 30670 + }, + { + "epoch": 11.33, + "learning_rate": 1.8685429240054196e-06, + "loss": 0.33, + "step": 30671 + }, + { + "epoch": 11.33, + "learning_rate": 1.868505973642074e-06, + "loss": 0.2257, + "step": 30672 + }, + { + "epoch": 11.33, + "learning_rate": 1.868469023278729e-06, + "loss": 0.299, + "step": 30673 + }, + { + "epoch": 11.33, + "learning_rate": 1.8684320729153837e-06, + "loss": 0.2707, + "step": 30674 + }, + { + "epoch": 11.33, + "learning_rate": 1.8683951225520385e-06, + "loss": 0.4883, + "step": 30675 + }, + { + "epoch": 11.33, + "learning_rate": 1.8683581721886932e-06, + "loss": 0.3901, + "step": 30676 + }, + { + "epoch": 11.33, + "learning_rate": 1.8683212218253482e-06, + "loss": 0.2697, + "step": 30677 + }, + { + "epoch": 11.33, + "learning_rate": 1.8682842714620027e-06, + "loss": 0.2686, + "step": 30678 + }, + { + "epoch": 11.33, + "learning_rate": 1.8682473210986573e-06, + "loss": 0.3243, + "step": 30679 + }, + { + "epoch": 11.33, + "learning_rate": 1.8682103707353124e-06, + "loss": 0.3014, + "step": 30680 + }, + { + "epoch": 11.33, + "learning_rate": 1.868173420371967e-06, + "loss": 0.2668, + "step": 30681 + }, + { + "epoch": 11.33, + "learning_rate": 1.8681364700086218e-06, + "loss": 0.3207, + "step": 30682 + }, + { + "epoch": 11.33, + "learning_rate": 1.8680995196452765e-06, + "loss": 0.2113, + "step": 30683 + }, + { + "epoch": 11.33, + "learning_rate": 1.8680625692819313e-06, + "loss": 0.4191, + "step": 30684 + }, + { + "epoch": 11.33, + "learning_rate": 1.868025618918586e-06, + "loss": 0.3027, + "step": 30685 + }, + { + "epoch": 11.33, + "learning_rate": 1.867988668555241e-06, + "loss": 0.2106, + "step": 30686 + }, + { + "epoch": 11.33, + "learning_rate": 1.8679517181918957e-06, + "loss": 0.2823, + "step": 30687 + }, + { + "epoch": 11.33, + "learning_rate": 1.8679147678285503e-06, + "loss": 0.2024, + "step": 30688 + }, + { + "epoch": 11.33, + "learning_rate": 1.8678778174652051e-06, + "loss": 0.4519, + "step": 30689 + }, + { + "epoch": 11.33, + "learning_rate": 1.8678408671018598e-06, + "loss": 0.3067, + "step": 30690 + }, + { + "epoch": 11.33, + "learning_rate": 1.8678039167385146e-06, + "loss": 0.3326, + "step": 30691 + }, + { + "epoch": 11.33, + "learning_rate": 1.8677669663751693e-06, + "loss": 0.2379, + "step": 30692 + }, + { + "epoch": 11.33, + "learning_rate": 1.8677300160118243e-06, + "loss": 0.3485, + "step": 30693 + }, + { + "epoch": 11.33, + "learning_rate": 1.8676930656484788e-06, + "loss": 0.3903, + "step": 30694 + }, + { + "epoch": 11.33, + "learning_rate": 1.8676561152851338e-06, + "loss": 0.3042, + "step": 30695 + }, + { + "epoch": 11.34, + "learning_rate": 1.8676191649217885e-06, + "loss": 0.3903, + "step": 30696 + }, + { + "epoch": 11.34, + "learning_rate": 1.8675822145584433e-06, + "loss": 0.2888, + "step": 30697 + }, + { + "epoch": 11.34, + "learning_rate": 1.867545264195098e-06, + "loss": 0.3478, + "step": 30698 + }, + { + "epoch": 11.34, + "learning_rate": 1.8675083138317526e-06, + "loss": 0.2641, + "step": 30699 + }, + { + "epoch": 11.34, + "learning_rate": 1.8674713634684074e-06, + "loss": 0.4023, + "step": 30700 + }, + { + "epoch": 11.34, + "learning_rate": 1.867434413105062e-06, + "loss": 0.3224, + "step": 30701 + }, + { + "epoch": 11.34, + "learning_rate": 1.8673974627417171e-06, + "loss": 0.3175, + "step": 30702 + }, + { + "epoch": 11.34, + "learning_rate": 1.8673605123783718e-06, + "loss": 0.3156, + "step": 30703 + }, + { + "epoch": 11.34, + "learning_rate": 1.8673235620150266e-06, + "loss": 0.2348, + "step": 30704 + }, + { + "epoch": 11.34, + "learning_rate": 1.8672866116516812e-06, + "loss": 0.2517, + "step": 30705 + }, + { + "epoch": 11.34, + "learning_rate": 1.867249661288336e-06, + "loss": 0.4621, + "step": 30706 + }, + { + "epoch": 11.34, + "learning_rate": 1.8672127109249907e-06, + "loss": 0.1947, + "step": 30707 + }, + { + "epoch": 11.34, + "learning_rate": 1.8671757605616454e-06, + "loss": 0.3523, + "step": 30708 + }, + { + "epoch": 11.34, + "learning_rate": 1.8671388101983004e-06, + "loss": 0.264, + "step": 30709 + }, + { + "epoch": 11.34, + "learning_rate": 1.867101859834955e-06, + "loss": 0.2711, + "step": 30710 + }, + { + "epoch": 11.34, + "learning_rate": 1.86706490947161e-06, + "loss": 0.2602, + "step": 30711 + }, + { + "epoch": 11.34, + "learning_rate": 1.8670279591082645e-06, + "loss": 0.3926, + "step": 30712 + }, + { + "epoch": 11.34, + "learning_rate": 1.8669910087449194e-06, + "loss": 0.4306, + "step": 30713 + }, + { + "epoch": 11.34, + "learning_rate": 1.866954058381574e-06, + "loss": 0.2633, + "step": 30714 + }, + { + "epoch": 11.34, + "learning_rate": 1.866917108018229e-06, + "loss": 0.2953, + "step": 30715 + }, + { + "epoch": 11.34, + "learning_rate": 1.8668801576548837e-06, + "loss": 0.4402, + "step": 30716 + }, + { + "epoch": 11.34, + "learning_rate": 1.8668432072915386e-06, + "loss": 0.2855, + "step": 30717 + }, + { + "epoch": 11.34, + "learning_rate": 1.8668062569281932e-06, + "loss": 0.3409, + "step": 30718 + }, + { + "epoch": 11.34, + "learning_rate": 1.8667693065648478e-06, + "loss": 0.2253, + "step": 30719 + }, + { + "epoch": 11.34, + "learning_rate": 1.8667323562015027e-06, + "loss": 0.2852, + "step": 30720 + }, + { + "epoch": 11.34, + "learning_rate": 1.8666954058381573e-06, + "loss": 0.2489, + "step": 30721 + }, + { + "epoch": 11.34, + "learning_rate": 1.8666584554748124e-06, + "loss": 0.3254, + "step": 30722 + }, + { + "epoch": 11.35, + "learning_rate": 1.8666215051114668e-06, + "loss": 0.2724, + "step": 30723 + }, + { + "epoch": 11.35, + "learning_rate": 1.8665845547481219e-06, + "loss": 0.3064, + "step": 30724 + }, + { + "epoch": 11.35, + "learning_rate": 1.8665476043847765e-06, + "loss": 0.3696, + "step": 30725 + }, + { + "epoch": 11.35, + "learning_rate": 1.8665106540214314e-06, + "loss": 0.2998, + "step": 30726 + }, + { + "epoch": 11.35, + "learning_rate": 1.866473703658086e-06, + "loss": 0.259, + "step": 30727 + }, + { + "epoch": 11.35, + "learning_rate": 1.8664367532947406e-06, + "loss": 0.3656, + "step": 30728 + }, + { + "epoch": 11.35, + "learning_rate": 1.8663998029313955e-06, + "loss": 0.2385, + "step": 30729 + }, + { + "epoch": 11.35, + "learning_rate": 1.8663628525680501e-06, + "loss": 0.4518, + "step": 30730 + }, + { + "epoch": 11.35, + "learning_rate": 1.8663259022047052e-06, + "loss": 0.3864, + "step": 30731 + }, + { + "epoch": 11.35, + "learning_rate": 1.8662889518413598e-06, + "loss": 0.3335, + "step": 30732 + }, + { + "epoch": 11.35, + "learning_rate": 1.8662520014780147e-06, + "loss": 0.3154, + "step": 30733 + }, + { + "epoch": 11.35, + "learning_rate": 1.8662150511146693e-06, + "loss": 0.1757, + "step": 30734 + }, + { + "epoch": 11.35, + "learning_rate": 1.8661781007513241e-06, + "loss": 0.324, + "step": 30735 + }, + { + "epoch": 11.35, + "learning_rate": 1.8661411503879788e-06, + "loss": 0.4067, + "step": 30736 + }, + { + "epoch": 11.35, + "learning_rate": 1.8661042000246338e-06, + "loss": 0.2216, + "step": 30737 + }, + { + "epoch": 11.35, + "learning_rate": 1.8660672496612885e-06, + "loss": 0.4074, + "step": 30738 + }, + { + "epoch": 11.35, + "learning_rate": 1.866030299297943e-06, + "loss": 0.223, + "step": 30739 + }, + { + "epoch": 11.35, + "learning_rate": 1.865993348934598e-06, + "loss": 0.316, + "step": 30740 + }, + { + "epoch": 11.35, + "learning_rate": 1.8659563985712526e-06, + "loss": 0.4642, + "step": 30741 + }, + { + "epoch": 11.35, + "learning_rate": 1.8659194482079074e-06, + "loss": 0.3096, + "step": 30742 + }, + { + "epoch": 11.35, + "learning_rate": 1.865882497844562e-06, + "loss": 0.2834, + "step": 30743 + }, + { + "epoch": 11.35, + "learning_rate": 1.8658455474812171e-06, + "loss": 0.2588, + "step": 30744 + }, + { + "epoch": 11.35, + "learning_rate": 1.8658085971178716e-06, + "loss": 0.4397, + "step": 30745 + }, + { + "epoch": 11.35, + "learning_rate": 1.8657716467545266e-06, + "loss": 0.3114, + "step": 30746 + }, + { + "epoch": 11.35, + "learning_rate": 1.8657346963911813e-06, + "loss": 0.3979, + "step": 30747 + }, + { + "epoch": 11.35, + "learning_rate": 1.865697746027836e-06, + "loss": 0.375, + "step": 30748 + }, + { + "epoch": 11.35, + "learning_rate": 1.8656607956644907e-06, + "loss": 0.3716, + "step": 30749 + }, + { + "epoch": 11.36, + "learning_rate": 1.8656238453011454e-06, + "loss": 0.4545, + "step": 30750 + }, + { + "epoch": 11.36, + "learning_rate": 1.8655868949378002e-06, + "loss": 0.2728, + "step": 30751 + }, + { + "epoch": 11.36, + "learning_rate": 1.8655499445744549e-06, + "loss": 0.4779, + "step": 30752 + }, + { + "epoch": 11.36, + "learning_rate": 1.86551299421111e-06, + "loss": 0.2809, + "step": 30753 + }, + { + "epoch": 11.36, + "learning_rate": 1.8654760438477646e-06, + "loss": 0.2412, + "step": 30754 + }, + { + "epoch": 11.36, + "learning_rate": 1.8654390934844194e-06, + "loss": 0.2811, + "step": 30755 + }, + { + "epoch": 11.36, + "learning_rate": 1.865402143121074e-06, + "loss": 0.3551, + "step": 30756 + }, + { + "epoch": 11.36, + "learning_rate": 1.865365192757729e-06, + "loss": 0.2627, + "step": 30757 + }, + { + "epoch": 11.36, + "learning_rate": 1.8653282423943835e-06, + "loss": 0.2655, + "step": 30758 + }, + { + "epoch": 11.36, + "learning_rate": 1.8652912920310382e-06, + "loss": 0.1951, + "step": 30759 + }, + { + "epoch": 11.36, + "learning_rate": 1.8652543416676932e-06, + "loss": 0.2879, + "step": 30760 + }, + { + "epoch": 11.36, + "learning_rate": 1.8652173913043479e-06, + "loss": 0.3401, + "step": 30761 + }, + { + "epoch": 11.36, + "learning_rate": 1.8651804409410027e-06, + "loss": 0.2962, + "step": 30762 + }, + { + "epoch": 11.36, + "learning_rate": 1.8651434905776574e-06, + "loss": 0.2723, + "step": 30763 + }, + { + "epoch": 11.36, + "learning_rate": 1.8651065402143122e-06, + "loss": 0.2577, + "step": 30764 + }, + { + "epoch": 11.36, + "learning_rate": 1.8650695898509668e-06, + "loss": 0.4667, + "step": 30765 + }, + { + "epoch": 11.36, + "learning_rate": 1.865032639487622e-06, + "loss": 0.2463, + "step": 30766 + }, + { + "epoch": 11.36, + "learning_rate": 1.8649956891242765e-06, + "loss": 0.2906, + "step": 30767 + }, + { + "epoch": 11.36, + "learning_rate": 1.864958738760931e-06, + "loss": 0.2459, + "step": 30768 + }, + { + "epoch": 11.36, + "learning_rate": 1.864921788397586e-06, + "loss": 0.2868, + "step": 30769 + }, + { + "epoch": 11.36, + "learning_rate": 1.8648848380342407e-06, + "loss": 0.2622, + "step": 30770 + }, + { + "epoch": 11.36, + "learning_rate": 1.8648478876708955e-06, + "loss": 0.4296, + "step": 30771 + }, + { + "epoch": 11.36, + "learning_rate": 1.8648109373075501e-06, + "loss": 0.331, + "step": 30772 + }, + { + "epoch": 11.36, + "learning_rate": 1.864773986944205e-06, + "loss": 0.3665, + "step": 30773 + }, + { + "epoch": 11.36, + "learning_rate": 1.8647370365808596e-06, + "loss": 0.3175, + "step": 30774 + }, + { + "epoch": 11.36, + "learning_rate": 1.8647000862175147e-06, + "loss": 0.3297, + "step": 30775 + }, + { + "epoch": 11.36, + "learning_rate": 1.8646631358541693e-06, + "loss": 0.2589, + "step": 30776 + }, + { + "epoch": 11.37, + "learning_rate": 1.8646261854908242e-06, + "loss": 0.3321, + "step": 30777 + }, + { + "epoch": 11.37, + "learning_rate": 1.8645892351274788e-06, + "loss": 0.4199, + "step": 30778 + }, + { + "epoch": 11.37, + "learning_rate": 1.8645522847641334e-06, + "loss": 0.2243, + "step": 30779 + }, + { + "epoch": 11.37, + "learning_rate": 1.8645153344007883e-06, + "loss": 0.2694, + "step": 30780 + }, + { + "epoch": 11.37, + "learning_rate": 1.864478384037443e-06, + "loss": 0.2819, + "step": 30781 + }, + { + "epoch": 11.37, + "learning_rate": 1.864441433674098e-06, + "loss": 0.3165, + "step": 30782 + }, + { + "epoch": 11.37, + "learning_rate": 1.8644044833107526e-06, + "loss": 0.2514, + "step": 30783 + }, + { + "epoch": 11.37, + "learning_rate": 1.8643675329474075e-06, + "loss": 0.2523, + "step": 30784 + }, + { + "epoch": 11.37, + "learning_rate": 1.864330582584062e-06, + "loss": 0.346, + "step": 30785 + }, + { + "epoch": 11.37, + "learning_rate": 1.864293632220717e-06, + "loss": 0.3301, + "step": 30786 + }, + { + "epoch": 11.37, + "learning_rate": 1.8642566818573716e-06, + "loss": 0.3654, + "step": 30787 + }, + { + "epoch": 11.37, + "learning_rate": 1.8642197314940262e-06, + "loss": 0.261, + "step": 30788 + }, + { + "epoch": 11.37, + "learning_rate": 1.8641827811306813e-06, + "loss": 0.3907, + "step": 30789 + }, + { + "epoch": 11.37, + "learning_rate": 1.8641458307673357e-06, + "loss": 0.277, + "step": 30790 + }, + { + "epoch": 11.37, + "learning_rate": 1.8641088804039908e-06, + "loss": 0.353, + "step": 30791 + }, + { + "epoch": 11.37, + "learning_rate": 1.8640719300406454e-06, + "loss": 0.4249, + "step": 30792 + }, + { + "epoch": 11.37, + "learning_rate": 1.8640349796773003e-06, + "loss": 0.2652, + "step": 30793 + }, + { + "epoch": 11.37, + "learning_rate": 1.8639980293139549e-06, + "loss": 0.445, + "step": 30794 + }, + { + "epoch": 11.37, + "learning_rate": 1.86396107895061e-06, + "loss": 0.3789, + "step": 30795 + }, + { + "epoch": 11.37, + "learning_rate": 1.8639241285872644e-06, + "loss": 0.3511, + "step": 30796 + }, + { + "epoch": 11.37, + "learning_rate": 1.8638871782239194e-06, + "loss": 0.4377, + "step": 30797 + }, + { + "epoch": 11.37, + "learning_rate": 1.863850227860574e-06, + "loss": 0.3141, + "step": 30798 + }, + { + "epoch": 11.37, + "learning_rate": 1.8638132774972287e-06, + "loss": 0.2379, + "step": 30799 + }, + { + "epoch": 11.37, + "learning_rate": 1.8637763271338836e-06, + "loss": 0.2855, + "step": 30800 + }, + { + "epoch": 11.37, + "learning_rate": 1.8637393767705382e-06, + "loss": 0.3182, + "step": 30801 + }, + { + "epoch": 11.37, + "learning_rate": 1.863702426407193e-06, + "loss": 0.2887, + "step": 30802 + }, + { + "epoch": 11.37, + "learning_rate": 1.8636654760438477e-06, + "loss": 0.3221, + "step": 30803 + }, + { + "epoch": 11.38, + "learning_rate": 1.8636285256805027e-06, + "loss": 0.2328, + "step": 30804 + }, + { + "epoch": 11.38, + "learning_rate": 1.8635915753171574e-06, + "loss": 0.3064, + "step": 30805 + }, + { + "epoch": 11.38, + "learning_rate": 1.8635546249538122e-06, + "loss": 0.2944, + "step": 30806 + }, + { + "epoch": 11.38, + "learning_rate": 1.8635176745904669e-06, + "loss": 0.2381, + "step": 30807 + }, + { + "epoch": 11.38, + "learning_rate": 1.8634807242271215e-06, + "loss": 0.3287, + "step": 30808 + }, + { + "epoch": 11.38, + "learning_rate": 1.8634437738637763e-06, + "loss": 0.3777, + "step": 30809 + }, + { + "epoch": 11.38, + "learning_rate": 1.863406823500431e-06, + "loss": 0.2535, + "step": 30810 + }, + { + "epoch": 11.38, + "learning_rate": 1.863369873137086e-06, + "loss": 0.2596, + "step": 30811 + }, + { + "epoch": 11.38, + "learning_rate": 1.8633329227737405e-06, + "loss": 0.3404, + "step": 30812 + }, + { + "epoch": 11.38, + "learning_rate": 1.8632959724103955e-06, + "loss": 0.3463, + "step": 30813 + }, + { + "epoch": 11.38, + "learning_rate": 1.8632590220470502e-06, + "loss": 0.2864, + "step": 30814 + }, + { + "epoch": 11.38, + "learning_rate": 1.863222071683705e-06, + "loss": 0.2486, + "step": 30815 + }, + { + "epoch": 11.38, + "learning_rate": 1.8631851213203596e-06, + "loss": 0.2807, + "step": 30816 + }, + { + "epoch": 11.38, + "learning_rate": 1.8631481709570147e-06, + "loss": 0.3418, + "step": 30817 + }, + { + "epoch": 11.38, + "learning_rate": 1.8631112205936691e-06, + "loss": 0.3891, + "step": 30818 + }, + { + "epoch": 11.38, + "learning_rate": 1.8630742702303238e-06, + "loss": 0.4404, + "step": 30819 + }, + { + "epoch": 11.38, + "learning_rate": 1.8630373198669788e-06, + "loss": 0.2367, + "step": 30820 + }, + { + "epoch": 11.38, + "learning_rate": 1.8630003695036335e-06, + "loss": 0.3335, + "step": 30821 + }, + { + "epoch": 11.38, + "learning_rate": 1.8629634191402883e-06, + "loss": 0.3062, + "step": 30822 + }, + { + "epoch": 11.38, + "learning_rate": 1.862926468776943e-06, + "loss": 0.3702, + "step": 30823 + }, + { + "epoch": 11.38, + "learning_rate": 1.8628895184135978e-06, + "loss": 0.3444, + "step": 30824 + }, + { + "epoch": 11.38, + "learning_rate": 1.8628525680502524e-06, + "loss": 0.188, + "step": 30825 + }, + { + "epoch": 11.38, + "learning_rate": 1.8628156176869075e-06, + "loss": 0.2576, + "step": 30826 + }, + { + "epoch": 11.38, + "learning_rate": 1.8627786673235621e-06, + "loss": 0.4122, + "step": 30827 + }, + { + "epoch": 11.38, + "learning_rate": 1.8627417169602168e-06, + "loss": 0.3488, + "step": 30828 + }, + { + "epoch": 11.38, + "learning_rate": 1.8627047665968716e-06, + "loss": 0.3481, + "step": 30829 + }, + { + "epoch": 11.38, + "learning_rate": 1.8626678162335263e-06, + "loss": 0.2288, + "step": 30830 + }, + { + "epoch": 11.39, + "learning_rate": 1.862630865870181e-06, + "loss": 0.358, + "step": 30831 + }, + { + "epoch": 11.39, + "learning_rate": 1.8625939155068357e-06, + "loss": 0.2138, + "step": 30832 + }, + { + "epoch": 11.39, + "learning_rate": 1.8625569651434908e-06, + "loss": 0.4286, + "step": 30833 + }, + { + "epoch": 11.39, + "learning_rate": 1.8625200147801454e-06, + "loss": 0.4062, + "step": 30834 + }, + { + "epoch": 11.39, + "learning_rate": 1.8624830644168003e-06, + "loss": 0.3281, + "step": 30835 + }, + { + "epoch": 11.39, + "learning_rate": 1.862446114053455e-06, + "loss": 0.3067, + "step": 30836 + }, + { + "epoch": 11.39, + "learning_rate": 1.8624091636901098e-06, + "loss": 0.275, + "step": 30837 + }, + { + "epoch": 11.39, + "learning_rate": 1.8623722133267644e-06, + "loss": 0.3455, + "step": 30838 + }, + { + "epoch": 11.39, + "learning_rate": 1.862335262963419e-06, + "loss": 0.2561, + "step": 30839 + }, + { + "epoch": 11.39, + "learning_rate": 1.862298312600074e-06, + "loss": 0.2744, + "step": 30840 + }, + { + "epoch": 11.39, + "learning_rate": 1.8622613622367285e-06, + "loss": 0.3812, + "step": 30841 + }, + { + "epoch": 11.39, + "learning_rate": 1.8622244118733836e-06, + "loss": 0.3992, + "step": 30842 + }, + { + "epoch": 11.39, + "learning_rate": 1.8621874615100382e-06, + "loss": 0.2336, + "step": 30843 + }, + { + "epoch": 11.39, + "learning_rate": 1.862150511146693e-06, + "loss": 0.2299, + "step": 30844 + }, + { + "epoch": 11.39, + "learning_rate": 1.8621135607833477e-06, + "loss": 0.272, + "step": 30845 + }, + { + "epoch": 11.39, + "learning_rate": 1.8620766104200026e-06, + "loss": 0.3497, + "step": 30846 + }, + { + "epoch": 11.39, + "learning_rate": 1.8620396600566572e-06, + "loss": 0.3345, + "step": 30847 + }, + { + "epoch": 11.39, + "learning_rate": 1.8620027096933118e-06, + "loss": 0.3195, + "step": 30848 + }, + { + "epoch": 11.39, + "learning_rate": 1.8619657593299669e-06, + "loss": 0.3666, + "step": 30849 + }, + { + "epoch": 11.39, + "learning_rate": 1.8619288089666215e-06, + "loss": 0.3463, + "step": 30850 + }, + { + "epoch": 11.39, + "learning_rate": 1.8618918586032764e-06, + "loss": 0.3183, + "step": 30851 + }, + { + "epoch": 11.39, + "learning_rate": 1.861854908239931e-06, + "loss": 0.279, + "step": 30852 + }, + { + "epoch": 11.39, + "learning_rate": 1.8618179578765859e-06, + "loss": 0.1731, + "step": 30853 + }, + { + "epoch": 11.39, + "learning_rate": 1.8617810075132405e-06, + "loss": 0.3309, + "step": 30854 + }, + { + "epoch": 11.39, + "learning_rate": 1.8617440571498955e-06, + "loss": 0.3216, + "step": 30855 + }, + { + "epoch": 11.39, + "learning_rate": 1.8617071067865502e-06, + "loss": 0.2503, + "step": 30856 + }, + { + "epoch": 11.39, + "learning_rate": 1.861670156423205e-06, + "loss": 0.429, + "step": 30857 + }, + { + "epoch": 11.4, + "learning_rate": 1.8616332060598597e-06, + "loss": 0.4185, + "step": 30858 + }, + { + "epoch": 11.4, + "learning_rate": 1.8615962556965143e-06, + "loss": 0.2967, + "step": 30859 + }, + { + "epoch": 11.4, + "learning_rate": 1.8615593053331692e-06, + "loss": 0.4331, + "step": 30860 + }, + { + "epoch": 11.4, + "learning_rate": 1.8615223549698238e-06, + "loss": 0.2001, + "step": 30861 + }, + { + "epoch": 11.4, + "learning_rate": 1.8614854046064789e-06, + "loss": 0.3216, + "step": 30862 + }, + { + "epoch": 11.4, + "learning_rate": 1.8614484542431333e-06, + "loss": 0.4089, + "step": 30863 + }, + { + "epoch": 11.4, + "learning_rate": 1.8614115038797883e-06, + "loss": 0.2576, + "step": 30864 + }, + { + "epoch": 11.4, + "learning_rate": 1.861374553516443e-06, + "loss": 0.4936, + "step": 30865 + }, + { + "epoch": 11.4, + "learning_rate": 1.8613376031530978e-06, + "loss": 0.4474, + "step": 30866 + }, + { + "epoch": 11.4, + "learning_rate": 1.8613006527897525e-06, + "loss": 0.3788, + "step": 30867 + }, + { + "epoch": 11.4, + "learning_rate": 1.861263702426407e-06, + "loss": 0.3296, + "step": 30868 + }, + { + "epoch": 11.4, + "learning_rate": 1.861226752063062e-06, + "loss": 0.3331, + "step": 30869 + }, + { + "epoch": 11.4, + "learning_rate": 1.8611898016997166e-06, + "loss": 0.2896, + "step": 30870 + }, + { + "epoch": 11.4, + "learning_rate": 1.8611528513363716e-06, + "loss": 0.3048, + "step": 30871 + }, + { + "epoch": 11.4, + "learning_rate": 1.8611159009730263e-06, + "loss": 0.2274, + "step": 30872 + }, + { + "epoch": 11.4, + "learning_rate": 1.8610789506096811e-06, + "loss": 0.328, + "step": 30873 + }, + { + "epoch": 11.4, + "learning_rate": 1.8610420002463358e-06, + "loss": 0.3523, + "step": 30874 + }, + { + "epoch": 11.4, + "learning_rate": 1.8610050498829906e-06, + "loss": 0.4163, + "step": 30875 + }, + { + "epoch": 11.4, + "learning_rate": 1.8609680995196452e-06, + "loss": 0.2865, + "step": 30876 + }, + { + "epoch": 11.4, + "learning_rate": 1.8609311491563003e-06, + "loss": 0.455, + "step": 30877 + }, + { + "epoch": 11.4, + "learning_rate": 1.860894198792955e-06, + "loss": 0.2737, + "step": 30878 + }, + { + "epoch": 11.4, + "learning_rate": 1.8608572484296094e-06, + "loss": 0.2916, + "step": 30879 + }, + { + "epoch": 11.4, + "learning_rate": 1.8608202980662644e-06, + "loss": 0.2239, + "step": 30880 + }, + { + "epoch": 11.4, + "learning_rate": 1.860783347702919e-06, + "loss": 0.3053, + "step": 30881 + }, + { + "epoch": 11.4, + "learning_rate": 1.860746397339574e-06, + "loss": 0.3357, + "step": 30882 + }, + { + "epoch": 11.4, + "learning_rate": 1.8607094469762285e-06, + "loss": 0.3185, + "step": 30883 + }, + { + "epoch": 11.4, + "learning_rate": 1.8606724966128836e-06, + "loss": 0.3196, + "step": 30884 + }, + { + "epoch": 11.41, + "learning_rate": 1.860635546249538e-06, + "loss": 0.2653, + "step": 30885 + }, + { + "epoch": 11.41, + "learning_rate": 1.860598595886193e-06, + "loss": 0.3391, + "step": 30886 + }, + { + "epoch": 11.41, + "learning_rate": 1.8605616455228477e-06, + "loss": 0.2818, + "step": 30887 + }, + { + "epoch": 11.41, + "learning_rate": 1.8605246951595024e-06, + "loss": 0.2982, + "step": 30888 + }, + { + "epoch": 11.41, + "learning_rate": 1.8604877447961572e-06, + "loss": 0.5029, + "step": 30889 + }, + { + "epoch": 11.41, + "learning_rate": 1.8604507944328118e-06, + "loss": 0.4287, + "step": 30890 + }, + { + "epoch": 11.41, + "learning_rate": 1.8604138440694667e-06, + "loss": 0.3266, + "step": 30891 + }, + { + "epoch": 11.41, + "learning_rate": 1.8603768937061213e-06, + "loss": 0.2752, + "step": 30892 + }, + { + "epoch": 11.41, + "learning_rate": 1.8603399433427764e-06, + "loss": 0.2575, + "step": 30893 + }, + { + "epoch": 11.41, + "learning_rate": 1.860302992979431e-06, + "loss": 0.3539, + "step": 30894 + }, + { + "epoch": 11.41, + "learning_rate": 1.8602660426160859e-06, + "loss": 0.2557, + "step": 30895 + }, + { + "epoch": 11.41, + "learning_rate": 1.8602290922527405e-06, + "loss": 0.3075, + "step": 30896 + }, + { + "epoch": 11.41, + "learning_rate": 1.8601921418893954e-06, + "loss": 0.2742, + "step": 30897 + }, + { + "epoch": 11.41, + "learning_rate": 1.86015519152605e-06, + "loss": 0.2541, + "step": 30898 + }, + { + "epoch": 11.41, + "learning_rate": 1.8601182411627046e-06, + "loss": 0.3548, + "step": 30899 + }, + { + "epoch": 11.41, + "learning_rate": 1.8600812907993597e-06, + "loss": 0.3324, + "step": 30900 + }, + { + "epoch": 11.41, + "learning_rate": 1.8600443404360143e-06, + "loss": 0.2722, + "step": 30901 + }, + { + "epoch": 11.41, + "learning_rate": 1.8600073900726692e-06, + "loss": 0.3142, + "step": 30902 + }, + { + "epoch": 11.41, + "learning_rate": 1.8599704397093238e-06, + "loss": 0.2544, + "step": 30903 + }, + { + "epoch": 11.41, + "learning_rate": 1.8599334893459787e-06, + "loss": 0.4113, + "step": 30904 + }, + { + "epoch": 11.41, + "learning_rate": 1.8598965389826333e-06, + "loss": 0.2367, + "step": 30905 + }, + { + "epoch": 11.41, + "learning_rate": 1.8598595886192884e-06, + "loss": 0.3171, + "step": 30906 + }, + { + "epoch": 11.41, + "learning_rate": 1.859822638255943e-06, + "loss": 0.4016, + "step": 30907 + }, + { + "epoch": 11.41, + "learning_rate": 1.8597856878925974e-06, + "loss": 0.4326, + "step": 30908 + }, + { + "epoch": 11.41, + "learning_rate": 1.8597487375292525e-06, + "loss": 0.4058, + "step": 30909 + }, + { + "epoch": 11.41, + "learning_rate": 1.8597117871659071e-06, + "loss": 0.319, + "step": 30910 + }, + { + "epoch": 11.41, + "learning_rate": 1.859674836802562e-06, + "loss": 0.3002, + "step": 30911 + }, + { + "epoch": 11.42, + "learning_rate": 1.8596378864392166e-06, + "loss": 0.3276, + "step": 30912 + }, + { + "epoch": 11.42, + "learning_rate": 1.8596009360758717e-06, + "loss": 0.3575, + "step": 30913 + }, + { + "epoch": 11.42, + "learning_rate": 1.859563985712526e-06, + "loss": 0.3188, + "step": 30914 + }, + { + "epoch": 11.42, + "learning_rate": 1.8595270353491811e-06, + "loss": 0.1919, + "step": 30915 + }, + { + "epoch": 11.42, + "learning_rate": 1.8594900849858358e-06, + "loss": 0.2079, + "step": 30916 + }, + { + "epoch": 11.42, + "learning_rate": 1.8594531346224904e-06, + "loss": 0.3375, + "step": 30917 + }, + { + "epoch": 11.42, + "learning_rate": 1.8594161842591453e-06, + "loss": 0.396, + "step": 30918 + }, + { + "epoch": 11.42, + "learning_rate": 1.8593792338958e-06, + "loss": 0.3611, + "step": 30919 + }, + { + "epoch": 11.42, + "learning_rate": 1.8593422835324548e-06, + "loss": 0.3604, + "step": 30920 + }, + { + "epoch": 11.42, + "learning_rate": 1.8593053331691094e-06, + "loss": 0.3129, + "step": 30921 + }, + { + "epoch": 11.42, + "learning_rate": 1.8592683828057644e-06, + "loss": 0.2679, + "step": 30922 + }, + { + "epoch": 11.42, + "learning_rate": 1.859231432442419e-06, + "loss": 0.2594, + "step": 30923 + }, + { + "epoch": 11.42, + "learning_rate": 1.859194482079074e-06, + "loss": 0.4473, + "step": 30924 + }, + { + "epoch": 11.42, + "learning_rate": 1.8591575317157286e-06, + "loss": 0.29, + "step": 30925 + }, + { + "epoch": 11.42, + "learning_rate": 1.8591205813523834e-06, + "loss": 0.2843, + "step": 30926 + }, + { + "epoch": 11.42, + "learning_rate": 1.859083630989038e-06, + "loss": 0.2787, + "step": 30927 + }, + { + "epoch": 11.42, + "learning_rate": 1.8590466806256927e-06, + "loss": 0.2014, + "step": 30928 + }, + { + "epoch": 11.42, + "learning_rate": 1.8590097302623478e-06, + "loss": 0.309, + "step": 30929 + }, + { + "epoch": 11.42, + "learning_rate": 1.8589727798990022e-06, + "loss": 0.3375, + "step": 30930 + }, + { + "epoch": 11.42, + "learning_rate": 1.8589358295356572e-06, + "loss": 0.4972, + "step": 30931 + }, + { + "epoch": 11.42, + "learning_rate": 1.8588988791723119e-06, + "loss": 0.5727, + "step": 30932 + }, + { + "epoch": 11.42, + "learning_rate": 1.8588619288089667e-06, + "loss": 0.2966, + "step": 30933 + }, + { + "epoch": 11.42, + "learning_rate": 1.8588249784456214e-06, + "loss": 0.3049, + "step": 30934 + }, + { + "epoch": 11.42, + "learning_rate": 1.8587880280822764e-06, + "loss": 0.3383, + "step": 30935 + }, + { + "epoch": 11.42, + "learning_rate": 1.8587510777189308e-06, + "loss": 0.2869, + "step": 30936 + }, + { + "epoch": 11.42, + "learning_rate": 1.8587141273555855e-06, + "loss": 0.3792, + "step": 30937 + }, + { + "epoch": 11.42, + "learning_rate": 1.8586771769922405e-06, + "loss": 0.5887, + "step": 30938 + }, + { + "epoch": 11.43, + "learning_rate": 1.8586402266288952e-06, + "loss": 0.2562, + "step": 30939 + }, + { + "epoch": 11.43, + "learning_rate": 1.85860327626555e-06, + "loss": 0.2965, + "step": 30940 + }, + { + "epoch": 11.43, + "learning_rate": 1.8585663259022047e-06, + "loss": 0.2853, + "step": 30941 + }, + { + "epoch": 11.43, + "learning_rate": 1.8585293755388595e-06, + "loss": 0.2775, + "step": 30942 + }, + { + "epoch": 11.43, + "learning_rate": 1.8584924251755141e-06, + "loss": 0.3259, + "step": 30943 + }, + { + "epoch": 11.43, + "learning_rate": 1.8584554748121692e-06, + "loss": 0.3132, + "step": 30944 + }, + { + "epoch": 11.43, + "learning_rate": 1.8584185244488238e-06, + "loss": 0.4823, + "step": 30945 + }, + { + "epoch": 11.43, + "learning_rate": 1.8583815740854787e-06, + "loss": 0.4735, + "step": 30946 + }, + { + "epoch": 11.43, + "learning_rate": 1.8583446237221333e-06, + "loss": 0.4957, + "step": 30947 + }, + { + "epoch": 11.43, + "learning_rate": 1.858307673358788e-06, + "loss": 0.4522, + "step": 30948 + }, + { + "epoch": 11.43, + "learning_rate": 1.8582707229954428e-06, + "loss": 0.391, + "step": 30949 + }, + { + "epoch": 11.43, + "learning_rate": 1.8582337726320974e-06, + "loss": 0.3575, + "step": 30950 + }, + { + "epoch": 11.43, + "learning_rate": 1.8581968222687525e-06, + "loss": 0.225, + "step": 30951 + }, + { + "epoch": 11.43, + "learning_rate": 1.8581598719054071e-06, + "loss": 0.3451, + "step": 30952 + }, + { + "epoch": 11.43, + "learning_rate": 1.858122921542062e-06, + "loss": 0.329, + "step": 30953 + }, + { + "epoch": 11.43, + "learning_rate": 1.8580859711787166e-06, + "loss": 0.2918, + "step": 30954 + }, + { + "epoch": 11.43, + "learning_rate": 1.8580490208153715e-06, + "loss": 0.2935, + "step": 30955 + }, + { + "epoch": 11.43, + "learning_rate": 1.8580120704520261e-06, + "loss": 0.3457, + "step": 30956 + }, + { + "epoch": 11.43, + "learning_rate": 1.8579751200886807e-06, + "loss": 0.2315, + "step": 30957 + }, + { + "epoch": 11.43, + "learning_rate": 1.8579381697253356e-06, + "loss": 0.2198, + "step": 30958 + }, + { + "epoch": 11.43, + "learning_rate": 1.8579012193619902e-06, + "loss": 0.276, + "step": 30959 + }, + { + "epoch": 11.43, + "learning_rate": 1.8578642689986453e-06, + "loss": 0.3074, + "step": 30960 + }, + { + "epoch": 11.43, + "learning_rate": 1.8578273186353e-06, + "loss": 0.298, + "step": 30961 + }, + { + "epoch": 11.43, + "learning_rate": 1.8577903682719548e-06, + "loss": 0.3313, + "step": 30962 + }, + { + "epoch": 11.43, + "learning_rate": 1.8577534179086094e-06, + "loss": 0.3404, + "step": 30963 + }, + { + "epoch": 11.43, + "learning_rate": 1.8577164675452643e-06, + "loss": 0.2185, + "step": 30964 + }, + { + "epoch": 11.43, + "learning_rate": 1.857679517181919e-06, + "loss": 0.385, + "step": 30965 + }, + { + "epoch": 11.44, + "learning_rate": 1.857642566818574e-06, + "loss": 0.3194, + "step": 30966 + }, + { + "epoch": 11.44, + "learning_rate": 1.8576056164552286e-06, + "loss": 0.319, + "step": 30967 + }, + { + "epoch": 11.44, + "learning_rate": 1.8575686660918832e-06, + "loss": 0.338, + "step": 30968 + }, + { + "epoch": 11.44, + "learning_rate": 1.857531715728538e-06, + "loss": 0.2623, + "step": 30969 + }, + { + "epoch": 11.44, + "learning_rate": 1.8574947653651927e-06, + "loss": 0.3051, + "step": 30970 + }, + { + "epoch": 11.44, + "learning_rate": 1.8574578150018476e-06, + "loss": 0.298, + "step": 30971 + }, + { + "epoch": 11.44, + "learning_rate": 1.8574208646385022e-06, + "loss": 0.3499, + "step": 30972 + }, + { + "epoch": 11.44, + "learning_rate": 1.8573839142751573e-06, + "loss": 0.2879, + "step": 30973 + }, + { + "epoch": 11.44, + "learning_rate": 1.857346963911812e-06, + "loss": 0.3691, + "step": 30974 + }, + { + "epoch": 11.44, + "learning_rate": 1.8573100135484667e-06, + "loss": 0.4754, + "step": 30975 + }, + { + "epoch": 11.44, + "learning_rate": 1.8572730631851214e-06, + "loss": 0.2467, + "step": 30976 + }, + { + "epoch": 11.44, + "learning_rate": 1.857236112821776e-06, + "loss": 0.2475, + "step": 30977 + }, + { + "epoch": 11.44, + "learning_rate": 1.8571991624584309e-06, + "loss": 0.323, + "step": 30978 + }, + { + "epoch": 11.44, + "learning_rate": 1.8571622120950855e-06, + "loss": 0.3251, + "step": 30979 + }, + { + "epoch": 11.44, + "learning_rate": 1.8571252617317406e-06, + "loss": 0.2655, + "step": 30980 + }, + { + "epoch": 11.44, + "learning_rate": 1.857088311368395e-06, + "loss": 0.3128, + "step": 30981 + }, + { + "epoch": 11.44, + "learning_rate": 1.85705136100505e-06, + "loss": 0.4327, + "step": 30982 + }, + { + "epoch": 11.44, + "learning_rate": 1.8570144106417047e-06, + "loss": 0.3624, + "step": 30983 + }, + { + "epoch": 11.44, + "learning_rate": 1.8569774602783595e-06, + "loss": 0.4683, + "step": 30984 + }, + { + "epoch": 11.44, + "learning_rate": 1.8569405099150142e-06, + "loss": 0.354, + "step": 30985 + }, + { + "epoch": 11.44, + "learning_rate": 1.8569035595516692e-06, + "loss": 0.284, + "step": 30986 + }, + { + "epoch": 11.44, + "learning_rate": 1.8568666091883237e-06, + "loss": 0.2281, + "step": 30987 + }, + { + "epoch": 11.44, + "learning_rate": 1.8568296588249783e-06, + "loss": 0.2438, + "step": 30988 + }, + { + "epoch": 11.44, + "learning_rate": 1.8567927084616333e-06, + "loss": 0.2476, + "step": 30989 + }, + { + "epoch": 11.44, + "learning_rate": 1.856755758098288e-06, + "loss": 0.2978, + "step": 30990 + }, + { + "epoch": 11.44, + "learning_rate": 1.8567188077349428e-06, + "loss": 0.1854, + "step": 30991 + }, + { + "epoch": 11.44, + "learning_rate": 1.8566818573715975e-06, + "loss": 0.3013, + "step": 30992 + }, + { + "epoch": 11.44, + "learning_rate": 1.8566449070082523e-06, + "loss": 0.2925, + "step": 30993 + }, + { + "epoch": 11.45, + "learning_rate": 1.856607956644907e-06, + "loss": 0.1659, + "step": 30994 + }, + { + "epoch": 11.45, + "learning_rate": 1.856571006281562e-06, + "loss": 0.3043, + "step": 30995 + }, + { + "epoch": 11.45, + "learning_rate": 1.8565340559182167e-06, + "loss": 0.3815, + "step": 30996 + }, + { + "epoch": 11.45, + "learning_rate": 1.856497105554871e-06, + "loss": 0.4379, + "step": 30997 + }, + { + "epoch": 11.45, + "learning_rate": 1.8564601551915261e-06, + "loss": 0.3878, + "step": 30998 + }, + { + "epoch": 11.45, + "learning_rate": 1.8564232048281808e-06, + "loss": 0.2522, + "step": 30999 + }, + { + "epoch": 11.45, + "learning_rate": 1.8563862544648356e-06, + "loss": 0.2853, + "step": 31000 + }, + { + "epoch": 11.45, + "learning_rate": 1.8563493041014903e-06, + "loss": 0.275, + "step": 31001 + }, + { + "epoch": 11.45, + "learning_rate": 1.8563123537381453e-06, + "loss": 0.3351, + "step": 31002 + }, + { + "epoch": 11.45, + "learning_rate": 1.8562754033747997e-06, + "loss": 0.3302, + "step": 31003 + }, + { + "epoch": 11.45, + "learning_rate": 1.8562384530114548e-06, + "loss": 0.2145, + "step": 31004 + }, + { + "epoch": 11.45, + "learning_rate": 1.8562015026481094e-06, + "loss": 0.2595, + "step": 31005 + }, + { + "epoch": 11.45, + "learning_rate": 1.8561645522847643e-06, + "loss": 0.4124, + "step": 31006 + }, + { + "epoch": 11.45, + "learning_rate": 1.856127601921419e-06, + "loss": 0.3713, + "step": 31007 + }, + { + "epoch": 11.45, + "learning_rate": 1.8560906515580736e-06, + "loss": 0.2798, + "step": 31008 + }, + { + "epoch": 11.45, + "learning_rate": 1.8560537011947284e-06, + "loss": 0.3066, + "step": 31009 + }, + { + "epoch": 11.45, + "learning_rate": 1.856016750831383e-06, + "loss": 0.2414, + "step": 31010 + }, + { + "epoch": 11.45, + "learning_rate": 1.855979800468038e-06, + "loss": 0.2933, + "step": 31011 + }, + { + "epoch": 11.45, + "learning_rate": 1.8559428501046927e-06, + "loss": 0.2861, + "step": 31012 + }, + { + "epoch": 11.45, + "learning_rate": 1.8559058997413476e-06, + "loss": 0.2386, + "step": 31013 + }, + { + "epoch": 11.45, + "learning_rate": 1.8558689493780022e-06, + "loss": 0.3328, + "step": 31014 + }, + { + "epoch": 11.45, + "learning_rate": 1.855831999014657e-06, + "loss": 0.3077, + "step": 31015 + }, + { + "epoch": 11.45, + "learning_rate": 1.8557950486513117e-06, + "loss": 0.3262, + "step": 31016 + }, + { + "epoch": 11.45, + "learning_rate": 1.8557580982879663e-06, + "loss": 0.2897, + "step": 31017 + }, + { + "epoch": 11.45, + "learning_rate": 1.8557211479246214e-06, + "loss": 0.3023, + "step": 31018 + }, + { + "epoch": 11.45, + "learning_rate": 1.855684197561276e-06, + "loss": 0.399, + "step": 31019 + }, + { + "epoch": 11.45, + "learning_rate": 1.8556472471979309e-06, + "loss": 0.3168, + "step": 31020 + }, + { + "epoch": 11.46, + "learning_rate": 1.8556102968345855e-06, + "loss": 0.251, + "step": 31021 + }, + { + "epoch": 11.46, + "learning_rate": 1.8555733464712404e-06, + "loss": 0.3332, + "step": 31022 + }, + { + "epoch": 11.46, + "learning_rate": 1.855536396107895e-06, + "loss": 0.2681, + "step": 31023 + }, + { + "epoch": 11.46, + "learning_rate": 1.85549944574455e-06, + "loss": 0.3035, + "step": 31024 + }, + { + "epoch": 11.46, + "learning_rate": 1.8554624953812047e-06, + "loss": 0.2352, + "step": 31025 + }, + { + "epoch": 11.46, + "learning_rate": 1.8554255450178596e-06, + "loss": 0.4064, + "step": 31026 + }, + { + "epoch": 11.46, + "learning_rate": 1.8553885946545142e-06, + "loss": 0.3021, + "step": 31027 + }, + { + "epoch": 11.46, + "learning_rate": 1.8553516442911688e-06, + "loss": 0.3492, + "step": 31028 + }, + { + "epoch": 11.46, + "learning_rate": 1.8553146939278237e-06, + "loss": 0.4092, + "step": 31029 + }, + { + "epoch": 11.46, + "learning_rate": 1.8552777435644783e-06, + "loss": 0.2956, + "step": 31030 + }, + { + "epoch": 11.46, + "learning_rate": 1.8552407932011334e-06, + "loss": 0.3662, + "step": 31031 + }, + { + "epoch": 11.46, + "learning_rate": 1.8552038428377878e-06, + "loss": 0.3903, + "step": 31032 + }, + { + "epoch": 11.46, + "learning_rate": 1.8551668924744429e-06, + "loss": 0.4737, + "step": 31033 + }, + { + "epoch": 11.46, + "learning_rate": 1.8551299421110975e-06, + "loss": 0.2589, + "step": 31034 + }, + { + "epoch": 11.46, + "learning_rate": 1.8550929917477523e-06, + "loss": 0.2725, + "step": 31035 + }, + { + "epoch": 11.46, + "learning_rate": 1.855056041384407e-06, + "loss": 0.3177, + "step": 31036 + }, + { + "epoch": 11.46, + "learning_rate": 1.8550190910210616e-06, + "loss": 0.4317, + "step": 31037 + }, + { + "epoch": 11.46, + "learning_rate": 1.8549821406577165e-06, + "loss": 0.2668, + "step": 31038 + }, + { + "epoch": 11.46, + "learning_rate": 1.854945190294371e-06, + "loss": 0.2676, + "step": 31039 + }, + { + "epoch": 11.46, + "learning_rate": 1.8549082399310262e-06, + "loss": 0.4065, + "step": 31040 + }, + { + "epoch": 11.46, + "learning_rate": 1.8548712895676808e-06, + "loss": 0.4333, + "step": 31041 + }, + { + "epoch": 11.46, + "learning_rate": 1.8548343392043356e-06, + "loss": 0.2813, + "step": 31042 + }, + { + "epoch": 11.46, + "learning_rate": 1.8547973888409903e-06, + "loss": 0.3053, + "step": 31043 + }, + { + "epoch": 11.46, + "learning_rate": 1.8547604384776451e-06, + "loss": 0.3851, + "step": 31044 + }, + { + "epoch": 11.46, + "learning_rate": 1.8547234881142998e-06, + "loss": 0.3947, + "step": 31045 + }, + { + "epoch": 11.46, + "learning_rate": 1.8546865377509548e-06, + "loss": 0.24, + "step": 31046 + }, + { + "epoch": 11.46, + "learning_rate": 1.8546495873876095e-06, + "loss": 0.3521, + "step": 31047 + }, + { + "epoch": 11.47, + "learning_rate": 1.8546126370242639e-06, + "loss": 0.3372, + "step": 31048 + }, + { + "epoch": 11.47, + "learning_rate": 1.854575686660919e-06, + "loss": 0.3381, + "step": 31049 + }, + { + "epoch": 11.47, + "learning_rate": 1.8545387362975736e-06, + "loss": 0.3506, + "step": 31050 + }, + { + "epoch": 11.47, + "learning_rate": 1.8545017859342284e-06, + "loss": 0.2841, + "step": 31051 + }, + { + "epoch": 11.47, + "learning_rate": 1.854464835570883e-06, + "loss": 0.2851, + "step": 31052 + }, + { + "epoch": 11.47, + "learning_rate": 1.8544278852075381e-06, + "loss": 0.2636, + "step": 31053 + }, + { + "epoch": 11.47, + "learning_rate": 1.8543909348441926e-06, + "loss": 0.3072, + "step": 31054 + }, + { + "epoch": 11.47, + "learning_rate": 1.8543539844808476e-06, + "loss": 0.2282, + "step": 31055 + }, + { + "epoch": 11.47, + "learning_rate": 1.8543170341175022e-06, + "loss": 0.4926, + "step": 31056 + }, + { + "epoch": 11.47, + "learning_rate": 1.8542800837541569e-06, + "loss": 0.3972, + "step": 31057 + }, + { + "epoch": 11.47, + "learning_rate": 1.8542431333908117e-06, + "loss": 0.3357, + "step": 31058 + }, + { + "epoch": 11.47, + "learning_rate": 1.8542061830274664e-06, + "loss": 0.5147, + "step": 31059 + }, + { + "epoch": 11.47, + "learning_rate": 1.8541692326641212e-06, + "loss": 0.3855, + "step": 31060 + }, + { + "epoch": 11.47, + "learning_rate": 1.8541322823007759e-06, + "loss": 0.2718, + "step": 31061 + }, + { + "epoch": 11.47, + "learning_rate": 1.854095331937431e-06, + "loss": 0.3233, + "step": 31062 + }, + { + "epoch": 11.47, + "learning_rate": 1.8540583815740856e-06, + "loss": 0.3341, + "step": 31063 + }, + { + "epoch": 11.47, + "learning_rate": 1.8540214312107404e-06, + "loss": 0.3793, + "step": 31064 + }, + { + "epoch": 11.47, + "learning_rate": 1.853984480847395e-06, + "loss": 0.3666, + "step": 31065 + }, + { + "epoch": 11.47, + "learning_rate": 1.8539475304840499e-06, + "loss": 0.4119, + "step": 31066 + }, + { + "epoch": 11.47, + "learning_rate": 1.8539105801207045e-06, + "loss": 0.1964, + "step": 31067 + }, + { + "epoch": 11.47, + "learning_rate": 1.8538736297573592e-06, + "loss": 0.4419, + "step": 31068 + }, + { + "epoch": 11.47, + "learning_rate": 1.8538366793940142e-06, + "loss": 0.4068, + "step": 31069 + }, + { + "epoch": 11.47, + "learning_rate": 1.8537997290306686e-06, + "loss": 0.2506, + "step": 31070 + }, + { + "epoch": 11.47, + "learning_rate": 1.8537627786673237e-06, + "loss": 0.3551, + "step": 31071 + }, + { + "epoch": 11.47, + "learning_rate": 1.8537258283039783e-06, + "loss": 0.3089, + "step": 31072 + }, + { + "epoch": 11.47, + "learning_rate": 1.8536888779406332e-06, + "loss": 0.2381, + "step": 31073 + }, + { + "epoch": 11.47, + "learning_rate": 1.8536519275772878e-06, + "loss": 0.3543, + "step": 31074 + }, + { + "epoch": 11.48, + "learning_rate": 1.8536149772139429e-06, + "loss": 0.2814, + "step": 31075 + }, + { + "epoch": 11.48, + "learning_rate": 1.8535780268505973e-06, + "loss": 0.2827, + "step": 31076 + }, + { + "epoch": 11.48, + "learning_rate": 1.853541076487252e-06, + "loss": 0.2671, + "step": 31077 + }, + { + "epoch": 11.48, + "learning_rate": 1.853504126123907e-06, + "loss": 0.3173, + "step": 31078 + }, + { + "epoch": 11.48, + "learning_rate": 1.8534671757605616e-06, + "loss": 0.2369, + "step": 31079 + }, + { + "epoch": 11.48, + "learning_rate": 1.8534302253972165e-06, + "loss": 0.3726, + "step": 31080 + }, + { + "epoch": 11.48, + "learning_rate": 1.8533932750338711e-06, + "loss": 0.3661, + "step": 31081 + }, + { + "epoch": 11.48, + "learning_rate": 1.853356324670526e-06, + "loss": 0.3139, + "step": 31082 + }, + { + "epoch": 11.48, + "learning_rate": 1.8533193743071806e-06, + "loss": 0.2058, + "step": 31083 + }, + { + "epoch": 11.48, + "learning_rate": 1.8532824239438357e-06, + "loss": 0.3279, + "step": 31084 + }, + { + "epoch": 11.48, + "learning_rate": 1.8532454735804903e-06, + "loss": 0.3197, + "step": 31085 + }, + { + "epoch": 11.48, + "learning_rate": 1.8532085232171452e-06, + "loss": 0.2429, + "step": 31086 + }, + { + "epoch": 11.48, + "learning_rate": 1.8531715728537998e-06, + "loss": 0.2831, + "step": 31087 + }, + { + "epoch": 11.48, + "learning_rate": 1.8531346224904544e-06, + "loss": 0.2892, + "step": 31088 + }, + { + "epoch": 11.48, + "learning_rate": 1.8530976721271093e-06, + "loss": 0.2013, + "step": 31089 + }, + { + "epoch": 11.48, + "learning_rate": 1.853060721763764e-06, + "loss": 0.3881, + "step": 31090 + }, + { + "epoch": 11.48, + "learning_rate": 1.853023771400419e-06, + "loss": 0.3784, + "step": 31091 + }, + { + "epoch": 11.48, + "learning_rate": 1.8529868210370736e-06, + "loss": 0.3906, + "step": 31092 + }, + { + "epoch": 11.48, + "learning_rate": 1.8529498706737285e-06, + "loss": 0.3121, + "step": 31093 + }, + { + "epoch": 11.48, + "learning_rate": 1.852912920310383e-06, + "loss": 0.2528, + "step": 31094 + }, + { + "epoch": 11.48, + "learning_rate": 1.852875969947038e-06, + "loss": 0.2819, + "step": 31095 + }, + { + "epoch": 11.48, + "learning_rate": 1.8528390195836926e-06, + "loss": 0.3112, + "step": 31096 + }, + { + "epoch": 11.48, + "learning_rate": 1.8528020692203472e-06, + "loss": 0.3953, + "step": 31097 + }, + { + "epoch": 11.48, + "learning_rate": 1.8527651188570023e-06, + "loss": 0.1654, + "step": 31098 + }, + { + "epoch": 11.48, + "learning_rate": 1.8527281684936567e-06, + "loss": 0.2599, + "step": 31099 + }, + { + "epoch": 11.48, + "learning_rate": 1.8526912181303118e-06, + "loss": 0.2051, + "step": 31100 + }, + { + "epoch": 11.48, + "learning_rate": 1.8526542677669664e-06, + "loss": 0.303, + "step": 31101 + }, + { + "epoch": 11.49, + "learning_rate": 1.8526173174036212e-06, + "loss": 0.3167, + "step": 31102 + }, + { + "epoch": 11.49, + "learning_rate": 1.8525803670402759e-06, + "loss": 0.2342, + "step": 31103 + }, + { + "epoch": 11.49, + "learning_rate": 1.852543416676931e-06, + "loss": 0.3162, + "step": 31104 + }, + { + "epoch": 11.49, + "learning_rate": 1.8525064663135854e-06, + "loss": 0.2683, + "step": 31105 + }, + { + "epoch": 11.49, + "learning_rate": 1.8524695159502404e-06, + "loss": 0.2779, + "step": 31106 + }, + { + "epoch": 11.49, + "learning_rate": 1.852432565586895e-06, + "loss": 0.287, + "step": 31107 + }, + { + "epoch": 11.49, + "learning_rate": 1.8523956152235497e-06, + "loss": 0.2938, + "step": 31108 + }, + { + "epoch": 11.49, + "learning_rate": 1.8523586648602045e-06, + "loss": 0.4058, + "step": 31109 + }, + { + "epoch": 11.49, + "learning_rate": 1.8523217144968592e-06, + "loss": 0.2556, + "step": 31110 + }, + { + "epoch": 11.49, + "learning_rate": 1.852284764133514e-06, + "loss": 0.3017, + "step": 31111 + }, + { + "epoch": 11.49, + "learning_rate": 1.8522478137701687e-06, + "loss": 0.1879, + "step": 31112 + }, + { + "epoch": 11.49, + "learning_rate": 1.8522108634068237e-06, + "loss": 0.3672, + "step": 31113 + }, + { + "epoch": 11.49, + "learning_rate": 1.8521739130434784e-06, + "loss": 0.2456, + "step": 31114 + }, + { + "epoch": 11.49, + "learning_rate": 1.8521369626801332e-06, + "loss": 0.3913, + "step": 31115 + }, + { + "epoch": 11.49, + "learning_rate": 1.8521000123167878e-06, + "loss": 0.3291, + "step": 31116 + }, + { + "epoch": 11.49, + "learning_rate": 1.8520630619534425e-06, + "loss": 0.3763, + "step": 31117 + }, + { + "epoch": 11.49, + "learning_rate": 1.8520261115900973e-06, + "loss": 0.369, + "step": 31118 + }, + { + "epoch": 11.49, + "learning_rate": 1.851989161226752e-06, + "loss": 0.3667, + "step": 31119 + }, + { + "epoch": 11.49, + "learning_rate": 1.851952210863407e-06, + "loss": 0.2713, + "step": 31120 + }, + { + "epoch": 11.49, + "learning_rate": 1.8519152605000615e-06, + "loss": 0.2312, + "step": 31121 + }, + { + "epoch": 11.49, + "learning_rate": 1.8518783101367165e-06, + "loss": 0.2758, + "step": 31122 + }, + { + "epoch": 11.49, + "learning_rate": 1.8518413597733711e-06, + "loss": 0.2794, + "step": 31123 + }, + { + "epoch": 11.49, + "learning_rate": 1.851804409410026e-06, + "loss": 0.2707, + "step": 31124 + }, + { + "epoch": 11.49, + "learning_rate": 1.8517674590466806e-06, + "loss": 0.2657, + "step": 31125 + }, + { + "epoch": 11.49, + "learning_rate": 1.8517305086833357e-06, + "loss": 0.2504, + "step": 31126 + }, + { + "epoch": 11.49, + "learning_rate": 1.8516935583199901e-06, + "loss": 0.3315, + "step": 31127 + }, + { + "epoch": 11.49, + "learning_rate": 1.8516566079566448e-06, + "loss": 0.2786, + "step": 31128 + }, + { + "epoch": 11.5, + "learning_rate": 1.8516196575932998e-06, + "loss": 0.3984, + "step": 31129 + }, + { + "epoch": 11.5, + "learning_rate": 1.8515827072299544e-06, + "loss": 0.4913, + "step": 31130 + }, + { + "epoch": 11.5, + "learning_rate": 1.8515457568666093e-06, + "loss": 0.3574, + "step": 31131 + }, + { + "epoch": 11.5, + "learning_rate": 1.851508806503264e-06, + "loss": 0.3284, + "step": 31132 + }, + { + "epoch": 11.5, + "learning_rate": 1.8514718561399188e-06, + "loss": 0.245, + "step": 31133 + }, + { + "epoch": 11.5, + "learning_rate": 1.8514349057765734e-06, + "loss": 0.38, + "step": 31134 + }, + { + "epoch": 11.5, + "learning_rate": 1.8513979554132285e-06, + "loss": 0.2081, + "step": 31135 + }, + { + "epoch": 11.5, + "learning_rate": 1.8513610050498831e-06, + "loss": 0.3441, + "step": 31136 + }, + { + "epoch": 11.5, + "learning_rate": 1.8513240546865378e-06, + "loss": 0.3307, + "step": 31137 + }, + { + "epoch": 11.5, + "learning_rate": 1.8512871043231926e-06, + "loss": 0.2261, + "step": 31138 + }, + { + "epoch": 11.5, + "learning_rate": 1.8512501539598472e-06, + "loss": 0.375, + "step": 31139 + }, + { + "epoch": 11.5, + "learning_rate": 1.851213203596502e-06, + "loss": 0.33, + "step": 31140 + }, + { + "epoch": 11.5, + "learning_rate": 1.8511762532331567e-06, + "loss": 0.2887, + "step": 31141 + }, + { + "epoch": 11.5, + "learning_rate": 1.8511393028698118e-06, + "loss": 0.3865, + "step": 31142 + }, + { + "epoch": 11.5, + "learning_rate": 1.8511023525064662e-06, + "loss": 0.341, + "step": 31143 + }, + { + "epoch": 11.5, + "learning_rate": 1.8510654021431213e-06, + "loss": 0.3568, + "step": 31144 + }, + { + "epoch": 11.5, + "learning_rate": 1.851028451779776e-06, + "loss": 0.3047, + "step": 31145 + }, + { + "epoch": 11.5, + "learning_rate": 1.8509915014164305e-06, + "loss": 0.2392, + "step": 31146 + }, + { + "epoch": 11.5, + "learning_rate": 1.8509545510530854e-06, + "loss": 0.448, + "step": 31147 + }, + { + "epoch": 11.5, + "learning_rate": 1.85091760068974e-06, + "loss": 0.2945, + "step": 31148 + }, + { + "epoch": 11.5, + "learning_rate": 1.8508806503263949e-06, + "loss": 0.2394, + "step": 31149 + }, + { + "epoch": 11.5, + "learning_rate": 1.8508436999630495e-06, + "loss": 0.4359, + "step": 31150 + }, + { + "epoch": 11.5, + "learning_rate": 1.8508067495997046e-06, + "loss": 0.312, + "step": 31151 + }, + { + "epoch": 11.5, + "learning_rate": 1.8507697992363592e-06, + "loss": 0.2375, + "step": 31152 + }, + { + "epoch": 11.5, + "learning_rate": 1.850732848873014e-06, + "loss": 0.3635, + "step": 31153 + }, + { + "epoch": 11.5, + "learning_rate": 1.8506958985096687e-06, + "loss": 0.3535, + "step": 31154 + }, + { + "epoch": 11.5, + "learning_rate": 1.8506589481463235e-06, + "loss": 0.3095, + "step": 31155 + }, + { + "epoch": 11.51, + "learning_rate": 1.8506219977829782e-06, + "loss": 0.2624, + "step": 31156 + }, + { + "epoch": 11.51, + "learning_rate": 1.8505850474196328e-06, + "loss": 0.3019, + "step": 31157 + }, + { + "epoch": 11.51, + "learning_rate": 1.8505480970562879e-06, + "loss": 0.2671, + "step": 31158 + }, + { + "epoch": 11.51, + "learning_rate": 1.8505111466929425e-06, + "loss": 0.3241, + "step": 31159 + }, + { + "epoch": 11.51, + "learning_rate": 1.8504741963295974e-06, + "loss": 0.4004, + "step": 31160 + }, + { + "epoch": 11.51, + "learning_rate": 1.850437245966252e-06, + "loss": 0.2837, + "step": 31161 + }, + { + "epoch": 11.51, + "learning_rate": 1.8504002956029068e-06, + "loss": 0.3186, + "step": 31162 + }, + { + "epoch": 11.51, + "learning_rate": 1.8503633452395615e-06, + "loss": 0.2593, + "step": 31163 + }, + { + "epoch": 11.51, + "learning_rate": 1.8503263948762165e-06, + "loss": 0.3148, + "step": 31164 + }, + { + "epoch": 11.51, + "learning_rate": 1.8502894445128712e-06, + "loss": 0.3193, + "step": 31165 + }, + { + "epoch": 11.51, + "learning_rate": 1.8502524941495256e-06, + "loss": 0.2273, + "step": 31166 + }, + { + "epoch": 11.51, + "learning_rate": 1.8502155437861807e-06, + "loss": 0.4295, + "step": 31167 + }, + { + "epoch": 11.51, + "learning_rate": 1.8501785934228353e-06, + "loss": 0.2935, + "step": 31168 + }, + { + "epoch": 11.51, + "learning_rate": 1.8501416430594901e-06, + "loss": 0.2875, + "step": 31169 + }, + { + "epoch": 11.51, + "learning_rate": 1.8501046926961448e-06, + "loss": 0.2897, + "step": 31170 + }, + { + "epoch": 11.51, + "learning_rate": 1.8500677423327998e-06, + "loss": 0.3601, + "step": 31171 + }, + { + "epoch": 11.51, + "learning_rate": 1.8500307919694543e-06, + "loss": 0.3776, + "step": 31172 + }, + { + "epoch": 11.51, + "learning_rate": 1.8499938416061093e-06, + "loss": 0.4068, + "step": 31173 + }, + { + "epoch": 11.51, + "learning_rate": 1.849956891242764e-06, + "loss": 0.3123, + "step": 31174 + }, + { + "epoch": 11.51, + "learning_rate": 1.8499199408794188e-06, + "loss": 0.3383, + "step": 31175 + }, + { + "epoch": 11.51, + "learning_rate": 1.8498829905160734e-06, + "loss": 0.3602, + "step": 31176 + }, + { + "epoch": 11.51, + "learning_rate": 1.849846040152728e-06, + "loss": 0.3176, + "step": 31177 + }, + { + "epoch": 11.51, + "learning_rate": 1.849809089789383e-06, + "loss": 0.3353, + "step": 31178 + }, + { + "epoch": 11.51, + "learning_rate": 1.8497721394260376e-06, + "loss": 0.2596, + "step": 31179 + }, + { + "epoch": 11.51, + "learning_rate": 1.8497351890626926e-06, + "loss": 0.3139, + "step": 31180 + }, + { + "epoch": 11.51, + "learning_rate": 1.8496982386993473e-06, + "loss": 0.3311, + "step": 31181 + }, + { + "epoch": 11.51, + "learning_rate": 1.8496612883360021e-06, + "loss": 0.3, + "step": 31182 + }, + { + "epoch": 11.52, + "learning_rate": 1.8496243379726567e-06, + "loss": 0.2819, + "step": 31183 + }, + { + "epoch": 11.52, + "learning_rate": 1.8495873876093116e-06, + "loss": 0.2664, + "step": 31184 + }, + { + "epoch": 11.52, + "learning_rate": 1.8495504372459662e-06, + "loss": 0.2449, + "step": 31185 + }, + { + "epoch": 11.52, + "learning_rate": 1.8495134868826209e-06, + "loss": 0.4547, + "step": 31186 + }, + { + "epoch": 11.52, + "learning_rate": 1.849476536519276e-06, + "loss": 0.2281, + "step": 31187 + }, + { + "epoch": 11.52, + "learning_rate": 1.8494395861559304e-06, + "loss": 0.2033, + "step": 31188 + }, + { + "epoch": 11.52, + "learning_rate": 1.8494026357925854e-06, + "loss": 0.3487, + "step": 31189 + }, + { + "epoch": 11.52, + "learning_rate": 1.84936568542924e-06, + "loss": 0.3143, + "step": 31190 + }, + { + "epoch": 11.52, + "learning_rate": 1.849328735065895e-06, + "loss": 0.2691, + "step": 31191 + }, + { + "epoch": 11.52, + "learning_rate": 1.8492917847025495e-06, + "loss": 0.4226, + "step": 31192 + }, + { + "epoch": 11.52, + "learning_rate": 1.8492548343392046e-06, + "loss": 0.3593, + "step": 31193 + }, + { + "epoch": 11.52, + "learning_rate": 1.849217883975859e-06, + "loss": 0.3662, + "step": 31194 + }, + { + "epoch": 11.52, + "learning_rate": 1.849180933612514e-06, + "loss": 0.2086, + "step": 31195 + }, + { + "epoch": 11.52, + "learning_rate": 1.8491439832491687e-06, + "loss": 0.284, + "step": 31196 + }, + { + "epoch": 11.52, + "learning_rate": 1.8491070328858233e-06, + "loss": 0.3139, + "step": 31197 + }, + { + "epoch": 11.52, + "learning_rate": 1.8490700825224782e-06, + "loss": 0.3083, + "step": 31198 + }, + { + "epoch": 11.52, + "learning_rate": 1.8490331321591328e-06, + "loss": 0.2991, + "step": 31199 + }, + { + "epoch": 11.52, + "learning_rate": 1.8489961817957877e-06, + "loss": 0.3229, + "step": 31200 + }, + { + "epoch": 11.52, + "learning_rate": 1.8489592314324423e-06, + "loss": 0.3524, + "step": 31201 + }, + { + "epoch": 11.52, + "learning_rate": 1.8489222810690974e-06, + "loss": 0.4081, + "step": 31202 + }, + { + "epoch": 11.52, + "learning_rate": 1.848885330705752e-06, + "loss": 0.2732, + "step": 31203 + }, + { + "epoch": 11.52, + "learning_rate": 1.8488483803424069e-06, + "loss": 0.271, + "step": 31204 + }, + { + "epoch": 11.52, + "learning_rate": 1.8488114299790615e-06, + "loss": 0.4096, + "step": 31205 + }, + { + "epoch": 11.52, + "learning_rate": 1.8487744796157161e-06, + "loss": 0.3462, + "step": 31206 + }, + { + "epoch": 11.52, + "learning_rate": 1.848737529252371e-06, + "loss": 0.3324, + "step": 31207 + }, + { + "epoch": 11.52, + "learning_rate": 1.8487005788890256e-06, + "loss": 0.4406, + "step": 31208 + }, + { + "epoch": 11.52, + "learning_rate": 1.8486636285256807e-06, + "loss": 0.2774, + "step": 31209 + }, + { + "epoch": 11.53, + "learning_rate": 1.8486266781623353e-06, + "loss": 0.351, + "step": 31210 + }, + { + "epoch": 11.53, + "learning_rate": 1.8485897277989902e-06, + "loss": 0.2681, + "step": 31211 + }, + { + "epoch": 11.53, + "learning_rate": 1.8485527774356448e-06, + "loss": 0.2115, + "step": 31212 + }, + { + "epoch": 11.53, + "learning_rate": 1.8485158270722997e-06, + "loss": 0.3638, + "step": 31213 + }, + { + "epoch": 11.53, + "learning_rate": 1.8484788767089543e-06, + "loss": 0.2605, + "step": 31214 + }, + { + "epoch": 11.53, + "learning_rate": 1.8484419263456093e-06, + "loss": 0.4331, + "step": 31215 + }, + { + "epoch": 11.53, + "learning_rate": 1.848404975982264e-06, + "loss": 0.3479, + "step": 31216 + }, + { + "epoch": 11.53, + "learning_rate": 1.8483680256189184e-06, + "loss": 0.2961, + "step": 31217 + }, + { + "epoch": 11.53, + "learning_rate": 1.8483310752555735e-06, + "loss": 0.2931, + "step": 31218 + }, + { + "epoch": 11.53, + "learning_rate": 1.848294124892228e-06, + "loss": 0.4358, + "step": 31219 + }, + { + "epoch": 11.53, + "learning_rate": 1.848257174528883e-06, + "loss": 0.3509, + "step": 31220 + }, + { + "epoch": 11.53, + "learning_rate": 1.8482202241655376e-06, + "loss": 0.2652, + "step": 31221 + }, + { + "epoch": 11.53, + "learning_rate": 1.8481832738021924e-06, + "loss": 0.3463, + "step": 31222 + }, + { + "epoch": 11.53, + "learning_rate": 1.848146323438847e-06, + "loss": 0.2043, + "step": 31223 + }, + { + "epoch": 11.53, + "learning_rate": 1.8481093730755021e-06, + "loss": 0.3282, + "step": 31224 + }, + { + "epoch": 11.53, + "learning_rate": 1.8480724227121568e-06, + "loss": 0.251, + "step": 31225 + }, + { + "epoch": 11.53, + "learning_rate": 1.8480354723488114e-06, + "loss": 0.2031, + "step": 31226 + }, + { + "epoch": 11.53, + "learning_rate": 1.8479985219854663e-06, + "loss": 0.2972, + "step": 31227 + }, + { + "epoch": 11.53, + "learning_rate": 1.8479615716221209e-06, + "loss": 0.5243, + "step": 31228 + }, + { + "epoch": 11.53, + "learning_rate": 1.8479246212587757e-06, + "loss": 0.2806, + "step": 31229 + }, + { + "epoch": 11.53, + "learning_rate": 1.8478876708954304e-06, + "loss": 0.3484, + "step": 31230 + }, + { + "epoch": 11.53, + "learning_rate": 1.8478507205320854e-06, + "loss": 0.2866, + "step": 31231 + }, + { + "epoch": 11.53, + "learning_rate": 1.84781377016874e-06, + "loss": 0.2973, + "step": 31232 + }, + { + "epoch": 11.53, + "learning_rate": 1.847776819805395e-06, + "loss": 0.3099, + "step": 31233 + }, + { + "epoch": 11.53, + "learning_rate": 1.8477398694420496e-06, + "loss": 0.3323, + "step": 31234 + }, + { + "epoch": 11.53, + "learning_rate": 1.8477029190787044e-06, + "loss": 0.4805, + "step": 31235 + }, + { + "epoch": 11.53, + "learning_rate": 1.847665968715359e-06, + "loss": 0.2786, + "step": 31236 + }, + { + "epoch": 11.54, + "learning_rate": 1.8476290183520137e-06, + "loss": 0.3037, + "step": 31237 + }, + { + "epoch": 11.54, + "learning_rate": 1.8475920679886687e-06, + "loss": 0.2029, + "step": 31238 + }, + { + "epoch": 11.54, + "learning_rate": 1.8475551176253232e-06, + "loss": 0.3365, + "step": 31239 + }, + { + "epoch": 11.54, + "learning_rate": 1.8475181672619782e-06, + "loss": 0.3446, + "step": 31240 + }, + { + "epoch": 11.54, + "learning_rate": 1.8474812168986329e-06, + "loss": 0.3828, + "step": 31241 + }, + { + "epoch": 11.54, + "learning_rate": 1.8474442665352877e-06, + "loss": 0.3339, + "step": 31242 + }, + { + "epoch": 11.54, + "learning_rate": 1.8474073161719423e-06, + "loss": 0.3156, + "step": 31243 + }, + { + "epoch": 11.54, + "learning_rate": 1.8473703658085974e-06, + "loss": 0.3678, + "step": 31244 + }, + { + "epoch": 11.54, + "learning_rate": 1.8473334154452518e-06, + "loss": 0.3744, + "step": 31245 + }, + { + "epoch": 11.54, + "learning_rate": 1.8472964650819065e-06, + "loss": 0.3016, + "step": 31246 + }, + { + "epoch": 11.54, + "learning_rate": 1.8472595147185615e-06, + "loss": 0.3714, + "step": 31247 + }, + { + "epoch": 11.54, + "learning_rate": 1.8472225643552162e-06, + "loss": 0.2767, + "step": 31248 + }, + { + "epoch": 11.54, + "learning_rate": 1.847185613991871e-06, + "loss": 0.328, + "step": 31249 + }, + { + "epoch": 11.54, + "learning_rate": 1.8471486636285256e-06, + "loss": 0.1979, + "step": 31250 + }, + { + "epoch": 11.54, + "learning_rate": 1.8471117132651805e-06, + "loss": 0.305, + "step": 31251 + }, + { + "epoch": 11.54, + "learning_rate": 1.8470747629018351e-06, + "loss": 0.4372, + "step": 31252 + }, + { + "epoch": 11.54, + "learning_rate": 1.8470378125384902e-06, + "loss": 0.3411, + "step": 31253 + }, + { + "epoch": 11.54, + "learning_rate": 1.8470008621751448e-06, + "loss": 0.2626, + "step": 31254 + }, + { + "epoch": 11.54, + "learning_rate": 1.8469639118117997e-06, + "loss": 0.2927, + "step": 31255 + }, + { + "epoch": 11.54, + "learning_rate": 1.8469269614484543e-06, + "loss": 0.361, + "step": 31256 + }, + { + "epoch": 11.54, + "learning_rate": 1.846890011085109e-06, + "loss": 0.2036, + "step": 31257 + }, + { + "epoch": 11.54, + "learning_rate": 1.8468530607217638e-06, + "loss": 0.3223, + "step": 31258 + }, + { + "epoch": 11.54, + "learning_rate": 1.8468161103584184e-06, + "loss": 0.1835, + "step": 31259 + }, + { + "epoch": 11.54, + "learning_rate": 1.8467791599950735e-06, + "loss": 0.5252, + "step": 31260 + }, + { + "epoch": 11.54, + "learning_rate": 1.846742209631728e-06, + "loss": 0.3759, + "step": 31261 + }, + { + "epoch": 11.54, + "learning_rate": 1.846705259268383e-06, + "loss": 0.3359, + "step": 31262 + }, + { + "epoch": 11.54, + "learning_rate": 1.8466683089050376e-06, + "loss": 0.2378, + "step": 31263 + }, + { + "epoch": 11.55, + "learning_rate": 1.8466313585416925e-06, + "loss": 0.3177, + "step": 31264 + }, + { + "epoch": 11.55, + "learning_rate": 1.846594408178347e-06, + "loss": 0.2308, + "step": 31265 + }, + { + "epoch": 11.55, + "learning_rate": 1.8465574578150017e-06, + "loss": 0.3838, + "step": 31266 + }, + { + "epoch": 11.55, + "learning_rate": 1.8465205074516566e-06, + "loss": 0.3748, + "step": 31267 + }, + { + "epoch": 11.55, + "learning_rate": 1.8464835570883112e-06, + "loss": 0.2949, + "step": 31268 + }, + { + "epoch": 11.55, + "learning_rate": 1.8464466067249663e-06, + "loss": 0.364, + "step": 31269 + }, + { + "epoch": 11.55, + "learning_rate": 1.846409656361621e-06, + "loss": 0.1482, + "step": 31270 + }, + { + "epoch": 11.55, + "learning_rate": 1.8463727059982758e-06, + "loss": 0.2948, + "step": 31271 + }, + { + "epoch": 11.55, + "learning_rate": 1.8463357556349304e-06, + "loss": 0.3099, + "step": 31272 + }, + { + "epoch": 11.55, + "learning_rate": 1.8462988052715852e-06, + "loss": 0.2088, + "step": 31273 + }, + { + "epoch": 11.55, + "learning_rate": 1.8462618549082399e-06, + "loss": 0.3431, + "step": 31274 + }, + { + "epoch": 11.55, + "learning_rate": 1.846224904544895e-06, + "loss": 0.2584, + "step": 31275 + }, + { + "epoch": 11.55, + "learning_rate": 1.8461879541815496e-06, + "loss": 0.3144, + "step": 31276 + }, + { + "epoch": 11.55, + "learning_rate": 1.8461510038182042e-06, + "loss": 0.3634, + "step": 31277 + }, + { + "epoch": 11.55, + "learning_rate": 1.846114053454859e-06, + "loss": 0.3669, + "step": 31278 + }, + { + "epoch": 11.55, + "learning_rate": 1.8460771030915137e-06, + "loss": 0.4092, + "step": 31279 + }, + { + "epoch": 11.55, + "learning_rate": 1.8460401527281685e-06, + "loss": 0.4921, + "step": 31280 + }, + { + "epoch": 11.55, + "learning_rate": 1.8460032023648232e-06, + "loss": 0.4178, + "step": 31281 + }, + { + "epoch": 11.55, + "learning_rate": 1.8459662520014782e-06, + "loss": 0.4309, + "step": 31282 + }, + { + "epoch": 11.55, + "learning_rate": 1.8459293016381329e-06, + "loss": 0.3471, + "step": 31283 + }, + { + "epoch": 11.55, + "learning_rate": 1.8458923512747877e-06, + "loss": 0.336, + "step": 31284 + }, + { + "epoch": 11.55, + "learning_rate": 1.8458554009114424e-06, + "loss": 0.3112, + "step": 31285 + }, + { + "epoch": 11.55, + "learning_rate": 1.845818450548097e-06, + "loss": 0.3789, + "step": 31286 + }, + { + "epoch": 11.55, + "learning_rate": 1.8457815001847519e-06, + "loss": 0.2638, + "step": 31287 + }, + { + "epoch": 11.55, + "learning_rate": 1.8457445498214065e-06, + "loss": 0.2412, + "step": 31288 + }, + { + "epoch": 11.55, + "learning_rate": 1.8457075994580615e-06, + "loss": 0.3207, + "step": 31289 + }, + { + "epoch": 11.55, + "learning_rate": 1.845670649094716e-06, + "loss": 0.2583, + "step": 31290 + }, + { + "epoch": 11.56, + "learning_rate": 1.845633698731371e-06, + "loss": 0.3684, + "step": 31291 + }, + { + "epoch": 11.56, + "learning_rate": 1.8455967483680257e-06, + "loss": 0.3838, + "step": 31292 + }, + { + "epoch": 11.56, + "learning_rate": 1.8455597980046805e-06, + "loss": 0.3622, + "step": 31293 + }, + { + "epoch": 11.56, + "learning_rate": 1.8455228476413352e-06, + "loss": 0.4053, + "step": 31294 + }, + { + "epoch": 11.56, + "learning_rate": 1.8454858972779902e-06, + "loss": 0.38, + "step": 31295 + }, + { + "epoch": 11.56, + "learning_rate": 1.8454489469146446e-06, + "loss": 0.3072, + "step": 31296 + }, + { + "epoch": 11.56, + "learning_rate": 1.8454119965512993e-06, + "loss": 0.2791, + "step": 31297 + }, + { + "epoch": 11.56, + "learning_rate": 1.8453750461879543e-06, + "loss": 0.3409, + "step": 31298 + }, + { + "epoch": 11.56, + "learning_rate": 1.845338095824609e-06, + "loss": 0.3055, + "step": 31299 + }, + { + "epoch": 11.56, + "learning_rate": 1.8453011454612638e-06, + "loss": 0.3618, + "step": 31300 + }, + { + "epoch": 11.56, + "learning_rate": 1.8452641950979185e-06, + "loss": 0.3274, + "step": 31301 + }, + { + "epoch": 11.56, + "learning_rate": 1.8452272447345733e-06, + "loss": 0.4408, + "step": 31302 + }, + { + "epoch": 11.56, + "learning_rate": 1.845190294371228e-06, + "loss": 0.2395, + "step": 31303 + }, + { + "epoch": 11.56, + "learning_rate": 1.845153344007883e-06, + "loss": 0.3081, + "step": 31304 + }, + { + "epoch": 11.56, + "learning_rate": 1.8451163936445376e-06, + "loss": 0.224, + "step": 31305 + }, + { + "epoch": 11.56, + "learning_rate": 1.845079443281192e-06, + "loss": 0.2806, + "step": 31306 + }, + { + "epoch": 11.56, + "learning_rate": 1.8450424929178471e-06, + "loss": 0.4192, + "step": 31307 + }, + { + "epoch": 11.56, + "learning_rate": 1.8450055425545018e-06, + "loss": 0.3977, + "step": 31308 + }, + { + "epoch": 11.56, + "learning_rate": 1.8449685921911566e-06, + "loss": 0.3396, + "step": 31309 + }, + { + "epoch": 11.56, + "learning_rate": 1.8449316418278112e-06, + "loss": 0.2996, + "step": 31310 + }, + { + "epoch": 11.56, + "learning_rate": 1.8448946914644663e-06, + "loss": 0.2885, + "step": 31311 + }, + { + "epoch": 11.56, + "learning_rate": 1.8448577411011207e-06, + "loss": 0.3145, + "step": 31312 + }, + { + "epoch": 11.56, + "learning_rate": 1.8448207907377758e-06, + "loss": 0.4859, + "step": 31313 + }, + { + "epoch": 11.56, + "learning_rate": 1.8447838403744304e-06, + "loss": 0.3443, + "step": 31314 + }, + { + "epoch": 11.56, + "learning_rate": 1.8447468900110853e-06, + "loss": 0.2895, + "step": 31315 + }, + { + "epoch": 11.56, + "learning_rate": 1.84470993964774e-06, + "loss": 0.3198, + "step": 31316 + }, + { + "epoch": 11.56, + "learning_rate": 1.8446729892843945e-06, + "loss": 0.3058, + "step": 31317 + }, + { + "epoch": 11.56, + "learning_rate": 1.8446360389210494e-06, + "loss": 0.2905, + "step": 31318 + }, + { + "epoch": 11.57, + "learning_rate": 1.844599088557704e-06, + "loss": 0.4259, + "step": 31319 + }, + { + "epoch": 11.57, + "learning_rate": 1.844562138194359e-06, + "loss": 0.2938, + "step": 31320 + }, + { + "epoch": 11.57, + "learning_rate": 1.8445251878310137e-06, + "loss": 0.251, + "step": 31321 + }, + { + "epoch": 11.57, + "learning_rate": 1.8444882374676686e-06, + "loss": 0.2744, + "step": 31322 + }, + { + "epoch": 11.57, + "learning_rate": 1.8444512871043232e-06, + "loss": 0.2235, + "step": 31323 + }, + { + "epoch": 11.57, + "learning_rate": 1.844414336740978e-06, + "loss": 0.3371, + "step": 31324 + }, + { + "epoch": 11.57, + "learning_rate": 1.8443773863776327e-06, + "loss": 0.1835, + "step": 31325 + }, + { + "epoch": 11.57, + "learning_rate": 1.8443404360142873e-06, + "loss": 0.2647, + "step": 31326 + }, + { + "epoch": 11.57, + "learning_rate": 1.8443034856509424e-06, + "loss": 0.3042, + "step": 31327 + }, + { + "epoch": 11.57, + "learning_rate": 1.844266535287597e-06, + "loss": 0.4163, + "step": 31328 + }, + { + "epoch": 11.57, + "learning_rate": 1.8442295849242519e-06, + "loss": 0.2448, + "step": 31329 + }, + { + "epoch": 11.57, + "learning_rate": 1.8441926345609065e-06, + "loss": 0.3079, + "step": 31330 + }, + { + "epoch": 11.57, + "learning_rate": 1.8441556841975614e-06, + "loss": 0.2762, + "step": 31331 + }, + { + "epoch": 11.57, + "learning_rate": 1.844118733834216e-06, + "loss": 0.3969, + "step": 31332 + }, + { + "epoch": 11.57, + "learning_rate": 1.844081783470871e-06, + "loss": 0.2274, + "step": 31333 + }, + { + "epoch": 11.57, + "learning_rate": 1.8440448331075255e-06, + "loss": 0.4174, + "step": 31334 + }, + { + "epoch": 11.57, + "learning_rate": 1.8440078827441805e-06, + "loss": 0.3792, + "step": 31335 + }, + { + "epoch": 11.57, + "learning_rate": 1.8439709323808352e-06, + "loss": 0.256, + "step": 31336 + }, + { + "epoch": 11.57, + "learning_rate": 1.8439339820174898e-06, + "loss": 0.3904, + "step": 31337 + }, + { + "epoch": 11.57, + "learning_rate": 1.8438970316541447e-06, + "loss": 0.2352, + "step": 31338 + }, + { + "epoch": 11.57, + "learning_rate": 1.8438600812907993e-06, + "loss": 0.2778, + "step": 31339 + }, + { + "epoch": 11.57, + "learning_rate": 1.8438231309274541e-06, + "loss": 0.3225, + "step": 31340 + }, + { + "epoch": 11.57, + "learning_rate": 1.8437861805641088e-06, + "loss": 0.3531, + "step": 31341 + }, + { + "epoch": 11.57, + "learning_rate": 1.8437492302007638e-06, + "loss": 0.2582, + "step": 31342 + }, + { + "epoch": 11.57, + "learning_rate": 1.8437122798374185e-06, + "loss": 0.247, + "step": 31343 + }, + { + "epoch": 11.57, + "learning_rate": 1.8436753294740733e-06, + "loss": 0.3279, + "step": 31344 + }, + { + "epoch": 11.57, + "learning_rate": 1.843638379110728e-06, + "loss": 0.281, + "step": 31345 + }, + { + "epoch": 11.58, + "learning_rate": 1.8436014287473826e-06, + "loss": 0.4255, + "step": 31346 + }, + { + "epoch": 11.58, + "learning_rate": 1.8435644783840374e-06, + "loss": 0.2902, + "step": 31347 + }, + { + "epoch": 11.58, + "learning_rate": 1.843527528020692e-06, + "loss": 0.3539, + "step": 31348 + }, + { + "epoch": 11.58, + "learning_rate": 1.8434905776573471e-06, + "loss": 0.3107, + "step": 31349 + }, + { + "epoch": 11.58, + "learning_rate": 1.8434536272940018e-06, + "loss": 0.279, + "step": 31350 + }, + { + "epoch": 11.58, + "learning_rate": 1.8434166769306566e-06, + "loss": 0.2325, + "step": 31351 + }, + { + "epoch": 11.58, + "learning_rate": 1.8433797265673113e-06, + "loss": 0.2864, + "step": 31352 + }, + { + "epoch": 11.58, + "learning_rate": 1.8433427762039661e-06, + "loss": 0.2522, + "step": 31353 + }, + { + "epoch": 11.58, + "learning_rate": 1.8433058258406208e-06, + "loss": 0.2463, + "step": 31354 + }, + { + "epoch": 11.58, + "learning_rate": 1.8432688754772758e-06, + "loss": 0.2113, + "step": 31355 + }, + { + "epoch": 11.58, + "learning_rate": 1.8432319251139304e-06, + "loss": 0.239, + "step": 31356 + }, + { + "epoch": 11.58, + "learning_rate": 1.8431949747505849e-06, + "loss": 0.3548, + "step": 31357 + }, + { + "epoch": 11.58, + "learning_rate": 1.84315802438724e-06, + "loss": 0.2755, + "step": 31358 + }, + { + "epoch": 11.58, + "learning_rate": 1.8431210740238946e-06, + "loss": 0.28, + "step": 31359 + }, + { + "epoch": 11.58, + "learning_rate": 1.8430841236605494e-06, + "loss": 0.2671, + "step": 31360 + }, + { + "epoch": 11.58, + "learning_rate": 1.843047173297204e-06, + "loss": 0.2583, + "step": 31361 + }, + { + "epoch": 11.58, + "learning_rate": 1.8430102229338591e-06, + "loss": 0.36, + "step": 31362 + }, + { + "epoch": 11.58, + "learning_rate": 1.8429732725705135e-06, + "loss": 0.2576, + "step": 31363 + }, + { + "epoch": 11.58, + "learning_rate": 1.8429363222071686e-06, + "loss": 0.4628, + "step": 31364 + }, + { + "epoch": 11.58, + "learning_rate": 1.8428993718438232e-06, + "loss": 0.2593, + "step": 31365 + }, + { + "epoch": 11.58, + "learning_rate": 1.8428624214804779e-06, + "loss": 0.3133, + "step": 31366 + }, + { + "epoch": 11.58, + "learning_rate": 1.8428254711171327e-06, + "loss": 0.3715, + "step": 31367 + }, + { + "epoch": 11.58, + "learning_rate": 1.8427885207537874e-06, + "loss": 0.3324, + "step": 31368 + }, + { + "epoch": 11.58, + "learning_rate": 1.8427515703904422e-06, + "loss": 0.3699, + "step": 31369 + }, + { + "epoch": 11.58, + "learning_rate": 1.8427146200270968e-06, + "loss": 0.5107, + "step": 31370 + }, + { + "epoch": 11.58, + "learning_rate": 1.842677669663752e-06, + "loss": 0.2172, + "step": 31371 + }, + { + "epoch": 11.58, + "learning_rate": 1.8426407193004065e-06, + "loss": 0.3631, + "step": 31372 + }, + { + "epoch": 11.59, + "learning_rate": 1.8426037689370614e-06, + "loss": 0.1892, + "step": 31373 + }, + { + "epoch": 11.59, + "learning_rate": 1.842566818573716e-06, + "loss": 0.2809, + "step": 31374 + }, + { + "epoch": 11.59, + "learning_rate": 1.8425298682103707e-06, + "loss": 0.3788, + "step": 31375 + }, + { + "epoch": 11.59, + "learning_rate": 1.8424929178470255e-06, + "loss": 0.4042, + "step": 31376 + }, + { + "epoch": 11.59, + "learning_rate": 1.8424559674836801e-06, + "loss": 0.4151, + "step": 31377 + }, + { + "epoch": 11.59, + "learning_rate": 1.8424190171203352e-06, + "loss": 0.2777, + "step": 31378 + }, + { + "epoch": 11.59, + "learning_rate": 1.8423820667569896e-06, + "loss": 0.3794, + "step": 31379 + }, + { + "epoch": 11.59, + "learning_rate": 1.8423451163936447e-06, + "loss": 0.4299, + "step": 31380 + }, + { + "epoch": 11.59, + "learning_rate": 1.8423081660302993e-06, + "loss": 0.326, + "step": 31381 + }, + { + "epoch": 11.59, + "learning_rate": 1.8422712156669542e-06, + "loss": 0.2992, + "step": 31382 + }, + { + "epoch": 11.59, + "learning_rate": 1.8422342653036088e-06, + "loss": 0.2323, + "step": 31383 + }, + { + "epoch": 11.59, + "learning_rate": 1.8421973149402639e-06, + "loss": 0.2876, + "step": 31384 + }, + { + "epoch": 11.59, + "learning_rate": 1.8421603645769183e-06, + "loss": 0.286, + "step": 31385 + }, + { + "epoch": 11.59, + "learning_rate": 1.842123414213573e-06, + "loss": 0.4652, + "step": 31386 + }, + { + "epoch": 11.59, + "learning_rate": 1.842086463850228e-06, + "loss": 0.3594, + "step": 31387 + }, + { + "epoch": 11.59, + "learning_rate": 1.8420495134868826e-06, + "loss": 0.2168, + "step": 31388 + }, + { + "epoch": 11.59, + "learning_rate": 1.8420125631235375e-06, + "loss": 0.257, + "step": 31389 + }, + { + "epoch": 11.59, + "learning_rate": 1.8419756127601921e-06, + "loss": 0.3409, + "step": 31390 + }, + { + "epoch": 11.59, + "learning_rate": 1.841938662396847e-06, + "loss": 0.3117, + "step": 31391 + }, + { + "epoch": 11.59, + "learning_rate": 1.8419017120335016e-06, + "loss": 0.4144, + "step": 31392 + }, + { + "epoch": 11.59, + "learning_rate": 1.8418647616701567e-06, + "loss": 0.316, + "step": 31393 + }, + { + "epoch": 11.59, + "learning_rate": 1.8418278113068113e-06, + "loss": 0.3866, + "step": 31394 + }, + { + "epoch": 11.59, + "learning_rate": 1.841790860943466e-06, + "loss": 0.2581, + "step": 31395 + }, + { + "epoch": 11.59, + "learning_rate": 1.8417539105801208e-06, + "loss": 0.2685, + "step": 31396 + }, + { + "epoch": 11.59, + "learning_rate": 1.8417169602167754e-06, + "loss": 0.3412, + "step": 31397 + }, + { + "epoch": 11.59, + "learning_rate": 1.8416800098534303e-06, + "loss": 0.3054, + "step": 31398 + }, + { + "epoch": 11.59, + "learning_rate": 1.841643059490085e-06, + "loss": 0.2788, + "step": 31399 + }, + { + "epoch": 11.6, + "learning_rate": 1.84160610912674e-06, + "loss": 0.2373, + "step": 31400 + }, + { + "epoch": 11.6, + "learning_rate": 1.8415691587633946e-06, + "loss": 0.4115, + "step": 31401 + }, + { + "epoch": 11.6, + "learning_rate": 1.8415322084000494e-06, + "loss": 0.4067, + "step": 31402 + }, + { + "epoch": 11.6, + "learning_rate": 1.841495258036704e-06, + "loss": 0.3109, + "step": 31403 + }, + { + "epoch": 11.6, + "learning_rate": 1.841458307673359e-06, + "loss": 0.2692, + "step": 31404 + }, + { + "epoch": 11.6, + "learning_rate": 1.8414213573100136e-06, + "loss": 0.3652, + "step": 31405 + }, + { + "epoch": 11.6, + "learning_rate": 1.8413844069466682e-06, + "loss": 0.3075, + "step": 31406 + }, + { + "epoch": 11.6, + "learning_rate": 1.841347456583323e-06, + "loss": 0.3587, + "step": 31407 + }, + { + "epoch": 11.6, + "learning_rate": 1.8413105062199777e-06, + "loss": 0.2815, + "step": 31408 + }, + { + "epoch": 11.6, + "learning_rate": 1.8412735558566327e-06, + "loss": 0.2253, + "step": 31409 + }, + { + "epoch": 11.6, + "learning_rate": 1.8412366054932874e-06, + "loss": 0.2713, + "step": 31410 + }, + { + "epoch": 11.6, + "learning_rate": 1.8411996551299422e-06, + "loss": 0.3314, + "step": 31411 + }, + { + "epoch": 11.6, + "learning_rate": 1.8411627047665969e-06, + "loss": 0.3855, + "step": 31412 + }, + { + "epoch": 11.6, + "learning_rate": 1.8411257544032517e-06, + "loss": 0.2537, + "step": 31413 + }, + { + "epoch": 11.6, + "learning_rate": 1.8410888040399063e-06, + "loss": 0.3395, + "step": 31414 + }, + { + "epoch": 11.6, + "learning_rate": 1.841051853676561e-06, + "loss": 0.3676, + "step": 31415 + }, + { + "epoch": 11.6, + "learning_rate": 1.841014903313216e-06, + "loss": 0.3133, + "step": 31416 + }, + { + "epoch": 11.6, + "learning_rate": 1.8409779529498707e-06, + "loss": 0.3183, + "step": 31417 + }, + { + "epoch": 11.6, + "learning_rate": 1.8409410025865255e-06, + "loss": 0.3192, + "step": 31418 + }, + { + "epoch": 11.6, + "learning_rate": 1.8409040522231802e-06, + "loss": 0.217, + "step": 31419 + }, + { + "epoch": 11.6, + "learning_rate": 1.840867101859835e-06, + "loss": 0.3802, + "step": 31420 + }, + { + "epoch": 11.6, + "learning_rate": 1.8408301514964897e-06, + "loss": 0.3634, + "step": 31421 + }, + { + "epoch": 11.6, + "learning_rate": 1.8407932011331447e-06, + "loss": 0.2581, + "step": 31422 + }, + { + "epoch": 11.6, + "learning_rate": 1.8407562507697993e-06, + "loss": 0.3143, + "step": 31423 + }, + { + "epoch": 11.6, + "learning_rate": 1.8407193004064542e-06, + "loss": 0.3194, + "step": 31424 + }, + { + "epoch": 11.6, + "learning_rate": 1.8406823500431088e-06, + "loss": 0.4082, + "step": 31425 + }, + { + "epoch": 11.6, + "learning_rate": 1.8406453996797635e-06, + "loss": 0.3089, + "step": 31426 + }, + { + "epoch": 11.61, + "learning_rate": 1.8406084493164183e-06, + "loss": 0.3024, + "step": 31427 + }, + { + "epoch": 11.61, + "learning_rate": 1.840571498953073e-06, + "loss": 0.2944, + "step": 31428 + }, + { + "epoch": 11.61, + "learning_rate": 1.840534548589728e-06, + "loss": 0.3792, + "step": 31429 + }, + { + "epoch": 11.61, + "learning_rate": 1.8404975982263824e-06, + "loss": 0.2655, + "step": 31430 + }, + { + "epoch": 11.61, + "learning_rate": 1.8404606478630375e-06, + "loss": 0.3768, + "step": 31431 + }, + { + "epoch": 11.61, + "learning_rate": 1.8404236974996921e-06, + "loss": 0.2011, + "step": 31432 + }, + { + "epoch": 11.61, + "learning_rate": 1.840386747136347e-06, + "loss": 0.3065, + "step": 31433 + }, + { + "epoch": 11.61, + "learning_rate": 1.8403497967730016e-06, + "loss": 0.1997, + "step": 31434 + }, + { + "epoch": 11.61, + "learning_rate": 1.8403128464096563e-06, + "loss": 0.3834, + "step": 31435 + }, + { + "epoch": 11.61, + "learning_rate": 1.840275896046311e-06, + "loss": 0.3067, + "step": 31436 + }, + { + "epoch": 11.61, + "learning_rate": 1.8402389456829657e-06, + "loss": 0.3168, + "step": 31437 + }, + { + "epoch": 11.61, + "learning_rate": 1.8402019953196208e-06, + "loss": 0.3791, + "step": 31438 + }, + { + "epoch": 11.61, + "learning_rate": 1.8401650449562754e-06, + "loss": 0.3114, + "step": 31439 + }, + { + "epoch": 11.61, + "learning_rate": 1.8401280945929303e-06, + "loss": 0.4206, + "step": 31440 + }, + { + "epoch": 11.61, + "learning_rate": 1.840091144229585e-06, + "loss": 0.2673, + "step": 31441 + }, + { + "epoch": 11.61, + "learning_rate": 1.8400541938662398e-06, + "loss": 0.3009, + "step": 31442 + }, + { + "epoch": 11.61, + "learning_rate": 1.8400172435028944e-06, + "loss": 0.3624, + "step": 31443 + }, + { + "epoch": 11.61, + "learning_rate": 1.8399802931395495e-06, + "loss": 0.3684, + "step": 31444 + }, + { + "epoch": 11.61, + "learning_rate": 1.839943342776204e-06, + "loss": 0.2828, + "step": 31445 + }, + { + "epoch": 11.61, + "learning_rate": 1.8399063924128585e-06, + "loss": 0.3722, + "step": 31446 + }, + { + "epoch": 11.61, + "learning_rate": 1.8398694420495136e-06, + "loss": 0.3068, + "step": 31447 + }, + { + "epoch": 11.61, + "learning_rate": 1.8398324916861682e-06, + "loss": 0.3068, + "step": 31448 + }, + { + "epoch": 11.61, + "learning_rate": 1.839795541322823e-06, + "loss": 0.2367, + "step": 31449 + }, + { + "epoch": 11.61, + "learning_rate": 1.8397585909594777e-06, + "loss": 0.3665, + "step": 31450 + }, + { + "epoch": 11.61, + "learning_rate": 1.8397216405961328e-06, + "loss": 0.3086, + "step": 31451 + }, + { + "epoch": 11.61, + "learning_rate": 1.8396846902327872e-06, + "loss": 0.3714, + "step": 31452 + }, + { + "epoch": 11.61, + "learning_rate": 1.8396477398694423e-06, + "loss": 0.3378, + "step": 31453 + }, + { + "epoch": 11.62, + "learning_rate": 1.8396107895060969e-06, + "loss": 0.3679, + "step": 31454 + }, + { + "epoch": 11.62, + "learning_rate": 1.8395738391427515e-06, + "loss": 0.269, + "step": 31455 + }, + { + "epoch": 11.62, + "learning_rate": 1.8395368887794064e-06, + "loss": 0.2827, + "step": 31456 + }, + { + "epoch": 11.62, + "learning_rate": 1.839499938416061e-06, + "loss": 0.3375, + "step": 31457 + }, + { + "epoch": 11.62, + "learning_rate": 1.8394629880527159e-06, + "loss": 0.2444, + "step": 31458 + }, + { + "epoch": 11.62, + "learning_rate": 1.8394260376893705e-06, + "loss": 0.1833, + "step": 31459 + }, + { + "epoch": 11.62, + "learning_rate": 1.8393890873260256e-06, + "loss": 0.4602, + "step": 31460 + }, + { + "epoch": 11.62, + "learning_rate": 1.8393521369626802e-06, + "loss": 0.3154, + "step": 31461 + }, + { + "epoch": 11.62, + "learning_rate": 1.839315186599335e-06, + "loss": 0.4161, + "step": 31462 + }, + { + "epoch": 11.62, + "learning_rate": 1.8392782362359897e-06, + "loss": 0.3224, + "step": 31463 + }, + { + "epoch": 11.62, + "learning_rate": 1.8392412858726445e-06, + "loss": 0.3285, + "step": 31464 + }, + { + "epoch": 11.62, + "learning_rate": 1.8392043355092992e-06, + "loss": 0.4166, + "step": 31465 + }, + { + "epoch": 11.62, + "learning_rate": 1.8391673851459538e-06, + "loss": 0.2753, + "step": 31466 + }, + { + "epoch": 11.62, + "learning_rate": 1.8391304347826089e-06, + "loss": 0.2594, + "step": 31467 + }, + { + "epoch": 11.62, + "learning_rate": 1.8390934844192635e-06, + "loss": 0.4184, + "step": 31468 + }, + { + "epoch": 11.62, + "learning_rate": 1.8390565340559183e-06, + "loss": 0.3078, + "step": 31469 + }, + { + "epoch": 11.62, + "learning_rate": 1.839019583692573e-06, + "loss": 0.3661, + "step": 31470 + }, + { + "epoch": 11.62, + "learning_rate": 1.8389826333292278e-06, + "loss": 0.3555, + "step": 31471 + }, + { + "epoch": 11.62, + "learning_rate": 1.8389456829658825e-06, + "loss": 0.2445, + "step": 31472 + }, + { + "epoch": 11.62, + "learning_rate": 1.8389087326025375e-06, + "loss": 0.2052, + "step": 31473 + }, + { + "epoch": 11.62, + "learning_rate": 1.8388717822391922e-06, + "loss": 0.2608, + "step": 31474 + }, + { + "epoch": 11.62, + "learning_rate": 1.8388348318758466e-06, + "loss": 0.2251, + "step": 31475 + }, + { + "epoch": 11.62, + "learning_rate": 1.8387978815125016e-06, + "loss": 0.2104, + "step": 31476 + }, + { + "epoch": 11.62, + "learning_rate": 1.8387609311491563e-06, + "loss": 0.3655, + "step": 31477 + }, + { + "epoch": 11.62, + "learning_rate": 1.8387239807858111e-06, + "loss": 0.3606, + "step": 31478 + }, + { + "epoch": 11.62, + "learning_rate": 1.8386870304224658e-06, + "loss": 0.3258, + "step": 31479 + }, + { + "epoch": 11.62, + "learning_rate": 1.8386500800591208e-06, + "loss": 0.3615, + "step": 31480 + }, + { + "epoch": 11.63, + "learning_rate": 1.8386131296957752e-06, + "loss": 0.2342, + "step": 31481 + }, + { + "epoch": 11.63, + "learning_rate": 1.8385761793324303e-06, + "loss": 0.3353, + "step": 31482 + }, + { + "epoch": 11.63, + "learning_rate": 1.838539228969085e-06, + "loss": 0.3232, + "step": 31483 + }, + { + "epoch": 11.63, + "learning_rate": 1.8385022786057398e-06, + "loss": 0.2701, + "step": 31484 + }, + { + "epoch": 11.63, + "learning_rate": 1.8384653282423944e-06, + "loss": 0.3796, + "step": 31485 + }, + { + "epoch": 11.63, + "learning_rate": 1.838428377879049e-06, + "loss": 0.4983, + "step": 31486 + }, + { + "epoch": 11.63, + "learning_rate": 1.838391427515704e-06, + "loss": 0.2139, + "step": 31487 + }, + { + "epoch": 11.63, + "learning_rate": 1.8383544771523586e-06, + "loss": 0.2433, + "step": 31488 + }, + { + "epoch": 11.63, + "learning_rate": 1.8383175267890136e-06, + "loss": 0.2793, + "step": 31489 + }, + { + "epoch": 11.63, + "learning_rate": 1.8382805764256682e-06, + "loss": 0.3803, + "step": 31490 + }, + { + "epoch": 11.63, + "learning_rate": 1.838243626062323e-06, + "loss": 0.444, + "step": 31491 + }, + { + "epoch": 11.63, + "learning_rate": 1.8382066756989777e-06, + "loss": 0.251, + "step": 31492 + }, + { + "epoch": 11.63, + "learning_rate": 1.8381697253356326e-06, + "loss": 0.4824, + "step": 31493 + }, + { + "epoch": 11.63, + "learning_rate": 1.8381327749722872e-06, + "loss": 0.3243, + "step": 31494 + }, + { + "epoch": 11.63, + "learning_rate": 1.8380958246089419e-06, + "loss": 0.246, + "step": 31495 + }, + { + "epoch": 11.63, + "learning_rate": 1.838058874245597e-06, + "loss": 0.4027, + "step": 31496 + }, + { + "epoch": 11.63, + "learning_rate": 1.8380219238822513e-06, + "loss": 0.2975, + "step": 31497 + }, + { + "epoch": 11.63, + "learning_rate": 1.8379849735189064e-06, + "loss": 0.2647, + "step": 31498 + }, + { + "epoch": 11.63, + "learning_rate": 1.837948023155561e-06, + "loss": 0.2952, + "step": 31499 + }, + { + "epoch": 11.63, + "learning_rate": 1.8379110727922159e-06, + "loss": 0.3293, + "step": 31500 + }, + { + "epoch": 11.63, + "learning_rate": 1.8378741224288705e-06, + "loss": 0.3319, + "step": 31501 + }, + { + "epoch": 11.63, + "learning_rate": 1.8378371720655256e-06, + "loss": 0.2847, + "step": 31502 + }, + { + "epoch": 11.63, + "learning_rate": 1.83780022170218e-06, + "loss": 0.3649, + "step": 31503 + }, + { + "epoch": 11.63, + "learning_rate": 1.837763271338835e-06, + "loss": 0.4262, + "step": 31504 + }, + { + "epoch": 11.63, + "learning_rate": 1.8377263209754897e-06, + "loss": 0.2412, + "step": 31505 + }, + { + "epoch": 11.63, + "learning_rate": 1.8376893706121443e-06, + "loss": 0.3011, + "step": 31506 + }, + { + "epoch": 11.63, + "learning_rate": 1.8376524202487992e-06, + "loss": 0.2011, + "step": 31507 + }, + { + "epoch": 11.64, + "learning_rate": 1.8376154698854538e-06, + "loss": 0.5463, + "step": 31508 + }, + { + "epoch": 11.64, + "learning_rate": 1.8375785195221087e-06, + "loss": 0.3743, + "step": 31509 + }, + { + "epoch": 11.64, + "learning_rate": 1.8375415691587633e-06, + "loss": 0.3808, + "step": 31510 + }, + { + "epoch": 11.64, + "learning_rate": 1.8375046187954184e-06, + "loss": 0.2448, + "step": 31511 + }, + { + "epoch": 11.64, + "learning_rate": 1.837467668432073e-06, + "loss": 0.2868, + "step": 31512 + }, + { + "epoch": 11.64, + "learning_rate": 1.8374307180687278e-06, + "loss": 0.3796, + "step": 31513 + }, + { + "epoch": 11.64, + "learning_rate": 1.8373937677053825e-06, + "loss": 0.3273, + "step": 31514 + }, + { + "epoch": 11.64, + "learning_rate": 1.8373568173420371e-06, + "loss": 0.274, + "step": 31515 + }, + { + "epoch": 11.64, + "learning_rate": 1.837319866978692e-06, + "loss": 0.2788, + "step": 31516 + }, + { + "epoch": 11.64, + "learning_rate": 1.8372829166153466e-06, + "loss": 0.4004, + "step": 31517 + }, + { + "epoch": 11.64, + "learning_rate": 1.8372459662520017e-06, + "loss": 0.3504, + "step": 31518 + }, + { + "epoch": 11.64, + "learning_rate": 1.837209015888656e-06, + "loss": 0.3202, + "step": 31519 + }, + { + "epoch": 11.64, + "learning_rate": 1.8371720655253112e-06, + "loss": 0.2335, + "step": 31520 + }, + { + "epoch": 11.64, + "learning_rate": 1.8371351151619658e-06, + "loss": 0.3849, + "step": 31521 + }, + { + "epoch": 11.64, + "learning_rate": 1.8370981647986206e-06, + "loss": 0.4337, + "step": 31522 + }, + { + "epoch": 11.64, + "learning_rate": 1.8370612144352753e-06, + "loss": 0.3268, + "step": 31523 + }, + { + "epoch": 11.64, + "learning_rate": 1.8370242640719303e-06, + "loss": 0.2277, + "step": 31524 + }, + { + "epoch": 11.64, + "learning_rate": 1.8369873137085848e-06, + "loss": 0.3247, + "step": 31525 + }, + { + "epoch": 11.64, + "learning_rate": 1.8369503633452394e-06, + "loss": 0.3806, + "step": 31526 + }, + { + "epoch": 11.64, + "learning_rate": 1.8369134129818945e-06, + "loss": 0.2411, + "step": 31527 + }, + { + "epoch": 11.64, + "learning_rate": 1.836876462618549e-06, + "loss": 0.2848, + "step": 31528 + }, + { + "epoch": 11.64, + "learning_rate": 1.836839512255204e-06, + "loss": 0.2282, + "step": 31529 + }, + { + "epoch": 11.64, + "learning_rate": 1.8368025618918586e-06, + "loss": 0.274, + "step": 31530 + }, + { + "epoch": 11.64, + "learning_rate": 1.8367656115285134e-06, + "loss": 0.2772, + "step": 31531 + }, + { + "epoch": 11.64, + "learning_rate": 1.836728661165168e-06, + "loss": 0.4696, + "step": 31532 + }, + { + "epoch": 11.64, + "learning_rate": 1.8366917108018231e-06, + "loss": 0.3284, + "step": 31533 + }, + { + "epoch": 11.64, + "learning_rate": 1.8366547604384778e-06, + "loss": 0.3046, + "step": 31534 + }, + { + "epoch": 11.65, + "learning_rate": 1.8366178100751324e-06, + "loss": 0.3965, + "step": 31535 + }, + { + "epoch": 11.65, + "learning_rate": 1.8365808597117872e-06, + "loss": 0.2341, + "step": 31536 + }, + { + "epoch": 11.65, + "learning_rate": 1.8365439093484419e-06, + "loss": 0.238, + "step": 31537 + }, + { + "epoch": 11.65, + "learning_rate": 1.8365069589850967e-06, + "loss": 0.2907, + "step": 31538 + }, + { + "epoch": 11.65, + "learning_rate": 1.8364700086217514e-06, + "loss": 0.321, + "step": 31539 + }, + { + "epoch": 11.65, + "learning_rate": 1.8364330582584064e-06, + "loss": 0.4029, + "step": 31540 + }, + { + "epoch": 11.65, + "learning_rate": 1.836396107895061e-06, + "loss": 0.3805, + "step": 31541 + }, + { + "epoch": 11.65, + "learning_rate": 1.836359157531716e-06, + "loss": 0.3828, + "step": 31542 + }, + { + "epoch": 11.65, + "learning_rate": 1.8363222071683705e-06, + "loss": 0.312, + "step": 31543 + }, + { + "epoch": 11.65, + "learning_rate": 1.8362852568050254e-06, + "loss": 0.4238, + "step": 31544 + }, + { + "epoch": 11.65, + "learning_rate": 1.83624830644168e-06, + "loss": 0.2118, + "step": 31545 + }, + { + "epoch": 11.65, + "learning_rate": 1.8362113560783347e-06, + "loss": 0.2398, + "step": 31546 + }, + { + "epoch": 11.65, + "learning_rate": 1.8361744057149897e-06, + "loss": 0.2661, + "step": 31547 + }, + { + "epoch": 11.65, + "learning_rate": 1.8361374553516441e-06, + "loss": 0.3917, + "step": 31548 + }, + { + "epoch": 11.65, + "learning_rate": 1.8361005049882992e-06, + "loss": 0.315, + "step": 31549 + }, + { + "epoch": 11.65, + "learning_rate": 1.8360635546249538e-06, + "loss": 0.3991, + "step": 31550 + }, + { + "epoch": 11.65, + "learning_rate": 1.8360266042616087e-06, + "loss": 0.4066, + "step": 31551 + }, + { + "epoch": 11.65, + "learning_rate": 1.8359896538982633e-06, + "loss": 0.2849, + "step": 31552 + }, + { + "epoch": 11.65, + "learning_rate": 1.8359527035349184e-06, + "loss": 0.3529, + "step": 31553 + }, + { + "epoch": 11.65, + "learning_rate": 1.8359157531715728e-06, + "loss": 0.4618, + "step": 31554 + }, + { + "epoch": 11.65, + "learning_rate": 1.8358788028082275e-06, + "loss": 0.2738, + "step": 31555 + }, + { + "epoch": 11.65, + "learning_rate": 1.8358418524448825e-06, + "loss": 0.3521, + "step": 31556 + }, + { + "epoch": 11.65, + "learning_rate": 1.8358049020815371e-06, + "loss": 0.2845, + "step": 31557 + }, + { + "epoch": 11.65, + "learning_rate": 1.835767951718192e-06, + "loss": 0.2561, + "step": 31558 + }, + { + "epoch": 11.65, + "learning_rate": 1.8357310013548466e-06, + "loss": 0.3891, + "step": 31559 + }, + { + "epoch": 11.65, + "learning_rate": 1.8356940509915015e-06, + "loss": 0.5301, + "step": 31560 + }, + { + "epoch": 11.65, + "learning_rate": 1.8356571006281561e-06, + "loss": 0.312, + "step": 31561 + }, + { + "epoch": 11.66, + "learning_rate": 1.8356201502648112e-06, + "loss": 0.4862, + "step": 31562 + }, + { + "epoch": 11.66, + "learning_rate": 1.8355831999014658e-06, + "loss": 0.2535, + "step": 31563 + }, + { + "epoch": 11.66, + "learning_rate": 1.8355462495381207e-06, + "loss": 0.2833, + "step": 31564 + }, + { + "epoch": 11.66, + "learning_rate": 1.8355092991747753e-06, + "loss": 0.3043, + "step": 31565 + }, + { + "epoch": 11.66, + "learning_rate": 1.83547234881143e-06, + "loss": 0.3545, + "step": 31566 + }, + { + "epoch": 11.66, + "learning_rate": 1.8354353984480848e-06, + "loss": 0.3283, + "step": 31567 + }, + { + "epoch": 11.66, + "learning_rate": 1.8353984480847394e-06, + "loss": 0.2762, + "step": 31568 + }, + { + "epoch": 11.66, + "learning_rate": 1.8353614977213945e-06, + "loss": 0.3381, + "step": 31569 + }, + { + "epoch": 11.66, + "learning_rate": 1.835324547358049e-06, + "loss": 0.353, + "step": 31570 + }, + { + "epoch": 11.66, + "learning_rate": 1.835287596994704e-06, + "loss": 0.2648, + "step": 31571 + }, + { + "epoch": 11.66, + "learning_rate": 1.8352506466313586e-06, + "loss": 0.244, + "step": 31572 + }, + { + "epoch": 11.66, + "learning_rate": 1.8352136962680134e-06, + "loss": 0.2834, + "step": 31573 + }, + { + "epoch": 11.66, + "learning_rate": 1.835176745904668e-06, + "loss": 0.3185, + "step": 31574 + }, + { + "epoch": 11.66, + "learning_rate": 1.8351397955413227e-06, + "loss": 0.3204, + "step": 31575 + }, + { + "epoch": 11.66, + "learning_rate": 1.8351028451779776e-06, + "loss": 0.2797, + "step": 31576 + }, + { + "epoch": 11.66, + "learning_rate": 1.8350658948146322e-06, + "loss": 0.316, + "step": 31577 + }, + { + "epoch": 11.66, + "learning_rate": 1.8350289444512873e-06, + "loss": 0.206, + "step": 31578 + }, + { + "epoch": 11.66, + "learning_rate": 1.834991994087942e-06, + "loss": 0.3173, + "step": 31579 + }, + { + "epoch": 11.66, + "learning_rate": 1.8349550437245967e-06, + "loss": 0.2618, + "step": 31580 + }, + { + "epoch": 11.66, + "learning_rate": 1.8349180933612514e-06, + "loss": 0.3423, + "step": 31581 + }, + { + "epoch": 11.66, + "learning_rate": 1.8348811429979062e-06, + "loss": 0.2662, + "step": 31582 + }, + { + "epoch": 11.66, + "learning_rate": 1.8348441926345609e-06, + "loss": 0.2316, + "step": 31583 + }, + { + "epoch": 11.66, + "learning_rate": 1.834807242271216e-06, + "loss": 0.3587, + "step": 31584 + }, + { + "epoch": 11.66, + "learning_rate": 1.8347702919078706e-06, + "loss": 0.4081, + "step": 31585 + }, + { + "epoch": 11.66, + "learning_rate": 1.8347333415445252e-06, + "loss": 0.345, + "step": 31586 + }, + { + "epoch": 11.66, + "learning_rate": 1.83469639118118e-06, + "loss": 0.3408, + "step": 31587 + }, + { + "epoch": 11.66, + "learning_rate": 1.8346594408178347e-06, + "loss": 0.3435, + "step": 31588 + }, + { + "epoch": 11.67, + "learning_rate": 1.8346224904544895e-06, + "loss": 0.3117, + "step": 31589 + }, + { + "epoch": 11.67, + "learning_rate": 1.8345855400911442e-06, + "loss": 0.3627, + "step": 31590 + }, + { + "epoch": 11.67, + "learning_rate": 1.8345485897277992e-06, + "loss": 0.2759, + "step": 31591 + }, + { + "epoch": 11.67, + "learning_rate": 1.8345116393644539e-06, + "loss": 0.2938, + "step": 31592 + }, + { + "epoch": 11.67, + "learning_rate": 1.8344746890011087e-06, + "loss": 0.2589, + "step": 31593 + }, + { + "epoch": 11.67, + "learning_rate": 1.8344377386377634e-06, + "loss": 0.3129, + "step": 31594 + }, + { + "epoch": 11.67, + "learning_rate": 1.834400788274418e-06, + "loss": 0.2701, + "step": 31595 + }, + { + "epoch": 11.67, + "learning_rate": 1.8343638379110728e-06, + "loss": 0.278, + "step": 31596 + }, + { + "epoch": 11.67, + "learning_rate": 1.8343268875477275e-06, + "loss": 0.3323, + "step": 31597 + }, + { + "epoch": 11.67, + "learning_rate": 1.8342899371843823e-06, + "loss": 0.2625, + "step": 31598 + }, + { + "epoch": 11.67, + "learning_rate": 1.834252986821037e-06, + "loss": 0.3746, + "step": 31599 + }, + { + "epoch": 11.67, + "learning_rate": 1.834216036457692e-06, + "loss": 0.2751, + "step": 31600 + }, + { + "epoch": 11.67, + "learning_rate": 1.8341790860943467e-06, + "loss": 0.2503, + "step": 31601 + }, + { + "epoch": 11.67, + "learning_rate": 1.8341421357310015e-06, + "loss": 0.24, + "step": 31602 + }, + { + "epoch": 11.67, + "learning_rate": 1.8341051853676561e-06, + "loss": 0.2578, + "step": 31603 + }, + { + "epoch": 11.67, + "learning_rate": 1.8340682350043108e-06, + "loss": 0.2923, + "step": 31604 + }, + { + "epoch": 11.67, + "learning_rate": 1.8340312846409656e-06, + "loss": 0.435, + "step": 31605 + }, + { + "epoch": 11.67, + "learning_rate": 1.8339943342776203e-06, + "loss": 0.3898, + "step": 31606 + }, + { + "epoch": 11.67, + "learning_rate": 1.8339573839142753e-06, + "loss": 0.3361, + "step": 31607 + }, + { + "epoch": 11.67, + "learning_rate": 1.83392043355093e-06, + "loss": 0.3264, + "step": 31608 + }, + { + "epoch": 11.67, + "learning_rate": 1.8338834831875848e-06, + "loss": 0.3991, + "step": 31609 + }, + { + "epoch": 11.67, + "learning_rate": 1.8338465328242394e-06, + "loss": 0.2878, + "step": 31610 + }, + { + "epoch": 11.67, + "learning_rate": 1.8338095824608943e-06, + "loss": 0.2959, + "step": 31611 + }, + { + "epoch": 11.67, + "learning_rate": 1.833772632097549e-06, + "loss": 0.4158, + "step": 31612 + }, + { + "epoch": 11.67, + "learning_rate": 1.833735681734204e-06, + "loss": 0.3105, + "step": 31613 + }, + { + "epoch": 11.67, + "learning_rate": 1.8336987313708586e-06, + "loss": 0.3128, + "step": 31614 + }, + { + "epoch": 11.67, + "learning_rate": 1.833661781007513e-06, + "loss": 0.374, + "step": 31615 + }, + { + "epoch": 11.68, + "learning_rate": 1.8336248306441681e-06, + "loss": 0.2165, + "step": 31616 + }, + { + "epoch": 11.68, + "learning_rate": 1.8335878802808227e-06, + "loss": 0.3731, + "step": 31617 + }, + { + "epoch": 11.68, + "learning_rate": 1.8335509299174776e-06, + "loss": 0.3342, + "step": 31618 + }, + { + "epoch": 11.68, + "learning_rate": 1.8335139795541322e-06, + "loss": 0.4308, + "step": 31619 + }, + { + "epoch": 11.68, + "learning_rate": 1.8334770291907873e-06, + "loss": 0.2671, + "step": 31620 + }, + { + "epoch": 11.68, + "learning_rate": 1.8334400788274417e-06, + "loss": 0.3625, + "step": 31621 + }, + { + "epoch": 11.68, + "learning_rate": 1.8334031284640968e-06, + "loss": 0.3865, + "step": 31622 + }, + { + "epoch": 11.68, + "learning_rate": 1.8333661781007514e-06, + "loss": 0.386, + "step": 31623 + }, + { + "epoch": 11.68, + "learning_rate": 1.833329227737406e-06, + "loss": 0.2185, + "step": 31624 + }, + { + "epoch": 11.68, + "learning_rate": 1.833292277374061e-06, + "loss": 0.2393, + "step": 31625 + }, + { + "epoch": 11.68, + "learning_rate": 1.8332553270107155e-06, + "loss": 0.3008, + "step": 31626 + }, + { + "epoch": 11.68, + "learning_rate": 1.8332183766473704e-06, + "loss": 0.3751, + "step": 31627 + }, + { + "epoch": 11.68, + "learning_rate": 1.833181426284025e-06, + "loss": 0.2072, + "step": 31628 + }, + { + "epoch": 11.68, + "learning_rate": 1.83314447592068e-06, + "loss": 0.2432, + "step": 31629 + }, + { + "epoch": 11.68, + "learning_rate": 1.8331075255573347e-06, + "loss": 0.309, + "step": 31630 + }, + { + "epoch": 11.68, + "learning_rate": 1.8330705751939896e-06, + "loss": 0.299, + "step": 31631 + }, + { + "epoch": 11.68, + "learning_rate": 1.8330336248306442e-06, + "loss": 0.2367, + "step": 31632 + }, + { + "epoch": 11.68, + "learning_rate": 1.832996674467299e-06, + "loss": 0.2346, + "step": 31633 + }, + { + "epoch": 11.68, + "learning_rate": 1.8329597241039537e-06, + "loss": 0.1522, + "step": 31634 + }, + { + "epoch": 11.68, + "learning_rate": 1.8329227737406083e-06, + "loss": 0.3674, + "step": 31635 + }, + { + "epoch": 11.68, + "learning_rate": 1.8328858233772634e-06, + "loss": 0.4174, + "step": 31636 + }, + { + "epoch": 11.68, + "learning_rate": 1.8328488730139178e-06, + "loss": 0.3392, + "step": 31637 + }, + { + "epoch": 11.68, + "learning_rate": 1.8328119226505729e-06, + "loss": 0.3295, + "step": 31638 + }, + { + "epoch": 11.68, + "learning_rate": 1.8327749722872275e-06, + "loss": 0.279, + "step": 31639 + }, + { + "epoch": 11.68, + "learning_rate": 1.8327380219238823e-06, + "loss": 0.2511, + "step": 31640 + }, + { + "epoch": 11.68, + "learning_rate": 1.832701071560537e-06, + "loss": 0.3822, + "step": 31641 + }, + { + "epoch": 11.68, + "learning_rate": 1.832664121197192e-06, + "loss": 0.3228, + "step": 31642 + }, + { + "epoch": 11.69, + "learning_rate": 1.8326271708338465e-06, + "loss": 0.2884, + "step": 31643 + }, + { + "epoch": 11.69, + "learning_rate": 1.832590220470501e-06, + "loss": 0.3049, + "step": 31644 + }, + { + "epoch": 11.69, + "learning_rate": 1.8325532701071562e-06, + "loss": 0.2504, + "step": 31645 + }, + { + "epoch": 11.69, + "learning_rate": 1.8325163197438108e-06, + "loss": 0.3366, + "step": 31646 + }, + { + "epoch": 11.69, + "learning_rate": 1.8324793693804656e-06, + "loss": 0.3832, + "step": 31647 + }, + { + "epoch": 11.69, + "learning_rate": 1.8324424190171203e-06, + "loss": 0.3124, + "step": 31648 + }, + { + "epoch": 11.69, + "learning_rate": 1.8324054686537751e-06, + "loss": 0.2817, + "step": 31649 + }, + { + "epoch": 11.69, + "learning_rate": 1.8323685182904298e-06, + "loss": 0.153, + "step": 31650 + }, + { + "epoch": 11.69, + "learning_rate": 1.8323315679270848e-06, + "loss": 0.2402, + "step": 31651 + }, + { + "epoch": 11.69, + "learning_rate": 1.8322946175637395e-06, + "loss": 0.3718, + "step": 31652 + }, + { + "epoch": 11.69, + "learning_rate": 1.8322576672003943e-06, + "loss": 0.2418, + "step": 31653 + }, + { + "epoch": 11.69, + "learning_rate": 1.832220716837049e-06, + "loss": 0.3539, + "step": 31654 + }, + { + "epoch": 11.69, + "learning_rate": 1.8321837664737036e-06, + "loss": 0.2408, + "step": 31655 + }, + { + "epoch": 11.69, + "learning_rate": 1.8321468161103584e-06, + "loss": 0.3672, + "step": 31656 + }, + { + "epoch": 11.69, + "learning_rate": 1.832109865747013e-06, + "loss": 0.3548, + "step": 31657 + }, + { + "epoch": 11.69, + "learning_rate": 1.8320729153836681e-06, + "loss": 0.2199, + "step": 31658 + }, + { + "epoch": 11.69, + "learning_rate": 1.8320359650203228e-06, + "loss": 0.1948, + "step": 31659 + }, + { + "epoch": 11.69, + "learning_rate": 1.8319990146569776e-06, + "loss": 0.2342, + "step": 31660 + }, + { + "epoch": 11.69, + "learning_rate": 1.8319620642936323e-06, + "loss": 0.3728, + "step": 31661 + }, + { + "epoch": 11.69, + "learning_rate": 1.831925113930287e-06, + "loss": 0.332, + "step": 31662 + }, + { + "epoch": 11.69, + "learning_rate": 1.8318881635669417e-06, + "loss": 0.3177, + "step": 31663 + }, + { + "epoch": 11.69, + "learning_rate": 1.8318512132035964e-06, + "loss": 0.2338, + "step": 31664 + }, + { + "epoch": 11.69, + "learning_rate": 1.8318142628402514e-06, + "loss": 0.2951, + "step": 31665 + }, + { + "epoch": 11.69, + "learning_rate": 1.8317773124769059e-06, + "loss": 0.332, + "step": 31666 + }, + { + "epoch": 11.69, + "learning_rate": 1.831740362113561e-06, + "loss": 0.3467, + "step": 31667 + }, + { + "epoch": 11.69, + "learning_rate": 1.8317034117502156e-06, + "loss": 0.2211, + "step": 31668 + }, + { + "epoch": 11.69, + "learning_rate": 1.8316664613868704e-06, + "loss": 0.3562, + "step": 31669 + }, + { + "epoch": 11.69, + "learning_rate": 1.831629511023525e-06, + "loss": 0.3997, + "step": 31670 + }, + { + "epoch": 11.7, + "learning_rate": 1.8315925606601799e-06, + "loss": 0.3984, + "step": 31671 + }, + { + "epoch": 11.7, + "learning_rate": 1.8315556102968345e-06, + "loss": 0.2636, + "step": 31672 + }, + { + "epoch": 11.7, + "learning_rate": 1.8315186599334896e-06, + "loss": 0.2739, + "step": 31673 + }, + { + "epoch": 11.7, + "learning_rate": 1.8314817095701442e-06, + "loss": 0.4535, + "step": 31674 + }, + { + "epoch": 11.7, + "learning_rate": 1.8314447592067989e-06, + "loss": 0.3752, + "step": 31675 + }, + { + "epoch": 11.7, + "learning_rate": 1.8314078088434537e-06, + "loss": 0.3069, + "step": 31676 + }, + { + "epoch": 11.7, + "learning_rate": 1.8313708584801083e-06, + "loss": 0.2671, + "step": 31677 + }, + { + "epoch": 11.7, + "learning_rate": 1.8313339081167632e-06, + "loss": 0.3136, + "step": 31678 + }, + { + "epoch": 11.7, + "learning_rate": 1.8312969577534178e-06, + "loss": 0.2909, + "step": 31679 + }, + { + "epoch": 11.7, + "learning_rate": 1.8312600073900729e-06, + "loss": 0.3921, + "step": 31680 + }, + { + "epoch": 11.7, + "learning_rate": 1.8312230570267275e-06, + "loss": 0.3008, + "step": 31681 + }, + { + "epoch": 11.7, + "learning_rate": 1.8311861066633824e-06, + "loss": 0.2588, + "step": 31682 + }, + { + "epoch": 11.7, + "learning_rate": 1.831149156300037e-06, + "loss": 0.3863, + "step": 31683 + }, + { + "epoch": 11.7, + "learning_rate": 1.8311122059366916e-06, + "loss": 0.3132, + "step": 31684 + }, + { + "epoch": 11.7, + "learning_rate": 1.8310752555733465e-06, + "loss": 0.2679, + "step": 31685 + }, + { + "epoch": 11.7, + "learning_rate": 1.8310383052100011e-06, + "loss": 0.3335, + "step": 31686 + }, + { + "epoch": 11.7, + "learning_rate": 1.8310013548466562e-06, + "loss": 0.3781, + "step": 31687 + }, + { + "epoch": 11.7, + "learning_rate": 1.8309644044833106e-06, + "loss": 0.1708, + "step": 31688 + }, + { + "epoch": 11.7, + "learning_rate": 1.8309274541199657e-06, + "loss": 0.3343, + "step": 31689 + }, + { + "epoch": 11.7, + "learning_rate": 1.8308905037566203e-06, + "loss": 0.2605, + "step": 31690 + }, + { + "epoch": 11.7, + "learning_rate": 1.8308535533932752e-06, + "loss": 0.3139, + "step": 31691 + }, + { + "epoch": 11.7, + "learning_rate": 1.8308166030299298e-06, + "loss": 0.26, + "step": 31692 + }, + { + "epoch": 11.7, + "learning_rate": 1.8307796526665849e-06, + "loss": 0.2062, + "step": 31693 + }, + { + "epoch": 11.7, + "learning_rate": 1.8307427023032393e-06, + "loss": 0.2855, + "step": 31694 + }, + { + "epoch": 11.7, + "learning_rate": 1.830705751939894e-06, + "loss": 0.2358, + "step": 31695 + }, + { + "epoch": 11.7, + "learning_rate": 1.830668801576549e-06, + "loss": 0.3204, + "step": 31696 + }, + { + "epoch": 11.7, + "learning_rate": 1.8306318512132036e-06, + "loss": 0.3746, + "step": 31697 + }, + { + "epoch": 11.71, + "learning_rate": 1.8305949008498585e-06, + "loss": 0.264, + "step": 31698 + }, + { + "epoch": 11.71, + "learning_rate": 1.830557950486513e-06, + "loss": 0.3953, + "step": 31699 + }, + { + "epoch": 11.71, + "learning_rate": 1.830521000123168e-06, + "loss": 0.3131, + "step": 31700 + }, + { + "epoch": 11.71, + "learning_rate": 1.8304840497598226e-06, + "loss": 0.4545, + "step": 31701 + }, + { + "epoch": 11.71, + "learning_rate": 1.8304470993964776e-06, + "loss": 0.2053, + "step": 31702 + }, + { + "epoch": 11.71, + "learning_rate": 1.8304101490331323e-06, + "loss": 0.201, + "step": 31703 + }, + { + "epoch": 11.71, + "learning_rate": 1.8303731986697867e-06, + "loss": 0.2674, + "step": 31704 + }, + { + "epoch": 11.71, + "learning_rate": 1.8303362483064418e-06, + "loss": 0.2241, + "step": 31705 + }, + { + "epoch": 11.71, + "learning_rate": 1.8302992979430964e-06, + "loss": 0.3079, + "step": 31706 + }, + { + "epoch": 11.71, + "learning_rate": 1.8302623475797512e-06, + "loss": 0.3658, + "step": 31707 + }, + { + "epoch": 11.71, + "learning_rate": 1.8302253972164059e-06, + "loss": 0.3938, + "step": 31708 + }, + { + "epoch": 11.71, + "learning_rate": 1.830188446853061e-06, + "loss": 0.2433, + "step": 31709 + }, + { + "epoch": 11.71, + "learning_rate": 1.8301514964897154e-06, + "loss": 0.3395, + "step": 31710 + }, + { + "epoch": 11.71, + "learning_rate": 1.8301145461263704e-06, + "loss": 0.3751, + "step": 31711 + }, + { + "epoch": 11.71, + "learning_rate": 1.830077595763025e-06, + "loss": 0.379, + "step": 31712 + }, + { + "epoch": 11.71, + "learning_rate": 1.83004064539968e-06, + "loss": 0.4266, + "step": 31713 + }, + { + "epoch": 11.71, + "learning_rate": 1.8300036950363345e-06, + "loss": 0.4253, + "step": 31714 + }, + { + "epoch": 11.71, + "learning_rate": 1.8299667446729892e-06, + "loss": 0.2982, + "step": 31715 + }, + { + "epoch": 11.71, + "learning_rate": 1.829929794309644e-06, + "loss": 0.242, + "step": 31716 + }, + { + "epoch": 11.71, + "learning_rate": 1.8298928439462987e-06, + "loss": 0.3436, + "step": 31717 + }, + { + "epoch": 11.71, + "learning_rate": 1.8298558935829537e-06, + "loss": 0.3315, + "step": 31718 + }, + { + "epoch": 11.71, + "learning_rate": 1.8298189432196084e-06, + "loss": 0.2517, + "step": 31719 + }, + { + "epoch": 11.71, + "learning_rate": 1.8297819928562632e-06, + "loss": 0.3379, + "step": 31720 + }, + { + "epoch": 11.71, + "learning_rate": 1.8297450424929179e-06, + "loss": 0.1911, + "step": 31721 + }, + { + "epoch": 11.71, + "learning_rate": 1.8297080921295727e-06, + "loss": 0.2729, + "step": 31722 + }, + { + "epoch": 11.71, + "learning_rate": 1.8296711417662273e-06, + "loss": 0.3658, + "step": 31723 + }, + { + "epoch": 11.71, + "learning_rate": 1.829634191402882e-06, + "loss": 0.3191, + "step": 31724 + }, + { + "epoch": 11.72, + "learning_rate": 1.829597241039537e-06, + "loss": 0.4523, + "step": 31725 + }, + { + "epoch": 11.72, + "learning_rate": 1.8295602906761917e-06, + "loss": 0.261, + "step": 31726 + }, + { + "epoch": 11.72, + "learning_rate": 1.8295233403128465e-06, + "loss": 0.335, + "step": 31727 + }, + { + "epoch": 11.72, + "learning_rate": 1.8294863899495012e-06, + "loss": 0.2626, + "step": 31728 + }, + { + "epoch": 11.72, + "learning_rate": 1.829449439586156e-06, + "loss": 0.399, + "step": 31729 + }, + { + "epoch": 11.72, + "learning_rate": 1.8294124892228106e-06, + "loss": 0.3273, + "step": 31730 + }, + { + "epoch": 11.72, + "learning_rate": 1.8293755388594657e-06, + "loss": 0.3212, + "step": 31731 + }, + { + "epoch": 11.72, + "learning_rate": 1.8293385884961203e-06, + "loss": 0.2984, + "step": 31732 + }, + { + "epoch": 11.72, + "learning_rate": 1.8293016381327752e-06, + "loss": 0.1908, + "step": 31733 + }, + { + "epoch": 11.72, + "learning_rate": 1.8292646877694298e-06, + "loss": 0.3769, + "step": 31734 + }, + { + "epoch": 11.72, + "learning_rate": 1.8292277374060845e-06, + "loss": 0.236, + "step": 31735 + }, + { + "epoch": 11.72, + "learning_rate": 1.8291907870427393e-06, + "loss": 0.3259, + "step": 31736 + }, + { + "epoch": 11.72, + "learning_rate": 1.829153836679394e-06, + "loss": 0.264, + "step": 31737 + }, + { + "epoch": 11.72, + "learning_rate": 1.829116886316049e-06, + "loss": 0.3201, + "step": 31738 + }, + { + "epoch": 11.72, + "learning_rate": 1.8290799359527034e-06, + "loss": 0.3462, + "step": 31739 + }, + { + "epoch": 11.72, + "learning_rate": 1.8290429855893585e-06, + "loss": 0.3061, + "step": 31740 + }, + { + "epoch": 11.72, + "learning_rate": 1.8290060352260131e-06, + "loss": 0.3235, + "step": 31741 + }, + { + "epoch": 11.72, + "learning_rate": 1.828969084862668e-06, + "loss": 0.3138, + "step": 31742 + }, + { + "epoch": 11.72, + "learning_rate": 1.8289321344993226e-06, + "loss": 0.2667, + "step": 31743 + }, + { + "epoch": 11.72, + "learning_rate": 1.8288951841359772e-06, + "loss": 0.3475, + "step": 31744 + }, + { + "epoch": 11.72, + "learning_rate": 1.828858233772632e-06, + "loss": 0.2712, + "step": 31745 + }, + { + "epoch": 11.72, + "learning_rate": 1.8288212834092867e-06, + "loss": 0.3469, + "step": 31746 + }, + { + "epoch": 11.72, + "learning_rate": 1.8287843330459418e-06, + "loss": 0.351, + "step": 31747 + }, + { + "epoch": 11.72, + "learning_rate": 1.8287473826825964e-06, + "loss": 0.2972, + "step": 31748 + }, + { + "epoch": 11.72, + "learning_rate": 1.8287104323192513e-06, + "loss": 0.2305, + "step": 31749 + }, + { + "epoch": 11.72, + "learning_rate": 1.828673481955906e-06, + "loss": 0.4443, + "step": 31750 + }, + { + "epoch": 11.72, + "learning_rate": 1.8286365315925608e-06, + "loss": 0.4263, + "step": 31751 + }, + { + "epoch": 11.73, + "learning_rate": 1.8285995812292154e-06, + "loss": 0.3225, + "step": 31752 + }, + { + "epoch": 11.73, + "learning_rate": 1.8285626308658705e-06, + "loss": 0.386, + "step": 31753 + }, + { + "epoch": 11.73, + "learning_rate": 1.828525680502525e-06, + "loss": 0.3167, + "step": 31754 + }, + { + "epoch": 11.73, + "learning_rate": 1.8284887301391795e-06, + "loss": 0.3079, + "step": 31755 + }, + { + "epoch": 11.73, + "learning_rate": 1.8284517797758346e-06, + "loss": 0.3268, + "step": 31756 + }, + { + "epoch": 11.73, + "learning_rate": 1.8284148294124892e-06, + "loss": 0.4263, + "step": 31757 + }, + { + "epoch": 11.73, + "learning_rate": 1.828377879049144e-06, + "loss": 0.2701, + "step": 31758 + }, + { + "epoch": 11.73, + "learning_rate": 1.8283409286857987e-06, + "loss": 0.3242, + "step": 31759 + }, + { + "epoch": 11.73, + "learning_rate": 1.8283039783224538e-06, + "loss": 0.3792, + "step": 31760 + }, + { + "epoch": 11.73, + "learning_rate": 1.8282670279591082e-06, + "loss": 0.3244, + "step": 31761 + }, + { + "epoch": 11.73, + "learning_rate": 1.8282300775957632e-06, + "loss": 0.2523, + "step": 31762 + }, + { + "epoch": 11.73, + "learning_rate": 1.8281931272324179e-06, + "loss": 0.2542, + "step": 31763 + }, + { + "epoch": 11.73, + "learning_rate": 1.8281561768690725e-06, + "loss": 0.2111, + "step": 31764 + }, + { + "epoch": 11.73, + "learning_rate": 1.8281192265057274e-06, + "loss": 0.3658, + "step": 31765 + }, + { + "epoch": 11.73, + "learning_rate": 1.828082276142382e-06, + "loss": 0.4386, + "step": 31766 + }, + { + "epoch": 11.73, + "learning_rate": 1.8280453257790368e-06, + "loss": 0.273, + "step": 31767 + }, + { + "epoch": 11.73, + "learning_rate": 1.8280083754156915e-06, + "loss": 0.206, + "step": 31768 + }, + { + "epoch": 11.73, + "learning_rate": 1.8279714250523465e-06, + "loss": 0.5424, + "step": 31769 + }, + { + "epoch": 11.73, + "learning_rate": 1.8279344746890012e-06, + "loss": 0.2672, + "step": 31770 + }, + { + "epoch": 11.73, + "learning_rate": 1.827897524325656e-06, + "loss": 0.1995, + "step": 31771 + }, + { + "epoch": 11.73, + "learning_rate": 1.8278605739623107e-06, + "loss": 0.3209, + "step": 31772 + }, + { + "epoch": 11.73, + "learning_rate": 1.8278236235989655e-06, + "loss": 0.2286, + "step": 31773 + }, + { + "epoch": 11.73, + "learning_rate": 1.8277866732356201e-06, + "loss": 0.1337, + "step": 31774 + }, + { + "epoch": 11.73, + "learning_rate": 1.8277497228722748e-06, + "loss": 0.3419, + "step": 31775 + }, + { + "epoch": 11.73, + "learning_rate": 1.8277127725089298e-06, + "loss": 0.428, + "step": 31776 + }, + { + "epoch": 11.73, + "learning_rate": 1.8276758221455845e-06, + "loss": 0.2668, + "step": 31777 + }, + { + "epoch": 11.73, + "learning_rate": 1.8276388717822393e-06, + "loss": 0.2682, + "step": 31778 + }, + { + "epoch": 11.74, + "learning_rate": 1.827601921418894e-06, + "loss": 0.3546, + "step": 31779 + }, + { + "epoch": 11.74, + "learning_rate": 1.8275649710555488e-06, + "loss": 0.2973, + "step": 31780 + }, + { + "epoch": 11.74, + "learning_rate": 1.8275280206922034e-06, + "loss": 0.3297, + "step": 31781 + }, + { + "epoch": 11.74, + "learning_rate": 1.8274910703288585e-06, + "loss": 0.2007, + "step": 31782 + }, + { + "epoch": 11.74, + "learning_rate": 1.827454119965513e-06, + "loss": 0.2706, + "step": 31783 + }, + { + "epoch": 11.74, + "learning_rate": 1.8274171696021676e-06, + "loss": 0.3685, + "step": 31784 + }, + { + "epoch": 11.74, + "learning_rate": 1.8273802192388226e-06, + "loss": 0.2505, + "step": 31785 + }, + { + "epoch": 11.74, + "learning_rate": 1.8273432688754773e-06, + "loss": 0.4689, + "step": 31786 + }, + { + "epoch": 11.74, + "learning_rate": 1.8273063185121321e-06, + "loss": 0.3033, + "step": 31787 + }, + { + "epoch": 11.74, + "learning_rate": 1.8272693681487868e-06, + "loss": 0.3131, + "step": 31788 + }, + { + "epoch": 11.74, + "learning_rate": 1.8272324177854416e-06, + "loss": 0.4384, + "step": 31789 + }, + { + "epoch": 11.74, + "learning_rate": 1.8271954674220962e-06, + "loss": 0.2719, + "step": 31790 + }, + { + "epoch": 11.74, + "learning_rate": 1.8271585170587513e-06, + "loss": 0.2519, + "step": 31791 + }, + { + "epoch": 11.74, + "learning_rate": 1.827121566695406e-06, + "loss": 0.4561, + "step": 31792 + }, + { + "epoch": 11.74, + "learning_rate": 1.8270846163320608e-06, + "loss": 0.2826, + "step": 31793 + }, + { + "epoch": 11.74, + "learning_rate": 1.8270476659687154e-06, + "loss": 0.2592, + "step": 31794 + }, + { + "epoch": 11.74, + "learning_rate": 1.82701071560537e-06, + "loss": 0.2647, + "step": 31795 + }, + { + "epoch": 11.74, + "learning_rate": 1.826973765242025e-06, + "loss": 0.4095, + "step": 31796 + }, + { + "epoch": 11.74, + "learning_rate": 1.8269368148786795e-06, + "loss": 0.2527, + "step": 31797 + }, + { + "epoch": 11.74, + "learning_rate": 1.8268998645153346e-06, + "loss": 0.5058, + "step": 31798 + }, + { + "epoch": 11.74, + "learning_rate": 1.8268629141519892e-06, + "loss": 0.3785, + "step": 31799 + }, + { + "epoch": 11.74, + "learning_rate": 1.826825963788644e-06, + "loss": 0.2982, + "step": 31800 + }, + { + "epoch": 11.74, + "learning_rate": 1.8267890134252987e-06, + "loss": 0.3766, + "step": 31801 + }, + { + "epoch": 11.74, + "learning_rate": 1.8267520630619536e-06, + "loss": 0.4684, + "step": 31802 + }, + { + "epoch": 11.74, + "learning_rate": 1.8267151126986082e-06, + "loss": 0.3325, + "step": 31803 + }, + { + "epoch": 11.74, + "learning_rate": 1.8266781623352628e-06, + "loss": 0.2717, + "step": 31804 + }, + { + "epoch": 11.74, + "learning_rate": 1.826641211971918e-06, + "loss": 0.3911, + "step": 31805 + }, + { + "epoch": 11.75, + "learning_rate": 1.8266042616085723e-06, + "loss": 0.3172, + "step": 31806 + }, + { + "epoch": 11.75, + "learning_rate": 1.8265673112452274e-06, + "loss": 0.2904, + "step": 31807 + }, + { + "epoch": 11.75, + "learning_rate": 1.826530360881882e-06, + "loss": 0.3972, + "step": 31808 + }, + { + "epoch": 11.75, + "learning_rate": 1.8264934105185369e-06, + "loss": 0.3541, + "step": 31809 + }, + { + "epoch": 11.75, + "learning_rate": 1.8264564601551915e-06, + "loss": 0.2368, + "step": 31810 + }, + { + "epoch": 11.75, + "learning_rate": 1.8264195097918466e-06, + "loss": 0.2921, + "step": 31811 + }, + { + "epoch": 11.75, + "learning_rate": 1.826382559428501e-06, + "loss": 0.3064, + "step": 31812 + }, + { + "epoch": 11.75, + "learning_rate": 1.826345609065156e-06, + "loss": 0.3214, + "step": 31813 + }, + { + "epoch": 11.75, + "learning_rate": 1.8263086587018107e-06, + "loss": 0.2687, + "step": 31814 + }, + { + "epoch": 11.75, + "learning_rate": 1.8262717083384653e-06, + "loss": 0.242, + "step": 31815 + }, + { + "epoch": 11.75, + "learning_rate": 1.8262347579751202e-06, + "loss": 0.4825, + "step": 31816 + }, + { + "epoch": 11.75, + "learning_rate": 1.8261978076117748e-06, + "loss": 0.3139, + "step": 31817 + }, + { + "epoch": 11.75, + "learning_rate": 1.8261608572484297e-06, + "loss": 0.378, + "step": 31818 + }, + { + "epoch": 11.75, + "learning_rate": 1.8261239068850843e-06, + "loss": 0.4023, + "step": 31819 + }, + { + "epoch": 11.75, + "learning_rate": 1.8260869565217394e-06, + "loss": 0.3605, + "step": 31820 + }, + { + "epoch": 11.75, + "learning_rate": 1.826050006158394e-06, + "loss": 0.3079, + "step": 31821 + }, + { + "epoch": 11.75, + "learning_rate": 1.8260130557950488e-06, + "loss": 0.3676, + "step": 31822 + }, + { + "epoch": 11.75, + "learning_rate": 1.8259761054317035e-06, + "loss": 0.4415, + "step": 31823 + }, + { + "epoch": 11.75, + "learning_rate": 1.8259391550683581e-06, + "loss": 0.2044, + "step": 31824 + }, + { + "epoch": 11.75, + "learning_rate": 1.825902204705013e-06, + "loss": 0.283, + "step": 31825 + }, + { + "epoch": 11.75, + "learning_rate": 1.8258652543416676e-06, + "loss": 0.3536, + "step": 31826 + }, + { + "epoch": 11.75, + "learning_rate": 1.8258283039783227e-06, + "loss": 0.3698, + "step": 31827 + }, + { + "epoch": 11.75, + "learning_rate": 1.825791353614977e-06, + "loss": 0.386, + "step": 31828 + }, + { + "epoch": 11.75, + "learning_rate": 1.8257544032516321e-06, + "loss": 0.3924, + "step": 31829 + }, + { + "epoch": 11.75, + "learning_rate": 1.8257174528882868e-06, + "loss": 0.1769, + "step": 31830 + }, + { + "epoch": 11.75, + "learning_rate": 1.8256805025249416e-06, + "loss": 0.4171, + "step": 31831 + }, + { + "epoch": 11.75, + "learning_rate": 1.8256435521615963e-06, + "loss": 0.4042, + "step": 31832 + }, + { + "epoch": 11.76, + "learning_rate": 1.825606601798251e-06, + "loss": 0.2759, + "step": 31833 + }, + { + "epoch": 11.76, + "learning_rate": 1.8255696514349057e-06, + "loss": 0.337, + "step": 31834 + }, + { + "epoch": 11.76, + "learning_rate": 1.8255327010715604e-06, + "loss": 0.3349, + "step": 31835 + }, + { + "epoch": 11.76, + "learning_rate": 1.8254957507082154e-06, + "loss": 0.3635, + "step": 31836 + }, + { + "epoch": 11.76, + "learning_rate": 1.82545880034487e-06, + "loss": 0.2545, + "step": 31837 + }, + { + "epoch": 11.76, + "learning_rate": 1.825421849981525e-06, + "loss": 0.2875, + "step": 31838 + }, + { + "epoch": 11.76, + "learning_rate": 1.8253848996181796e-06, + "loss": 0.3756, + "step": 31839 + }, + { + "epoch": 11.76, + "learning_rate": 1.8253479492548344e-06, + "loss": 0.2473, + "step": 31840 + }, + { + "epoch": 11.76, + "learning_rate": 1.825310998891489e-06, + "loss": 0.3268, + "step": 31841 + }, + { + "epoch": 11.76, + "learning_rate": 1.8252740485281441e-06, + "loss": 0.322, + "step": 31842 + }, + { + "epoch": 11.76, + "learning_rate": 1.8252370981647987e-06, + "loss": 0.4219, + "step": 31843 + }, + { + "epoch": 11.76, + "learning_rate": 1.8252001478014534e-06, + "loss": 0.3199, + "step": 31844 + }, + { + "epoch": 11.76, + "learning_rate": 1.8251631974381082e-06, + "loss": 0.2536, + "step": 31845 + }, + { + "epoch": 11.76, + "learning_rate": 1.8251262470747629e-06, + "loss": 0.3888, + "step": 31846 + }, + { + "epoch": 11.76, + "learning_rate": 1.8250892967114177e-06, + "loss": 0.2388, + "step": 31847 + }, + { + "epoch": 11.76, + "learning_rate": 1.8250523463480723e-06, + "loss": 0.3223, + "step": 31848 + }, + { + "epoch": 11.76, + "learning_rate": 1.8250153959847274e-06, + "loss": 0.3327, + "step": 31849 + }, + { + "epoch": 11.76, + "learning_rate": 1.824978445621382e-06, + "loss": 0.4882, + "step": 31850 + }, + { + "epoch": 11.76, + "learning_rate": 1.8249414952580369e-06, + "loss": 0.2521, + "step": 31851 + }, + { + "epoch": 11.76, + "learning_rate": 1.8249045448946915e-06, + "loss": 0.3227, + "step": 31852 + }, + { + "epoch": 11.76, + "learning_rate": 1.8248675945313462e-06, + "loss": 0.2919, + "step": 31853 + }, + { + "epoch": 11.76, + "learning_rate": 1.824830644168001e-06, + "loss": 0.2558, + "step": 31854 + }, + { + "epoch": 11.76, + "learning_rate": 1.8247936938046557e-06, + "loss": 0.2764, + "step": 31855 + }, + { + "epoch": 11.76, + "learning_rate": 1.8247567434413105e-06, + "loss": 0.2744, + "step": 31856 + }, + { + "epoch": 11.76, + "learning_rate": 1.8247197930779651e-06, + "loss": 0.234, + "step": 31857 + }, + { + "epoch": 11.76, + "learning_rate": 1.8246828427146202e-06, + "loss": 0.3161, + "step": 31858 + }, + { + "epoch": 11.76, + "learning_rate": 1.8246458923512748e-06, + "loss": 0.2317, + "step": 31859 + }, + { + "epoch": 11.77, + "learning_rate": 1.8246089419879297e-06, + "loss": 0.4829, + "step": 31860 + }, + { + "epoch": 11.77, + "learning_rate": 1.8245719916245843e-06, + "loss": 0.4486, + "step": 31861 + }, + { + "epoch": 11.77, + "learning_rate": 1.8245350412612392e-06, + "loss": 0.292, + "step": 31862 + }, + { + "epoch": 11.77, + "learning_rate": 1.8244980908978938e-06, + "loss": 0.3203, + "step": 31863 + }, + { + "epoch": 11.77, + "learning_rate": 1.8244611405345484e-06, + "loss": 0.3193, + "step": 31864 + }, + { + "epoch": 11.77, + "learning_rate": 1.8244241901712035e-06, + "loss": 0.2106, + "step": 31865 + }, + { + "epoch": 11.77, + "learning_rate": 1.8243872398078581e-06, + "loss": 0.269, + "step": 31866 + }, + { + "epoch": 11.77, + "learning_rate": 1.824350289444513e-06, + "loss": 0.4598, + "step": 31867 + }, + { + "epoch": 11.77, + "learning_rate": 1.8243133390811676e-06, + "loss": 0.2477, + "step": 31868 + }, + { + "epoch": 11.77, + "learning_rate": 1.8242763887178225e-06, + "loss": 0.2669, + "step": 31869 + }, + { + "epoch": 11.77, + "learning_rate": 1.824239438354477e-06, + "loss": 0.3562, + "step": 31870 + }, + { + "epoch": 11.77, + "learning_rate": 1.8242024879911322e-06, + "loss": 0.2253, + "step": 31871 + }, + { + "epoch": 11.77, + "learning_rate": 1.8241655376277868e-06, + "loss": 0.2802, + "step": 31872 + }, + { + "epoch": 11.77, + "learning_rate": 1.8241285872644412e-06, + "loss": 0.3528, + "step": 31873 + }, + { + "epoch": 11.77, + "learning_rate": 1.8240916369010963e-06, + "loss": 0.2213, + "step": 31874 + }, + { + "epoch": 11.77, + "learning_rate": 1.824054686537751e-06, + "loss": 0.3116, + "step": 31875 + }, + { + "epoch": 11.77, + "learning_rate": 1.8240177361744058e-06, + "loss": 0.2846, + "step": 31876 + }, + { + "epoch": 11.77, + "learning_rate": 1.8239807858110604e-06, + "loss": 0.2278, + "step": 31877 + }, + { + "epoch": 11.77, + "learning_rate": 1.8239438354477155e-06, + "loss": 0.2905, + "step": 31878 + }, + { + "epoch": 11.77, + "learning_rate": 1.8239068850843699e-06, + "loss": 0.3948, + "step": 31879 + }, + { + "epoch": 11.77, + "learning_rate": 1.823869934721025e-06, + "loss": 0.2381, + "step": 31880 + }, + { + "epoch": 11.77, + "learning_rate": 1.8238329843576796e-06, + "loss": 0.3098, + "step": 31881 + }, + { + "epoch": 11.77, + "learning_rate": 1.8237960339943344e-06, + "loss": 0.3812, + "step": 31882 + }, + { + "epoch": 11.77, + "learning_rate": 1.823759083630989e-06, + "loss": 0.3455, + "step": 31883 + }, + { + "epoch": 11.77, + "learning_rate": 1.8237221332676437e-06, + "loss": 0.3768, + "step": 31884 + }, + { + "epoch": 11.77, + "learning_rate": 1.8236851829042986e-06, + "loss": 0.2891, + "step": 31885 + }, + { + "epoch": 11.77, + "learning_rate": 1.8236482325409532e-06, + "loss": 0.309, + "step": 31886 + }, + { + "epoch": 11.78, + "learning_rate": 1.8236112821776083e-06, + "loss": 0.4047, + "step": 31887 + }, + { + "epoch": 11.78, + "learning_rate": 1.8235743318142629e-06, + "loss": 0.3687, + "step": 31888 + }, + { + "epoch": 11.78, + "learning_rate": 1.8235373814509177e-06, + "loss": 0.2533, + "step": 31889 + }, + { + "epoch": 11.78, + "learning_rate": 1.8235004310875724e-06, + "loss": 0.3065, + "step": 31890 + }, + { + "epoch": 11.78, + "learning_rate": 1.8234634807242272e-06, + "loss": 0.3428, + "step": 31891 + }, + { + "epoch": 11.78, + "learning_rate": 1.8234265303608819e-06, + "loss": 0.4102, + "step": 31892 + }, + { + "epoch": 11.78, + "learning_rate": 1.8233895799975365e-06, + "loss": 0.3053, + "step": 31893 + }, + { + "epoch": 11.78, + "learning_rate": 1.8233526296341916e-06, + "loss": 0.3671, + "step": 31894 + }, + { + "epoch": 11.78, + "learning_rate": 1.823315679270846e-06, + "loss": 0.3572, + "step": 31895 + }, + { + "epoch": 11.78, + "learning_rate": 1.823278728907501e-06, + "loss": 0.2258, + "step": 31896 + }, + { + "epoch": 11.78, + "learning_rate": 1.8232417785441557e-06, + "loss": 0.3196, + "step": 31897 + }, + { + "epoch": 11.78, + "learning_rate": 1.8232048281808105e-06, + "loss": 0.3022, + "step": 31898 + }, + { + "epoch": 11.78, + "learning_rate": 1.8231678778174652e-06, + "loss": 0.3282, + "step": 31899 + }, + { + "epoch": 11.78, + "learning_rate": 1.8231309274541202e-06, + "loss": 0.346, + "step": 31900 + }, + { + "epoch": 11.78, + "learning_rate": 1.8230939770907746e-06, + "loss": 0.3016, + "step": 31901 + }, + { + "epoch": 11.78, + "learning_rate": 1.8230570267274297e-06, + "loss": 0.3537, + "step": 31902 + }, + { + "epoch": 11.78, + "learning_rate": 1.8230200763640843e-06, + "loss": 0.2489, + "step": 31903 + }, + { + "epoch": 11.78, + "learning_rate": 1.822983126000739e-06, + "loss": 0.5229, + "step": 31904 + }, + { + "epoch": 11.78, + "learning_rate": 1.8229461756373938e-06, + "loss": 0.3613, + "step": 31905 + }, + { + "epoch": 11.78, + "learning_rate": 1.8229092252740485e-06, + "loss": 0.3621, + "step": 31906 + }, + { + "epoch": 11.78, + "learning_rate": 1.8228722749107033e-06, + "loss": 0.3886, + "step": 31907 + }, + { + "epoch": 11.78, + "learning_rate": 1.822835324547358e-06, + "loss": 0.3752, + "step": 31908 + }, + { + "epoch": 11.78, + "learning_rate": 1.822798374184013e-06, + "loss": 0.3299, + "step": 31909 + }, + { + "epoch": 11.78, + "learning_rate": 1.8227614238206676e-06, + "loss": 0.3062, + "step": 31910 + }, + { + "epoch": 11.78, + "learning_rate": 1.8227244734573225e-06, + "loss": 0.3969, + "step": 31911 + }, + { + "epoch": 11.78, + "learning_rate": 1.8226875230939771e-06, + "loss": 0.3505, + "step": 31912 + }, + { + "epoch": 11.78, + "learning_rate": 1.8226505727306318e-06, + "loss": 0.2777, + "step": 31913 + }, + { + "epoch": 11.79, + "learning_rate": 1.8226136223672866e-06, + "loss": 0.328, + "step": 31914 + }, + { + "epoch": 11.79, + "learning_rate": 1.8225766720039412e-06, + "loss": 0.398, + "step": 31915 + }, + { + "epoch": 11.79, + "learning_rate": 1.8225397216405963e-06, + "loss": 0.3263, + "step": 31916 + }, + { + "epoch": 11.79, + "learning_rate": 1.822502771277251e-06, + "loss": 0.2772, + "step": 31917 + }, + { + "epoch": 11.79, + "learning_rate": 1.8224658209139058e-06, + "loss": 0.2691, + "step": 31918 + }, + { + "epoch": 11.79, + "learning_rate": 1.8224288705505604e-06, + "loss": 0.3441, + "step": 31919 + }, + { + "epoch": 11.79, + "learning_rate": 1.8223919201872153e-06, + "loss": 0.2489, + "step": 31920 + }, + { + "epoch": 11.79, + "learning_rate": 1.82235496982387e-06, + "loss": 0.3399, + "step": 31921 + }, + { + "epoch": 11.79, + "learning_rate": 1.822318019460525e-06, + "loss": 0.4432, + "step": 31922 + }, + { + "epoch": 11.79, + "learning_rate": 1.8222810690971796e-06, + "loss": 0.2715, + "step": 31923 + }, + { + "epoch": 11.79, + "learning_rate": 1.822244118733834e-06, + "loss": 0.3015, + "step": 31924 + }, + { + "epoch": 11.79, + "learning_rate": 1.822207168370489e-06, + "loss": 0.3638, + "step": 31925 + }, + { + "epoch": 11.79, + "learning_rate": 1.8221702180071437e-06, + "loss": 0.3016, + "step": 31926 + }, + { + "epoch": 11.79, + "learning_rate": 1.8221332676437986e-06, + "loss": 0.2724, + "step": 31927 + }, + { + "epoch": 11.79, + "learning_rate": 1.8220963172804532e-06, + "loss": 0.4325, + "step": 31928 + }, + { + "epoch": 11.79, + "learning_rate": 1.8220593669171083e-06, + "loss": 0.4847, + "step": 31929 + }, + { + "epoch": 11.79, + "learning_rate": 1.8220224165537627e-06, + "loss": 0.3148, + "step": 31930 + }, + { + "epoch": 11.79, + "learning_rate": 1.8219854661904178e-06, + "loss": 0.298, + "step": 31931 + }, + { + "epoch": 11.79, + "learning_rate": 1.8219485158270724e-06, + "loss": 0.2421, + "step": 31932 + }, + { + "epoch": 11.79, + "learning_rate": 1.821911565463727e-06, + "loss": 0.3348, + "step": 31933 + }, + { + "epoch": 11.79, + "learning_rate": 1.8218746151003819e-06, + "loss": 0.2694, + "step": 31934 + }, + { + "epoch": 11.79, + "learning_rate": 1.8218376647370365e-06, + "loss": 0.3413, + "step": 31935 + }, + { + "epoch": 11.79, + "learning_rate": 1.8218007143736914e-06, + "loss": 0.2468, + "step": 31936 + }, + { + "epoch": 11.79, + "learning_rate": 1.821763764010346e-06, + "loss": 0.2419, + "step": 31937 + }, + { + "epoch": 11.79, + "learning_rate": 1.821726813647001e-06, + "loss": 0.3528, + "step": 31938 + }, + { + "epoch": 11.79, + "learning_rate": 1.8216898632836557e-06, + "loss": 0.2155, + "step": 31939 + }, + { + "epoch": 11.79, + "learning_rate": 1.8216529129203105e-06, + "loss": 0.3834, + "step": 31940 + }, + { + "epoch": 11.8, + "learning_rate": 1.8216159625569652e-06, + "loss": 0.4372, + "step": 31941 + }, + { + "epoch": 11.8, + "learning_rate": 1.82157901219362e-06, + "loss": 0.3022, + "step": 31942 + }, + { + "epoch": 11.8, + "learning_rate": 1.8215420618302747e-06, + "loss": 0.3492, + "step": 31943 + }, + { + "epoch": 11.8, + "learning_rate": 1.8215051114669293e-06, + "loss": 0.2068, + "step": 31944 + }, + { + "epoch": 11.8, + "learning_rate": 1.8214681611035844e-06, + "loss": 0.3732, + "step": 31945 + }, + { + "epoch": 11.8, + "learning_rate": 1.8214312107402388e-06, + "loss": 0.3523, + "step": 31946 + }, + { + "epoch": 11.8, + "learning_rate": 1.8213942603768938e-06, + "loss": 0.29, + "step": 31947 + }, + { + "epoch": 11.8, + "learning_rate": 1.8213573100135485e-06, + "loss": 0.2959, + "step": 31948 + }, + { + "epoch": 11.8, + "learning_rate": 1.8213203596502033e-06, + "loss": 0.3186, + "step": 31949 + }, + { + "epoch": 11.8, + "learning_rate": 1.821283409286858e-06, + "loss": 0.412, + "step": 31950 + }, + { + "epoch": 11.8, + "learning_rate": 1.821246458923513e-06, + "loss": 0.3954, + "step": 31951 + }, + { + "epoch": 11.8, + "learning_rate": 1.8212095085601675e-06, + "loss": 0.3029, + "step": 31952 + }, + { + "epoch": 11.8, + "learning_rate": 1.821172558196822e-06, + "loss": 0.3076, + "step": 31953 + }, + { + "epoch": 11.8, + "learning_rate": 1.8211356078334772e-06, + "loss": 0.3395, + "step": 31954 + }, + { + "epoch": 11.8, + "learning_rate": 1.8210986574701318e-06, + "loss": 0.3754, + "step": 31955 + }, + { + "epoch": 11.8, + "learning_rate": 1.8210617071067866e-06, + "loss": 0.3352, + "step": 31956 + }, + { + "epoch": 11.8, + "learning_rate": 1.8210247567434413e-06, + "loss": 0.3475, + "step": 31957 + }, + { + "epoch": 11.8, + "learning_rate": 1.8209878063800961e-06, + "loss": 0.2709, + "step": 31958 + }, + { + "epoch": 11.8, + "learning_rate": 1.8209508560167508e-06, + "loss": 0.2715, + "step": 31959 + }, + { + "epoch": 11.8, + "learning_rate": 1.8209139056534058e-06, + "loss": 0.2937, + "step": 31960 + }, + { + "epoch": 11.8, + "learning_rate": 1.8208769552900605e-06, + "loss": 0.2583, + "step": 31961 + }, + { + "epoch": 11.8, + "learning_rate": 1.8208400049267153e-06, + "loss": 0.335, + "step": 31962 + }, + { + "epoch": 11.8, + "learning_rate": 1.82080305456337e-06, + "loss": 0.2505, + "step": 31963 + }, + { + "epoch": 11.8, + "learning_rate": 1.8207661042000246e-06, + "loss": 0.3537, + "step": 31964 + }, + { + "epoch": 11.8, + "learning_rate": 1.8207291538366794e-06, + "loss": 0.3327, + "step": 31965 + }, + { + "epoch": 11.8, + "learning_rate": 1.820692203473334e-06, + "loss": 0.3462, + "step": 31966 + }, + { + "epoch": 11.8, + "learning_rate": 1.8206552531099891e-06, + "loss": 0.3106, + "step": 31967 + }, + { + "epoch": 11.81, + "learning_rate": 1.8206183027466435e-06, + "loss": 0.3686, + "step": 31968 + }, + { + "epoch": 11.81, + "learning_rate": 1.8205813523832986e-06, + "loss": 0.2974, + "step": 31969 + }, + { + "epoch": 11.81, + "learning_rate": 1.8205444020199532e-06, + "loss": 0.2401, + "step": 31970 + }, + { + "epoch": 11.81, + "learning_rate": 1.820507451656608e-06, + "loss": 0.2514, + "step": 31971 + }, + { + "epoch": 11.81, + "learning_rate": 1.8204705012932627e-06, + "loss": 0.2182, + "step": 31972 + }, + { + "epoch": 11.81, + "learning_rate": 1.8204335509299174e-06, + "loss": 0.3176, + "step": 31973 + }, + { + "epoch": 11.81, + "learning_rate": 1.8203966005665722e-06, + "loss": 0.3326, + "step": 31974 + }, + { + "epoch": 11.81, + "learning_rate": 1.8203596502032268e-06, + "loss": 0.4165, + "step": 31975 + }, + { + "epoch": 11.81, + "learning_rate": 1.820322699839882e-06, + "loss": 0.3267, + "step": 31976 + }, + { + "epoch": 11.81, + "learning_rate": 1.8202857494765365e-06, + "loss": 0.3115, + "step": 31977 + }, + { + "epoch": 11.81, + "learning_rate": 1.8202487991131914e-06, + "loss": 0.2752, + "step": 31978 + }, + { + "epoch": 11.81, + "learning_rate": 1.820211848749846e-06, + "loss": 0.2193, + "step": 31979 + }, + { + "epoch": 11.81, + "learning_rate": 1.8201748983865009e-06, + "loss": 0.4084, + "step": 31980 + }, + { + "epoch": 11.81, + "learning_rate": 1.8201379480231555e-06, + "loss": 0.3247, + "step": 31981 + }, + { + "epoch": 11.81, + "learning_rate": 1.8201009976598106e-06, + "loss": 0.3126, + "step": 31982 + }, + { + "epoch": 11.81, + "learning_rate": 1.8200640472964652e-06, + "loss": 0.4853, + "step": 31983 + }, + { + "epoch": 11.81, + "learning_rate": 1.8200270969331198e-06, + "loss": 0.238, + "step": 31984 + }, + { + "epoch": 11.81, + "learning_rate": 1.8199901465697747e-06, + "loss": 0.3692, + "step": 31985 + }, + { + "epoch": 11.81, + "learning_rate": 1.8199531962064293e-06, + "loss": 0.2547, + "step": 31986 + }, + { + "epoch": 11.81, + "learning_rate": 1.8199162458430842e-06, + "loss": 0.3735, + "step": 31987 + }, + { + "epoch": 11.81, + "learning_rate": 1.8198792954797388e-06, + "loss": 0.2566, + "step": 31988 + }, + { + "epoch": 11.81, + "learning_rate": 1.8198423451163939e-06, + "loss": 0.3082, + "step": 31989 + }, + { + "epoch": 11.81, + "learning_rate": 1.8198053947530485e-06, + "loss": 0.3136, + "step": 31990 + }, + { + "epoch": 11.81, + "learning_rate": 1.8197684443897034e-06, + "loss": 0.3357, + "step": 31991 + }, + { + "epoch": 11.81, + "learning_rate": 1.819731494026358e-06, + "loss": 0.3852, + "step": 31992 + }, + { + "epoch": 11.81, + "learning_rate": 1.8196945436630126e-06, + "loss": 0.3403, + "step": 31993 + }, + { + "epoch": 11.81, + "learning_rate": 1.8196575932996675e-06, + "loss": 0.2999, + "step": 31994 + }, + { + "epoch": 11.81, + "learning_rate": 1.8196206429363221e-06, + "loss": 0.3351, + "step": 31995 + }, + { + "epoch": 11.82, + "learning_rate": 1.8195836925729772e-06, + "loss": 0.4315, + "step": 31996 + }, + { + "epoch": 11.82, + "learning_rate": 1.8195467422096316e-06, + "loss": 0.2848, + "step": 31997 + }, + { + "epoch": 11.82, + "learning_rate": 1.8195097918462867e-06, + "loss": 0.3389, + "step": 31998 + }, + { + "epoch": 11.82, + "learning_rate": 1.8194728414829413e-06, + "loss": 0.3147, + "step": 31999 + }, + { + "epoch": 11.82, + "learning_rate": 1.8194358911195961e-06, + "loss": 0.4023, + "step": 32000 + }, + { + "epoch": 11.82, + "learning_rate": 1.8193989407562508e-06, + "loss": 0.42, + "step": 32001 + }, + { + "epoch": 11.82, + "learning_rate": 1.8193619903929058e-06, + "loss": 0.1823, + "step": 32002 + }, + { + "epoch": 11.82, + "learning_rate": 1.8193250400295603e-06, + "loss": 0.4711, + "step": 32003 + }, + { + "epoch": 11.82, + "learning_rate": 1.819288089666215e-06, + "loss": 0.3164, + "step": 32004 + }, + { + "epoch": 11.82, + "learning_rate": 1.81925113930287e-06, + "loss": 0.2726, + "step": 32005 + }, + { + "epoch": 11.82, + "learning_rate": 1.8192141889395246e-06, + "loss": 0.2468, + "step": 32006 + }, + { + "epoch": 11.82, + "learning_rate": 1.8191772385761794e-06, + "loss": 0.3429, + "step": 32007 + }, + { + "epoch": 11.82, + "learning_rate": 1.819140288212834e-06, + "loss": 0.3783, + "step": 32008 + }, + { + "epoch": 11.82, + "learning_rate": 1.819103337849489e-06, + "loss": 0.2486, + "step": 32009 + }, + { + "epoch": 11.82, + "learning_rate": 1.8190663874861436e-06, + "loss": 0.4295, + "step": 32010 + }, + { + "epoch": 11.82, + "learning_rate": 1.8190294371227986e-06, + "loss": 0.4071, + "step": 32011 + }, + { + "epoch": 11.82, + "learning_rate": 1.8189924867594533e-06, + "loss": 0.304, + "step": 32012 + }, + { + "epoch": 11.82, + "learning_rate": 1.8189555363961077e-06, + "loss": 0.3121, + "step": 32013 + }, + { + "epoch": 11.82, + "learning_rate": 1.8189185860327627e-06, + "loss": 0.3262, + "step": 32014 + }, + { + "epoch": 11.82, + "learning_rate": 1.8188816356694174e-06, + "loss": 0.2663, + "step": 32015 + }, + { + "epoch": 11.82, + "learning_rate": 1.8188446853060722e-06, + "loss": 0.3037, + "step": 32016 + }, + { + "epoch": 11.82, + "learning_rate": 1.8188077349427269e-06, + "loss": 0.4237, + "step": 32017 + }, + { + "epoch": 11.82, + "learning_rate": 1.818770784579382e-06, + "loss": 0.2978, + "step": 32018 + }, + { + "epoch": 11.82, + "learning_rate": 1.8187338342160364e-06, + "loss": 0.3973, + "step": 32019 + }, + { + "epoch": 11.82, + "learning_rate": 1.8186968838526914e-06, + "loss": 0.3469, + "step": 32020 + }, + { + "epoch": 11.82, + "learning_rate": 1.818659933489346e-06, + "loss": 0.3405, + "step": 32021 + }, + { + "epoch": 11.82, + "learning_rate": 1.818622983126001e-06, + "loss": 0.3534, + "step": 32022 + }, + { + "epoch": 11.83, + "learning_rate": 1.8185860327626555e-06, + "loss": 0.2901, + "step": 32023 + }, + { + "epoch": 11.83, + "learning_rate": 1.8185490823993102e-06, + "loss": 0.3624, + "step": 32024 + }, + { + "epoch": 11.83, + "learning_rate": 1.818512132035965e-06, + "loss": 0.3774, + "step": 32025 + }, + { + "epoch": 11.83, + "learning_rate": 1.8184751816726197e-06, + "loss": 0.2817, + "step": 32026 + }, + { + "epoch": 11.83, + "learning_rate": 1.8184382313092747e-06, + "loss": 0.2965, + "step": 32027 + }, + { + "epoch": 11.83, + "learning_rate": 1.8184012809459294e-06, + "loss": 0.3074, + "step": 32028 + }, + { + "epoch": 11.83, + "learning_rate": 1.8183643305825842e-06, + "loss": 0.3033, + "step": 32029 + }, + { + "epoch": 11.83, + "learning_rate": 1.8183273802192388e-06, + "loss": 0.2973, + "step": 32030 + }, + { + "epoch": 11.83, + "learning_rate": 1.8182904298558937e-06, + "loss": 0.3277, + "step": 32031 + }, + { + "epoch": 11.83, + "learning_rate": 1.8182534794925483e-06, + "loss": 0.3865, + "step": 32032 + }, + { + "epoch": 11.83, + "learning_rate": 1.818216529129203e-06, + "loss": 0.4193, + "step": 32033 + }, + { + "epoch": 11.83, + "learning_rate": 1.818179578765858e-06, + "loss": 0.2454, + "step": 32034 + }, + { + "epoch": 11.83, + "learning_rate": 1.8181426284025127e-06, + "loss": 0.2153, + "step": 32035 + }, + { + "epoch": 11.83, + "learning_rate": 1.8181056780391675e-06, + "loss": 0.3017, + "step": 32036 + }, + { + "epoch": 11.83, + "learning_rate": 1.8180687276758221e-06, + "loss": 0.2957, + "step": 32037 + }, + { + "epoch": 11.83, + "learning_rate": 1.818031777312477e-06, + "loss": 0.334, + "step": 32038 + }, + { + "epoch": 11.83, + "learning_rate": 1.8179948269491316e-06, + "loss": 0.3323, + "step": 32039 + }, + { + "epoch": 11.83, + "learning_rate": 1.8179578765857867e-06, + "loss": 0.2737, + "step": 32040 + }, + { + "epoch": 11.83, + "learning_rate": 1.8179209262224413e-06, + "loss": 0.2677, + "step": 32041 + }, + { + "epoch": 11.83, + "learning_rate": 1.8178839758590962e-06, + "loss": 0.2305, + "step": 32042 + }, + { + "epoch": 11.83, + "learning_rate": 1.8178470254957508e-06, + "loss": 0.3807, + "step": 32043 + }, + { + "epoch": 11.83, + "learning_rate": 1.8178100751324054e-06, + "loss": 0.398, + "step": 32044 + }, + { + "epoch": 11.83, + "learning_rate": 1.8177731247690603e-06, + "loss": 0.2907, + "step": 32045 + }, + { + "epoch": 11.83, + "learning_rate": 1.817736174405715e-06, + "loss": 0.3029, + "step": 32046 + }, + { + "epoch": 11.83, + "learning_rate": 1.8176992240423698e-06, + "loss": 0.2922, + "step": 32047 + }, + { + "epoch": 11.83, + "learning_rate": 1.8176622736790244e-06, + "loss": 0.3781, + "step": 32048 + }, + { + "epoch": 11.83, + "learning_rate": 1.8176253233156795e-06, + "loss": 0.2251, + "step": 32049 + }, + { + "epoch": 11.84, + "learning_rate": 1.8175883729523341e-06, + "loss": 0.3693, + "step": 32050 + }, + { + "epoch": 11.84, + "learning_rate": 1.817551422588989e-06, + "loss": 0.2405, + "step": 32051 + }, + { + "epoch": 11.84, + "learning_rate": 1.8175144722256436e-06, + "loss": 0.2252, + "step": 32052 + }, + { + "epoch": 11.84, + "learning_rate": 1.8174775218622982e-06, + "loss": 0.2523, + "step": 32053 + }, + { + "epoch": 11.84, + "learning_rate": 1.817440571498953e-06, + "loss": 0.3509, + "step": 32054 + }, + { + "epoch": 11.84, + "learning_rate": 1.8174036211356077e-06, + "loss": 0.293, + "step": 32055 + }, + { + "epoch": 11.84, + "learning_rate": 1.8173666707722628e-06, + "loss": 0.3907, + "step": 32056 + }, + { + "epoch": 11.84, + "learning_rate": 1.8173297204089174e-06, + "loss": 0.3056, + "step": 32057 + }, + { + "epoch": 11.84, + "learning_rate": 1.8172927700455723e-06, + "loss": 0.308, + "step": 32058 + }, + { + "epoch": 11.84, + "learning_rate": 1.8172558196822269e-06, + "loss": 0.3671, + "step": 32059 + }, + { + "epoch": 11.84, + "learning_rate": 1.8172188693188817e-06, + "loss": 0.3739, + "step": 32060 + }, + { + "epoch": 11.84, + "learning_rate": 1.8171819189555364e-06, + "loss": 0.217, + "step": 32061 + }, + { + "epoch": 11.84, + "learning_rate": 1.8171449685921914e-06, + "loss": 0.373, + "step": 32062 + }, + { + "epoch": 11.84, + "learning_rate": 1.817108018228846e-06, + "loss": 0.1641, + "step": 32063 + }, + { + "epoch": 11.84, + "learning_rate": 1.8170710678655005e-06, + "loss": 0.267, + "step": 32064 + }, + { + "epoch": 11.84, + "learning_rate": 1.8170341175021556e-06, + "loss": 0.3908, + "step": 32065 + }, + { + "epoch": 11.84, + "learning_rate": 1.8169971671388102e-06, + "loss": 0.1253, + "step": 32066 + }, + { + "epoch": 11.84, + "learning_rate": 1.816960216775465e-06, + "loss": 0.242, + "step": 32067 + }, + { + "epoch": 11.84, + "learning_rate": 1.8169232664121197e-06, + "loss": 0.2993, + "step": 32068 + }, + { + "epoch": 11.84, + "learning_rate": 1.8168863160487747e-06, + "loss": 0.2668, + "step": 32069 + }, + { + "epoch": 11.84, + "learning_rate": 1.8168493656854292e-06, + "loss": 0.259, + "step": 32070 + }, + { + "epoch": 11.84, + "learning_rate": 1.8168124153220842e-06, + "loss": 0.2924, + "step": 32071 + }, + { + "epoch": 11.84, + "learning_rate": 1.8167754649587389e-06, + "loss": 0.2331, + "step": 32072 + }, + { + "epoch": 11.84, + "learning_rate": 1.8167385145953935e-06, + "loss": 0.4182, + "step": 32073 + }, + { + "epoch": 11.84, + "learning_rate": 1.8167015642320483e-06, + "loss": 0.3134, + "step": 32074 + }, + { + "epoch": 11.84, + "learning_rate": 1.816664613868703e-06, + "loss": 0.2045, + "step": 32075 + }, + { + "epoch": 11.84, + "learning_rate": 1.8166276635053578e-06, + "loss": 0.3683, + "step": 32076 + }, + { + "epoch": 11.85, + "learning_rate": 1.8165907131420125e-06, + "loss": 0.2954, + "step": 32077 + }, + { + "epoch": 11.85, + "learning_rate": 1.8165537627786675e-06, + "loss": 0.2682, + "step": 32078 + }, + { + "epoch": 11.85, + "learning_rate": 1.8165168124153222e-06, + "loss": 0.3392, + "step": 32079 + }, + { + "epoch": 11.85, + "learning_rate": 1.816479862051977e-06, + "loss": 0.2484, + "step": 32080 + }, + { + "epoch": 11.85, + "learning_rate": 1.8164429116886316e-06, + "loss": 0.2362, + "step": 32081 + }, + { + "epoch": 11.85, + "learning_rate": 1.8164059613252863e-06, + "loss": 0.2896, + "step": 32082 + }, + { + "epoch": 11.85, + "learning_rate": 1.8163690109619411e-06, + "loss": 0.3603, + "step": 32083 + }, + { + "epoch": 11.85, + "learning_rate": 1.8163320605985958e-06, + "loss": 0.4135, + "step": 32084 + }, + { + "epoch": 11.85, + "learning_rate": 1.8162951102352508e-06, + "loss": 0.3035, + "step": 32085 + }, + { + "epoch": 11.85, + "learning_rate": 1.8162581598719053e-06, + "loss": 0.2958, + "step": 32086 + }, + { + "epoch": 11.85, + "learning_rate": 1.8162212095085603e-06, + "loss": 0.4853, + "step": 32087 + }, + { + "epoch": 11.85, + "learning_rate": 1.816184259145215e-06, + "loss": 0.282, + "step": 32088 + }, + { + "epoch": 11.85, + "learning_rate": 1.8161473087818698e-06, + "loss": 0.233, + "step": 32089 + }, + { + "epoch": 11.85, + "learning_rate": 1.8161103584185244e-06, + "loss": 0.3912, + "step": 32090 + }, + { + "epoch": 11.85, + "learning_rate": 1.8160734080551795e-06, + "loss": 0.2812, + "step": 32091 + }, + { + "epoch": 11.85, + "learning_rate": 1.816036457691834e-06, + "loss": 0.303, + "step": 32092 + }, + { + "epoch": 11.85, + "learning_rate": 1.8159995073284886e-06, + "loss": 0.3583, + "step": 32093 + }, + { + "epoch": 11.85, + "learning_rate": 1.8159625569651436e-06, + "loss": 0.3224, + "step": 32094 + }, + { + "epoch": 11.85, + "learning_rate": 1.8159256066017983e-06, + "loss": 0.1986, + "step": 32095 + }, + { + "epoch": 11.85, + "learning_rate": 1.815888656238453e-06, + "loss": 0.2897, + "step": 32096 + }, + { + "epoch": 11.85, + "learning_rate": 1.8158517058751077e-06, + "loss": 0.1506, + "step": 32097 + }, + { + "epoch": 11.85, + "learning_rate": 1.8158147555117626e-06, + "loss": 0.2613, + "step": 32098 + }, + { + "epoch": 11.85, + "learning_rate": 1.8157778051484172e-06, + "loss": 0.3419, + "step": 32099 + }, + { + "epoch": 11.85, + "learning_rate": 1.8157408547850723e-06, + "loss": 0.2656, + "step": 32100 + }, + { + "epoch": 11.85, + "learning_rate": 1.815703904421727e-06, + "loss": 0.3693, + "step": 32101 + }, + { + "epoch": 11.85, + "learning_rate": 1.8156669540583816e-06, + "loss": 0.3605, + "step": 32102 + }, + { + "epoch": 11.85, + "learning_rate": 1.8156300036950364e-06, + "loss": 0.4432, + "step": 32103 + }, + { + "epoch": 11.86, + "learning_rate": 1.815593053331691e-06, + "loss": 0.3326, + "step": 32104 + }, + { + "epoch": 11.86, + "learning_rate": 1.8155561029683459e-06, + "loss": 0.1842, + "step": 32105 + }, + { + "epoch": 11.86, + "learning_rate": 1.8155191526050005e-06, + "loss": 0.2818, + "step": 32106 + }, + { + "epoch": 11.86, + "learning_rate": 1.8154822022416556e-06, + "loss": 0.3831, + "step": 32107 + }, + { + "epoch": 11.86, + "learning_rate": 1.8154452518783102e-06, + "loss": 0.3049, + "step": 32108 + }, + { + "epoch": 11.86, + "learning_rate": 1.815408301514965e-06, + "loss": 0.4572, + "step": 32109 + }, + { + "epoch": 11.86, + "learning_rate": 1.8153713511516197e-06, + "loss": 0.2309, + "step": 32110 + }, + { + "epoch": 11.86, + "learning_rate": 1.8153344007882746e-06, + "loss": 0.3842, + "step": 32111 + }, + { + "epoch": 11.86, + "learning_rate": 1.8152974504249292e-06, + "loss": 0.3251, + "step": 32112 + }, + { + "epoch": 11.86, + "learning_rate": 1.8152605000615838e-06, + "loss": 0.261, + "step": 32113 + }, + { + "epoch": 11.86, + "learning_rate": 1.8152235496982389e-06, + "loss": 0.3134, + "step": 32114 + }, + { + "epoch": 11.86, + "learning_rate": 1.8151865993348933e-06, + "loss": 0.4174, + "step": 32115 + }, + { + "epoch": 11.86, + "learning_rate": 1.8151496489715484e-06, + "loss": 0.3193, + "step": 32116 + }, + { + "epoch": 11.86, + "learning_rate": 1.815112698608203e-06, + "loss": 0.3852, + "step": 32117 + }, + { + "epoch": 11.86, + "learning_rate": 1.8150757482448579e-06, + "loss": 0.3321, + "step": 32118 + }, + { + "epoch": 11.86, + "learning_rate": 1.8150387978815125e-06, + "loss": 0.3468, + "step": 32119 + }, + { + "epoch": 11.86, + "learning_rate": 1.8150018475181673e-06, + "loss": 0.2541, + "step": 32120 + }, + { + "epoch": 11.86, + "learning_rate": 1.814964897154822e-06, + "loss": 0.377, + "step": 32121 + }, + { + "epoch": 11.86, + "learning_rate": 1.8149279467914766e-06, + "loss": 0.3044, + "step": 32122 + }, + { + "epoch": 11.86, + "learning_rate": 1.8148909964281317e-06, + "loss": 0.3245, + "step": 32123 + }, + { + "epoch": 11.86, + "learning_rate": 1.8148540460647863e-06, + "loss": 0.3704, + "step": 32124 + }, + { + "epoch": 11.86, + "learning_rate": 1.8148170957014412e-06, + "loss": 0.2427, + "step": 32125 + }, + { + "epoch": 11.86, + "learning_rate": 1.8147801453380958e-06, + "loss": 0.3431, + "step": 32126 + }, + { + "epoch": 11.86, + "learning_rate": 1.8147431949747506e-06, + "loss": 0.1931, + "step": 32127 + }, + { + "epoch": 11.86, + "learning_rate": 1.8147062446114053e-06, + "loss": 0.3161, + "step": 32128 + }, + { + "epoch": 11.86, + "learning_rate": 1.8146692942480603e-06, + "loss": 0.324, + "step": 32129 + }, + { + "epoch": 11.86, + "learning_rate": 1.814632343884715e-06, + "loss": 0.3362, + "step": 32130 + }, + { + "epoch": 11.87, + "learning_rate": 1.8145953935213698e-06, + "loss": 0.2132, + "step": 32131 + }, + { + "epoch": 11.87, + "learning_rate": 1.8145584431580245e-06, + "loss": 0.2834, + "step": 32132 + }, + { + "epoch": 11.87, + "learning_rate": 1.814521492794679e-06, + "loss": 0.2761, + "step": 32133 + }, + { + "epoch": 11.87, + "learning_rate": 1.814484542431334e-06, + "loss": 0.3304, + "step": 32134 + }, + { + "epoch": 11.87, + "learning_rate": 1.8144475920679886e-06, + "loss": 0.3405, + "step": 32135 + }, + { + "epoch": 11.87, + "learning_rate": 1.8144106417046436e-06, + "loss": 0.3176, + "step": 32136 + }, + { + "epoch": 11.87, + "learning_rate": 1.814373691341298e-06, + "loss": 0.3239, + "step": 32137 + }, + { + "epoch": 11.87, + "learning_rate": 1.8143367409779531e-06, + "loss": 0.2945, + "step": 32138 + }, + { + "epoch": 11.87, + "learning_rate": 1.8142997906146078e-06, + "loss": 0.3508, + "step": 32139 + }, + { + "epoch": 11.87, + "learning_rate": 1.8142628402512626e-06, + "loss": 0.3208, + "step": 32140 + }, + { + "epoch": 11.87, + "learning_rate": 1.8142258898879172e-06, + "loss": 0.3152, + "step": 32141 + }, + { + "epoch": 11.87, + "learning_rate": 1.8141889395245719e-06, + "loss": 0.4343, + "step": 32142 + }, + { + "epoch": 11.87, + "learning_rate": 1.8141519891612267e-06, + "loss": 0.2859, + "step": 32143 + }, + { + "epoch": 11.87, + "learning_rate": 1.8141150387978814e-06, + "loss": 0.2847, + "step": 32144 + }, + { + "epoch": 11.87, + "learning_rate": 1.8140780884345364e-06, + "loss": 0.3102, + "step": 32145 + }, + { + "epoch": 11.87, + "learning_rate": 1.814041138071191e-06, + "loss": 0.3788, + "step": 32146 + }, + { + "epoch": 11.87, + "learning_rate": 1.814004187707846e-06, + "loss": 0.2863, + "step": 32147 + }, + { + "epoch": 11.87, + "learning_rate": 1.8139672373445005e-06, + "loss": 0.2749, + "step": 32148 + }, + { + "epoch": 11.87, + "learning_rate": 1.8139302869811554e-06, + "loss": 0.3882, + "step": 32149 + }, + { + "epoch": 11.87, + "learning_rate": 1.81389333661781e-06, + "loss": 0.2586, + "step": 32150 + }, + { + "epoch": 11.87, + "learning_rate": 1.813856386254465e-06, + "loss": 0.3624, + "step": 32151 + }, + { + "epoch": 11.87, + "learning_rate": 1.8138194358911197e-06, + "loss": 0.2223, + "step": 32152 + }, + { + "epoch": 11.87, + "learning_rate": 1.8137824855277742e-06, + "loss": 0.34, + "step": 32153 + }, + { + "epoch": 11.87, + "learning_rate": 1.8137455351644292e-06, + "loss": 0.3713, + "step": 32154 + }, + { + "epoch": 11.87, + "learning_rate": 1.8137085848010838e-06, + "loss": 0.2669, + "step": 32155 + }, + { + "epoch": 11.87, + "learning_rate": 1.8136716344377387e-06, + "loss": 0.514, + "step": 32156 + }, + { + "epoch": 11.87, + "learning_rate": 1.8136346840743933e-06, + "loss": 0.46, + "step": 32157 + }, + { + "epoch": 11.88, + "learning_rate": 1.8135977337110484e-06, + "loss": 0.2221, + "step": 32158 + }, + { + "epoch": 11.88, + "learning_rate": 1.8135607833477028e-06, + "loss": 0.3621, + "step": 32159 + }, + { + "epoch": 11.88, + "learning_rate": 1.8135238329843579e-06, + "loss": 0.3256, + "step": 32160 + }, + { + "epoch": 11.88, + "learning_rate": 1.8134868826210125e-06, + "loss": 0.1988, + "step": 32161 + }, + { + "epoch": 11.88, + "learning_rate": 1.8134499322576672e-06, + "loss": 0.2813, + "step": 32162 + }, + { + "epoch": 11.88, + "learning_rate": 1.813412981894322e-06, + "loss": 0.4222, + "step": 32163 + }, + { + "epoch": 11.88, + "learning_rate": 1.8133760315309766e-06, + "loss": 0.3279, + "step": 32164 + }, + { + "epoch": 11.88, + "learning_rate": 1.8133390811676315e-06, + "loss": 0.3622, + "step": 32165 + }, + { + "epoch": 11.88, + "learning_rate": 1.8133021308042861e-06, + "loss": 0.3356, + "step": 32166 + }, + { + "epoch": 11.88, + "learning_rate": 1.8132651804409412e-06, + "loss": 0.2672, + "step": 32167 + }, + { + "epoch": 11.88, + "learning_rate": 1.8132282300775958e-06, + "loss": 0.306, + "step": 32168 + }, + { + "epoch": 11.88, + "learning_rate": 1.8131912797142507e-06, + "loss": 0.3356, + "step": 32169 + }, + { + "epoch": 11.88, + "learning_rate": 1.8131543293509053e-06, + "loss": 0.2543, + "step": 32170 + }, + { + "epoch": 11.88, + "learning_rate": 1.8131173789875602e-06, + "loss": 0.3309, + "step": 32171 + }, + { + "epoch": 11.88, + "learning_rate": 1.8130804286242148e-06, + "loss": 0.1875, + "step": 32172 + }, + { + "epoch": 11.88, + "learning_rate": 1.8130434782608694e-06, + "loss": 0.3299, + "step": 32173 + }, + { + "epoch": 11.88, + "learning_rate": 1.8130065278975245e-06, + "loss": 0.2295, + "step": 32174 + }, + { + "epoch": 11.88, + "learning_rate": 1.8129695775341791e-06, + "loss": 0.4816, + "step": 32175 + }, + { + "epoch": 11.88, + "learning_rate": 1.812932627170834e-06, + "loss": 0.1726, + "step": 32176 + }, + { + "epoch": 11.88, + "learning_rate": 1.8128956768074886e-06, + "loss": 0.3236, + "step": 32177 + }, + { + "epoch": 11.88, + "learning_rate": 1.8128587264441435e-06, + "loss": 0.3268, + "step": 32178 + }, + { + "epoch": 11.88, + "learning_rate": 1.812821776080798e-06, + "loss": 0.4093, + "step": 32179 + }, + { + "epoch": 11.88, + "learning_rate": 1.8127848257174531e-06, + "loss": 0.1735, + "step": 32180 + }, + { + "epoch": 11.88, + "learning_rate": 1.8127478753541078e-06, + "loss": 0.295, + "step": 32181 + }, + { + "epoch": 11.88, + "learning_rate": 1.8127109249907622e-06, + "loss": 0.2821, + "step": 32182 + }, + { + "epoch": 11.88, + "learning_rate": 1.8126739746274173e-06, + "loss": 0.238, + "step": 32183 + }, + { + "epoch": 11.88, + "learning_rate": 1.812637024264072e-06, + "loss": 0.3733, + "step": 32184 + }, + { + "epoch": 11.89, + "learning_rate": 1.8126000739007268e-06, + "loss": 0.4353, + "step": 32185 + }, + { + "epoch": 11.89, + "learning_rate": 1.8125631235373814e-06, + "loss": 0.3485, + "step": 32186 + }, + { + "epoch": 11.89, + "learning_rate": 1.8125261731740365e-06, + "loss": 0.4655, + "step": 32187 + }, + { + "epoch": 11.89, + "learning_rate": 1.8124892228106909e-06, + "loss": 0.2373, + "step": 32188 + }, + { + "epoch": 11.89, + "learning_rate": 1.812452272447346e-06, + "loss": 0.3832, + "step": 32189 + }, + { + "epoch": 11.89, + "learning_rate": 1.8124153220840006e-06, + "loss": 0.4636, + "step": 32190 + }, + { + "epoch": 11.89, + "learning_rate": 1.8123783717206554e-06, + "loss": 0.2791, + "step": 32191 + }, + { + "epoch": 11.89, + "learning_rate": 1.81234142135731e-06, + "loss": 0.3264, + "step": 32192 + }, + { + "epoch": 11.89, + "learning_rate": 1.8123044709939647e-06, + "loss": 0.3016, + "step": 32193 + }, + { + "epoch": 11.89, + "learning_rate": 1.8122675206306195e-06, + "loss": 0.3855, + "step": 32194 + }, + { + "epoch": 11.89, + "learning_rate": 1.8122305702672742e-06, + "loss": 0.3949, + "step": 32195 + }, + { + "epoch": 11.89, + "learning_rate": 1.8121936199039292e-06, + "loss": 0.3765, + "step": 32196 + }, + { + "epoch": 11.89, + "learning_rate": 1.8121566695405839e-06, + "loss": 0.2973, + "step": 32197 + }, + { + "epoch": 11.89, + "learning_rate": 1.8121197191772387e-06, + "loss": 0.3266, + "step": 32198 + }, + { + "epoch": 11.89, + "learning_rate": 1.8120827688138934e-06, + "loss": 0.3255, + "step": 32199 + }, + { + "epoch": 11.89, + "learning_rate": 1.8120458184505482e-06, + "loss": 0.422, + "step": 32200 + }, + { + "epoch": 11.89, + "learning_rate": 1.8120088680872028e-06, + "loss": 0.2921, + "step": 32201 + }, + { + "epoch": 11.89, + "learning_rate": 1.8119719177238575e-06, + "loss": 0.2362, + "step": 32202 + }, + { + "epoch": 11.89, + "learning_rate": 1.8119349673605125e-06, + "loss": 0.3174, + "step": 32203 + }, + { + "epoch": 11.89, + "learning_rate": 1.811898016997167e-06, + "loss": 0.3715, + "step": 32204 + }, + { + "epoch": 11.89, + "learning_rate": 1.811861066633822e-06, + "loss": 0.5198, + "step": 32205 + }, + { + "epoch": 11.89, + "learning_rate": 1.8118241162704767e-06, + "loss": 0.312, + "step": 32206 + }, + { + "epoch": 11.89, + "learning_rate": 1.8117871659071315e-06, + "loss": 0.3198, + "step": 32207 + }, + { + "epoch": 11.89, + "learning_rate": 1.8117502155437861e-06, + "loss": 0.3104, + "step": 32208 + }, + { + "epoch": 11.89, + "learning_rate": 1.8117132651804412e-06, + "loss": 0.3323, + "step": 32209 + }, + { + "epoch": 11.89, + "learning_rate": 1.8116763148170956e-06, + "loss": 0.3137, + "step": 32210 + }, + { + "epoch": 11.89, + "learning_rate": 1.8116393644537507e-06, + "loss": 0.2927, + "step": 32211 + }, + { + "epoch": 11.9, + "learning_rate": 1.8116024140904053e-06, + "loss": 0.2769, + "step": 32212 + }, + { + "epoch": 11.9, + "learning_rate": 1.81156546372706e-06, + "loss": 0.4436, + "step": 32213 + }, + { + "epoch": 11.9, + "learning_rate": 1.8115285133637148e-06, + "loss": 0.2591, + "step": 32214 + }, + { + "epoch": 11.9, + "learning_rate": 1.8114915630003694e-06, + "loss": 0.4826, + "step": 32215 + }, + { + "epoch": 11.9, + "learning_rate": 1.8114546126370243e-06, + "loss": 0.4534, + "step": 32216 + }, + { + "epoch": 11.9, + "learning_rate": 1.811417662273679e-06, + "loss": 0.2717, + "step": 32217 + }, + { + "epoch": 11.9, + "learning_rate": 1.811380711910334e-06, + "loss": 0.3024, + "step": 32218 + }, + { + "epoch": 11.9, + "learning_rate": 1.8113437615469886e-06, + "loss": 0.3307, + "step": 32219 + }, + { + "epoch": 11.9, + "learning_rate": 1.8113068111836435e-06, + "loss": 0.2915, + "step": 32220 + }, + { + "epoch": 11.9, + "learning_rate": 1.8112698608202981e-06, + "loss": 0.337, + "step": 32221 + }, + { + "epoch": 11.9, + "learning_rate": 1.8112329104569527e-06, + "loss": 0.2513, + "step": 32222 + }, + { + "epoch": 11.9, + "learning_rate": 1.8111959600936076e-06, + "loss": 0.4289, + "step": 32223 + }, + { + "epoch": 11.9, + "learning_rate": 1.8111590097302622e-06, + "loss": 0.3873, + "step": 32224 + }, + { + "epoch": 11.9, + "learning_rate": 1.8111220593669173e-06, + "loss": 0.3408, + "step": 32225 + }, + { + "epoch": 11.9, + "learning_rate": 1.811085109003572e-06, + "loss": 0.3654, + "step": 32226 + }, + { + "epoch": 11.9, + "learning_rate": 1.8110481586402268e-06, + "loss": 0.257, + "step": 32227 + }, + { + "epoch": 11.9, + "learning_rate": 1.8110112082768814e-06, + "loss": 0.3629, + "step": 32228 + }, + { + "epoch": 11.9, + "learning_rate": 1.8109742579135363e-06, + "loss": 0.2816, + "step": 32229 + }, + { + "epoch": 11.9, + "learning_rate": 1.810937307550191e-06, + "loss": 0.2785, + "step": 32230 + }, + { + "epoch": 11.9, + "learning_rate": 1.810900357186846e-06, + "loss": 0.2615, + "step": 32231 + }, + { + "epoch": 11.9, + "learning_rate": 1.8108634068235004e-06, + "loss": 0.3535, + "step": 32232 + }, + { + "epoch": 11.9, + "learning_rate": 1.810826456460155e-06, + "loss": 0.4423, + "step": 32233 + }, + { + "epoch": 11.9, + "learning_rate": 1.81078950609681e-06, + "loss": 0.1771, + "step": 32234 + }, + { + "epoch": 11.9, + "learning_rate": 1.8107525557334647e-06, + "loss": 0.3075, + "step": 32235 + }, + { + "epoch": 11.9, + "learning_rate": 1.8107156053701196e-06, + "loss": 0.2856, + "step": 32236 + }, + { + "epoch": 11.9, + "learning_rate": 1.8106786550067742e-06, + "loss": 0.3457, + "step": 32237 + }, + { + "epoch": 11.9, + "learning_rate": 1.810641704643429e-06, + "loss": 0.2565, + "step": 32238 + }, + { + "epoch": 11.91, + "learning_rate": 1.8106047542800837e-06, + "loss": 0.4324, + "step": 32239 + }, + { + "epoch": 11.91, + "learning_rate": 1.8105678039167387e-06, + "loss": 0.2326, + "step": 32240 + }, + { + "epoch": 11.91, + "learning_rate": 1.8105308535533934e-06, + "loss": 0.3736, + "step": 32241 + }, + { + "epoch": 11.91, + "learning_rate": 1.810493903190048e-06, + "loss": 0.3127, + "step": 32242 + }, + { + "epoch": 11.91, + "learning_rate": 1.8104569528267029e-06, + "loss": 0.3155, + "step": 32243 + }, + { + "epoch": 11.91, + "learning_rate": 1.8104200024633575e-06, + "loss": 0.2722, + "step": 32244 + }, + { + "epoch": 11.91, + "learning_rate": 1.8103830521000124e-06, + "loss": 0.3082, + "step": 32245 + }, + { + "epoch": 11.91, + "learning_rate": 1.810346101736667e-06, + "loss": 0.3026, + "step": 32246 + }, + { + "epoch": 11.91, + "learning_rate": 1.810309151373322e-06, + "loss": 0.2543, + "step": 32247 + }, + { + "epoch": 11.91, + "learning_rate": 1.8102722010099767e-06, + "loss": 0.2817, + "step": 32248 + }, + { + "epoch": 11.91, + "learning_rate": 1.8102352506466315e-06, + "loss": 0.3393, + "step": 32249 + }, + { + "epoch": 11.91, + "learning_rate": 1.8101983002832862e-06, + "loss": 0.3437, + "step": 32250 + }, + { + "epoch": 11.91, + "learning_rate": 1.810161349919941e-06, + "loss": 0.3042, + "step": 32251 + }, + { + "epoch": 11.91, + "learning_rate": 1.8101243995565957e-06, + "loss": 0.2709, + "step": 32252 + }, + { + "epoch": 11.91, + "learning_rate": 1.8100874491932503e-06, + "loss": 0.3441, + "step": 32253 + }, + { + "epoch": 11.91, + "learning_rate": 1.8100504988299054e-06, + "loss": 0.5385, + "step": 32254 + }, + { + "epoch": 11.91, + "learning_rate": 1.8100135484665598e-06, + "loss": 0.3179, + "step": 32255 + }, + { + "epoch": 11.91, + "learning_rate": 1.8099765981032148e-06, + "loss": 0.2285, + "step": 32256 + }, + { + "epoch": 11.91, + "learning_rate": 1.8099396477398695e-06, + "loss": 0.2387, + "step": 32257 + }, + { + "epoch": 11.91, + "learning_rate": 1.8099026973765243e-06, + "loss": 0.3254, + "step": 32258 + }, + { + "epoch": 11.91, + "learning_rate": 1.809865747013179e-06, + "loss": 0.2548, + "step": 32259 + }, + { + "epoch": 11.91, + "learning_rate": 1.809828796649834e-06, + "loss": 0.4254, + "step": 32260 + }, + { + "epoch": 11.91, + "learning_rate": 1.8097918462864884e-06, + "loss": 0.2516, + "step": 32261 + }, + { + "epoch": 11.91, + "learning_rate": 1.809754895923143e-06, + "loss": 0.2435, + "step": 32262 + }, + { + "epoch": 11.91, + "learning_rate": 1.8097179455597981e-06, + "loss": 0.3157, + "step": 32263 + }, + { + "epoch": 11.91, + "learning_rate": 1.8096809951964528e-06, + "loss": 0.2528, + "step": 32264 + }, + { + "epoch": 11.91, + "learning_rate": 1.8096440448331076e-06, + "loss": 0.3528, + "step": 32265 + }, + { + "epoch": 11.92, + "learning_rate": 1.8096070944697623e-06, + "loss": 0.3483, + "step": 32266 + }, + { + "epoch": 11.92, + "learning_rate": 1.8095701441064171e-06, + "loss": 0.3409, + "step": 32267 + }, + { + "epoch": 11.92, + "learning_rate": 1.8095331937430717e-06, + "loss": 0.43, + "step": 32268 + }, + { + "epoch": 11.92, + "learning_rate": 1.8094962433797268e-06, + "loss": 0.2993, + "step": 32269 + }, + { + "epoch": 11.92, + "learning_rate": 1.8094592930163814e-06, + "loss": 0.2997, + "step": 32270 + }, + { + "epoch": 11.92, + "learning_rate": 1.8094223426530363e-06, + "loss": 0.3067, + "step": 32271 + }, + { + "epoch": 11.92, + "learning_rate": 1.809385392289691e-06, + "loss": 0.2482, + "step": 32272 + }, + { + "epoch": 11.92, + "learning_rate": 1.8093484419263456e-06, + "loss": 0.3477, + "step": 32273 + }, + { + "epoch": 11.92, + "learning_rate": 1.8093114915630004e-06, + "loss": 0.4485, + "step": 32274 + }, + { + "epoch": 11.92, + "learning_rate": 1.809274541199655e-06, + "loss": 0.2492, + "step": 32275 + }, + { + "epoch": 11.92, + "learning_rate": 1.80923759083631e-06, + "loss": 0.342, + "step": 32276 + }, + { + "epoch": 11.92, + "learning_rate": 1.8092006404729645e-06, + "loss": 0.4223, + "step": 32277 + }, + { + "epoch": 11.92, + "learning_rate": 1.8091636901096196e-06, + "loss": 0.3303, + "step": 32278 + }, + { + "epoch": 11.92, + "learning_rate": 1.8091267397462742e-06, + "loss": 0.4147, + "step": 32279 + }, + { + "epoch": 11.92, + "learning_rate": 1.809089789382929e-06, + "loss": 0.3913, + "step": 32280 + }, + { + "epoch": 11.92, + "learning_rate": 1.8090528390195837e-06, + "loss": 0.2785, + "step": 32281 + }, + { + "epoch": 11.92, + "learning_rate": 1.8090158886562383e-06, + "loss": 0.3732, + "step": 32282 + }, + { + "epoch": 11.92, + "learning_rate": 1.8089789382928932e-06, + "loss": 0.2318, + "step": 32283 + }, + { + "epoch": 11.92, + "learning_rate": 1.8089419879295478e-06, + "loss": 0.4229, + "step": 32284 + }, + { + "epoch": 11.92, + "learning_rate": 1.8089050375662029e-06, + "loss": 0.239, + "step": 32285 + }, + { + "epoch": 11.92, + "learning_rate": 1.8088680872028575e-06, + "loss": 0.2937, + "step": 32286 + }, + { + "epoch": 11.92, + "learning_rate": 1.8088311368395124e-06, + "loss": 0.2905, + "step": 32287 + }, + { + "epoch": 11.92, + "learning_rate": 1.808794186476167e-06, + "loss": 0.3474, + "step": 32288 + }, + { + "epoch": 11.92, + "learning_rate": 1.8087572361128219e-06, + "loss": 0.3613, + "step": 32289 + }, + { + "epoch": 11.92, + "learning_rate": 1.8087202857494765e-06, + "loss": 0.2618, + "step": 32290 + }, + { + "epoch": 11.92, + "learning_rate": 1.8086833353861316e-06, + "loss": 0.4239, + "step": 32291 + }, + { + "epoch": 11.92, + "learning_rate": 1.8086463850227862e-06, + "loss": 0.1715, + "step": 32292 + }, + { + "epoch": 11.93, + "learning_rate": 1.8086094346594408e-06, + "loss": 0.2603, + "step": 32293 + }, + { + "epoch": 11.93, + "learning_rate": 1.8085724842960957e-06, + "loss": 0.4251, + "step": 32294 + }, + { + "epoch": 11.93, + "learning_rate": 1.8085355339327503e-06, + "loss": 0.2667, + "step": 32295 + }, + { + "epoch": 11.93, + "learning_rate": 1.8084985835694052e-06, + "loss": 0.3394, + "step": 32296 + }, + { + "epoch": 11.93, + "learning_rate": 1.8084616332060598e-06, + "loss": 0.3057, + "step": 32297 + }, + { + "epoch": 11.93, + "learning_rate": 1.8084246828427149e-06, + "loss": 0.2586, + "step": 32298 + }, + { + "epoch": 11.93, + "learning_rate": 1.8083877324793695e-06, + "loss": 0.3368, + "step": 32299 + }, + { + "epoch": 11.93, + "learning_rate": 1.8083507821160243e-06, + "loss": 0.3943, + "step": 32300 + }, + { + "epoch": 11.93, + "learning_rate": 1.808313831752679e-06, + "loss": 0.3749, + "step": 32301 + }, + { + "epoch": 11.93, + "learning_rate": 1.8082768813893336e-06, + "loss": 0.3545, + "step": 32302 + }, + { + "epoch": 11.93, + "learning_rate": 1.8082399310259885e-06, + "loss": 0.2143, + "step": 32303 + }, + { + "epoch": 11.93, + "learning_rate": 1.808202980662643e-06, + "loss": 0.2194, + "step": 32304 + }, + { + "epoch": 11.93, + "learning_rate": 1.8081660302992982e-06, + "loss": 0.262, + "step": 32305 + }, + { + "epoch": 11.93, + "learning_rate": 1.8081290799359526e-06, + "loss": 0.3664, + "step": 32306 + }, + { + "epoch": 11.93, + "learning_rate": 1.8080921295726076e-06, + "loss": 0.4467, + "step": 32307 + }, + { + "epoch": 11.93, + "learning_rate": 1.8080551792092623e-06, + "loss": 0.3646, + "step": 32308 + }, + { + "epoch": 11.93, + "learning_rate": 1.8080182288459171e-06, + "loss": 0.2876, + "step": 32309 + }, + { + "epoch": 11.93, + "learning_rate": 1.8079812784825718e-06, + "loss": 0.3507, + "step": 32310 + }, + { + "epoch": 11.93, + "learning_rate": 1.8079443281192264e-06, + "loss": 0.3018, + "step": 32311 + }, + { + "epoch": 11.93, + "learning_rate": 1.8079073777558813e-06, + "loss": 0.4023, + "step": 32312 + }, + { + "epoch": 11.93, + "learning_rate": 1.8078704273925359e-06, + "loss": 0.3732, + "step": 32313 + }, + { + "epoch": 11.93, + "learning_rate": 1.807833477029191e-06, + "loss": 0.3069, + "step": 32314 + }, + { + "epoch": 11.93, + "learning_rate": 1.8077965266658456e-06, + "loss": 0.4066, + "step": 32315 + }, + { + "epoch": 11.93, + "learning_rate": 1.8077595763025004e-06, + "loss": 0.2527, + "step": 32316 + }, + { + "epoch": 11.93, + "learning_rate": 1.807722625939155e-06, + "loss": 0.2965, + "step": 32317 + }, + { + "epoch": 11.93, + "learning_rate": 1.80768567557581e-06, + "loss": 0.2166, + "step": 32318 + }, + { + "epoch": 11.93, + "learning_rate": 1.8076487252124646e-06, + "loss": 0.3616, + "step": 32319 + }, + { + "epoch": 11.94, + "learning_rate": 1.8076117748491196e-06, + "loss": 0.1961, + "step": 32320 + }, + { + "epoch": 11.94, + "learning_rate": 1.8075748244857743e-06, + "loss": 0.4218, + "step": 32321 + }, + { + "epoch": 11.94, + "learning_rate": 1.8075378741224287e-06, + "loss": 0.2871, + "step": 32322 + }, + { + "epoch": 11.94, + "learning_rate": 1.8075009237590837e-06, + "loss": 0.2819, + "step": 32323 + }, + { + "epoch": 11.94, + "learning_rate": 1.8074639733957384e-06, + "loss": 0.4055, + "step": 32324 + }, + { + "epoch": 11.94, + "learning_rate": 1.8074270230323932e-06, + "loss": 0.4384, + "step": 32325 + }, + { + "epoch": 11.94, + "learning_rate": 1.8073900726690479e-06, + "loss": 0.4161, + "step": 32326 + }, + { + "epoch": 11.94, + "learning_rate": 1.807353122305703e-06, + "loss": 0.3796, + "step": 32327 + }, + { + "epoch": 11.94, + "learning_rate": 1.8073161719423573e-06, + "loss": 0.2784, + "step": 32328 + }, + { + "epoch": 11.94, + "learning_rate": 1.8072792215790124e-06, + "loss": 0.2911, + "step": 32329 + }, + { + "epoch": 11.94, + "learning_rate": 1.807242271215667e-06, + "loss": 0.3687, + "step": 32330 + }, + { + "epoch": 11.94, + "learning_rate": 1.8072053208523217e-06, + "loss": 0.3208, + "step": 32331 + }, + { + "epoch": 11.94, + "learning_rate": 1.8071683704889765e-06, + "loss": 0.247, + "step": 32332 + }, + { + "epoch": 11.94, + "learning_rate": 1.8071314201256312e-06, + "loss": 0.4441, + "step": 32333 + }, + { + "epoch": 11.94, + "learning_rate": 1.807094469762286e-06, + "loss": 0.3586, + "step": 32334 + }, + { + "epoch": 11.94, + "learning_rate": 1.8070575193989406e-06, + "loss": 0.2483, + "step": 32335 + }, + { + "epoch": 11.94, + "learning_rate": 1.8070205690355957e-06, + "loss": 0.2857, + "step": 32336 + }, + { + "epoch": 11.94, + "learning_rate": 1.8069836186722503e-06, + "loss": 0.398, + "step": 32337 + }, + { + "epoch": 11.94, + "learning_rate": 1.8069466683089052e-06, + "loss": 0.4658, + "step": 32338 + }, + { + "epoch": 11.94, + "learning_rate": 1.8069097179455598e-06, + "loss": 0.2926, + "step": 32339 + }, + { + "epoch": 11.94, + "learning_rate": 1.8068727675822147e-06, + "loss": 0.4093, + "step": 32340 + }, + { + "epoch": 11.94, + "learning_rate": 1.8068358172188693e-06, + "loss": 0.3837, + "step": 32341 + }, + { + "epoch": 11.94, + "learning_rate": 1.806798866855524e-06, + "loss": 0.2907, + "step": 32342 + }, + { + "epoch": 11.94, + "learning_rate": 1.806761916492179e-06, + "loss": 0.3117, + "step": 32343 + }, + { + "epoch": 11.94, + "learning_rate": 1.8067249661288334e-06, + "loss": 0.3485, + "step": 32344 + }, + { + "epoch": 11.94, + "learning_rate": 1.8066880157654885e-06, + "loss": 0.3199, + "step": 32345 + }, + { + "epoch": 11.94, + "learning_rate": 1.8066510654021431e-06, + "loss": 0.3686, + "step": 32346 + }, + { + "epoch": 11.94, + "learning_rate": 1.806614115038798e-06, + "loss": 0.2876, + "step": 32347 + }, + { + "epoch": 11.95, + "learning_rate": 1.8065771646754526e-06, + "loss": 0.4254, + "step": 32348 + }, + { + "epoch": 11.95, + "learning_rate": 1.8065402143121077e-06, + "loss": 0.3307, + "step": 32349 + }, + { + "epoch": 11.95, + "learning_rate": 1.806503263948762e-06, + "loss": 0.2988, + "step": 32350 + }, + { + "epoch": 11.95, + "learning_rate": 1.8064663135854167e-06, + "loss": 0.3377, + "step": 32351 + }, + { + "epoch": 11.95, + "learning_rate": 1.8064293632220718e-06, + "loss": 0.3485, + "step": 32352 + }, + { + "epoch": 11.95, + "learning_rate": 1.8063924128587264e-06, + "loss": 0.251, + "step": 32353 + }, + { + "epoch": 11.95, + "learning_rate": 1.8063554624953813e-06, + "loss": 0.2847, + "step": 32354 + }, + { + "epoch": 11.95, + "learning_rate": 1.806318512132036e-06, + "loss": 0.3697, + "step": 32355 + }, + { + "epoch": 11.95, + "learning_rate": 1.8062815617686908e-06, + "loss": 0.2871, + "step": 32356 + }, + { + "epoch": 11.95, + "learning_rate": 1.8062446114053454e-06, + "loss": 0.2408, + "step": 32357 + }, + { + "epoch": 11.95, + "learning_rate": 1.8062076610420005e-06, + "loss": 0.2557, + "step": 32358 + }, + { + "epoch": 11.95, + "learning_rate": 1.806170710678655e-06, + "loss": 0.2598, + "step": 32359 + }, + { + "epoch": 11.95, + "learning_rate": 1.80613376031531e-06, + "loss": 0.2988, + "step": 32360 + }, + { + "epoch": 11.95, + "learning_rate": 1.8060968099519646e-06, + "loss": 0.3103, + "step": 32361 + }, + { + "epoch": 11.95, + "learning_rate": 1.8060598595886192e-06, + "loss": 0.294, + "step": 32362 + }, + { + "epoch": 11.95, + "learning_rate": 1.806022909225274e-06, + "loss": 0.4315, + "step": 32363 + }, + { + "epoch": 11.95, + "learning_rate": 1.8059859588619287e-06, + "loss": 0.288, + "step": 32364 + }, + { + "epoch": 11.95, + "learning_rate": 1.8059490084985838e-06, + "loss": 0.1911, + "step": 32365 + }, + { + "epoch": 11.95, + "learning_rate": 1.8059120581352384e-06, + "loss": 0.3221, + "step": 32366 + }, + { + "epoch": 11.95, + "learning_rate": 1.8058751077718932e-06, + "loss": 0.2464, + "step": 32367 + }, + { + "epoch": 11.95, + "learning_rate": 1.8058381574085479e-06, + "loss": 0.2929, + "step": 32368 + }, + { + "epoch": 11.95, + "learning_rate": 1.8058012070452027e-06, + "loss": 0.22, + "step": 32369 + }, + { + "epoch": 11.95, + "learning_rate": 1.8057642566818574e-06, + "loss": 0.3575, + "step": 32370 + }, + { + "epoch": 11.95, + "learning_rate": 1.805727306318512e-06, + "loss": 0.2787, + "step": 32371 + }, + { + "epoch": 11.95, + "learning_rate": 1.805690355955167e-06, + "loss": 0.1441, + "step": 32372 + }, + { + "epoch": 11.95, + "learning_rate": 1.8056534055918215e-06, + "loss": 0.3129, + "step": 32373 + }, + { + "epoch": 11.95, + "learning_rate": 1.8056164552284765e-06, + "loss": 0.2894, + "step": 32374 + }, + { + "epoch": 11.96, + "learning_rate": 1.8055795048651312e-06, + "loss": 0.3561, + "step": 32375 + }, + { + "epoch": 11.96, + "learning_rate": 1.805542554501786e-06, + "loss": 0.2719, + "step": 32376 + }, + { + "epoch": 11.96, + "learning_rate": 1.8055056041384407e-06, + "loss": 0.2308, + "step": 32377 + }, + { + "epoch": 11.96, + "learning_rate": 1.8054686537750957e-06, + "loss": 0.2932, + "step": 32378 + }, + { + "epoch": 11.96, + "learning_rate": 1.8054317034117502e-06, + "loss": 0.3453, + "step": 32379 + }, + { + "epoch": 11.96, + "learning_rate": 1.8053947530484052e-06, + "loss": 0.4896, + "step": 32380 + }, + { + "epoch": 11.96, + "learning_rate": 1.8053578026850598e-06, + "loss": 0.3467, + "step": 32381 + }, + { + "epoch": 11.96, + "learning_rate": 1.8053208523217145e-06, + "loss": 0.2519, + "step": 32382 + }, + { + "epoch": 11.96, + "learning_rate": 1.8052839019583693e-06, + "loss": 0.3831, + "step": 32383 + }, + { + "epoch": 11.96, + "learning_rate": 1.805246951595024e-06, + "loss": 0.4135, + "step": 32384 + }, + { + "epoch": 11.96, + "learning_rate": 1.8052100012316788e-06, + "loss": 0.3721, + "step": 32385 + }, + { + "epoch": 11.96, + "learning_rate": 1.8051730508683335e-06, + "loss": 0.2686, + "step": 32386 + }, + { + "epoch": 11.96, + "learning_rate": 1.8051361005049885e-06, + "loss": 0.2446, + "step": 32387 + }, + { + "epoch": 11.96, + "learning_rate": 1.8050991501416431e-06, + "loss": 0.3575, + "step": 32388 + }, + { + "epoch": 11.96, + "learning_rate": 1.805062199778298e-06, + "loss": 0.3144, + "step": 32389 + }, + { + "epoch": 11.96, + "learning_rate": 1.8050252494149526e-06, + "loss": 0.3379, + "step": 32390 + }, + { + "epoch": 11.96, + "learning_rate": 1.8049882990516073e-06, + "loss": 0.3184, + "step": 32391 + }, + { + "epoch": 11.96, + "learning_rate": 1.8049513486882621e-06, + "loss": 0.273, + "step": 32392 + }, + { + "epoch": 11.96, + "learning_rate": 1.8049143983249168e-06, + "loss": 0.4293, + "step": 32393 + }, + { + "epoch": 11.96, + "learning_rate": 1.8048774479615718e-06, + "loss": 0.2682, + "step": 32394 + }, + { + "epoch": 11.96, + "learning_rate": 1.8048404975982262e-06, + "loss": 0.3401, + "step": 32395 + }, + { + "epoch": 11.96, + "learning_rate": 1.8048035472348813e-06, + "loss": 0.289, + "step": 32396 + }, + { + "epoch": 11.96, + "learning_rate": 1.804766596871536e-06, + "loss": 0.2303, + "step": 32397 + }, + { + "epoch": 11.96, + "learning_rate": 1.8047296465081908e-06, + "loss": 0.2646, + "step": 32398 + }, + { + "epoch": 11.96, + "learning_rate": 1.8046926961448454e-06, + "loss": 0.2078, + "step": 32399 + }, + { + "epoch": 11.96, + "learning_rate": 1.8046557457815005e-06, + "loss": 0.28, + "step": 32400 + }, + { + "epoch": 11.96, + "learning_rate": 1.804618795418155e-06, + "loss": 0.3728, + "step": 32401 + }, + { + "epoch": 11.97, + "learning_rate": 1.8045818450548095e-06, + "loss": 0.2396, + "step": 32402 + }, + { + "epoch": 11.97, + "learning_rate": 1.8045448946914646e-06, + "loss": 0.4247, + "step": 32403 + }, + { + "epoch": 11.97, + "learning_rate": 1.8045079443281192e-06, + "loss": 0.3096, + "step": 32404 + }, + { + "epoch": 11.97, + "learning_rate": 1.804470993964774e-06, + "loss": 0.3013, + "step": 32405 + }, + { + "epoch": 11.97, + "learning_rate": 1.8044340436014287e-06, + "loss": 0.3525, + "step": 32406 + }, + { + "epoch": 11.97, + "learning_rate": 1.8043970932380836e-06, + "loss": 0.4545, + "step": 32407 + }, + { + "epoch": 11.97, + "learning_rate": 1.8043601428747382e-06, + "loss": 0.2425, + "step": 32408 + }, + { + "epoch": 11.97, + "learning_rate": 1.8043231925113933e-06, + "loss": 0.2365, + "step": 32409 + }, + { + "epoch": 11.97, + "learning_rate": 1.804286242148048e-06, + "loss": 0.2836, + "step": 32410 + }, + { + "epoch": 11.97, + "learning_rate": 1.8042492917847025e-06, + "loss": 0.3692, + "step": 32411 + }, + { + "epoch": 11.97, + "learning_rate": 1.8042123414213574e-06, + "loss": 0.3674, + "step": 32412 + }, + { + "epoch": 11.97, + "learning_rate": 1.804175391058012e-06, + "loss": 0.3728, + "step": 32413 + }, + { + "epoch": 11.97, + "learning_rate": 1.8041384406946669e-06, + "loss": 0.5043, + "step": 32414 + }, + { + "epoch": 11.97, + "learning_rate": 1.8041014903313215e-06, + "loss": 0.3703, + "step": 32415 + }, + { + "epoch": 11.97, + "learning_rate": 1.8040645399679766e-06, + "loss": 0.2567, + "step": 32416 + }, + { + "epoch": 11.97, + "learning_rate": 1.804027589604631e-06, + "loss": 0.312, + "step": 32417 + }, + { + "epoch": 11.97, + "learning_rate": 1.803990639241286e-06, + "loss": 0.2301, + "step": 32418 + }, + { + "epoch": 11.97, + "learning_rate": 1.8039536888779407e-06, + "loss": 0.2878, + "step": 32419 + }, + { + "epoch": 11.97, + "learning_rate": 1.8039167385145955e-06, + "loss": 0.4522, + "step": 32420 + }, + { + "epoch": 11.97, + "learning_rate": 1.8038797881512502e-06, + "loss": 0.2632, + "step": 32421 + }, + { + "epoch": 11.97, + "learning_rate": 1.8038428377879048e-06, + "loss": 0.3083, + "step": 32422 + }, + { + "epoch": 11.97, + "learning_rate": 1.8038058874245597e-06, + "loss": 0.1397, + "step": 32423 + }, + { + "epoch": 11.97, + "learning_rate": 1.8037689370612143e-06, + "loss": 0.3999, + "step": 32424 + }, + { + "epoch": 11.97, + "learning_rate": 1.8037319866978694e-06, + "loss": 0.3622, + "step": 32425 + }, + { + "epoch": 11.97, + "learning_rate": 1.803695036334524e-06, + "loss": 0.3912, + "step": 32426 + }, + { + "epoch": 11.97, + "learning_rate": 1.8036580859711788e-06, + "loss": 0.3602, + "step": 32427 + }, + { + "epoch": 11.97, + "learning_rate": 1.8036211356078335e-06, + "loss": 0.428, + "step": 32428 + }, + { + "epoch": 11.98, + "learning_rate": 1.8035841852444883e-06, + "loss": 0.3291, + "step": 32429 + }, + { + "epoch": 11.98, + "learning_rate": 1.803547234881143e-06, + "loss": 0.3545, + "step": 32430 + }, + { + "epoch": 11.98, + "learning_rate": 1.8035102845177976e-06, + "loss": 0.2204, + "step": 32431 + }, + { + "epoch": 11.98, + "learning_rate": 1.8034733341544527e-06, + "loss": 0.3069, + "step": 32432 + }, + { + "epoch": 11.98, + "learning_rate": 1.8034363837911073e-06, + "loss": 0.3303, + "step": 32433 + }, + { + "epoch": 11.98, + "learning_rate": 1.8033994334277621e-06, + "loss": 0.3281, + "step": 32434 + }, + { + "epoch": 11.98, + "learning_rate": 1.8033624830644168e-06, + "loss": 0.2942, + "step": 32435 + }, + { + "epoch": 11.98, + "learning_rate": 1.8033255327010716e-06, + "loss": 0.2542, + "step": 32436 + }, + { + "epoch": 11.98, + "learning_rate": 1.8032885823377263e-06, + "loss": 0.3706, + "step": 32437 + }, + { + "epoch": 11.98, + "learning_rate": 1.8032516319743813e-06, + "loss": 0.3248, + "step": 32438 + }, + { + "epoch": 11.98, + "learning_rate": 1.803214681611036e-06, + "loss": 0.3188, + "step": 32439 + }, + { + "epoch": 11.98, + "learning_rate": 1.8031777312476908e-06, + "loss": 0.339, + "step": 32440 + }, + { + "epoch": 11.98, + "learning_rate": 1.8031407808843454e-06, + "loss": 0.3688, + "step": 32441 + }, + { + "epoch": 11.98, + "learning_rate": 1.803103830521e-06, + "loss": 0.2261, + "step": 32442 + }, + { + "epoch": 11.98, + "learning_rate": 1.803066880157655e-06, + "loss": 0.2661, + "step": 32443 + }, + { + "epoch": 11.98, + "learning_rate": 1.8030299297943096e-06, + "loss": 0.4177, + "step": 32444 + }, + { + "epoch": 11.98, + "learning_rate": 1.8029929794309646e-06, + "loss": 0.2348, + "step": 32445 + }, + { + "epoch": 11.98, + "learning_rate": 1.802956029067619e-06, + "loss": 0.4038, + "step": 32446 + }, + { + "epoch": 11.98, + "learning_rate": 1.8029190787042741e-06, + "loss": 0.4081, + "step": 32447 + }, + { + "epoch": 11.98, + "learning_rate": 1.8028821283409287e-06, + "loss": 0.2278, + "step": 32448 + }, + { + "epoch": 11.98, + "learning_rate": 1.8028451779775836e-06, + "loss": 0.4195, + "step": 32449 + }, + { + "epoch": 11.98, + "learning_rate": 1.8028082276142382e-06, + "loss": 0.2923, + "step": 32450 + }, + { + "epoch": 11.98, + "learning_rate": 1.8027712772508929e-06, + "loss": 0.2718, + "step": 32451 + }, + { + "epoch": 11.98, + "learning_rate": 1.8027343268875477e-06, + "loss": 0.45, + "step": 32452 + }, + { + "epoch": 11.98, + "learning_rate": 1.8026973765242024e-06, + "loss": 0.2422, + "step": 32453 + }, + { + "epoch": 11.98, + "learning_rate": 1.8026604261608574e-06, + "loss": 0.4029, + "step": 32454 + }, + { + "epoch": 11.98, + "learning_rate": 1.802623475797512e-06, + "loss": 0.4054, + "step": 32455 + }, + { + "epoch": 11.99, + "learning_rate": 1.802586525434167e-06, + "loss": 0.3004, + "step": 32456 + }, + { + "epoch": 11.99, + "learning_rate": 1.8025495750708215e-06, + "loss": 0.235, + "step": 32457 + }, + { + "epoch": 11.99, + "learning_rate": 1.8025126247074764e-06, + "loss": 0.2721, + "step": 32458 + }, + { + "epoch": 11.99, + "learning_rate": 1.802475674344131e-06, + "loss": 0.2036, + "step": 32459 + }, + { + "epoch": 11.99, + "learning_rate": 1.802438723980786e-06, + "loss": 0.5329, + "step": 32460 + }, + { + "epoch": 11.99, + "learning_rate": 1.8024017736174407e-06, + "loss": 0.3684, + "step": 32461 + }, + { + "epoch": 11.99, + "learning_rate": 1.8023648232540951e-06, + "loss": 0.3345, + "step": 32462 + }, + { + "epoch": 11.99, + "learning_rate": 1.8023278728907502e-06, + "loss": 0.4093, + "step": 32463 + }, + { + "epoch": 11.99, + "learning_rate": 1.8022909225274048e-06, + "loss": 0.309, + "step": 32464 + }, + { + "epoch": 11.99, + "learning_rate": 1.8022539721640597e-06, + "loss": 0.3226, + "step": 32465 + }, + { + "epoch": 11.99, + "learning_rate": 1.8022170218007143e-06, + "loss": 0.305, + "step": 32466 + }, + { + "epoch": 11.99, + "learning_rate": 1.8021800714373694e-06, + "loss": 0.2775, + "step": 32467 + }, + { + "epoch": 11.99, + "learning_rate": 1.8021431210740238e-06, + "loss": 0.3692, + "step": 32468 + }, + { + "epoch": 11.99, + "learning_rate": 1.8021061707106789e-06, + "loss": 0.4956, + "step": 32469 + }, + { + "epoch": 11.99, + "learning_rate": 1.8020692203473335e-06, + "loss": 0.2378, + "step": 32470 + }, + { + "epoch": 11.99, + "learning_rate": 1.8020322699839881e-06, + "loss": 0.2565, + "step": 32471 + }, + { + "epoch": 11.99, + "learning_rate": 1.801995319620643e-06, + "loss": 0.3868, + "step": 32472 + }, + { + "epoch": 11.99, + "learning_rate": 1.8019583692572976e-06, + "loss": 0.2094, + "step": 32473 + }, + { + "epoch": 11.99, + "learning_rate": 1.8019214188939525e-06, + "loss": 0.4185, + "step": 32474 + }, + { + "epoch": 11.99, + "learning_rate": 1.8018844685306071e-06, + "loss": 0.2797, + "step": 32475 + }, + { + "epoch": 11.99, + "learning_rate": 1.8018475181672622e-06, + "loss": 0.3752, + "step": 32476 + }, + { + "epoch": 11.99, + "learning_rate": 1.8018105678039168e-06, + "loss": 0.362, + "step": 32477 + }, + { + "epoch": 11.99, + "learning_rate": 1.8017736174405717e-06, + "loss": 0.3085, + "step": 32478 + }, + { + "epoch": 11.99, + "learning_rate": 1.8017366670772263e-06, + "loss": 0.399, + "step": 32479 + }, + { + "epoch": 11.99, + "learning_rate": 1.8016997167138811e-06, + "loss": 0.2364, + "step": 32480 + }, + { + "epoch": 11.99, + "learning_rate": 1.8016627663505358e-06, + "loss": 0.2493, + "step": 32481 + }, + { + "epoch": 11.99, + "learning_rate": 1.8016258159871904e-06, + "loss": 0.3715, + "step": 32482 + }, + { + "epoch": 12.0, + "learning_rate": 1.8015888656238455e-06, + "loss": 0.1725, + "step": 32483 + }, + { + "epoch": 12.0, + "learning_rate": 1.8015519152605001e-06, + "loss": 0.2073, + "step": 32484 + }, + { + "epoch": 12.0, + "learning_rate": 1.801514964897155e-06, + "loss": 0.3752, + "step": 32485 + }, + { + "epoch": 12.0, + "learning_rate": 1.8014780145338096e-06, + "loss": 0.2299, + "step": 32486 + }, + { + "epoch": 12.0, + "learning_rate": 1.8014410641704644e-06, + "loss": 0.2651, + "step": 32487 + }, + { + "epoch": 12.0, + "learning_rate": 1.801404113807119e-06, + "loss": 0.3859, + "step": 32488 + }, + { + "epoch": 12.0, + "learning_rate": 1.8013671634437741e-06, + "loss": 0.2065, + "step": 32489 + }, + { + "epoch": 12.0, + "learning_rate": 1.8013302130804288e-06, + "loss": 0.3197, + "step": 32490 + }, + { + "epoch": 12.0, + "learning_rate": 1.8012932627170832e-06, + "loss": 0.327, + "step": 32491 + }, + { + "epoch": 12.0, + "learning_rate": 1.8012563123537383e-06, + "loss": 0.4483, + "step": 32492 + }, + { + "epoch": 12.0, + "learning_rate": 1.8012193619903929e-06, + "loss": 0.4224, + "step": 32493 + }, + { + "epoch": 12.0, + "learning_rate": 1.8011824116270477e-06, + "loss": 0.2432, + "step": 32494 + }, + { + "epoch": 12.0, + "learning_rate": 1.8011454612637024e-06, + "loss": 0.2114, + "step": 32495 + }, + { + "epoch": 12.0, + "learning_rate": 1.8011085109003572e-06, + "loss": 0.1077, + "step": 32496 + }, + { + "epoch": 12.0, + "eval_accuracy": 0.8277231380154223, + "eval_loss": 0.40464648604393005, + "eval_runtime": 25.2928, + "eval_samples_per_second": 1712.505, + "eval_steps_per_second": 26.767, + "step": 32496 + }, + { + "epoch": 12.0, + "learning_rate": 1.8010715605370119e-06, + "loss": 0.2891, + "step": 32497 + }, + { + "epoch": 12.0, + "learning_rate": 1.801034610173667e-06, + "loss": 0.1871, + "step": 32498 + }, + { + "epoch": 12.0, + "learning_rate": 1.8009976598103216e-06, + "loss": 0.2494, + "step": 32499 + }, + { + "epoch": 12.0, + "learning_rate": 1.8009607094469764e-06, + "loss": 0.3472, + "step": 32500 + }, + { + "epoch": 12.0, + "learning_rate": 1.800923759083631e-06, + "loss": 0.2917, + "step": 32501 + }, + { + "epoch": 12.0, + "learning_rate": 1.8008868087202857e-06, + "loss": 0.3738, + "step": 32502 + }, + { + "epoch": 12.0, + "learning_rate": 1.8008498583569405e-06, + "loss": 0.3234, + "step": 32503 + }, + { + "epoch": 12.0, + "learning_rate": 1.8008129079935952e-06, + "loss": 0.2546, + "step": 32504 + }, + { + "epoch": 12.0, + "learning_rate": 1.8007759576302502e-06, + "loss": 0.3342, + "step": 32505 + }, + { + "epoch": 12.0, + "learning_rate": 1.8007390072669049e-06, + "loss": 0.1839, + "step": 32506 + }, + { + "epoch": 12.0, + "learning_rate": 1.8007020569035597e-06, + "loss": 0.2865, + "step": 32507 + }, + { + "epoch": 12.0, + "learning_rate": 1.8006651065402143e-06, + "loss": 0.3009, + "step": 32508 + }, + { + "epoch": 12.0, + "learning_rate": 1.8006281561768692e-06, + "loss": 0.2928, + "step": 32509 + }, + { + "epoch": 12.01, + "learning_rate": 1.8005912058135238e-06, + "loss": 0.2279, + "step": 32510 + }, + { + "epoch": 12.01, + "learning_rate": 1.8005542554501785e-06, + "loss": 0.3224, + "step": 32511 + }, + { + "epoch": 12.01, + "learning_rate": 1.8005173050868335e-06, + "loss": 0.2457, + "step": 32512 + }, + { + "epoch": 12.01, + "learning_rate": 1.800480354723488e-06, + "loss": 0.3404, + "step": 32513 + }, + { + "epoch": 12.01, + "learning_rate": 1.800443404360143e-06, + "loss": 0.2477, + "step": 32514 + }, + { + "epoch": 12.01, + "learning_rate": 1.8004064539967976e-06, + "loss": 0.2802, + "step": 32515 + }, + { + "epoch": 12.01, + "learning_rate": 1.8003695036334525e-06, + "loss": 0.3129, + "step": 32516 + }, + { + "epoch": 12.01, + "learning_rate": 1.8003325532701071e-06, + "loss": 0.172, + "step": 32517 + }, + { + "epoch": 12.01, + "learning_rate": 1.8002956029067622e-06, + "loss": 0.3898, + "step": 32518 + }, + { + "epoch": 12.01, + "learning_rate": 1.8002586525434166e-06, + "loss": 0.3744, + "step": 32519 + }, + { + "epoch": 12.01, + "learning_rate": 1.8002217021800717e-06, + "loss": 0.3367, + "step": 32520 + }, + { + "epoch": 12.01, + "learning_rate": 1.8001847518167263e-06, + "loss": 0.6065, + "step": 32521 + }, + { + "epoch": 12.01, + "learning_rate": 1.800147801453381e-06, + "loss": 0.2364, + "step": 32522 + }, + { + "epoch": 12.01, + "learning_rate": 1.8001108510900358e-06, + "loss": 0.3158, + "step": 32523 + }, + { + "epoch": 12.01, + "learning_rate": 1.8000739007266904e-06, + "loss": 0.4223, + "step": 32524 + }, + { + "epoch": 12.01, + "learning_rate": 1.8000369503633453e-06, + "loss": 0.3758, + "step": 32525 + }, + { + "epoch": 12.01, + "learning_rate": 1.8e-06, + "loss": 0.3409, + "step": 32526 + }, + { + "epoch": 12.01, + "learning_rate": 1.799963049636655e-06, + "loss": 0.3187, + "step": 32527 + }, + { + "epoch": 12.01, + "learning_rate": 1.7999260992733096e-06, + "loss": 0.3338, + "step": 32528 + }, + { + "epoch": 12.01, + "learning_rate": 1.7998891489099645e-06, + "loss": 0.2784, + "step": 32529 + }, + { + "epoch": 12.01, + "learning_rate": 1.799852198546619e-06, + "loss": 0.2842, + "step": 32530 + }, + { + "epoch": 12.01, + "learning_rate": 1.7998152481832737e-06, + "loss": 0.3199, + "step": 32531 + }, + { + "epoch": 12.01, + "learning_rate": 1.7997782978199286e-06, + "loss": 0.261, + "step": 32532 + }, + { + "epoch": 12.01, + "learning_rate": 1.7997413474565832e-06, + "loss": 0.2614, + "step": 32533 + }, + { + "epoch": 12.01, + "learning_rate": 1.7997043970932383e-06, + "loss": 0.3182, + "step": 32534 + }, + { + "epoch": 12.01, + "learning_rate": 1.7996674467298927e-06, + "loss": 0.3126, + "step": 32535 + }, + { + "epoch": 12.01, + "learning_rate": 1.7996304963665478e-06, + "loss": 0.272, + "step": 32536 + }, + { + "epoch": 12.02, + "learning_rate": 1.7995935460032024e-06, + "loss": 0.3563, + "step": 32537 + }, + { + "epoch": 12.02, + "learning_rate": 1.7995565956398572e-06, + "loss": 0.276, + "step": 32538 + }, + { + "epoch": 12.02, + "learning_rate": 1.7995196452765119e-06, + "loss": 0.2174, + "step": 32539 + }, + { + "epoch": 12.02, + "learning_rate": 1.7994826949131665e-06, + "loss": 0.3791, + "step": 32540 + }, + { + "epoch": 12.02, + "learning_rate": 1.7994457445498214e-06, + "loss": 0.3326, + "step": 32541 + }, + { + "epoch": 12.02, + "learning_rate": 1.799408794186476e-06, + "loss": 0.3326, + "step": 32542 + }, + { + "epoch": 12.02, + "learning_rate": 1.799371843823131e-06, + "loss": 0.2186, + "step": 32543 + }, + { + "epoch": 12.02, + "learning_rate": 1.7993348934597857e-06, + "loss": 0.38, + "step": 32544 + }, + { + "epoch": 12.02, + "learning_rate": 1.7992979430964406e-06, + "loss": 0.3384, + "step": 32545 + }, + { + "epoch": 12.02, + "learning_rate": 1.7992609927330952e-06, + "loss": 0.3553, + "step": 32546 + }, + { + "epoch": 12.02, + "learning_rate": 1.79922404236975e-06, + "loss": 0.2475, + "step": 32547 + }, + { + "epoch": 12.02, + "learning_rate": 1.7991870920064047e-06, + "loss": 0.217, + "step": 32548 + }, + { + "epoch": 12.02, + "learning_rate": 1.7991501416430597e-06, + "loss": 0.4168, + "step": 32549 + }, + { + "epoch": 12.02, + "learning_rate": 1.7991131912797144e-06, + "loss": 0.3661, + "step": 32550 + }, + { + "epoch": 12.02, + "learning_rate": 1.799076240916369e-06, + "loss": 0.1732, + "step": 32551 + }, + { + "epoch": 12.02, + "learning_rate": 1.7990392905530239e-06, + "loss": 0.1973, + "step": 32552 + }, + { + "epoch": 12.02, + "learning_rate": 1.7990023401896785e-06, + "loss": 0.3612, + "step": 32553 + }, + { + "epoch": 12.02, + "learning_rate": 1.7989653898263333e-06, + "loss": 0.3382, + "step": 32554 + }, + { + "epoch": 12.02, + "learning_rate": 1.798928439462988e-06, + "loss": 0.2166, + "step": 32555 + }, + { + "epoch": 12.02, + "learning_rate": 1.798891489099643e-06, + "loss": 0.2783, + "step": 32556 + }, + { + "epoch": 12.02, + "learning_rate": 1.7988545387362977e-06, + "loss": 0.3233, + "step": 32557 + }, + { + "epoch": 12.02, + "learning_rate": 1.7988175883729525e-06, + "loss": 0.3293, + "step": 32558 + }, + { + "epoch": 12.02, + "learning_rate": 1.7987806380096072e-06, + "loss": 0.2847, + "step": 32559 + }, + { + "epoch": 12.02, + "learning_rate": 1.7987436876462618e-06, + "loss": 0.2244, + "step": 32560 + }, + { + "epoch": 12.02, + "learning_rate": 1.7987067372829166e-06, + "loss": 0.2847, + "step": 32561 + }, + { + "epoch": 12.02, + "learning_rate": 1.7986697869195713e-06, + "loss": 0.4057, + "step": 32562 + }, + { + "epoch": 12.02, + "learning_rate": 1.7986328365562263e-06, + "loss": 0.3183, + "step": 32563 + }, + { + "epoch": 12.03, + "learning_rate": 1.7985958861928808e-06, + "loss": 0.2678, + "step": 32564 + }, + { + "epoch": 12.03, + "learning_rate": 1.7985589358295358e-06, + "loss": 0.343, + "step": 32565 + }, + { + "epoch": 12.03, + "learning_rate": 1.7985219854661905e-06, + "loss": 0.2597, + "step": 32566 + }, + { + "epoch": 12.03, + "learning_rate": 1.7984850351028453e-06, + "loss": 0.2683, + "step": 32567 + }, + { + "epoch": 12.03, + "learning_rate": 1.7984480847395e-06, + "loss": 0.3557, + "step": 32568 + }, + { + "epoch": 12.03, + "learning_rate": 1.798411134376155e-06, + "loss": 0.352, + "step": 32569 + }, + { + "epoch": 12.03, + "learning_rate": 1.7983741840128094e-06, + "loss": 0.261, + "step": 32570 + }, + { + "epoch": 12.03, + "learning_rate": 1.798337233649464e-06, + "loss": 0.3995, + "step": 32571 + }, + { + "epoch": 12.03, + "learning_rate": 1.7983002832861191e-06, + "loss": 0.4245, + "step": 32572 + }, + { + "epoch": 12.03, + "learning_rate": 1.7982633329227738e-06, + "loss": 0.2907, + "step": 32573 + }, + { + "epoch": 12.03, + "learning_rate": 1.7982263825594286e-06, + "loss": 0.372, + "step": 32574 + }, + { + "epoch": 12.03, + "learning_rate": 1.7981894321960832e-06, + "loss": 0.256, + "step": 32575 + }, + { + "epoch": 12.03, + "learning_rate": 1.798152481832738e-06, + "loss": 0.1971, + "step": 32576 + }, + { + "epoch": 12.03, + "learning_rate": 1.7981155314693927e-06, + "loss": 0.2633, + "step": 32577 + }, + { + "epoch": 12.03, + "learning_rate": 1.7980785811060478e-06, + "loss": 0.2963, + "step": 32578 + }, + { + "epoch": 12.03, + "learning_rate": 1.7980416307427024e-06, + "loss": 0.3687, + "step": 32579 + }, + { + "epoch": 12.03, + "learning_rate": 1.7980046803793569e-06, + "loss": 0.3854, + "step": 32580 + }, + { + "epoch": 12.03, + "learning_rate": 1.797967730016012e-06, + "loss": 0.2289, + "step": 32581 + }, + { + "epoch": 12.03, + "learning_rate": 1.7979307796526665e-06, + "loss": 0.3409, + "step": 32582 + }, + { + "epoch": 12.03, + "learning_rate": 1.7978938292893214e-06, + "loss": 0.2301, + "step": 32583 + }, + { + "epoch": 12.03, + "learning_rate": 1.797856878925976e-06, + "loss": 0.3204, + "step": 32584 + }, + { + "epoch": 12.03, + "learning_rate": 1.797819928562631e-06, + "loss": 0.2871, + "step": 32585 + }, + { + "epoch": 12.03, + "learning_rate": 1.7977829781992855e-06, + "loss": 0.3079, + "step": 32586 + }, + { + "epoch": 12.03, + "learning_rate": 1.7977460278359406e-06, + "loss": 0.4494, + "step": 32587 + }, + { + "epoch": 12.03, + "learning_rate": 1.7977090774725952e-06, + "loss": 0.3489, + "step": 32588 + }, + { + "epoch": 12.03, + "learning_rate": 1.79767212710925e-06, + "loss": 0.3896, + "step": 32589 + }, + { + "epoch": 12.03, + "learning_rate": 1.7976351767459047e-06, + "loss": 0.4109, + "step": 32590 + }, + { + "epoch": 12.04, + "learning_rate": 1.7975982263825593e-06, + "loss": 0.2138, + "step": 32591 + }, + { + "epoch": 12.04, + "learning_rate": 1.7975612760192142e-06, + "loss": 0.3884, + "step": 32592 + }, + { + "epoch": 12.04, + "learning_rate": 1.7975243256558688e-06, + "loss": 0.2996, + "step": 32593 + }, + { + "epoch": 12.04, + "learning_rate": 1.7974873752925239e-06, + "loss": 0.2556, + "step": 32594 + }, + { + "epoch": 12.04, + "learning_rate": 1.7974504249291785e-06, + "loss": 0.3149, + "step": 32595 + }, + { + "epoch": 12.04, + "learning_rate": 1.7974134745658334e-06, + "loss": 0.3321, + "step": 32596 + }, + { + "epoch": 12.04, + "learning_rate": 1.797376524202488e-06, + "loss": 0.3934, + "step": 32597 + }, + { + "epoch": 12.04, + "learning_rate": 1.7973395738391428e-06, + "loss": 0.3558, + "step": 32598 + }, + { + "epoch": 12.04, + "learning_rate": 1.7973026234757975e-06, + "loss": 0.3363, + "step": 32599 + }, + { + "epoch": 12.04, + "learning_rate": 1.7972656731124521e-06, + "loss": 0.2799, + "step": 32600 + }, + { + "epoch": 12.04, + "learning_rate": 1.7972287227491072e-06, + "loss": 0.2843, + "step": 32601 + }, + { + "epoch": 12.04, + "learning_rate": 1.7971917723857618e-06, + "loss": 0.3157, + "step": 32602 + }, + { + "epoch": 12.04, + "learning_rate": 1.7971548220224167e-06, + "loss": 0.2315, + "step": 32603 + }, + { + "epoch": 12.04, + "learning_rate": 1.7971178716590713e-06, + "loss": 0.3299, + "step": 32604 + }, + { + "epoch": 12.04, + "learning_rate": 1.7970809212957261e-06, + "loss": 0.3131, + "step": 32605 + }, + { + "epoch": 12.04, + "learning_rate": 1.7970439709323808e-06, + "loss": 0.2178, + "step": 32606 + }, + { + "epoch": 12.04, + "learning_rate": 1.7970070205690358e-06, + "loss": 0.4133, + "step": 32607 + }, + { + "epoch": 12.04, + "learning_rate": 1.7969700702056903e-06, + "loss": 0.2381, + "step": 32608 + }, + { + "epoch": 12.04, + "learning_rate": 1.7969331198423453e-06, + "loss": 0.2786, + "step": 32609 + }, + { + "epoch": 12.04, + "learning_rate": 1.796896169479e-06, + "loss": 0.2268, + "step": 32610 + }, + { + "epoch": 12.04, + "learning_rate": 1.7968592191156546e-06, + "loss": 0.3234, + "step": 32611 + }, + { + "epoch": 12.04, + "learning_rate": 1.7968222687523095e-06, + "loss": 0.3793, + "step": 32612 + }, + { + "epoch": 12.04, + "learning_rate": 1.796785318388964e-06, + "loss": 0.2512, + "step": 32613 + }, + { + "epoch": 12.04, + "learning_rate": 1.796748368025619e-06, + "loss": 0.2551, + "step": 32614 + }, + { + "epoch": 12.04, + "learning_rate": 1.7967114176622736e-06, + "loss": 0.3471, + "step": 32615 + }, + { + "epoch": 12.04, + "learning_rate": 1.7966744672989286e-06, + "loss": 0.3338, + "step": 32616 + }, + { + "epoch": 12.04, + "learning_rate": 1.7966375169355833e-06, + "loss": 0.244, + "step": 32617 + }, + { + "epoch": 12.05, + "learning_rate": 1.7966005665722381e-06, + "loss": 0.2767, + "step": 32618 + }, + { + "epoch": 12.05, + "learning_rate": 1.7965636162088928e-06, + "loss": 0.2688, + "step": 32619 + }, + { + "epoch": 12.05, + "learning_rate": 1.7965266658455474e-06, + "loss": 0.3774, + "step": 32620 + }, + { + "epoch": 12.05, + "learning_rate": 1.7964897154822022e-06, + "loss": 0.2226, + "step": 32621 + }, + { + "epoch": 12.05, + "learning_rate": 1.7964527651188569e-06, + "loss": 0.3538, + "step": 32622 + }, + { + "epoch": 12.05, + "learning_rate": 1.796415814755512e-06, + "loss": 0.1843, + "step": 32623 + }, + { + "epoch": 12.05, + "learning_rate": 1.7963788643921666e-06, + "loss": 0.319, + "step": 32624 + }, + { + "epoch": 12.05, + "learning_rate": 1.7963419140288214e-06, + "loss": 0.1622, + "step": 32625 + }, + { + "epoch": 12.05, + "learning_rate": 1.796304963665476e-06, + "loss": 0.1986, + "step": 32626 + }, + { + "epoch": 12.05, + "learning_rate": 1.796268013302131e-06, + "loss": 0.432, + "step": 32627 + }, + { + "epoch": 12.05, + "learning_rate": 1.7962310629387855e-06, + "loss": 0.4053, + "step": 32628 + }, + { + "epoch": 12.05, + "learning_rate": 1.7961941125754406e-06, + "loss": 0.4024, + "step": 32629 + }, + { + "epoch": 12.05, + "learning_rate": 1.7961571622120952e-06, + "loss": 0.301, + "step": 32630 + }, + { + "epoch": 12.05, + "learning_rate": 1.7961202118487497e-06, + "loss": 0.2779, + "step": 32631 + }, + { + "epoch": 12.05, + "learning_rate": 1.7960832614854047e-06, + "loss": 0.3345, + "step": 32632 + }, + { + "epoch": 12.05, + "learning_rate": 1.7960463111220594e-06, + "loss": 0.1947, + "step": 32633 + }, + { + "epoch": 12.05, + "learning_rate": 1.7960093607587142e-06, + "loss": 0.3794, + "step": 32634 + }, + { + "epoch": 12.05, + "learning_rate": 1.7959724103953688e-06, + "loss": 0.3168, + "step": 32635 + }, + { + "epoch": 12.05, + "learning_rate": 1.795935460032024e-06, + "loss": 0.2556, + "step": 32636 + }, + { + "epoch": 12.05, + "learning_rate": 1.7958985096686783e-06, + "loss": 0.1621, + "step": 32637 + }, + { + "epoch": 12.05, + "learning_rate": 1.7958615593053334e-06, + "loss": 0.2454, + "step": 32638 + }, + { + "epoch": 12.05, + "learning_rate": 1.795824608941988e-06, + "loss": 0.2882, + "step": 32639 + }, + { + "epoch": 12.05, + "learning_rate": 1.7957876585786427e-06, + "loss": 0.2516, + "step": 32640 + }, + { + "epoch": 12.05, + "learning_rate": 1.7957507082152975e-06, + "loss": 0.2247, + "step": 32641 + }, + { + "epoch": 12.05, + "learning_rate": 1.7957137578519521e-06, + "loss": 0.1655, + "step": 32642 + }, + { + "epoch": 12.05, + "learning_rate": 1.795676807488607e-06, + "loss": 0.2396, + "step": 32643 + }, + { + "epoch": 12.05, + "learning_rate": 1.7956398571252616e-06, + "loss": 0.1428, + "step": 32644 + }, + { + "epoch": 12.06, + "learning_rate": 1.7956029067619167e-06, + "loss": 0.4348, + "step": 32645 + }, + { + "epoch": 12.06, + "learning_rate": 1.7955659563985713e-06, + "loss": 0.2386, + "step": 32646 + }, + { + "epoch": 12.06, + "learning_rate": 1.7955290060352262e-06, + "loss": 0.4214, + "step": 32647 + }, + { + "epoch": 12.06, + "learning_rate": 1.7954920556718808e-06, + "loss": 0.3965, + "step": 32648 + }, + { + "epoch": 12.06, + "learning_rate": 1.7954551053085357e-06, + "loss": 0.2837, + "step": 32649 + }, + { + "epoch": 12.06, + "learning_rate": 1.7954181549451903e-06, + "loss": 0.3008, + "step": 32650 + }, + { + "epoch": 12.06, + "learning_rate": 1.795381204581845e-06, + "loss": 0.3018, + "step": 32651 + }, + { + "epoch": 12.06, + "learning_rate": 1.7953442542185e-06, + "loss": 0.3343, + "step": 32652 + }, + { + "epoch": 12.06, + "learning_rate": 1.7953073038551544e-06, + "loss": 0.3479, + "step": 32653 + }, + { + "epoch": 12.06, + "learning_rate": 1.7952703534918095e-06, + "loss": 0.3032, + "step": 32654 + }, + { + "epoch": 12.06, + "learning_rate": 1.7952334031284641e-06, + "loss": 0.3152, + "step": 32655 + }, + { + "epoch": 12.06, + "learning_rate": 1.795196452765119e-06, + "loss": 0.4599, + "step": 32656 + }, + { + "epoch": 12.06, + "learning_rate": 1.7951595024017736e-06, + "loss": 0.2706, + "step": 32657 + }, + { + "epoch": 12.06, + "learning_rate": 1.7951225520384287e-06, + "loss": 0.2893, + "step": 32658 + }, + { + "epoch": 12.06, + "learning_rate": 1.795085601675083e-06, + "loss": 0.2897, + "step": 32659 + }, + { + "epoch": 12.06, + "learning_rate": 1.7950486513117377e-06, + "loss": 0.2979, + "step": 32660 + }, + { + "epoch": 12.06, + "learning_rate": 1.7950117009483928e-06, + "loss": 0.2767, + "step": 32661 + }, + { + "epoch": 12.06, + "learning_rate": 1.7949747505850474e-06, + "loss": 0.4192, + "step": 32662 + }, + { + "epoch": 12.06, + "learning_rate": 1.7949378002217023e-06, + "loss": 0.3096, + "step": 32663 + }, + { + "epoch": 12.06, + "learning_rate": 1.794900849858357e-06, + "loss": 0.2703, + "step": 32664 + }, + { + "epoch": 12.06, + "learning_rate": 1.7948638994950117e-06, + "loss": 0.2267, + "step": 32665 + }, + { + "epoch": 12.06, + "learning_rate": 1.7948269491316664e-06, + "loss": 0.3275, + "step": 32666 + }, + { + "epoch": 12.06, + "learning_rate": 1.7947899987683214e-06, + "loss": 0.3496, + "step": 32667 + }, + { + "epoch": 12.06, + "learning_rate": 1.794753048404976e-06, + "loss": 0.2792, + "step": 32668 + }, + { + "epoch": 12.06, + "learning_rate": 1.794716098041631e-06, + "loss": 0.2519, + "step": 32669 + }, + { + "epoch": 12.06, + "learning_rate": 1.7946791476782856e-06, + "loss": 0.3252, + "step": 32670 + }, + { + "epoch": 12.06, + "learning_rate": 1.7946421973149402e-06, + "loss": 0.2069, + "step": 32671 + }, + { + "epoch": 12.06, + "learning_rate": 1.794605246951595e-06, + "loss": 0.4758, + "step": 32672 + }, + { + "epoch": 12.07, + "learning_rate": 1.7945682965882497e-06, + "loss": 0.1461, + "step": 32673 + }, + { + "epoch": 12.07, + "learning_rate": 1.7945313462249047e-06, + "loss": 0.2584, + "step": 32674 + }, + { + "epoch": 12.07, + "learning_rate": 1.7944943958615594e-06, + "loss": 0.2364, + "step": 32675 + }, + { + "epoch": 12.07, + "learning_rate": 1.7944574454982142e-06, + "loss": 0.3039, + "step": 32676 + }, + { + "epoch": 12.07, + "learning_rate": 1.7944204951348689e-06, + "loss": 0.3278, + "step": 32677 + }, + { + "epoch": 12.07, + "learning_rate": 1.7943835447715237e-06, + "loss": 0.316, + "step": 32678 + }, + { + "epoch": 12.07, + "learning_rate": 1.7943465944081784e-06, + "loss": 0.1777, + "step": 32679 + }, + { + "epoch": 12.07, + "learning_rate": 1.794309644044833e-06, + "loss": 0.369, + "step": 32680 + }, + { + "epoch": 12.07, + "learning_rate": 1.7942726936814878e-06, + "loss": 0.2, + "step": 32681 + }, + { + "epoch": 12.07, + "learning_rate": 1.7942357433181425e-06, + "loss": 0.1866, + "step": 32682 + }, + { + "epoch": 12.07, + "learning_rate": 1.7941987929547975e-06, + "loss": 0.3126, + "step": 32683 + }, + { + "epoch": 12.07, + "learning_rate": 1.7941618425914522e-06, + "loss": 0.2884, + "step": 32684 + }, + { + "epoch": 12.07, + "learning_rate": 1.794124892228107e-06, + "loss": 0.2767, + "step": 32685 + }, + { + "epoch": 12.07, + "learning_rate": 1.7940879418647617e-06, + "loss": 0.4258, + "step": 32686 + }, + { + "epoch": 12.07, + "learning_rate": 1.7940509915014165e-06, + "loss": 0.2396, + "step": 32687 + }, + { + "epoch": 12.07, + "learning_rate": 1.7940140411380711e-06, + "loss": 0.3903, + "step": 32688 + }, + { + "epoch": 12.07, + "learning_rate": 1.7939770907747262e-06, + "loss": 0.2539, + "step": 32689 + }, + { + "epoch": 12.07, + "learning_rate": 1.7939401404113808e-06, + "loss": 0.268, + "step": 32690 + }, + { + "epoch": 12.07, + "learning_rate": 1.7939031900480355e-06, + "loss": 0.3783, + "step": 32691 + }, + { + "epoch": 12.07, + "learning_rate": 1.7938662396846903e-06, + "loss": 0.2892, + "step": 32692 + }, + { + "epoch": 12.07, + "learning_rate": 1.793829289321345e-06, + "loss": 0.2715, + "step": 32693 + }, + { + "epoch": 12.07, + "learning_rate": 1.7937923389579998e-06, + "loss": 0.2206, + "step": 32694 + }, + { + "epoch": 12.07, + "learning_rate": 1.7937553885946544e-06, + "loss": 0.273, + "step": 32695 + }, + { + "epoch": 12.07, + "learning_rate": 1.7937184382313095e-06, + "loss": 0.286, + "step": 32696 + }, + { + "epoch": 12.07, + "learning_rate": 1.7936814878679641e-06, + "loss": 0.2569, + "step": 32697 + }, + { + "epoch": 12.07, + "learning_rate": 1.793644537504619e-06, + "loss": 0.167, + "step": 32698 + }, + { + "epoch": 12.07, + "learning_rate": 1.7936075871412736e-06, + "loss": 0.2275, + "step": 32699 + }, + { + "epoch": 12.08, + "learning_rate": 1.7935706367779283e-06, + "loss": 0.3051, + "step": 32700 + }, + { + "epoch": 12.08, + "learning_rate": 1.793533686414583e-06, + "loss": 0.3523, + "step": 32701 + }, + { + "epoch": 12.08, + "learning_rate": 1.7934967360512377e-06, + "loss": 0.3575, + "step": 32702 + }, + { + "epoch": 12.08, + "learning_rate": 1.7934597856878928e-06, + "loss": 0.3275, + "step": 32703 + }, + { + "epoch": 12.08, + "learning_rate": 1.7934228353245472e-06, + "loss": 0.3375, + "step": 32704 + }, + { + "epoch": 12.08, + "learning_rate": 1.7933858849612023e-06, + "loss": 0.3691, + "step": 32705 + }, + { + "epoch": 12.08, + "learning_rate": 1.793348934597857e-06, + "loss": 0.2863, + "step": 32706 + }, + { + "epoch": 12.08, + "learning_rate": 1.7933119842345118e-06, + "loss": 0.335, + "step": 32707 + }, + { + "epoch": 12.08, + "learning_rate": 1.7932750338711664e-06, + "loss": 0.2686, + "step": 32708 + }, + { + "epoch": 12.08, + "learning_rate": 1.7932380835078215e-06, + "loss": 0.291, + "step": 32709 + }, + { + "epoch": 12.08, + "learning_rate": 1.7932011331444759e-06, + "loss": 0.2835, + "step": 32710 + }, + { + "epoch": 12.08, + "learning_rate": 1.7931641827811305e-06, + "loss": 0.3969, + "step": 32711 + }, + { + "epoch": 12.08, + "learning_rate": 1.7931272324177856e-06, + "loss": 0.3557, + "step": 32712 + }, + { + "epoch": 12.08, + "learning_rate": 1.7930902820544402e-06, + "loss": 0.3577, + "step": 32713 + }, + { + "epoch": 12.08, + "learning_rate": 1.793053331691095e-06, + "loss": 0.315, + "step": 32714 + }, + { + "epoch": 12.08, + "learning_rate": 1.7930163813277497e-06, + "loss": 0.3432, + "step": 32715 + }, + { + "epoch": 12.08, + "learning_rate": 1.7929794309644046e-06, + "loss": 0.2065, + "step": 32716 + }, + { + "epoch": 12.08, + "learning_rate": 1.7929424806010592e-06, + "loss": 0.3691, + "step": 32717 + }, + { + "epoch": 12.08, + "learning_rate": 1.7929055302377143e-06, + "loss": 0.3394, + "step": 32718 + }, + { + "epoch": 12.08, + "learning_rate": 1.7928685798743689e-06, + "loss": 0.2911, + "step": 32719 + }, + { + "epoch": 12.08, + "learning_rate": 1.7928316295110233e-06, + "loss": 0.4019, + "step": 32720 + }, + { + "epoch": 12.08, + "learning_rate": 1.7927946791476784e-06, + "loss": 0.2404, + "step": 32721 + }, + { + "epoch": 12.08, + "learning_rate": 1.792757728784333e-06, + "loss": 0.1165, + "step": 32722 + }, + { + "epoch": 12.08, + "learning_rate": 1.7927207784209879e-06, + "loss": 0.3994, + "step": 32723 + }, + { + "epoch": 12.08, + "learning_rate": 1.7926838280576425e-06, + "loss": 0.2365, + "step": 32724 + }, + { + "epoch": 12.08, + "learning_rate": 1.7926468776942976e-06, + "loss": 0.2488, + "step": 32725 + }, + { + "epoch": 12.08, + "learning_rate": 1.792609927330952e-06, + "loss": 0.2603, + "step": 32726 + }, + { + "epoch": 12.09, + "learning_rate": 1.792572976967607e-06, + "loss": 0.2412, + "step": 32727 + }, + { + "epoch": 12.09, + "learning_rate": 1.7925360266042617e-06, + "loss": 0.322, + "step": 32728 + }, + { + "epoch": 12.09, + "learning_rate": 1.7924990762409165e-06, + "loss": 0.3664, + "step": 32729 + }, + { + "epoch": 12.09, + "learning_rate": 1.7924621258775712e-06, + "loss": 0.2793, + "step": 32730 + }, + { + "epoch": 12.09, + "learning_rate": 1.7924251755142258e-06, + "loss": 0.3301, + "step": 32731 + }, + { + "epoch": 12.09, + "learning_rate": 1.7923882251508806e-06, + "loss": 0.2147, + "step": 32732 + }, + { + "epoch": 12.09, + "learning_rate": 1.7923512747875353e-06, + "loss": 0.3441, + "step": 32733 + }, + { + "epoch": 12.09, + "learning_rate": 1.7923143244241903e-06, + "loss": 0.2773, + "step": 32734 + }, + { + "epoch": 12.09, + "learning_rate": 1.792277374060845e-06, + "loss": 0.2226, + "step": 32735 + }, + { + "epoch": 12.09, + "learning_rate": 1.7922404236974998e-06, + "loss": 0.3107, + "step": 32736 + }, + { + "epoch": 12.09, + "learning_rate": 1.7922034733341545e-06, + "loss": 0.3468, + "step": 32737 + }, + { + "epoch": 12.09, + "learning_rate": 1.7921665229708093e-06, + "loss": 0.3316, + "step": 32738 + }, + { + "epoch": 12.09, + "learning_rate": 1.792129572607464e-06, + "loss": 0.3004, + "step": 32739 + }, + { + "epoch": 12.09, + "learning_rate": 1.7920926222441186e-06, + "loss": 0.4011, + "step": 32740 + }, + { + "epoch": 12.09, + "learning_rate": 1.7920556718807736e-06, + "loss": 0.4243, + "step": 32741 + }, + { + "epoch": 12.09, + "learning_rate": 1.7920187215174283e-06, + "loss": 0.3208, + "step": 32742 + }, + { + "epoch": 12.09, + "learning_rate": 1.7919817711540831e-06, + "loss": 0.3921, + "step": 32743 + }, + { + "epoch": 12.09, + "learning_rate": 1.7919448207907378e-06, + "loss": 0.3971, + "step": 32744 + }, + { + "epoch": 12.09, + "learning_rate": 1.7919078704273926e-06, + "loss": 0.3596, + "step": 32745 + }, + { + "epoch": 12.09, + "learning_rate": 1.7918709200640473e-06, + "loss": 0.354, + "step": 32746 + }, + { + "epoch": 12.09, + "learning_rate": 1.7918339697007023e-06, + "loss": 0.2198, + "step": 32747 + }, + { + "epoch": 12.09, + "learning_rate": 1.791797019337357e-06, + "loss": 0.295, + "step": 32748 + }, + { + "epoch": 12.09, + "learning_rate": 1.7917600689740118e-06, + "loss": 0.3125, + "step": 32749 + }, + { + "epoch": 12.09, + "learning_rate": 1.7917231186106664e-06, + "loss": 0.4139, + "step": 32750 + }, + { + "epoch": 12.09, + "learning_rate": 1.791686168247321e-06, + "loss": 0.3568, + "step": 32751 + }, + { + "epoch": 12.09, + "learning_rate": 1.791649217883976e-06, + "loss": 0.3517, + "step": 32752 + }, + { + "epoch": 12.09, + "learning_rate": 1.7916122675206306e-06, + "loss": 0.3747, + "step": 32753 + }, + { + "epoch": 12.1, + "learning_rate": 1.7915753171572856e-06, + "loss": 0.2776, + "step": 32754 + }, + { + "epoch": 12.1, + "learning_rate": 1.79153836679394e-06, + "loss": 0.353, + "step": 32755 + }, + { + "epoch": 12.1, + "learning_rate": 1.791501416430595e-06, + "loss": 0.2922, + "step": 32756 + }, + { + "epoch": 12.1, + "learning_rate": 1.7914644660672497e-06, + "loss": 0.2938, + "step": 32757 + }, + { + "epoch": 12.1, + "learning_rate": 1.7914275157039046e-06, + "loss": 0.3293, + "step": 32758 + }, + { + "epoch": 12.1, + "learning_rate": 1.7913905653405592e-06, + "loss": 0.3024, + "step": 32759 + }, + { + "epoch": 12.1, + "learning_rate": 1.7913536149772139e-06, + "loss": 0.2285, + "step": 32760 + }, + { + "epoch": 12.1, + "learning_rate": 1.7913166646138687e-06, + "loss": 0.1675, + "step": 32761 + }, + { + "epoch": 12.1, + "learning_rate": 1.7912797142505233e-06, + "loss": 0.2566, + "step": 32762 + }, + { + "epoch": 12.1, + "learning_rate": 1.7912427638871784e-06, + "loss": 0.4554, + "step": 32763 + }, + { + "epoch": 12.1, + "learning_rate": 1.791205813523833e-06, + "loss": 0.3141, + "step": 32764 + }, + { + "epoch": 12.1, + "learning_rate": 1.7911688631604879e-06, + "loss": 0.3447, + "step": 32765 + }, + { + "epoch": 12.1, + "learning_rate": 1.7911319127971425e-06, + "loss": 0.3007, + "step": 32766 + }, + { + "epoch": 12.1, + "learning_rate": 1.7910949624337974e-06, + "loss": 0.3225, + "step": 32767 + }, + { + "epoch": 12.1, + "learning_rate": 1.791058012070452e-06, + "loss": 0.2554, + "step": 32768 + }, + { + "epoch": 12.1, + "learning_rate": 1.7910210617071066e-06, + "loss": 0.255, + "step": 32769 + }, + { + "epoch": 12.1, + "learning_rate": 1.7909841113437617e-06, + "loss": 0.2494, + "step": 32770 + }, + { + "epoch": 12.1, + "learning_rate": 1.7909471609804161e-06, + "loss": 0.2641, + "step": 32771 + }, + { + "epoch": 12.1, + "learning_rate": 1.7909102106170712e-06, + "loss": 0.2945, + "step": 32772 + }, + { + "epoch": 12.1, + "learning_rate": 1.7908732602537258e-06, + "loss": 0.1818, + "step": 32773 + }, + { + "epoch": 12.1, + "learning_rate": 1.7908363098903807e-06, + "loss": 0.3627, + "step": 32774 + }, + { + "epoch": 12.1, + "learning_rate": 1.7907993595270353e-06, + "loss": 0.2326, + "step": 32775 + }, + { + "epoch": 12.1, + "learning_rate": 1.7907624091636904e-06, + "loss": 0.4658, + "step": 32776 + }, + { + "epoch": 12.1, + "learning_rate": 1.7907254588003448e-06, + "loss": 0.5164, + "step": 32777 + }, + { + "epoch": 12.1, + "learning_rate": 1.7906885084369999e-06, + "loss": 0.398, + "step": 32778 + }, + { + "epoch": 12.1, + "learning_rate": 1.7906515580736545e-06, + "loss": 0.326, + "step": 32779 + }, + { + "epoch": 12.1, + "learning_rate": 1.7906146077103091e-06, + "loss": 0.2053, + "step": 32780 + }, + { + "epoch": 12.11, + "learning_rate": 1.790577657346964e-06, + "loss": 0.4558, + "step": 32781 + }, + { + "epoch": 12.11, + "learning_rate": 1.7905407069836186e-06, + "loss": 0.2801, + "step": 32782 + }, + { + "epoch": 12.11, + "learning_rate": 1.7905037566202735e-06, + "loss": 0.3848, + "step": 32783 + }, + { + "epoch": 12.11, + "learning_rate": 1.790466806256928e-06, + "loss": 0.3555, + "step": 32784 + }, + { + "epoch": 12.11, + "learning_rate": 1.7904298558935832e-06, + "loss": 0.3954, + "step": 32785 + }, + { + "epoch": 12.11, + "learning_rate": 1.7903929055302378e-06, + "loss": 0.3434, + "step": 32786 + }, + { + "epoch": 12.11, + "learning_rate": 1.7903559551668926e-06, + "loss": 0.2507, + "step": 32787 + }, + { + "epoch": 12.11, + "learning_rate": 1.7903190048035473e-06, + "loss": 0.2265, + "step": 32788 + }, + { + "epoch": 12.11, + "learning_rate": 1.790282054440202e-06, + "loss": 0.2333, + "step": 32789 + }, + { + "epoch": 12.11, + "learning_rate": 1.7902451040768568e-06, + "loss": 0.4013, + "step": 32790 + }, + { + "epoch": 12.11, + "learning_rate": 1.7902081537135114e-06, + "loss": 0.3852, + "step": 32791 + }, + { + "epoch": 12.11, + "learning_rate": 1.7901712033501665e-06, + "loss": 0.278, + "step": 32792 + }, + { + "epoch": 12.11, + "learning_rate": 1.7901342529868209e-06, + "loss": 0.2357, + "step": 32793 + }, + { + "epoch": 12.11, + "learning_rate": 1.790097302623476e-06, + "loss": 0.2871, + "step": 32794 + }, + { + "epoch": 12.11, + "learning_rate": 1.7900603522601306e-06, + "loss": 0.255, + "step": 32795 + }, + { + "epoch": 12.11, + "learning_rate": 1.7900234018967854e-06, + "loss": 0.3098, + "step": 32796 + }, + { + "epoch": 12.11, + "learning_rate": 1.78998645153344e-06, + "loss": 0.2574, + "step": 32797 + }, + { + "epoch": 12.11, + "learning_rate": 1.7899495011700951e-06, + "loss": 0.4285, + "step": 32798 + }, + { + "epoch": 12.11, + "learning_rate": 1.7899125508067495e-06, + "loss": 0.2308, + "step": 32799 + }, + { + "epoch": 12.11, + "learning_rate": 1.7898756004434042e-06, + "loss": 0.2385, + "step": 32800 + }, + { + "epoch": 12.11, + "learning_rate": 1.7898386500800592e-06, + "loss": 0.3067, + "step": 32801 + }, + { + "epoch": 12.11, + "learning_rate": 1.7898016997167139e-06, + "loss": 0.2215, + "step": 32802 + }, + { + "epoch": 12.11, + "learning_rate": 1.7897647493533687e-06, + "loss": 0.2635, + "step": 32803 + }, + { + "epoch": 12.11, + "learning_rate": 1.7897277989900234e-06, + "loss": 0.3679, + "step": 32804 + }, + { + "epoch": 12.11, + "learning_rate": 1.7896908486266782e-06, + "loss": 0.3316, + "step": 32805 + }, + { + "epoch": 12.11, + "learning_rate": 1.7896538982633328e-06, + "loss": 0.3631, + "step": 32806 + }, + { + "epoch": 12.11, + "learning_rate": 1.789616947899988e-06, + "loss": 0.3999, + "step": 32807 + }, + { + "epoch": 12.12, + "learning_rate": 1.7895799975366425e-06, + "loss": 0.2809, + "step": 32808 + }, + { + "epoch": 12.12, + "learning_rate": 1.7895430471732972e-06, + "loss": 0.2748, + "step": 32809 + }, + { + "epoch": 12.12, + "learning_rate": 1.789506096809952e-06, + "loss": 0.2492, + "step": 32810 + }, + { + "epoch": 12.12, + "learning_rate": 1.7894691464466067e-06, + "loss": 0.3283, + "step": 32811 + }, + { + "epoch": 12.12, + "learning_rate": 1.7894321960832615e-06, + "loss": 0.3578, + "step": 32812 + }, + { + "epoch": 12.12, + "learning_rate": 1.7893952457199162e-06, + "loss": 0.2659, + "step": 32813 + }, + { + "epoch": 12.12, + "learning_rate": 1.7893582953565712e-06, + "loss": 0.2205, + "step": 32814 + }, + { + "epoch": 12.12, + "learning_rate": 1.7893213449932258e-06, + "loss": 0.4042, + "step": 32815 + }, + { + "epoch": 12.12, + "learning_rate": 1.7892843946298807e-06, + "loss": 0.2246, + "step": 32816 + }, + { + "epoch": 12.12, + "learning_rate": 1.7892474442665353e-06, + "loss": 0.2057, + "step": 32817 + }, + { + "epoch": 12.12, + "learning_rate": 1.7892104939031902e-06, + "loss": 0.1846, + "step": 32818 + }, + { + "epoch": 12.12, + "learning_rate": 1.7891735435398448e-06, + "loss": 0.3282, + "step": 32819 + }, + { + "epoch": 12.12, + "learning_rate": 1.7891365931764995e-06, + "loss": 0.3058, + "step": 32820 + }, + { + "epoch": 12.12, + "learning_rate": 1.7890996428131545e-06, + "loss": 0.2734, + "step": 32821 + }, + { + "epoch": 12.12, + "learning_rate": 1.789062692449809e-06, + "loss": 0.3656, + "step": 32822 + }, + { + "epoch": 12.12, + "learning_rate": 1.789025742086464e-06, + "loss": 0.238, + "step": 32823 + }, + { + "epoch": 12.12, + "learning_rate": 1.7889887917231186e-06, + "loss": 0.3871, + "step": 32824 + }, + { + "epoch": 12.12, + "learning_rate": 1.7889518413597735e-06, + "loss": 0.2674, + "step": 32825 + }, + { + "epoch": 12.12, + "learning_rate": 1.7889148909964281e-06, + "loss": 0.3178, + "step": 32826 + }, + { + "epoch": 12.12, + "learning_rate": 1.7888779406330832e-06, + "loss": 0.3407, + "step": 32827 + }, + { + "epoch": 12.12, + "learning_rate": 1.7888409902697376e-06, + "loss": 0.4297, + "step": 32828 + }, + { + "epoch": 12.12, + "learning_rate": 1.7888040399063922e-06, + "loss": 0.3287, + "step": 32829 + }, + { + "epoch": 12.12, + "learning_rate": 1.7887670895430473e-06, + "loss": 0.2753, + "step": 32830 + }, + { + "epoch": 12.12, + "learning_rate": 1.788730139179702e-06, + "loss": 0.3077, + "step": 32831 + }, + { + "epoch": 12.12, + "learning_rate": 1.7886931888163568e-06, + "loss": 0.323, + "step": 32832 + }, + { + "epoch": 12.12, + "learning_rate": 1.7886562384530114e-06, + "loss": 0.2816, + "step": 32833 + }, + { + "epoch": 12.12, + "learning_rate": 1.7886192880896663e-06, + "loss": 0.229, + "step": 32834 + }, + { + "epoch": 12.13, + "learning_rate": 1.788582337726321e-06, + "loss": 0.3038, + "step": 32835 + }, + { + "epoch": 12.13, + "learning_rate": 1.788545387362976e-06, + "loss": 0.318, + "step": 32836 + }, + { + "epoch": 12.13, + "learning_rate": 1.7885084369996306e-06, + "loss": 0.3684, + "step": 32837 + }, + { + "epoch": 12.13, + "learning_rate": 1.7884714866362854e-06, + "loss": 0.2986, + "step": 32838 + }, + { + "epoch": 12.13, + "learning_rate": 1.78843453627294e-06, + "loss": 0.2798, + "step": 32839 + }, + { + "epoch": 12.13, + "learning_rate": 1.7883975859095947e-06, + "loss": 0.292, + "step": 32840 + }, + { + "epoch": 12.13, + "learning_rate": 1.7883606355462496e-06, + "loss": 0.3132, + "step": 32841 + }, + { + "epoch": 12.13, + "learning_rate": 1.7883236851829042e-06, + "loss": 0.3039, + "step": 32842 + }, + { + "epoch": 12.13, + "learning_rate": 1.7882867348195593e-06, + "loss": 0.2655, + "step": 32843 + }, + { + "epoch": 12.13, + "learning_rate": 1.7882497844562137e-06, + "loss": 0.3564, + "step": 32844 + }, + { + "epoch": 12.13, + "learning_rate": 1.7882128340928688e-06, + "loss": 0.2276, + "step": 32845 + }, + { + "epoch": 12.13, + "learning_rate": 1.7881758837295234e-06, + "loss": 0.3213, + "step": 32846 + }, + { + "epoch": 12.13, + "learning_rate": 1.7881389333661782e-06, + "loss": 0.3214, + "step": 32847 + }, + { + "epoch": 12.13, + "learning_rate": 1.7881019830028329e-06, + "loss": 0.1797, + "step": 32848 + }, + { + "epoch": 12.13, + "learning_rate": 1.7880650326394875e-06, + "loss": 0.3536, + "step": 32849 + }, + { + "epoch": 12.13, + "learning_rate": 1.7880280822761424e-06, + "loss": 0.228, + "step": 32850 + }, + { + "epoch": 12.13, + "learning_rate": 1.787991131912797e-06, + "loss": 0.1469, + "step": 32851 + }, + { + "epoch": 12.13, + "learning_rate": 1.787954181549452e-06, + "loss": 0.2147, + "step": 32852 + }, + { + "epoch": 12.13, + "learning_rate": 1.7879172311861067e-06, + "loss": 0.2429, + "step": 32853 + }, + { + "epoch": 12.13, + "learning_rate": 1.7878802808227615e-06, + "loss": 0.3049, + "step": 32854 + }, + { + "epoch": 12.13, + "learning_rate": 1.7878433304594162e-06, + "loss": 0.2102, + "step": 32855 + }, + { + "epoch": 12.13, + "learning_rate": 1.787806380096071e-06, + "loss": 0.2633, + "step": 32856 + }, + { + "epoch": 12.13, + "learning_rate": 1.7877694297327257e-06, + "loss": 0.3474, + "step": 32857 + }, + { + "epoch": 12.13, + "learning_rate": 1.7877324793693807e-06, + "loss": 0.2716, + "step": 32858 + }, + { + "epoch": 12.13, + "learning_rate": 1.7876955290060354e-06, + "loss": 0.2392, + "step": 32859 + }, + { + "epoch": 12.13, + "learning_rate": 1.78765857864269e-06, + "loss": 0.4293, + "step": 32860 + }, + { + "epoch": 12.13, + "learning_rate": 1.7876216282793448e-06, + "loss": 0.4144, + "step": 32861 + }, + { + "epoch": 12.14, + "learning_rate": 1.7875846779159995e-06, + "loss": 0.4281, + "step": 32862 + }, + { + "epoch": 12.14, + "learning_rate": 1.7875477275526543e-06, + "loss": 0.1594, + "step": 32863 + }, + { + "epoch": 12.14, + "learning_rate": 1.787510777189309e-06, + "loss": 0.4037, + "step": 32864 + }, + { + "epoch": 12.14, + "learning_rate": 1.787473826825964e-06, + "loss": 0.2908, + "step": 32865 + }, + { + "epoch": 12.14, + "learning_rate": 1.7874368764626187e-06, + "loss": 0.2759, + "step": 32866 + }, + { + "epoch": 12.14, + "learning_rate": 1.7873999260992735e-06, + "loss": 0.3924, + "step": 32867 + }, + { + "epoch": 12.14, + "learning_rate": 1.7873629757359281e-06, + "loss": 0.3717, + "step": 32868 + }, + { + "epoch": 12.14, + "learning_rate": 1.7873260253725828e-06, + "loss": 0.3506, + "step": 32869 + }, + { + "epoch": 12.14, + "learning_rate": 1.7872890750092376e-06, + "loss": 0.3776, + "step": 32870 + }, + { + "epoch": 12.14, + "learning_rate": 1.7872521246458923e-06, + "loss": 0.3432, + "step": 32871 + }, + { + "epoch": 12.14, + "learning_rate": 1.7872151742825471e-06, + "loss": 0.2659, + "step": 32872 + }, + { + "epoch": 12.14, + "learning_rate": 1.7871782239192017e-06, + "loss": 0.2694, + "step": 32873 + }, + { + "epoch": 12.14, + "learning_rate": 1.7871412735558568e-06, + "loss": 0.2575, + "step": 32874 + }, + { + "epoch": 12.14, + "learning_rate": 1.7871043231925114e-06, + "loss": 0.3173, + "step": 32875 + }, + { + "epoch": 12.14, + "learning_rate": 1.7870673728291663e-06, + "loss": 0.25, + "step": 32876 + }, + { + "epoch": 12.14, + "learning_rate": 1.787030422465821e-06, + "loss": 0.303, + "step": 32877 + }, + { + "epoch": 12.14, + "learning_rate": 1.7869934721024758e-06, + "loss": 0.4291, + "step": 32878 + }, + { + "epoch": 12.14, + "learning_rate": 1.7869565217391304e-06, + "loss": 0.4367, + "step": 32879 + }, + { + "epoch": 12.14, + "learning_rate": 1.786919571375785e-06, + "loss": 0.2675, + "step": 32880 + }, + { + "epoch": 12.14, + "learning_rate": 1.7868826210124401e-06, + "loss": 0.3325, + "step": 32881 + }, + { + "epoch": 12.14, + "learning_rate": 1.7868456706490947e-06, + "loss": 0.3256, + "step": 32882 + }, + { + "epoch": 12.14, + "learning_rate": 1.7868087202857496e-06, + "loss": 0.2818, + "step": 32883 + }, + { + "epoch": 12.14, + "learning_rate": 1.7867717699224042e-06, + "loss": 0.3362, + "step": 32884 + }, + { + "epoch": 12.14, + "learning_rate": 1.786734819559059e-06, + "loss": 0.2901, + "step": 32885 + }, + { + "epoch": 12.14, + "learning_rate": 1.7866978691957137e-06, + "loss": 0.3963, + "step": 32886 + }, + { + "epoch": 12.14, + "learning_rate": 1.7866609188323688e-06, + "loss": 0.3726, + "step": 32887 + }, + { + "epoch": 12.14, + "learning_rate": 1.7866239684690234e-06, + "loss": 0.5117, + "step": 32888 + }, + { + "epoch": 12.15, + "learning_rate": 1.7865870181056778e-06, + "loss": 0.3099, + "step": 32889 + }, + { + "epoch": 12.15, + "learning_rate": 1.786550067742333e-06, + "loss": 0.3358, + "step": 32890 + }, + { + "epoch": 12.15, + "learning_rate": 1.7865131173789875e-06, + "loss": 0.3499, + "step": 32891 + }, + { + "epoch": 12.15, + "learning_rate": 1.7864761670156424e-06, + "loss": 0.2336, + "step": 32892 + }, + { + "epoch": 12.15, + "learning_rate": 1.786439216652297e-06, + "loss": 0.3976, + "step": 32893 + }, + { + "epoch": 12.15, + "learning_rate": 1.786402266288952e-06, + "loss": 0.2403, + "step": 32894 + }, + { + "epoch": 12.15, + "learning_rate": 1.7863653159256065e-06, + "loss": 0.3813, + "step": 32895 + }, + { + "epoch": 12.15, + "learning_rate": 1.7863283655622616e-06, + "loss": 0.2261, + "step": 32896 + }, + { + "epoch": 12.15, + "learning_rate": 1.7862914151989162e-06, + "loss": 0.3999, + "step": 32897 + }, + { + "epoch": 12.15, + "learning_rate": 1.786254464835571e-06, + "loss": 0.2086, + "step": 32898 + }, + { + "epoch": 12.15, + "learning_rate": 1.7862175144722257e-06, + "loss": 0.2497, + "step": 32899 + }, + { + "epoch": 12.15, + "learning_rate": 1.7861805641088803e-06, + "loss": 0.2692, + "step": 32900 + }, + { + "epoch": 12.15, + "learning_rate": 1.7861436137455352e-06, + "loss": 0.2683, + "step": 32901 + }, + { + "epoch": 12.15, + "learning_rate": 1.7861066633821898e-06, + "loss": 0.467, + "step": 32902 + }, + { + "epoch": 12.15, + "learning_rate": 1.7860697130188449e-06, + "loss": 0.3664, + "step": 32903 + }, + { + "epoch": 12.15, + "learning_rate": 1.7860327626554995e-06, + "loss": 0.4285, + "step": 32904 + }, + { + "epoch": 12.15, + "learning_rate": 1.7859958122921543e-06, + "loss": 0.3502, + "step": 32905 + }, + { + "epoch": 12.15, + "learning_rate": 1.785958861928809e-06, + "loss": 0.2722, + "step": 32906 + }, + { + "epoch": 12.15, + "learning_rate": 1.7859219115654638e-06, + "loss": 0.3114, + "step": 32907 + }, + { + "epoch": 12.15, + "learning_rate": 1.7858849612021185e-06, + "loss": 0.249, + "step": 32908 + }, + { + "epoch": 12.15, + "learning_rate": 1.7858480108387731e-06, + "loss": 0.3554, + "step": 32909 + }, + { + "epoch": 12.15, + "learning_rate": 1.7858110604754282e-06, + "loss": 0.3558, + "step": 32910 + }, + { + "epoch": 12.15, + "learning_rate": 1.7857741101120826e-06, + "loss": 0.227, + "step": 32911 + }, + { + "epoch": 12.15, + "learning_rate": 1.7857371597487377e-06, + "loss": 0.2808, + "step": 32912 + }, + { + "epoch": 12.15, + "learning_rate": 1.7857002093853923e-06, + "loss": 0.361, + "step": 32913 + }, + { + "epoch": 12.15, + "learning_rate": 1.7856632590220471e-06, + "loss": 0.3848, + "step": 32914 + }, + { + "epoch": 12.15, + "learning_rate": 1.7856263086587018e-06, + "loss": 0.3332, + "step": 32915 + }, + { + "epoch": 12.16, + "learning_rate": 1.7855893582953568e-06, + "loss": 0.283, + "step": 32916 + }, + { + "epoch": 12.16, + "learning_rate": 1.7855524079320113e-06, + "loss": 0.3393, + "step": 32917 + }, + { + "epoch": 12.16, + "learning_rate": 1.7855154575686663e-06, + "loss": 0.3491, + "step": 32918 + }, + { + "epoch": 12.16, + "learning_rate": 1.785478507205321e-06, + "loss": 0.268, + "step": 32919 + }, + { + "epoch": 12.16, + "learning_rate": 1.7854415568419756e-06, + "loss": 0.3463, + "step": 32920 + }, + { + "epoch": 12.16, + "learning_rate": 1.7854046064786304e-06, + "loss": 0.2596, + "step": 32921 + }, + { + "epoch": 12.16, + "learning_rate": 1.785367656115285e-06, + "loss": 0.3166, + "step": 32922 + }, + { + "epoch": 12.16, + "learning_rate": 1.78533070575194e-06, + "loss": 0.3075, + "step": 32923 + }, + { + "epoch": 12.16, + "learning_rate": 1.7852937553885946e-06, + "loss": 0.2506, + "step": 32924 + }, + { + "epoch": 12.16, + "learning_rate": 1.7852568050252496e-06, + "loss": 0.3105, + "step": 32925 + }, + { + "epoch": 12.16, + "learning_rate": 1.7852198546619043e-06, + "loss": 0.2754, + "step": 32926 + }, + { + "epoch": 12.16, + "learning_rate": 1.785182904298559e-06, + "loss": 0.2834, + "step": 32927 + }, + { + "epoch": 12.16, + "learning_rate": 1.7851459539352137e-06, + "loss": 0.2776, + "step": 32928 + }, + { + "epoch": 12.16, + "learning_rate": 1.7851090035718684e-06, + "loss": 0.3302, + "step": 32929 + }, + { + "epoch": 12.16, + "learning_rate": 1.7850720532085232e-06, + "loss": 0.2355, + "step": 32930 + }, + { + "epoch": 12.16, + "learning_rate": 1.7850351028451779e-06, + "loss": 0.2109, + "step": 32931 + }, + { + "epoch": 12.16, + "learning_rate": 1.784998152481833e-06, + "loss": 0.3266, + "step": 32932 + }, + { + "epoch": 12.16, + "learning_rate": 1.7849612021184876e-06, + "loss": 0.2691, + "step": 32933 + }, + { + "epoch": 12.16, + "learning_rate": 1.7849242517551424e-06, + "loss": 0.2472, + "step": 32934 + }, + { + "epoch": 12.16, + "learning_rate": 1.784887301391797e-06, + "loss": 0.3489, + "step": 32935 + }, + { + "epoch": 12.16, + "learning_rate": 1.7848503510284519e-06, + "loss": 0.3223, + "step": 32936 + }, + { + "epoch": 12.16, + "learning_rate": 1.7848134006651065e-06, + "loss": 0.1907, + "step": 32937 + }, + { + "epoch": 12.16, + "learning_rate": 1.7847764503017616e-06, + "loss": 0.3423, + "step": 32938 + }, + { + "epoch": 12.16, + "learning_rate": 1.7847394999384162e-06, + "loss": 0.2158, + "step": 32939 + }, + { + "epoch": 12.16, + "learning_rate": 1.7847025495750706e-06, + "loss": 0.3135, + "step": 32940 + }, + { + "epoch": 12.16, + "learning_rate": 1.7846655992117257e-06, + "loss": 0.443, + "step": 32941 + }, + { + "epoch": 12.16, + "learning_rate": 1.7846286488483803e-06, + "loss": 0.269, + "step": 32942 + }, + { + "epoch": 12.17, + "learning_rate": 1.7845916984850352e-06, + "loss": 0.3284, + "step": 32943 + }, + { + "epoch": 12.17, + "learning_rate": 1.7845547481216898e-06, + "loss": 0.2598, + "step": 32944 + }, + { + "epoch": 12.17, + "learning_rate": 1.7845177977583447e-06, + "loss": 0.4041, + "step": 32945 + }, + { + "epoch": 12.17, + "learning_rate": 1.7844808473949993e-06, + "loss": 0.2473, + "step": 32946 + }, + { + "epoch": 12.17, + "learning_rate": 1.7844438970316544e-06, + "loss": 0.3266, + "step": 32947 + }, + { + "epoch": 12.17, + "learning_rate": 1.784406946668309e-06, + "loss": 0.3383, + "step": 32948 + }, + { + "epoch": 12.17, + "learning_rate": 1.7843699963049636e-06, + "loss": 0.345, + "step": 32949 + }, + { + "epoch": 12.17, + "learning_rate": 1.7843330459416185e-06, + "loss": 0.407, + "step": 32950 + }, + { + "epoch": 12.17, + "learning_rate": 1.7842960955782731e-06, + "loss": 0.3208, + "step": 32951 + }, + { + "epoch": 12.17, + "learning_rate": 1.784259145214928e-06, + "loss": 0.3359, + "step": 32952 + }, + { + "epoch": 12.17, + "learning_rate": 1.7842221948515826e-06, + "loss": 0.159, + "step": 32953 + }, + { + "epoch": 12.17, + "learning_rate": 1.7841852444882377e-06, + "loss": 0.3105, + "step": 32954 + }, + { + "epoch": 12.17, + "learning_rate": 1.7841482941248923e-06, + "loss": 0.3245, + "step": 32955 + }, + { + "epoch": 12.17, + "learning_rate": 1.7841113437615472e-06, + "loss": 0.3391, + "step": 32956 + }, + { + "epoch": 12.17, + "learning_rate": 1.7840743933982018e-06, + "loss": 0.2117, + "step": 32957 + }, + { + "epoch": 12.17, + "learning_rate": 1.7840374430348566e-06, + "loss": 0.3686, + "step": 32958 + }, + { + "epoch": 12.17, + "learning_rate": 1.7840004926715113e-06, + "loss": 0.3807, + "step": 32959 + }, + { + "epoch": 12.17, + "learning_rate": 1.783963542308166e-06, + "loss": 0.1926, + "step": 32960 + }, + { + "epoch": 12.17, + "learning_rate": 1.783926591944821e-06, + "loss": 0.3708, + "step": 32961 + }, + { + "epoch": 12.17, + "learning_rate": 1.7838896415814754e-06, + "loss": 0.273, + "step": 32962 + }, + { + "epoch": 12.17, + "learning_rate": 1.7838526912181305e-06, + "loss": 0.3141, + "step": 32963 + }, + { + "epoch": 12.17, + "learning_rate": 1.783815740854785e-06, + "loss": 0.2859, + "step": 32964 + }, + { + "epoch": 12.17, + "learning_rate": 1.78377879049144e-06, + "loss": 0.2661, + "step": 32965 + }, + { + "epoch": 12.17, + "learning_rate": 1.7837418401280946e-06, + "loss": 0.2333, + "step": 32966 + }, + { + "epoch": 12.17, + "learning_rate": 1.7837048897647496e-06, + "loss": 0.3614, + "step": 32967 + }, + { + "epoch": 12.17, + "learning_rate": 1.783667939401404e-06, + "loss": 0.3449, + "step": 32968 + }, + { + "epoch": 12.17, + "learning_rate": 1.7836309890380587e-06, + "loss": 0.3135, + "step": 32969 + }, + { + "epoch": 12.18, + "learning_rate": 1.7835940386747138e-06, + "loss": 0.257, + "step": 32970 + }, + { + "epoch": 12.18, + "learning_rate": 1.7835570883113684e-06, + "loss": 0.2641, + "step": 32971 + }, + { + "epoch": 12.18, + "learning_rate": 1.7835201379480232e-06, + "loss": 0.3699, + "step": 32972 + }, + { + "epoch": 12.18, + "learning_rate": 1.7834831875846779e-06, + "loss": 0.2329, + "step": 32973 + }, + { + "epoch": 12.18, + "learning_rate": 1.7834462372213327e-06, + "loss": 0.3452, + "step": 32974 + }, + { + "epoch": 12.18, + "learning_rate": 1.7834092868579874e-06, + "loss": 0.3662, + "step": 32975 + }, + { + "epoch": 12.18, + "learning_rate": 1.7833723364946424e-06, + "loss": 0.3416, + "step": 32976 + }, + { + "epoch": 12.18, + "learning_rate": 1.783335386131297e-06, + "loss": 0.2479, + "step": 32977 + }, + { + "epoch": 12.18, + "learning_rate": 1.783298435767952e-06, + "loss": 0.3571, + "step": 32978 + }, + { + "epoch": 12.18, + "learning_rate": 1.7832614854046066e-06, + "loss": 0.3108, + "step": 32979 + }, + { + "epoch": 12.18, + "learning_rate": 1.7832245350412612e-06, + "loss": 0.2222, + "step": 32980 + }, + { + "epoch": 12.18, + "learning_rate": 1.783187584677916e-06, + "loss": 0.3889, + "step": 32981 + }, + { + "epoch": 12.18, + "learning_rate": 1.7831506343145707e-06, + "loss": 0.3282, + "step": 32982 + }, + { + "epoch": 12.18, + "learning_rate": 1.7831136839512257e-06, + "loss": 0.3416, + "step": 32983 + }, + { + "epoch": 12.18, + "learning_rate": 1.7830767335878802e-06, + "loss": 0.3643, + "step": 32984 + }, + { + "epoch": 12.18, + "learning_rate": 1.7830397832245352e-06, + "loss": 0.2376, + "step": 32985 + }, + { + "epoch": 12.18, + "learning_rate": 1.7830028328611899e-06, + "loss": 0.2673, + "step": 32986 + }, + { + "epoch": 12.18, + "learning_rate": 1.7829658824978447e-06, + "loss": 0.2424, + "step": 32987 + }, + { + "epoch": 12.18, + "learning_rate": 1.7829289321344993e-06, + "loss": 0.4553, + "step": 32988 + }, + { + "epoch": 12.18, + "learning_rate": 1.782891981771154e-06, + "loss": 0.3459, + "step": 32989 + }, + { + "epoch": 12.18, + "learning_rate": 1.7828550314078088e-06, + "loss": 0.2588, + "step": 32990 + }, + { + "epoch": 12.18, + "learning_rate": 1.7828180810444635e-06, + "loss": 0.3238, + "step": 32991 + }, + { + "epoch": 12.18, + "learning_rate": 1.7827811306811185e-06, + "loss": 0.3127, + "step": 32992 + }, + { + "epoch": 12.18, + "learning_rate": 1.7827441803177732e-06, + "loss": 0.4666, + "step": 32993 + }, + { + "epoch": 12.18, + "learning_rate": 1.782707229954428e-06, + "loss": 0.2888, + "step": 32994 + }, + { + "epoch": 12.18, + "learning_rate": 1.7826702795910826e-06, + "loss": 0.2489, + "step": 32995 + }, + { + "epoch": 12.18, + "learning_rate": 1.7826333292277375e-06, + "loss": 0.3081, + "step": 32996 + }, + { + "epoch": 12.19, + "learning_rate": 1.7825963788643921e-06, + "loss": 0.2005, + "step": 32997 + }, + { + "epoch": 12.19, + "learning_rate": 1.7825594285010468e-06, + "loss": 0.3254, + "step": 32998 + }, + { + "epoch": 12.19, + "learning_rate": 1.7825224781377018e-06, + "loss": 0.2728, + "step": 32999 + }, + { + "epoch": 12.19, + "learning_rate": 1.7824855277743565e-06, + "loss": 0.2827, + "step": 33000 + }, + { + "epoch": 12.19, + "learning_rate": 1.7824485774110113e-06, + "loss": 0.2633, + "step": 33001 + }, + { + "epoch": 12.19, + "learning_rate": 1.782411627047666e-06, + "loss": 0.2553, + "step": 33002 + }, + { + "epoch": 12.19, + "learning_rate": 1.7823746766843208e-06, + "loss": 0.284, + "step": 33003 + }, + { + "epoch": 12.19, + "learning_rate": 1.7823377263209754e-06, + "loss": 0.3038, + "step": 33004 + }, + { + "epoch": 12.19, + "learning_rate": 1.7823007759576305e-06, + "loss": 0.3191, + "step": 33005 + }, + { + "epoch": 12.19, + "learning_rate": 1.7822638255942851e-06, + "loss": 0.21, + "step": 33006 + }, + { + "epoch": 12.19, + "learning_rate": 1.78222687523094e-06, + "loss": 0.2739, + "step": 33007 + }, + { + "epoch": 12.19, + "learning_rate": 1.7821899248675946e-06, + "loss": 0.4284, + "step": 33008 + }, + { + "epoch": 12.19, + "learning_rate": 1.7821529745042492e-06, + "loss": 0.4514, + "step": 33009 + }, + { + "epoch": 12.19, + "learning_rate": 1.782116024140904e-06, + "loss": 0.3955, + "step": 33010 + }, + { + "epoch": 12.19, + "learning_rate": 1.7820790737775587e-06, + "loss": 0.2413, + "step": 33011 + }, + { + "epoch": 12.19, + "learning_rate": 1.7820421234142138e-06, + "loss": 0.2979, + "step": 33012 + }, + { + "epoch": 12.19, + "learning_rate": 1.7820051730508682e-06, + "loss": 0.2207, + "step": 33013 + }, + { + "epoch": 12.19, + "learning_rate": 1.7819682226875233e-06, + "loss": 0.375, + "step": 33014 + }, + { + "epoch": 12.19, + "learning_rate": 1.781931272324178e-06, + "loss": 0.3549, + "step": 33015 + }, + { + "epoch": 12.19, + "learning_rate": 1.7818943219608328e-06, + "loss": 0.3592, + "step": 33016 + }, + { + "epoch": 12.19, + "learning_rate": 1.7818573715974874e-06, + "loss": 0.2968, + "step": 33017 + }, + { + "epoch": 12.19, + "learning_rate": 1.781820421234142e-06, + "loss": 0.3375, + "step": 33018 + }, + { + "epoch": 12.19, + "learning_rate": 1.7817834708707969e-06, + "loss": 0.3811, + "step": 33019 + }, + { + "epoch": 12.19, + "learning_rate": 1.7817465205074515e-06, + "loss": 0.2953, + "step": 33020 + }, + { + "epoch": 12.19, + "learning_rate": 1.7817095701441066e-06, + "loss": 0.2466, + "step": 33021 + }, + { + "epoch": 12.19, + "learning_rate": 1.7816726197807612e-06, + "loss": 0.3546, + "step": 33022 + }, + { + "epoch": 12.19, + "learning_rate": 1.781635669417416e-06, + "loss": 0.2995, + "step": 33023 + }, + { + "epoch": 12.19, + "learning_rate": 1.7815987190540707e-06, + "loss": 0.2684, + "step": 33024 + }, + { + "epoch": 12.2, + "learning_rate": 1.7815617686907255e-06, + "loss": 0.2423, + "step": 33025 + }, + { + "epoch": 12.2, + "learning_rate": 1.7815248183273802e-06, + "loss": 0.3729, + "step": 33026 + }, + { + "epoch": 12.2, + "learning_rate": 1.7814878679640352e-06, + "loss": 0.3426, + "step": 33027 + }, + { + "epoch": 12.2, + "learning_rate": 1.7814509176006899e-06, + "loss": 0.2371, + "step": 33028 + }, + { + "epoch": 12.2, + "learning_rate": 1.7814139672373443e-06, + "loss": 0.2452, + "step": 33029 + }, + { + "epoch": 12.2, + "learning_rate": 1.7813770168739994e-06, + "loss": 0.2953, + "step": 33030 + }, + { + "epoch": 12.2, + "learning_rate": 1.781340066510654e-06, + "loss": 0.2327, + "step": 33031 + }, + { + "epoch": 12.2, + "learning_rate": 1.7813031161473088e-06, + "loss": 0.3184, + "step": 33032 + }, + { + "epoch": 12.2, + "learning_rate": 1.7812661657839635e-06, + "loss": 0.4583, + "step": 33033 + }, + { + "epoch": 12.2, + "learning_rate": 1.7812292154206185e-06, + "loss": 0.1801, + "step": 33034 + }, + { + "epoch": 12.2, + "learning_rate": 1.781192265057273e-06, + "loss": 0.3144, + "step": 33035 + }, + { + "epoch": 12.2, + "learning_rate": 1.781155314693928e-06, + "loss": 0.4365, + "step": 33036 + }, + { + "epoch": 12.2, + "learning_rate": 1.7811183643305827e-06, + "loss": 0.2884, + "step": 33037 + }, + { + "epoch": 12.2, + "learning_rate": 1.7810814139672373e-06, + "loss": 0.2812, + "step": 33038 + }, + { + "epoch": 12.2, + "learning_rate": 1.7810444636038921e-06, + "loss": 0.4361, + "step": 33039 + }, + { + "epoch": 12.2, + "learning_rate": 1.7810075132405468e-06, + "loss": 0.3089, + "step": 33040 + }, + { + "epoch": 12.2, + "learning_rate": 1.7809705628772016e-06, + "loss": 0.1765, + "step": 33041 + }, + { + "epoch": 12.2, + "learning_rate": 1.7809336125138563e-06, + "loss": 0.2881, + "step": 33042 + }, + { + "epoch": 12.2, + "learning_rate": 1.7808966621505113e-06, + "loss": 0.1897, + "step": 33043 + }, + { + "epoch": 12.2, + "learning_rate": 1.780859711787166e-06, + "loss": 0.3069, + "step": 33044 + }, + { + "epoch": 12.2, + "learning_rate": 1.7808227614238208e-06, + "loss": 0.1661, + "step": 33045 + }, + { + "epoch": 12.2, + "learning_rate": 1.7807858110604755e-06, + "loss": 0.1656, + "step": 33046 + }, + { + "epoch": 12.2, + "learning_rate": 1.7807488606971303e-06, + "loss": 0.2579, + "step": 33047 + }, + { + "epoch": 12.2, + "learning_rate": 1.780711910333785e-06, + "loss": 0.2443, + "step": 33048 + }, + { + "epoch": 12.2, + "learning_rate": 1.7806749599704396e-06, + "loss": 0.3107, + "step": 33049 + }, + { + "epoch": 12.2, + "learning_rate": 1.7806380096070946e-06, + "loss": 0.4557, + "step": 33050 + }, + { + "epoch": 12.2, + "learning_rate": 1.7806010592437493e-06, + "loss": 0.2558, + "step": 33051 + }, + { + "epoch": 12.21, + "learning_rate": 1.7805641088804041e-06, + "loss": 0.2786, + "step": 33052 + }, + { + "epoch": 12.21, + "learning_rate": 1.7805271585170588e-06, + "loss": 0.3374, + "step": 33053 + }, + { + "epoch": 12.21, + "learning_rate": 1.7804902081537136e-06, + "loss": 0.2492, + "step": 33054 + }, + { + "epoch": 12.21, + "learning_rate": 1.7804532577903682e-06, + "loss": 0.3928, + "step": 33055 + }, + { + "epoch": 12.21, + "learning_rate": 1.7804163074270233e-06, + "loss": 0.3476, + "step": 33056 + }, + { + "epoch": 12.21, + "learning_rate": 1.7803793570636777e-06, + "loss": 0.2469, + "step": 33057 + }, + { + "epoch": 12.21, + "learning_rate": 1.7803424067003324e-06, + "loss": 0.261, + "step": 33058 + }, + { + "epoch": 12.21, + "learning_rate": 1.7803054563369874e-06, + "loss": 0.1313, + "step": 33059 + }, + { + "epoch": 12.21, + "learning_rate": 1.780268505973642e-06, + "loss": 0.2231, + "step": 33060 + }, + { + "epoch": 12.21, + "learning_rate": 1.780231555610297e-06, + "loss": 0.2716, + "step": 33061 + }, + { + "epoch": 12.21, + "learning_rate": 1.7801946052469515e-06, + "loss": 0.27, + "step": 33062 + }, + { + "epoch": 12.21, + "learning_rate": 1.7801576548836064e-06, + "loss": 0.2887, + "step": 33063 + }, + { + "epoch": 12.21, + "learning_rate": 1.780120704520261e-06, + "loss": 0.3297, + "step": 33064 + }, + { + "epoch": 12.21, + "learning_rate": 1.780083754156916e-06, + "loss": 0.2282, + "step": 33065 + }, + { + "epoch": 12.21, + "learning_rate": 1.7800468037935707e-06, + "loss": 0.1738, + "step": 33066 + }, + { + "epoch": 12.21, + "learning_rate": 1.7800098534302256e-06, + "loss": 0.3351, + "step": 33067 + }, + { + "epoch": 12.21, + "learning_rate": 1.7799729030668802e-06, + "loss": 0.2912, + "step": 33068 + }, + { + "epoch": 12.21, + "learning_rate": 1.7799359527035348e-06, + "loss": 0.2661, + "step": 33069 + }, + { + "epoch": 12.21, + "learning_rate": 1.7798990023401897e-06, + "loss": 0.1745, + "step": 33070 + }, + { + "epoch": 12.21, + "learning_rate": 1.7798620519768443e-06, + "loss": 0.2803, + "step": 33071 + }, + { + "epoch": 12.21, + "learning_rate": 1.7798251016134994e-06, + "loss": 0.3608, + "step": 33072 + }, + { + "epoch": 12.21, + "learning_rate": 1.779788151250154e-06, + "loss": 0.3597, + "step": 33073 + }, + { + "epoch": 12.21, + "learning_rate": 1.7797512008868089e-06, + "loss": 0.365, + "step": 33074 + }, + { + "epoch": 12.21, + "learning_rate": 1.7797142505234635e-06, + "loss": 0.475, + "step": 33075 + }, + { + "epoch": 12.21, + "learning_rate": 1.7796773001601184e-06, + "loss": 0.2177, + "step": 33076 + }, + { + "epoch": 12.21, + "learning_rate": 1.779640349796773e-06, + "loss": 0.2683, + "step": 33077 + }, + { + "epoch": 12.21, + "learning_rate": 1.7796033994334276e-06, + "loss": 0.2046, + "step": 33078 + }, + { + "epoch": 12.22, + "learning_rate": 1.7795664490700827e-06, + "loss": 0.4494, + "step": 33079 + }, + { + "epoch": 12.22, + "learning_rate": 1.7795294987067371e-06, + "loss": 0.2783, + "step": 33080 + }, + { + "epoch": 12.22, + "learning_rate": 1.7794925483433922e-06, + "loss": 0.3397, + "step": 33081 + }, + { + "epoch": 12.22, + "learning_rate": 1.7794555979800468e-06, + "loss": 0.2771, + "step": 33082 + }, + { + "epoch": 12.22, + "learning_rate": 1.7794186476167017e-06, + "loss": 0.3297, + "step": 33083 + }, + { + "epoch": 12.22, + "learning_rate": 1.7793816972533563e-06, + "loss": 0.2756, + "step": 33084 + }, + { + "epoch": 12.22, + "learning_rate": 1.7793447468900114e-06, + "loss": 0.2651, + "step": 33085 + }, + { + "epoch": 12.22, + "learning_rate": 1.7793077965266658e-06, + "loss": 0.2538, + "step": 33086 + }, + { + "epoch": 12.22, + "learning_rate": 1.7792708461633208e-06, + "loss": 0.3304, + "step": 33087 + }, + { + "epoch": 12.22, + "learning_rate": 1.7792338957999755e-06, + "loss": 0.2948, + "step": 33088 + }, + { + "epoch": 12.22, + "learning_rate": 1.7791969454366301e-06, + "loss": 0.2903, + "step": 33089 + }, + { + "epoch": 12.22, + "learning_rate": 1.779159995073285e-06, + "loss": 0.2528, + "step": 33090 + }, + { + "epoch": 12.22, + "learning_rate": 1.7791230447099396e-06, + "loss": 0.3632, + "step": 33091 + }, + { + "epoch": 12.22, + "learning_rate": 1.7790860943465944e-06, + "loss": 0.4492, + "step": 33092 + }, + { + "epoch": 12.22, + "learning_rate": 1.779049143983249e-06, + "loss": 0.3418, + "step": 33093 + }, + { + "epoch": 12.22, + "learning_rate": 1.7790121936199041e-06, + "loss": 0.3996, + "step": 33094 + }, + { + "epoch": 12.22, + "learning_rate": 1.7789752432565588e-06, + "loss": 0.3177, + "step": 33095 + }, + { + "epoch": 12.22, + "learning_rate": 1.7789382928932136e-06, + "loss": 0.2796, + "step": 33096 + }, + { + "epoch": 12.22, + "learning_rate": 1.7789013425298683e-06, + "loss": 0.3571, + "step": 33097 + }, + { + "epoch": 12.22, + "learning_rate": 1.778864392166523e-06, + "loss": 0.3157, + "step": 33098 + }, + { + "epoch": 12.22, + "learning_rate": 1.7788274418031777e-06, + "loss": 0.2316, + "step": 33099 + }, + { + "epoch": 12.22, + "learning_rate": 1.7787904914398324e-06, + "loss": 0.3639, + "step": 33100 + }, + { + "epoch": 12.22, + "learning_rate": 1.7787535410764874e-06, + "loss": 0.3795, + "step": 33101 + }, + { + "epoch": 12.22, + "learning_rate": 1.7787165907131419e-06, + "loss": 0.2882, + "step": 33102 + }, + { + "epoch": 12.22, + "learning_rate": 1.778679640349797e-06, + "loss": 0.3186, + "step": 33103 + }, + { + "epoch": 12.22, + "learning_rate": 1.7786426899864516e-06, + "loss": 0.3017, + "step": 33104 + }, + { + "epoch": 12.22, + "learning_rate": 1.7786057396231064e-06, + "loss": 0.2589, + "step": 33105 + }, + { + "epoch": 12.23, + "learning_rate": 1.778568789259761e-06, + "loss": 0.2031, + "step": 33106 + }, + { + "epoch": 12.23, + "learning_rate": 1.7785318388964161e-06, + "loss": 0.2775, + "step": 33107 + }, + { + "epoch": 12.23, + "learning_rate": 1.7784948885330705e-06, + "loss": 0.3548, + "step": 33108 + }, + { + "epoch": 12.23, + "learning_rate": 1.7784579381697252e-06, + "loss": 0.2659, + "step": 33109 + }, + { + "epoch": 12.23, + "learning_rate": 1.7784209878063802e-06, + "loss": 0.3008, + "step": 33110 + }, + { + "epoch": 12.23, + "learning_rate": 1.7783840374430349e-06, + "loss": 0.2975, + "step": 33111 + }, + { + "epoch": 12.23, + "learning_rate": 1.7783470870796897e-06, + "loss": 0.2616, + "step": 33112 + }, + { + "epoch": 12.23, + "learning_rate": 1.7783101367163444e-06, + "loss": 0.2875, + "step": 33113 + }, + { + "epoch": 12.23, + "learning_rate": 1.7782731863529992e-06, + "loss": 0.3231, + "step": 33114 + }, + { + "epoch": 12.23, + "learning_rate": 1.7782362359896538e-06, + "loss": 0.2316, + "step": 33115 + }, + { + "epoch": 12.23, + "learning_rate": 1.778199285626309e-06, + "loss": 0.4125, + "step": 33116 + }, + { + "epoch": 12.23, + "learning_rate": 1.7781623352629635e-06, + "loss": 0.2178, + "step": 33117 + }, + { + "epoch": 12.23, + "learning_rate": 1.7781253848996182e-06, + "loss": 0.1474, + "step": 33118 + }, + { + "epoch": 12.23, + "learning_rate": 1.778088434536273e-06, + "loss": 0.3356, + "step": 33119 + }, + { + "epoch": 12.23, + "learning_rate": 1.7780514841729277e-06, + "loss": 0.2638, + "step": 33120 + }, + { + "epoch": 12.23, + "learning_rate": 1.7780145338095825e-06, + "loss": 0.3643, + "step": 33121 + }, + { + "epoch": 12.23, + "learning_rate": 1.7779775834462371e-06, + "loss": 0.2956, + "step": 33122 + }, + { + "epoch": 12.23, + "learning_rate": 1.7779406330828922e-06, + "loss": 0.2749, + "step": 33123 + }, + { + "epoch": 12.23, + "learning_rate": 1.7779036827195468e-06, + "loss": 0.248, + "step": 33124 + }, + { + "epoch": 12.23, + "learning_rate": 1.7778667323562017e-06, + "loss": 0.2017, + "step": 33125 + }, + { + "epoch": 12.23, + "learning_rate": 1.7778297819928563e-06, + "loss": 0.351, + "step": 33126 + }, + { + "epoch": 12.23, + "learning_rate": 1.7777928316295112e-06, + "loss": 0.3635, + "step": 33127 + }, + { + "epoch": 12.23, + "learning_rate": 1.7777558812661658e-06, + "loss": 0.329, + "step": 33128 + }, + { + "epoch": 12.23, + "learning_rate": 1.7777189309028204e-06, + "loss": 0.4171, + "step": 33129 + }, + { + "epoch": 12.23, + "learning_rate": 1.7776819805394755e-06, + "loss": 0.3039, + "step": 33130 + }, + { + "epoch": 12.23, + "learning_rate": 1.77764503017613e-06, + "loss": 0.2446, + "step": 33131 + }, + { + "epoch": 12.23, + "learning_rate": 1.777608079812785e-06, + "loss": 0.3389, + "step": 33132 + }, + { + "epoch": 12.24, + "learning_rate": 1.7775711294494396e-06, + "loss": 0.3808, + "step": 33133 + }, + { + "epoch": 12.24, + "learning_rate": 1.7775341790860945e-06, + "loss": 0.3165, + "step": 33134 + }, + { + "epoch": 12.24, + "learning_rate": 1.777497228722749e-06, + "loss": 0.2854, + "step": 33135 + }, + { + "epoch": 12.24, + "learning_rate": 1.777460278359404e-06, + "loss": 0.3635, + "step": 33136 + }, + { + "epoch": 12.24, + "learning_rate": 1.7774233279960586e-06, + "loss": 0.4419, + "step": 33137 + }, + { + "epoch": 12.24, + "learning_rate": 1.7773863776327132e-06, + "loss": 0.1166, + "step": 33138 + }, + { + "epoch": 12.24, + "learning_rate": 1.7773494272693683e-06, + "loss": 0.23, + "step": 33139 + }, + { + "epoch": 12.24, + "learning_rate": 1.777312476906023e-06, + "loss": 0.2967, + "step": 33140 + }, + { + "epoch": 12.24, + "learning_rate": 1.7772755265426778e-06, + "loss": 0.2782, + "step": 33141 + }, + { + "epoch": 12.24, + "learning_rate": 1.7772385761793324e-06, + "loss": 0.4319, + "step": 33142 + }, + { + "epoch": 12.24, + "learning_rate": 1.7772016258159873e-06, + "loss": 0.2179, + "step": 33143 + }, + { + "epoch": 12.24, + "learning_rate": 1.7771646754526419e-06, + "loss": 0.3814, + "step": 33144 + }, + { + "epoch": 12.24, + "learning_rate": 1.777127725089297e-06, + "loss": 0.2772, + "step": 33145 + }, + { + "epoch": 12.24, + "learning_rate": 1.7770907747259516e-06, + "loss": 0.3764, + "step": 33146 + }, + { + "epoch": 12.24, + "learning_rate": 1.7770538243626064e-06, + "loss": 0.2021, + "step": 33147 + }, + { + "epoch": 12.24, + "learning_rate": 1.777016873999261e-06, + "loss": 0.3663, + "step": 33148 + }, + { + "epoch": 12.24, + "learning_rate": 1.7769799236359157e-06, + "loss": 0.3297, + "step": 33149 + }, + { + "epoch": 12.24, + "learning_rate": 1.7769429732725706e-06, + "loss": 0.3477, + "step": 33150 + }, + { + "epoch": 12.24, + "learning_rate": 1.7769060229092252e-06, + "loss": 0.3387, + "step": 33151 + }, + { + "epoch": 12.24, + "learning_rate": 1.7768690725458803e-06, + "loss": 0.2479, + "step": 33152 + }, + { + "epoch": 12.24, + "learning_rate": 1.7768321221825347e-06, + "loss": 0.3023, + "step": 33153 + }, + { + "epoch": 12.24, + "learning_rate": 1.7767951718191897e-06, + "loss": 0.447, + "step": 33154 + }, + { + "epoch": 12.24, + "learning_rate": 1.7767582214558444e-06, + "loss": 0.2846, + "step": 33155 + }, + { + "epoch": 12.24, + "learning_rate": 1.7767212710924992e-06, + "loss": 0.1977, + "step": 33156 + }, + { + "epoch": 12.24, + "learning_rate": 1.7766843207291539e-06, + "loss": 0.3026, + "step": 33157 + }, + { + "epoch": 12.24, + "learning_rate": 1.7766473703658085e-06, + "loss": 0.3286, + "step": 33158 + }, + { + "epoch": 12.24, + "learning_rate": 1.7766104200024633e-06, + "loss": 0.1722, + "step": 33159 + }, + { + "epoch": 12.25, + "learning_rate": 1.776573469639118e-06, + "loss": 0.4167, + "step": 33160 + }, + { + "epoch": 12.25, + "learning_rate": 1.776536519275773e-06, + "loss": 0.4154, + "step": 33161 + }, + { + "epoch": 12.25, + "learning_rate": 1.7764995689124277e-06, + "loss": 0.266, + "step": 33162 + }, + { + "epoch": 12.25, + "learning_rate": 1.7764626185490825e-06, + "loss": 0.3601, + "step": 33163 + }, + { + "epoch": 12.25, + "learning_rate": 1.7764256681857372e-06, + "loss": 0.2401, + "step": 33164 + }, + { + "epoch": 12.25, + "learning_rate": 1.776388717822392e-06, + "loss": 0.3404, + "step": 33165 + }, + { + "epoch": 12.25, + "learning_rate": 1.7763517674590466e-06, + "loss": 0.2207, + "step": 33166 + }, + { + "epoch": 12.25, + "learning_rate": 1.7763148170957017e-06, + "loss": 0.3634, + "step": 33167 + }, + { + "epoch": 12.25, + "learning_rate": 1.7762778667323563e-06, + "loss": 0.2493, + "step": 33168 + }, + { + "epoch": 12.25, + "learning_rate": 1.7762409163690108e-06, + "loss": 0.2998, + "step": 33169 + }, + { + "epoch": 12.25, + "learning_rate": 1.7762039660056658e-06, + "loss": 0.303, + "step": 33170 + }, + { + "epoch": 12.25, + "learning_rate": 1.7761670156423205e-06, + "loss": 0.2491, + "step": 33171 + }, + { + "epoch": 12.25, + "learning_rate": 1.7761300652789753e-06, + "loss": 0.3187, + "step": 33172 + }, + { + "epoch": 12.25, + "learning_rate": 1.77609311491563e-06, + "loss": 0.2329, + "step": 33173 + }, + { + "epoch": 12.25, + "learning_rate": 1.776056164552285e-06, + "loss": 0.2355, + "step": 33174 + }, + { + "epoch": 12.25, + "learning_rate": 1.7760192141889394e-06, + "loss": 0.3286, + "step": 33175 + }, + { + "epoch": 12.25, + "learning_rate": 1.7759822638255945e-06, + "loss": 0.3624, + "step": 33176 + }, + { + "epoch": 12.25, + "learning_rate": 1.7759453134622491e-06, + "loss": 0.391, + "step": 33177 + }, + { + "epoch": 12.25, + "learning_rate": 1.7759083630989038e-06, + "loss": 0.3033, + "step": 33178 + }, + { + "epoch": 12.25, + "learning_rate": 1.7758714127355586e-06, + "loss": 0.2731, + "step": 33179 + }, + { + "epoch": 12.25, + "learning_rate": 1.7758344623722132e-06, + "loss": 0.4268, + "step": 33180 + }, + { + "epoch": 12.25, + "learning_rate": 1.775797512008868e-06, + "loss": 0.3605, + "step": 33181 + }, + { + "epoch": 12.25, + "learning_rate": 1.7757605616455227e-06, + "loss": 0.3156, + "step": 33182 + }, + { + "epoch": 12.25, + "learning_rate": 1.7757236112821778e-06, + "loss": 0.2923, + "step": 33183 + }, + { + "epoch": 12.25, + "learning_rate": 1.7756866609188324e-06, + "loss": 0.4426, + "step": 33184 + }, + { + "epoch": 12.25, + "learning_rate": 1.7756497105554873e-06, + "loss": 0.3559, + "step": 33185 + }, + { + "epoch": 12.25, + "learning_rate": 1.775612760192142e-06, + "loss": 0.3148, + "step": 33186 + }, + { + "epoch": 12.26, + "learning_rate": 1.7755758098287968e-06, + "loss": 0.3203, + "step": 33187 + }, + { + "epoch": 12.26, + "learning_rate": 1.7755388594654514e-06, + "loss": 0.3773, + "step": 33188 + }, + { + "epoch": 12.26, + "learning_rate": 1.775501909102106e-06, + "loss": 0.3493, + "step": 33189 + }, + { + "epoch": 12.26, + "learning_rate": 1.775464958738761e-06, + "loss": 0.2521, + "step": 33190 + }, + { + "epoch": 12.26, + "learning_rate": 1.7754280083754157e-06, + "loss": 0.374, + "step": 33191 + }, + { + "epoch": 12.26, + "learning_rate": 1.7753910580120706e-06, + "loss": 0.2861, + "step": 33192 + }, + { + "epoch": 12.26, + "learning_rate": 1.7753541076487252e-06, + "loss": 0.3247, + "step": 33193 + }, + { + "epoch": 12.26, + "learning_rate": 1.77531715728538e-06, + "loss": 0.2043, + "step": 33194 + }, + { + "epoch": 12.26, + "learning_rate": 1.7752802069220347e-06, + "loss": 0.3521, + "step": 33195 + }, + { + "epoch": 12.26, + "learning_rate": 1.7752432565586898e-06, + "loss": 0.2645, + "step": 33196 + }, + { + "epoch": 12.26, + "learning_rate": 1.7752063061953444e-06, + "loss": 0.3296, + "step": 33197 + }, + { + "epoch": 12.26, + "learning_rate": 1.7751693558319988e-06, + "loss": 0.2582, + "step": 33198 + }, + { + "epoch": 12.26, + "learning_rate": 1.7751324054686539e-06, + "loss": 0.2999, + "step": 33199 + }, + { + "epoch": 12.26, + "learning_rate": 1.7750954551053085e-06, + "loss": 0.2425, + "step": 33200 + }, + { + "epoch": 12.26, + "learning_rate": 1.7750585047419634e-06, + "loss": 0.2305, + "step": 33201 + }, + { + "epoch": 12.26, + "learning_rate": 1.775021554378618e-06, + "loss": 0.4156, + "step": 33202 + }, + { + "epoch": 12.26, + "learning_rate": 1.774984604015273e-06, + "loss": 0.3448, + "step": 33203 + }, + { + "epoch": 12.26, + "learning_rate": 1.7749476536519275e-06, + "loss": 0.3666, + "step": 33204 + }, + { + "epoch": 12.26, + "learning_rate": 1.7749107032885825e-06, + "loss": 0.2388, + "step": 33205 + }, + { + "epoch": 12.26, + "learning_rate": 1.7748737529252372e-06, + "loss": 0.2928, + "step": 33206 + }, + { + "epoch": 12.26, + "learning_rate": 1.774836802561892e-06, + "loss": 0.2365, + "step": 33207 + }, + { + "epoch": 12.26, + "learning_rate": 1.7747998521985467e-06, + "loss": 0.2438, + "step": 33208 + }, + { + "epoch": 12.26, + "learning_rate": 1.7747629018352013e-06, + "loss": 0.3407, + "step": 33209 + }, + { + "epoch": 12.26, + "learning_rate": 1.7747259514718562e-06, + "loss": 0.3274, + "step": 33210 + }, + { + "epoch": 12.26, + "learning_rate": 1.7746890011085108e-06, + "loss": 0.3602, + "step": 33211 + }, + { + "epoch": 12.26, + "learning_rate": 1.7746520507451659e-06, + "loss": 0.4733, + "step": 33212 + }, + { + "epoch": 12.26, + "learning_rate": 1.7746151003818205e-06, + "loss": 0.2939, + "step": 33213 + }, + { + "epoch": 12.27, + "learning_rate": 1.7745781500184753e-06, + "loss": 0.232, + "step": 33214 + }, + { + "epoch": 12.27, + "learning_rate": 1.77454119965513e-06, + "loss": 0.2281, + "step": 33215 + }, + { + "epoch": 12.27, + "learning_rate": 1.7745042492917848e-06, + "loss": 0.183, + "step": 33216 + }, + { + "epoch": 12.27, + "learning_rate": 1.7744672989284395e-06, + "loss": 0.3954, + "step": 33217 + }, + { + "epoch": 12.27, + "learning_rate": 1.774430348565094e-06, + "loss": 0.3381, + "step": 33218 + }, + { + "epoch": 12.27, + "learning_rate": 1.7743933982017492e-06, + "loss": 0.3614, + "step": 33219 + }, + { + "epoch": 12.27, + "learning_rate": 1.7743564478384036e-06, + "loss": 0.2641, + "step": 33220 + }, + { + "epoch": 12.27, + "learning_rate": 1.7743194974750586e-06, + "loss": 0.2858, + "step": 33221 + }, + { + "epoch": 12.27, + "learning_rate": 1.7742825471117133e-06, + "loss": 0.3909, + "step": 33222 + }, + { + "epoch": 12.27, + "learning_rate": 1.7742455967483681e-06, + "loss": 0.3045, + "step": 33223 + }, + { + "epoch": 12.27, + "learning_rate": 1.7742086463850228e-06, + "loss": 0.3494, + "step": 33224 + }, + { + "epoch": 12.27, + "learning_rate": 1.7741716960216778e-06, + "loss": 0.2927, + "step": 33225 + }, + { + "epoch": 12.27, + "learning_rate": 1.7741347456583322e-06, + "loss": 0.3691, + "step": 33226 + }, + { + "epoch": 12.27, + "learning_rate": 1.7740977952949869e-06, + "loss": 0.2755, + "step": 33227 + }, + { + "epoch": 12.27, + "learning_rate": 1.774060844931642e-06, + "loss": 0.2521, + "step": 33228 + }, + { + "epoch": 12.27, + "learning_rate": 1.7740238945682966e-06, + "loss": 0.2224, + "step": 33229 + }, + { + "epoch": 12.27, + "learning_rate": 1.7739869442049514e-06, + "loss": 0.3258, + "step": 33230 + }, + { + "epoch": 12.27, + "learning_rate": 1.773949993841606e-06, + "loss": 0.2863, + "step": 33231 + }, + { + "epoch": 12.27, + "learning_rate": 1.773913043478261e-06, + "loss": 0.2122, + "step": 33232 + }, + { + "epoch": 12.27, + "learning_rate": 1.7738760931149155e-06, + "loss": 0.2625, + "step": 33233 + }, + { + "epoch": 12.27, + "learning_rate": 1.7738391427515706e-06, + "loss": 0.316, + "step": 33234 + }, + { + "epoch": 12.27, + "learning_rate": 1.7738021923882252e-06, + "loss": 0.3641, + "step": 33235 + }, + { + "epoch": 12.27, + "learning_rate": 1.77376524202488e-06, + "loss": 0.424, + "step": 33236 + }, + { + "epoch": 12.27, + "learning_rate": 1.7737282916615347e-06, + "loss": 0.2703, + "step": 33237 + }, + { + "epoch": 12.27, + "learning_rate": 1.7736913412981894e-06, + "loss": 0.3159, + "step": 33238 + }, + { + "epoch": 12.27, + "learning_rate": 1.7736543909348442e-06, + "loss": 0.2483, + "step": 33239 + }, + { + "epoch": 12.27, + "learning_rate": 1.7736174405714988e-06, + "loss": 0.298, + "step": 33240 + }, + { + "epoch": 12.28, + "learning_rate": 1.773580490208154e-06, + "loss": 0.339, + "step": 33241 + }, + { + "epoch": 12.28, + "learning_rate": 1.7735435398448083e-06, + "loss": 0.2972, + "step": 33242 + }, + { + "epoch": 12.28, + "learning_rate": 1.7735065894814634e-06, + "loss": 0.3829, + "step": 33243 + }, + { + "epoch": 12.28, + "learning_rate": 1.773469639118118e-06, + "loss": 0.4465, + "step": 33244 + }, + { + "epoch": 12.28, + "learning_rate": 1.7734326887547729e-06, + "loss": 0.2358, + "step": 33245 + }, + { + "epoch": 12.28, + "learning_rate": 1.7733957383914275e-06, + "loss": 0.2378, + "step": 33246 + }, + { + "epoch": 12.28, + "learning_rate": 1.7733587880280821e-06, + "loss": 0.2109, + "step": 33247 + }, + { + "epoch": 12.28, + "learning_rate": 1.773321837664737e-06, + "loss": 0.2183, + "step": 33248 + }, + { + "epoch": 12.28, + "learning_rate": 1.7732848873013916e-06, + "loss": 0.3222, + "step": 33249 + }, + { + "epoch": 12.28, + "learning_rate": 1.7732479369380467e-06, + "loss": 0.2608, + "step": 33250 + }, + { + "epoch": 12.28, + "learning_rate": 1.7732109865747013e-06, + "loss": 0.3246, + "step": 33251 + }, + { + "epoch": 12.28, + "learning_rate": 1.7731740362113562e-06, + "loss": 0.3462, + "step": 33252 + }, + { + "epoch": 12.28, + "learning_rate": 1.7731370858480108e-06, + "loss": 0.4367, + "step": 33253 + }, + { + "epoch": 12.28, + "learning_rate": 1.7731001354846657e-06, + "loss": 0.1967, + "step": 33254 + }, + { + "epoch": 12.28, + "learning_rate": 1.7730631851213203e-06, + "loss": 0.2209, + "step": 33255 + }, + { + "epoch": 12.28, + "learning_rate": 1.7730262347579754e-06, + "loss": 0.3332, + "step": 33256 + }, + { + "epoch": 12.28, + "learning_rate": 1.77298928439463e-06, + "loss": 0.2683, + "step": 33257 + }, + { + "epoch": 12.28, + "learning_rate": 1.7729523340312846e-06, + "loss": 0.2644, + "step": 33258 + }, + { + "epoch": 12.28, + "learning_rate": 1.7729153836679395e-06, + "loss": 0.2282, + "step": 33259 + }, + { + "epoch": 12.28, + "learning_rate": 1.7728784333045941e-06, + "loss": 0.3089, + "step": 33260 + }, + { + "epoch": 12.28, + "learning_rate": 1.772841482941249e-06, + "loss": 0.4265, + "step": 33261 + }, + { + "epoch": 12.28, + "learning_rate": 1.7728045325779036e-06, + "loss": 0.3635, + "step": 33262 + }, + { + "epoch": 12.28, + "learning_rate": 1.7727675822145587e-06, + "loss": 0.3542, + "step": 33263 + }, + { + "epoch": 12.28, + "learning_rate": 1.7727306318512133e-06, + "loss": 0.4608, + "step": 33264 + }, + { + "epoch": 12.28, + "learning_rate": 1.7726936814878681e-06, + "loss": 0.426, + "step": 33265 + }, + { + "epoch": 12.28, + "learning_rate": 1.7726567311245228e-06, + "loss": 0.3163, + "step": 33266 + }, + { + "epoch": 12.28, + "learning_rate": 1.7726197807611774e-06, + "loss": 0.2893, + "step": 33267 + }, + { + "epoch": 12.29, + "learning_rate": 1.7725828303978323e-06, + "loss": 0.3186, + "step": 33268 + }, + { + "epoch": 12.29, + "learning_rate": 1.772545880034487e-06, + "loss": 0.3104, + "step": 33269 + }, + { + "epoch": 12.29, + "learning_rate": 1.772508929671142e-06, + "loss": 0.3929, + "step": 33270 + }, + { + "epoch": 12.29, + "learning_rate": 1.7724719793077964e-06, + "loss": 0.2477, + "step": 33271 + }, + { + "epoch": 12.29, + "learning_rate": 1.7724350289444514e-06, + "loss": 0.2846, + "step": 33272 + }, + { + "epoch": 12.29, + "learning_rate": 1.772398078581106e-06, + "loss": 0.4026, + "step": 33273 + }, + { + "epoch": 12.29, + "learning_rate": 1.772361128217761e-06, + "loss": 0.384, + "step": 33274 + }, + { + "epoch": 12.29, + "learning_rate": 1.7723241778544156e-06, + "loss": 0.2404, + "step": 33275 + }, + { + "epoch": 12.29, + "learning_rate": 1.7722872274910706e-06, + "loss": 0.372, + "step": 33276 + }, + { + "epoch": 12.29, + "learning_rate": 1.772250277127725e-06, + "loss": 0.3529, + "step": 33277 + }, + { + "epoch": 12.29, + "learning_rate": 1.7722133267643797e-06, + "loss": 0.3488, + "step": 33278 + }, + { + "epoch": 12.29, + "learning_rate": 1.7721763764010348e-06, + "loss": 0.2854, + "step": 33279 + }, + { + "epoch": 12.29, + "learning_rate": 1.7721394260376894e-06, + "loss": 0.2832, + "step": 33280 + }, + { + "epoch": 12.29, + "learning_rate": 1.7721024756743442e-06, + "loss": 0.2402, + "step": 33281 + }, + { + "epoch": 12.29, + "learning_rate": 1.7720655253109989e-06, + "loss": 0.1526, + "step": 33282 + }, + { + "epoch": 12.29, + "learning_rate": 1.7720285749476537e-06, + "loss": 0.3262, + "step": 33283 + }, + { + "epoch": 12.29, + "learning_rate": 1.7719916245843084e-06, + "loss": 0.3206, + "step": 33284 + }, + { + "epoch": 12.29, + "learning_rate": 1.7719546742209634e-06, + "loss": 0.3463, + "step": 33285 + }, + { + "epoch": 12.29, + "learning_rate": 1.771917723857618e-06, + "loss": 0.3024, + "step": 33286 + }, + { + "epoch": 12.29, + "learning_rate": 1.7718807734942725e-06, + "loss": 0.2448, + "step": 33287 + }, + { + "epoch": 12.29, + "learning_rate": 1.7718438231309275e-06, + "loss": 0.2216, + "step": 33288 + }, + { + "epoch": 12.29, + "learning_rate": 1.7718068727675822e-06, + "loss": 0.2372, + "step": 33289 + }, + { + "epoch": 12.29, + "learning_rate": 1.771769922404237e-06, + "loss": 0.2732, + "step": 33290 + }, + { + "epoch": 12.29, + "learning_rate": 1.7717329720408917e-06, + "loss": 0.4061, + "step": 33291 + }, + { + "epoch": 12.29, + "learning_rate": 1.7716960216775467e-06, + "loss": 0.4965, + "step": 33292 + }, + { + "epoch": 12.29, + "learning_rate": 1.7716590713142011e-06, + "loss": 0.3033, + "step": 33293 + }, + { + "epoch": 12.29, + "learning_rate": 1.7716221209508562e-06, + "loss": 0.3473, + "step": 33294 + }, + { + "epoch": 12.3, + "learning_rate": 1.7715851705875108e-06, + "loss": 0.1779, + "step": 33295 + }, + { + "epoch": 12.3, + "learning_rate": 1.7715482202241657e-06, + "loss": 0.3687, + "step": 33296 + }, + { + "epoch": 12.3, + "learning_rate": 1.7715112698608203e-06, + "loss": 0.423, + "step": 33297 + }, + { + "epoch": 12.3, + "learning_rate": 1.771474319497475e-06, + "loss": 0.3091, + "step": 33298 + }, + { + "epoch": 12.3, + "learning_rate": 1.7714373691341298e-06, + "loss": 0.401, + "step": 33299 + }, + { + "epoch": 12.3, + "learning_rate": 1.7714004187707844e-06, + "loss": 0.2549, + "step": 33300 + }, + { + "epoch": 12.3, + "learning_rate": 1.7713634684074395e-06, + "loss": 0.4328, + "step": 33301 + }, + { + "epoch": 12.3, + "learning_rate": 1.7713265180440941e-06, + "loss": 0.2678, + "step": 33302 + }, + { + "epoch": 12.3, + "learning_rate": 1.771289567680749e-06, + "loss": 0.2775, + "step": 33303 + }, + { + "epoch": 12.3, + "learning_rate": 1.7712526173174036e-06, + "loss": 0.4336, + "step": 33304 + }, + { + "epoch": 12.3, + "learning_rate": 1.7712156669540585e-06, + "loss": 0.2573, + "step": 33305 + }, + { + "epoch": 12.3, + "learning_rate": 1.7711787165907131e-06, + "loss": 0.2538, + "step": 33306 + }, + { + "epoch": 12.3, + "learning_rate": 1.7711417662273677e-06, + "loss": 0.2603, + "step": 33307 + }, + { + "epoch": 12.3, + "learning_rate": 1.7711048158640228e-06, + "loss": 0.4161, + "step": 33308 + }, + { + "epoch": 12.3, + "learning_rate": 1.7710678655006774e-06, + "loss": 0.2626, + "step": 33309 + }, + { + "epoch": 12.3, + "learning_rate": 1.7710309151373323e-06, + "loss": 0.32, + "step": 33310 + }, + { + "epoch": 12.3, + "learning_rate": 1.770993964773987e-06, + "loss": 0.3114, + "step": 33311 + }, + { + "epoch": 12.3, + "learning_rate": 1.7709570144106418e-06, + "loss": 0.2435, + "step": 33312 + }, + { + "epoch": 12.3, + "learning_rate": 1.7709200640472964e-06, + "loss": 0.2151, + "step": 33313 + }, + { + "epoch": 12.3, + "learning_rate": 1.7708831136839515e-06, + "loss": 0.2179, + "step": 33314 + }, + { + "epoch": 12.3, + "learning_rate": 1.7708461633206061e-06, + "loss": 0.3062, + "step": 33315 + }, + { + "epoch": 12.3, + "learning_rate": 1.770809212957261e-06, + "loss": 0.2581, + "step": 33316 + }, + { + "epoch": 12.3, + "learning_rate": 1.7707722625939156e-06, + "loss": 0.2473, + "step": 33317 + }, + { + "epoch": 12.3, + "learning_rate": 1.7707353122305702e-06, + "loss": 0.2682, + "step": 33318 + }, + { + "epoch": 12.3, + "learning_rate": 1.770698361867225e-06, + "loss": 0.2285, + "step": 33319 + }, + { + "epoch": 12.3, + "learning_rate": 1.7706614115038797e-06, + "loss": 0.2751, + "step": 33320 + }, + { + "epoch": 12.3, + "learning_rate": 1.7706244611405346e-06, + "loss": 0.2883, + "step": 33321 + }, + { + "epoch": 12.31, + "learning_rate": 1.7705875107771892e-06, + "loss": 0.3014, + "step": 33322 + }, + { + "epoch": 12.31, + "learning_rate": 1.7705505604138443e-06, + "loss": 0.3154, + "step": 33323 + }, + { + "epoch": 12.31, + "learning_rate": 1.770513610050499e-06, + "loss": 0.4351, + "step": 33324 + }, + { + "epoch": 12.31, + "learning_rate": 1.7704766596871537e-06, + "loss": 0.4107, + "step": 33325 + }, + { + "epoch": 12.31, + "learning_rate": 1.7704397093238084e-06, + "loss": 0.2457, + "step": 33326 + }, + { + "epoch": 12.31, + "learning_rate": 1.770402758960463e-06, + "loss": 0.3114, + "step": 33327 + }, + { + "epoch": 12.31, + "learning_rate": 1.7703658085971179e-06, + "loss": 0.3668, + "step": 33328 + }, + { + "epoch": 12.31, + "learning_rate": 1.7703288582337725e-06, + "loss": 0.3481, + "step": 33329 + }, + { + "epoch": 12.31, + "learning_rate": 1.7702919078704276e-06, + "loss": 0.3102, + "step": 33330 + }, + { + "epoch": 12.31, + "learning_rate": 1.7702549575070822e-06, + "loss": 0.2735, + "step": 33331 + }, + { + "epoch": 12.31, + "learning_rate": 1.770218007143737e-06, + "loss": 0.333, + "step": 33332 + }, + { + "epoch": 12.31, + "learning_rate": 1.7701810567803917e-06, + "loss": 0.2175, + "step": 33333 + }, + { + "epoch": 12.31, + "learning_rate": 1.7701441064170465e-06, + "loss": 0.2951, + "step": 33334 + }, + { + "epoch": 12.31, + "learning_rate": 1.7701071560537012e-06, + "loss": 0.2526, + "step": 33335 + }, + { + "epoch": 12.31, + "learning_rate": 1.7700702056903562e-06, + "loss": 0.3112, + "step": 33336 + }, + { + "epoch": 12.31, + "learning_rate": 1.7700332553270109e-06, + "loss": 0.3711, + "step": 33337 + }, + { + "epoch": 12.31, + "learning_rate": 1.7699963049636653e-06, + "loss": 0.3301, + "step": 33338 + }, + { + "epoch": 12.31, + "learning_rate": 1.7699593546003203e-06, + "loss": 0.3033, + "step": 33339 + }, + { + "epoch": 12.31, + "learning_rate": 1.769922404236975e-06, + "loss": 0.3597, + "step": 33340 + }, + { + "epoch": 12.31, + "learning_rate": 1.7698854538736298e-06, + "loss": 0.2998, + "step": 33341 + }, + { + "epoch": 12.31, + "learning_rate": 1.7698485035102845e-06, + "loss": 0.3734, + "step": 33342 + }, + { + "epoch": 12.31, + "learning_rate": 1.7698115531469395e-06, + "loss": 0.2363, + "step": 33343 + }, + { + "epoch": 12.31, + "learning_rate": 1.769774602783594e-06, + "loss": 0.2772, + "step": 33344 + }, + { + "epoch": 12.31, + "learning_rate": 1.769737652420249e-06, + "loss": 0.3358, + "step": 33345 + }, + { + "epoch": 12.31, + "learning_rate": 1.7697007020569037e-06, + "loss": 0.3166, + "step": 33346 + }, + { + "epoch": 12.31, + "learning_rate": 1.7696637516935583e-06, + "loss": 0.2307, + "step": 33347 + }, + { + "epoch": 12.31, + "learning_rate": 1.7696268013302131e-06, + "loss": 0.3413, + "step": 33348 + }, + { + "epoch": 12.31, + "learning_rate": 1.7695898509668678e-06, + "loss": 0.416, + "step": 33349 + }, + { + "epoch": 12.32, + "learning_rate": 1.7695529006035226e-06, + "loss": 0.3448, + "step": 33350 + }, + { + "epoch": 12.32, + "learning_rate": 1.7695159502401773e-06, + "loss": 0.2953, + "step": 33351 + }, + { + "epoch": 12.32, + "learning_rate": 1.7694789998768323e-06, + "loss": 0.3397, + "step": 33352 + }, + { + "epoch": 12.32, + "learning_rate": 1.769442049513487e-06, + "loss": 0.1928, + "step": 33353 + }, + { + "epoch": 12.32, + "learning_rate": 1.7694050991501418e-06, + "loss": 0.3526, + "step": 33354 + }, + { + "epoch": 12.32, + "learning_rate": 1.7693681487867964e-06, + "loss": 0.24, + "step": 33355 + }, + { + "epoch": 12.32, + "learning_rate": 1.7693311984234513e-06, + "loss": 0.2284, + "step": 33356 + }, + { + "epoch": 12.32, + "learning_rate": 1.769294248060106e-06, + "loss": 0.389, + "step": 33357 + }, + { + "epoch": 12.32, + "learning_rate": 1.7692572976967606e-06, + "loss": 0.2001, + "step": 33358 + }, + { + "epoch": 12.32, + "learning_rate": 1.7692203473334156e-06, + "loss": 0.2921, + "step": 33359 + }, + { + "epoch": 12.32, + "learning_rate": 1.76918339697007e-06, + "loss": 0.3163, + "step": 33360 + }, + { + "epoch": 12.32, + "learning_rate": 1.769146446606725e-06, + "loss": 0.3862, + "step": 33361 + }, + { + "epoch": 12.32, + "learning_rate": 1.7691094962433797e-06, + "loss": 0.2209, + "step": 33362 + }, + { + "epoch": 12.32, + "learning_rate": 1.7690725458800346e-06, + "loss": 0.2767, + "step": 33363 + }, + { + "epoch": 12.32, + "learning_rate": 1.7690355955166892e-06, + "loss": 0.4156, + "step": 33364 + }, + { + "epoch": 12.32, + "learning_rate": 1.7689986451533443e-06, + "loss": 0.2562, + "step": 33365 + }, + { + "epoch": 12.32, + "learning_rate": 1.7689616947899987e-06, + "loss": 0.2281, + "step": 33366 + }, + { + "epoch": 12.32, + "learning_rate": 1.7689247444266533e-06, + "loss": 0.3939, + "step": 33367 + }, + { + "epoch": 12.32, + "learning_rate": 1.7688877940633084e-06, + "loss": 0.3324, + "step": 33368 + }, + { + "epoch": 12.32, + "learning_rate": 1.768850843699963e-06, + "loss": 0.245, + "step": 33369 + }, + { + "epoch": 12.32, + "learning_rate": 1.7688138933366179e-06, + "loss": 0.2375, + "step": 33370 + }, + { + "epoch": 12.32, + "learning_rate": 1.7687769429732725e-06, + "loss": 0.3439, + "step": 33371 + }, + { + "epoch": 12.32, + "learning_rate": 1.7687399926099274e-06, + "loss": 0.3757, + "step": 33372 + }, + { + "epoch": 12.32, + "learning_rate": 1.768703042246582e-06, + "loss": 0.2308, + "step": 33373 + }, + { + "epoch": 12.32, + "learning_rate": 1.768666091883237e-06, + "loss": 0.2398, + "step": 33374 + }, + { + "epoch": 12.32, + "learning_rate": 1.7686291415198917e-06, + "loss": 0.3678, + "step": 33375 + }, + { + "epoch": 12.32, + "learning_rate": 1.7685921911565466e-06, + "loss": 0.2265, + "step": 33376 + }, + { + "epoch": 12.33, + "learning_rate": 1.7685552407932012e-06, + "loss": 0.3498, + "step": 33377 + }, + { + "epoch": 12.33, + "learning_rate": 1.7685182904298558e-06, + "loss": 0.1905, + "step": 33378 + }, + { + "epoch": 12.33, + "learning_rate": 1.7684813400665107e-06, + "loss": 0.254, + "step": 33379 + }, + { + "epoch": 12.33, + "learning_rate": 1.7684443897031653e-06, + "loss": 0.3691, + "step": 33380 + }, + { + "epoch": 12.33, + "learning_rate": 1.7684074393398204e-06, + "loss": 0.1996, + "step": 33381 + }, + { + "epoch": 12.33, + "learning_rate": 1.768370488976475e-06, + "loss": 0.4446, + "step": 33382 + }, + { + "epoch": 12.33, + "learning_rate": 1.7683335386131299e-06, + "loss": 0.2379, + "step": 33383 + }, + { + "epoch": 12.33, + "learning_rate": 1.7682965882497845e-06, + "loss": 0.4126, + "step": 33384 + }, + { + "epoch": 12.33, + "learning_rate": 1.7682596378864393e-06, + "loss": 0.2265, + "step": 33385 + }, + { + "epoch": 12.33, + "learning_rate": 1.768222687523094e-06, + "loss": 0.2337, + "step": 33386 + }, + { + "epoch": 12.33, + "learning_rate": 1.7681857371597486e-06, + "loss": 0.3168, + "step": 33387 + }, + { + "epoch": 12.33, + "learning_rate": 1.7681487867964037e-06, + "loss": 0.3619, + "step": 33388 + }, + { + "epoch": 12.33, + "learning_rate": 1.768111836433058e-06, + "loss": 0.2469, + "step": 33389 + }, + { + "epoch": 12.33, + "learning_rate": 1.7680748860697132e-06, + "loss": 0.3124, + "step": 33390 + }, + { + "epoch": 12.33, + "learning_rate": 1.7680379357063678e-06, + "loss": 0.2063, + "step": 33391 + }, + { + "epoch": 12.33, + "learning_rate": 1.7680009853430226e-06, + "loss": 0.4783, + "step": 33392 + }, + { + "epoch": 12.33, + "learning_rate": 1.7679640349796773e-06, + "loss": 0.2272, + "step": 33393 + }, + { + "epoch": 12.33, + "learning_rate": 1.7679270846163323e-06, + "loss": 0.2794, + "step": 33394 + }, + { + "epoch": 12.33, + "learning_rate": 1.7678901342529868e-06, + "loss": 0.2996, + "step": 33395 + }, + { + "epoch": 12.33, + "learning_rate": 1.7678531838896418e-06, + "loss": 0.3374, + "step": 33396 + }, + { + "epoch": 12.33, + "learning_rate": 1.7678162335262965e-06, + "loss": 0.288, + "step": 33397 + }, + { + "epoch": 12.33, + "learning_rate": 1.767779283162951e-06, + "loss": 0.3435, + "step": 33398 + }, + { + "epoch": 12.33, + "learning_rate": 1.767742332799606e-06, + "loss": 0.2325, + "step": 33399 + }, + { + "epoch": 12.33, + "learning_rate": 1.7677053824362606e-06, + "loss": 0.376, + "step": 33400 + }, + { + "epoch": 12.33, + "learning_rate": 1.7676684320729154e-06, + "loss": 0.2548, + "step": 33401 + }, + { + "epoch": 12.33, + "learning_rate": 1.76763148170957e-06, + "loss": 0.2087, + "step": 33402 + }, + { + "epoch": 12.33, + "learning_rate": 1.7675945313462251e-06, + "loss": 0.3602, + "step": 33403 + }, + { + "epoch": 12.34, + "learning_rate": 1.7675575809828798e-06, + "loss": 0.3371, + "step": 33404 + }, + { + "epoch": 12.34, + "learning_rate": 1.7675206306195346e-06, + "loss": 0.3924, + "step": 33405 + }, + { + "epoch": 12.34, + "learning_rate": 1.7674836802561892e-06, + "loss": 0.3633, + "step": 33406 + }, + { + "epoch": 12.34, + "learning_rate": 1.7674467298928439e-06, + "loss": 0.3095, + "step": 33407 + }, + { + "epoch": 12.34, + "learning_rate": 1.7674097795294987e-06, + "loss": 0.3152, + "step": 33408 + }, + { + "epoch": 12.34, + "learning_rate": 1.7673728291661534e-06, + "loss": 0.3261, + "step": 33409 + }, + { + "epoch": 12.34, + "learning_rate": 1.7673358788028084e-06, + "loss": 0.1587, + "step": 33410 + }, + { + "epoch": 12.34, + "learning_rate": 1.7672989284394629e-06, + "loss": 0.2823, + "step": 33411 + }, + { + "epoch": 12.34, + "learning_rate": 1.767261978076118e-06, + "loss": 0.3093, + "step": 33412 + }, + { + "epoch": 12.34, + "learning_rate": 1.7672250277127725e-06, + "loss": 0.4363, + "step": 33413 + }, + { + "epoch": 12.34, + "learning_rate": 1.7671880773494274e-06, + "loss": 0.3571, + "step": 33414 + }, + { + "epoch": 12.34, + "learning_rate": 1.767151126986082e-06, + "loss": 0.3355, + "step": 33415 + }, + { + "epoch": 12.34, + "learning_rate": 1.767114176622737e-06, + "loss": 0.3311, + "step": 33416 + }, + { + "epoch": 12.34, + "learning_rate": 1.7670772262593915e-06, + "loss": 0.3024, + "step": 33417 + }, + { + "epoch": 12.34, + "learning_rate": 1.7670402758960462e-06, + "loss": 0.29, + "step": 33418 + }, + { + "epoch": 12.34, + "learning_rate": 1.7670033255327012e-06, + "loss": 0.2531, + "step": 33419 + }, + { + "epoch": 12.34, + "learning_rate": 1.7669663751693559e-06, + "loss": 0.3308, + "step": 33420 + }, + { + "epoch": 12.34, + "learning_rate": 1.7669294248060107e-06, + "loss": 0.1788, + "step": 33421 + }, + { + "epoch": 12.34, + "learning_rate": 1.7668924744426653e-06, + "loss": 0.2561, + "step": 33422 + }, + { + "epoch": 12.34, + "learning_rate": 1.7668555240793202e-06, + "loss": 0.2342, + "step": 33423 + }, + { + "epoch": 12.34, + "learning_rate": 1.7668185737159748e-06, + "loss": 0.2739, + "step": 33424 + }, + { + "epoch": 12.34, + "learning_rate": 1.7667816233526299e-06, + "loss": 0.2562, + "step": 33425 + }, + { + "epoch": 12.34, + "learning_rate": 1.7667446729892845e-06, + "loss": 0.2854, + "step": 33426 + }, + { + "epoch": 12.34, + "learning_rate": 1.766707722625939e-06, + "loss": 0.2768, + "step": 33427 + }, + { + "epoch": 12.34, + "learning_rate": 1.766670772262594e-06, + "loss": 0.2534, + "step": 33428 + }, + { + "epoch": 12.34, + "learning_rate": 1.7666338218992486e-06, + "loss": 0.2358, + "step": 33429 + }, + { + "epoch": 12.34, + "learning_rate": 1.7665968715359035e-06, + "loss": 0.3131, + "step": 33430 + }, + { + "epoch": 12.35, + "learning_rate": 1.7665599211725581e-06, + "loss": 0.1633, + "step": 33431 + }, + { + "epoch": 12.35, + "learning_rate": 1.7665229708092132e-06, + "loss": 0.3356, + "step": 33432 + }, + { + "epoch": 12.35, + "learning_rate": 1.7664860204458676e-06, + "loss": 0.2618, + "step": 33433 + }, + { + "epoch": 12.35, + "learning_rate": 1.7664490700825227e-06, + "loss": 0.2681, + "step": 33434 + }, + { + "epoch": 12.35, + "learning_rate": 1.7664121197191773e-06, + "loss": 0.2646, + "step": 33435 + }, + { + "epoch": 12.35, + "learning_rate": 1.7663751693558322e-06, + "loss": 0.2017, + "step": 33436 + }, + { + "epoch": 12.35, + "learning_rate": 1.7663382189924868e-06, + "loss": 0.4008, + "step": 33437 + }, + { + "epoch": 12.35, + "learning_rate": 1.7663012686291414e-06, + "loss": 0.2213, + "step": 33438 + }, + { + "epoch": 12.35, + "learning_rate": 1.7662643182657963e-06, + "loss": 0.5297, + "step": 33439 + }, + { + "epoch": 12.35, + "learning_rate": 1.766227367902451e-06, + "loss": 0.3215, + "step": 33440 + }, + { + "epoch": 12.35, + "learning_rate": 1.766190417539106e-06, + "loss": 0.3034, + "step": 33441 + }, + { + "epoch": 12.35, + "learning_rate": 1.7661534671757606e-06, + "loss": 0.258, + "step": 33442 + }, + { + "epoch": 12.35, + "learning_rate": 1.7661165168124155e-06, + "loss": 0.2744, + "step": 33443 + }, + { + "epoch": 12.35, + "learning_rate": 1.76607956644907e-06, + "loss": 0.3252, + "step": 33444 + }, + { + "epoch": 12.35, + "learning_rate": 1.766042616085725e-06, + "loss": 0.4286, + "step": 33445 + }, + { + "epoch": 12.35, + "learning_rate": 1.7660056657223796e-06, + "loss": 0.2309, + "step": 33446 + }, + { + "epoch": 12.35, + "learning_rate": 1.7659687153590342e-06, + "loss": 0.2458, + "step": 33447 + }, + { + "epoch": 12.35, + "learning_rate": 1.7659317649956893e-06, + "loss": 0.4876, + "step": 33448 + }, + { + "epoch": 12.35, + "learning_rate": 1.765894814632344e-06, + "loss": 0.2445, + "step": 33449 + }, + { + "epoch": 12.35, + "learning_rate": 1.7658578642689988e-06, + "loss": 0.1996, + "step": 33450 + }, + { + "epoch": 12.35, + "learning_rate": 1.7658209139056534e-06, + "loss": 0.3987, + "step": 33451 + }, + { + "epoch": 12.35, + "learning_rate": 1.7657839635423082e-06, + "loss": 0.2801, + "step": 33452 + }, + { + "epoch": 12.35, + "learning_rate": 1.7657470131789629e-06, + "loss": 0.3515, + "step": 33453 + }, + { + "epoch": 12.35, + "learning_rate": 1.765710062815618e-06, + "loss": 0.3709, + "step": 33454 + }, + { + "epoch": 12.35, + "learning_rate": 1.7656731124522726e-06, + "loss": 0.2376, + "step": 33455 + }, + { + "epoch": 12.35, + "learning_rate": 1.765636162088927e-06, + "loss": 0.2716, + "step": 33456 + }, + { + "epoch": 12.35, + "learning_rate": 1.765599211725582e-06, + "loss": 0.2907, + "step": 33457 + }, + { + "epoch": 12.36, + "learning_rate": 1.7655622613622367e-06, + "loss": 0.4188, + "step": 33458 + }, + { + "epoch": 12.36, + "learning_rate": 1.7655253109988915e-06, + "loss": 0.1586, + "step": 33459 + }, + { + "epoch": 12.36, + "learning_rate": 1.7654883606355462e-06, + "loss": 0.1899, + "step": 33460 + }, + { + "epoch": 12.36, + "learning_rate": 1.7654514102722012e-06, + "loss": 0.2265, + "step": 33461 + }, + { + "epoch": 12.36, + "learning_rate": 1.7654144599088557e-06, + "loss": 0.2254, + "step": 33462 + }, + { + "epoch": 12.36, + "learning_rate": 1.7653775095455107e-06, + "loss": 0.2481, + "step": 33463 + }, + { + "epoch": 12.36, + "learning_rate": 1.7653405591821654e-06, + "loss": 0.2342, + "step": 33464 + }, + { + "epoch": 12.36, + "learning_rate": 1.7653036088188202e-06, + "loss": 0.3361, + "step": 33465 + }, + { + "epoch": 12.36, + "learning_rate": 1.7652666584554748e-06, + "loss": 0.2337, + "step": 33466 + }, + { + "epoch": 12.36, + "learning_rate": 1.7652297080921295e-06, + "loss": 0.2494, + "step": 33467 + }, + { + "epoch": 12.36, + "learning_rate": 1.7651927577287843e-06, + "loss": 0.2216, + "step": 33468 + }, + { + "epoch": 12.36, + "learning_rate": 1.765155807365439e-06, + "loss": 0.3902, + "step": 33469 + }, + { + "epoch": 12.36, + "learning_rate": 1.765118857002094e-06, + "loss": 0.2939, + "step": 33470 + }, + { + "epoch": 12.36, + "learning_rate": 1.7650819066387487e-06, + "loss": 0.3747, + "step": 33471 + }, + { + "epoch": 12.36, + "learning_rate": 1.7650449562754035e-06, + "loss": 0.4163, + "step": 33472 + }, + { + "epoch": 12.36, + "learning_rate": 1.7650080059120581e-06, + "loss": 0.3854, + "step": 33473 + }, + { + "epoch": 12.36, + "learning_rate": 1.764971055548713e-06, + "loss": 0.2277, + "step": 33474 + }, + { + "epoch": 12.36, + "learning_rate": 1.7649341051853676e-06, + "loss": 0.3875, + "step": 33475 + }, + { + "epoch": 12.36, + "learning_rate": 1.7648971548220223e-06, + "loss": 0.2027, + "step": 33476 + }, + { + "epoch": 12.36, + "learning_rate": 1.7648602044586773e-06, + "loss": 0.1786, + "step": 33477 + }, + { + "epoch": 12.36, + "learning_rate": 1.7648232540953318e-06, + "loss": 0.3108, + "step": 33478 + }, + { + "epoch": 12.36, + "learning_rate": 1.7647863037319868e-06, + "loss": 0.4211, + "step": 33479 + }, + { + "epoch": 12.36, + "learning_rate": 1.7647493533686414e-06, + "loss": 0.1747, + "step": 33480 + }, + { + "epoch": 12.36, + "learning_rate": 1.7647124030052963e-06, + "loss": 0.3009, + "step": 33481 + }, + { + "epoch": 12.36, + "learning_rate": 1.764675452641951e-06, + "loss": 0.2426, + "step": 33482 + }, + { + "epoch": 12.36, + "learning_rate": 1.764638502278606e-06, + "loss": 0.4968, + "step": 33483 + }, + { + "epoch": 12.36, + "learning_rate": 1.7646015519152604e-06, + "loss": 0.2455, + "step": 33484 + }, + { + "epoch": 12.37, + "learning_rate": 1.7645646015519155e-06, + "loss": 0.2478, + "step": 33485 + }, + { + "epoch": 12.37, + "learning_rate": 1.7645276511885701e-06, + "loss": 0.2542, + "step": 33486 + }, + { + "epoch": 12.37, + "learning_rate": 1.7644907008252248e-06, + "loss": 0.4156, + "step": 33487 + }, + { + "epoch": 12.37, + "learning_rate": 1.7644537504618796e-06, + "loss": 0.3653, + "step": 33488 + }, + { + "epoch": 12.37, + "learning_rate": 1.7644168000985342e-06, + "loss": 0.2865, + "step": 33489 + }, + { + "epoch": 12.37, + "learning_rate": 1.764379849735189e-06, + "loss": 0.2645, + "step": 33490 + }, + { + "epoch": 12.37, + "learning_rate": 1.7643428993718437e-06, + "loss": 0.2173, + "step": 33491 + }, + { + "epoch": 12.37, + "learning_rate": 1.7643059490084988e-06, + "loss": 0.2215, + "step": 33492 + }, + { + "epoch": 12.37, + "learning_rate": 1.7642689986451534e-06, + "loss": 0.5527, + "step": 33493 + }, + { + "epoch": 12.37, + "learning_rate": 1.7642320482818083e-06, + "loss": 0.3169, + "step": 33494 + }, + { + "epoch": 12.37, + "learning_rate": 1.764195097918463e-06, + "loss": 0.3388, + "step": 33495 + }, + { + "epoch": 12.37, + "learning_rate": 1.7641581475551175e-06, + "loss": 0.2865, + "step": 33496 + }, + { + "epoch": 12.37, + "learning_rate": 1.7641211971917724e-06, + "loss": 0.3771, + "step": 33497 + }, + { + "epoch": 12.37, + "learning_rate": 1.764084246828427e-06, + "loss": 0.3218, + "step": 33498 + }, + { + "epoch": 12.37, + "learning_rate": 1.764047296465082e-06, + "loss": 0.425, + "step": 33499 + }, + { + "epoch": 12.37, + "learning_rate": 1.7640103461017367e-06, + "loss": 0.2878, + "step": 33500 + }, + { + "epoch": 12.37, + "learning_rate": 1.7639733957383916e-06, + "loss": 0.2525, + "step": 33501 + }, + { + "epoch": 12.37, + "learning_rate": 1.7639364453750462e-06, + "loss": 0.3816, + "step": 33502 + }, + { + "epoch": 12.37, + "learning_rate": 1.763899495011701e-06, + "loss": 0.2734, + "step": 33503 + }, + { + "epoch": 12.37, + "learning_rate": 1.7638625446483557e-06, + "loss": 0.2892, + "step": 33504 + }, + { + "epoch": 12.37, + "learning_rate": 1.7638255942850107e-06, + "loss": 0.3811, + "step": 33505 + }, + { + "epoch": 12.37, + "learning_rate": 1.7637886439216652e-06, + "loss": 0.2155, + "step": 33506 + }, + { + "epoch": 12.37, + "learning_rate": 1.7637516935583198e-06, + "loss": 0.2797, + "step": 33507 + }, + { + "epoch": 12.37, + "learning_rate": 1.7637147431949749e-06, + "loss": 0.2051, + "step": 33508 + }, + { + "epoch": 12.37, + "learning_rate": 1.7636777928316295e-06, + "loss": 0.297, + "step": 33509 + }, + { + "epoch": 12.37, + "learning_rate": 1.7636408424682844e-06, + "loss": 0.2388, + "step": 33510 + }, + { + "epoch": 12.37, + "learning_rate": 1.763603892104939e-06, + "loss": 0.2418, + "step": 33511 + }, + { + "epoch": 12.38, + "learning_rate": 1.7635669417415938e-06, + "loss": 0.3281, + "step": 33512 + }, + { + "epoch": 12.38, + "learning_rate": 1.7635299913782485e-06, + "loss": 0.3336, + "step": 33513 + }, + { + "epoch": 12.38, + "learning_rate": 1.7634930410149035e-06, + "loss": 0.2885, + "step": 33514 + }, + { + "epoch": 12.38, + "learning_rate": 1.7634560906515582e-06, + "loss": 0.3833, + "step": 33515 + }, + { + "epoch": 12.38, + "learning_rate": 1.7634191402882128e-06, + "loss": 0.3391, + "step": 33516 + }, + { + "epoch": 12.38, + "learning_rate": 1.7633821899248677e-06, + "loss": 0.3079, + "step": 33517 + }, + { + "epoch": 12.38, + "learning_rate": 1.7633452395615223e-06, + "loss": 0.1713, + "step": 33518 + }, + { + "epoch": 12.38, + "learning_rate": 1.7633082891981771e-06, + "loss": 0.3092, + "step": 33519 + }, + { + "epoch": 12.38, + "learning_rate": 1.7632713388348318e-06, + "loss": 0.3713, + "step": 33520 + }, + { + "epoch": 12.38, + "learning_rate": 1.7632343884714868e-06, + "loss": 0.2957, + "step": 33521 + }, + { + "epoch": 12.38, + "learning_rate": 1.7631974381081415e-06, + "loss": 0.3272, + "step": 33522 + }, + { + "epoch": 12.38, + "learning_rate": 1.7631604877447963e-06, + "loss": 0.3142, + "step": 33523 + }, + { + "epoch": 12.38, + "learning_rate": 1.763123537381451e-06, + "loss": 0.3586, + "step": 33524 + }, + { + "epoch": 12.38, + "learning_rate": 1.7630865870181058e-06, + "loss": 0.2934, + "step": 33525 + }, + { + "epoch": 12.38, + "learning_rate": 1.7630496366547604e-06, + "loss": 0.3924, + "step": 33526 + }, + { + "epoch": 12.38, + "learning_rate": 1.763012686291415e-06, + "loss": 0.2934, + "step": 33527 + }, + { + "epoch": 12.38, + "learning_rate": 1.7629757359280701e-06, + "loss": 0.2619, + "step": 33528 + }, + { + "epoch": 12.38, + "learning_rate": 1.7629387855647246e-06, + "loss": 0.3171, + "step": 33529 + }, + { + "epoch": 12.38, + "learning_rate": 1.7629018352013796e-06, + "loss": 0.3501, + "step": 33530 + }, + { + "epoch": 12.38, + "learning_rate": 1.7628648848380343e-06, + "loss": 0.2609, + "step": 33531 + }, + { + "epoch": 12.38, + "learning_rate": 1.7628279344746891e-06, + "loss": 0.2995, + "step": 33532 + }, + { + "epoch": 12.38, + "learning_rate": 1.7627909841113437e-06, + "loss": 0.3185, + "step": 33533 + }, + { + "epoch": 12.38, + "learning_rate": 1.7627540337479988e-06, + "loss": 0.3128, + "step": 33534 + }, + { + "epoch": 12.38, + "learning_rate": 1.7627170833846532e-06, + "loss": 0.3352, + "step": 33535 + }, + { + "epoch": 12.38, + "learning_rate": 1.7626801330213079e-06, + "loss": 0.4007, + "step": 33536 + }, + { + "epoch": 12.38, + "learning_rate": 1.762643182657963e-06, + "loss": 0.3786, + "step": 33537 + }, + { + "epoch": 12.38, + "learning_rate": 1.7626062322946176e-06, + "loss": 0.407, + "step": 33538 + }, + { + "epoch": 12.39, + "learning_rate": 1.7625692819312724e-06, + "loss": 0.2779, + "step": 33539 + }, + { + "epoch": 12.39, + "learning_rate": 1.762532331567927e-06, + "loss": 0.2957, + "step": 33540 + }, + { + "epoch": 12.39, + "learning_rate": 1.762495381204582e-06, + "loss": 0.2814, + "step": 33541 + }, + { + "epoch": 12.39, + "learning_rate": 1.7624584308412365e-06, + "loss": 0.3349, + "step": 33542 + }, + { + "epoch": 12.39, + "learning_rate": 1.7624214804778916e-06, + "loss": 0.2725, + "step": 33543 + }, + { + "epoch": 12.39, + "learning_rate": 1.7623845301145462e-06, + "loss": 0.4856, + "step": 33544 + }, + { + "epoch": 12.39, + "learning_rate": 1.762347579751201e-06, + "loss": 0.2766, + "step": 33545 + }, + { + "epoch": 12.39, + "learning_rate": 1.7623106293878557e-06, + "loss": 0.3102, + "step": 33546 + }, + { + "epoch": 12.39, + "learning_rate": 1.7622736790245103e-06, + "loss": 0.3148, + "step": 33547 + }, + { + "epoch": 12.39, + "learning_rate": 1.7622367286611652e-06, + "loss": 0.3629, + "step": 33548 + }, + { + "epoch": 12.39, + "learning_rate": 1.7621997782978198e-06, + "loss": 0.3384, + "step": 33549 + }, + { + "epoch": 12.39, + "learning_rate": 1.7621628279344749e-06, + "loss": 0.2722, + "step": 33550 + }, + { + "epoch": 12.39, + "learning_rate": 1.7621258775711293e-06, + "loss": 0.4293, + "step": 33551 + }, + { + "epoch": 12.39, + "learning_rate": 1.7620889272077844e-06, + "loss": 0.2315, + "step": 33552 + }, + { + "epoch": 12.39, + "learning_rate": 1.762051976844439e-06, + "loss": 0.2938, + "step": 33553 + }, + { + "epoch": 12.39, + "learning_rate": 1.7620150264810939e-06, + "loss": 0.3589, + "step": 33554 + }, + { + "epoch": 12.39, + "learning_rate": 1.7619780761177485e-06, + "loss": 0.3071, + "step": 33555 + }, + { + "epoch": 12.39, + "learning_rate": 1.7619411257544031e-06, + "loss": 0.2235, + "step": 33556 + }, + { + "epoch": 12.39, + "learning_rate": 1.761904175391058e-06, + "loss": 0.2227, + "step": 33557 + }, + { + "epoch": 12.39, + "learning_rate": 1.7618672250277126e-06, + "loss": 0.277, + "step": 33558 + }, + { + "epoch": 12.39, + "learning_rate": 1.7618302746643677e-06, + "loss": 0.2939, + "step": 33559 + }, + { + "epoch": 12.39, + "learning_rate": 1.7617933243010223e-06, + "loss": 0.3625, + "step": 33560 + }, + { + "epoch": 12.39, + "learning_rate": 1.7617563739376772e-06, + "loss": 0.3022, + "step": 33561 + }, + { + "epoch": 12.39, + "learning_rate": 1.7617194235743318e-06, + "loss": 0.3553, + "step": 33562 + }, + { + "epoch": 12.39, + "learning_rate": 1.7616824732109866e-06, + "loss": 0.2267, + "step": 33563 + }, + { + "epoch": 12.39, + "learning_rate": 1.7616455228476413e-06, + "loss": 0.1803, + "step": 33564 + }, + { + "epoch": 12.39, + "learning_rate": 1.7616085724842963e-06, + "loss": 0.3606, + "step": 33565 + }, + { + "epoch": 12.4, + "learning_rate": 1.761571622120951e-06, + "loss": 0.3331, + "step": 33566 + }, + { + "epoch": 12.4, + "learning_rate": 1.7615346717576056e-06, + "loss": 0.3179, + "step": 33567 + }, + { + "epoch": 12.4, + "learning_rate": 1.7614977213942605e-06, + "loss": 0.2652, + "step": 33568 + }, + { + "epoch": 12.4, + "learning_rate": 1.761460771030915e-06, + "loss": 0.3088, + "step": 33569 + }, + { + "epoch": 12.4, + "learning_rate": 1.76142382066757e-06, + "loss": 0.2021, + "step": 33570 + }, + { + "epoch": 12.4, + "learning_rate": 1.7613868703042246e-06, + "loss": 0.1412, + "step": 33571 + }, + { + "epoch": 12.4, + "learning_rate": 1.7613499199408796e-06, + "loss": 0.3391, + "step": 33572 + }, + { + "epoch": 12.4, + "learning_rate": 1.7613129695775343e-06, + "loss": 0.2172, + "step": 33573 + }, + { + "epoch": 12.4, + "learning_rate": 1.7612760192141891e-06, + "loss": 0.2633, + "step": 33574 + }, + { + "epoch": 12.4, + "learning_rate": 1.7612390688508438e-06, + "loss": 0.2366, + "step": 33575 + }, + { + "epoch": 12.4, + "learning_rate": 1.7612021184874984e-06, + "loss": 0.263, + "step": 33576 + }, + { + "epoch": 12.4, + "learning_rate": 1.7611651681241533e-06, + "loss": 0.2197, + "step": 33577 + }, + { + "epoch": 12.4, + "learning_rate": 1.7611282177608079e-06, + "loss": 0.3118, + "step": 33578 + }, + { + "epoch": 12.4, + "learning_rate": 1.761091267397463e-06, + "loss": 0.1655, + "step": 33579 + }, + { + "epoch": 12.4, + "learning_rate": 1.7610543170341174e-06, + "loss": 0.2542, + "step": 33580 + }, + { + "epoch": 12.4, + "learning_rate": 1.7610173666707724e-06, + "loss": 0.3926, + "step": 33581 + }, + { + "epoch": 12.4, + "learning_rate": 1.760980416307427e-06, + "loss": 0.3016, + "step": 33582 + }, + { + "epoch": 12.4, + "learning_rate": 1.760943465944082e-06, + "loss": 0.281, + "step": 33583 + }, + { + "epoch": 12.4, + "learning_rate": 1.7609065155807366e-06, + "loss": 0.3736, + "step": 33584 + }, + { + "epoch": 12.4, + "learning_rate": 1.7608695652173914e-06, + "loss": 0.3143, + "step": 33585 + }, + { + "epoch": 12.4, + "learning_rate": 1.760832614854046e-06, + "loss": 0.2772, + "step": 33586 + }, + { + "epoch": 12.4, + "learning_rate": 1.7607956644907007e-06, + "loss": 0.3088, + "step": 33587 + }, + { + "epoch": 12.4, + "learning_rate": 1.7607587141273557e-06, + "loss": 0.4084, + "step": 33588 + }, + { + "epoch": 12.4, + "learning_rate": 1.7607217637640104e-06, + "loss": 0.2203, + "step": 33589 + }, + { + "epoch": 12.4, + "learning_rate": 1.7606848134006652e-06, + "loss": 0.4621, + "step": 33590 + }, + { + "epoch": 12.4, + "learning_rate": 1.7606478630373199e-06, + "loss": 0.2314, + "step": 33591 + }, + { + "epoch": 12.4, + "learning_rate": 1.7606109126739747e-06, + "loss": 0.3636, + "step": 33592 + }, + { + "epoch": 12.41, + "learning_rate": 1.7605739623106293e-06, + "loss": 0.302, + "step": 33593 + }, + { + "epoch": 12.41, + "learning_rate": 1.7605370119472844e-06, + "loss": 0.2578, + "step": 33594 + }, + { + "epoch": 12.41, + "learning_rate": 1.760500061583939e-06, + "loss": 0.3726, + "step": 33595 + }, + { + "epoch": 12.41, + "learning_rate": 1.7604631112205935e-06, + "loss": 0.3855, + "step": 33596 + }, + { + "epoch": 12.41, + "learning_rate": 1.7604261608572485e-06, + "loss": 0.2619, + "step": 33597 + }, + { + "epoch": 12.41, + "learning_rate": 1.7603892104939032e-06, + "loss": 0.2204, + "step": 33598 + }, + { + "epoch": 12.41, + "learning_rate": 1.760352260130558e-06, + "loss": 0.2165, + "step": 33599 + }, + { + "epoch": 12.41, + "learning_rate": 1.7603153097672126e-06, + "loss": 0.2955, + "step": 33600 + }, + { + "epoch": 12.41, + "learning_rate": 1.7602783594038677e-06, + "loss": 0.2585, + "step": 33601 + }, + { + "epoch": 12.41, + "learning_rate": 1.7602414090405221e-06, + "loss": 0.2724, + "step": 33602 + }, + { + "epoch": 12.41, + "learning_rate": 1.7602044586771772e-06, + "loss": 0.2663, + "step": 33603 + }, + { + "epoch": 12.41, + "learning_rate": 1.7601675083138318e-06, + "loss": 0.2241, + "step": 33604 + }, + { + "epoch": 12.41, + "learning_rate": 1.7601305579504867e-06, + "loss": 0.3007, + "step": 33605 + }, + { + "epoch": 12.41, + "learning_rate": 1.7600936075871413e-06, + "loss": 0.3073, + "step": 33606 + }, + { + "epoch": 12.41, + "learning_rate": 1.760056657223796e-06, + "loss": 0.4376, + "step": 33607 + }, + { + "epoch": 12.41, + "learning_rate": 1.7600197068604508e-06, + "loss": 0.3626, + "step": 33608 + }, + { + "epoch": 12.41, + "learning_rate": 1.7599827564971054e-06, + "loss": 0.5012, + "step": 33609 + }, + { + "epoch": 12.41, + "learning_rate": 1.7599458061337605e-06, + "loss": 0.2074, + "step": 33610 + }, + { + "epoch": 12.41, + "learning_rate": 1.7599088557704151e-06, + "loss": 0.3455, + "step": 33611 + }, + { + "epoch": 12.41, + "learning_rate": 1.75987190540707e-06, + "loss": 0.2109, + "step": 33612 + }, + { + "epoch": 12.41, + "learning_rate": 1.7598349550437246e-06, + "loss": 0.2374, + "step": 33613 + }, + { + "epoch": 12.41, + "learning_rate": 1.7597980046803795e-06, + "loss": 0.2179, + "step": 33614 + }, + { + "epoch": 12.41, + "learning_rate": 1.759761054317034e-06, + "loss": 0.2101, + "step": 33615 + }, + { + "epoch": 12.41, + "learning_rate": 1.7597241039536887e-06, + "loss": 0.2688, + "step": 33616 + }, + { + "epoch": 12.41, + "learning_rate": 1.7596871535903438e-06, + "loss": 0.3762, + "step": 33617 + }, + { + "epoch": 12.41, + "learning_rate": 1.7596502032269982e-06, + "loss": 0.2424, + "step": 33618 + }, + { + "epoch": 12.41, + "learning_rate": 1.7596132528636533e-06, + "loss": 0.324, + "step": 33619 + }, + { + "epoch": 12.42, + "learning_rate": 1.759576302500308e-06, + "loss": 0.2915, + "step": 33620 + }, + { + "epoch": 12.42, + "learning_rate": 1.7595393521369628e-06, + "loss": 0.3984, + "step": 33621 + }, + { + "epoch": 12.42, + "learning_rate": 1.7595024017736174e-06, + "loss": 0.1807, + "step": 33622 + }, + { + "epoch": 12.42, + "learning_rate": 1.7594654514102725e-06, + "loss": 0.311, + "step": 33623 + }, + { + "epoch": 12.42, + "learning_rate": 1.7594285010469269e-06, + "loss": 0.3294, + "step": 33624 + }, + { + "epoch": 12.42, + "learning_rate": 1.759391550683582e-06, + "loss": 0.3208, + "step": 33625 + }, + { + "epoch": 12.42, + "learning_rate": 1.7593546003202366e-06, + "loss": 0.2676, + "step": 33626 + }, + { + "epoch": 12.42, + "learning_rate": 1.7593176499568912e-06, + "loss": 0.2897, + "step": 33627 + }, + { + "epoch": 12.42, + "learning_rate": 1.759280699593546e-06, + "loss": 0.2792, + "step": 33628 + }, + { + "epoch": 12.42, + "learning_rate": 1.7592437492302007e-06, + "loss": 0.328, + "step": 33629 + }, + { + "epoch": 12.42, + "learning_rate": 1.7592067988668555e-06, + "loss": 0.2282, + "step": 33630 + }, + { + "epoch": 12.42, + "learning_rate": 1.7591698485035102e-06, + "loss": 0.1714, + "step": 33631 + }, + { + "epoch": 12.42, + "learning_rate": 1.7591328981401652e-06, + "loss": 0.2591, + "step": 33632 + }, + { + "epoch": 12.42, + "learning_rate": 1.7590959477768199e-06, + "loss": 0.3944, + "step": 33633 + }, + { + "epoch": 12.42, + "learning_rate": 1.7590589974134747e-06, + "loss": 0.2818, + "step": 33634 + }, + { + "epoch": 12.42, + "learning_rate": 1.7590220470501294e-06, + "loss": 0.2263, + "step": 33635 + }, + { + "epoch": 12.42, + "learning_rate": 1.758985096686784e-06, + "loss": 0.3492, + "step": 33636 + }, + { + "epoch": 12.42, + "learning_rate": 1.7589481463234389e-06, + "loss": 0.3665, + "step": 33637 + }, + { + "epoch": 12.42, + "learning_rate": 1.7589111959600935e-06, + "loss": 0.2807, + "step": 33638 + }, + { + "epoch": 12.42, + "learning_rate": 1.7588742455967485e-06, + "loss": 0.2985, + "step": 33639 + }, + { + "epoch": 12.42, + "learning_rate": 1.7588372952334032e-06, + "loss": 0.1696, + "step": 33640 + }, + { + "epoch": 12.42, + "learning_rate": 1.758800344870058e-06, + "loss": 0.3516, + "step": 33641 + }, + { + "epoch": 12.42, + "learning_rate": 1.7587633945067127e-06, + "loss": 0.2855, + "step": 33642 + }, + { + "epoch": 12.42, + "learning_rate": 1.7587264441433675e-06, + "loss": 0.4042, + "step": 33643 + }, + { + "epoch": 12.42, + "learning_rate": 1.7586894937800222e-06, + "loss": 0.2489, + "step": 33644 + }, + { + "epoch": 12.42, + "learning_rate": 1.7586525434166772e-06, + "loss": 0.2424, + "step": 33645 + }, + { + "epoch": 12.42, + "learning_rate": 1.7586155930533318e-06, + "loss": 0.2618, + "step": 33646 + }, + { + "epoch": 12.43, + "learning_rate": 1.7585786426899863e-06, + "loss": 0.3148, + "step": 33647 + }, + { + "epoch": 12.43, + "learning_rate": 1.7585416923266413e-06, + "loss": 0.2975, + "step": 33648 + }, + { + "epoch": 12.43, + "learning_rate": 1.758504741963296e-06, + "loss": 0.4125, + "step": 33649 + }, + { + "epoch": 12.43, + "learning_rate": 1.7584677915999508e-06, + "loss": 0.2381, + "step": 33650 + }, + { + "epoch": 12.43, + "learning_rate": 1.7584308412366055e-06, + "loss": 0.3482, + "step": 33651 + }, + { + "epoch": 12.43, + "learning_rate": 1.7583938908732605e-06, + "loss": 0.2818, + "step": 33652 + }, + { + "epoch": 12.43, + "learning_rate": 1.758356940509915e-06, + "loss": 0.2016, + "step": 33653 + }, + { + "epoch": 12.43, + "learning_rate": 1.75831999014657e-06, + "loss": 0.2487, + "step": 33654 + }, + { + "epoch": 12.43, + "learning_rate": 1.7582830397832246e-06, + "loss": 0.3299, + "step": 33655 + }, + { + "epoch": 12.43, + "learning_rate": 1.7582460894198793e-06, + "loss": 0.3934, + "step": 33656 + }, + { + "epoch": 12.43, + "learning_rate": 1.7582091390565341e-06, + "loss": 0.2334, + "step": 33657 + }, + { + "epoch": 12.43, + "learning_rate": 1.7581721886931888e-06, + "loss": 0.3812, + "step": 33658 + }, + { + "epoch": 12.43, + "learning_rate": 1.7581352383298436e-06, + "loss": 0.1934, + "step": 33659 + }, + { + "epoch": 12.43, + "learning_rate": 1.7580982879664982e-06, + "loss": 0.3625, + "step": 33660 + }, + { + "epoch": 12.43, + "learning_rate": 1.7580613376031533e-06, + "loss": 0.4087, + "step": 33661 + }, + { + "epoch": 12.43, + "learning_rate": 1.758024387239808e-06, + "loss": 0.2513, + "step": 33662 + }, + { + "epoch": 12.43, + "learning_rate": 1.7579874368764628e-06, + "loss": 0.2432, + "step": 33663 + }, + { + "epoch": 12.43, + "learning_rate": 1.7579504865131174e-06, + "loss": 0.3371, + "step": 33664 + }, + { + "epoch": 12.43, + "learning_rate": 1.7579135361497723e-06, + "loss": 0.2052, + "step": 33665 + }, + { + "epoch": 12.43, + "learning_rate": 1.757876585786427e-06, + "loss": 0.3935, + "step": 33666 + }, + { + "epoch": 12.43, + "learning_rate": 1.7578396354230815e-06, + "loss": 0.2766, + "step": 33667 + }, + { + "epoch": 12.43, + "learning_rate": 1.7578026850597366e-06, + "loss": 0.4575, + "step": 33668 + }, + { + "epoch": 12.43, + "learning_rate": 1.757765734696391e-06, + "loss": 0.4013, + "step": 33669 + }, + { + "epoch": 12.43, + "learning_rate": 1.757728784333046e-06, + "loss": 0.3938, + "step": 33670 + }, + { + "epoch": 12.43, + "learning_rate": 1.7576918339697007e-06, + "loss": 0.2968, + "step": 33671 + }, + { + "epoch": 12.43, + "learning_rate": 1.7576548836063556e-06, + "loss": 0.2994, + "step": 33672 + }, + { + "epoch": 12.43, + "learning_rate": 1.7576179332430102e-06, + "loss": 0.2399, + "step": 33673 + }, + { + "epoch": 12.44, + "learning_rate": 1.7575809828796653e-06, + "loss": 0.2541, + "step": 33674 + }, + { + "epoch": 12.44, + "learning_rate": 1.7575440325163197e-06, + "loss": 0.3088, + "step": 33675 + }, + { + "epoch": 12.44, + "learning_rate": 1.7575070821529743e-06, + "loss": 0.1925, + "step": 33676 + }, + { + "epoch": 12.44, + "learning_rate": 1.7574701317896294e-06, + "loss": 0.3151, + "step": 33677 + }, + { + "epoch": 12.44, + "learning_rate": 1.757433181426284e-06, + "loss": 0.339, + "step": 33678 + }, + { + "epoch": 12.44, + "learning_rate": 1.7573962310629389e-06, + "loss": 0.4248, + "step": 33679 + }, + { + "epoch": 12.44, + "learning_rate": 1.7573592806995935e-06, + "loss": 0.2855, + "step": 33680 + }, + { + "epoch": 12.44, + "learning_rate": 1.7573223303362484e-06, + "loss": 0.4121, + "step": 33681 + }, + { + "epoch": 12.44, + "learning_rate": 1.757285379972903e-06, + "loss": 0.3272, + "step": 33682 + }, + { + "epoch": 12.44, + "learning_rate": 1.757248429609558e-06, + "loss": 0.2984, + "step": 33683 + }, + { + "epoch": 12.44, + "learning_rate": 1.7572114792462127e-06, + "loss": 0.2108, + "step": 33684 + }, + { + "epoch": 12.44, + "learning_rate": 1.7571745288828675e-06, + "loss": 0.3032, + "step": 33685 + }, + { + "epoch": 12.44, + "learning_rate": 1.7571375785195222e-06, + "loss": 0.1655, + "step": 33686 + }, + { + "epoch": 12.44, + "learning_rate": 1.7571006281561768e-06, + "loss": 0.2685, + "step": 33687 + }, + { + "epoch": 12.44, + "learning_rate": 1.7570636777928317e-06, + "loss": 0.2157, + "step": 33688 + }, + { + "epoch": 12.44, + "learning_rate": 1.7570267274294863e-06, + "loss": 0.381, + "step": 33689 + }, + { + "epoch": 12.44, + "learning_rate": 1.7569897770661414e-06, + "loss": 0.2355, + "step": 33690 + }, + { + "epoch": 12.44, + "learning_rate": 1.7569528267027958e-06, + "loss": 0.2388, + "step": 33691 + }, + { + "epoch": 12.44, + "learning_rate": 1.7569158763394508e-06, + "loss": 0.3201, + "step": 33692 + }, + { + "epoch": 12.44, + "learning_rate": 1.7568789259761055e-06, + "loss": 0.2021, + "step": 33693 + }, + { + "epoch": 12.44, + "learning_rate": 1.7568419756127603e-06, + "loss": 0.2406, + "step": 33694 + }, + { + "epoch": 12.44, + "learning_rate": 1.756805025249415e-06, + "loss": 0.4227, + "step": 33695 + }, + { + "epoch": 12.44, + "learning_rate": 1.7567680748860696e-06, + "loss": 0.2978, + "step": 33696 + }, + { + "epoch": 12.44, + "learning_rate": 1.7567311245227244e-06, + "loss": 0.2626, + "step": 33697 + }, + { + "epoch": 12.44, + "learning_rate": 1.756694174159379e-06, + "loss": 0.3292, + "step": 33698 + }, + { + "epoch": 12.44, + "learning_rate": 1.7566572237960341e-06, + "loss": 0.3557, + "step": 33699 + }, + { + "epoch": 12.44, + "learning_rate": 1.7566202734326888e-06, + "loss": 0.3008, + "step": 33700 + }, + { + "epoch": 12.44, + "learning_rate": 1.7565833230693436e-06, + "loss": 0.2862, + "step": 33701 + }, + { + "epoch": 12.45, + "learning_rate": 1.7565463727059983e-06, + "loss": 0.2084, + "step": 33702 + }, + { + "epoch": 12.45, + "learning_rate": 1.7565094223426531e-06, + "loss": 0.2555, + "step": 33703 + }, + { + "epoch": 12.45, + "learning_rate": 1.7564724719793078e-06, + "loss": 0.2069, + "step": 33704 + }, + { + "epoch": 12.45, + "learning_rate": 1.7564355216159624e-06, + "loss": 0.2293, + "step": 33705 + }, + { + "epoch": 12.45, + "learning_rate": 1.7563985712526174e-06, + "loss": 0.2934, + "step": 33706 + }, + { + "epoch": 12.45, + "learning_rate": 1.756361620889272e-06, + "loss": 0.3162, + "step": 33707 + }, + { + "epoch": 12.45, + "learning_rate": 1.756324670525927e-06, + "loss": 0.2952, + "step": 33708 + }, + { + "epoch": 12.45, + "learning_rate": 1.7562877201625816e-06, + "loss": 0.2457, + "step": 33709 + }, + { + "epoch": 12.45, + "learning_rate": 1.7562507697992364e-06, + "loss": 0.3551, + "step": 33710 + }, + { + "epoch": 12.45, + "learning_rate": 1.756213819435891e-06, + "loss": 0.2347, + "step": 33711 + }, + { + "epoch": 12.45, + "learning_rate": 1.7561768690725461e-06, + "loss": 0.3044, + "step": 33712 + }, + { + "epoch": 12.45, + "learning_rate": 1.7561399187092007e-06, + "loss": 0.3106, + "step": 33713 + }, + { + "epoch": 12.45, + "learning_rate": 1.7561029683458556e-06, + "loss": 0.293, + "step": 33714 + }, + { + "epoch": 12.45, + "learning_rate": 1.7560660179825102e-06, + "loss": 0.2195, + "step": 33715 + }, + { + "epoch": 12.45, + "learning_rate": 1.7560290676191649e-06, + "loss": 0.46, + "step": 33716 + }, + { + "epoch": 12.45, + "learning_rate": 1.7559921172558197e-06, + "loss": 0.3562, + "step": 33717 + }, + { + "epoch": 12.45, + "learning_rate": 1.7559551668924744e-06, + "loss": 0.239, + "step": 33718 + }, + { + "epoch": 12.45, + "learning_rate": 1.7559182165291294e-06, + "loss": 0.3863, + "step": 33719 + }, + { + "epoch": 12.45, + "learning_rate": 1.7558812661657838e-06, + "loss": 0.2653, + "step": 33720 + }, + { + "epoch": 12.45, + "learning_rate": 1.755844315802439e-06, + "loss": 0.3375, + "step": 33721 + }, + { + "epoch": 12.45, + "learning_rate": 1.7558073654390935e-06, + "loss": 0.3913, + "step": 33722 + }, + { + "epoch": 12.45, + "learning_rate": 1.7557704150757484e-06, + "loss": 0.4608, + "step": 33723 + }, + { + "epoch": 12.45, + "learning_rate": 1.755733464712403e-06, + "loss": 0.3376, + "step": 33724 + }, + { + "epoch": 12.45, + "learning_rate": 1.7556965143490577e-06, + "loss": 0.267, + "step": 33725 + }, + { + "epoch": 12.45, + "learning_rate": 1.7556595639857125e-06, + "loss": 0.3067, + "step": 33726 + }, + { + "epoch": 12.45, + "learning_rate": 1.7556226136223671e-06, + "loss": 0.3339, + "step": 33727 + }, + { + "epoch": 12.45, + "learning_rate": 1.7555856632590222e-06, + "loss": 0.3325, + "step": 33728 + }, + { + "epoch": 12.46, + "learning_rate": 1.7555487128956768e-06, + "loss": 0.343, + "step": 33729 + }, + { + "epoch": 12.46, + "learning_rate": 1.7555117625323317e-06, + "loss": 0.2684, + "step": 33730 + }, + { + "epoch": 12.46, + "learning_rate": 1.7554748121689863e-06, + "loss": 0.2521, + "step": 33731 + }, + { + "epoch": 12.46, + "learning_rate": 1.7554378618056412e-06, + "loss": 0.2703, + "step": 33732 + }, + { + "epoch": 12.46, + "learning_rate": 1.7554009114422958e-06, + "loss": 0.2851, + "step": 33733 + }, + { + "epoch": 12.46, + "learning_rate": 1.7553639610789509e-06, + "loss": 0.3136, + "step": 33734 + }, + { + "epoch": 12.46, + "learning_rate": 1.7553270107156055e-06, + "loss": 0.3961, + "step": 33735 + }, + { + "epoch": 12.46, + "learning_rate": 1.75529006035226e-06, + "loss": 0.3423, + "step": 33736 + }, + { + "epoch": 12.46, + "learning_rate": 1.755253109988915e-06, + "loss": 0.3533, + "step": 33737 + }, + { + "epoch": 12.46, + "learning_rate": 1.7552161596255696e-06, + "loss": 0.2043, + "step": 33738 + }, + { + "epoch": 12.46, + "learning_rate": 1.7551792092622245e-06, + "loss": 0.4423, + "step": 33739 + }, + { + "epoch": 12.46, + "learning_rate": 1.7551422588988791e-06, + "loss": 0.3434, + "step": 33740 + }, + { + "epoch": 12.46, + "learning_rate": 1.7551053085355342e-06, + "loss": 0.336, + "step": 33741 + }, + { + "epoch": 12.46, + "learning_rate": 1.7550683581721886e-06, + "loss": 0.3964, + "step": 33742 + }, + { + "epoch": 12.46, + "learning_rate": 1.7550314078088437e-06, + "loss": 0.215, + "step": 33743 + }, + { + "epoch": 12.46, + "learning_rate": 1.7549944574454983e-06, + "loss": 0.2928, + "step": 33744 + }, + { + "epoch": 12.46, + "learning_rate": 1.754957507082153e-06, + "loss": 0.1145, + "step": 33745 + }, + { + "epoch": 12.46, + "learning_rate": 1.7549205567188078e-06, + "loss": 0.1418, + "step": 33746 + }, + { + "epoch": 12.46, + "learning_rate": 1.7548836063554624e-06, + "loss": 0.2325, + "step": 33747 + }, + { + "epoch": 12.46, + "learning_rate": 1.7548466559921173e-06, + "loss": 0.2871, + "step": 33748 + }, + { + "epoch": 12.46, + "learning_rate": 1.754809705628772e-06, + "loss": 0.2729, + "step": 33749 + }, + { + "epoch": 12.46, + "learning_rate": 1.754772755265427e-06, + "loss": 0.2486, + "step": 33750 + }, + { + "epoch": 12.46, + "learning_rate": 1.7547358049020816e-06, + "loss": 0.2955, + "step": 33751 + }, + { + "epoch": 12.46, + "learning_rate": 1.7546988545387364e-06, + "loss": 0.3279, + "step": 33752 + }, + { + "epoch": 12.46, + "learning_rate": 1.754661904175391e-06, + "loss": 0.3029, + "step": 33753 + }, + { + "epoch": 12.46, + "learning_rate": 1.754624953812046e-06, + "loss": 0.1776, + "step": 33754 + }, + { + "epoch": 12.46, + "learning_rate": 1.7545880034487006e-06, + "loss": 0.2001, + "step": 33755 + }, + { + "epoch": 12.47, + "learning_rate": 1.7545510530853552e-06, + "loss": 0.2878, + "step": 33756 + }, + { + "epoch": 12.47, + "learning_rate": 1.7545141027220103e-06, + "loss": 0.259, + "step": 33757 + }, + { + "epoch": 12.47, + "learning_rate": 1.754477152358665e-06, + "loss": 0.427, + "step": 33758 + }, + { + "epoch": 12.47, + "learning_rate": 1.7544402019953197e-06, + "loss": 0.1477, + "step": 33759 + }, + { + "epoch": 12.47, + "learning_rate": 1.7544032516319744e-06, + "loss": 0.1851, + "step": 33760 + }, + { + "epoch": 12.47, + "learning_rate": 1.7543663012686292e-06, + "loss": 0.4354, + "step": 33761 + }, + { + "epoch": 12.47, + "learning_rate": 1.7543293509052839e-06, + "loss": 0.4348, + "step": 33762 + }, + { + "epoch": 12.47, + "learning_rate": 1.754292400541939e-06, + "loss": 0.35, + "step": 33763 + }, + { + "epoch": 12.47, + "learning_rate": 1.7542554501785936e-06, + "loss": 0.2737, + "step": 33764 + }, + { + "epoch": 12.47, + "learning_rate": 1.754218499815248e-06, + "loss": 0.2893, + "step": 33765 + }, + { + "epoch": 12.47, + "learning_rate": 1.754181549451903e-06, + "loss": 0.2835, + "step": 33766 + }, + { + "epoch": 12.47, + "learning_rate": 1.7541445990885577e-06, + "loss": 0.2141, + "step": 33767 + }, + { + "epoch": 12.47, + "learning_rate": 1.7541076487252125e-06, + "loss": 0.4175, + "step": 33768 + }, + { + "epoch": 12.47, + "learning_rate": 1.7540706983618672e-06, + "loss": 0.2668, + "step": 33769 + }, + { + "epoch": 12.47, + "learning_rate": 1.754033747998522e-06, + "loss": 0.3713, + "step": 33770 + }, + { + "epoch": 12.47, + "learning_rate": 1.7539967976351767e-06, + "loss": 0.2307, + "step": 33771 + }, + { + "epoch": 12.47, + "learning_rate": 1.7539598472718317e-06, + "loss": 0.301, + "step": 33772 + }, + { + "epoch": 12.47, + "learning_rate": 1.7539228969084863e-06, + "loss": 0.2056, + "step": 33773 + }, + { + "epoch": 12.47, + "learning_rate": 1.7538859465451412e-06, + "loss": 0.2595, + "step": 33774 + }, + { + "epoch": 12.47, + "learning_rate": 1.7538489961817958e-06, + "loss": 0.2396, + "step": 33775 + }, + { + "epoch": 12.47, + "learning_rate": 1.7538120458184505e-06, + "loss": 0.3789, + "step": 33776 + }, + { + "epoch": 12.47, + "learning_rate": 1.7537750954551053e-06, + "loss": 0.306, + "step": 33777 + }, + { + "epoch": 12.47, + "learning_rate": 1.75373814509176e-06, + "loss": 0.3022, + "step": 33778 + }, + { + "epoch": 12.47, + "learning_rate": 1.753701194728415e-06, + "loss": 0.3033, + "step": 33779 + }, + { + "epoch": 12.47, + "learning_rate": 1.7536642443650696e-06, + "loss": 0.3043, + "step": 33780 + }, + { + "epoch": 12.47, + "learning_rate": 1.7536272940017245e-06, + "loss": 0.3538, + "step": 33781 + }, + { + "epoch": 12.47, + "learning_rate": 1.7535903436383791e-06, + "loss": 0.2978, + "step": 33782 + }, + { + "epoch": 12.48, + "learning_rate": 1.753553393275034e-06, + "loss": 0.3248, + "step": 33783 + }, + { + "epoch": 12.48, + "learning_rate": 1.7535164429116886e-06, + "loss": 0.3297, + "step": 33784 + }, + { + "epoch": 12.48, + "learning_rate": 1.7534794925483433e-06, + "loss": 0.1917, + "step": 33785 + }, + { + "epoch": 12.48, + "learning_rate": 1.7534425421849983e-06, + "loss": 0.1725, + "step": 33786 + }, + { + "epoch": 12.48, + "learning_rate": 1.7534055918216527e-06, + "loss": 0.3493, + "step": 33787 + }, + { + "epoch": 12.48, + "learning_rate": 1.7533686414583078e-06, + "loss": 0.3697, + "step": 33788 + }, + { + "epoch": 12.48, + "learning_rate": 1.7533316910949624e-06, + "loss": 0.3459, + "step": 33789 + }, + { + "epoch": 12.48, + "learning_rate": 1.7532947407316173e-06, + "loss": 0.316, + "step": 33790 + }, + { + "epoch": 12.48, + "learning_rate": 1.753257790368272e-06, + "loss": 0.3523, + "step": 33791 + }, + { + "epoch": 12.48, + "learning_rate": 1.753220840004927e-06, + "loss": 0.5058, + "step": 33792 + }, + { + "epoch": 12.48, + "learning_rate": 1.7531838896415814e-06, + "loss": 0.4009, + "step": 33793 + }, + { + "epoch": 12.48, + "learning_rate": 1.7531469392782365e-06, + "loss": 0.3696, + "step": 33794 + }, + { + "epoch": 12.48, + "learning_rate": 1.753109988914891e-06, + "loss": 0.3058, + "step": 33795 + }, + { + "epoch": 12.48, + "learning_rate": 1.7530730385515457e-06, + "loss": 0.2319, + "step": 33796 + }, + { + "epoch": 12.48, + "learning_rate": 1.7530360881882006e-06, + "loss": 0.382, + "step": 33797 + }, + { + "epoch": 12.48, + "learning_rate": 1.7529991378248552e-06, + "loss": 0.4102, + "step": 33798 + }, + { + "epoch": 12.48, + "learning_rate": 1.75296218746151e-06, + "loss": 0.4025, + "step": 33799 + }, + { + "epoch": 12.48, + "learning_rate": 1.7529252370981647e-06, + "loss": 0.3379, + "step": 33800 + }, + { + "epoch": 12.48, + "learning_rate": 1.7528882867348198e-06, + "loss": 0.2724, + "step": 33801 + }, + { + "epoch": 12.48, + "learning_rate": 1.7528513363714744e-06, + "loss": 0.4536, + "step": 33802 + }, + { + "epoch": 12.48, + "learning_rate": 1.7528143860081293e-06, + "loss": 0.4005, + "step": 33803 + }, + { + "epoch": 12.48, + "learning_rate": 1.7527774356447839e-06, + "loss": 0.3806, + "step": 33804 + }, + { + "epoch": 12.48, + "learning_rate": 1.7527404852814385e-06, + "loss": 0.2789, + "step": 33805 + }, + { + "epoch": 12.48, + "learning_rate": 1.7527035349180934e-06, + "loss": 0.2877, + "step": 33806 + }, + { + "epoch": 12.48, + "learning_rate": 1.752666584554748e-06, + "loss": 0.3121, + "step": 33807 + }, + { + "epoch": 12.48, + "learning_rate": 1.752629634191403e-06, + "loss": 0.2564, + "step": 33808 + }, + { + "epoch": 12.48, + "learning_rate": 1.7525926838280575e-06, + "loss": 0.2165, + "step": 33809 + }, + { + "epoch": 12.49, + "learning_rate": 1.7525557334647126e-06, + "loss": 0.272, + "step": 33810 + }, + { + "epoch": 12.49, + "learning_rate": 1.7525187831013672e-06, + "loss": 0.4585, + "step": 33811 + }, + { + "epoch": 12.49, + "learning_rate": 1.752481832738022e-06, + "loss": 0.2774, + "step": 33812 + }, + { + "epoch": 12.49, + "learning_rate": 1.7524448823746767e-06, + "loss": 0.3736, + "step": 33813 + }, + { + "epoch": 12.49, + "learning_rate": 1.7524079320113317e-06, + "loss": 0.2728, + "step": 33814 + }, + { + "epoch": 12.49, + "learning_rate": 1.7523709816479862e-06, + "loss": 0.3611, + "step": 33815 + }, + { + "epoch": 12.49, + "learning_rate": 1.7523340312846408e-06, + "loss": 0.3485, + "step": 33816 + }, + { + "epoch": 12.49, + "learning_rate": 1.7522970809212959e-06, + "loss": 0.2217, + "step": 33817 + }, + { + "epoch": 12.49, + "learning_rate": 1.7522601305579505e-06, + "loss": 0.1784, + "step": 33818 + }, + { + "epoch": 12.49, + "learning_rate": 1.7522231801946053e-06, + "loss": 0.2251, + "step": 33819 + }, + { + "epoch": 12.49, + "learning_rate": 1.75218622983126e-06, + "loss": 0.2294, + "step": 33820 + }, + { + "epoch": 12.49, + "learning_rate": 1.7521492794679148e-06, + "loss": 0.3231, + "step": 33821 + }, + { + "epoch": 12.49, + "learning_rate": 1.7521123291045695e-06, + "loss": 0.2421, + "step": 33822 + }, + { + "epoch": 12.49, + "learning_rate": 1.7520753787412245e-06, + "loss": 0.307, + "step": 33823 + }, + { + "epoch": 12.49, + "learning_rate": 1.7520384283778792e-06, + "loss": 0.268, + "step": 33824 + }, + { + "epoch": 12.49, + "learning_rate": 1.7520014780145338e-06, + "loss": 0.3544, + "step": 33825 + }, + { + "epoch": 12.49, + "learning_rate": 1.7519645276511886e-06, + "loss": 0.3438, + "step": 33826 + }, + { + "epoch": 12.49, + "learning_rate": 1.7519275772878433e-06, + "loss": 0.3376, + "step": 33827 + }, + { + "epoch": 12.49, + "learning_rate": 1.7518906269244981e-06, + "loss": 0.2566, + "step": 33828 + }, + { + "epoch": 12.49, + "learning_rate": 1.7518536765611528e-06, + "loss": 0.2645, + "step": 33829 + }, + { + "epoch": 12.49, + "learning_rate": 1.7518167261978078e-06, + "loss": 0.3719, + "step": 33830 + }, + { + "epoch": 12.49, + "learning_rate": 1.7517797758344625e-06, + "loss": 0.3841, + "step": 33831 + }, + { + "epoch": 12.49, + "learning_rate": 1.7517428254711173e-06, + "loss": 0.3009, + "step": 33832 + }, + { + "epoch": 12.49, + "learning_rate": 1.751705875107772e-06, + "loss": 0.2506, + "step": 33833 + }, + { + "epoch": 12.49, + "learning_rate": 1.7516689247444268e-06, + "loss": 0.212, + "step": 33834 + }, + { + "epoch": 12.49, + "learning_rate": 1.7516319743810814e-06, + "loss": 0.2623, + "step": 33835 + }, + { + "epoch": 12.49, + "learning_rate": 1.751595024017736e-06, + "loss": 0.3201, + "step": 33836 + }, + { + "epoch": 12.5, + "learning_rate": 1.7515580736543911e-06, + "loss": 0.2347, + "step": 33837 + }, + { + "epoch": 12.5, + "learning_rate": 1.7515211232910456e-06, + "loss": 0.2293, + "step": 33838 + }, + { + "epoch": 12.5, + "learning_rate": 1.7514841729277006e-06, + "loss": 0.2728, + "step": 33839 + }, + { + "epoch": 12.5, + "learning_rate": 1.7514472225643552e-06, + "loss": 0.3458, + "step": 33840 + }, + { + "epoch": 12.5, + "learning_rate": 1.75141027220101e-06, + "loss": 0.3525, + "step": 33841 + }, + { + "epoch": 12.5, + "learning_rate": 1.7513733218376647e-06, + "loss": 0.2175, + "step": 33842 + }, + { + "epoch": 12.5, + "learning_rate": 1.7513363714743198e-06, + "loss": 0.3461, + "step": 33843 + }, + { + "epoch": 12.5, + "learning_rate": 1.7512994211109742e-06, + "loss": 0.3415, + "step": 33844 + }, + { + "epoch": 12.5, + "learning_rate": 1.7512624707476289e-06, + "loss": 0.2957, + "step": 33845 + }, + { + "epoch": 12.5, + "learning_rate": 1.751225520384284e-06, + "loss": 0.2992, + "step": 33846 + }, + { + "epoch": 12.5, + "learning_rate": 1.7511885700209385e-06, + "loss": 0.3137, + "step": 33847 + }, + { + "epoch": 12.5, + "learning_rate": 1.7511516196575934e-06, + "loss": 0.2587, + "step": 33848 + }, + { + "epoch": 12.5, + "learning_rate": 1.751114669294248e-06, + "loss": 0.3128, + "step": 33849 + }, + { + "epoch": 12.5, + "learning_rate": 1.7510777189309029e-06, + "loss": 0.3053, + "step": 33850 + }, + { + "epoch": 12.5, + "learning_rate": 1.7510407685675575e-06, + "loss": 0.3796, + "step": 33851 + }, + { + "epoch": 12.5, + "learning_rate": 1.7510038182042126e-06, + "loss": 0.3129, + "step": 33852 + }, + { + "epoch": 12.5, + "learning_rate": 1.7509668678408672e-06, + "loss": 0.2323, + "step": 33853 + }, + { + "epoch": 12.5, + "learning_rate": 1.750929917477522e-06, + "loss": 0.2923, + "step": 33854 + }, + { + "epoch": 12.5, + "learning_rate": 1.7508929671141767e-06, + "loss": 0.3263, + "step": 33855 + }, + { + "epoch": 12.5, + "learning_rate": 1.7508560167508313e-06, + "loss": 0.3607, + "step": 33856 + }, + { + "epoch": 12.5, + "learning_rate": 1.7508190663874862e-06, + "loss": 0.3009, + "step": 33857 + }, + { + "epoch": 12.5, + "learning_rate": 1.7507821160241408e-06, + "loss": 0.3359, + "step": 33858 + }, + { + "epoch": 12.5, + "learning_rate": 1.7507451656607959e-06, + "loss": 0.3945, + "step": 33859 + }, + { + "epoch": 12.5, + "learning_rate": 1.7507082152974503e-06, + "loss": 0.3766, + "step": 33860 + }, + { + "epoch": 12.5, + "learning_rate": 1.7506712649341054e-06, + "loss": 0.3744, + "step": 33861 + }, + { + "epoch": 12.5, + "learning_rate": 1.75063431457076e-06, + "loss": 0.2167, + "step": 33862 + }, + { + "epoch": 12.5, + "learning_rate": 1.7505973642074148e-06, + "loss": 0.3087, + "step": 33863 + }, + { + "epoch": 12.51, + "learning_rate": 1.7505604138440695e-06, + "loss": 0.3193, + "step": 33864 + }, + { + "epoch": 12.51, + "learning_rate": 1.7505234634807241e-06, + "loss": 0.1874, + "step": 33865 + }, + { + "epoch": 12.51, + "learning_rate": 1.750486513117379e-06, + "loss": 0.2423, + "step": 33866 + }, + { + "epoch": 12.51, + "learning_rate": 1.7504495627540336e-06, + "loss": 0.3732, + "step": 33867 + }, + { + "epoch": 12.51, + "learning_rate": 1.7504126123906887e-06, + "loss": 0.2711, + "step": 33868 + }, + { + "epoch": 12.51, + "learning_rate": 1.7503756620273433e-06, + "loss": 0.3664, + "step": 33869 + }, + { + "epoch": 12.51, + "learning_rate": 1.7503387116639982e-06, + "loss": 0.4742, + "step": 33870 + }, + { + "epoch": 12.51, + "learning_rate": 1.7503017613006528e-06, + "loss": 0.3566, + "step": 33871 + }, + { + "epoch": 12.51, + "learning_rate": 1.7502648109373076e-06, + "loss": 0.2839, + "step": 33872 + }, + { + "epoch": 12.51, + "learning_rate": 1.7502278605739623e-06, + "loss": 0.2336, + "step": 33873 + }, + { + "epoch": 12.51, + "learning_rate": 1.7501909102106173e-06, + "loss": 0.2444, + "step": 33874 + }, + { + "epoch": 12.51, + "learning_rate": 1.750153959847272e-06, + "loss": 0.2803, + "step": 33875 + }, + { + "epoch": 12.51, + "learning_rate": 1.7501170094839266e-06, + "loss": 0.2555, + "step": 33876 + }, + { + "epoch": 12.51, + "learning_rate": 1.7500800591205815e-06, + "loss": 0.2953, + "step": 33877 + }, + { + "epoch": 12.51, + "learning_rate": 1.750043108757236e-06, + "loss": 0.2787, + "step": 33878 + }, + { + "epoch": 12.51, + "learning_rate": 1.750006158393891e-06, + "loss": 0.1519, + "step": 33879 + }, + { + "epoch": 12.51, + "learning_rate": 1.7499692080305456e-06, + "loss": 0.2825, + "step": 33880 + }, + { + "epoch": 12.51, + "learning_rate": 1.7499322576672006e-06, + "loss": 0.3605, + "step": 33881 + }, + { + "epoch": 12.51, + "learning_rate": 1.749895307303855e-06, + "loss": 0.3046, + "step": 33882 + }, + { + "epoch": 12.51, + "learning_rate": 1.7498583569405101e-06, + "loss": 0.213, + "step": 33883 + }, + { + "epoch": 12.51, + "learning_rate": 1.7498214065771648e-06, + "loss": 0.3206, + "step": 33884 + }, + { + "epoch": 12.51, + "learning_rate": 1.7497844562138194e-06, + "loss": 0.2837, + "step": 33885 + }, + { + "epoch": 12.51, + "learning_rate": 1.7497475058504742e-06, + "loss": 0.2858, + "step": 33886 + }, + { + "epoch": 12.51, + "learning_rate": 1.7497105554871289e-06, + "loss": 0.2503, + "step": 33887 + }, + { + "epoch": 12.51, + "learning_rate": 1.7496736051237837e-06, + "loss": 0.2132, + "step": 33888 + }, + { + "epoch": 12.51, + "learning_rate": 1.7496366547604384e-06, + "loss": 0.3472, + "step": 33889 + }, + { + "epoch": 12.51, + "learning_rate": 1.7495997043970934e-06, + "loss": 0.3376, + "step": 33890 + }, + { + "epoch": 12.52, + "learning_rate": 1.749562754033748e-06, + "loss": 0.4058, + "step": 33891 + }, + { + "epoch": 12.52, + "learning_rate": 1.749525803670403e-06, + "loss": 0.3781, + "step": 33892 + }, + { + "epoch": 12.52, + "learning_rate": 1.7494888533070575e-06, + "loss": 0.3036, + "step": 33893 + }, + { + "epoch": 12.52, + "learning_rate": 1.7494519029437124e-06, + "loss": 0.3797, + "step": 33894 + }, + { + "epoch": 12.52, + "learning_rate": 1.749414952580367e-06, + "loss": 0.2998, + "step": 33895 + }, + { + "epoch": 12.52, + "learning_rate": 1.7493780022170217e-06, + "loss": 0.2027, + "step": 33896 + }, + { + "epoch": 12.52, + "learning_rate": 1.7493410518536767e-06, + "loss": 0.3658, + "step": 33897 + }, + { + "epoch": 12.52, + "learning_rate": 1.7493041014903314e-06, + "loss": 0.2189, + "step": 33898 + }, + { + "epoch": 12.52, + "learning_rate": 1.7492671511269862e-06, + "loss": 0.396, + "step": 33899 + }, + { + "epoch": 12.52, + "learning_rate": 1.7492302007636408e-06, + "loss": 0.3894, + "step": 33900 + }, + { + "epoch": 12.52, + "learning_rate": 1.7491932504002957e-06, + "loss": 0.25, + "step": 33901 + }, + { + "epoch": 12.52, + "learning_rate": 1.7491563000369503e-06, + "loss": 0.2987, + "step": 33902 + }, + { + "epoch": 12.52, + "learning_rate": 1.7491193496736054e-06, + "loss": 0.2766, + "step": 33903 + }, + { + "epoch": 12.52, + "learning_rate": 1.74908239931026e-06, + "loss": 0.3253, + "step": 33904 + }, + { + "epoch": 12.52, + "learning_rate": 1.7490454489469145e-06, + "loss": 0.2944, + "step": 33905 + }, + { + "epoch": 12.52, + "learning_rate": 1.7490084985835695e-06, + "loss": 0.3517, + "step": 33906 + }, + { + "epoch": 12.52, + "learning_rate": 1.7489715482202241e-06, + "loss": 0.3014, + "step": 33907 + }, + { + "epoch": 12.52, + "learning_rate": 1.748934597856879e-06, + "loss": 0.4735, + "step": 33908 + }, + { + "epoch": 12.52, + "learning_rate": 1.7488976474935336e-06, + "loss": 0.2672, + "step": 33909 + }, + { + "epoch": 12.52, + "learning_rate": 1.7488606971301887e-06, + "loss": 0.3165, + "step": 33910 + }, + { + "epoch": 12.52, + "learning_rate": 1.7488237467668431e-06, + "loss": 0.3071, + "step": 33911 + }, + { + "epoch": 12.52, + "learning_rate": 1.7487867964034982e-06, + "loss": 0.4616, + "step": 33912 + }, + { + "epoch": 12.52, + "learning_rate": 1.7487498460401528e-06, + "loss": 0.2349, + "step": 33913 + }, + { + "epoch": 12.52, + "learning_rate": 1.7487128956768077e-06, + "loss": 0.2701, + "step": 33914 + }, + { + "epoch": 12.52, + "learning_rate": 1.7486759453134623e-06, + "loss": 0.2489, + "step": 33915 + }, + { + "epoch": 12.52, + "learning_rate": 1.748638994950117e-06, + "loss": 0.4476, + "step": 33916 + }, + { + "epoch": 12.52, + "learning_rate": 1.7486020445867718e-06, + "loss": 0.313, + "step": 33917 + }, + { + "epoch": 12.53, + "learning_rate": 1.7485650942234264e-06, + "loss": 0.3416, + "step": 33918 + }, + { + "epoch": 12.53, + "learning_rate": 1.7485281438600815e-06, + "loss": 0.1953, + "step": 33919 + }, + { + "epoch": 12.53, + "learning_rate": 1.7484911934967361e-06, + "loss": 0.231, + "step": 33920 + }, + { + "epoch": 12.53, + "learning_rate": 1.748454243133391e-06, + "loss": 0.2978, + "step": 33921 + }, + { + "epoch": 12.53, + "learning_rate": 1.7484172927700456e-06, + "loss": 0.2776, + "step": 33922 + }, + { + "epoch": 12.53, + "learning_rate": 1.7483803424067004e-06, + "loss": 0.3669, + "step": 33923 + }, + { + "epoch": 12.53, + "learning_rate": 1.748343392043355e-06, + "loss": 0.2701, + "step": 33924 + }, + { + "epoch": 12.53, + "learning_rate": 1.7483064416800097e-06, + "loss": 0.2451, + "step": 33925 + }, + { + "epoch": 12.53, + "learning_rate": 1.7482694913166648e-06, + "loss": 0.2766, + "step": 33926 + }, + { + "epoch": 12.53, + "learning_rate": 1.7482325409533192e-06, + "loss": 0.4013, + "step": 33927 + }, + { + "epoch": 12.53, + "learning_rate": 1.7481955905899743e-06, + "loss": 0.2957, + "step": 33928 + }, + { + "epoch": 12.53, + "learning_rate": 1.748158640226629e-06, + "loss": 0.2688, + "step": 33929 + }, + { + "epoch": 12.53, + "learning_rate": 1.7481216898632837e-06, + "loss": 0.2977, + "step": 33930 + }, + { + "epoch": 12.53, + "learning_rate": 1.7480847394999384e-06, + "loss": 0.4645, + "step": 33931 + }, + { + "epoch": 12.53, + "learning_rate": 1.7480477891365934e-06, + "loss": 0.2041, + "step": 33932 + }, + { + "epoch": 12.53, + "learning_rate": 1.7480108387732479e-06, + "loss": 0.2912, + "step": 33933 + }, + { + "epoch": 12.53, + "learning_rate": 1.7479738884099025e-06, + "loss": 0.1957, + "step": 33934 + }, + { + "epoch": 12.53, + "learning_rate": 1.7479369380465576e-06, + "loss": 0.1824, + "step": 33935 + }, + { + "epoch": 12.53, + "learning_rate": 1.7478999876832122e-06, + "loss": 0.3386, + "step": 33936 + }, + { + "epoch": 12.53, + "learning_rate": 1.747863037319867e-06, + "loss": 0.2281, + "step": 33937 + }, + { + "epoch": 12.53, + "learning_rate": 1.7478260869565217e-06, + "loss": 0.3636, + "step": 33938 + }, + { + "epoch": 12.53, + "learning_rate": 1.7477891365931765e-06, + "loss": 0.2482, + "step": 33939 + }, + { + "epoch": 12.53, + "learning_rate": 1.7477521862298312e-06, + "loss": 0.3797, + "step": 33940 + }, + { + "epoch": 12.53, + "learning_rate": 1.7477152358664862e-06, + "loss": 0.2625, + "step": 33941 + }, + { + "epoch": 12.53, + "learning_rate": 1.7476782855031409e-06, + "loss": 0.2135, + "step": 33942 + }, + { + "epoch": 12.53, + "learning_rate": 1.7476413351397957e-06, + "loss": 0.2752, + "step": 33943 + }, + { + "epoch": 12.53, + "learning_rate": 1.7476043847764504e-06, + "loss": 0.4441, + "step": 33944 + }, + { + "epoch": 12.54, + "learning_rate": 1.747567434413105e-06, + "loss": 0.4051, + "step": 33945 + }, + { + "epoch": 12.54, + "learning_rate": 1.7475304840497598e-06, + "loss": 0.3253, + "step": 33946 + }, + { + "epoch": 12.54, + "learning_rate": 1.7474935336864145e-06, + "loss": 0.439, + "step": 33947 + }, + { + "epoch": 12.54, + "learning_rate": 1.7474565833230695e-06, + "loss": 0.2387, + "step": 33948 + }, + { + "epoch": 12.54, + "learning_rate": 1.7474196329597242e-06, + "loss": 0.2595, + "step": 33949 + }, + { + "epoch": 12.54, + "learning_rate": 1.747382682596379e-06, + "loss": 0.3093, + "step": 33950 + }, + { + "epoch": 12.54, + "learning_rate": 1.7473457322330337e-06, + "loss": 0.2887, + "step": 33951 + }, + { + "epoch": 12.54, + "learning_rate": 1.7473087818696885e-06, + "loss": 0.1877, + "step": 33952 + }, + { + "epoch": 12.54, + "learning_rate": 1.7472718315063431e-06, + "loss": 0.3189, + "step": 33953 + }, + { + "epoch": 12.54, + "learning_rate": 1.7472348811429978e-06, + "loss": 0.3996, + "step": 33954 + }, + { + "epoch": 12.54, + "learning_rate": 1.7471979307796526e-06, + "loss": 0.2985, + "step": 33955 + }, + { + "epoch": 12.54, + "learning_rate": 1.7471609804163073e-06, + "loss": 0.4063, + "step": 33956 + }, + { + "epoch": 12.54, + "learning_rate": 1.7471240300529623e-06, + "loss": 0.3015, + "step": 33957 + }, + { + "epoch": 12.54, + "learning_rate": 1.747087079689617e-06, + "loss": 0.3519, + "step": 33958 + }, + { + "epoch": 12.54, + "learning_rate": 1.7470501293262718e-06, + "loss": 0.2434, + "step": 33959 + }, + { + "epoch": 12.54, + "learning_rate": 1.7470131789629264e-06, + "loss": 0.2978, + "step": 33960 + }, + { + "epoch": 12.54, + "learning_rate": 1.7469762285995813e-06, + "loss": 0.3166, + "step": 33961 + }, + { + "epoch": 12.54, + "learning_rate": 1.746939278236236e-06, + "loss": 0.1862, + "step": 33962 + }, + { + "epoch": 12.54, + "learning_rate": 1.746902327872891e-06, + "loss": 0.3295, + "step": 33963 + }, + { + "epoch": 12.54, + "learning_rate": 1.7468653775095456e-06, + "loss": 0.3034, + "step": 33964 + }, + { + "epoch": 12.54, + "learning_rate": 1.7468284271462003e-06, + "loss": 0.2827, + "step": 33965 + }, + { + "epoch": 12.54, + "learning_rate": 1.7467914767828551e-06, + "loss": 0.3135, + "step": 33966 + }, + { + "epoch": 12.54, + "learning_rate": 1.7467545264195097e-06, + "loss": 0.2787, + "step": 33967 + }, + { + "epoch": 12.54, + "learning_rate": 1.7467175760561646e-06, + "loss": 0.2717, + "step": 33968 + }, + { + "epoch": 12.54, + "learning_rate": 1.7466806256928192e-06, + "loss": 0.29, + "step": 33969 + }, + { + "epoch": 12.54, + "learning_rate": 1.7466436753294743e-06, + "loss": 0.3892, + "step": 33970 + }, + { + "epoch": 12.54, + "learning_rate": 1.746606724966129e-06, + "loss": 0.4329, + "step": 33971 + }, + { + "epoch": 12.55, + "learning_rate": 1.7465697746027838e-06, + "loss": 0.3085, + "step": 33972 + }, + { + "epoch": 12.55, + "learning_rate": 1.7465328242394384e-06, + "loss": 0.2421, + "step": 33973 + }, + { + "epoch": 12.55, + "learning_rate": 1.746495873876093e-06, + "loss": 0.3337, + "step": 33974 + }, + { + "epoch": 12.55, + "learning_rate": 1.746458923512748e-06, + "loss": 0.4173, + "step": 33975 + }, + { + "epoch": 12.55, + "learning_rate": 1.7464219731494025e-06, + "loss": 0.3017, + "step": 33976 + }, + { + "epoch": 12.55, + "learning_rate": 1.7463850227860576e-06, + "loss": 0.2681, + "step": 33977 + }, + { + "epoch": 12.55, + "learning_rate": 1.746348072422712e-06, + "loss": 0.3694, + "step": 33978 + }, + { + "epoch": 12.55, + "learning_rate": 1.746311122059367e-06, + "loss": 0.2874, + "step": 33979 + }, + { + "epoch": 12.55, + "learning_rate": 1.7462741716960217e-06, + "loss": 0.2089, + "step": 33980 + }, + { + "epoch": 12.55, + "learning_rate": 1.7462372213326766e-06, + "loss": 0.3183, + "step": 33981 + }, + { + "epoch": 12.55, + "learning_rate": 1.7462002709693312e-06, + "loss": 0.2124, + "step": 33982 + }, + { + "epoch": 12.55, + "learning_rate": 1.7461633206059863e-06, + "loss": 0.3372, + "step": 33983 + }, + { + "epoch": 12.55, + "learning_rate": 1.7461263702426407e-06, + "loss": 0.378, + "step": 33984 + }, + { + "epoch": 12.55, + "learning_rate": 1.7460894198792953e-06, + "loss": 0.276, + "step": 33985 + }, + { + "epoch": 12.55, + "learning_rate": 1.7460524695159504e-06, + "loss": 0.2047, + "step": 33986 + }, + { + "epoch": 12.55, + "learning_rate": 1.746015519152605e-06, + "loss": 0.2938, + "step": 33987 + }, + { + "epoch": 12.55, + "learning_rate": 1.7459785687892599e-06, + "loss": 0.2604, + "step": 33988 + }, + { + "epoch": 12.55, + "learning_rate": 1.7459416184259145e-06, + "loss": 0.3076, + "step": 33989 + }, + { + "epoch": 12.55, + "learning_rate": 1.7459046680625693e-06, + "loss": 0.2443, + "step": 33990 + }, + { + "epoch": 12.55, + "learning_rate": 1.745867717699224e-06, + "loss": 0.3136, + "step": 33991 + }, + { + "epoch": 12.55, + "learning_rate": 1.745830767335879e-06, + "loss": 0.3826, + "step": 33992 + }, + { + "epoch": 12.55, + "learning_rate": 1.7457938169725337e-06, + "loss": 0.2352, + "step": 33993 + }, + { + "epoch": 12.55, + "learning_rate": 1.745756866609188e-06, + "loss": 0.3372, + "step": 33994 + }, + { + "epoch": 12.55, + "learning_rate": 1.7457199162458432e-06, + "loss": 0.3369, + "step": 33995 + }, + { + "epoch": 12.55, + "learning_rate": 1.7456829658824978e-06, + "loss": 0.2637, + "step": 33996 + }, + { + "epoch": 12.55, + "learning_rate": 1.7456460155191526e-06, + "loss": 0.2878, + "step": 33997 + }, + { + "epoch": 12.55, + "learning_rate": 1.7456090651558073e-06, + "loss": 0.357, + "step": 33998 + }, + { + "epoch": 12.56, + "learning_rate": 1.7455721147924623e-06, + "loss": 0.296, + "step": 33999 + }, + { + "epoch": 12.56, + "learning_rate": 1.7455351644291168e-06, + "loss": 0.2138, + "step": 34000 + }, + { + "epoch": 12.56, + "learning_rate": 1.7454982140657718e-06, + "loss": 0.2465, + "step": 34001 + }, + { + "epoch": 12.56, + "learning_rate": 1.7454612637024265e-06, + "loss": 0.3205, + "step": 34002 + }, + { + "epoch": 12.56, + "learning_rate": 1.7454243133390813e-06, + "loss": 0.2909, + "step": 34003 + }, + { + "epoch": 12.56, + "learning_rate": 1.745387362975736e-06, + "loss": 0.4262, + "step": 34004 + }, + { + "epoch": 12.56, + "learning_rate": 1.7453504126123906e-06, + "loss": 0.2744, + "step": 34005 + }, + { + "epoch": 12.56, + "learning_rate": 1.7453134622490454e-06, + "loss": 0.1998, + "step": 34006 + }, + { + "epoch": 12.56, + "learning_rate": 1.7452765118857e-06, + "loss": 0.4855, + "step": 34007 + }, + { + "epoch": 12.56, + "learning_rate": 1.7452395615223551e-06, + "loss": 0.3113, + "step": 34008 + }, + { + "epoch": 12.56, + "learning_rate": 1.7452026111590098e-06, + "loss": 0.3892, + "step": 34009 + }, + { + "epoch": 12.56, + "learning_rate": 1.7451656607956646e-06, + "loss": 0.3299, + "step": 34010 + }, + { + "epoch": 12.56, + "learning_rate": 1.7451287104323193e-06, + "loss": 0.2911, + "step": 34011 + }, + { + "epoch": 12.56, + "learning_rate": 1.745091760068974e-06, + "loss": 0.2104, + "step": 34012 + }, + { + "epoch": 12.56, + "learning_rate": 1.7450548097056287e-06, + "loss": 0.3144, + "step": 34013 + }, + { + "epoch": 12.56, + "learning_rate": 1.7450178593422834e-06, + "loss": 0.4125, + "step": 34014 + }, + { + "epoch": 12.56, + "learning_rate": 1.7449809089789384e-06, + "loss": 0.1726, + "step": 34015 + }, + { + "epoch": 12.56, + "learning_rate": 1.744943958615593e-06, + "loss": 0.2815, + "step": 34016 + }, + { + "epoch": 12.56, + "learning_rate": 1.744907008252248e-06, + "loss": 0.5141, + "step": 34017 + }, + { + "epoch": 12.56, + "learning_rate": 1.7448700578889026e-06, + "loss": 0.2131, + "step": 34018 + }, + { + "epoch": 12.56, + "learning_rate": 1.7448331075255574e-06, + "loss": 0.2082, + "step": 34019 + }, + { + "epoch": 12.56, + "learning_rate": 1.744796157162212e-06, + "loss": 0.36, + "step": 34020 + }, + { + "epoch": 12.56, + "learning_rate": 1.744759206798867e-06, + "loss": 0.3167, + "step": 34021 + }, + { + "epoch": 12.56, + "learning_rate": 1.7447222564355217e-06, + "loss": 0.4039, + "step": 34022 + }, + { + "epoch": 12.56, + "learning_rate": 1.7446853060721766e-06, + "loss": 0.3181, + "step": 34023 + }, + { + "epoch": 12.56, + "learning_rate": 1.7446483557088312e-06, + "loss": 0.2076, + "step": 34024 + }, + { + "epoch": 12.56, + "learning_rate": 1.7446114053454859e-06, + "loss": 0.3607, + "step": 34025 + }, + { + "epoch": 12.56, + "learning_rate": 1.7445744549821407e-06, + "loss": 0.3689, + "step": 34026 + }, + { + "epoch": 12.57, + "learning_rate": 1.7445375046187953e-06, + "loss": 0.3878, + "step": 34027 + }, + { + "epoch": 12.57, + "learning_rate": 1.7445005542554504e-06, + "loss": 0.3089, + "step": 34028 + }, + { + "epoch": 12.57, + "learning_rate": 1.7444636038921048e-06, + "loss": 0.3287, + "step": 34029 + }, + { + "epoch": 12.57, + "learning_rate": 1.7444266535287599e-06, + "loss": 0.2952, + "step": 34030 + }, + { + "epoch": 12.57, + "learning_rate": 1.7443897031654145e-06, + "loss": 0.4228, + "step": 34031 + }, + { + "epoch": 12.57, + "learning_rate": 1.7443527528020694e-06, + "loss": 0.2502, + "step": 34032 + }, + { + "epoch": 12.57, + "learning_rate": 1.744315802438724e-06, + "loss": 0.1997, + "step": 34033 + }, + { + "epoch": 12.57, + "learning_rate": 1.7442788520753786e-06, + "loss": 0.2844, + "step": 34034 + }, + { + "epoch": 12.57, + "learning_rate": 1.7442419017120335e-06, + "loss": 0.3823, + "step": 34035 + }, + { + "epoch": 12.57, + "learning_rate": 1.7442049513486881e-06, + "loss": 0.3735, + "step": 34036 + }, + { + "epoch": 12.57, + "learning_rate": 1.7441680009853432e-06, + "loss": 0.3368, + "step": 34037 + }, + { + "epoch": 12.57, + "learning_rate": 1.7441310506219978e-06, + "loss": 0.3386, + "step": 34038 + }, + { + "epoch": 12.57, + "learning_rate": 1.7440941002586527e-06, + "loss": 0.4331, + "step": 34039 + }, + { + "epoch": 12.57, + "learning_rate": 1.7440571498953073e-06, + "loss": 0.2534, + "step": 34040 + }, + { + "epoch": 12.57, + "learning_rate": 1.7440201995319622e-06, + "loss": 0.2382, + "step": 34041 + }, + { + "epoch": 12.57, + "learning_rate": 1.7439832491686168e-06, + "loss": 0.3653, + "step": 34042 + }, + { + "epoch": 12.57, + "learning_rate": 1.7439462988052719e-06, + "loss": 0.4242, + "step": 34043 + }, + { + "epoch": 12.57, + "learning_rate": 1.7439093484419265e-06, + "loss": 0.3382, + "step": 34044 + }, + { + "epoch": 12.57, + "learning_rate": 1.743872398078581e-06, + "loss": 0.2276, + "step": 34045 + }, + { + "epoch": 12.57, + "learning_rate": 1.743835447715236e-06, + "loss": 0.1924, + "step": 34046 + }, + { + "epoch": 12.57, + "learning_rate": 1.7437984973518906e-06, + "loss": 0.3098, + "step": 34047 + }, + { + "epoch": 12.57, + "learning_rate": 1.7437615469885455e-06, + "loss": 0.2466, + "step": 34048 + }, + { + "epoch": 12.57, + "learning_rate": 1.7437245966252e-06, + "loss": 0.3694, + "step": 34049 + }, + { + "epoch": 12.57, + "learning_rate": 1.7436876462618552e-06, + "loss": 0.2394, + "step": 34050 + }, + { + "epoch": 12.57, + "learning_rate": 1.7436506958985096e-06, + "loss": 0.4232, + "step": 34051 + }, + { + "epoch": 12.57, + "learning_rate": 1.7436137455351646e-06, + "loss": 0.2868, + "step": 34052 + }, + { + "epoch": 12.57, + "learning_rate": 1.7435767951718193e-06, + "loss": 0.3435, + "step": 34053 + }, + { + "epoch": 12.58, + "learning_rate": 1.743539844808474e-06, + "loss": 0.3918, + "step": 34054 + }, + { + "epoch": 12.58, + "learning_rate": 1.7435028944451288e-06, + "loss": 0.2143, + "step": 34055 + }, + { + "epoch": 12.58, + "learning_rate": 1.7434659440817834e-06, + "loss": 0.3922, + "step": 34056 + }, + { + "epoch": 12.58, + "learning_rate": 1.7434289937184382e-06, + "loss": 0.3714, + "step": 34057 + }, + { + "epoch": 12.58, + "learning_rate": 1.7433920433550929e-06, + "loss": 0.2145, + "step": 34058 + }, + { + "epoch": 12.58, + "learning_rate": 1.743355092991748e-06, + "loss": 0.3086, + "step": 34059 + }, + { + "epoch": 12.58, + "learning_rate": 1.7433181426284026e-06, + "loss": 0.2374, + "step": 34060 + }, + { + "epoch": 12.58, + "learning_rate": 1.7432811922650574e-06, + "loss": 0.2859, + "step": 34061 + }, + { + "epoch": 12.58, + "learning_rate": 1.743244241901712e-06, + "loss": 0.4049, + "step": 34062 + }, + { + "epoch": 12.58, + "learning_rate": 1.743207291538367e-06, + "loss": 0.3239, + "step": 34063 + }, + { + "epoch": 12.58, + "learning_rate": 1.7431703411750215e-06, + "loss": 0.2503, + "step": 34064 + }, + { + "epoch": 12.58, + "learning_rate": 1.7431333908116762e-06, + "loss": 0.3147, + "step": 34065 + }, + { + "epoch": 12.58, + "learning_rate": 1.7430964404483312e-06, + "loss": 0.3741, + "step": 34066 + }, + { + "epoch": 12.58, + "learning_rate": 1.7430594900849857e-06, + "loss": 0.3602, + "step": 34067 + }, + { + "epoch": 12.58, + "learning_rate": 1.7430225397216407e-06, + "loss": 0.2142, + "step": 34068 + }, + { + "epoch": 12.58, + "learning_rate": 1.7429855893582954e-06, + "loss": 0.2656, + "step": 34069 + }, + { + "epoch": 12.58, + "learning_rate": 1.7429486389949502e-06, + "loss": 0.1823, + "step": 34070 + }, + { + "epoch": 12.58, + "learning_rate": 1.7429116886316049e-06, + "loss": 0.1948, + "step": 34071 + }, + { + "epoch": 12.58, + "learning_rate": 1.74287473826826e-06, + "loss": 0.3365, + "step": 34072 + }, + { + "epoch": 12.58, + "learning_rate": 1.7428377879049143e-06, + "loss": 0.3327, + "step": 34073 + }, + { + "epoch": 12.58, + "learning_rate": 1.742800837541569e-06, + "loss": 0.3487, + "step": 34074 + }, + { + "epoch": 12.58, + "learning_rate": 1.742763887178224e-06, + "loss": 0.2501, + "step": 34075 + }, + { + "epoch": 12.58, + "learning_rate": 1.7427269368148787e-06, + "loss": 0.3292, + "step": 34076 + }, + { + "epoch": 12.58, + "learning_rate": 1.7426899864515335e-06, + "loss": 0.3432, + "step": 34077 + }, + { + "epoch": 12.58, + "learning_rate": 1.7426530360881882e-06, + "loss": 0.2212, + "step": 34078 + }, + { + "epoch": 12.58, + "learning_rate": 1.742616085724843e-06, + "loss": 0.34, + "step": 34079 + }, + { + "epoch": 12.58, + "learning_rate": 1.7425791353614976e-06, + "loss": 0.4031, + "step": 34080 + }, + { + "epoch": 12.59, + "learning_rate": 1.7425421849981527e-06, + "loss": 0.3368, + "step": 34081 + }, + { + "epoch": 12.59, + "learning_rate": 1.7425052346348073e-06, + "loss": 0.4138, + "step": 34082 + }, + { + "epoch": 12.59, + "learning_rate": 1.7424682842714622e-06, + "loss": 0.3954, + "step": 34083 + }, + { + "epoch": 12.59, + "learning_rate": 1.7424313339081168e-06, + "loss": 0.218, + "step": 34084 + }, + { + "epoch": 12.59, + "learning_rate": 1.7423943835447715e-06, + "loss": 0.2812, + "step": 34085 + }, + { + "epoch": 12.59, + "learning_rate": 1.7423574331814263e-06, + "loss": 0.2751, + "step": 34086 + }, + { + "epoch": 12.59, + "learning_rate": 1.742320482818081e-06, + "loss": 0.2636, + "step": 34087 + }, + { + "epoch": 12.59, + "learning_rate": 1.742283532454736e-06, + "loss": 0.3018, + "step": 34088 + }, + { + "epoch": 12.59, + "learning_rate": 1.7422465820913906e-06, + "loss": 0.2303, + "step": 34089 + }, + { + "epoch": 12.59, + "learning_rate": 1.7422096317280455e-06, + "loss": 0.3366, + "step": 34090 + }, + { + "epoch": 12.59, + "learning_rate": 1.7421726813647001e-06, + "loss": 0.3427, + "step": 34091 + }, + { + "epoch": 12.59, + "learning_rate": 1.742135731001355e-06, + "loss": 0.3593, + "step": 34092 + }, + { + "epoch": 12.59, + "learning_rate": 1.7420987806380096e-06, + "loss": 0.4005, + "step": 34093 + }, + { + "epoch": 12.59, + "learning_rate": 1.7420618302746642e-06, + "loss": 0.2516, + "step": 34094 + }, + { + "epoch": 12.59, + "learning_rate": 1.7420248799113193e-06, + "loss": 0.4067, + "step": 34095 + }, + { + "epoch": 12.59, + "learning_rate": 1.7419879295479737e-06, + "loss": 0.2688, + "step": 34096 + }, + { + "epoch": 12.59, + "learning_rate": 1.7419509791846288e-06, + "loss": 0.363, + "step": 34097 + }, + { + "epoch": 12.59, + "learning_rate": 1.7419140288212834e-06, + "loss": 0.4891, + "step": 34098 + }, + { + "epoch": 12.59, + "learning_rate": 1.7418770784579383e-06, + "loss": 0.4306, + "step": 34099 + }, + { + "epoch": 12.59, + "learning_rate": 1.741840128094593e-06, + "loss": 0.2705, + "step": 34100 + }, + { + "epoch": 12.59, + "learning_rate": 1.741803177731248e-06, + "loss": 0.2724, + "step": 34101 + }, + { + "epoch": 12.59, + "learning_rate": 1.7417662273679024e-06, + "loss": 0.3121, + "step": 34102 + }, + { + "epoch": 12.59, + "learning_rate": 1.7417292770045575e-06, + "loss": 0.3854, + "step": 34103 + }, + { + "epoch": 12.59, + "learning_rate": 1.741692326641212e-06, + "loss": 0.2825, + "step": 34104 + }, + { + "epoch": 12.59, + "learning_rate": 1.7416553762778667e-06, + "loss": 0.2062, + "step": 34105 + }, + { + "epoch": 12.59, + "learning_rate": 1.7416184259145216e-06, + "loss": 0.2785, + "step": 34106 + }, + { + "epoch": 12.59, + "learning_rate": 1.7415814755511762e-06, + "loss": 0.3554, + "step": 34107 + }, + { + "epoch": 12.6, + "learning_rate": 1.741544525187831e-06, + "loss": 0.3392, + "step": 34108 + }, + { + "epoch": 12.6, + "learning_rate": 1.7415075748244857e-06, + "loss": 0.2709, + "step": 34109 + }, + { + "epoch": 12.6, + "learning_rate": 1.7414706244611408e-06, + "loss": 0.3045, + "step": 34110 + }, + { + "epoch": 12.6, + "learning_rate": 1.7414336740977954e-06, + "loss": 0.3133, + "step": 34111 + }, + { + "epoch": 12.6, + "learning_rate": 1.7413967237344502e-06, + "loss": 0.3282, + "step": 34112 + }, + { + "epoch": 12.6, + "learning_rate": 1.7413597733711049e-06, + "loss": 0.3993, + "step": 34113 + }, + { + "epoch": 12.6, + "learning_rate": 1.7413228230077595e-06, + "loss": 0.3662, + "step": 34114 + }, + { + "epoch": 12.6, + "learning_rate": 1.7412858726444144e-06, + "loss": 0.3757, + "step": 34115 + }, + { + "epoch": 12.6, + "learning_rate": 1.741248922281069e-06, + "loss": 0.2942, + "step": 34116 + }, + { + "epoch": 12.6, + "learning_rate": 1.741211971917724e-06, + "loss": 0.3555, + "step": 34117 + }, + { + "epoch": 12.6, + "learning_rate": 1.7411750215543785e-06, + "loss": 0.3402, + "step": 34118 + }, + { + "epoch": 12.6, + "learning_rate": 1.7411380711910335e-06, + "loss": 0.3053, + "step": 34119 + }, + { + "epoch": 12.6, + "learning_rate": 1.7411011208276882e-06, + "loss": 0.2386, + "step": 34120 + }, + { + "epoch": 12.6, + "learning_rate": 1.741064170464343e-06, + "loss": 0.3029, + "step": 34121 + }, + { + "epoch": 12.6, + "learning_rate": 1.7410272201009977e-06, + "loss": 0.3103, + "step": 34122 + }, + { + "epoch": 12.6, + "learning_rate": 1.7409902697376527e-06, + "loss": 0.2576, + "step": 34123 + }, + { + "epoch": 12.6, + "learning_rate": 1.7409533193743071e-06, + "loss": 0.2602, + "step": 34124 + }, + { + "epoch": 12.6, + "learning_rate": 1.7409163690109618e-06, + "loss": 0.3994, + "step": 34125 + }, + { + "epoch": 12.6, + "learning_rate": 1.7408794186476168e-06, + "loss": 0.2517, + "step": 34126 + }, + { + "epoch": 12.6, + "learning_rate": 1.7408424682842715e-06, + "loss": 0.2355, + "step": 34127 + }, + { + "epoch": 12.6, + "learning_rate": 1.7408055179209263e-06, + "loss": 0.3567, + "step": 34128 + }, + { + "epoch": 12.6, + "learning_rate": 1.740768567557581e-06, + "loss": 0.2229, + "step": 34129 + }, + { + "epoch": 12.6, + "learning_rate": 1.7407316171942358e-06, + "loss": 0.4488, + "step": 34130 + }, + { + "epoch": 12.6, + "learning_rate": 1.7406946668308904e-06, + "loss": 0.3883, + "step": 34131 + }, + { + "epoch": 12.6, + "learning_rate": 1.7406577164675455e-06, + "loss": 0.2857, + "step": 34132 + }, + { + "epoch": 12.6, + "learning_rate": 1.7406207661042001e-06, + "loss": 0.372, + "step": 34133 + }, + { + "epoch": 12.6, + "learning_rate": 1.7405838157408548e-06, + "loss": 0.2003, + "step": 34134 + }, + { + "epoch": 12.61, + "learning_rate": 1.7405468653775096e-06, + "loss": 0.3346, + "step": 34135 + }, + { + "epoch": 12.61, + "learning_rate": 1.7405099150141643e-06, + "loss": 0.1984, + "step": 34136 + }, + { + "epoch": 12.61, + "learning_rate": 1.7404729646508191e-06, + "loss": 0.2944, + "step": 34137 + }, + { + "epoch": 12.61, + "learning_rate": 1.7404360142874738e-06, + "loss": 0.2919, + "step": 34138 + }, + { + "epoch": 12.61, + "learning_rate": 1.7403990639241288e-06, + "loss": 0.3801, + "step": 34139 + }, + { + "epoch": 12.61, + "learning_rate": 1.7403621135607834e-06, + "loss": 0.2709, + "step": 34140 + }, + { + "epoch": 12.61, + "learning_rate": 1.7403251631974383e-06, + "loss": 0.2577, + "step": 34141 + }, + { + "epoch": 12.61, + "learning_rate": 1.740288212834093e-06, + "loss": 0.2351, + "step": 34142 + }, + { + "epoch": 12.61, + "learning_rate": 1.7402512624707478e-06, + "loss": 0.238, + "step": 34143 + }, + { + "epoch": 12.61, + "learning_rate": 1.7402143121074024e-06, + "loss": 0.2353, + "step": 34144 + }, + { + "epoch": 12.61, + "learning_rate": 1.740177361744057e-06, + "loss": 0.3376, + "step": 34145 + }, + { + "epoch": 12.61, + "learning_rate": 1.740140411380712e-06, + "loss": 0.4072, + "step": 34146 + }, + { + "epoch": 12.61, + "learning_rate": 1.7401034610173665e-06, + "loss": 0.3157, + "step": 34147 + }, + { + "epoch": 12.61, + "learning_rate": 1.7400665106540216e-06, + "loss": 0.25, + "step": 34148 + }, + { + "epoch": 12.61, + "learning_rate": 1.7400295602906762e-06, + "loss": 0.2397, + "step": 34149 + }, + { + "epoch": 12.61, + "learning_rate": 1.739992609927331e-06, + "loss": 0.3678, + "step": 34150 + }, + { + "epoch": 12.61, + "learning_rate": 1.7399556595639857e-06, + "loss": 0.2973, + "step": 34151 + }, + { + "epoch": 12.61, + "learning_rate": 1.7399187092006406e-06, + "loss": 0.3395, + "step": 34152 + }, + { + "epoch": 12.61, + "learning_rate": 1.7398817588372952e-06, + "loss": 0.2787, + "step": 34153 + }, + { + "epoch": 12.61, + "learning_rate": 1.7398448084739498e-06, + "loss": 0.4746, + "step": 34154 + }, + { + "epoch": 12.61, + "learning_rate": 1.739807858110605e-06, + "loss": 0.2941, + "step": 34155 + }, + { + "epoch": 12.61, + "learning_rate": 1.7397709077472595e-06, + "loss": 0.2724, + "step": 34156 + }, + { + "epoch": 12.61, + "learning_rate": 1.7397339573839144e-06, + "loss": 0.2229, + "step": 34157 + }, + { + "epoch": 12.61, + "learning_rate": 1.739697007020569e-06, + "loss": 0.233, + "step": 34158 + }, + { + "epoch": 12.61, + "learning_rate": 1.7396600566572239e-06, + "loss": 0.3063, + "step": 34159 + }, + { + "epoch": 12.61, + "learning_rate": 1.7396231062938785e-06, + "loss": 0.4023, + "step": 34160 + }, + { + "epoch": 12.61, + "learning_rate": 1.7395861559305336e-06, + "loss": 0.2955, + "step": 34161 + }, + { + "epoch": 12.62, + "learning_rate": 1.7395492055671882e-06, + "loss": 0.2574, + "step": 34162 + }, + { + "epoch": 12.62, + "learning_rate": 1.7395122552038426e-06, + "loss": 0.2878, + "step": 34163 + }, + { + "epoch": 12.62, + "learning_rate": 1.7394753048404977e-06, + "loss": 0.3248, + "step": 34164 + }, + { + "epoch": 12.62, + "learning_rate": 1.7394383544771523e-06, + "loss": 0.2374, + "step": 34165 + }, + { + "epoch": 12.62, + "learning_rate": 1.7394014041138072e-06, + "loss": 0.3513, + "step": 34166 + }, + { + "epoch": 12.62, + "learning_rate": 1.7393644537504618e-06, + "loss": 0.2488, + "step": 34167 + }, + { + "epoch": 12.62, + "learning_rate": 1.7393275033871169e-06, + "loss": 0.4256, + "step": 34168 + }, + { + "epoch": 12.62, + "learning_rate": 1.7392905530237713e-06, + "loss": 0.4036, + "step": 34169 + }, + { + "epoch": 12.62, + "learning_rate": 1.7392536026604264e-06, + "loss": 0.3226, + "step": 34170 + }, + { + "epoch": 12.62, + "learning_rate": 1.739216652297081e-06, + "loss": 0.2179, + "step": 34171 + }, + { + "epoch": 12.62, + "learning_rate": 1.7391797019337358e-06, + "loss": 0.4339, + "step": 34172 + }, + { + "epoch": 12.62, + "learning_rate": 1.7391427515703905e-06, + "loss": 0.3369, + "step": 34173 + }, + { + "epoch": 12.62, + "learning_rate": 1.7391058012070451e-06, + "loss": 0.5368, + "step": 34174 + }, + { + "epoch": 12.62, + "learning_rate": 1.7390688508437e-06, + "loss": 0.3994, + "step": 34175 + }, + { + "epoch": 12.62, + "learning_rate": 1.7390319004803546e-06, + "loss": 0.3537, + "step": 34176 + }, + { + "epoch": 12.62, + "learning_rate": 1.7389949501170097e-06, + "loss": 0.3804, + "step": 34177 + }, + { + "epoch": 12.62, + "learning_rate": 1.7389579997536643e-06, + "loss": 0.2617, + "step": 34178 + }, + { + "epoch": 12.62, + "learning_rate": 1.7389210493903191e-06, + "loss": 0.3003, + "step": 34179 + }, + { + "epoch": 12.62, + "learning_rate": 1.7388840990269738e-06, + "loss": 0.3824, + "step": 34180 + }, + { + "epoch": 12.62, + "learning_rate": 1.7388471486636286e-06, + "loss": 0.2141, + "step": 34181 + }, + { + "epoch": 12.62, + "learning_rate": 1.7388101983002833e-06, + "loss": 0.3152, + "step": 34182 + }, + { + "epoch": 12.62, + "learning_rate": 1.738773247936938e-06, + "loss": 0.3105, + "step": 34183 + }, + { + "epoch": 12.62, + "learning_rate": 1.738736297573593e-06, + "loss": 0.2687, + "step": 34184 + }, + { + "epoch": 12.62, + "learning_rate": 1.7386993472102474e-06, + "loss": 0.2051, + "step": 34185 + }, + { + "epoch": 12.62, + "learning_rate": 1.7386623968469024e-06, + "loss": 0.2147, + "step": 34186 + }, + { + "epoch": 12.62, + "learning_rate": 1.738625446483557e-06, + "loss": 0.1779, + "step": 34187 + }, + { + "epoch": 12.62, + "learning_rate": 1.738588496120212e-06, + "loss": 0.3278, + "step": 34188 + }, + { + "epoch": 12.63, + "learning_rate": 1.7385515457568666e-06, + "loss": 0.2625, + "step": 34189 + }, + { + "epoch": 12.63, + "learning_rate": 1.7385145953935216e-06, + "loss": 0.1296, + "step": 34190 + }, + { + "epoch": 12.63, + "learning_rate": 1.738477645030176e-06, + "loss": 0.1876, + "step": 34191 + }, + { + "epoch": 12.63, + "learning_rate": 1.738440694666831e-06, + "loss": 0.3637, + "step": 34192 + }, + { + "epoch": 12.63, + "learning_rate": 1.7384037443034857e-06, + "loss": 0.2572, + "step": 34193 + }, + { + "epoch": 12.63, + "learning_rate": 1.7383667939401404e-06, + "loss": 0.3844, + "step": 34194 + }, + { + "epoch": 12.63, + "learning_rate": 1.7383298435767952e-06, + "loss": 0.2526, + "step": 34195 + }, + { + "epoch": 12.63, + "learning_rate": 1.7382928932134499e-06, + "loss": 0.169, + "step": 34196 + }, + { + "epoch": 12.63, + "learning_rate": 1.7382559428501047e-06, + "loss": 0.3617, + "step": 34197 + }, + { + "epoch": 12.63, + "learning_rate": 1.7382189924867593e-06, + "loss": 0.1866, + "step": 34198 + }, + { + "epoch": 12.63, + "learning_rate": 1.7381820421234144e-06, + "loss": 0.2728, + "step": 34199 + }, + { + "epoch": 12.63, + "learning_rate": 1.738145091760069e-06, + "loss": 0.2346, + "step": 34200 + }, + { + "epoch": 12.63, + "learning_rate": 1.7381081413967239e-06, + "loss": 0.2947, + "step": 34201 + }, + { + "epoch": 12.63, + "learning_rate": 1.7380711910333785e-06, + "loss": 0.1624, + "step": 34202 + }, + { + "epoch": 12.63, + "learning_rate": 1.7380342406700332e-06, + "loss": 0.3117, + "step": 34203 + }, + { + "epoch": 12.63, + "learning_rate": 1.737997290306688e-06, + "loss": 0.3726, + "step": 34204 + }, + { + "epoch": 12.63, + "learning_rate": 1.7379603399433426e-06, + "loss": 0.4358, + "step": 34205 + }, + { + "epoch": 12.63, + "learning_rate": 1.7379233895799977e-06, + "loss": 0.3246, + "step": 34206 + }, + { + "epoch": 12.63, + "learning_rate": 1.7378864392166523e-06, + "loss": 0.3025, + "step": 34207 + }, + { + "epoch": 12.63, + "learning_rate": 1.7378494888533072e-06, + "loss": 0.336, + "step": 34208 + }, + { + "epoch": 12.63, + "learning_rate": 1.7378125384899618e-06, + "loss": 0.3166, + "step": 34209 + }, + { + "epoch": 12.63, + "learning_rate": 1.7377755881266167e-06, + "loss": 0.4241, + "step": 34210 + }, + { + "epoch": 12.63, + "learning_rate": 1.7377386377632713e-06, + "loss": 0.1674, + "step": 34211 + }, + { + "epoch": 12.63, + "learning_rate": 1.7377016873999264e-06, + "loss": 0.358, + "step": 34212 + }, + { + "epoch": 12.63, + "learning_rate": 1.737664737036581e-06, + "loss": 0.2762, + "step": 34213 + }, + { + "epoch": 12.63, + "learning_rate": 1.7376277866732354e-06, + "loss": 0.2747, + "step": 34214 + }, + { + "epoch": 12.63, + "learning_rate": 1.7375908363098905e-06, + "loss": 0.3626, + "step": 34215 + }, + { + "epoch": 12.64, + "learning_rate": 1.7375538859465451e-06, + "loss": 0.2867, + "step": 34216 + }, + { + "epoch": 12.64, + "learning_rate": 1.7375169355832e-06, + "loss": 0.2851, + "step": 34217 + }, + { + "epoch": 12.64, + "learning_rate": 1.7374799852198546e-06, + "loss": 0.3711, + "step": 34218 + }, + { + "epoch": 12.64, + "learning_rate": 1.7374430348565095e-06, + "loss": 0.2365, + "step": 34219 + }, + { + "epoch": 12.64, + "learning_rate": 1.737406084493164e-06, + "loss": 0.3483, + "step": 34220 + }, + { + "epoch": 12.64, + "learning_rate": 1.7373691341298192e-06, + "loss": 0.29, + "step": 34221 + }, + { + "epoch": 12.64, + "learning_rate": 1.7373321837664738e-06, + "loss": 0.3672, + "step": 34222 + }, + { + "epoch": 12.64, + "learning_rate": 1.7372952334031284e-06, + "loss": 0.4005, + "step": 34223 + }, + { + "epoch": 12.64, + "learning_rate": 1.7372582830397833e-06, + "loss": 0.2804, + "step": 34224 + }, + { + "epoch": 12.64, + "learning_rate": 1.737221332676438e-06, + "loss": 0.3596, + "step": 34225 + }, + { + "epoch": 12.64, + "learning_rate": 1.7371843823130928e-06, + "loss": 0.2723, + "step": 34226 + }, + { + "epoch": 12.64, + "learning_rate": 1.7371474319497474e-06, + "loss": 0.2291, + "step": 34227 + }, + { + "epoch": 12.64, + "learning_rate": 1.7371104815864025e-06, + "loss": 0.3398, + "step": 34228 + }, + { + "epoch": 12.64, + "learning_rate": 1.737073531223057e-06, + "loss": 0.3218, + "step": 34229 + }, + { + "epoch": 12.64, + "learning_rate": 1.737036580859712e-06, + "loss": 0.2142, + "step": 34230 + }, + { + "epoch": 12.64, + "learning_rate": 1.7369996304963666e-06, + "loss": 0.5683, + "step": 34231 + }, + { + "epoch": 12.64, + "learning_rate": 1.7369626801330214e-06, + "loss": 0.3011, + "step": 34232 + }, + { + "epoch": 12.64, + "learning_rate": 1.736925729769676e-06, + "loss": 0.205, + "step": 34233 + }, + { + "epoch": 12.64, + "learning_rate": 1.7368887794063307e-06, + "loss": 0.3306, + "step": 34234 + }, + { + "epoch": 12.64, + "learning_rate": 1.7368518290429858e-06, + "loss": 0.2512, + "step": 34235 + }, + { + "epoch": 12.64, + "learning_rate": 1.7368148786796402e-06, + "loss": 0.3097, + "step": 34236 + }, + { + "epoch": 12.64, + "learning_rate": 1.7367779283162953e-06, + "loss": 0.3788, + "step": 34237 + }, + { + "epoch": 12.64, + "learning_rate": 1.7367409779529499e-06, + "loss": 0.2785, + "step": 34238 + }, + { + "epoch": 12.64, + "learning_rate": 1.7367040275896047e-06, + "loss": 0.3406, + "step": 34239 + }, + { + "epoch": 12.64, + "learning_rate": 1.7366670772262594e-06, + "loss": 0.3179, + "step": 34240 + }, + { + "epoch": 12.64, + "learning_rate": 1.7366301268629144e-06, + "loss": 0.3869, + "step": 34241 + }, + { + "epoch": 12.64, + "learning_rate": 1.7365931764995689e-06, + "loss": 0.3403, + "step": 34242 + }, + { + "epoch": 12.65, + "learning_rate": 1.7365562261362235e-06, + "loss": 0.3963, + "step": 34243 + }, + { + "epoch": 12.65, + "learning_rate": 1.7365192757728786e-06, + "loss": 0.3441, + "step": 34244 + }, + { + "epoch": 12.65, + "learning_rate": 1.7364823254095332e-06, + "loss": 0.3215, + "step": 34245 + }, + { + "epoch": 12.65, + "learning_rate": 1.736445375046188e-06, + "loss": 0.238, + "step": 34246 + }, + { + "epoch": 12.65, + "learning_rate": 1.7364084246828427e-06, + "loss": 0.4367, + "step": 34247 + }, + { + "epoch": 12.65, + "learning_rate": 1.7363714743194975e-06, + "loss": 0.2678, + "step": 34248 + }, + { + "epoch": 12.65, + "learning_rate": 1.7363345239561522e-06, + "loss": 0.2234, + "step": 34249 + }, + { + "epoch": 12.65, + "learning_rate": 1.7362975735928072e-06, + "loss": 0.3533, + "step": 34250 + }, + { + "epoch": 12.65, + "learning_rate": 1.7362606232294619e-06, + "loss": 0.3091, + "step": 34251 + }, + { + "epoch": 12.65, + "learning_rate": 1.7362236728661167e-06, + "loss": 0.2148, + "step": 34252 + }, + { + "epoch": 12.65, + "learning_rate": 1.7361867225027713e-06, + "loss": 0.1648, + "step": 34253 + }, + { + "epoch": 12.65, + "learning_rate": 1.736149772139426e-06, + "loss": 0.3412, + "step": 34254 + }, + { + "epoch": 12.65, + "learning_rate": 1.7361128217760808e-06, + "loss": 0.2349, + "step": 34255 + }, + { + "epoch": 12.65, + "learning_rate": 1.7360758714127355e-06, + "loss": 0.2875, + "step": 34256 + }, + { + "epoch": 12.65, + "learning_rate": 1.7360389210493905e-06, + "loss": 0.3341, + "step": 34257 + }, + { + "epoch": 12.65, + "learning_rate": 1.736001970686045e-06, + "loss": 0.3494, + "step": 34258 + }, + { + "epoch": 12.65, + "learning_rate": 1.7359650203227e-06, + "loss": 0.3411, + "step": 34259 + }, + { + "epoch": 12.65, + "learning_rate": 1.7359280699593546e-06, + "loss": 0.2673, + "step": 34260 + }, + { + "epoch": 12.65, + "learning_rate": 1.7358911195960095e-06, + "loss": 0.1943, + "step": 34261 + }, + { + "epoch": 12.65, + "learning_rate": 1.7358541692326641e-06, + "loss": 0.3127, + "step": 34262 + }, + { + "epoch": 12.65, + "learning_rate": 1.7358172188693188e-06, + "loss": 0.3005, + "step": 34263 + }, + { + "epoch": 12.65, + "learning_rate": 1.7357802685059736e-06, + "loss": 0.3006, + "step": 34264 + }, + { + "epoch": 12.65, + "learning_rate": 1.7357433181426282e-06, + "loss": 0.424, + "step": 34265 + }, + { + "epoch": 12.65, + "learning_rate": 1.7357063677792833e-06, + "loss": 0.2678, + "step": 34266 + }, + { + "epoch": 12.65, + "learning_rate": 1.735669417415938e-06, + "loss": 0.2808, + "step": 34267 + }, + { + "epoch": 12.65, + "learning_rate": 1.7356324670525928e-06, + "loss": 0.2634, + "step": 34268 + }, + { + "epoch": 12.65, + "learning_rate": 1.7355955166892474e-06, + "loss": 0.2675, + "step": 34269 + }, + { + "epoch": 12.66, + "learning_rate": 1.7355585663259023e-06, + "loss": 0.3044, + "step": 34270 + }, + { + "epoch": 12.66, + "learning_rate": 1.735521615962557e-06, + "loss": 0.3108, + "step": 34271 + }, + { + "epoch": 12.66, + "learning_rate": 1.735484665599212e-06, + "loss": 0.4902, + "step": 34272 + }, + { + "epoch": 12.66, + "learning_rate": 1.7354477152358666e-06, + "loss": 0.3106, + "step": 34273 + }, + { + "epoch": 12.66, + "learning_rate": 1.7354107648725212e-06, + "loss": 0.4361, + "step": 34274 + }, + { + "epoch": 12.66, + "learning_rate": 1.735373814509176e-06, + "loss": 0.3322, + "step": 34275 + }, + { + "epoch": 12.66, + "learning_rate": 1.7353368641458307e-06, + "loss": 0.3564, + "step": 34276 + }, + { + "epoch": 12.66, + "learning_rate": 1.7352999137824856e-06, + "loss": 0.4415, + "step": 34277 + }, + { + "epoch": 12.66, + "learning_rate": 1.7352629634191402e-06, + "loss": 0.2481, + "step": 34278 + }, + { + "epoch": 12.66, + "learning_rate": 1.7352260130557953e-06, + "loss": 0.2866, + "step": 34279 + }, + { + "epoch": 12.66, + "learning_rate": 1.73518906269245e-06, + "loss": 0.1739, + "step": 34280 + }, + { + "epoch": 12.66, + "learning_rate": 1.7351521123291048e-06, + "loss": 0.4425, + "step": 34281 + }, + { + "epoch": 12.66, + "learning_rate": 1.7351151619657594e-06, + "loss": 0.3028, + "step": 34282 + }, + { + "epoch": 12.66, + "learning_rate": 1.735078211602414e-06, + "loss": 0.4299, + "step": 34283 + }, + { + "epoch": 12.66, + "learning_rate": 1.7350412612390689e-06, + "loss": 0.3184, + "step": 34284 + }, + { + "epoch": 12.66, + "learning_rate": 1.7350043108757235e-06, + "loss": 0.4226, + "step": 34285 + }, + { + "epoch": 12.66, + "learning_rate": 1.7349673605123786e-06, + "loss": 0.311, + "step": 34286 + }, + { + "epoch": 12.66, + "learning_rate": 1.734930410149033e-06, + "loss": 0.463, + "step": 34287 + }, + { + "epoch": 12.66, + "learning_rate": 1.734893459785688e-06, + "loss": 0.2015, + "step": 34288 + }, + { + "epoch": 12.66, + "learning_rate": 1.7348565094223427e-06, + "loss": 0.2842, + "step": 34289 + }, + { + "epoch": 12.66, + "learning_rate": 1.7348195590589975e-06, + "loss": 0.1833, + "step": 34290 + }, + { + "epoch": 12.66, + "learning_rate": 1.7347826086956522e-06, + "loss": 0.17, + "step": 34291 + }, + { + "epoch": 12.66, + "learning_rate": 1.7347456583323072e-06, + "loss": 0.3932, + "step": 34292 + }, + { + "epoch": 12.66, + "learning_rate": 1.7347087079689617e-06, + "loss": 0.2807, + "step": 34293 + }, + { + "epoch": 12.66, + "learning_rate": 1.7346717576056163e-06, + "loss": 0.273, + "step": 34294 + }, + { + "epoch": 12.66, + "learning_rate": 1.7346348072422714e-06, + "loss": 0.4422, + "step": 34295 + }, + { + "epoch": 12.66, + "learning_rate": 1.734597856878926e-06, + "loss": 0.338, + "step": 34296 + }, + { + "epoch": 12.67, + "learning_rate": 1.7345609065155808e-06, + "loss": 0.2682, + "step": 34297 + }, + { + "epoch": 12.67, + "learning_rate": 1.7345239561522355e-06, + "loss": 0.3174, + "step": 34298 + }, + { + "epoch": 12.67, + "learning_rate": 1.7344870057888903e-06, + "loss": 0.2054, + "step": 34299 + }, + { + "epoch": 12.67, + "learning_rate": 1.734450055425545e-06, + "loss": 0.3244, + "step": 34300 + }, + { + "epoch": 12.67, + "learning_rate": 1.7344131050622e-06, + "loss": 0.2462, + "step": 34301 + }, + { + "epoch": 12.67, + "learning_rate": 1.7343761546988547e-06, + "loss": 0.3624, + "step": 34302 + }, + { + "epoch": 12.67, + "learning_rate": 1.734339204335509e-06, + "loss": 0.3889, + "step": 34303 + }, + { + "epoch": 12.67, + "learning_rate": 1.7343022539721642e-06, + "loss": 0.4621, + "step": 34304 + }, + { + "epoch": 12.67, + "learning_rate": 1.7342653036088188e-06, + "loss": 0.2848, + "step": 34305 + }, + { + "epoch": 12.67, + "learning_rate": 1.7342283532454736e-06, + "loss": 0.2393, + "step": 34306 + }, + { + "epoch": 12.67, + "learning_rate": 1.7341914028821283e-06, + "loss": 0.3482, + "step": 34307 + }, + { + "epoch": 12.67, + "learning_rate": 1.7341544525187833e-06, + "loss": 0.3697, + "step": 34308 + }, + { + "epoch": 12.67, + "learning_rate": 1.7341175021554378e-06, + "loss": 0.2467, + "step": 34309 + }, + { + "epoch": 12.67, + "learning_rate": 1.7340805517920928e-06, + "loss": 0.3258, + "step": 34310 + }, + { + "epoch": 12.67, + "learning_rate": 1.7340436014287475e-06, + "loss": 0.2661, + "step": 34311 + }, + { + "epoch": 12.67, + "learning_rate": 1.7340066510654023e-06, + "loss": 0.3362, + "step": 34312 + }, + { + "epoch": 12.67, + "learning_rate": 1.733969700702057e-06, + "loss": 0.386, + "step": 34313 + }, + { + "epoch": 12.67, + "learning_rate": 1.7339327503387116e-06, + "loss": 0.2355, + "step": 34314 + }, + { + "epoch": 12.67, + "learning_rate": 1.7338957999753664e-06, + "loss": 0.3053, + "step": 34315 + }, + { + "epoch": 12.67, + "learning_rate": 1.733858849612021e-06, + "loss": 0.2621, + "step": 34316 + }, + { + "epoch": 12.67, + "learning_rate": 1.7338218992486761e-06, + "loss": 0.3815, + "step": 34317 + }, + { + "epoch": 12.67, + "learning_rate": 1.7337849488853308e-06, + "loss": 0.3526, + "step": 34318 + }, + { + "epoch": 12.67, + "learning_rate": 1.7337479985219856e-06, + "loss": 0.3692, + "step": 34319 + }, + { + "epoch": 12.67, + "learning_rate": 1.7337110481586402e-06, + "loss": 0.3293, + "step": 34320 + }, + { + "epoch": 12.67, + "learning_rate": 1.733674097795295e-06, + "loss": 0.3393, + "step": 34321 + }, + { + "epoch": 12.67, + "learning_rate": 1.7336371474319497e-06, + "loss": 0.3544, + "step": 34322 + }, + { + "epoch": 12.67, + "learning_rate": 1.7336001970686044e-06, + "loss": 0.4208, + "step": 34323 + }, + { + "epoch": 12.68, + "learning_rate": 1.7335632467052594e-06, + "loss": 0.2944, + "step": 34324 + }, + { + "epoch": 12.68, + "learning_rate": 1.733526296341914e-06, + "loss": 0.1932, + "step": 34325 + }, + { + "epoch": 12.68, + "learning_rate": 1.733489345978569e-06, + "loss": 0.2457, + "step": 34326 + }, + { + "epoch": 12.68, + "learning_rate": 1.7334523956152235e-06, + "loss": 0.2891, + "step": 34327 + }, + { + "epoch": 12.68, + "learning_rate": 1.7334154452518784e-06, + "loss": 0.2541, + "step": 34328 + }, + { + "epoch": 12.68, + "learning_rate": 1.733378494888533e-06, + "loss": 0.2023, + "step": 34329 + }, + { + "epoch": 12.68, + "learning_rate": 1.733341544525188e-06, + "loss": 0.3727, + "step": 34330 + }, + { + "epoch": 12.68, + "learning_rate": 1.7333045941618425e-06, + "loss": 0.2185, + "step": 34331 + }, + { + "epoch": 12.68, + "learning_rate": 1.7332676437984976e-06, + "loss": 0.2517, + "step": 34332 + }, + { + "epoch": 12.68, + "learning_rate": 1.7332306934351522e-06, + "loss": 0.3649, + "step": 34333 + }, + { + "epoch": 12.68, + "learning_rate": 1.7331937430718068e-06, + "loss": 0.2723, + "step": 34334 + }, + { + "epoch": 12.68, + "learning_rate": 1.7331567927084617e-06, + "loss": 0.2223, + "step": 34335 + }, + { + "epoch": 12.68, + "learning_rate": 1.7331198423451163e-06, + "loss": 0.3718, + "step": 34336 + }, + { + "epoch": 12.68, + "learning_rate": 1.7330828919817712e-06, + "loss": 0.4311, + "step": 34337 + }, + { + "epoch": 12.68, + "learning_rate": 1.7330459416184258e-06, + "loss": 0.401, + "step": 34338 + }, + { + "epoch": 12.68, + "learning_rate": 1.7330089912550809e-06, + "loss": 0.2726, + "step": 34339 + }, + { + "epoch": 12.68, + "learning_rate": 1.7329720408917355e-06, + "loss": 0.5079, + "step": 34340 + }, + { + "epoch": 12.68, + "learning_rate": 1.7329350905283904e-06, + "loss": 0.453, + "step": 34341 + }, + { + "epoch": 12.68, + "learning_rate": 1.732898140165045e-06, + "loss": 0.2503, + "step": 34342 + }, + { + "epoch": 12.68, + "learning_rate": 1.7328611898016996e-06, + "loss": 0.2647, + "step": 34343 + }, + { + "epoch": 12.68, + "learning_rate": 1.7328242394383545e-06, + "loss": 0.2932, + "step": 34344 + }, + { + "epoch": 12.68, + "learning_rate": 1.7327872890750091e-06, + "loss": 0.3397, + "step": 34345 + }, + { + "epoch": 12.68, + "learning_rate": 1.7327503387116642e-06, + "loss": 0.3546, + "step": 34346 + }, + { + "epoch": 12.68, + "learning_rate": 1.7327133883483188e-06, + "loss": 0.3123, + "step": 34347 + }, + { + "epoch": 12.68, + "learning_rate": 1.7326764379849737e-06, + "loss": 0.3908, + "step": 34348 + }, + { + "epoch": 12.68, + "learning_rate": 1.7326394876216283e-06, + "loss": 0.2323, + "step": 34349 + }, + { + "epoch": 12.68, + "learning_rate": 1.7326025372582831e-06, + "loss": 0.3211, + "step": 34350 + }, + { + "epoch": 12.69, + "learning_rate": 1.7325655868949378e-06, + "loss": 0.2371, + "step": 34351 + }, + { + "epoch": 12.69, + "learning_rate": 1.7325286365315928e-06, + "loss": 0.2552, + "step": 34352 + }, + { + "epoch": 12.69, + "learning_rate": 1.7324916861682475e-06, + "loss": 0.3083, + "step": 34353 + }, + { + "epoch": 12.69, + "learning_rate": 1.732454735804902e-06, + "loss": 0.2085, + "step": 34354 + }, + { + "epoch": 12.69, + "learning_rate": 1.732417785441557e-06, + "loss": 0.4196, + "step": 34355 + }, + { + "epoch": 12.69, + "learning_rate": 1.7323808350782116e-06, + "loss": 0.2097, + "step": 34356 + }, + { + "epoch": 12.69, + "learning_rate": 1.7323438847148664e-06, + "loss": 0.4212, + "step": 34357 + }, + { + "epoch": 12.69, + "learning_rate": 1.732306934351521e-06, + "loss": 0.3225, + "step": 34358 + }, + { + "epoch": 12.69, + "learning_rate": 1.7322699839881761e-06, + "loss": 0.256, + "step": 34359 + }, + { + "epoch": 12.69, + "learning_rate": 1.7322330336248306e-06, + "loss": 0.3129, + "step": 34360 + }, + { + "epoch": 12.69, + "learning_rate": 1.7321960832614856e-06, + "loss": 0.2074, + "step": 34361 + }, + { + "epoch": 12.69, + "learning_rate": 1.7321591328981403e-06, + "loss": 0.3256, + "step": 34362 + }, + { + "epoch": 12.69, + "learning_rate": 1.732122182534795e-06, + "loss": 0.2634, + "step": 34363 + }, + { + "epoch": 12.69, + "learning_rate": 1.7320852321714497e-06, + "loss": 0.2354, + "step": 34364 + }, + { + "epoch": 12.69, + "learning_rate": 1.7320482818081044e-06, + "loss": 0.4158, + "step": 34365 + }, + { + "epoch": 12.69, + "learning_rate": 1.7320113314447592e-06, + "loss": 0.3108, + "step": 34366 + }, + { + "epoch": 12.69, + "learning_rate": 1.7319743810814139e-06, + "loss": 0.3335, + "step": 34367 + }, + { + "epoch": 12.69, + "learning_rate": 1.731937430718069e-06, + "loss": 0.2958, + "step": 34368 + }, + { + "epoch": 12.69, + "learning_rate": 1.7319004803547236e-06, + "loss": 0.2698, + "step": 34369 + }, + { + "epoch": 12.69, + "learning_rate": 1.7318635299913784e-06, + "loss": 0.2937, + "step": 34370 + }, + { + "epoch": 12.69, + "learning_rate": 1.731826579628033e-06, + "loss": 0.2202, + "step": 34371 + }, + { + "epoch": 12.69, + "learning_rate": 1.731789629264688e-06, + "loss": 0.2573, + "step": 34372 + }, + { + "epoch": 12.69, + "learning_rate": 1.7317526789013425e-06, + "loss": 0.3825, + "step": 34373 + }, + { + "epoch": 12.69, + "learning_rate": 1.7317157285379972e-06, + "loss": 0.2369, + "step": 34374 + }, + { + "epoch": 12.69, + "learning_rate": 1.7316787781746522e-06, + "loss": 0.2427, + "step": 34375 + }, + { + "epoch": 12.69, + "learning_rate": 1.7316418278113067e-06, + "loss": 0.3261, + "step": 34376 + }, + { + "epoch": 12.69, + "learning_rate": 1.7316048774479617e-06, + "loss": 0.2056, + "step": 34377 + }, + { + "epoch": 12.69, + "learning_rate": 1.7315679270846164e-06, + "loss": 0.3694, + "step": 34378 + }, + { + "epoch": 12.7, + "learning_rate": 1.7315309767212712e-06, + "loss": 0.2819, + "step": 34379 + }, + { + "epoch": 12.7, + "learning_rate": 1.7314940263579258e-06, + "loss": 0.2323, + "step": 34380 + }, + { + "epoch": 12.7, + "learning_rate": 1.731457075994581e-06, + "loss": 0.4038, + "step": 34381 + }, + { + "epoch": 12.7, + "learning_rate": 1.7314201256312353e-06, + "loss": 0.2859, + "step": 34382 + }, + { + "epoch": 12.7, + "learning_rate": 1.73138317526789e-06, + "loss": 0.2574, + "step": 34383 + }, + { + "epoch": 12.7, + "learning_rate": 1.731346224904545e-06, + "loss": 0.3096, + "step": 34384 + }, + { + "epoch": 12.7, + "learning_rate": 1.7313092745411997e-06, + "loss": 0.3439, + "step": 34385 + }, + { + "epoch": 12.7, + "learning_rate": 1.7312723241778545e-06, + "loss": 0.3719, + "step": 34386 + }, + { + "epoch": 12.7, + "learning_rate": 1.7312353738145091e-06, + "loss": 0.2731, + "step": 34387 + }, + { + "epoch": 12.7, + "learning_rate": 1.731198423451164e-06, + "loss": 0.1526, + "step": 34388 + }, + { + "epoch": 12.7, + "learning_rate": 1.7311614730878186e-06, + "loss": 0.183, + "step": 34389 + }, + { + "epoch": 12.7, + "learning_rate": 1.7311245227244737e-06, + "loss": 0.1966, + "step": 34390 + }, + { + "epoch": 12.7, + "learning_rate": 1.7310875723611283e-06, + "loss": 0.365, + "step": 34391 + }, + { + "epoch": 12.7, + "learning_rate": 1.731050621997783e-06, + "loss": 0.5255, + "step": 34392 + }, + { + "epoch": 12.7, + "learning_rate": 1.7310136716344378e-06, + "loss": 0.3976, + "step": 34393 + }, + { + "epoch": 12.7, + "learning_rate": 1.7309767212710924e-06, + "loss": 0.3783, + "step": 34394 + }, + { + "epoch": 12.7, + "learning_rate": 1.7309397709077473e-06, + "loss": 0.2163, + "step": 34395 + }, + { + "epoch": 12.7, + "learning_rate": 1.730902820544402e-06, + "loss": 0.1968, + "step": 34396 + }, + { + "epoch": 12.7, + "learning_rate": 1.730865870181057e-06, + "loss": 0.2934, + "step": 34397 + }, + { + "epoch": 12.7, + "learning_rate": 1.7308289198177116e-06, + "loss": 0.3022, + "step": 34398 + }, + { + "epoch": 12.7, + "learning_rate": 1.7307919694543665e-06, + "loss": 0.2676, + "step": 34399 + }, + { + "epoch": 12.7, + "learning_rate": 1.7307550190910211e-06, + "loss": 0.4456, + "step": 34400 + }, + { + "epoch": 12.7, + "learning_rate": 1.730718068727676e-06, + "loss": 0.3039, + "step": 34401 + }, + { + "epoch": 12.7, + "learning_rate": 1.7306811183643306e-06, + "loss": 0.3466, + "step": 34402 + }, + { + "epoch": 12.7, + "learning_rate": 1.7306441680009852e-06, + "loss": 0.2626, + "step": 34403 + }, + { + "epoch": 12.7, + "learning_rate": 1.7306072176376403e-06, + "loss": 0.1858, + "step": 34404 + }, + { + "epoch": 12.7, + "learning_rate": 1.7305702672742947e-06, + "loss": 0.3305, + "step": 34405 + }, + { + "epoch": 12.71, + "learning_rate": 1.7305333169109498e-06, + "loss": 0.2807, + "step": 34406 + }, + { + "epoch": 12.71, + "learning_rate": 1.7304963665476044e-06, + "loss": 0.2946, + "step": 34407 + }, + { + "epoch": 12.71, + "learning_rate": 1.7304594161842593e-06, + "loss": 0.3314, + "step": 34408 + }, + { + "epoch": 12.71, + "learning_rate": 1.7304224658209139e-06, + "loss": 0.3367, + "step": 34409 + }, + { + "epoch": 12.71, + "learning_rate": 1.7303855154575687e-06, + "loss": 0.3055, + "step": 34410 + }, + { + "epoch": 12.71, + "learning_rate": 1.7303485650942234e-06, + "loss": 0.238, + "step": 34411 + }, + { + "epoch": 12.71, + "learning_rate": 1.730311614730878e-06, + "loss": 0.1713, + "step": 34412 + }, + { + "epoch": 12.71, + "learning_rate": 1.730274664367533e-06, + "loss": 0.3396, + "step": 34413 + }, + { + "epoch": 12.71, + "learning_rate": 1.7302377140041877e-06, + "loss": 0.2127, + "step": 34414 + }, + { + "epoch": 12.71, + "learning_rate": 1.7302007636408426e-06, + "loss": 0.2628, + "step": 34415 + }, + { + "epoch": 12.71, + "learning_rate": 1.7301638132774972e-06, + "loss": 0.2858, + "step": 34416 + }, + { + "epoch": 12.71, + "learning_rate": 1.730126862914152e-06, + "loss": 0.3792, + "step": 34417 + }, + { + "epoch": 12.71, + "learning_rate": 1.7300899125508067e-06, + "loss": 0.2795, + "step": 34418 + }, + { + "epoch": 12.71, + "learning_rate": 1.7300529621874617e-06, + "loss": 0.2589, + "step": 34419 + }, + { + "epoch": 12.71, + "learning_rate": 1.7300160118241164e-06, + "loss": 0.2198, + "step": 34420 + }, + { + "epoch": 12.71, + "learning_rate": 1.7299790614607712e-06, + "loss": 0.2279, + "step": 34421 + }, + { + "epoch": 12.71, + "learning_rate": 1.7299421110974259e-06, + "loss": 0.3206, + "step": 34422 + }, + { + "epoch": 12.71, + "learning_rate": 1.7299051607340805e-06, + "loss": 0.4751, + "step": 34423 + }, + { + "epoch": 12.71, + "learning_rate": 1.7298682103707353e-06, + "loss": 0.2411, + "step": 34424 + }, + { + "epoch": 12.71, + "learning_rate": 1.72983126000739e-06, + "loss": 0.4102, + "step": 34425 + }, + { + "epoch": 12.71, + "learning_rate": 1.729794309644045e-06, + "loss": 0.3588, + "step": 34426 + }, + { + "epoch": 12.71, + "learning_rate": 1.7297573592806995e-06, + "loss": 0.2477, + "step": 34427 + }, + { + "epoch": 12.71, + "learning_rate": 1.7297204089173545e-06, + "loss": 0.236, + "step": 34428 + }, + { + "epoch": 12.71, + "learning_rate": 1.7296834585540092e-06, + "loss": 0.1889, + "step": 34429 + }, + { + "epoch": 12.71, + "learning_rate": 1.729646508190664e-06, + "loss": 0.3604, + "step": 34430 + }, + { + "epoch": 12.71, + "learning_rate": 1.7296095578273186e-06, + "loss": 0.2911, + "step": 34431 + }, + { + "epoch": 12.71, + "learning_rate": 1.7295726074639733e-06, + "loss": 0.3799, + "step": 34432 + }, + { + "epoch": 12.72, + "learning_rate": 1.7295356571006281e-06, + "loss": 0.354, + "step": 34433 + }, + { + "epoch": 12.72, + "learning_rate": 1.7294987067372828e-06, + "loss": 0.32, + "step": 34434 + }, + { + "epoch": 12.72, + "learning_rate": 1.7294617563739378e-06, + "loss": 0.3191, + "step": 34435 + }, + { + "epoch": 12.72, + "learning_rate": 1.7294248060105925e-06, + "loss": 0.3102, + "step": 34436 + }, + { + "epoch": 12.72, + "learning_rate": 1.7293878556472473e-06, + "loss": 0.3205, + "step": 34437 + }, + { + "epoch": 12.72, + "learning_rate": 1.729350905283902e-06, + "loss": 0.3417, + "step": 34438 + }, + { + "epoch": 12.72, + "learning_rate": 1.7293139549205568e-06, + "loss": 0.2533, + "step": 34439 + }, + { + "epoch": 12.72, + "learning_rate": 1.7292770045572114e-06, + "loss": 0.1984, + "step": 34440 + }, + { + "epoch": 12.72, + "learning_rate": 1.7292400541938665e-06, + "loss": 0.3213, + "step": 34441 + }, + { + "epoch": 12.72, + "learning_rate": 1.7292031038305211e-06, + "loss": 0.3571, + "step": 34442 + }, + { + "epoch": 12.72, + "learning_rate": 1.7291661534671756e-06, + "loss": 0.3793, + "step": 34443 + }, + { + "epoch": 12.72, + "learning_rate": 1.7291292031038306e-06, + "loss": 0.1647, + "step": 34444 + }, + { + "epoch": 12.72, + "learning_rate": 1.7290922527404853e-06, + "loss": 0.3242, + "step": 34445 + }, + { + "epoch": 12.72, + "learning_rate": 1.72905530237714e-06, + "loss": 0.2832, + "step": 34446 + }, + { + "epoch": 12.72, + "learning_rate": 1.7290183520137947e-06, + "loss": 0.318, + "step": 34447 + }, + { + "epoch": 12.72, + "learning_rate": 1.7289814016504498e-06, + "loss": 0.2803, + "step": 34448 + }, + { + "epoch": 12.72, + "learning_rate": 1.7289444512871042e-06, + "loss": 0.2364, + "step": 34449 + }, + { + "epoch": 12.72, + "learning_rate": 1.7289075009237593e-06, + "loss": 0.1752, + "step": 34450 + }, + { + "epoch": 12.72, + "learning_rate": 1.728870550560414e-06, + "loss": 0.2457, + "step": 34451 + }, + { + "epoch": 12.72, + "learning_rate": 1.7288336001970686e-06, + "loss": 0.2405, + "step": 34452 + }, + { + "epoch": 12.72, + "learning_rate": 1.7287966498337234e-06, + "loss": 0.3066, + "step": 34453 + }, + { + "epoch": 12.72, + "learning_rate": 1.728759699470378e-06, + "loss": 0.2521, + "step": 34454 + }, + { + "epoch": 12.72, + "learning_rate": 1.7287227491070329e-06, + "loss": 0.3391, + "step": 34455 + }, + { + "epoch": 12.72, + "learning_rate": 1.7286857987436875e-06, + "loss": 0.4717, + "step": 34456 + }, + { + "epoch": 12.72, + "learning_rate": 1.7286488483803426e-06, + "loss": 0.278, + "step": 34457 + }, + { + "epoch": 12.72, + "learning_rate": 1.7286118980169972e-06, + "loss": 0.3715, + "step": 34458 + }, + { + "epoch": 12.72, + "learning_rate": 1.728574947653652e-06, + "loss": 0.2875, + "step": 34459 + }, + { + "epoch": 12.73, + "learning_rate": 1.7285379972903067e-06, + "loss": 0.2318, + "step": 34460 + }, + { + "epoch": 12.73, + "learning_rate": 1.7285010469269616e-06, + "loss": 0.2614, + "step": 34461 + }, + { + "epoch": 12.73, + "learning_rate": 1.7284640965636162e-06, + "loss": 0.2546, + "step": 34462 + }, + { + "epoch": 12.73, + "learning_rate": 1.7284271462002708e-06, + "loss": 0.2932, + "step": 34463 + }, + { + "epoch": 12.73, + "learning_rate": 1.7283901958369259e-06, + "loss": 0.309, + "step": 34464 + }, + { + "epoch": 12.73, + "learning_rate": 1.7283532454735805e-06, + "loss": 0.3206, + "step": 34465 + }, + { + "epoch": 12.73, + "learning_rate": 1.7283162951102354e-06, + "loss": 0.3023, + "step": 34466 + }, + { + "epoch": 12.73, + "learning_rate": 1.72827934474689e-06, + "loss": 0.2639, + "step": 34467 + }, + { + "epoch": 12.73, + "learning_rate": 1.7282423943835449e-06, + "loss": 0.2694, + "step": 34468 + }, + { + "epoch": 12.73, + "learning_rate": 1.7282054440201995e-06, + "loss": 0.4223, + "step": 34469 + }, + { + "epoch": 12.73, + "learning_rate": 1.7281684936568546e-06, + "loss": 0.3221, + "step": 34470 + }, + { + "epoch": 12.73, + "learning_rate": 1.7281315432935092e-06, + "loss": 0.2225, + "step": 34471 + }, + { + "epoch": 12.73, + "learning_rate": 1.7280945929301636e-06, + "loss": 0.2993, + "step": 34472 + }, + { + "epoch": 12.73, + "learning_rate": 1.7280576425668187e-06, + "loss": 0.3288, + "step": 34473 + }, + { + "epoch": 12.73, + "learning_rate": 1.7280206922034733e-06, + "loss": 0.396, + "step": 34474 + }, + { + "epoch": 12.73, + "learning_rate": 1.7279837418401282e-06, + "loss": 0.2683, + "step": 34475 + }, + { + "epoch": 12.73, + "learning_rate": 1.7279467914767828e-06, + "loss": 0.4325, + "step": 34476 + }, + { + "epoch": 12.73, + "learning_rate": 1.7279098411134379e-06, + "loss": 0.4766, + "step": 34477 + }, + { + "epoch": 12.73, + "learning_rate": 1.7278728907500923e-06, + "loss": 0.3733, + "step": 34478 + }, + { + "epoch": 12.73, + "learning_rate": 1.7278359403867473e-06, + "loss": 0.2445, + "step": 34479 + }, + { + "epoch": 12.73, + "learning_rate": 1.727798990023402e-06, + "loss": 0.4245, + "step": 34480 + }, + { + "epoch": 12.73, + "learning_rate": 1.7277620396600568e-06, + "loss": 0.2535, + "step": 34481 + }, + { + "epoch": 12.73, + "learning_rate": 1.7277250892967115e-06, + "loss": 0.248, + "step": 34482 + }, + { + "epoch": 12.73, + "learning_rate": 1.727688138933366e-06, + "loss": 0.2592, + "step": 34483 + }, + { + "epoch": 12.73, + "learning_rate": 1.727651188570021e-06, + "loss": 0.2435, + "step": 34484 + }, + { + "epoch": 12.73, + "learning_rate": 1.7276142382066756e-06, + "loss": 0.2988, + "step": 34485 + }, + { + "epoch": 12.73, + "learning_rate": 1.7275772878433306e-06, + "loss": 0.2379, + "step": 34486 + }, + { + "epoch": 12.74, + "learning_rate": 1.7275403374799853e-06, + "loss": 0.2173, + "step": 34487 + }, + { + "epoch": 12.74, + "learning_rate": 1.7275033871166401e-06, + "loss": 0.3234, + "step": 34488 + }, + { + "epoch": 12.74, + "learning_rate": 1.7274664367532948e-06, + "loss": 0.2552, + "step": 34489 + }, + { + "epoch": 12.74, + "learning_rate": 1.7274294863899496e-06, + "loss": 0.2952, + "step": 34490 + }, + { + "epoch": 12.74, + "learning_rate": 1.7273925360266042e-06, + "loss": 0.2441, + "step": 34491 + }, + { + "epoch": 12.74, + "learning_rate": 1.7273555856632589e-06, + "loss": 0.2861, + "step": 34492 + }, + { + "epoch": 12.74, + "learning_rate": 1.727318635299914e-06, + "loss": 0.5312, + "step": 34493 + }, + { + "epoch": 12.74, + "learning_rate": 1.7272816849365684e-06, + "loss": 0.3861, + "step": 34494 + }, + { + "epoch": 12.74, + "learning_rate": 1.7272447345732234e-06, + "loss": 0.3394, + "step": 34495 + }, + { + "epoch": 12.74, + "learning_rate": 1.727207784209878e-06, + "loss": 0.1862, + "step": 34496 + }, + { + "epoch": 12.74, + "learning_rate": 1.727170833846533e-06, + "loss": 0.3962, + "step": 34497 + }, + { + "epoch": 12.74, + "learning_rate": 1.7271338834831875e-06, + "loss": 0.261, + "step": 34498 + }, + { + "epoch": 12.74, + "learning_rate": 1.7270969331198426e-06, + "loss": 0.37, + "step": 34499 + }, + { + "epoch": 12.74, + "learning_rate": 1.727059982756497e-06, + "loss": 0.2414, + "step": 34500 + }, + { + "epoch": 12.74, + "learning_rate": 1.727023032393152e-06, + "loss": 0.3527, + "step": 34501 + }, + { + "epoch": 12.74, + "learning_rate": 1.7269860820298067e-06, + "loss": 0.285, + "step": 34502 + }, + { + "epoch": 12.74, + "learning_rate": 1.7269491316664614e-06, + "loss": 0.248, + "step": 34503 + }, + { + "epoch": 12.74, + "learning_rate": 1.7269121813031162e-06, + "loss": 0.2513, + "step": 34504 + }, + { + "epoch": 12.74, + "learning_rate": 1.7268752309397708e-06, + "loss": 0.3086, + "step": 34505 + }, + { + "epoch": 12.74, + "learning_rate": 1.7268382805764257e-06, + "loss": 0.2846, + "step": 34506 + }, + { + "epoch": 12.74, + "learning_rate": 1.7268013302130803e-06, + "loss": 0.3435, + "step": 34507 + }, + { + "epoch": 12.74, + "learning_rate": 1.7267643798497354e-06, + "loss": 0.2446, + "step": 34508 + }, + { + "epoch": 12.74, + "learning_rate": 1.72672742948639e-06, + "loss": 0.2232, + "step": 34509 + }, + { + "epoch": 12.74, + "learning_rate": 1.7266904791230449e-06, + "loss": 0.2657, + "step": 34510 + }, + { + "epoch": 12.74, + "learning_rate": 1.7266535287596995e-06, + "loss": 0.5173, + "step": 34511 + }, + { + "epoch": 12.74, + "learning_rate": 1.7266165783963542e-06, + "loss": 0.3536, + "step": 34512 + }, + { + "epoch": 12.74, + "learning_rate": 1.726579628033009e-06, + "loss": 0.2112, + "step": 34513 + }, + { + "epoch": 12.75, + "learning_rate": 1.7265426776696636e-06, + "loss": 0.2317, + "step": 34514 + }, + { + "epoch": 12.75, + "learning_rate": 1.7265057273063187e-06, + "loss": 0.3443, + "step": 34515 + }, + { + "epoch": 12.75, + "learning_rate": 1.7264687769429731e-06, + "loss": 0.3482, + "step": 34516 + }, + { + "epoch": 12.75, + "learning_rate": 1.7264318265796282e-06, + "loss": 0.3268, + "step": 34517 + }, + { + "epoch": 12.75, + "learning_rate": 1.7263948762162828e-06, + "loss": 0.4323, + "step": 34518 + }, + { + "epoch": 12.75, + "learning_rate": 1.7263579258529377e-06, + "loss": 0.393, + "step": 34519 + }, + { + "epoch": 12.75, + "learning_rate": 1.7263209754895923e-06, + "loss": 0.3022, + "step": 34520 + }, + { + "epoch": 12.75, + "learning_rate": 1.7262840251262474e-06, + "loss": 0.3281, + "step": 34521 + }, + { + "epoch": 12.75, + "learning_rate": 1.7262470747629018e-06, + "loss": 0.3448, + "step": 34522 + }, + { + "epoch": 12.75, + "learning_rate": 1.7262101243995564e-06, + "loss": 0.2478, + "step": 34523 + }, + { + "epoch": 12.75, + "learning_rate": 1.7261731740362115e-06, + "loss": 0.2968, + "step": 34524 + }, + { + "epoch": 12.75, + "learning_rate": 1.7261362236728661e-06, + "loss": 0.2067, + "step": 34525 + }, + { + "epoch": 12.75, + "learning_rate": 1.726099273309521e-06, + "loss": 0.3207, + "step": 34526 + }, + { + "epoch": 12.75, + "learning_rate": 1.7260623229461756e-06, + "loss": 0.2416, + "step": 34527 + }, + { + "epoch": 12.75, + "learning_rate": 1.7260253725828305e-06, + "loss": 0.424, + "step": 34528 + }, + { + "epoch": 12.75, + "learning_rate": 1.725988422219485e-06, + "loss": 0.2793, + "step": 34529 + }, + { + "epoch": 12.75, + "learning_rate": 1.7259514718561401e-06, + "loss": 0.2182, + "step": 34530 + }, + { + "epoch": 12.75, + "learning_rate": 1.7259145214927948e-06, + "loss": 0.2166, + "step": 34531 + }, + { + "epoch": 12.75, + "learning_rate": 1.7258775711294494e-06, + "loss": 0.3111, + "step": 34532 + }, + { + "epoch": 12.75, + "learning_rate": 1.7258406207661043e-06, + "loss": 0.2624, + "step": 34533 + }, + { + "epoch": 12.75, + "learning_rate": 1.725803670402759e-06, + "loss": 0.1955, + "step": 34534 + }, + { + "epoch": 12.75, + "learning_rate": 1.7257667200394138e-06, + "loss": 0.2872, + "step": 34535 + }, + { + "epoch": 12.75, + "learning_rate": 1.7257297696760684e-06, + "loss": 0.35, + "step": 34536 + }, + { + "epoch": 12.75, + "learning_rate": 1.7256928193127235e-06, + "loss": 0.3617, + "step": 34537 + }, + { + "epoch": 12.75, + "learning_rate": 1.725655868949378e-06, + "loss": 0.2325, + "step": 34538 + }, + { + "epoch": 12.75, + "learning_rate": 1.725618918586033e-06, + "loss": 0.2617, + "step": 34539 + }, + { + "epoch": 12.75, + "learning_rate": 1.7255819682226876e-06, + "loss": 0.2608, + "step": 34540 + }, + { + "epoch": 12.76, + "learning_rate": 1.7255450178593424e-06, + "loss": 0.3071, + "step": 34541 + }, + { + "epoch": 12.76, + "learning_rate": 1.725508067495997e-06, + "loss": 0.2031, + "step": 34542 + }, + { + "epoch": 12.76, + "learning_rate": 1.7254711171326517e-06, + "loss": 0.371, + "step": 34543 + }, + { + "epoch": 12.76, + "learning_rate": 1.7254341667693068e-06, + "loss": 0.2451, + "step": 34544 + }, + { + "epoch": 12.76, + "learning_rate": 1.7253972164059612e-06, + "loss": 0.3288, + "step": 34545 + }, + { + "epoch": 12.76, + "learning_rate": 1.7253602660426162e-06, + "loss": 0.2205, + "step": 34546 + }, + { + "epoch": 12.76, + "learning_rate": 1.7253233156792709e-06, + "loss": 0.3845, + "step": 34547 + }, + { + "epoch": 12.76, + "learning_rate": 1.7252863653159257e-06, + "loss": 0.2967, + "step": 34548 + }, + { + "epoch": 12.76, + "learning_rate": 1.7252494149525804e-06, + "loss": 0.2115, + "step": 34549 + }, + { + "epoch": 12.76, + "learning_rate": 1.7252124645892354e-06, + "loss": 0.1742, + "step": 34550 + }, + { + "epoch": 12.76, + "learning_rate": 1.7251755142258898e-06, + "loss": 0.3889, + "step": 34551 + }, + { + "epoch": 12.76, + "learning_rate": 1.7251385638625445e-06, + "loss": 0.2537, + "step": 34552 + }, + { + "epoch": 12.76, + "learning_rate": 1.7251016134991995e-06, + "loss": 0.3901, + "step": 34553 + }, + { + "epoch": 12.76, + "learning_rate": 1.7250646631358542e-06, + "loss": 0.2009, + "step": 34554 + }, + { + "epoch": 12.76, + "learning_rate": 1.725027712772509e-06, + "loss": 0.3426, + "step": 34555 + }, + { + "epoch": 12.76, + "learning_rate": 1.7249907624091637e-06, + "loss": 0.306, + "step": 34556 + }, + { + "epoch": 12.76, + "learning_rate": 1.7249538120458185e-06, + "loss": 0.2206, + "step": 34557 + }, + { + "epoch": 12.76, + "learning_rate": 1.7249168616824731e-06, + "loss": 0.3599, + "step": 34558 + }, + { + "epoch": 12.76, + "learning_rate": 1.7248799113191282e-06, + "loss": 0.2446, + "step": 34559 + }, + { + "epoch": 12.76, + "learning_rate": 1.7248429609557828e-06, + "loss": 0.1844, + "step": 34560 + }, + { + "epoch": 12.76, + "learning_rate": 1.7248060105924377e-06, + "loss": 0.3158, + "step": 34561 + }, + { + "epoch": 12.76, + "learning_rate": 1.7247690602290923e-06, + "loss": 0.3035, + "step": 34562 + }, + { + "epoch": 12.76, + "learning_rate": 1.724732109865747e-06, + "loss": 0.4264, + "step": 34563 + }, + { + "epoch": 12.76, + "learning_rate": 1.7246951595024018e-06, + "loss": 0.4791, + "step": 34564 + }, + { + "epoch": 12.76, + "learning_rate": 1.7246582091390564e-06, + "loss": 0.2233, + "step": 34565 + }, + { + "epoch": 12.76, + "learning_rate": 1.7246212587757115e-06, + "loss": 0.2832, + "step": 34566 + }, + { + "epoch": 12.76, + "learning_rate": 1.724584308412366e-06, + "loss": 0.29, + "step": 34567 + }, + { + "epoch": 12.77, + "learning_rate": 1.724547358049021e-06, + "loss": 0.3438, + "step": 34568 + }, + { + "epoch": 12.77, + "learning_rate": 1.7245104076856756e-06, + "loss": 0.3106, + "step": 34569 + }, + { + "epoch": 12.77, + "learning_rate": 1.7244734573223305e-06, + "loss": 0.3673, + "step": 34570 + }, + { + "epoch": 12.77, + "learning_rate": 1.7244365069589851e-06, + "loss": 0.2523, + "step": 34571 + }, + { + "epoch": 12.77, + "learning_rate": 1.7243995565956397e-06, + "loss": 0.3142, + "step": 34572 + }, + { + "epoch": 12.77, + "learning_rate": 1.7243626062322946e-06, + "loss": 0.2871, + "step": 34573 + }, + { + "epoch": 12.77, + "learning_rate": 1.7243256558689492e-06, + "loss": 0.3265, + "step": 34574 + }, + { + "epoch": 12.77, + "learning_rate": 1.7242887055056043e-06, + "loss": 0.2581, + "step": 34575 + }, + { + "epoch": 12.77, + "learning_rate": 1.724251755142259e-06, + "loss": 0.3449, + "step": 34576 + }, + { + "epoch": 12.77, + "learning_rate": 1.7242148047789138e-06, + "loss": 0.1979, + "step": 34577 + }, + { + "epoch": 12.77, + "learning_rate": 1.7241778544155684e-06, + "loss": 0.3418, + "step": 34578 + }, + { + "epoch": 12.77, + "learning_rate": 1.7241409040522233e-06, + "loss": 0.2101, + "step": 34579 + }, + { + "epoch": 12.77, + "learning_rate": 1.724103953688878e-06, + "loss": 0.283, + "step": 34580 + }, + { + "epoch": 12.77, + "learning_rate": 1.724067003325533e-06, + "loss": 0.3117, + "step": 34581 + }, + { + "epoch": 12.77, + "learning_rate": 1.7240300529621876e-06, + "loss": 0.2641, + "step": 34582 + }, + { + "epoch": 12.77, + "learning_rate": 1.7239931025988422e-06, + "loss": 0.3713, + "step": 34583 + }, + { + "epoch": 12.77, + "learning_rate": 1.723956152235497e-06, + "loss": 0.2597, + "step": 34584 + }, + { + "epoch": 12.77, + "learning_rate": 1.7239192018721517e-06, + "loss": 0.4651, + "step": 34585 + }, + { + "epoch": 12.77, + "learning_rate": 1.7238822515088066e-06, + "loss": 0.3528, + "step": 34586 + }, + { + "epoch": 12.77, + "learning_rate": 1.7238453011454612e-06, + "loss": 0.2467, + "step": 34587 + }, + { + "epoch": 12.77, + "learning_rate": 1.7238083507821163e-06, + "loss": 0.3631, + "step": 34588 + }, + { + "epoch": 12.77, + "learning_rate": 1.723771400418771e-06, + "loss": 0.1935, + "step": 34589 + }, + { + "epoch": 12.77, + "learning_rate": 1.7237344500554257e-06, + "loss": 0.237, + "step": 34590 + }, + { + "epoch": 12.77, + "learning_rate": 1.7236974996920804e-06, + "loss": 0.3217, + "step": 34591 + }, + { + "epoch": 12.77, + "learning_rate": 1.723660549328735e-06, + "loss": 0.2511, + "step": 34592 + }, + { + "epoch": 12.77, + "learning_rate": 1.7236235989653899e-06, + "loss": 0.2965, + "step": 34593 + }, + { + "epoch": 12.77, + "learning_rate": 1.7235866486020445e-06, + "loss": 0.2482, + "step": 34594 + }, + { + "epoch": 12.78, + "learning_rate": 1.7235496982386994e-06, + "loss": 0.306, + "step": 34595 + }, + { + "epoch": 12.78, + "learning_rate": 1.723512747875354e-06, + "loss": 0.3006, + "step": 34596 + }, + { + "epoch": 12.78, + "learning_rate": 1.723475797512009e-06, + "loss": 0.232, + "step": 34597 + }, + { + "epoch": 12.78, + "learning_rate": 1.7234388471486637e-06, + "loss": 0.2792, + "step": 34598 + }, + { + "epoch": 12.78, + "learning_rate": 1.7234018967853185e-06, + "loss": 0.4215, + "step": 34599 + }, + { + "epoch": 12.78, + "learning_rate": 1.7233649464219732e-06, + "loss": 0.3511, + "step": 34600 + }, + { + "epoch": 12.78, + "learning_rate": 1.723327996058628e-06, + "loss": 0.199, + "step": 34601 + }, + { + "epoch": 12.78, + "learning_rate": 1.7232910456952827e-06, + "loss": 0.3008, + "step": 34602 + }, + { + "epoch": 12.78, + "learning_rate": 1.7232540953319373e-06, + "loss": 0.3299, + "step": 34603 + }, + { + "epoch": 12.78, + "learning_rate": 1.7232171449685924e-06, + "loss": 0.4759, + "step": 34604 + }, + { + "epoch": 12.78, + "learning_rate": 1.723180194605247e-06, + "loss": 0.1773, + "step": 34605 + }, + { + "epoch": 12.78, + "learning_rate": 1.7231432442419018e-06, + "loss": 0.2794, + "step": 34606 + }, + { + "epoch": 12.78, + "learning_rate": 1.7231062938785565e-06, + "loss": 0.3789, + "step": 34607 + }, + { + "epoch": 12.78, + "learning_rate": 1.7230693435152113e-06, + "loss": 0.1718, + "step": 34608 + }, + { + "epoch": 12.78, + "learning_rate": 1.723032393151866e-06, + "loss": 0.1642, + "step": 34609 + }, + { + "epoch": 12.78, + "learning_rate": 1.722995442788521e-06, + "loss": 0.2937, + "step": 34610 + }, + { + "epoch": 12.78, + "learning_rate": 1.7229584924251757e-06, + "loss": 0.2919, + "step": 34611 + }, + { + "epoch": 12.78, + "learning_rate": 1.72292154206183e-06, + "loss": 0.2187, + "step": 34612 + }, + { + "epoch": 12.78, + "learning_rate": 1.7228845916984851e-06, + "loss": 0.3305, + "step": 34613 + }, + { + "epoch": 12.78, + "learning_rate": 1.7228476413351398e-06, + "loss": 0.3289, + "step": 34614 + }, + { + "epoch": 12.78, + "learning_rate": 1.7228106909717946e-06, + "loss": 0.3315, + "step": 34615 + }, + { + "epoch": 12.78, + "learning_rate": 1.7227737406084493e-06, + "loss": 0.1826, + "step": 34616 + }, + { + "epoch": 12.78, + "learning_rate": 1.7227367902451043e-06, + "loss": 0.2557, + "step": 34617 + }, + { + "epoch": 12.78, + "learning_rate": 1.7226998398817587e-06, + "loss": 0.2344, + "step": 34618 + }, + { + "epoch": 12.78, + "learning_rate": 1.7226628895184138e-06, + "loss": 0.2606, + "step": 34619 + }, + { + "epoch": 12.78, + "learning_rate": 1.7226259391550684e-06, + "loss": 0.3088, + "step": 34620 + }, + { + "epoch": 12.78, + "learning_rate": 1.722588988791723e-06, + "loss": 0.2641, + "step": 34621 + }, + { + "epoch": 12.79, + "learning_rate": 1.722552038428378e-06, + "loss": 0.3038, + "step": 34622 + }, + { + "epoch": 12.79, + "learning_rate": 1.7225150880650326e-06, + "loss": 0.2283, + "step": 34623 + }, + { + "epoch": 12.79, + "learning_rate": 1.7224781377016874e-06, + "loss": 0.4309, + "step": 34624 + }, + { + "epoch": 12.79, + "learning_rate": 1.722441187338342e-06, + "loss": 0.2315, + "step": 34625 + }, + { + "epoch": 12.79, + "learning_rate": 1.722404236974997e-06, + "loss": 0.3094, + "step": 34626 + }, + { + "epoch": 12.79, + "learning_rate": 1.7223672866116517e-06, + "loss": 0.2052, + "step": 34627 + }, + { + "epoch": 12.79, + "learning_rate": 1.7223303362483066e-06, + "loss": 0.2221, + "step": 34628 + }, + { + "epoch": 12.79, + "learning_rate": 1.7222933858849612e-06, + "loss": 0.285, + "step": 34629 + }, + { + "epoch": 12.79, + "learning_rate": 1.722256435521616e-06, + "loss": 0.3894, + "step": 34630 + }, + { + "epoch": 12.79, + "learning_rate": 1.7222194851582707e-06, + "loss": 0.249, + "step": 34631 + }, + { + "epoch": 12.79, + "learning_rate": 1.7221825347949253e-06, + "loss": 0.3631, + "step": 34632 + }, + { + "epoch": 12.79, + "learning_rate": 1.7221455844315804e-06, + "loss": 0.2651, + "step": 34633 + }, + { + "epoch": 12.79, + "learning_rate": 1.7221086340682348e-06, + "loss": 0.4486, + "step": 34634 + }, + { + "epoch": 12.79, + "learning_rate": 1.7220716837048899e-06, + "loss": 0.3316, + "step": 34635 + }, + { + "epoch": 12.79, + "learning_rate": 1.7220347333415445e-06, + "loss": 0.322, + "step": 34636 + }, + { + "epoch": 12.79, + "learning_rate": 1.7219977829781994e-06, + "loss": 0.4205, + "step": 34637 + }, + { + "epoch": 12.79, + "learning_rate": 1.721960832614854e-06, + "loss": 0.2517, + "step": 34638 + }, + { + "epoch": 12.79, + "learning_rate": 1.721923882251509e-06, + "loss": 0.3977, + "step": 34639 + }, + { + "epoch": 12.79, + "learning_rate": 1.7218869318881635e-06, + "loss": 0.2727, + "step": 34640 + }, + { + "epoch": 12.79, + "learning_rate": 1.7218499815248181e-06, + "loss": 0.28, + "step": 34641 + }, + { + "epoch": 12.79, + "learning_rate": 1.7218130311614732e-06, + "loss": 0.5743, + "step": 34642 + }, + { + "epoch": 12.79, + "learning_rate": 1.7217760807981278e-06, + "loss": 0.198, + "step": 34643 + }, + { + "epoch": 12.79, + "learning_rate": 1.7217391304347827e-06, + "loss": 0.2539, + "step": 34644 + }, + { + "epoch": 12.79, + "learning_rate": 1.7217021800714373e-06, + "loss": 0.377, + "step": 34645 + }, + { + "epoch": 12.79, + "learning_rate": 1.7216652297080922e-06, + "loss": 0.2887, + "step": 34646 + }, + { + "epoch": 12.79, + "learning_rate": 1.7216282793447468e-06, + "loss": 0.4976, + "step": 34647 + }, + { + "epoch": 12.79, + "learning_rate": 1.7215913289814019e-06, + "loss": 0.3777, + "step": 34648 + }, + { + "epoch": 12.8, + "learning_rate": 1.7215543786180565e-06, + "loss": 0.2209, + "step": 34649 + }, + { + "epoch": 12.8, + "learning_rate": 1.7215174282547113e-06, + "loss": 0.2617, + "step": 34650 + }, + { + "epoch": 12.8, + "learning_rate": 1.721480477891366e-06, + "loss": 0.2888, + "step": 34651 + }, + { + "epoch": 12.8, + "learning_rate": 1.7214435275280206e-06, + "loss": 0.283, + "step": 34652 + }, + { + "epoch": 12.8, + "learning_rate": 1.7214065771646755e-06, + "loss": 0.2441, + "step": 34653 + }, + { + "epoch": 12.8, + "learning_rate": 1.72136962680133e-06, + "loss": 0.3417, + "step": 34654 + }, + { + "epoch": 12.8, + "learning_rate": 1.7213326764379852e-06, + "loss": 0.2605, + "step": 34655 + }, + { + "epoch": 12.8, + "learning_rate": 1.7212957260746398e-06, + "loss": 0.1981, + "step": 34656 + }, + { + "epoch": 12.8, + "learning_rate": 1.7212587757112946e-06, + "loss": 0.4326, + "step": 34657 + }, + { + "epoch": 12.8, + "learning_rate": 1.7212218253479493e-06, + "loss": 0.3125, + "step": 34658 + }, + { + "epoch": 12.8, + "learning_rate": 1.7211848749846041e-06, + "loss": 0.4235, + "step": 34659 + }, + { + "epoch": 12.8, + "learning_rate": 1.7211479246212588e-06, + "loss": 0.391, + "step": 34660 + }, + { + "epoch": 12.8, + "learning_rate": 1.7211109742579134e-06, + "loss": 0.2801, + "step": 34661 + }, + { + "epoch": 12.8, + "learning_rate": 1.7210740238945685e-06, + "loss": 0.2594, + "step": 34662 + }, + { + "epoch": 12.8, + "learning_rate": 1.7210370735312229e-06, + "loss": 0.2522, + "step": 34663 + }, + { + "epoch": 12.8, + "learning_rate": 1.721000123167878e-06, + "loss": 0.3331, + "step": 34664 + }, + { + "epoch": 12.8, + "learning_rate": 1.7209631728045326e-06, + "loss": 0.1907, + "step": 34665 + }, + { + "epoch": 12.8, + "learning_rate": 1.7209262224411874e-06, + "loss": 0.3423, + "step": 34666 + }, + { + "epoch": 12.8, + "learning_rate": 1.720889272077842e-06, + "loss": 0.2041, + "step": 34667 + }, + { + "epoch": 12.8, + "learning_rate": 1.7208523217144971e-06, + "loss": 0.0915, + "step": 34668 + }, + { + "epoch": 12.8, + "learning_rate": 1.7208153713511516e-06, + "loss": 0.298, + "step": 34669 + }, + { + "epoch": 12.8, + "learning_rate": 1.7207784209878066e-06, + "loss": 0.2175, + "step": 34670 + }, + { + "epoch": 12.8, + "learning_rate": 1.7207414706244612e-06, + "loss": 0.3946, + "step": 34671 + }, + { + "epoch": 12.8, + "learning_rate": 1.7207045202611159e-06, + "loss": 0.3569, + "step": 34672 + }, + { + "epoch": 12.8, + "learning_rate": 1.7206675698977707e-06, + "loss": 0.4087, + "step": 34673 + }, + { + "epoch": 12.8, + "learning_rate": 1.7206306195344254e-06, + "loss": 0.309, + "step": 34674 + }, + { + "epoch": 12.8, + "learning_rate": 1.7205936691710802e-06, + "loss": 0.2674, + "step": 34675 + }, + { + "epoch": 12.81, + "learning_rate": 1.7205567188077349e-06, + "loss": 0.2465, + "step": 34676 + }, + { + "epoch": 12.81, + "learning_rate": 1.72051976844439e-06, + "loss": 0.3611, + "step": 34677 + }, + { + "epoch": 12.81, + "learning_rate": 1.7204828180810446e-06, + "loss": 0.2164, + "step": 34678 + }, + { + "epoch": 12.81, + "learning_rate": 1.7204458677176994e-06, + "loss": 0.3269, + "step": 34679 + }, + { + "epoch": 12.81, + "learning_rate": 1.720408917354354e-06, + "loss": 0.2401, + "step": 34680 + }, + { + "epoch": 12.81, + "learning_rate": 1.7203719669910087e-06, + "loss": 0.3295, + "step": 34681 + }, + { + "epoch": 12.81, + "learning_rate": 1.7203350166276635e-06, + "loss": 0.3078, + "step": 34682 + }, + { + "epoch": 12.81, + "learning_rate": 1.7202980662643182e-06, + "loss": 0.2346, + "step": 34683 + }, + { + "epoch": 12.81, + "learning_rate": 1.7202611159009732e-06, + "loss": 0.4391, + "step": 34684 + }, + { + "epoch": 12.81, + "learning_rate": 1.7202241655376276e-06, + "loss": 0.3117, + "step": 34685 + }, + { + "epoch": 12.81, + "learning_rate": 1.7201872151742827e-06, + "loss": 0.2603, + "step": 34686 + }, + { + "epoch": 12.81, + "learning_rate": 1.7201502648109373e-06, + "loss": 0.2518, + "step": 34687 + }, + { + "epoch": 12.81, + "learning_rate": 1.7201133144475922e-06, + "loss": 0.3108, + "step": 34688 + }, + { + "epoch": 12.81, + "learning_rate": 1.7200763640842468e-06, + "loss": 0.2429, + "step": 34689 + }, + { + "epoch": 12.81, + "learning_rate": 1.7200394137209019e-06, + "loss": 0.281, + "step": 34690 + }, + { + "epoch": 12.81, + "learning_rate": 1.7200024633575563e-06, + "loss": 0.3879, + "step": 34691 + }, + { + "epoch": 12.81, + "learning_rate": 1.719965512994211e-06, + "loss": 0.2847, + "step": 34692 + }, + { + "epoch": 12.81, + "learning_rate": 1.719928562630866e-06, + "loss": 0.2687, + "step": 34693 + }, + { + "epoch": 12.81, + "learning_rate": 1.7198916122675206e-06, + "loss": 0.3298, + "step": 34694 + }, + { + "epoch": 12.81, + "learning_rate": 1.7198546619041755e-06, + "loss": 0.25, + "step": 34695 + }, + { + "epoch": 12.81, + "learning_rate": 1.7198177115408301e-06, + "loss": 0.3525, + "step": 34696 + }, + { + "epoch": 12.81, + "learning_rate": 1.719780761177485e-06, + "loss": 0.2245, + "step": 34697 + }, + { + "epoch": 12.81, + "learning_rate": 1.7197438108141396e-06, + "loss": 0.3401, + "step": 34698 + }, + { + "epoch": 12.81, + "learning_rate": 1.7197068604507947e-06, + "loss": 0.4058, + "step": 34699 + }, + { + "epoch": 12.81, + "learning_rate": 1.7196699100874493e-06, + "loss": 0.2555, + "step": 34700 + }, + { + "epoch": 12.81, + "learning_rate": 1.719632959724104e-06, + "loss": 0.4162, + "step": 34701 + }, + { + "epoch": 12.81, + "learning_rate": 1.7195960093607588e-06, + "loss": 0.3178, + "step": 34702 + }, + { + "epoch": 12.81, + "learning_rate": 1.7195590589974134e-06, + "loss": 0.2354, + "step": 34703 + }, + { + "epoch": 12.82, + "learning_rate": 1.7195221086340683e-06, + "loss": 0.2595, + "step": 34704 + }, + { + "epoch": 12.82, + "learning_rate": 1.719485158270723e-06, + "loss": 0.2752, + "step": 34705 + }, + { + "epoch": 12.82, + "learning_rate": 1.719448207907378e-06, + "loss": 0.3408, + "step": 34706 + }, + { + "epoch": 12.82, + "learning_rate": 1.7194112575440324e-06, + "loss": 0.3498, + "step": 34707 + }, + { + "epoch": 12.82, + "learning_rate": 1.7193743071806875e-06, + "loss": 0.3519, + "step": 34708 + }, + { + "epoch": 12.82, + "learning_rate": 1.719337356817342e-06, + "loss": 0.3122, + "step": 34709 + }, + { + "epoch": 12.82, + "learning_rate": 1.719300406453997e-06, + "loss": 0.2196, + "step": 34710 + }, + { + "epoch": 12.82, + "learning_rate": 1.7192634560906516e-06, + "loss": 0.3833, + "step": 34711 + }, + { + "epoch": 12.82, + "learning_rate": 1.7192265057273062e-06, + "loss": 0.2848, + "step": 34712 + }, + { + "epoch": 12.82, + "learning_rate": 1.719189555363961e-06, + "loss": 0.3171, + "step": 34713 + }, + { + "epoch": 12.82, + "learning_rate": 1.7191526050006157e-06, + "loss": 0.2783, + "step": 34714 + }, + { + "epoch": 12.82, + "learning_rate": 1.7191156546372708e-06, + "loss": 0.3692, + "step": 34715 + }, + { + "epoch": 12.82, + "learning_rate": 1.7190787042739254e-06, + "loss": 0.3939, + "step": 34716 + }, + { + "epoch": 12.82, + "learning_rate": 1.7190417539105802e-06, + "loss": 0.3152, + "step": 34717 + }, + { + "epoch": 12.82, + "learning_rate": 1.7190048035472349e-06, + "loss": 0.2734, + "step": 34718 + }, + { + "epoch": 12.82, + "learning_rate": 1.7189678531838897e-06, + "loss": 0.5602, + "step": 34719 + }, + { + "epoch": 12.82, + "learning_rate": 1.7189309028205444e-06, + "loss": 0.2986, + "step": 34720 + }, + { + "epoch": 12.82, + "learning_rate": 1.718893952457199e-06, + "loss": 0.3315, + "step": 34721 + }, + { + "epoch": 12.82, + "learning_rate": 1.718857002093854e-06, + "loss": 0.386, + "step": 34722 + }, + { + "epoch": 12.82, + "learning_rate": 1.7188200517305087e-06, + "loss": 0.2234, + "step": 34723 + }, + { + "epoch": 12.82, + "learning_rate": 1.7187831013671635e-06, + "loss": 0.2245, + "step": 34724 + }, + { + "epoch": 12.82, + "learning_rate": 1.7187461510038182e-06, + "loss": 0.2891, + "step": 34725 + }, + { + "epoch": 12.82, + "learning_rate": 1.718709200640473e-06, + "loss": 0.366, + "step": 34726 + }, + { + "epoch": 12.82, + "learning_rate": 1.7186722502771277e-06, + "loss": 0.3882, + "step": 34727 + }, + { + "epoch": 12.82, + "learning_rate": 1.7186352999137827e-06, + "loss": 0.2773, + "step": 34728 + }, + { + "epoch": 12.82, + "learning_rate": 1.7185983495504374e-06, + "loss": 0.4126, + "step": 34729 + }, + { + "epoch": 12.82, + "learning_rate": 1.7185613991870922e-06, + "loss": 0.1752, + "step": 34730 + }, + { + "epoch": 12.83, + "learning_rate": 1.7185244488237468e-06, + "loss": 0.2548, + "step": 34731 + }, + { + "epoch": 12.83, + "learning_rate": 1.7184874984604015e-06, + "loss": 0.2228, + "step": 34732 + }, + { + "epoch": 12.83, + "learning_rate": 1.7184505480970563e-06, + "loss": 0.3441, + "step": 34733 + }, + { + "epoch": 12.83, + "learning_rate": 1.718413597733711e-06, + "loss": 0.4216, + "step": 34734 + }, + { + "epoch": 12.83, + "learning_rate": 1.718376647370366e-06, + "loss": 0.4017, + "step": 34735 + }, + { + "epoch": 12.83, + "learning_rate": 1.7183396970070205e-06, + "loss": 0.2209, + "step": 34736 + }, + { + "epoch": 12.83, + "learning_rate": 1.7183027466436755e-06, + "loss": 0.3106, + "step": 34737 + }, + { + "epoch": 12.83, + "learning_rate": 1.7182657962803301e-06, + "loss": 0.3532, + "step": 34738 + }, + { + "epoch": 12.83, + "learning_rate": 1.718228845916985e-06, + "loss": 0.2921, + "step": 34739 + }, + { + "epoch": 12.83, + "learning_rate": 1.7181918955536396e-06, + "loss": 0.2827, + "step": 34740 + }, + { + "epoch": 12.83, + "learning_rate": 1.7181549451902943e-06, + "loss": 0.3222, + "step": 34741 + }, + { + "epoch": 12.83, + "learning_rate": 1.7181179948269491e-06, + "loss": 0.1988, + "step": 34742 + }, + { + "epoch": 12.83, + "learning_rate": 1.7180810444636038e-06, + "loss": 0.2955, + "step": 34743 + }, + { + "epoch": 12.83, + "learning_rate": 1.7180440941002588e-06, + "loss": 0.3631, + "step": 34744 + }, + { + "epoch": 12.83, + "learning_rate": 1.7180071437369135e-06, + "loss": 0.3998, + "step": 34745 + }, + { + "epoch": 12.83, + "learning_rate": 1.7179701933735683e-06, + "loss": 0.3536, + "step": 34746 + }, + { + "epoch": 12.83, + "learning_rate": 1.717933243010223e-06, + "loss": 0.221, + "step": 34747 + }, + { + "epoch": 12.83, + "learning_rate": 1.7178962926468778e-06, + "loss": 0.218, + "step": 34748 + }, + { + "epoch": 12.83, + "learning_rate": 1.7178593422835324e-06, + "loss": 0.2962, + "step": 34749 + }, + { + "epoch": 12.83, + "learning_rate": 1.7178223919201875e-06, + "loss": 0.3047, + "step": 34750 + }, + { + "epoch": 12.83, + "learning_rate": 1.7177854415568421e-06, + "loss": 0.3811, + "step": 34751 + }, + { + "epoch": 12.83, + "learning_rate": 1.7177484911934965e-06, + "loss": 0.2286, + "step": 34752 + }, + { + "epoch": 12.83, + "learning_rate": 1.7177115408301516e-06, + "loss": 0.333, + "step": 34753 + }, + { + "epoch": 12.83, + "learning_rate": 1.7176745904668062e-06, + "loss": 0.3209, + "step": 34754 + }, + { + "epoch": 12.83, + "learning_rate": 1.717637640103461e-06, + "loss": 0.247, + "step": 34755 + }, + { + "epoch": 12.83, + "learning_rate": 1.7176006897401157e-06, + "loss": 0.324, + "step": 34756 + }, + { + "epoch": 12.83, + "learning_rate": 1.7175637393767708e-06, + "loss": 0.4565, + "step": 34757 + }, + { + "epoch": 12.84, + "learning_rate": 1.7175267890134252e-06, + "loss": 0.2668, + "step": 34758 + }, + { + "epoch": 12.84, + "learning_rate": 1.7174898386500803e-06, + "loss": 0.2327, + "step": 34759 + }, + { + "epoch": 12.84, + "learning_rate": 1.717452888286735e-06, + "loss": 0.3417, + "step": 34760 + }, + { + "epoch": 12.84, + "learning_rate": 1.7174159379233895e-06, + "loss": 0.4659, + "step": 34761 + }, + { + "epoch": 12.84, + "learning_rate": 1.7173789875600444e-06, + "loss": 0.2841, + "step": 34762 + }, + { + "epoch": 12.84, + "learning_rate": 1.717342037196699e-06, + "loss": 0.4573, + "step": 34763 + }, + { + "epoch": 12.84, + "learning_rate": 1.7173050868333539e-06, + "loss": 0.1953, + "step": 34764 + }, + { + "epoch": 12.84, + "learning_rate": 1.7172681364700085e-06, + "loss": 0.1441, + "step": 34765 + }, + { + "epoch": 12.84, + "learning_rate": 1.7172311861066636e-06, + "loss": 0.2408, + "step": 34766 + }, + { + "epoch": 12.84, + "learning_rate": 1.7171942357433182e-06, + "loss": 0.424, + "step": 34767 + }, + { + "epoch": 12.84, + "learning_rate": 1.717157285379973e-06, + "loss": 0.1848, + "step": 34768 + }, + { + "epoch": 12.84, + "learning_rate": 1.7171203350166277e-06, + "loss": 0.2846, + "step": 34769 + }, + { + "epoch": 12.84, + "learning_rate": 1.7170833846532825e-06, + "loss": 0.2491, + "step": 34770 + }, + { + "epoch": 12.84, + "learning_rate": 1.7170464342899372e-06, + "loss": 0.2221, + "step": 34771 + }, + { + "epoch": 12.84, + "learning_rate": 1.7170094839265918e-06, + "loss": 0.4394, + "step": 34772 + }, + { + "epoch": 12.84, + "learning_rate": 1.7169725335632469e-06, + "loss": 0.2828, + "step": 34773 + }, + { + "epoch": 12.84, + "learning_rate": 1.7169355831999015e-06, + "loss": 0.2544, + "step": 34774 + }, + { + "epoch": 12.84, + "learning_rate": 1.7168986328365564e-06, + "loss": 0.2999, + "step": 34775 + }, + { + "epoch": 12.84, + "learning_rate": 1.716861682473211e-06, + "loss": 0.202, + "step": 34776 + }, + { + "epoch": 12.84, + "learning_rate": 1.7168247321098658e-06, + "loss": 0.2998, + "step": 34777 + }, + { + "epoch": 12.84, + "learning_rate": 1.7167877817465205e-06, + "loss": 0.2743, + "step": 34778 + }, + { + "epoch": 12.84, + "learning_rate": 1.7167508313831755e-06, + "loss": 0.3272, + "step": 34779 + }, + { + "epoch": 12.84, + "learning_rate": 1.71671388101983e-06, + "loss": 0.2696, + "step": 34780 + }, + { + "epoch": 12.84, + "learning_rate": 1.7166769306564846e-06, + "loss": 0.4695, + "step": 34781 + }, + { + "epoch": 12.84, + "learning_rate": 1.7166399802931397e-06, + "loss": 0.2251, + "step": 34782 + }, + { + "epoch": 12.84, + "learning_rate": 1.7166030299297943e-06, + "loss": 0.3479, + "step": 34783 + }, + { + "epoch": 12.84, + "learning_rate": 1.7165660795664491e-06, + "loss": 0.2188, + "step": 34784 + }, + { + "epoch": 12.85, + "learning_rate": 1.7165291292031038e-06, + "loss": 0.2889, + "step": 34785 + }, + { + "epoch": 12.85, + "learning_rate": 1.7164921788397586e-06, + "loss": 0.2846, + "step": 34786 + }, + { + "epoch": 12.85, + "learning_rate": 1.7164552284764133e-06, + "loss": 0.2245, + "step": 34787 + }, + { + "epoch": 12.85, + "learning_rate": 1.7164182781130683e-06, + "loss": 0.3537, + "step": 34788 + }, + { + "epoch": 12.85, + "learning_rate": 1.716381327749723e-06, + "loss": 0.3344, + "step": 34789 + }, + { + "epoch": 12.85, + "learning_rate": 1.7163443773863778e-06, + "loss": 0.1798, + "step": 34790 + }, + { + "epoch": 12.85, + "learning_rate": 1.7163074270230324e-06, + "loss": 0.3203, + "step": 34791 + }, + { + "epoch": 12.85, + "learning_rate": 1.716270476659687e-06, + "loss": 0.2276, + "step": 34792 + }, + { + "epoch": 12.85, + "learning_rate": 1.716233526296342e-06, + "loss": 0.2463, + "step": 34793 + }, + { + "epoch": 12.85, + "learning_rate": 1.7161965759329966e-06, + "loss": 0.2125, + "step": 34794 + }, + { + "epoch": 12.85, + "learning_rate": 1.7161596255696516e-06, + "loss": 0.282, + "step": 34795 + }, + { + "epoch": 12.85, + "learning_rate": 1.7161226752063063e-06, + "loss": 0.3531, + "step": 34796 + }, + { + "epoch": 12.85, + "learning_rate": 1.7160857248429611e-06, + "loss": 0.2712, + "step": 34797 + }, + { + "epoch": 12.85, + "learning_rate": 1.7160487744796157e-06, + "loss": 0.3237, + "step": 34798 + }, + { + "epoch": 12.85, + "learning_rate": 1.7160118241162706e-06, + "loss": 0.2265, + "step": 34799 + }, + { + "epoch": 12.85, + "learning_rate": 1.7159748737529252e-06, + "loss": 0.2712, + "step": 34800 + }, + { + "epoch": 12.85, + "learning_rate": 1.7159379233895799e-06, + "loss": 0.213, + "step": 34801 + }, + { + "epoch": 12.85, + "learning_rate": 1.715900973026235e-06, + "loss": 0.3055, + "step": 34802 + }, + { + "epoch": 12.85, + "learning_rate": 1.7158640226628894e-06, + "loss": 0.3901, + "step": 34803 + }, + { + "epoch": 12.85, + "learning_rate": 1.7158270722995444e-06, + "loss": 0.2749, + "step": 34804 + }, + { + "epoch": 12.85, + "learning_rate": 1.715790121936199e-06, + "loss": 0.3238, + "step": 34805 + }, + { + "epoch": 12.85, + "learning_rate": 1.715753171572854e-06, + "loss": 0.3205, + "step": 34806 + }, + { + "epoch": 12.85, + "learning_rate": 1.7157162212095085e-06, + "loss": 0.3679, + "step": 34807 + }, + { + "epoch": 12.85, + "learning_rate": 1.7156792708461636e-06, + "loss": 0.2009, + "step": 34808 + }, + { + "epoch": 12.85, + "learning_rate": 1.715642320482818e-06, + "loss": 0.3583, + "step": 34809 + }, + { + "epoch": 12.85, + "learning_rate": 1.715605370119473e-06, + "loss": 0.2989, + "step": 34810 + }, + { + "epoch": 12.85, + "learning_rate": 1.7155684197561277e-06, + "loss": 0.2149, + "step": 34811 + }, + { + "epoch": 12.86, + "learning_rate": 1.7155314693927824e-06, + "loss": 0.3245, + "step": 34812 + }, + { + "epoch": 12.86, + "learning_rate": 1.7154945190294372e-06, + "loss": 0.255, + "step": 34813 + }, + { + "epoch": 12.86, + "learning_rate": 1.7154575686660918e-06, + "loss": 0.2601, + "step": 34814 + }, + { + "epoch": 12.86, + "learning_rate": 1.7154206183027467e-06, + "loss": 0.3164, + "step": 34815 + }, + { + "epoch": 12.86, + "learning_rate": 1.7153836679394013e-06, + "loss": 0.4038, + "step": 34816 + }, + { + "epoch": 12.86, + "learning_rate": 1.7153467175760564e-06, + "loss": 0.2165, + "step": 34817 + }, + { + "epoch": 12.86, + "learning_rate": 1.715309767212711e-06, + "loss": 0.4624, + "step": 34818 + }, + { + "epoch": 12.86, + "learning_rate": 1.7152728168493659e-06, + "loss": 0.2857, + "step": 34819 + }, + { + "epoch": 12.86, + "learning_rate": 1.7152358664860205e-06, + "loss": 0.4471, + "step": 34820 + }, + { + "epoch": 12.86, + "learning_rate": 1.7151989161226751e-06, + "loss": 0.2615, + "step": 34821 + }, + { + "epoch": 12.86, + "learning_rate": 1.71516196575933e-06, + "loss": 0.4707, + "step": 34822 + }, + { + "epoch": 12.86, + "learning_rate": 1.7151250153959846e-06, + "loss": 0.2961, + "step": 34823 + }, + { + "epoch": 12.86, + "learning_rate": 1.7150880650326397e-06, + "loss": 0.3545, + "step": 34824 + }, + { + "epoch": 12.86, + "learning_rate": 1.7150511146692941e-06, + "loss": 0.4145, + "step": 34825 + }, + { + "epoch": 12.86, + "learning_rate": 1.7150141643059492e-06, + "loss": 0.2446, + "step": 34826 + }, + { + "epoch": 12.86, + "learning_rate": 1.7149772139426038e-06, + "loss": 0.1988, + "step": 34827 + }, + { + "epoch": 12.86, + "learning_rate": 1.7149402635792587e-06, + "loss": 0.2478, + "step": 34828 + }, + { + "epoch": 12.86, + "learning_rate": 1.7149033132159133e-06, + "loss": 0.2527, + "step": 34829 + }, + { + "epoch": 12.86, + "learning_rate": 1.7148663628525683e-06, + "loss": 0.2972, + "step": 34830 + }, + { + "epoch": 12.86, + "learning_rate": 1.7148294124892228e-06, + "loss": 0.3344, + "step": 34831 + }, + { + "epoch": 12.86, + "learning_rate": 1.7147924621258774e-06, + "loss": 0.2174, + "step": 34832 + }, + { + "epoch": 12.86, + "learning_rate": 1.7147555117625325e-06, + "loss": 0.3389, + "step": 34833 + }, + { + "epoch": 12.86, + "learning_rate": 1.714718561399187e-06, + "loss": 0.3666, + "step": 34834 + }, + { + "epoch": 12.86, + "learning_rate": 1.714681611035842e-06, + "loss": 0.3907, + "step": 34835 + }, + { + "epoch": 12.86, + "learning_rate": 1.7146446606724966e-06, + "loss": 0.4048, + "step": 34836 + }, + { + "epoch": 12.86, + "learning_rate": 1.7146077103091514e-06, + "loss": 0.2016, + "step": 34837 + }, + { + "epoch": 12.86, + "learning_rate": 1.714570759945806e-06, + "loss": 0.3461, + "step": 34838 + }, + { + "epoch": 12.87, + "learning_rate": 1.7145338095824611e-06, + "loss": 0.3262, + "step": 34839 + }, + { + "epoch": 12.87, + "learning_rate": 1.7144968592191158e-06, + "loss": 0.3002, + "step": 34840 + }, + { + "epoch": 12.87, + "learning_rate": 1.7144599088557704e-06, + "loss": 0.3194, + "step": 34841 + }, + { + "epoch": 12.87, + "learning_rate": 1.7144229584924253e-06, + "loss": 0.3839, + "step": 34842 + }, + { + "epoch": 12.87, + "learning_rate": 1.7143860081290799e-06, + "loss": 0.293, + "step": 34843 + }, + { + "epoch": 12.87, + "learning_rate": 1.7143490577657347e-06, + "loss": 0.3426, + "step": 34844 + }, + { + "epoch": 12.87, + "learning_rate": 1.7143121074023894e-06, + "loss": 0.285, + "step": 34845 + }, + { + "epoch": 12.87, + "learning_rate": 1.7142751570390444e-06, + "loss": 0.2033, + "step": 34846 + }, + { + "epoch": 12.87, + "learning_rate": 1.714238206675699e-06, + "loss": 0.2523, + "step": 34847 + }, + { + "epoch": 12.87, + "learning_rate": 1.714201256312354e-06, + "loss": 0.2964, + "step": 34848 + }, + { + "epoch": 12.87, + "learning_rate": 1.7141643059490086e-06, + "loss": 0.3711, + "step": 34849 + }, + { + "epoch": 12.87, + "learning_rate": 1.7141273555856632e-06, + "loss": 0.1728, + "step": 34850 + }, + { + "epoch": 12.87, + "learning_rate": 1.714090405222318e-06, + "loss": 0.3035, + "step": 34851 + }, + { + "epoch": 12.87, + "learning_rate": 1.7140534548589727e-06, + "loss": 0.2722, + "step": 34852 + }, + { + "epoch": 12.87, + "learning_rate": 1.7140165044956277e-06, + "loss": 0.3059, + "step": 34853 + }, + { + "epoch": 12.87, + "learning_rate": 1.7139795541322822e-06, + "loss": 0.2516, + "step": 34854 + }, + { + "epoch": 12.87, + "learning_rate": 1.7139426037689372e-06, + "loss": 0.3963, + "step": 34855 + }, + { + "epoch": 12.87, + "learning_rate": 1.7139056534055919e-06, + "loss": 0.2537, + "step": 34856 + }, + { + "epoch": 12.87, + "learning_rate": 1.7138687030422467e-06, + "loss": 0.3012, + "step": 34857 + }, + { + "epoch": 12.87, + "learning_rate": 1.7138317526789013e-06, + "loss": 0.334, + "step": 34858 + }, + { + "epoch": 12.87, + "learning_rate": 1.7137948023155562e-06, + "loss": 0.2975, + "step": 34859 + }, + { + "epoch": 12.87, + "learning_rate": 1.7137578519522108e-06, + "loss": 0.4064, + "step": 34860 + }, + { + "epoch": 12.87, + "learning_rate": 1.7137209015888655e-06, + "loss": 0.2418, + "step": 34861 + }, + { + "epoch": 12.87, + "learning_rate": 1.7136839512255205e-06, + "loss": 0.2863, + "step": 34862 + }, + { + "epoch": 12.87, + "learning_rate": 1.7136470008621752e-06, + "loss": 0.293, + "step": 34863 + }, + { + "epoch": 12.87, + "learning_rate": 1.71361005049883e-06, + "loss": 0.2463, + "step": 34864 + }, + { + "epoch": 12.87, + "learning_rate": 1.7135731001354846e-06, + "loss": 0.1954, + "step": 34865 + }, + { + "epoch": 12.88, + "learning_rate": 1.7135361497721395e-06, + "loss": 0.4187, + "step": 34866 + }, + { + "epoch": 12.88, + "learning_rate": 1.7134991994087941e-06, + "loss": 0.3468, + "step": 34867 + }, + { + "epoch": 12.88, + "learning_rate": 1.7134622490454492e-06, + "loss": 0.2466, + "step": 34868 + }, + { + "epoch": 12.88, + "learning_rate": 1.7134252986821038e-06, + "loss": 0.3453, + "step": 34869 + }, + { + "epoch": 12.88, + "learning_rate": 1.7133883483187583e-06, + "loss": 0.3103, + "step": 34870 + }, + { + "epoch": 12.88, + "learning_rate": 1.7133513979554133e-06, + "loss": 0.3483, + "step": 34871 + }, + { + "epoch": 12.88, + "learning_rate": 1.713314447592068e-06, + "loss": 0.3707, + "step": 34872 + }, + { + "epoch": 12.88, + "learning_rate": 1.7132774972287228e-06, + "loss": 0.3687, + "step": 34873 + }, + { + "epoch": 12.88, + "learning_rate": 1.7132405468653774e-06, + "loss": 0.2631, + "step": 34874 + }, + { + "epoch": 12.88, + "learning_rate": 1.7132035965020325e-06, + "loss": 0.2892, + "step": 34875 + }, + { + "epoch": 12.88, + "learning_rate": 1.713166646138687e-06, + "loss": 0.2381, + "step": 34876 + }, + { + "epoch": 12.88, + "learning_rate": 1.713129695775342e-06, + "loss": 0.2204, + "step": 34877 + }, + { + "epoch": 12.88, + "learning_rate": 1.7130927454119966e-06, + "loss": 0.3105, + "step": 34878 + }, + { + "epoch": 12.88, + "learning_rate": 1.7130557950486515e-06, + "loss": 0.3492, + "step": 34879 + }, + { + "epoch": 12.88, + "learning_rate": 1.713018844685306e-06, + "loss": 0.3076, + "step": 34880 + }, + { + "epoch": 12.88, + "learning_rate": 1.7129818943219607e-06, + "loss": 0.24, + "step": 34881 + }, + { + "epoch": 12.88, + "learning_rate": 1.7129449439586156e-06, + "loss": 0.2436, + "step": 34882 + }, + { + "epoch": 12.88, + "learning_rate": 1.7129079935952702e-06, + "loss": 0.2119, + "step": 34883 + }, + { + "epoch": 12.88, + "learning_rate": 1.7128710432319253e-06, + "loss": 0.2815, + "step": 34884 + }, + { + "epoch": 12.88, + "learning_rate": 1.71283409286858e-06, + "loss": 0.3582, + "step": 34885 + }, + { + "epoch": 12.88, + "learning_rate": 1.7127971425052348e-06, + "loss": 0.301, + "step": 34886 + }, + { + "epoch": 12.88, + "learning_rate": 1.7127601921418894e-06, + "loss": 0.1888, + "step": 34887 + }, + { + "epoch": 12.88, + "learning_rate": 1.7127232417785442e-06, + "loss": 0.2781, + "step": 34888 + }, + { + "epoch": 12.88, + "learning_rate": 1.7126862914151989e-06, + "loss": 0.2197, + "step": 34889 + }, + { + "epoch": 12.88, + "learning_rate": 1.7126493410518535e-06, + "loss": 0.2846, + "step": 34890 + }, + { + "epoch": 12.88, + "learning_rate": 1.7126123906885086e-06, + "loss": 0.2135, + "step": 34891 + }, + { + "epoch": 12.88, + "learning_rate": 1.712575440325163e-06, + "loss": 0.3598, + "step": 34892 + }, + { + "epoch": 12.89, + "learning_rate": 1.712538489961818e-06, + "loss": 0.2403, + "step": 34893 + }, + { + "epoch": 12.89, + "learning_rate": 1.7125015395984727e-06, + "loss": 0.2472, + "step": 34894 + }, + { + "epoch": 12.89, + "learning_rate": 1.7124645892351276e-06, + "loss": 0.3991, + "step": 34895 + }, + { + "epoch": 12.89, + "learning_rate": 1.7124276388717822e-06, + "loss": 0.2775, + "step": 34896 + }, + { + "epoch": 12.89, + "learning_rate": 1.7123906885084372e-06, + "loss": 0.289, + "step": 34897 + }, + { + "epoch": 12.89, + "learning_rate": 1.7123537381450917e-06, + "loss": 0.3554, + "step": 34898 + }, + { + "epoch": 12.89, + "learning_rate": 1.7123167877817467e-06, + "loss": 0.2683, + "step": 34899 + }, + { + "epoch": 12.89, + "learning_rate": 1.7122798374184014e-06, + "loss": 0.2981, + "step": 34900 + }, + { + "epoch": 12.89, + "learning_rate": 1.712242887055056e-06, + "loss": 0.193, + "step": 34901 + }, + { + "epoch": 12.89, + "learning_rate": 1.7122059366917109e-06, + "loss": 0.1901, + "step": 34902 + }, + { + "epoch": 12.89, + "learning_rate": 1.7121689863283655e-06, + "loss": 0.303, + "step": 34903 + }, + { + "epoch": 12.89, + "learning_rate": 1.7121320359650203e-06, + "loss": 0.4769, + "step": 34904 + }, + { + "epoch": 12.89, + "learning_rate": 1.712095085601675e-06, + "loss": 0.411, + "step": 34905 + }, + { + "epoch": 12.89, + "learning_rate": 1.71205813523833e-06, + "loss": 0.3013, + "step": 34906 + }, + { + "epoch": 12.89, + "learning_rate": 1.7120211848749847e-06, + "loss": 0.3187, + "step": 34907 + }, + { + "epoch": 12.89, + "learning_rate": 1.7119842345116395e-06, + "loss": 0.2805, + "step": 34908 + }, + { + "epoch": 12.89, + "learning_rate": 1.7119472841482942e-06, + "loss": 0.2332, + "step": 34909 + }, + { + "epoch": 12.89, + "learning_rate": 1.7119103337849488e-06, + "loss": 0.2979, + "step": 34910 + }, + { + "epoch": 12.89, + "learning_rate": 1.7118733834216036e-06, + "loss": 0.2901, + "step": 34911 + }, + { + "epoch": 12.89, + "learning_rate": 1.7118364330582583e-06, + "loss": 0.3024, + "step": 34912 + }, + { + "epoch": 12.89, + "learning_rate": 1.7117994826949133e-06, + "loss": 0.4088, + "step": 34913 + }, + { + "epoch": 12.89, + "learning_rate": 1.711762532331568e-06, + "loss": 0.3055, + "step": 34914 + }, + { + "epoch": 12.89, + "learning_rate": 1.7117255819682228e-06, + "loss": 0.2605, + "step": 34915 + }, + { + "epoch": 12.89, + "learning_rate": 1.7116886316048775e-06, + "loss": 0.3123, + "step": 34916 + }, + { + "epoch": 12.89, + "learning_rate": 1.7116516812415323e-06, + "loss": 0.212, + "step": 34917 + }, + { + "epoch": 12.89, + "learning_rate": 1.711614730878187e-06, + "loss": 0.2538, + "step": 34918 + }, + { + "epoch": 12.89, + "learning_rate": 1.711577780514842e-06, + "loss": 0.3083, + "step": 34919 + }, + { + "epoch": 12.9, + "learning_rate": 1.7115408301514966e-06, + "loss": 0.3362, + "step": 34920 + }, + { + "epoch": 12.9, + "learning_rate": 1.711503879788151e-06, + "loss": 0.2786, + "step": 34921 + }, + { + "epoch": 12.9, + "learning_rate": 1.7114669294248061e-06, + "loss": 0.2654, + "step": 34922 + }, + { + "epoch": 12.9, + "learning_rate": 1.7114299790614608e-06, + "loss": 0.2686, + "step": 34923 + }, + { + "epoch": 12.9, + "learning_rate": 1.7113930286981156e-06, + "loss": 0.2094, + "step": 34924 + }, + { + "epoch": 12.9, + "learning_rate": 1.7113560783347702e-06, + "loss": 0.3773, + "step": 34925 + }, + { + "epoch": 12.9, + "learning_rate": 1.7113191279714253e-06, + "loss": 0.2025, + "step": 34926 + }, + { + "epoch": 12.9, + "learning_rate": 1.7112821776080797e-06, + "loss": 0.2659, + "step": 34927 + }, + { + "epoch": 12.9, + "learning_rate": 1.7112452272447348e-06, + "loss": 0.2471, + "step": 34928 + }, + { + "epoch": 12.9, + "learning_rate": 1.7112082768813894e-06, + "loss": 0.3997, + "step": 34929 + }, + { + "epoch": 12.9, + "learning_rate": 1.711171326518044e-06, + "loss": 0.1632, + "step": 34930 + }, + { + "epoch": 12.9, + "learning_rate": 1.711134376154699e-06, + "loss": 0.3408, + "step": 34931 + }, + { + "epoch": 12.9, + "learning_rate": 1.7110974257913535e-06, + "loss": 0.2179, + "step": 34932 + }, + { + "epoch": 12.9, + "learning_rate": 1.7110604754280084e-06, + "loss": 0.3561, + "step": 34933 + }, + { + "epoch": 12.9, + "learning_rate": 1.711023525064663e-06, + "loss": 0.2323, + "step": 34934 + }, + { + "epoch": 12.9, + "learning_rate": 1.710986574701318e-06, + "loss": 0.2238, + "step": 34935 + }, + { + "epoch": 12.9, + "learning_rate": 1.7109496243379727e-06, + "loss": 0.2434, + "step": 34936 + }, + { + "epoch": 12.9, + "learning_rate": 1.7109126739746276e-06, + "loss": 0.3497, + "step": 34937 + }, + { + "epoch": 12.9, + "learning_rate": 1.7108757236112822e-06, + "loss": 0.2971, + "step": 34938 + }, + { + "epoch": 12.9, + "learning_rate": 1.710838773247937e-06, + "loss": 0.4342, + "step": 34939 + }, + { + "epoch": 12.9, + "learning_rate": 1.7108018228845917e-06, + "loss": 0.3569, + "step": 34940 + }, + { + "epoch": 12.9, + "learning_rate": 1.7107648725212463e-06, + "loss": 0.3691, + "step": 34941 + }, + { + "epoch": 12.9, + "learning_rate": 1.7107279221579014e-06, + "loss": 0.3403, + "step": 34942 + }, + { + "epoch": 12.9, + "learning_rate": 1.7106909717945558e-06, + "loss": 0.198, + "step": 34943 + }, + { + "epoch": 12.9, + "learning_rate": 1.7106540214312109e-06, + "loss": 0.3181, + "step": 34944 + }, + { + "epoch": 12.9, + "learning_rate": 1.7106170710678655e-06, + "loss": 0.3989, + "step": 34945 + }, + { + "epoch": 12.9, + "learning_rate": 1.7105801207045204e-06, + "loss": 0.2732, + "step": 34946 + }, + { + "epoch": 12.91, + "learning_rate": 1.710543170341175e-06, + "loss": 0.3036, + "step": 34947 + }, + { + "epoch": 12.91, + "learning_rate": 1.71050621997783e-06, + "loss": 0.1465, + "step": 34948 + }, + { + "epoch": 12.91, + "learning_rate": 1.7104692696144845e-06, + "loss": 0.3787, + "step": 34949 + }, + { + "epoch": 12.91, + "learning_rate": 1.7104323192511391e-06, + "loss": 0.3517, + "step": 34950 + }, + { + "epoch": 12.91, + "learning_rate": 1.7103953688877942e-06, + "loss": 0.2961, + "step": 34951 + }, + { + "epoch": 12.91, + "learning_rate": 1.7103584185244488e-06, + "loss": 0.366, + "step": 34952 + }, + { + "epoch": 12.91, + "learning_rate": 1.7103214681611037e-06, + "loss": 0.3816, + "step": 34953 + }, + { + "epoch": 12.91, + "learning_rate": 1.7102845177977583e-06, + "loss": 0.3448, + "step": 34954 + }, + { + "epoch": 12.91, + "learning_rate": 1.7102475674344131e-06, + "loss": 0.3033, + "step": 34955 + }, + { + "epoch": 12.91, + "learning_rate": 1.7102106170710678e-06, + "loss": 0.3255, + "step": 34956 + }, + { + "epoch": 12.91, + "learning_rate": 1.7101736667077228e-06, + "loss": 0.4492, + "step": 34957 + }, + { + "epoch": 12.91, + "learning_rate": 1.7101367163443775e-06, + "loss": 0.3004, + "step": 34958 + }, + { + "epoch": 12.91, + "learning_rate": 1.7100997659810323e-06, + "loss": 0.3942, + "step": 34959 + }, + { + "epoch": 12.91, + "learning_rate": 1.710062815617687e-06, + "loss": 0.2305, + "step": 34960 + }, + { + "epoch": 12.91, + "learning_rate": 1.7100258652543416e-06, + "loss": 0.3442, + "step": 34961 + }, + { + "epoch": 12.91, + "learning_rate": 1.7099889148909965e-06, + "loss": 0.3628, + "step": 34962 + }, + { + "epoch": 12.91, + "learning_rate": 1.709951964527651e-06, + "loss": 0.4801, + "step": 34963 + }, + { + "epoch": 12.91, + "learning_rate": 1.7099150141643061e-06, + "loss": 0.2238, + "step": 34964 + }, + { + "epoch": 12.91, + "learning_rate": 1.7098780638009606e-06, + "loss": 0.2954, + "step": 34965 + }, + { + "epoch": 12.91, + "learning_rate": 1.7098411134376156e-06, + "loss": 0.2357, + "step": 34966 + }, + { + "epoch": 12.91, + "learning_rate": 1.7098041630742703e-06, + "loss": 0.3985, + "step": 34967 + }, + { + "epoch": 12.91, + "learning_rate": 1.7097672127109251e-06, + "loss": 0.2107, + "step": 34968 + }, + { + "epoch": 12.91, + "learning_rate": 1.7097302623475798e-06, + "loss": 0.2997, + "step": 34969 + }, + { + "epoch": 12.91, + "learning_rate": 1.7096933119842344e-06, + "loss": 0.2329, + "step": 34970 + }, + { + "epoch": 12.91, + "learning_rate": 1.7096563616208892e-06, + "loss": 0.3692, + "step": 34971 + }, + { + "epoch": 12.91, + "learning_rate": 1.7096194112575439e-06, + "loss": 0.3658, + "step": 34972 + }, + { + "epoch": 12.91, + "learning_rate": 1.709582460894199e-06, + "loss": 0.1851, + "step": 34973 + }, + { + "epoch": 12.92, + "learning_rate": 1.7095455105308536e-06, + "loss": 0.2145, + "step": 34974 + }, + { + "epoch": 12.92, + "learning_rate": 1.7095085601675084e-06, + "loss": 0.4298, + "step": 34975 + }, + { + "epoch": 12.92, + "learning_rate": 1.709471609804163e-06, + "loss": 0.1982, + "step": 34976 + }, + { + "epoch": 12.92, + "learning_rate": 1.709434659440818e-06, + "loss": 0.1984, + "step": 34977 + }, + { + "epoch": 12.92, + "learning_rate": 1.7093977090774725e-06, + "loss": 0.3013, + "step": 34978 + }, + { + "epoch": 12.92, + "learning_rate": 1.7093607587141276e-06, + "loss": 0.2014, + "step": 34979 + }, + { + "epoch": 12.92, + "learning_rate": 1.7093238083507822e-06, + "loss": 0.3205, + "step": 34980 + }, + { + "epoch": 12.92, + "learning_rate": 1.7092868579874369e-06, + "loss": 0.1709, + "step": 34981 + }, + { + "epoch": 12.92, + "learning_rate": 1.7092499076240917e-06, + "loss": 0.2205, + "step": 34982 + }, + { + "epoch": 12.92, + "learning_rate": 1.7092129572607464e-06, + "loss": 0.3015, + "step": 34983 + }, + { + "epoch": 12.92, + "learning_rate": 1.7091760068974012e-06, + "loss": 0.3281, + "step": 34984 + }, + { + "epoch": 12.92, + "learning_rate": 1.7091390565340558e-06, + "loss": 0.2328, + "step": 34985 + }, + { + "epoch": 12.92, + "learning_rate": 1.709102106170711e-06, + "loss": 0.2171, + "step": 34986 + }, + { + "epoch": 12.92, + "learning_rate": 1.7090651558073655e-06, + "loss": 0.2383, + "step": 34987 + }, + { + "epoch": 12.92, + "learning_rate": 1.7090282054440204e-06, + "loss": 0.2809, + "step": 34988 + }, + { + "epoch": 12.92, + "learning_rate": 1.708991255080675e-06, + "loss": 0.1722, + "step": 34989 + }, + { + "epoch": 12.92, + "learning_rate": 1.7089543047173297e-06, + "loss": 0.4104, + "step": 34990 + }, + { + "epoch": 12.92, + "learning_rate": 1.7089173543539845e-06, + "loss": 0.2575, + "step": 34991 + }, + { + "epoch": 12.92, + "learning_rate": 1.7088804039906391e-06, + "loss": 0.3271, + "step": 34992 + }, + { + "epoch": 12.92, + "learning_rate": 1.7088434536272942e-06, + "loss": 0.3118, + "step": 34993 + }, + { + "epoch": 12.92, + "learning_rate": 1.7088065032639486e-06, + "loss": 0.1817, + "step": 34994 + }, + { + "epoch": 12.92, + "learning_rate": 1.7087695529006037e-06, + "loss": 0.2544, + "step": 34995 + }, + { + "epoch": 12.92, + "learning_rate": 1.7087326025372583e-06, + "loss": 0.2933, + "step": 34996 + }, + { + "epoch": 12.92, + "learning_rate": 1.7086956521739132e-06, + "loss": 0.2749, + "step": 34997 + }, + { + "epoch": 12.92, + "learning_rate": 1.7086587018105678e-06, + "loss": 0.2331, + "step": 34998 + }, + { + "epoch": 12.92, + "learning_rate": 1.7086217514472229e-06, + "loss": 0.3378, + "step": 34999 + }, + { + "epoch": 12.92, + "learning_rate": 1.7085848010838773e-06, + "loss": 0.2863, + "step": 35000 + }, + { + "epoch": 12.93, + "learning_rate": 1.708547850720532e-06, + "loss": 0.305, + "step": 35001 + }, + { + "epoch": 12.93, + "learning_rate": 1.708510900357187e-06, + "loss": 0.3386, + "step": 35002 + }, + { + "epoch": 12.93, + "learning_rate": 1.7084739499938416e-06, + "loss": 0.4012, + "step": 35003 + }, + { + "epoch": 12.93, + "learning_rate": 1.7084369996304965e-06, + "loss": 0.3494, + "step": 35004 + }, + { + "epoch": 12.93, + "learning_rate": 1.7084000492671511e-06, + "loss": 0.2975, + "step": 35005 + }, + { + "epoch": 12.93, + "learning_rate": 1.708363098903806e-06, + "loss": 0.224, + "step": 35006 + }, + { + "epoch": 12.93, + "learning_rate": 1.7083261485404606e-06, + "loss": 0.3642, + "step": 35007 + }, + { + "epoch": 12.93, + "learning_rate": 1.7082891981771157e-06, + "loss": 0.3205, + "step": 35008 + }, + { + "epoch": 12.93, + "learning_rate": 1.7082522478137703e-06, + "loss": 0.323, + "step": 35009 + }, + { + "epoch": 12.93, + "learning_rate": 1.7082152974504247e-06, + "loss": 0.3796, + "step": 35010 + }, + { + "epoch": 12.93, + "learning_rate": 1.7081783470870798e-06, + "loss": 0.2577, + "step": 35011 + }, + { + "epoch": 12.93, + "learning_rate": 1.7081413967237344e-06, + "loss": 0.3159, + "step": 35012 + }, + { + "epoch": 12.93, + "learning_rate": 1.7081044463603893e-06, + "loss": 0.2907, + "step": 35013 + }, + { + "epoch": 12.93, + "learning_rate": 1.708067495997044e-06, + "loss": 0.279, + "step": 35014 + }, + { + "epoch": 12.93, + "learning_rate": 1.708030545633699e-06, + "loss": 0.3924, + "step": 35015 + }, + { + "epoch": 12.93, + "learning_rate": 1.7079935952703534e-06, + "loss": 0.2415, + "step": 35016 + }, + { + "epoch": 12.93, + "learning_rate": 1.7079566449070084e-06, + "loss": 0.2006, + "step": 35017 + }, + { + "epoch": 12.93, + "learning_rate": 1.707919694543663e-06, + "loss": 0.2756, + "step": 35018 + }, + { + "epoch": 12.93, + "learning_rate": 1.707882744180318e-06, + "loss": 0.2696, + "step": 35019 + }, + { + "epoch": 12.93, + "learning_rate": 1.7078457938169726e-06, + "loss": 0.2364, + "step": 35020 + }, + { + "epoch": 12.93, + "learning_rate": 1.7078088434536272e-06, + "loss": 0.3011, + "step": 35021 + }, + { + "epoch": 12.93, + "learning_rate": 1.707771893090282e-06, + "loss": 0.2915, + "step": 35022 + }, + { + "epoch": 12.93, + "learning_rate": 1.7077349427269367e-06, + "loss": 0.3348, + "step": 35023 + }, + { + "epoch": 12.93, + "learning_rate": 1.7076979923635917e-06, + "loss": 0.1699, + "step": 35024 + }, + { + "epoch": 12.93, + "learning_rate": 1.7076610420002464e-06, + "loss": 0.3262, + "step": 35025 + }, + { + "epoch": 12.93, + "learning_rate": 1.7076240916369012e-06, + "loss": 0.4923, + "step": 35026 + }, + { + "epoch": 12.93, + "learning_rate": 1.7075871412735559e-06, + "loss": 0.1872, + "step": 35027 + }, + { + "epoch": 12.94, + "learning_rate": 1.7075501909102107e-06, + "loss": 0.2924, + "step": 35028 + }, + { + "epoch": 12.94, + "learning_rate": 1.7075132405468654e-06, + "loss": 0.2846, + "step": 35029 + }, + { + "epoch": 12.94, + "learning_rate": 1.70747629018352e-06, + "loss": 0.2057, + "step": 35030 + }, + { + "epoch": 12.94, + "learning_rate": 1.707439339820175e-06, + "loss": 0.4121, + "step": 35031 + }, + { + "epoch": 12.94, + "learning_rate": 1.7074023894568297e-06, + "loss": 0.3261, + "step": 35032 + }, + { + "epoch": 12.94, + "learning_rate": 1.7073654390934845e-06, + "loss": 0.4342, + "step": 35033 + }, + { + "epoch": 12.94, + "learning_rate": 1.7073284887301392e-06, + "loss": 0.222, + "step": 35034 + }, + { + "epoch": 12.94, + "learning_rate": 1.707291538366794e-06, + "loss": 0.2852, + "step": 35035 + }, + { + "epoch": 12.94, + "learning_rate": 1.7072545880034487e-06, + "loss": 0.2848, + "step": 35036 + }, + { + "epoch": 12.94, + "learning_rate": 1.7072176376401037e-06, + "loss": 0.3337, + "step": 35037 + }, + { + "epoch": 12.94, + "learning_rate": 1.7071806872767583e-06, + "loss": 0.2441, + "step": 35038 + }, + { + "epoch": 12.94, + "learning_rate": 1.7071437369134132e-06, + "loss": 0.2236, + "step": 35039 + }, + { + "epoch": 12.94, + "learning_rate": 1.7071067865500678e-06, + "loss": 0.3635, + "step": 35040 + }, + { + "epoch": 12.94, + "learning_rate": 1.7070698361867225e-06, + "loss": 0.2362, + "step": 35041 + }, + { + "epoch": 12.94, + "learning_rate": 1.7070328858233773e-06, + "loss": 0.1971, + "step": 35042 + }, + { + "epoch": 12.94, + "learning_rate": 1.706995935460032e-06, + "loss": 0.271, + "step": 35043 + }, + { + "epoch": 12.94, + "learning_rate": 1.7069589850966868e-06, + "loss": 0.2402, + "step": 35044 + }, + { + "epoch": 12.94, + "learning_rate": 1.7069220347333414e-06, + "loss": 0.428, + "step": 35045 + }, + { + "epoch": 12.94, + "learning_rate": 1.7068850843699965e-06, + "loss": 0.3439, + "step": 35046 + }, + { + "epoch": 12.94, + "learning_rate": 1.7068481340066511e-06, + "loss": 0.3096, + "step": 35047 + }, + { + "epoch": 12.94, + "learning_rate": 1.706811183643306e-06, + "loss": 0.3161, + "step": 35048 + }, + { + "epoch": 12.94, + "learning_rate": 1.7067742332799606e-06, + "loss": 0.2675, + "step": 35049 + }, + { + "epoch": 12.94, + "learning_rate": 1.7067372829166153e-06, + "loss": 0.2393, + "step": 35050 + }, + { + "epoch": 12.94, + "learning_rate": 1.70670033255327e-06, + "loss": 0.3286, + "step": 35051 + }, + { + "epoch": 12.94, + "learning_rate": 1.7066633821899247e-06, + "loss": 0.4117, + "step": 35052 + }, + { + "epoch": 12.94, + "learning_rate": 1.7066264318265798e-06, + "loss": 0.2681, + "step": 35053 + }, + { + "epoch": 12.94, + "learning_rate": 1.7065894814632344e-06, + "loss": 0.3874, + "step": 35054 + }, + { + "epoch": 12.94, + "learning_rate": 1.7065525310998893e-06, + "loss": 0.4602, + "step": 35055 + }, + { + "epoch": 12.95, + "learning_rate": 1.706515580736544e-06, + "loss": 0.2963, + "step": 35056 + }, + { + "epoch": 12.95, + "learning_rate": 1.7064786303731988e-06, + "loss": 0.1715, + "step": 35057 + }, + { + "epoch": 12.95, + "learning_rate": 1.7064416800098534e-06, + "loss": 0.3237, + "step": 35058 + }, + { + "epoch": 12.95, + "learning_rate": 1.7064047296465085e-06, + "loss": 0.2494, + "step": 35059 + }, + { + "epoch": 12.95, + "learning_rate": 1.706367779283163e-06, + "loss": 0.2745, + "step": 35060 + }, + { + "epoch": 12.95, + "learning_rate": 1.7063308289198175e-06, + "loss": 0.4054, + "step": 35061 + }, + { + "epoch": 12.95, + "learning_rate": 1.7062938785564726e-06, + "loss": 0.453, + "step": 35062 + }, + { + "epoch": 12.95, + "learning_rate": 1.7062569281931272e-06, + "loss": 0.2308, + "step": 35063 + }, + { + "epoch": 12.95, + "learning_rate": 1.706219977829782e-06, + "loss": 0.2752, + "step": 35064 + }, + { + "epoch": 12.95, + "learning_rate": 1.7061830274664367e-06, + "loss": 0.3299, + "step": 35065 + }, + { + "epoch": 12.95, + "learning_rate": 1.7061460771030918e-06, + "loss": 0.348, + "step": 35066 + }, + { + "epoch": 12.95, + "learning_rate": 1.7061091267397462e-06, + "loss": 0.342, + "step": 35067 + }, + { + "epoch": 12.95, + "learning_rate": 1.7060721763764013e-06, + "loss": 0.2772, + "step": 35068 + }, + { + "epoch": 12.95, + "learning_rate": 1.7060352260130559e-06, + "loss": 0.2573, + "step": 35069 + }, + { + "epoch": 12.95, + "learning_rate": 1.7059982756497105e-06, + "loss": 0.2763, + "step": 35070 + }, + { + "epoch": 12.95, + "learning_rate": 1.7059613252863654e-06, + "loss": 0.2733, + "step": 35071 + }, + { + "epoch": 12.95, + "learning_rate": 1.70592437492302e-06, + "loss": 0.3107, + "step": 35072 + }, + { + "epoch": 12.95, + "learning_rate": 1.7058874245596749e-06, + "loss": 0.2114, + "step": 35073 + }, + { + "epoch": 12.95, + "learning_rate": 1.7058504741963295e-06, + "loss": 0.1898, + "step": 35074 + }, + { + "epoch": 12.95, + "learning_rate": 1.7058135238329846e-06, + "loss": 0.233, + "step": 35075 + }, + { + "epoch": 12.95, + "learning_rate": 1.7057765734696392e-06, + "loss": 0.2098, + "step": 35076 + }, + { + "epoch": 12.95, + "learning_rate": 1.705739623106294e-06, + "loss": 0.416, + "step": 35077 + }, + { + "epoch": 12.95, + "learning_rate": 1.7057026727429487e-06, + "loss": 0.1905, + "step": 35078 + }, + { + "epoch": 12.95, + "learning_rate": 1.7056657223796033e-06, + "loss": 0.2761, + "step": 35079 + }, + { + "epoch": 12.95, + "learning_rate": 1.7056287720162582e-06, + "loss": 0.2912, + "step": 35080 + }, + { + "epoch": 12.95, + "learning_rate": 1.7055918216529128e-06, + "loss": 0.4026, + "step": 35081 + }, + { + "epoch": 12.95, + "learning_rate": 1.7055548712895679e-06, + "loss": 0.2405, + "step": 35082 + }, + { + "epoch": 12.96, + "learning_rate": 1.7055179209262223e-06, + "loss": 0.3675, + "step": 35083 + }, + { + "epoch": 12.96, + "learning_rate": 1.7054809705628773e-06, + "loss": 0.2213, + "step": 35084 + }, + { + "epoch": 12.96, + "learning_rate": 1.705444020199532e-06, + "loss": 0.3589, + "step": 35085 + }, + { + "epoch": 12.96, + "learning_rate": 1.7054070698361868e-06, + "loss": 0.3595, + "step": 35086 + }, + { + "epoch": 12.96, + "learning_rate": 1.7053701194728415e-06, + "loss": 0.2978, + "step": 35087 + }, + { + "epoch": 12.96, + "learning_rate": 1.7053331691094965e-06, + "loss": 0.3759, + "step": 35088 + }, + { + "epoch": 12.96, + "learning_rate": 1.705296218746151e-06, + "loss": 0.3133, + "step": 35089 + }, + { + "epoch": 12.96, + "learning_rate": 1.7052592683828056e-06, + "loss": 0.4035, + "step": 35090 + }, + { + "epoch": 12.96, + "learning_rate": 1.7052223180194606e-06, + "loss": 0.3692, + "step": 35091 + }, + { + "epoch": 12.96, + "learning_rate": 1.7051853676561153e-06, + "loss": 0.2468, + "step": 35092 + }, + { + "epoch": 12.96, + "learning_rate": 1.7051484172927701e-06, + "loss": 0.37, + "step": 35093 + }, + { + "epoch": 12.96, + "learning_rate": 1.7051114669294248e-06, + "loss": 0.1794, + "step": 35094 + }, + { + "epoch": 12.96, + "learning_rate": 1.7050745165660796e-06, + "loss": 0.4623, + "step": 35095 + }, + { + "epoch": 12.96, + "learning_rate": 1.7050375662027343e-06, + "loss": 0.3493, + "step": 35096 + }, + { + "epoch": 12.96, + "learning_rate": 1.7050006158393893e-06, + "loss": 0.1924, + "step": 35097 + }, + { + "epoch": 12.96, + "learning_rate": 1.704963665476044e-06, + "loss": 0.2487, + "step": 35098 + }, + { + "epoch": 12.96, + "learning_rate": 1.7049267151126986e-06, + "loss": 0.1733, + "step": 35099 + }, + { + "epoch": 12.96, + "learning_rate": 1.7048897647493534e-06, + "loss": 0.3035, + "step": 35100 + }, + { + "epoch": 12.96, + "learning_rate": 1.704852814386008e-06, + "loss": 0.3714, + "step": 35101 + }, + { + "epoch": 12.96, + "learning_rate": 1.704815864022663e-06, + "loss": 0.3787, + "step": 35102 + }, + { + "epoch": 12.96, + "learning_rate": 1.7047789136593176e-06, + "loss": 0.3076, + "step": 35103 + }, + { + "epoch": 12.96, + "learning_rate": 1.7047419632959726e-06, + "loss": 0.2773, + "step": 35104 + }, + { + "epoch": 12.96, + "learning_rate": 1.7047050129326272e-06, + "loss": 0.3602, + "step": 35105 + }, + { + "epoch": 12.96, + "learning_rate": 1.704668062569282e-06, + "loss": 0.3286, + "step": 35106 + }, + { + "epoch": 12.96, + "learning_rate": 1.7046311122059367e-06, + "loss": 0.262, + "step": 35107 + }, + { + "epoch": 12.96, + "learning_rate": 1.7045941618425916e-06, + "loss": 0.225, + "step": 35108 + }, + { + "epoch": 12.96, + "learning_rate": 1.7045572114792462e-06, + "loss": 0.388, + "step": 35109 + }, + { + "epoch": 12.97, + "learning_rate": 1.7045202611159009e-06, + "loss": 0.2651, + "step": 35110 + }, + { + "epoch": 12.97, + "learning_rate": 1.704483310752556e-06, + "loss": 0.3783, + "step": 35111 + }, + { + "epoch": 12.97, + "learning_rate": 1.7044463603892103e-06, + "loss": 0.252, + "step": 35112 + }, + { + "epoch": 12.97, + "learning_rate": 1.7044094100258654e-06, + "loss": 0.4592, + "step": 35113 + }, + { + "epoch": 12.97, + "learning_rate": 1.70437245966252e-06, + "loss": 0.3111, + "step": 35114 + }, + { + "epoch": 12.97, + "learning_rate": 1.7043355092991749e-06, + "loss": 0.3841, + "step": 35115 + }, + { + "epoch": 12.97, + "learning_rate": 1.7042985589358295e-06, + "loss": 0.3365, + "step": 35116 + }, + { + "epoch": 12.97, + "learning_rate": 1.7042616085724846e-06, + "loss": 0.4117, + "step": 35117 + }, + { + "epoch": 12.97, + "learning_rate": 1.704224658209139e-06, + "loss": 0.3522, + "step": 35118 + }, + { + "epoch": 12.97, + "learning_rate": 1.7041877078457936e-06, + "loss": 0.374, + "step": 35119 + }, + { + "epoch": 12.97, + "learning_rate": 1.7041507574824487e-06, + "loss": 0.2939, + "step": 35120 + }, + { + "epoch": 12.97, + "learning_rate": 1.7041138071191033e-06, + "loss": 0.3418, + "step": 35121 + }, + { + "epoch": 12.97, + "learning_rate": 1.7040768567557582e-06, + "loss": 0.2077, + "step": 35122 + }, + { + "epoch": 12.97, + "learning_rate": 1.7040399063924128e-06, + "loss": 0.3128, + "step": 35123 + }, + { + "epoch": 12.97, + "learning_rate": 1.7040029560290677e-06, + "loss": 0.2871, + "step": 35124 + }, + { + "epoch": 12.97, + "learning_rate": 1.7039660056657223e-06, + "loss": 0.241, + "step": 35125 + }, + { + "epoch": 12.97, + "learning_rate": 1.7039290553023774e-06, + "loss": 0.4551, + "step": 35126 + }, + { + "epoch": 12.97, + "learning_rate": 1.703892104939032e-06, + "loss": 0.3593, + "step": 35127 + }, + { + "epoch": 12.97, + "learning_rate": 1.7038551545756869e-06, + "loss": 0.33, + "step": 35128 + }, + { + "epoch": 12.97, + "learning_rate": 1.7038182042123415e-06, + "loss": 0.1697, + "step": 35129 + }, + { + "epoch": 12.97, + "learning_rate": 1.7037812538489961e-06, + "loss": 0.2811, + "step": 35130 + }, + { + "epoch": 12.97, + "learning_rate": 1.703744303485651e-06, + "loss": 0.2841, + "step": 35131 + }, + { + "epoch": 12.97, + "learning_rate": 1.7037073531223056e-06, + "loss": 0.323, + "step": 35132 + }, + { + "epoch": 12.97, + "learning_rate": 1.7036704027589607e-06, + "loss": 0.4455, + "step": 35133 + }, + { + "epoch": 12.97, + "learning_rate": 1.703633452395615e-06, + "loss": 0.2874, + "step": 35134 + }, + { + "epoch": 12.97, + "learning_rate": 1.7035965020322702e-06, + "loss": 0.2961, + "step": 35135 + }, + { + "epoch": 12.97, + "learning_rate": 1.7035595516689248e-06, + "loss": 0.327, + "step": 35136 + }, + { + "epoch": 12.98, + "learning_rate": 1.7035226013055796e-06, + "loss": 0.4294, + "step": 35137 + }, + { + "epoch": 12.98, + "learning_rate": 1.7034856509422343e-06, + "loss": 0.3472, + "step": 35138 + }, + { + "epoch": 12.98, + "learning_rate": 1.703448700578889e-06, + "loss": 0.3243, + "step": 35139 + }, + { + "epoch": 12.98, + "learning_rate": 1.7034117502155438e-06, + "loss": 0.3719, + "step": 35140 + }, + { + "epoch": 12.98, + "learning_rate": 1.7033747998521984e-06, + "loss": 0.3784, + "step": 35141 + }, + { + "epoch": 12.98, + "learning_rate": 1.7033378494888535e-06, + "loss": 0.2939, + "step": 35142 + }, + { + "epoch": 12.98, + "learning_rate": 1.703300899125508e-06, + "loss": 0.3093, + "step": 35143 + }, + { + "epoch": 12.98, + "learning_rate": 1.703263948762163e-06, + "loss": 0.3209, + "step": 35144 + }, + { + "epoch": 12.98, + "learning_rate": 1.7032269983988176e-06, + "loss": 0.2517, + "step": 35145 + }, + { + "epoch": 12.98, + "learning_rate": 1.7031900480354724e-06, + "loss": 0.2718, + "step": 35146 + }, + { + "epoch": 12.98, + "learning_rate": 1.703153097672127e-06, + "loss": 0.3108, + "step": 35147 + }, + { + "epoch": 12.98, + "learning_rate": 1.7031161473087821e-06, + "loss": 0.2544, + "step": 35148 + }, + { + "epoch": 12.98, + "learning_rate": 1.7030791969454368e-06, + "loss": 0.2868, + "step": 35149 + }, + { + "epoch": 12.98, + "learning_rate": 1.7030422465820914e-06, + "loss": 0.2641, + "step": 35150 + }, + { + "epoch": 12.98, + "learning_rate": 1.7030052962187462e-06, + "loss": 0.3875, + "step": 35151 + }, + { + "epoch": 12.98, + "learning_rate": 1.7029683458554009e-06, + "loss": 0.3804, + "step": 35152 + }, + { + "epoch": 12.98, + "learning_rate": 1.7029313954920557e-06, + "loss": 0.2576, + "step": 35153 + }, + { + "epoch": 12.98, + "learning_rate": 1.7028944451287104e-06, + "loss": 0.2458, + "step": 35154 + }, + { + "epoch": 12.98, + "learning_rate": 1.7028574947653654e-06, + "loss": 0.2443, + "step": 35155 + }, + { + "epoch": 12.98, + "learning_rate": 1.7028205444020198e-06, + "loss": 0.2848, + "step": 35156 + }, + { + "epoch": 12.98, + "learning_rate": 1.702783594038675e-06, + "loss": 0.3007, + "step": 35157 + }, + { + "epoch": 12.98, + "learning_rate": 1.7027466436753295e-06, + "loss": 0.1931, + "step": 35158 + }, + { + "epoch": 12.98, + "learning_rate": 1.7027096933119842e-06, + "loss": 0.1928, + "step": 35159 + }, + { + "epoch": 12.98, + "learning_rate": 1.702672742948639e-06, + "loss": 0.3161, + "step": 35160 + }, + { + "epoch": 12.98, + "learning_rate": 1.7026357925852937e-06, + "loss": 0.3224, + "step": 35161 + }, + { + "epoch": 12.98, + "learning_rate": 1.7025988422219485e-06, + "loss": 0.3092, + "step": 35162 + }, + { + "epoch": 12.98, + "learning_rate": 1.7025618918586032e-06, + "loss": 0.2566, + "step": 35163 + }, + { + "epoch": 12.99, + "learning_rate": 1.7025249414952582e-06, + "loss": 0.2968, + "step": 35164 + }, + { + "epoch": 12.99, + "learning_rate": 1.7024879911319128e-06, + "loss": 0.3056, + "step": 35165 + }, + { + "epoch": 12.99, + "learning_rate": 1.7024510407685677e-06, + "loss": 0.3306, + "step": 35166 + }, + { + "epoch": 12.99, + "learning_rate": 1.7024140904052223e-06, + "loss": 0.1874, + "step": 35167 + }, + { + "epoch": 12.99, + "learning_rate": 1.7023771400418772e-06, + "loss": 0.2622, + "step": 35168 + }, + { + "epoch": 12.99, + "learning_rate": 1.7023401896785318e-06, + "loss": 0.3833, + "step": 35169 + }, + { + "epoch": 12.99, + "learning_rate": 1.7023032393151865e-06, + "loss": 0.2764, + "step": 35170 + }, + { + "epoch": 12.99, + "learning_rate": 1.7022662889518415e-06, + "loss": 0.2908, + "step": 35171 + }, + { + "epoch": 12.99, + "learning_rate": 1.7022293385884961e-06, + "loss": 0.394, + "step": 35172 + }, + { + "epoch": 12.99, + "learning_rate": 1.702192388225151e-06, + "loss": 0.2791, + "step": 35173 + }, + { + "epoch": 12.99, + "learning_rate": 1.7021554378618056e-06, + "loss": 0.2651, + "step": 35174 + }, + { + "epoch": 12.99, + "learning_rate": 1.7021184874984605e-06, + "loss": 0.2397, + "step": 35175 + }, + { + "epoch": 12.99, + "learning_rate": 1.7020815371351151e-06, + "loss": 0.3104, + "step": 35176 + }, + { + "epoch": 12.99, + "learning_rate": 1.7020445867717702e-06, + "loss": 0.2676, + "step": 35177 + }, + { + "epoch": 12.99, + "learning_rate": 1.7020076364084248e-06, + "loss": 0.3188, + "step": 35178 + }, + { + "epoch": 12.99, + "learning_rate": 1.7019706860450792e-06, + "loss": 0.2807, + "step": 35179 + }, + { + "epoch": 12.99, + "learning_rate": 1.7019337356817343e-06, + "loss": 0.2902, + "step": 35180 + }, + { + "epoch": 12.99, + "learning_rate": 1.701896785318389e-06, + "loss": 0.217, + "step": 35181 + }, + { + "epoch": 12.99, + "learning_rate": 1.7018598349550438e-06, + "loss": 0.358, + "step": 35182 + }, + { + "epoch": 12.99, + "learning_rate": 1.7018228845916984e-06, + "loss": 0.2284, + "step": 35183 + }, + { + "epoch": 12.99, + "learning_rate": 1.7017859342283535e-06, + "loss": 0.3, + "step": 35184 + }, + { + "epoch": 12.99, + "learning_rate": 1.701748983865008e-06, + "loss": 0.3632, + "step": 35185 + }, + { + "epoch": 12.99, + "learning_rate": 1.701712033501663e-06, + "loss": 0.4553, + "step": 35186 + }, + { + "epoch": 12.99, + "learning_rate": 1.7016750831383176e-06, + "loss": 0.214, + "step": 35187 + }, + { + "epoch": 12.99, + "learning_rate": 1.7016381327749724e-06, + "loss": 0.2654, + "step": 35188 + }, + { + "epoch": 12.99, + "learning_rate": 1.701601182411627e-06, + "loss": 0.3665, + "step": 35189 + }, + { + "epoch": 12.99, + "learning_rate": 1.7015642320482817e-06, + "loss": 0.3969, + "step": 35190 + }, + { + "epoch": 13.0, + "learning_rate": 1.7015272816849366e-06, + "loss": 0.2391, + "step": 35191 + }, + { + "epoch": 13.0, + "learning_rate": 1.7014903313215912e-06, + "loss": 0.2553, + "step": 35192 + }, + { + "epoch": 13.0, + "learning_rate": 1.7014533809582463e-06, + "loss": 0.2984, + "step": 35193 + }, + { + "epoch": 13.0, + "learning_rate": 1.701416430594901e-06, + "loss": 0.2069, + "step": 35194 + }, + { + "epoch": 13.0, + "learning_rate": 1.7013794802315558e-06, + "loss": 0.3142, + "step": 35195 + }, + { + "epoch": 13.0, + "learning_rate": 1.7013425298682104e-06, + "loss": 0.2932, + "step": 35196 + }, + { + "epoch": 13.0, + "learning_rate": 1.7013055795048652e-06, + "loss": 0.3427, + "step": 35197 + }, + { + "epoch": 13.0, + "learning_rate": 1.7012686291415199e-06, + "loss": 0.4438, + "step": 35198 + }, + { + "epoch": 13.0, + "learning_rate": 1.7012316787781745e-06, + "loss": 0.342, + "step": 35199 + }, + { + "epoch": 13.0, + "learning_rate": 1.7011947284148296e-06, + "loss": 0.2953, + "step": 35200 + }, + { + "epoch": 13.0, + "learning_rate": 1.701157778051484e-06, + "loss": 0.3103, + "step": 35201 + }, + { + "epoch": 13.0, + "learning_rate": 1.701120827688139e-06, + "loss": 0.3782, + "step": 35202 + }, + { + "epoch": 13.0, + "learning_rate": 1.7010838773247937e-06, + "loss": 0.3485, + "step": 35203 + }, + { + "epoch": 13.0, + "learning_rate": 1.7010469269614485e-06, + "loss": 0.0487, + "step": 35204 + }, + { + "epoch": 13.0, + "eval_accuracy": 0.8444152006279725, + "eval_loss": 0.3772488832473755, + "eval_runtime": 25.4774, + "eval_samples_per_second": 1700.096, + "eval_steps_per_second": 26.573, + "step": 35204 + }, + { + "epoch": 13.0, + "learning_rate": 1.7010099765981032e-06, + "loss": 0.3521, + "step": 35205 + }, + { + "epoch": 13.0, + "learning_rate": 1.7009730262347582e-06, + "loss": 0.2332, + "step": 35206 + }, + { + "epoch": 13.0, + "learning_rate": 1.7009360758714127e-06, + "loss": 0.3029, + "step": 35207 + }, + { + "epoch": 13.0, + "learning_rate": 1.7008991255080677e-06, + "loss": 0.3125, + "step": 35208 + }, + { + "epoch": 13.0, + "learning_rate": 1.7008621751447224e-06, + "loss": 0.3101, + "step": 35209 + }, + { + "epoch": 13.0, + "learning_rate": 1.700825224781377e-06, + "loss": 0.3218, + "step": 35210 + }, + { + "epoch": 13.0, + "learning_rate": 1.7007882744180318e-06, + "loss": 0.3334, + "step": 35211 + }, + { + "epoch": 13.0, + "learning_rate": 1.7007513240546865e-06, + "loss": 0.1408, + "step": 35212 + }, + { + "epoch": 13.0, + "learning_rate": 1.7007143736913413e-06, + "loss": 0.2494, + "step": 35213 + }, + { + "epoch": 13.0, + "learning_rate": 1.700677423327996e-06, + "loss": 0.3219, + "step": 35214 + }, + { + "epoch": 13.0, + "learning_rate": 1.700640472964651e-06, + "loss": 0.3405, + "step": 35215 + }, + { + "epoch": 13.0, + "learning_rate": 1.7006035226013057e-06, + "loss": 0.3549, + "step": 35216 + }, + { + "epoch": 13.0, + "learning_rate": 1.7005665722379605e-06, + "loss": 0.294, + "step": 35217 + }, + { + "epoch": 13.01, + "learning_rate": 1.7005296218746151e-06, + "loss": 0.1994, + "step": 35218 + }, + { + "epoch": 13.01, + "learning_rate": 1.7004926715112698e-06, + "loss": 0.29, + "step": 35219 + }, + { + "epoch": 13.01, + "learning_rate": 1.7004557211479246e-06, + "loss": 0.3229, + "step": 35220 + }, + { + "epoch": 13.01, + "learning_rate": 1.7004187707845793e-06, + "loss": 0.3326, + "step": 35221 + }, + { + "epoch": 13.01, + "learning_rate": 1.7003818204212343e-06, + "loss": 0.2811, + "step": 35222 + }, + { + "epoch": 13.01, + "learning_rate": 1.700344870057889e-06, + "loss": 0.2557, + "step": 35223 + }, + { + "epoch": 13.01, + "learning_rate": 1.7003079196945438e-06, + "loss": 0.2519, + "step": 35224 + }, + { + "epoch": 13.01, + "learning_rate": 1.7002709693311984e-06, + "loss": 0.4261, + "step": 35225 + }, + { + "epoch": 13.01, + "learning_rate": 1.7002340189678533e-06, + "loss": 0.3507, + "step": 35226 + }, + { + "epoch": 13.01, + "learning_rate": 1.700197068604508e-06, + "loss": 0.1628, + "step": 35227 + }, + { + "epoch": 13.01, + "learning_rate": 1.700160118241163e-06, + "loss": 0.2214, + "step": 35228 + }, + { + "epoch": 13.01, + "learning_rate": 1.7001231678778174e-06, + "loss": 0.3281, + "step": 35229 + }, + { + "epoch": 13.01, + "learning_rate": 1.700086217514472e-06, + "loss": 0.1762, + "step": 35230 + }, + { + "epoch": 13.01, + "learning_rate": 1.7000492671511271e-06, + "loss": 0.239, + "step": 35231 + }, + { + "epoch": 13.01, + "learning_rate": 1.7000123167877817e-06, + "loss": 0.2359, + "step": 35232 + }, + { + "epoch": 13.01, + "learning_rate": 1.6999753664244366e-06, + "loss": 0.1629, + "step": 35233 + }, + { + "epoch": 13.01, + "learning_rate": 1.6999384160610912e-06, + "loss": 0.3157, + "step": 35234 + }, + { + "epoch": 13.01, + "learning_rate": 1.699901465697746e-06, + "loss": 0.2149, + "step": 35235 + }, + { + "epoch": 13.01, + "learning_rate": 1.6998645153344007e-06, + "loss": 0.23, + "step": 35236 + }, + { + "epoch": 13.01, + "learning_rate": 1.6998275649710558e-06, + "loss": 0.2637, + "step": 35237 + }, + { + "epoch": 13.01, + "learning_rate": 1.6997906146077104e-06, + "loss": 0.308, + "step": 35238 + }, + { + "epoch": 13.01, + "learning_rate": 1.699753664244365e-06, + "loss": 0.2721, + "step": 35239 + }, + { + "epoch": 13.01, + "learning_rate": 1.69971671388102e-06, + "loss": 0.3445, + "step": 35240 + }, + { + "epoch": 13.01, + "learning_rate": 1.6996797635176745e-06, + "loss": 0.2206, + "step": 35241 + }, + { + "epoch": 13.01, + "learning_rate": 1.6996428131543294e-06, + "loss": 0.282, + "step": 35242 + }, + { + "epoch": 13.01, + "learning_rate": 1.699605862790984e-06, + "loss": 0.3838, + "step": 35243 + }, + { + "epoch": 13.01, + "learning_rate": 1.699568912427639e-06, + "loss": 0.2946, + "step": 35244 + }, + { + "epoch": 13.02, + "learning_rate": 1.6995319620642937e-06, + "loss": 0.3188, + "step": 35245 + }, + { + "epoch": 13.02, + "learning_rate": 1.6994950117009486e-06, + "loss": 0.2293, + "step": 35246 + }, + { + "epoch": 13.02, + "learning_rate": 1.6994580613376032e-06, + "loss": 0.3408, + "step": 35247 + }, + { + "epoch": 13.02, + "learning_rate": 1.699421110974258e-06, + "loss": 0.2119, + "step": 35248 + }, + { + "epoch": 13.02, + "learning_rate": 1.6993841606109127e-06, + "loss": 0.2432, + "step": 35249 + }, + { + "epoch": 13.02, + "learning_rate": 1.6993472102475673e-06, + "loss": 0.2696, + "step": 35250 + }, + { + "epoch": 13.02, + "learning_rate": 1.6993102598842224e-06, + "loss": 0.2978, + "step": 35251 + }, + { + "epoch": 13.02, + "learning_rate": 1.6992733095208768e-06, + "loss": 0.3917, + "step": 35252 + }, + { + "epoch": 13.02, + "learning_rate": 1.6992363591575319e-06, + "loss": 0.2133, + "step": 35253 + }, + { + "epoch": 13.02, + "learning_rate": 1.6991994087941865e-06, + "loss": 0.3231, + "step": 35254 + }, + { + "epoch": 13.02, + "learning_rate": 1.6991624584308413e-06, + "loss": 0.404, + "step": 35255 + }, + { + "epoch": 13.02, + "learning_rate": 1.699125508067496e-06, + "loss": 0.3352, + "step": 35256 + }, + { + "epoch": 13.02, + "learning_rate": 1.699088557704151e-06, + "loss": 0.226, + "step": 35257 + }, + { + "epoch": 13.02, + "learning_rate": 1.6990516073408055e-06, + "loss": 0.2464, + "step": 35258 + }, + { + "epoch": 13.02, + "learning_rate": 1.6990146569774601e-06, + "loss": 0.4108, + "step": 35259 + }, + { + "epoch": 13.02, + "learning_rate": 1.6989777066141152e-06, + "loss": 0.3055, + "step": 35260 + }, + { + "epoch": 13.02, + "learning_rate": 1.6989407562507698e-06, + "loss": 0.3264, + "step": 35261 + }, + { + "epoch": 13.02, + "learning_rate": 1.6989038058874247e-06, + "loss": 0.3299, + "step": 35262 + }, + { + "epoch": 13.02, + "learning_rate": 1.6988668555240793e-06, + "loss": 0.2011, + "step": 35263 + }, + { + "epoch": 13.02, + "learning_rate": 1.6988299051607341e-06, + "loss": 0.3296, + "step": 35264 + }, + { + "epoch": 13.02, + "learning_rate": 1.6987929547973888e-06, + "loss": 0.3572, + "step": 35265 + }, + { + "epoch": 13.02, + "learning_rate": 1.6987560044340438e-06, + "loss": 0.3502, + "step": 35266 + }, + { + "epoch": 13.02, + "learning_rate": 1.6987190540706985e-06, + "loss": 0.1691, + "step": 35267 + }, + { + "epoch": 13.02, + "learning_rate": 1.6986821037073533e-06, + "loss": 0.3636, + "step": 35268 + }, + { + "epoch": 13.02, + "learning_rate": 1.698645153344008e-06, + "loss": 0.2415, + "step": 35269 + }, + { + "epoch": 13.02, + "learning_rate": 1.6986082029806626e-06, + "loss": 0.2661, + "step": 35270 + }, + { + "epoch": 13.02, + "learning_rate": 1.6985712526173174e-06, + "loss": 0.3555, + "step": 35271 + }, + { + "epoch": 13.03, + "learning_rate": 1.698534302253972e-06, + "loss": 0.3214, + "step": 35272 + }, + { + "epoch": 13.03, + "learning_rate": 1.6984973518906271e-06, + "loss": 0.2541, + "step": 35273 + }, + { + "epoch": 13.03, + "learning_rate": 1.6984604015272816e-06, + "loss": 0.1944, + "step": 35274 + }, + { + "epoch": 13.03, + "learning_rate": 1.6984234511639366e-06, + "loss": 0.2642, + "step": 35275 + }, + { + "epoch": 13.03, + "learning_rate": 1.6983865008005913e-06, + "loss": 0.3497, + "step": 35276 + }, + { + "epoch": 13.03, + "learning_rate": 1.698349550437246e-06, + "loss": 0.3021, + "step": 35277 + }, + { + "epoch": 13.03, + "learning_rate": 1.6983126000739007e-06, + "loss": 0.3532, + "step": 35278 + }, + { + "epoch": 13.03, + "learning_rate": 1.6982756497105554e-06, + "loss": 0.3147, + "step": 35279 + }, + { + "epoch": 13.03, + "learning_rate": 1.6982386993472102e-06, + "loss": 0.2625, + "step": 35280 + }, + { + "epoch": 13.03, + "learning_rate": 1.6982017489838649e-06, + "loss": 0.2702, + "step": 35281 + }, + { + "epoch": 13.03, + "learning_rate": 1.69816479862052e-06, + "loss": 0.1938, + "step": 35282 + }, + { + "epoch": 13.03, + "learning_rate": 1.6981278482571746e-06, + "loss": 0.3775, + "step": 35283 + }, + { + "epoch": 13.03, + "learning_rate": 1.6980908978938294e-06, + "loss": 0.2962, + "step": 35284 + }, + { + "epoch": 13.03, + "learning_rate": 1.698053947530484e-06, + "loss": 0.3468, + "step": 35285 + }, + { + "epoch": 13.03, + "learning_rate": 1.6980169971671389e-06, + "loss": 0.2736, + "step": 35286 + }, + { + "epoch": 13.03, + "learning_rate": 1.6979800468037935e-06, + "loss": 0.4335, + "step": 35287 + }, + { + "epoch": 13.03, + "learning_rate": 1.6979430964404486e-06, + "loss": 0.3525, + "step": 35288 + }, + { + "epoch": 13.03, + "learning_rate": 1.6979061460771032e-06, + "loss": 0.3989, + "step": 35289 + }, + { + "epoch": 13.03, + "learning_rate": 1.6978691957137579e-06, + "loss": 0.2002, + "step": 35290 + }, + { + "epoch": 13.03, + "learning_rate": 1.6978322453504127e-06, + "loss": 0.3531, + "step": 35291 + }, + { + "epoch": 13.03, + "learning_rate": 1.6977952949870673e-06, + "loss": 0.2689, + "step": 35292 + }, + { + "epoch": 13.03, + "learning_rate": 1.6977583446237222e-06, + "loss": 0.2529, + "step": 35293 + }, + { + "epoch": 13.03, + "learning_rate": 1.6977213942603768e-06, + "loss": 0.276, + "step": 35294 + }, + { + "epoch": 13.03, + "learning_rate": 1.6976844438970319e-06, + "loss": 0.3732, + "step": 35295 + }, + { + "epoch": 13.03, + "learning_rate": 1.6976474935336865e-06, + "loss": 0.2633, + "step": 35296 + }, + { + "epoch": 13.03, + "learning_rate": 1.6976105431703414e-06, + "loss": 0.3789, + "step": 35297 + }, + { + "epoch": 13.03, + "learning_rate": 1.697573592806996e-06, + "loss": 0.3434, + "step": 35298 + }, + { + "epoch": 13.04, + "learning_rate": 1.6975366424436506e-06, + "loss": 0.3156, + "step": 35299 + }, + { + "epoch": 13.04, + "learning_rate": 1.6974996920803055e-06, + "loss": 0.4183, + "step": 35300 + }, + { + "epoch": 13.04, + "learning_rate": 1.6974627417169601e-06, + "loss": 0.4438, + "step": 35301 + }, + { + "epoch": 13.04, + "learning_rate": 1.6974257913536152e-06, + "loss": 0.2535, + "step": 35302 + }, + { + "epoch": 13.04, + "learning_rate": 1.6973888409902696e-06, + "loss": 0.2231, + "step": 35303 + }, + { + "epoch": 13.04, + "learning_rate": 1.6973518906269247e-06, + "loss": 0.2054, + "step": 35304 + }, + { + "epoch": 13.04, + "learning_rate": 1.6973149402635793e-06, + "loss": 0.2419, + "step": 35305 + }, + { + "epoch": 13.04, + "learning_rate": 1.6972779899002342e-06, + "loss": 0.3434, + "step": 35306 + }, + { + "epoch": 13.04, + "learning_rate": 1.6972410395368888e-06, + "loss": 0.2684, + "step": 35307 + }, + { + "epoch": 13.04, + "learning_rate": 1.6972040891735436e-06, + "loss": 0.3208, + "step": 35308 + }, + { + "epoch": 13.04, + "learning_rate": 1.6971671388101983e-06, + "loss": 0.3163, + "step": 35309 + }, + { + "epoch": 13.04, + "learning_rate": 1.697130188446853e-06, + "loss": 0.2684, + "step": 35310 + }, + { + "epoch": 13.04, + "learning_rate": 1.697093238083508e-06, + "loss": 0.2345, + "step": 35311 + }, + { + "epoch": 13.04, + "learning_rate": 1.6970562877201626e-06, + "loss": 0.2888, + "step": 35312 + }, + { + "epoch": 13.04, + "learning_rate": 1.6970193373568175e-06, + "loss": 0.3678, + "step": 35313 + }, + { + "epoch": 13.04, + "learning_rate": 1.696982386993472e-06, + "loss": 0.2611, + "step": 35314 + }, + { + "epoch": 13.04, + "learning_rate": 1.696945436630127e-06, + "loss": 0.2303, + "step": 35315 + }, + { + "epoch": 13.04, + "learning_rate": 1.6969084862667816e-06, + "loss": 0.2261, + "step": 35316 + }, + { + "epoch": 13.04, + "learning_rate": 1.6968715359034366e-06, + "loss": 0.2738, + "step": 35317 + }, + { + "epoch": 13.04, + "learning_rate": 1.6968345855400913e-06, + "loss": 0.2232, + "step": 35318 + }, + { + "epoch": 13.04, + "learning_rate": 1.6967976351767457e-06, + "loss": 0.3573, + "step": 35319 + }, + { + "epoch": 13.04, + "learning_rate": 1.6967606848134008e-06, + "loss": 0.2911, + "step": 35320 + }, + { + "epoch": 13.04, + "learning_rate": 1.6967237344500554e-06, + "loss": 0.2442, + "step": 35321 + }, + { + "epoch": 13.04, + "learning_rate": 1.6966867840867102e-06, + "loss": 0.3035, + "step": 35322 + }, + { + "epoch": 13.04, + "learning_rate": 1.6966498337233649e-06, + "loss": 0.2921, + "step": 35323 + }, + { + "epoch": 13.04, + "learning_rate": 1.69661288336002e-06, + "loss": 0.1877, + "step": 35324 + }, + { + "epoch": 13.04, + "learning_rate": 1.6965759329966744e-06, + "loss": 0.295, + "step": 35325 + }, + { + "epoch": 13.05, + "learning_rate": 1.6965389826333294e-06, + "loss": 0.2193, + "step": 35326 + }, + { + "epoch": 13.05, + "learning_rate": 1.696502032269984e-06, + "loss": 0.131, + "step": 35327 + }, + { + "epoch": 13.05, + "learning_rate": 1.6964650819066387e-06, + "loss": 0.3855, + "step": 35328 + }, + { + "epoch": 13.05, + "learning_rate": 1.6964281315432936e-06, + "loss": 0.2618, + "step": 35329 + }, + { + "epoch": 13.05, + "learning_rate": 1.6963911811799482e-06, + "loss": 0.3129, + "step": 35330 + }, + { + "epoch": 13.05, + "learning_rate": 1.696354230816603e-06, + "loss": 0.2817, + "step": 35331 + }, + { + "epoch": 13.05, + "learning_rate": 1.6963172804532577e-06, + "loss": 0.3264, + "step": 35332 + }, + { + "epoch": 13.05, + "learning_rate": 1.6962803300899127e-06, + "loss": 0.5442, + "step": 35333 + }, + { + "epoch": 13.05, + "learning_rate": 1.6962433797265674e-06, + "loss": 0.2639, + "step": 35334 + }, + { + "epoch": 13.05, + "learning_rate": 1.6962064293632222e-06, + "loss": 0.3427, + "step": 35335 + }, + { + "epoch": 13.05, + "learning_rate": 1.6961694789998769e-06, + "loss": 0.2529, + "step": 35336 + }, + { + "epoch": 13.05, + "learning_rate": 1.6961325286365317e-06, + "loss": 0.2086, + "step": 35337 + }, + { + "epoch": 13.05, + "learning_rate": 1.6960955782731863e-06, + "loss": 0.2132, + "step": 35338 + }, + { + "epoch": 13.05, + "learning_rate": 1.696058627909841e-06, + "loss": 0.3231, + "step": 35339 + }, + { + "epoch": 13.05, + "learning_rate": 1.696021677546496e-06, + "loss": 0.3038, + "step": 35340 + }, + { + "epoch": 13.05, + "learning_rate": 1.6959847271831505e-06, + "loss": 0.2499, + "step": 35341 + }, + { + "epoch": 13.05, + "learning_rate": 1.6959477768198055e-06, + "loss": 0.2708, + "step": 35342 + }, + { + "epoch": 13.05, + "learning_rate": 1.6959108264564602e-06, + "loss": 0.2377, + "step": 35343 + }, + { + "epoch": 13.05, + "learning_rate": 1.695873876093115e-06, + "loss": 0.2214, + "step": 35344 + }, + { + "epoch": 13.05, + "learning_rate": 1.6958369257297696e-06, + "loss": 0.3475, + "step": 35345 + }, + { + "epoch": 13.05, + "learning_rate": 1.6957999753664247e-06, + "loss": 0.1937, + "step": 35346 + }, + { + "epoch": 13.05, + "learning_rate": 1.6957630250030791e-06, + "loss": 0.2891, + "step": 35347 + }, + { + "epoch": 13.05, + "learning_rate": 1.6957260746397338e-06, + "loss": 0.2192, + "step": 35348 + }, + { + "epoch": 13.05, + "learning_rate": 1.6956891242763888e-06, + "loss": 0.2567, + "step": 35349 + }, + { + "epoch": 13.05, + "learning_rate": 1.6956521739130435e-06, + "loss": 0.2496, + "step": 35350 + }, + { + "epoch": 13.05, + "learning_rate": 1.6956152235496983e-06, + "loss": 0.2308, + "step": 35351 + }, + { + "epoch": 13.05, + "learning_rate": 1.695578273186353e-06, + "loss": 0.3048, + "step": 35352 + }, + { + "epoch": 13.06, + "learning_rate": 1.6955413228230078e-06, + "loss": 0.2213, + "step": 35353 + }, + { + "epoch": 13.06, + "learning_rate": 1.6955043724596624e-06, + "loss": 0.3121, + "step": 35354 + }, + { + "epoch": 13.06, + "learning_rate": 1.6954674220963175e-06, + "loss": 0.3622, + "step": 35355 + }, + { + "epoch": 13.06, + "learning_rate": 1.6954304717329721e-06, + "loss": 0.2469, + "step": 35356 + }, + { + "epoch": 13.06, + "learning_rate": 1.695393521369627e-06, + "loss": 0.3046, + "step": 35357 + }, + { + "epoch": 13.06, + "learning_rate": 1.6953565710062816e-06, + "loss": 0.2847, + "step": 35358 + }, + { + "epoch": 13.06, + "learning_rate": 1.6953196206429362e-06, + "loss": 0.2999, + "step": 35359 + }, + { + "epoch": 13.06, + "learning_rate": 1.695282670279591e-06, + "loss": 0.3203, + "step": 35360 + }, + { + "epoch": 13.06, + "learning_rate": 1.6952457199162457e-06, + "loss": 0.2206, + "step": 35361 + }, + { + "epoch": 13.06, + "learning_rate": 1.6952087695529008e-06, + "loss": 0.2043, + "step": 35362 + }, + { + "epoch": 13.06, + "learning_rate": 1.6951718191895554e-06, + "loss": 0.3232, + "step": 35363 + }, + { + "epoch": 13.06, + "learning_rate": 1.6951348688262103e-06, + "loss": 0.3819, + "step": 35364 + }, + { + "epoch": 13.06, + "learning_rate": 1.695097918462865e-06, + "loss": 0.3599, + "step": 35365 + }, + { + "epoch": 13.06, + "learning_rate": 1.6950609680995198e-06, + "loss": 0.2643, + "step": 35366 + }, + { + "epoch": 13.06, + "learning_rate": 1.6950240177361744e-06, + "loss": 0.3665, + "step": 35367 + }, + { + "epoch": 13.06, + "learning_rate": 1.694987067372829e-06, + "loss": 0.3566, + "step": 35368 + }, + { + "epoch": 13.06, + "learning_rate": 1.694950117009484e-06, + "loss": 0.3087, + "step": 35369 + }, + { + "epoch": 13.06, + "learning_rate": 1.6949131666461385e-06, + "loss": 0.3133, + "step": 35370 + }, + { + "epoch": 13.06, + "learning_rate": 1.6948762162827936e-06, + "loss": 0.2666, + "step": 35371 + }, + { + "epoch": 13.06, + "learning_rate": 1.6948392659194482e-06, + "loss": 0.2924, + "step": 35372 + }, + { + "epoch": 13.06, + "learning_rate": 1.694802315556103e-06, + "loss": 0.3826, + "step": 35373 + }, + { + "epoch": 13.06, + "learning_rate": 1.6947653651927577e-06, + "loss": 0.3778, + "step": 35374 + }, + { + "epoch": 13.06, + "learning_rate": 1.6947284148294128e-06, + "loss": 0.3382, + "step": 35375 + }, + { + "epoch": 13.06, + "learning_rate": 1.6946914644660672e-06, + "loss": 0.312, + "step": 35376 + }, + { + "epoch": 13.06, + "learning_rate": 1.6946545141027222e-06, + "loss": 0.2126, + "step": 35377 + }, + { + "epoch": 13.06, + "learning_rate": 1.6946175637393769e-06, + "loss": 0.3304, + "step": 35378 + }, + { + "epoch": 13.06, + "learning_rate": 1.6945806133760315e-06, + "loss": 0.3156, + "step": 35379 + }, + { + "epoch": 13.06, + "learning_rate": 1.6945436630126864e-06, + "loss": 0.2287, + "step": 35380 + }, + { + "epoch": 13.07, + "learning_rate": 1.694506712649341e-06, + "loss": 0.3464, + "step": 35381 + }, + { + "epoch": 13.07, + "learning_rate": 1.6944697622859958e-06, + "loss": 0.3354, + "step": 35382 + }, + { + "epoch": 13.07, + "learning_rate": 1.6944328119226505e-06, + "loss": 0.4894, + "step": 35383 + }, + { + "epoch": 13.07, + "learning_rate": 1.6943958615593055e-06, + "loss": 0.3508, + "step": 35384 + }, + { + "epoch": 13.07, + "learning_rate": 1.6943589111959602e-06, + "loss": 0.2845, + "step": 35385 + }, + { + "epoch": 13.07, + "learning_rate": 1.694321960832615e-06, + "loss": 0.3803, + "step": 35386 + }, + { + "epoch": 13.07, + "learning_rate": 1.6942850104692697e-06, + "loss": 0.1408, + "step": 35387 + }, + { + "epoch": 13.07, + "learning_rate": 1.6942480601059243e-06, + "loss": 0.2816, + "step": 35388 + }, + { + "epoch": 13.07, + "learning_rate": 1.6942111097425791e-06, + "loss": 0.3086, + "step": 35389 + }, + { + "epoch": 13.07, + "learning_rate": 1.6941741593792338e-06, + "loss": 0.4678, + "step": 35390 + }, + { + "epoch": 13.07, + "learning_rate": 1.6941372090158888e-06, + "loss": 0.2556, + "step": 35391 + }, + { + "epoch": 13.07, + "learning_rate": 1.6941002586525433e-06, + "loss": 0.285, + "step": 35392 + }, + { + "epoch": 13.07, + "learning_rate": 1.6940633082891983e-06, + "loss": 0.2106, + "step": 35393 + }, + { + "epoch": 13.07, + "learning_rate": 1.694026357925853e-06, + "loss": 0.2144, + "step": 35394 + }, + { + "epoch": 13.07, + "learning_rate": 1.6939894075625078e-06, + "loss": 0.3177, + "step": 35395 + }, + { + "epoch": 13.07, + "learning_rate": 1.6939524571991625e-06, + "loss": 0.1902, + "step": 35396 + }, + { + "epoch": 13.07, + "learning_rate": 1.6939155068358175e-06, + "loss": 0.369, + "step": 35397 + }, + { + "epoch": 13.07, + "learning_rate": 1.693878556472472e-06, + "loss": 0.1583, + "step": 35398 + }, + { + "epoch": 13.07, + "learning_rate": 1.6938416061091266e-06, + "loss": 0.2136, + "step": 35399 + }, + { + "epoch": 13.07, + "learning_rate": 1.6938046557457816e-06, + "loss": 0.3993, + "step": 35400 + }, + { + "epoch": 13.07, + "learning_rate": 1.6937677053824363e-06, + "loss": 0.3102, + "step": 35401 + }, + { + "epoch": 13.07, + "learning_rate": 1.6937307550190911e-06, + "loss": 0.1792, + "step": 35402 + }, + { + "epoch": 13.07, + "learning_rate": 1.6936938046557458e-06, + "loss": 0.3461, + "step": 35403 + }, + { + "epoch": 13.07, + "learning_rate": 1.6936568542924006e-06, + "loss": 0.275, + "step": 35404 + }, + { + "epoch": 13.07, + "learning_rate": 1.6936199039290552e-06, + "loss": 0.3552, + "step": 35405 + }, + { + "epoch": 13.07, + "learning_rate": 1.6935829535657103e-06, + "loss": 0.3179, + "step": 35406 + }, + { + "epoch": 13.07, + "learning_rate": 1.693546003202365e-06, + "loss": 0.2878, + "step": 35407 + }, + { + "epoch": 13.08, + "learning_rate": 1.6935090528390196e-06, + "loss": 0.2225, + "step": 35408 + }, + { + "epoch": 13.08, + "learning_rate": 1.6934721024756744e-06, + "loss": 0.188, + "step": 35409 + }, + { + "epoch": 13.08, + "learning_rate": 1.693435152112329e-06, + "loss": 0.3338, + "step": 35410 + }, + { + "epoch": 13.08, + "learning_rate": 1.693398201748984e-06, + "loss": 0.2922, + "step": 35411 + }, + { + "epoch": 13.08, + "learning_rate": 1.6933612513856385e-06, + "loss": 0.2173, + "step": 35412 + }, + { + "epoch": 13.08, + "learning_rate": 1.6933243010222936e-06, + "loss": 0.1836, + "step": 35413 + }, + { + "epoch": 13.08, + "learning_rate": 1.6932873506589482e-06, + "loss": 0.2309, + "step": 35414 + }, + { + "epoch": 13.08, + "learning_rate": 1.693250400295603e-06, + "loss": 0.2735, + "step": 35415 + }, + { + "epoch": 13.08, + "learning_rate": 1.6932134499322577e-06, + "loss": 0.1712, + "step": 35416 + }, + { + "epoch": 13.08, + "learning_rate": 1.6931764995689126e-06, + "loss": 0.1899, + "step": 35417 + }, + { + "epoch": 13.08, + "learning_rate": 1.6931395492055672e-06, + "loss": 0.21, + "step": 35418 + }, + { + "epoch": 13.08, + "learning_rate": 1.6931025988422218e-06, + "loss": 0.2432, + "step": 35419 + }, + { + "epoch": 13.08, + "learning_rate": 1.6930656484788767e-06, + "loss": 0.332, + "step": 35420 + }, + { + "epoch": 13.08, + "learning_rate": 1.6930286981155313e-06, + "loss": 0.2539, + "step": 35421 + }, + { + "epoch": 13.08, + "learning_rate": 1.6929917477521864e-06, + "loss": 0.21, + "step": 35422 + }, + { + "epoch": 13.08, + "learning_rate": 1.692954797388841e-06, + "loss": 0.2062, + "step": 35423 + }, + { + "epoch": 13.08, + "learning_rate": 1.6929178470254959e-06, + "loss": 0.2857, + "step": 35424 + }, + { + "epoch": 13.08, + "learning_rate": 1.6928808966621505e-06, + "loss": 0.3236, + "step": 35425 + }, + { + "epoch": 13.08, + "learning_rate": 1.6928439462988054e-06, + "loss": 0.3137, + "step": 35426 + }, + { + "epoch": 13.08, + "learning_rate": 1.69280699593546e-06, + "loss": 0.2975, + "step": 35427 + }, + { + "epoch": 13.08, + "learning_rate": 1.6927700455721146e-06, + "loss": 0.2106, + "step": 35428 + }, + { + "epoch": 13.08, + "learning_rate": 1.6927330952087697e-06, + "loss": 0.3183, + "step": 35429 + }, + { + "epoch": 13.08, + "learning_rate": 1.6926961448454243e-06, + "loss": 0.195, + "step": 35430 + }, + { + "epoch": 13.08, + "learning_rate": 1.6926591944820792e-06, + "loss": 0.2983, + "step": 35431 + }, + { + "epoch": 13.08, + "learning_rate": 1.6926222441187338e-06, + "loss": 0.22, + "step": 35432 + }, + { + "epoch": 13.08, + "learning_rate": 1.6925852937553887e-06, + "loss": 0.462, + "step": 35433 + }, + { + "epoch": 13.08, + "learning_rate": 1.6925483433920433e-06, + "loss": 0.3166, + "step": 35434 + }, + { + "epoch": 13.09, + "learning_rate": 1.6925113930286984e-06, + "loss": 0.3088, + "step": 35435 + }, + { + "epoch": 13.09, + "learning_rate": 1.692474442665353e-06, + "loss": 0.3256, + "step": 35436 + }, + { + "epoch": 13.09, + "learning_rate": 1.6924374923020078e-06, + "loss": 0.306, + "step": 35437 + }, + { + "epoch": 13.09, + "learning_rate": 1.6924005419386625e-06, + "loss": 0.2991, + "step": 35438 + }, + { + "epoch": 13.09, + "learning_rate": 1.6923635915753171e-06, + "loss": 0.3828, + "step": 35439 + }, + { + "epoch": 13.09, + "learning_rate": 1.692326641211972e-06, + "loss": 0.3713, + "step": 35440 + }, + { + "epoch": 13.09, + "learning_rate": 1.6922896908486266e-06, + "loss": 0.3168, + "step": 35441 + }, + { + "epoch": 13.09, + "learning_rate": 1.6922527404852817e-06, + "loss": 0.2445, + "step": 35442 + }, + { + "epoch": 13.09, + "learning_rate": 1.692215790121936e-06, + "loss": 0.197, + "step": 35443 + }, + { + "epoch": 13.09, + "learning_rate": 1.6921788397585911e-06, + "loss": 0.1336, + "step": 35444 + }, + { + "epoch": 13.09, + "learning_rate": 1.6921418893952458e-06, + "loss": 0.1947, + "step": 35445 + }, + { + "epoch": 13.09, + "learning_rate": 1.6921049390319006e-06, + "loss": 0.2487, + "step": 35446 + }, + { + "epoch": 13.09, + "learning_rate": 1.6920679886685553e-06, + "loss": 0.2186, + "step": 35447 + }, + { + "epoch": 13.09, + "learning_rate": 1.69203103830521e-06, + "loss": 0.2724, + "step": 35448 + }, + { + "epoch": 13.09, + "learning_rate": 1.6919940879418647e-06, + "loss": 0.4509, + "step": 35449 + }, + { + "epoch": 13.09, + "learning_rate": 1.6919571375785194e-06, + "loss": 0.2886, + "step": 35450 + }, + { + "epoch": 13.09, + "learning_rate": 1.6919201872151744e-06, + "loss": 0.2525, + "step": 35451 + }, + { + "epoch": 13.09, + "learning_rate": 1.691883236851829e-06, + "loss": 0.2516, + "step": 35452 + }, + { + "epoch": 13.09, + "learning_rate": 1.691846286488484e-06, + "loss": 0.1835, + "step": 35453 + }, + { + "epoch": 13.09, + "learning_rate": 1.6918093361251386e-06, + "loss": 0.2218, + "step": 35454 + }, + { + "epoch": 13.09, + "learning_rate": 1.6917723857617934e-06, + "loss": 0.2446, + "step": 35455 + }, + { + "epoch": 13.09, + "learning_rate": 1.691735435398448e-06, + "loss": 0.2897, + "step": 35456 + }, + { + "epoch": 13.09, + "learning_rate": 1.6916984850351031e-06, + "loss": 0.1838, + "step": 35457 + }, + { + "epoch": 13.09, + "learning_rate": 1.6916615346717577e-06, + "loss": 0.2426, + "step": 35458 + }, + { + "epoch": 13.09, + "learning_rate": 1.6916245843084122e-06, + "loss": 0.3023, + "step": 35459 + }, + { + "epoch": 13.09, + "learning_rate": 1.6915876339450672e-06, + "loss": 0.3939, + "step": 35460 + }, + { + "epoch": 13.09, + "learning_rate": 1.6915506835817219e-06, + "loss": 0.2857, + "step": 35461 + }, + { + "epoch": 13.1, + "learning_rate": 1.6915137332183767e-06, + "loss": 0.2089, + "step": 35462 + }, + { + "epoch": 13.1, + "learning_rate": 1.6914767828550313e-06, + "loss": 0.3243, + "step": 35463 + }, + { + "epoch": 13.1, + "learning_rate": 1.6914398324916864e-06, + "loss": 0.3345, + "step": 35464 + }, + { + "epoch": 13.1, + "learning_rate": 1.6914028821283408e-06, + "loss": 0.2892, + "step": 35465 + }, + { + "epoch": 13.1, + "learning_rate": 1.691365931764996e-06, + "loss": 0.2441, + "step": 35466 + }, + { + "epoch": 13.1, + "learning_rate": 1.6913289814016505e-06, + "loss": 0.2349, + "step": 35467 + }, + { + "epoch": 13.1, + "learning_rate": 1.6912920310383052e-06, + "loss": 0.3032, + "step": 35468 + }, + { + "epoch": 13.1, + "learning_rate": 1.69125508067496e-06, + "loss": 0.1814, + "step": 35469 + }, + { + "epoch": 13.1, + "learning_rate": 1.6912181303116147e-06, + "loss": 0.2936, + "step": 35470 + }, + { + "epoch": 13.1, + "learning_rate": 1.6911811799482695e-06, + "loss": 0.3978, + "step": 35471 + }, + { + "epoch": 13.1, + "learning_rate": 1.6911442295849241e-06, + "loss": 0.2591, + "step": 35472 + }, + { + "epoch": 13.1, + "learning_rate": 1.6911072792215792e-06, + "loss": 0.2373, + "step": 35473 + }, + { + "epoch": 13.1, + "learning_rate": 1.6910703288582338e-06, + "loss": 0.1835, + "step": 35474 + }, + { + "epoch": 13.1, + "learning_rate": 1.6910333784948887e-06, + "loss": 0.3133, + "step": 35475 + }, + { + "epoch": 13.1, + "learning_rate": 1.6909964281315433e-06, + "loss": 0.3322, + "step": 35476 + }, + { + "epoch": 13.1, + "learning_rate": 1.6909594777681982e-06, + "loss": 0.2732, + "step": 35477 + }, + { + "epoch": 13.1, + "learning_rate": 1.6909225274048528e-06, + "loss": 0.3022, + "step": 35478 + }, + { + "epoch": 13.1, + "learning_rate": 1.6908855770415074e-06, + "loss": 0.2409, + "step": 35479 + }, + { + "epoch": 13.1, + "learning_rate": 1.6908486266781625e-06, + "loss": 0.2875, + "step": 35480 + }, + { + "epoch": 13.1, + "learning_rate": 1.6908116763148171e-06, + "loss": 0.2493, + "step": 35481 + }, + { + "epoch": 13.1, + "learning_rate": 1.690774725951472e-06, + "loss": 0.2339, + "step": 35482 + }, + { + "epoch": 13.1, + "learning_rate": 1.6907377755881266e-06, + "loss": 0.2021, + "step": 35483 + }, + { + "epoch": 13.1, + "learning_rate": 1.6907008252247815e-06, + "loss": 0.3428, + "step": 35484 + }, + { + "epoch": 13.1, + "learning_rate": 1.690663874861436e-06, + "loss": 0.4117, + "step": 35485 + }, + { + "epoch": 13.1, + "learning_rate": 1.6906269244980912e-06, + "loss": 0.2804, + "step": 35486 + }, + { + "epoch": 13.1, + "learning_rate": 1.6905899741347458e-06, + "loss": 0.2028, + "step": 35487 + }, + { + "epoch": 13.1, + "learning_rate": 1.6905530237714002e-06, + "loss": 0.2041, + "step": 35488 + }, + { + "epoch": 13.11, + "learning_rate": 1.6905160734080553e-06, + "loss": 0.2198, + "step": 35489 + }, + { + "epoch": 13.11, + "learning_rate": 1.69047912304471e-06, + "loss": 0.2876, + "step": 35490 + }, + { + "epoch": 13.11, + "learning_rate": 1.6904421726813648e-06, + "loss": 0.2524, + "step": 35491 + }, + { + "epoch": 13.11, + "learning_rate": 1.6904052223180194e-06, + "loss": 0.2782, + "step": 35492 + }, + { + "epoch": 13.11, + "learning_rate": 1.6903682719546743e-06, + "loss": 0.3617, + "step": 35493 + }, + { + "epoch": 13.11, + "learning_rate": 1.6903313215913289e-06, + "loss": 0.2485, + "step": 35494 + }, + { + "epoch": 13.11, + "learning_rate": 1.690294371227984e-06, + "loss": 0.2536, + "step": 35495 + }, + { + "epoch": 13.11, + "learning_rate": 1.6902574208646386e-06, + "loss": 0.3141, + "step": 35496 + }, + { + "epoch": 13.11, + "learning_rate": 1.6902204705012934e-06, + "loss": 0.2628, + "step": 35497 + }, + { + "epoch": 13.11, + "learning_rate": 1.690183520137948e-06, + "loss": 0.1751, + "step": 35498 + }, + { + "epoch": 13.11, + "learning_rate": 1.6901465697746027e-06, + "loss": 0.2655, + "step": 35499 + }, + { + "epoch": 13.11, + "learning_rate": 1.6901096194112576e-06, + "loss": 0.4133, + "step": 35500 + }, + { + "epoch": 13.11, + "learning_rate": 1.6900726690479122e-06, + "loss": 0.2098, + "step": 35501 + }, + { + "epoch": 13.11, + "learning_rate": 1.6900357186845673e-06, + "loss": 0.2398, + "step": 35502 + }, + { + "epoch": 13.11, + "learning_rate": 1.6899987683212219e-06, + "loss": 0.2503, + "step": 35503 + }, + { + "epoch": 13.11, + "learning_rate": 1.6899618179578767e-06, + "loss": 0.3119, + "step": 35504 + }, + { + "epoch": 13.11, + "learning_rate": 1.6899248675945314e-06, + "loss": 0.3525, + "step": 35505 + }, + { + "epoch": 13.11, + "learning_rate": 1.6898879172311862e-06, + "loss": 0.4044, + "step": 35506 + }, + { + "epoch": 13.11, + "learning_rate": 1.6898509668678409e-06, + "loss": 0.2241, + "step": 35507 + }, + { + "epoch": 13.11, + "learning_rate": 1.6898140165044955e-06, + "loss": 0.221, + "step": 35508 + }, + { + "epoch": 13.11, + "learning_rate": 1.6897770661411506e-06, + "loss": 0.2543, + "step": 35509 + }, + { + "epoch": 13.11, + "learning_rate": 1.689740115777805e-06, + "loss": 0.2579, + "step": 35510 + }, + { + "epoch": 13.11, + "learning_rate": 1.68970316541446e-06, + "loss": 0.2245, + "step": 35511 + }, + { + "epoch": 13.11, + "learning_rate": 1.6896662150511147e-06, + "loss": 0.217, + "step": 35512 + }, + { + "epoch": 13.11, + "learning_rate": 1.6896292646877695e-06, + "loss": 0.4467, + "step": 35513 + }, + { + "epoch": 13.11, + "learning_rate": 1.6895923143244242e-06, + "loss": 0.2735, + "step": 35514 + }, + { + "epoch": 13.11, + "learning_rate": 1.6895553639610792e-06, + "loss": 0.2677, + "step": 35515 + }, + { + "epoch": 13.12, + "learning_rate": 1.6895184135977336e-06, + "loss": 0.2604, + "step": 35516 + }, + { + "epoch": 13.12, + "learning_rate": 1.6894814632343887e-06, + "loss": 0.1771, + "step": 35517 + }, + { + "epoch": 13.12, + "learning_rate": 1.6894445128710433e-06, + "loss": 0.3491, + "step": 35518 + }, + { + "epoch": 13.12, + "learning_rate": 1.689407562507698e-06, + "loss": 0.3789, + "step": 35519 + }, + { + "epoch": 13.12, + "learning_rate": 1.6893706121443528e-06, + "loss": 0.1893, + "step": 35520 + }, + { + "epoch": 13.12, + "learning_rate": 1.6893336617810075e-06, + "loss": 0.2576, + "step": 35521 + }, + { + "epoch": 13.12, + "learning_rate": 1.6892967114176623e-06, + "loss": 0.4037, + "step": 35522 + }, + { + "epoch": 13.12, + "learning_rate": 1.689259761054317e-06, + "loss": 0.3843, + "step": 35523 + }, + { + "epoch": 13.12, + "learning_rate": 1.689222810690972e-06, + "loss": 0.2717, + "step": 35524 + }, + { + "epoch": 13.12, + "learning_rate": 1.6891858603276266e-06, + "loss": 0.3098, + "step": 35525 + }, + { + "epoch": 13.12, + "learning_rate": 1.6891489099642815e-06, + "loss": 0.3477, + "step": 35526 + }, + { + "epoch": 13.12, + "learning_rate": 1.6891119596009361e-06, + "loss": 0.3443, + "step": 35527 + }, + { + "epoch": 13.12, + "learning_rate": 1.6890750092375908e-06, + "loss": 0.2988, + "step": 35528 + }, + { + "epoch": 13.12, + "learning_rate": 1.6890380588742456e-06, + "loss": 0.2593, + "step": 35529 + }, + { + "epoch": 13.12, + "learning_rate": 1.6890011085109002e-06, + "loss": 0.2216, + "step": 35530 + }, + { + "epoch": 13.12, + "learning_rate": 1.6889641581475553e-06, + "loss": 0.2871, + "step": 35531 + }, + { + "epoch": 13.12, + "learning_rate": 1.6889272077842097e-06, + "loss": 0.3469, + "step": 35532 + }, + { + "epoch": 13.12, + "learning_rate": 1.6888902574208648e-06, + "loss": 0.3746, + "step": 35533 + }, + { + "epoch": 13.12, + "learning_rate": 1.6888533070575194e-06, + "loss": 0.2175, + "step": 35534 + }, + { + "epoch": 13.12, + "learning_rate": 1.6888163566941743e-06, + "loss": 0.1851, + "step": 35535 + }, + { + "epoch": 13.12, + "learning_rate": 1.688779406330829e-06, + "loss": 0.2416, + "step": 35536 + }, + { + "epoch": 13.12, + "learning_rate": 1.688742455967484e-06, + "loss": 0.3239, + "step": 35537 + }, + { + "epoch": 13.12, + "learning_rate": 1.6887055056041384e-06, + "loss": 0.2589, + "step": 35538 + }, + { + "epoch": 13.12, + "learning_rate": 1.688668555240793e-06, + "loss": 0.2454, + "step": 35539 + }, + { + "epoch": 13.12, + "learning_rate": 1.688631604877448e-06, + "loss": 0.3547, + "step": 35540 + }, + { + "epoch": 13.12, + "learning_rate": 1.6885946545141027e-06, + "loss": 0.3184, + "step": 35541 + }, + { + "epoch": 13.12, + "learning_rate": 1.6885577041507576e-06, + "loss": 0.4126, + "step": 35542 + }, + { + "epoch": 13.13, + "learning_rate": 1.6885207537874122e-06, + "loss": 0.215, + "step": 35543 + }, + { + "epoch": 13.13, + "learning_rate": 1.688483803424067e-06, + "loss": 0.2296, + "step": 35544 + }, + { + "epoch": 13.13, + "learning_rate": 1.6884468530607217e-06, + "loss": 0.2796, + "step": 35545 + }, + { + "epoch": 13.13, + "learning_rate": 1.6884099026973768e-06, + "loss": 0.1848, + "step": 35546 + }, + { + "epoch": 13.13, + "learning_rate": 1.6883729523340314e-06, + "loss": 0.2555, + "step": 35547 + }, + { + "epoch": 13.13, + "learning_rate": 1.688336001970686e-06, + "loss": 0.3551, + "step": 35548 + }, + { + "epoch": 13.13, + "learning_rate": 1.6882990516073409e-06, + "loss": 0.1643, + "step": 35549 + }, + { + "epoch": 13.13, + "learning_rate": 1.6882621012439955e-06, + "loss": 0.2369, + "step": 35550 + }, + { + "epoch": 13.13, + "learning_rate": 1.6882251508806504e-06, + "loss": 0.3768, + "step": 35551 + }, + { + "epoch": 13.13, + "learning_rate": 1.688188200517305e-06, + "loss": 0.2718, + "step": 35552 + }, + { + "epoch": 13.13, + "learning_rate": 1.68815125015396e-06, + "loss": 0.2766, + "step": 35553 + }, + { + "epoch": 13.13, + "learning_rate": 1.6881142997906147e-06, + "loss": 0.318, + "step": 35554 + }, + { + "epoch": 13.13, + "learning_rate": 1.6880773494272695e-06, + "loss": 0.272, + "step": 35555 + }, + { + "epoch": 13.13, + "learning_rate": 1.6880403990639242e-06, + "loss": 0.1937, + "step": 35556 + }, + { + "epoch": 13.13, + "learning_rate": 1.6880034487005788e-06, + "loss": 0.271, + "step": 35557 + }, + { + "epoch": 13.13, + "learning_rate": 1.6879664983372337e-06, + "loss": 0.2732, + "step": 35558 + }, + { + "epoch": 13.13, + "learning_rate": 1.6879295479738883e-06, + "loss": 0.3287, + "step": 35559 + }, + { + "epoch": 13.13, + "learning_rate": 1.6878925976105434e-06, + "loss": 0.3147, + "step": 35560 + }, + { + "epoch": 13.13, + "learning_rate": 1.6878556472471978e-06, + "loss": 0.2719, + "step": 35561 + }, + { + "epoch": 13.13, + "learning_rate": 1.6878186968838529e-06, + "loss": 0.2627, + "step": 35562 + }, + { + "epoch": 13.13, + "learning_rate": 1.6877817465205075e-06, + "loss": 0.321, + "step": 35563 + }, + { + "epoch": 13.13, + "learning_rate": 1.6877447961571623e-06, + "loss": 0.3545, + "step": 35564 + }, + { + "epoch": 13.13, + "learning_rate": 1.687707845793817e-06, + "loss": 0.3169, + "step": 35565 + }, + { + "epoch": 13.13, + "learning_rate": 1.687670895430472e-06, + "loss": 0.1637, + "step": 35566 + }, + { + "epoch": 13.13, + "learning_rate": 1.6876339450671265e-06, + "loss": 0.2212, + "step": 35567 + }, + { + "epoch": 13.13, + "learning_rate": 1.687596994703781e-06, + "loss": 0.2362, + "step": 35568 + }, + { + "epoch": 13.13, + "learning_rate": 1.6875600443404362e-06, + "loss": 0.3206, + "step": 35569 + }, + { + "epoch": 13.14, + "learning_rate": 1.6875230939770908e-06, + "loss": 0.2607, + "step": 35570 + }, + { + "epoch": 13.14, + "learning_rate": 1.6874861436137456e-06, + "loss": 0.2248, + "step": 35571 + }, + { + "epoch": 13.14, + "learning_rate": 1.6874491932504003e-06, + "loss": 0.232, + "step": 35572 + }, + { + "epoch": 13.14, + "learning_rate": 1.6874122428870551e-06, + "loss": 0.3753, + "step": 35573 + }, + { + "epoch": 13.14, + "learning_rate": 1.6873752925237098e-06, + "loss": 0.1626, + "step": 35574 + }, + { + "epoch": 13.14, + "learning_rate": 1.6873383421603648e-06, + "loss": 0.379, + "step": 35575 + }, + { + "epoch": 13.14, + "learning_rate": 1.6873013917970195e-06, + "loss": 0.0811, + "step": 35576 + }, + { + "epoch": 13.14, + "learning_rate": 1.6872644414336739e-06, + "loss": 0.2181, + "step": 35577 + }, + { + "epoch": 13.14, + "learning_rate": 1.687227491070329e-06, + "loss": 0.3456, + "step": 35578 + }, + { + "epoch": 13.14, + "learning_rate": 1.6871905407069836e-06, + "loss": 0.2405, + "step": 35579 + }, + { + "epoch": 13.14, + "learning_rate": 1.6871535903436384e-06, + "loss": 0.241, + "step": 35580 + }, + { + "epoch": 13.14, + "learning_rate": 1.687116639980293e-06, + "loss": 0.2542, + "step": 35581 + }, + { + "epoch": 13.14, + "learning_rate": 1.6870796896169481e-06, + "loss": 0.2904, + "step": 35582 + }, + { + "epoch": 13.14, + "learning_rate": 1.6870427392536025e-06, + "loss": 0.2168, + "step": 35583 + }, + { + "epoch": 13.14, + "learning_rate": 1.6870057888902576e-06, + "loss": 0.4271, + "step": 35584 + }, + { + "epoch": 13.14, + "learning_rate": 1.6869688385269122e-06, + "loss": 0.2355, + "step": 35585 + }, + { + "epoch": 13.14, + "learning_rate": 1.686931888163567e-06, + "loss": 0.3442, + "step": 35586 + }, + { + "epoch": 13.14, + "learning_rate": 1.6868949378002217e-06, + "loss": 0.3314, + "step": 35587 + }, + { + "epoch": 13.14, + "learning_rate": 1.6868579874368764e-06, + "loss": 0.334, + "step": 35588 + }, + { + "epoch": 13.14, + "learning_rate": 1.6868210370735312e-06, + "loss": 0.3497, + "step": 35589 + }, + { + "epoch": 13.14, + "learning_rate": 1.6867840867101858e-06, + "loss": 0.3056, + "step": 35590 + }, + { + "epoch": 13.14, + "learning_rate": 1.686747136346841e-06, + "loss": 0.311, + "step": 35591 + }, + { + "epoch": 13.14, + "learning_rate": 1.6867101859834955e-06, + "loss": 0.1939, + "step": 35592 + }, + { + "epoch": 13.14, + "learning_rate": 1.6866732356201504e-06, + "loss": 0.2432, + "step": 35593 + }, + { + "epoch": 13.14, + "learning_rate": 1.686636285256805e-06, + "loss": 0.3051, + "step": 35594 + }, + { + "epoch": 13.14, + "learning_rate": 1.6865993348934599e-06, + "loss": 0.2352, + "step": 35595 + }, + { + "epoch": 13.14, + "learning_rate": 1.6865623845301145e-06, + "loss": 0.2821, + "step": 35596 + }, + { + "epoch": 13.15, + "learning_rate": 1.6865254341667691e-06, + "loss": 0.3582, + "step": 35597 + }, + { + "epoch": 13.15, + "learning_rate": 1.6864884838034242e-06, + "loss": 0.4148, + "step": 35598 + }, + { + "epoch": 13.15, + "learning_rate": 1.6864515334400788e-06, + "loss": 0.2642, + "step": 35599 + }, + { + "epoch": 13.15, + "learning_rate": 1.6864145830767337e-06, + "loss": 0.1966, + "step": 35600 + }, + { + "epoch": 13.15, + "learning_rate": 1.6863776327133883e-06, + "loss": 0.2191, + "step": 35601 + }, + { + "epoch": 13.15, + "learning_rate": 1.6863406823500432e-06, + "loss": 0.2682, + "step": 35602 + }, + { + "epoch": 13.15, + "learning_rate": 1.6863037319866978e-06, + "loss": 0.2405, + "step": 35603 + }, + { + "epoch": 13.15, + "learning_rate": 1.6862667816233529e-06, + "loss": 0.3885, + "step": 35604 + }, + { + "epoch": 13.15, + "learning_rate": 1.6862298312600073e-06, + "loss": 0.2975, + "step": 35605 + }, + { + "epoch": 13.15, + "learning_rate": 1.6861928808966624e-06, + "loss": 0.2735, + "step": 35606 + }, + { + "epoch": 13.15, + "learning_rate": 1.686155930533317e-06, + "loss": 0.215, + "step": 35607 + }, + { + "epoch": 13.15, + "learning_rate": 1.6861189801699716e-06, + "loss": 0.2885, + "step": 35608 + }, + { + "epoch": 13.15, + "learning_rate": 1.6860820298066265e-06, + "loss": 0.2383, + "step": 35609 + }, + { + "epoch": 13.15, + "learning_rate": 1.6860450794432811e-06, + "loss": 0.31, + "step": 35610 + }, + { + "epoch": 13.15, + "learning_rate": 1.686008129079936e-06, + "loss": 0.2097, + "step": 35611 + }, + { + "epoch": 13.15, + "learning_rate": 1.6859711787165906e-06, + "loss": 0.2628, + "step": 35612 + }, + { + "epoch": 13.15, + "learning_rate": 1.6859342283532457e-06, + "loss": 0.1889, + "step": 35613 + }, + { + "epoch": 13.15, + "learning_rate": 1.6858972779899003e-06, + "loss": 0.2035, + "step": 35614 + }, + { + "epoch": 13.15, + "learning_rate": 1.6858603276265551e-06, + "loss": 0.3551, + "step": 35615 + }, + { + "epoch": 13.15, + "learning_rate": 1.6858233772632098e-06, + "loss": 0.4204, + "step": 35616 + }, + { + "epoch": 13.15, + "learning_rate": 1.6857864268998644e-06, + "loss": 0.2549, + "step": 35617 + }, + { + "epoch": 13.15, + "learning_rate": 1.6857494765365193e-06, + "loss": 0.2541, + "step": 35618 + }, + { + "epoch": 13.15, + "learning_rate": 1.685712526173174e-06, + "loss": 0.2497, + "step": 35619 + }, + { + "epoch": 13.15, + "learning_rate": 1.685675575809829e-06, + "loss": 0.3697, + "step": 35620 + }, + { + "epoch": 13.15, + "learning_rate": 1.6856386254464836e-06, + "loss": 0.2556, + "step": 35621 + }, + { + "epoch": 13.15, + "learning_rate": 1.6856016750831384e-06, + "loss": 0.2328, + "step": 35622 + }, + { + "epoch": 13.15, + "learning_rate": 1.685564724719793e-06, + "loss": 0.3334, + "step": 35623 + }, + { + "epoch": 13.16, + "learning_rate": 1.685527774356448e-06, + "loss": 0.2572, + "step": 35624 + }, + { + "epoch": 13.16, + "learning_rate": 1.6854908239931026e-06, + "loss": 0.2669, + "step": 35625 + }, + { + "epoch": 13.16, + "learning_rate": 1.6854538736297576e-06, + "loss": 0.2141, + "step": 35626 + }, + { + "epoch": 13.16, + "learning_rate": 1.6854169232664123e-06, + "loss": 0.4443, + "step": 35627 + }, + { + "epoch": 13.16, + "learning_rate": 1.6853799729030667e-06, + "loss": 0.1665, + "step": 35628 + }, + { + "epoch": 13.16, + "learning_rate": 1.6853430225397218e-06, + "loss": 0.17, + "step": 35629 + }, + { + "epoch": 13.16, + "learning_rate": 1.6853060721763764e-06, + "loss": 0.3274, + "step": 35630 + }, + { + "epoch": 13.16, + "learning_rate": 1.6852691218130312e-06, + "loss": 0.4848, + "step": 35631 + }, + { + "epoch": 13.16, + "learning_rate": 1.6852321714496859e-06, + "loss": 0.4733, + "step": 35632 + }, + { + "epoch": 13.16, + "learning_rate": 1.685195221086341e-06, + "loss": 0.4003, + "step": 35633 + }, + { + "epoch": 13.16, + "learning_rate": 1.6851582707229954e-06, + "loss": 0.3321, + "step": 35634 + }, + { + "epoch": 13.16, + "learning_rate": 1.6851213203596504e-06, + "loss": 0.3361, + "step": 35635 + }, + { + "epoch": 13.16, + "learning_rate": 1.685084369996305e-06, + "loss": 0.1703, + "step": 35636 + }, + { + "epoch": 13.16, + "learning_rate": 1.6850474196329597e-06, + "loss": 0.2925, + "step": 35637 + }, + { + "epoch": 13.16, + "learning_rate": 1.6850104692696145e-06, + "loss": 0.3237, + "step": 35638 + }, + { + "epoch": 13.16, + "learning_rate": 1.6849735189062692e-06, + "loss": 0.3818, + "step": 35639 + }, + { + "epoch": 13.16, + "learning_rate": 1.684936568542924e-06, + "loss": 0.4172, + "step": 35640 + }, + { + "epoch": 13.16, + "learning_rate": 1.6848996181795787e-06, + "loss": 0.2624, + "step": 35641 + }, + { + "epoch": 13.16, + "learning_rate": 1.6848626678162337e-06, + "loss": 0.3918, + "step": 35642 + }, + { + "epoch": 13.16, + "learning_rate": 1.6848257174528884e-06, + "loss": 0.2166, + "step": 35643 + }, + { + "epoch": 13.16, + "learning_rate": 1.6847887670895432e-06, + "loss": 0.2885, + "step": 35644 + }, + { + "epoch": 13.16, + "learning_rate": 1.6847518167261978e-06, + "loss": 0.2761, + "step": 35645 + }, + { + "epoch": 13.16, + "learning_rate": 1.6847148663628527e-06, + "loss": 0.3125, + "step": 35646 + }, + { + "epoch": 13.16, + "learning_rate": 1.6846779159995073e-06, + "loss": 0.2333, + "step": 35647 + }, + { + "epoch": 13.16, + "learning_rate": 1.684640965636162e-06, + "loss": 0.3096, + "step": 35648 + }, + { + "epoch": 13.16, + "learning_rate": 1.684604015272817e-06, + "loss": 0.3429, + "step": 35649 + }, + { + "epoch": 13.16, + "learning_rate": 1.6845670649094714e-06, + "loss": 0.3024, + "step": 35650 + }, + { + "epoch": 13.17, + "learning_rate": 1.6845301145461265e-06, + "loss": 0.3593, + "step": 35651 + }, + { + "epoch": 13.17, + "learning_rate": 1.6844931641827811e-06, + "loss": 0.2543, + "step": 35652 + }, + { + "epoch": 13.17, + "learning_rate": 1.684456213819436e-06, + "loss": 0.2122, + "step": 35653 + }, + { + "epoch": 13.17, + "learning_rate": 1.6844192634560906e-06, + "loss": 0.3445, + "step": 35654 + }, + { + "epoch": 13.17, + "learning_rate": 1.6843823130927457e-06, + "loss": 0.1581, + "step": 35655 + }, + { + "epoch": 13.17, + "learning_rate": 1.6843453627294001e-06, + "loss": 0.2425, + "step": 35656 + }, + { + "epoch": 13.17, + "learning_rate": 1.6843084123660547e-06, + "loss": 0.3852, + "step": 35657 + }, + { + "epoch": 13.17, + "learning_rate": 1.6842714620027098e-06, + "loss": 0.3156, + "step": 35658 + }, + { + "epoch": 13.17, + "learning_rate": 1.6842345116393644e-06, + "loss": 0.3076, + "step": 35659 + }, + { + "epoch": 13.17, + "learning_rate": 1.6841975612760193e-06, + "loss": 0.24, + "step": 35660 + }, + { + "epoch": 13.17, + "learning_rate": 1.684160610912674e-06, + "loss": 0.2554, + "step": 35661 + }, + { + "epoch": 13.17, + "learning_rate": 1.6841236605493288e-06, + "loss": 0.3015, + "step": 35662 + }, + { + "epoch": 13.17, + "learning_rate": 1.6840867101859834e-06, + "loss": 0.3473, + "step": 35663 + }, + { + "epoch": 13.17, + "learning_rate": 1.6840497598226385e-06, + "loss": 0.2569, + "step": 35664 + }, + { + "epoch": 13.17, + "learning_rate": 1.6840128094592931e-06, + "loss": 0.2125, + "step": 35665 + }, + { + "epoch": 13.17, + "learning_rate": 1.683975859095948e-06, + "loss": 0.1834, + "step": 35666 + }, + { + "epoch": 13.17, + "learning_rate": 1.6839389087326026e-06, + "loss": 0.2305, + "step": 35667 + }, + { + "epoch": 13.17, + "learning_rate": 1.6839019583692572e-06, + "loss": 0.297, + "step": 35668 + }, + { + "epoch": 13.17, + "learning_rate": 1.683865008005912e-06, + "loss": 0.2298, + "step": 35669 + }, + { + "epoch": 13.17, + "learning_rate": 1.6838280576425667e-06, + "loss": 0.373, + "step": 35670 + }, + { + "epoch": 13.17, + "learning_rate": 1.6837911072792218e-06, + "loss": 0.3367, + "step": 35671 + }, + { + "epoch": 13.17, + "learning_rate": 1.6837541569158764e-06, + "loss": 0.1923, + "step": 35672 + }, + { + "epoch": 13.17, + "learning_rate": 1.6837172065525313e-06, + "loss": 0.4656, + "step": 35673 + }, + { + "epoch": 13.17, + "learning_rate": 1.683680256189186e-06, + "loss": 0.4252, + "step": 35674 + }, + { + "epoch": 13.17, + "learning_rate": 1.6836433058258407e-06, + "loss": 0.262, + "step": 35675 + }, + { + "epoch": 13.17, + "learning_rate": 1.6836063554624954e-06, + "loss": 0.3158, + "step": 35676 + }, + { + "epoch": 13.17, + "learning_rate": 1.68356940509915e-06, + "loss": 0.3368, + "step": 35677 + }, + { + "epoch": 13.18, + "learning_rate": 1.683532454735805e-06, + "loss": 0.2834, + "step": 35678 + }, + { + "epoch": 13.18, + "learning_rate": 1.6834955043724595e-06, + "loss": 0.3003, + "step": 35679 + }, + { + "epoch": 13.18, + "learning_rate": 1.6834585540091146e-06, + "loss": 0.3097, + "step": 35680 + }, + { + "epoch": 13.18, + "learning_rate": 1.6834216036457692e-06, + "loss": 0.2387, + "step": 35681 + }, + { + "epoch": 13.18, + "learning_rate": 1.683384653282424e-06, + "loss": 0.1734, + "step": 35682 + }, + { + "epoch": 13.18, + "learning_rate": 1.6833477029190787e-06, + "loss": 0.2226, + "step": 35683 + }, + { + "epoch": 13.18, + "learning_rate": 1.6833107525557335e-06, + "loss": 0.1875, + "step": 35684 + }, + { + "epoch": 13.18, + "learning_rate": 1.6832738021923882e-06, + "loss": 0.3158, + "step": 35685 + }, + { + "epoch": 13.18, + "learning_rate": 1.6832368518290432e-06, + "loss": 0.2304, + "step": 35686 + }, + { + "epoch": 13.18, + "learning_rate": 1.6831999014656979e-06, + "loss": 0.1353, + "step": 35687 + }, + { + "epoch": 13.18, + "learning_rate": 1.6831629511023525e-06, + "loss": 0.2609, + "step": 35688 + }, + { + "epoch": 13.18, + "learning_rate": 1.6831260007390073e-06, + "loss": 0.1953, + "step": 35689 + }, + { + "epoch": 13.18, + "learning_rate": 1.683089050375662e-06, + "loss": 0.281, + "step": 35690 + }, + { + "epoch": 13.18, + "learning_rate": 1.6830521000123168e-06, + "loss": 0.1852, + "step": 35691 + }, + { + "epoch": 13.18, + "learning_rate": 1.6830151496489715e-06, + "loss": 0.338, + "step": 35692 + }, + { + "epoch": 13.18, + "learning_rate": 1.6829781992856265e-06, + "loss": 0.3768, + "step": 35693 + }, + { + "epoch": 13.18, + "learning_rate": 1.6829412489222812e-06, + "loss": 0.3144, + "step": 35694 + }, + { + "epoch": 13.18, + "learning_rate": 1.682904298558936e-06, + "loss": 0.2653, + "step": 35695 + }, + { + "epoch": 13.18, + "learning_rate": 1.6828673481955906e-06, + "loss": 0.314, + "step": 35696 + }, + { + "epoch": 13.18, + "learning_rate": 1.6828303978322453e-06, + "loss": 0.2524, + "step": 35697 + }, + { + "epoch": 13.18, + "learning_rate": 1.6827934474689001e-06, + "loss": 0.2503, + "step": 35698 + }, + { + "epoch": 13.18, + "learning_rate": 1.6827564971055548e-06, + "loss": 0.4627, + "step": 35699 + }, + { + "epoch": 13.18, + "learning_rate": 1.6827195467422098e-06, + "loss": 0.2112, + "step": 35700 + }, + { + "epoch": 13.18, + "learning_rate": 1.6826825963788643e-06, + "loss": 0.307, + "step": 35701 + }, + { + "epoch": 13.18, + "learning_rate": 1.6826456460155193e-06, + "loss": 0.2458, + "step": 35702 + }, + { + "epoch": 13.18, + "learning_rate": 1.682608695652174e-06, + "loss": 0.2554, + "step": 35703 + }, + { + "epoch": 13.18, + "learning_rate": 1.6825717452888288e-06, + "loss": 0.2785, + "step": 35704 + }, + { + "epoch": 13.19, + "learning_rate": 1.6825347949254834e-06, + "loss": 0.2627, + "step": 35705 + }, + { + "epoch": 13.19, + "learning_rate": 1.6824978445621385e-06, + "loss": 0.3918, + "step": 35706 + }, + { + "epoch": 13.19, + "learning_rate": 1.682460894198793e-06, + "loss": 0.4134, + "step": 35707 + }, + { + "epoch": 13.19, + "learning_rate": 1.6824239438354476e-06, + "loss": 0.4186, + "step": 35708 + }, + { + "epoch": 13.19, + "learning_rate": 1.6823869934721026e-06, + "loss": 0.1468, + "step": 35709 + }, + { + "epoch": 13.19, + "learning_rate": 1.6823500431087573e-06, + "loss": 0.2134, + "step": 35710 + }, + { + "epoch": 13.19, + "learning_rate": 1.682313092745412e-06, + "loss": 0.2235, + "step": 35711 + }, + { + "epoch": 13.19, + "learning_rate": 1.6822761423820667e-06, + "loss": 0.335, + "step": 35712 + }, + { + "epoch": 13.19, + "learning_rate": 1.6822391920187216e-06, + "loss": 0.2195, + "step": 35713 + }, + { + "epoch": 13.19, + "learning_rate": 1.6822022416553762e-06, + "loss": 0.3269, + "step": 35714 + }, + { + "epoch": 13.19, + "learning_rate": 1.6821652912920313e-06, + "loss": 0.1075, + "step": 35715 + }, + { + "epoch": 13.19, + "learning_rate": 1.682128340928686e-06, + "loss": 0.3448, + "step": 35716 + }, + { + "epoch": 13.19, + "learning_rate": 1.6820913905653403e-06, + "loss": 0.3331, + "step": 35717 + }, + { + "epoch": 13.19, + "learning_rate": 1.6820544402019954e-06, + "loss": 0.2505, + "step": 35718 + }, + { + "epoch": 13.19, + "learning_rate": 1.68201748983865e-06, + "loss": 0.168, + "step": 35719 + }, + { + "epoch": 13.19, + "learning_rate": 1.6819805394753049e-06, + "loss": 0.3049, + "step": 35720 + }, + { + "epoch": 13.19, + "learning_rate": 1.6819435891119595e-06, + "loss": 0.2027, + "step": 35721 + }, + { + "epoch": 13.19, + "learning_rate": 1.6819066387486146e-06, + "loss": 0.4359, + "step": 35722 + }, + { + "epoch": 13.19, + "learning_rate": 1.681869688385269e-06, + "loss": 0.3488, + "step": 35723 + }, + { + "epoch": 13.19, + "learning_rate": 1.681832738021924e-06, + "loss": 0.203, + "step": 35724 + }, + { + "epoch": 13.19, + "learning_rate": 1.6817957876585787e-06, + "loss": 0.3856, + "step": 35725 + }, + { + "epoch": 13.19, + "learning_rate": 1.6817588372952336e-06, + "loss": 0.2934, + "step": 35726 + }, + { + "epoch": 13.19, + "learning_rate": 1.6817218869318882e-06, + "loss": 0.3274, + "step": 35727 + }, + { + "epoch": 13.19, + "learning_rate": 1.6816849365685428e-06, + "loss": 0.1893, + "step": 35728 + }, + { + "epoch": 13.19, + "learning_rate": 1.6816479862051977e-06, + "loss": 0.2776, + "step": 35729 + }, + { + "epoch": 13.19, + "learning_rate": 1.6816110358418523e-06, + "loss": 0.2715, + "step": 35730 + }, + { + "epoch": 13.19, + "learning_rate": 1.6815740854785074e-06, + "loss": 0.3997, + "step": 35731 + }, + { + "epoch": 13.19, + "learning_rate": 1.681537135115162e-06, + "loss": 0.3355, + "step": 35732 + }, + { + "epoch": 13.2, + "learning_rate": 1.6815001847518169e-06, + "loss": 0.2657, + "step": 35733 + }, + { + "epoch": 13.2, + "learning_rate": 1.6814632343884715e-06, + "loss": 0.2792, + "step": 35734 + }, + { + "epoch": 13.2, + "learning_rate": 1.6814262840251263e-06, + "loss": 0.1749, + "step": 35735 + }, + { + "epoch": 13.2, + "learning_rate": 1.681389333661781e-06, + "loss": 0.2525, + "step": 35736 + }, + { + "epoch": 13.2, + "learning_rate": 1.6813523832984356e-06, + "loss": 0.2461, + "step": 35737 + }, + { + "epoch": 13.2, + "learning_rate": 1.6813154329350907e-06, + "loss": 0.3193, + "step": 35738 + }, + { + "epoch": 13.2, + "learning_rate": 1.6812784825717453e-06, + "loss": 0.2667, + "step": 35739 + }, + { + "epoch": 13.2, + "learning_rate": 1.6812415322084002e-06, + "loss": 0.2861, + "step": 35740 + }, + { + "epoch": 13.2, + "learning_rate": 1.6812045818450548e-06, + "loss": 0.3394, + "step": 35741 + }, + { + "epoch": 13.2, + "learning_rate": 1.6811676314817096e-06, + "loss": 0.2544, + "step": 35742 + }, + { + "epoch": 13.2, + "learning_rate": 1.6811306811183643e-06, + "loss": 0.2779, + "step": 35743 + }, + { + "epoch": 13.2, + "learning_rate": 1.6810937307550193e-06, + "loss": 0.375, + "step": 35744 + }, + { + "epoch": 13.2, + "learning_rate": 1.681056780391674e-06, + "loss": 0.2096, + "step": 35745 + }, + { + "epoch": 13.2, + "learning_rate": 1.6810198300283288e-06, + "loss": 0.2648, + "step": 35746 + }, + { + "epoch": 13.2, + "learning_rate": 1.6809828796649835e-06, + "loss": 0.425, + "step": 35747 + }, + { + "epoch": 13.2, + "learning_rate": 1.680945929301638e-06, + "loss": 0.1897, + "step": 35748 + }, + { + "epoch": 13.2, + "learning_rate": 1.680908978938293e-06, + "loss": 0.4129, + "step": 35749 + }, + { + "epoch": 13.2, + "learning_rate": 1.6808720285749476e-06, + "loss": 0.2656, + "step": 35750 + }, + { + "epoch": 13.2, + "learning_rate": 1.6808350782116026e-06, + "loss": 0.1913, + "step": 35751 + }, + { + "epoch": 13.2, + "learning_rate": 1.680798127848257e-06, + "loss": 0.2881, + "step": 35752 + }, + { + "epoch": 13.2, + "learning_rate": 1.6807611774849121e-06, + "loss": 0.2444, + "step": 35753 + }, + { + "epoch": 13.2, + "learning_rate": 1.6807242271215668e-06, + "loss": 0.203, + "step": 35754 + }, + { + "epoch": 13.2, + "learning_rate": 1.6806872767582216e-06, + "loss": 0.4475, + "step": 35755 + }, + { + "epoch": 13.2, + "learning_rate": 1.6806503263948762e-06, + "loss": 0.358, + "step": 35756 + }, + { + "epoch": 13.2, + "learning_rate": 1.6806133760315309e-06, + "loss": 0.2778, + "step": 35757 + }, + { + "epoch": 13.2, + "learning_rate": 1.6805764256681857e-06, + "loss": 0.2338, + "step": 35758 + }, + { + "epoch": 13.2, + "learning_rate": 1.6805394753048404e-06, + "loss": 0.2545, + "step": 35759 + }, + { + "epoch": 13.21, + "learning_rate": 1.6805025249414954e-06, + "loss": 0.2769, + "step": 35760 + }, + { + "epoch": 13.21, + "learning_rate": 1.68046557457815e-06, + "loss": 0.2751, + "step": 35761 + }, + { + "epoch": 13.21, + "learning_rate": 1.680428624214805e-06, + "loss": 0.2982, + "step": 35762 + }, + { + "epoch": 13.21, + "learning_rate": 1.6803916738514595e-06, + "loss": 0.2573, + "step": 35763 + }, + { + "epoch": 13.21, + "learning_rate": 1.6803547234881144e-06, + "loss": 0.2686, + "step": 35764 + }, + { + "epoch": 13.21, + "learning_rate": 1.680317773124769e-06, + "loss": 0.2578, + "step": 35765 + }, + { + "epoch": 13.21, + "learning_rate": 1.680280822761424e-06, + "loss": 0.2425, + "step": 35766 + }, + { + "epoch": 13.21, + "learning_rate": 1.6802438723980787e-06, + "loss": 0.3564, + "step": 35767 + }, + { + "epoch": 13.21, + "learning_rate": 1.6802069220347332e-06, + "loss": 0.2966, + "step": 35768 + }, + { + "epoch": 13.21, + "learning_rate": 1.6801699716713882e-06, + "loss": 0.2454, + "step": 35769 + }, + { + "epoch": 13.21, + "learning_rate": 1.6801330213080429e-06, + "loss": 0.2293, + "step": 35770 + }, + { + "epoch": 13.21, + "learning_rate": 1.6800960709446977e-06, + "loss": 0.4584, + "step": 35771 + }, + { + "epoch": 13.21, + "learning_rate": 1.6800591205813523e-06, + "loss": 0.3314, + "step": 35772 + }, + { + "epoch": 13.21, + "learning_rate": 1.6800221702180074e-06, + "loss": 0.2677, + "step": 35773 + }, + { + "epoch": 13.21, + "learning_rate": 1.6799852198546618e-06, + "loss": 0.22, + "step": 35774 + }, + { + "epoch": 13.21, + "learning_rate": 1.6799482694913169e-06, + "loss": 0.2545, + "step": 35775 + }, + { + "epoch": 13.21, + "learning_rate": 1.6799113191279715e-06, + "loss": 0.3179, + "step": 35776 + }, + { + "epoch": 13.21, + "learning_rate": 1.6798743687646262e-06, + "loss": 0.2214, + "step": 35777 + }, + { + "epoch": 13.21, + "learning_rate": 1.679837418401281e-06, + "loss": 0.2478, + "step": 35778 + }, + { + "epoch": 13.21, + "learning_rate": 1.6798004680379356e-06, + "loss": 0.3448, + "step": 35779 + }, + { + "epoch": 13.21, + "learning_rate": 1.6797635176745905e-06, + "loss": 0.2572, + "step": 35780 + }, + { + "epoch": 13.21, + "learning_rate": 1.6797265673112451e-06, + "loss": 0.3875, + "step": 35781 + }, + { + "epoch": 13.21, + "learning_rate": 1.6796896169479002e-06, + "loss": 0.2295, + "step": 35782 + }, + { + "epoch": 13.21, + "learning_rate": 1.6796526665845548e-06, + "loss": 0.347, + "step": 35783 + }, + { + "epoch": 13.21, + "learning_rate": 1.6796157162212097e-06, + "loss": 0.4454, + "step": 35784 + }, + { + "epoch": 13.21, + "learning_rate": 1.6795787658578643e-06, + "loss": 0.2723, + "step": 35785 + }, + { + "epoch": 13.21, + "learning_rate": 1.679541815494519e-06, + "loss": 0.2447, + "step": 35786 + }, + { + "epoch": 13.22, + "learning_rate": 1.6795048651311738e-06, + "loss": 0.2582, + "step": 35787 + }, + { + "epoch": 13.22, + "learning_rate": 1.6794679147678284e-06, + "loss": 0.384, + "step": 35788 + }, + { + "epoch": 13.22, + "learning_rate": 1.6794309644044835e-06, + "loss": 0.273, + "step": 35789 + }, + { + "epoch": 13.22, + "learning_rate": 1.679394014041138e-06, + "loss": 0.2336, + "step": 35790 + }, + { + "epoch": 13.22, + "learning_rate": 1.679357063677793e-06, + "loss": 0.3217, + "step": 35791 + }, + { + "epoch": 13.22, + "learning_rate": 1.6793201133144476e-06, + "loss": 0.3355, + "step": 35792 + }, + { + "epoch": 13.22, + "learning_rate": 1.6792831629511025e-06, + "loss": 0.1574, + "step": 35793 + }, + { + "epoch": 13.22, + "learning_rate": 1.679246212587757e-06, + "loss": 0.2925, + "step": 35794 + }, + { + "epoch": 13.22, + "learning_rate": 1.6792092622244122e-06, + "loss": 0.1866, + "step": 35795 + }, + { + "epoch": 13.22, + "learning_rate": 1.6791723118610666e-06, + "loss": 0.2876, + "step": 35796 + }, + { + "epoch": 13.22, + "learning_rate": 1.6791353614977212e-06, + "loss": 0.3463, + "step": 35797 + }, + { + "epoch": 13.22, + "learning_rate": 1.6790984111343763e-06, + "loss": 0.2285, + "step": 35798 + }, + { + "epoch": 13.22, + "learning_rate": 1.679061460771031e-06, + "loss": 0.404, + "step": 35799 + }, + { + "epoch": 13.22, + "learning_rate": 1.6790245104076858e-06, + "loss": 0.3068, + "step": 35800 + }, + { + "epoch": 13.22, + "learning_rate": 1.6789875600443404e-06, + "loss": 0.2875, + "step": 35801 + }, + { + "epoch": 13.22, + "learning_rate": 1.6789506096809952e-06, + "loss": 0.2646, + "step": 35802 + }, + { + "epoch": 13.22, + "learning_rate": 1.6789136593176499e-06, + "loss": 0.3231, + "step": 35803 + }, + { + "epoch": 13.22, + "learning_rate": 1.678876708954305e-06, + "loss": 0.3568, + "step": 35804 + }, + { + "epoch": 13.22, + "learning_rate": 1.6788397585909596e-06, + "loss": 0.2379, + "step": 35805 + }, + { + "epoch": 13.22, + "learning_rate": 1.6788028082276142e-06, + "loss": 0.2274, + "step": 35806 + }, + { + "epoch": 13.22, + "learning_rate": 1.678765857864269e-06, + "loss": 0.3882, + "step": 35807 + }, + { + "epoch": 13.22, + "learning_rate": 1.6787289075009237e-06, + "loss": 0.2848, + "step": 35808 + }, + { + "epoch": 13.22, + "learning_rate": 1.6786919571375785e-06, + "loss": 0.4819, + "step": 35809 + }, + { + "epoch": 13.22, + "learning_rate": 1.6786550067742332e-06, + "loss": 0.3719, + "step": 35810 + }, + { + "epoch": 13.22, + "learning_rate": 1.6786180564108882e-06, + "loss": 0.2345, + "step": 35811 + }, + { + "epoch": 13.22, + "learning_rate": 1.6785811060475429e-06, + "loss": 0.2484, + "step": 35812 + }, + { + "epoch": 13.22, + "learning_rate": 1.6785441556841977e-06, + "loss": 0.3879, + "step": 35813 + }, + { + "epoch": 13.23, + "learning_rate": 1.6785072053208524e-06, + "loss": 0.2299, + "step": 35814 + }, + { + "epoch": 13.23, + "learning_rate": 1.6784702549575072e-06, + "loss": 0.1729, + "step": 35815 + }, + { + "epoch": 13.23, + "learning_rate": 1.6784333045941618e-06, + "loss": 0.2522, + "step": 35816 + }, + { + "epoch": 13.23, + "learning_rate": 1.6783963542308165e-06, + "loss": 0.2789, + "step": 35817 + }, + { + "epoch": 13.23, + "learning_rate": 1.6783594038674715e-06, + "loss": 0.2626, + "step": 35818 + }, + { + "epoch": 13.23, + "learning_rate": 1.678322453504126e-06, + "loss": 0.1633, + "step": 35819 + }, + { + "epoch": 13.23, + "learning_rate": 1.678285503140781e-06, + "loss": 0.1852, + "step": 35820 + }, + { + "epoch": 13.23, + "learning_rate": 1.6782485527774357e-06, + "loss": 0.2862, + "step": 35821 + }, + { + "epoch": 13.23, + "learning_rate": 1.6782116024140905e-06, + "loss": 0.3312, + "step": 35822 + }, + { + "epoch": 13.23, + "learning_rate": 1.6781746520507451e-06, + "loss": 0.1094, + "step": 35823 + }, + { + "epoch": 13.23, + "learning_rate": 1.6781377016874002e-06, + "loss": 0.2731, + "step": 35824 + }, + { + "epoch": 13.23, + "learning_rate": 1.6781007513240546e-06, + "loss": 0.3314, + "step": 35825 + }, + { + "epoch": 13.23, + "learning_rate": 1.6780638009607093e-06, + "loss": 0.2044, + "step": 35826 + }, + { + "epoch": 13.23, + "learning_rate": 1.6780268505973643e-06, + "loss": 0.2925, + "step": 35827 + }, + { + "epoch": 13.23, + "learning_rate": 1.677989900234019e-06, + "loss": 0.295, + "step": 35828 + }, + { + "epoch": 13.23, + "learning_rate": 1.6779529498706738e-06, + "loss": 0.2853, + "step": 35829 + }, + { + "epoch": 13.23, + "learning_rate": 1.6779159995073284e-06, + "loss": 0.4821, + "step": 35830 + }, + { + "epoch": 13.23, + "learning_rate": 1.6778790491439833e-06, + "loss": 0.3505, + "step": 35831 + }, + { + "epoch": 13.23, + "learning_rate": 1.677842098780638e-06, + "loss": 0.1888, + "step": 35832 + }, + { + "epoch": 13.23, + "learning_rate": 1.677805148417293e-06, + "loss": 0.4974, + "step": 35833 + }, + { + "epoch": 13.23, + "learning_rate": 1.6777681980539476e-06, + "loss": 0.2758, + "step": 35834 + }, + { + "epoch": 13.23, + "learning_rate": 1.6777312476906025e-06, + "loss": 0.3488, + "step": 35835 + }, + { + "epoch": 13.23, + "learning_rate": 1.6776942973272571e-06, + "loss": 0.2522, + "step": 35836 + }, + { + "epoch": 13.23, + "learning_rate": 1.6776573469639118e-06, + "loss": 0.1353, + "step": 35837 + }, + { + "epoch": 13.23, + "learning_rate": 1.6776203966005666e-06, + "loss": 0.279, + "step": 35838 + }, + { + "epoch": 13.23, + "learning_rate": 1.6775834462372212e-06, + "loss": 0.2637, + "step": 35839 + }, + { + "epoch": 13.23, + "learning_rate": 1.6775464958738763e-06, + "loss": 0.3501, + "step": 35840 + }, + { + "epoch": 13.24, + "learning_rate": 1.6775095455105307e-06, + "loss": 0.3641, + "step": 35841 + }, + { + "epoch": 13.24, + "learning_rate": 1.6774725951471858e-06, + "loss": 0.2345, + "step": 35842 + }, + { + "epoch": 13.24, + "learning_rate": 1.6774356447838404e-06, + "loss": 0.2083, + "step": 35843 + }, + { + "epoch": 13.24, + "learning_rate": 1.6773986944204953e-06, + "loss": 0.2808, + "step": 35844 + }, + { + "epoch": 13.24, + "learning_rate": 1.67736174405715e-06, + "loss": 0.1754, + "step": 35845 + }, + { + "epoch": 13.24, + "learning_rate": 1.6773247936938045e-06, + "loss": 0.2226, + "step": 35846 + }, + { + "epoch": 13.24, + "learning_rate": 1.6772878433304594e-06, + "loss": 0.4052, + "step": 35847 + }, + { + "epoch": 13.24, + "learning_rate": 1.677250892967114e-06, + "loss": 0.2427, + "step": 35848 + }, + { + "epoch": 13.24, + "learning_rate": 1.677213942603769e-06, + "loss": 0.2217, + "step": 35849 + }, + { + "epoch": 13.24, + "learning_rate": 1.6771769922404237e-06, + "loss": 0.2565, + "step": 35850 + }, + { + "epoch": 13.24, + "learning_rate": 1.6771400418770786e-06, + "loss": 0.1862, + "step": 35851 + }, + { + "epoch": 13.24, + "learning_rate": 1.6771030915137332e-06, + "loss": 0.1561, + "step": 35852 + }, + { + "epoch": 13.24, + "learning_rate": 1.677066141150388e-06, + "loss": 0.3114, + "step": 35853 + }, + { + "epoch": 13.24, + "learning_rate": 1.6770291907870427e-06, + "loss": 0.1803, + "step": 35854 + }, + { + "epoch": 13.24, + "learning_rate": 1.6769922404236977e-06, + "loss": 0.2945, + "step": 35855 + }, + { + "epoch": 13.24, + "learning_rate": 1.6769552900603524e-06, + "loss": 0.2439, + "step": 35856 + }, + { + "epoch": 13.24, + "learning_rate": 1.676918339697007e-06, + "loss": 0.2601, + "step": 35857 + }, + { + "epoch": 13.24, + "learning_rate": 1.6768813893336619e-06, + "loss": 0.3332, + "step": 35858 + }, + { + "epoch": 13.24, + "learning_rate": 1.6768444389703165e-06, + "loss": 0.2128, + "step": 35859 + }, + { + "epoch": 13.24, + "learning_rate": 1.6768074886069714e-06, + "loss": 0.2503, + "step": 35860 + }, + { + "epoch": 13.24, + "learning_rate": 1.676770538243626e-06, + "loss": 0.2441, + "step": 35861 + }, + { + "epoch": 13.24, + "learning_rate": 1.676733587880281e-06, + "loss": 0.2367, + "step": 35862 + }, + { + "epoch": 13.24, + "learning_rate": 1.6766966375169357e-06, + "loss": 0.3711, + "step": 35863 + }, + { + "epoch": 13.24, + "learning_rate": 1.6766596871535905e-06, + "loss": 0.367, + "step": 35864 + }, + { + "epoch": 13.24, + "learning_rate": 1.6766227367902452e-06, + "loss": 0.2374, + "step": 35865 + }, + { + "epoch": 13.24, + "learning_rate": 1.6765857864268998e-06, + "loss": 0.233, + "step": 35866 + }, + { + "epoch": 13.24, + "learning_rate": 1.6765488360635547e-06, + "loss": 0.1796, + "step": 35867 + }, + { + "epoch": 13.25, + "learning_rate": 1.6765118857002093e-06, + "loss": 0.2524, + "step": 35868 + }, + { + "epoch": 13.25, + "learning_rate": 1.6764749353368641e-06, + "loss": 0.3034, + "step": 35869 + }, + { + "epoch": 13.25, + "learning_rate": 1.6764379849735188e-06, + "loss": 0.3123, + "step": 35870 + }, + { + "epoch": 13.25, + "learning_rate": 1.6764010346101738e-06, + "loss": 0.3409, + "step": 35871 + }, + { + "epoch": 13.25, + "learning_rate": 1.6763640842468285e-06, + "loss": 0.1736, + "step": 35872 + }, + { + "epoch": 13.25, + "learning_rate": 1.6763271338834833e-06, + "loss": 0.4052, + "step": 35873 + }, + { + "epoch": 13.25, + "learning_rate": 1.676290183520138e-06, + "loss": 0.2711, + "step": 35874 + }, + { + "epoch": 13.25, + "learning_rate": 1.6762532331567928e-06, + "loss": 0.3495, + "step": 35875 + }, + { + "epoch": 13.25, + "learning_rate": 1.6762162827934474e-06, + "loss": 0.3323, + "step": 35876 + }, + { + "epoch": 13.25, + "learning_rate": 1.676179332430102e-06, + "loss": 0.3482, + "step": 35877 + }, + { + "epoch": 13.25, + "learning_rate": 1.6761423820667571e-06, + "loss": 0.2762, + "step": 35878 + }, + { + "epoch": 13.25, + "learning_rate": 1.6761054317034118e-06, + "loss": 0.2594, + "step": 35879 + }, + { + "epoch": 13.25, + "learning_rate": 1.6760684813400666e-06, + "loss": 0.4665, + "step": 35880 + }, + { + "epoch": 13.25, + "learning_rate": 1.6760315309767213e-06, + "loss": 0.3054, + "step": 35881 + }, + { + "epoch": 13.25, + "learning_rate": 1.6759945806133761e-06, + "loss": 0.26, + "step": 35882 + }, + { + "epoch": 13.25, + "learning_rate": 1.6759576302500307e-06, + "loss": 0.206, + "step": 35883 + }, + { + "epoch": 13.25, + "learning_rate": 1.6759206798866858e-06, + "loss": 0.2748, + "step": 35884 + }, + { + "epoch": 13.25, + "learning_rate": 1.6758837295233404e-06, + "loss": 0.3725, + "step": 35885 + }, + { + "epoch": 13.25, + "learning_rate": 1.6758467791599949e-06, + "loss": 0.2235, + "step": 35886 + }, + { + "epoch": 13.25, + "learning_rate": 1.67580982879665e-06, + "loss": 0.3039, + "step": 35887 + }, + { + "epoch": 13.25, + "learning_rate": 1.6757728784333046e-06, + "loss": 0.3933, + "step": 35888 + }, + { + "epoch": 13.25, + "learning_rate": 1.6757359280699594e-06, + "loss": 0.2265, + "step": 35889 + }, + { + "epoch": 13.25, + "learning_rate": 1.675698977706614e-06, + "loss": 0.3265, + "step": 35890 + }, + { + "epoch": 13.25, + "learning_rate": 1.6756620273432691e-06, + "loss": 0.4137, + "step": 35891 + }, + { + "epoch": 13.25, + "learning_rate": 1.6756250769799235e-06, + "loss": 0.2825, + "step": 35892 + }, + { + "epoch": 13.25, + "learning_rate": 1.6755881266165786e-06, + "loss": 0.3126, + "step": 35893 + }, + { + "epoch": 13.25, + "learning_rate": 1.6755511762532332e-06, + "loss": 0.2539, + "step": 35894 + }, + { + "epoch": 13.26, + "learning_rate": 1.675514225889888e-06, + "loss": 0.3365, + "step": 35895 + }, + { + "epoch": 13.26, + "learning_rate": 1.6754772755265427e-06, + "loss": 0.3692, + "step": 35896 + }, + { + "epoch": 13.26, + "learning_rate": 1.6754403251631973e-06, + "loss": 0.1936, + "step": 35897 + }, + { + "epoch": 13.26, + "learning_rate": 1.6754033747998522e-06, + "loss": 0.2764, + "step": 35898 + }, + { + "epoch": 13.26, + "learning_rate": 1.6753664244365068e-06, + "loss": 0.3947, + "step": 35899 + }, + { + "epoch": 13.26, + "learning_rate": 1.6753294740731619e-06, + "loss": 0.3229, + "step": 35900 + }, + { + "epoch": 13.26, + "learning_rate": 1.6752925237098165e-06, + "loss": 0.2365, + "step": 35901 + }, + { + "epoch": 13.26, + "learning_rate": 1.6752555733464714e-06, + "loss": 0.2545, + "step": 35902 + }, + { + "epoch": 13.26, + "learning_rate": 1.675218622983126e-06, + "loss": 0.4608, + "step": 35903 + }, + { + "epoch": 13.26, + "learning_rate": 1.6751816726197809e-06, + "loss": 0.2821, + "step": 35904 + }, + { + "epoch": 13.26, + "learning_rate": 1.6751447222564355e-06, + "loss": 0.3473, + "step": 35905 + }, + { + "epoch": 13.26, + "learning_rate": 1.6751077718930901e-06, + "loss": 0.2286, + "step": 35906 + }, + { + "epoch": 13.26, + "learning_rate": 1.6750708215297452e-06, + "loss": 0.213, + "step": 35907 + }, + { + "epoch": 13.26, + "learning_rate": 1.6750338711663996e-06, + "loss": 0.2305, + "step": 35908 + }, + { + "epoch": 13.26, + "learning_rate": 1.6749969208030547e-06, + "loss": 0.3572, + "step": 35909 + }, + { + "epoch": 13.26, + "learning_rate": 1.6749599704397093e-06, + "loss": 0.2419, + "step": 35910 + }, + { + "epoch": 13.26, + "learning_rate": 1.6749230200763642e-06, + "loss": 0.311, + "step": 35911 + }, + { + "epoch": 13.26, + "learning_rate": 1.6748860697130188e-06, + "loss": 0.2382, + "step": 35912 + }, + { + "epoch": 13.26, + "learning_rate": 1.6748491193496739e-06, + "loss": 0.2997, + "step": 35913 + }, + { + "epoch": 13.26, + "learning_rate": 1.6748121689863283e-06, + "loss": 0.3986, + "step": 35914 + }, + { + "epoch": 13.26, + "learning_rate": 1.6747752186229833e-06, + "loss": 0.2122, + "step": 35915 + }, + { + "epoch": 13.26, + "learning_rate": 1.674738268259638e-06, + "loss": 0.4164, + "step": 35916 + }, + { + "epoch": 13.26, + "learning_rate": 1.6747013178962926e-06, + "loss": 0.2223, + "step": 35917 + }, + { + "epoch": 13.26, + "learning_rate": 1.6746643675329475e-06, + "loss": 0.4872, + "step": 35918 + }, + { + "epoch": 13.26, + "learning_rate": 1.674627417169602e-06, + "loss": 0.2569, + "step": 35919 + }, + { + "epoch": 13.26, + "learning_rate": 1.674590466806257e-06, + "loss": 0.2623, + "step": 35920 + }, + { + "epoch": 13.26, + "learning_rate": 1.6745535164429116e-06, + "loss": 0.2126, + "step": 35921 + }, + { + "epoch": 13.27, + "learning_rate": 1.6745165660795666e-06, + "loss": 0.2957, + "step": 35922 + }, + { + "epoch": 13.27, + "learning_rate": 1.6744796157162213e-06, + "loss": 0.2831, + "step": 35923 + }, + { + "epoch": 13.27, + "learning_rate": 1.6744426653528761e-06, + "loss": 0.2573, + "step": 35924 + }, + { + "epoch": 13.27, + "learning_rate": 1.6744057149895308e-06, + "loss": 0.2441, + "step": 35925 + }, + { + "epoch": 13.27, + "learning_rate": 1.6743687646261854e-06, + "loss": 0.2367, + "step": 35926 + }, + { + "epoch": 13.27, + "learning_rate": 1.6743318142628403e-06, + "loss": 0.3425, + "step": 35927 + }, + { + "epoch": 13.27, + "learning_rate": 1.6742948638994949e-06, + "loss": 0.2062, + "step": 35928 + }, + { + "epoch": 13.27, + "learning_rate": 1.67425791353615e-06, + "loss": 0.1667, + "step": 35929 + }, + { + "epoch": 13.27, + "learning_rate": 1.6742209631728046e-06, + "loss": 0.4575, + "step": 35930 + }, + { + "epoch": 13.27, + "learning_rate": 1.6741840128094594e-06, + "loss": 0.241, + "step": 35931 + }, + { + "epoch": 13.27, + "learning_rate": 1.674147062446114e-06, + "loss": 0.2174, + "step": 35932 + }, + { + "epoch": 13.27, + "learning_rate": 1.674110112082769e-06, + "loss": 0.4305, + "step": 35933 + }, + { + "epoch": 13.27, + "learning_rate": 1.6740731617194236e-06, + "loss": 0.3315, + "step": 35934 + }, + { + "epoch": 13.27, + "learning_rate": 1.6740362113560786e-06, + "loss": 0.3174, + "step": 35935 + }, + { + "epoch": 13.27, + "learning_rate": 1.6739992609927333e-06, + "loss": 0.3029, + "step": 35936 + }, + { + "epoch": 13.27, + "learning_rate": 1.6739623106293877e-06, + "loss": 0.2113, + "step": 35937 + }, + { + "epoch": 13.27, + "learning_rate": 1.6739253602660427e-06, + "loss": 0.3597, + "step": 35938 + }, + { + "epoch": 13.27, + "learning_rate": 1.6738884099026974e-06, + "loss": 0.2514, + "step": 35939 + }, + { + "epoch": 13.27, + "learning_rate": 1.6738514595393522e-06, + "loss": 0.2261, + "step": 35940 + }, + { + "epoch": 13.27, + "learning_rate": 1.6738145091760069e-06, + "loss": 0.282, + "step": 35941 + }, + { + "epoch": 13.27, + "learning_rate": 1.673777558812662e-06, + "loss": 0.2227, + "step": 35942 + }, + { + "epoch": 13.27, + "learning_rate": 1.6737406084493163e-06, + "loss": 0.1902, + "step": 35943 + }, + { + "epoch": 13.27, + "learning_rate": 1.6737036580859714e-06, + "loss": 0.1713, + "step": 35944 + }, + { + "epoch": 13.27, + "learning_rate": 1.673666707722626e-06, + "loss": 0.2833, + "step": 35945 + }, + { + "epoch": 13.27, + "learning_rate": 1.6736297573592807e-06, + "loss": 0.3591, + "step": 35946 + }, + { + "epoch": 13.27, + "learning_rate": 1.6735928069959355e-06, + "loss": 0.2962, + "step": 35947 + }, + { + "epoch": 13.27, + "learning_rate": 1.6735558566325902e-06, + "loss": 0.2514, + "step": 35948 + }, + { + "epoch": 13.28, + "learning_rate": 1.673518906269245e-06, + "loss": 0.1429, + "step": 35949 + }, + { + "epoch": 13.28, + "learning_rate": 1.6734819559058996e-06, + "loss": 0.2684, + "step": 35950 + }, + { + "epoch": 13.28, + "learning_rate": 1.6734450055425547e-06, + "loss": 0.3029, + "step": 35951 + }, + { + "epoch": 13.28, + "learning_rate": 1.6734080551792093e-06, + "loss": 0.3126, + "step": 35952 + }, + { + "epoch": 13.28, + "learning_rate": 1.6733711048158642e-06, + "loss": 0.2737, + "step": 35953 + }, + { + "epoch": 13.28, + "learning_rate": 1.6733341544525188e-06, + "loss": 0.1936, + "step": 35954 + }, + { + "epoch": 13.28, + "learning_rate": 1.6732972040891737e-06, + "loss": 0.3625, + "step": 35955 + }, + { + "epoch": 13.28, + "learning_rate": 1.6732602537258283e-06, + "loss": 0.2406, + "step": 35956 + }, + { + "epoch": 13.28, + "learning_rate": 1.673223303362483e-06, + "loss": 0.2304, + "step": 35957 + }, + { + "epoch": 13.28, + "learning_rate": 1.673186352999138e-06, + "loss": 0.309, + "step": 35958 + }, + { + "epoch": 13.28, + "learning_rate": 1.6731494026357924e-06, + "loss": 0.3806, + "step": 35959 + }, + { + "epoch": 13.28, + "learning_rate": 1.6731124522724475e-06, + "loss": 0.226, + "step": 35960 + }, + { + "epoch": 13.28, + "learning_rate": 1.6730755019091021e-06, + "loss": 0.3139, + "step": 35961 + }, + { + "epoch": 13.28, + "learning_rate": 1.673038551545757e-06, + "loss": 0.4934, + "step": 35962 + }, + { + "epoch": 13.28, + "learning_rate": 1.6730016011824116e-06, + "loss": 0.2519, + "step": 35963 + }, + { + "epoch": 13.28, + "learning_rate": 1.6729646508190667e-06, + "loss": 0.1898, + "step": 35964 + }, + { + "epoch": 13.28, + "learning_rate": 1.672927700455721e-06, + "loss": 0.3321, + "step": 35965 + }, + { + "epoch": 13.28, + "learning_rate": 1.6728907500923757e-06, + "loss": 0.3128, + "step": 35966 + }, + { + "epoch": 13.28, + "learning_rate": 1.6728537997290308e-06, + "loss": 0.2296, + "step": 35967 + }, + { + "epoch": 13.28, + "learning_rate": 1.6728168493656854e-06, + "loss": 0.2503, + "step": 35968 + }, + { + "epoch": 13.28, + "learning_rate": 1.6727798990023403e-06, + "loss": 0.2756, + "step": 35969 + }, + { + "epoch": 13.28, + "learning_rate": 1.672742948638995e-06, + "loss": 0.2007, + "step": 35970 + }, + { + "epoch": 13.28, + "learning_rate": 1.6727059982756498e-06, + "loss": 0.3049, + "step": 35971 + }, + { + "epoch": 13.28, + "learning_rate": 1.6726690479123044e-06, + "loss": 0.2291, + "step": 35972 + }, + { + "epoch": 13.28, + "learning_rate": 1.6726320975489595e-06, + "loss": 0.2993, + "step": 35973 + }, + { + "epoch": 13.28, + "learning_rate": 1.672595147185614e-06, + "loss": 0.2698, + "step": 35974 + }, + { + "epoch": 13.28, + "learning_rate": 1.672558196822269e-06, + "loss": 0.1736, + "step": 35975 + }, + { + "epoch": 13.29, + "learning_rate": 1.6725212464589236e-06, + "loss": 0.2248, + "step": 35976 + }, + { + "epoch": 13.29, + "learning_rate": 1.6724842960955782e-06, + "loss": 0.1849, + "step": 35977 + }, + { + "epoch": 13.29, + "learning_rate": 1.672447345732233e-06, + "loss": 0.3886, + "step": 35978 + }, + { + "epoch": 13.29, + "learning_rate": 1.6724103953688877e-06, + "loss": 0.3518, + "step": 35979 + }, + { + "epoch": 13.29, + "learning_rate": 1.6723734450055428e-06, + "loss": 0.1923, + "step": 35980 + }, + { + "epoch": 13.29, + "learning_rate": 1.6723364946421972e-06, + "loss": 0.3511, + "step": 35981 + }, + { + "epoch": 13.29, + "learning_rate": 1.6722995442788522e-06, + "loss": 0.3824, + "step": 35982 + }, + { + "epoch": 13.29, + "learning_rate": 1.6722625939155069e-06, + "loss": 0.421, + "step": 35983 + }, + { + "epoch": 13.29, + "learning_rate": 1.6722256435521617e-06, + "loss": 0.2906, + "step": 35984 + }, + { + "epoch": 13.29, + "learning_rate": 1.6721886931888164e-06, + "loss": 0.2352, + "step": 35985 + }, + { + "epoch": 13.29, + "learning_rate": 1.672151742825471e-06, + "loss": 0.1505, + "step": 35986 + }, + { + "epoch": 13.29, + "learning_rate": 1.6721147924621259e-06, + "loss": 0.2672, + "step": 35987 + }, + { + "epoch": 13.29, + "learning_rate": 1.6720778420987805e-06, + "loss": 0.2497, + "step": 35988 + }, + { + "epoch": 13.29, + "learning_rate": 1.6720408917354355e-06, + "loss": 0.3175, + "step": 35989 + }, + { + "epoch": 13.29, + "learning_rate": 1.6720039413720902e-06, + "loss": 0.2995, + "step": 35990 + }, + { + "epoch": 13.29, + "learning_rate": 1.671966991008745e-06, + "loss": 0.2016, + "step": 35991 + }, + { + "epoch": 13.29, + "learning_rate": 1.6719300406453997e-06, + "loss": 0.3249, + "step": 35992 + }, + { + "epoch": 13.29, + "learning_rate": 1.6718930902820545e-06, + "loss": 0.2037, + "step": 35993 + }, + { + "epoch": 13.29, + "learning_rate": 1.6718561399187092e-06, + "loss": 0.4643, + "step": 35994 + }, + { + "epoch": 13.29, + "learning_rate": 1.6718191895553642e-06, + "loss": 0.2766, + "step": 35995 + }, + { + "epoch": 13.29, + "learning_rate": 1.6717822391920188e-06, + "loss": 0.2435, + "step": 35996 + }, + { + "epoch": 13.29, + "learning_rate": 1.6717452888286735e-06, + "loss": 0.2464, + "step": 35997 + }, + { + "epoch": 13.29, + "learning_rate": 1.6717083384653283e-06, + "loss": 0.2006, + "step": 35998 + }, + { + "epoch": 13.29, + "learning_rate": 1.671671388101983e-06, + "loss": 0.233, + "step": 35999 + }, + { + "epoch": 13.29, + "learning_rate": 1.6716344377386378e-06, + "loss": 0.2264, + "step": 36000 + }, + { + "epoch": 13.29, + "learning_rate": 1.6715974873752925e-06, + "loss": 0.2345, + "step": 36001 + }, + { + "epoch": 13.29, + "learning_rate": 1.6715605370119475e-06, + "loss": 0.3151, + "step": 36002 + }, + { + "epoch": 13.3, + "learning_rate": 1.6715235866486022e-06, + "loss": 0.3097, + "step": 36003 + }, + { + "epoch": 13.3, + "learning_rate": 1.671486636285257e-06, + "loss": 0.2542, + "step": 36004 + }, + { + "epoch": 13.3, + "learning_rate": 1.6714496859219116e-06, + "loss": 0.2945, + "step": 36005 + }, + { + "epoch": 13.3, + "learning_rate": 1.6714127355585663e-06, + "loss": 0.3282, + "step": 36006 + }, + { + "epoch": 13.3, + "learning_rate": 1.6713757851952211e-06, + "loss": 0.3044, + "step": 36007 + }, + { + "epoch": 13.3, + "learning_rate": 1.6713388348318758e-06, + "loss": 0.3053, + "step": 36008 + }, + { + "epoch": 13.3, + "learning_rate": 1.6713018844685308e-06, + "loss": 0.3209, + "step": 36009 + }, + { + "epoch": 13.3, + "learning_rate": 1.6712649341051852e-06, + "loss": 0.2391, + "step": 36010 + }, + { + "epoch": 13.3, + "learning_rate": 1.6712279837418403e-06, + "loss": 0.4663, + "step": 36011 + }, + { + "epoch": 13.3, + "learning_rate": 1.671191033378495e-06, + "loss": 0.2999, + "step": 36012 + }, + { + "epoch": 13.3, + "learning_rate": 1.6711540830151498e-06, + "loss": 0.255, + "step": 36013 + }, + { + "epoch": 13.3, + "learning_rate": 1.6711171326518044e-06, + "loss": 0.2175, + "step": 36014 + }, + { + "epoch": 13.3, + "learning_rate": 1.671080182288459e-06, + "loss": 0.2188, + "step": 36015 + }, + { + "epoch": 13.3, + "learning_rate": 1.671043231925114e-06, + "loss": 0.2769, + "step": 36016 + }, + { + "epoch": 13.3, + "learning_rate": 1.6710062815617685e-06, + "loss": 0.3326, + "step": 36017 + }, + { + "epoch": 13.3, + "learning_rate": 1.6709693311984236e-06, + "loss": 0.2227, + "step": 36018 + }, + { + "epoch": 13.3, + "learning_rate": 1.6709323808350782e-06, + "loss": 0.4042, + "step": 36019 + }, + { + "epoch": 13.3, + "learning_rate": 1.670895430471733e-06, + "loss": 0.2763, + "step": 36020 + }, + { + "epoch": 13.3, + "learning_rate": 1.6708584801083877e-06, + "loss": 0.2138, + "step": 36021 + }, + { + "epoch": 13.3, + "learning_rate": 1.6708215297450426e-06, + "loss": 0.1582, + "step": 36022 + }, + { + "epoch": 13.3, + "learning_rate": 1.6707845793816972e-06, + "loss": 0.3207, + "step": 36023 + }, + { + "epoch": 13.3, + "learning_rate": 1.6707476290183523e-06, + "loss": 0.2506, + "step": 36024 + }, + { + "epoch": 13.3, + "learning_rate": 1.670710678655007e-06, + "loss": 0.177, + "step": 36025 + }, + { + "epoch": 13.3, + "learning_rate": 1.6706737282916613e-06, + "loss": 0.2929, + "step": 36026 + }, + { + "epoch": 13.3, + "learning_rate": 1.6706367779283164e-06, + "loss": 0.2303, + "step": 36027 + }, + { + "epoch": 13.3, + "learning_rate": 1.670599827564971e-06, + "loss": 0.3146, + "step": 36028 + }, + { + "epoch": 13.3, + "learning_rate": 1.6705628772016259e-06, + "loss": 0.2853, + "step": 36029 + }, + { + "epoch": 13.31, + "learning_rate": 1.6705259268382805e-06, + "loss": 0.1899, + "step": 36030 + }, + { + "epoch": 13.31, + "learning_rate": 1.6704889764749356e-06, + "loss": 0.2388, + "step": 36031 + }, + { + "epoch": 13.31, + "learning_rate": 1.67045202611159e-06, + "loss": 0.2714, + "step": 36032 + }, + { + "epoch": 13.31, + "learning_rate": 1.670415075748245e-06, + "loss": 0.2787, + "step": 36033 + }, + { + "epoch": 13.31, + "learning_rate": 1.6703781253848997e-06, + "loss": 0.1867, + "step": 36034 + }, + { + "epoch": 13.31, + "learning_rate": 1.6703411750215543e-06, + "loss": 0.3068, + "step": 36035 + }, + { + "epoch": 13.31, + "learning_rate": 1.6703042246582092e-06, + "loss": 0.2367, + "step": 36036 + }, + { + "epoch": 13.31, + "learning_rate": 1.6702672742948638e-06, + "loss": 0.1431, + "step": 36037 + }, + { + "epoch": 13.31, + "learning_rate": 1.6702303239315187e-06, + "loss": 0.2194, + "step": 36038 + }, + { + "epoch": 13.31, + "learning_rate": 1.6701933735681733e-06, + "loss": 0.3037, + "step": 36039 + }, + { + "epoch": 13.31, + "learning_rate": 1.6701564232048284e-06, + "loss": 0.2895, + "step": 36040 + }, + { + "epoch": 13.31, + "learning_rate": 1.670119472841483e-06, + "loss": 0.2624, + "step": 36041 + }, + { + "epoch": 13.31, + "learning_rate": 1.6700825224781378e-06, + "loss": 0.2669, + "step": 36042 + }, + { + "epoch": 13.31, + "learning_rate": 1.6700455721147925e-06, + "loss": 0.2931, + "step": 36043 + }, + { + "epoch": 13.31, + "learning_rate": 1.6700086217514473e-06, + "loss": 0.2359, + "step": 36044 + }, + { + "epoch": 13.31, + "learning_rate": 1.669971671388102e-06, + "loss": 0.386, + "step": 36045 + }, + { + "epoch": 13.31, + "learning_rate": 1.6699347210247566e-06, + "loss": 0.3151, + "step": 36046 + }, + { + "epoch": 13.31, + "learning_rate": 1.6698977706614117e-06, + "loss": 0.2463, + "step": 36047 + }, + { + "epoch": 13.31, + "learning_rate": 1.6698608202980663e-06, + "loss": 0.2374, + "step": 36048 + }, + { + "epoch": 13.31, + "learning_rate": 1.6698238699347211e-06, + "loss": 0.4107, + "step": 36049 + }, + { + "epoch": 13.31, + "learning_rate": 1.6697869195713758e-06, + "loss": 0.3177, + "step": 36050 + }, + { + "epoch": 13.31, + "learning_rate": 1.6697499692080306e-06, + "loss": 0.252, + "step": 36051 + }, + { + "epoch": 13.31, + "learning_rate": 1.6697130188446853e-06, + "loss": 0.1611, + "step": 36052 + }, + { + "epoch": 13.31, + "learning_rate": 1.6696760684813403e-06, + "loss": 0.5156, + "step": 36053 + }, + { + "epoch": 13.31, + "learning_rate": 1.6696391181179948e-06, + "loss": 0.2476, + "step": 36054 + }, + { + "epoch": 13.31, + "learning_rate": 1.6696021677546494e-06, + "loss": 0.3425, + "step": 36055 + }, + { + "epoch": 13.31, + "learning_rate": 1.6695652173913044e-06, + "loss": 0.3264, + "step": 36056 + }, + { + "epoch": 13.31, + "learning_rate": 1.669528267027959e-06, + "loss": 0.1731, + "step": 36057 + }, + { + "epoch": 13.32, + "learning_rate": 1.669491316664614e-06, + "loss": 0.3148, + "step": 36058 + }, + { + "epoch": 13.32, + "learning_rate": 1.6694543663012686e-06, + "loss": 0.3579, + "step": 36059 + }, + { + "epoch": 13.32, + "learning_rate": 1.6694174159379234e-06, + "loss": 0.3595, + "step": 36060 + }, + { + "epoch": 13.32, + "learning_rate": 1.669380465574578e-06, + "loss": 0.2314, + "step": 36061 + }, + { + "epoch": 13.32, + "learning_rate": 1.6693435152112331e-06, + "loss": 0.1684, + "step": 36062 + }, + { + "epoch": 13.32, + "learning_rate": 1.6693065648478877e-06, + "loss": 0.3446, + "step": 36063 + }, + { + "epoch": 13.32, + "learning_rate": 1.6692696144845426e-06, + "loss": 0.3067, + "step": 36064 + }, + { + "epoch": 13.32, + "learning_rate": 1.6692326641211972e-06, + "loss": 0.2574, + "step": 36065 + }, + { + "epoch": 13.32, + "learning_rate": 1.6691957137578519e-06, + "loss": 0.1778, + "step": 36066 + }, + { + "epoch": 13.32, + "learning_rate": 1.6691587633945067e-06, + "loss": 0.3254, + "step": 36067 + }, + { + "epoch": 13.32, + "learning_rate": 1.6691218130311614e-06, + "loss": 0.3035, + "step": 36068 + }, + { + "epoch": 13.32, + "learning_rate": 1.6690848626678164e-06, + "loss": 0.2884, + "step": 36069 + }, + { + "epoch": 13.32, + "learning_rate": 1.669047912304471e-06, + "loss": 0.304, + "step": 36070 + }, + { + "epoch": 13.32, + "learning_rate": 1.669010961941126e-06, + "loss": 0.2815, + "step": 36071 + }, + { + "epoch": 13.32, + "learning_rate": 1.6689740115777805e-06, + "loss": 0.2973, + "step": 36072 + }, + { + "epoch": 13.32, + "learning_rate": 1.6689370612144354e-06, + "loss": 0.2032, + "step": 36073 + }, + { + "epoch": 13.32, + "learning_rate": 1.66890011085109e-06, + "loss": 0.331, + "step": 36074 + }, + { + "epoch": 13.32, + "learning_rate": 1.6688631604877447e-06, + "loss": 0.3473, + "step": 36075 + }, + { + "epoch": 13.32, + "learning_rate": 1.6688262101243997e-06, + "loss": 0.3119, + "step": 36076 + }, + { + "epoch": 13.32, + "learning_rate": 1.6687892597610541e-06, + "loss": 0.4199, + "step": 36077 + }, + { + "epoch": 13.32, + "learning_rate": 1.6687523093977092e-06, + "loss": 0.2784, + "step": 36078 + }, + { + "epoch": 13.32, + "learning_rate": 1.6687153590343638e-06, + "loss": 0.2021, + "step": 36079 + }, + { + "epoch": 13.32, + "learning_rate": 1.6686784086710187e-06, + "loss": 0.212, + "step": 36080 + }, + { + "epoch": 13.32, + "learning_rate": 1.6686414583076733e-06, + "loss": 0.1778, + "step": 36081 + }, + { + "epoch": 13.32, + "learning_rate": 1.6686045079443284e-06, + "loss": 0.2026, + "step": 36082 + }, + { + "epoch": 13.32, + "learning_rate": 1.6685675575809828e-06, + "loss": 0.2407, + "step": 36083 + }, + { + "epoch": 13.32, + "learning_rate": 1.6685306072176379e-06, + "loss": 0.2994, + "step": 36084 + }, + { + "epoch": 13.33, + "learning_rate": 1.6684936568542925e-06, + "loss": 0.2258, + "step": 36085 + }, + { + "epoch": 13.33, + "learning_rate": 1.6684567064909471e-06, + "loss": 0.2868, + "step": 36086 + }, + { + "epoch": 13.33, + "learning_rate": 1.668419756127602e-06, + "loss": 0.1869, + "step": 36087 + }, + { + "epoch": 13.33, + "learning_rate": 1.6683828057642566e-06, + "loss": 0.2877, + "step": 36088 + }, + { + "epoch": 13.33, + "learning_rate": 1.6683458554009115e-06, + "loss": 0.322, + "step": 36089 + }, + { + "epoch": 13.33, + "learning_rate": 1.6683089050375661e-06, + "loss": 0.3012, + "step": 36090 + }, + { + "epoch": 13.33, + "learning_rate": 1.6682719546742212e-06, + "loss": 0.2084, + "step": 36091 + }, + { + "epoch": 13.33, + "learning_rate": 1.6682350043108758e-06, + "loss": 0.4206, + "step": 36092 + }, + { + "epoch": 13.33, + "learning_rate": 1.6681980539475307e-06, + "loss": 0.1459, + "step": 36093 + }, + { + "epoch": 13.33, + "learning_rate": 1.6681611035841853e-06, + "loss": 0.3306, + "step": 36094 + }, + { + "epoch": 13.33, + "learning_rate": 1.66812415322084e-06, + "loss": 0.3316, + "step": 36095 + }, + { + "epoch": 13.33, + "learning_rate": 1.6680872028574948e-06, + "loss": 0.1731, + "step": 36096 + }, + { + "epoch": 13.33, + "learning_rate": 1.6680502524941494e-06, + "loss": 0.2365, + "step": 36097 + }, + { + "epoch": 13.33, + "learning_rate": 1.6680133021308045e-06, + "loss": 0.3468, + "step": 36098 + }, + { + "epoch": 13.33, + "learning_rate": 1.667976351767459e-06, + "loss": 0.324, + "step": 36099 + }, + { + "epoch": 13.33, + "learning_rate": 1.667939401404114e-06, + "loss": 0.3604, + "step": 36100 + }, + { + "epoch": 13.33, + "learning_rate": 1.6679024510407686e-06, + "loss": 0.4402, + "step": 36101 + }, + { + "epoch": 13.33, + "learning_rate": 1.6678655006774234e-06, + "loss": 0.2729, + "step": 36102 + }, + { + "epoch": 13.33, + "learning_rate": 1.667828550314078e-06, + "loss": 0.3009, + "step": 36103 + }, + { + "epoch": 13.33, + "learning_rate": 1.6677915999507331e-06, + "loss": 0.3374, + "step": 36104 + }, + { + "epoch": 13.33, + "learning_rate": 1.6677546495873876e-06, + "loss": 0.2129, + "step": 36105 + }, + { + "epoch": 13.33, + "learning_rate": 1.6677176992240422e-06, + "loss": 0.2123, + "step": 36106 + }, + { + "epoch": 13.33, + "learning_rate": 1.6676807488606973e-06, + "loss": 0.2575, + "step": 36107 + }, + { + "epoch": 13.33, + "learning_rate": 1.667643798497352e-06, + "loss": 0.1954, + "step": 36108 + }, + { + "epoch": 13.33, + "learning_rate": 1.6676068481340067e-06, + "loss": 0.32, + "step": 36109 + }, + { + "epoch": 13.33, + "learning_rate": 1.6675698977706614e-06, + "loss": 0.3375, + "step": 36110 + }, + { + "epoch": 13.33, + "learning_rate": 1.6675329474073162e-06, + "loss": 0.1956, + "step": 36111 + }, + { + "epoch": 13.34, + "learning_rate": 1.6674959970439709e-06, + "loss": 0.3304, + "step": 36112 + }, + { + "epoch": 13.34, + "learning_rate": 1.667459046680626e-06, + "loss": 0.3279, + "step": 36113 + }, + { + "epoch": 13.34, + "learning_rate": 1.6674220963172806e-06, + "loss": 0.3575, + "step": 36114 + }, + { + "epoch": 13.34, + "learning_rate": 1.6673851459539352e-06, + "loss": 0.448, + "step": 36115 + }, + { + "epoch": 13.34, + "learning_rate": 1.66734819559059e-06, + "loss": 0.2692, + "step": 36116 + }, + { + "epoch": 13.34, + "learning_rate": 1.6673112452272447e-06, + "loss": 0.1754, + "step": 36117 + }, + { + "epoch": 13.34, + "learning_rate": 1.6672742948638995e-06, + "loss": 0.2225, + "step": 36118 + }, + { + "epoch": 13.34, + "learning_rate": 1.6672373445005542e-06, + "loss": 0.2541, + "step": 36119 + }, + { + "epoch": 13.34, + "learning_rate": 1.6672003941372092e-06, + "loss": 0.33, + "step": 36120 + }, + { + "epoch": 13.34, + "learning_rate": 1.6671634437738639e-06, + "loss": 0.2766, + "step": 36121 + }, + { + "epoch": 13.34, + "learning_rate": 1.6671264934105187e-06, + "loss": 0.2439, + "step": 36122 + }, + { + "epoch": 13.34, + "learning_rate": 1.6670895430471733e-06, + "loss": 0.3078, + "step": 36123 + }, + { + "epoch": 13.34, + "learning_rate": 1.6670525926838282e-06, + "loss": 0.3586, + "step": 36124 + }, + { + "epoch": 13.34, + "learning_rate": 1.6670156423204828e-06, + "loss": 0.2438, + "step": 36125 + }, + { + "epoch": 13.34, + "learning_rate": 1.6669786919571375e-06, + "loss": 0.3276, + "step": 36126 + }, + { + "epoch": 13.34, + "learning_rate": 1.6669417415937925e-06, + "loss": 0.3281, + "step": 36127 + }, + { + "epoch": 13.34, + "learning_rate": 1.666904791230447e-06, + "loss": 0.237, + "step": 36128 + }, + { + "epoch": 13.34, + "learning_rate": 1.666867840867102e-06, + "loss": 0.2618, + "step": 36129 + }, + { + "epoch": 13.34, + "learning_rate": 1.6668308905037566e-06, + "loss": 0.1873, + "step": 36130 + }, + { + "epoch": 13.34, + "learning_rate": 1.6667939401404115e-06, + "loss": 0.2643, + "step": 36131 + }, + { + "epoch": 13.34, + "learning_rate": 1.6667569897770661e-06, + "loss": 0.1796, + "step": 36132 + }, + { + "epoch": 13.34, + "learning_rate": 1.666720039413721e-06, + "loss": 0.2581, + "step": 36133 + }, + { + "epoch": 13.34, + "learning_rate": 1.6666830890503756e-06, + "loss": 0.3524, + "step": 36134 + }, + { + "epoch": 13.34, + "learning_rate": 1.6666461386870303e-06, + "loss": 0.3468, + "step": 36135 + }, + { + "epoch": 13.34, + "learning_rate": 1.6666091883236853e-06, + "loss": 0.3356, + "step": 36136 + }, + { + "epoch": 13.34, + "learning_rate": 1.66657223796034e-06, + "loss": 0.2115, + "step": 36137 + }, + { + "epoch": 13.34, + "learning_rate": 1.6665352875969948e-06, + "loss": 0.2439, + "step": 36138 + }, + { + "epoch": 13.35, + "learning_rate": 1.6664983372336494e-06, + "loss": 0.4012, + "step": 36139 + }, + { + "epoch": 13.35, + "learning_rate": 1.6664613868703043e-06, + "loss": 0.2157, + "step": 36140 + }, + { + "epoch": 13.35, + "learning_rate": 1.666424436506959e-06, + "loss": 0.3383, + "step": 36141 + }, + { + "epoch": 13.35, + "learning_rate": 1.666387486143614e-06, + "loss": 0.2153, + "step": 36142 + }, + { + "epoch": 13.35, + "learning_rate": 1.6663505357802686e-06, + "loss": 0.2814, + "step": 36143 + }, + { + "epoch": 13.35, + "learning_rate": 1.6663135854169235e-06, + "loss": 0.2828, + "step": 36144 + }, + { + "epoch": 13.35, + "learning_rate": 1.666276635053578e-06, + "loss": 0.2507, + "step": 36145 + }, + { + "epoch": 13.35, + "learning_rate": 1.6662396846902327e-06, + "loss": 0.2891, + "step": 36146 + }, + { + "epoch": 13.35, + "learning_rate": 1.6662027343268876e-06, + "loss": 0.1836, + "step": 36147 + }, + { + "epoch": 13.35, + "learning_rate": 1.6661657839635422e-06, + "loss": 0.2451, + "step": 36148 + }, + { + "epoch": 13.35, + "learning_rate": 1.6661288336001973e-06, + "loss": 0.1634, + "step": 36149 + }, + { + "epoch": 13.35, + "learning_rate": 1.6660918832368517e-06, + "loss": 0.3156, + "step": 36150 + }, + { + "epoch": 13.35, + "learning_rate": 1.6660549328735068e-06, + "loss": 0.2536, + "step": 36151 + }, + { + "epoch": 13.35, + "learning_rate": 1.6660179825101614e-06, + "loss": 0.3111, + "step": 36152 + }, + { + "epoch": 13.35, + "learning_rate": 1.6659810321468163e-06, + "loss": 0.2336, + "step": 36153 + }, + { + "epoch": 13.35, + "learning_rate": 1.6659440817834709e-06, + "loss": 0.4082, + "step": 36154 + }, + { + "epoch": 13.35, + "learning_rate": 1.6659071314201255e-06, + "loss": 0.2227, + "step": 36155 + }, + { + "epoch": 13.35, + "learning_rate": 1.6658701810567804e-06, + "loss": 0.2754, + "step": 36156 + }, + { + "epoch": 13.35, + "learning_rate": 1.665833230693435e-06, + "loss": 0.3376, + "step": 36157 + }, + { + "epoch": 13.35, + "learning_rate": 1.66579628033009e-06, + "loss": 0.2643, + "step": 36158 + }, + { + "epoch": 13.35, + "learning_rate": 1.6657593299667447e-06, + "loss": 0.2482, + "step": 36159 + }, + { + "epoch": 13.35, + "learning_rate": 1.6657223796033996e-06, + "loss": 0.2758, + "step": 36160 + }, + { + "epoch": 13.35, + "learning_rate": 1.6656854292400542e-06, + "loss": 0.3116, + "step": 36161 + }, + { + "epoch": 13.35, + "learning_rate": 1.665648478876709e-06, + "loss": 0.1853, + "step": 36162 + }, + { + "epoch": 13.35, + "learning_rate": 1.6656115285133637e-06, + "loss": 0.27, + "step": 36163 + }, + { + "epoch": 13.35, + "learning_rate": 1.6655745781500187e-06, + "loss": 0.2666, + "step": 36164 + }, + { + "epoch": 13.35, + "learning_rate": 1.6655376277866734e-06, + "loss": 0.4296, + "step": 36165 + }, + { + "epoch": 13.36, + "learning_rate": 1.6655006774233278e-06, + "loss": 0.1876, + "step": 36166 + }, + { + "epoch": 13.36, + "learning_rate": 1.6654637270599829e-06, + "loss": 0.4052, + "step": 36167 + }, + { + "epoch": 13.36, + "learning_rate": 1.6654267766966375e-06, + "loss": 0.2729, + "step": 36168 + }, + { + "epoch": 13.36, + "learning_rate": 1.6653898263332923e-06, + "loss": 0.4034, + "step": 36169 + }, + { + "epoch": 13.36, + "learning_rate": 1.665352875969947e-06, + "loss": 0.2296, + "step": 36170 + }, + { + "epoch": 13.36, + "learning_rate": 1.665315925606602e-06, + "loss": 0.3818, + "step": 36171 + }, + { + "epoch": 13.36, + "learning_rate": 1.6652789752432565e-06, + "loss": 0.3695, + "step": 36172 + }, + { + "epoch": 13.36, + "learning_rate": 1.6652420248799115e-06, + "loss": 0.3409, + "step": 36173 + }, + { + "epoch": 13.36, + "learning_rate": 1.6652050745165662e-06, + "loss": 0.3906, + "step": 36174 + }, + { + "epoch": 13.36, + "learning_rate": 1.6651681241532208e-06, + "loss": 0.2225, + "step": 36175 + }, + { + "epoch": 13.36, + "learning_rate": 1.6651311737898756e-06, + "loss": 0.3222, + "step": 36176 + }, + { + "epoch": 13.36, + "learning_rate": 1.6650942234265303e-06, + "loss": 0.4028, + "step": 36177 + }, + { + "epoch": 13.36, + "learning_rate": 1.6650572730631851e-06, + "loss": 0.196, + "step": 36178 + }, + { + "epoch": 13.36, + "learning_rate": 1.6650203226998398e-06, + "loss": 0.3279, + "step": 36179 + }, + { + "epoch": 13.36, + "learning_rate": 1.6649833723364948e-06, + "loss": 0.3008, + "step": 36180 + }, + { + "epoch": 13.36, + "learning_rate": 1.6649464219731495e-06, + "loss": 0.2916, + "step": 36181 + }, + { + "epoch": 13.36, + "learning_rate": 1.6649094716098043e-06, + "loss": 0.2885, + "step": 36182 + }, + { + "epoch": 13.36, + "learning_rate": 1.664872521246459e-06, + "loss": 0.316, + "step": 36183 + }, + { + "epoch": 13.36, + "learning_rate": 1.6648355708831138e-06, + "loss": 0.27, + "step": 36184 + }, + { + "epoch": 13.36, + "learning_rate": 1.6647986205197684e-06, + "loss": 0.2994, + "step": 36185 + }, + { + "epoch": 13.36, + "learning_rate": 1.664761670156423e-06, + "loss": 0.3701, + "step": 36186 + }, + { + "epoch": 13.36, + "learning_rate": 1.6647247197930781e-06, + "loss": 0.353, + "step": 36187 + }, + { + "epoch": 13.36, + "learning_rate": 1.6646877694297328e-06, + "loss": 0.178, + "step": 36188 + }, + { + "epoch": 13.36, + "learning_rate": 1.6646508190663876e-06, + "loss": 0.3584, + "step": 36189 + }, + { + "epoch": 13.36, + "learning_rate": 1.6646138687030422e-06, + "loss": 0.2205, + "step": 36190 + }, + { + "epoch": 13.36, + "learning_rate": 1.664576918339697e-06, + "loss": 0.2459, + "step": 36191 + }, + { + "epoch": 13.36, + "learning_rate": 1.6645399679763517e-06, + "loss": 0.3212, + "step": 36192 + }, + { + "epoch": 13.37, + "learning_rate": 1.6645030176130068e-06, + "loss": 0.324, + "step": 36193 + }, + { + "epoch": 13.37, + "learning_rate": 1.6644660672496614e-06, + "loss": 0.3225, + "step": 36194 + }, + { + "epoch": 13.37, + "learning_rate": 1.6644291168863159e-06, + "loss": 0.2394, + "step": 36195 + }, + { + "epoch": 13.37, + "learning_rate": 1.664392166522971e-06, + "loss": 0.2502, + "step": 36196 + }, + { + "epoch": 13.37, + "learning_rate": 1.6643552161596255e-06, + "loss": 0.3527, + "step": 36197 + }, + { + "epoch": 13.37, + "learning_rate": 1.6643182657962804e-06, + "loss": 0.2581, + "step": 36198 + }, + { + "epoch": 13.37, + "learning_rate": 1.664281315432935e-06, + "loss": 0.2322, + "step": 36199 + }, + { + "epoch": 13.37, + "learning_rate": 1.66424436506959e-06, + "loss": 0.2311, + "step": 36200 + }, + { + "epoch": 13.37, + "learning_rate": 1.6642074147062445e-06, + "loss": 0.3566, + "step": 36201 + }, + { + "epoch": 13.37, + "learning_rate": 1.6641704643428996e-06, + "loss": 0.3401, + "step": 36202 + }, + { + "epoch": 13.37, + "learning_rate": 1.6641335139795542e-06, + "loss": 0.3587, + "step": 36203 + }, + { + "epoch": 13.37, + "learning_rate": 1.664096563616209e-06, + "loss": 0.3836, + "step": 36204 + }, + { + "epoch": 13.37, + "learning_rate": 1.6640596132528637e-06, + "loss": 0.2703, + "step": 36205 + }, + { + "epoch": 13.37, + "learning_rate": 1.6640226628895183e-06, + "loss": 0.2176, + "step": 36206 + }, + { + "epoch": 13.37, + "learning_rate": 1.6639857125261732e-06, + "loss": 0.2814, + "step": 36207 + }, + { + "epoch": 13.37, + "learning_rate": 1.6639487621628278e-06, + "loss": 0.3272, + "step": 36208 + }, + { + "epoch": 13.37, + "learning_rate": 1.6639118117994829e-06, + "loss": 0.2624, + "step": 36209 + }, + { + "epoch": 13.37, + "learning_rate": 1.6638748614361375e-06, + "loss": 0.293, + "step": 36210 + }, + { + "epoch": 13.37, + "learning_rate": 1.6638379110727924e-06, + "loss": 0.2217, + "step": 36211 + }, + { + "epoch": 13.37, + "learning_rate": 1.663800960709447e-06, + "loss": 0.2237, + "step": 36212 + }, + { + "epoch": 13.37, + "learning_rate": 1.6637640103461018e-06, + "loss": 0.3566, + "step": 36213 + }, + { + "epoch": 13.37, + "learning_rate": 1.6637270599827565e-06, + "loss": 0.2094, + "step": 36214 + }, + { + "epoch": 13.37, + "learning_rate": 1.6636901096194111e-06, + "loss": 0.3001, + "step": 36215 + }, + { + "epoch": 13.37, + "learning_rate": 1.6636531592560662e-06, + "loss": 0.2399, + "step": 36216 + }, + { + "epoch": 13.37, + "learning_rate": 1.6636162088927206e-06, + "loss": 0.2339, + "step": 36217 + }, + { + "epoch": 13.37, + "learning_rate": 1.6635792585293757e-06, + "loss": 0.3751, + "step": 36218 + }, + { + "epoch": 13.37, + "learning_rate": 1.6635423081660303e-06, + "loss": 0.271, + "step": 36219 + }, + { + "epoch": 13.38, + "learning_rate": 1.6635053578026852e-06, + "loss": 0.4313, + "step": 36220 + }, + { + "epoch": 13.38, + "learning_rate": 1.6634684074393398e-06, + "loss": 0.3063, + "step": 36221 + }, + { + "epoch": 13.38, + "learning_rate": 1.6634314570759948e-06, + "loss": 0.385, + "step": 36222 + }, + { + "epoch": 13.38, + "learning_rate": 1.6633945067126493e-06, + "loss": 0.2803, + "step": 36223 + }, + { + "epoch": 13.38, + "learning_rate": 1.6633575563493043e-06, + "loss": 0.1346, + "step": 36224 + }, + { + "epoch": 13.38, + "learning_rate": 1.663320605985959e-06, + "loss": 0.3695, + "step": 36225 + }, + { + "epoch": 13.38, + "learning_rate": 1.6632836556226136e-06, + "loss": 0.382, + "step": 36226 + }, + { + "epoch": 13.38, + "learning_rate": 1.6632467052592685e-06, + "loss": 0.3934, + "step": 36227 + }, + { + "epoch": 13.38, + "learning_rate": 1.663209754895923e-06, + "loss": 0.3271, + "step": 36228 + }, + { + "epoch": 13.38, + "learning_rate": 1.663172804532578e-06, + "loss": 0.4368, + "step": 36229 + }, + { + "epoch": 13.38, + "learning_rate": 1.6631358541692326e-06, + "loss": 0.4115, + "step": 36230 + }, + { + "epoch": 13.38, + "learning_rate": 1.6630989038058876e-06, + "loss": 0.1921, + "step": 36231 + }, + { + "epoch": 13.38, + "learning_rate": 1.6630619534425423e-06, + "loss": 0.3338, + "step": 36232 + }, + { + "epoch": 13.38, + "learning_rate": 1.6630250030791971e-06, + "loss": 0.3408, + "step": 36233 + }, + { + "epoch": 13.38, + "learning_rate": 1.6629880527158518e-06, + "loss": 0.2596, + "step": 36234 + }, + { + "epoch": 13.38, + "learning_rate": 1.6629511023525064e-06, + "loss": 0.2382, + "step": 36235 + }, + { + "epoch": 13.38, + "learning_rate": 1.6629141519891612e-06, + "loss": 0.2984, + "step": 36236 + }, + { + "epoch": 13.38, + "learning_rate": 1.6628772016258159e-06, + "loss": 0.2411, + "step": 36237 + }, + { + "epoch": 13.38, + "learning_rate": 1.662840251262471e-06, + "loss": 0.3897, + "step": 36238 + }, + { + "epoch": 13.38, + "learning_rate": 1.6628033008991256e-06, + "loss": 0.3189, + "step": 36239 + }, + { + "epoch": 13.38, + "learning_rate": 1.6627663505357804e-06, + "loss": 0.3049, + "step": 36240 + }, + { + "epoch": 13.38, + "learning_rate": 1.662729400172435e-06, + "loss": 0.2562, + "step": 36241 + }, + { + "epoch": 13.38, + "learning_rate": 1.66269244980909e-06, + "loss": 0.3021, + "step": 36242 + }, + { + "epoch": 13.38, + "learning_rate": 1.6626554994457445e-06, + "loss": 0.2799, + "step": 36243 + }, + { + "epoch": 13.38, + "learning_rate": 1.6626185490823992e-06, + "loss": 0.1993, + "step": 36244 + }, + { + "epoch": 13.38, + "learning_rate": 1.662581598719054e-06, + "loss": 0.2006, + "step": 36245 + }, + { + "epoch": 13.38, + "learning_rate": 1.6625446483557087e-06, + "loss": 0.2913, + "step": 36246 + }, + { + "epoch": 13.39, + "learning_rate": 1.6625076979923637e-06, + "loss": 0.0799, + "step": 36247 + }, + { + "epoch": 13.39, + "learning_rate": 1.6624707476290184e-06, + "loss": 0.2807, + "step": 36248 + }, + { + "epoch": 13.39, + "learning_rate": 1.6624337972656732e-06, + "loss": 0.1737, + "step": 36249 + }, + { + "epoch": 13.39, + "learning_rate": 1.6623968469023278e-06, + "loss": 0.3584, + "step": 36250 + }, + { + "epoch": 13.39, + "learning_rate": 1.6623598965389827e-06, + "loss": 0.2292, + "step": 36251 + }, + { + "epoch": 13.39, + "learning_rate": 1.6623229461756373e-06, + "loss": 0.3718, + "step": 36252 + }, + { + "epoch": 13.39, + "learning_rate": 1.6622859958122924e-06, + "loss": 0.2573, + "step": 36253 + }, + { + "epoch": 13.39, + "learning_rate": 1.662249045448947e-06, + "loss": 0.3102, + "step": 36254 + }, + { + "epoch": 13.39, + "learning_rate": 1.6622120950856017e-06, + "loss": 0.1555, + "step": 36255 + }, + { + "epoch": 13.39, + "learning_rate": 1.6621751447222565e-06, + "loss": 0.2554, + "step": 36256 + }, + { + "epoch": 13.39, + "learning_rate": 1.6621381943589111e-06, + "loss": 0.2548, + "step": 36257 + }, + { + "epoch": 13.39, + "learning_rate": 1.662101243995566e-06, + "loss": 0.2289, + "step": 36258 + }, + { + "epoch": 13.39, + "learning_rate": 1.6620642936322206e-06, + "loss": 0.5352, + "step": 36259 + }, + { + "epoch": 13.39, + "learning_rate": 1.6620273432688757e-06, + "loss": 0.2584, + "step": 36260 + }, + { + "epoch": 13.39, + "learning_rate": 1.6619903929055303e-06, + "loss": 0.2958, + "step": 36261 + }, + { + "epoch": 13.39, + "learning_rate": 1.6619534425421852e-06, + "loss": 0.2138, + "step": 36262 + }, + { + "epoch": 13.39, + "learning_rate": 1.6619164921788398e-06, + "loss": 0.2323, + "step": 36263 + }, + { + "epoch": 13.39, + "learning_rate": 1.6618795418154944e-06, + "loss": 0.3241, + "step": 36264 + }, + { + "epoch": 13.39, + "learning_rate": 1.6618425914521493e-06, + "loss": 0.3861, + "step": 36265 + }, + { + "epoch": 13.39, + "learning_rate": 1.661805641088804e-06, + "loss": 0.2768, + "step": 36266 + }, + { + "epoch": 13.39, + "learning_rate": 1.661768690725459e-06, + "loss": 0.2938, + "step": 36267 + }, + { + "epoch": 13.39, + "learning_rate": 1.6617317403621134e-06, + "loss": 0.2676, + "step": 36268 + }, + { + "epoch": 13.39, + "learning_rate": 1.6616947899987685e-06, + "loss": 0.4257, + "step": 36269 + }, + { + "epoch": 13.39, + "learning_rate": 1.6616578396354231e-06, + "loss": 0.3153, + "step": 36270 + }, + { + "epoch": 13.39, + "learning_rate": 1.661620889272078e-06, + "loss": 0.2817, + "step": 36271 + }, + { + "epoch": 13.39, + "learning_rate": 1.6615839389087326e-06, + "loss": 0.1922, + "step": 36272 + }, + { + "epoch": 13.39, + "learning_rate": 1.6615469885453877e-06, + "loss": 0.2018, + "step": 36273 + }, + { + "epoch": 13.4, + "learning_rate": 1.661510038182042e-06, + "loss": 0.2386, + "step": 36274 + }, + { + "epoch": 13.4, + "learning_rate": 1.6614730878186967e-06, + "loss": 0.293, + "step": 36275 + }, + { + "epoch": 13.4, + "learning_rate": 1.6614361374553518e-06, + "loss": 0.2777, + "step": 36276 + }, + { + "epoch": 13.4, + "learning_rate": 1.6613991870920064e-06, + "loss": 0.2106, + "step": 36277 + }, + { + "epoch": 13.4, + "learning_rate": 1.6613622367286613e-06, + "loss": 0.2495, + "step": 36278 + }, + { + "epoch": 13.4, + "learning_rate": 1.661325286365316e-06, + "loss": 0.3636, + "step": 36279 + }, + { + "epoch": 13.4, + "learning_rate": 1.6612883360019707e-06, + "loss": 0.3448, + "step": 36280 + }, + { + "epoch": 13.4, + "learning_rate": 1.6612513856386254e-06, + "loss": 0.2364, + "step": 36281 + }, + { + "epoch": 13.4, + "learning_rate": 1.6612144352752804e-06, + "loss": 0.2576, + "step": 36282 + }, + { + "epoch": 13.4, + "learning_rate": 1.661177484911935e-06, + "loss": 0.3799, + "step": 36283 + }, + { + "epoch": 13.4, + "learning_rate": 1.6611405345485895e-06, + "loss": 0.2457, + "step": 36284 + }, + { + "epoch": 13.4, + "learning_rate": 1.6611035841852446e-06, + "loss": 0.5336, + "step": 36285 + }, + { + "epoch": 13.4, + "learning_rate": 1.6610666338218992e-06, + "loss": 0.2958, + "step": 36286 + }, + { + "epoch": 13.4, + "learning_rate": 1.661029683458554e-06, + "loss": 0.3879, + "step": 36287 + }, + { + "epoch": 13.4, + "learning_rate": 1.6609927330952087e-06, + "loss": 0.2229, + "step": 36288 + }, + { + "epoch": 13.4, + "learning_rate": 1.6609557827318637e-06, + "loss": 0.2844, + "step": 36289 + }, + { + "epoch": 13.4, + "learning_rate": 1.6609188323685182e-06, + "loss": 0.2535, + "step": 36290 + }, + { + "epoch": 13.4, + "learning_rate": 1.6608818820051732e-06, + "loss": 0.3244, + "step": 36291 + }, + { + "epoch": 13.4, + "learning_rate": 1.6608449316418279e-06, + "loss": 0.304, + "step": 36292 + }, + { + "epoch": 13.4, + "learning_rate": 1.6608079812784827e-06, + "loss": 0.3679, + "step": 36293 + }, + { + "epoch": 13.4, + "learning_rate": 1.6607710309151374e-06, + "loss": 0.1495, + "step": 36294 + }, + { + "epoch": 13.4, + "learning_rate": 1.660734080551792e-06, + "loss": 0.2843, + "step": 36295 + }, + { + "epoch": 13.4, + "learning_rate": 1.6606971301884468e-06, + "loss": 0.2117, + "step": 36296 + }, + { + "epoch": 13.4, + "learning_rate": 1.6606601798251015e-06, + "loss": 0.2826, + "step": 36297 + }, + { + "epoch": 13.4, + "learning_rate": 1.6606232294617565e-06, + "loss": 0.2579, + "step": 36298 + }, + { + "epoch": 13.4, + "learning_rate": 1.6605862790984112e-06, + "loss": 0.2169, + "step": 36299 + }, + { + "epoch": 13.4, + "learning_rate": 1.660549328735066e-06, + "loss": 0.3213, + "step": 36300 + }, + { + "epoch": 13.41, + "learning_rate": 1.6605123783717207e-06, + "loss": 0.2668, + "step": 36301 + }, + { + "epoch": 13.41, + "learning_rate": 1.6604754280083755e-06, + "loss": 0.3026, + "step": 36302 + }, + { + "epoch": 13.41, + "learning_rate": 1.6604384776450301e-06, + "loss": 0.2416, + "step": 36303 + }, + { + "epoch": 13.41, + "learning_rate": 1.6604015272816848e-06, + "loss": 0.1814, + "step": 36304 + }, + { + "epoch": 13.41, + "learning_rate": 1.6603645769183398e-06, + "loss": 0.2895, + "step": 36305 + }, + { + "epoch": 13.41, + "learning_rate": 1.6603276265549945e-06, + "loss": 0.3439, + "step": 36306 + }, + { + "epoch": 13.41, + "learning_rate": 1.6602906761916493e-06, + "loss": 0.2427, + "step": 36307 + }, + { + "epoch": 13.41, + "learning_rate": 1.660253725828304e-06, + "loss": 0.3211, + "step": 36308 + }, + { + "epoch": 13.41, + "learning_rate": 1.6602167754649588e-06, + "loss": 0.3551, + "step": 36309 + }, + { + "epoch": 13.41, + "learning_rate": 1.6601798251016134e-06, + "loss": 0.3189, + "step": 36310 + }, + { + "epoch": 13.41, + "learning_rate": 1.6601428747382685e-06, + "loss": 0.3881, + "step": 36311 + }, + { + "epoch": 13.41, + "learning_rate": 1.6601059243749231e-06, + "loss": 0.2737, + "step": 36312 + }, + { + "epoch": 13.41, + "learning_rate": 1.660068974011578e-06, + "loss": 0.1685, + "step": 36313 + }, + { + "epoch": 13.41, + "learning_rate": 1.6600320236482326e-06, + "loss": 0.1871, + "step": 36314 + }, + { + "epoch": 13.41, + "learning_rate": 1.6599950732848873e-06, + "loss": 0.3257, + "step": 36315 + }, + { + "epoch": 13.41, + "learning_rate": 1.6599581229215421e-06, + "loss": 0.2778, + "step": 36316 + }, + { + "epoch": 13.41, + "learning_rate": 1.6599211725581967e-06, + "loss": 0.311, + "step": 36317 + }, + { + "epoch": 13.41, + "learning_rate": 1.6598842221948516e-06, + "loss": 0.3186, + "step": 36318 + }, + { + "epoch": 13.41, + "learning_rate": 1.6598472718315062e-06, + "loss": 0.4429, + "step": 36319 + }, + { + "epoch": 13.41, + "learning_rate": 1.6598103214681613e-06, + "loss": 0.2103, + "step": 36320 + }, + { + "epoch": 13.41, + "learning_rate": 1.659773371104816e-06, + "loss": 0.2191, + "step": 36321 + }, + { + "epoch": 13.41, + "learning_rate": 1.6597364207414708e-06, + "loss": 0.2125, + "step": 36322 + }, + { + "epoch": 13.41, + "learning_rate": 1.6596994703781254e-06, + "loss": 0.1623, + "step": 36323 + }, + { + "epoch": 13.41, + "learning_rate": 1.65966252001478e-06, + "loss": 0.3201, + "step": 36324 + }, + { + "epoch": 13.41, + "learning_rate": 1.6596255696514349e-06, + "loss": 0.3749, + "step": 36325 + }, + { + "epoch": 13.41, + "learning_rate": 1.6595886192880895e-06, + "loss": 0.2377, + "step": 36326 + }, + { + "epoch": 13.41, + "learning_rate": 1.6595516689247446e-06, + "loss": 0.2602, + "step": 36327 + }, + { + "epoch": 13.42, + "learning_rate": 1.6595147185613992e-06, + "loss": 0.2623, + "step": 36328 + }, + { + "epoch": 13.42, + "learning_rate": 1.659477768198054e-06, + "loss": 0.3348, + "step": 36329 + }, + { + "epoch": 13.42, + "learning_rate": 1.6594408178347087e-06, + "loss": 0.2406, + "step": 36330 + }, + { + "epoch": 13.42, + "learning_rate": 1.6594038674713636e-06, + "loss": 0.3662, + "step": 36331 + }, + { + "epoch": 13.42, + "learning_rate": 1.6593669171080182e-06, + "loss": 0.2853, + "step": 36332 + }, + { + "epoch": 13.42, + "learning_rate": 1.6593299667446733e-06, + "loss": 0.2103, + "step": 36333 + }, + { + "epoch": 13.42, + "learning_rate": 1.6592930163813279e-06, + "loss": 0.27, + "step": 36334 + }, + { + "epoch": 13.42, + "learning_rate": 1.6592560660179823e-06, + "loss": 0.3279, + "step": 36335 + }, + { + "epoch": 13.42, + "learning_rate": 1.6592191156546374e-06, + "loss": 0.2169, + "step": 36336 + }, + { + "epoch": 13.42, + "learning_rate": 1.659182165291292e-06, + "loss": 0.4054, + "step": 36337 + }, + { + "epoch": 13.42, + "learning_rate": 1.6591452149279469e-06, + "loss": 0.3095, + "step": 36338 + }, + { + "epoch": 13.42, + "learning_rate": 1.6591082645646015e-06, + "loss": 0.2482, + "step": 36339 + }, + { + "epoch": 13.42, + "learning_rate": 1.6590713142012566e-06, + "loss": 0.279, + "step": 36340 + }, + { + "epoch": 13.42, + "learning_rate": 1.659034363837911e-06, + "loss": 0.2906, + "step": 36341 + }, + { + "epoch": 13.42, + "learning_rate": 1.658997413474566e-06, + "loss": 0.2327, + "step": 36342 + }, + { + "epoch": 13.42, + "learning_rate": 1.6589604631112207e-06, + "loss": 0.3028, + "step": 36343 + }, + { + "epoch": 13.42, + "learning_rate": 1.6589235127478753e-06, + "loss": 0.2447, + "step": 36344 + }, + { + "epoch": 13.42, + "learning_rate": 1.6588865623845302e-06, + "loss": 0.2245, + "step": 36345 + }, + { + "epoch": 13.42, + "learning_rate": 1.6588496120211848e-06, + "loss": 0.253, + "step": 36346 + }, + { + "epoch": 13.42, + "learning_rate": 1.6588126616578396e-06, + "loss": 0.285, + "step": 36347 + }, + { + "epoch": 13.42, + "learning_rate": 1.6587757112944943e-06, + "loss": 0.3075, + "step": 36348 + }, + { + "epoch": 13.42, + "learning_rate": 1.6587387609311493e-06, + "loss": 0.3288, + "step": 36349 + }, + { + "epoch": 13.42, + "learning_rate": 1.658701810567804e-06, + "loss": 0.2312, + "step": 36350 + }, + { + "epoch": 13.42, + "learning_rate": 1.6586648602044588e-06, + "loss": 0.3907, + "step": 36351 + }, + { + "epoch": 13.42, + "learning_rate": 1.6586279098411135e-06, + "loss": 0.4333, + "step": 36352 + }, + { + "epoch": 13.42, + "learning_rate": 1.6585909594777683e-06, + "loss": 0.1317, + "step": 36353 + }, + { + "epoch": 13.42, + "learning_rate": 1.658554009114423e-06, + "loss": 0.2825, + "step": 36354 + }, + { + "epoch": 13.43, + "learning_rate": 1.6585170587510776e-06, + "loss": 0.2705, + "step": 36355 + }, + { + "epoch": 13.43, + "learning_rate": 1.6584801083877326e-06, + "loss": 0.2836, + "step": 36356 + }, + { + "epoch": 13.43, + "learning_rate": 1.658443158024387e-06, + "loss": 0.2816, + "step": 36357 + }, + { + "epoch": 13.43, + "learning_rate": 1.6584062076610421e-06, + "loss": 0.3249, + "step": 36358 + }, + { + "epoch": 13.43, + "learning_rate": 1.6583692572976968e-06, + "loss": 0.1881, + "step": 36359 + }, + { + "epoch": 13.43, + "learning_rate": 1.6583323069343516e-06, + "loss": 0.2254, + "step": 36360 + }, + { + "epoch": 13.43, + "learning_rate": 1.6582953565710063e-06, + "loss": 0.207, + "step": 36361 + }, + { + "epoch": 13.43, + "learning_rate": 1.6582584062076613e-06, + "loss": 0.3646, + "step": 36362 + }, + { + "epoch": 13.43, + "learning_rate": 1.6582214558443157e-06, + "loss": 0.2555, + "step": 36363 + }, + { + "epoch": 13.43, + "learning_rate": 1.6581845054809704e-06, + "loss": 0.3265, + "step": 36364 + }, + { + "epoch": 13.43, + "learning_rate": 1.6581475551176254e-06, + "loss": 0.2316, + "step": 36365 + }, + { + "epoch": 13.43, + "learning_rate": 1.65811060475428e-06, + "loss": 0.3964, + "step": 36366 + }, + { + "epoch": 13.43, + "learning_rate": 1.658073654390935e-06, + "loss": 0.1717, + "step": 36367 + }, + { + "epoch": 13.43, + "learning_rate": 1.6580367040275896e-06, + "loss": 0.1994, + "step": 36368 + }, + { + "epoch": 13.43, + "learning_rate": 1.6579997536642444e-06, + "loss": 0.2725, + "step": 36369 + }, + { + "epoch": 13.43, + "learning_rate": 1.657962803300899e-06, + "loss": 0.2973, + "step": 36370 + }, + { + "epoch": 13.43, + "learning_rate": 1.657925852937554e-06, + "loss": 0.2453, + "step": 36371 + }, + { + "epoch": 13.43, + "learning_rate": 1.6578889025742087e-06, + "loss": 0.2761, + "step": 36372 + }, + { + "epoch": 13.43, + "learning_rate": 1.6578519522108636e-06, + "loss": 0.3069, + "step": 36373 + }, + { + "epoch": 13.43, + "learning_rate": 1.6578150018475182e-06, + "loss": 0.1591, + "step": 36374 + }, + { + "epoch": 13.43, + "learning_rate": 1.6577780514841729e-06, + "loss": 0.1825, + "step": 36375 + }, + { + "epoch": 13.43, + "learning_rate": 1.6577411011208277e-06, + "loss": 0.327, + "step": 36376 + }, + { + "epoch": 13.43, + "learning_rate": 1.6577041507574823e-06, + "loss": 0.2174, + "step": 36377 + }, + { + "epoch": 13.43, + "learning_rate": 1.6576672003941374e-06, + "loss": 0.2452, + "step": 36378 + }, + { + "epoch": 13.43, + "learning_rate": 1.657630250030792e-06, + "loss": 0.311, + "step": 36379 + }, + { + "epoch": 13.43, + "learning_rate": 1.6575932996674469e-06, + "loss": 0.1955, + "step": 36380 + }, + { + "epoch": 13.43, + "learning_rate": 1.6575563493041015e-06, + "loss": 0.2184, + "step": 36381 + }, + { + "epoch": 13.44, + "learning_rate": 1.6575193989407564e-06, + "loss": 0.3242, + "step": 36382 + }, + { + "epoch": 13.44, + "learning_rate": 1.657482448577411e-06, + "loss": 0.2009, + "step": 36383 + }, + { + "epoch": 13.44, + "learning_rate": 1.6574454982140656e-06, + "loss": 0.1757, + "step": 36384 + }, + { + "epoch": 13.44, + "learning_rate": 1.6574085478507207e-06, + "loss": 0.3327, + "step": 36385 + }, + { + "epoch": 13.44, + "learning_rate": 1.6573715974873751e-06, + "loss": 0.3076, + "step": 36386 + }, + { + "epoch": 13.44, + "learning_rate": 1.6573346471240302e-06, + "loss": 0.2478, + "step": 36387 + }, + { + "epoch": 13.44, + "learning_rate": 1.6572976967606848e-06, + "loss": 0.3116, + "step": 36388 + }, + { + "epoch": 13.44, + "learning_rate": 1.6572607463973397e-06, + "loss": 0.353, + "step": 36389 + }, + { + "epoch": 13.44, + "learning_rate": 1.6572237960339943e-06, + "loss": 0.3615, + "step": 36390 + }, + { + "epoch": 13.44, + "learning_rate": 1.6571868456706494e-06, + "loss": 0.1706, + "step": 36391 + }, + { + "epoch": 13.44, + "learning_rate": 1.6571498953073038e-06, + "loss": 0.3081, + "step": 36392 + }, + { + "epoch": 13.44, + "learning_rate": 1.6571129449439589e-06, + "loss": 0.4519, + "step": 36393 + }, + { + "epoch": 13.44, + "learning_rate": 1.6570759945806135e-06, + "loss": 0.2166, + "step": 36394 + }, + { + "epoch": 13.44, + "learning_rate": 1.6570390442172681e-06, + "loss": 0.4439, + "step": 36395 + }, + { + "epoch": 13.44, + "learning_rate": 1.657002093853923e-06, + "loss": 0.2969, + "step": 36396 + }, + { + "epoch": 13.44, + "learning_rate": 1.6569651434905776e-06, + "loss": 0.2355, + "step": 36397 + }, + { + "epoch": 13.44, + "learning_rate": 1.6569281931272325e-06, + "loss": 0.3696, + "step": 36398 + }, + { + "epoch": 13.44, + "learning_rate": 1.656891242763887e-06, + "loss": 0.2177, + "step": 36399 + }, + { + "epoch": 13.44, + "learning_rate": 1.6568542924005422e-06, + "loss": 0.1652, + "step": 36400 + }, + { + "epoch": 13.44, + "learning_rate": 1.6568173420371968e-06, + "loss": 0.3765, + "step": 36401 + }, + { + "epoch": 13.44, + "learning_rate": 1.6567803916738516e-06, + "loss": 0.423, + "step": 36402 + }, + { + "epoch": 13.44, + "learning_rate": 1.6567434413105063e-06, + "loss": 0.2966, + "step": 36403 + }, + { + "epoch": 13.44, + "learning_rate": 1.656706490947161e-06, + "loss": 0.2828, + "step": 36404 + }, + { + "epoch": 13.44, + "learning_rate": 1.6566695405838158e-06, + "loss": 0.2246, + "step": 36405 + }, + { + "epoch": 13.44, + "learning_rate": 1.6566325902204704e-06, + "loss": 0.3767, + "step": 36406 + }, + { + "epoch": 13.44, + "learning_rate": 1.6565956398571255e-06, + "loss": 0.343, + "step": 36407 + }, + { + "epoch": 13.44, + "learning_rate": 1.6565586894937799e-06, + "loss": 0.2475, + "step": 36408 + }, + { + "epoch": 13.44, + "learning_rate": 1.656521739130435e-06, + "loss": 0.336, + "step": 36409 + }, + { + "epoch": 13.45, + "learning_rate": 1.6564847887670896e-06, + "loss": 0.2384, + "step": 36410 + }, + { + "epoch": 13.45, + "learning_rate": 1.6564478384037444e-06, + "loss": 0.2325, + "step": 36411 + }, + { + "epoch": 13.45, + "learning_rate": 1.656410888040399e-06, + "loss": 0.2433, + "step": 36412 + }, + { + "epoch": 13.45, + "learning_rate": 1.6563739376770541e-06, + "loss": 0.2393, + "step": 36413 + }, + { + "epoch": 13.45, + "learning_rate": 1.6563369873137085e-06, + "loss": 0.2779, + "step": 36414 + }, + { + "epoch": 13.45, + "learning_rate": 1.6563000369503632e-06, + "loss": 0.3243, + "step": 36415 + }, + { + "epoch": 13.45, + "learning_rate": 1.6562630865870182e-06, + "loss": 0.3652, + "step": 36416 + }, + { + "epoch": 13.45, + "learning_rate": 1.6562261362236729e-06, + "loss": 0.4089, + "step": 36417 + }, + { + "epoch": 13.45, + "learning_rate": 1.6561891858603277e-06, + "loss": 0.2063, + "step": 36418 + }, + { + "epoch": 13.45, + "learning_rate": 1.6561522354969824e-06, + "loss": 0.2663, + "step": 36419 + }, + { + "epoch": 13.45, + "learning_rate": 1.6561152851336372e-06, + "loss": 0.3887, + "step": 36420 + }, + { + "epoch": 13.45, + "learning_rate": 1.6560783347702919e-06, + "loss": 0.3385, + "step": 36421 + }, + { + "epoch": 13.45, + "learning_rate": 1.656041384406947e-06, + "loss": 0.2923, + "step": 36422 + }, + { + "epoch": 13.45, + "learning_rate": 1.6560044340436015e-06, + "loss": 0.2491, + "step": 36423 + }, + { + "epoch": 13.45, + "learning_rate": 1.6559674836802562e-06, + "loss": 0.3604, + "step": 36424 + }, + { + "epoch": 13.45, + "learning_rate": 1.655930533316911e-06, + "loss": 0.3049, + "step": 36425 + }, + { + "epoch": 13.45, + "learning_rate": 1.6558935829535657e-06, + "loss": 0.307, + "step": 36426 + }, + { + "epoch": 13.45, + "learning_rate": 1.6558566325902205e-06, + "loss": 0.3301, + "step": 36427 + }, + { + "epoch": 13.45, + "learning_rate": 1.6558196822268752e-06, + "loss": 0.34, + "step": 36428 + }, + { + "epoch": 13.45, + "learning_rate": 1.6557827318635302e-06, + "loss": 0.2879, + "step": 36429 + }, + { + "epoch": 13.45, + "learning_rate": 1.6557457815001846e-06, + "loss": 0.2182, + "step": 36430 + }, + { + "epoch": 13.45, + "learning_rate": 1.6557088311368397e-06, + "loss": 0.2667, + "step": 36431 + }, + { + "epoch": 13.45, + "learning_rate": 1.6556718807734943e-06, + "loss": 0.3849, + "step": 36432 + }, + { + "epoch": 13.45, + "learning_rate": 1.6556349304101492e-06, + "loss": 0.2834, + "step": 36433 + }, + { + "epoch": 13.45, + "learning_rate": 1.6555979800468038e-06, + "loss": 0.3135, + "step": 36434 + }, + { + "epoch": 13.45, + "learning_rate": 1.6555610296834585e-06, + "loss": 0.3096, + "step": 36435 + }, + { + "epoch": 13.45, + "learning_rate": 1.6555240793201133e-06, + "loss": 0.3035, + "step": 36436 + }, + { + "epoch": 13.46, + "learning_rate": 1.655487128956768e-06, + "loss": 0.1872, + "step": 36437 + }, + { + "epoch": 13.46, + "learning_rate": 1.655450178593423e-06, + "loss": 0.3945, + "step": 36438 + }, + { + "epoch": 13.46, + "learning_rate": 1.6554132282300776e-06, + "loss": 0.3384, + "step": 36439 + }, + { + "epoch": 13.46, + "learning_rate": 1.6553762778667325e-06, + "loss": 0.277, + "step": 36440 + }, + { + "epoch": 13.46, + "learning_rate": 1.6553393275033871e-06, + "loss": 0.2689, + "step": 36441 + }, + { + "epoch": 13.46, + "learning_rate": 1.655302377140042e-06, + "loss": 0.2419, + "step": 36442 + }, + { + "epoch": 13.46, + "learning_rate": 1.6552654267766966e-06, + "loss": 0.368, + "step": 36443 + }, + { + "epoch": 13.46, + "learning_rate": 1.6552284764133512e-06, + "loss": 0.3285, + "step": 36444 + }, + { + "epoch": 13.46, + "learning_rate": 1.6551915260500063e-06, + "loss": 0.2697, + "step": 36445 + }, + { + "epoch": 13.46, + "learning_rate": 1.655154575686661e-06, + "loss": 0.2014, + "step": 36446 + }, + { + "epoch": 13.46, + "learning_rate": 1.6551176253233158e-06, + "loss": 0.3164, + "step": 36447 + }, + { + "epoch": 13.46, + "learning_rate": 1.6550806749599704e-06, + "loss": 0.3079, + "step": 36448 + }, + { + "epoch": 13.46, + "learning_rate": 1.6550437245966253e-06, + "loss": 0.3432, + "step": 36449 + }, + { + "epoch": 13.46, + "learning_rate": 1.65500677423328e-06, + "loss": 0.2425, + "step": 36450 + }, + { + "epoch": 13.46, + "learning_rate": 1.654969823869935e-06, + "loss": 0.2674, + "step": 36451 + }, + { + "epoch": 13.46, + "learning_rate": 1.6549328735065896e-06, + "loss": 0.2661, + "step": 36452 + }, + { + "epoch": 13.46, + "learning_rate": 1.6548959231432445e-06, + "loss": 0.3086, + "step": 36453 + }, + { + "epoch": 13.46, + "learning_rate": 1.654858972779899e-06, + "loss": 0.2558, + "step": 36454 + }, + { + "epoch": 13.46, + "learning_rate": 1.6548220224165537e-06, + "loss": 0.1762, + "step": 36455 + }, + { + "epoch": 13.46, + "learning_rate": 1.6547850720532086e-06, + "loss": 0.3315, + "step": 36456 + }, + { + "epoch": 13.46, + "learning_rate": 1.6547481216898632e-06, + "loss": 0.3216, + "step": 36457 + }, + { + "epoch": 13.46, + "learning_rate": 1.6547111713265183e-06, + "loss": 0.1985, + "step": 36458 + }, + { + "epoch": 13.46, + "learning_rate": 1.6546742209631727e-06, + "loss": 0.3987, + "step": 36459 + }, + { + "epoch": 13.46, + "learning_rate": 1.6546372705998278e-06, + "loss": 0.2711, + "step": 36460 + }, + { + "epoch": 13.46, + "learning_rate": 1.6546003202364824e-06, + "loss": 0.3092, + "step": 36461 + }, + { + "epoch": 13.46, + "learning_rate": 1.6545633698731372e-06, + "loss": 0.207, + "step": 36462 + }, + { + "epoch": 13.46, + "learning_rate": 1.6545264195097919e-06, + "loss": 0.3013, + "step": 36463 + }, + { + "epoch": 13.47, + "learning_rate": 1.6544894691464465e-06, + "loss": 0.2598, + "step": 36464 + }, + { + "epoch": 13.47, + "learning_rate": 1.6544525187831014e-06, + "loss": 0.1523, + "step": 36465 + }, + { + "epoch": 13.47, + "learning_rate": 1.654415568419756e-06, + "loss": 0.3941, + "step": 36466 + }, + { + "epoch": 13.47, + "learning_rate": 1.654378618056411e-06, + "loss": 0.3905, + "step": 36467 + }, + { + "epoch": 13.47, + "learning_rate": 1.6543416676930657e-06, + "loss": 0.2047, + "step": 36468 + }, + { + "epoch": 13.47, + "learning_rate": 1.6543047173297205e-06, + "loss": 0.2381, + "step": 36469 + }, + { + "epoch": 13.47, + "learning_rate": 1.6542677669663752e-06, + "loss": 0.2589, + "step": 36470 + }, + { + "epoch": 13.47, + "learning_rate": 1.65423081660303e-06, + "loss": 0.3333, + "step": 36471 + }, + { + "epoch": 13.47, + "learning_rate": 1.6541938662396847e-06, + "loss": 0.2435, + "step": 36472 + }, + { + "epoch": 13.47, + "learning_rate": 1.6541569158763393e-06, + "loss": 0.2493, + "step": 36473 + }, + { + "epoch": 13.47, + "learning_rate": 1.6541199655129944e-06, + "loss": 0.3128, + "step": 36474 + }, + { + "epoch": 13.47, + "learning_rate": 1.6540830151496488e-06, + "loss": 0.4637, + "step": 36475 + }, + { + "epoch": 13.47, + "learning_rate": 1.6540460647863038e-06, + "loss": 0.2545, + "step": 36476 + }, + { + "epoch": 13.47, + "learning_rate": 1.6540091144229585e-06, + "loss": 0.1333, + "step": 36477 + }, + { + "epoch": 13.47, + "learning_rate": 1.6539721640596133e-06, + "loss": 0.2809, + "step": 36478 + }, + { + "epoch": 13.47, + "learning_rate": 1.653935213696268e-06, + "loss": 0.2843, + "step": 36479 + }, + { + "epoch": 13.47, + "learning_rate": 1.653898263332923e-06, + "loss": 0.201, + "step": 36480 + }, + { + "epoch": 13.47, + "learning_rate": 1.6538613129695774e-06, + "loss": 0.3587, + "step": 36481 + }, + { + "epoch": 13.47, + "learning_rate": 1.6538243626062325e-06, + "loss": 0.3754, + "step": 36482 + }, + { + "epoch": 13.47, + "learning_rate": 1.6537874122428871e-06, + "loss": 0.4401, + "step": 36483 + }, + { + "epoch": 13.47, + "learning_rate": 1.6537504618795418e-06, + "loss": 0.3275, + "step": 36484 + }, + { + "epoch": 13.47, + "learning_rate": 1.6537135115161966e-06, + "loss": 0.2505, + "step": 36485 + }, + { + "epoch": 13.47, + "learning_rate": 1.6536765611528513e-06, + "loss": 0.3508, + "step": 36486 + }, + { + "epoch": 13.47, + "learning_rate": 1.6536396107895061e-06, + "loss": 0.2699, + "step": 36487 + }, + { + "epoch": 13.47, + "learning_rate": 1.6536026604261607e-06, + "loss": 0.3068, + "step": 36488 + }, + { + "epoch": 13.47, + "learning_rate": 1.6535657100628158e-06, + "loss": 0.1819, + "step": 36489 + }, + { + "epoch": 13.47, + "learning_rate": 1.6535287596994704e-06, + "loss": 0.3825, + "step": 36490 + }, + { + "epoch": 13.48, + "learning_rate": 1.6534918093361253e-06, + "loss": 0.2021, + "step": 36491 + }, + { + "epoch": 13.48, + "learning_rate": 1.65345485897278e-06, + "loss": 0.4256, + "step": 36492 + }, + { + "epoch": 13.48, + "learning_rate": 1.6534179086094346e-06, + "loss": 0.2276, + "step": 36493 + }, + { + "epoch": 13.48, + "learning_rate": 1.6533809582460894e-06, + "loss": 0.4192, + "step": 36494 + }, + { + "epoch": 13.48, + "learning_rate": 1.653344007882744e-06, + "loss": 0.4158, + "step": 36495 + }, + { + "epoch": 13.48, + "learning_rate": 1.6533070575193991e-06, + "loss": 0.2683, + "step": 36496 + }, + { + "epoch": 13.48, + "learning_rate": 1.6532701071560537e-06, + "loss": 0.2781, + "step": 36497 + }, + { + "epoch": 13.48, + "learning_rate": 1.6532331567927086e-06, + "loss": 0.2203, + "step": 36498 + }, + { + "epoch": 13.48, + "learning_rate": 1.6531962064293632e-06, + "loss": 0.3268, + "step": 36499 + }, + { + "epoch": 13.48, + "learning_rate": 1.653159256066018e-06, + "loss": 0.2833, + "step": 36500 + }, + { + "epoch": 13.48, + "learning_rate": 1.6531223057026727e-06, + "loss": 0.2091, + "step": 36501 + }, + { + "epoch": 13.48, + "learning_rate": 1.6530853553393278e-06, + "loss": 0.3525, + "step": 36502 + }, + { + "epoch": 13.48, + "learning_rate": 1.6530484049759824e-06, + "loss": 0.3362, + "step": 36503 + }, + { + "epoch": 13.48, + "learning_rate": 1.6530114546126368e-06, + "loss": 0.327, + "step": 36504 + }, + { + "epoch": 13.48, + "learning_rate": 1.652974504249292e-06, + "loss": 0.4172, + "step": 36505 + }, + { + "epoch": 13.48, + "learning_rate": 1.6529375538859465e-06, + "loss": 0.2283, + "step": 36506 + }, + { + "epoch": 13.48, + "learning_rate": 1.6529006035226014e-06, + "loss": 0.1994, + "step": 36507 + }, + { + "epoch": 13.48, + "learning_rate": 1.652863653159256e-06, + "loss": 0.2595, + "step": 36508 + }, + { + "epoch": 13.48, + "learning_rate": 1.6528267027959109e-06, + "loss": 0.3513, + "step": 36509 + }, + { + "epoch": 13.48, + "learning_rate": 1.6527897524325655e-06, + "loss": 0.3126, + "step": 36510 + }, + { + "epoch": 13.48, + "learning_rate": 1.6527528020692206e-06, + "loss": 0.3409, + "step": 36511 + }, + { + "epoch": 13.48, + "learning_rate": 1.6527158517058752e-06, + "loss": 0.3267, + "step": 36512 + }, + { + "epoch": 13.48, + "learning_rate": 1.6526789013425298e-06, + "loss": 0.1927, + "step": 36513 + }, + { + "epoch": 13.48, + "learning_rate": 1.6526419509791847e-06, + "loss": 0.1927, + "step": 36514 + }, + { + "epoch": 13.48, + "learning_rate": 1.6526050006158393e-06, + "loss": 0.2393, + "step": 36515 + }, + { + "epoch": 13.48, + "learning_rate": 1.6525680502524942e-06, + "loss": 0.2389, + "step": 36516 + }, + { + "epoch": 13.48, + "learning_rate": 1.6525310998891488e-06, + "loss": 0.1907, + "step": 36517 + }, + { + "epoch": 13.49, + "learning_rate": 1.6524941495258039e-06, + "loss": 0.1899, + "step": 36518 + }, + { + "epoch": 13.49, + "learning_rate": 1.6524571991624585e-06, + "loss": 0.3081, + "step": 36519 + }, + { + "epoch": 13.49, + "learning_rate": 1.6524202487991134e-06, + "loss": 0.3525, + "step": 36520 + }, + { + "epoch": 13.49, + "learning_rate": 1.652383298435768e-06, + "loss": 0.3092, + "step": 36521 + }, + { + "epoch": 13.49, + "learning_rate": 1.6523463480724228e-06, + "loss": 0.2595, + "step": 36522 + }, + { + "epoch": 13.49, + "learning_rate": 1.6523093977090775e-06, + "loss": 0.1993, + "step": 36523 + }, + { + "epoch": 13.49, + "learning_rate": 1.6522724473457321e-06, + "loss": 0.2254, + "step": 36524 + }, + { + "epoch": 13.49, + "learning_rate": 1.6522354969823872e-06, + "loss": 0.2953, + "step": 36525 + }, + { + "epoch": 13.49, + "learning_rate": 1.6521985466190416e-06, + "loss": 0.2633, + "step": 36526 + }, + { + "epoch": 13.49, + "learning_rate": 1.6521615962556967e-06, + "loss": 0.2277, + "step": 36527 + }, + { + "epoch": 13.49, + "learning_rate": 1.6521246458923513e-06, + "loss": 0.3171, + "step": 36528 + }, + { + "epoch": 13.49, + "learning_rate": 1.6520876955290061e-06, + "loss": 0.2886, + "step": 36529 + }, + { + "epoch": 13.49, + "learning_rate": 1.6520507451656608e-06, + "loss": 0.2327, + "step": 36530 + }, + { + "epoch": 13.49, + "learning_rate": 1.6520137948023158e-06, + "loss": 0.2488, + "step": 36531 + }, + { + "epoch": 13.49, + "learning_rate": 1.6519768444389703e-06, + "loss": 0.1461, + "step": 36532 + }, + { + "epoch": 13.49, + "learning_rate": 1.651939894075625e-06, + "loss": 0.3552, + "step": 36533 + }, + { + "epoch": 13.49, + "learning_rate": 1.65190294371228e-06, + "loss": 0.3053, + "step": 36534 + }, + { + "epoch": 13.49, + "learning_rate": 1.6518659933489346e-06, + "loss": 0.2639, + "step": 36535 + }, + { + "epoch": 13.49, + "learning_rate": 1.6518290429855894e-06, + "loss": 0.2239, + "step": 36536 + }, + { + "epoch": 13.49, + "learning_rate": 1.651792092622244e-06, + "loss": 0.3371, + "step": 36537 + }, + { + "epoch": 13.49, + "learning_rate": 1.651755142258899e-06, + "loss": 0.3294, + "step": 36538 + }, + { + "epoch": 13.49, + "learning_rate": 1.6517181918955536e-06, + "loss": 0.3351, + "step": 36539 + }, + { + "epoch": 13.49, + "learning_rate": 1.6516812415322086e-06, + "loss": 0.3297, + "step": 36540 + }, + { + "epoch": 13.49, + "learning_rate": 1.6516442911688633e-06, + "loss": 0.1858, + "step": 36541 + }, + { + "epoch": 13.49, + "learning_rate": 1.651607340805518e-06, + "loss": 0.3565, + "step": 36542 + }, + { + "epoch": 13.49, + "learning_rate": 1.6515703904421727e-06, + "loss": 0.2287, + "step": 36543 + }, + { + "epoch": 13.49, + "learning_rate": 1.6515334400788274e-06, + "loss": 0.3937, + "step": 36544 + }, + { + "epoch": 13.5, + "learning_rate": 1.6514964897154822e-06, + "loss": 0.1761, + "step": 36545 + }, + { + "epoch": 13.5, + "learning_rate": 1.6514595393521369e-06, + "loss": 0.3059, + "step": 36546 + }, + { + "epoch": 13.5, + "learning_rate": 1.651422588988792e-06, + "loss": 0.3802, + "step": 36547 + }, + { + "epoch": 13.5, + "learning_rate": 1.6513856386254463e-06, + "loss": 0.1977, + "step": 36548 + }, + { + "epoch": 13.5, + "learning_rate": 1.6513486882621014e-06, + "loss": 0.2917, + "step": 36549 + }, + { + "epoch": 13.5, + "learning_rate": 1.651311737898756e-06, + "loss": 0.204, + "step": 36550 + }, + { + "epoch": 13.5, + "learning_rate": 1.6512747875354109e-06, + "loss": 0.3382, + "step": 36551 + }, + { + "epoch": 13.5, + "learning_rate": 1.6512378371720655e-06, + "loss": 0.4207, + "step": 36552 + }, + { + "epoch": 13.5, + "learning_rate": 1.6512008868087202e-06, + "loss": 0.2634, + "step": 36553 + }, + { + "epoch": 13.5, + "learning_rate": 1.651163936445375e-06, + "loss": 0.2928, + "step": 36554 + }, + { + "epoch": 13.5, + "learning_rate": 1.6511269860820296e-06, + "loss": 0.3042, + "step": 36555 + }, + { + "epoch": 13.5, + "learning_rate": 1.6510900357186847e-06, + "loss": 0.246, + "step": 36556 + }, + { + "epoch": 13.5, + "learning_rate": 1.6510530853553393e-06, + "loss": 0.3387, + "step": 36557 + }, + { + "epoch": 13.5, + "learning_rate": 1.6510161349919942e-06, + "loss": 0.3135, + "step": 36558 + }, + { + "epoch": 13.5, + "learning_rate": 1.6509791846286488e-06, + "loss": 0.3676, + "step": 36559 + }, + { + "epoch": 13.5, + "learning_rate": 1.6509422342653037e-06, + "loss": 0.4163, + "step": 36560 + }, + { + "epoch": 13.5, + "learning_rate": 1.6509052839019583e-06, + "loss": 0.3022, + "step": 36561 + }, + { + "epoch": 13.5, + "learning_rate": 1.6508683335386134e-06, + "loss": 0.2772, + "step": 36562 + }, + { + "epoch": 13.5, + "learning_rate": 1.650831383175268e-06, + "loss": 0.3075, + "step": 36563 + }, + { + "epoch": 13.5, + "learning_rate": 1.6507944328119226e-06, + "loss": 0.3141, + "step": 36564 + }, + { + "epoch": 13.5, + "learning_rate": 1.6507574824485775e-06, + "loss": 0.273, + "step": 36565 + }, + { + "epoch": 13.5, + "learning_rate": 1.6507205320852321e-06, + "loss": 0.1665, + "step": 36566 + }, + { + "epoch": 13.5, + "learning_rate": 1.650683581721887e-06, + "loss": 0.3122, + "step": 36567 + }, + { + "epoch": 13.5, + "learning_rate": 1.6506466313585416e-06, + "loss": 0.2639, + "step": 36568 + }, + { + "epoch": 13.5, + "learning_rate": 1.6506096809951967e-06, + "loss": 0.4422, + "step": 36569 + }, + { + "epoch": 13.5, + "learning_rate": 1.6505727306318513e-06, + "loss": 0.2816, + "step": 36570 + }, + { + "epoch": 13.5, + "learning_rate": 1.6505357802685062e-06, + "loss": 0.2222, + "step": 36571 + }, + { + "epoch": 13.51, + "learning_rate": 1.6504988299051608e-06, + "loss": 0.3218, + "step": 36572 + }, + { + "epoch": 13.51, + "learning_rate": 1.6504618795418154e-06, + "loss": 0.4825, + "step": 36573 + }, + { + "epoch": 13.51, + "learning_rate": 1.6504249291784703e-06, + "loss": 0.481, + "step": 36574 + }, + { + "epoch": 13.51, + "learning_rate": 1.650387978815125e-06, + "loss": 0.3419, + "step": 36575 + }, + { + "epoch": 13.51, + "learning_rate": 1.65035102845178e-06, + "loss": 0.1885, + "step": 36576 + }, + { + "epoch": 13.51, + "learning_rate": 1.6503140780884344e-06, + "loss": 0.3857, + "step": 36577 + }, + { + "epoch": 13.51, + "learning_rate": 1.6502771277250895e-06, + "loss": 0.197, + "step": 36578 + }, + { + "epoch": 13.51, + "learning_rate": 1.650240177361744e-06, + "loss": 0.3094, + "step": 36579 + }, + { + "epoch": 13.51, + "learning_rate": 1.650203226998399e-06, + "loss": 0.2443, + "step": 36580 + }, + { + "epoch": 13.51, + "learning_rate": 1.6501662766350536e-06, + "loss": 0.2188, + "step": 36581 + }, + { + "epoch": 13.51, + "learning_rate": 1.6501293262717084e-06, + "loss": 0.3048, + "step": 36582 + }, + { + "epoch": 13.51, + "learning_rate": 1.650092375908363e-06, + "loss": 0.2643, + "step": 36583 + }, + { + "epoch": 13.51, + "learning_rate": 1.6500554255450177e-06, + "loss": 0.4249, + "step": 36584 + }, + { + "epoch": 13.51, + "learning_rate": 1.6500184751816728e-06, + "loss": 0.3349, + "step": 36585 + }, + { + "epoch": 13.51, + "learning_rate": 1.6499815248183274e-06, + "loss": 0.2976, + "step": 36586 + }, + { + "epoch": 13.51, + "learning_rate": 1.6499445744549823e-06, + "loss": 0.2267, + "step": 36587 + }, + { + "epoch": 13.51, + "learning_rate": 1.6499076240916369e-06, + "loss": 0.2514, + "step": 36588 + }, + { + "epoch": 13.51, + "learning_rate": 1.6498706737282917e-06, + "loss": 0.2187, + "step": 36589 + }, + { + "epoch": 13.51, + "learning_rate": 1.6498337233649464e-06, + "loss": 0.2153, + "step": 36590 + }, + { + "epoch": 13.51, + "learning_rate": 1.6497967730016014e-06, + "loss": 0.2122, + "step": 36591 + }, + { + "epoch": 13.51, + "learning_rate": 1.649759822638256e-06, + "loss": 0.3835, + "step": 36592 + }, + { + "epoch": 13.51, + "learning_rate": 1.6497228722749105e-06, + "loss": 0.2955, + "step": 36593 + }, + { + "epoch": 13.51, + "learning_rate": 1.6496859219115656e-06, + "loss": 0.378, + "step": 36594 + }, + { + "epoch": 13.51, + "learning_rate": 1.6496489715482202e-06, + "loss": 0.2075, + "step": 36595 + }, + { + "epoch": 13.51, + "learning_rate": 1.649612021184875e-06, + "loss": 0.2769, + "step": 36596 + }, + { + "epoch": 13.51, + "learning_rate": 1.6495750708215297e-06, + "loss": 0.2824, + "step": 36597 + }, + { + "epoch": 13.51, + "learning_rate": 1.6495381204581847e-06, + "loss": 0.3425, + "step": 36598 + }, + { + "epoch": 13.52, + "learning_rate": 1.6495011700948392e-06, + "loss": 0.3202, + "step": 36599 + }, + { + "epoch": 13.52, + "learning_rate": 1.6494642197314942e-06, + "loss": 0.2787, + "step": 36600 + }, + { + "epoch": 13.52, + "learning_rate": 1.6494272693681489e-06, + "loss": 0.3534, + "step": 36601 + }, + { + "epoch": 13.52, + "learning_rate": 1.6493903190048037e-06, + "loss": 0.1822, + "step": 36602 + }, + { + "epoch": 13.52, + "learning_rate": 1.6493533686414583e-06, + "loss": 0.2626, + "step": 36603 + }, + { + "epoch": 13.52, + "learning_rate": 1.649316418278113e-06, + "loss": 0.3663, + "step": 36604 + }, + { + "epoch": 13.52, + "learning_rate": 1.6492794679147678e-06, + "loss": 0.2354, + "step": 36605 + }, + { + "epoch": 13.52, + "learning_rate": 1.6492425175514225e-06, + "loss": 0.2918, + "step": 36606 + }, + { + "epoch": 13.52, + "learning_rate": 1.6492055671880775e-06, + "loss": 0.3682, + "step": 36607 + }, + { + "epoch": 13.52, + "learning_rate": 1.6491686168247322e-06, + "loss": 0.2623, + "step": 36608 + }, + { + "epoch": 13.52, + "learning_rate": 1.649131666461387e-06, + "loss": 0.2905, + "step": 36609 + }, + { + "epoch": 13.52, + "learning_rate": 1.6490947160980416e-06, + "loss": 0.2917, + "step": 36610 + }, + { + "epoch": 13.52, + "learning_rate": 1.6490577657346965e-06, + "loss": 0.2953, + "step": 36611 + }, + { + "epoch": 13.52, + "learning_rate": 1.6490208153713511e-06, + "loss": 0.3177, + "step": 36612 + }, + { + "epoch": 13.52, + "learning_rate": 1.6489838650080058e-06, + "loss": 0.3612, + "step": 36613 + }, + { + "epoch": 13.52, + "learning_rate": 1.6489469146446608e-06, + "loss": 0.1752, + "step": 36614 + }, + { + "epoch": 13.52, + "learning_rate": 1.6489099642813152e-06, + "loss": 0.3568, + "step": 36615 + }, + { + "epoch": 13.52, + "learning_rate": 1.6488730139179703e-06, + "loss": 0.2899, + "step": 36616 + }, + { + "epoch": 13.52, + "learning_rate": 1.648836063554625e-06, + "loss": 0.2879, + "step": 36617 + }, + { + "epoch": 13.52, + "learning_rate": 1.6487991131912798e-06, + "loss": 0.2324, + "step": 36618 + }, + { + "epoch": 13.52, + "learning_rate": 1.6487621628279344e-06, + "loss": 0.2089, + "step": 36619 + }, + { + "epoch": 13.52, + "learning_rate": 1.6487252124645895e-06, + "loss": 0.3426, + "step": 36620 + }, + { + "epoch": 13.52, + "learning_rate": 1.648688262101244e-06, + "loss": 0.2118, + "step": 36621 + }, + { + "epoch": 13.52, + "learning_rate": 1.648651311737899e-06, + "loss": 0.4574, + "step": 36622 + }, + { + "epoch": 13.52, + "learning_rate": 1.6486143613745536e-06, + "loss": 0.1622, + "step": 36623 + }, + { + "epoch": 13.52, + "learning_rate": 1.6485774110112082e-06, + "loss": 0.3971, + "step": 36624 + }, + { + "epoch": 13.52, + "learning_rate": 1.648540460647863e-06, + "loss": 0.2475, + "step": 36625 + }, + { + "epoch": 13.53, + "learning_rate": 1.6485035102845177e-06, + "loss": 0.2493, + "step": 36626 + }, + { + "epoch": 13.53, + "learning_rate": 1.6484665599211726e-06, + "loss": 0.291, + "step": 36627 + }, + { + "epoch": 13.53, + "learning_rate": 1.6484296095578272e-06, + "loss": 0.175, + "step": 36628 + }, + { + "epoch": 13.53, + "learning_rate": 1.6483926591944823e-06, + "loss": 0.2031, + "step": 36629 + }, + { + "epoch": 13.53, + "learning_rate": 1.648355708831137e-06, + "loss": 0.357, + "step": 36630 + }, + { + "epoch": 13.53, + "learning_rate": 1.6483187584677918e-06, + "loss": 0.2888, + "step": 36631 + }, + { + "epoch": 13.53, + "learning_rate": 1.6482818081044464e-06, + "loss": 0.2525, + "step": 36632 + }, + { + "epoch": 13.53, + "learning_rate": 1.648244857741101e-06, + "loss": 0.2085, + "step": 36633 + }, + { + "epoch": 13.53, + "learning_rate": 1.6482079073777559e-06, + "loss": 0.4076, + "step": 36634 + }, + { + "epoch": 13.53, + "learning_rate": 1.6481709570144105e-06, + "loss": 0.2738, + "step": 36635 + }, + { + "epoch": 13.53, + "learning_rate": 1.6481340066510656e-06, + "loss": 0.243, + "step": 36636 + }, + { + "epoch": 13.53, + "learning_rate": 1.6480970562877202e-06, + "loss": 0.2958, + "step": 36637 + }, + { + "epoch": 13.53, + "learning_rate": 1.648060105924375e-06, + "loss": 0.2853, + "step": 36638 + }, + { + "epoch": 13.53, + "learning_rate": 1.6480231555610297e-06, + "loss": 0.3461, + "step": 36639 + }, + { + "epoch": 13.53, + "learning_rate": 1.6479862051976845e-06, + "loss": 0.31, + "step": 36640 + }, + { + "epoch": 13.53, + "learning_rate": 1.6479492548343392e-06, + "loss": 0.2844, + "step": 36641 + }, + { + "epoch": 13.53, + "learning_rate": 1.6479123044709942e-06, + "loss": 0.1708, + "step": 36642 + }, + { + "epoch": 13.53, + "learning_rate": 1.6478753541076489e-06, + "loss": 0.1781, + "step": 36643 + }, + { + "epoch": 13.53, + "learning_rate": 1.6478384037443033e-06, + "loss": 0.2042, + "step": 36644 + }, + { + "epoch": 13.53, + "learning_rate": 1.6478014533809584e-06, + "loss": 0.1813, + "step": 36645 + }, + { + "epoch": 13.53, + "learning_rate": 1.647764503017613e-06, + "loss": 0.1883, + "step": 36646 + }, + { + "epoch": 13.53, + "learning_rate": 1.6477275526542678e-06, + "loss": 0.251, + "step": 36647 + }, + { + "epoch": 13.53, + "learning_rate": 1.6476906022909225e-06, + "loss": 0.2614, + "step": 36648 + }, + { + "epoch": 13.53, + "learning_rate": 1.6476536519275775e-06, + "loss": 0.2965, + "step": 36649 + }, + { + "epoch": 13.53, + "learning_rate": 1.647616701564232e-06, + "loss": 0.2208, + "step": 36650 + }, + { + "epoch": 13.53, + "learning_rate": 1.647579751200887e-06, + "loss": 0.2343, + "step": 36651 + }, + { + "epoch": 13.53, + "learning_rate": 1.6475428008375417e-06, + "loss": 0.2628, + "step": 36652 + }, + { + "epoch": 13.54, + "learning_rate": 1.6475058504741963e-06, + "loss": 0.2305, + "step": 36653 + }, + { + "epoch": 13.54, + "learning_rate": 1.6474689001108512e-06, + "loss": 0.2824, + "step": 36654 + }, + { + "epoch": 13.54, + "learning_rate": 1.6474319497475058e-06, + "loss": 0.321, + "step": 36655 + }, + { + "epoch": 13.54, + "learning_rate": 1.6473949993841606e-06, + "loss": 0.3285, + "step": 36656 + }, + { + "epoch": 13.54, + "learning_rate": 1.6473580490208153e-06, + "loss": 0.344, + "step": 36657 + }, + { + "epoch": 13.54, + "learning_rate": 1.6473210986574703e-06, + "loss": 0.2608, + "step": 36658 + }, + { + "epoch": 13.54, + "learning_rate": 1.647284148294125e-06, + "loss": 0.2573, + "step": 36659 + }, + { + "epoch": 13.54, + "learning_rate": 1.6472471979307798e-06, + "loss": 0.2172, + "step": 36660 + }, + { + "epoch": 13.54, + "learning_rate": 1.6472102475674345e-06, + "loss": 0.2116, + "step": 36661 + }, + { + "epoch": 13.54, + "learning_rate": 1.6471732972040893e-06, + "loss": 0.2581, + "step": 36662 + }, + { + "epoch": 13.54, + "learning_rate": 1.647136346840744e-06, + "loss": 0.2125, + "step": 36663 + }, + { + "epoch": 13.54, + "learning_rate": 1.6470993964773986e-06, + "loss": 0.3608, + "step": 36664 + }, + { + "epoch": 13.54, + "learning_rate": 1.6470624461140536e-06, + "loss": 0.2454, + "step": 36665 + }, + { + "epoch": 13.54, + "learning_rate": 1.647025495750708e-06, + "loss": 0.2291, + "step": 36666 + }, + { + "epoch": 13.54, + "learning_rate": 1.6469885453873631e-06, + "loss": 0.3715, + "step": 36667 + }, + { + "epoch": 13.54, + "learning_rate": 1.6469515950240178e-06, + "loss": 0.1364, + "step": 36668 + }, + { + "epoch": 13.54, + "learning_rate": 1.6469146446606726e-06, + "loss": 0.2889, + "step": 36669 + }, + { + "epoch": 13.54, + "learning_rate": 1.6468776942973272e-06, + "loss": 0.3735, + "step": 36670 + }, + { + "epoch": 13.54, + "learning_rate": 1.6468407439339823e-06, + "loss": 0.2806, + "step": 36671 + }, + { + "epoch": 13.54, + "learning_rate": 1.6468037935706367e-06, + "loss": 0.2697, + "step": 36672 + }, + { + "epoch": 13.54, + "learning_rate": 1.6467668432072914e-06, + "loss": 0.2511, + "step": 36673 + }, + { + "epoch": 13.54, + "learning_rate": 1.6467298928439464e-06, + "loss": 0.2443, + "step": 36674 + }, + { + "epoch": 13.54, + "learning_rate": 1.646692942480601e-06, + "loss": 0.4203, + "step": 36675 + }, + { + "epoch": 13.54, + "learning_rate": 1.646655992117256e-06, + "loss": 0.2522, + "step": 36676 + }, + { + "epoch": 13.54, + "learning_rate": 1.6466190417539105e-06, + "loss": 0.3425, + "step": 36677 + }, + { + "epoch": 13.54, + "learning_rate": 1.6465820913905654e-06, + "loss": 0.353, + "step": 36678 + }, + { + "epoch": 13.54, + "learning_rate": 1.64654514102722e-06, + "loss": 0.2583, + "step": 36679 + }, + { + "epoch": 13.55, + "learning_rate": 1.646508190663875e-06, + "loss": 0.3011, + "step": 36680 + }, + { + "epoch": 13.55, + "learning_rate": 1.6464712403005297e-06, + "loss": 0.3775, + "step": 36681 + }, + { + "epoch": 13.55, + "learning_rate": 1.6464342899371846e-06, + "loss": 0.1311, + "step": 36682 + }, + { + "epoch": 13.55, + "learning_rate": 1.6463973395738392e-06, + "loss": 0.3183, + "step": 36683 + }, + { + "epoch": 13.55, + "learning_rate": 1.6463603892104938e-06, + "loss": 0.4253, + "step": 36684 + }, + { + "epoch": 13.55, + "learning_rate": 1.6463234388471487e-06, + "loss": 0.2654, + "step": 36685 + }, + { + "epoch": 13.55, + "learning_rate": 1.6462864884838033e-06, + "loss": 0.3286, + "step": 36686 + }, + { + "epoch": 13.55, + "learning_rate": 1.6462495381204584e-06, + "loss": 0.2017, + "step": 36687 + }, + { + "epoch": 13.55, + "learning_rate": 1.646212587757113e-06, + "loss": 0.253, + "step": 36688 + }, + { + "epoch": 13.55, + "learning_rate": 1.6461756373937679e-06, + "loss": 0.1759, + "step": 36689 + }, + { + "epoch": 13.55, + "learning_rate": 1.6461386870304225e-06, + "loss": 0.3473, + "step": 36690 + }, + { + "epoch": 13.55, + "learning_rate": 1.6461017366670774e-06, + "loss": 0.2508, + "step": 36691 + }, + { + "epoch": 13.55, + "learning_rate": 1.646064786303732e-06, + "loss": 0.304, + "step": 36692 + }, + { + "epoch": 13.55, + "learning_rate": 1.6460278359403866e-06, + "loss": 0.2863, + "step": 36693 + }, + { + "epoch": 13.55, + "learning_rate": 1.6459908855770415e-06, + "loss": 0.2602, + "step": 36694 + }, + { + "epoch": 13.55, + "learning_rate": 1.6459539352136961e-06, + "loss": 0.2504, + "step": 36695 + }, + { + "epoch": 13.55, + "learning_rate": 1.6459169848503512e-06, + "loss": 0.2296, + "step": 36696 + }, + { + "epoch": 13.55, + "learning_rate": 1.6458800344870058e-06, + "loss": 0.2115, + "step": 36697 + }, + { + "epoch": 13.55, + "learning_rate": 1.6458430841236607e-06, + "loss": 0.1961, + "step": 36698 + }, + { + "epoch": 13.55, + "learning_rate": 1.6458061337603153e-06, + "loss": 0.2914, + "step": 36699 + }, + { + "epoch": 13.55, + "learning_rate": 1.6457691833969701e-06, + "loss": 0.2466, + "step": 36700 + }, + { + "epoch": 13.55, + "learning_rate": 1.6457322330336248e-06, + "loss": 0.2979, + "step": 36701 + }, + { + "epoch": 13.55, + "learning_rate": 1.6456952826702794e-06, + "loss": 0.1924, + "step": 36702 + }, + { + "epoch": 13.55, + "learning_rate": 1.6456583323069345e-06, + "loss": 0.2389, + "step": 36703 + }, + { + "epoch": 13.55, + "learning_rate": 1.6456213819435891e-06, + "loss": 0.2576, + "step": 36704 + }, + { + "epoch": 13.55, + "learning_rate": 1.645584431580244e-06, + "loss": 0.312, + "step": 36705 + }, + { + "epoch": 13.55, + "learning_rate": 1.6455474812168986e-06, + "loss": 0.2491, + "step": 36706 + }, + { + "epoch": 13.56, + "learning_rate": 1.6455105308535534e-06, + "loss": 0.3472, + "step": 36707 + }, + { + "epoch": 13.56, + "learning_rate": 1.645473580490208e-06, + "loss": 0.2214, + "step": 36708 + }, + { + "epoch": 13.56, + "learning_rate": 1.6454366301268631e-06, + "loss": 0.192, + "step": 36709 + }, + { + "epoch": 13.56, + "learning_rate": 1.6453996797635178e-06, + "loss": 0.2393, + "step": 36710 + }, + { + "epoch": 13.56, + "learning_rate": 1.6453627294001726e-06, + "loss": 0.3259, + "step": 36711 + }, + { + "epoch": 13.56, + "learning_rate": 1.6453257790368273e-06, + "loss": 0.2115, + "step": 36712 + }, + { + "epoch": 13.56, + "learning_rate": 1.645288828673482e-06, + "loss": 0.2305, + "step": 36713 + }, + { + "epoch": 13.56, + "learning_rate": 1.6452518783101367e-06, + "loss": 0.3043, + "step": 36714 + }, + { + "epoch": 13.56, + "learning_rate": 1.6452149279467914e-06, + "loss": 0.2568, + "step": 36715 + }, + { + "epoch": 13.56, + "learning_rate": 1.6451779775834464e-06, + "loss": 0.274, + "step": 36716 + }, + { + "epoch": 13.56, + "learning_rate": 1.6451410272201009e-06, + "loss": 0.4089, + "step": 36717 + }, + { + "epoch": 13.56, + "learning_rate": 1.645104076856756e-06, + "loss": 0.3174, + "step": 36718 + }, + { + "epoch": 13.56, + "learning_rate": 1.6450671264934106e-06, + "loss": 0.3622, + "step": 36719 + }, + { + "epoch": 13.56, + "learning_rate": 1.6450301761300654e-06, + "loss": 0.2245, + "step": 36720 + }, + { + "epoch": 13.56, + "learning_rate": 1.64499322576672e-06, + "loss": 0.2508, + "step": 36721 + }, + { + "epoch": 13.56, + "learning_rate": 1.6449562754033747e-06, + "loss": 0.2551, + "step": 36722 + }, + { + "epoch": 13.56, + "learning_rate": 1.6449193250400295e-06, + "loss": 0.327, + "step": 36723 + }, + { + "epoch": 13.56, + "learning_rate": 1.6448823746766842e-06, + "loss": 0.355, + "step": 36724 + }, + { + "epoch": 13.56, + "learning_rate": 1.6448454243133392e-06, + "loss": 0.3991, + "step": 36725 + }, + { + "epoch": 13.56, + "learning_rate": 1.6448084739499939e-06, + "loss": 0.3128, + "step": 36726 + }, + { + "epoch": 13.56, + "learning_rate": 1.6447715235866487e-06, + "loss": 0.1047, + "step": 36727 + }, + { + "epoch": 13.56, + "learning_rate": 1.6447345732233034e-06, + "loss": 0.2275, + "step": 36728 + }, + { + "epoch": 13.56, + "learning_rate": 1.6446976228599582e-06, + "loss": 0.2302, + "step": 36729 + }, + { + "epoch": 13.56, + "learning_rate": 1.6446606724966128e-06, + "loss": 0.2614, + "step": 36730 + }, + { + "epoch": 13.56, + "learning_rate": 1.644623722133268e-06, + "loss": 0.2096, + "step": 36731 + }, + { + "epoch": 13.56, + "learning_rate": 1.6445867717699225e-06, + "loss": 0.4724, + "step": 36732 + }, + { + "epoch": 13.56, + "learning_rate": 1.644549821406577e-06, + "loss": 0.2738, + "step": 36733 + }, + { + "epoch": 13.56, + "learning_rate": 1.644512871043232e-06, + "loss": 0.4869, + "step": 36734 + }, + { + "epoch": 13.57, + "learning_rate": 1.6444759206798867e-06, + "loss": 0.287, + "step": 36735 + }, + { + "epoch": 13.57, + "learning_rate": 1.6444389703165415e-06, + "loss": 0.2511, + "step": 36736 + }, + { + "epoch": 13.57, + "learning_rate": 1.6444020199531961e-06, + "loss": 0.2958, + "step": 36737 + }, + { + "epoch": 13.57, + "learning_rate": 1.6443650695898512e-06, + "loss": 0.3226, + "step": 36738 + }, + { + "epoch": 13.57, + "learning_rate": 1.6443281192265056e-06, + "loss": 0.1969, + "step": 36739 + }, + { + "epoch": 13.57, + "learning_rate": 1.6442911688631607e-06, + "loss": 0.2123, + "step": 36740 + }, + { + "epoch": 13.57, + "learning_rate": 1.6442542184998153e-06, + "loss": 0.1469, + "step": 36741 + }, + { + "epoch": 13.57, + "learning_rate": 1.64421726813647e-06, + "loss": 0.3553, + "step": 36742 + }, + { + "epoch": 13.57, + "learning_rate": 1.6441803177731248e-06, + "loss": 0.1666, + "step": 36743 + }, + { + "epoch": 13.57, + "learning_rate": 1.6441433674097794e-06, + "loss": 0.2743, + "step": 36744 + }, + { + "epoch": 13.57, + "learning_rate": 1.6441064170464343e-06, + "loss": 0.2391, + "step": 36745 + }, + { + "epoch": 13.57, + "learning_rate": 1.644069466683089e-06, + "loss": 0.3318, + "step": 36746 + }, + { + "epoch": 13.57, + "learning_rate": 1.644032516319744e-06, + "loss": 0.3431, + "step": 36747 + }, + { + "epoch": 13.57, + "learning_rate": 1.6439955659563986e-06, + "loss": 0.3986, + "step": 36748 + }, + { + "epoch": 13.57, + "learning_rate": 1.6439586155930535e-06, + "loss": 0.4083, + "step": 36749 + }, + { + "epoch": 13.57, + "learning_rate": 1.6439216652297081e-06, + "loss": 0.3488, + "step": 36750 + }, + { + "epoch": 13.57, + "learning_rate": 1.643884714866363e-06, + "loss": 0.2748, + "step": 36751 + }, + { + "epoch": 13.57, + "learning_rate": 1.6438477645030176e-06, + "loss": 0.3689, + "step": 36752 + }, + { + "epoch": 13.57, + "learning_rate": 1.6438108141396722e-06, + "loss": 0.1954, + "step": 36753 + }, + { + "epoch": 13.57, + "learning_rate": 1.6437738637763273e-06, + "loss": 0.3359, + "step": 36754 + }, + { + "epoch": 13.57, + "learning_rate": 1.643736913412982e-06, + "loss": 0.3259, + "step": 36755 + }, + { + "epoch": 13.57, + "learning_rate": 1.6436999630496368e-06, + "loss": 0.3547, + "step": 36756 + }, + { + "epoch": 13.57, + "learning_rate": 1.6436630126862914e-06, + "loss": 0.3204, + "step": 36757 + }, + { + "epoch": 13.57, + "learning_rate": 1.6436260623229463e-06, + "loss": 0.3462, + "step": 36758 + }, + { + "epoch": 13.57, + "learning_rate": 1.6435891119596009e-06, + "loss": 0.2557, + "step": 36759 + }, + { + "epoch": 13.57, + "learning_rate": 1.643552161596256e-06, + "loss": 0.207, + "step": 36760 + }, + { + "epoch": 13.57, + "learning_rate": 1.6435152112329106e-06, + "loss": 0.2279, + "step": 36761 + }, + { + "epoch": 13.58, + "learning_rate": 1.643478260869565e-06, + "loss": 0.1891, + "step": 36762 + }, + { + "epoch": 13.58, + "learning_rate": 1.64344131050622e-06, + "loss": 0.2712, + "step": 36763 + }, + { + "epoch": 13.58, + "learning_rate": 1.6434043601428747e-06, + "loss": 0.3287, + "step": 36764 + }, + { + "epoch": 13.58, + "learning_rate": 1.6433674097795296e-06, + "loss": 0.363, + "step": 36765 + }, + { + "epoch": 13.58, + "learning_rate": 1.6433304594161842e-06, + "loss": 0.2724, + "step": 36766 + }, + { + "epoch": 13.58, + "learning_rate": 1.643293509052839e-06, + "loss": 0.3474, + "step": 36767 + }, + { + "epoch": 13.58, + "learning_rate": 1.6432565586894937e-06, + "loss": 0.2388, + "step": 36768 + }, + { + "epoch": 13.58, + "learning_rate": 1.6432196083261487e-06, + "loss": 0.2073, + "step": 36769 + }, + { + "epoch": 13.58, + "learning_rate": 1.6431826579628034e-06, + "loss": 0.3211, + "step": 36770 + }, + { + "epoch": 13.58, + "learning_rate": 1.6431457075994582e-06, + "loss": 0.3379, + "step": 36771 + }, + { + "epoch": 13.58, + "learning_rate": 1.6431087572361129e-06, + "loss": 0.3169, + "step": 36772 + }, + { + "epoch": 13.58, + "learning_rate": 1.6430718068727675e-06, + "loss": 0.2676, + "step": 36773 + }, + { + "epoch": 13.58, + "learning_rate": 1.6430348565094223e-06, + "loss": 0.3914, + "step": 36774 + }, + { + "epoch": 13.58, + "learning_rate": 1.642997906146077e-06, + "loss": 0.2825, + "step": 36775 + }, + { + "epoch": 13.58, + "learning_rate": 1.642960955782732e-06, + "loss": 0.3235, + "step": 36776 + }, + { + "epoch": 13.58, + "learning_rate": 1.6429240054193867e-06, + "loss": 0.2728, + "step": 36777 + }, + { + "epoch": 13.58, + "learning_rate": 1.6428870550560415e-06, + "loss": 0.1874, + "step": 36778 + }, + { + "epoch": 13.58, + "learning_rate": 1.6428501046926962e-06, + "loss": 0.2593, + "step": 36779 + }, + { + "epoch": 13.58, + "learning_rate": 1.642813154329351e-06, + "loss": 0.6108, + "step": 36780 + }, + { + "epoch": 13.58, + "learning_rate": 1.6427762039660056e-06, + "loss": 0.2562, + "step": 36781 + }, + { + "epoch": 13.58, + "learning_rate": 1.6427392536026603e-06, + "loss": 0.2375, + "step": 36782 + }, + { + "epoch": 13.58, + "learning_rate": 1.6427023032393153e-06, + "loss": 0.2439, + "step": 36783 + }, + { + "epoch": 13.58, + "learning_rate": 1.6426653528759698e-06, + "loss": 0.3535, + "step": 36784 + }, + { + "epoch": 13.58, + "learning_rate": 1.6426284025126248e-06, + "loss": 0.2597, + "step": 36785 + }, + { + "epoch": 13.58, + "learning_rate": 1.6425914521492795e-06, + "loss": 0.4129, + "step": 36786 + }, + { + "epoch": 13.58, + "learning_rate": 1.6425545017859343e-06, + "loss": 0.3346, + "step": 36787 + }, + { + "epoch": 13.58, + "learning_rate": 1.642517551422589e-06, + "loss": 0.2701, + "step": 36788 + }, + { + "epoch": 13.59, + "learning_rate": 1.642480601059244e-06, + "loss": 0.1644, + "step": 36789 + }, + { + "epoch": 13.59, + "learning_rate": 1.6424436506958984e-06, + "loss": 0.2317, + "step": 36790 + }, + { + "epoch": 13.59, + "learning_rate": 1.6424067003325535e-06, + "loss": 0.1555, + "step": 36791 + }, + { + "epoch": 13.59, + "learning_rate": 1.6423697499692081e-06, + "loss": 0.2704, + "step": 36792 + }, + { + "epoch": 13.59, + "learning_rate": 1.6423327996058628e-06, + "loss": 0.2174, + "step": 36793 + }, + { + "epoch": 13.59, + "learning_rate": 1.6422958492425176e-06, + "loss": 0.4072, + "step": 36794 + }, + { + "epoch": 13.59, + "learning_rate": 1.6422588988791723e-06, + "loss": 0.3285, + "step": 36795 + }, + { + "epoch": 13.59, + "learning_rate": 1.642221948515827e-06, + "loss": 0.2984, + "step": 36796 + }, + { + "epoch": 13.59, + "learning_rate": 1.6421849981524817e-06, + "loss": 0.4681, + "step": 36797 + }, + { + "epoch": 13.59, + "learning_rate": 1.6421480477891368e-06, + "loss": 0.3115, + "step": 36798 + }, + { + "epoch": 13.59, + "learning_rate": 1.6421110974257914e-06, + "loss": 0.4547, + "step": 36799 + }, + { + "epoch": 13.59, + "learning_rate": 1.6420741470624463e-06, + "loss": 0.4046, + "step": 36800 + }, + { + "epoch": 13.59, + "learning_rate": 1.642037196699101e-06, + "loss": 0.1803, + "step": 36801 + }, + { + "epoch": 13.59, + "learning_rate": 1.6420002463357556e-06, + "loss": 0.3199, + "step": 36802 + }, + { + "epoch": 13.59, + "learning_rate": 1.6419632959724104e-06, + "loss": 0.2215, + "step": 36803 + }, + { + "epoch": 13.59, + "learning_rate": 1.641926345609065e-06, + "loss": 0.1765, + "step": 36804 + }, + { + "epoch": 13.59, + "learning_rate": 1.64188939524572e-06, + "loss": 0.2775, + "step": 36805 + }, + { + "epoch": 13.59, + "learning_rate": 1.6418524448823745e-06, + "loss": 0.2579, + "step": 36806 + }, + { + "epoch": 13.59, + "learning_rate": 1.6418154945190296e-06, + "loss": 0.3037, + "step": 36807 + }, + { + "epoch": 13.59, + "learning_rate": 1.6417785441556842e-06, + "loss": 0.2228, + "step": 36808 + }, + { + "epoch": 13.59, + "learning_rate": 1.641741593792339e-06, + "loss": 0.3199, + "step": 36809 + }, + { + "epoch": 13.59, + "learning_rate": 1.6417046434289937e-06, + "loss": 0.2787, + "step": 36810 + }, + { + "epoch": 13.59, + "learning_rate": 1.6416676930656488e-06, + "loss": 0.2921, + "step": 36811 + }, + { + "epoch": 13.59, + "learning_rate": 1.6416307427023032e-06, + "loss": 0.2391, + "step": 36812 + }, + { + "epoch": 13.59, + "learning_rate": 1.6415937923389578e-06, + "loss": 0.2461, + "step": 36813 + }, + { + "epoch": 13.59, + "learning_rate": 1.6415568419756129e-06, + "loss": 0.3312, + "step": 36814 + }, + { + "epoch": 13.59, + "learning_rate": 1.6415198916122675e-06, + "loss": 0.2361, + "step": 36815 + }, + { + "epoch": 13.6, + "learning_rate": 1.6414829412489224e-06, + "loss": 0.2289, + "step": 36816 + }, + { + "epoch": 13.6, + "learning_rate": 1.641445990885577e-06, + "loss": 0.2098, + "step": 36817 + }, + { + "epoch": 13.6, + "learning_rate": 1.6414090405222319e-06, + "loss": 0.3076, + "step": 36818 + }, + { + "epoch": 13.6, + "learning_rate": 1.6413720901588865e-06, + "loss": 0.3515, + "step": 36819 + }, + { + "epoch": 13.6, + "learning_rate": 1.6413351397955416e-06, + "loss": 0.2759, + "step": 36820 + }, + { + "epoch": 13.6, + "learning_rate": 1.6412981894321962e-06, + "loss": 0.2817, + "step": 36821 + }, + { + "epoch": 13.6, + "learning_rate": 1.6412612390688508e-06, + "loss": 0.3012, + "step": 36822 + }, + { + "epoch": 13.6, + "learning_rate": 1.6412242887055057e-06, + "loss": 0.1952, + "step": 36823 + }, + { + "epoch": 13.6, + "learning_rate": 1.6411873383421603e-06, + "loss": 0.2988, + "step": 36824 + }, + { + "epoch": 13.6, + "learning_rate": 1.6411503879788152e-06, + "loss": 0.2318, + "step": 36825 + }, + { + "epoch": 13.6, + "learning_rate": 1.6411134376154698e-06, + "loss": 0.2666, + "step": 36826 + }, + { + "epoch": 13.6, + "learning_rate": 1.6410764872521249e-06, + "loss": 0.3483, + "step": 36827 + }, + { + "epoch": 13.6, + "learning_rate": 1.6410395368887795e-06, + "loss": 0.2675, + "step": 36828 + }, + { + "epoch": 13.6, + "learning_rate": 1.6410025865254343e-06, + "loss": 0.311, + "step": 36829 + }, + { + "epoch": 13.6, + "learning_rate": 1.640965636162089e-06, + "loss": 0.2652, + "step": 36830 + }, + { + "epoch": 13.6, + "learning_rate": 1.6409286857987438e-06, + "loss": 0.2606, + "step": 36831 + }, + { + "epoch": 13.6, + "learning_rate": 1.6408917354353985e-06, + "loss": 0.249, + "step": 36832 + }, + { + "epoch": 13.6, + "learning_rate": 1.640854785072053e-06, + "loss": 0.3708, + "step": 36833 + }, + { + "epoch": 13.6, + "learning_rate": 1.6408178347087082e-06, + "loss": 0.2558, + "step": 36834 + }, + { + "epoch": 13.6, + "learning_rate": 1.6407808843453626e-06, + "loss": 0.4768, + "step": 36835 + }, + { + "epoch": 13.6, + "learning_rate": 1.6407439339820176e-06, + "loss": 0.3303, + "step": 36836 + }, + { + "epoch": 13.6, + "learning_rate": 1.6407069836186723e-06, + "loss": 0.1952, + "step": 36837 + }, + { + "epoch": 13.6, + "learning_rate": 1.6406700332553271e-06, + "loss": 0.2396, + "step": 36838 + }, + { + "epoch": 13.6, + "learning_rate": 1.6406330828919818e-06, + "loss": 0.4072, + "step": 36839 + }, + { + "epoch": 13.6, + "learning_rate": 1.6405961325286368e-06, + "loss": 0.2742, + "step": 36840 + }, + { + "epoch": 13.6, + "learning_rate": 1.6405591821652912e-06, + "loss": 0.2486, + "step": 36841 + }, + { + "epoch": 13.6, + "learning_rate": 1.6405222318019459e-06, + "loss": 0.3495, + "step": 36842 + }, + { + "epoch": 13.61, + "learning_rate": 1.640485281438601e-06, + "loss": 0.2066, + "step": 36843 + }, + { + "epoch": 13.61, + "learning_rate": 1.6404483310752556e-06, + "loss": 0.2488, + "step": 36844 + }, + { + "epoch": 13.61, + "learning_rate": 1.6404113807119104e-06, + "loss": 0.3368, + "step": 36845 + }, + { + "epoch": 13.61, + "learning_rate": 1.640374430348565e-06, + "loss": 0.3327, + "step": 36846 + }, + { + "epoch": 13.61, + "learning_rate": 1.64033747998522e-06, + "loss": 0.2346, + "step": 36847 + }, + { + "epoch": 13.61, + "learning_rate": 1.6403005296218745e-06, + "loss": 0.2495, + "step": 36848 + }, + { + "epoch": 13.61, + "learning_rate": 1.6402635792585296e-06, + "loss": 0.2393, + "step": 36849 + }, + { + "epoch": 13.61, + "learning_rate": 1.6402266288951842e-06, + "loss": 0.3248, + "step": 36850 + }, + { + "epoch": 13.61, + "learning_rate": 1.640189678531839e-06, + "loss": 0.3219, + "step": 36851 + }, + { + "epoch": 13.61, + "learning_rate": 1.6401527281684937e-06, + "loss": 0.4162, + "step": 36852 + }, + { + "epoch": 13.61, + "learning_rate": 1.6401157778051484e-06, + "loss": 0.2667, + "step": 36853 + }, + { + "epoch": 13.61, + "learning_rate": 1.6400788274418032e-06, + "loss": 0.3631, + "step": 36854 + }, + { + "epoch": 13.61, + "learning_rate": 1.6400418770784578e-06, + "loss": 0.1786, + "step": 36855 + }, + { + "epoch": 13.61, + "learning_rate": 1.640004926715113e-06, + "loss": 0.2446, + "step": 36856 + }, + { + "epoch": 13.61, + "learning_rate": 1.6399679763517673e-06, + "loss": 0.2173, + "step": 36857 + }, + { + "epoch": 13.61, + "learning_rate": 1.6399310259884224e-06, + "loss": 0.3296, + "step": 36858 + }, + { + "epoch": 13.61, + "learning_rate": 1.639894075625077e-06, + "loss": 0.2599, + "step": 36859 + }, + { + "epoch": 13.61, + "learning_rate": 1.6398571252617319e-06, + "loss": 0.3239, + "step": 36860 + }, + { + "epoch": 13.61, + "learning_rate": 1.6398201748983865e-06, + "loss": 0.2449, + "step": 36861 + }, + { + "epoch": 13.61, + "learning_rate": 1.6397832245350412e-06, + "loss": 0.2981, + "step": 36862 + }, + { + "epoch": 13.61, + "learning_rate": 1.639746274171696e-06, + "loss": 0.5671, + "step": 36863 + }, + { + "epoch": 13.61, + "learning_rate": 1.6397093238083506e-06, + "loss": 0.278, + "step": 36864 + }, + { + "epoch": 13.61, + "learning_rate": 1.6396723734450057e-06, + "loss": 0.3062, + "step": 36865 + }, + { + "epoch": 13.61, + "learning_rate": 1.6396354230816603e-06, + "loss": 0.4869, + "step": 36866 + }, + { + "epoch": 13.61, + "learning_rate": 1.6395984727183152e-06, + "loss": 0.3969, + "step": 36867 + }, + { + "epoch": 13.61, + "learning_rate": 1.6395615223549698e-06, + "loss": 0.1956, + "step": 36868 + }, + { + "epoch": 13.61, + "learning_rate": 1.6395245719916247e-06, + "loss": 0.1665, + "step": 36869 + }, + { + "epoch": 13.62, + "learning_rate": 1.6394876216282793e-06, + "loss": 0.2091, + "step": 36870 + }, + { + "epoch": 13.62, + "learning_rate": 1.6394506712649344e-06, + "loss": 0.2875, + "step": 36871 + }, + { + "epoch": 13.62, + "learning_rate": 1.639413720901589e-06, + "loss": 0.2279, + "step": 36872 + }, + { + "epoch": 13.62, + "learning_rate": 1.6393767705382436e-06, + "loss": 0.2704, + "step": 36873 + }, + { + "epoch": 13.62, + "learning_rate": 1.6393398201748985e-06, + "loss": 0.2743, + "step": 36874 + }, + { + "epoch": 13.62, + "learning_rate": 1.6393028698115531e-06, + "loss": 0.2566, + "step": 36875 + }, + { + "epoch": 13.62, + "learning_rate": 1.639265919448208e-06, + "loss": 0.2519, + "step": 36876 + }, + { + "epoch": 13.62, + "learning_rate": 1.6392289690848626e-06, + "loss": 0.2616, + "step": 36877 + }, + { + "epoch": 13.62, + "learning_rate": 1.6391920187215177e-06, + "loss": 0.3074, + "step": 36878 + }, + { + "epoch": 13.62, + "learning_rate": 1.639155068358172e-06, + "loss": 0.1647, + "step": 36879 + }, + { + "epoch": 13.62, + "learning_rate": 1.6391181179948271e-06, + "loss": 0.3477, + "step": 36880 + }, + { + "epoch": 13.62, + "learning_rate": 1.6390811676314818e-06, + "loss": 0.445, + "step": 36881 + }, + { + "epoch": 13.62, + "learning_rate": 1.6390442172681364e-06, + "loss": 0.3105, + "step": 36882 + }, + { + "epoch": 13.62, + "learning_rate": 1.6390072669047913e-06, + "loss": 0.4096, + "step": 36883 + }, + { + "epoch": 13.62, + "learning_rate": 1.638970316541446e-06, + "loss": 0.2328, + "step": 36884 + }, + { + "epoch": 13.62, + "learning_rate": 1.6389333661781008e-06, + "loss": 0.2515, + "step": 36885 + }, + { + "epoch": 13.62, + "learning_rate": 1.6388964158147554e-06, + "loss": 0.4448, + "step": 36886 + }, + { + "epoch": 13.62, + "learning_rate": 1.6388594654514105e-06, + "loss": 0.2624, + "step": 36887 + }, + { + "epoch": 13.62, + "learning_rate": 1.638822515088065e-06, + "loss": 0.4928, + "step": 36888 + }, + { + "epoch": 13.62, + "learning_rate": 1.63878556472472e-06, + "loss": 0.2502, + "step": 36889 + }, + { + "epoch": 13.62, + "learning_rate": 1.6387486143613746e-06, + "loss": 0.2167, + "step": 36890 + }, + { + "epoch": 13.62, + "learning_rate": 1.6387116639980294e-06, + "loss": 0.2778, + "step": 36891 + }, + { + "epoch": 13.62, + "learning_rate": 1.638674713634684e-06, + "loss": 0.3538, + "step": 36892 + }, + { + "epoch": 13.62, + "learning_rate": 1.6386377632713387e-06, + "loss": 0.186, + "step": 36893 + }, + { + "epoch": 13.62, + "learning_rate": 1.6386008129079938e-06, + "loss": 0.2352, + "step": 36894 + }, + { + "epoch": 13.62, + "learning_rate": 1.6385638625446484e-06, + "loss": 0.2191, + "step": 36895 + }, + { + "epoch": 13.62, + "learning_rate": 1.6385269121813032e-06, + "loss": 0.2867, + "step": 36896 + }, + { + "epoch": 13.63, + "learning_rate": 1.6384899618179579e-06, + "loss": 0.2205, + "step": 36897 + }, + { + "epoch": 13.63, + "learning_rate": 1.6384530114546127e-06, + "loss": 0.3148, + "step": 36898 + }, + { + "epoch": 13.63, + "learning_rate": 1.6384160610912674e-06, + "loss": 0.3193, + "step": 36899 + }, + { + "epoch": 13.63, + "learning_rate": 1.6383791107279224e-06, + "loss": 0.3119, + "step": 36900 + }, + { + "epoch": 13.63, + "learning_rate": 1.638342160364577e-06, + "loss": 0.2205, + "step": 36901 + }, + { + "epoch": 13.63, + "learning_rate": 1.6383052100012315e-06, + "loss": 0.3782, + "step": 36902 + }, + { + "epoch": 13.63, + "learning_rate": 1.6382682596378865e-06, + "loss": 0.2044, + "step": 36903 + }, + { + "epoch": 13.63, + "learning_rate": 1.6382313092745412e-06, + "loss": 0.3041, + "step": 36904 + }, + { + "epoch": 13.63, + "learning_rate": 1.638194358911196e-06, + "loss": 0.248, + "step": 36905 + }, + { + "epoch": 13.63, + "learning_rate": 1.6381574085478507e-06, + "loss": 0.2958, + "step": 36906 + }, + { + "epoch": 13.63, + "learning_rate": 1.6381204581845057e-06, + "loss": 0.2804, + "step": 36907 + }, + { + "epoch": 13.63, + "learning_rate": 1.6380835078211601e-06, + "loss": 0.2348, + "step": 36908 + }, + { + "epoch": 13.63, + "learning_rate": 1.6380465574578152e-06, + "loss": 0.2982, + "step": 36909 + }, + { + "epoch": 13.63, + "learning_rate": 1.6380096070944698e-06, + "loss": 0.2968, + "step": 36910 + }, + { + "epoch": 13.63, + "learning_rate": 1.6379726567311247e-06, + "loss": 0.2817, + "step": 36911 + }, + { + "epoch": 13.63, + "learning_rate": 1.6379357063677793e-06, + "loss": 0.3701, + "step": 36912 + }, + { + "epoch": 13.63, + "learning_rate": 1.637898756004434e-06, + "loss": 0.2153, + "step": 36913 + }, + { + "epoch": 13.63, + "learning_rate": 1.6378618056410888e-06, + "loss": 0.2796, + "step": 36914 + }, + { + "epoch": 13.63, + "learning_rate": 1.6378248552777434e-06, + "loss": 0.2441, + "step": 36915 + }, + { + "epoch": 13.63, + "learning_rate": 1.6377879049143985e-06, + "loss": 0.2393, + "step": 36916 + }, + { + "epoch": 13.63, + "learning_rate": 1.6377509545510531e-06, + "loss": 0.2794, + "step": 36917 + }, + { + "epoch": 13.63, + "learning_rate": 1.637714004187708e-06, + "loss": 0.2642, + "step": 36918 + }, + { + "epoch": 13.63, + "learning_rate": 1.6376770538243626e-06, + "loss": 0.335, + "step": 36919 + }, + { + "epoch": 13.63, + "learning_rate": 1.6376401034610175e-06, + "loss": 0.3619, + "step": 36920 + }, + { + "epoch": 13.63, + "learning_rate": 1.6376031530976721e-06, + "loss": 0.2746, + "step": 36921 + }, + { + "epoch": 13.63, + "learning_rate": 1.6375662027343267e-06, + "loss": 0.3039, + "step": 36922 + }, + { + "epoch": 13.63, + "learning_rate": 1.6375292523709818e-06, + "loss": 0.2583, + "step": 36923 + }, + { + "epoch": 13.64, + "learning_rate": 1.6374923020076362e-06, + "loss": 0.3149, + "step": 36924 + }, + { + "epoch": 13.64, + "learning_rate": 1.6374553516442913e-06, + "loss": 0.2178, + "step": 36925 + }, + { + "epoch": 13.64, + "learning_rate": 1.637418401280946e-06, + "loss": 0.3112, + "step": 36926 + }, + { + "epoch": 13.64, + "learning_rate": 1.6373814509176008e-06, + "loss": 0.2345, + "step": 36927 + }, + { + "epoch": 13.64, + "learning_rate": 1.6373445005542554e-06, + "loss": 0.3213, + "step": 36928 + }, + { + "epoch": 13.64, + "learning_rate": 1.6373075501909105e-06, + "loss": 0.2612, + "step": 36929 + }, + { + "epoch": 13.64, + "learning_rate": 1.637270599827565e-06, + "loss": 0.3724, + "step": 36930 + }, + { + "epoch": 13.64, + "learning_rate": 1.63723364946422e-06, + "loss": 0.2889, + "step": 36931 + }, + { + "epoch": 13.64, + "learning_rate": 1.6371966991008746e-06, + "loss": 0.1971, + "step": 36932 + }, + { + "epoch": 13.64, + "learning_rate": 1.6371597487375292e-06, + "loss": 0.2739, + "step": 36933 + }, + { + "epoch": 13.64, + "learning_rate": 1.637122798374184e-06, + "loss": 0.293, + "step": 36934 + }, + { + "epoch": 13.64, + "learning_rate": 1.6370858480108387e-06, + "loss": 0.2766, + "step": 36935 + }, + { + "epoch": 13.64, + "learning_rate": 1.6370488976474936e-06, + "loss": 0.4118, + "step": 36936 + }, + { + "epoch": 13.64, + "learning_rate": 1.6370119472841482e-06, + "loss": 0.3414, + "step": 36937 + }, + { + "epoch": 13.64, + "learning_rate": 1.6369749969208033e-06, + "loss": 0.2703, + "step": 36938 + }, + { + "epoch": 13.64, + "learning_rate": 1.636938046557458e-06, + "loss": 0.3505, + "step": 36939 + }, + { + "epoch": 13.64, + "learning_rate": 1.6369010961941127e-06, + "loss": 0.2954, + "step": 36940 + }, + { + "epoch": 13.64, + "learning_rate": 1.6368641458307674e-06, + "loss": 0.204, + "step": 36941 + }, + { + "epoch": 13.64, + "learning_rate": 1.636827195467422e-06, + "loss": 0.3729, + "step": 36942 + }, + { + "epoch": 13.64, + "learning_rate": 1.6367902451040769e-06, + "loss": 0.3522, + "step": 36943 + }, + { + "epoch": 13.64, + "learning_rate": 1.6367532947407315e-06, + "loss": 0.336, + "step": 36944 + }, + { + "epoch": 13.64, + "learning_rate": 1.6367163443773866e-06, + "loss": 0.3128, + "step": 36945 + }, + { + "epoch": 13.64, + "learning_rate": 1.6366793940140412e-06, + "loss": 0.3765, + "step": 36946 + }, + { + "epoch": 13.64, + "learning_rate": 1.636642443650696e-06, + "loss": 0.2798, + "step": 36947 + }, + { + "epoch": 13.64, + "learning_rate": 1.6366054932873507e-06, + "loss": 0.3444, + "step": 36948 + }, + { + "epoch": 13.64, + "learning_rate": 1.6365685429240055e-06, + "loss": 0.1992, + "step": 36949 + }, + { + "epoch": 13.64, + "learning_rate": 1.6365315925606602e-06, + "loss": 0.279, + "step": 36950 + }, + { + "epoch": 13.65, + "learning_rate": 1.6364946421973148e-06, + "loss": 0.1658, + "step": 36951 + }, + { + "epoch": 13.65, + "learning_rate": 1.6364576918339699e-06, + "loss": 0.237, + "step": 36952 + }, + { + "epoch": 13.65, + "learning_rate": 1.6364207414706243e-06, + "loss": 0.2628, + "step": 36953 + }, + { + "epoch": 13.65, + "learning_rate": 1.6363837911072793e-06, + "loss": 0.3995, + "step": 36954 + }, + { + "epoch": 13.65, + "learning_rate": 1.636346840743934e-06, + "loss": 0.3502, + "step": 36955 + }, + { + "epoch": 13.65, + "learning_rate": 1.6363098903805888e-06, + "loss": 0.3567, + "step": 36956 + }, + { + "epoch": 13.65, + "learning_rate": 1.6362729400172435e-06, + "loss": 0.1795, + "step": 36957 + }, + { + "epoch": 13.65, + "learning_rate": 1.6362359896538983e-06, + "loss": 0.3014, + "step": 36958 + }, + { + "epoch": 13.65, + "learning_rate": 1.636199039290553e-06, + "loss": 0.3444, + "step": 36959 + }, + { + "epoch": 13.65, + "learning_rate": 1.636162088927208e-06, + "loss": 0.4401, + "step": 36960 + }, + { + "epoch": 13.65, + "learning_rate": 1.6361251385638627e-06, + "loss": 0.2897, + "step": 36961 + }, + { + "epoch": 13.65, + "learning_rate": 1.6360881882005173e-06, + "loss": 0.2241, + "step": 36962 + }, + { + "epoch": 13.65, + "learning_rate": 1.6360512378371721e-06, + "loss": 0.3573, + "step": 36963 + }, + { + "epoch": 13.65, + "learning_rate": 1.6360142874738268e-06, + "loss": 0.1994, + "step": 36964 + }, + { + "epoch": 13.65, + "learning_rate": 1.6359773371104816e-06, + "loss": 0.3559, + "step": 36965 + }, + { + "epoch": 13.65, + "learning_rate": 1.6359403867471363e-06, + "loss": 0.2129, + "step": 36966 + }, + { + "epoch": 13.65, + "learning_rate": 1.6359034363837913e-06, + "loss": 0.3163, + "step": 36967 + }, + { + "epoch": 13.65, + "learning_rate": 1.635866486020446e-06, + "loss": 0.2412, + "step": 36968 + }, + { + "epoch": 13.65, + "learning_rate": 1.6358295356571008e-06, + "loss": 0.2604, + "step": 36969 + }, + { + "epoch": 13.65, + "learning_rate": 1.6357925852937554e-06, + "loss": 0.2819, + "step": 36970 + }, + { + "epoch": 13.65, + "learning_rate": 1.63575563493041e-06, + "loss": 0.3596, + "step": 36971 + }, + { + "epoch": 13.65, + "learning_rate": 1.635718684567065e-06, + "loss": 0.263, + "step": 36972 + }, + { + "epoch": 13.65, + "learning_rate": 1.6356817342037196e-06, + "loss": 0.1807, + "step": 36973 + }, + { + "epoch": 13.65, + "learning_rate": 1.6356447838403746e-06, + "loss": 0.3862, + "step": 36974 + }, + { + "epoch": 13.65, + "learning_rate": 1.635607833477029e-06, + "loss": 0.3249, + "step": 36975 + }, + { + "epoch": 13.65, + "learning_rate": 1.635570883113684e-06, + "loss": 0.2632, + "step": 36976 + }, + { + "epoch": 13.65, + "learning_rate": 1.6355339327503387e-06, + "loss": 0.3269, + "step": 36977 + }, + { + "epoch": 13.66, + "learning_rate": 1.6354969823869936e-06, + "loss": 0.3912, + "step": 36978 + }, + { + "epoch": 13.66, + "learning_rate": 1.6354600320236482e-06, + "loss": 0.2615, + "step": 36979 + }, + { + "epoch": 13.66, + "learning_rate": 1.6354230816603033e-06, + "loss": 0.4177, + "step": 36980 + }, + { + "epoch": 13.66, + "learning_rate": 1.6353861312969577e-06, + "loss": 0.2688, + "step": 36981 + }, + { + "epoch": 13.66, + "learning_rate": 1.6353491809336123e-06, + "loss": 0.1333, + "step": 36982 + }, + { + "epoch": 13.66, + "learning_rate": 1.6353122305702674e-06, + "loss": 0.2622, + "step": 36983 + }, + { + "epoch": 13.66, + "learning_rate": 1.635275280206922e-06, + "loss": 0.3693, + "step": 36984 + }, + { + "epoch": 13.66, + "learning_rate": 1.6352383298435769e-06, + "loss": 0.2728, + "step": 36985 + }, + { + "epoch": 13.66, + "learning_rate": 1.6352013794802315e-06, + "loss": 0.2023, + "step": 36986 + }, + { + "epoch": 13.66, + "learning_rate": 1.6351644291168864e-06, + "loss": 0.3322, + "step": 36987 + }, + { + "epoch": 13.66, + "learning_rate": 1.635127478753541e-06, + "loss": 0.3328, + "step": 36988 + }, + { + "epoch": 13.66, + "learning_rate": 1.635090528390196e-06, + "loss": 0.2871, + "step": 36989 + }, + { + "epoch": 13.66, + "learning_rate": 1.6350535780268507e-06, + "loss": 0.3062, + "step": 36990 + }, + { + "epoch": 13.66, + "learning_rate": 1.6350166276635051e-06, + "loss": 0.3494, + "step": 36991 + }, + { + "epoch": 13.66, + "learning_rate": 1.6349796773001602e-06, + "loss": 0.2777, + "step": 36992 + }, + { + "epoch": 13.66, + "learning_rate": 1.6349427269368148e-06, + "loss": 0.3348, + "step": 36993 + }, + { + "epoch": 13.66, + "learning_rate": 1.6349057765734697e-06, + "loss": 0.3521, + "step": 36994 + }, + { + "epoch": 13.66, + "learning_rate": 1.6348688262101243e-06, + "loss": 0.2849, + "step": 36995 + }, + { + "epoch": 13.66, + "learning_rate": 1.6348318758467794e-06, + "loss": 0.2314, + "step": 36996 + }, + { + "epoch": 13.66, + "learning_rate": 1.6347949254834338e-06, + "loss": 0.4042, + "step": 36997 + }, + { + "epoch": 13.66, + "learning_rate": 1.6347579751200889e-06, + "loss": 0.2238, + "step": 36998 + }, + { + "epoch": 13.66, + "learning_rate": 1.6347210247567435e-06, + "loss": 0.1431, + "step": 36999 + }, + { + "epoch": 13.66, + "learning_rate": 1.6346840743933983e-06, + "loss": 0.3305, + "step": 37000 + }, + { + "epoch": 13.66, + "learning_rate": 1.634647124030053e-06, + "loss": 0.3953, + "step": 37001 + }, + { + "epoch": 13.66, + "learning_rate": 1.6346101736667076e-06, + "loss": 0.1689, + "step": 37002 + }, + { + "epoch": 13.66, + "learning_rate": 1.6345732233033625e-06, + "loss": 0.2501, + "step": 37003 + }, + { + "epoch": 13.66, + "learning_rate": 1.634536272940017e-06, + "loss": 0.4007, + "step": 37004 + }, + { + "epoch": 13.67, + "learning_rate": 1.6344993225766722e-06, + "loss": 0.3023, + "step": 37005 + }, + { + "epoch": 13.67, + "learning_rate": 1.6344623722133268e-06, + "loss": 0.3799, + "step": 37006 + }, + { + "epoch": 13.67, + "learning_rate": 1.6344254218499816e-06, + "loss": 0.3177, + "step": 37007 + }, + { + "epoch": 13.67, + "learning_rate": 1.6343884714866363e-06, + "loss": 0.3469, + "step": 37008 + }, + { + "epoch": 13.67, + "learning_rate": 1.6343515211232911e-06, + "loss": 0.2718, + "step": 37009 + }, + { + "epoch": 13.67, + "learning_rate": 1.6343145707599458e-06, + "loss": 0.1984, + "step": 37010 + }, + { + "epoch": 13.67, + "learning_rate": 1.6342776203966004e-06, + "loss": 0.1671, + "step": 37011 + }, + { + "epoch": 13.67, + "learning_rate": 1.6342406700332555e-06, + "loss": 0.2632, + "step": 37012 + }, + { + "epoch": 13.67, + "learning_rate": 1.63420371966991e-06, + "loss": 0.3631, + "step": 37013 + }, + { + "epoch": 13.67, + "learning_rate": 1.634166769306565e-06, + "loss": 0.4837, + "step": 37014 + }, + { + "epoch": 13.67, + "learning_rate": 1.6341298189432196e-06, + "loss": 0.3451, + "step": 37015 + }, + { + "epoch": 13.67, + "learning_rate": 1.6340928685798744e-06, + "loss": 0.3034, + "step": 37016 + }, + { + "epoch": 13.67, + "learning_rate": 1.634055918216529e-06, + "loss": 0.2315, + "step": 37017 + }, + { + "epoch": 13.67, + "learning_rate": 1.6340189678531841e-06, + "loss": 0.2624, + "step": 37018 + }, + { + "epoch": 13.67, + "learning_rate": 1.6339820174898388e-06, + "loss": 0.264, + "step": 37019 + }, + { + "epoch": 13.67, + "learning_rate": 1.6339450671264936e-06, + "loss": 0.1772, + "step": 37020 + }, + { + "epoch": 13.67, + "learning_rate": 1.6339081167631482e-06, + "loss": 0.2794, + "step": 37021 + }, + { + "epoch": 13.67, + "learning_rate": 1.6338711663998029e-06, + "loss": 0.2174, + "step": 37022 + }, + { + "epoch": 13.67, + "learning_rate": 1.6338342160364577e-06, + "loss": 0.2576, + "step": 37023 + }, + { + "epoch": 13.67, + "learning_rate": 1.6337972656731124e-06, + "loss": 0.2278, + "step": 37024 + }, + { + "epoch": 13.67, + "learning_rate": 1.6337603153097674e-06, + "loss": 0.2021, + "step": 37025 + }, + { + "epoch": 13.67, + "learning_rate": 1.6337233649464219e-06, + "loss": 0.2613, + "step": 37026 + }, + { + "epoch": 13.67, + "learning_rate": 1.633686414583077e-06, + "loss": 0.3571, + "step": 37027 + }, + { + "epoch": 13.67, + "learning_rate": 1.6336494642197316e-06, + "loss": 0.233, + "step": 37028 + }, + { + "epoch": 13.67, + "learning_rate": 1.6336125138563864e-06, + "loss": 0.3876, + "step": 37029 + }, + { + "epoch": 13.67, + "learning_rate": 1.633575563493041e-06, + "loss": 0.2244, + "step": 37030 + }, + { + "epoch": 13.67, + "learning_rate": 1.6335386131296957e-06, + "loss": 0.2974, + "step": 37031 + }, + { + "epoch": 13.68, + "learning_rate": 1.6335016627663505e-06, + "loss": 0.2684, + "step": 37032 + }, + { + "epoch": 13.68, + "learning_rate": 1.6334647124030052e-06, + "loss": 0.2566, + "step": 37033 + }, + { + "epoch": 13.68, + "learning_rate": 1.6334277620396602e-06, + "loss": 0.3269, + "step": 37034 + }, + { + "epoch": 13.68, + "learning_rate": 1.6333908116763149e-06, + "loss": 0.3185, + "step": 37035 + }, + { + "epoch": 13.68, + "learning_rate": 1.6333538613129697e-06, + "loss": 0.2271, + "step": 37036 + }, + { + "epoch": 13.68, + "learning_rate": 1.6333169109496243e-06, + "loss": 0.1517, + "step": 37037 + }, + { + "epoch": 13.68, + "learning_rate": 1.6332799605862792e-06, + "loss": 0.1231, + "step": 37038 + }, + { + "epoch": 13.68, + "learning_rate": 1.6332430102229338e-06, + "loss": 0.226, + "step": 37039 + }, + { + "epoch": 13.68, + "learning_rate": 1.6332060598595889e-06, + "loss": 0.2163, + "step": 37040 + }, + { + "epoch": 13.68, + "learning_rate": 1.6331691094962435e-06, + "loss": 0.3098, + "step": 37041 + }, + { + "epoch": 13.68, + "learning_rate": 1.633132159132898e-06, + "loss": 0.2964, + "step": 37042 + }, + { + "epoch": 13.68, + "learning_rate": 1.633095208769553e-06, + "loss": 0.2816, + "step": 37043 + }, + { + "epoch": 13.68, + "learning_rate": 1.6330582584062076e-06, + "loss": 0.2697, + "step": 37044 + }, + { + "epoch": 13.68, + "learning_rate": 1.6330213080428625e-06, + "loss": 0.3786, + "step": 37045 + }, + { + "epoch": 13.68, + "learning_rate": 1.6329843576795171e-06, + "loss": 0.2276, + "step": 37046 + }, + { + "epoch": 13.68, + "learning_rate": 1.6329474073161722e-06, + "loss": 0.1972, + "step": 37047 + }, + { + "epoch": 13.68, + "learning_rate": 1.6329104569528266e-06, + "loss": 0.2773, + "step": 37048 + }, + { + "epoch": 13.68, + "learning_rate": 1.6328735065894817e-06, + "loss": 0.2254, + "step": 37049 + }, + { + "epoch": 13.68, + "learning_rate": 1.6328365562261363e-06, + "loss": 0.2427, + "step": 37050 + }, + { + "epoch": 13.68, + "learning_rate": 1.632799605862791e-06, + "loss": 0.3042, + "step": 37051 + }, + { + "epoch": 13.68, + "learning_rate": 1.6327626554994458e-06, + "loss": 0.2479, + "step": 37052 + }, + { + "epoch": 13.68, + "learning_rate": 1.6327257051361004e-06, + "loss": 0.2528, + "step": 37053 + }, + { + "epoch": 13.68, + "learning_rate": 1.6326887547727553e-06, + "loss": 0.4241, + "step": 37054 + }, + { + "epoch": 13.68, + "learning_rate": 1.63265180440941e-06, + "loss": 0.1898, + "step": 37055 + }, + { + "epoch": 13.68, + "learning_rate": 1.632614854046065e-06, + "loss": 0.4123, + "step": 37056 + }, + { + "epoch": 13.68, + "learning_rate": 1.6325779036827196e-06, + "loss": 0.2121, + "step": 37057 + }, + { + "epoch": 13.68, + "learning_rate": 1.6325409533193745e-06, + "loss": 0.242, + "step": 37058 + }, + { + "epoch": 13.69, + "learning_rate": 1.632504002956029e-06, + "loss": 0.1783, + "step": 37059 + }, + { + "epoch": 13.69, + "learning_rate": 1.632467052592684e-06, + "loss": 0.3253, + "step": 37060 + }, + { + "epoch": 13.69, + "learning_rate": 1.6324301022293386e-06, + "loss": 0.1882, + "step": 37061 + }, + { + "epoch": 13.69, + "learning_rate": 1.6323931518659932e-06, + "loss": 0.2948, + "step": 37062 + }, + { + "epoch": 13.69, + "learning_rate": 1.6323562015026483e-06, + "loss": 0.2835, + "step": 37063 + }, + { + "epoch": 13.69, + "learning_rate": 1.6323192511393027e-06, + "loss": 0.2408, + "step": 37064 + }, + { + "epoch": 13.69, + "learning_rate": 1.6322823007759578e-06, + "loss": 0.3398, + "step": 37065 + }, + { + "epoch": 13.69, + "learning_rate": 1.6322453504126124e-06, + "loss": 0.3884, + "step": 37066 + }, + { + "epoch": 13.69, + "learning_rate": 1.6322084000492672e-06, + "loss": 0.2923, + "step": 37067 + }, + { + "epoch": 13.69, + "learning_rate": 1.6321714496859219e-06, + "loss": 0.3244, + "step": 37068 + }, + { + "epoch": 13.69, + "learning_rate": 1.632134499322577e-06, + "loss": 0.3504, + "step": 37069 + }, + { + "epoch": 13.69, + "learning_rate": 1.6320975489592314e-06, + "loss": 0.2676, + "step": 37070 + }, + { + "epoch": 13.69, + "learning_rate": 1.632060598595886e-06, + "loss": 0.231, + "step": 37071 + }, + { + "epoch": 13.69, + "learning_rate": 1.632023648232541e-06, + "loss": 0.2865, + "step": 37072 + }, + { + "epoch": 13.69, + "learning_rate": 1.6319866978691957e-06, + "loss": 0.2264, + "step": 37073 + }, + { + "epoch": 13.69, + "learning_rate": 1.6319497475058505e-06, + "loss": 0.3691, + "step": 37074 + }, + { + "epoch": 13.69, + "learning_rate": 1.6319127971425052e-06, + "loss": 0.2313, + "step": 37075 + }, + { + "epoch": 13.69, + "learning_rate": 1.63187584677916e-06, + "loss": 0.3042, + "step": 37076 + }, + { + "epoch": 13.69, + "learning_rate": 1.6318388964158147e-06, + "loss": 0.2233, + "step": 37077 + }, + { + "epoch": 13.69, + "learning_rate": 1.6318019460524697e-06, + "loss": 0.3075, + "step": 37078 + }, + { + "epoch": 13.69, + "learning_rate": 1.6317649956891244e-06, + "loss": 0.2993, + "step": 37079 + }, + { + "epoch": 13.69, + "learning_rate": 1.6317280453257792e-06, + "loss": 0.3795, + "step": 37080 + }, + { + "epoch": 13.69, + "learning_rate": 1.6316910949624338e-06, + "loss": 0.3348, + "step": 37081 + }, + { + "epoch": 13.69, + "learning_rate": 1.6316541445990885e-06, + "loss": 0.3081, + "step": 37082 + }, + { + "epoch": 13.69, + "learning_rate": 1.6316171942357433e-06, + "loss": 0.286, + "step": 37083 + }, + { + "epoch": 13.69, + "learning_rate": 1.631580243872398e-06, + "loss": 0.272, + "step": 37084 + }, + { + "epoch": 13.69, + "learning_rate": 1.631543293509053e-06, + "loss": 0.3644, + "step": 37085 + }, + { + "epoch": 13.69, + "learning_rate": 1.6315063431457077e-06, + "loss": 0.3825, + "step": 37086 + }, + { + "epoch": 13.7, + "learning_rate": 1.6314693927823625e-06, + "loss": 0.2524, + "step": 37087 + }, + { + "epoch": 13.7, + "learning_rate": 1.6314324424190171e-06, + "loss": 0.4782, + "step": 37088 + }, + { + "epoch": 13.7, + "learning_rate": 1.631395492055672e-06, + "loss": 0.2566, + "step": 37089 + }, + { + "epoch": 13.7, + "learning_rate": 1.6313585416923266e-06, + "loss": 0.4197, + "step": 37090 + }, + { + "epoch": 13.7, + "learning_rate": 1.6313215913289813e-06, + "loss": 0.2345, + "step": 37091 + }, + { + "epoch": 13.7, + "learning_rate": 1.6312846409656363e-06, + "loss": 0.1642, + "step": 37092 + }, + { + "epoch": 13.7, + "learning_rate": 1.6312476906022908e-06, + "loss": 0.2922, + "step": 37093 + }, + { + "epoch": 13.7, + "learning_rate": 1.6312107402389458e-06, + "loss": 0.4039, + "step": 37094 + }, + { + "epoch": 13.7, + "learning_rate": 1.6311737898756005e-06, + "loss": 0.297, + "step": 37095 + }, + { + "epoch": 13.7, + "learning_rate": 1.6311368395122553e-06, + "loss": 0.3824, + "step": 37096 + }, + { + "epoch": 13.7, + "learning_rate": 1.63109988914891e-06, + "loss": 0.3204, + "step": 37097 + }, + { + "epoch": 13.7, + "learning_rate": 1.631062938785565e-06, + "loss": 0.3911, + "step": 37098 + }, + { + "epoch": 13.7, + "learning_rate": 1.6310259884222194e-06, + "loss": 0.3332, + "step": 37099 + }, + { + "epoch": 13.7, + "learning_rate": 1.6309890380588745e-06, + "loss": 0.2267, + "step": 37100 + }, + { + "epoch": 13.7, + "learning_rate": 1.6309520876955291e-06, + "loss": 0.3851, + "step": 37101 + }, + { + "epoch": 13.7, + "learning_rate": 1.6309151373321838e-06, + "loss": 0.2987, + "step": 37102 + }, + { + "epoch": 13.7, + "learning_rate": 1.6308781869688386e-06, + "loss": 0.233, + "step": 37103 + }, + { + "epoch": 13.7, + "learning_rate": 1.6308412366054932e-06, + "loss": 0.2274, + "step": 37104 + }, + { + "epoch": 13.7, + "learning_rate": 1.630804286242148e-06, + "loss": 0.3856, + "step": 37105 + }, + { + "epoch": 13.7, + "learning_rate": 1.6307673358788027e-06, + "loss": 0.2333, + "step": 37106 + }, + { + "epoch": 13.7, + "learning_rate": 1.6307303855154578e-06, + "loss": 0.2402, + "step": 37107 + }, + { + "epoch": 13.7, + "learning_rate": 1.6306934351521124e-06, + "loss": 0.3281, + "step": 37108 + }, + { + "epoch": 13.7, + "learning_rate": 1.6306564847887673e-06, + "loss": 0.2671, + "step": 37109 + }, + { + "epoch": 13.7, + "learning_rate": 1.630619534425422e-06, + "loss": 0.2623, + "step": 37110 + }, + { + "epoch": 13.7, + "learning_rate": 1.6305825840620765e-06, + "loss": 0.1734, + "step": 37111 + }, + { + "epoch": 13.7, + "learning_rate": 1.6305456336987314e-06, + "loss": 0.1687, + "step": 37112 + }, + { + "epoch": 13.7, + "learning_rate": 1.630508683335386e-06, + "loss": 0.3359, + "step": 37113 + }, + { + "epoch": 13.71, + "learning_rate": 1.630471732972041e-06, + "loss": 0.3831, + "step": 37114 + }, + { + "epoch": 13.71, + "learning_rate": 1.6304347826086955e-06, + "loss": 0.1933, + "step": 37115 + }, + { + "epoch": 13.71, + "learning_rate": 1.6303978322453506e-06, + "loss": 0.1931, + "step": 37116 + }, + { + "epoch": 13.71, + "learning_rate": 1.6303608818820052e-06, + "loss": 0.3442, + "step": 37117 + }, + { + "epoch": 13.71, + "learning_rate": 1.63032393151866e-06, + "loss": 0.2319, + "step": 37118 + }, + { + "epoch": 13.71, + "learning_rate": 1.6302869811553147e-06, + "loss": 0.3544, + "step": 37119 + }, + { + "epoch": 13.71, + "learning_rate": 1.6302500307919698e-06, + "loss": 0.3038, + "step": 37120 + }, + { + "epoch": 13.71, + "learning_rate": 1.6302130804286242e-06, + "loss": 0.3165, + "step": 37121 + }, + { + "epoch": 13.71, + "learning_rate": 1.6301761300652788e-06, + "loss": 0.3773, + "step": 37122 + }, + { + "epoch": 13.71, + "learning_rate": 1.6301391797019339e-06, + "loss": 0.3215, + "step": 37123 + }, + { + "epoch": 13.71, + "learning_rate": 1.6301022293385885e-06, + "loss": 0.3067, + "step": 37124 + }, + { + "epoch": 13.71, + "learning_rate": 1.6300652789752434e-06, + "loss": 0.389, + "step": 37125 + }, + { + "epoch": 13.71, + "learning_rate": 1.630028328611898e-06, + "loss": 0.214, + "step": 37126 + }, + { + "epoch": 13.71, + "learning_rate": 1.6299913782485528e-06, + "loss": 0.492, + "step": 37127 + }, + { + "epoch": 13.71, + "learning_rate": 1.6299544278852075e-06, + "loss": 0.3236, + "step": 37128 + }, + { + "epoch": 13.71, + "learning_rate": 1.6299174775218625e-06, + "loss": 0.3067, + "step": 37129 + }, + { + "epoch": 13.71, + "learning_rate": 1.6298805271585172e-06, + "loss": 0.2463, + "step": 37130 + }, + { + "epoch": 13.71, + "learning_rate": 1.6298435767951718e-06, + "loss": 0.2554, + "step": 37131 + }, + { + "epoch": 13.71, + "learning_rate": 1.6298066264318267e-06, + "loss": 0.2103, + "step": 37132 + }, + { + "epoch": 13.71, + "learning_rate": 1.6297696760684813e-06, + "loss": 0.2994, + "step": 37133 + }, + { + "epoch": 13.71, + "learning_rate": 1.6297327257051361e-06, + "loss": 0.3199, + "step": 37134 + }, + { + "epoch": 13.71, + "learning_rate": 1.6296957753417908e-06, + "loss": 0.2224, + "step": 37135 + }, + { + "epoch": 13.71, + "learning_rate": 1.6296588249784458e-06, + "loss": 0.2801, + "step": 37136 + }, + { + "epoch": 13.71, + "learning_rate": 1.6296218746151005e-06, + "loss": 0.3233, + "step": 37137 + }, + { + "epoch": 13.71, + "learning_rate": 1.6295849242517553e-06, + "loss": 0.3181, + "step": 37138 + }, + { + "epoch": 13.71, + "learning_rate": 1.62954797388841e-06, + "loss": 0.3726, + "step": 37139 + }, + { + "epoch": 13.71, + "learning_rate": 1.6295110235250648e-06, + "loss": 0.2901, + "step": 37140 + }, + { + "epoch": 13.72, + "learning_rate": 1.6294740731617194e-06, + "loss": 0.2033, + "step": 37141 + }, + { + "epoch": 13.72, + "learning_rate": 1.629437122798374e-06, + "loss": 0.3227, + "step": 37142 + }, + { + "epoch": 13.72, + "learning_rate": 1.629400172435029e-06, + "loss": 0.4389, + "step": 37143 + }, + { + "epoch": 13.72, + "learning_rate": 1.6293632220716836e-06, + "loss": 0.2619, + "step": 37144 + }, + { + "epoch": 13.72, + "learning_rate": 1.6293262717083386e-06, + "loss": 0.3034, + "step": 37145 + }, + { + "epoch": 13.72, + "learning_rate": 1.6292893213449933e-06, + "loss": 0.3281, + "step": 37146 + }, + { + "epoch": 13.72, + "learning_rate": 1.6292523709816481e-06, + "loss": 0.2821, + "step": 37147 + }, + { + "epoch": 13.72, + "learning_rate": 1.6292154206183027e-06, + "loss": 0.3196, + "step": 37148 + }, + { + "epoch": 13.72, + "learning_rate": 1.6291784702549576e-06, + "loss": 0.2679, + "step": 37149 + }, + { + "epoch": 13.72, + "learning_rate": 1.6291415198916122e-06, + "loss": 0.2353, + "step": 37150 + }, + { + "epoch": 13.72, + "learning_rate": 1.6291045695282669e-06, + "loss": 0.3753, + "step": 37151 + }, + { + "epoch": 13.72, + "learning_rate": 1.629067619164922e-06, + "loss": 0.2269, + "step": 37152 + }, + { + "epoch": 13.72, + "learning_rate": 1.6290306688015766e-06, + "loss": 0.2173, + "step": 37153 + }, + { + "epoch": 13.72, + "learning_rate": 1.6289937184382314e-06, + "loss": 0.4901, + "step": 37154 + }, + { + "epoch": 13.72, + "learning_rate": 1.628956768074886e-06, + "loss": 0.2497, + "step": 37155 + }, + { + "epoch": 13.72, + "learning_rate": 1.628919817711541e-06, + "loss": 0.2856, + "step": 37156 + }, + { + "epoch": 13.72, + "learning_rate": 1.6288828673481955e-06, + "loss": 0.2479, + "step": 37157 + }, + { + "epoch": 13.72, + "learning_rate": 1.6288459169848506e-06, + "loss": 0.3185, + "step": 37158 + }, + { + "epoch": 13.72, + "learning_rate": 1.6288089666215052e-06, + "loss": 0.2873, + "step": 37159 + }, + { + "epoch": 13.72, + "learning_rate": 1.62877201625816e-06, + "loss": 0.3418, + "step": 37160 + }, + { + "epoch": 13.72, + "learning_rate": 1.6287350658948147e-06, + "loss": 0.3925, + "step": 37161 + }, + { + "epoch": 13.72, + "learning_rate": 1.6286981155314694e-06, + "loss": 0.2367, + "step": 37162 + }, + { + "epoch": 13.72, + "learning_rate": 1.6286611651681242e-06, + "loss": 0.3381, + "step": 37163 + }, + { + "epoch": 13.72, + "learning_rate": 1.6286242148047788e-06, + "loss": 0.3303, + "step": 37164 + }, + { + "epoch": 13.72, + "learning_rate": 1.628587264441434e-06, + "loss": 0.2478, + "step": 37165 + }, + { + "epoch": 13.72, + "learning_rate": 1.6285503140780883e-06, + "loss": 0.2656, + "step": 37166 + }, + { + "epoch": 13.72, + "learning_rate": 1.6285133637147434e-06, + "loss": 0.2868, + "step": 37167 + }, + { + "epoch": 13.73, + "learning_rate": 1.628476413351398e-06, + "loss": 0.2147, + "step": 37168 + }, + { + "epoch": 13.73, + "learning_rate": 1.6284394629880529e-06, + "loss": 0.362, + "step": 37169 + }, + { + "epoch": 13.73, + "learning_rate": 1.6284025126247075e-06, + "loss": 0.3018, + "step": 37170 + }, + { + "epoch": 13.73, + "learning_rate": 1.6283655622613621e-06, + "loss": 0.4344, + "step": 37171 + }, + { + "epoch": 13.73, + "learning_rate": 1.628328611898017e-06, + "loss": 0.3416, + "step": 37172 + }, + { + "epoch": 13.73, + "learning_rate": 1.6282916615346716e-06, + "loss": 0.3324, + "step": 37173 + }, + { + "epoch": 13.73, + "learning_rate": 1.6282547111713267e-06, + "loss": 0.2342, + "step": 37174 + }, + { + "epoch": 13.73, + "learning_rate": 1.6282177608079813e-06, + "loss": 0.2819, + "step": 37175 + }, + { + "epoch": 13.73, + "learning_rate": 1.6281808104446362e-06, + "loss": 0.3839, + "step": 37176 + }, + { + "epoch": 13.73, + "learning_rate": 1.6281438600812908e-06, + "loss": 0.3411, + "step": 37177 + }, + { + "epoch": 13.73, + "learning_rate": 1.6281069097179457e-06, + "loss": 0.1599, + "step": 37178 + }, + { + "epoch": 13.73, + "learning_rate": 1.6280699593546003e-06, + "loss": 0.4227, + "step": 37179 + }, + { + "epoch": 13.73, + "learning_rate": 1.628033008991255e-06, + "loss": 0.296, + "step": 37180 + }, + { + "epoch": 13.73, + "learning_rate": 1.62799605862791e-06, + "loss": 0.3096, + "step": 37181 + }, + { + "epoch": 13.73, + "learning_rate": 1.6279591082645644e-06, + "loss": 0.3474, + "step": 37182 + }, + { + "epoch": 13.73, + "learning_rate": 1.6279221579012195e-06, + "loss": 0.2408, + "step": 37183 + }, + { + "epoch": 13.73, + "learning_rate": 1.627885207537874e-06, + "loss": 0.3771, + "step": 37184 + }, + { + "epoch": 13.73, + "learning_rate": 1.627848257174529e-06, + "loss": 0.262, + "step": 37185 + }, + { + "epoch": 13.73, + "learning_rate": 1.6278113068111836e-06, + "loss": 0.4159, + "step": 37186 + }, + { + "epoch": 13.73, + "learning_rate": 1.6277743564478386e-06, + "loss": 0.2115, + "step": 37187 + }, + { + "epoch": 13.73, + "learning_rate": 1.627737406084493e-06, + "loss": 0.3631, + "step": 37188 + }, + { + "epoch": 13.73, + "learning_rate": 1.6277004557211481e-06, + "loss": 0.2922, + "step": 37189 + }, + { + "epoch": 13.73, + "learning_rate": 1.6276635053578028e-06, + "loss": 0.2921, + "step": 37190 + }, + { + "epoch": 13.73, + "learning_rate": 1.6276265549944574e-06, + "loss": 0.2457, + "step": 37191 + }, + { + "epoch": 13.73, + "learning_rate": 1.6275896046311123e-06, + "loss": 0.183, + "step": 37192 + }, + { + "epoch": 13.73, + "learning_rate": 1.6275526542677669e-06, + "loss": 0.2259, + "step": 37193 + }, + { + "epoch": 13.73, + "learning_rate": 1.6275157039044217e-06, + "loss": 0.4423, + "step": 37194 + }, + { + "epoch": 13.74, + "learning_rate": 1.6274787535410764e-06, + "loss": 0.2847, + "step": 37195 + }, + { + "epoch": 13.74, + "learning_rate": 1.6274418031777314e-06, + "loss": 0.2729, + "step": 37196 + }, + { + "epoch": 13.74, + "learning_rate": 1.627404852814386e-06, + "loss": 0.3294, + "step": 37197 + }, + { + "epoch": 13.74, + "learning_rate": 1.627367902451041e-06, + "loss": 0.1629, + "step": 37198 + }, + { + "epoch": 13.74, + "learning_rate": 1.6273309520876956e-06, + "loss": 0.2778, + "step": 37199 + }, + { + "epoch": 13.74, + "learning_rate": 1.6272940017243502e-06, + "loss": 0.294, + "step": 37200 + }, + { + "epoch": 13.74, + "learning_rate": 1.627257051361005e-06, + "loss": 0.2904, + "step": 37201 + }, + { + "epoch": 13.74, + "learning_rate": 1.6272201009976597e-06, + "loss": 0.2702, + "step": 37202 + }, + { + "epoch": 13.74, + "learning_rate": 1.6271831506343147e-06, + "loss": 0.27, + "step": 37203 + }, + { + "epoch": 13.74, + "learning_rate": 1.6271462002709694e-06, + "loss": 0.2835, + "step": 37204 + }, + { + "epoch": 13.74, + "learning_rate": 1.6271092499076242e-06, + "loss": 0.1833, + "step": 37205 + }, + { + "epoch": 13.74, + "learning_rate": 1.6270722995442789e-06, + "loss": 0.3108, + "step": 37206 + }, + { + "epoch": 13.74, + "learning_rate": 1.6270353491809337e-06, + "loss": 0.277, + "step": 37207 + }, + { + "epoch": 13.74, + "learning_rate": 1.6269983988175883e-06, + "loss": 0.2474, + "step": 37208 + }, + { + "epoch": 13.74, + "learning_rate": 1.6269614484542434e-06, + "loss": 0.3268, + "step": 37209 + }, + { + "epoch": 13.74, + "learning_rate": 1.626924498090898e-06, + "loss": 0.3143, + "step": 37210 + }, + { + "epoch": 13.74, + "learning_rate": 1.6268875477275525e-06, + "loss": 0.2291, + "step": 37211 + }, + { + "epoch": 13.74, + "learning_rate": 1.6268505973642075e-06, + "loss": 0.2387, + "step": 37212 + }, + { + "epoch": 13.74, + "learning_rate": 1.6268136470008622e-06, + "loss": 0.3715, + "step": 37213 + }, + { + "epoch": 13.74, + "learning_rate": 1.626776696637517e-06, + "loss": 0.435, + "step": 37214 + }, + { + "epoch": 13.74, + "learning_rate": 1.6267397462741716e-06, + "loss": 0.383, + "step": 37215 + }, + { + "epoch": 13.74, + "learning_rate": 1.6267027959108267e-06, + "loss": 0.3326, + "step": 37216 + }, + { + "epoch": 13.74, + "learning_rate": 1.6266658455474811e-06, + "loss": 0.2794, + "step": 37217 + }, + { + "epoch": 13.74, + "learning_rate": 1.6266288951841362e-06, + "loss": 0.2846, + "step": 37218 + }, + { + "epoch": 13.74, + "learning_rate": 1.6265919448207908e-06, + "loss": 0.3554, + "step": 37219 + }, + { + "epoch": 13.74, + "learning_rate": 1.6265549944574455e-06, + "loss": 0.3042, + "step": 37220 + }, + { + "epoch": 13.74, + "learning_rate": 1.6265180440941003e-06, + "loss": 0.2822, + "step": 37221 + }, + { + "epoch": 13.75, + "learning_rate": 1.626481093730755e-06, + "loss": 0.2842, + "step": 37222 + }, + { + "epoch": 13.75, + "learning_rate": 1.6264441433674098e-06, + "loss": 0.3371, + "step": 37223 + }, + { + "epoch": 13.75, + "learning_rate": 1.6264071930040644e-06, + "loss": 0.216, + "step": 37224 + }, + { + "epoch": 13.75, + "learning_rate": 1.6263702426407195e-06, + "loss": 0.3514, + "step": 37225 + }, + { + "epoch": 13.75, + "learning_rate": 1.6263332922773741e-06, + "loss": 0.3518, + "step": 37226 + }, + { + "epoch": 13.75, + "learning_rate": 1.626296341914029e-06, + "loss": 0.3145, + "step": 37227 + }, + { + "epoch": 13.75, + "learning_rate": 1.6262593915506836e-06, + "loss": 0.2783, + "step": 37228 + }, + { + "epoch": 13.75, + "learning_rate": 1.6262224411873385e-06, + "loss": 0.3479, + "step": 37229 + }, + { + "epoch": 13.75, + "learning_rate": 1.626185490823993e-06, + "loss": 0.3142, + "step": 37230 + }, + { + "epoch": 13.75, + "learning_rate": 1.6261485404606477e-06, + "loss": 0.2766, + "step": 37231 + }, + { + "epoch": 13.75, + "learning_rate": 1.6261115900973028e-06, + "loss": 0.3653, + "step": 37232 + }, + { + "epoch": 13.75, + "learning_rate": 1.6260746397339572e-06, + "loss": 0.3708, + "step": 37233 + }, + { + "epoch": 13.75, + "learning_rate": 1.6260376893706123e-06, + "loss": 0.2155, + "step": 37234 + }, + { + "epoch": 13.75, + "learning_rate": 1.626000739007267e-06, + "loss": 0.2725, + "step": 37235 + }, + { + "epoch": 13.75, + "learning_rate": 1.6259637886439218e-06, + "loss": 0.4551, + "step": 37236 + }, + { + "epoch": 13.75, + "learning_rate": 1.6259268382805764e-06, + "loss": 0.2844, + "step": 37237 + }, + { + "epoch": 13.75, + "learning_rate": 1.6258898879172315e-06, + "loss": 0.2965, + "step": 37238 + }, + { + "epoch": 13.75, + "learning_rate": 1.6258529375538859e-06, + "loss": 0.3195, + "step": 37239 + }, + { + "epoch": 13.75, + "learning_rate": 1.6258159871905405e-06, + "loss": 0.3106, + "step": 37240 + }, + { + "epoch": 13.75, + "learning_rate": 1.6257790368271956e-06, + "loss": 0.2691, + "step": 37241 + }, + { + "epoch": 13.75, + "learning_rate": 1.6257420864638502e-06, + "loss": 0.3163, + "step": 37242 + }, + { + "epoch": 13.75, + "learning_rate": 1.625705136100505e-06, + "loss": 0.2696, + "step": 37243 + }, + { + "epoch": 13.75, + "learning_rate": 1.6256681857371597e-06, + "loss": 0.3112, + "step": 37244 + }, + { + "epoch": 13.75, + "learning_rate": 1.6256312353738146e-06, + "loss": 0.3092, + "step": 37245 + }, + { + "epoch": 13.75, + "learning_rate": 1.6255942850104692e-06, + "loss": 0.277, + "step": 37246 + }, + { + "epoch": 13.75, + "learning_rate": 1.6255573346471242e-06, + "loss": 0.2702, + "step": 37247 + }, + { + "epoch": 13.75, + "learning_rate": 1.6255203842837789e-06, + "loss": 0.1583, + "step": 37248 + }, + { + "epoch": 13.76, + "learning_rate": 1.6254834339204337e-06, + "loss": 0.3466, + "step": 37249 + }, + { + "epoch": 13.76, + "learning_rate": 1.6254464835570884e-06, + "loss": 0.4114, + "step": 37250 + }, + { + "epoch": 13.76, + "learning_rate": 1.625409533193743e-06, + "loss": 0.2063, + "step": 37251 + }, + { + "epoch": 13.76, + "learning_rate": 1.6253725828303979e-06, + "loss": 0.2, + "step": 37252 + }, + { + "epoch": 13.76, + "learning_rate": 1.6253356324670525e-06, + "loss": 0.2435, + "step": 37253 + }, + { + "epoch": 13.76, + "learning_rate": 1.6252986821037075e-06, + "loss": 0.2425, + "step": 37254 + }, + { + "epoch": 13.76, + "learning_rate": 1.625261731740362e-06, + "loss": 0.1815, + "step": 37255 + }, + { + "epoch": 13.76, + "learning_rate": 1.625224781377017e-06, + "loss": 0.3869, + "step": 37256 + }, + { + "epoch": 13.76, + "learning_rate": 1.6251878310136717e-06, + "loss": 0.2476, + "step": 37257 + }, + { + "epoch": 13.76, + "learning_rate": 1.6251508806503265e-06, + "loss": 0.3603, + "step": 37258 + }, + { + "epoch": 13.76, + "learning_rate": 1.6251139302869812e-06, + "loss": 0.2249, + "step": 37259 + }, + { + "epoch": 13.76, + "learning_rate": 1.6250769799236358e-06, + "loss": 0.3432, + "step": 37260 + }, + { + "epoch": 13.76, + "learning_rate": 1.6250400295602906e-06, + "loss": 0.2355, + "step": 37261 + }, + { + "epoch": 13.76, + "learning_rate": 1.6250030791969453e-06, + "loss": 0.1747, + "step": 37262 + }, + { + "epoch": 13.76, + "learning_rate": 1.6249661288336003e-06, + "loss": 0.3855, + "step": 37263 + }, + { + "epoch": 13.76, + "learning_rate": 1.624929178470255e-06, + "loss": 0.278, + "step": 37264 + }, + { + "epoch": 13.76, + "learning_rate": 1.6248922281069098e-06, + "loss": 0.423, + "step": 37265 + }, + { + "epoch": 13.76, + "learning_rate": 1.6248552777435645e-06, + "loss": 0.3001, + "step": 37266 + }, + { + "epoch": 13.76, + "learning_rate": 1.6248183273802193e-06, + "loss": 0.2103, + "step": 37267 + }, + { + "epoch": 13.76, + "learning_rate": 1.624781377016874e-06, + "loss": 0.2185, + "step": 37268 + }, + { + "epoch": 13.76, + "learning_rate": 1.624744426653529e-06, + "loss": 0.255, + "step": 37269 + }, + { + "epoch": 13.76, + "learning_rate": 1.6247074762901836e-06, + "loss": 0.2621, + "step": 37270 + }, + { + "epoch": 13.76, + "learning_rate": 1.6246705259268383e-06, + "loss": 0.1812, + "step": 37271 + }, + { + "epoch": 13.76, + "learning_rate": 1.6246335755634931e-06, + "loss": 0.2383, + "step": 37272 + }, + { + "epoch": 13.76, + "learning_rate": 1.6245966252001478e-06, + "loss": 0.2036, + "step": 37273 + }, + { + "epoch": 13.76, + "learning_rate": 1.6245596748368026e-06, + "loss": 0.2643, + "step": 37274 + }, + { + "epoch": 13.76, + "learning_rate": 1.6245227244734572e-06, + "loss": 0.3218, + "step": 37275 + }, + { + "epoch": 13.77, + "learning_rate": 1.6244857741101123e-06, + "loss": 0.3812, + "step": 37276 + }, + { + "epoch": 13.77, + "learning_rate": 1.624448823746767e-06, + "loss": 0.3778, + "step": 37277 + }, + { + "epoch": 13.77, + "learning_rate": 1.6244118733834218e-06, + "loss": 0.3215, + "step": 37278 + }, + { + "epoch": 13.77, + "learning_rate": 1.6243749230200764e-06, + "loss": 0.3615, + "step": 37279 + }, + { + "epoch": 13.77, + "learning_rate": 1.624337972656731e-06, + "loss": 0.443, + "step": 37280 + }, + { + "epoch": 13.77, + "learning_rate": 1.624301022293386e-06, + "loss": 0.2487, + "step": 37281 + }, + { + "epoch": 13.77, + "learning_rate": 1.6242640719300405e-06, + "loss": 0.3571, + "step": 37282 + }, + { + "epoch": 13.77, + "learning_rate": 1.6242271215666956e-06, + "loss": 0.2735, + "step": 37283 + }, + { + "epoch": 13.77, + "learning_rate": 1.62419017120335e-06, + "loss": 0.3107, + "step": 37284 + }, + { + "epoch": 13.77, + "learning_rate": 1.624153220840005e-06, + "loss": 0.3905, + "step": 37285 + }, + { + "epoch": 13.77, + "learning_rate": 1.6241162704766597e-06, + "loss": 0.2521, + "step": 37286 + }, + { + "epoch": 13.77, + "learning_rate": 1.6240793201133146e-06, + "loss": 0.514, + "step": 37287 + }, + { + "epoch": 13.77, + "learning_rate": 1.6240423697499692e-06, + "loss": 0.3683, + "step": 37288 + }, + { + "epoch": 13.77, + "learning_rate": 1.6240054193866243e-06, + "loss": 0.3215, + "step": 37289 + }, + { + "epoch": 13.77, + "learning_rate": 1.6239684690232787e-06, + "loss": 0.2051, + "step": 37290 + }, + { + "epoch": 13.77, + "learning_rate": 1.6239315186599333e-06, + "loss": 0.2353, + "step": 37291 + }, + { + "epoch": 13.77, + "learning_rate": 1.6238945682965884e-06, + "loss": 0.2517, + "step": 37292 + }, + { + "epoch": 13.77, + "learning_rate": 1.623857617933243e-06, + "loss": 0.3016, + "step": 37293 + }, + { + "epoch": 13.77, + "learning_rate": 1.6238206675698979e-06, + "loss": 0.3018, + "step": 37294 + }, + { + "epoch": 13.77, + "learning_rate": 1.6237837172065525e-06, + "loss": 0.2848, + "step": 37295 + }, + { + "epoch": 13.77, + "learning_rate": 1.6237467668432074e-06, + "loss": 0.2609, + "step": 37296 + }, + { + "epoch": 13.77, + "learning_rate": 1.623709816479862e-06, + "loss": 0.3167, + "step": 37297 + }, + { + "epoch": 13.77, + "learning_rate": 1.623672866116517e-06, + "loss": 0.2374, + "step": 37298 + }, + { + "epoch": 13.77, + "learning_rate": 1.6236359157531717e-06, + "loss": 0.2054, + "step": 37299 + }, + { + "epoch": 13.77, + "learning_rate": 1.6235989653898261e-06, + "loss": 0.3247, + "step": 37300 + }, + { + "epoch": 13.77, + "learning_rate": 1.6235620150264812e-06, + "loss": 0.2808, + "step": 37301 + }, + { + "epoch": 13.77, + "learning_rate": 1.6235250646631358e-06, + "loss": 0.3282, + "step": 37302 + }, + { + "epoch": 13.78, + "learning_rate": 1.6234881142997907e-06, + "loss": 0.2729, + "step": 37303 + }, + { + "epoch": 13.78, + "learning_rate": 1.6234511639364453e-06, + "loss": 0.3544, + "step": 37304 + }, + { + "epoch": 13.78, + "learning_rate": 1.6234142135731004e-06, + "loss": 0.3203, + "step": 37305 + }, + { + "epoch": 13.78, + "learning_rate": 1.6233772632097548e-06, + "loss": 0.3366, + "step": 37306 + }, + { + "epoch": 13.78, + "learning_rate": 1.6233403128464098e-06, + "loss": 0.1773, + "step": 37307 + }, + { + "epoch": 13.78, + "learning_rate": 1.6233033624830645e-06, + "loss": 0.2506, + "step": 37308 + }, + { + "epoch": 13.78, + "learning_rate": 1.6232664121197193e-06, + "loss": 0.2444, + "step": 37309 + }, + { + "epoch": 13.78, + "learning_rate": 1.623229461756374e-06, + "loss": 0.3197, + "step": 37310 + }, + { + "epoch": 13.78, + "learning_rate": 1.6231925113930286e-06, + "loss": 0.3908, + "step": 37311 + }, + { + "epoch": 13.78, + "learning_rate": 1.6231555610296835e-06, + "loss": 0.3256, + "step": 37312 + }, + { + "epoch": 13.78, + "learning_rate": 1.623118610666338e-06, + "loss": 0.249, + "step": 37313 + }, + { + "epoch": 13.78, + "learning_rate": 1.6230816603029931e-06, + "loss": 0.3012, + "step": 37314 + }, + { + "epoch": 13.78, + "learning_rate": 1.6230447099396478e-06, + "loss": 0.3373, + "step": 37315 + }, + { + "epoch": 13.78, + "learning_rate": 1.6230077595763026e-06, + "loss": 0.3368, + "step": 37316 + }, + { + "epoch": 13.78, + "learning_rate": 1.6229708092129573e-06, + "loss": 0.1576, + "step": 37317 + }, + { + "epoch": 13.78, + "learning_rate": 1.6229338588496121e-06, + "loss": 0.2199, + "step": 37318 + }, + { + "epoch": 13.78, + "learning_rate": 1.6228969084862668e-06, + "loss": 0.2755, + "step": 37319 + }, + { + "epoch": 13.78, + "learning_rate": 1.6228599581229214e-06, + "loss": 0.2829, + "step": 37320 + }, + { + "epoch": 13.78, + "learning_rate": 1.6228230077595764e-06, + "loss": 0.2899, + "step": 37321 + }, + { + "epoch": 13.78, + "learning_rate": 1.622786057396231e-06, + "loss": 0.3095, + "step": 37322 + }, + { + "epoch": 13.78, + "learning_rate": 1.622749107032886e-06, + "loss": 0.2526, + "step": 37323 + }, + { + "epoch": 13.78, + "learning_rate": 1.6227121566695406e-06, + "loss": 0.2153, + "step": 37324 + }, + { + "epoch": 13.78, + "learning_rate": 1.6226752063061954e-06, + "loss": 0.3591, + "step": 37325 + }, + { + "epoch": 13.78, + "learning_rate": 1.62263825594285e-06, + "loss": 0.1881, + "step": 37326 + }, + { + "epoch": 13.78, + "learning_rate": 1.6226013055795051e-06, + "loss": 0.3074, + "step": 37327 + }, + { + "epoch": 13.78, + "learning_rate": 1.6225643552161595e-06, + "loss": 0.3675, + "step": 37328 + }, + { + "epoch": 13.78, + "learning_rate": 1.6225274048528146e-06, + "loss": 0.1776, + "step": 37329 + }, + { + "epoch": 13.79, + "learning_rate": 1.6224904544894692e-06, + "loss": 0.2153, + "step": 37330 + }, + { + "epoch": 13.79, + "learning_rate": 1.6224535041261239e-06, + "loss": 0.3355, + "step": 37331 + }, + { + "epoch": 13.79, + "learning_rate": 1.6224165537627787e-06, + "loss": 0.2531, + "step": 37332 + }, + { + "epoch": 13.79, + "learning_rate": 1.6223796033994334e-06, + "loss": 0.3594, + "step": 37333 + }, + { + "epoch": 13.79, + "learning_rate": 1.6223426530360882e-06, + "loss": 0.2569, + "step": 37334 + }, + { + "epoch": 13.79, + "learning_rate": 1.6223057026727428e-06, + "loss": 0.2602, + "step": 37335 + }, + { + "epoch": 13.79, + "learning_rate": 1.622268752309398e-06, + "loss": 0.3109, + "step": 37336 + }, + { + "epoch": 13.79, + "learning_rate": 1.6222318019460525e-06, + "loss": 0.3848, + "step": 37337 + }, + { + "epoch": 13.79, + "learning_rate": 1.6221948515827074e-06, + "loss": 0.3158, + "step": 37338 + }, + { + "epoch": 13.79, + "learning_rate": 1.622157901219362e-06, + "loss": 0.2298, + "step": 37339 + }, + { + "epoch": 13.79, + "learning_rate": 1.6221209508560167e-06, + "loss": 0.3498, + "step": 37340 + }, + { + "epoch": 13.79, + "learning_rate": 1.6220840004926715e-06, + "loss": 0.359, + "step": 37341 + }, + { + "epoch": 13.79, + "learning_rate": 1.6220470501293261e-06, + "loss": 0.212, + "step": 37342 + }, + { + "epoch": 13.79, + "learning_rate": 1.6220100997659812e-06, + "loss": 0.261, + "step": 37343 + }, + { + "epoch": 13.79, + "learning_rate": 1.6219731494026358e-06, + "loss": 0.302, + "step": 37344 + }, + { + "epoch": 13.79, + "learning_rate": 1.6219361990392907e-06, + "loss": 0.313, + "step": 37345 + }, + { + "epoch": 13.79, + "learning_rate": 1.6218992486759453e-06, + "loss": 0.4617, + "step": 37346 + }, + { + "epoch": 13.79, + "learning_rate": 1.6218622983126002e-06, + "loss": 0.2574, + "step": 37347 + }, + { + "epoch": 13.79, + "learning_rate": 1.6218253479492548e-06, + "loss": 0.361, + "step": 37348 + }, + { + "epoch": 13.79, + "learning_rate": 1.6217883975859099e-06, + "loss": 0.1852, + "step": 37349 + }, + { + "epoch": 13.79, + "learning_rate": 1.6217514472225645e-06, + "loss": 0.3555, + "step": 37350 + }, + { + "epoch": 13.79, + "learning_rate": 1.621714496859219e-06, + "loss": 0.3846, + "step": 37351 + }, + { + "epoch": 13.79, + "learning_rate": 1.621677546495874e-06, + "loss": 0.2214, + "step": 37352 + }, + { + "epoch": 13.79, + "learning_rate": 1.6216405961325286e-06, + "loss": 0.2445, + "step": 37353 + }, + { + "epoch": 13.79, + "learning_rate": 1.6216036457691835e-06, + "loss": 0.1893, + "step": 37354 + }, + { + "epoch": 13.79, + "learning_rate": 1.6215666954058381e-06, + "loss": 0.3106, + "step": 37355 + }, + { + "epoch": 13.79, + "learning_rate": 1.6215297450424932e-06, + "loss": 0.2057, + "step": 37356 + }, + { + "epoch": 13.8, + "learning_rate": 1.6214927946791476e-06, + "loss": 0.2487, + "step": 37357 + }, + { + "epoch": 13.8, + "learning_rate": 1.6214558443158027e-06, + "loss": 0.2098, + "step": 37358 + }, + { + "epoch": 13.8, + "learning_rate": 1.6214188939524573e-06, + "loss": 0.4013, + "step": 37359 + }, + { + "epoch": 13.8, + "learning_rate": 1.621381943589112e-06, + "loss": 0.3328, + "step": 37360 + }, + { + "epoch": 13.8, + "learning_rate": 1.6213449932257668e-06, + "loss": 0.4631, + "step": 37361 + }, + { + "epoch": 13.8, + "learning_rate": 1.6213080428624214e-06, + "loss": 0.2993, + "step": 37362 + }, + { + "epoch": 13.8, + "learning_rate": 1.6212710924990763e-06, + "loss": 0.3135, + "step": 37363 + }, + { + "epoch": 13.8, + "learning_rate": 1.621234142135731e-06, + "loss": 0.2767, + "step": 37364 + }, + { + "epoch": 13.8, + "learning_rate": 1.621197191772386e-06, + "loss": 0.3265, + "step": 37365 + }, + { + "epoch": 13.8, + "learning_rate": 1.6211602414090406e-06, + "loss": 0.2725, + "step": 37366 + }, + { + "epoch": 13.8, + "learning_rate": 1.6211232910456954e-06, + "loss": 0.2685, + "step": 37367 + }, + { + "epoch": 13.8, + "learning_rate": 1.62108634068235e-06, + "loss": 0.3687, + "step": 37368 + }, + { + "epoch": 13.8, + "learning_rate": 1.621049390319005e-06, + "loss": 0.4695, + "step": 37369 + }, + { + "epoch": 13.8, + "learning_rate": 1.6210124399556596e-06, + "loss": 0.3399, + "step": 37370 + }, + { + "epoch": 13.8, + "learning_rate": 1.6209754895923142e-06, + "loss": 0.24, + "step": 37371 + }, + { + "epoch": 13.8, + "learning_rate": 1.6209385392289693e-06, + "loss": 0.2374, + "step": 37372 + }, + { + "epoch": 13.8, + "learning_rate": 1.6209015888656237e-06, + "loss": 0.2543, + "step": 37373 + }, + { + "epoch": 13.8, + "learning_rate": 1.6208646385022787e-06, + "loss": 0.2869, + "step": 37374 + }, + { + "epoch": 13.8, + "learning_rate": 1.6208276881389334e-06, + "loss": 0.3041, + "step": 37375 + }, + { + "epoch": 13.8, + "learning_rate": 1.6207907377755882e-06, + "loss": 0.4749, + "step": 37376 + }, + { + "epoch": 13.8, + "learning_rate": 1.6207537874122429e-06, + "loss": 0.3236, + "step": 37377 + }, + { + "epoch": 13.8, + "learning_rate": 1.620716837048898e-06, + "loss": 0.3605, + "step": 37378 + }, + { + "epoch": 13.8, + "learning_rate": 1.6206798866855524e-06, + "loss": 0.3011, + "step": 37379 + }, + { + "epoch": 13.8, + "learning_rate": 1.620642936322207e-06, + "loss": 0.3374, + "step": 37380 + }, + { + "epoch": 13.8, + "learning_rate": 1.620605985958862e-06, + "loss": 0.262, + "step": 37381 + }, + { + "epoch": 13.8, + "learning_rate": 1.6205690355955167e-06, + "loss": 0.2851, + "step": 37382 + }, + { + "epoch": 13.8, + "learning_rate": 1.6205320852321715e-06, + "loss": 0.2411, + "step": 37383 + }, + { + "epoch": 13.81, + "learning_rate": 1.6204951348688262e-06, + "loss": 0.3768, + "step": 37384 + }, + { + "epoch": 13.81, + "learning_rate": 1.620458184505481e-06, + "loss": 0.2856, + "step": 37385 + }, + { + "epoch": 13.81, + "learning_rate": 1.6204212341421357e-06, + "loss": 0.3091, + "step": 37386 + }, + { + "epoch": 13.81, + "learning_rate": 1.6203842837787907e-06, + "loss": 0.1868, + "step": 37387 + }, + { + "epoch": 13.81, + "learning_rate": 1.6203473334154453e-06, + "loss": 0.2557, + "step": 37388 + }, + { + "epoch": 13.81, + "learning_rate": 1.6203103830521002e-06, + "loss": 0.253, + "step": 37389 + }, + { + "epoch": 13.81, + "learning_rate": 1.6202734326887548e-06, + "loss": 0.3995, + "step": 37390 + }, + { + "epoch": 13.81, + "learning_rate": 1.6202364823254095e-06, + "loss": 0.1746, + "step": 37391 + }, + { + "epoch": 13.81, + "learning_rate": 1.6201995319620643e-06, + "loss": 0.2368, + "step": 37392 + }, + { + "epoch": 13.81, + "learning_rate": 1.620162581598719e-06, + "loss": 0.1948, + "step": 37393 + }, + { + "epoch": 13.81, + "learning_rate": 1.620125631235374e-06, + "loss": 0.3268, + "step": 37394 + }, + { + "epoch": 13.81, + "learning_rate": 1.6200886808720287e-06, + "loss": 0.2403, + "step": 37395 + }, + { + "epoch": 13.81, + "learning_rate": 1.6200517305086835e-06, + "loss": 0.237, + "step": 37396 + }, + { + "epoch": 13.81, + "learning_rate": 1.6200147801453381e-06, + "loss": 0.2911, + "step": 37397 + }, + { + "epoch": 13.81, + "learning_rate": 1.619977829781993e-06, + "loss": 0.2949, + "step": 37398 + }, + { + "epoch": 13.81, + "learning_rate": 1.6199408794186476e-06, + "loss": 0.2822, + "step": 37399 + }, + { + "epoch": 13.81, + "learning_rate": 1.6199039290553023e-06, + "loss": 0.3538, + "step": 37400 + }, + { + "epoch": 13.81, + "learning_rate": 1.6198669786919573e-06, + "loss": 0.3608, + "step": 37401 + }, + { + "epoch": 13.81, + "learning_rate": 1.6198300283286117e-06, + "loss": 0.2637, + "step": 37402 + }, + { + "epoch": 13.81, + "learning_rate": 1.6197930779652668e-06, + "loss": 0.2842, + "step": 37403 + }, + { + "epoch": 13.81, + "learning_rate": 1.6197561276019214e-06, + "loss": 0.2246, + "step": 37404 + }, + { + "epoch": 13.81, + "learning_rate": 1.6197191772385763e-06, + "loss": 0.4297, + "step": 37405 + }, + { + "epoch": 13.81, + "learning_rate": 1.619682226875231e-06, + "loss": 0.306, + "step": 37406 + }, + { + "epoch": 13.81, + "learning_rate": 1.6196452765118858e-06, + "loss": 0.1622, + "step": 37407 + }, + { + "epoch": 13.81, + "learning_rate": 1.6196083261485404e-06, + "loss": 0.2533, + "step": 37408 + }, + { + "epoch": 13.81, + "learning_rate": 1.619571375785195e-06, + "loss": 0.2323, + "step": 37409 + }, + { + "epoch": 13.81, + "learning_rate": 1.61953442542185e-06, + "loss": 0.3339, + "step": 37410 + }, + { + "epoch": 13.81, + "learning_rate": 1.6194974750585047e-06, + "loss": 0.3021, + "step": 37411 + }, + { + "epoch": 13.82, + "learning_rate": 1.6194605246951596e-06, + "loss": 0.3438, + "step": 37412 + }, + { + "epoch": 13.82, + "learning_rate": 1.6194235743318142e-06, + "loss": 0.3063, + "step": 37413 + }, + { + "epoch": 13.82, + "learning_rate": 1.619386623968469e-06, + "loss": 0.2446, + "step": 37414 + }, + { + "epoch": 13.82, + "learning_rate": 1.6193496736051237e-06, + "loss": 0.2539, + "step": 37415 + }, + { + "epoch": 13.82, + "learning_rate": 1.6193127232417788e-06, + "loss": 0.2989, + "step": 37416 + }, + { + "epoch": 13.82, + "learning_rate": 1.6192757728784334e-06, + "loss": 0.1565, + "step": 37417 + }, + { + "epoch": 13.82, + "learning_rate": 1.6192388225150883e-06, + "loss": 0.228, + "step": 37418 + }, + { + "epoch": 13.82, + "learning_rate": 1.6192018721517429e-06, + "loss": 0.2508, + "step": 37419 + }, + { + "epoch": 13.82, + "learning_rate": 1.6191649217883975e-06, + "loss": 0.4028, + "step": 37420 + }, + { + "epoch": 13.82, + "learning_rate": 1.6191279714250524e-06, + "loss": 0.3671, + "step": 37421 + }, + { + "epoch": 13.82, + "learning_rate": 1.619091021061707e-06, + "loss": 0.2671, + "step": 37422 + }, + { + "epoch": 13.82, + "learning_rate": 1.619054070698362e-06, + "loss": 0.1957, + "step": 37423 + }, + { + "epoch": 13.82, + "learning_rate": 1.6190171203350165e-06, + "loss": 0.2626, + "step": 37424 + }, + { + "epoch": 13.82, + "learning_rate": 1.6189801699716716e-06, + "loss": 0.357, + "step": 37425 + }, + { + "epoch": 13.82, + "learning_rate": 1.6189432196083262e-06, + "loss": 0.261, + "step": 37426 + }, + { + "epoch": 13.82, + "learning_rate": 1.618906269244981e-06, + "loss": 0.4249, + "step": 37427 + }, + { + "epoch": 13.82, + "learning_rate": 1.6188693188816357e-06, + "loss": 0.2005, + "step": 37428 + }, + { + "epoch": 13.82, + "learning_rate": 1.6188323685182903e-06, + "loss": 0.2488, + "step": 37429 + }, + { + "epoch": 13.82, + "learning_rate": 1.6187954181549452e-06, + "loss": 0.2403, + "step": 37430 + }, + { + "epoch": 13.82, + "learning_rate": 1.6187584677915998e-06, + "loss": 0.3191, + "step": 37431 + }, + { + "epoch": 13.82, + "learning_rate": 1.6187215174282549e-06, + "loss": 0.4657, + "step": 37432 + }, + { + "epoch": 13.82, + "learning_rate": 1.6186845670649095e-06, + "loss": 0.1462, + "step": 37433 + }, + { + "epoch": 13.82, + "learning_rate": 1.6186476167015643e-06, + "loss": 0.2405, + "step": 37434 + }, + { + "epoch": 13.82, + "learning_rate": 1.618610666338219e-06, + "loss": 0.2962, + "step": 37435 + }, + { + "epoch": 13.82, + "learning_rate": 1.6185737159748738e-06, + "loss": 0.1678, + "step": 37436 + }, + { + "epoch": 13.82, + "learning_rate": 1.6185367656115285e-06, + "loss": 0.2807, + "step": 37437 + }, + { + "epoch": 13.82, + "learning_rate": 1.6184998152481835e-06, + "loss": 0.2394, + "step": 37438 + }, + { + "epoch": 13.83, + "learning_rate": 1.6184628648848382e-06, + "loss": 0.3502, + "step": 37439 + }, + { + "epoch": 13.83, + "learning_rate": 1.6184259145214926e-06, + "loss": 0.1593, + "step": 37440 + }, + { + "epoch": 13.83, + "learning_rate": 1.6183889641581476e-06, + "loss": 0.2693, + "step": 37441 + }, + { + "epoch": 13.83, + "learning_rate": 1.6183520137948023e-06, + "loss": 0.2463, + "step": 37442 + }, + { + "epoch": 13.83, + "learning_rate": 1.6183150634314571e-06, + "loss": 0.2905, + "step": 37443 + }, + { + "epoch": 13.83, + "learning_rate": 1.6182781130681118e-06, + "loss": 0.2832, + "step": 37444 + }, + { + "epoch": 13.83, + "learning_rate": 1.6182411627047668e-06, + "loss": 0.3173, + "step": 37445 + }, + { + "epoch": 13.83, + "learning_rate": 1.6182042123414213e-06, + "loss": 0.3732, + "step": 37446 + }, + { + "epoch": 13.83, + "learning_rate": 1.6181672619780763e-06, + "loss": 0.2087, + "step": 37447 + }, + { + "epoch": 13.83, + "learning_rate": 1.618130311614731e-06, + "loss": 0.2504, + "step": 37448 + }, + { + "epoch": 13.83, + "learning_rate": 1.6180933612513856e-06, + "loss": 0.2928, + "step": 37449 + }, + { + "epoch": 13.83, + "learning_rate": 1.6180564108880404e-06, + "loss": 0.2681, + "step": 37450 + }, + { + "epoch": 13.83, + "learning_rate": 1.618019460524695e-06, + "loss": 0.3007, + "step": 37451 + }, + { + "epoch": 13.83, + "learning_rate": 1.61798251016135e-06, + "loss": 0.3047, + "step": 37452 + }, + { + "epoch": 13.83, + "learning_rate": 1.6179455597980046e-06, + "loss": 0.2567, + "step": 37453 + }, + { + "epoch": 13.83, + "learning_rate": 1.6179086094346596e-06, + "loss": 0.3719, + "step": 37454 + }, + { + "epoch": 13.83, + "learning_rate": 1.6178716590713142e-06, + "loss": 0.4556, + "step": 37455 + }, + { + "epoch": 13.83, + "learning_rate": 1.617834708707969e-06, + "loss": 0.2624, + "step": 37456 + }, + { + "epoch": 13.83, + "learning_rate": 1.6177977583446237e-06, + "loss": 0.2596, + "step": 37457 + }, + { + "epoch": 13.83, + "learning_rate": 1.6177608079812786e-06, + "loss": 0.2596, + "step": 37458 + }, + { + "epoch": 13.83, + "learning_rate": 1.6177238576179332e-06, + "loss": 0.3058, + "step": 37459 + }, + { + "epoch": 13.83, + "learning_rate": 1.6176869072545879e-06, + "loss": 0.2462, + "step": 37460 + }, + { + "epoch": 13.83, + "learning_rate": 1.617649956891243e-06, + "loss": 0.3007, + "step": 37461 + }, + { + "epoch": 13.83, + "learning_rate": 1.6176130065278976e-06, + "loss": 0.3603, + "step": 37462 + }, + { + "epoch": 13.83, + "learning_rate": 1.6175760561645524e-06, + "loss": 0.3417, + "step": 37463 + }, + { + "epoch": 13.83, + "learning_rate": 1.617539105801207e-06, + "loss": 0.2193, + "step": 37464 + }, + { + "epoch": 13.83, + "learning_rate": 1.6175021554378619e-06, + "loss": 0.3303, + "step": 37465 + }, + { + "epoch": 13.84, + "learning_rate": 1.6174652050745165e-06, + "loss": 0.3931, + "step": 37466 + }, + { + "epoch": 13.84, + "learning_rate": 1.6174282547111716e-06, + "loss": 0.2864, + "step": 37467 + }, + { + "epoch": 13.84, + "learning_rate": 1.6173913043478262e-06, + "loss": 0.183, + "step": 37468 + }, + { + "epoch": 13.84, + "learning_rate": 1.6173543539844806e-06, + "loss": 0.2118, + "step": 37469 + }, + { + "epoch": 13.84, + "learning_rate": 1.6173174036211357e-06, + "loss": 0.3063, + "step": 37470 + }, + { + "epoch": 13.84, + "learning_rate": 1.6172804532577903e-06, + "loss": 0.3195, + "step": 37471 + }, + { + "epoch": 13.84, + "learning_rate": 1.6172435028944452e-06, + "loss": 0.276, + "step": 37472 + }, + { + "epoch": 13.84, + "learning_rate": 1.6172065525310998e-06, + "loss": 0.2156, + "step": 37473 + }, + { + "epoch": 13.84, + "learning_rate": 1.6171696021677549e-06, + "loss": 0.1949, + "step": 37474 + }, + { + "epoch": 13.84, + "learning_rate": 1.6171326518044093e-06, + "loss": 0.2766, + "step": 37475 + }, + { + "epoch": 13.84, + "learning_rate": 1.6170957014410644e-06, + "loss": 0.3659, + "step": 37476 + }, + { + "epoch": 13.84, + "learning_rate": 1.617058751077719e-06, + "loss": 0.2317, + "step": 37477 + }, + { + "epoch": 13.84, + "learning_rate": 1.6170218007143739e-06, + "loss": 0.2939, + "step": 37478 + }, + { + "epoch": 13.84, + "learning_rate": 1.6169848503510285e-06, + "loss": 0.2314, + "step": 37479 + }, + { + "epoch": 13.84, + "learning_rate": 1.6169478999876831e-06, + "loss": 0.2874, + "step": 37480 + }, + { + "epoch": 13.84, + "learning_rate": 1.616910949624338e-06, + "loss": 0.2965, + "step": 37481 + }, + { + "epoch": 13.84, + "learning_rate": 1.6168739992609926e-06, + "loss": 0.2634, + "step": 37482 + }, + { + "epoch": 13.84, + "learning_rate": 1.6168370488976477e-06, + "loss": 0.2046, + "step": 37483 + }, + { + "epoch": 13.84, + "learning_rate": 1.6168000985343023e-06, + "loss": 0.2033, + "step": 37484 + }, + { + "epoch": 13.84, + "learning_rate": 1.6167631481709572e-06, + "loss": 0.3272, + "step": 37485 + }, + { + "epoch": 13.84, + "learning_rate": 1.6167261978076118e-06, + "loss": 0.2536, + "step": 37486 + }, + { + "epoch": 13.84, + "learning_rate": 1.6166892474442666e-06, + "loss": 0.4152, + "step": 37487 + }, + { + "epoch": 13.84, + "learning_rate": 1.6166522970809213e-06, + "loss": 0.3138, + "step": 37488 + }, + { + "epoch": 13.84, + "learning_rate": 1.616615346717576e-06, + "loss": 0.246, + "step": 37489 + }, + { + "epoch": 13.84, + "learning_rate": 1.616578396354231e-06, + "loss": 0.2357, + "step": 37490 + }, + { + "epoch": 13.84, + "learning_rate": 1.6165414459908854e-06, + "loss": 0.2764, + "step": 37491 + }, + { + "epoch": 13.84, + "learning_rate": 1.6165044956275405e-06, + "loss": 0.2787, + "step": 37492 + }, + { + "epoch": 13.85, + "learning_rate": 1.616467545264195e-06, + "loss": 0.3526, + "step": 37493 + }, + { + "epoch": 13.85, + "learning_rate": 1.61643059490085e-06, + "loss": 0.2715, + "step": 37494 + }, + { + "epoch": 13.85, + "learning_rate": 1.6163936445375046e-06, + "loss": 0.3575, + "step": 37495 + }, + { + "epoch": 13.85, + "learning_rate": 1.6163566941741596e-06, + "loss": 0.2142, + "step": 37496 + }, + { + "epoch": 13.85, + "learning_rate": 1.616319743810814e-06, + "loss": 0.3649, + "step": 37497 + }, + { + "epoch": 13.85, + "learning_rate": 1.6162827934474691e-06, + "loss": 0.4003, + "step": 37498 + }, + { + "epoch": 13.85, + "learning_rate": 1.6162458430841238e-06, + "loss": 0.2404, + "step": 37499 + }, + { + "epoch": 13.85, + "learning_rate": 1.6162088927207784e-06, + "loss": 0.3746, + "step": 37500 + }, + { + "epoch": 13.85, + "learning_rate": 1.6161719423574332e-06, + "loss": 0.3372, + "step": 37501 + }, + { + "epoch": 13.85, + "learning_rate": 1.6161349919940879e-06, + "loss": 0.2479, + "step": 37502 + }, + { + "epoch": 13.85, + "learning_rate": 1.6160980416307427e-06, + "loss": 0.3324, + "step": 37503 + }, + { + "epoch": 13.85, + "learning_rate": 1.6160610912673974e-06, + "loss": 0.2947, + "step": 37504 + }, + { + "epoch": 13.85, + "learning_rate": 1.6160241409040524e-06, + "loss": 0.3651, + "step": 37505 + }, + { + "epoch": 13.85, + "learning_rate": 1.615987190540707e-06, + "loss": 0.371, + "step": 37506 + }, + { + "epoch": 13.85, + "learning_rate": 1.615950240177362e-06, + "loss": 0.4228, + "step": 37507 + }, + { + "epoch": 13.85, + "learning_rate": 1.6159132898140165e-06, + "loss": 0.2299, + "step": 37508 + }, + { + "epoch": 13.85, + "learning_rate": 1.6158763394506712e-06, + "loss": 0.2233, + "step": 37509 + }, + { + "epoch": 13.85, + "learning_rate": 1.615839389087326e-06, + "loss": 0.2782, + "step": 37510 + }, + { + "epoch": 13.85, + "learning_rate": 1.6158024387239807e-06, + "loss": 0.1646, + "step": 37511 + }, + { + "epoch": 13.85, + "learning_rate": 1.6157654883606357e-06, + "loss": 0.1763, + "step": 37512 + }, + { + "epoch": 13.85, + "learning_rate": 1.6157285379972904e-06, + "loss": 0.251, + "step": 37513 + }, + { + "epoch": 13.85, + "learning_rate": 1.6156915876339452e-06, + "loss": 0.3961, + "step": 37514 + }, + { + "epoch": 13.85, + "learning_rate": 1.6156546372705998e-06, + "loss": 0.2805, + "step": 37515 + }, + { + "epoch": 13.85, + "learning_rate": 1.6156176869072547e-06, + "loss": 0.3424, + "step": 37516 + }, + { + "epoch": 13.85, + "learning_rate": 1.6155807365439093e-06, + "loss": 0.2615, + "step": 37517 + }, + { + "epoch": 13.85, + "learning_rate": 1.6155437861805644e-06, + "loss": 0.1872, + "step": 37518 + }, + { + "epoch": 13.85, + "learning_rate": 1.6155068358172188e-06, + "loss": 0.2535, + "step": 37519 + }, + { + "epoch": 13.86, + "learning_rate": 1.6154698854538735e-06, + "loss": 0.2908, + "step": 37520 + }, + { + "epoch": 13.86, + "learning_rate": 1.6154329350905285e-06, + "loss": 0.3618, + "step": 37521 + }, + { + "epoch": 13.86, + "learning_rate": 1.6153959847271831e-06, + "loss": 0.42, + "step": 37522 + }, + { + "epoch": 13.86, + "learning_rate": 1.615359034363838e-06, + "loss": 0.1666, + "step": 37523 + }, + { + "epoch": 13.86, + "learning_rate": 1.6153220840004926e-06, + "loss": 0.2642, + "step": 37524 + }, + { + "epoch": 13.86, + "learning_rate": 1.6152851336371475e-06, + "loss": 0.1855, + "step": 37525 + }, + { + "epoch": 13.86, + "learning_rate": 1.6152481832738021e-06, + "loss": 0.1903, + "step": 37526 + }, + { + "epoch": 13.86, + "learning_rate": 1.6152112329104572e-06, + "loss": 0.1812, + "step": 37527 + }, + { + "epoch": 13.86, + "learning_rate": 1.6151742825471118e-06, + "loss": 0.2287, + "step": 37528 + }, + { + "epoch": 13.86, + "learning_rate": 1.6151373321837665e-06, + "loss": 0.2635, + "step": 37529 + }, + { + "epoch": 13.86, + "learning_rate": 1.6151003818204213e-06, + "loss": 0.3949, + "step": 37530 + }, + { + "epoch": 13.86, + "learning_rate": 1.615063431457076e-06, + "loss": 0.3603, + "step": 37531 + }, + { + "epoch": 13.86, + "learning_rate": 1.6150264810937308e-06, + "loss": 0.3264, + "step": 37532 + }, + { + "epoch": 13.86, + "learning_rate": 1.6149895307303854e-06, + "loss": 0.2567, + "step": 37533 + }, + { + "epoch": 13.86, + "learning_rate": 1.6149525803670405e-06, + "loss": 0.3122, + "step": 37534 + }, + { + "epoch": 13.86, + "learning_rate": 1.6149156300036951e-06, + "loss": 0.2897, + "step": 37535 + }, + { + "epoch": 13.86, + "learning_rate": 1.61487867964035e-06, + "loss": 0.3013, + "step": 37536 + }, + { + "epoch": 13.86, + "learning_rate": 1.6148417292770046e-06, + "loss": 0.2185, + "step": 37537 + }, + { + "epoch": 13.86, + "learning_rate": 1.6148047789136594e-06, + "loss": 0.2674, + "step": 37538 + }, + { + "epoch": 13.86, + "learning_rate": 1.614767828550314e-06, + "loss": 0.3663, + "step": 37539 + }, + { + "epoch": 13.86, + "learning_rate": 1.6147308781869687e-06, + "loss": 0.3331, + "step": 37540 + }, + { + "epoch": 13.86, + "learning_rate": 1.6146939278236238e-06, + "loss": 0.1618, + "step": 37541 + }, + { + "epoch": 13.86, + "learning_rate": 1.6146569774602782e-06, + "loss": 0.2575, + "step": 37542 + }, + { + "epoch": 13.86, + "learning_rate": 1.6146200270969333e-06, + "loss": 0.3058, + "step": 37543 + }, + { + "epoch": 13.86, + "learning_rate": 1.614583076733588e-06, + "loss": 0.3217, + "step": 37544 + }, + { + "epoch": 13.86, + "learning_rate": 1.6145461263702428e-06, + "loss": 0.2853, + "step": 37545 + }, + { + "epoch": 13.86, + "learning_rate": 1.6145091760068974e-06, + "loss": 0.2514, + "step": 37546 + }, + { + "epoch": 13.87, + "learning_rate": 1.6144722256435524e-06, + "loss": 0.1689, + "step": 37547 + }, + { + "epoch": 13.87, + "learning_rate": 1.6144352752802069e-06, + "loss": 0.332, + "step": 37548 + }, + { + "epoch": 13.87, + "learning_rate": 1.6143983249168615e-06, + "loss": 0.1832, + "step": 37549 + }, + { + "epoch": 13.87, + "learning_rate": 1.6143613745535166e-06, + "loss": 0.4038, + "step": 37550 + }, + { + "epoch": 13.87, + "learning_rate": 1.6143244241901712e-06, + "loss": 0.2412, + "step": 37551 + }, + { + "epoch": 13.87, + "learning_rate": 1.614287473826826e-06, + "loss": 0.2578, + "step": 37552 + }, + { + "epoch": 13.87, + "learning_rate": 1.6142505234634807e-06, + "loss": 0.1805, + "step": 37553 + }, + { + "epoch": 13.87, + "learning_rate": 1.6142135731001355e-06, + "loss": 0.3279, + "step": 37554 + }, + { + "epoch": 13.87, + "learning_rate": 1.6141766227367902e-06, + "loss": 0.3644, + "step": 37555 + }, + { + "epoch": 13.87, + "learning_rate": 1.6141396723734452e-06, + "loss": 0.2294, + "step": 37556 + }, + { + "epoch": 13.87, + "learning_rate": 1.6141027220100999e-06, + "loss": 0.2442, + "step": 37557 + }, + { + "epoch": 13.87, + "learning_rate": 1.6140657716467547e-06, + "loss": 0.2718, + "step": 37558 + }, + { + "epoch": 13.87, + "learning_rate": 1.6140288212834094e-06, + "loss": 0.2773, + "step": 37559 + }, + { + "epoch": 13.87, + "learning_rate": 1.613991870920064e-06, + "loss": 0.196, + "step": 37560 + }, + { + "epoch": 13.87, + "learning_rate": 1.6139549205567188e-06, + "loss": 0.371, + "step": 37561 + }, + { + "epoch": 13.87, + "learning_rate": 1.6139179701933735e-06, + "loss": 0.2801, + "step": 37562 + }, + { + "epoch": 13.87, + "learning_rate": 1.6138810198300285e-06, + "loss": 0.3052, + "step": 37563 + }, + { + "epoch": 13.87, + "learning_rate": 1.613844069466683e-06, + "loss": 0.2746, + "step": 37564 + }, + { + "epoch": 13.87, + "learning_rate": 1.613807119103338e-06, + "loss": 0.2987, + "step": 37565 + }, + { + "epoch": 13.87, + "learning_rate": 1.6137701687399927e-06, + "loss": 0.2736, + "step": 37566 + }, + { + "epoch": 13.87, + "learning_rate": 1.6137332183766475e-06, + "loss": 0.1896, + "step": 37567 + }, + { + "epoch": 13.87, + "learning_rate": 1.6136962680133021e-06, + "loss": 0.2111, + "step": 37568 + }, + { + "epoch": 13.87, + "learning_rate": 1.6136593176499568e-06, + "loss": 0.1962, + "step": 37569 + }, + { + "epoch": 13.87, + "learning_rate": 1.6136223672866116e-06, + "loss": 0.3437, + "step": 37570 + }, + { + "epoch": 13.87, + "learning_rate": 1.6135854169232663e-06, + "loss": 0.2505, + "step": 37571 + }, + { + "epoch": 13.87, + "learning_rate": 1.6135484665599213e-06, + "loss": 0.4111, + "step": 37572 + }, + { + "epoch": 13.87, + "learning_rate": 1.613511516196576e-06, + "loss": 0.3864, + "step": 37573 + }, + { + "epoch": 13.88, + "learning_rate": 1.6134745658332308e-06, + "loss": 0.2762, + "step": 37574 + }, + { + "epoch": 13.88, + "learning_rate": 1.6134376154698854e-06, + "loss": 0.2587, + "step": 37575 + }, + { + "epoch": 13.88, + "learning_rate": 1.6134006651065403e-06, + "loss": 0.2489, + "step": 37576 + }, + { + "epoch": 13.88, + "learning_rate": 1.613363714743195e-06, + "loss": 0.4017, + "step": 37577 + }, + { + "epoch": 13.88, + "learning_rate": 1.61332676437985e-06, + "loss": 0.2572, + "step": 37578 + }, + { + "epoch": 13.88, + "learning_rate": 1.6132898140165046e-06, + "loss": 0.3117, + "step": 37579 + }, + { + "epoch": 13.88, + "learning_rate": 1.6132528636531593e-06, + "loss": 0.2774, + "step": 37580 + }, + { + "epoch": 13.88, + "learning_rate": 1.6132159132898141e-06, + "loss": 0.203, + "step": 37581 + }, + { + "epoch": 13.88, + "learning_rate": 1.6131789629264687e-06, + "loss": 0.2978, + "step": 37582 + }, + { + "epoch": 13.88, + "learning_rate": 1.6131420125631236e-06, + "loss": 0.348, + "step": 37583 + }, + { + "epoch": 13.88, + "learning_rate": 1.6131050621997782e-06, + "loss": 0.4904, + "step": 37584 + }, + { + "epoch": 13.88, + "learning_rate": 1.6130681118364333e-06, + "loss": 0.1619, + "step": 37585 + }, + { + "epoch": 13.88, + "learning_rate": 1.613031161473088e-06, + "loss": 0.2365, + "step": 37586 + }, + { + "epoch": 13.88, + "learning_rate": 1.6129942111097428e-06, + "loss": 0.2699, + "step": 37587 + }, + { + "epoch": 13.88, + "learning_rate": 1.6129572607463974e-06, + "loss": 0.3145, + "step": 37588 + }, + { + "epoch": 13.88, + "learning_rate": 1.612920310383052e-06, + "loss": 0.2644, + "step": 37589 + }, + { + "epoch": 13.88, + "learning_rate": 1.612883360019707e-06, + "loss": 0.388, + "step": 37590 + }, + { + "epoch": 13.88, + "learning_rate": 1.6128464096563615e-06, + "loss": 0.2954, + "step": 37591 + }, + { + "epoch": 13.88, + "learning_rate": 1.6128094592930164e-06, + "loss": 0.3481, + "step": 37592 + }, + { + "epoch": 13.88, + "learning_rate": 1.612772508929671e-06, + "loss": 0.2978, + "step": 37593 + }, + { + "epoch": 13.88, + "learning_rate": 1.612735558566326e-06, + "loss": 0.3164, + "step": 37594 + }, + { + "epoch": 13.88, + "learning_rate": 1.6126986082029807e-06, + "loss": 0.3475, + "step": 37595 + }, + { + "epoch": 13.88, + "learning_rate": 1.6126616578396356e-06, + "loss": 0.2028, + "step": 37596 + }, + { + "epoch": 13.88, + "learning_rate": 1.6126247074762902e-06, + "loss": 0.3562, + "step": 37597 + }, + { + "epoch": 13.88, + "learning_rate": 1.612587757112945e-06, + "loss": 0.3318, + "step": 37598 + }, + { + "epoch": 13.88, + "learning_rate": 1.6125508067495997e-06, + "loss": 0.4415, + "step": 37599 + }, + { + "epoch": 13.88, + "learning_rate": 1.6125138563862543e-06, + "loss": 0.2176, + "step": 37600 + }, + { + "epoch": 13.89, + "learning_rate": 1.6124769060229094e-06, + "loss": 0.3144, + "step": 37601 + }, + { + "epoch": 13.89, + "learning_rate": 1.612439955659564e-06, + "loss": 0.148, + "step": 37602 + }, + { + "epoch": 13.89, + "learning_rate": 1.6124030052962189e-06, + "loss": 0.262, + "step": 37603 + }, + { + "epoch": 13.89, + "learning_rate": 1.6123660549328735e-06, + "loss": 0.2774, + "step": 37604 + }, + { + "epoch": 13.89, + "learning_rate": 1.6123291045695283e-06, + "loss": 0.3195, + "step": 37605 + }, + { + "epoch": 13.89, + "learning_rate": 1.612292154206183e-06, + "loss": 0.2882, + "step": 37606 + }, + { + "epoch": 13.89, + "learning_rate": 1.612255203842838e-06, + "loss": 0.2755, + "step": 37607 + }, + { + "epoch": 13.89, + "learning_rate": 1.6122182534794927e-06, + "loss": 0.2749, + "step": 37608 + }, + { + "epoch": 13.89, + "learning_rate": 1.6121813031161471e-06, + "loss": 0.3565, + "step": 37609 + }, + { + "epoch": 13.89, + "learning_rate": 1.6121443527528022e-06, + "loss": 0.4021, + "step": 37610 + }, + { + "epoch": 13.89, + "learning_rate": 1.6121074023894568e-06, + "loss": 0.2378, + "step": 37611 + }, + { + "epoch": 13.89, + "learning_rate": 1.6120704520261117e-06, + "loss": 0.3393, + "step": 37612 + }, + { + "epoch": 13.89, + "learning_rate": 1.6120335016627663e-06, + "loss": 0.4544, + "step": 37613 + }, + { + "epoch": 13.89, + "learning_rate": 1.6119965512994213e-06, + "loss": 0.4922, + "step": 37614 + }, + { + "epoch": 13.89, + "learning_rate": 1.6119596009360758e-06, + "loss": 0.209, + "step": 37615 + }, + { + "epoch": 13.89, + "learning_rate": 1.6119226505727308e-06, + "loss": 0.4239, + "step": 37616 + }, + { + "epoch": 13.89, + "learning_rate": 1.6118857002093855e-06, + "loss": 0.3191, + "step": 37617 + }, + { + "epoch": 13.89, + "learning_rate": 1.6118487498460403e-06, + "loss": 0.2872, + "step": 37618 + }, + { + "epoch": 13.89, + "learning_rate": 1.611811799482695e-06, + "loss": 0.2863, + "step": 37619 + }, + { + "epoch": 13.89, + "learning_rate": 1.6117748491193496e-06, + "loss": 0.2548, + "step": 37620 + }, + { + "epoch": 13.89, + "learning_rate": 1.6117378987560044e-06, + "loss": 0.3048, + "step": 37621 + }, + { + "epoch": 13.89, + "learning_rate": 1.611700948392659e-06, + "loss": 0.2871, + "step": 37622 + }, + { + "epoch": 13.89, + "learning_rate": 1.6116639980293141e-06, + "loss": 0.233, + "step": 37623 + }, + { + "epoch": 13.89, + "learning_rate": 1.6116270476659688e-06, + "loss": 0.202, + "step": 37624 + }, + { + "epoch": 13.89, + "learning_rate": 1.6115900973026236e-06, + "loss": 0.2865, + "step": 37625 + }, + { + "epoch": 13.89, + "learning_rate": 1.6115531469392783e-06, + "loss": 0.422, + "step": 37626 + }, + { + "epoch": 13.89, + "learning_rate": 1.611516196575933e-06, + "loss": 0.4178, + "step": 37627 + }, + { + "epoch": 13.9, + "learning_rate": 1.6114792462125877e-06, + "loss": 0.3473, + "step": 37628 + }, + { + "epoch": 13.9, + "learning_rate": 1.6114422958492424e-06, + "loss": 0.3614, + "step": 37629 + }, + { + "epoch": 13.9, + "learning_rate": 1.6114053454858974e-06, + "loss": 0.2719, + "step": 37630 + }, + { + "epoch": 13.9, + "learning_rate": 1.6113683951225519e-06, + "loss": 0.3135, + "step": 37631 + }, + { + "epoch": 13.9, + "learning_rate": 1.611331444759207e-06, + "loss": 0.2012, + "step": 37632 + }, + { + "epoch": 13.9, + "learning_rate": 1.6112944943958616e-06, + "loss": 0.3023, + "step": 37633 + }, + { + "epoch": 13.9, + "learning_rate": 1.6112575440325164e-06, + "loss": 0.3171, + "step": 37634 + }, + { + "epoch": 13.9, + "learning_rate": 1.611220593669171e-06, + "loss": 0.1362, + "step": 37635 + }, + { + "epoch": 13.9, + "learning_rate": 1.611183643305826e-06, + "loss": 0.2419, + "step": 37636 + }, + { + "epoch": 13.9, + "learning_rate": 1.6111466929424805e-06, + "loss": 0.2166, + "step": 37637 + }, + { + "epoch": 13.9, + "learning_rate": 1.6111097425791352e-06, + "loss": 0.2761, + "step": 37638 + }, + { + "epoch": 13.9, + "learning_rate": 1.6110727922157902e-06, + "loss": 0.1713, + "step": 37639 + }, + { + "epoch": 13.9, + "learning_rate": 1.6110358418524449e-06, + "loss": 0.3098, + "step": 37640 + }, + { + "epoch": 13.9, + "learning_rate": 1.6109988914890997e-06, + "loss": 0.2635, + "step": 37641 + }, + { + "epoch": 13.9, + "learning_rate": 1.6109619411257543e-06, + "loss": 0.2101, + "step": 37642 + }, + { + "epoch": 13.9, + "learning_rate": 1.6109249907624092e-06, + "loss": 0.3272, + "step": 37643 + }, + { + "epoch": 13.9, + "learning_rate": 1.6108880403990638e-06, + "loss": 0.3235, + "step": 37644 + }, + { + "epoch": 13.9, + "learning_rate": 1.6108510900357189e-06, + "loss": 0.3496, + "step": 37645 + }, + { + "epoch": 13.9, + "learning_rate": 1.6108141396723735e-06, + "loss": 0.2736, + "step": 37646 + }, + { + "epoch": 13.9, + "learning_rate": 1.6107771893090284e-06, + "loss": 0.3351, + "step": 37647 + }, + { + "epoch": 13.9, + "learning_rate": 1.610740238945683e-06, + "loss": 0.4031, + "step": 37648 + }, + { + "epoch": 13.9, + "learning_rate": 1.6107032885823376e-06, + "loss": 0.3013, + "step": 37649 + }, + { + "epoch": 13.9, + "learning_rate": 1.6106663382189925e-06, + "loss": 0.3009, + "step": 37650 + }, + { + "epoch": 13.9, + "learning_rate": 1.6106293878556471e-06, + "loss": 0.3512, + "step": 37651 + }, + { + "epoch": 13.9, + "learning_rate": 1.6105924374923022e-06, + "loss": 0.2246, + "step": 37652 + }, + { + "epoch": 13.9, + "learning_rate": 1.6105554871289568e-06, + "loss": 0.4094, + "step": 37653 + }, + { + "epoch": 13.9, + "learning_rate": 1.6105185367656117e-06, + "loss": 0.281, + "step": 37654 + }, + { + "epoch": 13.91, + "learning_rate": 1.6104815864022663e-06, + "loss": 0.1988, + "step": 37655 + }, + { + "epoch": 13.91, + "learning_rate": 1.6104446360389212e-06, + "loss": 0.3104, + "step": 37656 + }, + { + "epoch": 13.91, + "learning_rate": 1.6104076856755758e-06, + "loss": 0.2035, + "step": 37657 + }, + { + "epoch": 13.91, + "learning_rate": 1.6103707353122304e-06, + "loss": 0.2712, + "step": 37658 + }, + { + "epoch": 13.91, + "learning_rate": 1.6103337849488855e-06, + "loss": 0.322, + "step": 37659 + }, + { + "epoch": 13.91, + "learning_rate": 1.61029683458554e-06, + "loss": 0.2421, + "step": 37660 + }, + { + "epoch": 13.91, + "learning_rate": 1.610259884222195e-06, + "loss": 0.1872, + "step": 37661 + }, + { + "epoch": 13.91, + "learning_rate": 1.6102229338588496e-06, + "loss": 0.402, + "step": 37662 + }, + { + "epoch": 13.91, + "learning_rate": 1.6101859834955045e-06, + "loss": 0.3006, + "step": 37663 + }, + { + "epoch": 13.91, + "learning_rate": 1.610149033132159e-06, + "loss": 0.1612, + "step": 37664 + }, + { + "epoch": 13.91, + "learning_rate": 1.6101120827688142e-06, + "loss": 0.4086, + "step": 37665 + }, + { + "epoch": 13.91, + "learning_rate": 1.6100751324054686e-06, + "loss": 0.2995, + "step": 37666 + }, + { + "epoch": 13.91, + "learning_rate": 1.6100381820421236e-06, + "loss": 0.1686, + "step": 37667 + }, + { + "epoch": 13.91, + "learning_rate": 1.6100012316787783e-06, + "loss": 0.2667, + "step": 37668 + }, + { + "epoch": 13.91, + "learning_rate": 1.609964281315433e-06, + "loss": 0.2571, + "step": 37669 + }, + { + "epoch": 13.91, + "learning_rate": 1.6099273309520878e-06, + "loss": 0.2978, + "step": 37670 + }, + { + "epoch": 13.91, + "learning_rate": 1.6098903805887424e-06, + "loss": 0.2639, + "step": 37671 + }, + { + "epoch": 13.91, + "learning_rate": 1.6098534302253972e-06, + "loss": 0.1981, + "step": 37672 + }, + { + "epoch": 13.91, + "learning_rate": 1.6098164798620519e-06, + "loss": 0.2848, + "step": 37673 + }, + { + "epoch": 13.91, + "learning_rate": 1.609779529498707e-06, + "loss": 0.4112, + "step": 37674 + }, + { + "epoch": 13.91, + "learning_rate": 1.6097425791353616e-06, + "loss": 0.2746, + "step": 37675 + }, + { + "epoch": 13.91, + "learning_rate": 1.6097056287720164e-06, + "loss": 0.2084, + "step": 37676 + }, + { + "epoch": 13.91, + "learning_rate": 1.609668678408671e-06, + "loss": 0.3629, + "step": 37677 + }, + { + "epoch": 13.91, + "learning_rate": 1.6096317280453257e-06, + "loss": 0.3226, + "step": 37678 + }, + { + "epoch": 13.91, + "learning_rate": 1.6095947776819806e-06, + "loss": 0.1923, + "step": 37679 + }, + { + "epoch": 13.91, + "learning_rate": 1.6095578273186352e-06, + "loss": 0.1951, + "step": 37680 + }, + { + "epoch": 13.91, + "learning_rate": 1.6095208769552902e-06, + "loss": 0.2381, + "step": 37681 + }, + { + "epoch": 13.92, + "learning_rate": 1.6094839265919447e-06, + "loss": 0.2895, + "step": 37682 + }, + { + "epoch": 13.92, + "learning_rate": 1.6094469762285997e-06, + "loss": 0.4754, + "step": 37683 + }, + { + "epoch": 13.92, + "learning_rate": 1.6094100258652544e-06, + "loss": 0.2245, + "step": 37684 + }, + { + "epoch": 13.92, + "learning_rate": 1.6093730755019092e-06, + "loss": 0.1966, + "step": 37685 + }, + { + "epoch": 13.92, + "learning_rate": 1.6093361251385639e-06, + "loss": 0.297, + "step": 37686 + }, + { + "epoch": 13.92, + "learning_rate": 1.609299174775219e-06, + "loss": 0.2234, + "step": 37687 + }, + { + "epoch": 13.92, + "learning_rate": 1.6092622244118733e-06, + "loss": 0.3429, + "step": 37688 + }, + { + "epoch": 13.92, + "learning_rate": 1.609225274048528e-06, + "loss": 0.1586, + "step": 37689 + }, + { + "epoch": 13.92, + "learning_rate": 1.609188323685183e-06, + "loss": 0.3792, + "step": 37690 + }, + { + "epoch": 13.92, + "learning_rate": 1.6091513733218377e-06, + "loss": 0.3482, + "step": 37691 + }, + { + "epoch": 13.92, + "learning_rate": 1.6091144229584925e-06, + "loss": 0.3918, + "step": 37692 + }, + { + "epoch": 13.92, + "learning_rate": 1.6090774725951472e-06, + "loss": 0.2009, + "step": 37693 + }, + { + "epoch": 13.92, + "learning_rate": 1.609040522231802e-06, + "loss": 0.3468, + "step": 37694 + }, + { + "epoch": 13.92, + "learning_rate": 1.6090035718684566e-06, + "loss": 0.2194, + "step": 37695 + }, + { + "epoch": 13.92, + "learning_rate": 1.6089666215051117e-06, + "loss": 0.2546, + "step": 37696 + }, + { + "epoch": 13.92, + "learning_rate": 1.6089296711417663e-06, + "loss": 0.2695, + "step": 37697 + }, + { + "epoch": 13.92, + "learning_rate": 1.608892720778421e-06, + "loss": 0.2944, + "step": 37698 + }, + { + "epoch": 13.92, + "learning_rate": 1.6088557704150758e-06, + "loss": 0.2637, + "step": 37699 + }, + { + "epoch": 13.92, + "learning_rate": 1.6088188200517305e-06, + "loss": 0.236, + "step": 37700 + }, + { + "epoch": 13.92, + "learning_rate": 1.6087818696883853e-06, + "loss": 0.325, + "step": 37701 + }, + { + "epoch": 13.92, + "learning_rate": 1.60874491932504e-06, + "loss": 0.2825, + "step": 37702 + }, + { + "epoch": 13.92, + "learning_rate": 1.608707968961695e-06, + "loss": 0.2901, + "step": 37703 + }, + { + "epoch": 13.92, + "learning_rate": 1.6086710185983494e-06, + "loss": 0.2463, + "step": 37704 + }, + { + "epoch": 13.92, + "learning_rate": 1.6086340682350045e-06, + "loss": 0.2928, + "step": 37705 + }, + { + "epoch": 13.92, + "learning_rate": 1.6085971178716591e-06, + "loss": 0.2436, + "step": 37706 + }, + { + "epoch": 13.92, + "learning_rate": 1.608560167508314e-06, + "loss": 0.2807, + "step": 37707 + }, + { + "epoch": 13.92, + "learning_rate": 1.6085232171449686e-06, + "loss": 0.2741, + "step": 37708 + }, + { + "epoch": 13.93, + "learning_rate": 1.6084862667816232e-06, + "loss": 0.2516, + "step": 37709 + }, + { + "epoch": 13.93, + "learning_rate": 1.608449316418278e-06, + "loss": 0.2658, + "step": 37710 + }, + { + "epoch": 13.93, + "learning_rate": 1.6084123660549327e-06, + "loss": 0.273, + "step": 37711 + }, + { + "epoch": 13.93, + "learning_rate": 1.6083754156915878e-06, + "loss": 0.3346, + "step": 37712 + }, + { + "epoch": 13.93, + "learning_rate": 1.6083384653282424e-06, + "loss": 0.2634, + "step": 37713 + }, + { + "epoch": 13.93, + "learning_rate": 1.6083015149648973e-06, + "loss": 0.3408, + "step": 37714 + }, + { + "epoch": 13.93, + "learning_rate": 1.608264564601552e-06, + "loss": 0.2024, + "step": 37715 + }, + { + "epoch": 13.93, + "learning_rate": 1.6082276142382068e-06, + "loss": 0.2924, + "step": 37716 + }, + { + "epoch": 13.93, + "learning_rate": 1.6081906638748614e-06, + "loss": 0.3324, + "step": 37717 + }, + { + "epoch": 13.93, + "learning_rate": 1.608153713511516e-06, + "loss": 0.3751, + "step": 37718 + }, + { + "epoch": 13.93, + "learning_rate": 1.608116763148171e-06, + "loss": 0.3123, + "step": 37719 + }, + { + "epoch": 13.93, + "learning_rate": 1.6080798127848257e-06, + "loss": 0.2044, + "step": 37720 + }, + { + "epoch": 13.93, + "learning_rate": 1.6080428624214806e-06, + "loss": 0.2524, + "step": 37721 + }, + { + "epoch": 13.93, + "learning_rate": 1.6080059120581352e-06, + "loss": 0.3932, + "step": 37722 + }, + { + "epoch": 13.93, + "learning_rate": 1.60796896169479e-06, + "loss": 0.1957, + "step": 37723 + }, + { + "epoch": 13.93, + "learning_rate": 1.6079320113314447e-06, + "loss": 0.2614, + "step": 37724 + }, + { + "epoch": 13.93, + "learning_rate": 1.6078950609680998e-06, + "loss": 0.1425, + "step": 37725 + }, + { + "epoch": 13.93, + "learning_rate": 1.6078581106047544e-06, + "loss": 0.2856, + "step": 37726 + }, + { + "epoch": 13.93, + "learning_rate": 1.6078211602414092e-06, + "loss": 0.2432, + "step": 37727 + }, + { + "epoch": 13.93, + "learning_rate": 1.6077842098780639e-06, + "loss": 0.2583, + "step": 37728 + }, + { + "epoch": 13.93, + "learning_rate": 1.6077472595147185e-06, + "loss": 0.3019, + "step": 37729 + }, + { + "epoch": 13.93, + "learning_rate": 1.6077103091513734e-06, + "loss": 0.2964, + "step": 37730 + }, + { + "epoch": 13.93, + "learning_rate": 1.607673358788028e-06, + "loss": 0.4271, + "step": 37731 + }, + { + "epoch": 13.93, + "learning_rate": 1.607636408424683e-06, + "loss": 0.2656, + "step": 37732 + }, + { + "epoch": 13.93, + "learning_rate": 1.6075994580613375e-06, + "loss": 0.2912, + "step": 37733 + }, + { + "epoch": 13.93, + "learning_rate": 1.6075625076979925e-06, + "loss": 0.2828, + "step": 37734 + }, + { + "epoch": 13.93, + "learning_rate": 1.6075255573346472e-06, + "loss": 0.2974, + "step": 37735 + }, + { + "epoch": 13.94, + "learning_rate": 1.607488606971302e-06, + "loss": 0.3453, + "step": 37736 + }, + { + "epoch": 13.94, + "learning_rate": 1.6074516566079567e-06, + "loss": 0.2736, + "step": 37737 + }, + { + "epoch": 13.94, + "learning_rate": 1.6074147062446113e-06, + "loss": 0.2261, + "step": 37738 + }, + { + "epoch": 13.94, + "learning_rate": 1.6073777558812661e-06, + "loss": 0.2393, + "step": 37739 + }, + { + "epoch": 13.94, + "learning_rate": 1.6073408055179208e-06, + "loss": 0.2843, + "step": 37740 + }, + { + "epoch": 13.94, + "learning_rate": 1.6073038551545758e-06, + "loss": 0.2759, + "step": 37741 + }, + { + "epoch": 13.94, + "learning_rate": 1.6072669047912305e-06, + "loss": 0.3717, + "step": 37742 + }, + { + "epoch": 13.94, + "learning_rate": 1.6072299544278853e-06, + "loss": 0.2006, + "step": 37743 + }, + { + "epoch": 13.94, + "learning_rate": 1.60719300406454e-06, + "loss": 0.2095, + "step": 37744 + }, + { + "epoch": 13.94, + "learning_rate": 1.6071560537011948e-06, + "loss": 0.2187, + "step": 37745 + }, + { + "epoch": 13.94, + "learning_rate": 1.6071191033378494e-06, + "loss": 0.2812, + "step": 37746 + }, + { + "epoch": 13.94, + "learning_rate": 1.6070821529745045e-06, + "loss": 0.3484, + "step": 37747 + }, + { + "epoch": 13.94, + "learning_rate": 1.6070452026111591e-06, + "loss": 0.2606, + "step": 37748 + }, + { + "epoch": 13.94, + "learning_rate": 1.6070082522478136e-06, + "loss": 0.2008, + "step": 37749 + }, + { + "epoch": 13.94, + "learning_rate": 1.6069713018844686e-06, + "loss": 0.2917, + "step": 37750 + }, + { + "epoch": 13.94, + "learning_rate": 1.6069343515211233e-06, + "loss": 0.182, + "step": 37751 + }, + { + "epoch": 13.94, + "learning_rate": 1.6068974011577781e-06, + "loss": 0.3285, + "step": 37752 + }, + { + "epoch": 13.94, + "learning_rate": 1.6068604507944328e-06, + "loss": 0.2854, + "step": 37753 + }, + { + "epoch": 13.94, + "learning_rate": 1.6068235004310878e-06, + "loss": 0.189, + "step": 37754 + }, + { + "epoch": 13.94, + "learning_rate": 1.6067865500677422e-06, + "loss": 0.2709, + "step": 37755 + }, + { + "epoch": 13.94, + "learning_rate": 1.6067495997043973e-06, + "loss": 0.2181, + "step": 37756 + }, + { + "epoch": 13.94, + "learning_rate": 1.606712649341052e-06, + "loss": 0.2788, + "step": 37757 + }, + { + "epoch": 13.94, + "learning_rate": 1.6066756989777066e-06, + "loss": 0.2626, + "step": 37758 + }, + { + "epoch": 13.94, + "learning_rate": 1.6066387486143614e-06, + "loss": 0.2942, + "step": 37759 + }, + { + "epoch": 13.94, + "learning_rate": 1.606601798251016e-06, + "loss": 0.2493, + "step": 37760 + }, + { + "epoch": 13.94, + "learning_rate": 1.606564847887671e-06, + "loss": 0.2905, + "step": 37761 + }, + { + "epoch": 13.94, + "learning_rate": 1.6065278975243255e-06, + "loss": 0.3674, + "step": 37762 + }, + { + "epoch": 13.94, + "learning_rate": 1.6064909471609806e-06, + "loss": 0.3448, + "step": 37763 + }, + { + "epoch": 13.95, + "learning_rate": 1.6064539967976352e-06, + "loss": 0.2609, + "step": 37764 + }, + { + "epoch": 13.95, + "learning_rate": 1.60641704643429e-06, + "loss": 0.3243, + "step": 37765 + }, + { + "epoch": 13.95, + "learning_rate": 1.6063800960709447e-06, + "loss": 0.3355, + "step": 37766 + }, + { + "epoch": 13.95, + "learning_rate": 1.6063431457075996e-06, + "loss": 0.2996, + "step": 37767 + }, + { + "epoch": 13.95, + "learning_rate": 1.6063061953442542e-06, + "loss": 0.2342, + "step": 37768 + }, + { + "epoch": 13.95, + "learning_rate": 1.6062692449809088e-06, + "loss": 0.4324, + "step": 37769 + }, + { + "epoch": 13.95, + "learning_rate": 1.606232294617564e-06, + "loss": 0.3676, + "step": 37770 + }, + { + "epoch": 13.95, + "learning_rate": 1.6061953442542185e-06, + "loss": 0.3718, + "step": 37771 + }, + { + "epoch": 13.95, + "learning_rate": 1.6061583938908734e-06, + "loss": 0.3271, + "step": 37772 + }, + { + "epoch": 13.95, + "learning_rate": 1.606121443527528e-06, + "loss": 0.4163, + "step": 37773 + }, + { + "epoch": 13.95, + "learning_rate": 1.6060844931641829e-06, + "loss": 0.3263, + "step": 37774 + }, + { + "epoch": 13.95, + "learning_rate": 1.6060475428008375e-06, + "loss": 0.1956, + "step": 37775 + }, + { + "epoch": 13.95, + "learning_rate": 1.6060105924374926e-06, + "loss": 0.2608, + "step": 37776 + }, + { + "epoch": 13.95, + "learning_rate": 1.6059736420741472e-06, + "loss": 0.169, + "step": 37777 + }, + { + "epoch": 13.95, + "learning_rate": 1.6059366917108016e-06, + "loss": 0.3488, + "step": 37778 + }, + { + "epoch": 13.95, + "learning_rate": 1.6058997413474567e-06, + "loss": 0.3276, + "step": 37779 + }, + { + "epoch": 13.95, + "learning_rate": 1.6058627909841113e-06, + "loss": 0.4279, + "step": 37780 + }, + { + "epoch": 13.95, + "learning_rate": 1.6058258406207662e-06, + "loss": 0.2759, + "step": 37781 + }, + { + "epoch": 13.95, + "learning_rate": 1.6057888902574208e-06, + "loss": 0.1977, + "step": 37782 + }, + { + "epoch": 13.95, + "learning_rate": 1.6057519398940757e-06, + "loss": 0.2693, + "step": 37783 + }, + { + "epoch": 13.95, + "learning_rate": 1.6057149895307303e-06, + "loss": 0.2553, + "step": 37784 + }, + { + "epoch": 13.95, + "learning_rate": 1.6056780391673854e-06, + "loss": 0.3631, + "step": 37785 + }, + { + "epoch": 13.95, + "learning_rate": 1.60564108880404e-06, + "loss": 0.2767, + "step": 37786 + }, + { + "epoch": 13.95, + "learning_rate": 1.6056041384406948e-06, + "loss": 0.4807, + "step": 37787 + }, + { + "epoch": 13.95, + "learning_rate": 1.6055671880773495e-06, + "loss": 0.2685, + "step": 37788 + }, + { + "epoch": 13.95, + "learning_rate": 1.6055302377140041e-06, + "loss": 0.3354, + "step": 37789 + }, + { + "epoch": 13.95, + "learning_rate": 1.605493287350659e-06, + "loss": 0.2573, + "step": 37790 + }, + { + "epoch": 13.96, + "learning_rate": 1.6054563369873136e-06, + "loss": 0.343, + "step": 37791 + }, + { + "epoch": 13.96, + "learning_rate": 1.6054193866239687e-06, + "loss": 0.2656, + "step": 37792 + }, + { + "epoch": 13.96, + "learning_rate": 1.6053824362606233e-06, + "loss": 0.2713, + "step": 37793 + }, + { + "epoch": 13.96, + "learning_rate": 1.6053454858972781e-06, + "loss": 0.2568, + "step": 37794 + }, + { + "epoch": 13.96, + "learning_rate": 1.6053085355339328e-06, + "loss": 0.406, + "step": 37795 + }, + { + "epoch": 13.96, + "learning_rate": 1.6052715851705876e-06, + "loss": 0.2626, + "step": 37796 + }, + { + "epoch": 13.96, + "learning_rate": 1.6052346348072423e-06, + "loss": 0.2888, + "step": 37797 + }, + { + "epoch": 13.96, + "learning_rate": 1.605197684443897e-06, + "loss": 0.274, + "step": 37798 + }, + { + "epoch": 13.96, + "learning_rate": 1.605160734080552e-06, + "loss": 0.2469, + "step": 37799 + }, + { + "epoch": 13.96, + "learning_rate": 1.6051237837172064e-06, + "loss": 0.2722, + "step": 37800 + }, + { + "epoch": 13.96, + "learning_rate": 1.6050868333538614e-06, + "loss": 0.3423, + "step": 37801 + }, + { + "epoch": 13.96, + "learning_rate": 1.605049882990516e-06, + "loss": 0.352, + "step": 37802 + }, + { + "epoch": 13.96, + "learning_rate": 1.605012932627171e-06, + "loss": 0.3268, + "step": 37803 + }, + { + "epoch": 13.96, + "learning_rate": 1.6049759822638256e-06, + "loss": 0.1962, + "step": 37804 + }, + { + "epoch": 13.96, + "learning_rate": 1.6049390319004806e-06, + "loss": 0.3146, + "step": 37805 + }, + { + "epoch": 13.96, + "learning_rate": 1.604902081537135e-06, + "loss": 0.3312, + "step": 37806 + }, + { + "epoch": 13.96, + "learning_rate": 1.6048651311737901e-06, + "loss": 0.2004, + "step": 37807 + }, + { + "epoch": 13.96, + "learning_rate": 1.6048281808104447e-06, + "loss": 0.2419, + "step": 37808 + }, + { + "epoch": 13.96, + "learning_rate": 1.6047912304470994e-06, + "loss": 0.2947, + "step": 37809 + }, + { + "epoch": 13.96, + "learning_rate": 1.6047542800837542e-06, + "loss": 0.2401, + "step": 37810 + }, + { + "epoch": 13.96, + "learning_rate": 1.6047173297204089e-06, + "loss": 0.3379, + "step": 37811 + }, + { + "epoch": 13.96, + "learning_rate": 1.6046803793570637e-06, + "loss": 0.2127, + "step": 37812 + }, + { + "epoch": 13.96, + "learning_rate": 1.6046434289937183e-06, + "loss": 0.4589, + "step": 37813 + }, + { + "epoch": 13.96, + "learning_rate": 1.6046064786303734e-06, + "loss": 0.3244, + "step": 37814 + }, + { + "epoch": 13.96, + "learning_rate": 1.604569528267028e-06, + "loss": 0.1525, + "step": 37815 + }, + { + "epoch": 13.96, + "learning_rate": 1.6045325779036829e-06, + "loss": 0.2122, + "step": 37816 + }, + { + "epoch": 13.96, + "learning_rate": 1.6044956275403375e-06, + "loss": 0.2394, + "step": 37817 + }, + { + "epoch": 13.97, + "learning_rate": 1.6044586771769922e-06, + "loss": 0.2401, + "step": 37818 + }, + { + "epoch": 13.97, + "learning_rate": 1.604421726813647e-06, + "loss": 0.3612, + "step": 37819 + }, + { + "epoch": 13.97, + "learning_rate": 1.6043847764503017e-06, + "loss": 0.3974, + "step": 37820 + }, + { + "epoch": 13.97, + "learning_rate": 1.6043478260869567e-06, + "loss": 0.2877, + "step": 37821 + }, + { + "epoch": 13.97, + "learning_rate": 1.6043108757236111e-06, + "loss": 0.3839, + "step": 37822 + }, + { + "epoch": 13.97, + "learning_rate": 1.6042739253602662e-06, + "loss": 0.2831, + "step": 37823 + }, + { + "epoch": 13.97, + "learning_rate": 1.6042369749969208e-06, + "loss": 0.287, + "step": 37824 + }, + { + "epoch": 13.97, + "learning_rate": 1.6042000246335757e-06, + "loss": 0.3013, + "step": 37825 + }, + { + "epoch": 13.97, + "learning_rate": 1.6041630742702303e-06, + "loss": 0.3122, + "step": 37826 + }, + { + "epoch": 13.97, + "learning_rate": 1.6041261239068854e-06, + "loss": 0.1666, + "step": 37827 + }, + { + "epoch": 13.97, + "learning_rate": 1.6040891735435398e-06, + "loss": 0.2066, + "step": 37828 + }, + { + "epoch": 13.97, + "learning_rate": 1.6040522231801944e-06, + "loss": 0.1994, + "step": 37829 + }, + { + "epoch": 13.97, + "learning_rate": 1.6040152728168495e-06, + "loss": 0.2321, + "step": 37830 + }, + { + "epoch": 13.97, + "learning_rate": 1.6039783224535041e-06, + "loss": 0.436, + "step": 37831 + }, + { + "epoch": 13.97, + "learning_rate": 1.603941372090159e-06, + "loss": 0.2453, + "step": 37832 + }, + { + "epoch": 13.97, + "learning_rate": 1.6039044217268136e-06, + "loss": 0.3505, + "step": 37833 + }, + { + "epoch": 13.97, + "learning_rate": 1.6038674713634685e-06, + "loss": 0.3553, + "step": 37834 + }, + { + "epoch": 13.97, + "learning_rate": 1.603830521000123e-06, + "loss": 0.4425, + "step": 37835 + }, + { + "epoch": 13.97, + "learning_rate": 1.6037935706367782e-06, + "loss": 0.13, + "step": 37836 + }, + { + "epoch": 13.97, + "learning_rate": 1.6037566202734328e-06, + "loss": 0.2271, + "step": 37837 + }, + { + "epoch": 13.97, + "learning_rate": 1.6037196699100874e-06, + "loss": 0.3036, + "step": 37838 + }, + { + "epoch": 13.97, + "learning_rate": 1.6036827195467423e-06, + "loss": 0.2325, + "step": 37839 + }, + { + "epoch": 13.97, + "learning_rate": 1.603645769183397e-06, + "loss": 0.3073, + "step": 37840 + }, + { + "epoch": 13.97, + "learning_rate": 1.6036088188200518e-06, + "loss": 0.2398, + "step": 37841 + }, + { + "epoch": 13.97, + "learning_rate": 1.6035718684567064e-06, + "loss": 0.2706, + "step": 37842 + }, + { + "epoch": 13.97, + "learning_rate": 1.6035349180933615e-06, + "loss": 0.3278, + "step": 37843 + }, + { + "epoch": 13.97, + "learning_rate": 1.603497967730016e-06, + "loss": 0.2816, + "step": 37844 + }, + { + "epoch": 13.98, + "learning_rate": 1.603461017366671e-06, + "loss": 0.3075, + "step": 37845 + }, + { + "epoch": 13.98, + "learning_rate": 1.6034240670033256e-06, + "loss": 0.3956, + "step": 37846 + }, + { + "epoch": 13.98, + "learning_rate": 1.6033871166399804e-06, + "loss": 0.3121, + "step": 37847 + }, + { + "epoch": 13.98, + "learning_rate": 1.603350166276635e-06, + "loss": 0.2632, + "step": 37848 + }, + { + "epoch": 13.98, + "learning_rate": 1.6033132159132897e-06, + "loss": 0.3161, + "step": 37849 + }, + { + "epoch": 13.98, + "learning_rate": 1.6032762655499448e-06, + "loss": 0.2707, + "step": 37850 + }, + { + "epoch": 13.98, + "learning_rate": 1.6032393151865992e-06, + "loss": 0.2595, + "step": 37851 + }, + { + "epoch": 13.98, + "learning_rate": 1.6032023648232543e-06, + "loss": 0.2493, + "step": 37852 + }, + { + "epoch": 13.98, + "learning_rate": 1.6031654144599089e-06, + "loss": 0.3921, + "step": 37853 + }, + { + "epoch": 13.98, + "learning_rate": 1.6031284640965637e-06, + "loss": 0.3348, + "step": 37854 + }, + { + "epoch": 13.98, + "learning_rate": 1.6030915137332184e-06, + "loss": 0.3102, + "step": 37855 + }, + { + "epoch": 13.98, + "learning_rate": 1.6030545633698732e-06, + "loss": 0.2011, + "step": 37856 + }, + { + "epoch": 13.98, + "learning_rate": 1.6030176130065279e-06, + "loss": 0.3086, + "step": 37857 + }, + { + "epoch": 13.98, + "learning_rate": 1.6029806626431825e-06, + "loss": 0.3066, + "step": 37858 + }, + { + "epoch": 13.98, + "learning_rate": 1.6029437122798376e-06, + "loss": 0.2453, + "step": 37859 + }, + { + "epoch": 13.98, + "learning_rate": 1.6029067619164922e-06, + "loss": 0.3519, + "step": 37860 + }, + { + "epoch": 13.98, + "learning_rate": 1.602869811553147e-06, + "loss": 0.2079, + "step": 37861 + }, + { + "epoch": 13.98, + "learning_rate": 1.6028328611898017e-06, + "loss": 0.2984, + "step": 37862 + }, + { + "epoch": 13.98, + "learning_rate": 1.6027959108264565e-06, + "loss": 0.2204, + "step": 37863 + }, + { + "epoch": 13.98, + "learning_rate": 1.6027589604631112e-06, + "loss": 0.1963, + "step": 37864 + }, + { + "epoch": 13.98, + "learning_rate": 1.6027220100997662e-06, + "loss": 0.3061, + "step": 37865 + }, + { + "epoch": 13.98, + "learning_rate": 1.6026850597364209e-06, + "loss": 0.312, + "step": 37866 + }, + { + "epoch": 13.98, + "learning_rate": 1.6026481093730753e-06, + "loss": 0.2693, + "step": 37867 + }, + { + "epoch": 13.98, + "learning_rate": 1.6026111590097303e-06, + "loss": 0.1854, + "step": 37868 + }, + { + "epoch": 13.98, + "learning_rate": 1.602574208646385e-06, + "loss": 0.3205, + "step": 37869 + }, + { + "epoch": 13.98, + "learning_rate": 1.6025372582830398e-06, + "loss": 0.2487, + "step": 37870 + }, + { + "epoch": 13.98, + "learning_rate": 1.6025003079196945e-06, + "loss": 0.3099, + "step": 37871 + }, + { + "epoch": 13.99, + "learning_rate": 1.6024633575563495e-06, + "loss": 0.2068, + "step": 37872 + }, + { + "epoch": 13.99, + "learning_rate": 1.602426407193004e-06, + "loss": 0.3436, + "step": 37873 + }, + { + "epoch": 13.99, + "learning_rate": 1.602389456829659e-06, + "loss": 0.1824, + "step": 37874 + }, + { + "epoch": 13.99, + "learning_rate": 1.6023525064663136e-06, + "loss": 0.2688, + "step": 37875 + }, + { + "epoch": 13.99, + "learning_rate": 1.6023155561029685e-06, + "loss": 0.3177, + "step": 37876 + }, + { + "epoch": 13.99, + "learning_rate": 1.6022786057396231e-06, + "loss": 0.4302, + "step": 37877 + }, + { + "epoch": 13.99, + "learning_rate": 1.6022416553762778e-06, + "loss": 0.2986, + "step": 37878 + }, + { + "epoch": 13.99, + "learning_rate": 1.6022047050129326e-06, + "loss": 0.2763, + "step": 37879 + }, + { + "epoch": 13.99, + "learning_rate": 1.6021677546495872e-06, + "loss": 0.3757, + "step": 37880 + }, + { + "epoch": 13.99, + "learning_rate": 1.6021308042862423e-06, + "loss": 0.1994, + "step": 37881 + }, + { + "epoch": 13.99, + "learning_rate": 1.602093853922897e-06, + "loss": 0.2602, + "step": 37882 + }, + { + "epoch": 13.99, + "learning_rate": 1.6020569035595518e-06, + "loss": 0.3454, + "step": 37883 + }, + { + "epoch": 13.99, + "learning_rate": 1.6020199531962064e-06, + "loss": 0.2806, + "step": 37884 + }, + { + "epoch": 13.99, + "learning_rate": 1.6019830028328613e-06, + "loss": 0.24, + "step": 37885 + }, + { + "epoch": 13.99, + "learning_rate": 1.601946052469516e-06, + "loss": 0.2072, + "step": 37886 + }, + { + "epoch": 13.99, + "learning_rate": 1.6019091021061706e-06, + "loss": 0.1723, + "step": 37887 + }, + { + "epoch": 13.99, + "learning_rate": 1.6018721517428256e-06, + "loss": 0.3162, + "step": 37888 + }, + { + "epoch": 13.99, + "learning_rate": 1.60183520137948e-06, + "loss": 0.156, + "step": 37889 + }, + { + "epoch": 13.99, + "learning_rate": 1.601798251016135e-06, + "loss": 0.328, + "step": 37890 + }, + { + "epoch": 13.99, + "learning_rate": 1.6017613006527897e-06, + "loss": 0.3157, + "step": 37891 + }, + { + "epoch": 13.99, + "learning_rate": 1.6017243502894446e-06, + "loss": 0.2925, + "step": 37892 + }, + { + "epoch": 13.99, + "learning_rate": 1.6016873999260992e-06, + "loss": 0.326, + "step": 37893 + }, + { + "epoch": 13.99, + "learning_rate": 1.6016504495627543e-06, + "loss": 0.1693, + "step": 37894 + }, + { + "epoch": 13.99, + "learning_rate": 1.6016134991994087e-06, + "loss": 0.2837, + "step": 37895 + }, + { + "epoch": 13.99, + "learning_rate": 1.6015765488360638e-06, + "loss": 0.2688, + "step": 37896 + }, + { + "epoch": 13.99, + "learning_rate": 1.6015395984727184e-06, + "loss": 0.274, + "step": 37897 + }, + { + "epoch": 13.99, + "learning_rate": 1.601502648109373e-06, + "loss": 0.3523, + "step": 37898 + }, + { + "epoch": 14.0, + "learning_rate": 1.6014656977460279e-06, + "loss": 0.3638, + "step": 37899 + }, + { + "epoch": 14.0, + "learning_rate": 1.6014287473826825e-06, + "loss": 0.372, + "step": 37900 + }, + { + "epoch": 14.0, + "learning_rate": 1.6013917970193374e-06, + "loss": 0.2699, + "step": 37901 + }, + { + "epoch": 14.0, + "learning_rate": 1.601354846655992e-06, + "loss": 0.2279, + "step": 37902 + }, + { + "epoch": 14.0, + "learning_rate": 1.601317896292647e-06, + "loss": 0.1129, + "step": 37903 + }, + { + "epoch": 14.0, + "learning_rate": 1.6012809459293017e-06, + "loss": 0.3097, + "step": 37904 + }, + { + "epoch": 14.0, + "learning_rate": 1.6012439955659565e-06, + "loss": 0.2622, + "step": 37905 + }, + { + "epoch": 14.0, + "learning_rate": 1.6012070452026112e-06, + "loss": 0.3607, + "step": 37906 + }, + { + "epoch": 14.0, + "learning_rate": 1.6011700948392658e-06, + "loss": 0.4192, + "step": 37907 + }, + { + "epoch": 14.0, + "learning_rate": 1.6011331444759207e-06, + "loss": 0.2283, + "step": 37908 + }, + { + "epoch": 14.0, + "learning_rate": 1.6010961941125753e-06, + "loss": 0.2722, + "step": 37909 + }, + { + "epoch": 14.0, + "learning_rate": 1.6010592437492304e-06, + "loss": 0.29, + "step": 37910 + }, + { + "epoch": 14.0, + "learning_rate": 1.601022293385885e-06, + "loss": 0.3046, + "step": 37911 + }, + { + "epoch": 14.0, + "learning_rate": 1.6009853430225399e-06, + "loss": 0.1595, + "step": 37912 + }, + { + "epoch": 14.0, + "eval_accuracy": 0.8485478136399317, + "eval_loss": 0.36726006865501404, + "eval_runtime": 25.3034, + "eval_samples_per_second": 1711.784, + "eval_steps_per_second": 26.755, + "step": 37912 + }, + { + "epoch": 14.0, + "learning_rate": 1.6009483926591945e-06, + "loss": 0.2119, + "step": 37913 + }, + { + "epoch": 14.0, + "learning_rate": 1.6009114422958493e-06, + "loss": 0.3415, + "step": 37914 + }, + { + "epoch": 14.0, + "learning_rate": 1.600874491932504e-06, + "loss": 0.2978, + "step": 37915 + }, + { + "epoch": 14.0, + "learning_rate": 1.600837541569159e-06, + "loss": 0.3514, + "step": 37916 + }, + { + "epoch": 14.0, + "learning_rate": 1.6008005912058137e-06, + "loss": 0.2889, + "step": 37917 + }, + { + "epoch": 14.0, + "learning_rate": 1.600763640842468e-06, + "loss": 0.2316, + "step": 37918 + }, + { + "epoch": 14.0, + "learning_rate": 1.6007266904791232e-06, + "loss": 0.26, + "step": 37919 + }, + { + "epoch": 14.0, + "learning_rate": 1.6006897401157778e-06, + "loss": 0.2574, + "step": 37920 + }, + { + "epoch": 14.0, + "learning_rate": 1.6006527897524326e-06, + "loss": 0.3876, + "step": 37921 + }, + { + "epoch": 14.0, + "learning_rate": 1.6006158393890873e-06, + "loss": 0.2126, + "step": 37922 + }, + { + "epoch": 14.0, + "learning_rate": 1.6005788890257423e-06, + "loss": 0.2181, + "step": 37923 + }, + { + "epoch": 14.0, + "learning_rate": 1.6005419386623968e-06, + "loss": 0.2663, + "step": 37924 + }, + { + "epoch": 14.0, + "learning_rate": 1.6005049882990518e-06, + "loss": 0.2834, + "step": 37925 + }, + { + "epoch": 14.01, + "learning_rate": 1.6004680379357065e-06, + "loss": 0.1635, + "step": 37926 + }, + { + "epoch": 14.01, + "learning_rate": 1.600431087572361e-06, + "loss": 0.2955, + "step": 37927 + }, + { + "epoch": 14.01, + "learning_rate": 1.600394137209016e-06, + "loss": 0.3373, + "step": 37928 + }, + { + "epoch": 14.01, + "learning_rate": 1.6003571868456706e-06, + "loss": 0.2752, + "step": 37929 + }, + { + "epoch": 14.01, + "learning_rate": 1.6003202364823254e-06, + "loss": 0.4808, + "step": 37930 + }, + { + "epoch": 14.01, + "learning_rate": 1.60028328611898e-06, + "loss": 0.2408, + "step": 37931 + }, + { + "epoch": 14.01, + "learning_rate": 1.6002463357556351e-06, + "loss": 0.3107, + "step": 37932 + }, + { + "epoch": 14.01, + "learning_rate": 1.6002093853922898e-06, + "loss": 0.2543, + "step": 37933 + }, + { + "epoch": 14.01, + "learning_rate": 1.6001724350289446e-06, + "loss": 0.2582, + "step": 37934 + }, + { + "epoch": 14.01, + "learning_rate": 1.6001354846655992e-06, + "loss": 0.4202, + "step": 37935 + }, + { + "epoch": 14.01, + "learning_rate": 1.600098534302254e-06, + "loss": 0.2852, + "step": 37936 + }, + { + "epoch": 14.01, + "learning_rate": 1.6000615839389087e-06, + "loss": 0.3355, + "step": 37937 + }, + { + "epoch": 14.01, + "learning_rate": 1.6000246335755634e-06, + "loss": 0.2581, + "step": 37938 + }, + { + "epoch": 14.01, + "learning_rate": 1.5999876832122184e-06, + "loss": 0.335, + "step": 37939 + }, + { + "epoch": 14.01, + "learning_rate": 1.5999507328488728e-06, + "loss": 0.2739, + "step": 37940 + }, + { + "epoch": 14.01, + "learning_rate": 1.599913782485528e-06, + "loss": 0.2154, + "step": 37941 + }, + { + "epoch": 14.01, + "learning_rate": 1.5998768321221825e-06, + "loss": 0.2558, + "step": 37942 + }, + { + "epoch": 14.01, + "learning_rate": 1.5998398817588374e-06, + "loss": 0.3843, + "step": 37943 + }, + { + "epoch": 14.01, + "learning_rate": 1.599802931395492e-06, + "loss": 0.2941, + "step": 37944 + }, + { + "epoch": 14.01, + "learning_rate": 1.599765981032147e-06, + "loss": 0.3096, + "step": 37945 + }, + { + "epoch": 14.01, + "learning_rate": 1.5997290306688015e-06, + "loss": 0.314, + "step": 37946 + }, + { + "epoch": 14.01, + "learning_rate": 1.5996920803054561e-06, + "loss": 0.1843, + "step": 37947 + }, + { + "epoch": 14.01, + "learning_rate": 1.5996551299421112e-06, + "loss": 0.33, + "step": 37948 + }, + { + "epoch": 14.01, + "learning_rate": 1.5996181795787658e-06, + "loss": 0.2437, + "step": 37949 + }, + { + "epoch": 14.01, + "learning_rate": 1.5995812292154207e-06, + "loss": 0.3602, + "step": 37950 + }, + { + "epoch": 14.01, + "learning_rate": 1.5995442788520753e-06, + "loss": 0.261, + "step": 37951 + }, + { + "epoch": 14.01, + "learning_rate": 1.5995073284887302e-06, + "loss": 0.2179, + "step": 37952 + }, + { + "epoch": 14.02, + "learning_rate": 1.5994703781253848e-06, + "loss": 0.3098, + "step": 37953 + }, + { + "epoch": 14.02, + "learning_rate": 1.5994334277620399e-06, + "loss": 0.2109, + "step": 37954 + }, + { + "epoch": 14.02, + "learning_rate": 1.5993964773986945e-06, + "loss": 0.2034, + "step": 37955 + }, + { + "epoch": 14.02, + "learning_rate": 1.5993595270353494e-06, + "loss": 0.3968, + "step": 37956 + }, + { + "epoch": 14.02, + "learning_rate": 1.599322576672004e-06, + "loss": 0.3174, + "step": 37957 + }, + { + "epoch": 14.02, + "learning_rate": 1.5992856263086586e-06, + "loss": 0.1767, + "step": 37958 + }, + { + "epoch": 14.02, + "learning_rate": 1.5992486759453135e-06, + "loss": 0.2689, + "step": 37959 + }, + { + "epoch": 14.02, + "learning_rate": 1.5992117255819681e-06, + "loss": 0.2221, + "step": 37960 + }, + { + "epoch": 14.02, + "learning_rate": 1.5991747752186232e-06, + "loss": 0.4161, + "step": 37961 + }, + { + "epoch": 14.02, + "learning_rate": 1.5991378248552778e-06, + "loss": 0.1598, + "step": 37962 + }, + { + "epoch": 14.02, + "learning_rate": 1.5991008744919327e-06, + "loss": 0.2845, + "step": 37963 + }, + { + "epoch": 14.02, + "learning_rate": 1.5990639241285873e-06, + "loss": 0.2319, + "step": 37964 + }, + { + "epoch": 14.02, + "learning_rate": 1.5990269737652421e-06, + "loss": 0.2311, + "step": 37965 + }, + { + "epoch": 14.02, + "learning_rate": 1.5989900234018968e-06, + "loss": 0.1874, + "step": 37966 + }, + { + "epoch": 14.02, + "learning_rate": 1.5989530730385514e-06, + "loss": 0.2491, + "step": 37967 + }, + { + "epoch": 14.02, + "learning_rate": 1.5989161226752063e-06, + "loss": 0.3087, + "step": 37968 + }, + { + "epoch": 14.02, + "learning_rate": 1.598879172311861e-06, + "loss": 0.2368, + "step": 37969 + }, + { + "epoch": 14.02, + "learning_rate": 1.598842221948516e-06, + "loss": 0.2539, + "step": 37970 + }, + { + "epoch": 14.02, + "learning_rate": 1.5988052715851706e-06, + "loss": 0.3394, + "step": 37971 + }, + { + "epoch": 14.02, + "learning_rate": 1.5987683212218254e-06, + "loss": 0.1847, + "step": 37972 + }, + { + "epoch": 14.02, + "learning_rate": 1.59873137085848e-06, + "loss": 0.2787, + "step": 37973 + }, + { + "epoch": 14.02, + "learning_rate": 1.598694420495135e-06, + "loss": 0.3242, + "step": 37974 + }, + { + "epoch": 14.02, + "learning_rate": 1.5986574701317896e-06, + "loss": 0.252, + "step": 37975 + }, + { + "epoch": 14.02, + "learning_rate": 1.5986205197684446e-06, + "loss": 0.4955, + "step": 37976 + }, + { + "epoch": 14.02, + "learning_rate": 1.5985835694050993e-06, + "loss": 0.2263, + "step": 37977 + }, + { + "epoch": 14.02, + "learning_rate": 1.598546619041754e-06, + "loss": 0.1937, + "step": 37978 + }, + { + "epoch": 14.02, + "learning_rate": 1.5985096686784087e-06, + "loss": 0.2046, + "step": 37979 + }, + { + "epoch": 14.03, + "learning_rate": 1.5984727183150634e-06, + "loss": 0.2094, + "step": 37980 + }, + { + "epoch": 14.03, + "learning_rate": 1.5984357679517182e-06, + "loss": 0.3865, + "step": 37981 + }, + { + "epoch": 14.03, + "learning_rate": 1.5983988175883729e-06, + "loss": 0.3431, + "step": 37982 + }, + { + "epoch": 14.03, + "learning_rate": 1.598361867225028e-06, + "loss": 0.3185, + "step": 37983 + }, + { + "epoch": 14.03, + "learning_rate": 1.5983249168616826e-06, + "loss": 0.4363, + "step": 37984 + }, + { + "epoch": 14.03, + "learning_rate": 1.5982879664983374e-06, + "loss": 0.3215, + "step": 37985 + }, + { + "epoch": 14.03, + "learning_rate": 1.598251016134992e-06, + "loss": 0.2216, + "step": 37986 + }, + { + "epoch": 14.03, + "learning_rate": 1.5982140657716467e-06, + "loss": 0.382, + "step": 37987 + }, + { + "epoch": 14.03, + "learning_rate": 1.5981771154083015e-06, + "loss": 0.2165, + "step": 37988 + }, + { + "epoch": 14.03, + "learning_rate": 1.5981401650449562e-06, + "loss": 0.2295, + "step": 37989 + }, + { + "epoch": 14.03, + "learning_rate": 1.5981032146816112e-06, + "loss": 0.237, + "step": 37990 + }, + { + "epoch": 14.03, + "learning_rate": 1.5980662643182657e-06, + "loss": 0.2103, + "step": 37991 + }, + { + "epoch": 14.03, + "learning_rate": 1.5980293139549207e-06, + "loss": 0.26, + "step": 37992 + }, + { + "epoch": 14.03, + "learning_rate": 1.5979923635915754e-06, + "loss": 0.1837, + "step": 37993 + }, + { + "epoch": 14.03, + "learning_rate": 1.5979554132282302e-06, + "loss": 0.2677, + "step": 37994 + }, + { + "epoch": 14.03, + "learning_rate": 1.5979184628648848e-06, + "loss": 0.429, + "step": 37995 + }, + { + "epoch": 14.03, + "learning_rate": 1.59788151250154e-06, + "loss": 0.1963, + "step": 37996 + }, + { + "epoch": 14.03, + "learning_rate": 1.5978445621381943e-06, + "loss": 0.2247, + "step": 37997 + }, + { + "epoch": 14.03, + "learning_rate": 1.597807611774849e-06, + "loss": 0.2109, + "step": 37998 + }, + { + "epoch": 14.03, + "learning_rate": 1.597770661411504e-06, + "loss": 0.2721, + "step": 37999 + }, + { + "epoch": 14.03, + "learning_rate": 1.5977337110481587e-06, + "loss": 0.3099, + "step": 38000 + }, + { + "epoch": 14.03, + "learning_rate": 1.5976967606848135e-06, + "loss": 0.166, + "step": 38001 + }, + { + "epoch": 14.03, + "learning_rate": 1.5976598103214681e-06, + "loss": 0.3037, + "step": 38002 + }, + { + "epoch": 14.03, + "learning_rate": 1.597622859958123e-06, + "loss": 0.2604, + "step": 38003 + }, + { + "epoch": 14.03, + "learning_rate": 1.5975859095947776e-06, + "loss": 0.3516, + "step": 38004 + }, + { + "epoch": 14.03, + "learning_rate": 1.5975489592314327e-06, + "loss": 0.2385, + "step": 38005 + }, + { + "epoch": 14.03, + "learning_rate": 1.5975120088680873e-06, + "loss": 0.2711, + "step": 38006 + }, + { + "epoch": 14.04, + "learning_rate": 1.5974750585047417e-06, + "loss": 0.2723, + "step": 38007 + }, + { + "epoch": 14.04, + "learning_rate": 1.5974381081413968e-06, + "loss": 0.3337, + "step": 38008 + }, + { + "epoch": 14.04, + "learning_rate": 1.5974011577780514e-06, + "loss": 0.2331, + "step": 38009 + }, + { + "epoch": 14.04, + "learning_rate": 1.5973642074147063e-06, + "loss": 0.261, + "step": 38010 + }, + { + "epoch": 14.04, + "learning_rate": 1.597327257051361e-06, + "loss": 0.2122, + "step": 38011 + }, + { + "epoch": 14.04, + "learning_rate": 1.597290306688016e-06, + "loss": 0.2866, + "step": 38012 + }, + { + "epoch": 14.04, + "learning_rate": 1.5972533563246704e-06, + "loss": 0.2348, + "step": 38013 + }, + { + "epoch": 14.04, + "learning_rate": 1.5972164059613255e-06, + "loss": 0.3301, + "step": 38014 + }, + { + "epoch": 14.04, + "learning_rate": 1.5971794555979801e-06, + "loss": 0.2153, + "step": 38015 + }, + { + "epoch": 14.04, + "learning_rate": 1.597142505234635e-06, + "loss": 0.3324, + "step": 38016 + }, + { + "epoch": 14.04, + "learning_rate": 1.5971055548712896e-06, + "loss": 0.3668, + "step": 38017 + }, + { + "epoch": 14.04, + "learning_rate": 1.5970686045079442e-06, + "loss": 0.2553, + "step": 38018 + }, + { + "epoch": 14.04, + "learning_rate": 1.597031654144599e-06, + "loss": 0.2714, + "step": 38019 + }, + { + "epoch": 14.04, + "learning_rate": 1.5969947037812537e-06, + "loss": 0.1851, + "step": 38020 + }, + { + "epoch": 14.04, + "learning_rate": 1.5969577534179088e-06, + "loss": 0.222, + "step": 38021 + }, + { + "epoch": 14.04, + "learning_rate": 1.5969208030545634e-06, + "loss": 0.3946, + "step": 38022 + }, + { + "epoch": 14.04, + "learning_rate": 1.5968838526912183e-06, + "loss": 0.3141, + "step": 38023 + }, + { + "epoch": 14.04, + "learning_rate": 1.596846902327873e-06, + "loss": 0.2398, + "step": 38024 + }, + { + "epoch": 14.04, + "learning_rate": 1.5968099519645277e-06, + "loss": 0.2782, + "step": 38025 + }, + { + "epoch": 14.04, + "learning_rate": 1.5967730016011824e-06, + "loss": 0.2271, + "step": 38026 + }, + { + "epoch": 14.04, + "learning_rate": 1.596736051237837e-06, + "loss": 0.1727, + "step": 38027 + }, + { + "epoch": 14.04, + "learning_rate": 1.596699100874492e-06, + "loss": 0.2653, + "step": 38028 + }, + { + "epoch": 14.04, + "learning_rate": 1.5966621505111467e-06, + "loss": 0.4072, + "step": 38029 + }, + { + "epoch": 14.04, + "learning_rate": 1.5966252001478016e-06, + "loss": 0.2754, + "step": 38030 + }, + { + "epoch": 14.04, + "learning_rate": 1.5965882497844562e-06, + "loss": 0.2019, + "step": 38031 + }, + { + "epoch": 14.04, + "learning_rate": 1.596551299421111e-06, + "loss": 0.2088, + "step": 38032 + }, + { + "epoch": 14.04, + "learning_rate": 1.5965143490577657e-06, + "loss": 0.3486, + "step": 38033 + }, + { + "epoch": 14.05, + "learning_rate": 1.5964773986944207e-06, + "loss": 0.1905, + "step": 38034 + }, + { + "epoch": 14.05, + "learning_rate": 1.5964404483310754e-06, + "loss": 0.2232, + "step": 38035 + }, + { + "epoch": 14.05, + "learning_rate": 1.5964034979677302e-06, + "loss": 0.2129, + "step": 38036 + }, + { + "epoch": 14.05, + "learning_rate": 1.5963665476043849e-06, + "loss": 0.2499, + "step": 38037 + }, + { + "epoch": 14.05, + "learning_rate": 1.5963295972410395e-06, + "loss": 0.2857, + "step": 38038 + }, + { + "epoch": 14.05, + "learning_rate": 1.5962926468776943e-06, + "loss": 0.2373, + "step": 38039 + }, + { + "epoch": 14.05, + "learning_rate": 1.596255696514349e-06, + "loss": 0.1849, + "step": 38040 + }, + { + "epoch": 14.05, + "learning_rate": 1.596218746151004e-06, + "loss": 0.4113, + "step": 38041 + }, + { + "epoch": 14.05, + "learning_rate": 1.5961817957876585e-06, + "loss": 0.2723, + "step": 38042 + }, + { + "epoch": 14.05, + "learning_rate": 1.5961448454243135e-06, + "loss": 0.1655, + "step": 38043 + }, + { + "epoch": 14.05, + "learning_rate": 1.5961078950609682e-06, + "loss": 0.2125, + "step": 38044 + }, + { + "epoch": 14.05, + "learning_rate": 1.596070944697623e-06, + "loss": 0.1554, + "step": 38045 + }, + { + "epoch": 14.05, + "learning_rate": 1.5960339943342776e-06, + "loss": 0.2195, + "step": 38046 + }, + { + "epoch": 14.05, + "learning_rate": 1.5959970439709323e-06, + "loss": 0.3455, + "step": 38047 + }, + { + "epoch": 14.05, + "learning_rate": 1.5959600936075871e-06, + "loss": 0.2089, + "step": 38048 + }, + { + "epoch": 14.05, + "learning_rate": 1.5959231432442418e-06, + "loss": 0.2467, + "step": 38049 + }, + { + "epoch": 14.05, + "learning_rate": 1.5958861928808968e-06, + "loss": 0.3853, + "step": 38050 + }, + { + "epoch": 14.05, + "learning_rate": 1.5958492425175515e-06, + "loss": 0.2113, + "step": 38051 + }, + { + "epoch": 14.05, + "learning_rate": 1.5958122921542063e-06, + "loss": 0.2896, + "step": 38052 + }, + { + "epoch": 14.05, + "learning_rate": 1.595775341790861e-06, + "loss": 0.216, + "step": 38053 + }, + { + "epoch": 14.05, + "learning_rate": 1.5957383914275158e-06, + "loss": 0.202, + "step": 38054 + }, + { + "epoch": 14.05, + "learning_rate": 1.5957014410641704e-06, + "loss": 0.2118, + "step": 38055 + }, + { + "epoch": 14.05, + "learning_rate": 1.5956644907008255e-06, + "loss": 0.3119, + "step": 38056 + }, + { + "epoch": 14.05, + "learning_rate": 1.5956275403374801e-06, + "loss": 0.1685, + "step": 38057 + }, + { + "epoch": 14.05, + "learning_rate": 1.5955905899741346e-06, + "loss": 0.3437, + "step": 38058 + }, + { + "epoch": 14.05, + "learning_rate": 1.5955536396107896e-06, + "loss": 0.2721, + "step": 38059 + }, + { + "epoch": 14.05, + "learning_rate": 1.5955166892474443e-06, + "loss": 0.2555, + "step": 38060 + }, + { + "epoch": 14.06, + "learning_rate": 1.595479738884099e-06, + "loss": 0.4837, + "step": 38061 + }, + { + "epoch": 14.06, + "learning_rate": 1.5954427885207537e-06, + "loss": 0.166, + "step": 38062 + }, + { + "epoch": 14.06, + "learning_rate": 1.5954058381574088e-06, + "loss": 0.1698, + "step": 38063 + }, + { + "epoch": 14.06, + "learning_rate": 1.5953688877940632e-06, + "loss": 0.2648, + "step": 38064 + }, + { + "epoch": 14.06, + "learning_rate": 1.5953319374307183e-06, + "loss": 0.2274, + "step": 38065 + }, + { + "epoch": 14.06, + "learning_rate": 1.595294987067373e-06, + "loss": 0.3048, + "step": 38066 + }, + { + "epoch": 14.06, + "learning_rate": 1.5952580367040276e-06, + "loss": 0.2989, + "step": 38067 + }, + { + "epoch": 14.06, + "learning_rate": 1.5952210863406824e-06, + "loss": 0.2652, + "step": 38068 + }, + { + "epoch": 14.06, + "learning_rate": 1.595184135977337e-06, + "loss": 0.3878, + "step": 38069 + }, + { + "epoch": 14.06, + "learning_rate": 1.5951471856139919e-06, + "loss": 0.4001, + "step": 38070 + }, + { + "epoch": 14.06, + "learning_rate": 1.5951102352506465e-06, + "loss": 0.1278, + "step": 38071 + }, + { + "epoch": 14.06, + "learning_rate": 1.5950732848873016e-06, + "loss": 0.2771, + "step": 38072 + }, + { + "epoch": 14.06, + "learning_rate": 1.5950363345239562e-06, + "loss": 0.2589, + "step": 38073 + }, + { + "epoch": 14.06, + "learning_rate": 1.594999384160611e-06, + "loss": 0.3362, + "step": 38074 + }, + { + "epoch": 14.06, + "learning_rate": 1.5949624337972657e-06, + "loss": 0.2584, + "step": 38075 + }, + { + "epoch": 14.06, + "learning_rate": 1.5949254834339206e-06, + "loss": 0.2179, + "step": 38076 + }, + { + "epoch": 14.06, + "learning_rate": 1.5948885330705752e-06, + "loss": 0.277, + "step": 38077 + }, + { + "epoch": 14.06, + "learning_rate": 1.5948515827072298e-06, + "loss": 0.3598, + "step": 38078 + }, + { + "epoch": 14.06, + "learning_rate": 1.5948146323438849e-06, + "loss": 0.2241, + "step": 38079 + }, + { + "epoch": 14.06, + "learning_rate": 1.5947776819805393e-06, + "loss": 0.2685, + "step": 38080 + }, + { + "epoch": 14.06, + "learning_rate": 1.5947407316171944e-06, + "loss": 0.2792, + "step": 38081 + }, + { + "epoch": 14.06, + "learning_rate": 1.594703781253849e-06, + "loss": 0.4003, + "step": 38082 + }, + { + "epoch": 14.06, + "learning_rate": 1.5946668308905039e-06, + "loss": 0.2803, + "step": 38083 + }, + { + "epoch": 14.06, + "learning_rate": 1.5946298805271585e-06, + "loss": 0.2386, + "step": 38084 + }, + { + "epoch": 14.06, + "learning_rate": 1.5945929301638136e-06, + "loss": 0.1997, + "step": 38085 + }, + { + "epoch": 14.06, + "learning_rate": 1.594555979800468e-06, + "loss": 0.18, + "step": 38086 + }, + { + "epoch": 14.06, + "learning_rate": 1.5945190294371226e-06, + "loss": 0.2585, + "step": 38087 + }, + { + "epoch": 14.06, + "learning_rate": 1.5944820790737777e-06, + "loss": 0.2989, + "step": 38088 + }, + { + "epoch": 14.07, + "learning_rate": 1.5944451287104323e-06, + "loss": 0.2996, + "step": 38089 + }, + { + "epoch": 14.07, + "learning_rate": 1.5944081783470872e-06, + "loss": 0.2091, + "step": 38090 + }, + { + "epoch": 14.07, + "learning_rate": 1.5943712279837418e-06, + "loss": 0.2321, + "step": 38091 + }, + { + "epoch": 14.07, + "learning_rate": 1.5943342776203966e-06, + "loss": 0.215, + "step": 38092 + }, + { + "epoch": 14.07, + "learning_rate": 1.5942973272570513e-06, + "loss": 0.3081, + "step": 38093 + }, + { + "epoch": 14.07, + "learning_rate": 1.5942603768937063e-06, + "loss": 0.1574, + "step": 38094 + }, + { + "epoch": 14.07, + "learning_rate": 1.594223426530361e-06, + "loss": 0.4228, + "step": 38095 + }, + { + "epoch": 14.07, + "learning_rate": 1.5941864761670156e-06, + "loss": 0.359, + "step": 38096 + }, + { + "epoch": 14.07, + "learning_rate": 1.5941495258036705e-06, + "loss": 0.3114, + "step": 38097 + }, + { + "epoch": 14.07, + "learning_rate": 1.594112575440325e-06, + "loss": 0.2593, + "step": 38098 + }, + { + "epoch": 14.07, + "learning_rate": 1.59407562507698e-06, + "loss": 0.252, + "step": 38099 + }, + { + "epoch": 14.07, + "learning_rate": 1.5940386747136346e-06, + "loss": 0.1417, + "step": 38100 + }, + { + "epoch": 14.07, + "learning_rate": 1.5940017243502896e-06, + "loss": 0.2498, + "step": 38101 + }, + { + "epoch": 14.07, + "learning_rate": 1.5939647739869443e-06, + "loss": 0.1795, + "step": 38102 + }, + { + "epoch": 14.07, + "learning_rate": 1.5939278236235991e-06, + "loss": 0.2213, + "step": 38103 + }, + { + "epoch": 14.07, + "learning_rate": 1.5938908732602538e-06, + "loss": 0.3187, + "step": 38104 + }, + { + "epoch": 14.07, + "learning_rate": 1.5938539228969086e-06, + "loss": 0.3415, + "step": 38105 + }, + { + "epoch": 14.07, + "learning_rate": 1.5938169725335632e-06, + "loss": 0.2223, + "step": 38106 + }, + { + "epoch": 14.07, + "learning_rate": 1.5937800221702179e-06, + "loss": 0.1991, + "step": 38107 + }, + { + "epoch": 14.07, + "learning_rate": 1.593743071806873e-06, + "loss": 0.2823, + "step": 38108 + }, + { + "epoch": 14.07, + "learning_rate": 1.5937061214435274e-06, + "loss": 0.3135, + "step": 38109 + }, + { + "epoch": 14.07, + "learning_rate": 1.5936691710801824e-06, + "loss": 0.3063, + "step": 38110 + }, + { + "epoch": 14.07, + "learning_rate": 1.593632220716837e-06, + "loss": 0.2452, + "step": 38111 + }, + { + "epoch": 14.07, + "learning_rate": 1.593595270353492e-06, + "loss": 0.4066, + "step": 38112 + }, + { + "epoch": 14.07, + "learning_rate": 1.5935583199901465e-06, + "loss": 0.3162, + "step": 38113 + }, + { + "epoch": 14.07, + "learning_rate": 1.5935213696268016e-06, + "loss": 0.2141, + "step": 38114 + }, + { + "epoch": 14.07, + "learning_rate": 1.593484419263456e-06, + "loss": 0.182, + "step": 38115 + }, + { + "epoch": 14.08, + "learning_rate": 1.5934474689001107e-06, + "loss": 0.2889, + "step": 38116 + }, + { + "epoch": 14.08, + "learning_rate": 1.5934105185367657e-06, + "loss": 0.3512, + "step": 38117 + }, + { + "epoch": 14.08, + "learning_rate": 1.5933735681734204e-06, + "loss": 0.2059, + "step": 38118 + }, + { + "epoch": 14.08, + "learning_rate": 1.5933366178100752e-06, + "loss": 0.3607, + "step": 38119 + }, + { + "epoch": 14.08, + "learning_rate": 1.5932996674467299e-06, + "loss": 0.1405, + "step": 38120 + }, + { + "epoch": 14.08, + "learning_rate": 1.5932627170833847e-06, + "loss": 0.312, + "step": 38121 + }, + { + "epoch": 14.08, + "learning_rate": 1.5932257667200393e-06, + "loss": 0.3803, + "step": 38122 + }, + { + "epoch": 14.08, + "learning_rate": 1.5931888163566944e-06, + "loss": 0.3355, + "step": 38123 + }, + { + "epoch": 14.08, + "learning_rate": 1.593151865993349e-06, + "loss": 0.1953, + "step": 38124 + }, + { + "epoch": 14.08, + "learning_rate": 1.5931149156300039e-06, + "loss": 0.2469, + "step": 38125 + }, + { + "epoch": 14.08, + "learning_rate": 1.5930779652666585e-06, + "loss": 0.207, + "step": 38126 + }, + { + "epoch": 14.08, + "learning_rate": 1.5930410149033132e-06, + "loss": 0.198, + "step": 38127 + }, + { + "epoch": 14.08, + "learning_rate": 1.593004064539968e-06, + "loss": 0.2928, + "step": 38128 + }, + { + "epoch": 14.08, + "learning_rate": 1.5929671141766226e-06, + "loss": 0.2683, + "step": 38129 + }, + { + "epoch": 14.08, + "learning_rate": 1.5929301638132777e-06, + "loss": 0.2489, + "step": 38130 + }, + { + "epoch": 14.08, + "learning_rate": 1.5928932134499321e-06, + "loss": 0.1636, + "step": 38131 + }, + { + "epoch": 14.08, + "learning_rate": 1.5928562630865872e-06, + "loss": 0.2287, + "step": 38132 + }, + { + "epoch": 14.08, + "learning_rate": 1.5928193127232418e-06, + "loss": 0.3409, + "step": 38133 + }, + { + "epoch": 14.08, + "learning_rate": 1.5927823623598967e-06, + "loss": 0.3466, + "step": 38134 + }, + { + "epoch": 14.08, + "learning_rate": 1.5927454119965513e-06, + "loss": 0.3324, + "step": 38135 + }, + { + "epoch": 14.08, + "learning_rate": 1.592708461633206e-06, + "loss": 0.1681, + "step": 38136 + }, + { + "epoch": 14.08, + "learning_rate": 1.5926715112698608e-06, + "loss": 0.1882, + "step": 38137 + }, + { + "epoch": 14.08, + "learning_rate": 1.5926345609065154e-06, + "loss": 0.3545, + "step": 38138 + }, + { + "epoch": 14.08, + "learning_rate": 1.5925976105431705e-06, + "loss": 0.29, + "step": 38139 + }, + { + "epoch": 14.08, + "learning_rate": 1.5925606601798251e-06, + "loss": 0.426, + "step": 38140 + }, + { + "epoch": 14.08, + "learning_rate": 1.59252370981648e-06, + "loss": 0.2445, + "step": 38141 + }, + { + "epoch": 14.08, + "learning_rate": 1.5924867594531346e-06, + "loss": 0.3069, + "step": 38142 + }, + { + "epoch": 14.09, + "learning_rate": 1.5924498090897895e-06, + "loss": 0.2975, + "step": 38143 + }, + { + "epoch": 14.09, + "learning_rate": 1.592412858726444e-06, + "loss": 0.2022, + "step": 38144 + }, + { + "epoch": 14.09, + "learning_rate": 1.5923759083630992e-06, + "loss": 0.3597, + "step": 38145 + }, + { + "epoch": 14.09, + "learning_rate": 1.5923389579997538e-06, + "loss": 0.2807, + "step": 38146 + }, + { + "epoch": 14.09, + "learning_rate": 1.5923020076364084e-06, + "loss": 0.3078, + "step": 38147 + }, + { + "epoch": 14.09, + "learning_rate": 1.5922650572730633e-06, + "loss": 0.301, + "step": 38148 + }, + { + "epoch": 14.09, + "learning_rate": 1.592228106909718e-06, + "loss": 0.2989, + "step": 38149 + }, + { + "epoch": 14.09, + "learning_rate": 1.5921911565463728e-06, + "loss": 0.2437, + "step": 38150 + }, + { + "epoch": 14.09, + "learning_rate": 1.5921542061830274e-06, + "loss": 0.2991, + "step": 38151 + }, + { + "epoch": 14.09, + "learning_rate": 1.5921172558196825e-06, + "loss": 0.2152, + "step": 38152 + }, + { + "epoch": 14.09, + "learning_rate": 1.5920803054563369e-06, + "loss": 0.2475, + "step": 38153 + }, + { + "epoch": 14.09, + "learning_rate": 1.592043355092992e-06, + "loss": 0.3065, + "step": 38154 + }, + { + "epoch": 14.09, + "learning_rate": 1.5920064047296466e-06, + "loss": 0.269, + "step": 38155 + }, + { + "epoch": 14.09, + "learning_rate": 1.5919694543663012e-06, + "loss": 0.2498, + "step": 38156 + }, + { + "epoch": 14.09, + "learning_rate": 1.591932504002956e-06, + "loss": 0.2354, + "step": 38157 + }, + { + "epoch": 14.09, + "learning_rate": 1.5918955536396107e-06, + "loss": 0.2339, + "step": 38158 + }, + { + "epoch": 14.09, + "learning_rate": 1.5918586032762655e-06, + "loss": 0.2498, + "step": 38159 + }, + { + "epoch": 14.09, + "learning_rate": 1.5918216529129202e-06, + "loss": 0.3803, + "step": 38160 + }, + { + "epoch": 14.09, + "learning_rate": 1.5917847025495752e-06, + "loss": 0.3661, + "step": 38161 + }, + { + "epoch": 14.09, + "learning_rate": 1.5917477521862299e-06, + "loss": 0.3991, + "step": 38162 + }, + { + "epoch": 14.09, + "learning_rate": 1.5917108018228847e-06, + "loss": 0.2578, + "step": 38163 + }, + { + "epoch": 14.09, + "learning_rate": 1.5916738514595394e-06, + "loss": 0.217, + "step": 38164 + }, + { + "epoch": 14.09, + "learning_rate": 1.5916369010961942e-06, + "loss": 0.2197, + "step": 38165 + }, + { + "epoch": 14.09, + "learning_rate": 1.5915999507328488e-06, + "loss": 0.2185, + "step": 38166 + }, + { + "epoch": 14.09, + "learning_rate": 1.5915630003695035e-06, + "loss": 0.2289, + "step": 38167 + }, + { + "epoch": 14.09, + "learning_rate": 1.5915260500061585e-06, + "loss": 0.2048, + "step": 38168 + }, + { + "epoch": 14.09, + "learning_rate": 1.5914890996428132e-06, + "loss": 0.2264, + "step": 38169 + }, + { + "epoch": 14.1, + "learning_rate": 1.591452149279468e-06, + "loss": 0.2562, + "step": 38170 + }, + { + "epoch": 14.1, + "learning_rate": 1.5914151989161227e-06, + "loss": 0.2762, + "step": 38171 + }, + { + "epoch": 14.1, + "learning_rate": 1.5913782485527775e-06, + "loss": 0.2565, + "step": 38172 + }, + { + "epoch": 14.1, + "learning_rate": 1.5913412981894321e-06, + "loss": 0.2635, + "step": 38173 + }, + { + "epoch": 14.1, + "learning_rate": 1.5913043478260872e-06, + "loss": 0.2971, + "step": 38174 + }, + { + "epoch": 14.1, + "learning_rate": 1.5912673974627418e-06, + "loss": 0.3121, + "step": 38175 + }, + { + "epoch": 14.1, + "learning_rate": 1.5912304470993963e-06, + "loss": 0.2116, + "step": 38176 + }, + { + "epoch": 14.1, + "learning_rate": 1.5911934967360513e-06, + "loss": 0.271, + "step": 38177 + }, + { + "epoch": 14.1, + "learning_rate": 1.591156546372706e-06, + "loss": 0.2768, + "step": 38178 + }, + { + "epoch": 14.1, + "learning_rate": 1.5911195960093608e-06, + "loss": 0.2114, + "step": 38179 + }, + { + "epoch": 14.1, + "learning_rate": 1.5910826456460154e-06, + "loss": 0.3036, + "step": 38180 + }, + { + "epoch": 14.1, + "learning_rate": 1.5910456952826705e-06, + "loss": 0.363, + "step": 38181 + }, + { + "epoch": 14.1, + "learning_rate": 1.591008744919325e-06, + "loss": 0.3605, + "step": 38182 + }, + { + "epoch": 14.1, + "learning_rate": 1.59097179455598e-06, + "loss": 0.4593, + "step": 38183 + }, + { + "epoch": 14.1, + "learning_rate": 1.5909348441926346e-06, + "loss": 0.1748, + "step": 38184 + }, + { + "epoch": 14.1, + "learning_rate": 1.5908978938292895e-06, + "loss": 0.2429, + "step": 38185 + }, + { + "epoch": 14.1, + "learning_rate": 1.5908609434659441e-06, + "loss": 0.2709, + "step": 38186 + }, + { + "epoch": 14.1, + "learning_rate": 1.5908239931025988e-06, + "loss": 0.3097, + "step": 38187 + }, + { + "epoch": 14.1, + "learning_rate": 1.5907870427392536e-06, + "loss": 0.3435, + "step": 38188 + }, + { + "epoch": 14.1, + "learning_rate": 1.5907500923759082e-06, + "loss": 0.314, + "step": 38189 + }, + { + "epoch": 14.1, + "learning_rate": 1.5907131420125633e-06, + "loss": 0.3061, + "step": 38190 + }, + { + "epoch": 14.1, + "learning_rate": 1.590676191649218e-06, + "loss": 0.2902, + "step": 38191 + }, + { + "epoch": 14.1, + "learning_rate": 1.5906392412858728e-06, + "loss": 0.2619, + "step": 38192 + }, + { + "epoch": 14.1, + "learning_rate": 1.5906022909225274e-06, + "loss": 0.2754, + "step": 38193 + }, + { + "epoch": 14.1, + "learning_rate": 1.5905653405591823e-06, + "loss": 0.23, + "step": 38194 + }, + { + "epoch": 14.1, + "learning_rate": 1.590528390195837e-06, + "loss": 0.2385, + "step": 38195 + }, + { + "epoch": 14.1, + "learning_rate": 1.5904914398324915e-06, + "loss": 0.3926, + "step": 38196 + }, + { + "epoch": 14.11, + "learning_rate": 1.5904544894691466e-06, + "loss": 0.2981, + "step": 38197 + }, + { + "epoch": 14.11, + "learning_rate": 1.590417539105801e-06, + "loss": 0.2513, + "step": 38198 + }, + { + "epoch": 14.11, + "learning_rate": 1.590380588742456e-06, + "loss": 0.2482, + "step": 38199 + }, + { + "epoch": 14.11, + "learning_rate": 1.5903436383791107e-06, + "loss": 0.3117, + "step": 38200 + }, + { + "epoch": 14.11, + "learning_rate": 1.5903066880157656e-06, + "loss": 0.2187, + "step": 38201 + }, + { + "epoch": 14.11, + "learning_rate": 1.5902697376524202e-06, + "loss": 0.2808, + "step": 38202 + }, + { + "epoch": 14.11, + "learning_rate": 1.5902327872890753e-06, + "loss": 0.394, + "step": 38203 + }, + { + "epoch": 14.11, + "learning_rate": 1.5901958369257297e-06, + "loss": 0.2881, + "step": 38204 + }, + { + "epoch": 14.11, + "learning_rate": 1.5901588865623847e-06, + "loss": 0.3681, + "step": 38205 + }, + { + "epoch": 14.11, + "learning_rate": 1.5901219361990394e-06, + "loss": 0.3521, + "step": 38206 + }, + { + "epoch": 14.11, + "learning_rate": 1.590084985835694e-06, + "loss": 0.2944, + "step": 38207 + }, + { + "epoch": 14.11, + "learning_rate": 1.5900480354723489e-06, + "loss": 0.2993, + "step": 38208 + }, + { + "epoch": 14.11, + "learning_rate": 1.5900110851090035e-06, + "loss": 0.2166, + "step": 38209 + }, + { + "epoch": 14.11, + "learning_rate": 1.5899741347456584e-06, + "loss": 0.1899, + "step": 38210 + }, + { + "epoch": 14.11, + "learning_rate": 1.589937184382313e-06, + "loss": 0.325, + "step": 38211 + }, + { + "epoch": 14.11, + "learning_rate": 1.589900234018968e-06, + "loss": 0.2865, + "step": 38212 + }, + { + "epoch": 14.11, + "learning_rate": 1.5898632836556227e-06, + "loss": 0.369, + "step": 38213 + }, + { + "epoch": 14.11, + "learning_rate": 1.5898263332922775e-06, + "loss": 0.315, + "step": 38214 + }, + { + "epoch": 14.11, + "learning_rate": 1.5897893829289322e-06, + "loss": 0.3355, + "step": 38215 + }, + { + "epoch": 14.11, + "learning_rate": 1.5897524325655868e-06, + "loss": 0.2187, + "step": 38216 + }, + { + "epoch": 14.11, + "learning_rate": 1.5897154822022417e-06, + "loss": 0.3153, + "step": 38217 + }, + { + "epoch": 14.11, + "learning_rate": 1.5896785318388963e-06, + "loss": 0.2518, + "step": 38218 + }, + { + "epoch": 14.11, + "learning_rate": 1.5896415814755514e-06, + "loss": 0.2429, + "step": 38219 + }, + { + "epoch": 14.11, + "learning_rate": 1.589604631112206e-06, + "loss": 0.2685, + "step": 38220 + }, + { + "epoch": 14.11, + "learning_rate": 1.5895676807488608e-06, + "loss": 0.1878, + "step": 38221 + }, + { + "epoch": 14.11, + "learning_rate": 1.5895307303855155e-06, + "loss": 0.2856, + "step": 38222 + }, + { + "epoch": 14.11, + "learning_rate": 1.5894937800221703e-06, + "loss": 0.2051, + "step": 38223 + }, + { + "epoch": 14.12, + "learning_rate": 1.589456829658825e-06, + "loss": 0.3034, + "step": 38224 + }, + { + "epoch": 14.12, + "learning_rate": 1.58941987929548e-06, + "loss": 0.3242, + "step": 38225 + }, + { + "epoch": 14.12, + "learning_rate": 1.5893829289321347e-06, + "loss": 0.263, + "step": 38226 + }, + { + "epoch": 14.12, + "learning_rate": 1.589345978568789e-06, + "loss": 0.3471, + "step": 38227 + }, + { + "epoch": 14.12, + "learning_rate": 1.5893090282054441e-06, + "loss": 0.2419, + "step": 38228 + }, + { + "epoch": 14.12, + "learning_rate": 1.5892720778420988e-06, + "loss": 0.2014, + "step": 38229 + }, + { + "epoch": 14.12, + "learning_rate": 1.5892351274787536e-06, + "loss": 0.195, + "step": 38230 + }, + { + "epoch": 14.12, + "learning_rate": 1.5891981771154083e-06, + "loss": 0.1723, + "step": 38231 + }, + { + "epoch": 14.12, + "learning_rate": 1.5891612267520631e-06, + "loss": 0.2701, + "step": 38232 + }, + { + "epoch": 14.12, + "learning_rate": 1.5891242763887177e-06, + "loss": 0.3358, + "step": 38233 + }, + { + "epoch": 14.12, + "learning_rate": 1.5890873260253728e-06, + "loss": 0.3289, + "step": 38234 + }, + { + "epoch": 14.12, + "learning_rate": 1.5890503756620274e-06, + "loss": 0.2209, + "step": 38235 + }, + { + "epoch": 14.12, + "learning_rate": 1.589013425298682e-06, + "loss": 0.3345, + "step": 38236 + }, + { + "epoch": 14.12, + "learning_rate": 1.588976474935337e-06, + "loss": 0.2104, + "step": 38237 + }, + { + "epoch": 14.12, + "learning_rate": 1.5889395245719916e-06, + "loss": 0.2773, + "step": 38238 + }, + { + "epoch": 14.12, + "learning_rate": 1.5889025742086464e-06, + "loss": 0.3867, + "step": 38239 + }, + { + "epoch": 14.12, + "learning_rate": 1.588865623845301e-06, + "loss": 0.3893, + "step": 38240 + }, + { + "epoch": 14.12, + "learning_rate": 1.5888286734819561e-06, + "loss": 0.287, + "step": 38241 + }, + { + "epoch": 14.12, + "learning_rate": 1.5887917231186107e-06, + "loss": 0.3467, + "step": 38242 + }, + { + "epoch": 14.12, + "learning_rate": 1.5887547727552656e-06, + "loss": 0.2484, + "step": 38243 + }, + { + "epoch": 14.12, + "learning_rate": 1.5887178223919202e-06, + "loss": 0.2629, + "step": 38244 + }, + { + "epoch": 14.12, + "learning_rate": 1.588680872028575e-06, + "loss": 0.1949, + "step": 38245 + }, + { + "epoch": 14.12, + "learning_rate": 1.5886439216652297e-06, + "loss": 0.3313, + "step": 38246 + }, + { + "epoch": 14.12, + "learning_rate": 1.5886069713018843e-06, + "loss": 0.2874, + "step": 38247 + }, + { + "epoch": 14.12, + "learning_rate": 1.5885700209385394e-06, + "loss": 0.2953, + "step": 38248 + }, + { + "epoch": 14.12, + "learning_rate": 1.5885330705751938e-06, + "loss": 0.339, + "step": 38249 + }, + { + "epoch": 14.12, + "learning_rate": 1.5884961202118489e-06, + "loss": 0.287, + "step": 38250 + }, + { + "epoch": 14.13, + "learning_rate": 1.5884591698485035e-06, + "loss": 0.2702, + "step": 38251 + }, + { + "epoch": 14.13, + "learning_rate": 1.5884222194851584e-06, + "loss": 0.2304, + "step": 38252 + }, + { + "epoch": 14.13, + "learning_rate": 1.588385269121813e-06, + "loss": 0.2813, + "step": 38253 + }, + { + "epoch": 14.13, + "learning_rate": 1.588348318758468e-06, + "loss": 0.2239, + "step": 38254 + }, + { + "epoch": 14.13, + "learning_rate": 1.5883113683951225e-06, + "loss": 0.2291, + "step": 38255 + }, + { + "epoch": 14.13, + "learning_rate": 1.5882744180317771e-06, + "loss": 0.2771, + "step": 38256 + }, + { + "epoch": 14.13, + "learning_rate": 1.5882374676684322e-06, + "loss": 0.2609, + "step": 38257 + }, + { + "epoch": 14.13, + "learning_rate": 1.5882005173050868e-06, + "loss": 0.2331, + "step": 38258 + }, + { + "epoch": 14.13, + "learning_rate": 1.5881635669417417e-06, + "loss": 0.2595, + "step": 38259 + }, + { + "epoch": 14.13, + "learning_rate": 1.5881266165783963e-06, + "loss": 0.2931, + "step": 38260 + }, + { + "epoch": 14.13, + "learning_rate": 1.5880896662150512e-06, + "loss": 0.3415, + "step": 38261 + }, + { + "epoch": 14.13, + "learning_rate": 1.5880527158517058e-06, + "loss": 0.2426, + "step": 38262 + }, + { + "epoch": 14.13, + "learning_rate": 1.5880157654883609e-06, + "loss": 0.1763, + "step": 38263 + }, + { + "epoch": 14.13, + "learning_rate": 1.5879788151250155e-06, + "loss": 0.1358, + "step": 38264 + }, + { + "epoch": 14.13, + "learning_rate": 1.5879418647616703e-06, + "loss": 0.2327, + "step": 38265 + }, + { + "epoch": 14.13, + "learning_rate": 1.587904914398325e-06, + "loss": 0.2103, + "step": 38266 + }, + { + "epoch": 14.13, + "learning_rate": 1.5878679640349796e-06, + "loss": 0.2547, + "step": 38267 + }, + { + "epoch": 14.13, + "learning_rate": 1.5878310136716345e-06, + "loss": 0.2338, + "step": 38268 + }, + { + "epoch": 14.13, + "learning_rate": 1.587794063308289e-06, + "loss": 0.1497, + "step": 38269 + }, + { + "epoch": 14.13, + "learning_rate": 1.5877571129449442e-06, + "loss": 0.277, + "step": 38270 + }, + { + "epoch": 14.13, + "learning_rate": 1.5877201625815986e-06, + "loss": 0.4078, + "step": 38271 + }, + { + "epoch": 14.13, + "learning_rate": 1.5876832122182536e-06, + "loss": 0.3253, + "step": 38272 + }, + { + "epoch": 14.13, + "learning_rate": 1.5876462618549083e-06, + "loss": 0.302, + "step": 38273 + }, + { + "epoch": 14.13, + "learning_rate": 1.5876093114915631e-06, + "loss": 0.3195, + "step": 38274 + }, + { + "epoch": 14.13, + "learning_rate": 1.5875723611282178e-06, + "loss": 0.255, + "step": 38275 + }, + { + "epoch": 14.13, + "learning_rate": 1.5875354107648724e-06, + "loss": 0.2123, + "step": 38276 + }, + { + "epoch": 14.13, + "learning_rate": 1.5874984604015273e-06, + "loss": 0.1871, + "step": 38277 + }, + { + "epoch": 14.14, + "learning_rate": 1.5874615100381819e-06, + "loss": 0.2297, + "step": 38278 + }, + { + "epoch": 14.14, + "learning_rate": 1.587424559674837e-06, + "loss": 0.2834, + "step": 38279 + }, + { + "epoch": 14.14, + "learning_rate": 1.5873876093114916e-06, + "loss": 0.25, + "step": 38280 + }, + { + "epoch": 14.14, + "learning_rate": 1.5873506589481464e-06, + "loss": 0.256, + "step": 38281 + }, + { + "epoch": 14.14, + "learning_rate": 1.587313708584801e-06, + "loss": 0.2736, + "step": 38282 + }, + { + "epoch": 14.14, + "learning_rate": 1.587276758221456e-06, + "loss": 0.1815, + "step": 38283 + }, + { + "epoch": 14.14, + "learning_rate": 1.5872398078581106e-06, + "loss": 0.3456, + "step": 38284 + }, + { + "epoch": 14.14, + "learning_rate": 1.5872028574947656e-06, + "loss": 0.2334, + "step": 38285 + }, + { + "epoch": 14.14, + "learning_rate": 1.5871659071314203e-06, + "loss": 0.191, + "step": 38286 + }, + { + "epoch": 14.14, + "learning_rate": 1.5871289567680749e-06, + "loss": 0.3384, + "step": 38287 + }, + { + "epoch": 14.14, + "learning_rate": 1.5870920064047297e-06, + "loss": 0.2348, + "step": 38288 + }, + { + "epoch": 14.14, + "learning_rate": 1.5870550560413844e-06, + "loss": 0.258, + "step": 38289 + }, + { + "epoch": 14.14, + "learning_rate": 1.5870181056780392e-06, + "loss": 0.2334, + "step": 38290 + }, + { + "epoch": 14.14, + "learning_rate": 1.5869811553146939e-06, + "loss": 0.2056, + "step": 38291 + }, + { + "epoch": 14.14, + "learning_rate": 1.586944204951349e-06, + "loss": 0.3633, + "step": 38292 + }, + { + "epoch": 14.14, + "learning_rate": 1.5869072545880036e-06, + "loss": 0.4481, + "step": 38293 + }, + { + "epoch": 14.14, + "learning_rate": 1.5868703042246584e-06, + "loss": 0.3015, + "step": 38294 + }, + { + "epoch": 14.14, + "learning_rate": 1.586833353861313e-06, + "loss": 0.3915, + "step": 38295 + }, + { + "epoch": 14.14, + "learning_rate": 1.5867964034979677e-06, + "loss": 0.2212, + "step": 38296 + }, + { + "epoch": 14.14, + "learning_rate": 1.5867594531346225e-06, + "loss": 0.3415, + "step": 38297 + }, + { + "epoch": 14.14, + "learning_rate": 1.5867225027712772e-06, + "loss": 0.1681, + "step": 38298 + }, + { + "epoch": 14.14, + "learning_rate": 1.5866855524079322e-06, + "loss": 0.233, + "step": 38299 + }, + { + "epoch": 14.14, + "learning_rate": 1.5866486020445866e-06, + "loss": 0.2036, + "step": 38300 + }, + { + "epoch": 14.14, + "learning_rate": 1.5866116516812417e-06, + "loss": 0.3821, + "step": 38301 + }, + { + "epoch": 14.14, + "learning_rate": 1.5865747013178963e-06, + "loss": 0.2546, + "step": 38302 + }, + { + "epoch": 14.14, + "learning_rate": 1.5865377509545512e-06, + "loss": 0.2485, + "step": 38303 + }, + { + "epoch": 14.14, + "learning_rate": 1.5865008005912058e-06, + "loss": 0.2546, + "step": 38304 + }, + { + "epoch": 14.15, + "learning_rate": 1.5864638502278609e-06, + "loss": 0.3342, + "step": 38305 + }, + { + "epoch": 14.15, + "learning_rate": 1.5864268998645153e-06, + "loss": 0.3113, + "step": 38306 + }, + { + "epoch": 14.15, + "learning_rate": 1.58638994950117e-06, + "loss": 0.2027, + "step": 38307 + }, + { + "epoch": 14.15, + "learning_rate": 1.586352999137825e-06, + "loss": 0.239, + "step": 38308 + }, + { + "epoch": 14.15, + "learning_rate": 1.5863160487744796e-06, + "loss": 0.24, + "step": 38309 + }, + { + "epoch": 14.15, + "learning_rate": 1.5862790984111345e-06, + "loss": 0.2651, + "step": 38310 + }, + { + "epoch": 14.15, + "learning_rate": 1.5862421480477891e-06, + "loss": 0.207, + "step": 38311 + }, + { + "epoch": 14.15, + "learning_rate": 1.586205197684444e-06, + "loss": 0.4259, + "step": 38312 + }, + { + "epoch": 14.15, + "learning_rate": 1.5861682473210986e-06, + "loss": 0.1715, + "step": 38313 + }, + { + "epoch": 14.15, + "learning_rate": 1.5861312969577537e-06, + "loss": 0.217, + "step": 38314 + }, + { + "epoch": 14.15, + "learning_rate": 1.5860943465944083e-06, + "loss": 0.3279, + "step": 38315 + }, + { + "epoch": 14.15, + "learning_rate": 1.5860573962310627e-06, + "loss": 0.225, + "step": 38316 + }, + { + "epoch": 14.15, + "learning_rate": 1.5860204458677178e-06, + "loss": 0.325, + "step": 38317 + }, + { + "epoch": 14.15, + "learning_rate": 1.5859834955043724e-06, + "loss": 0.2649, + "step": 38318 + }, + { + "epoch": 14.15, + "learning_rate": 1.5859465451410273e-06, + "loss": 0.1741, + "step": 38319 + }, + { + "epoch": 14.15, + "learning_rate": 1.585909594777682e-06, + "loss": 0.219, + "step": 38320 + }, + { + "epoch": 14.15, + "learning_rate": 1.585872644414337e-06, + "loss": 0.2865, + "step": 38321 + }, + { + "epoch": 14.15, + "learning_rate": 1.5858356940509914e-06, + "loss": 0.4065, + "step": 38322 + }, + { + "epoch": 14.15, + "learning_rate": 1.5857987436876465e-06, + "loss": 0.2499, + "step": 38323 + }, + { + "epoch": 14.15, + "learning_rate": 1.585761793324301e-06, + "loss": 0.3693, + "step": 38324 + }, + { + "epoch": 14.15, + "learning_rate": 1.585724842960956e-06, + "loss": 0.2611, + "step": 38325 + }, + { + "epoch": 14.15, + "learning_rate": 1.5856878925976106e-06, + "loss": 0.2111, + "step": 38326 + }, + { + "epoch": 14.15, + "learning_rate": 1.5856509422342652e-06, + "loss": 0.2016, + "step": 38327 + }, + { + "epoch": 14.15, + "learning_rate": 1.58561399187092e-06, + "loss": 0.4019, + "step": 38328 + }, + { + "epoch": 14.15, + "learning_rate": 1.5855770415075747e-06, + "loss": 0.3367, + "step": 38329 + }, + { + "epoch": 14.15, + "learning_rate": 1.5855400911442298e-06, + "loss": 0.2368, + "step": 38330 + }, + { + "epoch": 14.15, + "learning_rate": 1.5855031407808844e-06, + "loss": 0.2817, + "step": 38331 + }, + { + "epoch": 14.16, + "learning_rate": 1.5854661904175392e-06, + "loss": 0.2833, + "step": 38332 + }, + { + "epoch": 14.16, + "learning_rate": 1.5854292400541939e-06, + "loss": 0.3001, + "step": 38333 + }, + { + "epoch": 14.16, + "learning_rate": 1.5853922896908487e-06, + "loss": 0.3405, + "step": 38334 + }, + { + "epoch": 14.16, + "learning_rate": 1.5853553393275034e-06, + "loss": 0.3899, + "step": 38335 + }, + { + "epoch": 14.16, + "learning_rate": 1.585318388964158e-06, + "loss": 0.2813, + "step": 38336 + }, + { + "epoch": 14.16, + "learning_rate": 1.585281438600813e-06, + "loss": 0.1639, + "step": 38337 + }, + { + "epoch": 14.16, + "learning_rate": 1.5852444882374675e-06, + "loss": 0.2417, + "step": 38338 + }, + { + "epoch": 14.16, + "learning_rate": 1.5852075378741225e-06, + "loss": 0.2487, + "step": 38339 + }, + { + "epoch": 14.16, + "learning_rate": 1.5851705875107772e-06, + "loss": 0.1352, + "step": 38340 + }, + { + "epoch": 14.16, + "learning_rate": 1.585133637147432e-06, + "loss": 0.3315, + "step": 38341 + }, + { + "epoch": 14.16, + "learning_rate": 1.5850966867840867e-06, + "loss": 0.3756, + "step": 38342 + }, + { + "epoch": 14.16, + "learning_rate": 1.5850597364207417e-06, + "loss": 0.3079, + "step": 38343 + }, + { + "epoch": 14.16, + "learning_rate": 1.5850227860573962e-06, + "loss": 0.2314, + "step": 38344 + }, + { + "epoch": 14.16, + "learning_rate": 1.5849858356940508e-06, + "loss": 0.3656, + "step": 38345 + }, + { + "epoch": 14.16, + "learning_rate": 1.5849488853307058e-06, + "loss": 0.1182, + "step": 38346 + }, + { + "epoch": 14.16, + "learning_rate": 1.5849119349673605e-06, + "loss": 0.254, + "step": 38347 + }, + { + "epoch": 14.16, + "learning_rate": 1.5848749846040153e-06, + "loss": 0.197, + "step": 38348 + }, + { + "epoch": 14.16, + "learning_rate": 1.58483803424067e-06, + "loss": 0.2182, + "step": 38349 + }, + { + "epoch": 14.16, + "learning_rate": 1.5848010838773248e-06, + "loss": 0.2938, + "step": 38350 + }, + { + "epoch": 14.16, + "learning_rate": 1.5847641335139795e-06, + "loss": 0.2639, + "step": 38351 + }, + { + "epoch": 14.16, + "learning_rate": 1.5847271831506345e-06, + "loss": 0.2277, + "step": 38352 + }, + { + "epoch": 14.16, + "learning_rate": 1.5846902327872892e-06, + "loss": 0.3347, + "step": 38353 + }, + { + "epoch": 14.16, + "learning_rate": 1.584653282423944e-06, + "loss": 0.2707, + "step": 38354 + }, + { + "epoch": 14.16, + "learning_rate": 1.5846163320605986e-06, + "loss": 0.3113, + "step": 38355 + }, + { + "epoch": 14.16, + "learning_rate": 1.5845793816972533e-06, + "loss": 0.2874, + "step": 38356 + }, + { + "epoch": 14.16, + "learning_rate": 1.5845424313339081e-06, + "loss": 0.346, + "step": 38357 + }, + { + "epoch": 14.16, + "learning_rate": 1.5845054809705628e-06, + "loss": 0.2281, + "step": 38358 + }, + { + "epoch": 14.17, + "learning_rate": 1.5844685306072178e-06, + "loss": 0.2863, + "step": 38359 + }, + { + "epoch": 14.17, + "learning_rate": 1.5844315802438725e-06, + "loss": 0.2783, + "step": 38360 + }, + { + "epoch": 14.17, + "learning_rate": 1.5843946298805273e-06, + "loss": 0.3907, + "step": 38361 + }, + { + "epoch": 14.17, + "learning_rate": 1.584357679517182e-06, + "loss": 0.2357, + "step": 38362 + }, + { + "epoch": 14.17, + "learning_rate": 1.5843207291538368e-06, + "loss": 0.3159, + "step": 38363 + }, + { + "epoch": 14.17, + "learning_rate": 1.5842837787904914e-06, + "loss": 0.2253, + "step": 38364 + }, + { + "epoch": 14.17, + "learning_rate": 1.584246828427146e-06, + "loss": 0.2938, + "step": 38365 + }, + { + "epoch": 14.17, + "learning_rate": 1.5842098780638011e-06, + "loss": 0.1737, + "step": 38366 + }, + { + "epoch": 14.17, + "learning_rate": 1.5841729277004555e-06, + "loss": 0.3113, + "step": 38367 + }, + { + "epoch": 14.17, + "learning_rate": 1.5841359773371106e-06, + "loss": 0.3205, + "step": 38368 + }, + { + "epoch": 14.17, + "learning_rate": 1.5840990269737652e-06, + "loss": 0.2763, + "step": 38369 + }, + { + "epoch": 14.17, + "learning_rate": 1.58406207661042e-06, + "loss": 0.2805, + "step": 38370 + }, + { + "epoch": 14.17, + "learning_rate": 1.5840251262470747e-06, + "loss": 0.3695, + "step": 38371 + }, + { + "epoch": 14.17, + "learning_rate": 1.5839881758837298e-06, + "loss": 0.1767, + "step": 38372 + }, + { + "epoch": 14.17, + "learning_rate": 1.5839512255203842e-06, + "loss": 0.3852, + "step": 38373 + }, + { + "epoch": 14.17, + "learning_rate": 1.5839142751570393e-06, + "loss": 0.4163, + "step": 38374 + }, + { + "epoch": 14.17, + "learning_rate": 1.583877324793694e-06, + "loss": 0.1922, + "step": 38375 + }, + { + "epoch": 14.17, + "learning_rate": 1.5838403744303485e-06, + "loss": 0.3484, + "step": 38376 + }, + { + "epoch": 14.17, + "learning_rate": 1.5838034240670034e-06, + "loss": 0.2244, + "step": 38377 + }, + { + "epoch": 14.17, + "learning_rate": 1.583766473703658e-06, + "loss": 0.3015, + "step": 38378 + }, + { + "epoch": 14.17, + "learning_rate": 1.5837295233403129e-06, + "loss": 0.281, + "step": 38379 + }, + { + "epoch": 14.17, + "learning_rate": 1.5836925729769675e-06, + "loss": 0.287, + "step": 38380 + }, + { + "epoch": 14.17, + "learning_rate": 1.5836556226136226e-06, + "loss": 0.309, + "step": 38381 + }, + { + "epoch": 14.17, + "learning_rate": 1.5836186722502772e-06, + "loss": 0.2533, + "step": 38382 + }, + { + "epoch": 14.17, + "learning_rate": 1.583581721886932e-06, + "loss": 0.3734, + "step": 38383 + }, + { + "epoch": 14.17, + "learning_rate": 1.5835447715235867e-06, + "loss": 0.3459, + "step": 38384 + }, + { + "epoch": 14.17, + "learning_rate": 1.5835078211602413e-06, + "loss": 0.378, + "step": 38385 + }, + { + "epoch": 14.18, + "learning_rate": 1.5834708707968962e-06, + "loss": 0.2296, + "step": 38386 + }, + { + "epoch": 14.18, + "learning_rate": 1.5834339204335508e-06, + "loss": 0.3781, + "step": 38387 + }, + { + "epoch": 14.18, + "learning_rate": 1.5833969700702059e-06, + "loss": 0.3338, + "step": 38388 + }, + { + "epoch": 14.18, + "learning_rate": 1.5833600197068603e-06, + "loss": 0.239, + "step": 38389 + }, + { + "epoch": 14.18, + "learning_rate": 1.5833230693435154e-06, + "loss": 0.2453, + "step": 38390 + }, + { + "epoch": 14.18, + "learning_rate": 1.58328611898017e-06, + "loss": 0.2114, + "step": 38391 + }, + { + "epoch": 14.18, + "learning_rate": 1.5832491686168248e-06, + "loss": 0.2654, + "step": 38392 + }, + { + "epoch": 14.18, + "learning_rate": 1.5832122182534795e-06, + "loss": 0.2512, + "step": 38393 + }, + { + "epoch": 14.18, + "learning_rate": 1.5831752678901345e-06, + "loss": 0.2445, + "step": 38394 + }, + { + "epoch": 14.18, + "learning_rate": 1.583138317526789e-06, + "loss": 0.2412, + "step": 38395 + }, + { + "epoch": 14.18, + "learning_rate": 1.5831013671634436e-06, + "loss": 0.3154, + "step": 38396 + }, + { + "epoch": 14.18, + "learning_rate": 1.5830644168000987e-06, + "loss": 0.3474, + "step": 38397 + }, + { + "epoch": 14.18, + "learning_rate": 1.5830274664367533e-06, + "loss": 0.2004, + "step": 38398 + }, + { + "epoch": 14.18, + "learning_rate": 1.5829905160734081e-06, + "loss": 0.135, + "step": 38399 + }, + { + "epoch": 14.18, + "learning_rate": 1.5829535657100628e-06, + "loss": 0.2644, + "step": 38400 + }, + { + "epoch": 14.18, + "learning_rate": 1.5829166153467176e-06, + "loss": 0.2337, + "step": 38401 + }, + { + "epoch": 14.18, + "learning_rate": 1.5828796649833723e-06, + "loss": 0.2406, + "step": 38402 + }, + { + "epoch": 14.18, + "learning_rate": 1.5828427146200273e-06, + "loss": 0.2964, + "step": 38403 + }, + { + "epoch": 14.18, + "learning_rate": 1.582805764256682e-06, + "loss": 0.2735, + "step": 38404 + }, + { + "epoch": 14.18, + "learning_rate": 1.5827688138933366e-06, + "loss": 0.2578, + "step": 38405 + }, + { + "epoch": 14.18, + "learning_rate": 1.5827318635299914e-06, + "loss": 0.2636, + "step": 38406 + }, + { + "epoch": 14.18, + "learning_rate": 1.582694913166646e-06, + "loss": 0.3125, + "step": 38407 + }, + { + "epoch": 14.18, + "learning_rate": 1.582657962803301e-06, + "loss": 0.2948, + "step": 38408 + }, + { + "epoch": 14.18, + "learning_rate": 1.5826210124399556e-06, + "loss": 0.3209, + "step": 38409 + }, + { + "epoch": 14.18, + "learning_rate": 1.5825840620766106e-06, + "loss": 0.1765, + "step": 38410 + }, + { + "epoch": 14.18, + "learning_rate": 1.5825471117132653e-06, + "loss": 0.2877, + "step": 38411 + }, + { + "epoch": 14.18, + "learning_rate": 1.5825101613499201e-06, + "loss": 0.3594, + "step": 38412 + }, + { + "epoch": 14.19, + "learning_rate": 1.5824732109865747e-06, + "loss": 0.3294, + "step": 38413 + }, + { + "epoch": 14.19, + "learning_rate": 1.5824362606232296e-06, + "loss": 0.1901, + "step": 38414 + }, + { + "epoch": 14.19, + "learning_rate": 1.5823993102598842e-06, + "loss": 0.1978, + "step": 38415 + }, + { + "epoch": 14.19, + "learning_rate": 1.5823623598965389e-06, + "loss": 0.2553, + "step": 38416 + }, + { + "epoch": 14.19, + "learning_rate": 1.5823254095331937e-06, + "loss": 0.2117, + "step": 38417 + }, + { + "epoch": 14.19, + "learning_rate": 1.5822884591698484e-06, + "loss": 0.2857, + "step": 38418 + }, + { + "epoch": 14.19, + "learning_rate": 1.5822515088065034e-06, + "loss": 0.2011, + "step": 38419 + }, + { + "epoch": 14.19, + "learning_rate": 1.582214558443158e-06, + "loss": 0.3083, + "step": 38420 + }, + { + "epoch": 14.19, + "learning_rate": 1.582177608079813e-06, + "loss": 0.2947, + "step": 38421 + }, + { + "epoch": 14.19, + "learning_rate": 1.5821406577164675e-06, + "loss": 0.2095, + "step": 38422 + }, + { + "epoch": 14.19, + "learning_rate": 1.5821037073531224e-06, + "loss": 0.2467, + "step": 38423 + }, + { + "epoch": 14.19, + "learning_rate": 1.582066756989777e-06, + "loss": 0.2367, + "step": 38424 + }, + { + "epoch": 14.19, + "learning_rate": 1.5820298066264317e-06, + "loss": 0.3111, + "step": 38425 + }, + { + "epoch": 14.19, + "learning_rate": 1.5819928562630867e-06, + "loss": 0.2711, + "step": 38426 + }, + { + "epoch": 14.19, + "learning_rate": 1.5819559058997414e-06, + "loss": 0.2149, + "step": 38427 + }, + { + "epoch": 14.19, + "learning_rate": 1.5819189555363962e-06, + "loss": 0.1811, + "step": 38428 + }, + { + "epoch": 14.19, + "learning_rate": 1.5818820051730508e-06, + "loss": 0.218, + "step": 38429 + }, + { + "epoch": 14.19, + "learning_rate": 1.5818450548097057e-06, + "loss": 0.3081, + "step": 38430 + }, + { + "epoch": 14.19, + "learning_rate": 1.5818081044463603e-06, + "loss": 0.3181, + "step": 38431 + }, + { + "epoch": 14.19, + "learning_rate": 1.5817711540830154e-06, + "loss": 0.3038, + "step": 38432 + }, + { + "epoch": 14.19, + "learning_rate": 1.58173420371967e-06, + "loss": 0.2444, + "step": 38433 + }, + { + "epoch": 14.19, + "learning_rate": 1.5816972533563249e-06, + "loss": 0.2298, + "step": 38434 + }, + { + "epoch": 14.19, + "learning_rate": 1.5816603029929795e-06, + "loss": 0.2847, + "step": 38435 + }, + { + "epoch": 14.19, + "learning_rate": 1.5816233526296341e-06, + "loss": 0.241, + "step": 38436 + }, + { + "epoch": 14.19, + "learning_rate": 1.581586402266289e-06, + "loss": 0.2717, + "step": 38437 + }, + { + "epoch": 14.19, + "learning_rate": 1.5815494519029436e-06, + "loss": 0.2611, + "step": 38438 + }, + { + "epoch": 14.19, + "learning_rate": 1.5815125015395987e-06, + "loss": 0.2044, + "step": 38439 + }, + { + "epoch": 14.19, + "learning_rate": 1.5814755511762531e-06, + "loss": 0.1865, + "step": 38440 + }, + { + "epoch": 14.2, + "learning_rate": 1.5814386008129082e-06, + "loss": 0.2348, + "step": 38441 + }, + { + "epoch": 14.2, + "learning_rate": 1.5814016504495628e-06, + "loss": 0.3007, + "step": 38442 + }, + { + "epoch": 14.2, + "learning_rate": 1.5813647000862177e-06, + "loss": 0.322, + "step": 38443 + }, + { + "epoch": 14.2, + "learning_rate": 1.5813277497228723e-06, + "loss": 0.255, + "step": 38444 + }, + { + "epoch": 14.2, + "learning_rate": 1.581290799359527e-06, + "loss": 0.2498, + "step": 38445 + }, + { + "epoch": 14.2, + "learning_rate": 1.5812538489961818e-06, + "loss": 0.2767, + "step": 38446 + }, + { + "epoch": 14.2, + "learning_rate": 1.5812168986328364e-06, + "loss": 0.177, + "step": 38447 + }, + { + "epoch": 14.2, + "learning_rate": 1.5811799482694915e-06, + "loss": 0.2949, + "step": 38448 + }, + { + "epoch": 14.2, + "learning_rate": 1.5811429979061461e-06, + "loss": 0.2901, + "step": 38449 + }, + { + "epoch": 14.2, + "learning_rate": 1.581106047542801e-06, + "loss": 0.3358, + "step": 38450 + }, + { + "epoch": 14.2, + "learning_rate": 1.5810690971794556e-06, + "loss": 0.3283, + "step": 38451 + }, + { + "epoch": 14.2, + "learning_rate": 1.5810321468161104e-06, + "loss": 0.173, + "step": 38452 + }, + { + "epoch": 14.2, + "learning_rate": 1.580995196452765e-06, + "loss": 0.2715, + "step": 38453 + }, + { + "epoch": 14.2, + "learning_rate": 1.5809582460894201e-06, + "loss": 0.2967, + "step": 38454 + }, + { + "epoch": 14.2, + "learning_rate": 1.5809212957260748e-06, + "loss": 0.3181, + "step": 38455 + }, + { + "epoch": 14.2, + "learning_rate": 1.5808843453627292e-06, + "loss": 0.2411, + "step": 38456 + }, + { + "epoch": 14.2, + "learning_rate": 1.5808473949993843e-06, + "loss": 0.3948, + "step": 38457 + }, + { + "epoch": 14.2, + "learning_rate": 1.580810444636039e-06, + "loss": 0.2899, + "step": 38458 + }, + { + "epoch": 14.2, + "learning_rate": 1.5807734942726937e-06, + "loss": 0.3227, + "step": 38459 + }, + { + "epoch": 14.2, + "learning_rate": 1.5807365439093484e-06, + "loss": 0.3419, + "step": 38460 + }, + { + "epoch": 14.2, + "learning_rate": 1.5806995935460034e-06, + "loss": 0.3288, + "step": 38461 + }, + { + "epoch": 14.2, + "learning_rate": 1.5806626431826579e-06, + "loss": 0.246, + "step": 38462 + }, + { + "epoch": 14.2, + "learning_rate": 1.580625692819313e-06, + "loss": 0.2367, + "step": 38463 + }, + { + "epoch": 14.2, + "learning_rate": 1.5805887424559676e-06, + "loss": 0.2869, + "step": 38464 + }, + { + "epoch": 14.2, + "learning_rate": 1.5805517920926222e-06, + "loss": 0.3965, + "step": 38465 + }, + { + "epoch": 14.2, + "learning_rate": 1.580514841729277e-06, + "loss": 0.2846, + "step": 38466 + }, + { + "epoch": 14.2, + "learning_rate": 1.5804778913659317e-06, + "loss": 0.2267, + "step": 38467 + }, + { + "epoch": 14.21, + "learning_rate": 1.5804409410025865e-06, + "loss": 0.4284, + "step": 38468 + }, + { + "epoch": 14.21, + "learning_rate": 1.5804039906392412e-06, + "loss": 0.3263, + "step": 38469 + }, + { + "epoch": 14.21, + "learning_rate": 1.5803670402758962e-06, + "loss": 0.2565, + "step": 38470 + }, + { + "epoch": 14.21, + "learning_rate": 1.5803300899125509e-06, + "loss": 0.3586, + "step": 38471 + }, + { + "epoch": 14.21, + "learning_rate": 1.5802931395492057e-06, + "loss": 0.2934, + "step": 38472 + }, + { + "epoch": 14.21, + "learning_rate": 1.5802561891858603e-06, + "loss": 0.2676, + "step": 38473 + }, + { + "epoch": 14.21, + "learning_rate": 1.5802192388225152e-06, + "loss": 0.2087, + "step": 38474 + }, + { + "epoch": 14.21, + "learning_rate": 1.5801822884591698e-06, + "loss": 0.1828, + "step": 38475 + }, + { + "epoch": 14.21, + "learning_rate": 1.5801453380958245e-06, + "loss": 0.2149, + "step": 38476 + }, + { + "epoch": 14.21, + "learning_rate": 1.5801083877324795e-06, + "loss": 0.1665, + "step": 38477 + }, + { + "epoch": 14.21, + "learning_rate": 1.5800714373691342e-06, + "loss": 0.351, + "step": 38478 + }, + { + "epoch": 14.21, + "learning_rate": 1.580034487005789e-06, + "loss": 0.1403, + "step": 38479 + }, + { + "epoch": 14.21, + "learning_rate": 1.5799975366424436e-06, + "loss": 0.2083, + "step": 38480 + }, + { + "epoch": 14.21, + "learning_rate": 1.5799605862790985e-06, + "loss": 0.2767, + "step": 38481 + }, + { + "epoch": 14.21, + "learning_rate": 1.5799236359157531e-06, + "loss": 0.2803, + "step": 38482 + }, + { + "epoch": 14.21, + "learning_rate": 1.5798866855524082e-06, + "loss": 0.3054, + "step": 38483 + }, + { + "epoch": 14.21, + "learning_rate": 1.5798497351890628e-06, + "loss": 0.2776, + "step": 38484 + }, + { + "epoch": 14.21, + "learning_rate": 1.5798127848257173e-06, + "loss": 0.2347, + "step": 38485 + }, + { + "epoch": 14.21, + "learning_rate": 1.5797758344623723e-06, + "loss": 0.2849, + "step": 38486 + }, + { + "epoch": 14.21, + "learning_rate": 1.579738884099027e-06, + "loss": 0.3133, + "step": 38487 + }, + { + "epoch": 14.21, + "learning_rate": 1.5797019337356818e-06, + "loss": 0.1867, + "step": 38488 + }, + { + "epoch": 14.21, + "learning_rate": 1.5796649833723364e-06, + "loss": 0.3027, + "step": 38489 + }, + { + "epoch": 14.21, + "learning_rate": 1.5796280330089915e-06, + "loss": 0.2994, + "step": 38490 + }, + { + "epoch": 14.21, + "learning_rate": 1.579591082645646e-06, + "loss": 0.33, + "step": 38491 + }, + { + "epoch": 14.21, + "learning_rate": 1.579554132282301e-06, + "loss": 0.3463, + "step": 38492 + }, + { + "epoch": 14.21, + "learning_rate": 1.5795171819189556e-06, + "loss": 0.2983, + "step": 38493 + }, + { + "epoch": 14.21, + "learning_rate": 1.5794802315556105e-06, + "loss": 0.1892, + "step": 38494 + }, + { + "epoch": 14.22, + "learning_rate": 1.579443281192265e-06, + "loss": 0.2755, + "step": 38495 + }, + { + "epoch": 14.22, + "learning_rate": 1.5794063308289197e-06, + "loss": 0.1584, + "step": 38496 + }, + { + "epoch": 14.22, + "learning_rate": 1.5793693804655746e-06, + "loss": 0.5083, + "step": 38497 + }, + { + "epoch": 14.22, + "learning_rate": 1.5793324301022292e-06, + "loss": 0.2956, + "step": 38498 + }, + { + "epoch": 14.22, + "learning_rate": 1.5792954797388843e-06, + "loss": 0.2883, + "step": 38499 + }, + { + "epoch": 14.22, + "learning_rate": 1.579258529375539e-06, + "loss": 0.2149, + "step": 38500 + }, + { + "epoch": 14.22, + "learning_rate": 1.5792215790121938e-06, + "loss": 0.2725, + "step": 38501 + }, + { + "epoch": 14.22, + "learning_rate": 1.5791846286488484e-06, + "loss": 0.2924, + "step": 38502 + }, + { + "epoch": 14.22, + "learning_rate": 1.5791476782855033e-06, + "loss": 0.2405, + "step": 38503 + }, + { + "epoch": 14.22, + "learning_rate": 1.5791107279221579e-06, + "loss": 0.2558, + "step": 38504 + }, + { + "epoch": 14.22, + "learning_rate": 1.5790737775588125e-06, + "loss": 0.1884, + "step": 38505 + }, + { + "epoch": 14.22, + "learning_rate": 1.5790368271954676e-06, + "loss": 0.4224, + "step": 38506 + }, + { + "epoch": 14.22, + "learning_rate": 1.578999876832122e-06, + "loss": 0.3256, + "step": 38507 + }, + { + "epoch": 14.22, + "learning_rate": 1.578962926468777e-06, + "loss": 0.2041, + "step": 38508 + }, + { + "epoch": 14.22, + "learning_rate": 1.5789259761054317e-06, + "loss": 0.4485, + "step": 38509 + }, + { + "epoch": 14.22, + "learning_rate": 1.5788890257420866e-06, + "loss": 0.4231, + "step": 38510 + }, + { + "epoch": 14.22, + "learning_rate": 1.5788520753787412e-06, + "loss": 0.2539, + "step": 38511 + }, + { + "epoch": 14.22, + "learning_rate": 1.5788151250153962e-06, + "loss": 0.2097, + "step": 38512 + }, + { + "epoch": 14.22, + "learning_rate": 1.5787781746520507e-06, + "loss": 0.2426, + "step": 38513 + }, + { + "epoch": 14.22, + "learning_rate": 1.5787412242887057e-06, + "loss": 0.2066, + "step": 38514 + }, + { + "epoch": 14.22, + "learning_rate": 1.5787042739253604e-06, + "loss": 0.2309, + "step": 38515 + }, + { + "epoch": 14.22, + "learning_rate": 1.578667323562015e-06, + "loss": 0.2108, + "step": 38516 + }, + { + "epoch": 14.22, + "learning_rate": 1.5786303731986699e-06, + "loss": 0.1748, + "step": 38517 + }, + { + "epoch": 14.22, + "learning_rate": 1.5785934228353245e-06, + "loss": 0.322, + "step": 38518 + }, + { + "epoch": 14.22, + "learning_rate": 1.5785564724719793e-06, + "loss": 0.2306, + "step": 38519 + }, + { + "epoch": 14.22, + "learning_rate": 1.578519522108634e-06, + "loss": 0.1924, + "step": 38520 + }, + { + "epoch": 14.22, + "learning_rate": 1.578482571745289e-06, + "loss": 0.3883, + "step": 38521 + }, + { + "epoch": 14.23, + "learning_rate": 1.5784456213819437e-06, + "loss": 0.2909, + "step": 38522 + }, + { + "epoch": 14.23, + "learning_rate": 1.5784086710185985e-06, + "loss": 0.265, + "step": 38523 + }, + { + "epoch": 14.23, + "learning_rate": 1.5783717206552532e-06, + "loss": 0.369, + "step": 38524 + }, + { + "epoch": 14.23, + "learning_rate": 1.5783347702919078e-06, + "loss": 0.2803, + "step": 38525 + }, + { + "epoch": 14.23, + "learning_rate": 1.5782978199285626e-06, + "loss": 0.1761, + "step": 38526 + }, + { + "epoch": 14.23, + "learning_rate": 1.5782608695652173e-06, + "loss": 0.3132, + "step": 38527 + }, + { + "epoch": 14.23, + "learning_rate": 1.5782239192018723e-06, + "loss": 0.2982, + "step": 38528 + }, + { + "epoch": 14.23, + "learning_rate": 1.5781869688385268e-06, + "loss": 0.3742, + "step": 38529 + }, + { + "epoch": 14.23, + "learning_rate": 1.5781500184751818e-06, + "loss": 0.3416, + "step": 38530 + }, + { + "epoch": 14.23, + "learning_rate": 1.5781130681118365e-06, + "loss": 0.1531, + "step": 38531 + }, + { + "epoch": 14.23, + "learning_rate": 1.5780761177484913e-06, + "loss": 0.2349, + "step": 38532 + }, + { + "epoch": 14.23, + "learning_rate": 1.578039167385146e-06, + "loss": 0.3313, + "step": 38533 + }, + { + "epoch": 14.23, + "learning_rate": 1.578002217021801e-06, + "loss": 0.3311, + "step": 38534 + }, + { + "epoch": 14.23, + "learning_rate": 1.5779652666584554e-06, + "loss": 0.2053, + "step": 38535 + }, + { + "epoch": 14.23, + "learning_rate": 1.57792831629511e-06, + "loss": 0.2859, + "step": 38536 + }, + { + "epoch": 14.23, + "learning_rate": 1.5778913659317651e-06, + "loss": 0.4861, + "step": 38537 + }, + { + "epoch": 14.23, + "learning_rate": 1.5778544155684198e-06, + "loss": 0.1676, + "step": 38538 + }, + { + "epoch": 14.23, + "learning_rate": 1.5778174652050746e-06, + "loss": 0.288, + "step": 38539 + }, + { + "epoch": 14.23, + "learning_rate": 1.5777805148417292e-06, + "loss": 0.1721, + "step": 38540 + }, + { + "epoch": 14.23, + "learning_rate": 1.577743564478384e-06, + "loss": 0.2291, + "step": 38541 + }, + { + "epoch": 14.23, + "learning_rate": 1.5777066141150387e-06, + "loss": 0.2344, + "step": 38542 + }, + { + "epoch": 14.23, + "learning_rate": 1.5776696637516938e-06, + "loss": 0.2433, + "step": 38543 + }, + { + "epoch": 14.23, + "learning_rate": 1.5776327133883484e-06, + "loss": 0.2109, + "step": 38544 + }, + { + "epoch": 14.23, + "learning_rate": 1.577595763025003e-06, + "loss": 0.2951, + "step": 38545 + }, + { + "epoch": 14.23, + "learning_rate": 1.577558812661658e-06, + "loss": 0.3705, + "step": 38546 + }, + { + "epoch": 14.23, + "learning_rate": 1.5775218622983125e-06, + "loss": 0.2699, + "step": 38547 + }, + { + "epoch": 14.23, + "learning_rate": 1.5774849119349674e-06, + "loss": 0.2696, + "step": 38548 + }, + { + "epoch": 14.24, + "learning_rate": 1.577447961571622e-06, + "loss": 0.2999, + "step": 38549 + }, + { + "epoch": 14.24, + "learning_rate": 1.577411011208277e-06, + "loss": 0.2433, + "step": 38550 + }, + { + "epoch": 14.24, + "learning_rate": 1.5773740608449317e-06, + "loss": 0.2487, + "step": 38551 + }, + { + "epoch": 14.24, + "learning_rate": 1.5773371104815866e-06, + "loss": 0.2127, + "step": 38552 + }, + { + "epoch": 14.24, + "learning_rate": 1.5773001601182412e-06, + "loss": 0.2993, + "step": 38553 + }, + { + "epoch": 14.24, + "learning_rate": 1.577263209754896e-06, + "loss": 0.211, + "step": 38554 + }, + { + "epoch": 14.24, + "learning_rate": 1.5772262593915507e-06, + "loss": 0.1989, + "step": 38555 + }, + { + "epoch": 14.24, + "learning_rate": 1.5771893090282053e-06, + "loss": 0.3838, + "step": 38556 + }, + { + "epoch": 14.24, + "learning_rate": 1.5771523586648604e-06, + "loss": 0.3325, + "step": 38557 + }, + { + "epoch": 14.24, + "learning_rate": 1.5771154083015148e-06, + "loss": 0.1701, + "step": 38558 + }, + { + "epoch": 14.24, + "learning_rate": 1.5770784579381699e-06, + "loss": 0.286, + "step": 38559 + }, + { + "epoch": 14.24, + "learning_rate": 1.5770415075748245e-06, + "loss": 0.2304, + "step": 38560 + }, + { + "epoch": 14.24, + "learning_rate": 1.5770045572114794e-06, + "loss": 0.1592, + "step": 38561 + }, + { + "epoch": 14.24, + "learning_rate": 1.576967606848134e-06, + "loss": 0.23, + "step": 38562 + }, + { + "epoch": 14.24, + "learning_rate": 1.576930656484789e-06, + "loss": 0.2168, + "step": 38563 + }, + { + "epoch": 14.24, + "learning_rate": 1.5768937061214435e-06, + "loss": 0.2765, + "step": 38564 + }, + { + "epoch": 14.24, + "learning_rate": 1.5768567557580981e-06, + "loss": 0.3417, + "step": 38565 + }, + { + "epoch": 14.24, + "learning_rate": 1.5768198053947532e-06, + "loss": 0.2731, + "step": 38566 + }, + { + "epoch": 14.24, + "learning_rate": 1.5767828550314078e-06, + "loss": 0.2203, + "step": 38567 + }, + { + "epoch": 14.24, + "learning_rate": 1.5767459046680627e-06, + "loss": 0.3402, + "step": 38568 + }, + { + "epoch": 14.24, + "learning_rate": 1.5767089543047173e-06, + "loss": 0.2418, + "step": 38569 + }, + { + "epoch": 14.24, + "learning_rate": 1.5766720039413722e-06, + "loss": 0.3979, + "step": 38570 + }, + { + "epoch": 14.24, + "learning_rate": 1.5766350535780268e-06, + "loss": 0.337, + "step": 38571 + }, + { + "epoch": 14.24, + "learning_rate": 1.5765981032146818e-06, + "loss": 0.2244, + "step": 38572 + }, + { + "epoch": 14.24, + "learning_rate": 1.5765611528513365e-06, + "loss": 0.4185, + "step": 38573 + }, + { + "epoch": 14.24, + "learning_rate": 1.576524202487991e-06, + "loss": 0.1959, + "step": 38574 + }, + { + "epoch": 14.24, + "learning_rate": 1.576487252124646e-06, + "loss": 0.3255, + "step": 38575 + }, + { + "epoch": 14.25, + "learning_rate": 1.5764503017613006e-06, + "loss": 0.2916, + "step": 38576 + }, + { + "epoch": 14.25, + "learning_rate": 1.5764133513979555e-06, + "loss": 0.205, + "step": 38577 + }, + { + "epoch": 14.25, + "learning_rate": 1.57637640103461e-06, + "loss": 0.3022, + "step": 38578 + }, + { + "epoch": 14.25, + "learning_rate": 1.5763394506712651e-06, + "loss": 0.1543, + "step": 38579 + }, + { + "epoch": 14.25, + "learning_rate": 1.5763025003079196e-06, + "loss": 0.2735, + "step": 38580 + }, + { + "epoch": 14.25, + "learning_rate": 1.5762655499445746e-06, + "loss": 0.3391, + "step": 38581 + }, + { + "epoch": 14.25, + "learning_rate": 1.5762285995812293e-06, + "loss": 0.3685, + "step": 38582 + }, + { + "epoch": 14.25, + "learning_rate": 1.5761916492178841e-06, + "loss": 0.327, + "step": 38583 + }, + { + "epoch": 14.25, + "learning_rate": 1.5761546988545388e-06, + "loss": 0.2252, + "step": 38584 + }, + { + "epoch": 14.25, + "learning_rate": 1.5761177484911934e-06, + "loss": 0.2357, + "step": 38585 + }, + { + "epoch": 14.25, + "learning_rate": 1.5760807981278482e-06, + "loss": 0.3269, + "step": 38586 + }, + { + "epoch": 14.25, + "learning_rate": 1.5760438477645029e-06, + "loss": 0.3552, + "step": 38587 + }, + { + "epoch": 14.25, + "learning_rate": 1.576006897401158e-06, + "loss": 0.34, + "step": 38588 + }, + { + "epoch": 14.25, + "learning_rate": 1.5759699470378126e-06, + "loss": 0.3229, + "step": 38589 + }, + { + "epoch": 14.25, + "learning_rate": 1.5759329966744674e-06, + "loss": 0.2177, + "step": 38590 + }, + { + "epoch": 14.25, + "learning_rate": 1.575896046311122e-06, + "loss": 0.3348, + "step": 38591 + }, + { + "epoch": 14.25, + "learning_rate": 1.575859095947777e-06, + "loss": 0.2936, + "step": 38592 + }, + { + "epoch": 14.25, + "learning_rate": 1.5758221455844315e-06, + "loss": 0.363, + "step": 38593 + }, + { + "epoch": 14.25, + "learning_rate": 1.5757851952210862e-06, + "loss": 0.2937, + "step": 38594 + }, + { + "epoch": 14.25, + "learning_rate": 1.5757482448577412e-06, + "loss": 0.2524, + "step": 38595 + }, + { + "epoch": 14.25, + "learning_rate": 1.5757112944943959e-06, + "loss": 0.2581, + "step": 38596 + }, + { + "epoch": 14.25, + "learning_rate": 1.5756743441310507e-06, + "loss": 0.2022, + "step": 38597 + }, + { + "epoch": 14.25, + "learning_rate": 1.5756373937677054e-06, + "loss": 0.1931, + "step": 38598 + }, + { + "epoch": 14.25, + "learning_rate": 1.5756004434043602e-06, + "loss": 0.234, + "step": 38599 + }, + { + "epoch": 14.25, + "learning_rate": 1.5755634930410148e-06, + "loss": 0.221, + "step": 38600 + }, + { + "epoch": 14.25, + "learning_rate": 1.57552654267767e-06, + "loss": 0.2538, + "step": 38601 + }, + { + "epoch": 14.25, + "learning_rate": 1.5754895923143243e-06, + "loss": 0.3395, + "step": 38602 + }, + { + "epoch": 14.26, + "learning_rate": 1.5754526419509794e-06, + "loss": 0.2557, + "step": 38603 + }, + { + "epoch": 14.26, + "learning_rate": 1.575415691587634e-06, + "loss": 0.1695, + "step": 38604 + }, + { + "epoch": 14.26, + "learning_rate": 1.5753787412242887e-06, + "loss": 0.3616, + "step": 38605 + }, + { + "epoch": 14.26, + "learning_rate": 1.5753417908609435e-06, + "loss": 0.2278, + "step": 38606 + }, + { + "epoch": 14.26, + "learning_rate": 1.5753048404975981e-06, + "loss": 0.3245, + "step": 38607 + }, + { + "epoch": 14.26, + "learning_rate": 1.575267890134253e-06, + "loss": 0.3438, + "step": 38608 + }, + { + "epoch": 14.26, + "learning_rate": 1.5752309397709076e-06, + "loss": 0.2288, + "step": 38609 + }, + { + "epoch": 14.26, + "learning_rate": 1.5751939894075627e-06, + "loss": 0.2469, + "step": 38610 + }, + { + "epoch": 14.26, + "learning_rate": 1.5751570390442173e-06, + "loss": 0.2565, + "step": 38611 + }, + { + "epoch": 14.26, + "learning_rate": 1.5751200886808722e-06, + "loss": 0.2621, + "step": 38612 + }, + { + "epoch": 14.26, + "learning_rate": 1.5750831383175268e-06, + "loss": 0.1854, + "step": 38613 + }, + { + "epoch": 14.26, + "learning_rate": 1.5750461879541814e-06, + "loss": 0.2333, + "step": 38614 + }, + { + "epoch": 14.26, + "learning_rate": 1.5750092375908363e-06, + "loss": 0.3002, + "step": 38615 + }, + { + "epoch": 14.26, + "learning_rate": 1.574972287227491e-06, + "loss": 0.2706, + "step": 38616 + }, + { + "epoch": 14.26, + "learning_rate": 1.574935336864146e-06, + "loss": 0.2688, + "step": 38617 + }, + { + "epoch": 14.26, + "learning_rate": 1.5748983865008006e-06, + "loss": 0.2442, + "step": 38618 + }, + { + "epoch": 14.26, + "learning_rate": 1.5748614361374555e-06, + "loss": 0.2342, + "step": 38619 + }, + { + "epoch": 14.26, + "learning_rate": 1.5748244857741101e-06, + "loss": 0.2097, + "step": 38620 + }, + { + "epoch": 14.26, + "learning_rate": 1.574787535410765e-06, + "loss": 0.342, + "step": 38621 + }, + { + "epoch": 14.26, + "learning_rate": 1.5747505850474196e-06, + "loss": 0.2639, + "step": 38622 + }, + { + "epoch": 14.26, + "learning_rate": 1.5747136346840747e-06, + "loss": 0.3435, + "step": 38623 + }, + { + "epoch": 14.26, + "learning_rate": 1.5746766843207293e-06, + "loss": 0.2149, + "step": 38624 + }, + { + "epoch": 14.26, + "learning_rate": 1.5746397339573837e-06, + "loss": 0.2576, + "step": 38625 + }, + { + "epoch": 14.26, + "learning_rate": 1.5746027835940388e-06, + "loss": 0.1487, + "step": 38626 + }, + { + "epoch": 14.26, + "learning_rate": 1.5745658332306934e-06, + "loss": 0.3043, + "step": 38627 + }, + { + "epoch": 14.26, + "learning_rate": 1.5745288828673483e-06, + "loss": 0.1789, + "step": 38628 + }, + { + "epoch": 14.26, + "learning_rate": 1.574491932504003e-06, + "loss": 0.313, + "step": 38629 + }, + { + "epoch": 14.27, + "learning_rate": 1.574454982140658e-06, + "loss": 0.2934, + "step": 38630 + }, + { + "epoch": 14.27, + "learning_rate": 1.5744180317773124e-06, + "loss": 0.1984, + "step": 38631 + }, + { + "epoch": 14.27, + "learning_rate": 1.5743810814139674e-06, + "loss": 0.2625, + "step": 38632 + }, + { + "epoch": 14.27, + "learning_rate": 1.574344131050622e-06, + "loss": 0.3177, + "step": 38633 + }, + { + "epoch": 14.27, + "learning_rate": 1.5743071806872767e-06, + "loss": 0.1782, + "step": 38634 + }, + { + "epoch": 14.27, + "learning_rate": 1.5742702303239316e-06, + "loss": 0.22, + "step": 38635 + }, + { + "epoch": 14.27, + "learning_rate": 1.5742332799605862e-06, + "loss": 0.3011, + "step": 38636 + }, + { + "epoch": 14.27, + "learning_rate": 1.574196329597241e-06, + "loss": 0.3459, + "step": 38637 + }, + { + "epoch": 14.27, + "learning_rate": 1.5741593792338957e-06, + "loss": 0.2041, + "step": 38638 + }, + { + "epoch": 14.27, + "learning_rate": 1.5741224288705507e-06, + "loss": 0.2501, + "step": 38639 + }, + { + "epoch": 14.27, + "learning_rate": 1.5740854785072054e-06, + "loss": 0.2599, + "step": 38640 + }, + { + "epoch": 14.27, + "learning_rate": 1.5740485281438602e-06, + "loss": 0.2784, + "step": 38641 + }, + { + "epoch": 14.27, + "learning_rate": 1.5740115777805149e-06, + "loss": 0.2613, + "step": 38642 + }, + { + "epoch": 14.27, + "learning_rate": 1.5739746274171697e-06, + "loss": 0.1931, + "step": 38643 + }, + { + "epoch": 14.27, + "learning_rate": 1.5739376770538244e-06, + "loss": 0.1971, + "step": 38644 + }, + { + "epoch": 14.27, + "learning_rate": 1.573900726690479e-06, + "loss": 0.3324, + "step": 38645 + }, + { + "epoch": 14.27, + "learning_rate": 1.573863776327134e-06, + "loss": 0.2361, + "step": 38646 + }, + { + "epoch": 14.27, + "learning_rate": 1.5738268259637885e-06, + "loss": 0.3487, + "step": 38647 + }, + { + "epoch": 14.27, + "learning_rate": 1.5737898756004435e-06, + "loss": 0.2957, + "step": 38648 + }, + { + "epoch": 14.27, + "learning_rate": 1.5737529252370982e-06, + "loss": 0.1559, + "step": 38649 + }, + { + "epoch": 14.27, + "learning_rate": 1.573715974873753e-06, + "loss": 0.3312, + "step": 38650 + }, + { + "epoch": 14.27, + "learning_rate": 1.5736790245104077e-06, + "loss": 0.4766, + "step": 38651 + }, + { + "epoch": 14.27, + "learning_rate": 1.5736420741470627e-06, + "loss": 0.1721, + "step": 38652 + }, + { + "epoch": 14.27, + "learning_rate": 1.5736051237837171e-06, + "loss": 0.2727, + "step": 38653 + }, + { + "epoch": 14.27, + "learning_rate": 1.5735681734203718e-06, + "loss": 0.2673, + "step": 38654 + }, + { + "epoch": 14.27, + "learning_rate": 1.5735312230570268e-06, + "loss": 0.3646, + "step": 38655 + }, + { + "epoch": 14.27, + "learning_rate": 1.5734942726936815e-06, + "loss": 0.3588, + "step": 38656 + }, + { + "epoch": 14.28, + "learning_rate": 1.5734573223303363e-06, + "loss": 0.3535, + "step": 38657 + }, + { + "epoch": 14.28, + "learning_rate": 1.573420371966991e-06, + "loss": 0.2002, + "step": 38658 + }, + { + "epoch": 14.28, + "learning_rate": 1.5733834216036458e-06, + "loss": 0.5085, + "step": 38659 + }, + { + "epoch": 14.28, + "learning_rate": 1.5733464712403004e-06, + "loss": 0.3287, + "step": 38660 + }, + { + "epoch": 14.28, + "learning_rate": 1.5733095208769555e-06, + "loss": 0.2189, + "step": 38661 + }, + { + "epoch": 14.28, + "learning_rate": 1.5732725705136101e-06, + "loss": 0.1678, + "step": 38662 + }, + { + "epoch": 14.28, + "learning_rate": 1.573235620150265e-06, + "loss": 0.2069, + "step": 38663 + }, + { + "epoch": 14.28, + "learning_rate": 1.5731986697869196e-06, + "loss": 0.3353, + "step": 38664 + }, + { + "epoch": 14.28, + "learning_rate": 1.5731617194235743e-06, + "loss": 0.3133, + "step": 38665 + }, + { + "epoch": 14.28, + "learning_rate": 1.5731247690602291e-06, + "loss": 0.2858, + "step": 38666 + }, + { + "epoch": 14.28, + "learning_rate": 1.5730878186968837e-06, + "loss": 0.2135, + "step": 38667 + }, + { + "epoch": 14.28, + "learning_rate": 1.5730508683335388e-06, + "loss": 0.3007, + "step": 38668 + }, + { + "epoch": 14.28, + "learning_rate": 1.5730139179701934e-06, + "loss": 0.2483, + "step": 38669 + }, + { + "epoch": 14.28, + "learning_rate": 1.5729769676068483e-06, + "loss": 0.2156, + "step": 38670 + }, + { + "epoch": 14.28, + "learning_rate": 1.572940017243503e-06, + "loss": 0.4055, + "step": 38671 + }, + { + "epoch": 14.28, + "learning_rate": 1.5729030668801578e-06, + "loss": 0.2285, + "step": 38672 + }, + { + "epoch": 14.28, + "learning_rate": 1.5728661165168124e-06, + "loss": 0.238, + "step": 38673 + }, + { + "epoch": 14.28, + "learning_rate": 1.572829166153467e-06, + "loss": 0.2552, + "step": 38674 + }, + { + "epoch": 14.28, + "learning_rate": 1.572792215790122e-06, + "loss": 0.1729, + "step": 38675 + }, + { + "epoch": 14.28, + "learning_rate": 1.5727552654267765e-06, + "loss": 0.2167, + "step": 38676 + }, + { + "epoch": 14.28, + "learning_rate": 1.5727183150634316e-06, + "loss": 0.2244, + "step": 38677 + }, + { + "epoch": 14.28, + "learning_rate": 1.5726813647000862e-06, + "loss": 0.3072, + "step": 38678 + }, + { + "epoch": 14.28, + "learning_rate": 1.572644414336741e-06, + "loss": 0.2826, + "step": 38679 + }, + { + "epoch": 14.28, + "learning_rate": 1.5726074639733957e-06, + "loss": 0.2386, + "step": 38680 + }, + { + "epoch": 14.28, + "learning_rate": 1.5725705136100506e-06, + "loss": 0.268, + "step": 38681 + }, + { + "epoch": 14.28, + "learning_rate": 1.5725335632467052e-06, + "loss": 0.3577, + "step": 38682 + }, + { + "epoch": 14.28, + "learning_rate": 1.5724966128833603e-06, + "loss": 0.2795, + "step": 38683 + }, + { + "epoch": 14.29, + "learning_rate": 1.5724596625200149e-06, + "loss": 0.2516, + "step": 38684 + }, + { + "epoch": 14.29, + "learning_rate": 1.5724227121566695e-06, + "loss": 0.3258, + "step": 38685 + }, + { + "epoch": 14.29, + "learning_rate": 1.5723857617933244e-06, + "loss": 0.2553, + "step": 38686 + }, + { + "epoch": 14.29, + "learning_rate": 1.572348811429979e-06, + "loss": 0.2809, + "step": 38687 + }, + { + "epoch": 14.29, + "learning_rate": 1.5723118610666339e-06, + "loss": 0.3258, + "step": 38688 + }, + { + "epoch": 14.29, + "learning_rate": 1.5722749107032885e-06, + "loss": 0.2651, + "step": 38689 + }, + { + "epoch": 14.29, + "learning_rate": 1.5722379603399436e-06, + "loss": 0.2992, + "step": 38690 + }, + { + "epoch": 14.29, + "learning_rate": 1.5722010099765982e-06, + "loss": 0.3163, + "step": 38691 + }, + { + "epoch": 14.29, + "learning_rate": 1.572164059613253e-06, + "loss": 0.3045, + "step": 38692 + }, + { + "epoch": 14.29, + "learning_rate": 1.5721271092499077e-06, + "loss": 0.2126, + "step": 38693 + }, + { + "epoch": 14.29, + "learning_rate": 1.5720901588865623e-06, + "loss": 0.2714, + "step": 38694 + }, + { + "epoch": 14.29, + "learning_rate": 1.5720532085232172e-06, + "loss": 0.4319, + "step": 38695 + }, + { + "epoch": 14.29, + "learning_rate": 1.5720162581598718e-06, + "loss": 0.3394, + "step": 38696 + }, + { + "epoch": 14.29, + "learning_rate": 1.5719793077965269e-06, + "loss": 0.2261, + "step": 38697 + }, + { + "epoch": 14.29, + "learning_rate": 1.5719423574331813e-06, + "loss": 0.357, + "step": 38698 + }, + { + "epoch": 14.29, + "learning_rate": 1.5719054070698363e-06, + "loss": 0.298, + "step": 38699 + }, + { + "epoch": 14.29, + "learning_rate": 1.571868456706491e-06, + "loss": 0.2137, + "step": 38700 + }, + { + "epoch": 14.29, + "learning_rate": 1.5718315063431458e-06, + "loss": 0.263, + "step": 38701 + }, + { + "epoch": 14.29, + "learning_rate": 1.5717945559798005e-06, + "loss": 0.195, + "step": 38702 + }, + { + "epoch": 14.29, + "learning_rate": 1.5717576056164555e-06, + "loss": 0.1702, + "step": 38703 + }, + { + "epoch": 14.29, + "learning_rate": 1.57172065525311e-06, + "loss": 0.315, + "step": 38704 + }, + { + "epoch": 14.29, + "learning_rate": 1.5716837048897646e-06, + "loss": 0.238, + "step": 38705 + }, + { + "epoch": 14.29, + "learning_rate": 1.5716467545264196e-06, + "loss": 0.2829, + "step": 38706 + }, + { + "epoch": 14.29, + "learning_rate": 1.5716098041630743e-06, + "loss": 0.1834, + "step": 38707 + }, + { + "epoch": 14.29, + "learning_rate": 1.5715728537997291e-06, + "loss": 0.2124, + "step": 38708 + }, + { + "epoch": 14.29, + "learning_rate": 1.5715359034363838e-06, + "loss": 0.2755, + "step": 38709 + }, + { + "epoch": 14.29, + "learning_rate": 1.5714989530730386e-06, + "loss": 0.1533, + "step": 38710 + }, + { + "epoch": 14.3, + "learning_rate": 1.5714620027096933e-06, + "loss": 0.2493, + "step": 38711 + }, + { + "epoch": 14.3, + "learning_rate": 1.5714250523463483e-06, + "loss": 0.2484, + "step": 38712 + }, + { + "epoch": 14.3, + "learning_rate": 1.571388101983003e-06, + "loss": 0.2728, + "step": 38713 + }, + { + "epoch": 14.3, + "learning_rate": 1.5713511516196574e-06, + "loss": 0.2708, + "step": 38714 + }, + { + "epoch": 14.3, + "learning_rate": 1.5713142012563124e-06, + "loss": 0.4589, + "step": 38715 + }, + { + "epoch": 14.3, + "learning_rate": 1.571277250892967e-06, + "loss": 0.2313, + "step": 38716 + }, + { + "epoch": 14.3, + "learning_rate": 1.571240300529622e-06, + "loss": 0.2229, + "step": 38717 + }, + { + "epoch": 14.3, + "learning_rate": 1.5712033501662766e-06, + "loss": 0.2305, + "step": 38718 + }, + { + "epoch": 14.3, + "learning_rate": 1.5711663998029316e-06, + "loss": 0.322, + "step": 38719 + }, + { + "epoch": 14.3, + "learning_rate": 1.571129449439586e-06, + "loss": 0.304, + "step": 38720 + }, + { + "epoch": 14.3, + "learning_rate": 1.571092499076241e-06, + "loss": 0.2266, + "step": 38721 + }, + { + "epoch": 14.3, + "learning_rate": 1.5710555487128957e-06, + "loss": 0.1825, + "step": 38722 + }, + { + "epoch": 14.3, + "learning_rate": 1.5710185983495506e-06, + "loss": 0.2813, + "step": 38723 + }, + { + "epoch": 14.3, + "learning_rate": 1.5709816479862052e-06, + "loss": 0.3066, + "step": 38724 + }, + { + "epoch": 14.3, + "learning_rate": 1.5709446976228599e-06, + "loss": 0.3028, + "step": 38725 + }, + { + "epoch": 14.3, + "learning_rate": 1.5709077472595147e-06, + "loss": 0.2848, + "step": 38726 + }, + { + "epoch": 14.3, + "learning_rate": 1.5708707968961693e-06, + "loss": 0.2657, + "step": 38727 + }, + { + "epoch": 14.3, + "learning_rate": 1.5708338465328244e-06, + "loss": 0.2387, + "step": 38728 + }, + { + "epoch": 14.3, + "learning_rate": 1.570796896169479e-06, + "loss": 0.3091, + "step": 38729 + }, + { + "epoch": 14.3, + "learning_rate": 1.5707599458061339e-06, + "loss": 0.1759, + "step": 38730 + }, + { + "epoch": 14.3, + "learning_rate": 1.5707229954427885e-06, + "loss": 0.4085, + "step": 38731 + }, + { + "epoch": 14.3, + "learning_rate": 1.5706860450794434e-06, + "loss": 0.3207, + "step": 38732 + }, + { + "epoch": 14.3, + "learning_rate": 1.570649094716098e-06, + "loss": 0.2985, + "step": 38733 + }, + { + "epoch": 14.3, + "learning_rate": 1.5706121443527526e-06, + "loss": 0.3039, + "step": 38734 + }, + { + "epoch": 14.3, + "learning_rate": 1.5705751939894077e-06, + "loss": 0.2258, + "step": 38735 + }, + { + "epoch": 14.3, + "learning_rate": 1.5705382436260623e-06, + "loss": 0.5056, + "step": 38736 + }, + { + "epoch": 14.3, + "learning_rate": 1.5705012932627172e-06, + "loss": 0.2093, + "step": 38737 + }, + { + "epoch": 14.31, + "learning_rate": 1.5704643428993718e-06, + "loss": 0.2117, + "step": 38738 + }, + { + "epoch": 14.31, + "learning_rate": 1.5704273925360267e-06, + "loss": 0.3479, + "step": 38739 + }, + { + "epoch": 14.31, + "learning_rate": 1.5703904421726813e-06, + "loss": 0.2325, + "step": 38740 + }, + { + "epoch": 14.31, + "learning_rate": 1.5703534918093364e-06, + "loss": 0.2598, + "step": 38741 + }, + { + "epoch": 14.31, + "learning_rate": 1.570316541445991e-06, + "loss": 0.2416, + "step": 38742 + }, + { + "epoch": 14.31, + "learning_rate": 1.5702795910826459e-06, + "loss": 0.273, + "step": 38743 + }, + { + "epoch": 14.31, + "learning_rate": 1.5702426407193005e-06, + "loss": 0.1754, + "step": 38744 + }, + { + "epoch": 14.31, + "learning_rate": 1.5702056903559551e-06, + "loss": 0.1943, + "step": 38745 + }, + { + "epoch": 14.31, + "learning_rate": 1.57016873999261e-06, + "loss": 0.2463, + "step": 38746 + }, + { + "epoch": 14.31, + "learning_rate": 1.5701317896292646e-06, + "loss": 0.3651, + "step": 38747 + }, + { + "epoch": 14.31, + "learning_rate": 1.5700948392659197e-06, + "loss": 0.254, + "step": 38748 + }, + { + "epoch": 14.31, + "learning_rate": 1.570057888902574e-06, + "loss": 0.2591, + "step": 38749 + }, + { + "epoch": 14.31, + "learning_rate": 1.5700209385392292e-06, + "loss": 0.2702, + "step": 38750 + }, + { + "epoch": 14.31, + "learning_rate": 1.5699839881758838e-06, + "loss": 0.2388, + "step": 38751 + }, + { + "epoch": 14.31, + "learning_rate": 1.5699470378125386e-06, + "loss": 0.177, + "step": 38752 + }, + { + "epoch": 14.31, + "learning_rate": 1.5699100874491933e-06, + "loss": 0.3821, + "step": 38753 + }, + { + "epoch": 14.31, + "learning_rate": 1.569873137085848e-06, + "loss": 0.2962, + "step": 38754 + }, + { + "epoch": 14.31, + "learning_rate": 1.5698361867225028e-06, + "loss": 0.2147, + "step": 38755 + }, + { + "epoch": 14.31, + "learning_rate": 1.5697992363591574e-06, + "loss": 0.274, + "step": 38756 + }, + { + "epoch": 14.31, + "learning_rate": 1.5697622859958125e-06, + "loss": 0.2549, + "step": 38757 + }, + { + "epoch": 14.31, + "learning_rate": 1.569725335632467e-06, + "loss": 0.2301, + "step": 38758 + }, + { + "epoch": 14.31, + "learning_rate": 1.569688385269122e-06, + "loss": 0.1397, + "step": 38759 + }, + { + "epoch": 14.31, + "learning_rate": 1.5696514349057766e-06, + "loss": 0.316, + "step": 38760 + }, + { + "epoch": 14.31, + "learning_rate": 1.5696144845424314e-06, + "loss": 0.3489, + "step": 38761 + }, + { + "epoch": 14.31, + "learning_rate": 1.569577534179086e-06, + "loss": 0.3337, + "step": 38762 + }, + { + "epoch": 14.31, + "learning_rate": 1.5695405838157411e-06, + "loss": 0.2271, + "step": 38763 + }, + { + "epoch": 14.31, + "learning_rate": 1.5695036334523958e-06, + "loss": 0.2451, + "step": 38764 + }, + { + "epoch": 14.31, + "learning_rate": 1.5694666830890502e-06, + "loss": 0.1599, + "step": 38765 + }, + { + "epoch": 14.32, + "learning_rate": 1.5694297327257052e-06, + "loss": 0.2892, + "step": 38766 + }, + { + "epoch": 14.32, + "learning_rate": 1.5693927823623599e-06, + "loss": 0.2742, + "step": 38767 + }, + { + "epoch": 14.32, + "learning_rate": 1.5693558319990147e-06, + "loss": 0.3252, + "step": 38768 + }, + { + "epoch": 14.32, + "learning_rate": 1.5693188816356694e-06, + "loss": 0.1826, + "step": 38769 + }, + { + "epoch": 14.32, + "learning_rate": 1.5692819312723244e-06, + "loss": 0.3836, + "step": 38770 + }, + { + "epoch": 14.32, + "learning_rate": 1.5692449809089788e-06, + "loss": 0.3314, + "step": 38771 + }, + { + "epoch": 14.32, + "learning_rate": 1.569208030545634e-06, + "loss": 0.3682, + "step": 38772 + }, + { + "epoch": 14.32, + "learning_rate": 1.5691710801822885e-06, + "loss": 0.3008, + "step": 38773 + }, + { + "epoch": 14.32, + "learning_rate": 1.5691341298189432e-06, + "loss": 0.2975, + "step": 38774 + }, + { + "epoch": 14.32, + "learning_rate": 1.569097179455598e-06, + "loss": 0.3973, + "step": 38775 + }, + { + "epoch": 14.32, + "learning_rate": 1.5690602290922527e-06, + "loss": 0.3408, + "step": 38776 + }, + { + "epoch": 14.32, + "learning_rate": 1.5690232787289075e-06, + "loss": 0.1807, + "step": 38777 + }, + { + "epoch": 14.32, + "learning_rate": 1.5689863283655622e-06, + "loss": 0.2817, + "step": 38778 + }, + { + "epoch": 14.32, + "learning_rate": 1.5689493780022172e-06, + "loss": 0.2665, + "step": 38779 + }, + { + "epoch": 14.32, + "learning_rate": 1.5689124276388718e-06, + "loss": 0.3105, + "step": 38780 + }, + { + "epoch": 14.32, + "learning_rate": 1.5688754772755267e-06, + "loss": 0.2573, + "step": 38781 + }, + { + "epoch": 14.32, + "learning_rate": 1.5688385269121813e-06, + "loss": 0.2441, + "step": 38782 + }, + { + "epoch": 14.32, + "learning_rate": 1.5688015765488362e-06, + "loss": 0.2852, + "step": 38783 + }, + { + "epoch": 14.32, + "learning_rate": 1.5687646261854908e-06, + "loss": 0.187, + "step": 38784 + }, + { + "epoch": 14.32, + "learning_rate": 1.5687276758221455e-06, + "loss": 0.1853, + "step": 38785 + }, + { + "epoch": 14.32, + "learning_rate": 1.5686907254588005e-06, + "loss": 0.3515, + "step": 38786 + }, + { + "epoch": 14.32, + "learning_rate": 1.5686537750954552e-06, + "loss": 0.3353, + "step": 38787 + }, + { + "epoch": 14.32, + "learning_rate": 1.56861682473211e-06, + "loss": 0.1911, + "step": 38788 + }, + { + "epoch": 14.32, + "learning_rate": 1.5685798743687646e-06, + "loss": 0.2592, + "step": 38789 + }, + { + "epoch": 14.32, + "learning_rate": 1.5685429240054195e-06, + "loss": 0.2747, + "step": 38790 + }, + { + "epoch": 14.32, + "learning_rate": 1.5685059736420741e-06, + "loss": 0.2866, + "step": 38791 + }, + { + "epoch": 14.32, + "learning_rate": 1.5684690232787292e-06, + "loss": 0.2652, + "step": 38792 + }, + { + "epoch": 14.33, + "learning_rate": 1.5684320729153836e-06, + "loss": 0.2882, + "step": 38793 + }, + { + "epoch": 14.33, + "learning_rate": 1.5683951225520382e-06, + "loss": 0.1871, + "step": 38794 + }, + { + "epoch": 14.33, + "learning_rate": 1.5683581721886933e-06, + "loss": 0.303, + "step": 38795 + }, + { + "epoch": 14.33, + "learning_rate": 1.568321221825348e-06, + "loss": 0.3331, + "step": 38796 + }, + { + "epoch": 14.33, + "learning_rate": 1.5682842714620028e-06, + "loss": 0.2633, + "step": 38797 + }, + { + "epoch": 14.33, + "learning_rate": 1.5682473210986574e-06, + "loss": 0.2277, + "step": 38798 + }, + { + "epoch": 14.33, + "learning_rate": 1.5682103707353123e-06, + "loss": 0.1527, + "step": 38799 + }, + { + "epoch": 14.33, + "learning_rate": 1.568173420371967e-06, + "loss": 0.1787, + "step": 38800 + }, + { + "epoch": 14.33, + "learning_rate": 1.568136470008622e-06, + "loss": 0.1552, + "step": 38801 + }, + { + "epoch": 14.33, + "learning_rate": 1.5680995196452766e-06, + "loss": 0.2317, + "step": 38802 + }, + { + "epoch": 14.33, + "learning_rate": 1.5680625692819312e-06, + "loss": 0.2591, + "step": 38803 + }, + { + "epoch": 14.33, + "learning_rate": 1.568025618918586e-06, + "loss": 0.3166, + "step": 38804 + }, + { + "epoch": 14.33, + "learning_rate": 1.5679886685552407e-06, + "loss": 0.2623, + "step": 38805 + }, + { + "epoch": 14.33, + "learning_rate": 1.5679517181918956e-06, + "loss": 0.2377, + "step": 38806 + }, + { + "epoch": 14.33, + "learning_rate": 1.5679147678285502e-06, + "loss": 0.1534, + "step": 38807 + }, + { + "epoch": 14.33, + "learning_rate": 1.5678778174652053e-06, + "loss": 0.3122, + "step": 38808 + }, + { + "epoch": 14.33, + "learning_rate": 1.56784086710186e-06, + "loss": 0.1516, + "step": 38809 + }, + { + "epoch": 14.33, + "learning_rate": 1.5678039167385148e-06, + "loss": 0.2522, + "step": 38810 + }, + { + "epoch": 14.33, + "learning_rate": 1.5677669663751694e-06, + "loss": 0.2438, + "step": 38811 + }, + { + "epoch": 14.33, + "learning_rate": 1.5677300160118242e-06, + "loss": 0.3118, + "step": 38812 + }, + { + "epoch": 14.33, + "learning_rate": 1.5676930656484789e-06, + "loss": 0.173, + "step": 38813 + }, + { + "epoch": 14.33, + "learning_rate": 1.5676561152851335e-06, + "loss": 0.2911, + "step": 38814 + }, + { + "epoch": 14.33, + "learning_rate": 1.5676191649217886e-06, + "loss": 0.2351, + "step": 38815 + }, + { + "epoch": 14.33, + "learning_rate": 1.567582214558443e-06, + "loss": 0.1972, + "step": 38816 + }, + { + "epoch": 14.33, + "learning_rate": 1.567545264195098e-06, + "loss": 0.2866, + "step": 38817 + }, + { + "epoch": 14.33, + "learning_rate": 1.5675083138317527e-06, + "loss": 0.2921, + "step": 38818 + }, + { + "epoch": 14.33, + "learning_rate": 1.5674713634684075e-06, + "loss": 0.1992, + "step": 38819 + }, + { + "epoch": 14.34, + "learning_rate": 1.5674344131050622e-06, + "loss": 0.366, + "step": 38820 + }, + { + "epoch": 14.34, + "learning_rate": 1.5673974627417172e-06, + "loss": 0.1977, + "step": 38821 + }, + { + "epoch": 14.34, + "learning_rate": 1.5673605123783717e-06, + "loss": 0.3591, + "step": 38822 + }, + { + "epoch": 14.34, + "learning_rate": 1.5673235620150263e-06, + "loss": 0.2348, + "step": 38823 + }, + { + "epoch": 14.34, + "learning_rate": 1.5672866116516814e-06, + "loss": 0.2988, + "step": 38824 + }, + { + "epoch": 14.34, + "learning_rate": 1.567249661288336e-06, + "loss": 0.2927, + "step": 38825 + }, + { + "epoch": 14.34, + "learning_rate": 1.5672127109249908e-06, + "loss": 0.3034, + "step": 38826 + }, + { + "epoch": 14.34, + "learning_rate": 1.5671757605616455e-06, + "loss": 0.3829, + "step": 38827 + }, + { + "epoch": 14.34, + "learning_rate": 1.5671388101983003e-06, + "loss": 0.3184, + "step": 38828 + }, + { + "epoch": 14.34, + "learning_rate": 1.567101859834955e-06, + "loss": 0.3071, + "step": 38829 + }, + { + "epoch": 14.34, + "learning_rate": 1.56706490947161e-06, + "loss": 0.4191, + "step": 38830 + }, + { + "epoch": 14.34, + "learning_rate": 1.5670279591082647e-06, + "loss": 0.2401, + "step": 38831 + }, + { + "epoch": 14.34, + "learning_rate": 1.5669910087449195e-06, + "loss": 0.2128, + "step": 38832 + }, + { + "epoch": 14.34, + "learning_rate": 1.5669540583815741e-06, + "loss": 0.243, + "step": 38833 + }, + { + "epoch": 14.34, + "learning_rate": 1.5669171080182288e-06, + "loss": 0.2972, + "step": 38834 + }, + { + "epoch": 14.34, + "learning_rate": 1.5668801576548836e-06, + "loss": 0.2145, + "step": 38835 + }, + { + "epoch": 14.34, + "learning_rate": 1.5668432072915383e-06, + "loss": 0.2325, + "step": 38836 + }, + { + "epoch": 14.34, + "learning_rate": 1.5668062569281933e-06, + "loss": 0.2443, + "step": 38837 + }, + { + "epoch": 14.34, + "learning_rate": 1.5667693065648477e-06, + "loss": 0.294, + "step": 38838 + }, + { + "epoch": 14.34, + "learning_rate": 1.5667323562015028e-06, + "loss": 0.2839, + "step": 38839 + }, + { + "epoch": 14.34, + "learning_rate": 1.5666954058381574e-06, + "loss": 0.3142, + "step": 38840 + }, + { + "epoch": 14.34, + "learning_rate": 1.5666584554748123e-06, + "loss": 0.2983, + "step": 38841 + }, + { + "epoch": 14.34, + "learning_rate": 1.566621505111467e-06, + "loss": 0.3215, + "step": 38842 + }, + { + "epoch": 14.34, + "learning_rate": 1.5665845547481216e-06, + "loss": 0.311, + "step": 38843 + }, + { + "epoch": 14.34, + "learning_rate": 1.5665476043847764e-06, + "loss": 0.417, + "step": 38844 + }, + { + "epoch": 14.34, + "learning_rate": 1.566510654021431e-06, + "loss": 0.285, + "step": 38845 + }, + { + "epoch": 14.34, + "learning_rate": 1.5664737036580861e-06, + "loss": 0.3943, + "step": 38846 + }, + { + "epoch": 14.35, + "learning_rate": 1.5664367532947407e-06, + "loss": 0.2769, + "step": 38847 + }, + { + "epoch": 14.35, + "learning_rate": 1.5663998029313956e-06, + "loss": 0.2942, + "step": 38848 + }, + { + "epoch": 14.35, + "learning_rate": 1.5663628525680502e-06, + "loss": 0.3202, + "step": 38849 + }, + { + "epoch": 14.35, + "learning_rate": 1.566325902204705e-06, + "loss": 0.2608, + "step": 38850 + }, + { + "epoch": 14.35, + "learning_rate": 1.5662889518413597e-06, + "loss": 0.3085, + "step": 38851 + }, + { + "epoch": 14.35, + "learning_rate": 1.5662520014780148e-06, + "loss": 0.2636, + "step": 38852 + }, + { + "epoch": 14.35, + "learning_rate": 1.5662150511146694e-06, + "loss": 0.2641, + "step": 38853 + }, + { + "epoch": 14.35, + "learning_rate": 1.566178100751324e-06, + "loss": 0.4052, + "step": 38854 + }, + { + "epoch": 14.35, + "learning_rate": 1.566141150387979e-06, + "loss": 0.2596, + "step": 38855 + }, + { + "epoch": 14.35, + "learning_rate": 1.5661042000246335e-06, + "loss": 0.3904, + "step": 38856 + }, + { + "epoch": 14.35, + "learning_rate": 1.5660672496612884e-06, + "loss": 0.3537, + "step": 38857 + }, + { + "epoch": 14.35, + "learning_rate": 1.566030299297943e-06, + "loss": 0.1913, + "step": 38858 + }, + { + "epoch": 14.35, + "learning_rate": 1.565993348934598e-06, + "loss": 0.2183, + "step": 38859 + }, + { + "epoch": 14.35, + "learning_rate": 1.5659563985712527e-06, + "loss": 0.2303, + "step": 38860 + }, + { + "epoch": 14.35, + "learning_rate": 1.5659194482079076e-06, + "loss": 0.2464, + "step": 38861 + }, + { + "epoch": 14.35, + "learning_rate": 1.5658824978445622e-06, + "loss": 0.3399, + "step": 38862 + }, + { + "epoch": 14.35, + "learning_rate": 1.5658455474812168e-06, + "loss": 0.2584, + "step": 38863 + }, + { + "epoch": 14.35, + "learning_rate": 1.5658085971178717e-06, + "loss": 0.3482, + "step": 38864 + }, + { + "epoch": 14.35, + "learning_rate": 1.5657716467545263e-06, + "loss": 0.3024, + "step": 38865 + }, + { + "epoch": 14.35, + "learning_rate": 1.5657346963911812e-06, + "loss": 0.21, + "step": 38866 + }, + { + "epoch": 14.35, + "learning_rate": 1.5656977460278358e-06, + "loss": 0.3048, + "step": 38867 + }, + { + "epoch": 14.35, + "learning_rate": 1.5656607956644909e-06, + "loss": 0.2648, + "step": 38868 + }, + { + "epoch": 14.35, + "learning_rate": 1.5656238453011455e-06, + "loss": 0.3189, + "step": 38869 + }, + { + "epoch": 14.35, + "learning_rate": 1.5655868949378004e-06, + "loss": 0.2795, + "step": 38870 + }, + { + "epoch": 14.35, + "learning_rate": 1.565549944574455e-06, + "loss": 0.1556, + "step": 38871 + }, + { + "epoch": 14.35, + "learning_rate": 1.5655129942111098e-06, + "loss": 0.32, + "step": 38872 + }, + { + "epoch": 14.35, + "learning_rate": 1.5654760438477645e-06, + "loss": 0.2347, + "step": 38873 + }, + { + "epoch": 14.36, + "learning_rate": 1.5654390934844191e-06, + "loss": 0.3291, + "step": 38874 + }, + { + "epoch": 14.36, + "learning_rate": 1.5654021431210742e-06, + "loss": 0.2115, + "step": 38875 + }, + { + "epoch": 14.36, + "learning_rate": 1.5653651927577288e-06, + "loss": 0.2448, + "step": 38876 + }, + { + "epoch": 14.36, + "learning_rate": 1.5653282423943837e-06, + "loss": 0.3754, + "step": 38877 + }, + { + "epoch": 14.36, + "learning_rate": 1.5652912920310383e-06, + "loss": 0.3842, + "step": 38878 + }, + { + "epoch": 14.36, + "learning_rate": 1.5652543416676931e-06, + "loss": 0.3883, + "step": 38879 + }, + { + "epoch": 14.36, + "learning_rate": 1.5652173913043478e-06, + "loss": 0.2016, + "step": 38880 + }, + { + "epoch": 14.36, + "learning_rate": 1.5651804409410028e-06, + "loss": 0.3537, + "step": 38881 + }, + { + "epoch": 14.36, + "learning_rate": 1.5651434905776575e-06, + "loss": 0.4166, + "step": 38882 + }, + { + "epoch": 14.36, + "learning_rate": 1.565106540214312e-06, + "loss": 0.2376, + "step": 38883 + }, + { + "epoch": 14.36, + "learning_rate": 1.565069589850967e-06, + "loss": 0.2824, + "step": 38884 + }, + { + "epoch": 14.36, + "learning_rate": 1.5650326394876216e-06, + "loss": 0.2916, + "step": 38885 + }, + { + "epoch": 14.36, + "learning_rate": 1.5649956891242764e-06, + "loss": 0.3256, + "step": 38886 + }, + { + "epoch": 14.36, + "learning_rate": 1.564958738760931e-06, + "loss": 0.2801, + "step": 38887 + }, + { + "epoch": 14.36, + "learning_rate": 1.5649217883975861e-06, + "loss": 0.2265, + "step": 38888 + }, + { + "epoch": 14.36, + "learning_rate": 1.5648848380342406e-06, + "loss": 0.2956, + "step": 38889 + }, + { + "epoch": 14.36, + "learning_rate": 1.5648478876708956e-06, + "loss": 0.2859, + "step": 38890 + }, + { + "epoch": 14.36, + "learning_rate": 1.5648109373075503e-06, + "loss": 0.2634, + "step": 38891 + }, + { + "epoch": 14.36, + "learning_rate": 1.564773986944205e-06, + "loss": 0.1874, + "step": 38892 + }, + { + "epoch": 14.36, + "learning_rate": 1.5647370365808597e-06, + "loss": 0.3258, + "step": 38893 + }, + { + "epoch": 14.36, + "learning_rate": 1.5647000862175144e-06, + "loss": 0.383, + "step": 38894 + }, + { + "epoch": 14.36, + "learning_rate": 1.5646631358541692e-06, + "loss": 0.253, + "step": 38895 + }, + { + "epoch": 14.36, + "learning_rate": 1.5646261854908239e-06, + "loss": 0.3192, + "step": 38896 + }, + { + "epoch": 14.36, + "learning_rate": 1.564589235127479e-06, + "loss": 0.1149, + "step": 38897 + }, + { + "epoch": 14.36, + "learning_rate": 1.5645522847641336e-06, + "loss": 0.2965, + "step": 38898 + }, + { + "epoch": 14.36, + "learning_rate": 1.5645153344007884e-06, + "loss": 0.3316, + "step": 38899 + }, + { + "epoch": 14.36, + "learning_rate": 1.564478384037443e-06, + "loss": 0.3395, + "step": 38900 + }, + { + "epoch": 14.37, + "learning_rate": 1.5644414336740979e-06, + "loss": 0.2875, + "step": 38901 + }, + { + "epoch": 14.37, + "learning_rate": 1.5644044833107525e-06, + "loss": 0.2226, + "step": 38902 + }, + { + "epoch": 14.37, + "learning_rate": 1.5643675329474072e-06, + "loss": 0.3685, + "step": 38903 + }, + { + "epoch": 14.37, + "learning_rate": 1.5643305825840622e-06, + "loss": 0.1812, + "step": 38904 + }, + { + "epoch": 14.37, + "learning_rate": 1.5642936322207166e-06, + "loss": 0.28, + "step": 38905 + }, + { + "epoch": 14.37, + "learning_rate": 1.5642566818573717e-06, + "loss": 0.1574, + "step": 38906 + }, + { + "epoch": 14.37, + "learning_rate": 1.5642197314940263e-06, + "loss": 0.1719, + "step": 38907 + }, + { + "epoch": 14.37, + "learning_rate": 1.5641827811306812e-06, + "loss": 0.2461, + "step": 38908 + }, + { + "epoch": 14.37, + "learning_rate": 1.5641458307673358e-06, + "loss": 0.2515, + "step": 38909 + }, + { + "epoch": 14.37, + "learning_rate": 1.5641088804039909e-06, + "loss": 0.1857, + "step": 38910 + }, + { + "epoch": 14.37, + "learning_rate": 1.5640719300406453e-06, + "loss": 0.239, + "step": 38911 + }, + { + "epoch": 14.37, + "learning_rate": 1.5640349796773004e-06, + "loss": 0.2768, + "step": 38912 + }, + { + "epoch": 14.37, + "learning_rate": 1.563998029313955e-06, + "loss": 0.2285, + "step": 38913 + }, + { + "epoch": 14.37, + "learning_rate": 1.5639610789506096e-06, + "loss": 0.3451, + "step": 38914 + }, + { + "epoch": 14.37, + "learning_rate": 1.5639241285872645e-06, + "loss": 0.1275, + "step": 38915 + }, + { + "epoch": 14.37, + "learning_rate": 1.5638871782239191e-06, + "loss": 0.3406, + "step": 38916 + }, + { + "epoch": 14.37, + "learning_rate": 1.563850227860574e-06, + "loss": 0.4737, + "step": 38917 + }, + { + "epoch": 14.37, + "learning_rate": 1.5638132774972286e-06, + "loss": 0.3863, + "step": 38918 + }, + { + "epoch": 14.37, + "learning_rate": 1.5637763271338837e-06, + "loss": 0.2228, + "step": 38919 + }, + { + "epoch": 14.37, + "learning_rate": 1.5637393767705383e-06, + "loss": 0.2931, + "step": 38920 + }, + { + "epoch": 14.37, + "learning_rate": 1.5637024264071932e-06, + "loss": 0.2809, + "step": 38921 + }, + { + "epoch": 14.37, + "learning_rate": 1.5636654760438478e-06, + "loss": 0.202, + "step": 38922 + }, + { + "epoch": 14.37, + "learning_rate": 1.5636285256805024e-06, + "loss": 0.3216, + "step": 38923 + }, + { + "epoch": 14.37, + "learning_rate": 1.5635915753171573e-06, + "loss": 0.3875, + "step": 38924 + }, + { + "epoch": 14.37, + "learning_rate": 1.563554624953812e-06, + "loss": 0.3221, + "step": 38925 + }, + { + "epoch": 14.37, + "learning_rate": 1.563517674590467e-06, + "loss": 0.1948, + "step": 38926 + }, + { + "epoch": 14.37, + "learning_rate": 1.5634807242271216e-06, + "loss": 0.2477, + "step": 38927 + }, + { + "epoch": 14.38, + "learning_rate": 1.5634437738637765e-06, + "loss": 0.3456, + "step": 38928 + }, + { + "epoch": 14.38, + "learning_rate": 1.563406823500431e-06, + "loss": 0.2038, + "step": 38929 + }, + { + "epoch": 14.38, + "learning_rate": 1.563369873137086e-06, + "loss": 0.2097, + "step": 38930 + }, + { + "epoch": 14.38, + "learning_rate": 1.5633329227737406e-06, + "loss": 0.1855, + "step": 38931 + }, + { + "epoch": 14.38, + "learning_rate": 1.5632959724103956e-06, + "loss": 0.2293, + "step": 38932 + }, + { + "epoch": 14.38, + "learning_rate": 1.5632590220470503e-06, + "loss": 0.1678, + "step": 38933 + }, + { + "epoch": 14.38, + "learning_rate": 1.5632220716837047e-06, + "loss": 0.131, + "step": 38934 + }, + { + "epoch": 14.38, + "learning_rate": 1.5631851213203598e-06, + "loss": 0.2209, + "step": 38935 + }, + { + "epoch": 14.38, + "learning_rate": 1.5631481709570144e-06, + "loss": 0.3028, + "step": 38936 + }, + { + "epoch": 14.38, + "learning_rate": 1.5631112205936693e-06, + "loss": 0.3718, + "step": 38937 + }, + { + "epoch": 14.38, + "learning_rate": 1.5630742702303239e-06, + "loss": 0.1931, + "step": 38938 + }, + { + "epoch": 14.38, + "learning_rate": 1.563037319866979e-06, + "loss": 0.1399, + "step": 38939 + }, + { + "epoch": 14.38, + "learning_rate": 1.5630003695036334e-06, + "loss": 0.353, + "step": 38940 + }, + { + "epoch": 14.38, + "learning_rate": 1.5629634191402884e-06, + "loss": 0.3903, + "step": 38941 + }, + { + "epoch": 14.38, + "learning_rate": 1.562926468776943e-06, + "loss": 0.1553, + "step": 38942 + }, + { + "epoch": 14.38, + "learning_rate": 1.5628895184135977e-06, + "loss": 0.1893, + "step": 38943 + }, + { + "epoch": 14.38, + "learning_rate": 1.5628525680502526e-06, + "loss": 0.2534, + "step": 38944 + }, + { + "epoch": 14.38, + "learning_rate": 1.5628156176869072e-06, + "loss": 0.2115, + "step": 38945 + }, + { + "epoch": 14.38, + "learning_rate": 1.562778667323562e-06, + "loss": 0.357, + "step": 38946 + }, + { + "epoch": 14.38, + "learning_rate": 1.5627417169602167e-06, + "loss": 0.3654, + "step": 38947 + }, + { + "epoch": 14.38, + "learning_rate": 1.5627047665968717e-06, + "loss": 0.3226, + "step": 38948 + }, + { + "epoch": 14.38, + "learning_rate": 1.5626678162335264e-06, + "loss": 0.4655, + "step": 38949 + }, + { + "epoch": 14.38, + "learning_rate": 1.5626308658701812e-06, + "loss": 0.2564, + "step": 38950 + }, + { + "epoch": 14.38, + "learning_rate": 1.5625939155068359e-06, + "loss": 0.171, + "step": 38951 + }, + { + "epoch": 14.38, + "learning_rate": 1.5625569651434907e-06, + "loss": 0.2865, + "step": 38952 + }, + { + "epoch": 14.38, + "learning_rate": 1.5625200147801453e-06, + "loss": 0.2965, + "step": 38953 + }, + { + "epoch": 14.38, + "learning_rate": 1.5624830644168e-06, + "loss": 0.2177, + "step": 38954 + }, + { + "epoch": 14.39, + "learning_rate": 1.562446114053455e-06, + "loss": 0.2551, + "step": 38955 + }, + { + "epoch": 14.39, + "learning_rate": 1.5624091636901095e-06, + "loss": 0.1743, + "step": 38956 + }, + { + "epoch": 14.39, + "learning_rate": 1.5623722133267645e-06, + "loss": 0.1823, + "step": 38957 + }, + { + "epoch": 14.39, + "learning_rate": 1.5623352629634192e-06, + "loss": 0.232, + "step": 38958 + }, + { + "epoch": 14.39, + "learning_rate": 1.562298312600074e-06, + "loss": 0.1563, + "step": 38959 + }, + { + "epoch": 14.39, + "learning_rate": 1.5622613622367286e-06, + "loss": 0.2263, + "step": 38960 + }, + { + "epoch": 14.39, + "learning_rate": 1.5622244118733837e-06, + "loss": 0.2134, + "step": 38961 + }, + { + "epoch": 14.39, + "learning_rate": 1.5621874615100381e-06, + "loss": 0.2128, + "step": 38962 + }, + { + "epoch": 14.39, + "learning_rate": 1.5621505111466928e-06, + "loss": 0.1885, + "step": 38963 + }, + { + "epoch": 14.39, + "learning_rate": 1.5621135607833478e-06, + "loss": 0.2416, + "step": 38964 + }, + { + "epoch": 14.39, + "learning_rate": 1.5620766104200025e-06, + "loss": 0.2673, + "step": 38965 + }, + { + "epoch": 14.39, + "learning_rate": 1.5620396600566573e-06, + "loss": 0.2609, + "step": 38966 + }, + { + "epoch": 14.39, + "learning_rate": 1.562002709693312e-06, + "loss": 0.1776, + "step": 38967 + }, + { + "epoch": 14.39, + "learning_rate": 1.5619657593299668e-06, + "loss": 0.2397, + "step": 38968 + }, + { + "epoch": 14.39, + "learning_rate": 1.5619288089666214e-06, + "loss": 0.2192, + "step": 38969 + }, + { + "epoch": 14.39, + "learning_rate": 1.5618918586032765e-06, + "loss": 0.2181, + "step": 38970 + }, + { + "epoch": 14.39, + "learning_rate": 1.5618549082399311e-06, + "loss": 0.2875, + "step": 38971 + }, + { + "epoch": 14.39, + "learning_rate": 1.561817957876586e-06, + "loss": 0.2459, + "step": 38972 + }, + { + "epoch": 14.39, + "learning_rate": 1.5617810075132406e-06, + "loss": 0.2184, + "step": 38973 + }, + { + "epoch": 14.39, + "learning_rate": 1.5617440571498952e-06, + "loss": 0.238, + "step": 38974 + }, + { + "epoch": 14.39, + "learning_rate": 1.56170710678655e-06, + "loss": 0.2883, + "step": 38975 + }, + { + "epoch": 14.39, + "learning_rate": 1.5616701564232047e-06, + "loss": 0.3085, + "step": 38976 + }, + { + "epoch": 14.39, + "learning_rate": 1.5616332060598598e-06, + "loss": 0.1979, + "step": 38977 + }, + { + "epoch": 14.39, + "learning_rate": 1.5615962556965142e-06, + "loss": 0.2235, + "step": 38978 + }, + { + "epoch": 14.39, + "learning_rate": 1.5615593053331693e-06, + "loss": 0.3776, + "step": 38979 + }, + { + "epoch": 14.39, + "learning_rate": 1.561522354969824e-06, + "loss": 0.2621, + "step": 38980 + }, + { + "epoch": 14.39, + "learning_rate": 1.5614854046064788e-06, + "loss": 0.387, + "step": 38981 + }, + { + "epoch": 14.4, + "learning_rate": 1.5614484542431334e-06, + "loss": 0.1791, + "step": 38982 + }, + { + "epoch": 14.4, + "learning_rate": 1.561411503879788e-06, + "loss": 0.1542, + "step": 38983 + }, + { + "epoch": 14.4, + "learning_rate": 1.5613745535164429e-06, + "loss": 0.1402, + "step": 38984 + }, + { + "epoch": 14.4, + "learning_rate": 1.5613376031530975e-06, + "loss": 0.2737, + "step": 38985 + }, + { + "epoch": 14.4, + "learning_rate": 1.5613006527897526e-06, + "loss": 0.2276, + "step": 38986 + }, + { + "epoch": 14.4, + "learning_rate": 1.5612637024264072e-06, + "loss": 0.2515, + "step": 38987 + }, + { + "epoch": 14.4, + "learning_rate": 1.561226752063062e-06, + "loss": 0.189, + "step": 38988 + }, + { + "epoch": 14.4, + "learning_rate": 1.5611898016997167e-06, + "loss": 0.3316, + "step": 38989 + }, + { + "epoch": 14.4, + "learning_rate": 1.5611528513363715e-06, + "loss": 0.2757, + "step": 38990 + }, + { + "epoch": 14.4, + "learning_rate": 1.5611159009730262e-06, + "loss": 0.3399, + "step": 38991 + }, + { + "epoch": 14.4, + "learning_rate": 1.5610789506096812e-06, + "loss": 0.2989, + "step": 38992 + }, + { + "epoch": 14.4, + "learning_rate": 1.5610420002463359e-06, + "loss": 0.3162, + "step": 38993 + }, + { + "epoch": 14.4, + "learning_rate": 1.5610050498829905e-06, + "loss": 0.3402, + "step": 38994 + }, + { + "epoch": 14.4, + "learning_rate": 1.5609680995196454e-06, + "loss": 0.2779, + "step": 38995 + }, + { + "epoch": 14.4, + "learning_rate": 1.5609311491563e-06, + "loss": 0.2982, + "step": 38996 + }, + { + "epoch": 14.4, + "learning_rate": 1.5608941987929548e-06, + "loss": 0.2703, + "step": 38997 + }, + { + "epoch": 14.4, + "learning_rate": 1.5608572484296095e-06, + "loss": 0.2096, + "step": 38998 + }, + { + "epoch": 14.4, + "learning_rate": 1.5608202980662645e-06, + "loss": 0.1782, + "step": 38999 + }, + { + "epoch": 14.4, + "learning_rate": 1.5607833477029192e-06, + "loss": 0.2337, + "step": 39000 + }, + { + "epoch": 14.4, + "learning_rate": 1.560746397339574e-06, + "loss": 0.2443, + "step": 39001 + }, + { + "epoch": 14.4, + "learning_rate": 1.5607094469762287e-06, + "loss": 0.2204, + "step": 39002 + }, + { + "epoch": 14.4, + "learning_rate": 1.5606724966128833e-06, + "loss": 0.2701, + "step": 39003 + }, + { + "epoch": 14.4, + "learning_rate": 1.5606355462495381e-06, + "loss": 0.2268, + "step": 39004 + }, + { + "epoch": 14.4, + "learning_rate": 1.5605985958861928e-06, + "loss": 0.3462, + "step": 39005 + }, + { + "epoch": 14.4, + "learning_rate": 1.5605616455228478e-06, + "loss": 0.2982, + "step": 39006 + }, + { + "epoch": 14.4, + "learning_rate": 1.5605246951595023e-06, + "loss": 0.2585, + "step": 39007 + }, + { + "epoch": 14.4, + "learning_rate": 1.5604877447961573e-06, + "loss": 0.2818, + "step": 39008 + }, + { + "epoch": 14.41, + "learning_rate": 1.560450794432812e-06, + "loss": 0.2898, + "step": 39009 + }, + { + "epoch": 14.41, + "learning_rate": 1.5604138440694668e-06, + "loss": 0.3425, + "step": 39010 + }, + { + "epoch": 14.41, + "learning_rate": 1.5603768937061215e-06, + "loss": 0.3008, + "step": 39011 + }, + { + "epoch": 14.41, + "learning_rate": 1.5603399433427765e-06, + "loss": 0.2475, + "step": 39012 + }, + { + "epoch": 14.41, + "learning_rate": 1.560302992979431e-06, + "loss": 0.166, + "step": 39013 + }, + { + "epoch": 14.41, + "learning_rate": 1.5602660426160856e-06, + "loss": 0.3379, + "step": 39014 + }, + { + "epoch": 14.41, + "learning_rate": 1.5602290922527406e-06, + "loss": 0.2567, + "step": 39015 + }, + { + "epoch": 14.41, + "learning_rate": 1.5601921418893953e-06, + "loss": 0.2172, + "step": 39016 + }, + { + "epoch": 14.41, + "learning_rate": 1.5601551915260501e-06, + "loss": 0.1362, + "step": 39017 + }, + { + "epoch": 14.41, + "learning_rate": 1.5601182411627048e-06, + "loss": 0.2163, + "step": 39018 + }, + { + "epoch": 14.41, + "learning_rate": 1.5600812907993596e-06, + "loss": 0.2384, + "step": 39019 + }, + { + "epoch": 14.41, + "learning_rate": 1.5600443404360142e-06, + "loss": 0.2398, + "step": 39020 + }, + { + "epoch": 14.41, + "learning_rate": 1.5600073900726693e-06, + "loss": 0.3721, + "step": 39021 + }, + { + "epoch": 14.41, + "learning_rate": 1.559970439709324e-06, + "loss": 0.2287, + "step": 39022 + }, + { + "epoch": 14.41, + "learning_rate": 1.5599334893459784e-06, + "loss": 0.2515, + "step": 39023 + }, + { + "epoch": 14.41, + "learning_rate": 1.5598965389826334e-06, + "loss": 0.3301, + "step": 39024 + }, + { + "epoch": 14.41, + "learning_rate": 1.559859588619288e-06, + "loss": 0.3235, + "step": 39025 + }, + { + "epoch": 14.41, + "learning_rate": 1.559822638255943e-06, + "loss": 0.1768, + "step": 39026 + }, + { + "epoch": 14.41, + "learning_rate": 1.5597856878925975e-06, + "loss": 0.4311, + "step": 39027 + }, + { + "epoch": 14.41, + "learning_rate": 1.5597487375292526e-06, + "loss": 0.22, + "step": 39028 + }, + { + "epoch": 14.41, + "learning_rate": 1.559711787165907e-06, + "loss": 0.1628, + "step": 39029 + }, + { + "epoch": 14.41, + "learning_rate": 1.559674836802562e-06, + "loss": 0.2815, + "step": 39030 + }, + { + "epoch": 14.41, + "learning_rate": 1.5596378864392167e-06, + "loss": 0.2927, + "step": 39031 + }, + { + "epoch": 14.41, + "learning_rate": 1.5596009360758714e-06, + "loss": 0.2786, + "step": 39032 + }, + { + "epoch": 14.41, + "learning_rate": 1.5595639857125262e-06, + "loss": 0.2956, + "step": 39033 + }, + { + "epoch": 14.41, + "learning_rate": 1.5595270353491808e-06, + "loss": 0.1655, + "step": 39034 + }, + { + "epoch": 14.41, + "learning_rate": 1.5594900849858357e-06, + "loss": 0.3938, + "step": 39035 + }, + { + "epoch": 14.42, + "learning_rate": 1.5594531346224903e-06, + "loss": 0.2849, + "step": 39036 + }, + { + "epoch": 14.42, + "learning_rate": 1.5594161842591454e-06, + "loss": 0.1995, + "step": 39037 + }, + { + "epoch": 14.42, + "learning_rate": 1.5593792338958e-06, + "loss": 0.2163, + "step": 39038 + }, + { + "epoch": 14.42, + "learning_rate": 1.5593422835324549e-06, + "loss": 0.2945, + "step": 39039 + }, + { + "epoch": 14.42, + "learning_rate": 1.5593053331691095e-06, + "loss": 0.2767, + "step": 39040 + }, + { + "epoch": 14.42, + "learning_rate": 1.5592683828057644e-06, + "loss": 0.2851, + "step": 39041 + }, + { + "epoch": 14.42, + "learning_rate": 1.559231432442419e-06, + "loss": 0.3346, + "step": 39042 + }, + { + "epoch": 14.42, + "learning_rate": 1.5591944820790736e-06, + "loss": 0.2656, + "step": 39043 + }, + { + "epoch": 14.42, + "learning_rate": 1.5591575317157287e-06, + "loss": 0.2071, + "step": 39044 + }, + { + "epoch": 14.42, + "learning_rate": 1.5591205813523833e-06, + "loss": 0.2352, + "step": 39045 + }, + { + "epoch": 14.42, + "learning_rate": 1.5590836309890382e-06, + "loss": 0.3056, + "step": 39046 + }, + { + "epoch": 14.42, + "learning_rate": 1.5590466806256928e-06, + "loss": 0.1576, + "step": 39047 + }, + { + "epoch": 14.42, + "learning_rate": 1.5590097302623477e-06, + "loss": 0.2276, + "step": 39048 + }, + { + "epoch": 14.42, + "learning_rate": 1.5589727798990023e-06, + "loss": 0.3892, + "step": 39049 + }, + { + "epoch": 14.42, + "learning_rate": 1.5589358295356574e-06, + "loss": 0.3898, + "step": 39050 + }, + { + "epoch": 14.42, + "learning_rate": 1.558898879172312e-06, + "loss": 0.287, + "step": 39051 + }, + { + "epoch": 14.42, + "learning_rate": 1.5588619288089664e-06, + "loss": 0.2318, + "step": 39052 + }, + { + "epoch": 14.42, + "learning_rate": 1.5588249784456215e-06, + "loss": 0.3845, + "step": 39053 + }, + { + "epoch": 14.42, + "learning_rate": 1.5587880280822761e-06, + "loss": 0.3312, + "step": 39054 + }, + { + "epoch": 14.42, + "learning_rate": 1.558751077718931e-06, + "loss": 0.2239, + "step": 39055 + }, + { + "epoch": 14.42, + "learning_rate": 1.5587141273555856e-06, + "loss": 0.2089, + "step": 39056 + }, + { + "epoch": 14.42, + "learning_rate": 1.5586771769922404e-06, + "loss": 0.3609, + "step": 39057 + }, + { + "epoch": 14.42, + "learning_rate": 1.558640226628895e-06, + "loss": 0.3732, + "step": 39058 + }, + { + "epoch": 14.42, + "learning_rate": 1.5586032762655501e-06, + "loss": 0.3241, + "step": 39059 + }, + { + "epoch": 14.42, + "learning_rate": 1.5585663259022048e-06, + "loss": 0.278, + "step": 39060 + }, + { + "epoch": 14.42, + "learning_rate": 1.5585293755388596e-06, + "loss": 0.2887, + "step": 39061 + }, + { + "epoch": 14.42, + "learning_rate": 1.5584924251755143e-06, + "loss": 0.251, + "step": 39062 + }, + { + "epoch": 14.43, + "learning_rate": 1.558455474812169e-06, + "loss": 0.1962, + "step": 39063 + }, + { + "epoch": 14.43, + "learning_rate": 1.5584185244488237e-06, + "loss": 0.54, + "step": 39064 + }, + { + "epoch": 14.43, + "learning_rate": 1.5583815740854784e-06, + "loss": 0.3354, + "step": 39065 + }, + { + "epoch": 14.43, + "learning_rate": 1.5583446237221334e-06, + "loss": 0.3075, + "step": 39066 + }, + { + "epoch": 14.43, + "learning_rate": 1.558307673358788e-06, + "loss": 0.1483, + "step": 39067 + }, + { + "epoch": 14.43, + "learning_rate": 1.558270722995443e-06, + "loss": 0.3241, + "step": 39068 + }, + { + "epoch": 14.43, + "learning_rate": 1.5582337726320976e-06, + "loss": 0.1529, + "step": 39069 + }, + { + "epoch": 14.43, + "learning_rate": 1.5581968222687524e-06, + "loss": 0.2533, + "step": 39070 + }, + { + "epoch": 14.43, + "learning_rate": 1.558159871905407e-06, + "loss": 0.4172, + "step": 39071 + }, + { + "epoch": 14.43, + "learning_rate": 1.5581229215420617e-06, + "loss": 0.2231, + "step": 39072 + }, + { + "epoch": 14.43, + "learning_rate": 1.5580859711787167e-06, + "loss": 0.2386, + "step": 39073 + }, + { + "epoch": 14.43, + "learning_rate": 1.5580490208153712e-06, + "loss": 0.3265, + "step": 39074 + }, + { + "epoch": 14.43, + "learning_rate": 1.5580120704520262e-06, + "loss": 0.3703, + "step": 39075 + }, + { + "epoch": 14.43, + "learning_rate": 1.5579751200886809e-06, + "loss": 0.1982, + "step": 39076 + }, + { + "epoch": 14.43, + "learning_rate": 1.5579381697253357e-06, + "loss": 0.3232, + "step": 39077 + }, + { + "epoch": 14.43, + "learning_rate": 1.5579012193619904e-06, + "loss": 0.3405, + "step": 39078 + }, + { + "epoch": 14.43, + "learning_rate": 1.5578642689986454e-06, + "loss": 0.3335, + "step": 39079 + }, + { + "epoch": 14.43, + "learning_rate": 1.5578273186352998e-06, + "loss": 0.3682, + "step": 39080 + }, + { + "epoch": 14.43, + "learning_rate": 1.557790368271955e-06, + "loss": 0.3453, + "step": 39081 + }, + { + "epoch": 14.43, + "learning_rate": 1.5577534179086095e-06, + "loss": 0.2754, + "step": 39082 + }, + { + "epoch": 14.43, + "learning_rate": 1.5577164675452642e-06, + "loss": 0.3381, + "step": 39083 + }, + { + "epoch": 14.43, + "learning_rate": 1.557679517181919e-06, + "loss": 0.2479, + "step": 39084 + }, + { + "epoch": 14.43, + "learning_rate": 1.5576425668185737e-06, + "loss": 0.239, + "step": 39085 + }, + { + "epoch": 14.43, + "learning_rate": 1.5576056164552285e-06, + "loss": 0.2983, + "step": 39086 + }, + { + "epoch": 14.43, + "learning_rate": 1.5575686660918831e-06, + "loss": 0.2556, + "step": 39087 + }, + { + "epoch": 14.43, + "learning_rate": 1.5575317157285382e-06, + "loss": 0.27, + "step": 39088 + }, + { + "epoch": 14.43, + "learning_rate": 1.5574947653651928e-06, + "loss": 0.217, + "step": 39089 + }, + { + "epoch": 14.44, + "learning_rate": 1.5574578150018477e-06, + "loss": 0.2734, + "step": 39090 + }, + { + "epoch": 14.44, + "learning_rate": 1.5574208646385023e-06, + "loss": 0.2898, + "step": 39091 + }, + { + "epoch": 14.44, + "learning_rate": 1.557383914275157e-06, + "loss": 0.2595, + "step": 39092 + }, + { + "epoch": 14.44, + "learning_rate": 1.5573469639118118e-06, + "loss": 0.3129, + "step": 39093 + }, + { + "epoch": 14.44, + "learning_rate": 1.5573100135484664e-06, + "loss": 0.3741, + "step": 39094 + }, + { + "epoch": 14.44, + "learning_rate": 1.5572730631851215e-06, + "loss": 0.3874, + "step": 39095 + }, + { + "epoch": 14.44, + "learning_rate": 1.557236112821776e-06, + "loss": 0.1948, + "step": 39096 + }, + { + "epoch": 14.44, + "learning_rate": 1.557199162458431e-06, + "loss": 0.2379, + "step": 39097 + }, + { + "epoch": 14.44, + "learning_rate": 1.5571622120950856e-06, + "loss": 0.3076, + "step": 39098 + }, + { + "epoch": 14.44, + "learning_rate": 1.5571252617317405e-06, + "loss": 0.2695, + "step": 39099 + }, + { + "epoch": 14.44, + "learning_rate": 1.5570883113683951e-06, + "loss": 0.2297, + "step": 39100 + }, + { + "epoch": 14.44, + "learning_rate": 1.5570513610050502e-06, + "loss": 0.4035, + "step": 39101 + }, + { + "epoch": 14.44, + "learning_rate": 1.5570144106417046e-06, + "loss": 0.3102, + "step": 39102 + }, + { + "epoch": 14.44, + "learning_rate": 1.5569774602783592e-06, + "loss": 0.2765, + "step": 39103 + }, + { + "epoch": 14.44, + "learning_rate": 1.5569405099150143e-06, + "loss": 0.1937, + "step": 39104 + }, + { + "epoch": 14.44, + "learning_rate": 1.556903559551669e-06, + "loss": 0.2897, + "step": 39105 + }, + { + "epoch": 14.44, + "learning_rate": 1.5568666091883238e-06, + "loss": 0.2612, + "step": 39106 + }, + { + "epoch": 14.44, + "learning_rate": 1.5568296588249784e-06, + "loss": 0.2857, + "step": 39107 + }, + { + "epoch": 14.44, + "learning_rate": 1.5567927084616333e-06, + "loss": 0.2378, + "step": 39108 + }, + { + "epoch": 14.44, + "learning_rate": 1.5567557580982879e-06, + "loss": 0.4022, + "step": 39109 + }, + { + "epoch": 14.44, + "learning_rate": 1.556718807734943e-06, + "loss": 0.1804, + "step": 39110 + }, + { + "epoch": 14.44, + "learning_rate": 1.5566818573715976e-06, + "loss": 0.2055, + "step": 39111 + }, + { + "epoch": 14.44, + "learning_rate": 1.5566449070082522e-06, + "loss": 0.2441, + "step": 39112 + }, + { + "epoch": 14.44, + "learning_rate": 1.556607956644907e-06, + "loss": 0.2618, + "step": 39113 + }, + { + "epoch": 14.44, + "learning_rate": 1.5565710062815617e-06, + "loss": 0.3559, + "step": 39114 + }, + { + "epoch": 14.44, + "learning_rate": 1.5565340559182166e-06, + "loss": 0.1916, + "step": 39115 + }, + { + "epoch": 14.44, + "learning_rate": 1.5564971055548712e-06, + "loss": 0.2723, + "step": 39116 + }, + { + "epoch": 14.44, + "learning_rate": 1.5564601551915263e-06, + "loss": 0.3731, + "step": 39117 + }, + { + "epoch": 14.45, + "learning_rate": 1.5564232048281809e-06, + "loss": 0.3016, + "step": 39118 + }, + { + "epoch": 14.45, + "learning_rate": 1.5563862544648357e-06, + "loss": 0.1883, + "step": 39119 + }, + { + "epoch": 14.45, + "learning_rate": 1.5563493041014904e-06, + "loss": 0.2725, + "step": 39120 + }, + { + "epoch": 14.45, + "learning_rate": 1.5563123537381452e-06, + "loss": 0.3454, + "step": 39121 + }, + { + "epoch": 14.45, + "learning_rate": 1.5562754033747999e-06, + "loss": 0.2426, + "step": 39122 + }, + { + "epoch": 14.45, + "learning_rate": 1.5562384530114545e-06, + "loss": 0.3368, + "step": 39123 + }, + { + "epoch": 14.45, + "learning_rate": 1.5562015026481096e-06, + "loss": 0.2864, + "step": 39124 + }, + { + "epoch": 14.45, + "learning_rate": 1.556164552284764e-06, + "loss": 0.2718, + "step": 39125 + }, + { + "epoch": 14.45, + "learning_rate": 1.556127601921419e-06, + "loss": 0.3216, + "step": 39126 + }, + { + "epoch": 14.45, + "learning_rate": 1.5560906515580737e-06, + "loss": 0.2021, + "step": 39127 + }, + { + "epoch": 14.45, + "learning_rate": 1.5560537011947285e-06, + "loss": 0.2153, + "step": 39128 + }, + { + "epoch": 14.45, + "learning_rate": 1.5560167508313832e-06, + "loss": 0.4485, + "step": 39129 + }, + { + "epoch": 14.45, + "learning_rate": 1.555979800468038e-06, + "loss": 0.2557, + "step": 39130 + }, + { + "epoch": 14.45, + "learning_rate": 1.5559428501046926e-06, + "loss": 0.2762, + "step": 39131 + }, + { + "epoch": 14.45, + "learning_rate": 1.5559058997413473e-06, + "loss": 0.2836, + "step": 39132 + }, + { + "epoch": 14.45, + "learning_rate": 1.5558689493780023e-06, + "loss": 0.3281, + "step": 39133 + }, + { + "epoch": 14.45, + "learning_rate": 1.555831999014657e-06, + "loss": 0.3115, + "step": 39134 + }, + { + "epoch": 14.45, + "learning_rate": 1.5557950486513118e-06, + "loss": 0.2453, + "step": 39135 + }, + { + "epoch": 14.45, + "learning_rate": 1.5557580982879665e-06, + "loss": 0.3373, + "step": 39136 + }, + { + "epoch": 14.45, + "learning_rate": 1.5557211479246213e-06, + "loss": 0.0848, + "step": 39137 + }, + { + "epoch": 14.45, + "learning_rate": 1.555684197561276e-06, + "loss": 0.1472, + "step": 39138 + }, + { + "epoch": 14.45, + "learning_rate": 1.555647247197931e-06, + "loss": 0.2354, + "step": 39139 + }, + { + "epoch": 14.45, + "learning_rate": 1.5556102968345856e-06, + "loss": 0.4215, + "step": 39140 + }, + { + "epoch": 14.45, + "learning_rate": 1.5555733464712405e-06, + "loss": 0.2922, + "step": 39141 + }, + { + "epoch": 14.45, + "learning_rate": 1.5555363961078951e-06, + "loss": 0.2798, + "step": 39142 + }, + { + "epoch": 14.45, + "learning_rate": 1.5554994457445498e-06, + "loss": 0.2375, + "step": 39143 + }, + { + "epoch": 14.45, + "learning_rate": 1.5554624953812046e-06, + "loss": 0.348, + "step": 39144 + }, + { + "epoch": 14.46, + "learning_rate": 1.5554255450178593e-06, + "loss": 0.2692, + "step": 39145 + }, + { + "epoch": 14.46, + "learning_rate": 1.5553885946545143e-06, + "loss": 0.2607, + "step": 39146 + }, + { + "epoch": 14.46, + "learning_rate": 1.5553516442911687e-06, + "loss": 0.1781, + "step": 39147 + }, + { + "epoch": 14.46, + "learning_rate": 1.5553146939278238e-06, + "loss": 0.3084, + "step": 39148 + }, + { + "epoch": 14.46, + "learning_rate": 1.5552777435644784e-06, + "loss": 0.3074, + "step": 39149 + }, + { + "epoch": 14.46, + "learning_rate": 1.5552407932011333e-06, + "loss": 0.2901, + "step": 39150 + }, + { + "epoch": 14.46, + "learning_rate": 1.555203842837788e-06, + "loss": 0.3649, + "step": 39151 + }, + { + "epoch": 14.46, + "learning_rate": 1.5551668924744426e-06, + "loss": 0.3367, + "step": 39152 + }, + { + "epoch": 14.46, + "learning_rate": 1.5551299421110974e-06, + "loss": 0.2929, + "step": 39153 + }, + { + "epoch": 14.46, + "learning_rate": 1.555092991747752e-06, + "loss": 0.3725, + "step": 39154 + }, + { + "epoch": 14.46, + "learning_rate": 1.555056041384407e-06, + "loss": 0.2639, + "step": 39155 + }, + { + "epoch": 14.46, + "learning_rate": 1.5550190910210617e-06, + "loss": 0.2017, + "step": 39156 + }, + { + "epoch": 14.46, + "learning_rate": 1.5549821406577166e-06, + "loss": 0.326, + "step": 39157 + }, + { + "epoch": 14.46, + "learning_rate": 1.5549451902943712e-06, + "loss": 0.2804, + "step": 39158 + }, + { + "epoch": 14.46, + "learning_rate": 1.554908239931026e-06, + "loss": 0.223, + "step": 39159 + }, + { + "epoch": 14.46, + "learning_rate": 1.5548712895676807e-06, + "loss": 0.2746, + "step": 39160 + }, + { + "epoch": 14.46, + "learning_rate": 1.5548343392043358e-06, + "loss": 0.1664, + "step": 39161 + }, + { + "epoch": 14.46, + "learning_rate": 1.5547973888409904e-06, + "loss": 0.3666, + "step": 39162 + }, + { + "epoch": 14.46, + "learning_rate": 1.5547604384776448e-06, + "loss": 0.3262, + "step": 39163 + }, + { + "epoch": 14.46, + "learning_rate": 1.5547234881142999e-06, + "loss": 0.3374, + "step": 39164 + }, + { + "epoch": 14.46, + "learning_rate": 1.5546865377509545e-06, + "loss": 0.2414, + "step": 39165 + }, + { + "epoch": 14.46, + "learning_rate": 1.5546495873876094e-06, + "loss": 0.2671, + "step": 39166 + }, + { + "epoch": 14.46, + "learning_rate": 1.554612637024264e-06, + "loss": 0.2257, + "step": 39167 + }, + { + "epoch": 14.46, + "learning_rate": 1.554575686660919e-06, + "loss": 0.2252, + "step": 39168 + }, + { + "epoch": 14.46, + "learning_rate": 1.5545387362975735e-06, + "loss": 0.2642, + "step": 39169 + }, + { + "epoch": 14.46, + "learning_rate": 1.5545017859342286e-06, + "loss": 0.271, + "step": 39170 + }, + { + "epoch": 14.46, + "learning_rate": 1.5544648355708832e-06, + "loss": 0.2662, + "step": 39171 + }, + { + "epoch": 14.47, + "learning_rate": 1.5544278852075378e-06, + "loss": 0.2314, + "step": 39172 + }, + { + "epoch": 14.47, + "learning_rate": 1.5543909348441927e-06, + "loss": 0.2625, + "step": 39173 + }, + { + "epoch": 14.47, + "learning_rate": 1.5543539844808473e-06, + "loss": 0.2181, + "step": 39174 + }, + { + "epoch": 14.47, + "learning_rate": 1.5543170341175022e-06, + "loss": 0.1326, + "step": 39175 + }, + { + "epoch": 14.47, + "learning_rate": 1.5542800837541568e-06, + "loss": 0.2338, + "step": 39176 + }, + { + "epoch": 14.47, + "learning_rate": 1.5542431333908119e-06, + "loss": 0.2692, + "step": 39177 + }, + { + "epoch": 14.47, + "learning_rate": 1.5542061830274665e-06, + "loss": 0.3542, + "step": 39178 + }, + { + "epoch": 14.47, + "learning_rate": 1.5541692326641213e-06, + "loss": 0.1711, + "step": 39179 + }, + { + "epoch": 14.47, + "learning_rate": 1.554132282300776e-06, + "loss": 0.1937, + "step": 39180 + }, + { + "epoch": 14.47, + "learning_rate": 1.5540953319374308e-06, + "loss": 0.292, + "step": 39181 + }, + { + "epoch": 14.47, + "learning_rate": 1.5540583815740855e-06, + "loss": 0.2819, + "step": 39182 + }, + { + "epoch": 14.47, + "learning_rate": 1.55402143121074e-06, + "loss": 0.1085, + "step": 39183 + }, + { + "epoch": 14.47, + "learning_rate": 1.5539844808473952e-06, + "loss": 0.1705, + "step": 39184 + }, + { + "epoch": 14.47, + "learning_rate": 1.5539475304840498e-06, + "loss": 0.3285, + "step": 39185 + }, + { + "epoch": 14.47, + "learning_rate": 1.5539105801207046e-06, + "loss": 0.2585, + "step": 39186 + }, + { + "epoch": 14.47, + "learning_rate": 1.5538736297573593e-06, + "loss": 0.2192, + "step": 39187 + }, + { + "epoch": 14.47, + "learning_rate": 1.5538366793940141e-06, + "loss": 0.1415, + "step": 39188 + }, + { + "epoch": 14.47, + "learning_rate": 1.5537997290306688e-06, + "loss": 0.2329, + "step": 39189 + }, + { + "epoch": 14.47, + "learning_rate": 1.5537627786673238e-06, + "loss": 0.2889, + "step": 39190 + }, + { + "epoch": 14.47, + "learning_rate": 1.5537258283039785e-06, + "loss": 0.3024, + "step": 39191 + }, + { + "epoch": 14.47, + "learning_rate": 1.5536888779406329e-06, + "loss": 0.3974, + "step": 39192 + }, + { + "epoch": 14.47, + "learning_rate": 1.553651927577288e-06, + "loss": 0.261, + "step": 39193 + }, + { + "epoch": 14.47, + "learning_rate": 1.5536149772139426e-06, + "loss": 0.3512, + "step": 39194 + }, + { + "epoch": 14.47, + "learning_rate": 1.5535780268505974e-06, + "loss": 0.249, + "step": 39195 + }, + { + "epoch": 14.47, + "learning_rate": 1.553541076487252e-06, + "loss": 0.223, + "step": 39196 + }, + { + "epoch": 14.47, + "learning_rate": 1.5535041261239071e-06, + "loss": 0.4337, + "step": 39197 + }, + { + "epoch": 14.47, + "learning_rate": 1.5534671757605615e-06, + "loss": 0.3287, + "step": 39198 + }, + { + "epoch": 14.48, + "learning_rate": 1.5534302253972166e-06, + "loss": 0.2491, + "step": 39199 + }, + { + "epoch": 14.48, + "learning_rate": 1.5533932750338712e-06, + "loss": 0.2397, + "step": 39200 + }, + { + "epoch": 14.48, + "learning_rate": 1.553356324670526e-06, + "loss": 0.2609, + "step": 39201 + }, + { + "epoch": 14.48, + "learning_rate": 1.5533193743071807e-06, + "loss": 0.3357, + "step": 39202 + }, + { + "epoch": 14.48, + "learning_rate": 1.5532824239438354e-06, + "loss": 0.2565, + "step": 39203 + }, + { + "epoch": 14.48, + "learning_rate": 1.5532454735804902e-06, + "loss": 0.2798, + "step": 39204 + }, + { + "epoch": 14.48, + "learning_rate": 1.5532085232171448e-06, + "loss": 0.2695, + "step": 39205 + }, + { + "epoch": 14.48, + "learning_rate": 1.5531715728538e-06, + "loss": 0.2809, + "step": 39206 + }, + { + "epoch": 14.48, + "learning_rate": 1.5531346224904545e-06, + "loss": 0.2052, + "step": 39207 + }, + { + "epoch": 14.48, + "learning_rate": 1.5530976721271094e-06, + "loss": 0.1775, + "step": 39208 + }, + { + "epoch": 14.48, + "learning_rate": 1.553060721763764e-06, + "loss": 0.2524, + "step": 39209 + }, + { + "epoch": 14.48, + "learning_rate": 1.5530237714004189e-06, + "loss": 0.3016, + "step": 39210 + }, + { + "epoch": 14.48, + "learning_rate": 1.5529868210370735e-06, + "loss": 0.3374, + "step": 39211 + }, + { + "epoch": 14.48, + "learning_rate": 1.5529498706737282e-06, + "loss": 0.347, + "step": 39212 + }, + { + "epoch": 14.48, + "learning_rate": 1.5529129203103832e-06, + "loss": 0.2188, + "step": 39213 + }, + { + "epoch": 14.48, + "learning_rate": 1.5528759699470376e-06, + "loss": 0.2742, + "step": 39214 + }, + { + "epoch": 14.48, + "learning_rate": 1.5528390195836927e-06, + "loss": 0.1727, + "step": 39215 + }, + { + "epoch": 14.48, + "learning_rate": 1.5528020692203473e-06, + "loss": 0.3609, + "step": 39216 + }, + { + "epoch": 14.48, + "learning_rate": 1.5527651188570022e-06, + "loss": 0.2741, + "step": 39217 + }, + { + "epoch": 14.48, + "learning_rate": 1.5527281684936568e-06, + "loss": 0.2444, + "step": 39218 + }, + { + "epoch": 14.48, + "learning_rate": 1.5526912181303119e-06, + "loss": 0.2454, + "step": 39219 + }, + { + "epoch": 14.48, + "learning_rate": 1.5526542677669663e-06, + "loss": 0.2885, + "step": 39220 + }, + { + "epoch": 14.48, + "learning_rate": 1.5526173174036214e-06, + "loss": 0.2803, + "step": 39221 + }, + { + "epoch": 14.48, + "learning_rate": 1.552580367040276e-06, + "loss": 0.156, + "step": 39222 + }, + { + "epoch": 14.48, + "learning_rate": 1.5525434166769306e-06, + "loss": 0.2021, + "step": 39223 + }, + { + "epoch": 14.48, + "learning_rate": 1.5525064663135855e-06, + "loss": 0.3419, + "step": 39224 + }, + { + "epoch": 14.48, + "learning_rate": 1.5524695159502401e-06, + "loss": 0.231, + "step": 39225 + }, + { + "epoch": 14.49, + "learning_rate": 1.552432565586895e-06, + "loss": 0.2051, + "step": 39226 + }, + { + "epoch": 14.49, + "learning_rate": 1.5523956152235496e-06, + "loss": 0.4742, + "step": 39227 + }, + { + "epoch": 14.49, + "learning_rate": 1.5523586648602047e-06, + "loss": 0.3304, + "step": 39228 + }, + { + "epoch": 14.49, + "learning_rate": 1.5523217144968593e-06, + "loss": 0.3757, + "step": 39229 + }, + { + "epoch": 14.49, + "learning_rate": 1.5522847641335141e-06, + "loss": 0.3104, + "step": 39230 + }, + { + "epoch": 14.49, + "learning_rate": 1.5522478137701688e-06, + "loss": 0.2478, + "step": 39231 + }, + { + "epoch": 14.49, + "learning_rate": 1.5522108634068234e-06, + "loss": 0.276, + "step": 39232 + }, + { + "epoch": 14.49, + "learning_rate": 1.5521739130434783e-06, + "loss": 0.3399, + "step": 39233 + }, + { + "epoch": 14.49, + "learning_rate": 1.552136962680133e-06, + "loss": 0.2553, + "step": 39234 + }, + { + "epoch": 14.49, + "learning_rate": 1.552100012316788e-06, + "loss": 0.2458, + "step": 39235 + }, + { + "epoch": 14.49, + "learning_rate": 1.5520630619534426e-06, + "loss": 0.2795, + "step": 39236 + }, + { + "epoch": 14.49, + "learning_rate": 1.5520261115900974e-06, + "loss": 0.3314, + "step": 39237 + }, + { + "epoch": 14.49, + "learning_rate": 1.551989161226752e-06, + "loss": 0.4144, + "step": 39238 + }, + { + "epoch": 14.49, + "learning_rate": 1.551952210863407e-06, + "loss": 0.3954, + "step": 39239 + }, + { + "epoch": 14.49, + "learning_rate": 1.5519152605000616e-06, + "loss": 0.2433, + "step": 39240 + }, + { + "epoch": 14.49, + "learning_rate": 1.5518783101367166e-06, + "loss": 0.3621, + "step": 39241 + }, + { + "epoch": 14.49, + "learning_rate": 1.551841359773371e-06, + "loss": 0.24, + "step": 39242 + }, + { + "epoch": 14.49, + "learning_rate": 1.5518044094100257e-06, + "loss": 0.2736, + "step": 39243 + }, + { + "epoch": 14.49, + "learning_rate": 1.5517674590466808e-06, + "loss": 0.1986, + "step": 39244 + }, + { + "epoch": 14.49, + "learning_rate": 1.5517305086833354e-06, + "loss": 0.3189, + "step": 39245 + }, + { + "epoch": 14.49, + "learning_rate": 1.5516935583199902e-06, + "loss": 0.1762, + "step": 39246 + }, + { + "epoch": 14.49, + "learning_rate": 1.5516566079566449e-06, + "loss": 0.1702, + "step": 39247 + }, + { + "epoch": 14.49, + "learning_rate": 1.5516196575932997e-06, + "loss": 0.3718, + "step": 39248 + }, + { + "epoch": 14.49, + "learning_rate": 1.5515827072299544e-06, + "loss": 0.3373, + "step": 39249 + }, + { + "epoch": 14.49, + "learning_rate": 1.5515457568666094e-06, + "loss": 0.2695, + "step": 39250 + }, + { + "epoch": 14.49, + "learning_rate": 1.551508806503264e-06, + "loss": 0.2112, + "step": 39251 + }, + { + "epoch": 14.49, + "learning_rate": 1.5514718561399187e-06, + "loss": 0.2805, + "step": 39252 + }, + { + "epoch": 14.5, + "learning_rate": 1.5514349057765735e-06, + "loss": 0.186, + "step": 39253 + }, + { + "epoch": 14.5, + "learning_rate": 1.5513979554132282e-06, + "loss": 0.2315, + "step": 39254 + }, + { + "epoch": 14.5, + "learning_rate": 1.551361005049883e-06, + "loss": 0.3094, + "step": 39255 + }, + { + "epoch": 14.5, + "learning_rate": 1.5513240546865377e-06, + "loss": 0.2483, + "step": 39256 + }, + { + "epoch": 14.5, + "learning_rate": 1.5512871043231927e-06, + "loss": 0.2155, + "step": 39257 + }, + { + "epoch": 14.5, + "learning_rate": 1.5512501539598474e-06, + "loss": 0.2325, + "step": 39258 + }, + { + "epoch": 14.5, + "learning_rate": 1.5512132035965022e-06, + "loss": 0.3049, + "step": 39259 + }, + { + "epoch": 14.5, + "learning_rate": 1.5511762532331568e-06, + "loss": 0.2059, + "step": 39260 + }, + { + "epoch": 14.5, + "learning_rate": 1.5511393028698115e-06, + "loss": 0.2301, + "step": 39261 + }, + { + "epoch": 14.5, + "learning_rate": 1.5511023525064663e-06, + "loss": 0.356, + "step": 39262 + }, + { + "epoch": 14.5, + "learning_rate": 1.551065402143121e-06, + "loss": 0.2987, + "step": 39263 + }, + { + "epoch": 14.5, + "learning_rate": 1.551028451779776e-06, + "loss": 0.2556, + "step": 39264 + }, + { + "epoch": 14.5, + "learning_rate": 1.5509915014164304e-06, + "loss": 0.2037, + "step": 39265 + }, + { + "epoch": 14.5, + "learning_rate": 1.5509545510530855e-06, + "loss": 0.3202, + "step": 39266 + }, + { + "epoch": 14.5, + "learning_rate": 1.5509176006897401e-06, + "loss": 0.163, + "step": 39267 + }, + { + "epoch": 14.5, + "learning_rate": 1.550880650326395e-06, + "loss": 0.2266, + "step": 39268 + }, + { + "epoch": 14.5, + "learning_rate": 1.5508436999630496e-06, + "loss": 0.2255, + "step": 39269 + }, + { + "epoch": 14.5, + "learning_rate": 1.5508067495997047e-06, + "loss": 0.2942, + "step": 39270 + }, + { + "epoch": 14.5, + "learning_rate": 1.5507697992363591e-06, + "loss": 0.3031, + "step": 39271 + }, + { + "epoch": 14.5, + "learning_rate": 1.5507328488730137e-06, + "loss": 0.342, + "step": 39272 + }, + { + "epoch": 14.5, + "learning_rate": 1.5506958985096688e-06, + "loss": 0.2542, + "step": 39273 + }, + { + "epoch": 14.5, + "learning_rate": 1.5506589481463234e-06, + "loss": 0.2034, + "step": 39274 + }, + { + "epoch": 14.5, + "learning_rate": 1.5506219977829783e-06, + "loss": 0.2547, + "step": 39275 + }, + { + "epoch": 14.5, + "learning_rate": 1.550585047419633e-06, + "loss": 0.2507, + "step": 39276 + }, + { + "epoch": 14.5, + "learning_rate": 1.5505480970562878e-06, + "loss": 0.3092, + "step": 39277 + }, + { + "epoch": 14.5, + "learning_rate": 1.5505111466929424e-06, + "loss": 0.424, + "step": 39278 + }, + { + "epoch": 14.5, + "learning_rate": 1.5504741963295975e-06, + "loss": 0.2532, + "step": 39279 + }, + { + "epoch": 14.51, + "learning_rate": 1.5504372459662521e-06, + "loss": 0.2534, + "step": 39280 + }, + { + "epoch": 14.51, + "learning_rate": 1.5504002956029065e-06, + "loss": 0.3073, + "step": 39281 + }, + { + "epoch": 14.51, + "learning_rate": 1.5503633452395616e-06, + "loss": 0.3134, + "step": 39282 + }, + { + "epoch": 14.51, + "learning_rate": 1.5503263948762162e-06, + "loss": 0.325, + "step": 39283 + }, + { + "epoch": 14.51, + "learning_rate": 1.550289444512871e-06, + "loss": 0.2672, + "step": 39284 + }, + { + "epoch": 14.51, + "learning_rate": 1.5502524941495257e-06, + "loss": 0.2768, + "step": 39285 + }, + { + "epoch": 14.51, + "learning_rate": 1.5502155437861808e-06, + "loss": 0.138, + "step": 39286 + }, + { + "epoch": 14.51, + "learning_rate": 1.5501785934228352e-06, + "loss": 0.2524, + "step": 39287 + }, + { + "epoch": 14.51, + "learning_rate": 1.5501416430594903e-06, + "loss": 0.3316, + "step": 39288 + }, + { + "epoch": 14.51, + "learning_rate": 1.550104692696145e-06, + "loss": 0.2526, + "step": 39289 + }, + { + "epoch": 14.51, + "learning_rate": 1.5500677423327997e-06, + "loss": 0.255, + "step": 39290 + }, + { + "epoch": 14.51, + "learning_rate": 1.5500307919694544e-06, + "loss": 0.2015, + "step": 39291 + }, + { + "epoch": 14.51, + "learning_rate": 1.549993841606109e-06, + "loss": 0.2221, + "step": 39292 + }, + { + "epoch": 14.51, + "learning_rate": 1.5499568912427639e-06, + "loss": 0.1987, + "step": 39293 + }, + { + "epoch": 14.51, + "learning_rate": 1.5499199408794185e-06, + "loss": 0.328, + "step": 39294 + }, + { + "epoch": 14.51, + "learning_rate": 1.5498829905160736e-06, + "loss": 0.1857, + "step": 39295 + }, + { + "epoch": 14.51, + "learning_rate": 1.5498460401527282e-06, + "loss": 0.2195, + "step": 39296 + }, + { + "epoch": 14.51, + "learning_rate": 1.549809089789383e-06, + "loss": 0.2261, + "step": 39297 + }, + { + "epoch": 14.51, + "learning_rate": 1.5497721394260377e-06, + "loss": 0.2303, + "step": 39298 + }, + { + "epoch": 14.51, + "learning_rate": 1.5497351890626925e-06, + "loss": 0.2168, + "step": 39299 + }, + { + "epoch": 14.51, + "learning_rate": 1.5496982386993472e-06, + "loss": 0.3659, + "step": 39300 + }, + { + "epoch": 14.51, + "learning_rate": 1.5496612883360018e-06, + "loss": 0.2195, + "step": 39301 + }, + { + "epoch": 14.51, + "learning_rate": 1.5496243379726569e-06, + "loss": 0.1528, + "step": 39302 + }, + { + "epoch": 14.51, + "learning_rate": 1.5495873876093115e-06, + "loss": 0.1443, + "step": 39303 + }, + { + "epoch": 14.51, + "learning_rate": 1.5495504372459663e-06, + "loss": 0.3021, + "step": 39304 + }, + { + "epoch": 14.51, + "learning_rate": 1.549513486882621e-06, + "loss": 0.3479, + "step": 39305 + }, + { + "epoch": 14.51, + "learning_rate": 1.5494765365192758e-06, + "loss": 0.1811, + "step": 39306 + }, + { + "epoch": 14.52, + "learning_rate": 1.5494395861559305e-06, + "loss": 0.1971, + "step": 39307 + }, + { + "epoch": 14.52, + "learning_rate": 1.5494026357925855e-06, + "loss": 0.2522, + "step": 39308 + }, + { + "epoch": 14.52, + "learning_rate": 1.5493656854292402e-06, + "loss": 0.3161, + "step": 39309 + }, + { + "epoch": 14.52, + "learning_rate": 1.549328735065895e-06, + "loss": 0.182, + "step": 39310 + }, + { + "epoch": 14.52, + "learning_rate": 1.5492917847025497e-06, + "loss": 0.2331, + "step": 39311 + }, + { + "epoch": 14.52, + "learning_rate": 1.5492548343392043e-06, + "loss": 0.391, + "step": 39312 + }, + { + "epoch": 14.52, + "learning_rate": 1.5492178839758591e-06, + "loss": 0.2803, + "step": 39313 + }, + { + "epoch": 14.52, + "learning_rate": 1.5491809336125138e-06, + "loss": 0.1916, + "step": 39314 + }, + { + "epoch": 14.52, + "learning_rate": 1.5491439832491688e-06, + "loss": 0.2336, + "step": 39315 + }, + { + "epoch": 14.52, + "learning_rate": 1.5491070328858233e-06, + "loss": 0.2025, + "step": 39316 + }, + { + "epoch": 14.52, + "learning_rate": 1.5490700825224783e-06, + "loss": 0.2793, + "step": 39317 + }, + { + "epoch": 14.52, + "learning_rate": 1.549033132159133e-06, + "loss": 0.4739, + "step": 39318 + }, + { + "epoch": 14.52, + "learning_rate": 1.5489961817957878e-06, + "loss": 0.1615, + "step": 39319 + }, + { + "epoch": 14.52, + "learning_rate": 1.5489592314324424e-06, + "loss": 0.2762, + "step": 39320 + }, + { + "epoch": 14.52, + "learning_rate": 1.548922281069097e-06, + "loss": 0.2978, + "step": 39321 + }, + { + "epoch": 14.52, + "learning_rate": 1.548885330705752e-06, + "loss": 0.2778, + "step": 39322 + }, + { + "epoch": 14.52, + "learning_rate": 1.5488483803424066e-06, + "loss": 0.144, + "step": 39323 + }, + { + "epoch": 14.52, + "learning_rate": 1.5488114299790616e-06, + "loss": 0.3432, + "step": 39324 + }, + { + "epoch": 14.52, + "learning_rate": 1.5487744796157163e-06, + "loss": 0.3542, + "step": 39325 + }, + { + "epoch": 14.52, + "learning_rate": 1.548737529252371e-06, + "loss": 0.3368, + "step": 39326 + }, + { + "epoch": 14.52, + "learning_rate": 1.5487005788890257e-06, + "loss": 0.3633, + "step": 39327 + }, + { + "epoch": 14.52, + "learning_rate": 1.5486636285256806e-06, + "loss": 0.3271, + "step": 39328 + }, + { + "epoch": 14.52, + "learning_rate": 1.5486266781623352e-06, + "loss": 0.2659, + "step": 39329 + }, + { + "epoch": 14.52, + "learning_rate": 1.5485897277989903e-06, + "loss": 0.2488, + "step": 39330 + }, + { + "epoch": 14.52, + "learning_rate": 1.548552777435645e-06, + "loss": 0.3329, + "step": 39331 + }, + { + "epoch": 14.52, + "learning_rate": 1.5485158270722993e-06, + "loss": 0.3099, + "step": 39332 + }, + { + "epoch": 14.52, + "learning_rate": 1.5484788767089544e-06, + "loss": 0.3261, + "step": 39333 + }, + { + "epoch": 14.53, + "learning_rate": 1.548441926345609e-06, + "loss": 0.2237, + "step": 39334 + }, + { + "epoch": 14.53, + "learning_rate": 1.5484049759822639e-06, + "loss": 0.1931, + "step": 39335 + }, + { + "epoch": 14.53, + "learning_rate": 1.5483680256189185e-06, + "loss": 0.3782, + "step": 39336 + }, + { + "epoch": 14.53, + "learning_rate": 1.5483310752555736e-06, + "loss": 0.2969, + "step": 39337 + }, + { + "epoch": 14.53, + "learning_rate": 1.548294124892228e-06, + "loss": 0.3191, + "step": 39338 + }, + { + "epoch": 14.53, + "learning_rate": 1.548257174528883e-06, + "loss": 0.2067, + "step": 39339 + }, + { + "epoch": 14.53, + "learning_rate": 1.5482202241655377e-06, + "loss": 0.2636, + "step": 39340 + }, + { + "epoch": 14.53, + "learning_rate": 1.5481832738021923e-06, + "loss": 0.1603, + "step": 39341 + }, + { + "epoch": 14.53, + "learning_rate": 1.5481463234388472e-06, + "loss": 0.2911, + "step": 39342 + }, + { + "epoch": 14.53, + "learning_rate": 1.5481093730755018e-06, + "loss": 0.2474, + "step": 39343 + }, + { + "epoch": 14.53, + "learning_rate": 1.5480724227121567e-06, + "loss": 0.1586, + "step": 39344 + }, + { + "epoch": 14.53, + "learning_rate": 1.5480354723488113e-06, + "loss": 0.1738, + "step": 39345 + }, + { + "epoch": 14.53, + "learning_rate": 1.5479985219854664e-06, + "loss": 0.148, + "step": 39346 + }, + { + "epoch": 14.53, + "learning_rate": 1.547961571622121e-06, + "loss": 0.2452, + "step": 39347 + }, + { + "epoch": 14.53, + "learning_rate": 1.5479246212587759e-06, + "loss": 0.2197, + "step": 39348 + }, + { + "epoch": 14.53, + "learning_rate": 1.5478876708954305e-06, + "loss": 0.2411, + "step": 39349 + }, + { + "epoch": 14.53, + "learning_rate": 1.5478507205320853e-06, + "loss": 0.3217, + "step": 39350 + }, + { + "epoch": 14.53, + "learning_rate": 1.54781377016874e-06, + "loss": 0.3031, + "step": 39351 + }, + { + "epoch": 14.53, + "learning_rate": 1.5477768198053946e-06, + "loss": 0.3454, + "step": 39352 + }, + { + "epoch": 14.53, + "learning_rate": 1.5477398694420497e-06, + "loss": 0.2369, + "step": 39353 + }, + { + "epoch": 14.53, + "learning_rate": 1.547702919078704e-06, + "loss": 0.1804, + "step": 39354 + }, + { + "epoch": 14.53, + "learning_rate": 1.5476659687153592e-06, + "loss": 0.2503, + "step": 39355 + }, + { + "epoch": 14.53, + "learning_rate": 1.5476290183520138e-06, + "loss": 0.2252, + "step": 39356 + }, + { + "epoch": 14.53, + "learning_rate": 1.5475920679886686e-06, + "loss": 0.2798, + "step": 39357 + }, + { + "epoch": 14.53, + "learning_rate": 1.5475551176253233e-06, + "loss": 0.2521, + "step": 39358 + }, + { + "epoch": 14.53, + "learning_rate": 1.5475181672619783e-06, + "loss": 0.1957, + "step": 39359 + }, + { + "epoch": 14.53, + "learning_rate": 1.5474812168986328e-06, + "loss": 0.2791, + "step": 39360 + }, + { + "epoch": 14.54, + "learning_rate": 1.5474442665352874e-06, + "loss": 0.3109, + "step": 39361 + }, + { + "epoch": 14.54, + "learning_rate": 1.5474073161719425e-06, + "loss": 0.1975, + "step": 39362 + }, + { + "epoch": 14.54, + "learning_rate": 1.547370365808597e-06, + "loss": 0.3938, + "step": 39363 + }, + { + "epoch": 14.54, + "learning_rate": 1.547333415445252e-06, + "loss": 0.2943, + "step": 39364 + }, + { + "epoch": 14.54, + "learning_rate": 1.5472964650819066e-06, + "loss": 0.2932, + "step": 39365 + }, + { + "epoch": 14.54, + "learning_rate": 1.5472595147185614e-06, + "loss": 0.3487, + "step": 39366 + }, + { + "epoch": 14.54, + "learning_rate": 1.547222564355216e-06, + "loss": 0.2786, + "step": 39367 + }, + { + "epoch": 14.54, + "learning_rate": 1.5471856139918711e-06, + "loss": 0.2651, + "step": 39368 + }, + { + "epoch": 14.54, + "learning_rate": 1.5471486636285258e-06, + "loss": 0.379, + "step": 39369 + }, + { + "epoch": 14.54, + "learning_rate": 1.5471117132651806e-06, + "loss": 0.1754, + "step": 39370 + }, + { + "epoch": 14.54, + "learning_rate": 1.5470747629018352e-06, + "loss": 0.3166, + "step": 39371 + }, + { + "epoch": 14.54, + "learning_rate": 1.5470378125384899e-06, + "loss": 0.2658, + "step": 39372 + }, + { + "epoch": 14.54, + "learning_rate": 1.5470008621751447e-06, + "loss": 0.1968, + "step": 39373 + }, + { + "epoch": 14.54, + "learning_rate": 1.5469639118117994e-06, + "loss": 0.195, + "step": 39374 + }, + { + "epoch": 14.54, + "learning_rate": 1.5469269614484544e-06, + "loss": 0.1789, + "step": 39375 + }, + { + "epoch": 14.54, + "learning_rate": 1.546890011085109e-06, + "loss": 0.2074, + "step": 39376 + }, + { + "epoch": 14.54, + "learning_rate": 1.546853060721764e-06, + "loss": 0.2847, + "step": 39377 + }, + { + "epoch": 14.54, + "learning_rate": 1.5468161103584186e-06, + "loss": 0.3666, + "step": 39378 + }, + { + "epoch": 14.54, + "learning_rate": 1.5467791599950734e-06, + "loss": 0.3234, + "step": 39379 + }, + { + "epoch": 14.54, + "learning_rate": 1.546742209631728e-06, + "loss": 0.3553, + "step": 39380 + }, + { + "epoch": 14.54, + "learning_rate": 1.5467052592683827e-06, + "loss": 0.3783, + "step": 39381 + }, + { + "epoch": 14.54, + "learning_rate": 1.5466683089050377e-06, + "loss": 0.3617, + "step": 39382 + }, + { + "epoch": 14.54, + "learning_rate": 1.5466313585416922e-06, + "loss": 0.2675, + "step": 39383 + }, + { + "epoch": 14.54, + "learning_rate": 1.5465944081783472e-06, + "loss": 0.4737, + "step": 39384 + }, + { + "epoch": 14.54, + "learning_rate": 1.5465574578150019e-06, + "loss": 0.4147, + "step": 39385 + }, + { + "epoch": 14.54, + "learning_rate": 1.5465205074516567e-06, + "loss": 0.2759, + "step": 39386 + }, + { + "epoch": 14.54, + "learning_rate": 1.5464835570883113e-06, + "loss": 0.2678, + "step": 39387 + }, + { + "epoch": 14.55, + "learning_rate": 1.5464466067249664e-06, + "loss": 0.2348, + "step": 39388 + }, + { + "epoch": 14.55, + "learning_rate": 1.5464096563616208e-06, + "loss": 0.2547, + "step": 39389 + }, + { + "epoch": 14.55, + "learning_rate": 1.5463727059982759e-06, + "loss": 0.3032, + "step": 39390 + }, + { + "epoch": 14.55, + "learning_rate": 1.5463357556349305e-06, + "loss": 0.3686, + "step": 39391 + }, + { + "epoch": 14.55, + "learning_rate": 1.5462988052715852e-06, + "loss": 0.2128, + "step": 39392 + }, + { + "epoch": 14.55, + "learning_rate": 1.54626185490824e-06, + "loss": 0.2683, + "step": 39393 + }, + { + "epoch": 14.55, + "learning_rate": 1.5462249045448946e-06, + "loss": 0.3584, + "step": 39394 + }, + { + "epoch": 14.55, + "learning_rate": 1.5461879541815495e-06, + "loss": 0.1808, + "step": 39395 + }, + { + "epoch": 14.55, + "learning_rate": 1.5461510038182041e-06, + "loss": 0.3284, + "step": 39396 + }, + { + "epoch": 14.55, + "learning_rate": 1.5461140534548592e-06, + "loss": 0.3, + "step": 39397 + }, + { + "epoch": 14.55, + "learning_rate": 1.5460771030915138e-06, + "loss": 0.3183, + "step": 39398 + }, + { + "epoch": 14.55, + "learning_rate": 1.5460401527281687e-06, + "loss": 0.3481, + "step": 39399 + }, + { + "epoch": 14.55, + "learning_rate": 1.5460032023648233e-06, + "loss": 0.2659, + "step": 39400 + }, + { + "epoch": 14.55, + "learning_rate": 1.545966252001478e-06, + "loss": 0.2478, + "step": 39401 + }, + { + "epoch": 14.55, + "learning_rate": 1.5459293016381328e-06, + "loss": 0.2342, + "step": 39402 + }, + { + "epoch": 14.55, + "learning_rate": 1.5458923512747874e-06, + "loss": 0.2562, + "step": 39403 + }, + { + "epoch": 14.55, + "learning_rate": 1.5458554009114425e-06, + "loss": 0.1179, + "step": 39404 + }, + { + "epoch": 14.55, + "learning_rate": 1.545818450548097e-06, + "loss": 0.26, + "step": 39405 + }, + { + "epoch": 14.55, + "learning_rate": 1.545781500184752e-06, + "loss": 0.2032, + "step": 39406 + }, + { + "epoch": 14.55, + "learning_rate": 1.5457445498214066e-06, + "loss": 0.3161, + "step": 39407 + }, + { + "epoch": 14.55, + "learning_rate": 1.5457075994580615e-06, + "loss": 0.2974, + "step": 39408 + }, + { + "epoch": 14.55, + "learning_rate": 1.545670649094716e-06, + "loss": 0.2641, + "step": 39409 + }, + { + "epoch": 14.55, + "learning_rate": 1.5456336987313712e-06, + "loss": 0.3421, + "step": 39410 + }, + { + "epoch": 14.55, + "learning_rate": 1.5455967483680256e-06, + "loss": 0.1889, + "step": 39411 + }, + { + "epoch": 14.55, + "learning_rate": 1.5455597980046802e-06, + "loss": 0.2983, + "step": 39412 + }, + { + "epoch": 14.55, + "learning_rate": 1.5455228476413353e-06, + "loss": 0.2726, + "step": 39413 + }, + { + "epoch": 14.55, + "learning_rate": 1.54548589727799e-06, + "loss": 0.3428, + "step": 39414 + }, + { + "epoch": 14.56, + "learning_rate": 1.5454489469146448e-06, + "loss": 0.297, + "step": 39415 + }, + { + "epoch": 14.56, + "learning_rate": 1.5454119965512994e-06, + "loss": 0.3806, + "step": 39416 + }, + { + "epoch": 14.56, + "learning_rate": 1.5453750461879542e-06, + "loss": 0.3213, + "step": 39417 + }, + { + "epoch": 14.56, + "learning_rate": 1.5453380958246089e-06, + "loss": 0.2385, + "step": 39418 + }, + { + "epoch": 14.56, + "learning_rate": 1.545301145461264e-06, + "loss": 0.3038, + "step": 39419 + }, + { + "epoch": 14.56, + "learning_rate": 1.5452641950979186e-06, + "loss": 0.2536, + "step": 39420 + }, + { + "epoch": 14.56, + "learning_rate": 1.5452272447345732e-06, + "loss": 0.2912, + "step": 39421 + }, + { + "epoch": 14.56, + "learning_rate": 1.545190294371228e-06, + "loss": 0.2478, + "step": 39422 + }, + { + "epoch": 14.56, + "learning_rate": 1.5451533440078827e-06, + "loss": 0.1937, + "step": 39423 + }, + { + "epoch": 14.56, + "learning_rate": 1.5451163936445375e-06, + "loss": 0.2751, + "step": 39424 + }, + { + "epoch": 14.56, + "learning_rate": 1.5450794432811922e-06, + "loss": 0.3571, + "step": 39425 + }, + { + "epoch": 14.56, + "learning_rate": 1.5450424929178472e-06, + "loss": 0.375, + "step": 39426 + }, + { + "epoch": 14.56, + "learning_rate": 1.5450055425545017e-06, + "loss": 0.1702, + "step": 39427 + }, + { + "epoch": 14.56, + "learning_rate": 1.5449685921911567e-06, + "loss": 0.2454, + "step": 39428 + }, + { + "epoch": 14.56, + "learning_rate": 1.5449316418278114e-06, + "loss": 0.1946, + "step": 39429 + }, + { + "epoch": 14.56, + "learning_rate": 1.5448946914644662e-06, + "loss": 0.2952, + "step": 39430 + }, + { + "epoch": 14.56, + "learning_rate": 1.5448577411011208e-06, + "loss": 0.193, + "step": 39431 + }, + { + "epoch": 14.56, + "learning_rate": 1.5448207907377755e-06, + "loss": 0.3054, + "step": 39432 + }, + { + "epoch": 14.56, + "learning_rate": 1.5447838403744303e-06, + "loss": 0.3496, + "step": 39433 + }, + { + "epoch": 14.56, + "learning_rate": 1.544746890011085e-06, + "loss": 0.2542, + "step": 39434 + }, + { + "epoch": 14.56, + "learning_rate": 1.54470993964774e-06, + "loss": 0.186, + "step": 39435 + }, + { + "epoch": 14.56, + "learning_rate": 1.5446729892843947e-06, + "loss": 0.2457, + "step": 39436 + }, + { + "epoch": 14.56, + "learning_rate": 1.5446360389210495e-06, + "loss": 0.2537, + "step": 39437 + }, + { + "epoch": 14.56, + "learning_rate": 1.5445990885577041e-06, + "loss": 0.3829, + "step": 39438 + }, + { + "epoch": 14.56, + "learning_rate": 1.544562138194359e-06, + "loss": 0.2507, + "step": 39439 + }, + { + "epoch": 14.56, + "learning_rate": 1.5445251878310136e-06, + "loss": 0.2874, + "step": 39440 + }, + { + "epoch": 14.56, + "learning_rate": 1.5444882374676683e-06, + "loss": 0.2881, + "step": 39441 + }, + { + "epoch": 14.56, + "learning_rate": 1.5444512871043233e-06, + "loss": 0.1957, + "step": 39442 + }, + { + "epoch": 14.57, + "learning_rate": 1.544414336740978e-06, + "loss": 0.3097, + "step": 39443 + }, + { + "epoch": 14.57, + "learning_rate": 1.5443773863776328e-06, + "loss": 0.453, + "step": 39444 + }, + { + "epoch": 14.57, + "learning_rate": 1.5443404360142875e-06, + "loss": 0.3061, + "step": 39445 + }, + { + "epoch": 14.57, + "learning_rate": 1.5443034856509423e-06, + "loss": 0.2768, + "step": 39446 + }, + { + "epoch": 14.57, + "learning_rate": 1.544266535287597e-06, + "loss": 0.2523, + "step": 39447 + }, + { + "epoch": 14.57, + "learning_rate": 1.544229584924252e-06, + "loss": 0.306, + "step": 39448 + }, + { + "epoch": 14.57, + "learning_rate": 1.5441926345609066e-06, + "loss": 0.1874, + "step": 39449 + }, + { + "epoch": 14.57, + "learning_rate": 1.5441556841975615e-06, + "loss": 0.359, + "step": 39450 + }, + { + "epoch": 14.57, + "learning_rate": 1.5441187338342161e-06, + "loss": 0.211, + "step": 39451 + }, + { + "epoch": 14.57, + "learning_rate": 1.5440817834708708e-06, + "loss": 0.2634, + "step": 39452 + }, + { + "epoch": 14.57, + "learning_rate": 1.5440448331075256e-06, + "loss": 0.1633, + "step": 39453 + }, + { + "epoch": 14.57, + "learning_rate": 1.5440078827441802e-06, + "loss": 0.2849, + "step": 39454 + }, + { + "epoch": 14.57, + "learning_rate": 1.5439709323808353e-06, + "loss": 0.3753, + "step": 39455 + }, + { + "epoch": 14.57, + "learning_rate": 1.5439339820174897e-06, + "loss": 0.2113, + "step": 39456 + }, + { + "epoch": 14.57, + "learning_rate": 1.5438970316541448e-06, + "loss": 0.1546, + "step": 39457 + }, + { + "epoch": 14.57, + "learning_rate": 1.5438600812907994e-06, + "loss": 0.38, + "step": 39458 + }, + { + "epoch": 14.57, + "learning_rate": 1.5438231309274543e-06, + "loss": 0.1617, + "step": 39459 + }, + { + "epoch": 14.57, + "learning_rate": 1.543786180564109e-06, + "loss": 0.271, + "step": 39460 + }, + { + "epoch": 14.57, + "learning_rate": 1.5437492302007635e-06, + "loss": 0.3494, + "step": 39461 + }, + { + "epoch": 14.57, + "learning_rate": 1.5437122798374184e-06, + "loss": 0.2756, + "step": 39462 + }, + { + "epoch": 14.57, + "learning_rate": 1.543675329474073e-06, + "loss": 0.1736, + "step": 39463 + }, + { + "epoch": 14.57, + "learning_rate": 1.543638379110728e-06, + "loss": 0.2658, + "step": 39464 + }, + { + "epoch": 14.57, + "learning_rate": 1.5436014287473827e-06, + "loss": 0.371, + "step": 39465 + }, + { + "epoch": 14.57, + "learning_rate": 1.5435644783840376e-06, + "loss": 0.158, + "step": 39466 + }, + { + "epoch": 14.57, + "learning_rate": 1.5435275280206922e-06, + "loss": 0.2723, + "step": 39467 + }, + { + "epoch": 14.57, + "learning_rate": 1.543490577657347e-06, + "loss": 0.3935, + "step": 39468 + }, + { + "epoch": 14.57, + "learning_rate": 1.5434536272940017e-06, + "loss": 0.191, + "step": 39469 + }, + { + "epoch": 14.58, + "learning_rate": 1.5434166769306567e-06, + "loss": 0.2584, + "step": 39470 + }, + { + "epoch": 14.58, + "learning_rate": 1.5433797265673114e-06, + "loss": 0.1872, + "step": 39471 + }, + { + "epoch": 14.58, + "learning_rate": 1.5433427762039658e-06, + "loss": 0.2745, + "step": 39472 + }, + { + "epoch": 14.58, + "learning_rate": 1.5433058258406209e-06, + "loss": 0.2715, + "step": 39473 + }, + { + "epoch": 14.58, + "learning_rate": 1.5432688754772755e-06, + "loss": 0.2507, + "step": 39474 + }, + { + "epoch": 14.58, + "learning_rate": 1.5432319251139304e-06, + "loss": 0.2601, + "step": 39475 + }, + { + "epoch": 14.58, + "learning_rate": 1.543194974750585e-06, + "loss": 0.3995, + "step": 39476 + }, + { + "epoch": 14.58, + "learning_rate": 1.54315802438724e-06, + "loss": 0.2838, + "step": 39477 + }, + { + "epoch": 14.58, + "learning_rate": 1.5431210740238945e-06, + "loss": 0.1852, + "step": 39478 + }, + { + "epoch": 14.58, + "learning_rate": 1.5430841236605495e-06, + "loss": 0.2578, + "step": 39479 + }, + { + "epoch": 14.58, + "learning_rate": 1.5430471732972042e-06, + "loss": 0.1959, + "step": 39480 + }, + { + "epoch": 14.58, + "learning_rate": 1.5430102229338588e-06, + "loss": 0.2591, + "step": 39481 + }, + { + "epoch": 14.58, + "learning_rate": 1.5429732725705137e-06, + "loss": 0.3825, + "step": 39482 + }, + { + "epoch": 14.58, + "learning_rate": 1.5429363222071683e-06, + "loss": 0.2204, + "step": 39483 + }, + { + "epoch": 14.58, + "learning_rate": 1.5428993718438231e-06, + "loss": 0.308, + "step": 39484 + }, + { + "epoch": 14.58, + "learning_rate": 1.5428624214804778e-06, + "loss": 0.3149, + "step": 39485 + }, + { + "epoch": 14.58, + "learning_rate": 1.5428254711171328e-06, + "loss": 0.2052, + "step": 39486 + }, + { + "epoch": 14.58, + "learning_rate": 1.5427885207537875e-06, + "loss": 0.1488, + "step": 39487 + }, + { + "epoch": 14.58, + "learning_rate": 1.5427515703904423e-06, + "loss": 0.4115, + "step": 39488 + }, + { + "epoch": 14.58, + "learning_rate": 1.542714620027097e-06, + "loss": 0.269, + "step": 39489 + }, + { + "epoch": 14.58, + "learning_rate": 1.5426776696637516e-06, + "loss": 0.273, + "step": 39490 + }, + { + "epoch": 14.58, + "learning_rate": 1.5426407193004064e-06, + "loss": 0.3265, + "step": 39491 + }, + { + "epoch": 14.58, + "learning_rate": 1.542603768937061e-06, + "loss": 0.3124, + "step": 39492 + }, + { + "epoch": 14.58, + "learning_rate": 1.5425668185737161e-06, + "loss": 0.2736, + "step": 39493 + }, + { + "epoch": 14.58, + "learning_rate": 1.5425298682103708e-06, + "loss": 0.27, + "step": 39494 + }, + { + "epoch": 14.58, + "learning_rate": 1.5424929178470256e-06, + "loss": 0.3565, + "step": 39495 + }, + { + "epoch": 14.58, + "learning_rate": 1.5424559674836803e-06, + "loss": 0.4911, + "step": 39496 + }, + { + "epoch": 14.59, + "learning_rate": 1.5424190171203351e-06, + "loss": 0.3319, + "step": 39497 + }, + { + "epoch": 14.59, + "learning_rate": 1.5423820667569897e-06, + "loss": 0.1764, + "step": 39498 + }, + { + "epoch": 14.59, + "learning_rate": 1.5423451163936448e-06, + "loss": 0.2429, + "step": 39499 + }, + { + "epoch": 14.59, + "learning_rate": 1.5423081660302994e-06, + "loss": 0.3845, + "step": 39500 + }, + { + "epoch": 14.59, + "learning_rate": 1.5422712156669539e-06, + "loss": 0.3097, + "step": 39501 + }, + { + "epoch": 14.59, + "learning_rate": 1.542234265303609e-06, + "loss": 0.3123, + "step": 39502 + }, + { + "epoch": 14.59, + "learning_rate": 1.5421973149402636e-06, + "loss": 0.2218, + "step": 39503 + }, + { + "epoch": 14.59, + "learning_rate": 1.5421603645769184e-06, + "loss": 0.2511, + "step": 39504 + }, + { + "epoch": 14.59, + "learning_rate": 1.542123414213573e-06, + "loss": 0.233, + "step": 39505 + }, + { + "epoch": 14.59, + "learning_rate": 1.542086463850228e-06, + "loss": 0.1654, + "step": 39506 + }, + { + "epoch": 14.59, + "learning_rate": 1.5420495134868825e-06, + "loss": 0.3105, + "step": 39507 + }, + { + "epoch": 14.59, + "learning_rate": 1.5420125631235376e-06, + "loss": 0.3605, + "step": 39508 + }, + { + "epoch": 14.59, + "learning_rate": 1.5419756127601922e-06, + "loss": 0.3946, + "step": 39509 + }, + { + "epoch": 14.59, + "learning_rate": 1.5419386623968469e-06, + "loss": 0.4846, + "step": 39510 + }, + { + "epoch": 14.59, + "learning_rate": 1.5419017120335017e-06, + "loss": 0.1953, + "step": 39511 + }, + { + "epoch": 14.59, + "learning_rate": 1.5418647616701564e-06, + "loss": 0.2799, + "step": 39512 + }, + { + "epoch": 14.59, + "learning_rate": 1.5418278113068112e-06, + "loss": 0.3929, + "step": 39513 + }, + { + "epoch": 14.59, + "learning_rate": 1.5417908609434658e-06, + "loss": 0.2593, + "step": 39514 + }, + { + "epoch": 14.59, + "learning_rate": 1.541753910580121e-06, + "loss": 0.3348, + "step": 39515 + }, + { + "epoch": 14.59, + "learning_rate": 1.5417169602167755e-06, + "loss": 0.2933, + "step": 39516 + }, + { + "epoch": 14.59, + "learning_rate": 1.5416800098534304e-06, + "loss": 0.2987, + "step": 39517 + }, + { + "epoch": 14.59, + "learning_rate": 1.541643059490085e-06, + "loss": 0.4439, + "step": 39518 + }, + { + "epoch": 14.59, + "learning_rate": 1.5416061091267399e-06, + "loss": 0.3653, + "step": 39519 + }, + { + "epoch": 14.59, + "learning_rate": 1.5415691587633945e-06, + "loss": 0.1918, + "step": 39520 + }, + { + "epoch": 14.59, + "learning_rate": 1.5415322084000491e-06, + "loss": 0.2261, + "step": 39521 + }, + { + "epoch": 14.59, + "learning_rate": 1.5414952580367042e-06, + "loss": 0.1664, + "step": 39522 + }, + { + "epoch": 14.59, + "learning_rate": 1.5414583076733586e-06, + "loss": 0.2328, + "step": 39523 + }, + { + "epoch": 14.6, + "learning_rate": 1.5414213573100137e-06, + "loss": 0.2769, + "step": 39524 + }, + { + "epoch": 14.6, + "learning_rate": 1.5413844069466683e-06, + "loss": 0.2038, + "step": 39525 + }, + { + "epoch": 14.6, + "learning_rate": 1.5413474565833232e-06, + "loss": 0.3586, + "step": 39526 + }, + { + "epoch": 14.6, + "learning_rate": 1.5413105062199778e-06, + "loss": 0.2275, + "step": 39527 + }, + { + "epoch": 14.6, + "learning_rate": 1.5412735558566329e-06, + "loss": 0.2347, + "step": 39528 + }, + { + "epoch": 14.6, + "learning_rate": 1.5412366054932873e-06, + "loss": 0.2544, + "step": 39529 + }, + { + "epoch": 14.6, + "learning_rate": 1.541199655129942e-06, + "loss": 0.1911, + "step": 39530 + }, + { + "epoch": 14.6, + "learning_rate": 1.541162704766597e-06, + "loss": 0.3237, + "step": 39531 + }, + { + "epoch": 14.6, + "learning_rate": 1.5411257544032516e-06, + "loss": 0.3952, + "step": 39532 + }, + { + "epoch": 14.6, + "learning_rate": 1.5410888040399065e-06, + "loss": 0.2893, + "step": 39533 + }, + { + "epoch": 14.6, + "learning_rate": 1.541051853676561e-06, + "loss": 0.3517, + "step": 39534 + }, + { + "epoch": 14.6, + "learning_rate": 1.541014903313216e-06, + "loss": 0.3171, + "step": 39535 + }, + { + "epoch": 14.6, + "learning_rate": 1.5409779529498706e-06, + "loss": 0.1211, + "step": 39536 + }, + { + "epoch": 14.6, + "learning_rate": 1.5409410025865256e-06, + "loss": 0.1667, + "step": 39537 + }, + { + "epoch": 14.6, + "learning_rate": 1.5409040522231803e-06, + "loss": 0.3618, + "step": 39538 + }, + { + "epoch": 14.6, + "learning_rate": 1.5408671018598351e-06, + "loss": 0.2426, + "step": 39539 + }, + { + "epoch": 14.6, + "learning_rate": 1.5408301514964898e-06, + "loss": 0.2022, + "step": 39540 + }, + { + "epoch": 14.6, + "learning_rate": 1.5407932011331444e-06, + "loss": 0.2776, + "step": 39541 + }, + { + "epoch": 14.6, + "learning_rate": 1.5407562507697993e-06, + "loss": 0.2757, + "step": 39542 + }, + { + "epoch": 14.6, + "learning_rate": 1.5407193004064539e-06, + "loss": 0.1912, + "step": 39543 + }, + { + "epoch": 14.6, + "learning_rate": 1.540682350043109e-06, + "loss": 0.2606, + "step": 39544 + }, + { + "epoch": 14.6, + "learning_rate": 1.5406453996797634e-06, + "loss": 0.2754, + "step": 39545 + }, + { + "epoch": 14.6, + "learning_rate": 1.5406084493164184e-06, + "loss": 0.193, + "step": 39546 + }, + { + "epoch": 14.6, + "learning_rate": 1.540571498953073e-06, + "loss": 0.3103, + "step": 39547 + }, + { + "epoch": 14.6, + "learning_rate": 1.540534548589728e-06, + "loss": 0.263, + "step": 39548 + }, + { + "epoch": 14.6, + "learning_rate": 1.5404975982263826e-06, + "loss": 0.1829, + "step": 39549 + }, + { + "epoch": 14.6, + "learning_rate": 1.5404606478630372e-06, + "loss": 0.3014, + "step": 39550 + }, + { + "epoch": 14.61, + "learning_rate": 1.540423697499692e-06, + "loss": 0.3435, + "step": 39551 + }, + { + "epoch": 14.61, + "learning_rate": 1.5403867471363467e-06, + "loss": 0.2136, + "step": 39552 + }, + { + "epoch": 14.61, + "learning_rate": 1.5403497967730017e-06, + "loss": 0.304, + "step": 39553 + }, + { + "epoch": 14.61, + "learning_rate": 1.5403128464096564e-06, + "loss": 0.3375, + "step": 39554 + }, + { + "epoch": 14.61, + "learning_rate": 1.5402758960463112e-06, + "loss": 0.3996, + "step": 39555 + }, + { + "epoch": 14.61, + "learning_rate": 1.5402389456829659e-06, + "loss": 0.343, + "step": 39556 + }, + { + "epoch": 14.61, + "learning_rate": 1.5402019953196207e-06, + "loss": 0.2232, + "step": 39557 + }, + { + "epoch": 14.61, + "learning_rate": 1.5401650449562753e-06, + "loss": 0.4114, + "step": 39558 + }, + { + "epoch": 14.61, + "learning_rate": 1.5401280945929304e-06, + "loss": 0.2528, + "step": 39559 + }, + { + "epoch": 14.61, + "learning_rate": 1.540091144229585e-06, + "loss": 0.3397, + "step": 39560 + }, + { + "epoch": 14.61, + "learning_rate": 1.5400541938662397e-06, + "loss": 0.2264, + "step": 39561 + }, + { + "epoch": 14.61, + "learning_rate": 1.5400172435028945e-06, + "loss": 0.1723, + "step": 39562 + }, + { + "epoch": 14.61, + "learning_rate": 1.5399802931395492e-06, + "loss": 0.381, + "step": 39563 + }, + { + "epoch": 14.61, + "learning_rate": 1.539943342776204e-06, + "loss": 0.2091, + "step": 39564 + }, + { + "epoch": 14.61, + "learning_rate": 1.5399063924128586e-06, + "loss": 0.2735, + "step": 39565 + }, + { + "epoch": 14.61, + "learning_rate": 1.5398694420495137e-06, + "loss": 0.2656, + "step": 39566 + }, + { + "epoch": 14.61, + "learning_rate": 1.5398324916861683e-06, + "loss": 0.3552, + "step": 39567 + }, + { + "epoch": 14.61, + "learning_rate": 1.5397955413228232e-06, + "loss": 0.2403, + "step": 39568 + }, + { + "epoch": 14.61, + "learning_rate": 1.5397585909594778e-06, + "loss": 0.3016, + "step": 39569 + }, + { + "epoch": 14.61, + "learning_rate": 1.5397216405961325e-06, + "loss": 0.2859, + "step": 39570 + }, + { + "epoch": 14.61, + "learning_rate": 1.5396846902327873e-06, + "loss": 0.3548, + "step": 39571 + }, + { + "epoch": 14.61, + "learning_rate": 1.539647739869442e-06, + "loss": 0.3192, + "step": 39572 + }, + { + "epoch": 14.61, + "learning_rate": 1.539610789506097e-06, + "loss": 0.2934, + "step": 39573 + }, + { + "epoch": 14.61, + "learning_rate": 1.5395738391427514e-06, + "loss": 0.194, + "step": 39574 + }, + { + "epoch": 14.61, + "learning_rate": 1.5395368887794065e-06, + "loss": 0.2099, + "step": 39575 + }, + { + "epoch": 14.61, + "learning_rate": 1.5394999384160611e-06, + "loss": 0.344, + "step": 39576 + }, + { + "epoch": 14.61, + "learning_rate": 1.539462988052716e-06, + "loss": 0.1624, + "step": 39577 + }, + { + "epoch": 14.62, + "learning_rate": 1.5394260376893706e-06, + "loss": 0.1649, + "step": 39578 + }, + { + "epoch": 14.62, + "learning_rate": 1.5393890873260257e-06, + "loss": 0.3265, + "step": 39579 + }, + { + "epoch": 14.62, + "learning_rate": 1.53935213696268e-06, + "loss": 0.2838, + "step": 39580 + }, + { + "epoch": 14.62, + "learning_rate": 1.5393151865993347e-06, + "loss": 0.2745, + "step": 39581 + }, + { + "epoch": 14.62, + "learning_rate": 1.5392782362359898e-06, + "loss": 0.217, + "step": 39582 + }, + { + "epoch": 14.62, + "learning_rate": 1.5392412858726444e-06, + "loss": 0.3583, + "step": 39583 + }, + { + "epoch": 14.62, + "learning_rate": 1.5392043355092993e-06, + "loss": 0.3106, + "step": 39584 + }, + { + "epoch": 14.62, + "learning_rate": 1.539167385145954e-06, + "loss": 0.2036, + "step": 39585 + }, + { + "epoch": 14.62, + "learning_rate": 1.5391304347826088e-06, + "loss": 0.2808, + "step": 39586 + }, + { + "epoch": 14.62, + "learning_rate": 1.5390934844192634e-06, + "loss": 0.1219, + "step": 39587 + }, + { + "epoch": 14.62, + "learning_rate": 1.5390565340559185e-06, + "loss": 0.2279, + "step": 39588 + }, + { + "epoch": 14.62, + "learning_rate": 1.539019583692573e-06, + "loss": 0.3139, + "step": 39589 + }, + { + "epoch": 14.62, + "learning_rate": 1.5389826333292275e-06, + "loss": 0.3081, + "step": 39590 + }, + { + "epoch": 14.62, + "learning_rate": 1.5389456829658826e-06, + "loss": 0.2388, + "step": 39591 + }, + { + "epoch": 14.62, + "learning_rate": 1.5389087326025372e-06, + "loss": 0.3578, + "step": 39592 + }, + { + "epoch": 14.62, + "learning_rate": 1.538871782239192e-06, + "loss": 0.4002, + "step": 39593 + }, + { + "epoch": 14.62, + "learning_rate": 1.5388348318758467e-06, + "loss": 0.2258, + "step": 39594 + }, + { + "epoch": 14.62, + "learning_rate": 1.5387978815125018e-06, + "loss": 0.2992, + "step": 39595 + }, + { + "epoch": 14.62, + "learning_rate": 1.5387609311491562e-06, + "loss": 0.3666, + "step": 39596 + }, + { + "epoch": 14.62, + "learning_rate": 1.5387239807858112e-06, + "loss": 0.2257, + "step": 39597 + }, + { + "epoch": 14.62, + "learning_rate": 1.5386870304224659e-06, + "loss": 0.2037, + "step": 39598 + }, + { + "epoch": 14.62, + "learning_rate": 1.5386500800591207e-06, + "loss": 0.1468, + "step": 39599 + }, + { + "epoch": 14.62, + "learning_rate": 1.5386131296957754e-06, + "loss": 0.3305, + "step": 39600 + }, + { + "epoch": 14.62, + "learning_rate": 1.53857617933243e-06, + "loss": 0.1249, + "step": 39601 + }, + { + "epoch": 14.62, + "learning_rate": 1.5385392289690849e-06, + "loss": 0.3803, + "step": 39602 + }, + { + "epoch": 14.62, + "learning_rate": 1.5385022786057395e-06, + "loss": 0.2463, + "step": 39603 + }, + { + "epoch": 14.62, + "learning_rate": 1.5384653282423945e-06, + "loss": 0.3506, + "step": 39604 + }, + { + "epoch": 14.63, + "learning_rate": 1.5384283778790492e-06, + "loss": 0.3527, + "step": 39605 + }, + { + "epoch": 14.63, + "learning_rate": 1.538391427515704e-06, + "loss": 0.3172, + "step": 39606 + }, + { + "epoch": 14.63, + "learning_rate": 1.5383544771523587e-06, + "loss": 0.2214, + "step": 39607 + }, + { + "epoch": 14.63, + "learning_rate": 1.5383175267890135e-06, + "loss": 0.3354, + "step": 39608 + }, + { + "epoch": 14.63, + "learning_rate": 1.5382805764256682e-06, + "loss": 0.33, + "step": 39609 + }, + { + "epoch": 14.63, + "learning_rate": 1.5382436260623228e-06, + "loss": 0.1153, + "step": 39610 + }, + { + "epoch": 14.63, + "learning_rate": 1.5382066756989779e-06, + "loss": 0.2913, + "step": 39611 + }, + { + "epoch": 14.63, + "learning_rate": 1.5381697253356325e-06, + "loss": 0.271, + "step": 39612 + }, + { + "epoch": 14.63, + "learning_rate": 1.5381327749722873e-06, + "loss": 0.2827, + "step": 39613 + }, + { + "epoch": 14.63, + "learning_rate": 1.538095824608942e-06, + "loss": 0.2395, + "step": 39614 + }, + { + "epoch": 14.63, + "learning_rate": 1.5380588742455968e-06, + "loss": 0.3056, + "step": 39615 + }, + { + "epoch": 14.63, + "learning_rate": 1.5380219238822515e-06, + "loss": 0.3, + "step": 39616 + }, + { + "epoch": 14.63, + "learning_rate": 1.5379849735189065e-06, + "loss": 0.2403, + "step": 39617 + }, + { + "epoch": 14.63, + "learning_rate": 1.537948023155561e-06, + "loss": 0.2325, + "step": 39618 + }, + { + "epoch": 14.63, + "learning_rate": 1.537911072792216e-06, + "loss": 0.2322, + "step": 39619 + }, + { + "epoch": 14.63, + "learning_rate": 1.5378741224288706e-06, + "loss": 0.2083, + "step": 39620 + }, + { + "epoch": 14.63, + "learning_rate": 1.5378371720655253e-06, + "loss": 0.2479, + "step": 39621 + }, + { + "epoch": 14.63, + "learning_rate": 1.5378002217021801e-06, + "loss": 0.2622, + "step": 39622 + }, + { + "epoch": 14.63, + "learning_rate": 1.5377632713388348e-06, + "loss": 0.2894, + "step": 39623 + }, + { + "epoch": 14.63, + "learning_rate": 1.5377263209754896e-06, + "loss": 0.2524, + "step": 39624 + }, + { + "epoch": 14.63, + "learning_rate": 1.5376893706121442e-06, + "loss": 0.3228, + "step": 39625 + }, + { + "epoch": 14.63, + "learning_rate": 1.5376524202487993e-06, + "loss": 0.3041, + "step": 39626 + }, + { + "epoch": 14.63, + "learning_rate": 1.537615469885454e-06, + "loss": 0.3845, + "step": 39627 + }, + { + "epoch": 14.63, + "learning_rate": 1.5375785195221088e-06, + "loss": 0.2449, + "step": 39628 + }, + { + "epoch": 14.63, + "learning_rate": 1.5375415691587634e-06, + "loss": 0.3188, + "step": 39629 + }, + { + "epoch": 14.63, + "learning_rate": 1.537504618795418e-06, + "loss": 0.2145, + "step": 39630 + }, + { + "epoch": 14.63, + "learning_rate": 1.537467668432073e-06, + "loss": 0.2442, + "step": 39631 + }, + { + "epoch": 14.64, + "learning_rate": 1.5374307180687275e-06, + "loss": 0.2515, + "step": 39632 + }, + { + "epoch": 14.64, + "learning_rate": 1.5373937677053826e-06, + "loss": 0.2954, + "step": 39633 + }, + { + "epoch": 14.64, + "learning_rate": 1.5373568173420372e-06, + "loss": 0.267, + "step": 39634 + }, + { + "epoch": 14.64, + "learning_rate": 1.537319866978692e-06, + "loss": 0.3058, + "step": 39635 + }, + { + "epoch": 14.64, + "learning_rate": 1.5372829166153467e-06, + "loss": 0.2603, + "step": 39636 + }, + { + "epoch": 14.64, + "learning_rate": 1.5372459662520016e-06, + "loss": 0.1845, + "step": 39637 + }, + { + "epoch": 14.64, + "learning_rate": 1.5372090158886562e-06, + "loss": 0.3276, + "step": 39638 + }, + { + "epoch": 14.64, + "learning_rate": 1.5371720655253113e-06, + "loss": 0.3476, + "step": 39639 + }, + { + "epoch": 14.64, + "learning_rate": 1.537135115161966e-06, + "loss": 0.2443, + "step": 39640 + }, + { + "epoch": 14.64, + "learning_rate": 1.5370981647986203e-06, + "loss": 0.2915, + "step": 39641 + }, + { + "epoch": 14.64, + "learning_rate": 1.5370612144352754e-06, + "loss": 0.2491, + "step": 39642 + }, + { + "epoch": 14.64, + "learning_rate": 1.53702426407193e-06, + "loss": 0.1605, + "step": 39643 + }, + { + "epoch": 14.64, + "learning_rate": 1.5369873137085849e-06, + "loss": 0.1771, + "step": 39644 + }, + { + "epoch": 14.64, + "learning_rate": 1.5369503633452395e-06, + "loss": 0.2903, + "step": 39645 + }, + { + "epoch": 14.64, + "learning_rate": 1.5369134129818946e-06, + "loss": 0.2804, + "step": 39646 + }, + { + "epoch": 14.64, + "learning_rate": 1.536876462618549e-06, + "loss": 0.239, + "step": 39647 + }, + { + "epoch": 14.64, + "learning_rate": 1.536839512255204e-06, + "loss": 0.2611, + "step": 39648 + }, + { + "epoch": 14.64, + "learning_rate": 1.5368025618918587e-06, + "loss": 0.4499, + "step": 39649 + }, + { + "epoch": 14.64, + "learning_rate": 1.5367656115285133e-06, + "loss": 0.2362, + "step": 39650 + }, + { + "epoch": 14.64, + "learning_rate": 1.5367286611651682e-06, + "loss": 0.1716, + "step": 39651 + }, + { + "epoch": 14.64, + "learning_rate": 1.5366917108018228e-06, + "loss": 0.2806, + "step": 39652 + }, + { + "epoch": 14.64, + "learning_rate": 1.5366547604384777e-06, + "loss": 0.3412, + "step": 39653 + }, + { + "epoch": 14.64, + "learning_rate": 1.5366178100751323e-06, + "loss": 0.1637, + "step": 39654 + }, + { + "epoch": 14.64, + "learning_rate": 1.5365808597117874e-06, + "loss": 0.2261, + "step": 39655 + }, + { + "epoch": 14.64, + "learning_rate": 1.536543909348442e-06, + "loss": 0.2898, + "step": 39656 + }, + { + "epoch": 14.64, + "learning_rate": 1.5365069589850968e-06, + "loss": 0.2779, + "step": 39657 + }, + { + "epoch": 14.64, + "learning_rate": 1.5364700086217515e-06, + "loss": 0.3191, + "step": 39658 + }, + { + "epoch": 14.65, + "learning_rate": 1.5364330582584063e-06, + "loss": 0.3546, + "step": 39659 + }, + { + "epoch": 14.65, + "learning_rate": 1.536396107895061e-06, + "loss": 0.2485, + "step": 39660 + }, + { + "epoch": 14.65, + "learning_rate": 1.5363591575317156e-06, + "loss": 0.2994, + "step": 39661 + }, + { + "epoch": 14.65, + "learning_rate": 1.5363222071683707e-06, + "loss": 0.2567, + "step": 39662 + }, + { + "epoch": 14.65, + "learning_rate": 1.536285256805025e-06, + "loss": 0.2469, + "step": 39663 + }, + { + "epoch": 14.65, + "learning_rate": 1.5362483064416801e-06, + "loss": 0.357, + "step": 39664 + }, + { + "epoch": 14.65, + "learning_rate": 1.5362113560783348e-06, + "loss": 0.3548, + "step": 39665 + }, + { + "epoch": 14.65, + "learning_rate": 1.5361744057149896e-06, + "loss": 0.2327, + "step": 39666 + }, + { + "epoch": 14.65, + "learning_rate": 1.5361374553516443e-06, + "loss": 0.2829, + "step": 39667 + }, + { + "epoch": 14.65, + "learning_rate": 1.5361005049882993e-06, + "loss": 0.1664, + "step": 39668 + }, + { + "epoch": 14.65, + "learning_rate": 1.5360635546249538e-06, + "loss": 0.1929, + "step": 39669 + }, + { + "epoch": 14.65, + "learning_rate": 1.5360266042616084e-06, + "loss": 0.2821, + "step": 39670 + }, + { + "epoch": 14.65, + "learning_rate": 1.5359896538982634e-06, + "loss": 0.3676, + "step": 39671 + }, + { + "epoch": 14.65, + "learning_rate": 1.535952703534918e-06, + "loss": 0.3216, + "step": 39672 + }, + { + "epoch": 14.65, + "learning_rate": 1.535915753171573e-06, + "loss": 0.2325, + "step": 39673 + }, + { + "epoch": 14.65, + "learning_rate": 1.5358788028082276e-06, + "loss": 0.2559, + "step": 39674 + }, + { + "epoch": 14.65, + "learning_rate": 1.5358418524448824e-06, + "loss": 0.2994, + "step": 39675 + }, + { + "epoch": 14.65, + "learning_rate": 1.535804902081537e-06, + "loss": 0.275, + "step": 39676 + }, + { + "epoch": 14.65, + "learning_rate": 1.5357679517181921e-06, + "loss": 0.3925, + "step": 39677 + }, + { + "epoch": 14.65, + "learning_rate": 1.5357310013548468e-06, + "loss": 0.381, + "step": 39678 + }, + { + "epoch": 14.65, + "learning_rate": 1.5356940509915016e-06, + "loss": 0.3444, + "step": 39679 + }, + { + "epoch": 14.65, + "learning_rate": 1.5356571006281562e-06, + "loss": 0.2224, + "step": 39680 + }, + { + "epoch": 14.65, + "learning_rate": 1.5356201502648109e-06, + "loss": 0.3175, + "step": 39681 + }, + { + "epoch": 14.65, + "learning_rate": 1.5355831999014657e-06, + "loss": 0.3218, + "step": 39682 + }, + { + "epoch": 14.65, + "learning_rate": 1.5355462495381204e-06, + "loss": 0.3583, + "step": 39683 + }, + { + "epoch": 14.65, + "learning_rate": 1.5355092991747754e-06, + "loss": 0.2902, + "step": 39684 + }, + { + "epoch": 14.65, + "learning_rate": 1.53547234881143e-06, + "loss": 0.3015, + "step": 39685 + }, + { + "epoch": 14.66, + "learning_rate": 1.535435398448085e-06, + "loss": 0.2101, + "step": 39686 + }, + { + "epoch": 14.66, + "learning_rate": 1.5353984480847395e-06, + "loss": 0.1868, + "step": 39687 + }, + { + "epoch": 14.66, + "learning_rate": 1.5353614977213944e-06, + "loss": 0.2679, + "step": 39688 + }, + { + "epoch": 14.66, + "learning_rate": 1.535324547358049e-06, + "loss": 0.3363, + "step": 39689 + }, + { + "epoch": 14.66, + "learning_rate": 1.5352875969947037e-06, + "loss": 0.2997, + "step": 39690 + }, + { + "epoch": 14.66, + "learning_rate": 1.5352506466313585e-06, + "loss": 0.3312, + "step": 39691 + }, + { + "epoch": 14.66, + "learning_rate": 1.5352136962680131e-06, + "loss": 0.2806, + "step": 39692 + }, + { + "epoch": 14.66, + "learning_rate": 1.5351767459046682e-06, + "loss": 0.3592, + "step": 39693 + }, + { + "epoch": 14.66, + "learning_rate": 1.5351397955413228e-06, + "loss": 0.1811, + "step": 39694 + }, + { + "epoch": 14.66, + "learning_rate": 1.5351028451779777e-06, + "loss": 0.2753, + "step": 39695 + }, + { + "epoch": 14.66, + "learning_rate": 1.5350658948146323e-06, + "loss": 0.2944, + "step": 39696 + }, + { + "epoch": 14.66, + "learning_rate": 1.5350289444512872e-06, + "loss": 0.3792, + "step": 39697 + }, + { + "epoch": 14.66, + "learning_rate": 1.5349919940879418e-06, + "loss": 0.2778, + "step": 39698 + }, + { + "epoch": 14.66, + "learning_rate": 1.5349550437245969e-06, + "loss": 0.2302, + "step": 39699 + }, + { + "epoch": 14.66, + "learning_rate": 1.5349180933612515e-06, + "loss": 0.3469, + "step": 39700 + }, + { + "epoch": 14.66, + "learning_rate": 1.5348811429979061e-06, + "loss": 0.2595, + "step": 39701 + }, + { + "epoch": 14.66, + "learning_rate": 1.534844192634561e-06, + "loss": 0.2581, + "step": 39702 + }, + { + "epoch": 14.66, + "learning_rate": 1.5348072422712156e-06, + "loss": 0.3226, + "step": 39703 + }, + { + "epoch": 14.66, + "learning_rate": 1.5347702919078705e-06, + "loss": 0.2575, + "step": 39704 + }, + { + "epoch": 14.66, + "learning_rate": 1.5347333415445251e-06, + "loss": 0.2059, + "step": 39705 + }, + { + "epoch": 14.66, + "learning_rate": 1.5346963911811802e-06, + "loss": 0.3074, + "step": 39706 + }, + { + "epoch": 14.66, + "learning_rate": 1.5346594408178348e-06, + "loss": 0.2609, + "step": 39707 + }, + { + "epoch": 14.66, + "learning_rate": 1.5346224904544897e-06, + "loss": 0.293, + "step": 39708 + }, + { + "epoch": 14.66, + "learning_rate": 1.5345855400911443e-06, + "loss": 0.236, + "step": 39709 + }, + { + "epoch": 14.66, + "learning_rate": 1.534548589727799e-06, + "loss": 0.1766, + "step": 39710 + }, + { + "epoch": 14.66, + "learning_rate": 1.5345116393644538e-06, + "loss": 0.3583, + "step": 39711 + }, + { + "epoch": 14.66, + "learning_rate": 1.5344746890011084e-06, + "loss": 0.2559, + "step": 39712 + }, + { + "epoch": 14.67, + "learning_rate": 1.5344377386377635e-06, + "loss": 0.2292, + "step": 39713 + }, + { + "epoch": 14.67, + "learning_rate": 1.534400788274418e-06, + "loss": 0.2133, + "step": 39714 + }, + { + "epoch": 14.67, + "learning_rate": 1.534363837911073e-06, + "loss": 0.1932, + "step": 39715 + }, + { + "epoch": 14.67, + "learning_rate": 1.5343268875477276e-06, + "loss": 0.28, + "step": 39716 + }, + { + "epoch": 14.67, + "learning_rate": 1.5342899371843824e-06, + "loss": 0.2045, + "step": 39717 + }, + { + "epoch": 14.67, + "learning_rate": 1.534252986821037e-06, + "loss": 0.3204, + "step": 39718 + }, + { + "epoch": 14.67, + "learning_rate": 1.5342160364576917e-06, + "loss": 0.2501, + "step": 39719 + }, + { + "epoch": 14.67, + "learning_rate": 1.5341790860943466e-06, + "loss": 0.3519, + "step": 39720 + }, + { + "epoch": 14.67, + "learning_rate": 1.5341421357310012e-06, + "loss": 0.2991, + "step": 39721 + }, + { + "epoch": 14.67, + "learning_rate": 1.5341051853676563e-06, + "loss": 0.2401, + "step": 39722 + }, + { + "epoch": 14.67, + "learning_rate": 1.534068235004311e-06, + "loss": 0.3219, + "step": 39723 + }, + { + "epoch": 14.67, + "learning_rate": 1.5340312846409657e-06, + "loss": 0.2401, + "step": 39724 + }, + { + "epoch": 14.67, + "learning_rate": 1.5339943342776204e-06, + "loss": 0.2545, + "step": 39725 + }, + { + "epoch": 14.67, + "learning_rate": 1.5339573839142752e-06, + "loss": 0.2617, + "step": 39726 + }, + { + "epoch": 14.67, + "learning_rate": 1.5339204335509299e-06, + "loss": 0.2746, + "step": 39727 + }, + { + "epoch": 14.67, + "learning_rate": 1.533883483187585e-06, + "loss": 0.4635, + "step": 39728 + }, + { + "epoch": 14.67, + "learning_rate": 1.5338465328242396e-06, + "loss": 0.2919, + "step": 39729 + }, + { + "epoch": 14.67, + "learning_rate": 1.533809582460894e-06, + "loss": 0.2267, + "step": 39730 + }, + { + "epoch": 14.67, + "learning_rate": 1.533772632097549e-06, + "loss": 0.2814, + "step": 39731 + }, + { + "epoch": 14.67, + "learning_rate": 1.5337356817342037e-06, + "loss": 0.1832, + "step": 39732 + }, + { + "epoch": 14.67, + "learning_rate": 1.5336987313708585e-06, + "loss": 0.2366, + "step": 39733 + }, + { + "epoch": 14.67, + "learning_rate": 1.5336617810075132e-06, + "loss": 0.328, + "step": 39734 + }, + { + "epoch": 14.67, + "learning_rate": 1.5336248306441682e-06, + "loss": 0.2749, + "step": 39735 + }, + { + "epoch": 14.67, + "learning_rate": 1.5335878802808227e-06, + "loss": 0.3796, + "step": 39736 + }, + { + "epoch": 14.67, + "learning_rate": 1.5335509299174777e-06, + "loss": 0.1512, + "step": 39737 + }, + { + "epoch": 14.67, + "learning_rate": 1.5335139795541323e-06, + "loss": 0.189, + "step": 39738 + }, + { + "epoch": 14.67, + "learning_rate": 1.533477029190787e-06, + "loss": 0.2027, + "step": 39739 + }, + { + "epoch": 14.68, + "learning_rate": 1.5334400788274418e-06, + "loss": 0.2631, + "step": 39740 + }, + { + "epoch": 14.68, + "learning_rate": 1.5334031284640965e-06, + "loss": 0.2893, + "step": 39741 + }, + { + "epoch": 14.68, + "learning_rate": 1.5333661781007513e-06, + "loss": 0.1535, + "step": 39742 + }, + { + "epoch": 14.68, + "learning_rate": 1.533329227737406e-06, + "loss": 0.1802, + "step": 39743 + }, + { + "epoch": 14.68, + "learning_rate": 1.533292277374061e-06, + "loss": 0.3307, + "step": 39744 + }, + { + "epoch": 14.68, + "learning_rate": 1.5332553270107157e-06, + "loss": 0.318, + "step": 39745 + }, + { + "epoch": 14.68, + "learning_rate": 1.5332183766473705e-06, + "loss": 0.163, + "step": 39746 + }, + { + "epoch": 14.68, + "learning_rate": 1.5331814262840251e-06, + "loss": 0.224, + "step": 39747 + }, + { + "epoch": 14.68, + "learning_rate": 1.53314447592068e-06, + "loss": 0.3266, + "step": 39748 + }, + { + "epoch": 14.68, + "learning_rate": 1.5331075255573346e-06, + "loss": 0.231, + "step": 39749 + }, + { + "epoch": 14.68, + "learning_rate": 1.5330705751939893e-06, + "loss": 0.2424, + "step": 39750 + }, + { + "epoch": 14.68, + "learning_rate": 1.5330336248306443e-06, + "loss": 0.3153, + "step": 39751 + }, + { + "epoch": 14.68, + "learning_rate": 1.532996674467299e-06, + "loss": 0.3136, + "step": 39752 + }, + { + "epoch": 14.68, + "learning_rate": 1.5329597241039538e-06, + "loss": 0.2885, + "step": 39753 + }, + { + "epoch": 14.68, + "learning_rate": 1.5329227737406084e-06, + "loss": 0.3329, + "step": 39754 + }, + { + "epoch": 14.68, + "learning_rate": 1.5328858233772633e-06, + "loss": 0.3229, + "step": 39755 + }, + { + "epoch": 14.68, + "learning_rate": 1.532848873013918e-06, + "loss": 0.1925, + "step": 39756 + }, + { + "epoch": 14.68, + "learning_rate": 1.532811922650573e-06, + "loss": 0.2696, + "step": 39757 + }, + { + "epoch": 14.68, + "learning_rate": 1.5327749722872276e-06, + "loss": 0.3465, + "step": 39758 + }, + { + "epoch": 14.68, + "learning_rate": 1.532738021923882e-06, + "loss": 0.3494, + "step": 39759 + }, + { + "epoch": 14.68, + "learning_rate": 1.532701071560537e-06, + "loss": 0.3406, + "step": 39760 + }, + { + "epoch": 14.68, + "learning_rate": 1.5326641211971917e-06, + "loss": 0.3799, + "step": 39761 + }, + { + "epoch": 14.68, + "learning_rate": 1.5326271708338466e-06, + "loss": 0.3774, + "step": 39762 + }, + { + "epoch": 14.68, + "learning_rate": 1.5325902204705012e-06, + "loss": 0.3397, + "step": 39763 + }, + { + "epoch": 14.68, + "learning_rate": 1.5325532701071563e-06, + "loss": 0.2479, + "step": 39764 + }, + { + "epoch": 14.68, + "learning_rate": 1.5325163197438107e-06, + "loss": 0.1878, + "step": 39765 + }, + { + "epoch": 14.68, + "learning_rate": 1.5324793693804658e-06, + "loss": 0.242, + "step": 39766 + }, + { + "epoch": 14.69, + "learning_rate": 1.5324424190171204e-06, + "loss": 0.2602, + "step": 39767 + }, + { + "epoch": 14.69, + "learning_rate": 1.5324054686537753e-06, + "loss": 0.2744, + "step": 39768 + }, + { + "epoch": 14.69, + "learning_rate": 1.5323685182904299e-06, + "loss": 0.1559, + "step": 39769 + }, + { + "epoch": 14.69, + "learning_rate": 1.5323315679270845e-06, + "loss": 0.1949, + "step": 39770 + }, + { + "epoch": 14.69, + "learning_rate": 1.5322946175637394e-06, + "loss": 0.2365, + "step": 39771 + }, + { + "epoch": 14.69, + "learning_rate": 1.532257667200394e-06, + "loss": 0.2869, + "step": 39772 + }, + { + "epoch": 14.69, + "learning_rate": 1.532220716837049e-06, + "loss": 0.2635, + "step": 39773 + }, + { + "epoch": 14.69, + "learning_rate": 1.5321837664737037e-06, + "loss": 0.4074, + "step": 39774 + }, + { + "epoch": 14.69, + "learning_rate": 1.5321468161103586e-06, + "loss": 0.2061, + "step": 39775 + }, + { + "epoch": 14.69, + "learning_rate": 1.5321098657470132e-06, + "loss": 0.2467, + "step": 39776 + }, + { + "epoch": 14.69, + "learning_rate": 1.532072915383668e-06, + "loss": 0.2206, + "step": 39777 + }, + { + "epoch": 14.69, + "learning_rate": 1.5320359650203227e-06, + "loss": 0.3212, + "step": 39778 + }, + { + "epoch": 14.69, + "learning_rate": 1.5319990146569773e-06, + "loss": 0.2552, + "step": 39779 + }, + { + "epoch": 14.69, + "learning_rate": 1.5319620642936324e-06, + "loss": 0.3074, + "step": 39780 + }, + { + "epoch": 14.69, + "learning_rate": 1.5319251139302868e-06, + "loss": 0.2394, + "step": 39781 + }, + { + "epoch": 14.69, + "learning_rate": 1.5318881635669419e-06, + "loss": 0.2226, + "step": 39782 + }, + { + "epoch": 14.69, + "learning_rate": 1.5318512132035965e-06, + "loss": 0.2701, + "step": 39783 + }, + { + "epoch": 14.69, + "learning_rate": 1.5318142628402513e-06, + "loss": 0.2032, + "step": 39784 + }, + { + "epoch": 14.69, + "learning_rate": 1.531777312476906e-06, + "loss": 0.3174, + "step": 39785 + }, + { + "epoch": 14.69, + "learning_rate": 1.531740362113561e-06, + "loss": 0.223, + "step": 39786 + }, + { + "epoch": 14.69, + "learning_rate": 1.5317034117502155e-06, + "loss": 0.2995, + "step": 39787 + }, + { + "epoch": 14.69, + "learning_rate": 1.5316664613868705e-06, + "loss": 0.2491, + "step": 39788 + }, + { + "epoch": 14.69, + "learning_rate": 1.5316295110235252e-06, + "loss": 0.2493, + "step": 39789 + }, + { + "epoch": 14.69, + "learning_rate": 1.5315925606601798e-06, + "loss": 0.2607, + "step": 39790 + }, + { + "epoch": 14.69, + "learning_rate": 1.5315556102968346e-06, + "loss": 0.2548, + "step": 39791 + }, + { + "epoch": 14.69, + "learning_rate": 1.5315186599334893e-06, + "loss": 0.2455, + "step": 39792 + }, + { + "epoch": 14.69, + "learning_rate": 1.5314817095701441e-06, + "loss": 0.1894, + "step": 39793 + }, + { + "epoch": 14.69, + "learning_rate": 1.5314447592067988e-06, + "loss": 0.2643, + "step": 39794 + }, + { + "epoch": 14.7, + "learning_rate": 1.5314078088434538e-06, + "loss": 0.4011, + "step": 39795 + }, + { + "epoch": 14.7, + "learning_rate": 1.5313708584801085e-06, + "loss": 0.3906, + "step": 39796 + }, + { + "epoch": 14.7, + "learning_rate": 1.5313339081167633e-06, + "loss": 0.2847, + "step": 39797 + }, + { + "epoch": 14.7, + "learning_rate": 1.531296957753418e-06, + "loss": 0.4426, + "step": 39798 + }, + { + "epoch": 14.7, + "learning_rate": 1.5312600073900726e-06, + "loss": 0.23, + "step": 39799 + }, + { + "epoch": 14.7, + "learning_rate": 1.5312230570267274e-06, + "loss": 0.342, + "step": 39800 + }, + { + "epoch": 14.7, + "learning_rate": 1.531186106663382e-06, + "loss": 0.1747, + "step": 39801 + }, + { + "epoch": 14.7, + "learning_rate": 1.5311491563000371e-06, + "loss": 0.3313, + "step": 39802 + }, + { + "epoch": 14.7, + "learning_rate": 1.5311122059366916e-06, + "loss": 0.3788, + "step": 39803 + }, + { + "epoch": 14.7, + "learning_rate": 1.5310752555733466e-06, + "loss": 0.2071, + "step": 39804 + }, + { + "epoch": 14.7, + "learning_rate": 1.5310383052100012e-06, + "loss": 0.1587, + "step": 39805 + }, + { + "epoch": 14.7, + "learning_rate": 1.531001354846656e-06, + "loss": 0.4194, + "step": 39806 + }, + { + "epoch": 14.7, + "learning_rate": 1.5309644044833107e-06, + "loss": 0.2565, + "step": 39807 + }, + { + "epoch": 14.7, + "learning_rate": 1.5309274541199658e-06, + "loss": 0.2244, + "step": 39808 + }, + { + "epoch": 14.7, + "learning_rate": 1.5308905037566202e-06, + "loss": 0.3813, + "step": 39809 + }, + { + "epoch": 14.7, + "learning_rate": 1.5308535533932749e-06, + "loss": 0.1526, + "step": 39810 + }, + { + "epoch": 14.7, + "learning_rate": 1.53081660302993e-06, + "loss": 0.2393, + "step": 39811 + }, + { + "epoch": 14.7, + "learning_rate": 1.5307796526665846e-06, + "loss": 0.3192, + "step": 39812 + }, + { + "epoch": 14.7, + "learning_rate": 1.5307427023032394e-06, + "loss": 0.1929, + "step": 39813 + }, + { + "epoch": 14.7, + "learning_rate": 1.530705751939894e-06, + "loss": 0.2329, + "step": 39814 + }, + { + "epoch": 14.7, + "learning_rate": 1.5306688015765489e-06, + "loss": 0.2684, + "step": 39815 + }, + { + "epoch": 14.7, + "learning_rate": 1.5306318512132035e-06, + "loss": 0.1486, + "step": 39816 + }, + { + "epoch": 14.7, + "learning_rate": 1.5305949008498586e-06, + "loss": 0.1974, + "step": 39817 + }, + { + "epoch": 14.7, + "learning_rate": 1.5305579504865132e-06, + "loss": 0.1179, + "step": 39818 + }, + { + "epoch": 14.7, + "learning_rate": 1.5305210001231679e-06, + "loss": 0.4421, + "step": 39819 + }, + { + "epoch": 14.7, + "learning_rate": 1.5304840497598227e-06, + "loss": 0.3593, + "step": 39820 + }, + { + "epoch": 14.7, + "learning_rate": 1.5304470993964773e-06, + "loss": 0.1929, + "step": 39821 + }, + { + "epoch": 14.71, + "learning_rate": 1.5304101490331322e-06, + "loss": 0.3208, + "step": 39822 + }, + { + "epoch": 14.71, + "learning_rate": 1.5303731986697868e-06, + "loss": 0.2344, + "step": 39823 + }, + { + "epoch": 14.71, + "learning_rate": 1.5303362483064419e-06, + "loss": 0.2613, + "step": 39824 + }, + { + "epoch": 14.71, + "learning_rate": 1.5302992979430965e-06, + "loss": 0.2399, + "step": 39825 + }, + { + "epoch": 14.71, + "learning_rate": 1.5302623475797514e-06, + "loss": 0.2313, + "step": 39826 + }, + { + "epoch": 14.71, + "learning_rate": 1.530225397216406e-06, + "loss": 0.2117, + "step": 39827 + }, + { + "epoch": 14.71, + "learning_rate": 1.5301884468530609e-06, + "loss": 0.2297, + "step": 39828 + }, + { + "epoch": 14.71, + "learning_rate": 1.5301514964897155e-06, + "loss": 0.3029, + "step": 39829 + }, + { + "epoch": 14.71, + "learning_rate": 1.5301145461263701e-06, + "loss": 0.2146, + "step": 39830 + }, + { + "epoch": 14.71, + "learning_rate": 1.5300775957630252e-06, + "loss": 0.2136, + "step": 39831 + }, + { + "epoch": 14.71, + "learning_rate": 1.5300406453996796e-06, + "loss": 0.2526, + "step": 39832 + }, + { + "epoch": 14.71, + "learning_rate": 1.5300036950363347e-06, + "loss": 0.2987, + "step": 39833 + }, + { + "epoch": 14.71, + "learning_rate": 1.5299667446729893e-06, + "loss": 0.5058, + "step": 39834 + }, + { + "epoch": 14.71, + "learning_rate": 1.5299297943096442e-06, + "loss": 0.2247, + "step": 39835 + }, + { + "epoch": 14.71, + "learning_rate": 1.5298928439462988e-06, + "loss": 0.3411, + "step": 39836 + }, + { + "epoch": 14.71, + "learning_rate": 1.5298558935829538e-06, + "loss": 0.3028, + "step": 39837 + }, + { + "epoch": 14.71, + "learning_rate": 1.5298189432196083e-06, + "loss": 0.3198, + "step": 39838 + }, + { + "epoch": 14.71, + "learning_rate": 1.529781992856263e-06, + "loss": 0.182, + "step": 39839 + }, + { + "epoch": 14.71, + "learning_rate": 1.529745042492918e-06, + "loss": 0.2311, + "step": 39840 + }, + { + "epoch": 14.71, + "learning_rate": 1.5297080921295726e-06, + "loss": 0.3186, + "step": 39841 + }, + { + "epoch": 14.71, + "learning_rate": 1.5296711417662275e-06, + "loss": 0.2844, + "step": 39842 + }, + { + "epoch": 14.71, + "learning_rate": 1.529634191402882e-06, + "loss": 0.2322, + "step": 39843 + }, + { + "epoch": 14.71, + "learning_rate": 1.529597241039537e-06, + "loss": 0.267, + "step": 39844 + }, + { + "epoch": 14.71, + "learning_rate": 1.5295602906761916e-06, + "loss": 0.1771, + "step": 39845 + }, + { + "epoch": 14.71, + "learning_rate": 1.5295233403128466e-06, + "loss": 0.2778, + "step": 39846 + }, + { + "epoch": 14.71, + "learning_rate": 1.5294863899495013e-06, + "loss": 0.2325, + "step": 39847 + }, + { + "epoch": 14.71, + "learning_rate": 1.5294494395861561e-06, + "loss": 0.2176, + "step": 39848 + }, + { + "epoch": 14.72, + "learning_rate": 1.5294124892228108e-06, + "loss": 0.2304, + "step": 39849 + }, + { + "epoch": 14.72, + "learning_rate": 1.5293755388594654e-06, + "loss": 0.2934, + "step": 39850 + }, + { + "epoch": 14.72, + "learning_rate": 1.5293385884961202e-06, + "loss": 0.224, + "step": 39851 + }, + { + "epoch": 14.72, + "learning_rate": 1.5293016381327749e-06, + "loss": 0.1448, + "step": 39852 + }, + { + "epoch": 14.72, + "learning_rate": 1.52926468776943e-06, + "loss": 0.2646, + "step": 39853 + }, + { + "epoch": 14.72, + "learning_rate": 1.5292277374060844e-06, + "loss": 0.2296, + "step": 39854 + }, + { + "epoch": 14.72, + "learning_rate": 1.5291907870427394e-06, + "loss": 0.2122, + "step": 39855 + }, + { + "epoch": 14.72, + "learning_rate": 1.529153836679394e-06, + "loss": 0.1731, + "step": 39856 + }, + { + "epoch": 14.72, + "learning_rate": 1.529116886316049e-06, + "loss": 0.32, + "step": 39857 + }, + { + "epoch": 14.72, + "learning_rate": 1.5290799359527035e-06, + "loss": 0.2346, + "step": 39858 + }, + { + "epoch": 14.72, + "learning_rate": 1.5290429855893582e-06, + "loss": 0.1925, + "step": 39859 + }, + { + "epoch": 14.72, + "learning_rate": 1.529006035226013e-06, + "loss": 0.1733, + "step": 39860 + }, + { + "epoch": 14.72, + "learning_rate": 1.5289690848626677e-06, + "loss": 0.4212, + "step": 39861 + }, + { + "epoch": 14.72, + "learning_rate": 1.5289321344993227e-06, + "loss": 0.2147, + "step": 39862 + }, + { + "epoch": 14.72, + "learning_rate": 1.5288951841359774e-06, + "loss": 0.4238, + "step": 39863 + }, + { + "epoch": 14.72, + "learning_rate": 1.5288582337726322e-06, + "loss": 0.2304, + "step": 39864 + }, + { + "epoch": 14.72, + "learning_rate": 1.5288212834092868e-06, + "loss": 0.3379, + "step": 39865 + }, + { + "epoch": 14.72, + "learning_rate": 1.5287843330459417e-06, + "loss": 0.2467, + "step": 39866 + }, + { + "epoch": 14.72, + "learning_rate": 1.5287473826825963e-06, + "loss": 0.3147, + "step": 39867 + }, + { + "epoch": 14.72, + "learning_rate": 1.5287104323192514e-06, + "loss": 0.1754, + "step": 39868 + }, + { + "epoch": 14.72, + "learning_rate": 1.528673481955906e-06, + "loss": 0.2767, + "step": 39869 + }, + { + "epoch": 14.72, + "learning_rate": 1.5286365315925607e-06, + "loss": 0.2191, + "step": 39870 + }, + { + "epoch": 14.72, + "learning_rate": 1.5285995812292155e-06, + "loss": 0.3179, + "step": 39871 + }, + { + "epoch": 14.72, + "learning_rate": 1.5285626308658701e-06, + "loss": 0.2185, + "step": 39872 + }, + { + "epoch": 14.72, + "learning_rate": 1.528525680502525e-06, + "loss": 0.3595, + "step": 39873 + }, + { + "epoch": 14.72, + "learning_rate": 1.5284887301391796e-06, + "loss": 0.2566, + "step": 39874 + }, + { + "epoch": 14.72, + "learning_rate": 1.5284517797758347e-06, + "loss": 0.2618, + "step": 39875 + }, + { + "epoch": 14.73, + "learning_rate": 1.5284148294124891e-06, + "loss": 0.2726, + "step": 39876 + }, + { + "epoch": 14.73, + "learning_rate": 1.5283778790491442e-06, + "loss": 0.2774, + "step": 39877 + }, + { + "epoch": 14.73, + "learning_rate": 1.5283409286857988e-06, + "loss": 0.2654, + "step": 39878 + }, + { + "epoch": 14.73, + "learning_rate": 1.5283039783224534e-06, + "loss": 0.3606, + "step": 39879 + }, + { + "epoch": 14.73, + "learning_rate": 1.5282670279591083e-06, + "loss": 0.1886, + "step": 39880 + }, + { + "epoch": 14.73, + "learning_rate": 1.528230077595763e-06, + "loss": 0.2206, + "step": 39881 + }, + { + "epoch": 14.73, + "learning_rate": 1.5281931272324178e-06, + "loss": 0.3883, + "step": 39882 + }, + { + "epoch": 14.73, + "learning_rate": 1.5281561768690724e-06, + "loss": 0.2664, + "step": 39883 + }, + { + "epoch": 14.73, + "learning_rate": 1.5281192265057275e-06, + "loss": 0.2459, + "step": 39884 + }, + { + "epoch": 14.73, + "learning_rate": 1.5280822761423821e-06, + "loss": 0.2914, + "step": 39885 + }, + { + "epoch": 14.73, + "learning_rate": 1.528045325779037e-06, + "loss": 0.2196, + "step": 39886 + }, + { + "epoch": 14.73, + "learning_rate": 1.5280083754156916e-06, + "loss": 0.1817, + "step": 39887 + }, + { + "epoch": 14.73, + "learning_rate": 1.5279714250523464e-06, + "loss": 0.2282, + "step": 39888 + }, + { + "epoch": 14.73, + "learning_rate": 1.527934474689001e-06, + "loss": 0.2789, + "step": 39889 + }, + { + "epoch": 14.73, + "learning_rate": 1.5278975243256557e-06, + "loss": 0.3646, + "step": 39890 + }, + { + "epoch": 14.73, + "learning_rate": 1.5278605739623108e-06, + "loss": 0.1971, + "step": 39891 + }, + { + "epoch": 14.73, + "learning_rate": 1.5278236235989654e-06, + "loss": 0.2611, + "step": 39892 + }, + { + "epoch": 14.73, + "learning_rate": 1.5277866732356203e-06, + "loss": 0.2816, + "step": 39893 + }, + { + "epoch": 14.73, + "learning_rate": 1.527749722872275e-06, + "loss": 0.3722, + "step": 39894 + }, + { + "epoch": 14.73, + "learning_rate": 1.5277127725089298e-06, + "loss": 0.3487, + "step": 39895 + }, + { + "epoch": 14.73, + "learning_rate": 1.5276758221455844e-06, + "loss": 0.2369, + "step": 39896 + }, + { + "epoch": 14.73, + "learning_rate": 1.5276388717822394e-06, + "loss": 0.3634, + "step": 39897 + }, + { + "epoch": 14.73, + "learning_rate": 1.527601921418894e-06, + "loss": 0.2805, + "step": 39898 + }, + { + "epoch": 14.73, + "learning_rate": 1.5275649710555485e-06, + "loss": 0.244, + "step": 39899 + }, + { + "epoch": 14.73, + "learning_rate": 1.5275280206922036e-06, + "loss": 0.1498, + "step": 39900 + }, + { + "epoch": 14.73, + "learning_rate": 1.5274910703288582e-06, + "loss": 0.2303, + "step": 39901 + }, + { + "epoch": 14.73, + "learning_rate": 1.527454119965513e-06, + "loss": 0.1903, + "step": 39902 + }, + { + "epoch": 14.74, + "learning_rate": 1.5274171696021677e-06, + "loss": 0.3452, + "step": 39903 + }, + { + "epoch": 14.74, + "learning_rate": 1.5273802192388227e-06, + "loss": 0.4435, + "step": 39904 + }, + { + "epoch": 14.74, + "learning_rate": 1.5273432688754772e-06, + "loss": 0.151, + "step": 39905 + }, + { + "epoch": 14.74, + "learning_rate": 1.5273063185121322e-06, + "loss": 0.3551, + "step": 39906 + }, + { + "epoch": 14.74, + "learning_rate": 1.5272693681487869e-06, + "loss": 0.2354, + "step": 39907 + }, + { + "epoch": 14.74, + "learning_rate": 1.5272324177854417e-06, + "loss": 0.3079, + "step": 39908 + }, + { + "epoch": 14.74, + "learning_rate": 1.5271954674220964e-06, + "loss": 0.2261, + "step": 39909 + }, + { + "epoch": 14.74, + "learning_rate": 1.527158517058751e-06, + "loss": 0.2986, + "step": 39910 + }, + { + "epoch": 14.74, + "learning_rate": 1.5271215666954058e-06, + "loss": 0.2467, + "step": 39911 + }, + { + "epoch": 14.74, + "learning_rate": 1.5270846163320605e-06, + "loss": 0.1804, + "step": 39912 + }, + { + "epoch": 14.74, + "learning_rate": 1.5270476659687155e-06, + "loss": 0.3897, + "step": 39913 + }, + { + "epoch": 14.74, + "learning_rate": 1.5270107156053702e-06, + "loss": 0.4405, + "step": 39914 + }, + { + "epoch": 14.74, + "learning_rate": 1.526973765242025e-06, + "loss": 0.1721, + "step": 39915 + }, + { + "epoch": 14.74, + "learning_rate": 1.5269368148786797e-06, + "loss": 0.2394, + "step": 39916 + }, + { + "epoch": 14.74, + "learning_rate": 1.5268998645153345e-06, + "loss": 0.265, + "step": 39917 + }, + { + "epoch": 14.74, + "learning_rate": 1.5268629141519891e-06, + "loss": 0.3214, + "step": 39918 + }, + { + "epoch": 14.74, + "learning_rate": 1.5268259637886438e-06, + "loss": 0.331, + "step": 39919 + }, + { + "epoch": 14.74, + "learning_rate": 1.5267890134252988e-06, + "loss": 0.2202, + "step": 39920 + }, + { + "epoch": 14.74, + "learning_rate": 1.5267520630619533e-06, + "loss": 0.2876, + "step": 39921 + }, + { + "epoch": 14.74, + "learning_rate": 1.5267151126986083e-06, + "loss": 0.2047, + "step": 39922 + }, + { + "epoch": 14.74, + "learning_rate": 1.526678162335263e-06, + "loss": 0.4315, + "step": 39923 + }, + { + "epoch": 14.74, + "learning_rate": 1.5266412119719178e-06, + "loss": 0.289, + "step": 39924 + }, + { + "epoch": 14.74, + "learning_rate": 1.5266042616085724e-06, + "loss": 0.2763, + "step": 39925 + }, + { + "epoch": 14.74, + "learning_rate": 1.5265673112452275e-06, + "loss": 0.3213, + "step": 39926 + }, + { + "epoch": 14.74, + "learning_rate": 1.526530360881882e-06, + "loss": 0.2227, + "step": 39927 + }, + { + "epoch": 14.74, + "learning_rate": 1.526493410518537e-06, + "loss": 0.267, + "step": 39928 + }, + { + "epoch": 14.74, + "learning_rate": 1.5264564601551916e-06, + "loss": 0.2981, + "step": 39929 + }, + { + "epoch": 14.75, + "learning_rate": 1.5264195097918463e-06, + "loss": 0.2808, + "step": 39930 + }, + { + "epoch": 14.75, + "learning_rate": 1.5263825594285011e-06, + "loss": 0.3222, + "step": 39931 + }, + { + "epoch": 14.75, + "learning_rate": 1.5263456090651557e-06, + "loss": 0.1692, + "step": 39932 + }, + { + "epoch": 14.75, + "learning_rate": 1.5263086587018106e-06, + "loss": 0.313, + "step": 39933 + }, + { + "epoch": 14.75, + "learning_rate": 1.5262717083384652e-06, + "loss": 0.2452, + "step": 39934 + }, + { + "epoch": 14.75, + "learning_rate": 1.5262347579751203e-06, + "loss": 0.1911, + "step": 39935 + }, + { + "epoch": 14.75, + "learning_rate": 1.526197807611775e-06, + "loss": 0.2534, + "step": 39936 + }, + { + "epoch": 14.75, + "learning_rate": 1.5261608572484298e-06, + "loss": 0.3209, + "step": 39937 + }, + { + "epoch": 14.75, + "learning_rate": 1.5261239068850844e-06, + "loss": 0.2496, + "step": 39938 + }, + { + "epoch": 14.75, + "learning_rate": 1.526086956521739e-06, + "loss": 0.4512, + "step": 39939 + }, + { + "epoch": 14.75, + "learning_rate": 1.526050006158394e-06, + "loss": 0.2806, + "step": 39940 + }, + { + "epoch": 14.75, + "learning_rate": 1.5260130557950485e-06, + "loss": 0.2343, + "step": 39941 + }, + { + "epoch": 14.75, + "learning_rate": 1.5259761054317036e-06, + "loss": 0.2056, + "step": 39942 + }, + { + "epoch": 14.75, + "learning_rate": 1.5259391550683582e-06, + "loss": 0.3075, + "step": 39943 + }, + { + "epoch": 14.75, + "learning_rate": 1.525902204705013e-06, + "loss": 0.3034, + "step": 39944 + }, + { + "epoch": 14.75, + "learning_rate": 1.5258652543416677e-06, + "loss": 0.315, + "step": 39945 + }, + { + "epoch": 14.75, + "learning_rate": 1.5258283039783226e-06, + "loss": 0.1951, + "step": 39946 + }, + { + "epoch": 14.75, + "learning_rate": 1.5257913536149772e-06, + "loss": 0.3648, + "step": 39947 + }, + { + "epoch": 14.75, + "learning_rate": 1.5257544032516323e-06, + "loss": 0.2535, + "step": 39948 + }, + { + "epoch": 14.75, + "learning_rate": 1.525717452888287e-06, + "loss": 0.1444, + "step": 39949 + }, + { + "epoch": 14.75, + "learning_rate": 1.5256805025249413e-06, + "loss": 0.2321, + "step": 39950 + }, + { + "epoch": 14.75, + "learning_rate": 1.5256435521615964e-06, + "loss": 0.252, + "step": 39951 + }, + { + "epoch": 14.75, + "learning_rate": 1.525606601798251e-06, + "loss": 0.1954, + "step": 39952 + }, + { + "epoch": 14.75, + "learning_rate": 1.5255696514349059e-06, + "loss": 0.2862, + "step": 39953 + }, + { + "epoch": 14.75, + "learning_rate": 1.5255327010715605e-06, + "loss": 0.2358, + "step": 39954 + }, + { + "epoch": 14.75, + "learning_rate": 1.5254957507082153e-06, + "loss": 0.2745, + "step": 39955 + }, + { + "epoch": 14.75, + "learning_rate": 1.52545880034487e-06, + "loss": 0.354, + "step": 39956 + }, + { + "epoch": 14.76, + "learning_rate": 1.525421849981525e-06, + "loss": 0.3001, + "step": 39957 + }, + { + "epoch": 14.76, + "learning_rate": 1.5253848996181797e-06, + "loss": 0.1761, + "step": 39958 + }, + { + "epoch": 14.76, + "learning_rate": 1.5253479492548343e-06, + "loss": 0.4153, + "step": 39959 + }, + { + "epoch": 14.76, + "learning_rate": 1.5253109988914892e-06, + "loss": 0.2341, + "step": 39960 + }, + { + "epoch": 14.76, + "learning_rate": 1.5252740485281438e-06, + "loss": 0.2266, + "step": 39961 + }, + { + "epoch": 14.76, + "learning_rate": 1.5252370981647987e-06, + "loss": 0.1862, + "step": 39962 + }, + { + "epoch": 14.76, + "learning_rate": 1.5252001478014533e-06, + "loss": 0.2822, + "step": 39963 + }, + { + "epoch": 14.76, + "learning_rate": 1.5251631974381083e-06, + "loss": 0.1785, + "step": 39964 + }, + { + "epoch": 14.76, + "learning_rate": 1.525126247074763e-06, + "loss": 0.2316, + "step": 39965 + }, + { + "epoch": 14.76, + "learning_rate": 1.5250892967114178e-06, + "loss": 0.1792, + "step": 39966 + }, + { + "epoch": 14.76, + "learning_rate": 1.5250523463480725e-06, + "loss": 0.2553, + "step": 39967 + }, + { + "epoch": 14.76, + "learning_rate": 1.525015395984727e-06, + "loss": 0.2546, + "step": 39968 + }, + { + "epoch": 14.76, + "learning_rate": 1.524978445621382e-06, + "loss": 0.3896, + "step": 39969 + }, + { + "epoch": 14.76, + "learning_rate": 1.5249414952580366e-06, + "loss": 0.226, + "step": 39970 + }, + { + "epoch": 14.76, + "learning_rate": 1.5249045448946916e-06, + "loss": 0.2381, + "step": 39971 + }, + { + "epoch": 14.76, + "learning_rate": 1.524867594531346e-06, + "loss": 0.2683, + "step": 39972 + }, + { + "epoch": 14.76, + "learning_rate": 1.5248306441680011e-06, + "loss": 0.3386, + "step": 39973 + }, + { + "epoch": 14.76, + "learning_rate": 1.5247936938046558e-06, + "loss": 0.2044, + "step": 39974 + }, + { + "epoch": 14.76, + "learning_rate": 1.5247567434413106e-06, + "loss": 0.5003, + "step": 39975 + }, + { + "epoch": 14.76, + "learning_rate": 1.5247197930779653e-06, + "loss": 0.2168, + "step": 39976 + }, + { + "epoch": 14.76, + "learning_rate": 1.5246828427146203e-06, + "loss": 0.2169, + "step": 39977 + }, + { + "epoch": 14.76, + "learning_rate": 1.5246458923512747e-06, + "loss": 0.226, + "step": 39978 + }, + { + "epoch": 14.76, + "learning_rate": 1.5246089419879294e-06, + "loss": 0.3444, + "step": 39979 + }, + { + "epoch": 14.76, + "learning_rate": 1.5245719916245844e-06, + "loss": 0.2887, + "step": 39980 + }, + { + "epoch": 14.76, + "learning_rate": 1.524535041261239e-06, + "loss": 0.3134, + "step": 39981 + }, + { + "epoch": 14.76, + "learning_rate": 1.524498090897894e-06, + "loss": 0.2797, + "step": 39982 + }, + { + "epoch": 14.76, + "learning_rate": 1.5244611405345486e-06, + "loss": 0.3236, + "step": 39983 + }, + { + "epoch": 14.77, + "learning_rate": 1.5244241901712034e-06, + "loss": 0.1424, + "step": 39984 + }, + { + "epoch": 14.77, + "learning_rate": 1.524387239807858e-06, + "loss": 0.2999, + "step": 39985 + }, + { + "epoch": 14.77, + "learning_rate": 1.524350289444513e-06, + "loss": 0.3057, + "step": 39986 + }, + { + "epoch": 14.77, + "learning_rate": 1.5243133390811677e-06, + "loss": 0.2341, + "step": 39987 + }, + { + "epoch": 14.77, + "learning_rate": 1.5242763887178222e-06, + "loss": 0.268, + "step": 39988 + }, + { + "epoch": 14.77, + "learning_rate": 1.5242394383544772e-06, + "loss": 0.258, + "step": 39989 + }, + { + "epoch": 14.77, + "learning_rate": 1.5242024879911319e-06, + "loss": 0.238, + "step": 39990 + }, + { + "epoch": 14.77, + "learning_rate": 1.5241655376277867e-06, + "loss": 0.2156, + "step": 39991 + }, + { + "epoch": 14.77, + "learning_rate": 1.5241285872644413e-06, + "loss": 0.3117, + "step": 39992 + }, + { + "epoch": 14.77, + "learning_rate": 1.5240916369010964e-06, + "loss": 0.1831, + "step": 39993 + }, + { + "epoch": 14.77, + "learning_rate": 1.5240546865377508e-06, + "loss": 0.2523, + "step": 39994 + }, + { + "epoch": 14.77, + "learning_rate": 1.5240177361744059e-06, + "loss": 0.2405, + "step": 39995 + }, + { + "epoch": 14.77, + "learning_rate": 1.5239807858110605e-06, + "loss": 0.3184, + "step": 39996 + }, + { + "epoch": 14.77, + "learning_rate": 1.5239438354477154e-06, + "loss": 0.4886, + "step": 39997 + }, + { + "epoch": 14.77, + "learning_rate": 1.52390688508437e-06, + "loss": 0.2725, + "step": 39998 + }, + { + "epoch": 14.77, + "learning_rate": 1.5238699347210246e-06, + "loss": 0.2034, + "step": 39999 + }, + { + "epoch": 14.77, + "learning_rate": 1.5238329843576795e-06, + "loss": 0.2001, + "step": 40000 + }, + { + "epoch": 14.77, + "learning_rate": 1.5237960339943341e-06, + "loss": 0.2749, + "step": 40001 + }, + { + "epoch": 14.77, + "learning_rate": 1.5237590836309892e-06, + "loss": 0.1782, + "step": 40002 + }, + { + "epoch": 14.77, + "learning_rate": 1.5237221332676438e-06, + "loss": 0.2567, + "step": 40003 + }, + { + "epoch": 14.77, + "learning_rate": 1.5236851829042987e-06, + "loss": 0.2303, + "step": 40004 + }, + { + "epoch": 14.77, + "learning_rate": 1.5236482325409533e-06, + "loss": 0.2578, + "step": 40005 + }, + { + "epoch": 14.77, + "learning_rate": 1.5236112821776082e-06, + "loss": 0.2373, + "step": 40006 + }, + { + "epoch": 14.77, + "learning_rate": 1.5235743318142628e-06, + "loss": 0.1299, + "step": 40007 + }, + { + "epoch": 14.77, + "learning_rate": 1.5235373814509174e-06, + "loss": 0.2082, + "step": 40008 + }, + { + "epoch": 14.77, + "learning_rate": 1.5235004310875725e-06, + "loss": 0.2325, + "step": 40009 + }, + { + "epoch": 14.77, + "learning_rate": 1.5234634807242271e-06, + "loss": 0.2153, + "step": 40010 + }, + { + "epoch": 14.78, + "learning_rate": 1.523426530360882e-06, + "loss": 0.2026, + "step": 40011 + }, + { + "epoch": 14.78, + "learning_rate": 1.5233895799975366e-06, + "loss": 0.219, + "step": 40012 + }, + { + "epoch": 14.78, + "learning_rate": 1.5233526296341915e-06, + "loss": 0.3444, + "step": 40013 + }, + { + "epoch": 14.78, + "learning_rate": 1.523315679270846e-06, + "loss": 0.2304, + "step": 40014 + }, + { + "epoch": 14.78, + "learning_rate": 1.5232787289075012e-06, + "loss": 0.3116, + "step": 40015 + }, + { + "epoch": 14.78, + "learning_rate": 1.5232417785441558e-06, + "loss": 0.3174, + "step": 40016 + }, + { + "epoch": 14.78, + "learning_rate": 1.5232048281808106e-06, + "loss": 0.4153, + "step": 40017 + }, + { + "epoch": 14.78, + "learning_rate": 1.5231678778174653e-06, + "loss": 0.2724, + "step": 40018 + }, + { + "epoch": 14.78, + "learning_rate": 1.52313092745412e-06, + "loss": 0.191, + "step": 40019 + }, + { + "epoch": 14.78, + "learning_rate": 1.5230939770907748e-06, + "loss": 0.2543, + "step": 40020 + }, + { + "epoch": 14.78, + "learning_rate": 1.5230570267274294e-06, + "loss": 0.2685, + "step": 40021 + }, + { + "epoch": 14.78, + "learning_rate": 1.5230200763640845e-06, + "loss": 0.2452, + "step": 40022 + }, + { + "epoch": 14.78, + "learning_rate": 1.5229831260007389e-06, + "loss": 0.2836, + "step": 40023 + }, + { + "epoch": 14.78, + "learning_rate": 1.522946175637394e-06, + "loss": 0.2668, + "step": 40024 + }, + { + "epoch": 14.78, + "learning_rate": 1.5229092252740486e-06, + "loss": 0.2208, + "step": 40025 + }, + { + "epoch": 14.78, + "learning_rate": 1.5228722749107034e-06, + "loss": 0.2791, + "step": 40026 + }, + { + "epoch": 14.78, + "learning_rate": 1.522835324547358e-06, + "loss": 0.1826, + "step": 40027 + }, + { + "epoch": 14.78, + "learning_rate": 1.5227983741840127e-06, + "loss": 0.169, + "step": 40028 + }, + { + "epoch": 14.78, + "learning_rate": 1.5227614238206675e-06, + "loss": 0.3241, + "step": 40029 + }, + { + "epoch": 14.78, + "learning_rate": 1.5227244734573222e-06, + "loss": 0.5291, + "step": 40030 + }, + { + "epoch": 14.78, + "learning_rate": 1.5226875230939772e-06, + "loss": 0.2608, + "step": 40031 + }, + { + "epoch": 14.78, + "learning_rate": 1.5226505727306319e-06, + "loss": 0.2148, + "step": 40032 + }, + { + "epoch": 14.78, + "learning_rate": 1.5226136223672867e-06, + "loss": 0.3198, + "step": 40033 + }, + { + "epoch": 14.78, + "learning_rate": 1.5225766720039414e-06, + "loss": 0.1993, + "step": 40034 + }, + { + "epoch": 14.78, + "learning_rate": 1.5225397216405962e-06, + "loss": 0.3914, + "step": 40035 + }, + { + "epoch": 14.78, + "learning_rate": 1.5225027712772509e-06, + "loss": 0.2917, + "step": 40036 + }, + { + "epoch": 14.78, + "learning_rate": 1.522465820913906e-06, + "loss": 0.3237, + "step": 40037 + }, + { + "epoch": 14.79, + "learning_rate": 1.5224288705505605e-06, + "loss": 0.237, + "step": 40038 + }, + { + "epoch": 14.79, + "learning_rate": 1.522391920187215e-06, + "loss": 0.2353, + "step": 40039 + }, + { + "epoch": 14.79, + "learning_rate": 1.52235496982387e-06, + "loss": 0.2283, + "step": 40040 + }, + { + "epoch": 14.79, + "learning_rate": 1.5223180194605247e-06, + "loss": 0.2597, + "step": 40041 + }, + { + "epoch": 14.79, + "learning_rate": 1.5222810690971795e-06, + "loss": 0.2319, + "step": 40042 + }, + { + "epoch": 14.79, + "learning_rate": 1.5222441187338342e-06, + "loss": 0.2329, + "step": 40043 + }, + { + "epoch": 14.79, + "learning_rate": 1.5222071683704892e-06, + "loss": 0.2275, + "step": 40044 + }, + { + "epoch": 14.79, + "learning_rate": 1.5221702180071436e-06, + "loss": 0.2469, + "step": 40045 + }, + { + "epoch": 14.79, + "learning_rate": 1.5221332676437987e-06, + "loss": 0.1575, + "step": 40046 + }, + { + "epoch": 14.79, + "learning_rate": 1.5220963172804533e-06, + "loss": 0.378, + "step": 40047 + }, + { + "epoch": 14.79, + "learning_rate": 1.522059366917108e-06, + "loss": 0.3605, + "step": 40048 + }, + { + "epoch": 14.79, + "learning_rate": 1.5220224165537628e-06, + "loss": 0.161, + "step": 40049 + }, + { + "epoch": 14.79, + "learning_rate": 1.5219854661904175e-06, + "loss": 0.3055, + "step": 40050 + }, + { + "epoch": 14.79, + "learning_rate": 1.5219485158270723e-06, + "loss": 0.3991, + "step": 40051 + }, + { + "epoch": 14.79, + "learning_rate": 1.521911565463727e-06, + "loss": 0.3051, + "step": 40052 + }, + { + "epoch": 14.79, + "learning_rate": 1.521874615100382e-06, + "loss": 0.2108, + "step": 40053 + }, + { + "epoch": 14.79, + "learning_rate": 1.5218376647370366e-06, + "loss": 0.2403, + "step": 40054 + }, + { + "epoch": 14.79, + "learning_rate": 1.5218007143736915e-06, + "loss": 0.2759, + "step": 40055 + }, + { + "epoch": 14.79, + "learning_rate": 1.5217637640103461e-06, + "loss": 0.2797, + "step": 40056 + }, + { + "epoch": 14.79, + "learning_rate": 1.521726813647001e-06, + "loss": 0.2354, + "step": 40057 + }, + { + "epoch": 14.79, + "learning_rate": 1.5216898632836556e-06, + "loss": 0.2199, + "step": 40058 + }, + { + "epoch": 14.79, + "learning_rate": 1.5216529129203102e-06, + "loss": 0.2182, + "step": 40059 + }, + { + "epoch": 14.79, + "learning_rate": 1.5216159625569653e-06, + "loss": 0.2669, + "step": 40060 + }, + { + "epoch": 14.79, + "learning_rate": 1.52157901219362e-06, + "loss": 0.1738, + "step": 40061 + }, + { + "epoch": 14.79, + "learning_rate": 1.5215420618302748e-06, + "loss": 0.2042, + "step": 40062 + }, + { + "epoch": 14.79, + "learning_rate": 1.5215051114669294e-06, + "loss": 0.1467, + "step": 40063 + }, + { + "epoch": 14.79, + "learning_rate": 1.5214681611035843e-06, + "loss": 0.164, + "step": 40064 + }, + { + "epoch": 14.8, + "learning_rate": 1.521431210740239e-06, + "loss": 0.2893, + "step": 40065 + }, + { + "epoch": 14.8, + "learning_rate": 1.521394260376894e-06, + "loss": 0.2572, + "step": 40066 + }, + { + "epoch": 14.8, + "learning_rate": 1.5213573100135484e-06, + "loss": 0.1822, + "step": 40067 + }, + { + "epoch": 14.8, + "learning_rate": 1.521320359650203e-06, + "loss": 0.1852, + "step": 40068 + }, + { + "epoch": 14.8, + "learning_rate": 1.521283409286858e-06, + "loss": 0.3334, + "step": 40069 + }, + { + "epoch": 14.8, + "learning_rate": 1.5212464589235127e-06, + "loss": 0.3373, + "step": 40070 + }, + { + "epoch": 14.8, + "learning_rate": 1.5212095085601676e-06, + "loss": 0.2454, + "step": 40071 + }, + { + "epoch": 14.8, + "learning_rate": 1.5211725581968222e-06, + "loss": 0.3542, + "step": 40072 + }, + { + "epoch": 14.8, + "learning_rate": 1.521135607833477e-06, + "loss": 0.1611, + "step": 40073 + }, + { + "epoch": 14.8, + "learning_rate": 1.5210986574701317e-06, + "loss": 0.1505, + "step": 40074 + }, + { + "epoch": 14.8, + "learning_rate": 1.5210617071067868e-06, + "loss": 0.2459, + "step": 40075 + }, + { + "epoch": 14.8, + "learning_rate": 1.5210247567434414e-06, + "loss": 0.2516, + "step": 40076 + }, + { + "epoch": 14.8, + "learning_rate": 1.5209878063800962e-06, + "loss": 0.222, + "step": 40077 + }, + { + "epoch": 14.8, + "learning_rate": 1.5209508560167509e-06, + "loss": 0.2184, + "step": 40078 + }, + { + "epoch": 14.8, + "learning_rate": 1.5209139056534055e-06, + "loss": 0.1222, + "step": 40079 + }, + { + "epoch": 14.8, + "learning_rate": 1.5208769552900604e-06, + "loss": 0.2064, + "step": 40080 + }, + { + "epoch": 14.8, + "learning_rate": 1.520840004926715e-06, + "loss": 0.3631, + "step": 40081 + }, + { + "epoch": 14.8, + "learning_rate": 1.52080305456337e-06, + "loss": 0.2938, + "step": 40082 + }, + { + "epoch": 14.8, + "learning_rate": 1.5207661042000247e-06, + "loss": 0.3586, + "step": 40083 + }, + { + "epoch": 14.8, + "learning_rate": 1.5207291538366795e-06, + "loss": 0.2814, + "step": 40084 + }, + { + "epoch": 14.8, + "learning_rate": 1.5206922034733342e-06, + "loss": 0.4084, + "step": 40085 + }, + { + "epoch": 14.8, + "learning_rate": 1.520655253109989e-06, + "loss": 0.3433, + "step": 40086 + }, + { + "epoch": 14.8, + "learning_rate": 1.5206183027466437e-06, + "loss": 0.1965, + "step": 40087 + }, + { + "epoch": 14.8, + "learning_rate": 1.5205813523832983e-06, + "loss": 0.2266, + "step": 40088 + }, + { + "epoch": 14.8, + "learning_rate": 1.5205444020199534e-06, + "loss": 0.2471, + "step": 40089 + }, + { + "epoch": 14.8, + "learning_rate": 1.5205074516566078e-06, + "loss": 0.3971, + "step": 40090 + }, + { + "epoch": 14.8, + "learning_rate": 1.5204705012932628e-06, + "loss": 0.2683, + "step": 40091 + }, + { + "epoch": 14.81, + "learning_rate": 1.5204335509299175e-06, + "loss": 0.2513, + "step": 40092 + }, + { + "epoch": 14.81, + "learning_rate": 1.5203966005665723e-06, + "loss": 0.193, + "step": 40093 + }, + { + "epoch": 14.81, + "learning_rate": 1.520359650203227e-06, + "loss": 0.3305, + "step": 40094 + }, + { + "epoch": 14.81, + "learning_rate": 1.520322699839882e-06, + "loss": 0.2379, + "step": 40095 + }, + { + "epoch": 14.81, + "learning_rate": 1.5202857494765364e-06, + "loss": 0.2301, + "step": 40096 + }, + { + "epoch": 14.81, + "learning_rate": 1.5202487991131915e-06, + "loss": 0.2649, + "step": 40097 + }, + { + "epoch": 14.81, + "learning_rate": 1.5202118487498461e-06, + "loss": 0.3197, + "step": 40098 + }, + { + "epoch": 14.81, + "learning_rate": 1.5201748983865008e-06, + "loss": 0.1997, + "step": 40099 + }, + { + "epoch": 14.81, + "learning_rate": 1.5201379480231556e-06, + "loss": 0.2747, + "step": 40100 + }, + { + "epoch": 14.81, + "learning_rate": 1.5201009976598103e-06, + "loss": 0.3187, + "step": 40101 + }, + { + "epoch": 14.81, + "learning_rate": 1.5200640472964651e-06, + "loss": 0.2628, + "step": 40102 + }, + { + "epoch": 14.81, + "learning_rate": 1.5200270969331198e-06, + "loss": 0.3638, + "step": 40103 + }, + { + "epoch": 14.81, + "learning_rate": 1.5199901465697748e-06, + "loss": 0.3044, + "step": 40104 + }, + { + "epoch": 14.81, + "learning_rate": 1.5199531962064294e-06, + "loss": 0.1942, + "step": 40105 + }, + { + "epoch": 14.81, + "learning_rate": 1.5199162458430843e-06, + "loss": 0.2715, + "step": 40106 + }, + { + "epoch": 14.81, + "learning_rate": 1.519879295479739e-06, + "loss": 0.1763, + "step": 40107 + }, + { + "epoch": 14.81, + "learning_rate": 1.5198423451163936e-06, + "loss": 0.3135, + "step": 40108 + }, + { + "epoch": 14.81, + "learning_rate": 1.5198053947530484e-06, + "loss": 0.2812, + "step": 40109 + }, + { + "epoch": 14.81, + "learning_rate": 1.519768444389703e-06, + "loss": 0.3804, + "step": 40110 + }, + { + "epoch": 14.81, + "learning_rate": 1.5197314940263581e-06, + "loss": 0.1676, + "step": 40111 + }, + { + "epoch": 14.81, + "learning_rate": 1.5196945436630125e-06, + "loss": 0.3455, + "step": 40112 + }, + { + "epoch": 14.81, + "learning_rate": 1.5196575932996676e-06, + "loss": 0.3077, + "step": 40113 + }, + { + "epoch": 14.81, + "learning_rate": 1.5196206429363222e-06, + "loss": 0.3931, + "step": 40114 + }, + { + "epoch": 14.81, + "learning_rate": 1.519583692572977e-06, + "loss": 0.25, + "step": 40115 + }, + { + "epoch": 14.81, + "learning_rate": 1.5195467422096317e-06, + "loss": 0.2112, + "step": 40116 + }, + { + "epoch": 14.81, + "learning_rate": 1.5195097918462868e-06, + "loss": 0.2913, + "step": 40117 + }, + { + "epoch": 14.81, + "learning_rate": 1.5194728414829412e-06, + "loss": 0.3206, + "step": 40118 + }, + { + "epoch": 14.81, + "learning_rate": 1.5194358911195958e-06, + "loss": 0.2252, + "step": 40119 + }, + { + "epoch": 14.82, + "learning_rate": 1.519398940756251e-06, + "loss": 0.2566, + "step": 40120 + }, + { + "epoch": 14.82, + "learning_rate": 1.5193619903929055e-06, + "loss": 0.2891, + "step": 40121 + }, + { + "epoch": 14.82, + "learning_rate": 1.5193250400295604e-06, + "loss": 0.3418, + "step": 40122 + }, + { + "epoch": 14.82, + "learning_rate": 1.519288089666215e-06, + "loss": 0.2375, + "step": 40123 + }, + { + "epoch": 14.82, + "learning_rate": 1.5192511393028699e-06, + "loss": 0.1904, + "step": 40124 + }, + { + "epoch": 14.82, + "learning_rate": 1.5192141889395245e-06, + "loss": 0.3076, + "step": 40125 + }, + { + "epoch": 14.82, + "learning_rate": 1.5191772385761796e-06, + "loss": 0.2457, + "step": 40126 + }, + { + "epoch": 14.82, + "learning_rate": 1.5191402882128342e-06, + "loss": 0.2815, + "step": 40127 + }, + { + "epoch": 14.82, + "learning_rate": 1.5191033378494888e-06, + "loss": 0.1836, + "step": 40128 + }, + { + "epoch": 14.82, + "learning_rate": 1.5190663874861437e-06, + "loss": 0.2245, + "step": 40129 + }, + { + "epoch": 14.82, + "learning_rate": 1.5190294371227983e-06, + "loss": 0.2832, + "step": 40130 + }, + { + "epoch": 14.82, + "learning_rate": 1.5189924867594532e-06, + "loss": 0.2874, + "step": 40131 + }, + { + "epoch": 14.82, + "learning_rate": 1.5189555363961078e-06, + "loss": 0.2185, + "step": 40132 + }, + { + "epoch": 14.82, + "learning_rate": 1.5189185860327629e-06, + "loss": 0.3375, + "step": 40133 + }, + { + "epoch": 14.82, + "learning_rate": 1.5188816356694175e-06, + "loss": 0.2369, + "step": 40134 + }, + { + "epoch": 14.82, + "learning_rate": 1.5188446853060724e-06, + "loss": 0.225, + "step": 40135 + }, + { + "epoch": 14.82, + "learning_rate": 1.518807734942727e-06, + "loss": 0.257, + "step": 40136 + }, + { + "epoch": 14.82, + "learning_rate": 1.5187707845793818e-06, + "loss": 0.2926, + "step": 40137 + }, + { + "epoch": 14.82, + "learning_rate": 1.5187338342160365e-06, + "loss": 0.2592, + "step": 40138 + }, + { + "epoch": 14.82, + "learning_rate": 1.5186968838526911e-06, + "loss": 0.3146, + "step": 40139 + }, + { + "epoch": 14.82, + "learning_rate": 1.518659933489346e-06, + "loss": 0.265, + "step": 40140 + }, + { + "epoch": 14.82, + "learning_rate": 1.5186229831260006e-06, + "loss": 0.256, + "step": 40141 + }, + { + "epoch": 14.82, + "learning_rate": 1.5185860327626557e-06, + "loss": 0.2864, + "step": 40142 + }, + { + "epoch": 14.82, + "learning_rate": 1.5185490823993103e-06, + "loss": 0.1684, + "step": 40143 + }, + { + "epoch": 14.82, + "learning_rate": 1.5185121320359651e-06, + "loss": 0.3259, + "step": 40144 + }, + { + "epoch": 14.82, + "learning_rate": 1.5184751816726198e-06, + "loss": 0.308, + "step": 40145 + }, + { + "epoch": 14.82, + "learning_rate": 1.5184382313092746e-06, + "loss": 0.4481, + "step": 40146 + }, + { + "epoch": 14.83, + "learning_rate": 1.5184012809459293e-06, + "loss": 0.2772, + "step": 40147 + }, + { + "epoch": 14.83, + "learning_rate": 1.518364330582584e-06, + "loss": 0.1979, + "step": 40148 + }, + { + "epoch": 14.83, + "learning_rate": 1.518327380219239e-06, + "loss": 0.2304, + "step": 40149 + }, + { + "epoch": 14.83, + "learning_rate": 1.5182904298558936e-06, + "loss": 0.2826, + "step": 40150 + }, + { + "epoch": 14.83, + "learning_rate": 1.5182534794925484e-06, + "loss": 0.2825, + "step": 40151 + }, + { + "epoch": 14.83, + "learning_rate": 1.518216529129203e-06, + "loss": 0.2765, + "step": 40152 + }, + { + "epoch": 14.83, + "learning_rate": 1.518179578765858e-06, + "loss": 0.1591, + "step": 40153 + }, + { + "epoch": 14.83, + "learning_rate": 1.5181426284025126e-06, + "loss": 0.264, + "step": 40154 + }, + { + "epoch": 14.83, + "learning_rate": 1.5181056780391676e-06, + "loss": 0.1724, + "step": 40155 + }, + { + "epoch": 14.83, + "learning_rate": 1.5180687276758223e-06, + "loss": 0.2425, + "step": 40156 + }, + { + "epoch": 14.83, + "learning_rate": 1.5180317773124771e-06, + "loss": 0.289, + "step": 40157 + }, + { + "epoch": 14.83, + "learning_rate": 1.5179948269491317e-06, + "loss": 0.2312, + "step": 40158 + }, + { + "epoch": 14.83, + "learning_rate": 1.5179578765857864e-06, + "loss": 0.2227, + "step": 40159 + }, + { + "epoch": 14.83, + "learning_rate": 1.5179209262224412e-06, + "loss": 0.2063, + "step": 40160 + }, + { + "epoch": 14.83, + "learning_rate": 1.5178839758590959e-06, + "loss": 0.2713, + "step": 40161 + }, + { + "epoch": 14.83, + "learning_rate": 1.517847025495751e-06, + "loss": 0.3878, + "step": 40162 + }, + { + "epoch": 14.83, + "learning_rate": 1.5178100751324053e-06, + "loss": 0.2424, + "step": 40163 + }, + { + "epoch": 14.83, + "learning_rate": 1.5177731247690604e-06, + "loss": 0.1814, + "step": 40164 + }, + { + "epoch": 14.83, + "learning_rate": 1.517736174405715e-06, + "loss": 0.3243, + "step": 40165 + }, + { + "epoch": 14.83, + "learning_rate": 1.5176992240423699e-06, + "loss": 0.1883, + "step": 40166 + }, + { + "epoch": 14.83, + "learning_rate": 1.5176622736790245e-06, + "loss": 0.2366, + "step": 40167 + }, + { + "epoch": 14.83, + "learning_rate": 1.5176253233156792e-06, + "loss": 0.2835, + "step": 40168 + }, + { + "epoch": 14.83, + "learning_rate": 1.517588372952334e-06, + "loss": 0.2853, + "step": 40169 + }, + { + "epoch": 14.83, + "learning_rate": 1.5175514225889887e-06, + "loss": 0.297, + "step": 40170 + }, + { + "epoch": 14.83, + "learning_rate": 1.5175144722256437e-06, + "loss": 0.2023, + "step": 40171 + }, + { + "epoch": 14.83, + "learning_rate": 1.5174775218622983e-06, + "loss": 0.1456, + "step": 40172 + }, + { + "epoch": 14.83, + "learning_rate": 1.5174405714989532e-06, + "loss": 0.2537, + "step": 40173 + }, + { + "epoch": 14.84, + "learning_rate": 1.5174036211356078e-06, + "loss": 0.2474, + "step": 40174 + }, + { + "epoch": 14.84, + "learning_rate": 1.5173666707722627e-06, + "loss": 0.2562, + "step": 40175 + }, + { + "epoch": 14.84, + "learning_rate": 1.5173297204089173e-06, + "loss": 0.2282, + "step": 40176 + }, + { + "epoch": 14.84, + "learning_rate": 1.5172927700455724e-06, + "loss": 0.3281, + "step": 40177 + }, + { + "epoch": 14.84, + "learning_rate": 1.517255819682227e-06, + "loss": 0.2164, + "step": 40178 + }, + { + "epoch": 14.84, + "learning_rate": 1.5172188693188814e-06, + "loss": 0.2378, + "step": 40179 + }, + { + "epoch": 14.84, + "learning_rate": 1.5171819189555365e-06, + "loss": 0.2068, + "step": 40180 + }, + { + "epoch": 14.84, + "learning_rate": 1.5171449685921911e-06, + "loss": 0.2851, + "step": 40181 + }, + { + "epoch": 14.84, + "learning_rate": 1.517108018228846e-06, + "loss": 0.1771, + "step": 40182 + }, + { + "epoch": 14.84, + "learning_rate": 1.5170710678655006e-06, + "loss": 0.2538, + "step": 40183 + }, + { + "epoch": 14.84, + "learning_rate": 1.5170341175021557e-06, + "loss": 0.1669, + "step": 40184 + }, + { + "epoch": 14.84, + "learning_rate": 1.51699716713881e-06, + "loss": 0.2627, + "step": 40185 + }, + { + "epoch": 14.84, + "learning_rate": 1.5169602167754652e-06, + "loss": 0.198, + "step": 40186 + }, + { + "epoch": 14.84, + "learning_rate": 1.5169232664121198e-06, + "loss": 0.3385, + "step": 40187 + }, + { + "epoch": 14.84, + "learning_rate": 1.5168863160487744e-06, + "loss": 0.3768, + "step": 40188 + }, + { + "epoch": 14.84, + "learning_rate": 1.5168493656854293e-06, + "loss": 0.2943, + "step": 40189 + }, + { + "epoch": 14.84, + "learning_rate": 1.516812415322084e-06, + "loss": 0.1898, + "step": 40190 + }, + { + "epoch": 14.84, + "learning_rate": 1.5167754649587388e-06, + "loss": 0.1976, + "step": 40191 + }, + { + "epoch": 14.84, + "learning_rate": 1.5167385145953934e-06, + "loss": 0.3369, + "step": 40192 + }, + { + "epoch": 14.84, + "learning_rate": 1.5167015642320485e-06, + "loss": 0.2554, + "step": 40193 + }, + { + "epoch": 14.84, + "learning_rate": 1.516664613868703e-06, + "loss": 0.1679, + "step": 40194 + }, + { + "epoch": 14.84, + "learning_rate": 1.516627663505358e-06, + "loss": 0.3232, + "step": 40195 + }, + { + "epoch": 14.84, + "learning_rate": 1.5165907131420126e-06, + "loss": 0.3297, + "step": 40196 + }, + { + "epoch": 14.84, + "learning_rate": 1.5165537627786672e-06, + "loss": 0.2646, + "step": 40197 + }, + { + "epoch": 14.84, + "learning_rate": 1.516516812415322e-06, + "loss": 0.3272, + "step": 40198 + }, + { + "epoch": 14.84, + "learning_rate": 1.5164798620519767e-06, + "loss": 0.3495, + "step": 40199 + }, + { + "epoch": 14.84, + "learning_rate": 1.5164429116886318e-06, + "loss": 0.1724, + "step": 40200 + }, + { + "epoch": 14.85, + "learning_rate": 1.5164059613252864e-06, + "loss": 0.3077, + "step": 40201 + }, + { + "epoch": 14.85, + "learning_rate": 1.5163690109619413e-06, + "loss": 0.2593, + "step": 40202 + }, + { + "epoch": 14.85, + "learning_rate": 1.5163320605985959e-06, + "loss": 0.1467, + "step": 40203 + }, + { + "epoch": 14.85, + "learning_rate": 1.5162951102352507e-06, + "loss": 0.2833, + "step": 40204 + }, + { + "epoch": 14.85, + "learning_rate": 1.5162581598719054e-06, + "loss": 0.2058, + "step": 40205 + }, + { + "epoch": 14.85, + "learning_rate": 1.5162212095085604e-06, + "loss": 0.3317, + "step": 40206 + }, + { + "epoch": 14.85, + "learning_rate": 1.516184259145215e-06, + "loss": 0.3221, + "step": 40207 + }, + { + "epoch": 14.85, + "learning_rate": 1.5161473087818695e-06, + "loss": 0.2747, + "step": 40208 + }, + { + "epoch": 14.85, + "learning_rate": 1.5161103584185246e-06, + "loss": 0.2464, + "step": 40209 + }, + { + "epoch": 14.85, + "learning_rate": 1.5160734080551792e-06, + "loss": 0.2858, + "step": 40210 + }, + { + "epoch": 14.85, + "learning_rate": 1.516036457691834e-06, + "loss": 0.2348, + "step": 40211 + }, + { + "epoch": 14.85, + "learning_rate": 1.5159995073284887e-06, + "loss": 0.2913, + "step": 40212 + }, + { + "epoch": 14.85, + "learning_rate": 1.5159625569651437e-06, + "loss": 0.3462, + "step": 40213 + }, + { + "epoch": 14.85, + "learning_rate": 1.5159256066017982e-06, + "loss": 0.284, + "step": 40214 + }, + { + "epoch": 14.85, + "learning_rate": 1.5158886562384532e-06, + "loss": 0.2788, + "step": 40215 + }, + { + "epoch": 14.85, + "learning_rate": 1.5158517058751079e-06, + "loss": 0.4011, + "step": 40216 + }, + { + "epoch": 14.85, + "learning_rate": 1.5158147555117625e-06, + "loss": 0.3126, + "step": 40217 + }, + { + "epoch": 14.85, + "learning_rate": 1.5157778051484173e-06, + "loss": 0.3496, + "step": 40218 + }, + { + "epoch": 14.85, + "learning_rate": 1.515740854785072e-06, + "loss": 0.2772, + "step": 40219 + }, + { + "epoch": 14.85, + "learning_rate": 1.5157039044217268e-06, + "loss": 0.2456, + "step": 40220 + }, + { + "epoch": 14.85, + "learning_rate": 1.5156669540583815e-06, + "loss": 0.297, + "step": 40221 + }, + { + "epoch": 14.85, + "learning_rate": 1.5156300036950365e-06, + "loss": 0.4077, + "step": 40222 + }, + { + "epoch": 14.85, + "learning_rate": 1.5155930533316912e-06, + "loss": 0.2459, + "step": 40223 + }, + { + "epoch": 14.85, + "learning_rate": 1.515556102968346e-06, + "loss": 0.3523, + "step": 40224 + }, + { + "epoch": 14.85, + "learning_rate": 1.5155191526050006e-06, + "loss": 0.2976, + "step": 40225 + }, + { + "epoch": 14.85, + "learning_rate": 1.5154822022416555e-06, + "loss": 0.2622, + "step": 40226 + }, + { + "epoch": 14.85, + "learning_rate": 1.5154452518783101e-06, + "loss": 0.3169, + "step": 40227 + }, + { + "epoch": 14.86, + "learning_rate": 1.5154083015149648e-06, + "loss": 0.236, + "step": 40228 + }, + { + "epoch": 14.86, + "learning_rate": 1.5153713511516198e-06, + "loss": 0.1821, + "step": 40229 + }, + { + "epoch": 14.86, + "learning_rate": 1.5153344007882742e-06, + "loss": 0.1677, + "step": 40230 + }, + { + "epoch": 14.86, + "learning_rate": 1.5152974504249293e-06, + "loss": 0.2836, + "step": 40231 + }, + { + "epoch": 14.86, + "learning_rate": 1.515260500061584e-06, + "loss": 0.2631, + "step": 40232 + }, + { + "epoch": 14.86, + "learning_rate": 1.5152235496982388e-06, + "loss": 0.2993, + "step": 40233 + }, + { + "epoch": 14.86, + "learning_rate": 1.5151865993348934e-06, + "loss": 0.3385, + "step": 40234 + }, + { + "epoch": 14.86, + "learning_rate": 1.5151496489715485e-06, + "loss": 0.2524, + "step": 40235 + }, + { + "epoch": 14.86, + "learning_rate": 1.515112698608203e-06, + "loss": 0.2771, + "step": 40236 + }, + { + "epoch": 14.86, + "learning_rate": 1.5150757482448576e-06, + "loss": 0.1963, + "step": 40237 + }, + { + "epoch": 14.86, + "learning_rate": 1.5150387978815126e-06, + "loss": 0.2304, + "step": 40238 + }, + { + "epoch": 14.86, + "learning_rate": 1.5150018475181672e-06, + "loss": 0.2005, + "step": 40239 + }, + { + "epoch": 14.86, + "learning_rate": 1.514964897154822e-06, + "loss": 0.21, + "step": 40240 + }, + { + "epoch": 14.86, + "learning_rate": 1.5149279467914767e-06, + "loss": 0.3038, + "step": 40241 + }, + { + "epoch": 14.86, + "learning_rate": 1.5148909964281316e-06, + "loss": 0.186, + "step": 40242 + }, + { + "epoch": 14.86, + "learning_rate": 1.5148540460647862e-06, + "loss": 0.2684, + "step": 40243 + }, + { + "epoch": 14.86, + "learning_rate": 1.5148170957014413e-06, + "loss": 0.1596, + "step": 40244 + }, + { + "epoch": 14.86, + "learning_rate": 1.514780145338096e-06, + "loss": 0.2753, + "step": 40245 + }, + { + "epoch": 14.86, + "learning_rate": 1.5147431949747508e-06, + "loss": 0.2717, + "step": 40246 + }, + { + "epoch": 14.86, + "learning_rate": 1.5147062446114054e-06, + "loss": 0.2748, + "step": 40247 + }, + { + "epoch": 14.86, + "learning_rate": 1.51466929424806e-06, + "loss": 0.1962, + "step": 40248 + }, + { + "epoch": 14.86, + "learning_rate": 1.5146323438847149e-06, + "loss": 0.1715, + "step": 40249 + }, + { + "epoch": 14.86, + "learning_rate": 1.5145953935213695e-06, + "loss": 0.2028, + "step": 40250 + }, + { + "epoch": 14.86, + "learning_rate": 1.5145584431580246e-06, + "loss": 0.2695, + "step": 40251 + }, + { + "epoch": 14.86, + "learning_rate": 1.514521492794679e-06, + "loss": 0.2234, + "step": 40252 + }, + { + "epoch": 14.86, + "learning_rate": 1.514484542431334e-06, + "loss": 0.2708, + "step": 40253 + }, + { + "epoch": 14.86, + "learning_rate": 1.5144475920679887e-06, + "loss": 0.2293, + "step": 40254 + }, + { + "epoch": 14.87, + "learning_rate": 1.5144106417046435e-06, + "loss": 0.2879, + "step": 40255 + }, + { + "epoch": 14.87, + "learning_rate": 1.5143736913412982e-06, + "loss": 0.2479, + "step": 40256 + }, + { + "epoch": 14.87, + "learning_rate": 1.5143367409779528e-06, + "loss": 0.2098, + "step": 40257 + }, + { + "epoch": 14.87, + "learning_rate": 1.5142997906146077e-06, + "loss": 0.3606, + "step": 40258 + }, + { + "epoch": 14.87, + "learning_rate": 1.5142628402512623e-06, + "loss": 0.3195, + "step": 40259 + }, + { + "epoch": 14.87, + "learning_rate": 1.5142258898879174e-06, + "loss": 0.3054, + "step": 40260 + }, + { + "epoch": 14.87, + "learning_rate": 1.514188939524572e-06, + "loss": 0.3392, + "step": 40261 + }, + { + "epoch": 14.87, + "learning_rate": 1.5141519891612268e-06, + "loss": 0.2696, + "step": 40262 + }, + { + "epoch": 14.87, + "learning_rate": 1.5141150387978815e-06, + "loss": 0.3611, + "step": 40263 + }, + { + "epoch": 14.87, + "learning_rate": 1.5140780884345363e-06, + "loss": 0.3213, + "step": 40264 + }, + { + "epoch": 14.87, + "learning_rate": 1.514041138071191e-06, + "loss": 0.2633, + "step": 40265 + }, + { + "epoch": 14.87, + "learning_rate": 1.514004187707846e-06, + "loss": 0.2984, + "step": 40266 + }, + { + "epoch": 14.87, + "learning_rate": 1.5139672373445007e-06, + "loss": 0.1108, + "step": 40267 + }, + { + "epoch": 14.87, + "learning_rate": 1.5139302869811553e-06, + "loss": 0.2378, + "step": 40268 + }, + { + "epoch": 14.87, + "learning_rate": 1.5138933366178102e-06, + "loss": 0.3631, + "step": 40269 + }, + { + "epoch": 14.87, + "learning_rate": 1.5138563862544648e-06, + "loss": 0.2291, + "step": 40270 + }, + { + "epoch": 14.87, + "learning_rate": 1.5138194358911196e-06, + "loss": 0.2846, + "step": 40271 + }, + { + "epoch": 14.87, + "learning_rate": 1.5137824855277743e-06, + "loss": 0.2367, + "step": 40272 + }, + { + "epoch": 14.87, + "learning_rate": 1.5137455351644293e-06, + "loss": 0.2281, + "step": 40273 + }, + { + "epoch": 14.87, + "learning_rate": 1.513708584801084e-06, + "loss": 0.1694, + "step": 40274 + }, + { + "epoch": 14.87, + "learning_rate": 1.5136716344377388e-06, + "loss": 0.3758, + "step": 40275 + }, + { + "epoch": 14.87, + "learning_rate": 1.5136346840743935e-06, + "loss": 0.242, + "step": 40276 + }, + { + "epoch": 14.87, + "learning_rate": 1.513597733711048e-06, + "loss": 0.2476, + "step": 40277 + }, + { + "epoch": 14.87, + "learning_rate": 1.513560783347703e-06, + "loss": 0.3156, + "step": 40278 + }, + { + "epoch": 14.87, + "learning_rate": 1.5135238329843576e-06, + "loss": 0.1255, + "step": 40279 + }, + { + "epoch": 14.87, + "learning_rate": 1.5134868826210126e-06, + "loss": 0.1521, + "step": 40280 + }, + { + "epoch": 14.87, + "learning_rate": 1.513449932257667e-06, + "loss": 0.2355, + "step": 40281 + }, + { + "epoch": 14.88, + "learning_rate": 1.5134129818943221e-06, + "loss": 0.2272, + "step": 40282 + }, + { + "epoch": 14.88, + "learning_rate": 1.5133760315309768e-06, + "loss": 0.1451, + "step": 40283 + }, + { + "epoch": 14.88, + "learning_rate": 1.5133390811676316e-06, + "loss": 0.208, + "step": 40284 + }, + { + "epoch": 14.88, + "learning_rate": 1.5133021308042862e-06, + "loss": 0.3511, + "step": 40285 + }, + { + "epoch": 14.88, + "learning_rate": 1.5132651804409413e-06, + "loss": 0.3273, + "step": 40286 + }, + { + "epoch": 14.88, + "learning_rate": 1.5132282300775957e-06, + "loss": 0.2815, + "step": 40287 + }, + { + "epoch": 14.88, + "learning_rate": 1.5131912797142504e-06, + "loss": 0.2137, + "step": 40288 + }, + { + "epoch": 14.88, + "learning_rate": 1.5131543293509054e-06, + "loss": 0.4343, + "step": 40289 + }, + { + "epoch": 14.88, + "learning_rate": 1.51311737898756e-06, + "loss": 0.2789, + "step": 40290 + }, + { + "epoch": 14.88, + "learning_rate": 1.513080428624215e-06, + "loss": 0.2221, + "step": 40291 + }, + { + "epoch": 14.88, + "learning_rate": 1.5130434782608695e-06, + "loss": 0.2129, + "step": 40292 + }, + { + "epoch": 14.88, + "learning_rate": 1.5130065278975244e-06, + "loss": 0.3056, + "step": 40293 + }, + { + "epoch": 14.88, + "learning_rate": 1.512969577534179e-06, + "loss": 0.2186, + "step": 40294 + }, + { + "epoch": 14.88, + "learning_rate": 1.512932627170834e-06, + "loss": 0.2382, + "step": 40295 + }, + { + "epoch": 14.88, + "learning_rate": 1.5128956768074887e-06, + "loss": 0.2933, + "step": 40296 + }, + { + "epoch": 14.88, + "learning_rate": 1.5128587264441431e-06, + "loss": 0.2867, + "step": 40297 + }, + { + "epoch": 14.88, + "learning_rate": 1.5128217760807982e-06, + "loss": 0.3175, + "step": 40298 + }, + { + "epoch": 14.88, + "learning_rate": 1.5127848257174528e-06, + "loss": 0.2915, + "step": 40299 + }, + { + "epoch": 14.88, + "learning_rate": 1.5127478753541077e-06, + "loss": 0.2615, + "step": 40300 + }, + { + "epoch": 14.88, + "learning_rate": 1.5127109249907623e-06, + "loss": 0.3339, + "step": 40301 + }, + { + "epoch": 14.88, + "learning_rate": 1.5126739746274174e-06, + "loss": 0.3257, + "step": 40302 + }, + { + "epoch": 14.88, + "learning_rate": 1.5126370242640718e-06, + "loss": 0.2089, + "step": 40303 + }, + { + "epoch": 14.88, + "learning_rate": 1.5126000739007269e-06, + "loss": 0.2334, + "step": 40304 + }, + { + "epoch": 14.88, + "learning_rate": 1.5125631235373815e-06, + "loss": 0.3243, + "step": 40305 + }, + { + "epoch": 14.88, + "learning_rate": 1.5125261731740364e-06, + "loss": 0.4051, + "step": 40306 + }, + { + "epoch": 14.88, + "learning_rate": 1.512489222810691e-06, + "loss": 0.2426, + "step": 40307 + }, + { + "epoch": 14.88, + "learning_rate": 1.5124522724473456e-06, + "loss": 0.3935, + "step": 40308 + }, + { + "epoch": 14.89, + "learning_rate": 1.5124153220840005e-06, + "loss": 0.2036, + "step": 40309 + }, + { + "epoch": 14.89, + "learning_rate": 1.5123783717206551e-06, + "loss": 0.2898, + "step": 40310 + }, + { + "epoch": 14.89, + "learning_rate": 1.5123414213573102e-06, + "loss": 0.333, + "step": 40311 + }, + { + "epoch": 14.89, + "learning_rate": 1.5123044709939648e-06, + "loss": 0.2662, + "step": 40312 + }, + { + "epoch": 14.89, + "learning_rate": 1.5122675206306197e-06, + "loss": 0.3271, + "step": 40313 + }, + { + "epoch": 14.89, + "learning_rate": 1.5122305702672743e-06, + "loss": 0.5376, + "step": 40314 + }, + { + "epoch": 14.89, + "learning_rate": 1.5121936199039291e-06, + "loss": 0.2958, + "step": 40315 + }, + { + "epoch": 14.89, + "learning_rate": 1.5121566695405838e-06, + "loss": 0.2216, + "step": 40316 + }, + { + "epoch": 14.89, + "learning_rate": 1.5121197191772384e-06, + "loss": 0.3018, + "step": 40317 + }, + { + "epoch": 14.89, + "learning_rate": 1.5120827688138935e-06, + "loss": 0.2754, + "step": 40318 + }, + { + "epoch": 14.89, + "learning_rate": 1.5120458184505481e-06, + "loss": 0.3577, + "step": 40319 + }, + { + "epoch": 14.89, + "learning_rate": 1.512008868087203e-06, + "loss": 0.3141, + "step": 40320 + }, + { + "epoch": 14.89, + "learning_rate": 1.5119719177238576e-06, + "loss": 0.281, + "step": 40321 + }, + { + "epoch": 14.89, + "learning_rate": 1.5119349673605124e-06, + "loss": 0.3289, + "step": 40322 + }, + { + "epoch": 14.89, + "learning_rate": 1.511898016997167e-06, + "loss": 0.417, + "step": 40323 + }, + { + "epoch": 14.89, + "learning_rate": 1.5118610666338221e-06, + "loss": 0.1642, + "step": 40324 + }, + { + "epoch": 14.89, + "learning_rate": 1.5118241162704768e-06, + "loss": 0.4605, + "step": 40325 + }, + { + "epoch": 14.89, + "learning_rate": 1.5117871659071316e-06, + "loss": 0.2518, + "step": 40326 + }, + { + "epoch": 14.89, + "learning_rate": 1.5117502155437863e-06, + "loss": 0.1887, + "step": 40327 + }, + { + "epoch": 14.89, + "learning_rate": 1.511713265180441e-06, + "loss": 0.305, + "step": 40328 + }, + { + "epoch": 14.89, + "learning_rate": 1.5116763148170957e-06, + "loss": 0.2014, + "step": 40329 + }, + { + "epoch": 14.89, + "learning_rate": 1.5116393644537504e-06, + "loss": 0.3678, + "step": 40330 + }, + { + "epoch": 14.89, + "learning_rate": 1.5116024140904052e-06, + "loss": 0.3164, + "step": 40331 + }, + { + "epoch": 14.89, + "learning_rate": 1.5115654637270599e-06, + "loss": 0.2716, + "step": 40332 + }, + { + "epoch": 14.89, + "learning_rate": 1.511528513363715e-06, + "loss": 0.3293, + "step": 40333 + }, + { + "epoch": 14.89, + "learning_rate": 1.5114915630003696e-06, + "loss": 0.3763, + "step": 40334 + }, + { + "epoch": 14.89, + "learning_rate": 1.5114546126370244e-06, + "loss": 0.2558, + "step": 40335 + }, + { + "epoch": 14.9, + "learning_rate": 1.511417662273679e-06, + "loss": 0.2432, + "step": 40336 + }, + { + "epoch": 14.9, + "learning_rate": 1.5113807119103337e-06, + "loss": 0.2059, + "step": 40337 + }, + { + "epoch": 14.9, + "learning_rate": 1.5113437615469885e-06, + "loss": 0.2439, + "step": 40338 + }, + { + "epoch": 14.9, + "learning_rate": 1.5113068111836432e-06, + "loss": 0.1494, + "step": 40339 + }, + { + "epoch": 14.9, + "learning_rate": 1.5112698608202982e-06, + "loss": 0.3793, + "step": 40340 + }, + { + "epoch": 14.9, + "learning_rate": 1.5112329104569529e-06, + "loss": 0.3724, + "step": 40341 + }, + { + "epoch": 14.9, + "learning_rate": 1.5111959600936077e-06, + "loss": 0.3146, + "step": 40342 + }, + { + "epoch": 14.9, + "learning_rate": 1.5111590097302624e-06, + "loss": 0.3186, + "step": 40343 + }, + { + "epoch": 14.9, + "learning_rate": 1.5111220593669172e-06, + "loss": 0.2133, + "step": 40344 + }, + { + "epoch": 14.9, + "learning_rate": 1.5110851090035718e-06, + "loss": 0.3137, + "step": 40345 + }, + { + "epoch": 14.9, + "learning_rate": 1.511048158640227e-06, + "loss": 0.2658, + "step": 40346 + }, + { + "epoch": 14.9, + "learning_rate": 1.5110112082768815e-06, + "loss": 0.315, + "step": 40347 + }, + { + "epoch": 14.9, + "learning_rate": 1.510974257913536e-06, + "loss": 0.2205, + "step": 40348 + }, + { + "epoch": 14.9, + "learning_rate": 1.510937307550191e-06, + "loss": 0.5031, + "step": 40349 + }, + { + "epoch": 14.9, + "learning_rate": 1.5109003571868457e-06, + "loss": 0.2915, + "step": 40350 + }, + { + "epoch": 14.9, + "learning_rate": 1.5108634068235005e-06, + "loss": 0.2745, + "step": 40351 + }, + { + "epoch": 14.9, + "learning_rate": 1.5108264564601551e-06, + "loss": 0.2319, + "step": 40352 + }, + { + "epoch": 14.9, + "learning_rate": 1.5107895060968102e-06, + "loss": 0.4444, + "step": 40353 + }, + { + "epoch": 14.9, + "learning_rate": 1.5107525557334646e-06, + "loss": 0.1394, + "step": 40354 + }, + { + "epoch": 14.9, + "learning_rate": 1.5107156053701197e-06, + "loss": 0.2725, + "step": 40355 + }, + { + "epoch": 14.9, + "learning_rate": 1.5106786550067743e-06, + "loss": 0.2731, + "step": 40356 + }, + { + "epoch": 14.9, + "learning_rate": 1.510641704643429e-06, + "loss": 0.1953, + "step": 40357 + }, + { + "epoch": 14.9, + "learning_rate": 1.5106047542800838e-06, + "loss": 0.2424, + "step": 40358 + }, + { + "epoch": 14.9, + "learning_rate": 1.5105678039167384e-06, + "loss": 0.1821, + "step": 40359 + }, + { + "epoch": 14.9, + "learning_rate": 1.5105308535533933e-06, + "loss": 0.2542, + "step": 40360 + }, + { + "epoch": 14.9, + "learning_rate": 1.510493903190048e-06, + "loss": 0.3006, + "step": 40361 + }, + { + "epoch": 14.9, + "learning_rate": 1.510456952826703e-06, + "loss": 0.2056, + "step": 40362 + }, + { + "epoch": 14.91, + "learning_rate": 1.5104200024633576e-06, + "loss": 0.2385, + "step": 40363 + }, + { + "epoch": 14.91, + "learning_rate": 1.5103830521000125e-06, + "loss": 0.3429, + "step": 40364 + }, + { + "epoch": 14.91, + "learning_rate": 1.5103461017366671e-06, + "loss": 0.1759, + "step": 40365 + }, + { + "epoch": 14.91, + "learning_rate": 1.510309151373322e-06, + "loss": 0.3283, + "step": 40366 + }, + { + "epoch": 14.91, + "learning_rate": 1.5102722010099766e-06, + "loss": 0.2907, + "step": 40367 + }, + { + "epoch": 14.91, + "learning_rate": 1.5102352506466312e-06, + "loss": 0.2163, + "step": 40368 + }, + { + "epoch": 14.91, + "learning_rate": 1.5101983002832863e-06, + "loss": 0.4098, + "step": 40369 + }, + { + "epoch": 14.91, + "learning_rate": 1.5101613499199407e-06, + "loss": 0.2781, + "step": 40370 + }, + { + "epoch": 14.91, + "learning_rate": 1.5101243995565958e-06, + "loss": 0.3768, + "step": 40371 + }, + { + "epoch": 14.91, + "learning_rate": 1.5100874491932504e-06, + "loss": 0.2693, + "step": 40372 + }, + { + "epoch": 14.91, + "learning_rate": 1.5100504988299053e-06, + "loss": 0.3381, + "step": 40373 + }, + { + "epoch": 14.91, + "learning_rate": 1.51001354846656e-06, + "loss": 0.2307, + "step": 40374 + }, + { + "epoch": 14.91, + "learning_rate": 1.509976598103215e-06, + "loss": 0.3827, + "step": 40375 + }, + { + "epoch": 14.91, + "learning_rate": 1.5099396477398694e-06, + "loss": 0.2251, + "step": 40376 + }, + { + "epoch": 14.91, + "learning_rate": 1.509902697376524e-06, + "loss": 0.1715, + "step": 40377 + }, + { + "epoch": 14.91, + "learning_rate": 1.509865747013179e-06, + "loss": 0.2192, + "step": 40378 + }, + { + "epoch": 14.91, + "learning_rate": 1.5098287966498337e-06, + "loss": 0.2622, + "step": 40379 + }, + { + "epoch": 14.91, + "learning_rate": 1.5097918462864886e-06, + "loss": 0.3453, + "step": 40380 + }, + { + "epoch": 14.91, + "learning_rate": 1.5097548959231432e-06, + "loss": 0.2373, + "step": 40381 + }, + { + "epoch": 14.91, + "learning_rate": 1.509717945559798e-06, + "loss": 0.3598, + "step": 40382 + }, + { + "epoch": 14.91, + "learning_rate": 1.5096809951964527e-06, + "loss": 0.288, + "step": 40383 + }, + { + "epoch": 14.91, + "learning_rate": 1.5096440448331077e-06, + "loss": 0.4111, + "step": 40384 + }, + { + "epoch": 14.91, + "learning_rate": 1.5096070944697624e-06, + "loss": 0.2825, + "step": 40385 + }, + { + "epoch": 14.91, + "learning_rate": 1.5095701441064172e-06, + "loss": 0.252, + "step": 40386 + }, + { + "epoch": 14.91, + "learning_rate": 1.5095331937430719e-06, + "loss": 0.2432, + "step": 40387 + }, + { + "epoch": 14.91, + "learning_rate": 1.5094962433797265e-06, + "loss": 0.2224, + "step": 40388 + }, + { + "epoch": 14.91, + "learning_rate": 1.5094592930163813e-06, + "loss": 0.1887, + "step": 40389 + }, + { + "epoch": 14.92, + "learning_rate": 1.509422342653036e-06, + "loss": 0.1946, + "step": 40390 + }, + { + "epoch": 14.92, + "learning_rate": 1.509385392289691e-06, + "loss": 0.1884, + "step": 40391 + }, + { + "epoch": 14.92, + "learning_rate": 1.5093484419263457e-06, + "loss": 0.1939, + "step": 40392 + }, + { + "epoch": 14.92, + "learning_rate": 1.5093114915630005e-06, + "loss": 0.2211, + "step": 40393 + }, + { + "epoch": 14.92, + "learning_rate": 1.5092745411996552e-06, + "loss": 0.3295, + "step": 40394 + }, + { + "epoch": 14.92, + "learning_rate": 1.50923759083631e-06, + "loss": 0.3913, + "step": 40395 + }, + { + "epoch": 14.92, + "learning_rate": 1.5092006404729646e-06, + "loss": 0.1787, + "step": 40396 + }, + { + "epoch": 14.92, + "learning_rate": 1.5091636901096193e-06, + "loss": 0.18, + "step": 40397 + }, + { + "epoch": 14.92, + "learning_rate": 1.5091267397462743e-06, + "loss": 0.2548, + "step": 40398 + }, + { + "epoch": 14.92, + "learning_rate": 1.5090897893829288e-06, + "loss": 0.2938, + "step": 40399 + }, + { + "epoch": 14.92, + "learning_rate": 1.5090528390195838e-06, + "loss": 0.3131, + "step": 40400 + }, + { + "epoch": 14.92, + "learning_rate": 1.5090158886562385e-06, + "loss": 0.1246, + "step": 40401 + }, + { + "epoch": 14.92, + "learning_rate": 1.5089789382928933e-06, + "loss": 0.2682, + "step": 40402 + }, + { + "epoch": 14.92, + "learning_rate": 1.508941987929548e-06, + "loss": 0.1757, + "step": 40403 + }, + { + "epoch": 14.92, + "learning_rate": 1.5089050375662028e-06, + "loss": 0.3529, + "step": 40404 + }, + { + "epoch": 14.92, + "learning_rate": 1.5088680872028574e-06, + "loss": 0.3497, + "step": 40405 + }, + { + "epoch": 14.92, + "learning_rate": 1.5088311368395125e-06, + "loss": 0.1927, + "step": 40406 + }, + { + "epoch": 14.92, + "learning_rate": 1.5087941864761671e-06, + "loss": 0.2271, + "step": 40407 + }, + { + "epoch": 14.92, + "learning_rate": 1.5087572361128218e-06, + "loss": 0.2851, + "step": 40408 + }, + { + "epoch": 14.92, + "learning_rate": 1.5087202857494766e-06, + "loss": 0.331, + "step": 40409 + }, + { + "epoch": 14.92, + "learning_rate": 1.5086833353861313e-06, + "loss": 0.2089, + "step": 40410 + }, + { + "epoch": 14.92, + "learning_rate": 1.508646385022786e-06, + "loss": 0.2871, + "step": 40411 + }, + { + "epoch": 14.92, + "learning_rate": 1.5086094346594407e-06, + "loss": 0.4266, + "step": 40412 + }, + { + "epoch": 14.92, + "learning_rate": 1.5085724842960958e-06, + "loss": 0.4414, + "step": 40413 + }, + { + "epoch": 14.92, + "learning_rate": 1.5085355339327504e-06, + "loss": 0.2278, + "step": 40414 + }, + { + "epoch": 14.92, + "learning_rate": 1.5084985835694053e-06, + "loss": 0.3816, + "step": 40415 + }, + { + "epoch": 14.92, + "learning_rate": 1.50846163320606e-06, + "loss": 0.1502, + "step": 40416 + }, + { + "epoch": 14.93, + "learning_rate": 1.5084246828427146e-06, + "loss": 0.3104, + "step": 40417 + }, + { + "epoch": 14.93, + "learning_rate": 1.5083877324793694e-06, + "loss": 0.3088, + "step": 40418 + }, + { + "epoch": 14.93, + "learning_rate": 1.508350782116024e-06, + "loss": 0.2396, + "step": 40419 + }, + { + "epoch": 14.93, + "learning_rate": 1.508313831752679e-06, + "loss": 0.2971, + "step": 40420 + }, + { + "epoch": 14.93, + "learning_rate": 1.5082768813893335e-06, + "loss": 0.35, + "step": 40421 + }, + { + "epoch": 14.93, + "learning_rate": 1.5082399310259886e-06, + "loss": 0.3276, + "step": 40422 + }, + { + "epoch": 14.93, + "learning_rate": 1.5082029806626432e-06, + "loss": 0.3426, + "step": 40423 + }, + { + "epoch": 14.93, + "learning_rate": 1.508166030299298e-06, + "loss": 0.3253, + "step": 40424 + }, + { + "epoch": 14.93, + "learning_rate": 1.5081290799359527e-06, + "loss": 0.2166, + "step": 40425 + }, + { + "epoch": 14.93, + "learning_rate": 1.5080921295726073e-06, + "loss": 0.2979, + "step": 40426 + }, + { + "epoch": 14.93, + "learning_rate": 1.5080551792092622e-06, + "loss": 0.2276, + "step": 40427 + }, + { + "epoch": 14.93, + "learning_rate": 1.5080182288459168e-06, + "loss": 0.3357, + "step": 40428 + }, + { + "epoch": 14.93, + "learning_rate": 1.5079812784825719e-06, + "loss": 0.1876, + "step": 40429 + }, + { + "epoch": 14.93, + "learning_rate": 1.5079443281192265e-06, + "loss": 0.2707, + "step": 40430 + }, + { + "epoch": 14.93, + "learning_rate": 1.5079073777558814e-06, + "loss": 0.1139, + "step": 40431 + }, + { + "epoch": 14.93, + "learning_rate": 1.507870427392536e-06, + "loss": 0.3017, + "step": 40432 + }, + { + "epoch": 14.93, + "learning_rate": 1.5078334770291909e-06, + "loss": 0.2075, + "step": 40433 + }, + { + "epoch": 14.93, + "learning_rate": 1.5077965266658455e-06, + "loss": 0.4209, + "step": 40434 + }, + { + "epoch": 14.93, + "learning_rate": 1.5077595763025006e-06, + "loss": 0.1641, + "step": 40435 + }, + { + "epoch": 14.93, + "learning_rate": 1.5077226259391552e-06, + "loss": 0.3506, + "step": 40436 + }, + { + "epoch": 14.93, + "learning_rate": 1.5076856755758096e-06, + "loss": 0.2168, + "step": 40437 + }, + { + "epoch": 14.93, + "learning_rate": 1.5076487252124647e-06, + "loss": 0.236, + "step": 40438 + }, + { + "epoch": 14.93, + "learning_rate": 1.5076117748491193e-06, + "loss": 0.3162, + "step": 40439 + }, + { + "epoch": 14.93, + "learning_rate": 1.5075748244857742e-06, + "loss": 0.2233, + "step": 40440 + }, + { + "epoch": 14.93, + "learning_rate": 1.5075378741224288e-06, + "loss": 0.3101, + "step": 40441 + }, + { + "epoch": 14.93, + "learning_rate": 1.5075009237590839e-06, + "loss": 0.1764, + "step": 40442 + }, + { + "epoch": 14.93, + "learning_rate": 1.5074639733957383e-06, + "loss": 0.217, + "step": 40443 + }, + { + "epoch": 14.94, + "learning_rate": 1.5074270230323933e-06, + "loss": 0.1964, + "step": 40444 + }, + { + "epoch": 14.94, + "learning_rate": 1.507390072669048e-06, + "loss": 0.3421, + "step": 40445 + }, + { + "epoch": 14.94, + "learning_rate": 1.5073531223057026e-06, + "loss": 0.365, + "step": 40446 + }, + { + "epoch": 14.94, + "learning_rate": 1.5073161719423575e-06, + "loss": 0.4209, + "step": 40447 + }, + { + "epoch": 14.94, + "learning_rate": 1.507279221579012e-06, + "loss": 0.2826, + "step": 40448 + }, + { + "epoch": 14.94, + "learning_rate": 1.507242271215667e-06, + "loss": 0.1928, + "step": 40449 + }, + { + "epoch": 14.94, + "learning_rate": 1.5072053208523216e-06, + "loss": 0.2213, + "step": 40450 + }, + { + "epoch": 14.94, + "learning_rate": 1.5071683704889766e-06, + "loss": 0.3109, + "step": 40451 + }, + { + "epoch": 14.94, + "learning_rate": 1.5071314201256313e-06, + "loss": 0.259, + "step": 40452 + }, + { + "epoch": 14.94, + "learning_rate": 1.5070944697622861e-06, + "loss": 0.3404, + "step": 40453 + }, + { + "epoch": 14.94, + "learning_rate": 1.5070575193989408e-06, + "loss": 0.2365, + "step": 40454 + }, + { + "epoch": 14.94, + "learning_rate": 1.5070205690355956e-06, + "loss": 0.3772, + "step": 40455 + }, + { + "epoch": 14.94, + "learning_rate": 1.5069836186722502e-06, + "loss": 0.2323, + "step": 40456 + }, + { + "epoch": 14.94, + "learning_rate": 1.5069466683089049e-06, + "loss": 0.1583, + "step": 40457 + }, + { + "epoch": 14.94, + "learning_rate": 1.50690971794556e-06, + "loss": 0.46, + "step": 40458 + }, + { + "epoch": 14.94, + "learning_rate": 1.5068727675822146e-06, + "loss": 0.2121, + "step": 40459 + }, + { + "epoch": 14.94, + "learning_rate": 1.5068358172188694e-06, + "loss": 0.2137, + "step": 40460 + }, + { + "epoch": 14.94, + "learning_rate": 1.506798866855524e-06, + "loss": 0.3431, + "step": 40461 + }, + { + "epoch": 14.94, + "learning_rate": 1.506761916492179e-06, + "loss": 0.2182, + "step": 40462 + }, + { + "epoch": 14.94, + "learning_rate": 1.5067249661288335e-06, + "loss": 0.268, + "step": 40463 + }, + { + "epoch": 14.94, + "learning_rate": 1.5066880157654886e-06, + "loss": 0.2856, + "step": 40464 + }, + { + "epoch": 14.94, + "learning_rate": 1.5066510654021432e-06, + "loss": 0.4036, + "step": 40465 + }, + { + "epoch": 14.94, + "learning_rate": 1.5066141150387977e-06, + "loss": 0.3372, + "step": 40466 + }, + { + "epoch": 14.94, + "learning_rate": 1.5065771646754527e-06, + "loss": 0.1988, + "step": 40467 + }, + { + "epoch": 14.94, + "learning_rate": 1.5065402143121074e-06, + "loss": 0.4462, + "step": 40468 + }, + { + "epoch": 14.94, + "learning_rate": 1.5065032639487622e-06, + "loss": 0.1654, + "step": 40469 + }, + { + "epoch": 14.94, + "learning_rate": 1.5064663135854169e-06, + "loss": 0.1712, + "step": 40470 + }, + { + "epoch": 14.94, + "learning_rate": 1.506429363222072e-06, + "loss": 0.2234, + "step": 40471 + }, + { + "epoch": 14.95, + "learning_rate": 1.5063924128587263e-06, + "loss": 0.2643, + "step": 40472 + }, + { + "epoch": 14.95, + "learning_rate": 1.5063554624953814e-06, + "loss": 0.387, + "step": 40473 + }, + { + "epoch": 14.95, + "learning_rate": 1.506318512132036e-06, + "loss": 0.3468, + "step": 40474 + }, + { + "epoch": 14.95, + "learning_rate": 1.5062815617686909e-06, + "loss": 0.2408, + "step": 40475 + }, + { + "epoch": 14.95, + "learning_rate": 1.5062446114053455e-06, + "loss": 0.3035, + "step": 40476 + }, + { + "epoch": 14.95, + "learning_rate": 1.5062076610420002e-06, + "loss": 0.3422, + "step": 40477 + }, + { + "epoch": 14.95, + "learning_rate": 1.506170710678655e-06, + "loss": 0.2139, + "step": 40478 + }, + { + "epoch": 14.95, + "learning_rate": 1.5061337603153096e-06, + "loss": 0.2389, + "step": 40479 + }, + { + "epoch": 14.95, + "learning_rate": 1.5060968099519647e-06, + "loss": 0.1987, + "step": 40480 + }, + { + "epoch": 14.95, + "learning_rate": 1.5060598595886193e-06, + "loss": 0.3905, + "step": 40481 + }, + { + "epoch": 14.95, + "learning_rate": 1.5060229092252742e-06, + "loss": 0.2555, + "step": 40482 + }, + { + "epoch": 14.95, + "learning_rate": 1.5059859588619288e-06, + "loss": 0.1601, + "step": 40483 + }, + { + "epoch": 14.95, + "learning_rate": 1.5059490084985837e-06, + "loss": 0.2439, + "step": 40484 + }, + { + "epoch": 14.95, + "learning_rate": 1.5059120581352383e-06, + "loss": 0.2032, + "step": 40485 + }, + { + "epoch": 14.95, + "learning_rate": 1.505875107771893e-06, + "loss": 0.2751, + "step": 40486 + }, + { + "epoch": 14.95, + "learning_rate": 1.505838157408548e-06, + "loss": 0.1851, + "step": 40487 + }, + { + "epoch": 14.95, + "learning_rate": 1.5058012070452024e-06, + "loss": 0.2208, + "step": 40488 + }, + { + "epoch": 14.95, + "learning_rate": 1.5057642566818575e-06, + "loss": 0.2226, + "step": 40489 + }, + { + "epoch": 14.95, + "learning_rate": 1.5057273063185121e-06, + "loss": 0.2283, + "step": 40490 + }, + { + "epoch": 14.95, + "learning_rate": 1.505690355955167e-06, + "loss": 0.2792, + "step": 40491 + }, + { + "epoch": 14.95, + "learning_rate": 1.5056534055918216e-06, + "loss": 0.2291, + "step": 40492 + }, + { + "epoch": 14.95, + "learning_rate": 1.5056164552284767e-06, + "loss": 0.1836, + "step": 40493 + }, + { + "epoch": 14.95, + "learning_rate": 1.505579504865131e-06, + "loss": 0.2391, + "step": 40494 + }, + { + "epoch": 14.95, + "learning_rate": 1.5055425545017861e-06, + "loss": 0.2141, + "step": 40495 + }, + { + "epoch": 14.95, + "learning_rate": 1.5055056041384408e-06, + "loss": 0.2028, + "step": 40496 + }, + { + "epoch": 14.95, + "learning_rate": 1.5054686537750954e-06, + "loss": 0.2931, + "step": 40497 + }, + { + "epoch": 14.95, + "learning_rate": 1.5054317034117503e-06, + "loss": 0.1862, + "step": 40498 + }, + { + "epoch": 14.96, + "learning_rate": 1.505394753048405e-06, + "loss": 0.2333, + "step": 40499 + }, + { + "epoch": 14.96, + "learning_rate": 1.5053578026850598e-06, + "loss": 0.2478, + "step": 40500 + }, + { + "epoch": 14.96, + "learning_rate": 1.5053208523217144e-06, + "loss": 0.2775, + "step": 40501 + }, + { + "epoch": 14.96, + "learning_rate": 1.5052839019583695e-06, + "loss": 0.2375, + "step": 40502 + }, + { + "epoch": 14.96, + "learning_rate": 1.505246951595024e-06, + "loss": 0.2078, + "step": 40503 + }, + { + "epoch": 14.96, + "learning_rate": 1.505210001231679e-06, + "loss": 0.2774, + "step": 40504 + }, + { + "epoch": 14.96, + "learning_rate": 1.5051730508683336e-06, + "loss": 0.2495, + "step": 40505 + }, + { + "epoch": 14.96, + "learning_rate": 1.5051361005049882e-06, + "loss": 0.3543, + "step": 40506 + }, + { + "epoch": 14.96, + "learning_rate": 1.505099150141643e-06, + "loss": 0.2731, + "step": 40507 + }, + { + "epoch": 14.96, + "learning_rate": 1.5050621997782977e-06, + "loss": 0.2594, + "step": 40508 + }, + { + "epoch": 14.96, + "learning_rate": 1.5050252494149528e-06, + "loss": 0.1419, + "step": 40509 + }, + { + "epoch": 14.96, + "learning_rate": 1.5049882990516074e-06, + "loss": 0.2334, + "step": 40510 + }, + { + "epoch": 14.96, + "learning_rate": 1.5049513486882622e-06, + "loss": 0.2394, + "step": 40511 + }, + { + "epoch": 14.96, + "learning_rate": 1.5049143983249169e-06, + "loss": 0.2916, + "step": 40512 + }, + { + "epoch": 14.96, + "learning_rate": 1.5048774479615717e-06, + "loss": 0.2027, + "step": 40513 + }, + { + "epoch": 14.96, + "learning_rate": 1.5048404975982264e-06, + "loss": 0.3782, + "step": 40514 + }, + { + "epoch": 14.96, + "learning_rate": 1.5048035472348814e-06, + "loss": 0.3945, + "step": 40515 + }, + { + "epoch": 14.96, + "learning_rate": 1.5047665968715358e-06, + "loss": 0.2892, + "step": 40516 + }, + { + "epoch": 14.96, + "learning_rate": 1.5047296465081905e-06, + "loss": 0.2199, + "step": 40517 + }, + { + "epoch": 14.96, + "learning_rate": 1.5046926961448455e-06, + "loss": 0.3138, + "step": 40518 + }, + { + "epoch": 14.96, + "learning_rate": 1.5046557457815002e-06, + "loss": 0.1696, + "step": 40519 + }, + { + "epoch": 14.96, + "learning_rate": 1.504618795418155e-06, + "loss": 0.2393, + "step": 40520 + }, + { + "epoch": 14.96, + "learning_rate": 1.5045818450548097e-06, + "loss": 0.2564, + "step": 40521 + }, + { + "epoch": 14.96, + "learning_rate": 1.5045448946914645e-06, + "loss": 0.268, + "step": 40522 + }, + { + "epoch": 14.96, + "learning_rate": 1.5045079443281191e-06, + "loss": 0.268, + "step": 40523 + }, + { + "epoch": 14.96, + "learning_rate": 1.5044709939647742e-06, + "loss": 0.328, + "step": 40524 + }, + { + "epoch": 14.96, + "learning_rate": 1.5044340436014288e-06, + "loss": 0.2418, + "step": 40525 + }, + { + "epoch": 14.97, + "learning_rate": 1.5043970932380835e-06, + "loss": 0.1454, + "step": 40526 + }, + { + "epoch": 14.97, + "learning_rate": 1.5043601428747383e-06, + "loss": 0.2876, + "step": 40527 + }, + { + "epoch": 14.97, + "learning_rate": 1.504323192511393e-06, + "loss": 0.1455, + "step": 40528 + }, + { + "epoch": 14.97, + "learning_rate": 1.5042862421480478e-06, + "loss": 0.359, + "step": 40529 + }, + { + "epoch": 14.97, + "learning_rate": 1.5042492917847024e-06, + "loss": 0.207, + "step": 40530 + }, + { + "epoch": 14.97, + "learning_rate": 1.5042123414213575e-06, + "loss": 0.311, + "step": 40531 + }, + { + "epoch": 14.97, + "learning_rate": 1.5041753910580121e-06, + "loss": 0.1831, + "step": 40532 + }, + { + "epoch": 14.97, + "learning_rate": 1.504138440694667e-06, + "loss": 0.3329, + "step": 40533 + }, + { + "epoch": 14.97, + "learning_rate": 1.5041014903313216e-06, + "loss": 0.2614, + "step": 40534 + }, + { + "epoch": 14.97, + "learning_rate": 1.5040645399679765e-06, + "loss": 0.3769, + "step": 40535 + }, + { + "epoch": 14.97, + "learning_rate": 1.5040275896046311e-06, + "loss": 0.2273, + "step": 40536 + }, + { + "epoch": 14.97, + "learning_rate": 1.5039906392412858e-06, + "loss": 0.4265, + "step": 40537 + }, + { + "epoch": 14.97, + "learning_rate": 1.5039536888779408e-06, + "loss": 0.2932, + "step": 40538 + }, + { + "epoch": 14.97, + "learning_rate": 1.5039167385145952e-06, + "loss": 0.1432, + "step": 40539 + }, + { + "epoch": 14.97, + "learning_rate": 1.5038797881512503e-06, + "loss": 0.3353, + "step": 40540 + }, + { + "epoch": 14.97, + "learning_rate": 1.503842837787905e-06, + "loss": 0.3003, + "step": 40541 + }, + { + "epoch": 14.97, + "learning_rate": 1.5038058874245598e-06, + "loss": 0.1513, + "step": 40542 + }, + { + "epoch": 14.97, + "learning_rate": 1.5037689370612144e-06, + "loss": 0.2065, + "step": 40543 + }, + { + "epoch": 14.97, + "learning_rate": 1.5037319866978695e-06, + "loss": 0.2151, + "step": 40544 + }, + { + "epoch": 14.97, + "learning_rate": 1.503695036334524e-06, + "loss": 0.3051, + "step": 40545 + }, + { + "epoch": 14.97, + "learning_rate": 1.5036580859711785e-06, + "loss": 0.3306, + "step": 40546 + }, + { + "epoch": 14.97, + "learning_rate": 1.5036211356078336e-06, + "loss": 0.2927, + "step": 40547 + }, + { + "epoch": 14.97, + "learning_rate": 1.5035841852444882e-06, + "loss": 0.2011, + "step": 40548 + }, + { + "epoch": 14.97, + "learning_rate": 1.503547234881143e-06, + "loss": 0.4463, + "step": 40549 + }, + { + "epoch": 14.97, + "learning_rate": 1.5035102845177977e-06, + "loss": 0.1989, + "step": 40550 + }, + { + "epoch": 14.97, + "learning_rate": 1.5034733341544526e-06, + "loss": 0.1518, + "step": 40551 + }, + { + "epoch": 14.97, + "learning_rate": 1.5034363837911072e-06, + "loss": 0.2325, + "step": 40552 + }, + { + "epoch": 14.98, + "learning_rate": 1.5033994334277623e-06, + "loss": 0.339, + "step": 40553 + }, + { + "epoch": 14.98, + "learning_rate": 1.503362483064417e-06, + "loss": 0.3848, + "step": 40554 + }, + { + "epoch": 14.98, + "learning_rate": 1.5033255327010717e-06, + "loss": 0.2783, + "step": 40555 + }, + { + "epoch": 14.98, + "learning_rate": 1.5032885823377264e-06, + "loss": 0.3522, + "step": 40556 + }, + { + "epoch": 14.98, + "learning_rate": 1.503251631974381e-06, + "loss": 0.2464, + "step": 40557 + }, + { + "epoch": 14.98, + "learning_rate": 1.5032146816110359e-06, + "loss": 0.3762, + "step": 40558 + }, + { + "epoch": 14.98, + "learning_rate": 1.5031777312476905e-06, + "loss": 0.1117, + "step": 40559 + }, + { + "epoch": 14.98, + "learning_rate": 1.5031407808843456e-06, + "loss": 0.2504, + "step": 40560 + }, + { + "epoch": 14.98, + "learning_rate": 1.503103830521e-06, + "loss": 0.2308, + "step": 40561 + }, + { + "epoch": 14.98, + "learning_rate": 1.503066880157655e-06, + "loss": 0.2369, + "step": 40562 + }, + { + "epoch": 14.98, + "learning_rate": 1.5030299297943097e-06, + "loss": 0.3549, + "step": 40563 + }, + { + "epoch": 14.98, + "learning_rate": 1.5029929794309645e-06, + "loss": 0.3234, + "step": 40564 + }, + { + "epoch": 14.98, + "learning_rate": 1.5029560290676192e-06, + "loss": 0.3694, + "step": 40565 + }, + { + "epoch": 14.98, + "learning_rate": 1.5029190787042738e-06, + "loss": 0.2114, + "step": 40566 + }, + { + "epoch": 14.98, + "learning_rate": 1.5028821283409287e-06, + "loss": 0.3536, + "step": 40567 + }, + { + "epoch": 14.98, + "learning_rate": 1.5028451779775833e-06, + "loss": 0.2604, + "step": 40568 + }, + { + "epoch": 14.98, + "learning_rate": 1.5028082276142384e-06, + "loss": 0.24, + "step": 40569 + }, + { + "epoch": 14.98, + "learning_rate": 1.502771277250893e-06, + "loss": 0.2253, + "step": 40570 + }, + { + "epoch": 14.98, + "learning_rate": 1.5027343268875478e-06, + "loss": 0.2354, + "step": 40571 + }, + { + "epoch": 14.98, + "learning_rate": 1.5026973765242025e-06, + "loss": 0.2172, + "step": 40572 + }, + { + "epoch": 14.98, + "learning_rate": 1.5026604261608573e-06, + "loss": 0.2079, + "step": 40573 + }, + { + "epoch": 14.98, + "learning_rate": 1.502623475797512e-06, + "loss": 0.2581, + "step": 40574 + }, + { + "epoch": 14.98, + "learning_rate": 1.502586525434167e-06, + "loss": 0.4067, + "step": 40575 + }, + { + "epoch": 14.98, + "learning_rate": 1.5025495750708217e-06, + "loss": 0.3255, + "step": 40576 + }, + { + "epoch": 14.98, + "learning_rate": 1.5025126247074763e-06, + "loss": 0.3331, + "step": 40577 + }, + { + "epoch": 14.98, + "learning_rate": 1.5024756743441311e-06, + "loss": 0.2481, + "step": 40578 + }, + { + "epoch": 14.98, + "learning_rate": 1.5024387239807858e-06, + "loss": 0.284, + "step": 40579 + }, + { + "epoch": 14.99, + "learning_rate": 1.5024017736174406e-06, + "loss": 0.2461, + "step": 40580 + }, + { + "epoch": 14.99, + "learning_rate": 1.5023648232540953e-06, + "loss": 0.1553, + "step": 40581 + }, + { + "epoch": 14.99, + "learning_rate": 1.5023278728907503e-06, + "loss": 0.1854, + "step": 40582 + }, + { + "epoch": 14.99, + "learning_rate": 1.502290922527405e-06, + "loss": 0.205, + "step": 40583 + }, + { + "epoch": 14.99, + "learning_rate": 1.5022539721640598e-06, + "loss": 0.2285, + "step": 40584 + }, + { + "epoch": 14.99, + "learning_rate": 1.5022170218007144e-06, + "loss": 0.303, + "step": 40585 + }, + { + "epoch": 14.99, + "learning_rate": 1.502180071437369e-06, + "loss": 0.2853, + "step": 40586 + }, + { + "epoch": 14.99, + "learning_rate": 1.502143121074024e-06, + "loss": 0.233, + "step": 40587 + }, + { + "epoch": 14.99, + "learning_rate": 1.5021061707106786e-06, + "loss": 0.1783, + "step": 40588 + }, + { + "epoch": 14.99, + "learning_rate": 1.5020692203473336e-06, + "loss": 0.2054, + "step": 40589 + }, + { + "epoch": 14.99, + "learning_rate": 1.502032269983988e-06, + "loss": 0.5246, + "step": 40590 + }, + { + "epoch": 14.99, + "learning_rate": 1.5019953196206431e-06, + "loss": 0.2372, + "step": 40591 + }, + { + "epoch": 14.99, + "learning_rate": 1.5019583692572977e-06, + "loss": 0.2504, + "step": 40592 + }, + { + "epoch": 14.99, + "learning_rate": 1.5019214188939526e-06, + "loss": 0.2101, + "step": 40593 + }, + { + "epoch": 14.99, + "learning_rate": 1.5018844685306072e-06, + "loss": 0.2554, + "step": 40594 + }, + { + "epoch": 14.99, + "learning_rate": 1.501847518167262e-06, + "loss": 0.1924, + "step": 40595 + }, + { + "epoch": 14.99, + "learning_rate": 1.5018105678039167e-06, + "loss": 0.2876, + "step": 40596 + }, + { + "epoch": 14.99, + "learning_rate": 1.5017736174405713e-06, + "loss": 0.2679, + "step": 40597 + }, + { + "epoch": 14.99, + "learning_rate": 1.5017366670772264e-06, + "loss": 0.272, + "step": 40598 + }, + { + "epoch": 14.99, + "learning_rate": 1.501699716713881e-06, + "loss": 0.3306, + "step": 40599 + }, + { + "epoch": 14.99, + "learning_rate": 1.5016627663505359e-06, + "loss": 0.1601, + "step": 40600 + }, + { + "epoch": 14.99, + "learning_rate": 1.5016258159871905e-06, + "loss": 0.3178, + "step": 40601 + }, + { + "epoch": 14.99, + "learning_rate": 1.5015888656238454e-06, + "loss": 0.2695, + "step": 40602 + }, + { + "epoch": 14.99, + "learning_rate": 1.5015519152605e-06, + "loss": 0.2289, + "step": 40603 + }, + { + "epoch": 14.99, + "learning_rate": 1.501514964897155e-06, + "loss": 0.1823, + "step": 40604 + }, + { + "epoch": 14.99, + "learning_rate": 1.5014780145338097e-06, + "loss": 0.2321, + "step": 40605 + }, + { + "epoch": 14.99, + "learning_rate": 1.5014410641704641e-06, + "loss": 0.2784, + "step": 40606 + }, + { + "epoch": 15.0, + "learning_rate": 1.5014041138071192e-06, + "loss": 0.3269, + "step": 40607 + }, + { + "epoch": 15.0, + "learning_rate": 1.5013671634437738e-06, + "loss": 0.2613, + "step": 40608 + }, + { + "epoch": 15.0, + "learning_rate": 1.5013302130804287e-06, + "loss": 0.2254, + "step": 40609 + }, + { + "epoch": 15.0, + "learning_rate": 1.5012932627170833e-06, + "loss": 0.257, + "step": 40610 + }, + { + "epoch": 15.0, + "learning_rate": 1.5012563123537384e-06, + "loss": 0.2341, + "step": 40611 + }, + { + "epoch": 15.0, + "learning_rate": 1.5012193619903928e-06, + "loss": 0.2255, + "step": 40612 + }, + { + "epoch": 15.0, + "learning_rate": 1.5011824116270479e-06, + "loss": 0.203, + "step": 40613 + }, + { + "epoch": 15.0, + "learning_rate": 1.5011454612637025e-06, + "loss": 0.2113, + "step": 40614 + }, + { + "epoch": 15.0, + "learning_rate": 1.5011085109003573e-06, + "loss": 0.2189, + "step": 40615 + }, + { + "epoch": 15.0, + "learning_rate": 1.501071560537012e-06, + "loss": 0.3511, + "step": 40616 + }, + { + "epoch": 15.0, + "learning_rate": 1.5010346101736666e-06, + "loss": 0.2565, + "step": 40617 + }, + { + "epoch": 15.0, + "learning_rate": 1.5009976598103215e-06, + "loss": 0.3033, + "step": 40618 + }, + { + "epoch": 15.0, + "learning_rate": 1.500960709446976e-06, + "loss": 0.3359, + "step": 40619 + }, + { + "epoch": 15.0, + "learning_rate": 1.5009237590836312e-06, + "loss": 0.1488, + "step": 40620 + }, + { + "epoch": 15.0, + "eval_accuracy": 0.850371704298841, + "eval_loss": 0.3722574710845947, + "eval_runtime": 25.3684, + "eval_samples_per_second": 1707.402, + "eval_steps_per_second": 26.687, + "step": 40620 + }, + { + "epoch": 15.0, + "learning_rate": 1.5008868087202858e-06, + "loss": 0.2532, + "step": 40621 + }, + { + "epoch": 15.0, + "learning_rate": 1.5008498583569406e-06, + "loss": 0.2482, + "step": 40622 + }, + { + "epoch": 15.0, + "learning_rate": 1.5008129079935953e-06, + "loss": 0.3293, + "step": 40623 + }, + { + "epoch": 15.0, + "learning_rate": 1.5007759576302501e-06, + "loss": 0.2325, + "step": 40624 + }, + { + "epoch": 15.0, + "learning_rate": 1.5007390072669048e-06, + "loss": 0.3203, + "step": 40625 + }, + { + "epoch": 15.0, + "learning_rate": 1.5007020569035594e-06, + "loss": 0.2187, + "step": 40626 + }, + { + "epoch": 15.0, + "learning_rate": 1.5006651065402145e-06, + "loss": 0.3509, + "step": 40627 + }, + { + "epoch": 15.0, + "learning_rate": 1.5006281561768689e-06, + "loss": 0.1518, + "step": 40628 + }, + { + "epoch": 15.0, + "learning_rate": 1.500591205813524e-06, + "loss": 0.3216, + "step": 40629 + }, + { + "epoch": 15.0, + "learning_rate": 1.5005542554501786e-06, + "loss": 0.2493, + "step": 40630 + }, + { + "epoch": 15.0, + "learning_rate": 1.5005173050868334e-06, + "loss": 0.3432, + "step": 40631 + }, + { + "epoch": 15.0, + "learning_rate": 1.500480354723488e-06, + "loss": 0.202, + "step": 40632 + }, + { + "epoch": 15.0, + "learning_rate": 1.5004434043601431e-06, + "loss": 0.3009, + "step": 40633 + }, + { + "epoch": 15.01, + "learning_rate": 1.5004064539967976e-06, + "loss": 0.2698, + "step": 40634 + }, + { + "epoch": 15.01, + "learning_rate": 1.5003695036334526e-06, + "loss": 0.113, + "step": 40635 + }, + { + "epoch": 15.01, + "learning_rate": 1.5003325532701073e-06, + "loss": 0.2555, + "step": 40636 + }, + { + "epoch": 15.01, + "learning_rate": 1.5002956029067619e-06, + "loss": 0.3587, + "step": 40637 + }, + { + "epoch": 15.01, + "learning_rate": 1.5002586525434167e-06, + "loss": 0.2763, + "step": 40638 + }, + { + "epoch": 15.01, + "learning_rate": 1.5002217021800714e-06, + "loss": 0.2062, + "step": 40639 + }, + { + "epoch": 15.01, + "learning_rate": 1.5001847518167262e-06, + "loss": 0.2624, + "step": 40640 + }, + { + "epoch": 15.01, + "learning_rate": 1.5001478014533809e-06, + "loss": 0.2331, + "step": 40641 + }, + { + "epoch": 15.01, + "learning_rate": 1.500110851090036e-06, + "loss": 0.1219, + "step": 40642 + }, + { + "epoch": 15.01, + "learning_rate": 1.5000739007266906e-06, + "loss": 0.2465, + "step": 40643 + }, + { + "epoch": 15.01, + "learning_rate": 1.5000369503633454e-06, + "loss": 0.43, + "step": 40644 + }, + { + "epoch": 15.01, + "learning_rate": 1.5e-06, + "loss": 0.1844, + "step": 40645 + }, + { + "epoch": 15.01, + "learning_rate": 1.4999630496366549e-06, + "loss": 0.2877, + "step": 40646 + }, + { + "epoch": 15.01, + "learning_rate": 1.4999260992733095e-06, + "loss": 0.2282, + "step": 40647 + }, + { + "epoch": 15.01, + "learning_rate": 1.4998891489099644e-06, + "loss": 0.2621, + "step": 40648 + }, + { + "epoch": 15.01, + "learning_rate": 1.4998521985466192e-06, + "loss": 0.2249, + "step": 40649 + }, + { + "epoch": 15.01, + "learning_rate": 1.4998152481832739e-06, + "loss": 0.2159, + "step": 40650 + }, + { + "epoch": 15.01, + "learning_rate": 1.4997782978199285e-06, + "loss": 0.2333, + "step": 40651 + }, + { + "epoch": 15.01, + "learning_rate": 1.4997413474565833e-06, + "loss": 0.1885, + "step": 40652 + }, + { + "epoch": 15.01, + "learning_rate": 1.4997043970932382e-06, + "loss": 0.2518, + "step": 40653 + }, + { + "epoch": 15.01, + "learning_rate": 1.4996674467298928e-06, + "loss": 0.1525, + "step": 40654 + }, + { + "epoch": 15.01, + "learning_rate": 1.4996304963665477e-06, + "loss": 0.1894, + "step": 40655 + }, + { + "epoch": 15.01, + "learning_rate": 1.4995935460032025e-06, + "loss": 0.3088, + "step": 40656 + }, + { + "epoch": 15.01, + "learning_rate": 1.4995565956398572e-06, + "loss": 0.1366, + "step": 40657 + }, + { + "epoch": 15.01, + "learning_rate": 1.499519645276512e-06, + "loss": 0.3271, + "step": 40658 + }, + { + "epoch": 15.01, + "learning_rate": 1.4994826949131669e-06, + "loss": 0.337, + "step": 40659 + }, + { + "epoch": 15.01, + "learning_rate": 1.4994457445498213e-06, + "loss": 0.2527, + "step": 40660 + }, + { + "epoch": 15.02, + "learning_rate": 1.4994087941864761e-06, + "loss": 0.1904, + "step": 40661 + }, + { + "epoch": 15.02, + "learning_rate": 1.499371843823131e-06, + "loss": 0.2465, + "step": 40662 + }, + { + "epoch": 15.02, + "learning_rate": 1.4993348934597856e-06, + "loss": 0.1582, + "step": 40663 + }, + { + "epoch": 15.02, + "learning_rate": 1.4992979430964405e-06, + "loss": 0.2626, + "step": 40664 + }, + { + "epoch": 15.02, + "learning_rate": 1.4992609927330953e-06, + "loss": 0.1747, + "step": 40665 + }, + { + "epoch": 15.02, + "learning_rate": 1.49922404236975e-06, + "loss": 0.272, + "step": 40666 + }, + { + "epoch": 15.02, + "learning_rate": 1.4991870920064048e-06, + "loss": 0.2801, + "step": 40667 + }, + { + "epoch": 15.02, + "learning_rate": 1.4991501416430596e-06, + "loss": 0.1903, + "step": 40668 + }, + { + "epoch": 15.02, + "learning_rate": 1.4991131912797143e-06, + "loss": 0.2148, + "step": 40669 + }, + { + "epoch": 15.02, + "learning_rate": 1.499076240916369e-06, + "loss": 0.1903, + "step": 40670 + }, + { + "epoch": 15.02, + "learning_rate": 1.4990392905530238e-06, + "loss": 0.2902, + "step": 40671 + }, + { + "epoch": 15.02, + "learning_rate": 1.4990023401896786e-06, + "loss": 0.1867, + "step": 40672 + }, + { + "epoch": 15.02, + "learning_rate": 1.4989653898263332e-06, + "loss": 0.4002, + "step": 40673 + }, + { + "epoch": 15.02, + "learning_rate": 1.498928439462988e-06, + "loss": 0.2226, + "step": 40674 + }, + { + "epoch": 15.02, + "learning_rate": 1.498891489099643e-06, + "loss": 0.2699, + "step": 40675 + }, + { + "epoch": 15.02, + "learning_rate": 1.4988545387362976e-06, + "loss": 0.2429, + "step": 40676 + }, + { + "epoch": 15.02, + "learning_rate": 1.4988175883729524e-06, + "loss": 0.2176, + "step": 40677 + }, + { + "epoch": 15.02, + "learning_rate": 1.4987806380096073e-06, + "loss": 0.337, + "step": 40678 + }, + { + "epoch": 15.02, + "learning_rate": 1.498743687646262e-06, + "loss": 0.1835, + "step": 40679 + }, + { + "epoch": 15.02, + "learning_rate": 1.4987067372829165e-06, + "loss": 0.2758, + "step": 40680 + }, + { + "epoch": 15.02, + "learning_rate": 1.4986697869195714e-06, + "loss": 0.1562, + "step": 40681 + }, + { + "epoch": 15.02, + "learning_rate": 1.498632836556226e-06, + "loss": 0.2002, + "step": 40682 + }, + { + "epoch": 15.02, + "learning_rate": 1.4985958861928809e-06, + "loss": 0.4367, + "step": 40683 + }, + { + "epoch": 15.02, + "learning_rate": 1.4985589358295357e-06, + "loss": 0.2722, + "step": 40684 + }, + { + "epoch": 15.02, + "learning_rate": 1.4985219854661904e-06, + "loss": 0.3163, + "step": 40685 + }, + { + "epoch": 15.02, + "learning_rate": 1.4984850351028452e-06, + "loss": 0.2893, + "step": 40686 + }, + { + "epoch": 15.02, + "learning_rate": 1.4984480847395e-06, + "loss": 0.1867, + "step": 40687 + }, + { + "epoch": 15.03, + "learning_rate": 1.4984111343761547e-06, + "loss": 0.1966, + "step": 40688 + }, + { + "epoch": 15.03, + "learning_rate": 1.4983741840128095e-06, + "loss": 0.2042, + "step": 40689 + }, + { + "epoch": 15.03, + "learning_rate": 1.4983372336494642e-06, + "loss": 0.2099, + "step": 40690 + }, + { + "epoch": 15.03, + "learning_rate": 1.498300283286119e-06, + "loss": 0.2496, + "step": 40691 + }, + { + "epoch": 15.03, + "learning_rate": 1.4982633329227737e-06, + "loss": 0.3477, + "step": 40692 + }, + { + "epoch": 15.03, + "learning_rate": 1.4982263825594285e-06, + "loss": 0.1523, + "step": 40693 + }, + { + "epoch": 15.03, + "learning_rate": 1.4981894321960834e-06, + "loss": 0.2246, + "step": 40694 + }, + { + "epoch": 15.03, + "learning_rate": 1.498152481832738e-06, + "loss": 0.3402, + "step": 40695 + }, + { + "epoch": 15.03, + "learning_rate": 1.4981155314693928e-06, + "loss": 0.2638, + "step": 40696 + }, + { + "epoch": 15.03, + "learning_rate": 1.4980785811060477e-06, + "loss": 0.3849, + "step": 40697 + }, + { + "epoch": 15.03, + "learning_rate": 1.4980416307427023e-06, + "loss": 0.2506, + "step": 40698 + }, + { + "epoch": 15.03, + "learning_rate": 1.4980046803793572e-06, + "loss": 0.3088, + "step": 40699 + }, + { + "epoch": 15.03, + "learning_rate": 1.4979677300160118e-06, + "loss": 0.2524, + "step": 40700 + }, + { + "epoch": 15.03, + "learning_rate": 1.4979307796526665e-06, + "loss": 0.2464, + "step": 40701 + }, + { + "epoch": 15.03, + "learning_rate": 1.4978938292893213e-06, + "loss": 0.2869, + "step": 40702 + }, + { + "epoch": 15.03, + "learning_rate": 1.4978568789259762e-06, + "loss": 0.1841, + "step": 40703 + }, + { + "epoch": 15.03, + "learning_rate": 1.4978199285626308e-06, + "loss": 0.3576, + "step": 40704 + }, + { + "epoch": 15.03, + "learning_rate": 1.4977829781992856e-06, + "loss": 0.3808, + "step": 40705 + }, + { + "epoch": 15.03, + "learning_rate": 1.4977460278359405e-06, + "loss": 0.1322, + "step": 40706 + }, + { + "epoch": 15.03, + "learning_rate": 1.4977090774725951e-06, + "loss": 0.3318, + "step": 40707 + }, + { + "epoch": 15.03, + "learning_rate": 1.49767212710925e-06, + "loss": 0.2487, + "step": 40708 + }, + { + "epoch": 15.03, + "learning_rate": 1.4976351767459048e-06, + "loss": 0.2958, + "step": 40709 + }, + { + "epoch": 15.03, + "learning_rate": 1.4975982263825595e-06, + "loss": 0.2361, + "step": 40710 + }, + { + "epoch": 15.03, + "learning_rate": 1.497561276019214e-06, + "loss": 0.2272, + "step": 40711 + }, + { + "epoch": 15.03, + "learning_rate": 1.497524325655869e-06, + "loss": 0.2434, + "step": 40712 + }, + { + "epoch": 15.03, + "learning_rate": 1.4974873752925238e-06, + "loss": 0.433, + "step": 40713 + }, + { + "epoch": 15.03, + "learning_rate": 1.4974504249291784e-06, + "loss": 0.4682, + "step": 40714 + }, + { + "epoch": 15.04, + "learning_rate": 1.4974134745658333e-06, + "loss": 0.2519, + "step": 40715 + }, + { + "epoch": 15.04, + "learning_rate": 1.4973765242024881e-06, + "loss": 0.28, + "step": 40716 + }, + { + "epoch": 15.04, + "learning_rate": 1.4973395738391428e-06, + "loss": 0.2693, + "step": 40717 + }, + { + "epoch": 15.04, + "learning_rate": 1.4973026234757976e-06, + "loss": 0.2475, + "step": 40718 + }, + { + "epoch": 15.04, + "learning_rate": 1.4972656731124525e-06, + "loss": 0.2343, + "step": 40719 + }, + { + "epoch": 15.04, + "learning_rate": 1.497228722749107e-06, + "loss": 0.229, + "step": 40720 + }, + { + "epoch": 15.04, + "learning_rate": 1.4971917723857617e-06, + "loss": 0.1583, + "step": 40721 + }, + { + "epoch": 15.04, + "learning_rate": 1.4971548220224166e-06, + "loss": 0.1865, + "step": 40722 + }, + { + "epoch": 15.04, + "learning_rate": 1.4971178716590714e-06, + "loss": 0.237, + "step": 40723 + }, + { + "epoch": 15.04, + "learning_rate": 1.497080921295726e-06, + "loss": 0.2007, + "step": 40724 + }, + { + "epoch": 15.04, + "learning_rate": 1.497043970932381e-06, + "loss": 0.3365, + "step": 40725 + }, + { + "epoch": 15.04, + "learning_rate": 1.4970070205690358e-06, + "loss": 0.1503, + "step": 40726 + }, + { + "epoch": 15.04, + "learning_rate": 1.4969700702056904e-06, + "loss": 0.2183, + "step": 40727 + }, + { + "epoch": 15.04, + "learning_rate": 1.4969331198423452e-06, + "loss": 0.1639, + "step": 40728 + }, + { + "epoch": 15.04, + "learning_rate": 1.496896169479e-06, + "loss": 0.3214, + "step": 40729 + }, + { + "epoch": 15.04, + "learning_rate": 1.4968592191156545e-06, + "loss": 0.2355, + "step": 40730 + }, + { + "epoch": 15.04, + "learning_rate": 1.4968222687523094e-06, + "loss": 0.1778, + "step": 40731 + }, + { + "epoch": 15.04, + "learning_rate": 1.4967853183889642e-06, + "loss": 0.1664, + "step": 40732 + }, + { + "epoch": 15.04, + "learning_rate": 1.4967483680256188e-06, + "loss": 0.3458, + "step": 40733 + }, + { + "epoch": 15.04, + "learning_rate": 1.4967114176622737e-06, + "loss": 0.1699, + "step": 40734 + }, + { + "epoch": 15.04, + "learning_rate": 1.4966744672989285e-06, + "loss": 0.2706, + "step": 40735 + }, + { + "epoch": 15.04, + "learning_rate": 1.4966375169355832e-06, + "loss": 0.2195, + "step": 40736 + }, + { + "epoch": 15.04, + "learning_rate": 1.496600566572238e-06, + "loss": 0.3923, + "step": 40737 + }, + { + "epoch": 15.04, + "learning_rate": 1.4965636162088929e-06, + "loss": 0.1543, + "step": 40738 + }, + { + "epoch": 15.04, + "learning_rate": 1.4965266658455475e-06, + "loss": 0.1899, + "step": 40739 + }, + { + "epoch": 15.04, + "learning_rate": 1.4964897154822021e-06, + "loss": 0.1814, + "step": 40740 + }, + { + "epoch": 15.04, + "learning_rate": 1.496452765118857e-06, + "loss": 0.2735, + "step": 40741 + }, + { + "epoch": 15.05, + "learning_rate": 1.4964158147555118e-06, + "loss": 0.3727, + "step": 40742 + }, + { + "epoch": 15.05, + "learning_rate": 1.4963788643921665e-06, + "loss": 0.2819, + "step": 40743 + }, + { + "epoch": 15.05, + "learning_rate": 1.4963419140288213e-06, + "loss": 0.3293, + "step": 40744 + }, + { + "epoch": 15.05, + "learning_rate": 1.4963049636654762e-06, + "loss": 0.2641, + "step": 40745 + }, + { + "epoch": 15.05, + "learning_rate": 1.4962680133021308e-06, + "loss": 0.2087, + "step": 40746 + }, + { + "epoch": 15.05, + "learning_rate": 1.4962310629387857e-06, + "loss": 0.2404, + "step": 40747 + }, + { + "epoch": 15.05, + "learning_rate": 1.4961941125754405e-06, + "loss": 0.366, + "step": 40748 + }, + { + "epoch": 15.05, + "learning_rate": 1.4961571622120951e-06, + "loss": 0.1267, + "step": 40749 + }, + { + "epoch": 15.05, + "learning_rate": 1.4961202118487498e-06, + "loss": 0.3015, + "step": 40750 + }, + { + "epoch": 15.05, + "learning_rate": 1.4960832614854046e-06, + "loss": 0.1927, + "step": 40751 + }, + { + "epoch": 15.05, + "learning_rate": 1.4960463111220593e-06, + "loss": 0.3621, + "step": 40752 + }, + { + "epoch": 15.05, + "learning_rate": 1.4960093607587141e-06, + "loss": 0.2704, + "step": 40753 + }, + { + "epoch": 15.05, + "learning_rate": 1.495972410395369e-06, + "loss": 0.338, + "step": 40754 + }, + { + "epoch": 15.05, + "learning_rate": 1.4959354600320236e-06, + "loss": 0.2992, + "step": 40755 + }, + { + "epoch": 15.05, + "learning_rate": 1.4958985096686784e-06, + "loss": 0.2211, + "step": 40756 + }, + { + "epoch": 15.05, + "learning_rate": 1.4958615593053333e-06, + "loss": 0.3795, + "step": 40757 + }, + { + "epoch": 15.05, + "learning_rate": 1.495824608941988e-06, + "loss": 0.2948, + "step": 40758 + }, + { + "epoch": 15.05, + "learning_rate": 1.4957876585786428e-06, + "loss": 0.2244, + "step": 40759 + }, + { + "epoch": 15.05, + "learning_rate": 1.4957507082152974e-06, + "loss": 0.2524, + "step": 40760 + }, + { + "epoch": 15.05, + "learning_rate": 1.4957137578519523e-06, + "loss": 0.3471, + "step": 40761 + }, + { + "epoch": 15.05, + "learning_rate": 1.495676807488607e-06, + "loss": 0.1816, + "step": 40762 + }, + { + "epoch": 15.05, + "learning_rate": 1.4956398571252617e-06, + "loss": 0.3172, + "step": 40763 + }, + { + "epoch": 15.05, + "learning_rate": 1.4956029067619166e-06, + "loss": 0.3149, + "step": 40764 + }, + { + "epoch": 15.05, + "learning_rate": 1.4955659563985712e-06, + "loss": 0.2578, + "step": 40765 + }, + { + "epoch": 15.05, + "learning_rate": 1.495529006035226e-06, + "loss": 0.3259, + "step": 40766 + }, + { + "epoch": 15.05, + "learning_rate": 1.495492055671881e-06, + "loss": 0.2254, + "step": 40767 + }, + { + "epoch": 15.05, + "learning_rate": 1.4954551053085356e-06, + "loss": 0.1717, + "step": 40768 + }, + { + "epoch": 15.06, + "learning_rate": 1.4954181549451904e-06, + "loss": 0.2659, + "step": 40769 + }, + { + "epoch": 15.06, + "learning_rate": 1.495381204581845e-06, + "loss": 0.3576, + "step": 40770 + }, + { + "epoch": 15.06, + "learning_rate": 1.4953442542185e-06, + "loss": 0.1739, + "step": 40771 + }, + { + "epoch": 15.06, + "learning_rate": 1.4953073038551545e-06, + "loss": 0.1923, + "step": 40772 + }, + { + "epoch": 15.06, + "learning_rate": 1.4952703534918094e-06, + "loss": 0.2459, + "step": 40773 + }, + { + "epoch": 15.06, + "learning_rate": 1.4952334031284642e-06, + "loss": 0.2737, + "step": 40774 + }, + { + "epoch": 15.06, + "learning_rate": 1.4951964527651189e-06, + "loss": 0.4499, + "step": 40775 + }, + { + "epoch": 15.06, + "learning_rate": 1.4951595024017737e-06, + "loss": 0.2169, + "step": 40776 + }, + { + "epoch": 15.06, + "learning_rate": 1.4951225520384284e-06, + "loss": 0.2006, + "step": 40777 + }, + { + "epoch": 15.06, + "learning_rate": 1.4950856016750832e-06, + "loss": 0.3216, + "step": 40778 + }, + { + "epoch": 15.06, + "learning_rate": 1.495048651311738e-06, + "loss": 0.2538, + "step": 40779 + }, + { + "epoch": 15.06, + "learning_rate": 1.4950117009483927e-06, + "loss": 0.2456, + "step": 40780 + }, + { + "epoch": 15.06, + "learning_rate": 1.4949747505850473e-06, + "loss": 0.3536, + "step": 40781 + }, + { + "epoch": 15.06, + "learning_rate": 1.4949378002217022e-06, + "loss": 0.2563, + "step": 40782 + }, + { + "epoch": 15.06, + "learning_rate": 1.494900849858357e-06, + "loss": 0.2217, + "step": 40783 + }, + { + "epoch": 15.06, + "learning_rate": 1.4948638994950117e-06, + "loss": 0.3116, + "step": 40784 + }, + { + "epoch": 15.06, + "learning_rate": 1.4948269491316665e-06, + "loss": 0.2051, + "step": 40785 + }, + { + "epoch": 15.06, + "learning_rate": 1.4947899987683214e-06, + "loss": 0.2978, + "step": 40786 + }, + { + "epoch": 15.06, + "learning_rate": 1.494753048404976e-06, + "loss": 0.2866, + "step": 40787 + }, + { + "epoch": 15.06, + "learning_rate": 1.4947160980416308e-06, + "loss": 0.3418, + "step": 40788 + }, + { + "epoch": 15.06, + "learning_rate": 1.4946791476782857e-06, + "loss": 0.2806, + "step": 40789 + }, + { + "epoch": 15.06, + "learning_rate": 1.4946421973149403e-06, + "loss": 0.1479, + "step": 40790 + }, + { + "epoch": 15.06, + "learning_rate": 1.494605246951595e-06, + "loss": 0.2575, + "step": 40791 + }, + { + "epoch": 15.06, + "learning_rate": 1.4945682965882498e-06, + "loss": 0.3402, + "step": 40792 + }, + { + "epoch": 15.06, + "learning_rate": 1.4945313462249047e-06, + "loss": 0.1912, + "step": 40793 + }, + { + "epoch": 15.06, + "learning_rate": 1.4944943958615593e-06, + "loss": 0.1998, + "step": 40794 + }, + { + "epoch": 15.06, + "learning_rate": 1.4944574454982141e-06, + "loss": 0.2062, + "step": 40795 + }, + { + "epoch": 15.06, + "learning_rate": 1.494420495134869e-06, + "loss": 0.2523, + "step": 40796 + }, + { + "epoch": 15.07, + "learning_rate": 1.4943835447715236e-06, + "loss": 0.3578, + "step": 40797 + }, + { + "epoch": 15.07, + "learning_rate": 1.4943465944081785e-06, + "loss": 0.2369, + "step": 40798 + }, + { + "epoch": 15.07, + "learning_rate": 1.4943096440448333e-06, + "loss": 0.2846, + "step": 40799 + }, + { + "epoch": 15.07, + "learning_rate": 1.4942726936814877e-06, + "loss": 0.1781, + "step": 40800 + }, + { + "epoch": 15.07, + "learning_rate": 1.4942357433181426e-06, + "loss": 0.2424, + "step": 40801 + }, + { + "epoch": 15.07, + "learning_rate": 1.4941987929547974e-06, + "loss": 0.2947, + "step": 40802 + }, + { + "epoch": 15.07, + "learning_rate": 1.494161842591452e-06, + "loss": 0.2227, + "step": 40803 + }, + { + "epoch": 15.07, + "learning_rate": 1.494124892228107e-06, + "loss": 0.3097, + "step": 40804 + }, + { + "epoch": 15.07, + "learning_rate": 1.4940879418647618e-06, + "loss": 0.2454, + "step": 40805 + }, + { + "epoch": 15.07, + "learning_rate": 1.4940509915014164e-06, + "loss": 0.2616, + "step": 40806 + }, + { + "epoch": 15.07, + "learning_rate": 1.4940140411380713e-06, + "loss": 0.1988, + "step": 40807 + }, + { + "epoch": 15.07, + "learning_rate": 1.493977090774726e-06, + "loss": 0.3666, + "step": 40808 + }, + { + "epoch": 15.07, + "learning_rate": 1.4939401404113807e-06, + "loss": 0.2718, + "step": 40809 + }, + { + "epoch": 15.07, + "learning_rate": 1.4939031900480354e-06, + "loss": 0.2792, + "step": 40810 + }, + { + "epoch": 15.07, + "learning_rate": 1.4938662396846902e-06, + "loss": 0.2465, + "step": 40811 + }, + { + "epoch": 15.07, + "learning_rate": 1.493829289321345e-06, + "loss": 0.1963, + "step": 40812 + }, + { + "epoch": 15.07, + "learning_rate": 1.4937923389579997e-06, + "loss": 0.2659, + "step": 40813 + }, + { + "epoch": 15.07, + "learning_rate": 1.4937553885946546e-06, + "loss": 0.2178, + "step": 40814 + }, + { + "epoch": 15.07, + "learning_rate": 1.4937184382313094e-06, + "loss": 0.2726, + "step": 40815 + }, + { + "epoch": 15.07, + "learning_rate": 1.493681487867964e-06, + "loss": 0.3151, + "step": 40816 + }, + { + "epoch": 15.07, + "learning_rate": 1.4936445375046189e-06, + "loss": 0.2919, + "step": 40817 + }, + { + "epoch": 15.07, + "learning_rate": 1.4936075871412737e-06, + "loss": 0.3435, + "step": 40818 + }, + { + "epoch": 15.07, + "learning_rate": 1.4935706367779284e-06, + "loss": 0.3063, + "step": 40819 + }, + { + "epoch": 15.07, + "learning_rate": 1.493533686414583e-06, + "loss": 0.2942, + "step": 40820 + }, + { + "epoch": 15.07, + "learning_rate": 1.4934967360512379e-06, + "loss": 0.1908, + "step": 40821 + }, + { + "epoch": 15.07, + "learning_rate": 1.4934597856878925e-06, + "loss": 0.2601, + "step": 40822 + }, + { + "epoch": 15.07, + "learning_rate": 1.4934228353245473e-06, + "loss": 0.2128, + "step": 40823 + }, + { + "epoch": 15.08, + "learning_rate": 1.4933858849612022e-06, + "loss": 0.3249, + "step": 40824 + }, + { + "epoch": 15.08, + "learning_rate": 1.4933489345978568e-06, + "loss": 0.2287, + "step": 40825 + }, + { + "epoch": 15.08, + "learning_rate": 1.4933119842345117e-06, + "loss": 0.2674, + "step": 40826 + }, + { + "epoch": 15.08, + "learning_rate": 1.4932750338711665e-06, + "loss": 0.2243, + "step": 40827 + }, + { + "epoch": 15.08, + "learning_rate": 1.4932380835078212e-06, + "loss": 0.2649, + "step": 40828 + }, + { + "epoch": 15.08, + "learning_rate": 1.493201133144476e-06, + "loss": 0.2643, + "step": 40829 + }, + { + "epoch": 15.08, + "learning_rate": 1.4931641827811306e-06, + "loss": 0.3947, + "step": 40830 + }, + { + "epoch": 15.08, + "learning_rate": 1.4931272324177855e-06, + "loss": 0.2477, + "step": 40831 + }, + { + "epoch": 15.08, + "learning_rate": 1.4930902820544401e-06, + "loss": 0.2102, + "step": 40832 + }, + { + "epoch": 15.08, + "learning_rate": 1.493053331691095e-06, + "loss": 0.2195, + "step": 40833 + }, + { + "epoch": 15.08, + "learning_rate": 1.4930163813277498e-06, + "loss": 0.3324, + "step": 40834 + }, + { + "epoch": 15.08, + "learning_rate": 1.4929794309644045e-06, + "loss": 0.1352, + "step": 40835 + }, + { + "epoch": 15.08, + "learning_rate": 1.4929424806010593e-06, + "loss": 0.3792, + "step": 40836 + }, + { + "epoch": 15.08, + "learning_rate": 1.4929055302377142e-06, + "loss": 0.3437, + "step": 40837 + }, + { + "epoch": 15.08, + "learning_rate": 1.4928685798743688e-06, + "loss": 0.1889, + "step": 40838 + }, + { + "epoch": 15.08, + "learning_rate": 1.4928316295110236e-06, + "loss": 0.234, + "step": 40839 + }, + { + "epoch": 15.08, + "learning_rate": 1.4927946791476783e-06, + "loss": 0.1948, + "step": 40840 + }, + { + "epoch": 15.08, + "learning_rate": 1.4927577287843331e-06, + "loss": 0.157, + "step": 40841 + }, + { + "epoch": 15.08, + "learning_rate": 1.4927207784209878e-06, + "loss": 0.4795, + "step": 40842 + }, + { + "epoch": 15.08, + "learning_rate": 1.4926838280576426e-06, + "loss": 0.1737, + "step": 40843 + }, + { + "epoch": 15.08, + "learning_rate": 1.4926468776942975e-06, + "loss": 0.3563, + "step": 40844 + }, + { + "epoch": 15.08, + "learning_rate": 1.492609927330952e-06, + "loss": 0.262, + "step": 40845 + }, + { + "epoch": 15.08, + "learning_rate": 1.492572976967607e-06, + "loss": 0.1924, + "step": 40846 + }, + { + "epoch": 15.08, + "learning_rate": 1.4925360266042618e-06, + "loss": 0.278, + "step": 40847 + }, + { + "epoch": 15.08, + "learning_rate": 1.4924990762409164e-06, + "loss": 0.291, + "step": 40848 + }, + { + "epoch": 15.08, + "learning_rate": 1.4924621258775713e-06, + "loss": 0.2749, + "step": 40849 + }, + { + "epoch": 15.08, + "learning_rate": 1.492425175514226e-06, + "loss": 0.2514, + "step": 40850 + }, + { + "epoch": 15.09, + "learning_rate": 1.4923882251508806e-06, + "loss": 0.3754, + "step": 40851 + }, + { + "epoch": 15.09, + "learning_rate": 1.4923512747875354e-06, + "loss": 0.3017, + "step": 40852 + }, + { + "epoch": 15.09, + "learning_rate": 1.4923143244241903e-06, + "loss": 0.3606, + "step": 40853 + }, + { + "epoch": 15.09, + "learning_rate": 1.4922773740608449e-06, + "loss": 0.244, + "step": 40854 + }, + { + "epoch": 15.09, + "learning_rate": 1.4922404236974997e-06, + "loss": 0.1661, + "step": 40855 + }, + { + "epoch": 15.09, + "learning_rate": 1.4922034733341546e-06, + "loss": 0.285, + "step": 40856 + }, + { + "epoch": 15.09, + "learning_rate": 1.4921665229708092e-06, + "loss": 0.3327, + "step": 40857 + }, + { + "epoch": 15.09, + "learning_rate": 1.492129572607464e-06, + "loss": 0.2795, + "step": 40858 + }, + { + "epoch": 15.09, + "learning_rate": 1.492092622244119e-06, + "loss": 0.2282, + "step": 40859 + }, + { + "epoch": 15.09, + "learning_rate": 1.4920556718807736e-06, + "loss": 0.1533, + "step": 40860 + }, + { + "epoch": 15.09, + "learning_rate": 1.4920187215174282e-06, + "loss": 0.2532, + "step": 40861 + }, + { + "epoch": 15.09, + "learning_rate": 1.491981771154083e-06, + "loss": 0.3477, + "step": 40862 + }, + { + "epoch": 15.09, + "learning_rate": 1.4919448207907379e-06, + "loss": 0.3213, + "step": 40863 + }, + { + "epoch": 15.09, + "learning_rate": 1.4919078704273925e-06, + "loss": 0.2862, + "step": 40864 + }, + { + "epoch": 15.09, + "learning_rate": 1.4918709200640474e-06, + "loss": 0.2389, + "step": 40865 + }, + { + "epoch": 15.09, + "learning_rate": 1.4918339697007022e-06, + "loss": 0.2422, + "step": 40866 + }, + { + "epoch": 15.09, + "learning_rate": 1.4917970193373569e-06, + "loss": 0.3106, + "step": 40867 + }, + { + "epoch": 15.09, + "learning_rate": 1.4917600689740117e-06, + "loss": 0.1704, + "step": 40868 + }, + { + "epoch": 15.09, + "learning_rate": 1.4917231186106666e-06, + "loss": 0.2899, + "step": 40869 + }, + { + "epoch": 15.09, + "learning_rate": 1.491686168247321e-06, + "loss": 0.2228, + "step": 40870 + }, + { + "epoch": 15.09, + "learning_rate": 1.4916492178839758e-06, + "loss": 0.2528, + "step": 40871 + }, + { + "epoch": 15.09, + "learning_rate": 1.4916122675206307e-06, + "loss": 0.2951, + "step": 40872 + }, + { + "epoch": 15.09, + "learning_rate": 1.4915753171572853e-06, + "loss": 0.3607, + "step": 40873 + }, + { + "epoch": 15.09, + "learning_rate": 1.4915383667939402e-06, + "loss": 0.3101, + "step": 40874 + }, + { + "epoch": 15.09, + "learning_rate": 1.491501416430595e-06, + "loss": 0.3213, + "step": 40875 + }, + { + "epoch": 15.09, + "learning_rate": 1.4914644660672496e-06, + "loss": 0.2216, + "step": 40876 + }, + { + "epoch": 15.09, + "learning_rate": 1.4914275157039045e-06, + "loss": 0.309, + "step": 40877 + }, + { + "epoch": 15.1, + "learning_rate": 1.4913905653405593e-06, + "loss": 0.2633, + "step": 40878 + }, + { + "epoch": 15.1, + "learning_rate": 1.491353614977214e-06, + "loss": 0.2392, + "step": 40879 + }, + { + "epoch": 15.1, + "learning_rate": 1.4913166646138686e-06, + "loss": 0.4247, + "step": 40880 + }, + { + "epoch": 15.1, + "learning_rate": 1.4912797142505235e-06, + "loss": 0.1518, + "step": 40881 + }, + { + "epoch": 15.1, + "learning_rate": 1.4912427638871783e-06, + "loss": 0.4052, + "step": 40882 + }, + { + "epoch": 15.1, + "learning_rate": 1.491205813523833e-06, + "loss": 0.2303, + "step": 40883 + }, + { + "epoch": 15.1, + "learning_rate": 1.4911688631604878e-06, + "loss": 0.1829, + "step": 40884 + }, + { + "epoch": 15.1, + "learning_rate": 1.4911319127971426e-06, + "loss": 0.2412, + "step": 40885 + }, + { + "epoch": 15.1, + "learning_rate": 1.4910949624337973e-06, + "loss": 0.2118, + "step": 40886 + }, + { + "epoch": 15.1, + "learning_rate": 1.4910580120704521e-06, + "loss": 0.2149, + "step": 40887 + }, + { + "epoch": 15.1, + "learning_rate": 1.491021061707107e-06, + "loss": 0.2512, + "step": 40888 + }, + { + "epoch": 15.1, + "learning_rate": 1.4909841113437614e-06, + "loss": 0.2353, + "step": 40889 + }, + { + "epoch": 15.1, + "learning_rate": 1.4909471609804162e-06, + "loss": 0.1912, + "step": 40890 + }, + { + "epoch": 15.1, + "learning_rate": 1.490910210617071e-06, + "loss": 0.2479, + "step": 40891 + }, + { + "epoch": 15.1, + "learning_rate": 1.4908732602537257e-06, + "loss": 0.2547, + "step": 40892 + }, + { + "epoch": 15.1, + "learning_rate": 1.4908363098903806e-06, + "loss": 0.1908, + "step": 40893 + }, + { + "epoch": 15.1, + "learning_rate": 1.4907993595270354e-06, + "loss": 0.2872, + "step": 40894 + }, + { + "epoch": 15.1, + "learning_rate": 1.49076240916369e-06, + "loss": 0.2596, + "step": 40895 + }, + { + "epoch": 15.1, + "learning_rate": 1.490725458800345e-06, + "loss": 0.2475, + "step": 40896 + }, + { + "epoch": 15.1, + "learning_rate": 1.4906885084369998e-06, + "loss": 0.23, + "step": 40897 + }, + { + "epoch": 15.1, + "learning_rate": 1.4906515580736544e-06, + "loss": 0.3983, + "step": 40898 + }, + { + "epoch": 15.1, + "learning_rate": 1.490614607710309e-06, + "loss": 0.384, + "step": 40899 + }, + { + "epoch": 15.1, + "learning_rate": 1.4905776573469639e-06, + "loss": 0.242, + "step": 40900 + }, + { + "epoch": 15.1, + "learning_rate": 1.4905407069836187e-06, + "loss": 0.3228, + "step": 40901 + }, + { + "epoch": 15.1, + "learning_rate": 1.4905037566202734e-06, + "loss": 0.3247, + "step": 40902 + }, + { + "epoch": 15.1, + "learning_rate": 1.4904668062569282e-06, + "loss": 0.3365, + "step": 40903 + }, + { + "epoch": 15.1, + "learning_rate": 1.490429855893583e-06, + "loss": 0.3534, + "step": 40904 + }, + { + "epoch": 15.11, + "learning_rate": 1.4903929055302377e-06, + "loss": 0.1857, + "step": 40905 + }, + { + "epoch": 15.11, + "learning_rate": 1.4903559551668925e-06, + "loss": 0.2422, + "step": 40906 + }, + { + "epoch": 15.11, + "learning_rate": 1.4903190048035474e-06, + "loss": 0.2399, + "step": 40907 + }, + { + "epoch": 15.11, + "learning_rate": 1.490282054440202e-06, + "loss": 0.2387, + "step": 40908 + }, + { + "epoch": 15.11, + "learning_rate": 1.4902451040768567e-06, + "loss": 0.3671, + "step": 40909 + }, + { + "epoch": 15.11, + "learning_rate": 1.4902081537135115e-06, + "loss": 0.4254, + "step": 40910 + }, + { + "epoch": 15.11, + "learning_rate": 1.4901712033501664e-06, + "loss": 0.2741, + "step": 40911 + }, + { + "epoch": 15.11, + "learning_rate": 1.490134252986821e-06, + "loss": 0.2008, + "step": 40912 + }, + { + "epoch": 15.11, + "learning_rate": 1.4900973026234758e-06, + "loss": 0.2335, + "step": 40913 + }, + { + "epoch": 15.11, + "learning_rate": 1.4900603522601307e-06, + "loss": 0.3561, + "step": 40914 + }, + { + "epoch": 15.11, + "learning_rate": 1.4900234018967853e-06, + "loss": 0.3434, + "step": 40915 + }, + { + "epoch": 15.11, + "learning_rate": 1.4899864515334402e-06, + "loss": 0.1551, + "step": 40916 + }, + { + "epoch": 15.11, + "learning_rate": 1.489949501170095e-06, + "loss": 0.277, + "step": 40917 + }, + { + "epoch": 15.11, + "learning_rate": 1.4899125508067497e-06, + "loss": 0.268, + "step": 40918 + }, + { + "epoch": 15.11, + "learning_rate": 1.4898756004434043e-06, + "loss": 0.2091, + "step": 40919 + }, + { + "epoch": 15.11, + "learning_rate": 1.4898386500800592e-06, + "loss": 0.4166, + "step": 40920 + }, + { + "epoch": 15.11, + "learning_rate": 1.4898016997167138e-06, + "loss": 0.3454, + "step": 40921 + }, + { + "epoch": 15.11, + "learning_rate": 1.4897647493533686e-06, + "loss": 0.2819, + "step": 40922 + }, + { + "epoch": 15.11, + "learning_rate": 1.4897277989900235e-06, + "loss": 0.1748, + "step": 40923 + }, + { + "epoch": 15.11, + "learning_rate": 1.4896908486266781e-06, + "loss": 0.1919, + "step": 40924 + }, + { + "epoch": 15.11, + "learning_rate": 1.489653898263333e-06, + "loss": 0.3103, + "step": 40925 + }, + { + "epoch": 15.11, + "learning_rate": 1.4896169478999878e-06, + "loss": 0.2634, + "step": 40926 + }, + { + "epoch": 15.11, + "learning_rate": 1.4895799975366425e-06, + "loss": 0.2468, + "step": 40927 + }, + { + "epoch": 15.11, + "learning_rate": 1.4895430471732973e-06, + "loss": 0.2489, + "step": 40928 + }, + { + "epoch": 15.11, + "learning_rate": 1.489506096809952e-06, + "loss": 0.2643, + "step": 40929 + }, + { + "epoch": 15.11, + "learning_rate": 1.4894691464466068e-06, + "loss": 0.1772, + "step": 40930 + }, + { + "epoch": 15.11, + "learning_rate": 1.4894321960832614e-06, + "loss": 0.1936, + "step": 40931 + }, + { + "epoch": 15.12, + "learning_rate": 1.4893952457199163e-06, + "loss": 0.2879, + "step": 40932 + }, + { + "epoch": 15.12, + "learning_rate": 1.4893582953565711e-06, + "loss": 0.1769, + "step": 40933 + }, + { + "epoch": 15.12, + "learning_rate": 1.4893213449932258e-06, + "loss": 0.3574, + "step": 40934 + }, + { + "epoch": 15.12, + "learning_rate": 1.4892843946298806e-06, + "loss": 0.1825, + "step": 40935 + }, + { + "epoch": 15.12, + "learning_rate": 1.4892474442665355e-06, + "loss": 0.3736, + "step": 40936 + }, + { + "epoch": 15.12, + "learning_rate": 1.48921049390319e-06, + "loss": 0.2964, + "step": 40937 + }, + { + "epoch": 15.12, + "learning_rate": 1.489173543539845e-06, + "loss": 0.2605, + "step": 40938 + }, + { + "epoch": 15.12, + "learning_rate": 1.4891365931764996e-06, + "loss": 0.2457, + "step": 40939 + }, + { + "epoch": 15.12, + "learning_rate": 1.4890996428131542e-06, + "loss": 0.1996, + "step": 40940 + }, + { + "epoch": 15.12, + "learning_rate": 1.489062692449809e-06, + "loss": 0.2087, + "step": 40941 + }, + { + "epoch": 15.12, + "learning_rate": 1.489025742086464e-06, + "loss": 0.3567, + "step": 40942 + }, + { + "epoch": 15.12, + "learning_rate": 1.4889887917231185e-06, + "loss": 0.2562, + "step": 40943 + }, + { + "epoch": 15.12, + "learning_rate": 1.4889518413597734e-06, + "loss": 0.2688, + "step": 40944 + }, + { + "epoch": 15.12, + "learning_rate": 1.4889148909964282e-06, + "loss": 0.3089, + "step": 40945 + }, + { + "epoch": 15.12, + "learning_rate": 1.4888779406330829e-06, + "loss": 0.1809, + "step": 40946 + }, + { + "epoch": 15.12, + "learning_rate": 1.4888409902697377e-06, + "loss": 0.3521, + "step": 40947 + }, + { + "epoch": 15.12, + "learning_rate": 1.4888040399063926e-06, + "loss": 0.3455, + "step": 40948 + }, + { + "epoch": 15.12, + "learning_rate": 1.4887670895430472e-06, + "loss": 0.2365, + "step": 40949 + }, + { + "epoch": 15.12, + "learning_rate": 1.4887301391797018e-06, + "loss": 0.4514, + "step": 40950 + }, + { + "epoch": 15.12, + "learning_rate": 1.4886931888163567e-06, + "loss": 0.2522, + "step": 40951 + }, + { + "epoch": 15.12, + "learning_rate": 1.4886562384530115e-06, + "loss": 0.3409, + "step": 40952 + }, + { + "epoch": 15.12, + "learning_rate": 1.4886192880896662e-06, + "loss": 0.3905, + "step": 40953 + }, + { + "epoch": 15.12, + "learning_rate": 1.488582337726321e-06, + "loss": 0.2191, + "step": 40954 + }, + { + "epoch": 15.12, + "learning_rate": 1.4885453873629759e-06, + "loss": 0.3002, + "step": 40955 + }, + { + "epoch": 15.12, + "learning_rate": 1.4885084369996305e-06, + "loss": 0.2094, + "step": 40956 + }, + { + "epoch": 15.12, + "learning_rate": 1.4884714866362854e-06, + "loss": 0.3616, + "step": 40957 + }, + { + "epoch": 15.12, + "learning_rate": 1.4884345362729402e-06, + "loss": 0.1898, + "step": 40958 + }, + { + "epoch": 15.13, + "learning_rate": 1.4883975859095948e-06, + "loss": 0.143, + "step": 40959 + }, + { + "epoch": 15.13, + "learning_rate": 1.4883606355462495e-06, + "loss": 0.4191, + "step": 40960 + }, + { + "epoch": 15.13, + "learning_rate": 1.4883236851829043e-06, + "loss": 0.1431, + "step": 40961 + }, + { + "epoch": 15.13, + "learning_rate": 1.4882867348195592e-06, + "loss": 0.142, + "step": 40962 + }, + { + "epoch": 15.13, + "learning_rate": 1.4882497844562138e-06, + "loss": 0.1432, + "step": 40963 + }, + { + "epoch": 15.13, + "learning_rate": 1.4882128340928687e-06, + "loss": 0.2318, + "step": 40964 + }, + { + "epoch": 15.13, + "learning_rate": 1.4881758837295233e-06, + "loss": 0.3788, + "step": 40965 + }, + { + "epoch": 15.13, + "learning_rate": 1.4881389333661781e-06, + "loss": 0.3616, + "step": 40966 + }, + { + "epoch": 15.13, + "learning_rate": 1.488101983002833e-06, + "loss": 0.1592, + "step": 40967 + }, + { + "epoch": 15.13, + "learning_rate": 1.4880650326394876e-06, + "loss": 0.2083, + "step": 40968 + }, + { + "epoch": 15.13, + "learning_rate": 1.4880280822761423e-06, + "loss": 0.2171, + "step": 40969 + }, + { + "epoch": 15.13, + "learning_rate": 1.4879911319127971e-06, + "loss": 0.2452, + "step": 40970 + }, + { + "epoch": 15.13, + "learning_rate": 1.487954181549452e-06, + "loss": 0.3525, + "step": 40971 + }, + { + "epoch": 15.13, + "learning_rate": 1.4879172311861066e-06, + "loss": 0.1629, + "step": 40972 + }, + { + "epoch": 15.13, + "learning_rate": 1.4878802808227614e-06, + "loss": 0.2584, + "step": 40973 + }, + { + "epoch": 15.13, + "learning_rate": 1.4878433304594163e-06, + "loss": 0.2071, + "step": 40974 + }, + { + "epoch": 15.13, + "learning_rate": 1.487806380096071e-06, + "loss": 0.3025, + "step": 40975 + }, + { + "epoch": 15.13, + "learning_rate": 1.4877694297327258e-06, + "loss": 0.4655, + "step": 40976 + }, + { + "epoch": 15.13, + "learning_rate": 1.4877324793693806e-06, + "loss": 0.1593, + "step": 40977 + }, + { + "epoch": 15.13, + "learning_rate": 1.4876955290060353e-06, + "loss": 0.271, + "step": 40978 + }, + { + "epoch": 15.13, + "learning_rate": 1.48765857864269e-06, + "loss": 0.2077, + "step": 40979 + }, + { + "epoch": 15.13, + "learning_rate": 1.4876216282793447e-06, + "loss": 0.1719, + "step": 40980 + }, + { + "epoch": 15.13, + "learning_rate": 1.4875846779159996e-06, + "loss": 0.1854, + "step": 40981 + }, + { + "epoch": 15.13, + "learning_rate": 1.4875477275526542e-06, + "loss": 0.3475, + "step": 40982 + }, + { + "epoch": 15.13, + "learning_rate": 1.487510777189309e-06, + "loss": 0.2503, + "step": 40983 + }, + { + "epoch": 15.13, + "learning_rate": 1.487473826825964e-06, + "loss": 0.2001, + "step": 40984 + }, + { + "epoch": 15.13, + "learning_rate": 1.4874368764626186e-06, + "loss": 0.2542, + "step": 40985 + }, + { + "epoch": 15.14, + "learning_rate": 1.4873999260992734e-06, + "loss": 0.1467, + "step": 40986 + }, + { + "epoch": 15.14, + "learning_rate": 1.4873629757359283e-06, + "loss": 0.2425, + "step": 40987 + }, + { + "epoch": 15.14, + "learning_rate": 1.487326025372583e-06, + "loss": 0.3344, + "step": 40988 + }, + { + "epoch": 15.14, + "learning_rate": 1.4872890750092375e-06, + "loss": 0.1975, + "step": 40989 + }, + { + "epoch": 15.14, + "learning_rate": 1.4872521246458924e-06, + "loss": 0.2731, + "step": 40990 + }, + { + "epoch": 15.14, + "learning_rate": 1.487215174282547e-06, + "loss": 0.3205, + "step": 40991 + }, + { + "epoch": 15.14, + "learning_rate": 1.4871782239192019e-06, + "loss": 0.3768, + "step": 40992 + }, + { + "epoch": 15.14, + "learning_rate": 1.4871412735558567e-06, + "loss": 0.1934, + "step": 40993 + }, + { + "epoch": 15.14, + "learning_rate": 1.4871043231925114e-06, + "loss": 0.3177, + "step": 40994 + }, + { + "epoch": 15.14, + "learning_rate": 1.4870673728291662e-06, + "loss": 0.2754, + "step": 40995 + }, + { + "epoch": 15.14, + "learning_rate": 1.487030422465821e-06, + "loss": 0.2171, + "step": 40996 + }, + { + "epoch": 15.14, + "learning_rate": 1.4869934721024757e-06, + "loss": 0.2687, + "step": 40997 + }, + { + "epoch": 15.14, + "learning_rate": 1.4869565217391305e-06, + "loss": 0.1842, + "step": 40998 + }, + { + "epoch": 15.14, + "learning_rate": 1.4869195713757852e-06, + "loss": 0.199, + "step": 40999 + }, + { + "epoch": 15.14, + "learning_rate": 1.48688262101244e-06, + "loss": 0.2364, + "step": 41000 + }, + { + "epoch": 15.14, + "learning_rate": 1.4868456706490947e-06, + "loss": 0.1909, + "step": 41001 + }, + { + "epoch": 15.14, + "learning_rate": 1.4868087202857495e-06, + "loss": 0.1791, + "step": 41002 + }, + { + "epoch": 15.14, + "learning_rate": 1.4867717699224044e-06, + "loss": 0.2482, + "step": 41003 + }, + { + "epoch": 15.14, + "learning_rate": 1.486734819559059e-06, + "loss": 0.1614, + "step": 41004 + }, + { + "epoch": 15.14, + "learning_rate": 1.4866978691957138e-06, + "loss": 0.1548, + "step": 41005 + }, + { + "epoch": 15.14, + "learning_rate": 1.4866609188323687e-06, + "loss": 0.2432, + "step": 41006 + }, + { + "epoch": 15.14, + "learning_rate": 1.4866239684690233e-06, + "loss": 0.2733, + "step": 41007 + }, + { + "epoch": 15.14, + "learning_rate": 1.4865870181056782e-06, + "loss": 0.2664, + "step": 41008 + }, + { + "epoch": 15.14, + "learning_rate": 1.4865500677423328e-06, + "loss": 0.3409, + "step": 41009 + }, + { + "epoch": 15.14, + "learning_rate": 1.4865131173789874e-06, + "loss": 0.3077, + "step": 41010 + }, + { + "epoch": 15.14, + "learning_rate": 1.4864761670156423e-06, + "loss": 0.2212, + "step": 41011 + }, + { + "epoch": 15.14, + "learning_rate": 1.4864392166522971e-06, + "loss": 0.3001, + "step": 41012 + }, + { + "epoch": 15.15, + "learning_rate": 1.4864022662889518e-06, + "loss": 0.1971, + "step": 41013 + }, + { + "epoch": 15.15, + "learning_rate": 1.4863653159256066e-06, + "loss": 0.2467, + "step": 41014 + }, + { + "epoch": 15.15, + "learning_rate": 1.4863283655622615e-06, + "loss": 0.2339, + "step": 41015 + }, + { + "epoch": 15.15, + "learning_rate": 1.4862914151989161e-06, + "loss": 0.2234, + "step": 41016 + }, + { + "epoch": 15.15, + "learning_rate": 1.486254464835571e-06, + "loss": 0.2516, + "step": 41017 + }, + { + "epoch": 15.15, + "learning_rate": 1.4862175144722258e-06, + "loss": 0.2437, + "step": 41018 + }, + { + "epoch": 15.15, + "learning_rate": 1.4861805641088804e-06, + "loss": 0.1977, + "step": 41019 + }, + { + "epoch": 15.15, + "learning_rate": 1.486143613745535e-06, + "loss": 0.2204, + "step": 41020 + }, + { + "epoch": 15.15, + "learning_rate": 1.48610666338219e-06, + "loss": 0.2375, + "step": 41021 + }, + { + "epoch": 15.15, + "learning_rate": 1.4860697130188448e-06, + "loss": 0.4713, + "step": 41022 + }, + { + "epoch": 15.15, + "learning_rate": 1.4860327626554994e-06, + "loss": 0.3241, + "step": 41023 + }, + { + "epoch": 15.15, + "learning_rate": 1.4859958122921543e-06, + "loss": 0.2591, + "step": 41024 + }, + { + "epoch": 15.15, + "learning_rate": 1.485958861928809e-06, + "loss": 0.3148, + "step": 41025 + }, + { + "epoch": 15.15, + "learning_rate": 1.4859219115654637e-06, + "loss": 0.2518, + "step": 41026 + }, + { + "epoch": 15.15, + "learning_rate": 1.4858849612021186e-06, + "loss": 0.2542, + "step": 41027 + }, + { + "epoch": 15.15, + "learning_rate": 1.4858480108387734e-06, + "loss": 0.3761, + "step": 41028 + }, + { + "epoch": 15.15, + "learning_rate": 1.485811060475428e-06, + "loss": 0.3915, + "step": 41029 + }, + { + "epoch": 15.15, + "learning_rate": 1.4857741101120827e-06, + "loss": 0.5194, + "step": 41030 + }, + { + "epoch": 15.15, + "learning_rate": 1.4857371597487376e-06, + "loss": 0.2226, + "step": 41031 + }, + { + "epoch": 15.15, + "learning_rate": 1.4857002093853924e-06, + "loss": 0.3117, + "step": 41032 + }, + { + "epoch": 15.15, + "learning_rate": 1.485663259022047e-06, + "loss": 0.2292, + "step": 41033 + }, + { + "epoch": 15.15, + "learning_rate": 1.4856263086587019e-06, + "loss": 0.1615, + "step": 41034 + }, + { + "epoch": 15.15, + "learning_rate": 1.4855893582953567e-06, + "loss": 0.2779, + "step": 41035 + }, + { + "epoch": 15.15, + "learning_rate": 1.4855524079320114e-06, + "loss": 0.2549, + "step": 41036 + }, + { + "epoch": 15.15, + "learning_rate": 1.4855154575686662e-06, + "loss": 0.1437, + "step": 41037 + }, + { + "epoch": 15.15, + "learning_rate": 1.485478507205321e-06, + "loss": 0.1628, + "step": 41038 + }, + { + "epoch": 15.15, + "learning_rate": 1.4854415568419755e-06, + "loss": 0.3057, + "step": 41039 + }, + { + "epoch": 15.16, + "learning_rate": 1.4854046064786303e-06, + "loss": 0.2539, + "step": 41040 + }, + { + "epoch": 15.16, + "learning_rate": 1.4853676561152852e-06, + "loss": 0.2468, + "step": 41041 + }, + { + "epoch": 15.16, + "learning_rate": 1.4853307057519398e-06, + "loss": 0.2444, + "step": 41042 + }, + { + "epoch": 15.16, + "learning_rate": 1.4852937553885947e-06, + "loss": 0.1787, + "step": 41043 + }, + { + "epoch": 15.16, + "learning_rate": 1.4852568050252495e-06, + "loss": 0.2781, + "step": 41044 + }, + { + "epoch": 15.16, + "learning_rate": 1.4852198546619042e-06, + "loss": 0.258, + "step": 41045 + }, + { + "epoch": 15.16, + "learning_rate": 1.485182904298559e-06, + "loss": 0.2291, + "step": 41046 + }, + { + "epoch": 15.16, + "learning_rate": 1.4851459539352139e-06, + "loss": 0.2925, + "step": 41047 + }, + { + "epoch": 15.16, + "learning_rate": 1.4851090035718685e-06, + "loss": 0.3404, + "step": 41048 + }, + { + "epoch": 15.16, + "learning_rate": 1.4850720532085231e-06, + "loss": 0.2289, + "step": 41049 + }, + { + "epoch": 15.16, + "learning_rate": 1.485035102845178e-06, + "loss": 0.3047, + "step": 41050 + }, + { + "epoch": 15.16, + "learning_rate": 1.4849981524818328e-06, + "loss": 0.2366, + "step": 41051 + }, + { + "epoch": 15.16, + "learning_rate": 1.4849612021184875e-06, + "loss": 0.2038, + "step": 41052 + }, + { + "epoch": 15.16, + "learning_rate": 1.4849242517551423e-06, + "loss": 0.2544, + "step": 41053 + }, + { + "epoch": 15.16, + "learning_rate": 1.4848873013917972e-06, + "loss": 0.1987, + "step": 41054 + }, + { + "epoch": 15.16, + "learning_rate": 1.4848503510284518e-06, + "loss": 0.2786, + "step": 41055 + }, + { + "epoch": 15.16, + "learning_rate": 1.4848134006651066e-06, + "loss": 0.1928, + "step": 41056 + }, + { + "epoch": 15.16, + "learning_rate": 1.4847764503017615e-06, + "loss": 0.1908, + "step": 41057 + }, + { + "epoch": 15.16, + "learning_rate": 1.4847394999384161e-06, + "loss": 0.3091, + "step": 41058 + }, + { + "epoch": 15.16, + "learning_rate": 1.4847025495750708e-06, + "loss": 0.236, + "step": 41059 + }, + { + "epoch": 15.16, + "learning_rate": 1.4846655992117256e-06, + "loss": 0.2391, + "step": 41060 + }, + { + "epoch": 15.16, + "learning_rate": 1.4846286488483803e-06, + "loss": 0.2676, + "step": 41061 + }, + { + "epoch": 15.16, + "learning_rate": 1.484591698485035e-06, + "loss": 0.2566, + "step": 41062 + }, + { + "epoch": 15.16, + "learning_rate": 1.48455474812169e-06, + "loss": 0.3475, + "step": 41063 + }, + { + "epoch": 15.16, + "learning_rate": 1.4845177977583446e-06, + "loss": 0.2851, + "step": 41064 + }, + { + "epoch": 15.16, + "learning_rate": 1.4844808473949994e-06, + "loss": 0.2612, + "step": 41065 + }, + { + "epoch": 15.16, + "learning_rate": 1.4844438970316543e-06, + "loss": 0.2057, + "step": 41066 + }, + { + "epoch": 15.17, + "learning_rate": 1.484406946668309e-06, + "loss": 0.2323, + "step": 41067 + }, + { + "epoch": 15.17, + "learning_rate": 1.4843699963049638e-06, + "loss": 0.3419, + "step": 41068 + }, + { + "epoch": 15.17, + "learning_rate": 1.4843330459416184e-06, + "loss": 0.3007, + "step": 41069 + }, + { + "epoch": 15.17, + "learning_rate": 1.4842960955782733e-06, + "loss": 0.2017, + "step": 41070 + }, + { + "epoch": 15.17, + "learning_rate": 1.4842591452149279e-06, + "loss": 0.2921, + "step": 41071 + }, + { + "epoch": 15.17, + "learning_rate": 1.4842221948515827e-06, + "loss": 0.1875, + "step": 41072 + }, + { + "epoch": 15.17, + "learning_rate": 1.4841852444882376e-06, + "loss": 0.2963, + "step": 41073 + }, + { + "epoch": 15.17, + "learning_rate": 1.4841482941248922e-06, + "loss": 0.2894, + "step": 41074 + }, + { + "epoch": 15.17, + "learning_rate": 1.484111343761547e-06, + "loss": 0.1984, + "step": 41075 + }, + { + "epoch": 15.17, + "learning_rate": 1.484074393398202e-06, + "loss": 0.2559, + "step": 41076 + }, + { + "epoch": 15.17, + "learning_rate": 1.4840374430348566e-06, + "loss": 0.228, + "step": 41077 + }, + { + "epoch": 15.17, + "learning_rate": 1.4840004926715114e-06, + "loss": 0.3574, + "step": 41078 + }, + { + "epoch": 15.17, + "learning_rate": 1.483963542308166e-06, + "loss": 0.2667, + "step": 41079 + }, + { + "epoch": 15.17, + "learning_rate": 1.4839265919448207e-06, + "loss": 0.1413, + "step": 41080 + }, + { + "epoch": 15.17, + "learning_rate": 1.4838896415814755e-06, + "loss": 0.336, + "step": 41081 + }, + { + "epoch": 15.17, + "learning_rate": 1.4838526912181304e-06, + "loss": 0.3178, + "step": 41082 + }, + { + "epoch": 15.17, + "learning_rate": 1.483815740854785e-06, + "loss": 0.2406, + "step": 41083 + }, + { + "epoch": 15.17, + "learning_rate": 1.4837787904914399e-06, + "loss": 0.1949, + "step": 41084 + }, + { + "epoch": 15.17, + "learning_rate": 1.4837418401280947e-06, + "loss": 0.2144, + "step": 41085 + }, + { + "epoch": 15.17, + "learning_rate": 1.4837048897647493e-06, + "loss": 0.1569, + "step": 41086 + }, + { + "epoch": 15.17, + "learning_rate": 1.4836679394014042e-06, + "loss": 0.1866, + "step": 41087 + }, + { + "epoch": 15.17, + "learning_rate": 1.483630989038059e-06, + "loss": 0.3373, + "step": 41088 + }, + { + "epoch": 15.17, + "learning_rate": 1.4835940386747137e-06, + "loss": 0.202, + "step": 41089 + }, + { + "epoch": 15.17, + "learning_rate": 1.4835570883113683e-06, + "loss": 0.1696, + "step": 41090 + }, + { + "epoch": 15.17, + "learning_rate": 1.4835201379480232e-06, + "loss": 0.2177, + "step": 41091 + }, + { + "epoch": 15.17, + "learning_rate": 1.483483187584678e-06, + "loss": 0.2468, + "step": 41092 + }, + { + "epoch": 15.17, + "learning_rate": 1.4834462372213326e-06, + "loss": 0.2698, + "step": 41093 + }, + { + "epoch": 15.18, + "learning_rate": 1.4834092868579875e-06, + "loss": 0.3168, + "step": 41094 + }, + { + "epoch": 15.18, + "learning_rate": 1.4833723364946423e-06, + "loss": 0.3203, + "step": 41095 + }, + { + "epoch": 15.18, + "learning_rate": 1.483335386131297e-06, + "loss": 0.2254, + "step": 41096 + }, + { + "epoch": 15.18, + "learning_rate": 1.4832984357679518e-06, + "loss": 0.2067, + "step": 41097 + }, + { + "epoch": 15.18, + "learning_rate": 1.4832614854046067e-06, + "loss": 0.3436, + "step": 41098 + }, + { + "epoch": 15.18, + "learning_rate": 1.4832245350412613e-06, + "loss": 0.2084, + "step": 41099 + }, + { + "epoch": 15.18, + "learning_rate": 1.483187584677916e-06, + "loss": 0.1528, + "step": 41100 + }, + { + "epoch": 15.18, + "learning_rate": 1.4831506343145708e-06, + "loss": 0.2036, + "step": 41101 + }, + { + "epoch": 15.18, + "learning_rate": 1.4831136839512256e-06, + "loss": 0.2896, + "step": 41102 + }, + { + "epoch": 15.18, + "learning_rate": 1.4830767335878803e-06, + "loss": 0.2785, + "step": 41103 + }, + { + "epoch": 15.18, + "learning_rate": 1.4830397832245351e-06, + "loss": 0.2656, + "step": 41104 + }, + { + "epoch": 15.18, + "learning_rate": 1.48300283286119e-06, + "loss": 0.1885, + "step": 41105 + }, + { + "epoch": 15.18, + "learning_rate": 1.4829658824978446e-06, + "loss": 0.2218, + "step": 41106 + }, + { + "epoch": 15.18, + "learning_rate": 1.4829289321344995e-06, + "loss": 0.1355, + "step": 41107 + }, + { + "epoch": 15.18, + "learning_rate": 1.4828919817711543e-06, + "loss": 0.273, + "step": 41108 + }, + { + "epoch": 15.18, + "learning_rate": 1.4828550314078087e-06, + "loss": 0.2976, + "step": 41109 + }, + { + "epoch": 15.18, + "learning_rate": 1.4828180810444636e-06, + "loss": 0.1358, + "step": 41110 + }, + { + "epoch": 15.18, + "learning_rate": 1.4827811306811184e-06, + "loss": 0.2784, + "step": 41111 + }, + { + "epoch": 15.18, + "learning_rate": 1.482744180317773e-06, + "loss": 0.274, + "step": 41112 + }, + { + "epoch": 15.18, + "learning_rate": 1.482707229954428e-06, + "loss": 0.3294, + "step": 41113 + }, + { + "epoch": 15.18, + "learning_rate": 1.4826702795910828e-06, + "loss": 0.2589, + "step": 41114 + }, + { + "epoch": 15.18, + "learning_rate": 1.4826333292277374e-06, + "loss": 0.2405, + "step": 41115 + }, + { + "epoch": 15.18, + "learning_rate": 1.4825963788643922e-06, + "loss": 0.2339, + "step": 41116 + }, + { + "epoch": 15.18, + "learning_rate": 1.482559428501047e-06, + "loss": 0.2608, + "step": 41117 + }, + { + "epoch": 15.18, + "learning_rate": 1.4825224781377017e-06, + "loss": 0.2606, + "step": 41118 + }, + { + "epoch": 15.18, + "learning_rate": 1.4824855277743564e-06, + "loss": 0.321, + "step": 41119 + }, + { + "epoch": 15.18, + "learning_rate": 1.4824485774110112e-06, + "loss": 0.2594, + "step": 41120 + }, + { + "epoch": 15.19, + "learning_rate": 1.482411627047666e-06, + "loss": 0.248, + "step": 41121 + }, + { + "epoch": 15.19, + "learning_rate": 1.4823746766843207e-06, + "loss": 0.1765, + "step": 41122 + }, + { + "epoch": 15.19, + "learning_rate": 1.4823377263209755e-06, + "loss": 0.2911, + "step": 41123 + }, + { + "epoch": 15.19, + "learning_rate": 1.4823007759576304e-06, + "loss": 0.2262, + "step": 41124 + }, + { + "epoch": 15.19, + "learning_rate": 1.482263825594285e-06, + "loss": 0.293, + "step": 41125 + }, + { + "epoch": 15.19, + "learning_rate": 1.4822268752309399e-06, + "loss": 0.2669, + "step": 41126 + }, + { + "epoch": 15.19, + "learning_rate": 1.4821899248675947e-06, + "loss": 0.2006, + "step": 41127 + }, + { + "epoch": 15.19, + "learning_rate": 1.4821529745042492e-06, + "loss": 0.1227, + "step": 41128 + }, + { + "epoch": 15.19, + "learning_rate": 1.482116024140904e-06, + "loss": 0.282, + "step": 41129 + }, + { + "epoch": 15.19, + "learning_rate": 1.4820790737775588e-06, + "loss": 0.2774, + "step": 41130 + }, + { + "epoch": 15.19, + "learning_rate": 1.4820421234142135e-06, + "loss": 0.1401, + "step": 41131 + }, + { + "epoch": 15.19, + "learning_rate": 1.4820051730508683e-06, + "loss": 0.339, + "step": 41132 + }, + { + "epoch": 15.19, + "learning_rate": 1.4819682226875232e-06, + "loss": 0.2943, + "step": 41133 + }, + { + "epoch": 15.19, + "learning_rate": 1.4819312723241778e-06, + "loss": 0.3089, + "step": 41134 + }, + { + "epoch": 15.19, + "learning_rate": 1.4818943219608327e-06, + "loss": 0.3338, + "step": 41135 + }, + { + "epoch": 15.19, + "learning_rate": 1.4818573715974875e-06, + "loss": 0.2085, + "step": 41136 + }, + { + "epoch": 15.19, + "learning_rate": 1.4818204212341421e-06, + "loss": 0.2678, + "step": 41137 + }, + { + "epoch": 15.19, + "learning_rate": 1.4817834708707968e-06, + "loss": 0.163, + "step": 41138 + }, + { + "epoch": 15.19, + "learning_rate": 1.4817465205074516e-06, + "loss": 0.2696, + "step": 41139 + }, + { + "epoch": 15.19, + "learning_rate": 1.4817095701441065e-06, + "loss": 0.1622, + "step": 41140 + }, + { + "epoch": 15.19, + "learning_rate": 1.4816726197807611e-06, + "loss": 0.2344, + "step": 41141 + }, + { + "epoch": 15.19, + "learning_rate": 1.481635669417416e-06, + "loss": 0.3186, + "step": 41142 + }, + { + "epoch": 15.19, + "learning_rate": 1.4815987190540708e-06, + "loss": 0.121, + "step": 41143 + }, + { + "epoch": 15.19, + "learning_rate": 1.4815617686907255e-06, + "loss": 0.2567, + "step": 41144 + }, + { + "epoch": 15.19, + "learning_rate": 1.4815248183273803e-06, + "loss": 0.3678, + "step": 41145 + }, + { + "epoch": 15.19, + "learning_rate": 1.4814878679640351e-06, + "loss": 0.1984, + "step": 41146 + }, + { + "epoch": 15.19, + "learning_rate": 1.4814509176006898e-06, + "loss": 0.2814, + "step": 41147 + }, + { + "epoch": 15.19, + "learning_rate": 1.4814139672373444e-06, + "loss": 0.3357, + "step": 41148 + }, + { + "epoch": 15.2, + "learning_rate": 1.4813770168739993e-06, + "loss": 0.3224, + "step": 41149 + }, + { + "epoch": 15.2, + "learning_rate": 1.4813400665106541e-06, + "loss": 0.1737, + "step": 41150 + }, + { + "epoch": 15.2, + "learning_rate": 1.4813031161473088e-06, + "loss": 0.237, + "step": 41151 + }, + { + "epoch": 15.2, + "learning_rate": 1.4812661657839636e-06, + "loss": 0.2893, + "step": 41152 + }, + { + "epoch": 15.2, + "learning_rate": 1.4812292154206182e-06, + "loss": 0.1748, + "step": 41153 + }, + { + "epoch": 15.2, + "learning_rate": 1.481192265057273e-06, + "loss": 0.2314, + "step": 41154 + }, + { + "epoch": 15.2, + "learning_rate": 1.481155314693928e-06, + "loss": 0.293, + "step": 41155 + }, + { + "epoch": 15.2, + "learning_rate": 1.4811183643305826e-06, + "loss": 0.2178, + "step": 41156 + }, + { + "epoch": 15.2, + "learning_rate": 1.4810814139672374e-06, + "loss": 0.4184, + "step": 41157 + }, + { + "epoch": 15.2, + "learning_rate": 1.481044463603892e-06, + "loss": 0.3727, + "step": 41158 + }, + { + "epoch": 15.2, + "learning_rate": 1.481007513240547e-06, + "loss": 0.3732, + "step": 41159 + }, + { + "epoch": 15.2, + "learning_rate": 1.4809705628772015e-06, + "loss": 0.2178, + "step": 41160 + }, + { + "epoch": 15.2, + "learning_rate": 1.4809336125138564e-06, + "loss": 0.2044, + "step": 41161 + }, + { + "epoch": 15.2, + "learning_rate": 1.4808966621505112e-06, + "loss": 0.2912, + "step": 41162 + }, + { + "epoch": 15.2, + "learning_rate": 1.4808597117871659e-06, + "loss": 0.2491, + "step": 41163 + }, + { + "epoch": 15.2, + "learning_rate": 1.4808227614238207e-06, + "loss": 0.28, + "step": 41164 + }, + { + "epoch": 15.2, + "learning_rate": 1.4807858110604756e-06, + "loss": 0.1309, + "step": 41165 + }, + { + "epoch": 15.2, + "learning_rate": 1.4807488606971302e-06, + "loss": 0.2243, + "step": 41166 + }, + { + "epoch": 15.2, + "learning_rate": 1.480711910333785e-06, + "loss": 0.1576, + "step": 41167 + }, + { + "epoch": 15.2, + "learning_rate": 1.4806749599704397e-06, + "loss": 0.2254, + "step": 41168 + }, + { + "epoch": 15.2, + "learning_rate": 1.4806380096070945e-06, + "loss": 0.2628, + "step": 41169 + }, + { + "epoch": 15.2, + "learning_rate": 1.4806010592437492e-06, + "loss": 0.1081, + "step": 41170 + }, + { + "epoch": 15.2, + "learning_rate": 1.480564108880404e-06, + "loss": 0.1832, + "step": 41171 + }, + { + "epoch": 15.2, + "learning_rate": 1.4805271585170589e-06, + "loss": 0.2248, + "step": 41172 + }, + { + "epoch": 15.2, + "learning_rate": 1.4804902081537135e-06, + "loss": 0.2095, + "step": 41173 + }, + { + "epoch": 15.2, + "learning_rate": 1.4804532577903684e-06, + "loss": 0.2041, + "step": 41174 + }, + { + "epoch": 15.2, + "learning_rate": 1.4804163074270232e-06, + "loss": 0.1945, + "step": 41175 + }, + { + "epoch": 15.21, + "learning_rate": 1.4803793570636778e-06, + "loss": 0.3084, + "step": 41176 + }, + { + "epoch": 15.21, + "learning_rate": 1.4803424067003327e-06, + "loss": 0.2691, + "step": 41177 + }, + { + "epoch": 15.21, + "learning_rate": 1.4803054563369873e-06, + "loss": 0.2676, + "step": 41178 + }, + { + "epoch": 15.21, + "learning_rate": 1.480268505973642e-06, + "loss": 0.1721, + "step": 41179 + }, + { + "epoch": 15.21, + "learning_rate": 1.4802315556102968e-06, + "loss": 0.2674, + "step": 41180 + }, + { + "epoch": 15.21, + "learning_rate": 1.4801946052469517e-06, + "loss": 0.3363, + "step": 41181 + }, + { + "epoch": 15.21, + "learning_rate": 1.4801576548836063e-06, + "loss": 0.309, + "step": 41182 + }, + { + "epoch": 15.21, + "learning_rate": 1.4801207045202611e-06, + "loss": 0.2703, + "step": 41183 + }, + { + "epoch": 15.21, + "learning_rate": 1.480083754156916e-06, + "loss": 0.3033, + "step": 41184 + }, + { + "epoch": 15.21, + "learning_rate": 1.4800468037935706e-06, + "loss": 0.2066, + "step": 41185 + }, + { + "epoch": 15.21, + "learning_rate": 1.4800098534302255e-06, + "loss": 0.3278, + "step": 41186 + }, + { + "epoch": 15.21, + "learning_rate": 1.4799729030668803e-06, + "loss": 0.2868, + "step": 41187 + }, + { + "epoch": 15.21, + "learning_rate": 1.479935952703535e-06, + "loss": 0.1497, + "step": 41188 + }, + { + "epoch": 15.21, + "learning_rate": 1.4798990023401896e-06, + "loss": 0.3297, + "step": 41189 + }, + { + "epoch": 15.21, + "learning_rate": 1.4798620519768444e-06, + "loss": 0.3268, + "step": 41190 + }, + { + "epoch": 15.21, + "learning_rate": 1.4798251016134993e-06, + "loss": 0.3089, + "step": 41191 + }, + { + "epoch": 15.21, + "learning_rate": 1.479788151250154e-06, + "loss": 0.2601, + "step": 41192 + }, + { + "epoch": 15.21, + "learning_rate": 1.4797512008868088e-06, + "loss": 0.1793, + "step": 41193 + }, + { + "epoch": 15.21, + "learning_rate": 1.4797142505234636e-06, + "loss": 0.2243, + "step": 41194 + }, + { + "epoch": 15.21, + "learning_rate": 1.4796773001601183e-06, + "loss": 0.1596, + "step": 41195 + }, + { + "epoch": 15.21, + "learning_rate": 1.4796403497967731e-06, + "loss": 0.2931, + "step": 41196 + }, + { + "epoch": 15.21, + "learning_rate": 1.479603399433428e-06, + "loss": 0.1736, + "step": 41197 + }, + { + "epoch": 15.21, + "learning_rate": 1.4795664490700824e-06, + "loss": 0.2612, + "step": 41198 + }, + { + "epoch": 15.21, + "learning_rate": 1.4795294987067372e-06, + "loss": 0.2837, + "step": 41199 + }, + { + "epoch": 15.21, + "learning_rate": 1.479492548343392e-06, + "loss": 0.1606, + "step": 41200 + }, + { + "epoch": 15.21, + "learning_rate": 1.4794555979800467e-06, + "loss": 0.2841, + "step": 41201 + }, + { + "epoch": 15.21, + "learning_rate": 1.4794186476167016e-06, + "loss": 0.1854, + "step": 41202 + }, + { + "epoch": 15.22, + "learning_rate": 1.4793816972533564e-06, + "loss": 0.1792, + "step": 41203 + }, + { + "epoch": 15.22, + "learning_rate": 1.479344746890011e-06, + "loss": 0.3568, + "step": 41204 + }, + { + "epoch": 15.22, + "learning_rate": 1.479307796526666e-06, + "loss": 0.1911, + "step": 41205 + }, + { + "epoch": 15.22, + "learning_rate": 1.4792708461633207e-06, + "loss": 0.2653, + "step": 41206 + }, + { + "epoch": 15.22, + "learning_rate": 1.4792338957999754e-06, + "loss": 0.2584, + "step": 41207 + }, + { + "epoch": 15.22, + "learning_rate": 1.47919694543663e-06, + "loss": 0.1873, + "step": 41208 + }, + { + "epoch": 15.22, + "learning_rate": 1.4791599950732849e-06, + "loss": 0.2834, + "step": 41209 + }, + { + "epoch": 15.22, + "learning_rate": 1.4791230447099397e-06, + "loss": 0.228, + "step": 41210 + }, + { + "epoch": 15.22, + "learning_rate": 1.4790860943465944e-06, + "loss": 0.2571, + "step": 41211 + }, + { + "epoch": 15.22, + "learning_rate": 1.4790491439832492e-06, + "loss": 0.1663, + "step": 41212 + }, + { + "epoch": 15.22, + "learning_rate": 1.479012193619904e-06, + "loss": 0.3029, + "step": 41213 + }, + { + "epoch": 15.22, + "learning_rate": 1.4789752432565587e-06, + "loss": 0.1689, + "step": 41214 + }, + { + "epoch": 15.22, + "learning_rate": 1.4789382928932135e-06, + "loss": 0.3299, + "step": 41215 + }, + { + "epoch": 15.22, + "learning_rate": 1.4789013425298684e-06, + "loss": 0.2473, + "step": 41216 + }, + { + "epoch": 15.22, + "learning_rate": 1.478864392166523e-06, + "loss": 0.2235, + "step": 41217 + }, + { + "epoch": 15.22, + "learning_rate": 1.4788274418031777e-06, + "loss": 0.2741, + "step": 41218 + }, + { + "epoch": 15.22, + "learning_rate": 1.4787904914398325e-06, + "loss": 0.3379, + "step": 41219 + }, + { + "epoch": 15.22, + "learning_rate": 1.4787535410764874e-06, + "loss": 0.2704, + "step": 41220 + }, + { + "epoch": 15.22, + "learning_rate": 1.478716590713142e-06, + "loss": 0.3306, + "step": 41221 + }, + { + "epoch": 15.22, + "learning_rate": 1.4786796403497968e-06, + "loss": 0.184, + "step": 41222 + }, + { + "epoch": 15.22, + "learning_rate": 1.4786426899864517e-06, + "loss": 0.2527, + "step": 41223 + }, + { + "epoch": 15.22, + "learning_rate": 1.4786057396231063e-06, + "loss": 0.3229, + "step": 41224 + }, + { + "epoch": 15.22, + "learning_rate": 1.4785687892597612e-06, + "loss": 0.3177, + "step": 41225 + }, + { + "epoch": 15.22, + "learning_rate": 1.478531838896416e-06, + "loss": 0.3214, + "step": 41226 + }, + { + "epoch": 15.22, + "learning_rate": 1.4784948885330707e-06, + "loss": 0.3387, + "step": 41227 + }, + { + "epoch": 15.22, + "learning_rate": 1.4784579381697253e-06, + "loss": 0.2043, + "step": 41228 + }, + { + "epoch": 15.22, + "learning_rate": 1.4784209878063801e-06, + "loss": 0.2825, + "step": 41229 + }, + { + "epoch": 15.23, + "learning_rate": 1.4783840374430348e-06, + "loss": 0.2232, + "step": 41230 + }, + { + "epoch": 15.23, + "learning_rate": 1.4783470870796896e-06, + "loss": 0.3302, + "step": 41231 + }, + { + "epoch": 15.23, + "learning_rate": 1.4783101367163445e-06, + "loss": 0.3096, + "step": 41232 + }, + { + "epoch": 15.23, + "learning_rate": 1.4782731863529991e-06, + "loss": 0.1401, + "step": 41233 + }, + { + "epoch": 15.23, + "learning_rate": 1.478236235989654e-06, + "loss": 0.2093, + "step": 41234 + }, + { + "epoch": 15.23, + "learning_rate": 1.4781992856263088e-06, + "loss": 0.1348, + "step": 41235 + }, + { + "epoch": 15.23, + "learning_rate": 1.4781623352629634e-06, + "loss": 0.1979, + "step": 41236 + }, + { + "epoch": 15.23, + "learning_rate": 1.4781253848996183e-06, + "loss": 0.1725, + "step": 41237 + }, + { + "epoch": 15.23, + "learning_rate": 1.478088434536273e-06, + "loss": 0.3495, + "step": 41238 + }, + { + "epoch": 15.23, + "learning_rate": 1.4780514841729278e-06, + "loss": 0.2724, + "step": 41239 + }, + { + "epoch": 15.23, + "learning_rate": 1.4780145338095824e-06, + "loss": 0.296, + "step": 41240 + }, + { + "epoch": 15.23, + "learning_rate": 1.4779775834462373e-06, + "loss": 0.2578, + "step": 41241 + }, + { + "epoch": 15.23, + "learning_rate": 1.477940633082892e-06, + "loss": 0.2251, + "step": 41242 + }, + { + "epoch": 15.23, + "learning_rate": 1.4779036827195467e-06, + "loss": 0.1845, + "step": 41243 + }, + { + "epoch": 15.23, + "learning_rate": 1.4778667323562016e-06, + "loss": 0.3167, + "step": 41244 + }, + { + "epoch": 15.23, + "learning_rate": 1.4778297819928564e-06, + "loss": 0.2403, + "step": 41245 + }, + { + "epoch": 15.23, + "learning_rate": 1.477792831629511e-06, + "loss": 0.235, + "step": 41246 + }, + { + "epoch": 15.23, + "learning_rate": 1.477755881266166e-06, + "loss": 0.4235, + "step": 41247 + }, + { + "epoch": 15.23, + "learning_rate": 1.4777189309028206e-06, + "loss": 0.2823, + "step": 41248 + }, + { + "epoch": 15.23, + "learning_rate": 1.4776819805394752e-06, + "loss": 0.2474, + "step": 41249 + }, + { + "epoch": 15.23, + "learning_rate": 1.47764503017613e-06, + "loss": 0.3037, + "step": 41250 + }, + { + "epoch": 15.23, + "learning_rate": 1.4776080798127849e-06, + "loss": 0.1287, + "step": 41251 + }, + { + "epoch": 15.23, + "learning_rate": 1.4775711294494395e-06, + "loss": 0.2775, + "step": 41252 + }, + { + "epoch": 15.23, + "learning_rate": 1.4775341790860944e-06, + "loss": 0.2266, + "step": 41253 + }, + { + "epoch": 15.23, + "learning_rate": 1.4774972287227492e-06, + "loss": 0.2215, + "step": 41254 + }, + { + "epoch": 15.23, + "learning_rate": 1.4774602783594039e-06, + "loss": 0.2012, + "step": 41255 + }, + { + "epoch": 15.23, + "learning_rate": 1.4774233279960587e-06, + "loss": 0.2799, + "step": 41256 + }, + { + "epoch": 15.24, + "learning_rate": 1.4773863776327136e-06, + "loss": 0.3909, + "step": 41257 + }, + { + "epoch": 15.24, + "learning_rate": 1.4773494272693682e-06, + "loss": 0.3346, + "step": 41258 + }, + { + "epoch": 15.24, + "learning_rate": 1.4773124769060228e-06, + "loss": 0.5032, + "step": 41259 + }, + { + "epoch": 15.24, + "learning_rate": 1.4772755265426777e-06, + "loss": 0.3171, + "step": 41260 + }, + { + "epoch": 15.24, + "learning_rate": 1.4772385761793325e-06, + "loss": 0.3123, + "step": 41261 + }, + { + "epoch": 15.24, + "learning_rate": 1.4772016258159872e-06, + "loss": 0.2927, + "step": 41262 + }, + { + "epoch": 15.24, + "learning_rate": 1.477164675452642e-06, + "loss": 0.1762, + "step": 41263 + }, + { + "epoch": 15.24, + "learning_rate": 1.4771277250892969e-06, + "loss": 0.1761, + "step": 41264 + }, + { + "epoch": 15.24, + "learning_rate": 1.4770907747259515e-06, + "loss": 0.2599, + "step": 41265 + }, + { + "epoch": 15.24, + "learning_rate": 1.4770538243626063e-06, + "loss": 0.2551, + "step": 41266 + }, + { + "epoch": 15.24, + "learning_rate": 1.4770168739992612e-06, + "loss": 0.2808, + "step": 41267 + }, + { + "epoch": 15.24, + "learning_rate": 1.4769799236359156e-06, + "loss": 0.2458, + "step": 41268 + }, + { + "epoch": 15.24, + "learning_rate": 1.4769429732725705e-06, + "loss": 0.1683, + "step": 41269 + }, + { + "epoch": 15.24, + "learning_rate": 1.4769060229092253e-06, + "loss": 0.3044, + "step": 41270 + }, + { + "epoch": 15.24, + "learning_rate": 1.47686907254588e-06, + "loss": 0.3101, + "step": 41271 + }, + { + "epoch": 15.24, + "learning_rate": 1.4768321221825348e-06, + "loss": 0.2656, + "step": 41272 + }, + { + "epoch": 15.24, + "learning_rate": 1.4767951718191896e-06, + "loss": 0.2158, + "step": 41273 + }, + { + "epoch": 15.24, + "learning_rate": 1.4767582214558443e-06, + "loss": 0.1475, + "step": 41274 + }, + { + "epoch": 15.24, + "learning_rate": 1.4767212710924991e-06, + "loss": 0.2661, + "step": 41275 + }, + { + "epoch": 15.24, + "learning_rate": 1.476684320729154e-06, + "loss": 0.2334, + "step": 41276 + }, + { + "epoch": 15.24, + "learning_rate": 1.4766473703658086e-06, + "loss": 0.2528, + "step": 41277 + }, + { + "epoch": 15.24, + "learning_rate": 1.4766104200024633e-06, + "loss": 0.1885, + "step": 41278 + }, + { + "epoch": 15.24, + "learning_rate": 1.476573469639118e-06, + "loss": 0.1682, + "step": 41279 + }, + { + "epoch": 15.24, + "learning_rate": 1.476536519275773e-06, + "loss": 0.2479, + "step": 41280 + }, + { + "epoch": 15.24, + "learning_rate": 1.4764995689124276e-06, + "loss": 0.3024, + "step": 41281 + }, + { + "epoch": 15.24, + "learning_rate": 1.4764626185490824e-06, + "loss": 0.2548, + "step": 41282 + }, + { + "epoch": 15.24, + "learning_rate": 1.4764256681857373e-06, + "loss": 0.3617, + "step": 41283 + }, + { + "epoch": 15.25, + "learning_rate": 1.476388717822392e-06, + "loss": 0.1892, + "step": 41284 + }, + { + "epoch": 15.25, + "learning_rate": 1.4763517674590468e-06, + "loss": 0.2783, + "step": 41285 + }, + { + "epoch": 15.25, + "learning_rate": 1.4763148170957016e-06, + "loss": 0.2121, + "step": 41286 + }, + { + "epoch": 15.25, + "learning_rate": 1.4762778667323562e-06, + "loss": 0.3333, + "step": 41287 + }, + { + "epoch": 15.25, + "learning_rate": 1.4762409163690109e-06, + "loss": 0.1883, + "step": 41288 + }, + { + "epoch": 15.25, + "learning_rate": 1.4762039660056657e-06, + "loss": 0.2724, + "step": 41289 + }, + { + "epoch": 15.25, + "learning_rate": 1.4761670156423206e-06, + "loss": 0.2129, + "step": 41290 + }, + { + "epoch": 15.25, + "learning_rate": 1.4761300652789752e-06, + "loss": 0.2177, + "step": 41291 + }, + { + "epoch": 15.25, + "learning_rate": 1.47609311491563e-06, + "loss": 0.1522, + "step": 41292 + }, + { + "epoch": 15.25, + "learning_rate": 1.476056164552285e-06, + "loss": 0.2242, + "step": 41293 + }, + { + "epoch": 15.25, + "learning_rate": 1.4760192141889396e-06, + "loss": 0.4358, + "step": 41294 + }, + { + "epoch": 15.25, + "learning_rate": 1.4759822638255944e-06, + "loss": 0.2623, + "step": 41295 + }, + { + "epoch": 15.25, + "learning_rate": 1.4759453134622492e-06, + "loss": 0.1355, + "step": 41296 + }, + { + "epoch": 15.25, + "learning_rate": 1.4759083630989039e-06, + "loss": 0.2939, + "step": 41297 + }, + { + "epoch": 15.25, + "learning_rate": 1.4758714127355585e-06, + "loss": 0.2206, + "step": 41298 + }, + { + "epoch": 15.25, + "learning_rate": 1.4758344623722134e-06, + "loss": 0.2082, + "step": 41299 + }, + { + "epoch": 15.25, + "learning_rate": 1.475797512008868e-06, + "loss": 0.1895, + "step": 41300 + }, + { + "epoch": 15.25, + "learning_rate": 1.4757605616455229e-06, + "loss": 0.2814, + "step": 41301 + }, + { + "epoch": 15.25, + "learning_rate": 1.4757236112821777e-06, + "loss": 0.1047, + "step": 41302 + }, + { + "epoch": 15.25, + "learning_rate": 1.4756866609188323e-06, + "loss": 0.2174, + "step": 41303 + }, + { + "epoch": 15.25, + "learning_rate": 1.4756497105554872e-06, + "loss": 0.3717, + "step": 41304 + }, + { + "epoch": 15.25, + "learning_rate": 1.475612760192142e-06, + "loss": 0.2693, + "step": 41305 + }, + { + "epoch": 15.25, + "learning_rate": 1.4755758098287967e-06, + "loss": 0.1584, + "step": 41306 + }, + { + "epoch": 15.25, + "learning_rate": 1.4755388594654515e-06, + "loss": 0.2297, + "step": 41307 + }, + { + "epoch": 15.25, + "learning_rate": 1.4755019091021062e-06, + "loss": 0.1371, + "step": 41308 + }, + { + "epoch": 15.25, + "learning_rate": 1.475464958738761e-06, + "loss": 0.2002, + "step": 41309 + }, + { + "epoch": 15.25, + "learning_rate": 1.4754280083754156e-06, + "loss": 0.2632, + "step": 41310 + }, + { + "epoch": 15.26, + "learning_rate": 1.4753910580120705e-06, + "loss": 0.2393, + "step": 41311 + }, + { + "epoch": 15.26, + "learning_rate": 1.4753541076487253e-06, + "loss": 0.3077, + "step": 41312 + }, + { + "epoch": 15.26, + "learning_rate": 1.47531715728538e-06, + "loss": 0.3649, + "step": 41313 + }, + { + "epoch": 15.26, + "learning_rate": 1.4752802069220348e-06, + "loss": 0.312, + "step": 41314 + }, + { + "epoch": 15.26, + "learning_rate": 1.4752432565586897e-06, + "loss": 0.2862, + "step": 41315 + }, + { + "epoch": 15.26, + "learning_rate": 1.4752063061953443e-06, + "loss": 0.1926, + "step": 41316 + }, + { + "epoch": 15.26, + "learning_rate": 1.4751693558319992e-06, + "loss": 0.3569, + "step": 41317 + }, + { + "epoch": 15.26, + "learning_rate": 1.4751324054686538e-06, + "loss": 0.2711, + "step": 41318 + }, + { + "epoch": 15.26, + "learning_rate": 1.4750954551053084e-06, + "loss": 0.4011, + "step": 41319 + }, + { + "epoch": 15.26, + "learning_rate": 1.4750585047419633e-06, + "loss": 0.2275, + "step": 41320 + }, + { + "epoch": 15.26, + "learning_rate": 1.4750215543786181e-06, + "loss": 0.3076, + "step": 41321 + }, + { + "epoch": 15.26, + "learning_rate": 1.4749846040152728e-06, + "loss": 0.2785, + "step": 41322 + }, + { + "epoch": 15.26, + "learning_rate": 1.4749476536519276e-06, + "loss": 0.2434, + "step": 41323 + }, + { + "epoch": 15.26, + "learning_rate": 1.4749107032885825e-06, + "loss": 0.2821, + "step": 41324 + }, + { + "epoch": 15.26, + "learning_rate": 1.474873752925237e-06, + "loss": 0.1936, + "step": 41325 + }, + { + "epoch": 15.26, + "learning_rate": 1.474836802561892e-06, + "loss": 0.2238, + "step": 41326 + }, + { + "epoch": 15.26, + "learning_rate": 1.4747998521985468e-06, + "loss": 0.2214, + "step": 41327 + }, + { + "epoch": 15.26, + "learning_rate": 1.4747629018352014e-06, + "loss": 0.1139, + "step": 41328 + }, + { + "epoch": 15.26, + "learning_rate": 1.474725951471856e-06, + "loss": 0.1289, + "step": 41329 + }, + { + "epoch": 15.26, + "learning_rate": 1.474689001108511e-06, + "loss": 0.2572, + "step": 41330 + }, + { + "epoch": 15.26, + "learning_rate": 1.4746520507451658e-06, + "loss": 0.1471, + "step": 41331 + }, + { + "epoch": 15.26, + "learning_rate": 1.4746151003818204e-06, + "loss": 0.2513, + "step": 41332 + }, + { + "epoch": 15.26, + "learning_rate": 1.4745781500184752e-06, + "loss": 0.336, + "step": 41333 + }, + { + "epoch": 15.26, + "learning_rate": 1.47454119965513e-06, + "loss": 0.2123, + "step": 41334 + }, + { + "epoch": 15.26, + "learning_rate": 1.4745042492917847e-06, + "loss": 0.2773, + "step": 41335 + }, + { + "epoch": 15.26, + "learning_rate": 1.4744672989284396e-06, + "loss": 0.2219, + "step": 41336 + }, + { + "epoch": 15.26, + "learning_rate": 1.4744303485650944e-06, + "loss": 0.2946, + "step": 41337 + }, + { + "epoch": 15.27, + "learning_rate": 1.4743933982017488e-06, + "loss": 0.2424, + "step": 41338 + }, + { + "epoch": 15.27, + "learning_rate": 1.4743564478384037e-06, + "loss": 0.2751, + "step": 41339 + }, + { + "epoch": 15.27, + "learning_rate": 1.4743194974750585e-06, + "loss": 0.1805, + "step": 41340 + }, + { + "epoch": 15.27, + "learning_rate": 1.4742825471117132e-06, + "loss": 0.2002, + "step": 41341 + }, + { + "epoch": 15.27, + "learning_rate": 1.474245596748368e-06, + "loss": 0.2641, + "step": 41342 + }, + { + "epoch": 15.27, + "learning_rate": 1.4742086463850229e-06, + "loss": 0.2642, + "step": 41343 + }, + { + "epoch": 15.27, + "learning_rate": 1.4741716960216775e-06, + "loss": 0.2574, + "step": 41344 + }, + { + "epoch": 15.27, + "learning_rate": 1.4741347456583324e-06, + "loss": 0.3205, + "step": 41345 + }, + { + "epoch": 15.27, + "learning_rate": 1.4740977952949872e-06, + "loss": 0.237, + "step": 41346 + }, + { + "epoch": 15.27, + "learning_rate": 1.4740608449316418e-06, + "loss": 0.1867, + "step": 41347 + }, + { + "epoch": 15.27, + "learning_rate": 1.4740238945682965e-06, + "loss": 0.3291, + "step": 41348 + }, + { + "epoch": 15.27, + "learning_rate": 1.4739869442049513e-06, + "loss": 0.2669, + "step": 41349 + }, + { + "epoch": 15.27, + "learning_rate": 1.4739499938416062e-06, + "loss": 0.1899, + "step": 41350 + }, + { + "epoch": 15.27, + "learning_rate": 1.4739130434782608e-06, + "loss": 0.1845, + "step": 41351 + }, + { + "epoch": 15.27, + "learning_rate": 1.4738760931149157e-06, + "loss": 0.2217, + "step": 41352 + }, + { + "epoch": 15.27, + "learning_rate": 1.4738391427515705e-06, + "loss": 0.3717, + "step": 41353 + }, + { + "epoch": 15.27, + "learning_rate": 1.4738021923882251e-06, + "loss": 0.2634, + "step": 41354 + }, + { + "epoch": 15.27, + "learning_rate": 1.47376524202488e-06, + "loss": 0.2957, + "step": 41355 + }, + { + "epoch": 15.27, + "learning_rate": 1.4737282916615348e-06, + "loss": 0.4122, + "step": 41356 + }, + { + "epoch": 15.27, + "learning_rate": 1.4736913412981895e-06, + "loss": 0.3054, + "step": 41357 + }, + { + "epoch": 15.27, + "learning_rate": 1.4736543909348441e-06, + "loss": 0.291, + "step": 41358 + }, + { + "epoch": 15.27, + "learning_rate": 1.473617440571499e-06, + "loss": 0.2574, + "step": 41359 + }, + { + "epoch": 15.27, + "learning_rate": 1.4735804902081538e-06, + "loss": 0.2134, + "step": 41360 + }, + { + "epoch": 15.27, + "learning_rate": 1.4735435398448085e-06, + "loss": 0.3744, + "step": 41361 + }, + { + "epoch": 15.27, + "learning_rate": 1.4735065894814633e-06, + "loss": 0.202, + "step": 41362 + }, + { + "epoch": 15.27, + "learning_rate": 1.4734696391181181e-06, + "loss": 0.1091, + "step": 41363 + }, + { + "epoch": 15.27, + "learning_rate": 1.4734326887547728e-06, + "loss": 0.3488, + "step": 41364 + }, + { + "epoch": 15.28, + "learning_rate": 1.4733957383914276e-06, + "loss": 0.2548, + "step": 41365 + }, + { + "epoch": 15.28, + "learning_rate": 1.4733587880280825e-06, + "loss": 0.2241, + "step": 41366 + }, + { + "epoch": 15.28, + "learning_rate": 1.473321837664737e-06, + "loss": 0.3014, + "step": 41367 + }, + { + "epoch": 15.28, + "learning_rate": 1.4732848873013918e-06, + "loss": 0.1269, + "step": 41368 + }, + { + "epoch": 15.28, + "learning_rate": 1.4732479369380466e-06, + "loss": 0.1354, + "step": 41369 + }, + { + "epoch": 15.28, + "learning_rate": 1.4732109865747012e-06, + "loss": 0.2356, + "step": 41370 + }, + { + "epoch": 15.28, + "learning_rate": 1.473174036211356e-06, + "loss": 0.2641, + "step": 41371 + }, + { + "epoch": 15.28, + "learning_rate": 1.473137085848011e-06, + "loss": 0.2899, + "step": 41372 + }, + { + "epoch": 15.28, + "learning_rate": 1.4731001354846656e-06, + "loss": 0.2959, + "step": 41373 + }, + { + "epoch": 15.28, + "learning_rate": 1.4730631851213204e-06, + "loss": 0.3064, + "step": 41374 + }, + { + "epoch": 15.28, + "learning_rate": 1.4730262347579753e-06, + "loss": 0.3497, + "step": 41375 + }, + { + "epoch": 15.28, + "learning_rate": 1.47298928439463e-06, + "loss": 0.1911, + "step": 41376 + }, + { + "epoch": 15.28, + "learning_rate": 1.4729523340312845e-06, + "loss": 0.3398, + "step": 41377 + }, + { + "epoch": 15.28, + "learning_rate": 1.4729153836679394e-06, + "loss": 0.2131, + "step": 41378 + }, + { + "epoch": 15.28, + "learning_rate": 1.4728784333045942e-06, + "loss": 0.2602, + "step": 41379 + }, + { + "epoch": 15.28, + "learning_rate": 1.4728414829412489e-06, + "loss": 0.3172, + "step": 41380 + }, + { + "epoch": 15.28, + "learning_rate": 1.4728045325779037e-06, + "loss": 0.1673, + "step": 41381 + }, + { + "epoch": 15.28, + "learning_rate": 1.4727675822145586e-06, + "loss": 0.2256, + "step": 41382 + }, + { + "epoch": 15.28, + "learning_rate": 1.4727306318512132e-06, + "loss": 0.3414, + "step": 41383 + }, + { + "epoch": 15.28, + "learning_rate": 1.472693681487868e-06, + "loss": 0.2671, + "step": 41384 + }, + { + "epoch": 15.28, + "learning_rate": 1.472656731124523e-06, + "loss": 0.2387, + "step": 41385 + }, + { + "epoch": 15.28, + "learning_rate": 1.4726197807611775e-06, + "loss": 0.3076, + "step": 41386 + }, + { + "epoch": 15.28, + "learning_rate": 1.4725828303978322e-06, + "loss": 0.326, + "step": 41387 + }, + { + "epoch": 15.28, + "learning_rate": 1.472545880034487e-06, + "loss": 0.2954, + "step": 41388 + }, + { + "epoch": 15.28, + "learning_rate": 1.4725089296711417e-06, + "loss": 0.238, + "step": 41389 + }, + { + "epoch": 15.28, + "learning_rate": 1.4724719793077965e-06, + "loss": 0.307, + "step": 41390 + }, + { + "epoch": 15.28, + "learning_rate": 1.4724350289444514e-06, + "loss": 0.4079, + "step": 41391 + }, + { + "epoch": 15.29, + "learning_rate": 1.472398078581106e-06, + "loss": 0.2603, + "step": 41392 + }, + { + "epoch": 15.29, + "learning_rate": 1.4723611282177608e-06, + "loss": 0.2006, + "step": 41393 + }, + { + "epoch": 15.29, + "learning_rate": 1.4723241778544157e-06, + "loss": 0.2432, + "step": 41394 + }, + { + "epoch": 15.29, + "learning_rate": 1.4722872274910703e-06, + "loss": 0.2868, + "step": 41395 + }, + { + "epoch": 15.29, + "learning_rate": 1.4722502771277252e-06, + "loss": 0.185, + "step": 41396 + }, + { + "epoch": 15.29, + "learning_rate": 1.4722133267643798e-06, + "loss": 0.2393, + "step": 41397 + }, + { + "epoch": 15.29, + "learning_rate": 1.4721763764010347e-06, + "loss": 0.4083, + "step": 41398 + }, + { + "epoch": 15.29, + "learning_rate": 1.4721394260376893e-06, + "loss": 0.1203, + "step": 41399 + }, + { + "epoch": 15.29, + "learning_rate": 1.4721024756743441e-06, + "loss": 0.4544, + "step": 41400 + }, + { + "epoch": 15.29, + "learning_rate": 1.472065525310999e-06, + "loss": 0.2455, + "step": 41401 + }, + { + "epoch": 15.29, + "learning_rate": 1.4720285749476536e-06, + "loss": 0.2832, + "step": 41402 + }, + { + "epoch": 15.29, + "learning_rate": 1.4719916245843085e-06, + "loss": 0.4112, + "step": 41403 + }, + { + "epoch": 15.29, + "learning_rate": 1.4719546742209633e-06, + "loss": 0.2483, + "step": 41404 + }, + { + "epoch": 15.29, + "learning_rate": 1.471917723857618e-06, + "loss": 0.2981, + "step": 41405 + }, + { + "epoch": 15.29, + "learning_rate": 1.4718807734942728e-06, + "loss": 0.2558, + "step": 41406 + }, + { + "epoch": 15.29, + "learning_rate": 1.4718438231309274e-06, + "loss": 0.2213, + "step": 41407 + }, + { + "epoch": 15.29, + "learning_rate": 1.4718068727675823e-06, + "loss": 0.2104, + "step": 41408 + }, + { + "epoch": 15.29, + "learning_rate": 1.471769922404237e-06, + "loss": 0.2721, + "step": 41409 + }, + { + "epoch": 15.29, + "learning_rate": 1.4717329720408918e-06, + "loss": 0.1489, + "step": 41410 + }, + { + "epoch": 15.29, + "learning_rate": 1.4716960216775466e-06, + "loss": 0.241, + "step": 41411 + }, + { + "epoch": 15.29, + "learning_rate": 1.4716590713142013e-06, + "loss": 0.2594, + "step": 41412 + }, + { + "epoch": 15.29, + "learning_rate": 1.4716221209508561e-06, + "loss": 0.2724, + "step": 41413 + }, + { + "epoch": 15.29, + "learning_rate": 1.471585170587511e-06, + "loss": 0.2551, + "step": 41414 + }, + { + "epoch": 15.29, + "learning_rate": 1.4715482202241656e-06, + "loss": 0.2927, + "step": 41415 + }, + { + "epoch": 15.29, + "learning_rate": 1.4715112698608204e-06, + "loss": 0.2255, + "step": 41416 + }, + { + "epoch": 15.29, + "learning_rate": 1.471474319497475e-06, + "loss": 0.287, + "step": 41417 + }, + { + "epoch": 15.29, + "learning_rate": 1.4714373691341297e-06, + "loss": 0.3619, + "step": 41418 + }, + { + "epoch": 15.3, + "learning_rate": 1.4714004187707846e-06, + "loss": 0.3089, + "step": 41419 + }, + { + "epoch": 15.3, + "learning_rate": 1.4713634684074394e-06, + "loss": 0.2375, + "step": 41420 + }, + { + "epoch": 15.3, + "learning_rate": 1.471326518044094e-06, + "loss": 0.1973, + "step": 41421 + }, + { + "epoch": 15.3, + "learning_rate": 1.471289567680749e-06, + "loss": 0.274, + "step": 41422 + }, + { + "epoch": 15.3, + "learning_rate": 1.4712526173174037e-06, + "loss": 0.2633, + "step": 41423 + }, + { + "epoch": 15.3, + "learning_rate": 1.4712156669540584e-06, + "loss": 0.3073, + "step": 41424 + }, + { + "epoch": 15.3, + "learning_rate": 1.4711787165907132e-06, + "loss": 0.3262, + "step": 41425 + }, + { + "epoch": 15.3, + "learning_rate": 1.471141766227368e-06, + "loss": 0.2537, + "step": 41426 + }, + { + "epoch": 15.3, + "learning_rate": 1.4711048158640227e-06, + "loss": 0.2663, + "step": 41427 + }, + { + "epoch": 15.3, + "learning_rate": 1.4710678655006774e-06, + "loss": 0.2713, + "step": 41428 + }, + { + "epoch": 15.3, + "learning_rate": 1.4710309151373322e-06, + "loss": 0.1963, + "step": 41429 + }, + { + "epoch": 15.3, + "learning_rate": 1.470993964773987e-06, + "loss": 0.2191, + "step": 41430 + }, + { + "epoch": 15.3, + "learning_rate": 1.4709570144106417e-06, + "loss": 0.2442, + "step": 41431 + }, + { + "epoch": 15.3, + "learning_rate": 1.4709200640472965e-06, + "loss": 0.2528, + "step": 41432 + }, + { + "epoch": 15.3, + "learning_rate": 1.4708831136839514e-06, + "loss": 0.2455, + "step": 41433 + }, + { + "epoch": 15.3, + "learning_rate": 1.470846163320606e-06, + "loss": 0.174, + "step": 41434 + }, + { + "epoch": 15.3, + "learning_rate": 1.4708092129572609e-06, + "loss": 0.2042, + "step": 41435 + }, + { + "epoch": 15.3, + "learning_rate": 1.4707722625939157e-06, + "loss": 0.1661, + "step": 41436 + }, + { + "epoch": 15.3, + "learning_rate": 1.4707353122305701e-06, + "loss": 0.2135, + "step": 41437 + }, + { + "epoch": 15.3, + "learning_rate": 1.470698361867225e-06, + "loss": 0.1654, + "step": 41438 + }, + { + "epoch": 15.3, + "learning_rate": 1.4706614115038798e-06, + "loss": 0.1728, + "step": 41439 + }, + { + "epoch": 15.3, + "learning_rate": 1.4706244611405345e-06, + "loss": 0.3166, + "step": 41440 + }, + { + "epoch": 15.3, + "learning_rate": 1.4705875107771893e-06, + "loss": 0.147, + "step": 41441 + }, + { + "epoch": 15.3, + "learning_rate": 1.4705505604138442e-06, + "loss": 0.2666, + "step": 41442 + }, + { + "epoch": 15.3, + "learning_rate": 1.4705136100504988e-06, + "loss": 0.1999, + "step": 41443 + }, + { + "epoch": 15.3, + "learning_rate": 1.4704766596871537e-06, + "loss": 0.3052, + "step": 41444 + }, + { + "epoch": 15.3, + "learning_rate": 1.4704397093238085e-06, + "loss": 0.3618, + "step": 41445 + }, + { + "epoch": 15.31, + "learning_rate": 1.4704027589604631e-06, + "loss": 0.168, + "step": 41446 + }, + { + "epoch": 15.31, + "learning_rate": 1.4703658085971178e-06, + "loss": 0.3252, + "step": 41447 + }, + { + "epoch": 15.31, + "learning_rate": 1.4703288582337726e-06, + "loss": 0.1774, + "step": 41448 + }, + { + "epoch": 15.31, + "learning_rate": 1.4702919078704275e-06, + "loss": 0.3518, + "step": 41449 + }, + { + "epoch": 15.31, + "learning_rate": 1.470254957507082e-06, + "loss": 0.2829, + "step": 41450 + }, + { + "epoch": 15.31, + "learning_rate": 1.470218007143737e-06, + "loss": 0.2194, + "step": 41451 + }, + { + "epoch": 15.31, + "learning_rate": 1.4701810567803918e-06, + "loss": 0.2778, + "step": 41452 + }, + { + "epoch": 15.31, + "learning_rate": 1.4701441064170464e-06, + "loss": 0.4221, + "step": 41453 + }, + { + "epoch": 15.31, + "learning_rate": 1.4701071560537013e-06, + "loss": 0.1862, + "step": 41454 + }, + { + "epoch": 15.31, + "learning_rate": 1.4700702056903561e-06, + "loss": 0.2892, + "step": 41455 + }, + { + "epoch": 15.31, + "learning_rate": 1.4700332553270108e-06, + "loss": 0.1484, + "step": 41456 + }, + { + "epoch": 15.31, + "learning_rate": 1.4699963049636654e-06, + "loss": 0.1161, + "step": 41457 + }, + { + "epoch": 15.31, + "learning_rate": 1.4699593546003203e-06, + "loss": 0.2891, + "step": 41458 + }, + { + "epoch": 15.31, + "learning_rate": 1.4699224042369749e-06, + "loss": 0.2785, + "step": 41459 + }, + { + "epoch": 15.31, + "learning_rate": 1.4698854538736297e-06, + "loss": 0.1727, + "step": 41460 + }, + { + "epoch": 15.31, + "learning_rate": 1.4698485035102846e-06, + "loss": 0.2867, + "step": 41461 + }, + { + "epoch": 15.31, + "learning_rate": 1.4698115531469392e-06, + "loss": 0.3065, + "step": 41462 + }, + { + "epoch": 15.31, + "learning_rate": 1.469774602783594e-06, + "loss": 0.2972, + "step": 41463 + }, + { + "epoch": 15.31, + "learning_rate": 1.469737652420249e-06, + "loss": 0.2361, + "step": 41464 + }, + { + "epoch": 15.31, + "learning_rate": 1.4697007020569036e-06, + "loss": 0.1698, + "step": 41465 + }, + { + "epoch": 15.31, + "learning_rate": 1.4696637516935584e-06, + "loss": 0.2899, + "step": 41466 + }, + { + "epoch": 15.31, + "learning_rate": 1.469626801330213e-06, + "loss": 0.258, + "step": 41467 + }, + { + "epoch": 15.31, + "learning_rate": 1.4695898509668679e-06, + "loss": 0.2403, + "step": 41468 + }, + { + "epoch": 15.31, + "learning_rate": 1.4695529006035225e-06, + "loss": 0.2535, + "step": 41469 + }, + { + "epoch": 15.31, + "learning_rate": 1.4695159502401774e-06, + "loss": 0.3536, + "step": 41470 + }, + { + "epoch": 15.31, + "learning_rate": 1.4694789998768322e-06, + "loss": 0.304, + "step": 41471 + }, + { + "epoch": 15.31, + "learning_rate": 1.4694420495134869e-06, + "loss": 0.3026, + "step": 41472 + }, + { + "epoch": 15.31, + "learning_rate": 1.4694050991501417e-06, + "loss": 0.2713, + "step": 41473 + }, + { + "epoch": 15.32, + "learning_rate": 1.4693681487867966e-06, + "loss": 0.2447, + "step": 41474 + }, + { + "epoch": 15.32, + "learning_rate": 1.4693311984234512e-06, + "loss": 0.2029, + "step": 41475 + }, + { + "epoch": 15.32, + "learning_rate": 1.469294248060106e-06, + "loss": 0.2903, + "step": 41476 + }, + { + "epoch": 15.32, + "learning_rate": 1.4692572976967607e-06, + "loss": 0.4394, + "step": 41477 + }, + { + "epoch": 15.32, + "learning_rate": 1.4692203473334155e-06, + "loss": 0.3645, + "step": 41478 + }, + { + "epoch": 15.32, + "learning_rate": 1.4691833969700702e-06, + "loss": 0.4081, + "step": 41479 + }, + { + "epoch": 15.32, + "learning_rate": 1.469146446606725e-06, + "loss": 0.1409, + "step": 41480 + }, + { + "epoch": 15.32, + "learning_rate": 1.4691094962433799e-06, + "loss": 0.2434, + "step": 41481 + }, + { + "epoch": 15.32, + "learning_rate": 1.4690725458800345e-06, + "loss": 0.2367, + "step": 41482 + }, + { + "epoch": 15.32, + "learning_rate": 1.4690355955166893e-06, + "loss": 0.2285, + "step": 41483 + }, + { + "epoch": 15.32, + "learning_rate": 1.4689986451533442e-06, + "loss": 0.3181, + "step": 41484 + }, + { + "epoch": 15.32, + "learning_rate": 1.4689616947899988e-06, + "loss": 0.2079, + "step": 41485 + }, + { + "epoch": 15.32, + "learning_rate": 1.4689247444266537e-06, + "loss": 0.2376, + "step": 41486 + }, + { + "epoch": 15.32, + "learning_rate": 1.4688877940633083e-06, + "loss": 0.188, + "step": 41487 + }, + { + "epoch": 15.32, + "learning_rate": 1.468850843699963e-06, + "loss": 0.3013, + "step": 41488 + }, + { + "epoch": 15.32, + "learning_rate": 1.4688138933366178e-06, + "loss": 0.1968, + "step": 41489 + }, + { + "epoch": 15.32, + "learning_rate": 1.4687769429732726e-06, + "loss": 0.2097, + "step": 41490 + }, + { + "epoch": 15.32, + "learning_rate": 1.4687399926099273e-06, + "loss": 0.1857, + "step": 41491 + }, + { + "epoch": 15.32, + "learning_rate": 1.4687030422465821e-06, + "loss": 0.2978, + "step": 41492 + }, + { + "epoch": 15.32, + "learning_rate": 1.468666091883237e-06, + "loss": 0.1955, + "step": 41493 + }, + { + "epoch": 15.32, + "learning_rate": 1.4686291415198916e-06, + "loss": 0.1414, + "step": 41494 + }, + { + "epoch": 15.32, + "learning_rate": 1.4685921911565465e-06, + "loss": 0.1927, + "step": 41495 + }, + { + "epoch": 15.32, + "learning_rate": 1.4685552407932013e-06, + "loss": 0.1423, + "step": 41496 + }, + { + "epoch": 15.32, + "learning_rate": 1.468518290429856e-06, + "loss": 0.1544, + "step": 41497 + }, + { + "epoch": 15.32, + "learning_rate": 1.4684813400665106e-06, + "loss": 0.2008, + "step": 41498 + }, + { + "epoch": 15.32, + "learning_rate": 1.4684443897031654e-06, + "loss": 0.2082, + "step": 41499 + }, + { + "epoch": 15.32, + "learning_rate": 1.4684074393398203e-06, + "loss": 0.3871, + "step": 41500 + }, + { + "epoch": 15.33, + "learning_rate": 1.468370488976475e-06, + "loss": 0.3501, + "step": 41501 + }, + { + "epoch": 15.33, + "learning_rate": 1.4683335386131298e-06, + "loss": 0.3168, + "step": 41502 + }, + { + "epoch": 15.33, + "learning_rate": 1.4682965882497846e-06, + "loss": 0.3586, + "step": 41503 + }, + { + "epoch": 15.33, + "learning_rate": 1.4682596378864392e-06, + "loss": 0.1633, + "step": 41504 + }, + { + "epoch": 15.33, + "learning_rate": 1.468222687523094e-06, + "loss": 0.211, + "step": 41505 + }, + { + "epoch": 15.33, + "learning_rate": 1.468185737159749e-06, + "loss": 0.2919, + "step": 41506 + }, + { + "epoch": 15.33, + "learning_rate": 1.4681487867964034e-06, + "loss": 0.2956, + "step": 41507 + }, + { + "epoch": 15.33, + "learning_rate": 1.4681118364330582e-06, + "loss": 0.1887, + "step": 41508 + }, + { + "epoch": 15.33, + "learning_rate": 1.468074886069713e-06, + "loss": 0.1822, + "step": 41509 + }, + { + "epoch": 15.33, + "learning_rate": 1.4680379357063677e-06, + "loss": 0.3674, + "step": 41510 + }, + { + "epoch": 15.33, + "learning_rate": 1.4680009853430226e-06, + "loss": 0.1812, + "step": 41511 + }, + { + "epoch": 15.33, + "learning_rate": 1.4679640349796774e-06, + "loss": 0.2914, + "step": 41512 + }, + { + "epoch": 15.33, + "learning_rate": 1.467927084616332e-06, + "loss": 0.3131, + "step": 41513 + }, + { + "epoch": 15.33, + "learning_rate": 1.4678901342529869e-06, + "loss": 0.2382, + "step": 41514 + }, + { + "epoch": 15.33, + "learning_rate": 1.4678531838896417e-06, + "loss": 0.2055, + "step": 41515 + }, + { + "epoch": 15.33, + "learning_rate": 1.4678162335262964e-06, + "loss": 0.1538, + "step": 41516 + }, + { + "epoch": 15.33, + "learning_rate": 1.467779283162951e-06, + "loss": 0.1866, + "step": 41517 + }, + { + "epoch": 15.33, + "learning_rate": 1.4677423327996059e-06, + "loss": 0.2408, + "step": 41518 + }, + { + "epoch": 15.33, + "learning_rate": 1.4677053824362607e-06, + "loss": 0.1489, + "step": 41519 + }, + { + "epoch": 15.33, + "learning_rate": 1.4676684320729153e-06, + "loss": 0.2231, + "step": 41520 + }, + { + "epoch": 15.33, + "learning_rate": 1.4676314817095702e-06, + "loss": 0.2146, + "step": 41521 + }, + { + "epoch": 15.33, + "learning_rate": 1.467594531346225e-06, + "loss": 0.2552, + "step": 41522 + }, + { + "epoch": 15.33, + "learning_rate": 1.4675575809828797e-06, + "loss": 0.1652, + "step": 41523 + }, + { + "epoch": 15.33, + "learning_rate": 1.4675206306195345e-06, + "loss": 0.2071, + "step": 41524 + }, + { + "epoch": 15.33, + "learning_rate": 1.4674836802561894e-06, + "loss": 0.2698, + "step": 41525 + }, + { + "epoch": 15.33, + "learning_rate": 1.467446729892844e-06, + "loss": 0.3931, + "step": 41526 + }, + { + "epoch": 15.33, + "learning_rate": 1.4674097795294986e-06, + "loss": 0.3072, + "step": 41527 + }, + { + "epoch": 15.34, + "learning_rate": 1.4673728291661535e-06, + "loss": 0.2432, + "step": 41528 + }, + { + "epoch": 15.34, + "learning_rate": 1.4673358788028081e-06, + "loss": 0.331, + "step": 41529 + }, + { + "epoch": 15.34, + "learning_rate": 1.467298928439463e-06, + "loss": 0.3402, + "step": 41530 + }, + { + "epoch": 15.34, + "learning_rate": 1.4672619780761178e-06, + "loss": 0.2029, + "step": 41531 + }, + { + "epoch": 15.34, + "learning_rate": 1.4672250277127725e-06, + "loss": 0.3944, + "step": 41532 + }, + { + "epoch": 15.34, + "learning_rate": 1.4671880773494273e-06, + "loss": 0.2848, + "step": 41533 + }, + { + "epoch": 15.34, + "learning_rate": 1.4671511269860822e-06, + "loss": 0.2669, + "step": 41534 + }, + { + "epoch": 15.34, + "learning_rate": 1.4671141766227368e-06, + "loss": 0.2112, + "step": 41535 + }, + { + "epoch": 15.34, + "learning_rate": 1.4670772262593916e-06, + "loss": 0.1655, + "step": 41536 + }, + { + "epoch": 15.34, + "learning_rate": 1.4670402758960463e-06, + "loss": 0.2579, + "step": 41537 + }, + { + "epoch": 15.34, + "learning_rate": 1.4670033255327011e-06, + "loss": 0.2968, + "step": 41538 + }, + { + "epoch": 15.34, + "learning_rate": 1.4669663751693558e-06, + "loss": 0.1766, + "step": 41539 + }, + { + "epoch": 15.34, + "learning_rate": 1.4669294248060106e-06, + "loss": 0.2454, + "step": 41540 + }, + { + "epoch": 15.34, + "learning_rate": 1.4668924744426655e-06, + "loss": 0.2887, + "step": 41541 + }, + { + "epoch": 15.34, + "learning_rate": 1.46685552407932e-06, + "loss": 0.185, + "step": 41542 + }, + { + "epoch": 15.34, + "learning_rate": 1.466818573715975e-06, + "loss": 0.1677, + "step": 41543 + }, + { + "epoch": 15.34, + "learning_rate": 1.4667816233526298e-06, + "loss": 0.4264, + "step": 41544 + }, + { + "epoch": 15.34, + "learning_rate": 1.4667446729892844e-06, + "loss": 0.2724, + "step": 41545 + }, + { + "epoch": 15.34, + "learning_rate": 1.4667077226259393e-06, + "loss": 0.2802, + "step": 41546 + }, + { + "epoch": 15.34, + "learning_rate": 1.466670772262594e-06, + "loss": 0.2364, + "step": 41547 + }, + { + "epoch": 15.34, + "learning_rate": 1.4666338218992488e-06, + "loss": 0.3246, + "step": 41548 + }, + { + "epoch": 15.34, + "learning_rate": 1.4665968715359034e-06, + "loss": 0.2871, + "step": 41549 + }, + { + "epoch": 15.34, + "learning_rate": 1.4665599211725582e-06, + "loss": 0.2986, + "step": 41550 + }, + { + "epoch": 15.34, + "learning_rate": 1.466522970809213e-06, + "loss": 0.3815, + "step": 41551 + }, + { + "epoch": 15.34, + "learning_rate": 1.4664860204458677e-06, + "loss": 0.3749, + "step": 41552 + }, + { + "epoch": 15.34, + "learning_rate": 1.4664490700825226e-06, + "loss": 0.1526, + "step": 41553 + }, + { + "epoch": 15.34, + "learning_rate": 1.4664121197191774e-06, + "loss": 0.2464, + "step": 41554 + }, + { + "epoch": 15.35, + "learning_rate": 1.466375169355832e-06, + "loss": 0.3388, + "step": 41555 + }, + { + "epoch": 15.35, + "learning_rate": 1.466338218992487e-06, + "loss": 0.2331, + "step": 41556 + }, + { + "epoch": 15.35, + "learning_rate": 1.4663012686291415e-06, + "loss": 0.3443, + "step": 41557 + }, + { + "epoch": 15.35, + "learning_rate": 1.4662643182657962e-06, + "loss": 0.3001, + "step": 41558 + }, + { + "epoch": 15.35, + "learning_rate": 1.466227367902451e-06, + "loss": 0.4377, + "step": 41559 + }, + { + "epoch": 15.35, + "learning_rate": 1.4661904175391059e-06, + "loss": 0.19, + "step": 41560 + }, + { + "epoch": 15.35, + "learning_rate": 1.4661534671757605e-06, + "loss": 0.2881, + "step": 41561 + }, + { + "epoch": 15.35, + "learning_rate": 1.4661165168124154e-06, + "loss": 0.2546, + "step": 41562 + }, + { + "epoch": 15.35, + "learning_rate": 1.4660795664490702e-06, + "loss": 0.2204, + "step": 41563 + }, + { + "epoch": 15.35, + "learning_rate": 1.4660426160857248e-06, + "loss": 0.253, + "step": 41564 + }, + { + "epoch": 15.35, + "learning_rate": 1.4660056657223797e-06, + "loss": 0.3423, + "step": 41565 + }, + { + "epoch": 15.35, + "learning_rate": 1.4659687153590345e-06, + "loss": 0.2985, + "step": 41566 + }, + { + "epoch": 15.35, + "learning_rate": 1.4659317649956892e-06, + "loss": 0.2145, + "step": 41567 + }, + { + "epoch": 15.35, + "learning_rate": 1.4658948146323438e-06, + "loss": 0.3967, + "step": 41568 + }, + { + "epoch": 15.35, + "learning_rate": 1.4658578642689987e-06, + "loss": 0.3411, + "step": 41569 + }, + { + "epoch": 15.35, + "learning_rate": 1.4658209139056535e-06, + "loss": 0.2961, + "step": 41570 + }, + { + "epoch": 15.35, + "learning_rate": 1.4657839635423081e-06, + "loss": 0.3878, + "step": 41571 + }, + { + "epoch": 15.35, + "learning_rate": 1.465747013178963e-06, + "loss": 0.2094, + "step": 41572 + }, + { + "epoch": 15.35, + "learning_rate": 1.4657100628156178e-06, + "loss": 0.2511, + "step": 41573 + }, + { + "epoch": 15.35, + "learning_rate": 1.4656731124522725e-06, + "loss": 0.2111, + "step": 41574 + }, + { + "epoch": 15.35, + "learning_rate": 1.4656361620889273e-06, + "loss": 0.2015, + "step": 41575 + }, + { + "epoch": 15.35, + "learning_rate": 1.465599211725582e-06, + "loss": 0.278, + "step": 41576 + }, + { + "epoch": 15.35, + "learning_rate": 1.4655622613622366e-06, + "loss": 0.3309, + "step": 41577 + }, + { + "epoch": 15.35, + "learning_rate": 1.4655253109988915e-06, + "loss": 0.285, + "step": 41578 + }, + { + "epoch": 15.35, + "learning_rate": 1.4654883606355463e-06, + "loss": 0.1748, + "step": 41579 + }, + { + "epoch": 15.35, + "learning_rate": 1.465451410272201e-06, + "loss": 0.2946, + "step": 41580 + }, + { + "epoch": 15.35, + "learning_rate": 1.4654144599088558e-06, + "loss": 0.3721, + "step": 41581 + }, + { + "epoch": 15.36, + "learning_rate": 1.4653775095455106e-06, + "loss": 0.1932, + "step": 41582 + }, + { + "epoch": 15.36, + "learning_rate": 1.4653405591821653e-06, + "loss": 0.29, + "step": 41583 + }, + { + "epoch": 15.36, + "learning_rate": 1.4653036088188201e-06, + "loss": 0.2436, + "step": 41584 + }, + { + "epoch": 15.36, + "learning_rate": 1.465266658455475e-06, + "loss": 0.3297, + "step": 41585 + }, + { + "epoch": 15.36, + "learning_rate": 1.4652297080921296e-06, + "loss": 0.2366, + "step": 41586 + }, + { + "epoch": 15.36, + "learning_rate": 1.4651927577287842e-06, + "loss": 0.2094, + "step": 41587 + }, + { + "epoch": 15.36, + "learning_rate": 1.465155807365439e-06, + "loss": 0.3475, + "step": 41588 + }, + { + "epoch": 15.36, + "learning_rate": 1.465118857002094e-06, + "loss": 0.2401, + "step": 41589 + }, + { + "epoch": 15.36, + "learning_rate": 1.4650819066387486e-06, + "loss": 0.2456, + "step": 41590 + }, + { + "epoch": 15.36, + "learning_rate": 1.4650449562754034e-06, + "loss": 0.2268, + "step": 41591 + }, + { + "epoch": 15.36, + "learning_rate": 1.4650080059120583e-06, + "loss": 0.2576, + "step": 41592 + }, + { + "epoch": 15.36, + "learning_rate": 1.464971055548713e-06, + "loss": 0.2641, + "step": 41593 + }, + { + "epoch": 15.36, + "learning_rate": 1.4649341051853678e-06, + "loss": 0.3378, + "step": 41594 + }, + { + "epoch": 15.36, + "learning_rate": 1.4648971548220226e-06, + "loss": 0.2505, + "step": 41595 + }, + { + "epoch": 15.36, + "learning_rate": 1.4648602044586772e-06, + "loss": 0.2675, + "step": 41596 + }, + { + "epoch": 15.36, + "learning_rate": 1.4648232540953319e-06, + "loss": 0.3157, + "step": 41597 + }, + { + "epoch": 15.36, + "learning_rate": 1.4647863037319867e-06, + "loss": 0.2769, + "step": 41598 + }, + { + "epoch": 15.36, + "learning_rate": 1.4647493533686416e-06, + "loss": 0.2594, + "step": 41599 + }, + { + "epoch": 15.36, + "learning_rate": 1.4647124030052962e-06, + "loss": 0.1675, + "step": 41600 + }, + { + "epoch": 15.36, + "learning_rate": 1.464675452641951e-06, + "loss": 0.1884, + "step": 41601 + }, + { + "epoch": 15.36, + "learning_rate": 1.4646385022786057e-06, + "loss": 0.3041, + "step": 41602 + }, + { + "epoch": 15.36, + "learning_rate": 1.4646015519152605e-06, + "loss": 0.2144, + "step": 41603 + }, + { + "epoch": 15.36, + "learning_rate": 1.4645646015519154e-06, + "loss": 0.2428, + "step": 41604 + }, + { + "epoch": 15.36, + "learning_rate": 1.46452765118857e-06, + "loss": 0.2367, + "step": 41605 + }, + { + "epoch": 15.36, + "learning_rate": 1.4644907008252247e-06, + "loss": 0.3305, + "step": 41606 + }, + { + "epoch": 15.36, + "learning_rate": 1.4644537504618795e-06, + "loss": 0.2641, + "step": 41607 + }, + { + "epoch": 15.36, + "learning_rate": 1.4644168000985344e-06, + "loss": 0.1676, + "step": 41608 + }, + { + "epoch": 15.37, + "learning_rate": 1.464379849735189e-06, + "loss": 0.2742, + "step": 41609 + }, + { + "epoch": 15.37, + "learning_rate": 1.4643428993718438e-06, + "loss": 0.4021, + "step": 41610 + }, + { + "epoch": 15.37, + "learning_rate": 1.4643059490084987e-06, + "loss": 0.3238, + "step": 41611 + }, + { + "epoch": 15.37, + "learning_rate": 1.4642689986451533e-06, + "loss": 0.1869, + "step": 41612 + }, + { + "epoch": 15.37, + "learning_rate": 1.4642320482818082e-06, + "loss": 0.3153, + "step": 41613 + }, + { + "epoch": 15.37, + "learning_rate": 1.464195097918463e-06, + "loss": 0.412, + "step": 41614 + }, + { + "epoch": 15.37, + "learning_rate": 1.4641581475551177e-06, + "loss": 0.3057, + "step": 41615 + }, + { + "epoch": 15.37, + "learning_rate": 1.4641211971917723e-06, + "loss": 0.2146, + "step": 41616 + }, + { + "epoch": 15.37, + "learning_rate": 1.4640842468284271e-06, + "loss": 0.2098, + "step": 41617 + }, + { + "epoch": 15.37, + "learning_rate": 1.464047296465082e-06, + "loss": 0.2087, + "step": 41618 + }, + { + "epoch": 15.37, + "learning_rate": 1.4640103461017366e-06, + "loss": 0.3373, + "step": 41619 + }, + { + "epoch": 15.37, + "learning_rate": 1.4639733957383915e-06, + "loss": 0.2093, + "step": 41620 + }, + { + "epoch": 15.37, + "learning_rate": 1.4639364453750463e-06, + "loss": 0.1574, + "step": 41621 + }, + { + "epoch": 15.37, + "learning_rate": 1.463899495011701e-06, + "loss": 0.1697, + "step": 41622 + }, + { + "epoch": 15.37, + "learning_rate": 1.4638625446483558e-06, + "loss": 0.2728, + "step": 41623 + }, + { + "epoch": 15.37, + "learning_rate": 1.4638255942850107e-06, + "loss": 0.2087, + "step": 41624 + }, + { + "epoch": 15.37, + "learning_rate": 1.4637886439216653e-06, + "loss": 0.5147, + "step": 41625 + }, + { + "epoch": 15.37, + "learning_rate": 1.46375169355832e-06, + "loss": 0.1829, + "step": 41626 + }, + { + "epoch": 15.37, + "learning_rate": 1.4637147431949748e-06, + "loss": 0.3346, + "step": 41627 + }, + { + "epoch": 15.37, + "learning_rate": 1.4636777928316294e-06, + "loss": 0.3118, + "step": 41628 + }, + { + "epoch": 15.37, + "learning_rate": 1.4636408424682843e-06, + "loss": 0.3897, + "step": 41629 + }, + { + "epoch": 15.37, + "learning_rate": 1.4636038921049391e-06, + "loss": 0.318, + "step": 41630 + }, + { + "epoch": 15.37, + "learning_rate": 1.4635669417415937e-06, + "loss": 0.3114, + "step": 41631 + }, + { + "epoch": 15.37, + "learning_rate": 1.4635299913782486e-06, + "loss": 0.2035, + "step": 41632 + }, + { + "epoch": 15.37, + "learning_rate": 1.4634930410149034e-06, + "loss": 0.285, + "step": 41633 + }, + { + "epoch": 15.37, + "learning_rate": 1.463456090651558e-06, + "loss": 0.2008, + "step": 41634 + }, + { + "epoch": 15.37, + "learning_rate": 1.463419140288213e-06, + "loss": 0.2985, + "step": 41635 + }, + { + "epoch": 15.38, + "learning_rate": 1.4633821899248676e-06, + "loss": 0.1598, + "step": 41636 + }, + { + "epoch": 15.38, + "learning_rate": 1.4633452395615224e-06, + "loss": 0.194, + "step": 41637 + }, + { + "epoch": 15.38, + "learning_rate": 1.463308289198177e-06, + "loss": 0.3599, + "step": 41638 + }, + { + "epoch": 15.38, + "learning_rate": 1.463271338834832e-06, + "loss": 0.1895, + "step": 41639 + }, + { + "epoch": 15.38, + "learning_rate": 1.4632343884714867e-06, + "loss": 0.2943, + "step": 41640 + }, + { + "epoch": 15.38, + "learning_rate": 1.4631974381081414e-06, + "loss": 0.1779, + "step": 41641 + }, + { + "epoch": 15.38, + "learning_rate": 1.4631604877447962e-06, + "loss": 0.3131, + "step": 41642 + }, + { + "epoch": 15.38, + "learning_rate": 1.463123537381451e-06, + "loss": 0.2054, + "step": 41643 + }, + { + "epoch": 15.38, + "learning_rate": 1.4630865870181057e-06, + "loss": 0.2584, + "step": 41644 + }, + { + "epoch": 15.38, + "learning_rate": 1.4630496366547606e-06, + "loss": 0.2329, + "step": 41645 + }, + { + "epoch": 15.38, + "learning_rate": 1.4630126862914152e-06, + "loss": 0.2353, + "step": 41646 + }, + { + "epoch": 15.38, + "learning_rate": 1.4629757359280698e-06, + "loss": 0.2142, + "step": 41647 + }, + { + "epoch": 15.38, + "learning_rate": 1.4629387855647247e-06, + "loss": 0.279, + "step": 41648 + }, + { + "epoch": 15.38, + "learning_rate": 1.4629018352013795e-06, + "loss": 0.2637, + "step": 41649 + }, + { + "epoch": 15.38, + "learning_rate": 1.4628648848380342e-06, + "loss": 0.2085, + "step": 41650 + }, + { + "epoch": 15.38, + "learning_rate": 1.462827934474689e-06, + "loss": 0.3121, + "step": 41651 + }, + { + "epoch": 15.38, + "learning_rate": 1.4627909841113439e-06, + "loss": 0.2988, + "step": 41652 + }, + { + "epoch": 15.38, + "learning_rate": 1.4627540337479985e-06, + "loss": 0.1656, + "step": 41653 + }, + { + "epoch": 15.38, + "learning_rate": 1.4627170833846533e-06, + "loss": 0.3032, + "step": 41654 + }, + { + "epoch": 15.38, + "learning_rate": 1.4626801330213082e-06, + "loss": 0.2924, + "step": 41655 + }, + { + "epoch": 15.38, + "learning_rate": 1.4626431826579628e-06, + "loss": 0.2374, + "step": 41656 + }, + { + "epoch": 15.38, + "learning_rate": 1.4626062322946175e-06, + "loss": 0.1958, + "step": 41657 + }, + { + "epoch": 15.38, + "learning_rate": 1.4625692819312723e-06, + "loss": 0.4042, + "step": 41658 + }, + { + "epoch": 15.38, + "learning_rate": 1.4625323315679272e-06, + "loss": 0.2627, + "step": 41659 + }, + { + "epoch": 15.38, + "learning_rate": 1.4624953812045818e-06, + "loss": 0.2663, + "step": 41660 + }, + { + "epoch": 15.38, + "learning_rate": 1.4624584308412367e-06, + "loss": 0.3157, + "step": 41661 + }, + { + "epoch": 15.38, + "learning_rate": 1.4624214804778915e-06, + "loss": 0.2091, + "step": 41662 + }, + { + "epoch": 15.39, + "learning_rate": 1.4623845301145461e-06, + "loss": 0.2759, + "step": 41663 + }, + { + "epoch": 15.39, + "learning_rate": 1.462347579751201e-06, + "loss": 0.2256, + "step": 41664 + }, + { + "epoch": 15.39, + "learning_rate": 1.4623106293878558e-06, + "loss": 0.2741, + "step": 41665 + }, + { + "epoch": 15.39, + "learning_rate": 1.4622736790245105e-06, + "loss": 0.3401, + "step": 41666 + }, + { + "epoch": 15.39, + "learning_rate": 1.462236728661165e-06, + "loss": 0.2478, + "step": 41667 + }, + { + "epoch": 15.39, + "learning_rate": 1.46219977829782e-06, + "loss": 0.3565, + "step": 41668 + }, + { + "epoch": 15.39, + "learning_rate": 1.4621628279344748e-06, + "loss": 0.2617, + "step": 41669 + }, + { + "epoch": 15.39, + "learning_rate": 1.4621258775711294e-06, + "loss": 0.2542, + "step": 41670 + }, + { + "epoch": 15.39, + "learning_rate": 1.4620889272077843e-06, + "loss": 0.4706, + "step": 41671 + }, + { + "epoch": 15.39, + "learning_rate": 1.4620519768444391e-06, + "loss": 0.2623, + "step": 41672 + }, + { + "epoch": 15.39, + "learning_rate": 1.4620150264810938e-06, + "loss": 0.269, + "step": 41673 + }, + { + "epoch": 15.39, + "learning_rate": 1.4619780761177486e-06, + "loss": 0.1853, + "step": 41674 + }, + { + "epoch": 15.39, + "learning_rate": 1.4619411257544035e-06, + "loss": 0.2213, + "step": 41675 + }, + { + "epoch": 15.39, + "learning_rate": 1.4619041753910579e-06, + "loss": 0.2362, + "step": 41676 + }, + { + "epoch": 15.39, + "learning_rate": 1.4618672250277127e-06, + "loss": 0.3412, + "step": 41677 + }, + { + "epoch": 15.39, + "learning_rate": 1.4618302746643676e-06, + "loss": 0.2245, + "step": 41678 + }, + { + "epoch": 15.39, + "learning_rate": 1.4617933243010222e-06, + "loss": 0.1616, + "step": 41679 + }, + { + "epoch": 15.39, + "learning_rate": 1.461756373937677e-06, + "loss": 0.2103, + "step": 41680 + }, + { + "epoch": 15.39, + "learning_rate": 1.461719423574332e-06, + "loss": 0.2131, + "step": 41681 + }, + { + "epoch": 15.39, + "learning_rate": 1.4616824732109866e-06, + "loss": 0.2201, + "step": 41682 + }, + { + "epoch": 15.39, + "learning_rate": 1.4616455228476414e-06, + "loss": 0.2868, + "step": 41683 + }, + { + "epoch": 15.39, + "learning_rate": 1.4616085724842963e-06, + "loss": 0.1902, + "step": 41684 + }, + { + "epoch": 15.39, + "learning_rate": 1.4615716221209509e-06, + "loss": 0.1785, + "step": 41685 + }, + { + "epoch": 15.39, + "learning_rate": 1.4615346717576055e-06, + "loss": 0.2377, + "step": 41686 + }, + { + "epoch": 15.39, + "learning_rate": 1.4614977213942604e-06, + "loss": 0.2905, + "step": 41687 + }, + { + "epoch": 15.39, + "learning_rate": 1.4614607710309152e-06, + "loss": 0.2572, + "step": 41688 + }, + { + "epoch": 15.39, + "learning_rate": 1.4614238206675699e-06, + "loss": 0.2797, + "step": 41689 + }, + { + "epoch": 15.4, + "learning_rate": 1.4613868703042247e-06, + "loss": 0.1547, + "step": 41690 + }, + { + "epoch": 15.4, + "learning_rate": 1.4613499199408796e-06, + "loss": 0.3227, + "step": 41691 + }, + { + "epoch": 15.4, + "learning_rate": 1.4613129695775342e-06, + "loss": 0.1912, + "step": 41692 + }, + { + "epoch": 15.4, + "learning_rate": 1.461276019214189e-06, + "loss": 0.2246, + "step": 41693 + }, + { + "epoch": 15.4, + "learning_rate": 1.4612390688508439e-06, + "loss": 0.3147, + "step": 41694 + }, + { + "epoch": 15.4, + "learning_rate": 1.4612021184874985e-06, + "loss": 0.2555, + "step": 41695 + }, + { + "epoch": 15.4, + "learning_rate": 1.4611651681241532e-06, + "loss": 0.3063, + "step": 41696 + }, + { + "epoch": 15.4, + "learning_rate": 1.461128217760808e-06, + "loss": 0.3071, + "step": 41697 + }, + { + "epoch": 15.4, + "learning_rate": 1.4610912673974626e-06, + "loss": 0.2303, + "step": 41698 + }, + { + "epoch": 15.4, + "learning_rate": 1.4610543170341175e-06, + "loss": 0.2266, + "step": 41699 + }, + { + "epoch": 15.4, + "learning_rate": 1.4610173666707723e-06, + "loss": 0.247, + "step": 41700 + }, + { + "epoch": 15.4, + "learning_rate": 1.460980416307427e-06, + "loss": 0.2778, + "step": 41701 + }, + { + "epoch": 15.4, + "learning_rate": 1.4609434659440818e-06, + "loss": 0.2467, + "step": 41702 + }, + { + "epoch": 15.4, + "learning_rate": 1.4609065155807367e-06, + "loss": 0.1361, + "step": 41703 + }, + { + "epoch": 15.4, + "learning_rate": 1.4608695652173913e-06, + "loss": 0.2097, + "step": 41704 + }, + { + "epoch": 15.4, + "learning_rate": 1.4608326148540462e-06, + "loss": 0.2496, + "step": 41705 + }, + { + "epoch": 15.4, + "learning_rate": 1.4607956644907008e-06, + "loss": 0.2664, + "step": 41706 + }, + { + "epoch": 15.4, + "learning_rate": 1.4607587141273556e-06, + "loss": 0.1659, + "step": 41707 + }, + { + "epoch": 15.4, + "learning_rate": 1.4607217637640103e-06, + "loss": 0.2068, + "step": 41708 + }, + { + "epoch": 15.4, + "learning_rate": 1.4606848134006651e-06, + "loss": 0.2193, + "step": 41709 + }, + { + "epoch": 15.4, + "learning_rate": 1.46064786303732e-06, + "loss": 0.2214, + "step": 41710 + }, + { + "epoch": 15.4, + "learning_rate": 1.4606109126739746e-06, + "loss": 0.1508, + "step": 41711 + }, + { + "epoch": 15.4, + "learning_rate": 1.4605739623106295e-06, + "loss": 0.3137, + "step": 41712 + }, + { + "epoch": 15.4, + "learning_rate": 1.4605370119472843e-06, + "loss": 0.1302, + "step": 41713 + }, + { + "epoch": 15.4, + "learning_rate": 1.460500061583939e-06, + "loss": 0.276, + "step": 41714 + }, + { + "epoch": 15.4, + "learning_rate": 1.4604631112205938e-06, + "loss": 0.1958, + "step": 41715 + }, + { + "epoch": 15.4, + "learning_rate": 1.4604261608572484e-06, + "loss": 0.231, + "step": 41716 + }, + { + "epoch": 15.41, + "learning_rate": 1.460389210493903e-06, + "loss": 0.3867, + "step": 41717 + }, + { + "epoch": 15.41, + "learning_rate": 1.460352260130558e-06, + "loss": 0.3285, + "step": 41718 + }, + { + "epoch": 15.41, + "learning_rate": 1.4603153097672128e-06, + "loss": 0.2779, + "step": 41719 + }, + { + "epoch": 15.41, + "learning_rate": 1.4602783594038674e-06, + "loss": 0.2957, + "step": 41720 + }, + { + "epoch": 15.41, + "learning_rate": 1.4602414090405222e-06, + "loss": 0.2858, + "step": 41721 + }, + { + "epoch": 15.41, + "learning_rate": 1.460204458677177e-06, + "loss": 0.289, + "step": 41722 + }, + { + "epoch": 15.41, + "learning_rate": 1.4601675083138317e-06, + "loss": 0.2488, + "step": 41723 + }, + { + "epoch": 15.41, + "learning_rate": 1.4601305579504866e-06, + "loss": 0.1585, + "step": 41724 + }, + { + "epoch": 15.41, + "learning_rate": 1.4600936075871414e-06, + "loss": 0.4291, + "step": 41725 + }, + { + "epoch": 15.41, + "learning_rate": 1.460056657223796e-06, + "loss": 0.32, + "step": 41726 + }, + { + "epoch": 15.41, + "learning_rate": 1.4600197068604507e-06, + "loss": 0.2182, + "step": 41727 + }, + { + "epoch": 15.41, + "learning_rate": 1.4599827564971056e-06, + "loss": 0.1853, + "step": 41728 + }, + { + "epoch": 15.41, + "learning_rate": 1.4599458061337604e-06, + "loss": 0.22, + "step": 41729 + }, + { + "epoch": 15.41, + "learning_rate": 1.459908855770415e-06, + "loss": 0.1872, + "step": 41730 + }, + { + "epoch": 15.41, + "learning_rate": 1.4598719054070699e-06, + "loss": 0.1995, + "step": 41731 + }, + { + "epoch": 15.41, + "learning_rate": 1.4598349550437247e-06, + "loss": 0.1717, + "step": 41732 + }, + { + "epoch": 15.41, + "learning_rate": 1.4597980046803794e-06, + "loss": 0.3732, + "step": 41733 + }, + { + "epoch": 15.41, + "learning_rate": 1.4597610543170342e-06, + "loss": 0.3108, + "step": 41734 + }, + { + "epoch": 15.41, + "learning_rate": 1.459724103953689e-06, + "loss": 0.2503, + "step": 41735 + }, + { + "epoch": 15.41, + "learning_rate": 1.4596871535903437e-06, + "loss": 0.3124, + "step": 41736 + }, + { + "epoch": 15.41, + "learning_rate": 1.4596502032269983e-06, + "loss": 0.2997, + "step": 41737 + }, + { + "epoch": 15.41, + "learning_rate": 1.4596132528636532e-06, + "loss": 0.3163, + "step": 41738 + }, + { + "epoch": 15.41, + "learning_rate": 1.459576302500308e-06, + "loss": 0.1812, + "step": 41739 + }, + { + "epoch": 15.41, + "learning_rate": 1.4595393521369627e-06, + "loss": 0.2076, + "step": 41740 + }, + { + "epoch": 15.41, + "learning_rate": 1.4595024017736175e-06, + "loss": 0.2132, + "step": 41741 + }, + { + "epoch": 15.41, + "learning_rate": 1.4594654514102724e-06, + "loss": 0.2405, + "step": 41742 + }, + { + "epoch": 15.41, + "learning_rate": 1.459428501046927e-06, + "loss": 0.2677, + "step": 41743 + }, + { + "epoch": 15.42, + "learning_rate": 1.4593915506835819e-06, + "loss": 0.2533, + "step": 41744 + }, + { + "epoch": 15.42, + "learning_rate": 1.4593546003202367e-06, + "loss": 0.2896, + "step": 41745 + }, + { + "epoch": 15.42, + "learning_rate": 1.4593176499568911e-06, + "loss": 0.2149, + "step": 41746 + }, + { + "epoch": 15.42, + "learning_rate": 1.459280699593546e-06, + "loss": 0.258, + "step": 41747 + }, + { + "epoch": 15.42, + "learning_rate": 1.4592437492302008e-06, + "loss": 0.3349, + "step": 41748 + }, + { + "epoch": 15.42, + "learning_rate": 1.4592067988668555e-06, + "loss": 0.2002, + "step": 41749 + }, + { + "epoch": 15.42, + "learning_rate": 1.4591698485035103e-06, + "loss": 0.2521, + "step": 41750 + }, + { + "epoch": 15.42, + "learning_rate": 1.4591328981401652e-06, + "loss": 0.2861, + "step": 41751 + }, + { + "epoch": 15.42, + "learning_rate": 1.4590959477768198e-06, + "loss": 0.2271, + "step": 41752 + }, + { + "epoch": 15.42, + "learning_rate": 1.4590589974134746e-06, + "loss": 0.2475, + "step": 41753 + }, + { + "epoch": 15.42, + "learning_rate": 1.4590220470501295e-06, + "loss": 0.1516, + "step": 41754 + }, + { + "epoch": 15.42, + "learning_rate": 1.4589850966867841e-06, + "loss": 0.1861, + "step": 41755 + }, + { + "epoch": 15.42, + "learning_rate": 1.4589481463234388e-06, + "loss": 0.317, + "step": 41756 + }, + { + "epoch": 15.42, + "learning_rate": 1.4589111959600936e-06, + "loss": 0.1883, + "step": 41757 + }, + { + "epoch": 15.42, + "learning_rate": 1.4588742455967485e-06, + "loss": 0.2196, + "step": 41758 + }, + { + "epoch": 15.42, + "learning_rate": 1.458837295233403e-06, + "loss": 0.3062, + "step": 41759 + }, + { + "epoch": 15.42, + "learning_rate": 1.458800344870058e-06, + "loss": 0.2922, + "step": 41760 + }, + { + "epoch": 15.42, + "learning_rate": 1.4587633945067128e-06, + "loss": 0.1781, + "step": 41761 + }, + { + "epoch": 15.42, + "learning_rate": 1.4587264441433674e-06, + "loss": 0.2958, + "step": 41762 + }, + { + "epoch": 15.42, + "learning_rate": 1.4586894937800223e-06, + "loss": 0.2507, + "step": 41763 + }, + { + "epoch": 15.42, + "learning_rate": 1.4586525434166771e-06, + "loss": 0.1962, + "step": 41764 + }, + { + "epoch": 15.42, + "learning_rate": 1.4586155930533318e-06, + "loss": 0.2726, + "step": 41765 + }, + { + "epoch": 15.42, + "learning_rate": 1.4585786426899864e-06, + "loss": 0.2425, + "step": 41766 + }, + { + "epoch": 15.42, + "learning_rate": 1.4585416923266412e-06, + "loss": 0.2289, + "step": 41767 + }, + { + "epoch": 15.42, + "learning_rate": 1.4585047419632959e-06, + "loss": 0.1794, + "step": 41768 + }, + { + "epoch": 15.42, + "learning_rate": 1.4584677915999507e-06, + "loss": 0.1737, + "step": 41769 + }, + { + "epoch": 15.42, + "learning_rate": 1.4584308412366056e-06, + "loss": 0.2512, + "step": 41770 + }, + { + "epoch": 15.43, + "learning_rate": 1.4583938908732602e-06, + "loss": 0.2386, + "step": 41771 + }, + { + "epoch": 15.43, + "learning_rate": 1.458356940509915e-06, + "loss": 0.1486, + "step": 41772 + }, + { + "epoch": 15.43, + "learning_rate": 1.45831999014657e-06, + "loss": 0.3414, + "step": 41773 + }, + { + "epoch": 15.43, + "learning_rate": 1.4582830397832245e-06, + "loss": 0.2939, + "step": 41774 + }, + { + "epoch": 15.43, + "learning_rate": 1.4582460894198794e-06, + "loss": 0.3318, + "step": 41775 + }, + { + "epoch": 15.43, + "learning_rate": 1.458209139056534e-06, + "loss": 0.3275, + "step": 41776 + }, + { + "epoch": 15.43, + "learning_rate": 1.4581721886931889e-06, + "loss": 0.18, + "step": 41777 + }, + { + "epoch": 15.43, + "learning_rate": 1.4581352383298435e-06, + "loss": 0.3193, + "step": 41778 + }, + { + "epoch": 15.43, + "learning_rate": 1.4580982879664984e-06, + "loss": 0.3103, + "step": 41779 + }, + { + "epoch": 15.43, + "learning_rate": 1.4580613376031532e-06, + "loss": 0.3131, + "step": 41780 + }, + { + "epoch": 15.43, + "learning_rate": 1.4580243872398078e-06, + "loss": 0.2432, + "step": 41781 + }, + { + "epoch": 15.43, + "learning_rate": 1.4579874368764627e-06, + "loss": 0.2882, + "step": 41782 + }, + { + "epoch": 15.43, + "learning_rate": 1.4579504865131175e-06, + "loss": 0.1927, + "step": 41783 + }, + { + "epoch": 15.43, + "learning_rate": 1.4579135361497722e-06, + "loss": 0.2073, + "step": 41784 + }, + { + "epoch": 15.43, + "learning_rate": 1.457876585786427e-06, + "loss": 0.3039, + "step": 41785 + }, + { + "epoch": 15.43, + "learning_rate": 1.4578396354230817e-06, + "loss": 0.2581, + "step": 41786 + }, + { + "epoch": 15.43, + "learning_rate": 1.4578026850597365e-06, + "loss": 0.2131, + "step": 41787 + }, + { + "epoch": 15.43, + "learning_rate": 1.4577657346963911e-06, + "loss": 0.291, + "step": 41788 + }, + { + "epoch": 15.43, + "learning_rate": 1.457728784333046e-06, + "loss": 0.2322, + "step": 41789 + }, + { + "epoch": 15.43, + "learning_rate": 1.4576918339697006e-06, + "loss": 0.251, + "step": 41790 + }, + { + "epoch": 15.43, + "learning_rate": 1.4576548836063555e-06, + "loss": 0.2813, + "step": 41791 + }, + { + "epoch": 15.43, + "learning_rate": 1.4576179332430103e-06, + "loss": 0.2581, + "step": 41792 + }, + { + "epoch": 15.43, + "learning_rate": 1.457580982879665e-06, + "loss": 0.1942, + "step": 41793 + }, + { + "epoch": 15.43, + "learning_rate": 1.4575440325163198e-06, + "loss": 0.2205, + "step": 41794 + }, + { + "epoch": 15.43, + "learning_rate": 1.4575070821529747e-06, + "loss": 0.2507, + "step": 41795 + }, + { + "epoch": 15.43, + "learning_rate": 1.4574701317896293e-06, + "loss": 0.2932, + "step": 41796 + }, + { + "epoch": 15.43, + "learning_rate": 1.457433181426284e-06, + "loss": 0.4158, + "step": 41797 + }, + { + "epoch": 15.44, + "learning_rate": 1.4573962310629388e-06, + "loss": 0.309, + "step": 41798 + }, + { + "epoch": 15.44, + "learning_rate": 1.4573592806995936e-06, + "loss": 0.4371, + "step": 41799 + }, + { + "epoch": 15.44, + "learning_rate": 1.4573223303362483e-06, + "loss": 0.323, + "step": 41800 + }, + { + "epoch": 15.44, + "learning_rate": 1.4572853799729031e-06, + "loss": 0.2293, + "step": 41801 + }, + { + "epoch": 15.44, + "learning_rate": 1.457248429609558e-06, + "loss": 0.2297, + "step": 41802 + }, + { + "epoch": 15.44, + "learning_rate": 1.4572114792462126e-06, + "loss": 0.1554, + "step": 41803 + }, + { + "epoch": 15.44, + "learning_rate": 1.4571745288828674e-06, + "loss": 0.316, + "step": 41804 + }, + { + "epoch": 15.44, + "learning_rate": 1.457137578519522e-06, + "loss": 0.2571, + "step": 41805 + }, + { + "epoch": 15.44, + "learning_rate": 1.457100628156177e-06, + "loss": 0.238, + "step": 41806 + }, + { + "epoch": 15.44, + "learning_rate": 1.4570636777928316e-06, + "loss": 0.4048, + "step": 41807 + }, + { + "epoch": 15.44, + "learning_rate": 1.4570267274294864e-06, + "loss": 0.2264, + "step": 41808 + }, + { + "epoch": 15.44, + "learning_rate": 1.4569897770661413e-06, + "loss": 0.1449, + "step": 41809 + }, + { + "epoch": 15.44, + "learning_rate": 1.456952826702796e-06, + "loss": 0.18, + "step": 41810 + }, + { + "epoch": 15.44, + "learning_rate": 1.4569158763394508e-06, + "loss": 0.3203, + "step": 41811 + }, + { + "epoch": 15.44, + "learning_rate": 1.4568789259761056e-06, + "loss": 0.183, + "step": 41812 + }, + { + "epoch": 15.44, + "learning_rate": 1.4568419756127602e-06, + "loss": 0.4408, + "step": 41813 + }, + { + "epoch": 15.44, + "learning_rate": 1.456805025249415e-06, + "loss": 0.2166, + "step": 41814 + }, + { + "epoch": 15.44, + "learning_rate": 1.4567680748860697e-06, + "loss": 0.2776, + "step": 41815 + }, + { + "epoch": 15.44, + "learning_rate": 1.4567311245227244e-06, + "loss": 0.1717, + "step": 41816 + }, + { + "epoch": 15.44, + "learning_rate": 1.4566941741593792e-06, + "loss": 0.2646, + "step": 41817 + }, + { + "epoch": 15.44, + "learning_rate": 1.456657223796034e-06, + "loss": 0.2353, + "step": 41818 + }, + { + "epoch": 15.44, + "learning_rate": 1.4566202734326887e-06, + "loss": 0.2489, + "step": 41819 + }, + { + "epoch": 15.44, + "learning_rate": 1.4565833230693435e-06, + "loss": 0.2421, + "step": 41820 + }, + { + "epoch": 15.44, + "learning_rate": 1.4565463727059984e-06, + "loss": 0.2759, + "step": 41821 + }, + { + "epoch": 15.44, + "learning_rate": 1.456509422342653e-06, + "loss": 0.2317, + "step": 41822 + }, + { + "epoch": 15.44, + "learning_rate": 1.4564724719793079e-06, + "loss": 0.3396, + "step": 41823 + }, + { + "epoch": 15.44, + "learning_rate": 1.4564355216159627e-06, + "loss": 0.1964, + "step": 41824 + }, + { + "epoch": 15.44, + "learning_rate": 1.4563985712526174e-06, + "loss": 0.2416, + "step": 41825 + }, + { + "epoch": 15.45, + "learning_rate": 1.456361620889272e-06, + "loss": 0.1997, + "step": 41826 + }, + { + "epoch": 15.45, + "learning_rate": 1.4563246705259268e-06, + "loss": 0.481, + "step": 41827 + }, + { + "epoch": 15.45, + "learning_rate": 1.4562877201625817e-06, + "loss": 0.2113, + "step": 41828 + }, + { + "epoch": 15.45, + "learning_rate": 1.4562507697992363e-06, + "loss": 0.2647, + "step": 41829 + }, + { + "epoch": 15.45, + "learning_rate": 1.4562138194358912e-06, + "loss": 0.2274, + "step": 41830 + }, + { + "epoch": 15.45, + "learning_rate": 1.456176869072546e-06, + "loss": 0.2717, + "step": 41831 + }, + { + "epoch": 15.45, + "learning_rate": 1.4561399187092007e-06, + "loss": 0.3388, + "step": 41832 + }, + { + "epoch": 15.45, + "learning_rate": 1.4561029683458555e-06, + "loss": 0.2079, + "step": 41833 + }, + { + "epoch": 15.45, + "learning_rate": 1.4560660179825104e-06, + "loss": 0.2066, + "step": 41834 + }, + { + "epoch": 15.45, + "learning_rate": 1.4560290676191648e-06, + "loss": 0.288, + "step": 41835 + }, + { + "epoch": 15.45, + "learning_rate": 1.4559921172558196e-06, + "loss": 0.2007, + "step": 41836 + }, + { + "epoch": 15.45, + "learning_rate": 1.4559551668924745e-06, + "loss": 0.3494, + "step": 41837 + }, + { + "epoch": 15.45, + "learning_rate": 1.4559182165291291e-06, + "loss": 0.4526, + "step": 41838 + }, + { + "epoch": 15.45, + "learning_rate": 1.455881266165784e-06, + "loss": 0.329, + "step": 41839 + }, + { + "epoch": 15.45, + "learning_rate": 1.4558443158024388e-06, + "loss": 0.2403, + "step": 41840 + }, + { + "epoch": 15.45, + "learning_rate": 1.4558073654390934e-06, + "loss": 0.2654, + "step": 41841 + }, + { + "epoch": 15.45, + "learning_rate": 1.4557704150757483e-06, + "loss": 0.2457, + "step": 41842 + }, + { + "epoch": 15.45, + "learning_rate": 1.4557334647124031e-06, + "loss": 0.2588, + "step": 41843 + }, + { + "epoch": 15.45, + "learning_rate": 1.4556965143490578e-06, + "loss": 0.2999, + "step": 41844 + }, + { + "epoch": 15.45, + "learning_rate": 1.4556595639857124e-06, + "loss": 0.2784, + "step": 41845 + }, + { + "epoch": 15.45, + "learning_rate": 1.4556226136223673e-06, + "loss": 0.3299, + "step": 41846 + }, + { + "epoch": 15.45, + "learning_rate": 1.4555856632590221e-06, + "loss": 0.197, + "step": 41847 + }, + { + "epoch": 15.45, + "learning_rate": 1.4555487128956767e-06, + "loss": 0.2237, + "step": 41848 + }, + { + "epoch": 15.45, + "learning_rate": 1.4555117625323316e-06, + "loss": 0.2844, + "step": 41849 + }, + { + "epoch": 15.45, + "learning_rate": 1.4554748121689864e-06, + "loss": 0.2401, + "step": 41850 + }, + { + "epoch": 15.45, + "learning_rate": 1.455437861805641e-06, + "loss": 0.1739, + "step": 41851 + }, + { + "epoch": 15.45, + "learning_rate": 1.455400911442296e-06, + "loss": 0.3291, + "step": 41852 + }, + { + "epoch": 15.46, + "learning_rate": 1.4553639610789508e-06, + "loss": 0.2312, + "step": 41853 + }, + { + "epoch": 15.46, + "learning_rate": 1.4553270107156054e-06, + "loss": 0.2226, + "step": 41854 + }, + { + "epoch": 15.46, + "learning_rate": 1.45529006035226e-06, + "loss": 0.2081, + "step": 41855 + }, + { + "epoch": 15.46, + "learning_rate": 1.455253109988915e-06, + "loss": 0.2629, + "step": 41856 + }, + { + "epoch": 15.46, + "learning_rate": 1.4552161596255697e-06, + "loss": 0.2685, + "step": 41857 + }, + { + "epoch": 15.46, + "learning_rate": 1.4551792092622244e-06, + "loss": 0.2331, + "step": 41858 + }, + { + "epoch": 15.46, + "learning_rate": 1.4551422588988792e-06, + "loss": 0.2219, + "step": 41859 + }, + { + "epoch": 15.46, + "learning_rate": 1.455105308535534e-06, + "loss": 0.208, + "step": 41860 + }, + { + "epoch": 15.46, + "learning_rate": 1.4550683581721887e-06, + "loss": 0.2753, + "step": 41861 + }, + { + "epoch": 15.46, + "learning_rate": 1.4550314078088436e-06, + "loss": 0.3276, + "step": 41862 + }, + { + "epoch": 15.46, + "learning_rate": 1.4549944574454984e-06, + "loss": 0.2403, + "step": 41863 + }, + { + "epoch": 15.46, + "learning_rate": 1.454957507082153e-06, + "loss": 0.3022, + "step": 41864 + }, + { + "epoch": 15.46, + "learning_rate": 1.4549205567188077e-06, + "loss": 0.2826, + "step": 41865 + }, + { + "epoch": 15.46, + "learning_rate": 1.4548836063554625e-06, + "loss": 0.2385, + "step": 41866 + }, + { + "epoch": 15.46, + "learning_rate": 1.4548466559921172e-06, + "loss": 0.2568, + "step": 41867 + }, + { + "epoch": 15.46, + "learning_rate": 1.454809705628772e-06, + "loss": 0.2242, + "step": 41868 + }, + { + "epoch": 15.46, + "learning_rate": 1.4547727552654269e-06, + "loss": 0.2395, + "step": 41869 + }, + { + "epoch": 15.46, + "learning_rate": 1.4547358049020815e-06, + "loss": 0.2094, + "step": 41870 + }, + { + "epoch": 15.46, + "learning_rate": 1.4546988545387363e-06, + "loss": 0.2368, + "step": 41871 + }, + { + "epoch": 15.46, + "learning_rate": 1.4546619041753912e-06, + "loss": 0.3721, + "step": 41872 + }, + { + "epoch": 15.46, + "learning_rate": 1.4546249538120458e-06, + "loss": 0.4499, + "step": 41873 + }, + { + "epoch": 15.46, + "learning_rate": 1.4545880034487007e-06, + "loss": 0.2941, + "step": 41874 + }, + { + "epoch": 15.46, + "learning_rate": 1.4545510530853553e-06, + "loss": 0.2069, + "step": 41875 + }, + { + "epoch": 15.46, + "learning_rate": 1.4545141027220102e-06, + "loss": 0.1722, + "step": 41876 + }, + { + "epoch": 15.46, + "learning_rate": 1.4544771523586648e-06, + "loss": 0.3293, + "step": 41877 + }, + { + "epoch": 15.46, + "learning_rate": 1.4544402019953197e-06, + "loss": 0.1719, + "step": 41878 + }, + { + "epoch": 15.46, + "learning_rate": 1.4544032516319745e-06, + "loss": 0.3049, + "step": 41879 + }, + { + "epoch": 15.47, + "learning_rate": 1.4543663012686291e-06, + "loss": 0.2725, + "step": 41880 + }, + { + "epoch": 15.47, + "learning_rate": 1.454329350905284e-06, + "loss": 0.2084, + "step": 41881 + }, + { + "epoch": 15.47, + "learning_rate": 1.4542924005419388e-06, + "loss": 0.1575, + "step": 41882 + }, + { + "epoch": 15.47, + "learning_rate": 1.4542554501785935e-06, + "loss": 0.2288, + "step": 41883 + }, + { + "epoch": 15.47, + "learning_rate": 1.4542184998152483e-06, + "loss": 0.2556, + "step": 41884 + }, + { + "epoch": 15.47, + "learning_rate": 1.454181549451903e-06, + "loss": 0.2931, + "step": 41885 + }, + { + "epoch": 15.47, + "learning_rate": 1.4541445990885576e-06, + "loss": 0.1456, + "step": 41886 + }, + { + "epoch": 15.47, + "learning_rate": 1.4541076487252124e-06, + "loss": 0.157, + "step": 41887 + }, + { + "epoch": 15.47, + "learning_rate": 1.4540706983618673e-06, + "loss": 0.5353, + "step": 41888 + }, + { + "epoch": 15.47, + "learning_rate": 1.454033747998522e-06, + "loss": 0.2007, + "step": 41889 + }, + { + "epoch": 15.47, + "learning_rate": 1.4539967976351768e-06, + "loss": 0.3332, + "step": 41890 + }, + { + "epoch": 15.47, + "learning_rate": 1.4539598472718316e-06, + "loss": 0.2081, + "step": 41891 + }, + { + "epoch": 15.47, + "learning_rate": 1.4539228969084863e-06, + "loss": 0.2778, + "step": 41892 + }, + { + "epoch": 15.47, + "learning_rate": 1.453885946545141e-06, + "loss": 0.2914, + "step": 41893 + }, + { + "epoch": 15.47, + "learning_rate": 1.453848996181796e-06, + "loss": 0.2428, + "step": 41894 + }, + { + "epoch": 15.47, + "learning_rate": 1.4538120458184506e-06, + "loss": 0.2178, + "step": 41895 + }, + { + "epoch": 15.47, + "learning_rate": 1.4537750954551052e-06, + "loss": 0.2559, + "step": 41896 + }, + { + "epoch": 15.47, + "learning_rate": 1.45373814509176e-06, + "loss": 0.2133, + "step": 41897 + }, + { + "epoch": 15.47, + "learning_rate": 1.453701194728415e-06, + "loss": 0.1629, + "step": 41898 + }, + { + "epoch": 15.47, + "learning_rate": 1.4536642443650696e-06, + "loss": 0.2883, + "step": 41899 + }, + { + "epoch": 15.47, + "learning_rate": 1.4536272940017244e-06, + "loss": 0.2514, + "step": 41900 + }, + { + "epoch": 15.47, + "learning_rate": 1.4535903436383793e-06, + "loss": 0.2043, + "step": 41901 + }, + { + "epoch": 15.47, + "learning_rate": 1.4535533932750339e-06, + "loss": 0.3962, + "step": 41902 + }, + { + "epoch": 15.47, + "learning_rate": 1.4535164429116887e-06, + "loss": 0.2859, + "step": 41903 + }, + { + "epoch": 15.47, + "learning_rate": 1.4534794925483436e-06, + "loss": 0.2719, + "step": 41904 + }, + { + "epoch": 15.47, + "learning_rate": 1.453442542184998e-06, + "loss": 0.2102, + "step": 41905 + }, + { + "epoch": 15.47, + "learning_rate": 1.4534055918216529e-06, + "loss": 0.2374, + "step": 41906 + }, + { + "epoch": 15.48, + "learning_rate": 1.4533686414583077e-06, + "loss": 0.2897, + "step": 41907 + }, + { + "epoch": 15.48, + "learning_rate": 1.4533316910949623e-06, + "loss": 0.349, + "step": 41908 + }, + { + "epoch": 15.48, + "learning_rate": 1.4532947407316172e-06, + "loss": 0.237, + "step": 41909 + }, + { + "epoch": 15.48, + "learning_rate": 1.453257790368272e-06, + "loss": 0.1434, + "step": 41910 + }, + { + "epoch": 15.48, + "learning_rate": 1.4532208400049267e-06, + "loss": 0.2574, + "step": 41911 + }, + { + "epoch": 15.48, + "learning_rate": 1.4531838896415815e-06, + "loss": 0.1842, + "step": 41912 + }, + { + "epoch": 15.48, + "learning_rate": 1.4531469392782364e-06, + "loss": 0.2515, + "step": 41913 + }, + { + "epoch": 15.48, + "learning_rate": 1.453109988914891e-06, + "loss": 0.2931, + "step": 41914 + }, + { + "epoch": 15.48, + "learning_rate": 1.4530730385515456e-06, + "loss": 0.1237, + "step": 41915 + }, + { + "epoch": 15.48, + "learning_rate": 1.4530360881882005e-06, + "loss": 0.3179, + "step": 41916 + }, + { + "epoch": 15.48, + "learning_rate": 1.4529991378248553e-06, + "loss": 0.1961, + "step": 41917 + }, + { + "epoch": 15.48, + "learning_rate": 1.45296218746151e-06, + "loss": 0.2241, + "step": 41918 + }, + { + "epoch": 15.48, + "learning_rate": 1.4529252370981648e-06, + "loss": 0.1539, + "step": 41919 + }, + { + "epoch": 15.48, + "learning_rate": 1.4528882867348197e-06, + "loss": 0.2815, + "step": 41920 + }, + { + "epoch": 15.48, + "learning_rate": 1.4528513363714743e-06, + "loss": 0.257, + "step": 41921 + }, + { + "epoch": 15.48, + "learning_rate": 1.4528143860081292e-06, + "loss": 0.2354, + "step": 41922 + }, + { + "epoch": 15.48, + "learning_rate": 1.452777435644784e-06, + "loss": 0.1935, + "step": 41923 + }, + { + "epoch": 15.48, + "learning_rate": 1.4527404852814386e-06, + "loss": 0.2707, + "step": 41924 + }, + { + "epoch": 15.48, + "learning_rate": 1.4527035349180933e-06, + "loss": 0.241, + "step": 41925 + }, + { + "epoch": 15.48, + "learning_rate": 1.4526665845547481e-06, + "loss": 0.2536, + "step": 41926 + }, + { + "epoch": 15.48, + "learning_rate": 1.452629634191403e-06, + "loss": 0.2521, + "step": 41927 + }, + { + "epoch": 15.48, + "learning_rate": 1.4525926838280576e-06, + "loss": 0.1899, + "step": 41928 + }, + { + "epoch": 15.48, + "learning_rate": 1.4525557334647125e-06, + "loss": 0.2661, + "step": 41929 + }, + { + "epoch": 15.48, + "learning_rate": 1.4525187831013673e-06, + "loss": 0.2019, + "step": 41930 + }, + { + "epoch": 15.48, + "learning_rate": 1.452481832738022e-06, + "loss": 0.3598, + "step": 41931 + }, + { + "epoch": 15.48, + "learning_rate": 1.4524448823746768e-06, + "loss": 0.2418, + "step": 41932 + }, + { + "epoch": 15.48, + "learning_rate": 1.4524079320113316e-06, + "loss": 0.2302, + "step": 41933 + }, + { + "epoch": 15.49, + "learning_rate": 1.4523709816479863e-06, + "loss": 0.1997, + "step": 41934 + }, + { + "epoch": 15.49, + "learning_rate": 1.452334031284641e-06, + "loss": 0.2913, + "step": 41935 + }, + { + "epoch": 15.49, + "learning_rate": 1.4522970809212958e-06, + "loss": 0.3308, + "step": 41936 + }, + { + "epoch": 15.49, + "learning_rate": 1.4522601305579504e-06, + "loss": 0.2241, + "step": 41937 + }, + { + "epoch": 15.49, + "learning_rate": 1.4522231801946052e-06, + "loss": 0.2856, + "step": 41938 + }, + { + "epoch": 15.49, + "learning_rate": 1.45218622983126e-06, + "loss": 0.3828, + "step": 41939 + }, + { + "epoch": 15.49, + "learning_rate": 1.4521492794679147e-06, + "loss": 0.1633, + "step": 41940 + }, + { + "epoch": 15.49, + "learning_rate": 1.4521123291045696e-06, + "loss": 0.355, + "step": 41941 + }, + { + "epoch": 15.49, + "learning_rate": 1.4520753787412244e-06, + "loss": 0.1563, + "step": 41942 + }, + { + "epoch": 15.49, + "learning_rate": 1.452038428377879e-06, + "loss": 0.3048, + "step": 41943 + }, + { + "epoch": 15.49, + "learning_rate": 1.452001478014534e-06, + "loss": 0.2136, + "step": 41944 + }, + { + "epoch": 15.49, + "learning_rate": 1.4519645276511886e-06, + "loss": 0.3232, + "step": 41945 + }, + { + "epoch": 15.49, + "learning_rate": 1.4519275772878434e-06, + "loss": 0.2322, + "step": 41946 + }, + { + "epoch": 15.49, + "learning_rate": 1.451890626924498e-06, + "loss": 0.3799, + "step": 41947 + }, + { + "epoch": 15.49, + "learning_rate": 1.4518536765611529e-06, + "loss": 0.2169, + "step": 41948 + }, + { + "epoch": 15.49, + "learning_rate": 1.4518167261978077e-06, + "loss": 0.2143, + "step": 41949 + }, + { + "epoch": 15.49, + "learning_rate": 1.4517797758344624e-06, + "loss": 0.2532, + "step": 41950 + }, + { + "epoch": 15.49, + "learning_rate": 1.4517428254711172e-06, + "loss": 0.2812, + "step": 41951 + }, + { + "epoch": 15.49, + "learning_rate": 1.451705875107772e-06, + "loss": 0.2498, + "step": 41952 + }, + { + "epoch": 15.49, + "learning_rate": 1.4516689247444267e-06, + "loss": 0.2047, + "step": 41953 + }, + { + "epoch": 15.49, + "learning_rate": 1.4516319743810815e-06, + "loss": 0.2124, + "step": 41954 + }, + { + "epoch": 15.49, + "learning_rate": 1.4515950240177362e-06, + "loss": 0.2339, + "step": 41955 + }, + { + "epoch": 15.49, + "learning_rate": 1.4515580736543908e-06, + "loss": 0.201, + "step": 41956 + }, + { + "epoch": 15.49, + "learning_rate": 1.4515211232910457e-06, + "loss": 0.2034, + "step": 41957 + }, + { + "epoch": 15.49, + "learning_rate": 1.4514841729277005e-06, + "loss": 0.2668, + "step": 41958 + }, + { + "epoch": 15.49, + "learning_rate": 1.4514472225643552e-06, + "loss": 0.2637, + "step": 41959 + }, + { + "epoch": 15.49, + "learning_rate": 1.45141027220101e-06, + "loss": 0.294, + "step": 41960 + }, + { + "epoch": 15.5, + "learning_rate": 1.4513733218376649e-06, + "loss": 0.226, + "step": 41961 + }, + { + "epoch": 15.5, + "learning_rate": 1.4513363714743195e-06, + "loss": 0.2002, + "step": 41962 + }, + { + "epoch": 15.5, + "learning_rate": 1.4512994211109743e-06, + "loss": 0.2506, + "step": 41963 + }, + { + "epoch": 15.5, + "learning_rate": 1.4512624707476292e-06, + "loss": 0.2724, + "step": 41964 + }, + { + "epoch": 15.5, + "learning_rate": 1.4512255203842838e-06, + "loss": 0.286, + "step": 41965 + }, + { + "epoch": 15.5, + "learning_rate": 1.4511885700209385e-06, + "loss": 0.1626, + "step": 41966 + }, + { + "epoch": 15.5, + "learning_rate": 1.4511516196575933e-06, + "loss": 0.3025, + "step": 41967 + }, + { + "epoch": 15.5, + "learning_rate": 1.4511146692942482e-06, + "loss": 0.2604, + "step": 41968 + }, + { + "epoch": 15.5, + "learning_rate": 1.4510777189309028e-06, + "loss": 0.2164, + "step": 41969 + }, + { + "epoch": 15.5, + "learning_rate": 1.4510407685675576e-06, + "loss": 0.1998, + "step": 41970 + }, + { + "epoch": 15.5, + "learning_rate": 1.4510038182042125e-06, + "loss": 0.1201, + "step": 41971 + }, + { + "epoch": 15.5, + "learning_rate": 1.4509668678408671e-06, + "loss": 0.215, + "step": 41972 + }, + { + "epoch": 15.5, + "learning_rate": 1.450929917477522e-06, + "loss": 0.3029, + "step": 41973 + }, + { + "epoch": 15.5, + "learning_rate": 1.4508929671141768e-06, + "loss": 0.2641, + "step": 41974 + }, + { + "epoch": 15.5, + "learning_rate": 1.4508560167508312e-06, + "loss": 0.2201, + "step": 41975 + }, + { + "epoch": 15.5, + "learning_rate": 1.450819066387486e-06, + "loss": 0.2691, + "step": 41976 + }, + { + "epoch": 15.5, + "learning_rate": 1.450782116024141e-06, + "loss": 0.2181, + "step": 41977 + }, + { + "epoch": 15.5, + "learning_rate": 1.4507451656607956e-06, + "loss": 0.3011, + "step": 41978 + }, + { + "epoch": 15.5, + "learning_rate": 1.4507082152974504e-06, + "loss": 0.192, + "step": 41979 + }, + { + "epoch": 15.5, + "learning_rate": 1.4506712649341053e-06, + "loss": 0.2074, + "step": 41980 + }, + { + "epoch": 15.5, + "learning_rate": 1.45063431457076e-06, + "loss": 0.2092, + "step": 41981 + }, + { + "epoch": 15.5, + "learning_rate": 1.4505973642074148e-06, + "loss": 0.3674, + "step": 41982 + }, + { + "epoch": 15.5, + "learning_rate": 1.4505604138440696e-06, + "loss": 0.2997, + "step": 41983 + }, + { + "epoch": 15.5, + "learning_rate": 1.4505234634807242e-06, + "loss": 0.2016, + "step": 41984 + }, + { + "epoch": 15.5, + "learning_rate": 1.4504865131173789e-06, + "loss": 0.2918, + "step": 41985 + }, + { + "epoch": 15.5, + "learning_rate": 1.4504495627540337e-06, + "loss": 0.2332, + "step": 41986 + }, + { + "epoch": 15.5, + "learning_rate": 1.4504126123906886e-06, + "loss": 0.1191, + "step": 41987 + }, + { + "epoch": 15.51, + "learning_rate": 1.4503756620273432e-06, + "loss": 0.2481, + "step": 41988 + }, + { + "epoch": 15.51, + "learning_rate": 1.450338711663998e-06, + "loss": 0.2256, + "step": 41989 + }, + { + "epoch": 15.51, + "learning_rate": 1.450301761300653e-06, + "loss": 0.2008, + "step": 41990 + }, + { + "epoch": 15.51, + "learning_rate": 1.4502648109373075e-06, + "loss": 0.2597, + "step": 41991 + }, + { + "epoch": 15.51, + "learning_rate": 1.4502278605739624e-06, + "loss": 0.2006, + "step": 41992 + }, + { + "epoch": 15.51, + "learning_rate": 1.4501909102106172e-06, + "loss": 0.1826, + "step": 41993 + }, + { + "epoch": 15.51, + "learning_rate": 1.4501539598472719e-06, + "loss": 0.2273, + "step": 41994 + }, + { + "epoch": 15.51, + "learning_rate": 1.4501170094839265e-06, + "loss": 0.3021, + "step": 41995 + }, + { + "epoch": 15.51, + "learning_rate": 1.4500800591205814e-06, + "loss": 0.2053, + "step": 41996 + }, + { + "epoch": 15.51, + "learning_rate": 1.4500431087572362e-06, + "loss": 0.2582, + "step": 41997 + }, + { + "epoch": 15.51, + "learning_rate": 1.4500061583938908e-06, + "loss": 0.3242, + "step": 41998 + }, + { + "epoch": 15.51, + "learning_rate": 1.4499692080305457e-06, + "loss": 0.2444, + "step": 41999 + }, + { + "epoch": 15.51, + "learning_rate": 1.4499322576672005e-06, + "loss": 0.3517, + "step": 42000 + }, + { + "epoch": 15.51, + "learning_rate": 1.4498953073038552e-06, + "loss": 0.2911, + "step": 42001 + }, + { + "epoch": 15.51, + "learning_rate": 1.44985835694051e-06, + "loss": 0.2917, + "step": 42002 + }, + { + "epoch": 15.51, + "learning_rate": 1.4498214065771649e-06, + "loss": 0.21, + "step": 42003 + }, + { + "epoch": 15.51, + "learning_rate": 1.4497844562138195e-06, + "loss": 0.1648, + "step": 42004 + }, + { + "epoch": 15.51, + "learning_rate": 1.4497475058504741e-06, + "loss": 0.1847, + "step": 42005 + }, + { + "epoch": 15.51, + "learning_rate": 1.449710555487129e-06, + "loss": 0.2647, + "step": 42006 + }, + { + "epoch": 15.51, + "learning_rate": 1.4496736051237836e-06, + "loss": 0.2812, + "step": 42007 + }, + { + "epoch": 15.51, + "learning_rate": 1.4496366547604385e-06, + "loss": 0.5371, + "step": 42008 + }, + { + "epoch": 15.51, + "learning_rate": 1.4495997043970933e-06, + "loss": 0.2365, + "step": 42009 + }, + { + "epoch": 15.51, + "learning_rate": 1.449562754033748e-06, + "loss": 0.4158, + "step": 42010 + }, + { + "epoch": 15.51, + "learning_rate": 1.4495258036704028e-06, + "loss": 0.2084, + "step": 42011 + }, + { + "epoch": 15.51, + "learning_rate": 1.4494888533070577e-06, + "loss": 0.2678, + "step": 42012 + }, + { + "epoch": 15.51, + "learning_rate": 1.4494519029437123e-06, + "loss": 0.1906, + "step": 42013 + }, + { + "epoch": 15.51, + "learning_rate": 1.4494149525803671e-06, + "loss": 0.2188, + "step": 42014 + }, + { + "epoch": 15.52, + "learning_rate": 1.4493780022170218e-06, + "loss": 0.2187, + "step": 42015 + }, + { + "epoch": 15.52, + "learning_rate": 1.4493410518536766e-06, + "loss": 0.202, + "step": 42016 + }, + { + "epoch": 15.52, + "learning_rate": 1.4493041014903313e-06, + "loss": 0.2474, + "step": 42017 + }, + { + "epoch": 15.52, + "learning_rate": 1.4492671511269861e-06, + "loss": 0.24, + "step": 42018 + }, + { + "epoch": 15.52, + "learning_rate": 1.449230200763641e-06, + "loss": 0.231, + "step": 42019 + }, + { + "epoch": 15.52, + "learning_rate": 1.4491932504002956e-06, + "loss": 0.312, + "step": 42020 + }, + { + "epoch": 15.52, + "learning_rate": 1.4491563000369504e-06, + "loss": 0.1597, + "step": 42021 + }, + { + "epoch": 15.52, + "learning_rate": 1.4491193496736053e-06, + "loss": 0.2459, + "step": 42022 + }, + { + "epoch": 15.52, + "learning_rate": 1.44908239931026e-06, + "loss": 0.3029, + "step": 42023 + }, + { + "epoch": 15.52, + "learning_rate": 1.4490454489469148e-06, + "loss": 0.2573, + "step": 42024 + }, + { + "epoch": 15.52, + "learning_rate": 1.4490084985835694e-06, + "loss": 0.2523, + "step": 42025 + }, + { + "epoch": 15.52, + "learning_rate": 1.448971548220224e-06, + "loss": 0.1412, + "step": 42026 + }, + { + "epoch": 15.52, + "learning_rate": 1.448934597856879e-06, + "loss": 0.2632, + "step": 42027 + }, + { + "epoch": 15.52, + "learning_rate": 1.4488976474935338e-06, + "loss": 0.2183, + "step": 42028 + }, + { + "epoch": 15.52, + "learning_rate": 1.4488606971301884e-06, + "loss": 0.1797, + "step": 42029 + }, + { + "epoch": 15.52, + "learning_rate": 1.4488237467668432e-06, + "loss": 0.2311, + "step": 42030 + }, + { + "epoch": 15.52, + "learning_rate": 1.448786796403498e-06, + "loss": 0.302, + "step": 42031 + }, + { + "epoch": 15.52, + "learning_rate": 1.4487498460401527e-06, + "loss": 0.1419, + "step": 42032 + }, + { + "epoch": 15.52, + "learning_rate": 1.4487128956768076e-06, + "loss": 0.2321, + "step": 42033 + }, + { + "epoch": 15.52, + "learning_rate": 1.4486759453134624e-06, + "loss": 0.2123, + "step": 42034 + }, + { + "epoch": 15.52, + "learning_rate": 1.448638994950117e-06, + "loss": 0.2987, + "step": 42035 + }, + { + "epoch": 15.52, + "learning_rate": 1.4486020445867717e-06, + "loss": 0.2877, + "step": 42036 + }, + { + "epoch": 15.52, + "learning_rate": 1.4485650942234265e-06, + "loss": 0.3734, + "step": 42037 + }, + { + "epoch": 15.52, + "learning_rate": 1.4485281438600814e-06, + "loss": 0.3445, + "step": 42038 + }, + { + "epoch": 15.52, + "learning_rate": 1.448491193496736e-06, + "loss": 0.3769, + "step": 42039 + }, + { + "epoch": 15.52, + "learning_rate": 1.4484542431333909e-06, + "loss": 0.3541, + "step": 42040 + }, + { + "epoch": 15.52, + "learning_rate": 1.4484172927700457e-06, + "loss": 0.2588, + "step": 42041 + }, + { + "epoch": 15.53, + "learning_rate": 1.4483803424067004e-06, + "loss": 0.2173, + "step": 42042 + }, + { + "epoch": 15.53, + "learning_rate": 1.4483433920433552e-06, + "loss": 0.2571, + "step": 42043 + }, + { + "epoch": 15.53, + "learning_rate": 1.4483064416800098e-06, + "loss": 0.2233, + "step": 42044 + }, + { + "epoch": 15.53, + "learning_rate": 1.4482694913166647e-06, + "loss": 0.4653, + "step": 42045 + }, + { + "epoch": 15.53, + "learning_rate": 1.4482325409533193e-06, + "loss": 0.2577, + "step": 42046 + }, + { + "epoch": 15.53, + "learning_rate": 1.4481955905899742e-06, + "loss": 0.2215, + "step": 42047 + }, + { + "epoch": 15.53, + "learning_rate": 1.448158640226629e-06, + "loss": 0.2477, + "step": 42048 + }, + { + "epoch": 15.53, + "learning_rate": 1.4481216898632837e-06, + "loss": 0.2345, + "step": 42049 + }, + { + "epoch": 15.53, + "learning_rate": 1.4480847394999385e-06, + "loss": 0.1751, + "step": 42050 + }, + { + "epoch": 15.53, + "learning_rate": 1.4480477891365934e-06, + "loss": 0.3099, + "step": 42051 + }, + { + "epoch": 15.53, + "learning_rate": 1.448010838773248e-06, + "loss": 0.3616, + "step": 42052 + }, + { + "epoch": 15.53, + "learning_rate": 1.4479738884099028e-06, + "loss": 0.096, + "step": 42053 + }, + { + "epoch": 15.53, + "learning_rate": 1.4479369380465575e-06, + "loss": 0.2308, + "step": 42054 + }, + { + "epoch": 15.53, + "learning_rate": 1.4478999876832121e-06, + "loss": 0.3951, + "step": 42055 + }, + { + "epoch": 15.53, + "learning_rate": 1.447863037319867e-06, + "loss": 0.2749, + "step": 42056 + }, + { + "epoch": 15.53, + "learning_rate": 1.4478260869565218e-06, + "loss": 0.135, + "step": 42057 + }, + { + "epoch": 15.53, + "learning_rate": 1.4477891365931764e-06, + "loss": 0.3113, + "step": 42058 + }, + { + "epoch": 15.53, + "learning_rate": 1.4477521862298313e-06, + "loss": 0.2223, + "step": 42059 + }, + { + "epoch": 15.53, + "learning_rate": 1.4477152358664861e-06, + "loss": 0.2205, + "step": 42060 + }, + { + "epoch": 15.53, + "learning_rate": 1.4476782855031408e-06, + "loss": 0.2296, + "step": 42061 + }, + { + "epoch": 15.53, + "learning_rate": 1.4476413351397956e-06, + "loss": 0.2479, + "step": 42062 + }, + { + "epoch": 15.53, + "learning_rate": 1.4476043847764505e-06, + "loss": 0.2669, + "step": 42063 + }, + { + "epoch": 15.53, + "learning_rate": 1.4475674344131051e-06, + "loss": 0.236, + "step": 42064 + }, + { + "epoch": 15.53, + "learning_rate": 1.4475304840497597e-06, + "loss": 0.2115, + "step": 42065 + }, + { + "epoch": 15.53, + "learning_rate": 1.4474935336864146e-06, + "loss": 0.3614, + "step": 42066 + }, + { + "epoch": 15.53, + "learning_rate": 1.4474565833230694e-06, + "loss": 0.2439, + "step": 42067 + }, + { + "epoch": 15.53, + "learning_rate": 1.447419632959724e-06, + "loss": 0.2327, + "step": 42068 + }, + { + "epoch": 15.54, + "learning_rate": 1.447382682596379e-06, + "loss": 0.2389, + "step": 42069 + }, + { + "epoch": 15.54, + "learning_rate": 1.4473457322330338e-06, + "loss": 0.2552, + "step": 42070 + }, + { + "epoch": 15.54, + "learning_rate": 1.4473087818696884e-06, + "loss": 0.2779, + "step": 42071 + }, + { + "epoch": 15.54, + "learning_rate": 1.4472718315063433e-06, + "loss": 0.2693, + "step": 42072 + }, + { + "epoch": 15.54, + "learning_rate": 1.4472348811429981e-06, + "loss": 0.2143, + "step": 42073 + }, + { + "epoch": 15.54, + "learning_rate": 1.4471979307796525e-06, + "loss": 0.2655, + "step": 42074 + }, + { + "epoch": 15.54, + "learning_rate": 1.4471609804163074e-06, + "loss": 0.2555, + "step": 42075 + }, + { + "epoch": 15.54, + "learning_rate": 1.4471240300529622e-06, + "loss": 0.3167, + "step": 42076 + }, + { + "epoch": 15.54, + "learning_rate": 1.4470870796896169e-06, + "loss": 0.2048, + "step": 42077 + }, + { + "epoch": 15.54, + "learning_rate": 1.4470501293262717e-06, + "loss": 0.3247, + "step": 42078 + }, + { + "epoch": 15.54, + "learning_rate": 1.4470131789629266e-06, + "loss": 0.231, + "step": 42079 + }, + { + "epoch": 15.54, + "learning_rate": 1.4469762285995812e-06, + "loss": 0.1105, + "step": 42080 + }, + { + "epoch": 15.54, + "learning_rate": 1.446939278236236e-06, + "loss": 0.3388, + "step": 42081 + }, + { + "epoch": 15.54, + "learning_rate": 1.446902327872891e-06, + "loss": 0.3535, + "step": 42082 + }, + { + "epoch": 15.54, + "learning_rate": 1.4468653775095455e-06, + "loss": 0.3189, + "step": 42083 + }, + { + "epoch": 15.54, + "learning_rate": 1.4468284271462002e-06, + "loss": 0.2013, + "step": 42084 + }, + { + "epoch": 15.54, + "learning_rate": 1.446791476782855e-06, + "loss": 0.2516, + "step": 42085 + }, + { + "epoch": 15.54, + "learning_rate": 1.4467545264195099e-06, + "loss": 0.1876, + "step": 42086 + }, + { + "epoch": 15.54, + "learning_rate": 1.4467175760561645e-06, + "loss": 0.3224, + "step": 42087 + }, + { + "epoch": 15.54, + "learning_rate": 1.4466806256928193e-06, + "loss": 0.158, + "step": 42088 + }, + { + "epoch": 15.54, + "learning_rate": 1.4466436753294742e-06, + "loss": 0.2764, + "step": 42089 + }, + { + "epoch": 15.54, + "learning_rate": 1.4466067249661288e-06, + "loss": 0.216, + "step": 42090 + }, + { + "epoch": 15.54, + "learning_rate": 1.4465697746027837e-06, + "loss": 0.2314, + "step": 42091 + }, + { + "epoch": 15.54, + "learning_rate": 1.4465328242394385e-06, + "loss": 0.2408, + "step": 42092 + }, + { + "epoch": 15.54, + "learning_rate": 1.4464958738760932e-06, + "loss": 0.2294, + "step": 42093 + }, + { + "epoch": 15.54, + "learning_rate": 1.4464589235127478e-06, + "loss": 0.3351, + "step": 42094 + }, + { + "epoch": 15.54, + "learning_rate": 1.4464219731494027e-06, + "loss": 0.234, + "step": 42095 + }, + { + "epoch": 15.55, + "learning_rate": 1.4463850227860573e-06, + "loss": 0.2773, + "step": 42096 + }, + { + "epoch": 15.55, + "learning_rate": 1.4463480724227121e-06, + "loss": 0.1938, + "step": 42097 + }, + { + "epoch": 15.55, + "learning_rate": 1.446311122059367e-06, + "loss": 0.3446, + "step": 42098 + }, + { + "epoch": 15.55, + "learning_rate": 1.4462741716960216e-06, + "loss": 0.3587, + "step": 42099 + }, + { + "epoch": 15.55, + "learning_rate": 1.4462372213326765e-06, + "loss": 0.1754, + "step": 42100 + }, + { + "epoch": 15.55, + "learning_rate": 1.4462002709693313e-06, + "loss": 0.188, + "step": 42101 + }, + { + "epoch": 15.55, + "learning_rate": 1.446163320605986e-06, + "loss": 0.4048, + "step": 42102 + }, + { + "epoch": 15.55, + "learning_rate": 1.4461263702426408e-06, + "loss": 0.4194, + "step": 42103 + }, + { + "epoch": 15.55, + "learning_rate": 1.4460894198792954e-06, + "loss": 0.2278, + "step": 42104 + }, + { + "epoch": 15.55, + "learning_rate": 1.4460524695159503e-06, + "loss": 0.2089, + "step": 42105 + }, + { + "epoch": 15.55, + "learning_rate": 1.446015519152605e-06, + "loss": 0.1362, + "step": 42106 + }, + { + "epoch": 15.55, + "learning_rate": 1.4459785687892598e-06, + "loss": 0.2849, + "step": 42107 + }, + { + "epoch": 15.55, + "learning_rate": 1.4459416184259146e-06, + "loss": 0.3022, + "step": 42108 + }, + { + "epoch": 15.55, + "learning_rate": 1.4459046680625693e-06, + "loss": 0.2638, + "step": 42109 + }, + { + "epoch": 15.55, + "learning_rate": 1.445867717699224e-06, + "loss": 0.1568, + "step": 42110 + }, + { + "epoch": 15.55, + "learning_rate": 1.445830767335879e-06, + "loss": 0.2656, + "step": 42111 + }, + { + "epoch": 15.55, + "learning_rate": 1.4457938169725336e-06, + "loss": 0.3679, + "step": 42112 + }, + { + "epoch": 15.55, + "learning_rate": 1.4457568666091884e-06, + "loss": 0.3181, + "step": 42113 + }, + { + "epoch": 15.55, + "learning_rate": 1.445719916245843e-06, + "loss": 0.132, + "step": 42114 + }, + { + "epoch": 15.55, + "learning_rate": 1.445682965882498e-06, + "loss": 0.2651, + "step": 42115 + }, + { + "epoch": 15.55, + "learning_rate": 1.4456460155191526e-06, + "loss": 0.1488, + "step": 42116 + }, + { + "epoch": 15.55, + "learning_rate": 1.4456090651558074e-06, + "loss": 0.1386, + "step": 42117 + }, + { + "epoch": 15.55, + "learning_rate": 1.4455721147924623e-06, + "loss": 0.171, + "step": 42118 + }, + { + "epoch": 15.55, + "learning_rate": 1.4455351644291169e-06, + "loss": 0.3072, + "step": 42119 + }, + { + "epoch": 15.55, + "learning_rate": 1.4454982140657717e-06, + "loss": 0.1828, + "step": 42120 + }, + { + "epoch": 15.55, + "learning_rate": 1.4454612637024266e-06, + "loss": 0.2548, + "step": 42121 + }, + { + "epoch": 15.55, + "learning_rate": 1.4454243133390812e-06, + "loss": 0.2676, + "step": 42122 + }, + { + "epoch": 15.56, + "learning_rate": 1.445387362975736e-06, + "loss": 0.3426, + "step": 42123 + }, + { + "epoch": 15.56, + "learning_rate": 1.4453504126123907e-06, + "loss": 0.2083, + "step": 42124 + }, + { + "epoch": 15.56, + "learning_rate": 1.4453134622490453e-06, + "loss": 0.3221, + "step": 42125 + }, + { + "epoch": 15.56, + "learning_rate": 1.4452765118857002e-06, + "loss": 0.1974, + "step": 42126 + }, + { + "epoch": 15.56, + "learning_rate": 1.445239561522355e-06, + "loss": 0.1473, + "step": 42127 + }, + { + "epoch": 15.56, + "learning_rate": 1.4452026111590097e-06, + "loss": 0.2358, + "step": 42128 + }, + { + "epoch": 15.56, + "learning_rate": 1.4451656607956645e-06, + "loss": 0.1443, + "step": 42129 + }, + { + "epoch": 15.56, + "learning_rate": 1.4451287104323194e-06, + "loss": 0.3323, + "step": 42130 + }, + { + "epoch": 15.56, + "learning_rate": 1.445091760068974e-06, + "loss": 0.3149, + "step": 42131 + }, + { + "epoch": 15.56, + "learning_rate": 1.4450548097056289e-06, + "loss": 0.4443, + "step": 42132 + }, + { + "epoch": 15.56, + "learning_rate": 1.4450178593422837e-06, + "loss": 0.1876, + "step": 42133 + }, + { + "epoch": 15.56, + "learning_rate": 1.4449809089789383e-06, + "loss": 0.2764, + "step": 42134 + }, + { + "epoch": 15.56, + "learning_rate": 1.444943958615593e-06, + "loss": 0.3645, + "step": 42135 + }, + { + "epoch": 15.56, + "learning_rate": 1.4449070082522478e-06, + "loss": 0.313, + "step": 42136 + }, + { + "epoch": 15.56, + "learning_rate": 1.4448700578889027e-06, + "loss": 0.3212, + "step": 42137 + }, + { + "epoch": 15.56, + "learning_rate": 1.4448331075255573e-06, + "loss": 0.1565, + "step": 42138 + }, + { + "epoch": 15.56, + "learning_rate": 1.4447961571622122e-06, + "loss": 0.2084, + "step": 42139 + }, + { + "epoch": 15.56, + "learning_rate": 1.444759206798867e-06, + "loss": 0.2293, + "step": 42140 + }, + { + "epoch": 15.56, + "learning_rate": 1.4447222564355216e-06, + "loss": 0.2877, + "step": 42141 + }, + { + "epoch": 15.56, + "learning_rate": 1.4446853060721765e-06, + "loss": 0.2953, + "step": 42142 + }, + { + "epoch": 15.56, + "learning_rate": 1.4446483557088313e-06, + "loss": 0.413, + "step": 42143 + }, + { + "epoch": 15.56, + "learning_rate": 1.4446114053454858e-06, + "loss": 0.125, + "step": 42144 + }, + { + "epoch": 15.56, + "learning_rate": 1.4445744549821406e-06, + "loss": 0.2433, + "step": 42145 + }, + { + "epoch": 15.56, + "learning_rate": 1.4445375046187955e-06, + "loss": 0.2254, + "step": 42146 + }, + { + "epoch": 15.56, + "learning_rate": 1.44450055425545e-06, + "loss": 0.2458, + "step": 42147 + }, + { + "epoch": 15.56, + "learning_rate": 1.444463603892105e-06, + "loss": 0.3688, + "step": 42148 + }, + { + "epoch": 15.56, + "learning_rate": 1.4444266535287598e-06, + "loss": 0.1575, + "step": 42149 + }, + { + "epoch": 15.56, + "learning_rate": 1.4443897031654144e-06, + "loss": 0.4212, + "step": 42150 + }, + { + "epoch": 15.57, + "learning_rate": 1.4443527528020693e-06, + "loss": 0.3639, + "step": 42151 + }, + { + "epoch": 15.57, + "learning_rate": 1.4443158024387241e-06, + "loss": 0.5548, + "step": 42152 + }, + { + "epoch": 15.57, + "learning_rate": 1.4442788520753788e-06, + "loss": 0.1387, + "step": 42153 + }, + { + "epoch": 15.57, + "learning_rate": 1.4442419017120334e-06, + "loss": 0.2273, + "step": 42154 + }, + { + "epoch": 15.57, + "learning_rate": 1.4442049513486882e-06, + "loss": 0.2765, + "step": 42155 + }, + { + "epoch": 15.57, + "learning_rate": 1.444168000985343e-06, + "loss": 0.2736, + "step": 42156 + }, + { + "epoch": 15.57, + "learning_rate": 1.4441310506219977e-06, + "loss": 0.2107, + "step": 42157 + }, + { + "epoch": 15.57, + "learning_rate": 1.4440941002586526e-06, + "loss": 0.3231, + "step": 42158 + }, + { + "epoch": 15.57, + "learning_rate": 1.4440571498953074e-06, + "loss": 0.2519, + "step": 42159 + }, + { + "epoch": 15.57, + "learning_rate": 1.444020199531962e-06, + "loss": 0.1884, + "step": 42160 + }, + { + "epoch": 15.57, + "learning_rate": 1.443983249168617e-06, + "loss": 0.2496, + "step": 42161 + }, + { + "epoch": 15.57, + "learning_rate": 1.4439462988052718e-06, + "loss": 0.2497, + "step": 42162 + }, + { + "epoch": 15.57, + "learning_rate": 1.4439093484419264e-06, + "loss": 0.2287, + "step": 42163 + }, + { + "epoch": 15.57, + "learning_rate": 1.443872398078581e-06, + "loss": 0.153, + "step": 42164 + }, + { + "epoch": 15.57, + "learning_rate": 1.4438354477152359e-06, + "loss": 0.2568, + "step": 42165 + }, + { + "epoch": 15.57, + "learning_rate": 1.4437984973518905e-06, + "loss": 0.1767, + "step": 42166 + }, + { + "epoch": 15.57, + "learning_rate": 1.4437615469885454e-06, + "loss": 0.1299, + "step": 42167 + }, + { + "epoch": 15.57, + "learning_rate": 1.4437245966252002e-06, + "loss": 0.2706, + "step": 42168 + }, + { + "epoch": 15.57, + "learning_rate": 1.4436876462618549e-06, + "loss": 0.1677, + "step": 42169 + }, + { + "epoch": 15.57, + "learning_rate": 1.4436506958985097e-06, + "loss": 0.1971, + "step": 42170 + }, + { + "epoch": 15.57, + "learning_rate": 1.4436137455351645e-06, + "loss": 0.2998, + "step": 42171 + }, + { + "epoch": 15.57, + "learning_rate": 1.4435767951718192e-06, + "loss": 0.2781, + "step": 42172 + }, + { + "epoch": 15.57, + "learning_rate": 1.443539844808474e-06, + "loss": 0.1417, + "step": 42173 + }, + { + "epoch": 15.57, + "learning_rate": 1.4435028944451287e-06, + "loss": 0.2232, + "step": 42174 + }, + { + "epoch": 15.57, + "learning_rate": 1.4434659440817835e-06, + "loss": 0.3653, + "step": 42175 + }, + { + "epoch": 15.57, + "learning_rate": 1.4434289937184382e-06, + "loss": 0.2704, + "step": 42176 + }, + { + "epoch": 15.57, + "learning_rate": 1.443392043355093e-06, + "loss": 0.2557, + "step": 42177 + }, + { + "epoch": 15.58, + "learning_rate": 1.4433550929917479e-06, + "loss": 0.2468, + "step": 42178 + }, + { + "epoch": 15.58, + "learning_rate": 1.4433181426284025e-06, + "loss": 0.3567, + "step": 42179 + }, + { + "epoch": 15.58, + "learning_rate": 1.4432811922650573e-06, + "loss": 0.262, + "step": 42180 + }, + { + "epoch": 15.58, + "learning_rate": 1.4432442419017122e-06, + "loss": 0.3251, + "step": 42181 + }, + { + "epoch": 15.58, + "learning_rate": 1.4432072915383668e-06, + "loss": 0.2646, + "step": 42182 + }, + { + "epoch": 15.58, + "learning_rate": 1.4431703411750217e-06, + "loss": 0.2968, + "step": 42183 + }, + { + "epoch": 15.58, + "learning_rate": 1.4431333908116763e-06, + "loss": 0.1813, + "step": 42184 + }, + { + "epoch": 15.58, + "learning_rate": 1.4430964404483312e-06, + "loss": 0.174, + "step": 42185 + }, + { + "epoch": 15.58, + "learning_rate": 1.4430594900849858e-06, + "loss": 0.4016, + "step": 42186 + }, + { + "epoch": 15.58, + "learning_rate": 1.4430225397216406e-06, + "loss": 0.1652, + "step": 42187 + }, + { + "epoch": 15.58, + "learning_rate": 1.4429855893582955e-06, + "loss": 0.3063, + "step": 42188 + }, + { + "epoch": 15.58, + "learning_rate": 1.4429486389949501e-06, + "loss": 0.3233, + "step": 42189 + }, + { + "epoch": 15.58, + "learning_rate": 1.442911688631605e-06, + "loss": 0.2206, + "step": 42190 + }, + { + "epoch": 15.58, + "learning_rate": 1.4428747382682598e-06, + "loss": 0.1556, + "step": 42191 + }, + { + "epoch": 15.58, + "learning_rate": 1.4428377879049145e-06, + "loss": 0.181, + "step": 42192 + }, + { + "epoch": 15.58, + "learning_rate": 1.4428008375415693e-06, + "loss": 0.2865, + "step": 42193 + }, + { + "epoch": 15.58, + "learning_rate": 1.442763887178224e-06, + "loss": 0.3032, + "step": 42194 + }, + { + "epoch": 15.58, + "learning_rate": 1.4427269368148786e-06, + "loss": 0.2405, + "step": 42195 + }, + { + "epoch": 15.58, + "learning_rate": 1.4426899864515334e-06, + "loss": 0.2722, + "step": 42196 + }, + { + "epoch": 15.58, + "learning_rate": 1.4426530360881883e-06, + "loss": 0.14, + "step": 42197 + }, + { + "epoch": 15.58, + "learning_rate": 1.442616085724843e-06, + "loss": 0.3365, + "step": 42198 + }, + { + "epoch": 15.58, + "learning_rate": 1.4425791353614978e-06, + "loss": 0.223, + "step": 42199 + }, + { + "epoch": 15.58, + "learning_rate": 1.4425421849981526e-06, + "loss": 0.2131, + "step": 42200 + }, + { + "epoch": 15.58, + "learning_rate": 1.4425052346348072e-06, + "loss": 0.135, + "step": 42201 + }, + { + "epoch": 15.58, + "learning_rate": 1.442468284271462e-06, + "loss": 0.2047, + "step": 42202 + }, + { + "epoch": 15.58, + "learning_rate": 1.442431333908117e-06, + "loss": 0.3355, + "step": 42203 + }, + { + "epoch": 15.58, + "learning_rate": 1.4423943835447716e-06, + "loss": 0.2405, + "step": 42204 + }, + { + "epoch": 15.59, + "learning_rate": 1.4423574331814262e-06, + "loss": 0.1751, + "step": 42205 + }, + { + "epoch": 15.59, + "learning_rate": 1.442320482818081e-06, + "loss": 0.2408, + "step": 42206 + }, + { + "epoch": 15.59, + "learning_rate": 1.442283532454736e-06, + "loss": 0.1589, + "step": 42207 + }, + { + "epoch": 15.59, + "learning_rate": 1.4422465820913905e-06, + "loss": 0.1451, + "step": 42208 + }, + { + "epoch": 15.59, + "learning_rate": 1.4422096317280454e-06, + "loss": 0.1952, + "step": 42209 + }, + { + "epoch": 15.59, + "learning_rate": 1.4421726813647002e-06, + "loss": 0.3947, + "step": 42210 + }, + { + "epoch": 15.59, + "learning_rate": 1.4421357310013549e-06, + "loss": 0.2222, + "step": 42211 + }, + { + "epoch": 15.59, + "learning_rate": 1.4420987806380097e-06, + "loss": 0.2887, + "step": 42212 + }, + { + "epoch": 15.59, + "learning_rate": 1.4420618302746646e-06, + "loss": 0.1663, + "step": 42213 + }, + { + "epoch": 15.59, + "learning_rate": 1.442024879911319e-06, + "loss": 0.1297, + "step": 42214 + }, + { + "epoch": 15.59, + "learning_rate": 1.4419879295479738e-06, + "loss": 0.2336, + "step": 42215 + }, + { + "epoch": 15.59, + "learning_rate": 1.4419509791846287e-06, + "loss": 0.2299, + "step": 42216 + }, + { + "epoch": 15.59, + "learning_rate": 1.4419140288212833e-06, + "loss": 0.2348, + "step": 42217 + }, + { + "epoch": 15.59, + "learning_rate": 1.4418770784579382e-06, + "loss": 0.2041, + "step": 42218 + }, + { + "epoch": 15.59, + "learning_rate": 1.441840128094593e-06, + "loss": 0.2222, + "step": 42219 + }, + { + "epoch": 15.59, + "learning_rate": 1.4418031777312477e-06, + "loss": 0.193, + "step": 42220 + }, + { + "epoch": 15.59, + "learning_rate": 1.4417662273679025e-06, + "loss": 0.2858, + "step": 42221 + }, + { + "epoch": 15.59, + "learning_rate": 1.4417292770045574e-06, + "loss": 0.135, + "step": 42222 + }, + { + "epoch": 15.59, + "learning_rate": 1.441692326641212e-06, + "loss": 0.2263, + "step": 42223 + }, + { + "epoch": 15.59, + "learning_rate": 1.4416553762778666e-06, + "loss": 0.2181, + "step": 42224 + }, + { + "epoch": 15.59, + "learning_rate": 1.4416184259145215e-06, + "loss": 0.3591, + "step": 42225 + }, + { + "epoch": 15.59, + "learning_rate": 1.4415814755511763e-06, + "loss": 0.3191, + "step": 42226 + }, + { + "epoch": 15.59, + "learning_rate": 1.441544525187831e-06, + "loss": 0.3205, + "step": 42227 + }, + { + "epoch": 15.59, + "learning_rate": 1.4415075748244858e-06, + "loss": 0.3115, + "step": 42228 + }, + { + "epoch": 15.59, + "learning_rate": 1.4414706244611407e-06, + "loss": 0.2331, + "step": 42229 + }, + { + "epoch": 15.59, + "learning_rate": 1.4414336740977953e-06, + "loss": 0.2119, + "step": 42230 + }, + { + "epoch": 15.59, + "learning_rate": 1.4413967237344501e-06, + "loss": 0.2995, + "step": 42231 + }, + { + "epoch": 15.6, + "learning_rate": 1.441359773371105e-06, + "loss": 0.2306, + "step": 42232 + }, + { + "epoch": 15.6, + "learning_rate": 1.4413228230077596e-06, + "loss": 0.1752, + "step": 42233 + }, + { + "epoch": 15.6, + "learning_rate": 1.4412858726444143e-06, + "loss": 0.2113, + "step": 42234 + }, + { + "epoch": 15.6, + "learning_rate": 1.4412489222810691e-06, + "loss": 0.2852, + "step": 42235 + }, + { + "epoch": 15.6, + "learning_rate": 1.441211971917724e-06, + "loss": 0.2456, + "step": 42236 + }, + { + "epoch": 15.6, + "learning_rate": 1.4411750215543786e-06, + "loss": 0.1992, + "step": 42237 + }, + { + "epoch": 15.6, + "learning_rate": 1.4411380711910334e-06, + "loss": 0.175, + "step": 42238 + }, + { + "epoch": 15.6, + "learning_rate": 1.441101120827688e-06, + "loss": 0.2664, + "step": 42239 + }, + { + "epoch": 15.6, + "learning_rate": 1.441064170464343e-06, + "loss": 0.3898, + "step": 42240 + }, + { + "epoch": 15.6, + "learning_rate": 1.4410272201009978e-06, + "loss": 0.1685, + "step": 42241 + }, + { + "epoch": 15.6, + "learning_rate": 1.4409902697376524e-06, + "loss": 0.4335, + "step": 42242 + }, + { + "epoch": 15.6, + "learning_rate": 1.4409533193743073e-06, + "loss": 0.2296, + "step": 42243 + }, + { + "epoch": 15.6, + "learning_rate": 1.440916369010962e-06, + "loss": 0.4312, + "step": 42244 + }, + { + "epoch": 15.6, + "learning_rate": 1.4408794186476168e-06, + "loss": 0.188, + "step": 42245 + }, + { + "epoch": 15.6, + "learning_rate": 1.4408424682842714e-06, + "loss": 0.3332, + "step": 42246 + }, + { + "epoch": 15.6, + "learning_rate": 1.4408055179209262e-06, + "loss": 0.2028, + "step": 42247 + }, + { + "epoch": 15.6, + "learning_rate": 1.440768567557581e-06, + "loss": 0.2557, + "step": 42248 + }, + { + "epoch": 15.6, + "learning_rate": 1.4407316171942357e-06, + "loss": 0.2119, + "step": 42249 + }, + { + "epoch": 15.6, + "learning_rate": 1.4406946668308906e-06, + "loss": 0.3619, + "step": 42250 + }, + { + "epoch": 15.6, + "learning_rate": 1.4406577164675454e-06, + "loss": 0.3483, + "step": 42251 + }, + { + "epoch": 15.6, + "learning_rate": 1.4406207661042e-06, + "loss": 0.2314, + "step": 42252 + }, + { + "epoch": 15.6, + "learning_rate": 1.440583815740855e-06, + "loss": 0.3219, + "step": 42253 + }, + { + "epoch": 15.6, + "learning_rate": 1.4405468653775095e-06, + "loss": 0.2026, + "step": 42254 + }, + { + "epoch": 15.6, + "learning_rate": 1.4405099150141644e-06, + "loss": 0.2726, + "step": 42255 + }, + { + "epoch": 15.6, + "learning_rate": 1.440472964650819e-06, + "loss": 0.2143, + "step": 42256 + }, + { + "epoch": 15.6, + "learning_rate": 1.4404360142874739e-06, + "loss": 0.3391, + "step": 42257 + }, + { + "epoch": 15.6, + "learning_rate": 1.4403990639241287e-06, + "loss": 0.2922, + "step": 42258 + }, + { + "epoch": 15.61, + "learning_rate": 1.4403621135607834e-06, + "loss": 0.2409, + "step": 42259 + }, + { + "epoch": 15.61, + "learning_rate": 1.4403251631974382e-06, + "loss": 0.2612, + "step": 42260 + }, + { + "epoch": 15.61, + "learning_rate": 1.440288212834093e-06, + "loss": 0.3161, + "step": 42261 + }, + { + "epoch": 15.61, + "learning_rate": 1.4402512624707477e-06, + "loss": 0.2752, + "step": 42262 + }, + { + "epoch": 15.61, + "learning_rate": 1.4402143121074025e-06, + "loss": 0.3338, + "step": 42263 + }, + { + "epoch": 15.61, + "learning_rate": 1.4401773617440572e-06, + "loss": 0.208, + "step": 42264 + }, + { + "epoch": 15.61, + "learning_rate": 1.4401404113807118e-06, + "loss": 0.2445, + "step": 42265 + }, + { + "epoch": 15.61, + "learning_rate": 1.4401034610173667e-06, + "loss": 0.3247, + "step": 42266 + }, + { + "epoch": 15.61, + "learning_rate": 1.4400665106540215e-06, + "loss": 0.2113, + "step": 42267 + }, + { + "epoch": 15.61, + "learning_rate": 1.4400295602906761e-06, + "loss": 0.2852, + "step": 42268 + }, + { + "epoch": 15.61, + "learning_rate": 1.439992609927331e-06, + "loss": 0.1443, + "step": 42269 + }, + { + "epoch": 15.61, + "learning_rate": 1.4399556595639858e-06, + "loss": 0.2702, + "step": 42270 + }, + { + "epoch": 15.61, + "learning_rate": 1.4399187092006405e-06, + "loss": 0.3078, + "step": 42271 + }, + { + "epoch": 15.61, + "learning_rate": 1.4398817588372953e-06, + "loss": 0.2211, + "step": 42272 + }, + { + "epoch": 15.61, + "learning_rate": 1.43984480847395e-06, + "loss": 0.1608, + "step": 42273 + }, + { + "epoch": 15.61, + "learning_rate": 1.4398078581106048e-06, + "loss": 0.2027, + "step": 42274 + }, + { + "epoch": 15.61, + "learning_rate": 1.4397709077472594e-06, + "loss": 0.2178, + "step": 42275 + }, + { + "epoch": 15.61, + "learning_rate": 1.4397339573839143e-06, + "loss": 0.415, + "step": 42276 + }, + { + "epoch": 15.61, + "learning_rate": 1.4396970070205691e-06, + "loss": 0.2883, + "step": 42277 + }, + { + "epoch": 15.61, + "learning_rate": 1.4396600566572238e-06, + "loss": 0.2301, + "step": 42278 + }, + { + "epoch": 15.61, + "learning_rate": 1.4396231062938786e-06, + "loss": 0.3469, + "step": 42279 + }, + { + "epoch": 15.61, + "learning_rate": 1.4395861559305335e-06, + "loss": 0.3296, + "step": 42280 + }, + { + "epoch": 15.61, + "learning_rate": 1.4395492055671881e-06, + "loss": 0.2442, + "step": 42281 + }, + { + "epoch": 15.61, + "learning_rate": 1.439512255203843e-06, + "loss": 0.3192, + "step": 42282 + }, + { + "epoch": 15.61, + "learning_rate": 1.4394753048404976e-06, + "loss": 0.322, + "step": 42283 + }, + { + "epoch": 15.61, + "learning_rate": 1.4394383544771522e-06, + "loss": 0.3177, + "step": 42284 + }, + { + "epoch": 15.61, + "learning_rate": 1.439401404113807e-06, + "loss": 0.3118, + "step": 42285 + }, + { + "epoch": 15.62, + "learning_rate": 1.439364453750462e-06, + "loss": 0.2268, + "step": 42286 + }, + { + "epoch": 15.62, + "learning_rate": 1.4393275033871166e-06, + "loss": 0.318, + "step": 42287 + }, + { + "epoch": 15.62, + "learning_rate": 1.4392905530237714e-06, + "loss": 0.1291, + "step": 42288 + }, + { + "epoch": 15.62, + "learning_rate": 1.4392536026604263e-06, + "loss": 0.3505, + "step": 42289 + }, + { + "epoch": 15.62, + "learning_rate": 1.439216652297081e-06, + "loss": 0.1841, + "step": 42290 + }, + { + "epoch": 15.62, + "learning_rate": 1.4391797019337357e-06, + "loss": 0.2816, + "step": 42291 + }, + { + "epoch": 15.62, + "learning_rate": 1.4391427515703906e-06, + "loss": 0.2146, + "step": 42292 + }, + { + "epoch": 15.62, + "learning_rate": 1.4391058012070452e-06, + "loss": 0.3299, + "step": 42293 + }, + { + "epoch": 15.62, + "learning_rate": 1.4390688508436999e-06, + "loss": 0.1628, + "step": 42294 + }, + { + "epoch": 15.62, + "learning_rate": 1.4390319004803547e-06, + "loss": 0.2762, + "step": 42295 + }, + { + "epoch": 15.62, + "learning_rate": 1.4389949501170096e-06, + "loss": 0.32, + "step": 42296 + }, + { + "epoch": 15.62, + "learning_rate": 1.4389579997536642e-06, + "loss": 0.2045, + "step": 42297 + }, + { + "epoch": 15.62, + "learning_rate": 1.438921049390319e-06, + "loss": 0.2574, + "step": 42298 + }, + { + "epoch": 15.62, + "learning_rate": 1.4388840990269739e-06, + "loss": 0.15, + "step": 42299 + }, + { + "epoch": 15.62, + "learning_rate": 1.4388471486636285e-06, + "loss": 0.1976, + "step": 42300 + }, + { + "epoch": 15.62, + "learning_rate": 1.4388101983002834e-06, + "loss": 0.1667, + "step": 42301 + }, + { + "epoch": 15.62, + "learning_rate": 1.4387732479369382e-06, + "loss": 0.2897, + "step": 42302 + }, + { + "epoch": 15.62, + "learning_rate": 1.4387362975735929e-06, + "loss": 0.2848, + "step": 42303 + }, + { + "epoch": 15.62, + "learning_rate": 1.4386993472102475e-06, + "loss": 0.1851, + "step": 42304 + }, + { + "epoch": 15.62, + "learning_rate": 1.4386623968469023e-06, + "loss": 0.3083, + "step": 42305 + }, + { + "epoch": 15.62, + "learning_rate": 1.4386254464835572e-06, + "loss": 0.3051, + "step": 42306 + }, + { + "epoch": 15.62, + "learning_rate": 1.4385884961202118e-06, + "loss": 0.2374, + "step": 42307 + }, + { + "epoch": 15.62, + "learning_rate": 1.4385515457568667e-06, + "loss": 0.2911, + "step": 42308 + }, + { + "epoch": 15.62, + "learning_rate": 1.4385145953935215e-06, + "loss": 0.3122, + "step": 42309 + }, + { + "epoch": 15.62, + "learning_rate": 1.4384776450301762e-06, + "loss": 0.3749, + "step": 42310 + }, + { + "epoch": 15.62, + "learning_rate": 1.438440694666831e-06, + "loss": 0.2286, + "step": 42311 + }, + { + "epoch": 15.62, + "learning_rate": 1.4384037443034859e-06, + "loss": 0.297, + "step": 42312 + }, + { + "epoch": 15.63, + "learning_rate": 1.4383667939401403e-06, + "loss": 0.2266, + "step": 42313 + }, + { + "epoch": 15.63, + "learning_rate": 1.4383298435767951e-06, + "loss": 0.3198, + "step": 42314 + }, + { + "epoch": 15.63, + "learning_rate": 1.43829289321345e-06, + "loss": 0.3447, + "step": 42315 + }, + { + "epoch": 15.63, + "learning_rate": 1.4382559428501046e-06, + "loss": 0.2155, + "step": 42316 + }, + { + "epoch": 15.63, + "learning_rate": 1.4382189924867595e-06, + "loss": 0.3156, + "step": 42317 + }, + { + "epoch": 15.63, + "learning_rate": 1.4381820421234143e-06, + "loss": 0.1393, + "step": 42318 + }, + { + "epoch": 15.63, + "learning_rate": 1.438145091760069e-06, + "loss": 0.2677, + "step": 42319 + }, + { + "epoch": 15.63, + "learning_rate": 1.4381081413967238e-06, + "loss": 0.2537, + "step": 42320 + }, + { + "epoch": 15.63, + "learning_rate": 1.4380711910333786e-06, + "loss": 0.2989, + "step": 42321 + }, + { + "epoch": 15.63, + "learning_rate": 1.4380342406700333e-06, + "loss": 0.241, + "step": 42322 + }, + { + "epoch": 15.63, + "learning_rate": 1.437997290306688e-06, + "loss": 0.3323, + "step": 42323 + }, + { + "epoch": 15.63, + "learning_rate": 1.4379603399433428e-06, + "loss": 0.2457, + "step": 42324 + }, + { + "epoch": 15.63, + "learning_rate": 1.4379233895799976e-06, + "loss": 0.3787, + "step": 42325 + }, + { + "epoch": 15.63, + "learning_rate": 1.4378864392166523e-06, + "loss": 0.2541, + "step": 42326 + }, + { + "epoch": 15.63, + "learning_rate": 1.437849488853307e-06, + "loss": 0.2549, + "step": 42327 + }, + { + "epoch": 15.63, + "learning_rate": 1.437812538489962e-06, + "loss": 0.3196, + "step": 42328 + }, + { + "epoch": 15.63, + "learning_rate": 1.4377755881266166e-06, + "loss": 0.3123, + "step": 42329 + }, + { + "epoch": 15.63, + "learning_rate": 1.4377386377632714e-06, + "loss": 0.2038, + "step": 42330 + }, + { + "epoch": 15.63, + "learning_rate": 1.4377016873999263e-06, + "loss": 0.29, + "step": 42331 + }, + { + "epoch": 15.63, + "learning_rate": 1.437664737036581e-06, + "loss": 0.2206, + "step": 42332 + }, + { + "epoch": 15.63, + "learning_rate": 1.4376277866732356e-06, + "loss": 0.145, + "step": 42333 + }, + { + "epoch": 15.63, + "learning_rate": 1.4375908363098904e-06, + "loss": 0.235, + "step": 42334 + }, + { + "epoch": 15.63, + "learning_rate": 1.437553885946545e-06, + "loss": 0.2806, + "step": 42335 + }, + { + "epoch": 15.63, + "learning_rate": 1.4375169355831999e-06, + "loss": 0.3976, + "step": 42336 + }, + { + "epoch": 15.63, + "learning_rate": 1.4374799852198547e-06, + "loss": 0.326, + "step": 42337 + }, + { + "epoch": 15.63, + "learning_rate": 1.4374430348565094e-06, + "loss": 0.2086, + "step": 42338 + }, + { + "epoch": 15.63, + "learning_rate": 1.4374060844931642e-06, + "loss": 0.2893, + "step": 42339 + }, + { + "epoch": 15.64, + "learning_rate": 1.437369134129819e-06, + "loss": 0.2371, + "step": 42340 + }, + { + "epoch": 15.64, + "learning_rate": 1.4373321837664737e-06, + "loss": 0.2465, + "step": 42341 + }, + { + "epoch": 15.64, + "learning_rate": 1.4372952334031286e-06, + "loss": 0.2387, + "step": 42342 + }, + { + "epoch": 15.64, + "learning_rate": 1.4372582830397832e-06, + "loss": 0.2805, + "step": 42343 + }, + { + "epoch": 15.64, + "learning_rate": 1.437221332676438e-06, + "loss": 0.258, + "step": 42344 + }, + { + "epoch": 15.64, + "learning_rate": 1.4371843823130927e-06, + "loss": 0.1996, + "step": 42345 + }, + { + "epoch": 15.64, + "learning_rate": 1.4371474319497475e-06, + "loss": 0.248, + "step": 42346 + }, + { + "epoch": 15.64, + "learning_rate": 1.4371104815864024e-06, + "loss": 0.2596, + "step": 42347 + }, + { + "epoch": 15.64, + "learning_rate": 1.437073531223057e-06, + "loss": 0.2869, + "step": 42348 + }, + { + "epoch": 15.64, + "learning_rate": 1.4370365808597119e-06, + "loss": 0.193, + "step": 42349 + }, + { + "epoch": 15.64, + "learning_rate": 1.4369996304963667e-06, + "loss": 0.2643, + "step": 42350 + }, + { + "epoch": 15.64, + "learning_rate": 1.4369626801330213e-06, + "loss": 0.1719, + "step": 42351 + }, + { + "epoch": 15.64, + "learning_rate": 1.4369257297696762e-06, + "loss": 0.2682, + "step": 42352 + }, + { + "epoch": 15.64, + "learning_rate": 1.4368887794063308e-06, + "loss": 0.2276, + "step": 42353 + }, + { + "epoch": 15.64, + "learning_rate": 1.4368518290429855e-06, + "loss": 0.2571, + "step": 42354 + }, + { + "epoch": 15.64, + "learning_rate": 1.4368148786796403e-06, + "loss": 0.2736, + "step": 42355 + }, + { + "epoch": 15.64, + "learning_rate": 1.4367779283162952e-06, + "loss": 0.3625, + "step": 42356 + }, + { + "epoch": 15.64, + "learning_rate": 1.4367409779529498e-06, + "loss": 0.247, + "step": 42357 + }, + { + "epoch": 15.64, + "learning_rate": 1.4367040275896046e-06, + "loss": 0.1927, + "step": 42358 + }, + { + "epoch": 15.64, + "learning_rate": 1.4366670772262595e-06, + "loss": 0.3773, + "step": 42359 + }, + { + "epoch": 15.64, + "learning_rate": 1.4366301268629141e-06, + "loss": 0.1529, + "step": 42360 + }, + { + "epoch": 15.64, + "learning_rate": 1.436593176499569e-06, + "loss": 0.2858, + "step": 42361 + }, + { + "epoch": 15.64, + "learning_rate": 1.4365562261362238e-06, + "loss": 0.267, + "step": 42362 + }, + { + "epoch": 15.64, + "learning_rate": 1.4365192757728785e-06, + "loss": 0.2833, + "step": 42363 + }, + { + "epoch": 15.64, + "learning_rate": 1.436482325409533e-06, + "loss": 0.2096, + "step": 42364 + }, + { + "epoch": 15.64, + "learning_rate": 1.436445375046188e-06, + "loss": 0.2188, + "step": 42365 + }, + { + "epoch": 15.64, + "learning_rate": 1.4364084246828428e-06, + "loss": 0.3625, + "step": 42366 + }, + { + "epoch": 15.65, + "learning_rate": 1.4363714743194974e-06, + "loss": 0.4076, + "step": 42367 + }, + { + "epoch": 15.65, + "learning_rate": 1.4363345239561523e-06, + "loss": 0.2104, + "step": 42368 + }, + { + "epoch": 15.65, + "learning_rate": 1.4362975735928071e-06, + "loss": 0.286, + "step": 42369 + }, + { + "epoch": 15.65, + "learning_rate": 1.4362606232294618e-06, + "loss": 0.2176, + "step": 42370 + }, + { + "epoch": 15.65, + "learning_rate": 1.4362236728661166e-06, + "loss": 0.3023, + "step": 42371 + }, + { + "epoch": 15.65, + "learning_rate": 1.4361867225027715e-06, + "loss": 0.1598, + "step": 42372 + }, + { + "epoch": 15.65, + "learning_rate": 1.436149772139426e-06, + "loss": 0.2204, + "step": 42373 + }, + { + "epoch": 15.65, + "learning_rate": 1.4361128217760807e-06, + "loss": 0.2313, + "step": 42374 + }, + { + "epoch": 15.65, + "learning_rate": 1.4360758714127356e-06, + "loss": 0.1535, + "step": 42375 + }, + { + "epoch": 15.65, + "learning_rate": 1.4360389210493904e-06, + "loss": 0.2847, + "step": 42376 + }, + { + "epoch": 15.65, + "learning_rate": 1.436001970686045e-06, + "loss": 0.2702, + "step": 42377 + }, + { + "epoch": 15.65, + "learning_rate": 1.4359650203227e-06, + "loss": 0.3015, + "step": 42378 + }, + { + "epoch": 15.65, + "learning_rate": 1.4359280699593548e-06, + "loss": 0.2195, + "step": 42379 + }, + { + "epoch": 15.65, + "learning_rate": 1.4358911195960094e-06, + "loss": 0.2209, + "step": 42380 + }, + { + "epoch": 15.65, + "learning_rate": 1.4358541692326642e-06, + "loss": 0.2752, + "step": 42381 + }, + { + "epoch": 15.65, + "learning_rate": 1.435817218869319e-06, + "loss": 0.1781, + "step": 42382 + }, + { + "epoch": 15.65, + "learning_rate": 1.4357802685059735e-06, + "loss": 0.1719, + "step": 42383 + }, + { + "epoch": 15.65, + "learning_rate": 1.4357433181426284e-06, + "loss": 0.2665, + "step": 42384 + }, + { + "epoch": 15.65, + "learning_rate": 1.4357063677792832e-06, + "loss": 0.195, + "step": 42385 + }, + { + "epoch": 15.65, + "learning_rate": 1.4356694174159379e-06, + "loss": 0.2205, + "step": 42386 + }, + { + "epoch": 15.65, + "learning_rate": 1.4356324670525927e-06, + "loss": 0.2747, + "step": 42387 + }, + { + "epoch": 15.65, + "learning_rate": 1.4355955166892475e-06, + "loss": 0.2292, + "step": 42388 + }, + { + "epoch": 15.65, + "learning_rate": 1.4355585663259022e-06, + "loss": 0.2201, + "step": 42389 + }, + { + "epoch": 15.65, + "learning_rate": 1.435521615962557e-06, + "loss": 0.218, + "step": 42390 + }, + { + "epoch": 15.65, + "learning_rate": 1.4354846655992119e-06, + "loss": 0.2399, + "step": 42391 + }, + { + "epoch": 15.65, + "learning_rate": 1.4354477152358665e-06, + "loss": 0.3228, + "step": 42392 + }, + { + "epoch": 15.65, + "learning_rate": 1.4354107648725212e-06, + "loss": 0.4045, + "step": 42393 + }, + { + "epoch": 15.66, + "learning_rate": 1.435373814509176e-06, + "loss": 0.2744, + "step": 42394 + }, + { + "epoch": 15.66, + "learning_rate": 1.4353368641458308e-06, + "loss": 0.1816, + "step": 42395 + }, + { + "epoch": 15.66, + "learning_rate": 1.4352999137824855e-06, + "loss": 0.2173, + "step": 42396 + }, + { + "epoch": 15.66, + "learning_rate": 1.4352629634191403e-06, + "loss": 0.303, + "step": 42397 + }, + { + "epoch": 15.66, + "learning_rate": 1.4352260130557952e-06, + "loss": 0.1532, + "step": 42398 + }, + { + "epoch": 15.66, + "learning_rate": 1.4351890626924498e-06, + "loss": 0.205, + "step": 42399 + }, + { + "epoch": 15.66, + "learning_rate": 1.4351521123291047e-06, + "loss": 0.1772, + "step": 42400 + }, + { + "epoch": 15.66, + "learning_rate": 1.4351151619657595e-06, + "loss": 0.2908, + "step": 42401 + }, + { + "epoch": 15.66, + "learning_rate": 1.4350782116024142e-06, + "loss": 0.2755, + "step": 42402 + }, + { + "epoch": 15.66, + "learning_rate": 1.4350412612390688e-06, + "loss": 0.1726, + "step": 42403 + }, + { + "epoch": 15.66, + "learning_rate": 1.4350043108757236e-06, + "loss": 0.2115, + "step": 42404 + }, + { + "epoch": 15.66, + "learning_rate": 1.4349673605123783e-06, + "loss": 0.1628, + "step": 42405 + }, + { + "epoch": 15.66, + "learning_rate": 1.4349304101490331e-06, + "loss": 0.2962, + "step": 42406 + }, + { + "epoch": 15.66, + "learning_rate": 1.434893459785688e-06, + "loss": 0.275, + "step": 42407 + }, + { + "epoch": 15.66, + "learning_rate": 1.4348565094223426e-06, + "loss": 0.4123, + "step": 42408 + }, + { + "epoch": 15.66, + "learning_rate": 1.4348195590589975e-06, + "loss": 0.3738, + "step": 42409 + }, + { + "epoch": 15.66, + "learning_rate": 1.4347826086956523e-06, + "loss": 0.3645, + "step": 42410 + }, + { + "epoch": 15.66, + "learning_rate": 1.434745658332307e-06, + "loss": 0.2915, + "step": 42411 + }, + { + "epoch": 15.66, + "learning_rate": 1.4347087079689618e-06, + "loss": 0.2226, + "step": 42412 + }, + { + "epoch": 15.66, + "learning_rate": 1.4346717576056164e-06, + "loss": 0.2593, + "step": 42413 + }, + { + "epoch": 15.66, + "learning_rate": 1.4346348072422713e-06, + "loss": 0.1229, + "step": 42414 + }, + { + "epoch": 15.66, + "learning_rate": 1.434597856878926e-06, + "loss": 0.286, + "step": 42415 + }, + { + "epoch": 15.66, + "learning_rate": 1.4345609065155808e-06, + "loss": 0.2412, + "step": 42416 + }, + { + "epoch": 15.66, + "learning_rate": 1.4345239561522356e-06, + "loss": 0.2769, + "step": 42417 + }, + { + "epoch": 15.66, + "learning_rate": 1.4344870057888902e-06, + "loss": 0.3335, + "step": 42418 + }, + { + "epoch": 15.66, + "learning_rate": 1.434450055425545e-06, + "loss": 0.2689, + "step": 42419 + }, + { + "epoch": 15.66, + "learning_rate": 1.4344131050622e-06, + "loss": 0.2468, + "step": 42420 + }, + { + "epoch": 15.67, + "learning_rate": 1.4343761546988546e-06, + "loss": 0.2665, + "step": 42421 + }, + { + "epoch": 15.67, + "learning_rate": 1.4343392043355094e-06, + "loss": 0.1901, + "step": 42422 + }, + { + "epoch": 15.67, + "learning_rate": 1.434302253972164e-06, + "loss": 0.2845, + "step": 42423 + }, + { + "epoch": 15.67, + "learning_rate": 1.434265303608819e-06, + "loss": 0.2087, + "step": 42424 + }, + { + "epoch": 15.67, + "learning_rate": 1.4342283532454735e-06, + "loss": 0.3442, + "step": 42425 + }, + { + "epoch": 15.67, + "learning_rate": 1.4341914028821284e-06, + "loss": 0.2107, + "step": 42426 + }, + { + "epoch": 15.67, + "learning_rate": 1.434154452518783e-06, + "loss": 0.2193, + "step": 42427 + }, + { + "epoch": 15.67, + "learning_rate": 1.4341175021554379e-06, + "loss": 0.2847, + "step": 42428 + }, + { + "epoch": 15.67, + "learning_rate": 1.4340805517920927e-06, + "loss": 0.2199, + "step": 42429 + }, + { + "epoch": 15.67, + "learning_rate": 1.4340436014287474e-06, + "loss": 0.2396, + "step": 42430 + }, + { + "epoch": 15.67, + "learning_rate": 1.4340066510654022e-06, + "loss": 0.1672, + "step": 42431 + }, + { + "epoch": 15.67, + "learning_rate": 1.433969700702057e-06, + "loss": 0.3006, + "step": 42432 + }, + { + "epoch": 15.67, + "learning_rate": 1.4339327503387117e-06, + "loss": 0.2909, + "step": 42433 + }, + { + "epoch": 15.67, + "learning_rate": 1.4338957999753663e-06, + "loss": 0.3666, + "step": 42434 + }, + { + "epoch": 15.67, + "learning_rate": 1.4338588496120212e-06, + "loss": 0.2941, + "step": 42435 + }, + { + "epoch": 15.67, + "learning_rate": 1.433821899248676e-06, + "loss": 0.3108, + "step": 42436 + }, + { + "epoch": 15.67, + "learning_rate": 1.4337849488853307e-06, + "loss": 0.2854, + "step": 42437 + }, + { + "epoch": 15.67, + "learning_rate": 1.4337479985219855e-06, + "loss": 0.2711, + "step": 42438 + }, + { + "epoch": 15.67, + "learning_rate": 1.4337110481586404e-06, + "loss": 0.2302, + "step": 42439 + }, + { + "epoch": 15.67, + "learning_rate": 1.433674097795295e-06, + "loss": 0.3309, + "step": 42440 + }, + { + "epoch": 15.67, + "learning_rate": 1.4336371474319498e-06, + "loss": 0.2737, + "step": 42441 + }, + { + "epoch": 15.67, + "learning_rate": 1.4336001970686047e-06, + "loss": 0.1532, + "step": 42442 + }, + { + "epoch": 15.67, + "learning_rate": 1.4335632467052593e-06, + "loss": 0.1702, + "step": 42443 + }, + { + "epoch": 15.67, + "learning_rate": 1.433526296341914e-06, + "loss": 0.2004, + "step": 42444 + }, + { + "epoch": 15.67, + "learning_rate": 1.4334893459785688e-06, + "loss": 0.3469, + "step": 42445 + }, + { + "epoch": 15.67, + "learning_rate": 1.4334523956152237e-06, + "loss": 0.1882, + "step": 42446 + }, + { + "epoch": 15.67, + "learning_rate": 1.4334154452518783e-06, + "loss": 0.3083, + "step": 42447 + }, + { + "epoch": 15.68, + "learning_rate": 1.4333784948885331e-06, + "loss": 0.2681, + "step": 42448 + }, + { + "epoch": 15.68, + "learning_rate": 1.433341544525188e-06, + "loss": 0.3656, + "step": 42449 + }, + { + "epoch": 15.68, + "learning_rate": 1.4333045941618426e-06, + "loss": 0.152, + "step": 42450 + }, + { + "epoch": 15.68, + "learning_rate": 1.4332676437984975e-06, + "loss": 0.2163, + "step": 42451 + }, + { + "epoch": 15.68, + "learning_rate": 1.4332306934351523e-06, + "loss": 0.1525, + "step": 42452 + }, + { + "epoch": 15.68, + "learning_rate": 1.4331937430718068e-06, + "loss": 0.2048, + "step": 42453 + }, + { + "epoch": 15.68, + "learning_rate": 1.4331567927084616e-06, + "loss": 0.2877, + "step": 42454 + }, + { + "epoch": 15.68, + "learning_rate": 1.4331198423451164e-06, + "loss": 0.2522, + "step": 42455 + }, + { + "epoch": 15.68, + "learning_rate": 1.433082891981771e-06, + "loss": 0.2388, + "step": 42456 + }, + { + "epoch": 15.68, + "learning_rate": 1.433045941618426e-06, + "loss": 0.2485, + "step": 42457 + }, + { + "epoch": 15.68, + "learning_rate": 1.4330089912550808e-06, + "loss": 0.1726, + "step": 42458 + }, + { + "epoch": 15.68, + "learning_rate": 1.4329720408917354e-06, + "loss": 0.1905, + "step": 42459 + }, + { + "epoch": 15.68, + "learning_rate": 1.4329350905283903e-06, + "loss": 0.3664, + "step": 42460 + }, + { + "epoch": 15.68, + "learning_rate": 1.4328981401650451e-06, + "loss": 0.3472, + "step": 42461 + }, + { + "epoch": 15.68, + "learning_rate": 1.4328611898016997e-06, + "loss": 0.263, + "step": 42462 + }, + { + "epoch": 15.68, + "learning_rate": 1.4328242394383544e-06, + "loss": 0.265, + "step": 42463 + }, + { + "epoch": 15.68, + "learning_rate": 1.4327872890750092e-06, + "loss": 0.213, + "step": 42464 + }, + { + "epoch": 15.68, + "learning_rate": 1.432750338711664e-06, + "loss": 0.3461, + "step": 42465 + }, + { + "epoch": 15.68, + "learning_rate": 1.4327133883483187e-06, + "loss": 0.1978, + "step": 42466 + }, + { + "epoch": 15.68, + "learning_rate": 1.4326764379849736e-06, + "loss": 0.2736, + "step": 42467 + }, + { + "epoch": 15.68, + "learning_rate": 1.4326394876216284e-06, + "loss": 0.2445, + "step": 42468 + }, + { + "epoch": 15.68, + "learning_rate": 1.432602537258283e-06, + "loss": 0.254, + "step": 42469 + }, + { + "epoch": 15.68, + "learning_rate": 1.432565586894938e-06, + "loss": 0.3704, + "step": 42470 + }, + { + "epoch": 15.68, + "learning_rate": 1.4325286365315927e-06, + "loss": 0.2357, + "step": 42471 + }, + { + "epoch": 15.68, + "learning_rate": 1.4324916861682474e-06, + "loss": 0.1913, + "step": 42472 + }, + { + "epoch": 15.68, + "learning_rate": 1.432454735804902e-06, + "loss": 0.2099, + "step": 42473 + }, + { + "epoch": 15.68, + "learning_rate": 1.4324177854415569e-06, + "loss": 0.1552, + "step": 42474 + }, + { + "epoch": 15.69, + "learning_rate": 1.4323808350782115e-06, + "loss": 0.2712, + "step": 42475 + }, + { + "epoch": 15.69, + "learning_rate": 1.4323438847148664e-06, + "loss": 0.3273, + "step": 42476 + }, + { + "epoch": 15.69, + "learning_rate": 1.4323069343515212e-06, + "loss": 0.1644, + "step": 42477 + }, + { + "epoch": 15.69, + "learning_rate": 1.4322699839881758e-06, + "loss": 0.1893, + "step": 42478 + }, + { + "epoch": 15.69, + "learning_rate": 1.4322330336248307e-06, + "loss": 0.3456, + "step": 42479 + }, + { + "epoch": 15.69, + "learning_rate": 1.4321960832614855e-06, + "loss": 0.3524, + "step": 42480 + }, + { + "epoch": 15.69, + "learning_rate": 1.4321591328981402e-06, + "loss": 0.2647, + "step": 42481 + }, + { + "epoch": 15.69, + "learning_rate": 1.432122182534795e-06, + "loss": 0.2042, + "step": 42482 + }, + { + "epoch": 15.69, + "learning_rate": 1.4320852321714497e-06, + "loss": 0.2889, + "step": 42483 + }, + { + "epoch": 15.69, + "learning_rate": 1.4320482818081045e-06, + "loss": 0.274, + "step": 42484 + }, + { + "epoch": 15.69, + "learning_rate": 1.4320113314447591e-06, + "loss": 0.3181, + "step": 42485 + }, + { + "epoch": 15.69, + "learning_rate": 1.431974381081414e-06, + "loss": 0.1966, + "step": 42486 + }, + { + "epoch": 15.69, + "learning_rate": 1.4319374307180688e-06, + "loss": 0.2712, + "step": 42487 + }, + { + "epoch": 15.69, + "learning_rate": 1.4319004803547235e-06, + "loss": 0.4186, + "step": 42488 + }, + { + "epoch": 15.69, + "learning_rate": 1.4318635299913783e-06, + "loss": 0.1866, + "step": 42489 + }, + { + "epoch": 15.69, + "learning_rate": 1.4318265796280332e-06, + "loss": 0.3054, + "step": 42490 + }, + { + "epoch": 15.69, + "learning_rate": 1.4317896292646878e-06, + "loss": 0.2042, + "step": 42491 + }, + { + "epoch": 15.69, + "learning_rate": 1.4317526789013427e-06, + "loss": 0.3111, + "step": 42492 + }, + { + "epoch": 15.69, + "learning_rate": 1.4317157285379973e-06, + "loss": 0.266, + "step": 42493 + }, + { + "epoch": 15.69, + "learning_rate": 1.4316787781746521e-06, + "loss": 0.1952, + "step": 42494 + }, + { + "epoch": 15.69, + "learning_rate": 1.4316418278113068e-06, + "loss": 0.3341, + "step": 42495 + }, + { + "epoch": 15.69, + "learning_rate": 1.4316048774479616e-06, + "loss": 0.2464, + "step": 42496 + }, + { + "epoch": 15.69, + "learning_rate": 1.4315679270846165e-06, + "loss": 0.301, + "step": 42497 + }, + { + "epoch": 15.69, + "learning_rate": 1.4315309767212711e-06, + "loss": 0.4362, + "step": 42498 + }, + { + "epoch": 15.69, + "learning_rate": 1.431494026357926e-06, + "loss": 0.3677, + "step": 42499 + }, + { + "epoch": 15.69, + "learning_rate": 1.4314570759945808e-06, + "loss": 0.1783, + "step": 42500 + }, + { + "epoch": 15.69, + "learning_rate": 1.4314201256312354e-06, + "loss": 0.2692, + "step": 42501 + }, + { + "epoch": 15.69, + "learning_rate": 1.43138317526789e-06, + "loss": 0.3216, + "step": 42502 + }, + { + "epoch": 15.7, + "learning_rate": 1.431346224904545e-06, + "loss": 0.1581, + "step": 42503 + }, + { + "epoch": 15.7, + "learning_rate": 1.4313092745411996e-06, + "loss": 0.1785, + "step": 42504 + }, + { + "epoch": 15.7, + "learning_rate": 1.4312723241778544e-06, + "loss": 0.1398, + "step": 42505 + }, + { + "epoch": 15.7, + "learning_rate": 1.4312353738145093e-06, + "loss": 0.2237, + "step": 42506 + }, + { + "epoch": 15.7, + "learning_rate": 1.431198423451164e-06, + "loss": 0.3518, + "step": 42507 + }, + { + "epoch": 15.7, + "learning_rate": 1.4311614730878187e-06, + "loss": 0.1583, + "step": 42508 + }, + { + "epoch": 15.7, + "learning_rate": 1.4311245227244736e-06, + "loss": 0.2084, + "step": 42509 + }, + { + "epoch": 15.7, + "learning_rate": 1.4310875723611282e-06, + "loss": 0.2641, + "step": 42510 + }, + { + "epoch": 15.7, + "learning_rate": 1.431050621997783e-06, + "loss": 0.2977, + "step": 42511 + }, + { + "epoch": 15.7, + "learning_rate": 1.4310136716344377e-06, + "loss": 0.2494, + "step": 42512 + }, + { + "epoch": 15.7, + "learning_rate": 1.4309767212710926e-06, + "loss": 0.3144, + "step": 42513 + }, + { + "epoch": 15.7, + "learning_rate": 1.4309397709077472e-06, + "loss": 0.2996, + "step": 42514 + }, + { + "epoch": 15.7, + "learning_rate": 1.430902820544402e-06, + "loss": 0.1473, + "step": 42515 + }, + { + "epoch": 15.7, + "learning_rate": 1.4308658701810569e-06, + "loss": 0.379, + "step": 42516 + }, + { + "epoch": 15.7, + "learning_rate": 1.4308289198177115e-06, + "loss": 0.2181, + "step": 42517 + }, + { + "epoch": 15.7, + "learning_rate": 1.4307919694543664e-06, + "loss": 0.3424, + "step": 42518 + }, + { + "epoch": 15.7, + "learning_rate": 1.4307550190910212e-06, + "loss": 0.1268, + "step": 42519 + }, + { + "epoch": 15.7, + "learning_rate": 1.4307180687276759e-06, + "loss": 0.1849, + "step": 42520 + }, + { + "epoch": 15.7, + "learning_rate": 1.4306811183643307e-06, + "loss": 0.3064, + "step": 42521 + }, + { + "epoch": 15.7, + "learning_rate": 1.4306441680009853e-06, + "loss": 0.3166, + "step": 42522 + }, + { + "epoch": 15.7, + "learning_rate": 1.43060721763764e-06, + "loss": 0.3491, + "step": 42523 + }, + { + "epoch": 15.7, + "learning_rate": 1.4305702672742948e-06, + "loss": 0.3453, + "step": 42524 + }, + { + "epoch": 15.7, + "learning_rate": 1.4305333169109497e-06, + "loss": 0.3308, + "step": 42525 + }, + { + "epoch": 15.7, + "learning_rate": 1.4304963665476043e-06, + "loss": 0.245, + "step": 42526 + }, + { + "epoch": 15.7, + "learning_rate": 1.4304594161842592e-06, + "loss": 0.1864, + "step": 42527 + }, + { + "epoch": 15.7, + "learning_rate": 1.430422465820914e-06, + "loss": 0.2437, + "step": 42528 + }, + { + "epoch": 15.7, + "learning_rate": 1.4303855154575686e-06, + "loss": 0.3553, + "step": 42529 + }, + { + "epoch": 15.71, + "learning_rate": 1.4303485650942235e-06, + "loss": 0.2524, + "step": 42530 + }, + { + "epoch": 15.71, + "learning_rate": 1.4303116147308783e-06, + "loss": 0.2348, + "step": 42531 + }, + { + "epoch": 15.71, + "learning_rate": 1.430274664367533e-06, + "loss": 0.187, + "step": 42532 + }, + { + "epoch": 15.71, + "learning_rate": 1.4302377140041876e-06, + "loss": 0.3138, + "step": 42533 + }, + { + "epoch": 15.71, + "learning_rate": 1.4302007636408425e-06, + "loss": 0.1926, + "step": 42534 + }, + { + "epoch": 15.71, + "learning_rate": 1.4301638132774973e-06, + "loss": 0.1695, + "step": 42535 + }, + { + "epoch": 15.71, + "learning_rate": 1.430126862914152e-06, + "loss": 0.3011, + "step": 42536 + }, + { + "epoch": 15.71, + "learning_rate": 1.4300899125508068e-06, + "loss": 0.2413, + "step": 42537 + }, + { + "epoch": 15.71, + "learning_rate": 1.4300529621874616e-06, + "loss": 0.2189, + "step": 42538 + }, + { + "epoch": 15.71, + "learning_rate": 1.4300160118241163e-06, + "loss": 0.2479, + "step": 42539 + }, + { + "epoch": 15.71, + "learning_rate": 1.4299790614607711e-06, + "loss": 0.2064, + "step": 42540 + }, + { + "epoch": 15.71, + "learning_rate": 1.429942111097426e-06, + "loss": 0.3836, + "step": 42541 + }, + { + "epoch": 15.71, + "learning_rate": 1.4299051607340804e-06, + "loss": 0.2282, + "step": 42542 + }, + { + "epoch": 15.71, + "learning_rate": 1.4298682103707353e-06, + "loss": 0.2391, + "step": 42543 + }, + { + "epoch": 15.71, + "learning_rate": 1.42983126000739e-06, + "loss": 0.2555, + "step": 42544 + }, + { + "epoch": 15.71, + "learning_rate": 1.4297943096440447e-06, + "loss": 0.2396, + "step": 42545 + }, + { + "epoch": 15.71, + "learning_rate": 1.4297573592806996e-06, + "loss": 0.3715, + "step": 42546 + }, + { + "epoch": 15.71, + "learning_rate": 1.4297204089173544e-06, + "loss": 0.2688, + "step": 42547 + }, + { + "epoch": 15.71, + "learning_rate": 1.429683458554009e-06, + "loss": 0.2362, + "step": 42548 + }, + { + "epoch": 15.71, + "learning_rate": 1.429646508190664e-06, + "loss": 0.3018, + "step": 42549 + }, + { + "epoch": 15.71, + "learning_rate": 1.4296095578273188e-06, + "loss": 0.3214, + "step": 42550 + }, + { + "epoch": 15.71, + "learning_rate": 1.4295726074639734e-06, + "loss": 0.3362, + "step": 42551 + }, + { + "epoch": 15.71, + "learning_rate": 1.429535657100628e-06, + "loss": 0.1909, + "step": 42552 + }, + { + "epoch": 15.71, + "learning_rate": 1.4294987067372829e-06, + "loss": 0.2754, + "step": 42553 + }, + { + "epoch": 15.71, + "learning_rate": 1.4294617563739377e-06, + "loss": 0.2653, + "step": 42554 + }, + { + "epoch": 15.71, + "learning_rate": 1.4294248060105924e-06, + "loss": 0.2894, + "step": 42555 + }, + { + "epoch": 15.71, + "learning_rate": 1.4293878556472472e-06, + "loss": 0.2116, + "step": 42556 + }, + { + "epoch": 15.72, + "learning_rate": 1.429350905283902e-06, + "loss": 0.2733, + "step": 42557 + }, + { + "epoch": 15.72, + "learning_rate": 1.4293139549205567e-06, + "loss": 0.3429, + "step": 42558 + }, + { + "epoch": 15.72, + "learning_rate": 1.4292770045572116e-06, + "loss": 0.3201, + "step": 42559 + }, + { + "epoch": 15.72, + "learning_rate": 1.4292400541938664e-06, + "loss": 0.1935, + "step": 42560 + }, + { + "epoch": 15.72, + "learning_rate": 1.429203103830521e-06, + "loss": 0.267, + "step": 42561 + }, + { + "epoch": 15.72, + "learning_rate": 1.4291661534671757e-06, + "loss": 0.3194, + "step": 42562 + }, + { + "epoch": 15.72, + "learning_rate": 1.4291292031038305e-06, + "loss": 0.1804, + "step": 42563 + }, + { + "epoch": 15.72, + "learning_rate": 1.4290922527404854e-06, + "loss": 0.2121, + "step": 42564 + }, + { + "epoch": 15.72, + "learning_rate": 1.42905530237714e-06, + "loss": 0.1331, + "step": 42565 + }, + { + "epoch": 15.72, + "learning_rate": 1.4290183520137949e-06, + "loss": 0.273, + "step": 42566 + }, + { + "epoch": 15.72, + "learning_rate": 1.4289814016504497e-06, + "loss": 0.1657, + "step": 42567 + }, + { + "epoch": 15.72, + "learning_rate": 1.4289444512871043e-06, + "loss": 0.1563, + "step": 42568 + }, + { + "epoch": 15.72, + "learning_rate": 1.4289075009237592e-06, + "loss": 0.2308, + "step": 42569 + }, + { + "epoch": 15.72, + "learning_rate": 1.428870550560414e-06, + "loss": 0.332, + "step": 42570 + }, + { + "epoch": 15.72, + "learning_rate": 1.4288336001970687e-06, + "loss": 0.1929, + "step": 42571 + }, + { + "epoch": 15.72, + "learning_rate": 1.4287966498337233e-06, + "loss": 0.2046, + "step": 42572 + }, + { + "epoch": 15.72, + "learning_rate": 1.4287596994703782e-06, + "loss": 0.2353, + "step": 42573 + }, + { + "epoch": 15.72, + "learning_rate": 1.4287227491070328e-06, + "loss": 0.5068, + "step": 42574 + }, + { + "epoch": 15.72, + "learning_rate": 1.4286857987436876e-06, + "loss": 0.1491, + "step": 42575 + }, + { + "epoch": 15.72, + "learning_rate": 1.4286488483803425e-06, + "loss": 0.0996, + "step": 42576 + }, + { + "epoch": 15.72, + "learning_rate": 1.4286118980169971e-06, + "loss": 0.2362, + "step": 42577 + }, + { + "epoch": 15.72, + "learning_rate": 1.428574947653652e-06, + "loss": 0.2978, + "step": 42578 + }, + { + "epoch": 15.72, + "learning_rate": 1.4285379972903068e-06, + "loss": 0.3749, + "step": 42579 + }, + { + "epoch": 15.72, + "learning_rate": 1.4285010469269615e-06, + "loss": 0.2531, + "step": 42580 + }, + { + "epoch": 15.72, + "learning_rate": 1.4284640965636163e-06, + "loss": 0.2408, + "step": 42581 + }, + { + "epoch": 15.72, + "learning_rate": 1.428427146200271e-06, + "loss": 0.2429, + "step": 42582 + }, + { + "epoch": 15.72, + "learning_rate": 1.4283901958369258e-06, + "loss": 0.1649, + "step": 42583 + }, + { + "epoch": 15.73, + "learning_rate": 1.4283532454735804e-06, + "loss": 0.2581, + "step": 42584 + }, + { + "epoch": 15.73, + "learning_rate": 1.4283162951102353e-06, + "loss": 0.2842, + "step": 42585 + }, + { + "epoch": 15.73, + "learning_rate": 1.4282793447468901e-06, + "loss": 0.2615, + "step": 42586 + }, + { + "epoch": 15.73, + "learning_rate": 1.4282423943835448e-06, + "loss": 0.1513, + "step": 42587 + }, + { + "epoch": 15.73, + "learning_rate": 1.4282054440201996e-06, + "loss": 0.3724, + "step": 42588 + }, + { + "epoch": 15.73, + "learning_rate": 1.4281684936568545e-06, + "loss": 0.2129, + "step": 42589 + }, + { + "epoch": 15.73, + "learning_rate": 1.428131543293509e-06, + "loss": 0.2179, + "step": 42590 + }, + { + "epoch": 15.73, + "learning_rate": 1.428094592930164e-06, + "loss": 0.311, + "step": 42591 + }, + { + "epoch": 15.73, + "learning_rate": 1.4280576425668186e-06, + "loss": 0.404, + "step": 42592 + }, + { + "epoch": 15.73, + "learning_rate": 1.4280206922034732e-06, + "loss": 0.1961, + "step": 42593 + }, + { + "epoch": 15.73, + "learning_rate": 1.427983741840128e-06, + "loss": 0.3015, + "step": 42594 + }, + { + "epoch": 15.73, + "learning_rate": 1.427946791476783e-06, + "loss": 0.2377, + "step": 42595 + }, + { + "epoch": 15.73, + "learning_rate": 1.4279098411134375e-06, + "loss": 0.2234, + "step": 42596 + }, + { + "epoch": 15.73, + "learning_rate": 1.4278728907500924e-06, + "loss": 0.3195, + "step": 42597 + }, + { + "epoch": 15.73, + "learning_rate": 1.4278359403867472e-06, + "loss": 0.2806, + "step": 42598 + }, + { + "epoch": 15.73, + "learning_rate": 1.4277989900234019e-06, + "loss": 0.2984, + "step": 42599 + }, + { + "epoch": 15.73, + "learning_rate": 1.4277620396600567e-06, + "loss": 0.2275, + "step": 42600 + }, + { + "epoch": 15.73, + "learning_rate": 1.4277250892967116e-06, + "loss": 0.3322, + "step": 42601 + }, + { + "epoch": 15.73, + "learning_rate": 1.4276881389333662e-06, + "loss": 0.3709, + "step": 42602 + }, + { + "epoch": 15.73, + "learning_rate": 1.4276511885700209e-06, + "loss": 0.2207, + "step": 42603 + }, + { + "epoch": 15.73, + "learning_rate": 1.4276142382066757e-06, + "loss": 0.2614, + "step": 42604 + }, + { + "epoch": 15.73, + "learning_rate": 1.4275772878433305e-06, + "loss": 0.2381, + "step": 42605 + }, + { + "epoch": 15.73, + "learning_rate": 1.4275403374799852e-06, + "loss": 0.2193, + "step": 42606 + }, + { + "epoch": 15.73, + "learning_rate": 1.42750338711664e-06, + "loss": 0.2111, + "step": 42607 + }, + { + "epoch": 15.73, + "learning_rate": 1.4274664367532949e-06, + "loss": 0.206, + "step": 42608 + }, + { + "epoch": 15.73, + "learning_rate": 1.4274294863899495e-06, + "loss": 0.2253, + "step": 42609 + }, + { + "epoch": 15.73, + "learning_rate": 1.4273925360266044e-06, + "loss": 0.2168, + "step": 42610 + }, + { + "epoch": 15.74, + "learning_rate": 1.4273555856632592e-06, + "loss": 0.2997, + "step": 42611 + }, + { + "epoch": 15.74, + "learning_rate": 1.4273186352999136e-06, + "loss": 0.2712, + "step": 42612 + }, + { + "epoch": 15.74, + "learning_rate": 1.4272816849365685e-06, + "loss": 0.1957, + "step": 42613 + }, + { + "epoch": 15.74, + "learning_rate": 1.4272447345732233e-06, + "loss": 0.1992, + "step": 42614 + }, + { + "epoch": 15.74, + "learning_rate": 1.427207784209878e-06, + "loss": 0.2777, + "step": 42615 + }, + { + "epoch": 15.74, + "learning_rate": 1.4271708338465328e-06, + "loss": 0.2792, + "step": 42616 + }, + { + "epoch": 15.74, + "learning_rate": 1.4271338834831877e-06, + "loss": 0.3269, + "step": 42617 + }, + { + "epoch": 15.74, + "learning_rate": 1.4270969331198423e-06, + "loss": 0.2023, + "step": 42618 + }, + { + "epoch": 15.74, + "learning_rate": 1.4270599827564972e-06, + "loss": 0.3344, + "step": 42619 + }, + { + "epoch": 15.74, + "learning_rate": 1.427023032393152e-06, + "loss": 0.3001, + "step": 42620 + }, + { + "epoch": 15.74, + "learning_rate": 1.4269860820298066e-06, + "loss": 0.2237, + "step": 42621 + }, + { + "epoch": 15.74, + "learning_rate": 1.4269491316664613e-06, + "loss": 0.1908, + "step": 42622 + }, + { + "epoch": 15.74, + "learning_rate": 1.4269121813031161e-06, + "loss": 0.303, + "step": 42623 + }, + { + "epoch": 15.74, + "learning_rate": 1.426875230939771e-06, + "loss": 0.2435, + "step": 42624 + }, + { + "epoch": 15.74, + "learning_rate": 1.4268382805764256e-06, + "loss": 0.1987, + "step": 42625 + }, + { + "epoch": 15.74, + "learning_rate": 1.4268013302130805e-06, + "loss": 0.2407, + "step": 42626 + }, + { + "epoch": 15.74, + "learning_rate": 1.4267643798497353e-06, + "loss": 0.3636, + "step": 42627 + }, + { + "epoch": 15.74, + "learning_rate": 1.42672742948639e-06, + "loss": 0.2407, + "step": 42628 + }, + { + "epoch": 15.74, + "learning_rate": 1.4266904791230448e-06, + "loss": 0.4592, + "step": 42629 + }, + { + "epoch": 15.74, + "learning_rate": 1.4266535287596996e-06, + "loss": 0.1401, + "step": 42630 + }, + { + "epoch": 15.74, + "learning_rate": 1.4266165783963543e-06, + "loss": 0.2153, + "step": 42631 + }, + { + "epoch": 15.74, + "learning_rate": 1.426579628033009e-06, + "loss": 0.2191, + "step": 42632 + }, + { + "epoch": 15.74, + "learning_rate": 1.4265426776696638e-06, + "loss": 0.3704, + "step": 42633 + }, + { + "epoch": 15.74, + "learning_rate": 1.4265057273063186e-06, + "loss": 0.1865, + "step": 42634 + }, + { + "epoch": 15.74, + "learning_rate": 1.4264687769429732e-06, + "loss": 0.2845, + "step": 42635 + }, + { + "epoch": 15.74, + "learning_rate": 1.426431826579628e-06, + "loss": 0.178, + "step": 42636 + }, + { + "epoch": 15.74, + "learning_rate": 1.426394876216283e-06, + "loss": 0.31, + "step": 42637 + }, + { + "epoch": 15.75, + "learning_rate": 1.4263579258529376e-06, + "loss": 0.2873, + "step": 42638 + }, + { + "epoch": 15.75, + "learning_rate": 1.4263209754895924e-06, + "loss": 0.2951, + "step": 42639 + }, + { + "epoch": 15.75, + "learning_rate": 1.4262840251262473e-06, + "loss": 0.253, + "step": 42640 + }, + { + "epoch": 15.75, + "learning_rate": 1.426247074762902e-06, + "loss": 0.2133, + "step": 42641 + }, + { + "epoch": 15.75, + "learning_rate": 1.4262101243995565e-06, + "loss": 0.2244, + "step": 42642 + }, + { + "epoch": 15.75, + "learning_rate": 1.4261731740362114e-06, + "loss": 0.2452, + "step": 42643 + }, + { + "epoch": 15.75, + "learning_rate": 1.426136223672866e-06, + "loss": 0.3192, + "step": 42644 + }, + { + "epoch": 15.75, + "learning_rate": 1.4260992733095209e-06, + "loss": 0.2213, + "step": 42645 + }, + { + "epoch": 15.75, + "learning_rate": 1.4260623229461757e-06, + "loss": 0.2587, + "step": 42646 + }, + { + "epoch": 15.75, + "learning_rate": 1.4260253725828304e-06, + "loss": 0.1918, + "step": 42647 + }, + { + "epoch": 15.75, + "learning_rate": 1.4259884222194852e-06, + "loss": 0.2996, + "step": 42648 + }, + { + "epoch": 15.75, + "learning_rate": 1.42595147185614e-06, + "loss": 0.2502, + "step": 42649 + }, + { + "epoch": 15.75, + "learning_rate": 1.4259145214927947e-06, + "loss": 0.1831, + "step": 42650 + }, + { + "epoch": 15.75, + "learning_rate": 1.4258775711294495e-06, + "loss": 0.1344, + "step": 42651 + }, + { + "epoch": 15.75, + "learning_rate": 1.4258406207661042e-06, + "loss": 0.2511, + "step": 42652 + }, + { + "epoch": 15.75, + "learning_rate": 1.425803670402759e-06, + "loss": 0.3508, + "step": 42653 + }, + { + "epoch": 15.75, + "learning_rate": 1.4257667200394137e-06, + "loss": 0.3748, + "step": 42654 + }, + { + "epoch": 15.75, + "learning_rate": 1.4257297696760685e-06, + "loss": 0.3231, + "step": 42655 + }, + { + "epoch": 15.75, + "learning_rate": 1.4256928193127234e-06, + "loss": 0.2332, + "step": 42656 + }, + { + "epoch": 15.75, + "learning_rate": 1.425655868949378e-06, + "loss": 0.2363, + "step": 42657 + }, + { + "epoch": 15.75, + "learning_rate": 1.4256189185860328e-06, + "loss": 0.2786, + "step": 42658 + }, + { + "epoch": 15.75, + "learning_rate": 1.4255819682226877e-06, + "loss": 0.2323, + "step": 42659 + }, + { + "epoch": 15.75, + "learning_rate": 1.4255450178593423e-06, + "loss": 0.1698, + "step": 42660 + }, + { + "epoch": 15.75, + "learning_rate": 1.4255080674959972e-06, + "loss": 0.2838, + "step": 42661 + }, + { + "epoch": 15.75, + "learning_rate": 1.4254711171326518e-06, + "loss": 0.2973, + "step": 42662 + }, + { + "epoch": 15.75, + "learning_rate": 1.4254341667693064e-06, + "loss": 0.2541, + "step": 42663 + }, + { + "epoch": 15.75, + "learning_rate": 1.4253972164059613e-06, + "loss": 0.1592, + "step": 42664 + }, + { + "epoch": 15.76, + "learning_rate": 1.4253602660426161e-06, + "loss": 0.2484, + "step": 42665 + }, + { + "epoch": 15.76, + "learning_rate": 1.4253233156792708e-06, + "loss": 0.1585, + "step": 42666 + }, + { + "epoch": 15.76, + "learning_rate": 1.4252863653159256e-06, + "loss": 0.3233, + "step": 42667 + }, + { + "epoch": 15.76, + "learning_rate": 1.4252494149525805e-06, + "loss": 0.3999, + "step": 42668 + }, + { + "epoch": 15.76, + "learning_rate": 1.4252124645892351e-06, + "loss": 0.217, + "step": 42669 + }, + { + "epoch": 15.76, + "learning_rate": 1.42517551422589e-06, + "loss": 0.2968, + "step": 42670 + }, + { + "epoch": 15.76, + "learning_rate": 1.4251385638625448e-06, + "loss": 0.3583, + "step": 42671 + }, + { + "epoch": 15.76, + "learning_rate": 1.4251016134991994e-06, + "loss": 0.2128, + "step": 42672 + }, + { + "epoch": 15.76, + "learning_rate": 1.425064663135854e-06, + "loss": 0.3595, + "step": 42673 + }, + { + "epoch": 15.76, + "learning_rate": 1.425027712772509e-06, + "loss": 0.307, + "step": 42674 + }, + { + "epoch": 15.76, + "learning_rate": 1.4249907624091638e-06, + "loss": 0.2892, + "step": 42675 + }, + { + "epoch": 15.76, + "learning_rate": 1.4249538120458184e-06, + "loss": 0.3226, + "step": 42676 + }, + { + "epoch": 15.76, + "learning_rate": 1.4249168616824733e-06, + "loss": 0.1166, + "step": 42677 + }, + { + "epoch": 15.76, + "learning_rate": 1.4248799113191281e-06, + "loss": 0.3157, + "step": 42678 + }, + { + "epoch": 15.76, + "learning_rate": 1.4248429609557827e-06, + "loss": 0.2309, + "step": 42679 + }, + { + "epoch": 15.76, + "learning_rate": 1.4248060105924376e-06, + "loss": 0.2362, + "step": 42680 + }, + { + "epoch": 15.76, + "learning_rate": 1.4247690602290924e-06, + "loss": 0.266, + "step": 42681 + }, + { + "epoch": 15.76, + "learning_rate": 1.424732109865747e-06, + "loss": 0.3483, + "step": 42682 + }, + { + "epoch": 15.76, + "learning_rate": 1.4246951595024017e-06, + "loss": 0.2197, + "step": 42683 + }, + { + "epoch": 15.76, + "learning_rate": 1.4246582091390566e-06, + "loss": 0.1296, + "step": 42684 + }, + { + "epoch": 15.76, + "learning_rate": 1.4246212587757114e-06, + "loss": 0.2076, + "step": 42685 + }, + { + "epoch": 15.76, + "learning_rate": 1.424584308412366e-06, + "loss": 0.2876, + "step": 42686 + }, + { + "epoch": 15.76, + "learning_rate": 1.424547358049021e-06, + "loss": 0.2736, + "step": 42687 + }, + { + "epoch": 15.76, + "learning_rate": 1.4245104076856757e-06, + "loss": 0.1941, + "step": 42688 + }, + { + "epoch": 15.76, + "learning_rate": 1.4244734573223304e-06, + "loss": 0.2512, + "step": 42689 + }, + { + "epoch": 15.76, + "learning_rate": 1.4244365069589852e-06, + "loss": 0.2821, + "step": 42690 + }, + { + "epoch": 15.76, + "learning_rate": 1.4243995565956399e-06, + "loss": 0.1773, + "step": 42691 + }, + { + "epoch": 15.77, + "learning_rate": 1.4243626062322945e-06, + "loss": 0.2019, + "step": 42692 + }, + { + "epoch": 15.77, + "learning_rate": 1.4243256558689494e-06, + "loss": 0.3023, + "step": 42693 + }, + { + "epoch": 15.77, + "learning_rate": 1.4242887055056042e-06, + "loss": 0.3847, + "step": 42694 + }, + { + "epoch": 15.77, + "learning_rate": 1.4242517551422588e-06, + "loss": 0.3086, + "step": 42695 + }, + { + "epoch": 15.77, + "learning_rate": 1.4242148047789137e-06, + "loss": 0.2597, + "step": 42696 + }, + { + "epoch": 15.77, + "learning_rate": 1.4241778544155685e-06, + "loss": 0.2503, + "step": 42697 + }, + { + "epoch": 15.77, + "learning_rate": 1.4241409040522232e-06, + "loss": 0.2908, + "step": 42698 + }, + { + "epoch": 15.77, + "learning_rate": 1.424103953688878e-06, + "loss": 0.2884, + "step": 42699 + }, + { + "epoch": 15.77, + "learning_rate": 1.4240670033255329e-06, + "loss": 0.189, + "step": 42700 + }, + { + "epoch": 15.77, + "learning_rate": 1.4240300529621875e-06, + "loss": 0.1692, + "step": 42701 + }, + { + "epoch": 15.77, + "learning_rate": 1.4239931025988421e-06, + "loss": 0.3754, + "step": 42702 + }, + { + "epoch": 15.77, + "learning_rate": 1.423956152235497e-06, + "loss": 0.1355, + "step": 42703 + }, + { + "epoch": 15.77, + "learning_rate": 1.4239192018721518e-06, + "loss": 0.2361, + "step": 42704 + }, + { + "epoch": 15.77, + "learning_rate": 1.4238822515088065e-06, + "loss": 0.3494, + "step": 42705 + }, + { + "epoch": 15.77, + "learning_rate": 1.4238453011454613e-06, + "loss": 0.2848, + "step": 42706 + }, + { + "epoch": 15.77, + "learning_rate": 1.4238083507821162e-06, + "loss": 0.2907, + "step": 42707 + }, + { + "epoch": 15.77, + "learning_rate": 1.4237714004187708e-06, + "loss": 0.2008, + "step": 42708 + }, + { + "epoch": 15.77, + "learning_rate": 1.4237344500554257e-06, + "loss": 0.2764, + "step": 42709 + }, + { + "epoch": 15.77, + "learning_rate": 1.4236974996920805e-06, + "loss": 0.1899, + "step": 42710 + }, + { + "epoch": 15.77, + "learning_rate": 1.4236605493287351e-06, + "loss": 0.2067, + "step": 42711 + }, + { + "epoch": 15.77, + "learning_rate": 1.4236235989653898e-06, + "loss": 0.3347, + "step": 42712 + }, + { + "epoch": 15.77, + "learning_rate": 1.4235866486020446e-06, + "loss": 0.252, + "step": 42713 + }, + { + "epoch": 15.77, + "learning_rate": 1.4235496982386993e-06, + "loss": 0.4257, + "step": 42714 + }, + { + "epoch": 15.77, + "learning_rate": 1.4235127478753541e-06, + "loss": 0.3063, + "step": 42715 + }, + { + "epoch": 15.77, + "learning_rate": 1.423475797512009e-06, + "loss": 0.1852, + "step": 42716 + }, + { + "epoch": 15.77, + "learning_rate": 1.4234388471486636e-06, + "loss": 0.1672, + "step": 42717 + }, + { + "epoch": 15.77, + "learning_rate": 1.4234018967853184e-06, + "loss": 0.199, + "step": 42718 + }, + { + "epoch": 15.78, + "learning_rate": 1.4233649464219733e-06, + "loss": 0.2914, + "step": 42719 + }, + { + "epoch": 15.78, + "learning_rate": 1.423327996058628e-06, + "loss": 0.155, + "step": 42720 + }, + { + "epoch": 15.78, + "learning_rate": 1.4232910456952828e-06, + "loss": 0.3201, + "step": 42721 + }, + { + "epoch": 15.78, + "learning_rate": 1.4232540953319374e-06, + "loss": 0.2406, + "step": 42722 + }, + { + "epoch": 15.78, + "learning_rate": 1.4232171449685923e-06, + "loss": 0.2918, + "step": 42723 + }, + { + "epoch": 15.78, + "learning_rate": 1.423180194605247e-06, + "loss": 0.2988, + "step": 42724 + }, + { + "epoch": 15.78, + "learning_rate": 1.4231432442419017e-06, + "loss": 0.1758, + "step": 42725 + }, + { + "epoch": 15.78, + "learning_rate": 1.4231062938785566e-06, + "loss": 0.1958, + "step": 42726 + }, + { + "epoch": 15.78, + "learning_rate": 1.4230693435152112e-06, + "loss": 0.2544, + "step": 42727 + }, + { + "epoch": 15.78, + "learning_rate": 1.423032393151866e-06, + "loss": 0.1697, + "step": 42728 + }, + { + "epoch": 15.78, + "learning_rate": 1.422995442788521e-06, + "loss": 0.2466, + "step": 42729 + }, + { + "epoch": 15.78, + "learning_rate": 1.4229584924251756e-06, + "loss": 0.2919, + "step": 42730 + }, + { + "epoch": 15.78, + "learning_rate": 1.4229215420618304e-06, + "loss": 0.2917, + "step": 42731 + }, + { + "epoch": 15.78, + "learning_rate": 1.422884591698485e-06, + "loss": 0.374, + "step": 42732 + }, + { + "epoch": 15.78, + "learning_rate": 1.4228476413351397e-06, + "loss": 0.2975, + "step": 42733 + }, + { + "epoch": 15.78, + "learning_rate": 1.4228106909717945e-06, + "loss": 0.2344, + "step": 42734 + }, + { + "epoch": 15.78, + "learning_rate": 1.4227737406084494e-06, + "loss": 0.3147, + "step": 42735 + }, + { + "epoch": 15.78, + "learning_rate": 1.422736790245104e-06, + "loss": 0.1881, + "step": 42736 + }, + { + "epoch": 15.78, + "learning_rate": 1.4226998398817589e-06, + "loss": 0.2296, + "step": 42737 + }, + { + "epoch": 15.78, + "learning_rate": 1.4226628895184137e-06, + "loss": 0.2026, + "step": 42738 + }, + { + "epoch": 15.78, + "learning_rate": 1.4226259391550683e-06, + "loss": 0.3343, + "step": 42739 + }, + { + "epoch": 15.78, + "learning_rate": 1.4225889887917232e-06, + "loss": 0.325, + "step": 42740 + }, + { + "epoch": 15.78, + "learning_rate": 1.4225520384283778e-06, + "loss": 0.3311, + "step": 42741 + }, + { + "epoch": 15.78, + "learning_rate": 1.4225150880650327e-06, + "loss": 0.2154, + "step": 42742 + }, + { + "epoch": 15.78, + "learning_rate": 1.4224781377016873e-06, + "loss": 0.3008, + "step": 42743 + }, + { + "epoch": 15.78, + "learning_rate": 1.4224411873383422e-06, + "loss": 0.2178, + "step": 42744 + }, + { + "epoch": 15.78, + "learning_rate": 1.422404236974997e-06, + "loss": 0.2009, + "step": 42745 + }, + { + "epoch": 15.79, + "learning_rate": 1.4223672866116516e-06, + "loss": 0.1649, + "step": 42746 + }, + { + "epoch": 15.79, + "learning_rate": 1.4223303362483065e-06, + "loss": 0.2443, + "step": 42747 + }, + { + "epoch": 15.79, + "learning_rate": 1.4222933858849613e-06, + "loss": 0.2883, + "step": 42748 + }, + { + "epoch": 15.79, + "learning_rate": 1.422256435521616e-06, + "loss": 0.2735, + "step": 42749 + }, + { + "epoch": 15.79, + "learning_rate": 1.4222194851582708e-06, + "loss": 0.2393, + "step": 42750 + }, + { + "epoch": 15.79, + "learning_rate": 1.4221825347949255e-06, + "loss": 0.2781, + "step": 42751 + }, + { + "epoch": 15.79, + "learning_rate": 1.4221455844315803e-06, + "loss": 0.3042, + "step": 42752 + }, + { + "epoch": 15.79, + "learning_rate": 1.422108634068235e-06, + "loss": 0.1912, + "step": 42753 + }, + { + "epoch": 15.79, + "learning_rate": 1.4220716837048898e-06, + "loss": 0.3161, + "step": 42754 + }, + { + "epoch": 15.79, + "learning_rate": 1.4220347333415446e-06, + "loss": 0.3155, + "step": 42755 + }, + { + "epoch": 15.79, + "learning_rate": 1.4219977829781993e-06, + "loss": 0.2775, + "step": 42756 + }, + { + "epoch": 15.79, + "learning_rate": 1.4219608326148541e-06, + "loss": 0.266, + "step": 42757 + }, + { + "epoch": 15.79, + "learning_rate": 1.421923882251509e-06, + "loss": 0.3078, + "step": 42758 + }, + { + "epoch": 15.79, + "learning_rate": 1.4218869318881636e-06, + "loss": 0.2263, + "step": 42759 + }, + { + "epoch": 15.79, + "learning_rate": 1.4218499815248185e-06, + "loss": 0.2921, + "step": 42760 + }, + { + "epoch": 15.79, + "learning_rate": 1.421813031161473e-06, + "loss": 0.3263, + "step": 42761 + }, + { + "epoch": 15.79, + "learning_rate": 1.4217760807981277e-06, + "loss": 0.3424, + "step": 42762 + }, + { + "epoch": 15.79, + "learning_rate": 1.4217391304347826e-06, + "loss": 0.2012, + "step": 42763 + }, + { + "epoch": 15.79, + "learning_rate": 1.4217021800714374e-06, + "loss": 0.4119, + "step": 42764 + }, + { + "epoch": 15.79, + "learning_rate": 1.421665229708092e-06, + "loss": 0.3722, + "step": 42765 + }, + { + "epoch": 15.79, + "learning_rate": 1.421628279344747e-06, + "loss": 0.2443, + "step": 42766 + }, + { + "epoch": 15.79, + "learning_rate": 1.4215913289814018e-06, + "loss": 0.1463, + "step": 42767 + }, + { + "epoch": 15.79, + "learning_rate": 1.4215543786180564e-06, + "loss": 0.2804, + "step": 42768 + }, + { + "epoch": 15.79, + "learning_rate": 1.4215174282547113e-06, + "loss": 0.2271, + "step": 42769 + }, + { + "epoch": 15.79, + "learning_rate": 1.421480477891366e-06, + "loss": 0.4188, + "step": 42770 + }, + { + "epoch": 15.79, + "learning_rate": 1.4214435275280207e-06, + "loss": 0.2588, + "step": 42771 + }, + { + "epoch": 15.79, + "learning_rate": 1.4214065771646754e-06, + "loss": 0.4149, + "step": 42772 + }, + { + "epoch": 15.8, + "learning_rate": 1.4213696268013302e-06, + "loss": 0.1782, + "step": 42773 + }, + { + "epoch": 15.8, + "learning_rate": 1.421332676437985e-06, + "loss": 0.3253, + "step": 42774 + }, + { + "epoch": 15.8, + "learning_rate": 1.4212957260746397e-06, + "loss": 0.2007, + "step": 42775 + }, + { + "epoch": 15.8, + "learning_rate": 1.4212587757112946e-06, + "loss": 0.352, + "step": 42776 + }, + { + "epoch": 15.8, + "learning_rate": 1.4212218253479494e-06, + "loss": 0.3917, + "step": 42777 + }, + { + "epoch": 15.8, + "learning_rate": 1.421184874984604e-06, + "loss": 0.2508, + "step": 42778 + }, + { + "epoch": 15.8, + "learning_rate": 1.4211479246212589e-06, + "loss": 0.2061, + "step": 42779 + }, + { + "epoch": 15.8, + "learning_rate": 1.4211109742579137e-06, + "loss": 0.2902, + "step": 42780 + }, + { + "epoch": 15.8, + "learning_rate": 1.4210740238945682e-06, + "loss": 0.3038, + "step": 42781 + }, + { + "epoch": 15.8, + "learning_rate": 1.421037073531223e-06, + "loss": 0.3401, + "step": 42782 + }, + { + "epoch": 15.8, + "learning_rate": 1.4210001231678779e-06, + "loss": 0.4283, + "step": 42783 + }, + { + "epoch": 15.8, + "learning_rate": 1.4209631728045325e-06, + "loss": 0.2388, + "step": 42784 + }, + { + "epoch": 15.8, + "learning_rate": 1.4209262224411873e-06, + "loss": 0.268, + "step": 42785 + }, + { + "epoch": 15.8, + "learning_rate": 1.4208892720778422e-06, + "loss": 0.1439, + "step": 42786 + }, + { + "epoch": 15.8, + "learning_rate": 1.4208523217144968e-06, + "loss": 0.1728, + "step": 42787 + }, + { + "epoch": 15.8, + "learning_rate": 1.4208153713511517e-06, + "loss": 0.1618, + "step": 42788 + }, + { + "epoch": 15.8, + "learning_rate": 1.4207784209878065e-06, + "loss": 0.2535, + "step": 42789 + }, + { + "epoch": 15.8, + "learning_rate": 1.4207414706244612e-06, + "loss": 0.3061, + "step": 42790 + }, + { + "epoch": 15.8, + "learning_rate": 1.4207045202611158e-06, + "loss": 0.3066, + "step": 42791 + }, + { + "epoch": 15.8, + "learning_rate": 1.4206675698977706e-06, + "loss": 0.2888, + "step": 42792 + }, + { + "epoch": 15.8, + "learning_rate": 1.4206306195344255e-06, + "loss": 0.2025, + "step": 42793 + }, + { + "epoch": 15.8, + "learning_rate": 1.4205936691710801e-06, + "loss": 0.3164, + "step": 42794 + }, + { + "epoch": 15.8, + "learning_rate": 1.420556718807735e-06, + "loss": 0.1655, + "step": 42795 + }, + { + "epoch": 15.8, + "learning_rate": 1.4205197684443898e-06, + "loss": 0.1577, + "step": 42796 + }, + { + "epoch": 15.8, + "learning_rate": 1.4204828180810445e-06, + "loss": 0.2819, + "step": 42797 + }, + { + "epoch": 15.8, + "learning_rate": 1.4204458677176993e-06, + "loss": 0.2518, + "step": 42798 + }, + { + "epoch": 15.8, + "learning_rate": 1.4204089173543542e-06, + "loss": 0.107, + "step": 42799 + }, + { + "epoch": 15.81, + "learning_rate": 1.4203719669910088e-06, + "loss": 0.2765, + "step": 42800 + }, + { + "epoch": 15.81, + "learning_rate": 1.4203350166276634e-06, + "loss": 0.2618, + "step": 42801 + }, + { + "epoch": 15.81, + "learning_rate": 1.4202980662643183e-06, + "loss": 0.1724, + "step": 42802 + }, + { + "epoch": 15.81, + "learning_rate": 1.420261115900973e-06, + "loss": 0.1703, + "step": 42803 + }, + { + "epoch": 15.81, + "learning_rate": 1.4202241655376278e-06, + "loss": 0.2766, + "step": 42804 + }, + { + "epoch": 15.81, + "learning_rate": 1.4201872151742826e-06, + "loss": 0.2469, + "step": 42805 + }, + { + "epoch": 15.81, + "learning_rate": 1.4201502648109372e-06, + "loss": 0.1873, + "step": 42806 + }, + { + "epoch": 15.81, + "learning_rate": 1.420113314447592e-06, + "loss": 0.1855, + "step": 42807 + }, + { + "epoch": 15.81, + "learning_rate": 1.420076364084247e-06, + "loss": 0.3013, + "step": 42808 + }, + { + "epoch": 15.81, + "learning_rate": 1.4200394137209016e-06, + "loss": 0.2, + "step": 42809 + }, + { + "epoch": 15.81, + "learning_rate": 1.4200024633575564e-06, + "loss": 0.1689, + "step": 42810 + }, + { + "epoch": 15.81, + "learning_rate": 1.419965512994211e-06, + "loss": 0.1893, + "step": 42811 + }, + { + "epoch": 15.81, + "learning_rate": 1.419928562630866e-06, + "loss": 0.3124, + "step": 42812 + }, + { + "epoch": 15.81, + "learning_rate": 1.4198916122675205e-06, + "loss": 0.2068, + "step": 42813 + }, + { + "epoch": 15.81, + "learning_rate": 1.4198546619041754e-06, + "loss": 0.266, + "step": 42814 + }, + { + "epoch": 15.81, + "learning_rate": 1.4198177115408302e-06, + "loss": 0.2005, + "step": 42815 + }, + { + "epoch": 15.81, + "learning_rate": 1.4197807611774849e-06, + "loss": 0.3305, + "step": 42816 + }, + { + "epoch": 15.81, + "learning_rate": 1.4197438108141397e-06, + "loss": 0.1662, + "step": 42817 + }, + { + "epoch": 15.81, + "learning_rate": 1.4197068604507946e-06, + "loss": 0.1606, + "step": 42818 + }, + { + "epoch": 15.81, + "learning_rate": 1.4196699100874492e-06, + "loss": 0.2772, + "step": 42819 + }, + { + "epoch": 15.81, + "learning_rate": 1.419632959724104e-06, + "loss": 0.2333, + "step": 42820 + }, + { + "epoch": 15.81, + "learning_rate": 1.4195960093607587e-06, + "loss": 0.2789, + "step": 42821 + }, + { + "epoch": 15.81, + "learning_rate": 1.4195590589974135e-06, + "loss": 0.2266, + "step": 42822 + }, + { + "epoch": 15.81, + "learning_rate": 1.4195221086340682e-06, + "loss": 0.1947, + "step": 42823 + }, + { + "epoch": 15.81, + "learning_rate": 1.419485158270723e-06, + "loss": 0.2944, + "step": 42824 + }, + { + "epoch": 15.81, + "learning_rate": 1.4194482079073779e-06, + "loss": 0.2557, + "step": 42825 + }, + { + "epoch": 15.81, + "learning_rate": 1.4194112575440325e-06, + "loss": 0.2005, + "step": 42826 + }, + { + "epoch": 15.81, + "learning_rate": 1.4193743071806874e-06, + "loss": 0.2762, + "step": 42827 + }, + { + "epoch": 15.82, + "learning_rate": 1.4193373568173422e-06, + "loss": 0.1935, + "step": 42828 + }, + { + "epoch": 15.82, + "learning_rate": 1.4193004064539968e-06, + "loss": 0.256, + "step": 42829 + }, + { + "epoch": 15.82, + "learning_rate": 1.4192634560906517e-06, + "loss": 0.2916, + "step": 42830 + }, + { + "epoch": 15.82, + "learning_rate": 1.4192265057273063e-06, + "loss": 0.3412, + "step": 42831 + }, + { + "epoch": 15.82, + "learning_rate": 1.419189555363961e-06, + "loss": 0.3384, + "step": 42832 + }, + { + "epoch": 15.82, + "learning_rate": 1.4191526050006158e-06, + "loss": 0.1895, + "step": 42833 + }, + { + "epoch": 15.82, + "learning_rate": 1.4191156546372707e-06, + "loss": 0.2071, + "step": 42834 + }, + { + "epoch": 15.82, + "learning_rate": 1.4190787042739253e-06, + "loss": 0.1659, + "step": 42835 + }, + { + "epoch": 15.82, + "learning_rate": 1.4190417539105802e-06, + "loss": 0.2147, + "step": 42836 + }, + { + "epoch": 15.82, + "learning_rate": 1.419004803547235e-06, + "loss": 0.2987, + "step": 42837 + }, + { + "epoch": 15.82, + "learning_rate": 1.4189678531838896e-06, + "loss": 0.224, + "step": 42838 + }, + { + "epoch": 15.82, + "learning_rate": 1.4189309028205445e-06, + "loss": 0.2888, + "step": 42839 + }, + { + "epoch": 15.82, + "learning_rate": 1.4188939524571993e-06, + "loss": 0.2354, + "step": 42840 + }, + { + "epoch": 15.82, + "learning_rate": 1.418857002093854e-06, + "loss": 0.3611, + "step": 42841 + }, + { + "epoch": 15.82, + "learning_rate": 1.4188200517305086e-06, + "loss": 0.145, + "step": 42842 + }, + { + "epoch": 15.82, + "learning_rate": 1.4187831013671635e-06, + "loss": 0.1999, + "step": 42843 + }, + { + "epoch": 15.82, + "learning_rate": 1.4187461510038183e-06, + "loss": 0.2145, + "step": 42844 + }, + { + "epoch": 15.82, + "learning_rate": 1.418709200640473e-06, + "loss": 0.2893, + "step": 42845 + }, + { + "epoch": 15.82, + "learning_rate": 1.4186722502771278e-06, + "loss": 0.3717, + "step": 42846 + }, + { + "epoch": 15.82, + "learning_rate": 1.4186352999137826e-06, + "loss": 0.1809, + "step": 42847 + }, + { + "epoch": 15.82, + "learning_rate": 1.4185983495504373e-06, + "loss": 0.2502, + "step": 42848 + }, + { + "epoch": 15.82, + "learning_rate": 1.4185613991870921e-06, + "loss": 0.2882, + "step": 42849 + }, + { + "epoch": 15.82, + "learning_rate": 1.418524448823747e-06, + "loss": 0.2043, + "step": 42850 + }, + { + "epoch": 15.82, + "learning_rate": 1.4184874984604014e-06, + "loss": 0.3094, + "step": 42851 + }, + { + "epoch": 15.82, + "learning_rate": 1.4184505480970562e-06, + "loss": 0.3866, + "step": 42852 + }, + { + "epoch": 15.82, + "learning_rate": 1.418413597733711e-06, + "loss": 0.1163, + "step": 42853 + }, + { + "epoch": 15.82, + "learning_rate": 1.4183766473703657e-06, + "loss": 0.2965, + "step": 42854 + }, + { + "epoch": 15.83, + "learning_rate": 1.4183396970070206e-06, + "loss": 0.2161, + "step": 42855 + }, + { + "epoch": 15.83, + "learning_rate": 1.4183027466436754e-06, + "loss": 0.1633, + "step": 42856 + }, + { + "epoch": 15.83, + "learning_rate": 1.41826579628033e-06, + "loss": 0.2775, + "step": 42857 + }, + { + "epoch": 15.83, + "learning_rate": 1.418228845916985e-06, + "loss": 0.2151, + "step": 42858 + }, + { + "epoch": 15.83, + "learning_rate": 1.4181918955536398e-06, + "loss": 0.3968, + "step": 42859 + }, + { + "epoch": 15.83, + "learning_rate": 1.4181549451902944e-06, + "loss": 0.2083, + "step": 42860 + }, + { + "epoch": 15.83, + "learning_rate": 1.418117994826949e-06, + "loss": 0.2673, + "step": 42861 + }, + { + "epoch": 15.83, + "learning_rate": 1.4180810444636039e-06, + "loss": 0.3288, + "step": 42862 + }, + { + "epoch": 15.83, + "learning_rate": 1.4180440941002587e-06, + "loss": 0.2609, + "step": 42863 + }, + { + "epoch": 15.83, + "learning_rate": 1.4180071437369134e-06, + "loss": 0.271, + "step": 42864 + }, + { + "epoch": 15.83, + "learning_rate": 1.4179701933735682e-06, + "loss": 0.2079, + "step": 42865 + }, + { + "epoch": 15.83, + "learning_rate": 1.417933243010223e-06, + "loss": 0.1498, + "step": 42866 + }, + { + "epoch": 15.83, + "learning_rate": 1.4178962926468777e-06, + "loss": 0.3796, + "step": 42867 + }, + { + "epoch": 15.83, + "learning_rate": 1.4178593422835325e-06, + "loss": 0.3506, + "step": 42868 + }, + { + "epoch": 15.83, + "learning_rate": 1.4178223919201874e-06, + "loss": 0.207, + "step": 42869 + }, + { + "epoch": 15.83, + "learning_rate": 1.417785441556842e-06, + "loss": 0.2039, + "step": 42870 + }, + { + "epoch": 15.83, + "learning_rate": 1.4177484911934967e-06, + "loss": 0.2258, + "step": 42871 + }, + { + "epoch": 15.83, + "learning_rate": 1.4177115408301515e-06, + "loss": 0.2316, + "step": 42872 + }, + { + "epoch": 15.83, + "learning_rate": 1.4176745904668064e-06, + "loss": 0.2915, + "step": 42873 + }, + { + "epoch": 15.83, + "learning_rate": 1.417637640103461e-06, + "loss": 0.1029, + "step": 42874 + }, + { + "epoch": 15.83, + "learning_rate": 1.4176006897401158e-06, + "loss": 0.2503, + "step": 42875 + }, + { + "epoch": 15.83, + "learning_rate": 1.4175637393767705e-06, + "loss": 0.1938, + "step": 42876 + }, + { + "epoch": 15.83, + "learning_rate": 1.4175267890134253e-06, + "loss": 0.1638, + "step": 42877 + }, + { + "epoch": 15.83, + "learning_rate": 1.4174898386500802e-06, + "loss": 0.3389, + "step": 42878 + }, + { + "epoch": 15.83, + "learning_rate": 1.4174528882867348e-06, + "loss": 0.4094, + "step": 42879 + }, + { + "epoch": 15.83, + "learning_rate": 1.4174159379233897e-06, + "loss": 0.1922, + "step": 42880 + }, + { + "epoch": 15.83, + "learning_rate": 1.4173789875600443e-06, + "loss": 0.432, + "step": 42881 + }, + { + "epoch": 15.84, + "learning_rate": 1.4173420371966991e-06, + "loss": 0.35, + "step": 42882 + }, + { + "epoch": 15.84, + "learning_rate": 1.4173050868333538e-06, + "loss": 0.2087, + "step": 42883 + }, + { + "epoch": 15.84, + "learning_rate": 1.4172681364700086e-06, + "loss": 0.2787, + "step": 42884 + }, + { + "epoch": 15.84, + "learning_rate": 1.4172311861066635e-06, + "loss": 0.2167, + "step": 42885 + }, + { + "epoch": 15.84, + "learning_rate": 1.4171942357433181e-06, + "loss": 0.1955, + "step": 42886 + }, + { + "epoch": 15.84, + "learning_rate": 1.417157285379973e-06, + "loss": 0.3238, + "step": 42887 + }, + { + "epoch": 15.84, + "learning_rate": 1.4171203350166278e-06, + "loss": 0.3163, + "step": 42888 + }, + { + "epoch": 15.84, + "learning_rate": 1.4170833846532824e-06, + "loss": 0.2363, + "step": 42889 + }, + { + "epoch": 15.84, + "learning_rate": 1.4170464342899373e-06, + "loss": 0.218, + "step": 42890 + }, + { + "epoch": 15.84, + "learning_rate": 1.417009483926592e-06, + "loss": 0.2262, + "step": 42891 + }, + { + "epoch": 15.84, + "learning_rate": 1.4169725335632468e-06, + "loss": 0.2894, + "step": 42892 + }, + { + "epoch": 15.84, + "learning_rate": 1.4169355831999014e-06, + "loss": 0.2961, + "step": 42893 + }, + { + "epoch": 15.84, + "learning_rate": 1.4168986328365563e-06, + "loss": 0.2572, + "step": 42894 + }, + { + "epoch": 15.84, + "learning_rate": 1.4168616824732111e-06, + "loss": 0.1176, + "step": 42895 + }, + { + "epoch": 15.84, + "learning_rate": 1.4168247321098657e-06, + "loss": 0.2183, + "step": 42896 + }, + { + "epoch": 15.84, + "learning_rate": 1.4167877817465206e-06, + "loss": 0.1121, + "step": 42897 + }, + { + "epoch": 15.84, + "learning_rate": 1.4167508313831754e-06, + "loss": 0.0984, + "step": 42898 + }, + { + "epoch": 15.84, + "learning_rate": 1.41671388101983e-06, + "loss": 0.2343, + "step": 42899 + }, + { + "epoch": 15.84, + "learning_rate": 1.416676930656485e-06, + "loss": 0.1791, + "step": 42900 + }, + { + "epoch": 15.84, + "learning_rate": 1.4166399802931396e-06, + "loss": 0.2239, + "step": 42901 + }, + { + "epoch": 15.84, + "learning_rate": 1.4166030299297942e-06, + "loss": 0.275, + "step": 42902 + }, + { + "epoch": 15.84, + "learning_rate": 1.416566079566449e-06, + "loss": 0.2444, + "step": 42903 + }, + { + "epoch": 15.84, + "learning_rate": 1.416529129203104e-06, + "loss": 0.3838, + "step": 42904 + }, + { + "epoch": 15.84, + "learning_rate": 1.4164921788397585e-06, + "loss": 0.3033, + "step": 42905 + }, + { + "epoch": 15.84, + "learning_rate": 1.4164552284764134e-06, + "loss": 0.379, + "step": 42906 + }, + { + "epoch": 15.84, + "learning_rate": 1.4164182781130682e-06, + "loss": 0.1644, + "step": 42907 + }, + { + "epoch": 15.84, + "learning_rate": 1.4163813277497229e-06, + "loss": 0.267, + "step": 42908 + }, + { + "epoch": 15.85, + "learning_rate": 1.4163443773863777e-06, + "loss": 0.3268, + "step": 42909 + }, + { + "epoch": 15.85, + "learning_rate": 1.4163074270230326e-06, + "loss": 0.1904, + "step": 42910 + }, + { + "epoch": 15.85, + "learning_rate": 1.4162704766596872e-06, + "loss": 0.2359, + "step": 42911 + }, + { + "epoch": 15.85, + "learning_rate": 1.4162335262963418e-06, + "loss": 0.3342, + "step": 42912 + }, + { + "epoch": 15.85, + "learning_rate": 1.4161965759329967e-06, + "loss": 0.3689, + "step": 42913 + }, + { + "epoch": 15.85, + "learning_rate": 1.4161596255696515e-06, + "loss": 0.2987, + "step": 42914 + }, + { + "epoch": 15.85, + "learning_rate": 1.4161226752063062e-06, + "loss": 0.1717, + "step": 42915 + }, + { + "epoch": 15.85, + "learning_rate": 1.416085724842961e-06, + "loss": 0.3225, + "step": 42916 + }, + { + "epoch": 15.85, + "learning_rate": 1.4160487744796159e-06, + "loss": 0.2282, + "step": 42917 + }, + { + "epoch": 15.85, + "learning_rate": 1.4160118241162705e-06, + "loss": 0.2285, + "step": 42918 + }, + { + "epoch": 15.85, + "learning_rate": 1.4159748737529254e-06, + "loss": 0.2556, + "step": 42919 + }, + { + "epoch": 15.85, + "learning_rate": 1.4159379233895802e-06, + "loss": 0.2013, + "step": 42920 + }, + { + "epoch": 15.85, + "learning_rate": 1.4159009730262346e-06, + "loss": 0.3267, + "step": 42921 + }, + { + "epoch": 15.85, + "learning_rate": 1.4158640226628895e-06, + "loss": 0.3451, + "step": 42922 + }, + { + "epoch": 15.85, + "learning_rate": 1.4158270722995443e-06, + "loss": 0.2828, + "step": 42923 + }, + { + "epoch": 15.85, + "learning_rate": 1.415790121936199e-06, + "loss": 0.2183, + "step": 42924 + }, + { + "epoch": 15.85, + "learning_rate": 1.4157531715728538e-06, + "loss": 0.1244, + "step": 42925 + }, + { + "epoch": 15.85, + "learning_rate": 1.4157162212095087e-06, + "loss": 0.136, + "step": 42926 + }, + { + "epoch": 15.85, + "learning_rate": 1.4156792708461633e-06, + "loss": 0.2645, + "step": 42927 + }, + { + "epoch": 15.85, + "learning_rate": 1.4156423204828181e-06, + "loss": 0.141, + "step": 42928 + }, + { + "epoch": 15.85, + "learning_rate": 1.415605370119473e-06, + "loss": 0.1885, + "step": 42929 + }, + { + "epoch": 15.85, + "learning_rate": 1.4155684197561276e-06, + "loss": 0.1009, + "step": 42930 + }, + { + "epoch": 15.85, + "learning_rate": 1.4155314693927823e-06, + "loss": 0.2547, + "step": 42931 + }, + { + "epoch": 15.85, + "learning_rate": 1.4154945190294371e-06, + "loss": 0.327, + "step": 42932 + }, + { + "epoch": 15.85, + "learning_rate": 1.415457568666092e-06, + "loss": 0.2142, + "step": 42933 + }, + { + "epoch": 15.85, + "learning_rate": 1.4154206183027466e-06, + "loss": 0.2535, + "step": 42934 + }, + { + "epoch": 15.85, + "learning_rate": 1.4153836679394014e-06, + "loss": 0.2623, + "step": 42935 + }, + { + "epoch": 15.86, + "learning_rate": 1.4153467175760563e-06, + "loss": 0.3029, + "step": 42936 + }, + { + "epoch": 15.86, + "learning_rate": 1.415309767212711e-06, + "loss": 0.1495, + "step": 42937 + }, + { + "epoch": 15.86, + "learning_rate": 1.4152728168493658e-06, + "loss": 0.2409, + "step": 42938 + }, + { + "epoch": 15.86, + "learning_rate": 1.4152358664860206e-06, + "loss": 0.1702, + "step": 42939 + }, + { + "epoch": 15.86, + "learning_rate": 1.4151989161226753e-06, + "loss": 0.2547, + "step": 42940 + }, + { + "epoch": 15.86, + "learning_rate": 1.4151619657593299e-06, + "loss": 0.3273, + "step": 42941 + }, + { + "epoch": 15.86, + "learning_rate": 1.4151250153959847e-06, + "loss": 0.3127, + "step": 42942 + }, + { + "epoch": 15.86, + "learning_rate": 1.4150880650326396e-06, + "loss": 0.276, + "step": 42943 + }, + { + "epoch": 15.86, + "learning_rate": 1.4150511146692942e-06, + "loss": 0.2612, + "step": 42944 + }, + { + "epoch": 15.86, + "learning_rate": 1.415014164305949e-06, + "loss": 0.1756, + "step": 42945 + }, + { + "epoch": 15.86, + "learning_rate": 1.414977213942604e-06, + "loss": 0.331, + "step": 42946 + }, + { + "epoch": 15.86, + "learning_rate": 1.4149402635792586e-06, + "loss": 0.1921, + "step": 42947 + }, + { + "epoch": 15.86, + "learning_rate": 1.4149033132159134e-06, + "loss": 0.2727, + "step": 42948 + }, + { + "epoch": 15.86, + "learning_rate": 1.4148663628525683e-06, + "loss": 0.229, + "step": 42949 + }, + { + "epoch": 15.86, + "learning_rate": 1.4148294124892229e-06, + "loss": 0.2228, + "step": 42950 + }, + { + "epoch": 15.86, + "learning_rate": 1.4147924621258775e-06, + "loss": 0.1818, + "step": 42951 + }, + { + "epoch": 15.86, + "learning_rate": 1.4147555117625324e-06, + "loss": 0.3569, + "step": 42952 + }, + { + "epoch": 15.86, + "learning_rate": 1.414718561399187e-06, + "loss": 0.1955, + "step": 42953 + }, + { + "epoch": 15.86, + "learning_rate": 1.4146816110358419e-06, + "loss": 0.3405, + "step": 42954 + }, + { + "epoch": 15.86, + "learning_rate": 1.4146446606724967e-06, + "loss": 0.1114, + "step": 42955 + }, + { + "epoch": 15.86, + "learning_rate": 1.4146077103091513e-06, + "loss": 0.1884, + "step": 42956 + }, + { + "epoch": 15.86, + "learning_rate": 1.4145707599458062e-06, + "loss": 0.3258, + "step": 42957 + }, + { + "epoch": 15.86, + "learning_rate": 1.414533809582461e-06, + "loss": 0.2824, + "step": 42958 + }, + { + "epoch": 15.86, + "learning_rate": 1.4144968592191157e-06, + "loss": 0.3253, + "step": 42959 + }, + { + "epoch": 15.86, + "learning_rate": 1.4144599088557705e-06, + "loss": 0.1747, + "step": 42960 + }, + { + "epoch": 15.86, + "learning_rate": 1.4144229584924252e-06, + "loss": 0.2292, + "step": 42961 + }, + { + "epoch": 15.86, + "learning_rate": 1.41438600812908e-06, + "loss": 0.2095, + "step": 42962 + }, + { + "epoch": 15.87, + "learning_rate": 1.4143490577657346e-06, + "loss": 0.2981, + "step": 42963 + }, + { + "epoch": 15.87, + "learning_rate": 1.4143121074023895e-06, + "loss": 0.2943, + "step": 42964 + }, + { + "epoch": 15.87, + "learning_rate": 1.4142751570390443e-06, + "loss": 0.2203, + "step": 42965 + }, + { + "epoch": 15.87, + "learning_rate": 1.414238206675699e-06, + "loss": 0.2127, + "step": 42966 + }, + { + "epoch": 15.87, + "learning_rate": 1.4142012563123538e-06, + "loss": 0.3137, + "step": 42967 + }, + { + "epoch": 15.87, + "learning_rate": 1.4141643059490087e-06, + "loss": 0.3032, + "step": 42968 + }, + { + "epoch": 15.87, + "learning_rate": 1.4141273555856633e-06, + "loss": 0.3783, + "step": 42969 + }, + { + "epoch": 15.87, + "learning_rate": 1.414090405222318e-06, + "loss": 0.3512, + "step": 42970 + }, + { + "epoch": 15.87, + "learning_rate": 1.4140534548589728e-06, + "loss": 0.2928, + "step": 42971 + }, + { + "epoch": 15.87, + "learning_rate": 1.4140165044956274e-06, + "loss": 0.2058, + "step": 42972 + }, + { + "epoch": 15.87, + "learning_rate": 1.4139795541322823e-06, + "loss": 0.3728, + "step": 42973 + }, + { + "epoch": 15.87, + "learning_rate": 1.4139426037689371e-06, + "loss": 0.4066, + "step": 42974 + }, + { + "epoch": 15.87, + "learning_rate": 1.4139056534055918e-06, + "loss": 0.3552, + "step": 42975 + }, + { + "epoch": 15.87, + "learning_rate": 1.4138687030422466e-06, + "loss": 0.1549, + "step": 42976 + }, + { + "epoch": 15.87, + "learning_rate": 1.4138317526789015e-06, + "loss": 0.1519, + "step": 42977 + }, + { + "epoch": 15.87, + "learning_rate": 1.413794802315556e-06, + "loss": 0.1651, + "step": 42978 + }, + { + "epoch": 15.87, + "learning_rate": 1.413757851952211e-06, + "loss": 0.2146, + "step": 42979 + }, + { + "epoch": 15.87, + "learning_rate": 1.4137209015888656e-06, + "loss": 0.1682, + "step": 42980 + }, + { + "epoch": 15.87, + "learning_rate": 1.4136839512255204e-06, + "loss": 0.2363, + "step": 42981 + }, + { + "epoch": 15.87, + "learning_rate": 1.413647000862175e-06, + "loss": 0.4038, + "step": 42982 + }, + { + "epoch": 15.87, + "learning_rate": 1.41361005049883e-06, + "loss": 0.1664, + "step": 42983 + }, + { + "epoch": 15.87, + "learning_rate": 1.4135731001354848e-06, + "loss": 0.3288, + "step": 42984 + }, + { + "epoch": 15.87, + "learning_rate": 1.4135361497721394e-06, + "loss": 0.3093, + "step": 42985 + }, + { + "epoch": 15.87, + "learning_rate": 1.4134991994087943e-06, + "loss": 0.2835, + "step": 42986 + }, + { + "epoch": 15.87, + "learning_rate": 1.413462249045449e-06, + "loss": 0.3829, + "step": 42987 + }, + { + "epoch": 15.87, + "learning_rate": 1.4134252986821037e-06, + "loss": 0.293, + "step": 42988 + }, + { + "epoch": 15.87, + "learning_rate": 1.4133883483187586e-06, + "loss": 0.2822, + "step": 42989 + }, + { + "epoch": 15.88, + "learning_rate": 1.4133513979554132e-06, + "loss": 0.2437, + "step": 42990 + }, + { + "epoch": 15.88, + "learning_rate": 1.4133144475920679e-06, + "loss": 0.2708, + "step": 42991 + }, + { + "epoch": 15.88, + "learning_rate": 1.4132774972287227e-06, + "loss": 0.2587, + "step": 42992 + }, + { + "epoch": 15.88, + "learning_rate": 1.4132405468653776e-06, + "loss": 0.3112, + "step": 42993 + }, + { + "epoch": 15.88, + "learning_rate": 1.4132035965020322e-06, + "loss": 0.3534, + "step": 42994 + }, + { + "epoch": 15.88, + "learning_rate": 1.413166646138687e-06, + "loss": 0.18, + "step": 42995 + }, + { + "epoch": 15.88, + "learning_rate": 1.4131296957753419e-06, + "loss": 0.2012, + "step": 42996 + }, + { + "epoch": 15.88, + "learning_rate": 1.4130927454119965e-06, + "loss": 0.14, + "step": 42997 + }, + { + "epoch": 15.88, + "learning_rate": 1.4130557950486514e-06, + "loss": 0.1857, + "step": 42998 + }, + { + "epoch": 15.88, + "learning_rate": 1.4130188446853062e-06, + "loss": 0.2416, + "step": 42999 + }, + { + "epoch": 15.88, + "learning_rate": 1.4129818943219609e-06, + "loss": 0.2867, + "step": 43000 + }, + { + "epoch": 15.88, + "learning_rate": 1.4129449439586155e-06, + "loss": 0.3293, + "step": 43001 + }, + { + "epoch": 15.88, + "learning_rate": 1.4129079935952703e-06, + "loss": 0.17, + "step": 43002 + }, + { + "epoch": 15.88, + "learning_rate": 1.4128710432319252e-06, + "loss": 0.2614, + "step": 43003 + }, + { + "epoch": 15.88, + "learning_rate": 1.4128340928685798e-06, + "loss": 0.244, + "step": 43004 + }, + { + "epoch": 15.88, + "learning_rate": 1.4127971425052347e-06, + "loss": 0.2397, + "step": 43005 + }, + { + "epoch": 15.88, + "learning_rate": 1.4127601921418895e-06, + "loss": 0.1942, + "step": 43006 + }, + { + "epoch": 15.88, + "learning_rate": 1.4127232417785442e-06, + "loss": 0.3262, + "step": 43007 + }, + { + "epoch": 15.88, + "learning_rate": 1.412686291415199e-06, + "loss": 0.1932, + "step": 43008 + }, + { + "epoch": 15.88, + "learning_rate": 1.4126493410518539e-06, + "loss": 0.2253, + "step": 43009 + }, + { + "epoch": 15.88, + "learning_rate": 1.4126123906885085e-06, + "loss": 0.2788, + "step": 43010 + }, + { + "epoch": 15.88, + "learning_rate": 1.4125754403251631e-06, + "loss": 0.3452, + "step": 43011 + }, + { + "epoch": 15.88, + "learning_rate": 1.412538489961818e-06, + "loss": 0.1877, + "step": 43012 + }, + { + "epoch": 15.88, + "learning_rate": 1.4125015395984728e-06, + "loss": 0.339, + "step": 43013 + }, + { + "epoch": 15.88, + "learning_rate": 1.4124645892351275e-06, + "loss": 0.2826, + "step": 43014 + }, + { + "epoch": 15.88, + "learning_rate": 1.4124276388717823e-06, + "loss": 0.301, + "step": 43015 + }, + { + "epoch": 15.88, + "learning_rate": 1.4123906885084372e-06, + "loss": 0.1916, + "step": 43016 + }, + { + "epoch": 15.89, + "learning_rate": 1.4123537381450918e-06, + "loss": 0.3458, + "step": 43017 + }, + { + "epoch": 15.89, + "learning_rate": 1.4123167877817466e-06, + "loss": 0.1462, + "step": 43018 + }, + { + "epoch": 15.89, + "learning_rate": 1.4122798374184015e-06, + "loss": 0.2529, + "step": 43019 + }, + { + "epoch": 15.89, + "learning_rate": 1.412242887055056e-06, + "loss": 0.2036, + "step": 43020 + }, + { + "epoch": 15.89, + "learning_rate": 1.4122059366917108e-06, + "loss": 0.139, + "step": 43021 + }, + { + "epoch": 15.89, + "learning_rate": 1.4121689863283656e-06, + "loss": 0.2185, + "step": 43022 + }, + { + "epoch": 15.89, + "learning_rate": 1.4121320359650202e-06, + "loss": 0.2032, + "step": 43023 + }, + { + "epoch": 15.89, + "learning_rate": 1.412095085601675e-06, + "loss": 0.2027, + "step": 43024 + }, + { + "epoch": 15.89, + "learning_rate": 1.41205813523833e-06, + "loss": 0.1616, + "step": 43025 + }, + { + "epoch": 15.89, + "learning_rate": 1.4120211848749846e-06, + "loss": 0.4703, + "step": 43026 + }, + { + "epoch": 15.89, + "learning_rate": 1.4119842345116394e-06, + "loss": 0.2108, + "step": 43027 + }, + { + "epoch": 15.89, + "learning_rate": 1.4119472841482943e-06, + "loss": 0.1618, + "step": 43028 + }, + { + "epoch": 15.89, + "learning_rate": 1.411910333784949e-06, + "loss": 0.2055, + "step": 43029 + }, + { + "epoch": 15.89, + "learning_rate": 1.4118733834216035e-06, + "loss": 0.1973, + "step": 43030 + }, + { + "epoch": 15.89, + "learning_rate": 1.4118364330582584e-06, + "loss": 0.3116, + "step": 43031 + }, + { + "epoch": 15.89, + "learning_rate": 1.4117994826949132e-06, + "loss": 0.3006, + "step": 43032 + }, + { + "epoch": 15.89, + "learning_rate": 1.4117625323315679e-06, + "loss": 0.1837, + "step": 43033 + }, + { + "epoch": 15.89, + "learning_rate": 1.4117255819682227e-06, + "loss": 0.3102, + "step": 43034 + }, + { + "epoch": 15.89, + "learning_rate": 1.4116886316048776e-06, + "loss": 0.2237, + "step": 43035 + }, + { + "epoch": 15.89, + "learning_rate": 1.4116516812415322e-06, + "loss": 0.2484, + "step": 43036 + }, + { + "epoch": 15.89, + "learning_rate": 1.411614730878187e-06, + "loss": 0.1418, + "step": 43037 + }, + { + "epoch": 15.89, + "learning_rate": 1.411577780514842e-06, + "loss": 0.2813, + "step": 43038 + }, + { + "epoch": 15.89, + "learning_rate": 1.4115408301514965e-06, + "loss": 0.2329, + "step": 43039 + }, + { + "epoch": 15.89, + "learning_rate": 1.4115038797881512e-06, + "loss": 0.2868, + "step": 43040 + }, + { + "epoch": 15.89, + "learning_rate": 1.411466929424806e-06, + "loss": 0.2138, + "step": 43041 + }, + { + "epoch": 15.89, + "learning_rate": 1.4114299790614607e-06, + "loss": 0.1534, + "step": 43042 + }, + { + "epoch": 15.89, + "learning_rate": 1.4113930286981155e-06, + "loss": 0.3181, + "step": 43043 + }, + { + "epoch": 15.9, + "learning_rate": 1.4113560783347704e-06, + "loss": 0.369, + "step": 43044 + }, + { + "epoch": 15.9, + "learning_rate": 1.411319127971425e-06, + "loss": 0.3401, + "step": 43045 + }, + { + "epoch": 15.9, + "learning_rate": 1.4112821776080798e-06, + "loss": 0.2394, + "step": 43046 + }, + { + "epoch": 15.9, + "learning_rate": 1.4112452272447347e-06, + "loss": 0.22, + "step": 43047 + }, + { + "epoch": 15.9, + "learning_rate": 1.4112082768813893e-06, + "loss": 0.2863, + "step": 43048 + }, + { + "epoch": 15.9, + "learning_rate": 1.4111713265180442e-06, + "loss": 0.1734, + "step": 43049 + }, + { + "epoch": 15.9, + "learning_rate": 1.4111343761546988e-06, + "loss": 0.2302, + "step": 43050 + }, + { + "epoch": 15.9, + "learning_rate": 1.4110974257913537e-06, + "loss": 0.2214, + "step": 43051 + }, + { + "epoch": 15.9, + "learning_rate": 1.4110604754280083e-06, + "loss": 0.4375, + "step": 43052 + }, + { + "epoch": 15.9, + "learning_rate": 1.4110235250646632e-06, + "loss": 0.2378, + "step": 43053 + }, + { + "epoch": 15.9, + "learning_rate": 1.410986574701318e-06, + "loss": 0.1644, + "step": 43054 + }, + { + "epoch": 15.9, + "learning_rate": 1.4109496243379726e-06, + "loss": 0.1793, + "step": 43055 + }, + { + "epoch": 15.9, + "learning_rate": 1.4109126739746275e-06, + "loss": 0.2611, + "step": 43056 + }, + { + "epoch": 15.9, + "learning_rate": 1.4108757236112823e-06, + "loss": 0.2972, + "step": 43057 + }, + { + "epoch": 15.9, + "learning_rate": 1.410838773247937e-06, + "loss": 0.1843, + "step": 43058 + }, + { + "epoch": 15.9, + "learning_rate": 1.4108018228845918e-06, + "loss": 0.1266, + "step": 43059 + }, + { + "epoch": 15.9, + "learning_rate": 1.4107648725212465e-06, + "loss": 0.2196, + "step": 43060 + }, + { + "epoch": 15.9, + "learning_rate": 1.4107279221579013e-06, + "loss": 0.2057, + "step": 43061 + }, + { + "epoch": 15.9, + "learning_rate": 1.410690971794556e-06, + "loss": 0.2631, + "step": 43062 + }, + { + "epoch": 15.9, + "learning_rate": 1.4106540214312108e-06, + "loss": 0.2388, + "step": 43063 + }, + { + "epoch": 15.9, + "learning_rate": 1.4106170710678654e-06, + "loss": 0.3171, + "step": 43064 + }, + { + "epoch": 15.9, + "learning_rate": 1.4105801207045203e-06, + "loss": 0.2777, + "step": 43065 + }, + { + "epoch": 15.9, + "learning_rate": 1.4105431703411751e-06, + "loss": 0.2039, + "step": 43066 + }, + { + "epoch": 15.9, + "learning_rate": 1.4105062199778298e-06, + "loss": 0.3813, + "step": 43067 + }, + { + "epoch": 15.9, + "learning_rate": 1.4104692696144846e-06, + "loss": 0.2246, + "step": 43068 + }, + { + "epoch": 15.9, + "learning_rate": 1.4104323192511395e-06, + "loss": 0.2612, + "step": 43069 + }, + { + "epoch": 15.9, + "learning_rate": 1.410395368887794e-06, + "loss": 0.3081, + "step": 43070 + }, + { + "epoch": 15.91, + "learning_rate": 1.4103584185244487e-06, + "loss": 0.2637, + "step": 43071 + }, + { + "epoch": 15.91, + "learning_rate": 1.4103214681611036e-06, + "loss": 0.248, + "step": 43072 + }, + { + "epoch": 15.91, + "learning_rate": 1.4102845177977584e-06, + "loss": 0.227, + "step": 43073 + }, + { + "epoch": 15.91, + "learning_rate": 1.410247567434413e-06, + "loss": 0.1667, + "step": 43074 + }, + { + "epoch": 15.91, + "learning_rate": 1.410210617071068e-06, + "loss": 0.3625, + "step": 43075 + }, + { + "epoch": 15.91, + "learning_rate": 1.4101736667077228e-06, + "loss": 0.2677, + "step": 43076 + }, + { + "epoch": 15.91, + "learning_rate": 1.4101367163443774e-06, + "loss": 0.4058, + "step": 43077 + }, + { + "epoch": 15.91, + "learning_rate": 1.4100997659810322e-06, + "loss": 0.185, + "step": 43078 + }, + { + "epoch": 15.91, + "learning_rate": 1.410062815617687e-06, + "loss": 0.238, + "step": 43079 + }, + { + "epoch": 15.91, + "learning_rate": 1.4100258652543417e-06, + "loss": 0.2409, + "step": 43080 + }, + { + "epoch": 15.91, + "learning_rate": 1.4099889148909964e-06, + "loss": 0.2949, + "step": 43081 + }, + { + "epoch": 15.91, + "learning_rate": 1.4099519645276512e-06, + "loss": 0.2226, + "step": 43082 + }, + { + "epoch": 15.91, + "learning_rate": 1.409915014164306e-06, + "loss": 0.1999, + "step": 43083 + }, + { + "epoch": 15.91, + "learning_rate": 1.4098780638009607e-06, + "loss": 0.2606, + "step": 43084 + }, + { + "epoch": 15.91, + "learning_rate": 1.4098411134376155e-06, + "loss": 0.2612, + "step": 43085 + }, + { + "epoch": 15.91, + "learning_rate": 1.4098041630742704e-06, + "loss": 0.2619, + "step": 43086 + }, + { + "epoch": 15.91, + "learning_rate": 1.409767212710925e-06, + "loss": 0.2598, + "step": 43087 + }, + { + "epoch": 15.91, + "learning_rate": 1.4097302623475799e-06, + "loss": 0.2789, + "step": 43088 + }, + { + "epoch": 15.91, + "learning_rate": 1.4096933119842347e-06, + "loss": 0.2063, + "step": 43089 + }, + { + "epoch": 15.91, + "learning_rate": 1.4096563616208891e-06, + "loss": 0.2297, + "step": 43090 + }, + { + "epoch": 15.91, + "learning_rate": 1.409619411257544e-06, + "loss": 0.2496, + "step": 43091 + }, + { + "epoch": 15.91, + "learning_rate": 1.4095824608941988e-06, + "loss": 0.3093, + "step": 43092 + }, + { + "epoch": 15.91, + "learning_rate": 1.4095455105308535e-06, + "loss": 0.1621, + "step": 43093 + }, + { + "epoch": 15.91, + "learning_rate": 1.4095085601675083e-06, + "loss": 0.36, + "step": 43094 + }, + { + "epoch": 15.91, + "learning_rate": 1.4094716098041632e-06, + "loss": 0.4267, + "step": 43095 + }, + { + "epoch": 15.91, + "learning_rate": 1.4094346594408178e-06, + "loss": 0.188, + "step": 43096 + }, + { + "epoch": 15.91, + "learning_rate": 1.4093977090774727e-06, + "loss": 0.174, + "step": 43097 + }, + { + "epoch": 15.92, + "learning_rate": 1.4093607587141275e-06, + "loss": 0.2735, + "step": 43098 + }, + { + "epoch": 15.92, + "learning_rate": 1.4093238083507821e-06, + "loss": 0.2296, + "step": 43099 + }, + { + "epoch": 15.92, + "learning_rate": 1.4092868579874368e-06, + "loss": 0.2289, + "step": 43100 + }, + { + "epoch": 15.92, + "learning_rate": 1.4092499076240916e-06, + "loss": 0.2615, + "step": 43101 + }, + { + "epoch": 15.92, + "learning_rate": 1.4092129572607465e-06, + "loss": 0.4227, + "step": 43102 + }, + { + "epoch": 15.92, + "learning_rate": 1.4091760068974011e-06, + "loss": 0.2287, + "step": 43103 + }, + { + "epoch": 15.92, + "learning_rate": 1.409139056534056e-06, + "loss": 0.3353, + "step": 43104 + }, + { + "epoch": 15.92, + "learning_rate": 1.4091021061707108e-06, + "loss": 0.248, + "step": 43105 + }, + { + "epoch": 15.92, + "learning_rate": 1.4090651558073654e-06, + "loss": 0.2565, + "step": 43106 + }, + { + "epoch": 15.92, + "learning_rate": 1.4090282054440203e-06, + "loss": 0.2646, + "step": 43107 + }, + { + "epoch": 15.92, + "learning_rate": 1.4089912550806751e-06, + "loss": 0.2749, + "step": 43108 + }, + { + "epoch": 15.92, + "learning_rate": 1.4089543047173298e-06, + "loss": 0.1453, + "step": 43109 + }, + { + "epoch": 15.92, + "learning_rate": 1.4089173543539844e-06, + "loss": 0.1887, + "step": 43110 + }, + { + "epoch": 15.92, + "learning_rate": 1.4088804039906393e-06, + "loss": 0.2254, + "step": 43111 + }, + { + "epoch": 15.92, + "learning_rate": 1.408843453627294e-06, + "loss": 0.2353, + "step": 43112 + }, + { + "epoch": 15.92, + "learning_rate": 1.4088065032639487e-06, + "loss": 0.288, + "step": 43113 + }, + { + "epoch": 15.92, + "learning_rate": 1.4087695529006036e-06, + "loss": 0.217, + "step": 43114 + }, + { + "epoch": 15.92, + "learning_rate": 1.4087326025372582e-06, + "loss": 0.2582, + "step": 43115 + }, + { + "epoch": 15.92, + "learning_rate": 1.408695652173913e-06, + "loss": 0.2064, + "step": 43116 + }, + { + "epoch": 15.92, + "learning_rate": 1.408658701810568e-06, + "loss": 0.193, + "step": 43117 + }, + { + "epoch": 15.92, + "learning_rate": 1.4086217514472226e-06, + "loss": 0.1675, + "step": 43118 + }, + { + "epoch": 15.92, + "learning_rate": 1.4085848010838774e-06, + "loss": 0.1864, + "step": 43119 + }, + { + "epoch": 15.92, + "learning_rate": 1.408547850720532e-06, + "loss": 0.2498, + "step": 43120 + }, + { + "epoch": 15.92, + "learning_rate": 1.408510900357187e-06, + "loss": 0.2355, + "step": 43121 + }, + { + "epoch": 15.92, + "learning_rate": 1.4084739499938415e-06, + "loss": 0.2074, + "step": 43122 + }, + { + "epoch": 15.92, + "learning_rate": 1.4084369996304964e-06, + "loss": 0.3459, + "step": 43123 + }, + { + "epoch": 15.92, + "learning_rate": 1.4084000492671512e-06, + "loss": 0.2152, + "step": 43124 + }, + { + "epoch": 15.93, + "learning_rate": 1.4083630989038059e-06, + "loss": 0.2026, + "step": 43125 + }, + { + "epoch": 15.93, + "learning_rate": 1.4083261485404607e-06, + "loss": 0.244, + "step": 43126 + }, + { + "epoch": 15.93, + "learning_rate": 1.4082891981771156e-06, + "loss": 0.207, + "step": 43127 + }, + { + "epoch": 15.93, + "learning_rate": 1.4082522478137702e-06, + "loss": 0.1943, + "step": 43128 + }, + { + "epoch": 15.93, + "learning_rate": 1.408215297450425e-06, + "loss": 0.3376, + "step": 43129 + }, + { + "epoch": 15.93, + "learning_rate": 1.4081783470870797e-06, + "loss": 0.2237, + "step": 43130 + }, + { + "epoch": 15.93, + "learning_rate": 1.4081413967237345e-06, + "loss": 0.2605, + "step": 43131 + }, + { + "epoch": 15.93, + "learning_rate": 1.4081044463603892e-06, + "loss": 0.3022, + "step": 43132 + }, + { + "epoch": 15.93, + "learning_rate": 1.408067495997044e-06, + "loss": 0.3464, + "step": 43133 + }, + { + "epoch": 15.93, + "learning_rate": 1.4080305456336989e-06, + "loss": 0.1483, + "step": 43134 + }, + { + "epoch": 15.93, + "learning_rate": 1.4079935952703535e-06, + "loss": 0.3462, + "step": 43135 + }, + { + "epoch": 15.93, + "learning_rate": 1.4079566449070084e-06, + "loss": 0.2257, + "step": 43136 + }, + { + "epoch": 15.93, + "learning_rate": 1.4079196945436632e-06, + "loss": 0.2363, + "step": 43137 + }, + { + "epoch": 15.93, + "learning_rate": 1.4078827441803178e-06, + "loss": 0.2679, + "step": 43138 + }, + { + "epoch": 15.93, + "learning_rate": 1.4078457938169727e-06, + "loss": 0.2752, + "step": 43139 + }, + { + "epoch": 15.93, + "learning_rate": 1.4078088434536273e-06, + "loss": 0.3866, + "step": 43140 + }, + { + "epoch": 15.93, + "learning_rate": 1.407771893090282e-06, + "loss": 0.2803, + "step": 43141 + }, + { + "epoch": 15.93, + "learning_rate": 1.4077349427269368e-06, + "loss": 0.2509, + "step": 43142 + }, + { + "epoch": 15.93, + "learning_rate": 1.4076979923635917e-06, + "loss": 0.2965, + "step": 43143 + }, + { + "epoch": 15.93, + "learning_rate": 1.4076610420002463e-06, + "loss": 0.3583, + "step": 43144 + }, + { + "epoch": 15.93, + "learning_rate": 1.4076240916369011e-06, + "loss": 0.1976, + "step": 43145 + }, + { + "epoch": 15.93, + "learning_rate": 1.407587141273556e-06, + "loss": 0.2158, + "step": 43146 + }, + { + "epoch": 15.93, + "learning_rate": 1.4075501909102106e-06, + "loss": 0.3248, + "step": 43147 + }, + { + "epoch": 15.93, + "learning_rate": 1.4075132405468655e-06, + "loss": 0.287, + "step": 43148 + }, + { + "epoch": 15.93, + "learning_rate": 1.4074762901835203e-06, + "loss": 0.2181, + "step": 43149 + }, + { + "epoch": 15.93, + "learning_rate": 1.407439339820175e-06, + "loss": 0.2777, + "step": 43150 + }, + { + "epoch": 15.93, + "learning_rate": 1.4074023894568296e-06, + "loss": 0.2775, + "step": 43151 + }, + { + "epoch": 15.94, + "learning_rate": 1.4073654390934844e-06, + "loss": 0.2617, + "step": 43152 + }, + { + "epoch": 15.94, + "learning_rate": 1.4073284887301393e-06, + "loss": 0.1269, + "step": 43153 + }, + { + "epoch": 15.94, + "learning_rate": 1.407291538366794e-06, + "loss": 0.2396, + "step": 43154 + }, + { + "epoch": 15.94, + "learning_rate": 1.4072545880034488e-06, + "loss": 0.3014, + "step": 43155 + }, + { + "epoch": 15.94, + "learning_rate": 1.4072176376401036e-06, + "loss": 0.2267, + "step": 43156 + }, + { + "epoch": 15.94, + "learning_rate": 1.4071806872767583e-06, + "loss": 0.295, + "step": 43157 + }, + { + "epoch": 15.94, + "learning_rate": 1.407143736913413e-06, + "loss": 0.2768, + "step": 43158 + }, + { + "epoch": 15.94, + "learning_rate": 1.407106786550068e-06, + "loss": 0.2032, + "step": 43159 + }, + { + "epoch": 15.94, + "learning_rate": 1.4070698361867224e-06, + "loss": 0.3006, + "step": 43160 + }, + { + "epoch": 15.94, + "learning_rate": 1.4070328858233772e-06, + "loss": 0.2671, + "step": 43161 + }, + { + "epoch": 15.94, + "learning_rate": 1.406995935460032e-06, + "loss": 0.4275, + "step": 43162 + }, + { + "epoch": 15.94, + "learning_rate": 1.4069589850966867e-06, + "loss": 0.327, + "step": 43163 + }, + { + "epoch": 15.94, + "learning_rate": 1.4069220347333416e-06, + "loss": 0.3162, + "step": 43164 + }, + { + "epoch": 15.94, + "learning_rate": 1.4068850843699964e-06, + "loss": 0.1588, + "step": 43165 + }, + { + "epoch": 15.94, + "learning_rate": 1.406848134006651e-06, + "loss": 0.1308, + "step": 43166 + }, + { + "epoch": 15.94, + "learning_rate": 1.4068111836433059e-06, + "loss": 0.1219, + "step": 43167 + }, + { + "epoch": 15.94, + "learning_rate": 1.4067742332799607e-06, + "loss": 0.3942, + "step": 43168 + }, + { + "epoch": 15.94, + "learning_rate": 1.4067372829166154e-06, + "loss": 0.2559, + "step": 43169 + }, + { + "epoch": 15.94, + "learning_rate": 1.40670033255327e-06, + "loss": 0.2235, + "step": 43170 + }, + { + "epoch": 15.94, + "learning_rate": 1.4066633821899249e-06, + "loss": 0.1908, + "step": 43171 + }, + { + "epoch": 15.94, + "learning_rate": 1.4066264318265797e-06, + "loss": 0.2436, + "step": 43172 + }, + { + "epoch": 15.94, + "learning_rate": 1.4065894814632343e-06, + "loss": 0.2951, + "step": 43173 + }, + { + "epoch": 15.94, + "learning_rate": 1.4065525310998892e-06, + "loss": 0.2772, + "step": 43174 + }, + { + "epoch": 15.94, + "learning_rate": 1.406515580736544e-06, + "loss": 0.2582, + "step": 43175 + }, + { + "epoch": 15.94, + "learning_rate": 1.4064786303731987e-06, + "loss": 0.3331, + "step": 43176 + }, + { + "epoch": 15.94, + "learning_rate": 1.4064416800098535e-06, + "loss": 0.2506, + "step": 43177 + }, + { + "epoch": 15.94, + "learning_rate": 1.4064047296465084e-06, + "loss": 0.3183, + "step": 43178 + }, + { + "epoch": 15.94, + "learning_rate": 1.406367779283163e-06, + "loss": 0.2127, + "step": 43179 + }, + { + "epoch": 15.95, + "learning_rate": 1.4063308289198176e-06, + "loss": 0.1968, + "step": 43180 + }, + { + "epoch": 15.95, + "learning_rate": 1.4062938785564725e-06, + "loss": 0.1975, + "step": 43181 + }, + { + "epoch": 15.95, + "learning_rate": 1.4062569281931271e-06, + "loss": 0.1513, + "step": 43182 + }, + { + "epoch": 15.95, + "learning_rate": 1.406219977829782e-06, + "loss": 0.3549, + "step": 43183 + }, + { + "epoch": 15.95, + "learning_rate": 1.4061830274664368e-06, + "loss": 0.3486, + "step": 43184 + }, + { + "epoch": 15.95, + "learning_rate": 1.4061460771030915e-06, + "loss": 0.176, + "step": 43185 + }, + { + "epoch": 15.95, + "learning_rate": 1.4061091267397463e-06, + "loss": 0.228, + "step": 43186 + }, + { + "epoch": 15.95, + "learning_rate": 1.4060721763764012e-06, + "loss": 0.1702, + "step": 43187 + }, + { + "epoch": 15.95, + "learning_rate": 1.4060352260130558e-06, + "loss": 0.2616, + "step": 43188 + }, + { + "epoch": 15.95, + "learning_rate": 1.4059982756497106e-06, + "loss": 0.2297, + "step": 43189 + }, + { + "epoch": 15.95, + "learning_rate": 1.4059613252863653e-06, + "loss": 0.2517, + "step": 43190 + }, + { + "epoch": 15.95, + "learning_rate": 1.4059243749230201e-06, + "loss": 0.2898, + "step": 43191 + }, + { + "epoch": 15.95, + "learning_rate": 1.4058874245596748e-06, + "loss": 0.1247, + "step": 43192 + }, + { + "epoch": 15.95, + "learning_rate": 1.4058504741963296e-06, + "loss": 0.1462, + "step": 43193 + }, + { + "epoch": 15.95, + "learning_rate": 1.4058135238329845e-06, + "loss": 0.2007, + "step": 43194 + }, + { + "epoch": 15.95, + "learning_rate": 1.405776573469639e-06, + "loss": 0.2712, + "step": 43195 + }, + { + "epoch": 15.95, + "learning_rate": 1.405739623106294e-06, + "loss": 0.1934, + "step": 43196 + }, + { + "epoch": 15.95, + "learning_rate": 1.4057026727429488e-06, + "loss": 0.2113, + "step": 43197 + }, + { + "epoch": 15.95, + "learning_rate": 1.4056657223796034e-06, + "loss": 0.1824, + "step": 43198 + }, + { + "epoch": 15.95, + "learning_rate": 1.405628772016258e-06, + "loss": 0.1156, + "step": 43199 + }, + { + "epoch": 15.95, + "learning_rate": 1.405591821652913e-06, + "loss": 0.1993, + "step": 43200 + }, + { + "epoch": 15.95, + "learning_rate": 1.4055548712895678e-06, + "loss": 0.2566, + "step": 43201 + }, + { + "epoch": 15.95, + "learning_rate": 1.4055179209262224e-06, + "loss": 0.224, + "step": 43202 + }, + { + "epoch": 15.95, + "learning_rate": 1.4054809705628773e-06, + "loss": 0.3215, + "step": 43203 + }, + { + "epoch": 15.95, + "learning_rate": 1.405444020199532e-06, + "loss": 0.3608, + "step": 43204 + }, + { + "epoch": 15.95, + "learning_rate": 1.4054070698361867e-06, + "loss": 0.1931, + "step": 43205 + }, + { + "epoch": 15.95, + "learning_rate": 1.4053701194728416e-06, + "loss": 0.3043, + "step": 43206 + }, + { + "epoch": 15.96, + "learning_rate": 1.4053331691094964e-06, + "loss": 0.3524, + "step": 43207 + }, + { + "epoch": 15.96, + "learning_rate": 1.405296218746151e-06, + "loss": 0.1851, + "step": 43208 + }, + { + "epoch": 15.96, + "learning_rate": 1.4052592683828057e-06, + "loss": 0.234, + "step": 43209 + }, + { + "epoch": 15.96, + "learning_rate": 1.4052223180194606e-06, + "loss": 0.1553, + "step": 43210 + }, + { + "epoch": 15.96, + "learning_rate": 1.4051853676561152e-06, + "loss": 0.2266, + "step": 43211 + }, + { + "epoch": 15.96, + "learning_rate": 1.40514841729277e-06, + "loss": 0.2687, + "step": 43212 + }, + { + "epoch": 15.96, + "learning_rate": 1.4051114669294249e-06, + "loss": 0.2803, + "step": 43213 + }, + { + "epoch": 15.96, + "learning_rate": 1.4050745165660795e-06, + "loss": 0.3121, + "step": 43214 + }, + { + "epoch": 15.96, + "learning_rate": 1.4050375662027344e-06, + "loss": 0.3295, + "step": 43215 + }, + { + "epoch": 15.96, + "learning_rate": 1.4050006158393892e-06, + "loss": 0.2892, + "step": 43216 + }, + { + "epoch": 15.96, + "learning_rate": 1.4049636654760439e-06, + "loss": 0.2098, + "step": 43217 + }, + { + "epoch": 15.96, + "learning_rate": 1.4049267151126987e-06, + "loss": 0.2653, + "step": 43218 + }, + { + "epoch": 15.96, + "learning_rate": 1.4048897647493533e-06, + "loss": 0.2574, + "step": 43219 + }, + { + "epoch": 15.96, + "learning_rate": 1.4048528143860082e-06, + "loss": 0.2331, + "step": 43220 + }, + { + "epoch": 15.96, + "learning_rate": 1.4048158640226628e-06, + "loss": 0.1542, + "step": 43221 + }, + { + "epoch": 15.96, + "learning_rate": 1.4047789136593177e-06, + "loss": 0.1667, + "step": 43222 + }, + { + "epoch": 15.96, + "learning_rate": 1.4047419632959725e-06, + "loss": 0.2687, + "step": 43223 + }, + { + "epoch": 15.96, + "learning_rate": 1.4047050129326272e-06, + "loss": 0.2265, + "step": 43224 + }, + { + "epoch": 15.96, + "learning_rate": 1.404668062569282e-06, + "loss": 0.3055, + "step": 43225 + }, + { + "epoch": 15.96, + "learning_rate": 1.4046311122059369e-06, + "loss": 0.2043, + "step": 43226 + }, + { + "epoch": 15.96, + "learning_rate": 1.4045941618425915e-06, + "loss": 0.4395, + "step": 43227 + }, + { + "epoch": 15.96, + "learning_rate": 1.4045572114792463e-06, + "loss": 0.2327, + "step": 43228 + }, + { + "epoch": 15.96, + "learning_rate": 1.404520261115901e-06, + "loss": 0.169, + "step": 43229 + }, + { + "epoch": 15.96, + "learning_rate": 1.4044833107525556e-06, + "loss": 0.2356, + "step": 43230 + }, + { + "epoch": 15.96, + "learning_rate": 1.4044463603892105e-06, + "loss": 0.1719, + "step": 43231 + }, + { + "epoch": 15.96, + "learning_rate": 1.4044094100258653e-06, + "loss": 0.1814, + "step": 43232 + }, + { + "epoch": 15.96, + "learning_rate": 1.40437245966252e-06, + "loss": 0.2729, + "step": 43233 + }, + { + "epoch": 15.97, + "learning_rate": 1.4043355092991748e-06, + "loss": 0.182, + "step": 43234 + }, + { + "epoch": 15.97, + "learning_rate": 1.4042985589358296e-06, + "loss": 0.295, + "step": 43235 + }, + { + "epoch": 15.97, + "learning_rate": 1.4042616085724843e-06, + "loss": 0.2356, + "step": 43236 + }, + { + "epoch": 15.97, + "learning_rate": 1.4042246582091391e-06, + "loss": 0.2126, + "step": 43237 + }, + { + "epoch": 15.97, + "learning_rate": 1.404187707845794e-06, + "loss": 0.1988, + "step": 43238 + }, + { + "epoch": 15.97, + "learning_rate": 1.4041507574824486e-06, + "loss": 0.324, + "step": 43239 + }, + { + "epoch": 15.97, + "learning_rate": 1.4041138071191032e-06, + "loss": 0.305, + "step": 43240 + }, + { + "epoch": 15.97, + "learning_rate": 1.404076856755758e-06, + "loss": 0.2392, + "step": 43241 + }, + { + "epoch": 15.97, + "learning_rate": 1.404039906392413e-06, + "loss": 0.2524, + "step": 43242 + }, + { + "epoch": 15.97, + "learning_rate": 1.4040029560290676e-06, + "loss": 0.3344, + "step": 43243 + }, + { + "epoch": 15.97, + "learning_rate": 1.4039660056657224e-06, + "loss": 0.2345, + "step": 43244 + }, + { + "epoch": 15.97, + "learning_rate": 1.4039290553023773e-06, + "loss": 0.1445, + "step": 43245 + }, + { + "epoch": 15.97, + "learning_rate": 1.403892104939032e-06, + "loss": 0.2729, + "step": 43246 + }, + { + "epoch": 15.97, + "learning_rate": 1.4038551545756868e-06, + "loss": 0.3147, + "step": 43247 + }, + { + "epoch": 15.97, + "learning_rate": 1.4038182042123416e-06, + "loss": 0.342, + "step": 43248 + }, + { + "epoch": 15.97, + "learning_rate": 1.403781253848996e-06, + "loss": 0.2205, + "step": 43249 + }, + { + "epoch": 15.97, + "learning_rate": 1.4037443034856509e-06, + "loss": 0.303, + "step": 43250 + }, + { + "epoch": 15.97, + "learning_rate": 1.4037073531223057e-06, + "loss": 0.2671, + "step": 43251 + }, + { + "epoch": 15.97, + "learning_rate": 1.4036704027589604e-06, + "loss": 0.3359, + "step": 43252 + }, + { + "epoch": 15.97, + "learning_rate": 1.4036334523956152e-06, + "loss": 0.2753, + "step": 43253 + }, + { + "epoch": 15.97, + "learning_rate": 1.40359650203227e-06, + "loss": 0.2288, + "step": 43254 + }, + { + "epoch": 15.97, + "learning_rate": 1.4035595516689247e-06, + "loss": 0.4476, + "step": 43255 + }, + { + "epoch": 15.97, + "learning_rate": 1.4035226013055795e-06, + "loss": 0.1974, + "step": 43256 + }, + { + "epoch": 15.97, + "learning_rate": 1.4034856509422344e-06, + "loss": 0.2416, + "step": 43257 + }, + { + "epoch": 15.97, + "learning_rate": 1.403448700578889e-06, + "loss": 0.1945, + "step": 43258 + }, + { + "epoch": 15.97, + "learning_rate": 1.4034117502155437e-06, + "loss": 0.2789, + "step": 43259 + }, + { + "epoch": 15.97, + "learning_rate": 1.4033747998521985e-06, + "loss": 0.2938, + "step": 43260 + }, + { + "epoch": 15.98, + "learning_rate": 1.4033378494888534e-06, + "loss": 0.1847, + "step": 43261 + }, + { + "epoch": 15.98, + "learning_rate": 1.403300899125508e-06, + "loss": 0.2391, + "step": 43262 + }, + { + "epoch": 15.98, + "learning_rate": 1.4032639487621628e-06, + "loss": 0.2933, + "step": 43263 + }, + { + "epoch": 15.98, + "learning_rate": 1.4032269983988177e-06, + "loss": 0.2396, + "step": 43264 + }, + { + "epoch": 15.98, + "learning_rate": 1.4031900480354723e-06, + "loss": 0.4423, + "step": 43265 + }, + { + "epoch": 15.98, + "learning_rate": 1.4031530976721272e-06, + "loss": 0.2661, + "step": 43266 + }, + { + "epoch": 15.98, + "learning_rate": 1.403116147308782e-06, + "loss": 0.2123, + "step": 43267 + }, + { + "epoch": 15.98, + "learning_rate": 1.4030791969454367e-06, + "loss": 0.3207, + "step": 43268 + }, + { + "epoch": 15.98, + "learning_rate": 1.4030422465820913e-06, + "loss": 0.3697, + "step": 43269 + }, + { + "epoch": 15.98, + "learning_rate": 1.4030052962187462e-06, + "loss": 0.2623, + "step": 43270 + }, + { + "epoch": 15.98, + "learning_rate": 1.402968345855401e-06, + "loss": 0.2737, + "step": 43271 + }, + { + "epoch": 15.98, + "learning_rate": 1.4029313954920556e-06, + "loss": 0.3632, + "step": 43272 + }, + { + "epoch": 15.98, + "learning_rate": 1.4028944451287105e-06, + "loss": 0.1997, + "step": 43273 + }, + { + "epoch": 15.98, + "learning_rate": 1.4028574947653653e-06, + "loss": 0.3145, + "step": 43274 + }, + { + "epoch": 15.98, + "learning_rate": 1.40282054440202e-06, + "loss": 0.354, + "step": 43275 + }, + { + "epoch": 15.98, + "learning_rate": 1.4027835940386748e-06, + "loss": 0.1899, + "step": 43276 + }, + { + "epoch": 15.98, + "learning_rate": 1.4027466436753297e-06, + "loss": 0.3186, + "step": 43277 + }, + { + "epoch": 15.98, + "learning_rate": 1.4027096933119843e-06, + "loss": 0.3019, + "step": 43278 + }, + { + "epoch": 15.98, + "learning_rate": 1.402672742948639e-06, + "loss": 0.305, + "step": 43279 + }, + { + "epoch": 15.98, + "learning_rate": 1.4026357925852938e-06, + "loss": 0.2956, + "step": 43280 + }, + { + "epoch": 15.98, + "learning_rate": 1.4025988422219484e-06, + "loss": 0.2599, + "step": 43281 + }, + { + "epoch": 15.98, + "learning_rate": 1.4025618918586033e-06, + "loss": 0.354, + "step": 43282 + }, + { + "epoch": 15.98, + "learning_rate": 1.4025249414952581e-06, + "loss": 0.29, + "step": 43283 + }, + { + "epoch": 15.98, + "learning_rate": 1.4024879911319128e-06, + "loss": 0.2086, + "step": 43284 + }, + { + "epoch": 15.98, + "learning_rate": 1.4024510407685676e-06, + "loss": 0.2784, + "step": 43285 + }, + { + "epoch": 15.98, + "learning_rate": 1.4024140904052225e-06, + "loss": 0.1974, + "step": 43286 + }, + { + "epoch": 15.98, + "learning_rate": 1.402377140041877e-06, + "loss": 0.1844, + "step": 43287 + }, + { + "epoch": 15.99, + "learning_rate": 1.402340189678532e-06, + "loss": 0.1826, + "step": 43288 + }, + { + "epoch": 15.99, + "learning_rate": 1.4023032393151866e-06, + "loss": 0.1344, + "step": 43289 + }, + { + "epoch": 15.99, + "learning_rate": 1.4022662889518414e-06, + "loss": 0.2398, + "step": 43290 + }, + { + "epoch": 15.99, + "learning_rate": 1.402229338588496e-06, + "loss": 0.2559, + "step": 43291 + }, + { + "epoch": 15.99, + "learning_rate": 1.402192388225151e-06, + "loss": 0.2612, + "step": 43292 + }, + { + "epoch": 15.99, + "learning_rate": 1.4021554378618058e-06, + "loss": 0.3418, + "step": 43293 + }, + { + "epoch": 15.99, + "learning_rate": 1.4021184874984604e-06, + "loss": 0.248, + "step": 43294 + }, + { + "epoch": 15.99, + "learning_rate": 1.4020815371351152e-06, + "loss": 0.1847, + "step": 43295 + }, + { + "epoch": 15.99, + "learning_rate": 1.40204458677177e-06, + "loss": 0.2453, + "step": 43296 + }, + { + "epoch": 15.99, + "learning_rate": 1.4020076364084247e-06, + "loss": 0.1803, + "step": 43297 + }, + { + "epoch": 15.99, + "learning_rate": 1.4019706860450796e-06, + "loss": 0.2465, + "step": 43298 + }, + { + "epoch": 15.99, + "learning_rate": 1.4019337356817342e-06, + "loss": 0.2613, + "step": 43299 + }, + { + "epoch": 15.99, + "learning_rate": 1.4018967853183888e-06, + "loss": 0.2782, + "step": 43300 + }, + { + "epoch": 15.99, + "learning_rate": 1.4018598349550437e-06, + "loss": 0.2078, + "step": 43301 + }, + { + "epoch": 15.99, + "learning_rate": 1.4018228845916985e-06, + "loss": 0.4025, + "step": 43302 + }, + { + "epoch": 15.99, + "learning_rate": 1.4017859342283532e-06, + "loss": 0.456, + "step": 43303 + }, + { + "epoch": 15.99, + "learning_rate": 1.401748983865008e-06, + "loss": 0.3641, + "step": 43304 + }, + { + "epoch": 15.99, + "learning_rate": 1.4017120335016629e-06, + "loss": 0.2639, + "step": 43305 + }, + { + "epoch": 15.99, + "learning_rate": 1.4016750831383175e-06, + "loss": 0.2676, + "step": 43306 + }, + { + "epoch": 15.99, + "learning_rate": 1.4016381327749724e-06, + "loss": 0.2716, + "step": 43307 + }, + { + "epoch": 15.99, + "learning_rate": 1.4016011824116272e-06, + "loss": 0.1703, + "step": 43308 + }, + { + "epoch": 15.99, + "learning_rate": 1.4015642320482818e-06, + "loss": 0.2136, + "step": 43309 + }, + { + "epoch": 15.99, + "learning_rate": 1.4015272816849365e-06, + "loss": 0.1838, + "step": 43310 + }, + { + "epoch": 15.99, + "learning_rate": 1.4014903313215913e-06, + "loss": 0.2728, + "step": 43311 + }, + { + "epoch": 15.99, + "learning_rate": 1.4014533809582462e-06, + "loss": 0.2495, + "step": 43312 + }, + { + "epoch": 15.99, + "learning_rate": 1.4014164305949008e-06, + "loss": 0.1478, + "step": 43313 + }, + { + "epoch": 15.99, + "learning_rate": 1.4013794802315557e-06, + "loss": 0.1881, + "step": 43314 + }, + { + "epoch": 16.0, + "learning_rate": 1.4013425298682105e-06, + "loss": 0.2917, + "step": 43315 + }, + { + "epoch": 16.0, + "learning_rate": 1.4013055795048651e-06, + "loss": 0.3206, + "step": 43316 + }, + { + "epoch": 16.0, + "learning_rate": 1.40126862914152e-06, + "loss": 0.2437, + "step": 43317 + }, + { + "epoch": 16.0, + "learning_rate": 1.4012316787781748e-06, + "loss": 0.2675, + "step": 43318 + }, + { + "epoch": 16.0, + "learning_rate": 1.4011947284148295e-06, + "loss": 0.3253, + "step": 43319 + }, + { + "epoch": 16.0, + "learning_rate": 1.4011577780514841e-06, + "loss": 0.3853, + "step": 43320 + }, + { + "epoch": 16.0, + "learning_rate": 1.401120827688139e-06, + "loss": 0.2302, + "step": 43321 + }, + { + "epoch": 16.0, + "learning_rate": 1.4010838773247938e-06, + "loss": 0.2063, + "step": 43322 + }, + { + "epoch": 16.0, + "learning_rate": 1.4010469269614484e-06, + "loss": 0.24, + "step": 43323 + }, + { + "epoch": 16.0, + "learning_rate": 1.4010099765981033e-06, + "loss": 0.1774, + "step": 43324 + }, + { + "epoch": 16.0, + "learning_rate": 1.4009730262347581e-06, + "loss": 0.3387, + "step": 43325 + }, + { + "epoch": 16.0, + "learning_rate": 1.4009360758714128e-06, + "loss": 0.2004, + "step": 43326 + }, + { + "epoch": 16.0, + "learning_rate": 1.4008991255080676e-06, + "loss": 0.3632, + "step": 43327 + }, + { + "epoch": 16.0, + "learning_rate": 1.4008621751447223e-06, + "loss": 0.2502, + "step": 43328 + }, + { + "epoch": 16.0, + "eval_accuracy": 0.8516876760400794, + "eval_loss": 0.3740586042404175, + "eval_runtime": 25.4075, + "eval_samples_per_second": 1704.77, + "eval_steps_per_second": 26.646, + "step": 43328 + }, + { + "epoch": 16.0, + "learning_rate": 1.400825224781377e-06, + "loss": 0.2139, + "step": 43329 + }, + { + "epoch": 16.0, + "learning_rate": 1.4007882744180317e-06, + "loss": 0.3616, + "step": 43330 + }, + { + "epoch": 16.0, + "learning_rate": 1.4007513240546866e-06, + "loss": 0.2155, + "step": 43331 + }, + { + "epoch": 16.0, + "learning_rate": 1.4007143736913412e-06, + "loss": 0.1906, + "step": 43332 + }, + { + "epoch": 16.0, + "learning_rate": 1.400677423327996e-06, + "loss": 0.316, + "step": 43333 + }, + { + "epoch": 16.0, + "learning_rate": 1.400640472964651e-06, + "loss": 0.1899, + "step": 43334 + }, + { + "epoch": 16.0, + "learning_rate": 1.4006035226013056e-06, + "loss": 0.2505, + "step": 43335 + }, + { + "epoch": 16.0, + "learning_rate": 1.4005665722379604e-06, + "loss": 0.1918, + "step": 43336 + }, + { + "epoch": 16.0, + "learning_rate": 1.4005296218746153e-06, + "loss": 0.2571, + "step": 43337 + }, + { + "epoch": 16.0, + "learning_rate": 1.40049267151127e-06, + "loss": 0.1633, + "step": 43338 + }, + { + "epoch": 16.0, + "learning_rate": 1.4004557211479245e-06, + "loss": 0.2665, + "step": 43339 + }, + { + "epoch": 16.0, + "learning_rate": 1.4004187707845794e-06, + "loss": 0.1618, + "step": 43340 + }, + { + "epoch": 16.0, + "learning_rate": 1.4003818204212342e-06, + "loss": 0.208, + "step": 43341 + }, + { + "epoch": 16.01, + "learning_rate": 1.4003448700578889e-06, + "loss": 0.2306, + "step": 43342 + }, + { + "epoch": 16.01, + "learning_rate": 1.4003079196945437e-06, + "loss": 0.2894, + "step": 43343 + }, + { + "epoch": 16.01, + "learning_rate": 1.4002709693311986e-06, + "loss": 0.2641, + "step": 43344 + }, + { + "epoch": 16.01, + "learning_rate": 1.4002340189678532e-06, + "loss": 0.1852, + "step": 43345 + }, + { + "epoch": 16.01, + "learning_rate": 1.400197068604508e-06, + "loss": 0.276, + "step": 43346 + }, + { + "epoch": 16.01, + "learning_rate": 1.400160118241163e-06, + "loss": 0.4075, + "step": 43347 + }, + { + "epoch": 16.01, + "learning_rate": 1.4001231678778175e-06, + "loss": 0.2124, + "step": 43348 + }, + { + "epoch": 16.01, + "learning_rate": 1.4000862175144722e-06, + "loss": 0.3072, + "step": 43349 + }, + { + "epoch": 16.01, + "learning_rate": 1.400049267151127e-06, + "loss": 0.1532, + "step": 43350 + }, + { + "epoch": 16.01, + "learning_rate": 1.4000123167877817e-06, + "loss": 0.206, + "step": 43351 + }, + { + "epoch": 16.01, + "learning_rate": 1.3999753664244365e-06, + "loss": 0.3231, + "step": 43352 + }, + { + "epoch": 16.01, + "learning_rate": 1.3999384160610914e-06, + "loss": 0.2885, + "step": 43353 + }, + { + "epoch": 16.01, + "learning_rate": 1.399901465697746e-06, + "loss": 0.2274, + "step": 43354 + }, + { + "epoch": 16.01, + "learning_rate": 1.3998645153344008e-06, + "loss": 0.1152, + "step": 43355 + }, + { + "epoch": 16.01, + "learning_rate": 1.3998275649710557e-06, + "loss": 0.3509, + "step": 43356 + }, + { + "epoch": 16.01, + "learning_rate": 1.3997906146077103e-06, + "loss": 0.2698, + "step": 43357 + }, + { + "epoch": 16.01, + "learning_rate": 1.3997536642443652e-06, + "loss": 0.2126, + "step": 43358 + }, + { + "epoch": 16.01, + "learning_rate": 1.3997167138810198e-06, + "loss": 0.2181, + "step": 43359 + }, + { + "epoch": 16.01, + "learning_rate": 1.3996797635176747e-06, + "loss": 0.2372, + "step": 43360 + }, + { + "epoch": 16.01, + "learning_rate": 1.3996428131543293e-06, + "loss": 0.1939, + "step": 43361 + }, + { + "epoch": 16.01, + "learning_rate": 1.3996058627909841e-06, + "loss": 0.2428, + "step": 43362 + }, + { + "epoch": 16.01, + "learning_rate": 1.399568912427639e-06, + "loss": 0.215, + "step": 43363 + }, + { + "epoch": 16.01, + "learning_rate": 1.3995319620642936e-06, + "loss": 0.2033, + "step": 43364 + }, + { + "epoch": 16.01, + "learning_rate": 1.3994950117009485e-06, + "loss": 0.2675, + "step": 43365 + }, + { + "epoch": 16.01, + "learning_rate": 1.3994580613376033e-06, + "loss": 0.3259, + "step": 43366 + }, + { + "epoch": 16.01, + "learning_rate": 1.399421110974258e-06, + "loss": 0.1846, + "step": 43367 + }, + { + "epoch": 16.01, + "learning_rate": 1.3993841606109128e-06, + "loss": 0.1662, + "step": 43368 + }, + { + "epoch": 16.02, + "learning_rate": 1.3993472102475674e-06, + "loss": 0.2565, + "step": 43369 + }, + { + "epoch": 16.02, + "learning_rate": 1.399310259884222e-06, + "loss": 0.2371, + "step": 43370 + }, + { + "epoch": 16.02, + "learning_rate": 1.399273309520877e-06, + "loss": 0.2192, + "step": 43371 + }, + { + "epoch": 16.02, + "learning_rate": 1.3992363591575318e-06, + "loss": 0.2673, + "step": 43372 + }, + { + "epoch": 16.02, + "learning_rate": 1.3991994087941864e-06, + "loss": 0.3304, + "step": 43373 + }, + { + "epoch": 16.02, + "learning_rate": 1.3991624584308413e-06, + "loss": 0.2717, + "step": 43374 + }, + { + "epoch": 16.02, + "learning_rate": 1.399125508067496e-06, + "loss": 0.2536, + "step": 43375 + }, + { + "epoch": 16.02, + "learning_rate": 1.3990885577041507e-06, + "loss": 0.3156, + "step": 43376 + }, + { + "epoch": 16.02, + "learning_rate": 1.3990516073408056e-06, + "loss": 0.1638, + "step": 43377 + }, + { + "epoch": 16.02, + "learning_rate": 1.3990146569774604e-06, + "loss": 0.2192, + "step": 43378 + }, + { + "epoch": 16.02, + "learning_rate": 1.398977706614115e-06, + "loss": 0.2253, + "step": 43379 + }, + { + "epoch": 16.02, + "learning_rate": 1.3989407562507697e-06, + "loss": 0.1857, + "step": 43380 + }, + { + "epoch": 16.02, + "learning_rate": 1.3989038058874246e-06, + "loss": 0.2557, + "step": 43381 + }, + { + "epoch": 16.02, + "learning_rate": 1.3988668555240794e-06, + "loss": 0.1918, + "step": 43382 + }, + { + "epoch": 16.02, + "learning_rate": 1.398829905160734e-06, + "loss": 0.295, + "step": 43383 + }, + { + "epoch": 16.02, + "learning_rate": 1.3987929547973889e-06, + "loss": 0.2261, + "step": 43384 + }, + { + "epoch": 16.02, + "learning_rate": 1.3987560044340437e-06, + "loss": 0.2616, + "step": 43385 + }, + { + "epoch": 16.02, + "learning_rate": 1.3987190540706984e-06, + "loss": 0.3289, + "step": 43386 + }, + { + "epoch": 16.02, + "learning_rate": 1.3986821037073532e-06, + "loss": 0.3097, + "step": 43387 + }, + { + "epoch": 16.02, + "learning_rate": 1.398645153344008e-06, + "loss": 0.2614, + "step": 43388 + }, + { + "epoch": 16.02, + "learning_rate": 1.3986082029806627e-06, + "loss": 0.2986, + "step": 43389 + }, + { + "epoch": 16.02, + "learning_rate": 1.3985712526173173e-06, + "loss": 0.3804, + "step": 43390 + }, + { + "epoch": 16.02, + "learning_rate": 1.3985343022539722e-06, + "loss": 0.2422, + "step": 43391 + }, + { + "epoch": 16.02, + "learning_rate": 1.398497351890627e-06, + "loss": 0.1678, + "step": 43392 + }, + { + "epoch": 16.02, + "learning_rate": 1.3984604015272817e-06, + "loss": 0.1949, + "step": 43393 + }, + { + "epoch": 16.02, + "learning_rate": 1.3984234511639365e-06, + "loss": 0.2197, + "step": 43394 + }, + { + "epoch": 16.02, + "learning_rate": 1.3983865008005914e-06, + "loss": 0.2257, + "step": 43395 + }, + { + "epoch": 16.03, + "learning_rate": 1.398349550437246e-06, + "loss": 0.1857, + "step": 43396 + }, + { + "epoch": 16.03, + "learning_rate": 1.3983126000739009e-06, + "loss": 0.2287, + "step": 43397 + }, + { + "epoch": 16.03, + "learning_rate": 1.3982756497105557e-06, + "loss": 0.2932, + "step": 43398 + }, + { + "epoch": 16.03, + "learning_rate": 1.3982386993472101e-06, + "loss": 0.2193, + "step": 43399 + }, + { + "epoch": 16.03, + "learning_rate": 1.398201748983865e-06, + "loss": 0.2023, + "step": 43400 + }, + { + "epoch": 16.03, + "learning_rate": 1.3981647986205198e-06, + "loss": 0.1838, + "step": 43401 + }, + { + "epoch": 16.03, + "learning_rate": 1.3981278482571745e-06, + "loss": 0.1159, + "step": 43402 + }, + { + "epoch": 16.03, + "learning_rate": 1.3980908978938293e-06, + "loss": 0.0871, + "step": 43403 + }, + { + "epoch": 16.03, + "learning_rate": 1.3980539475304842e-06, + "loss": 0.3016, + "step": 43404 + }, + { + "epoch": 16.03, + "learning_rate": 1.3980169971671388e-06, + "loss": 0.2396, + "step": 43405 + }, + { + "epoch": 16.03, + "learning_rate": 1.3979800468037936e-06, + "loss": 0.2393, + "step": 43406 + }, + { + "epoch": 16.03, + "learning_rate": 1.3979430964404485e-06, + "loss": 0.2576, + "step": 43407 + }, + { + "epoch": 16.03, + "learning_rate": 1.3979061460771031e-06, + "loss": 0.175, + "step": 43408 + }, + { + "epoch": 16.03, + "learning_rate": 1.3978691957137578e-06, + "loss": 0.2226, + "step": 43409 + }, + { + "epoch": 16.03, + "learning_rate": 1.3978322453504126e-06, + "loss": 0.1555, + "step": 43410 + }, + { + "epoch": 16.03, + "learning_rate": 1.3977952949870675e-06, + "loss": 0.3733, + "step": 43411 + }, + { + "epoch": 16.03, + "learning_rate": 1.397758344623722e-06, + "loss": 0.1992, + "step": 43412 + }, + { + "epoch": 16.03, + "learning_rate": 1.397721394260377e-06, + "loss": 0.3083, + "step": 43413 + }, + { + "epoch": 16.03, + "learning_rate": 1.3976844438970318e-06, + "loss": 0.2137, + "step": 43414 + }, + { + "epoch": 16.03, + "learning_rate": 1.3976474935336864e-06, + "loss": 0.1675, + "step": 43415 + }, + { + "epoch": 16.03, + "learning_rate": 1.3976105431703413e-06, + "loss": 0.3172, + "step": 43416 + }, + { + "epoch": 16.03, + "learning_rate": 1.3975735928069961e-06, + "loss": 0.3733, + "step": 43417 + }, + { + "epoch": 16.03, + "learning_rate": 1.3975366424436508e-06, + "loss": 0.3199, + "step": 43418 + }, + { + "epoch": 16.03, + "learning_rate": 1.3974996920803054e-06, + "loss": 0.2004, + "step": 43419 + }, + { + "epoch": 16.03, + "learning_rate": 1.3974627417169602e-06, + "loss": 0.27, + "step": 43420 + }, + { + "epoch": 16.03, + "learning_rate": 1.3974257913536149e-06, + "loss": 0.2208, + "step": 43421 + }, + { + "epoch": 16.03, + "learning_rate": 1.3973888409902697e-06, + "loss": 0.3275, + "step": 43422 + }, + { + "epoch": 16.04, + "learning_rate": 1.3973518906269246e-06, + "loss": 0.1875, + "step": 43423 + }, + { + "epoch": 16.04, + "learning_rate": 1.3973149402635792e-06, + "loss": 0.2909, + "step": 43424 + }, + { + "epoch": 16.04, + "learning_rate": 1.397277989900234e-06, + "loss": 0.1879, + "step": 43425 + }, + { + "epoch": 16.04, + "learning_rate": 1.397241039536889e-06, + "loss": 0.2624, + "step": 43426 + }, + { + "epoch": 16.04, + "learning_rate": 1.3972040891735436e-06, + "loss": 0.2492, + "step": 43427 + }, + { + "epoch": 16.04, + "learning_rate": 1.3971671388101984e-06, + "loss": 0.2946, + "step": 43428 + }, + { + "epoch": 16.04, + "learning_rate": 1.397130188446853e-06, + "loss": 0.2906, + "step": 43429 + }, + { + "epoch": 16.04, + "learning_rate": 1.3970932380835079e-06, + "loss": 0.3496, + "step": 43430 + }, + { + "epoch": 16.04, + "learning_rate": 1.3970562877201625e-06, + "loss": 0.3417, + "step": 43431 + }, + { + "epoch": 16.04, + "learning_rate": 1.3970193373568174e-06, + "loss": 0.2341, + "step": 43432 + }, + { + "epoch": 16.04, + "learning_rate": 1.3969823869934722e-06, + "loss": 0.3162, + "step": 43433 + }, + { + "epoch": 16.04, + "learning_rate": 1.3969454366301269e-06, + "loss": 0.231, + "step": 43434 + }, + { + "epoch": 16.04, + "learning_rate": 1.3969084862667817e-06, + "loss": 0.2423, + "step": 43435 + }, + { + "epoch": 16.04, + "learning_rate": 1.3968715359034366e-06, + "loss": 0.3679, + "step": 43436 + }, + { + "epoch": 16.04, + "learning_rate": 1.3968345855400912e-06, + "loss": 0.3051, + "step": 43437 + }, + { + "epoch": 16.04, + "learning_rate": 1.3967976351767458e-06, + "loss": 0.4103, + "step": 43438 + }, + { + "epoch": 16.04, + "learning_rate": 1.3967606848134007e-06, + "loss": 0.3275, + "step": 43439 + }, + { + "epoch": 16.04, + "learning_rate": 1.3967237344500553e-06, + "loss": 0.302, + "step": 43440 + }, + { + "epoch": 16.04, + "learning_rate": 1.3966867840867102e-06, + "loss": 0.2157, + "step": 43441 + }, + { + "epoch": 16.04, + "learning_rate": 1.396649833723365e-06, + "loss": 0.1883, + "step": 43442 + }, + { + "epoch": 16.04, + "learning_rate": 1.3966128833600196e-06, + "loss": 0.2463, + "step": 43443 + }, + { + "epoch": 16.04, + "learning_rate": 1.3965759329966745e-06, + "loss": 0.2281, + "step": 43444 + }, + { + "epoch": 16.04, + "learning_rate": 1.3965389826333293e-06, + "loss": 0.1875, + "step": 43445 + }, + { + "epoch": 16.04, + "learning_rate": 1.396502032269984e-06, + "loss": 0.1979, + "step": 43446 + }, + { + "epoch": 16.04, + "learning_rate": 1.3964650819066388e-06, + "loss": 0.1511, + "step": 43447 + }, + { + "epoch": 16.04, + "learning_rate": 1.3964281315432935e-06, + "loss": 0.2692, + "step": 43448 + }, + { + "epoch": 16.04, + "learning_rate": 1.3963911811799483e-06, + "loss": 0.2912, + "step": 43449 + }, + { + "epoch": 16.05, + "learning_rate": 1.396354230816603e-06, + "loss": 0.2748, + "step": 43450 + }, + { + "epoch": 16.05, + "learning_rate": 1.3963172804532578e-06, + "loss": 0.2716, + "step": 43451 + }, + { + "epoch": 16.05, + "learning_rate": 1.3962803300899126e-06, + "loss": 0.3139, + "step": 43452 + }, + { + "epoch": 16.05, + "learning_rate": 1.3962433797265673e-06, + "loss": 0.1901, + "step": 43453 + }, + { + "epoch": 16.05, + "learning_rate": 1.3962064293632221e-06, + "loss": 0.2265, + "step": 43454 + }, + { + "epoch": 16.05, + "learning_rate": 1.396169478999877e-06, + "loss": 0.2058, + "step": 43455 + }, + { + "epoch": 16.05, + "learning_rate": 1.3961325286365316e-06, + "loss": 0.2698, + "step": 43456 + }, + { + "epoch": 16.05, + "learning_rate": 1.3960955782731865e-06, + "loss": 0.2412, + "step": 43457 + }, + { + "epoch": 16.05, + "learning_rate": 1.396058627909841e-06, + "loss": 0.2597, + "step": 43458 + }, + { + "epoch": 16.05, + "learning_rate": 1.396021677546496e-06, + "loss": 0.2108, + "step": 43459 + }, + { + "epoch": 16.05, + "learning_rate": 1.3959847271831506e-06, + "loss": 0.2491, + "step": 43460 + }, + { + "epoch": 16.05, + "learning_rate": 1.3959477768198054e-06, + "loss": 0.2376, + "step": 43461 + }, + { + "epoch": 16.05, + "learning_rate": 1.3959108264564603e-06, + "loss": 0.3004, + "step": 43462 + }, + { + "epoch": 16.05, + "learning_rate": 1.395873876093115e-06, + "loss": 0.2326, + "step": 43463 + }, + { + "epoch": 16.05, + "learning_rate": 1.3958369257297698e-06, + "loss": 0.3101, + "step": 43464 + }, + { + "epoch": 16.05, + "learning_rate": 1.3957999753664246e-06, + "loss": 0.2677, + "step": 43465 + }, + { + "epoch": 16.05, + "learning_rate": 1.3957630250030792e-06, + "loss": 0.3683, + "step": 43466 + }, + { + "epoch": 16.05, + "learning_rate": 1.395726074639734e-06, + "loss": 0.2793, + "step": 43467 + }, + { + "epoch": 16.05, + "learning_rate": 1.3956891242763887e-06, + "loss": 0.2077, + "step": 43468 + }, + { + "epoch": 16.05, + "learning_rate": 1.3956521739130434e-06, + "loss": 0.1953, + "step": 43469 + }, + { + "epoch": 16.05, + "learning_rate": 1.3956152235496982e-06, + "loss": 0.2951, + "step": 43470 + }, + { + "epoch": 16.05, + "learning_rate": 1.395578273186353e-06, + "loss": 0.2601, + "step": 43471 + }, + { + "epoch": 16.05, + "learning_rate": 1.3955413228230077e-06, + "loss": 0.2413, + "step": 43472 + }, + { + "epoch": 16.05, + "learning_rate": 1.3955043724596625e-06, + "loss": 0.1692, + "step": 43473 + }, + { + "epoch": 16.05, + "learning_rate": 1.3954674220963174e-06, + "loss": 0.2762, + "step": 43474 + }, + { + "epoch": 16.05, + "learning_rate": 1.395430471732972e-06, + "loss": 0.2278, + "step": 43475 + }, + { + "epoch": 16.05, + "learning_rate": 1.3953935213696269e-06, + "loss": 0.3313, + "step": 43476 + }, + { + "epoch": 16.06, + "learning_rate": 1.3953565710062817e-06, + "loss": 0.193, + "step": 43477 + }, + { + "epoch": 16.06, + "learning_rate": 1.3953196206429364e-06, + "loss": 0.2096, + "step": 43478 + }, + { + "epoch": 16.06, + "learning_rate": 1.395282670279591e-06, + "loss": 0.3203, + "step": 43479 + }, + { + "epoch": 16.06, + "learning_rate": 1.3952457199162458e-06, + "loss": 0.1924, + "step": 43480 + }, + { + "epoch": 16.06, + "learning_rate": 1.3952087695529007e-06, + "loss": 0.2258, + "step": 43481 + }, + { + "epoch": 16.06, + "learning_rate": 1.3951718191895553e-06, + "loss": 0.3497, + "step": 43482 + }, + { + "epoch": 16.06, + "learning_rate": 1.3951348688262102e-06, + "loss": 0.2826, + "step": 43483 + }, + { + "epoch": 16.06, + "learning_rate": 1.395097918462865e-06, + "loss": 0.1809, + "step": 43484 + }, + { + "epoch": 16.06, + "learning_rate": 1.3950609680995197e-06, + "loss": 0.2439, + "step": 43485 + }, + { + "epoch": 16.06, + "learning_rate": 1.3950240177361745e-06, + "loss": 0.2384, + "step": 43486 + }, + { + "epoch": 16.06, + "learning_rate": 1.3949870673728294e-06, + "loss": 0.2241, + "step": 43487 + }, + { + "epoch": 16.06, + "learning_rate": 1.3949501170094838e-06, + "loss": 0.2501, + "step": 43488 + }, + { + "epoch": 16.06, + "learning_rate": 1.3949131666461386e-06, + "loss": 0.1903, + "step": 43489 + }, + { + "epoch": 16.06, + "learning_rate": 1.3948762162827935e-06, + "loss": 0.1822, + "step": 43490 + }, + { + "epoch": 16.06, + "learning_rate": 1.3948392659194481e-06, + "loss": 0.3238, + "step": 43491 + }, + { + "epoch": 16.06, + "learning_rate": 1.394802315556103e-06, + "loss": 0.1246, + "step": 43492 + }, + { + "epoch": 16.06, + "learning_rate": 1.3947653651927578e-06, + "loss": 0.1109, + "step": 43493 + }, + { + "epoch": 16.06, + "learning_rate": 1.3947284148294125e-06, + "loss": 0.1819, + "step": 43494 + }, + { + "epoch": 16.06, + "learning_rate": 1.3946914644660673e-06, + "loss": 0.1795, + "step": 43495 + }, + { + "epoch": 16.06, + "learning_rate": 1.3946545141027221e-06, + "loss": 0.1642, + "step": 43496 + }, + { + "epoch": 16.06, + "learning_rate": 1.3946175637393768e-06, + "loss": 0.2078, + "step": 43497 + }, + { + "epoch": 16.06, + "learning_rate": 1.3945806133760314e-06, + "loss": 0.2341, + "step": 43498 + }, + { + "epoch": 16.06, + "learning_rate": 1.3945436630126863e-06, + "loss": 0.2335, + "step": 43499 + }, + { + "epoch": 16.06, + "learning_rate": 1.3945067126493411e-06, + "loss": 0.2074, + "step": 43500 + }, + { + "epoch": 16.06, + "learning_rate": 1.3944697622859958e-06, + "loss": 0.3432, + "step": 43501 + }, + { + "epoch": 16.06, + "learning_rate": 1.3944328119226506e-06, + "loss": 0.1065, + "step": 43502 + }, + { + "epoch": 16.06, + "learning_rate": 1.3943958615593055e-06, + "loss": 0.2206, + "step": 43503 + }, + { + "epoch": 16.06, + "learning_rate": 1.39435891119596e-06, + "loss": 0.2352, + "step": 43504 + }, + { + "epoch": 16.07, + "learning_rate": 1.394321960832615e-06, + "loss": 0.2764, + "step": 43505 + }, + { + "epoch": 16.07, + "learning_rate": 1.3942850104692698e-06, + "loss": 0.3154, + "step": 43506 + }, + { + "epoch": 16.07, + "learning_rate": 1.3942480601059244e-06, + "loss": 0.3993, + "step": 43507 + }, + { + "epoch": 16.07, + "learning_rate": 1.394211109742579e-06, + "loss": 0.1756, + "step": 43508 + }, + { + "epoch": 16.07, + "learning_rate": 1.394174159379234e-06, + "loss": 0.1475, + "step": 43509 + }, + { + "epoch": 16.07, + "learning_rate": 1.3941372090158888e-06, + "loss": 0.2576, + "step": 43510 + }, + { + "epoch": 16.07, + "learning_rate": 1.3941002586525434e-06, + "loss": 0.2503, + "step": 43511 + }, + { + "epoch": 16.07, + "learning_rate": 1.3940633082891982e-06, + "loss": 0.222, + "step": 43512 + }, + { + "epoch": 16.07, + "learning_rate": 1.3940263579258529e-06, + "loss": 0.2831, + "step": 43513 + }, + { + "epoch": 16.07, + "learning_rate": 1.3939894075625077e-06, + "loss": 0.1523, + "step": 43514 + }, + { + "epoch": 16.07, + "learning_rate": 1.3939524571991626e-06, + "loss": 0.2867, + "step": 43515 + }, + { + "epoch": 16.07, + "learning_rate": 1.3939155068358172e-06, + "loss": 0.3635, + "step": 43516 + }, + { + "epoch": 16.07, + "learning_rate": 1.393878556472472e-06, + "loss": 0.2872, + "step": 43517 + }, + { + "epoch": 16.07, + "learning_rate": 1.3938416061091267e-06, + "loss": 0.2938, + "step": 43518 + }, + { + "epoch": 16.07, + "learning_rate": 1.3938046557457815e-06, + "loss": 0.3056, + "step": 43519 + }, + { + "epoch": 16.07, + "learning_rate": 1.3937677053824362e-06, + "loss": 0.2697, + "step": 43520 + }, + { + "epoch": 16.07, + "learning_rate": 1.393730755019091e-06, + "loss": 0.2535, + "step": 43521 + }, + { + "epoch": 16.07, + "learning_rate": 1.3936938046557459e-06, + "loss": 0.1902, + "step": 43522 + }, + { + "epoch": 16.07, + "learning_rate": 1.3936568542924005e-06, + "loss": 0.2526, + "step": 43523 + }, + { + "epoch": 16.07, + "learning_rate": 1.3936199039290554e-06, + "loss": 0.2704, + "step": 43524 + }, + { + "epoch": 16.07, + "learning_rate": 1.3935829535657102e-06, + "loss": 0.1952, + "step": 43525 + }, + { + "epoch": 16.07, + "learning_rate": 1.3935460032023648e-06, + "loss": 0.1389, + "step": 43526 + }, + { + "epoch": 16.07, + "learning_rate": 1.3935090528390197e-06, + "loss": 0.2377, + "step": 43527 + }, + { + "epoch": 16.07, + "learning_rate": 1.3934721024756743e-06, + "loss": 0.2253, + "step": 43528 + }, + { + "epoch": 16.07, + "learning_rate": 1.3934351521123292e-06, + "loss": 0.2071, + "step": 43529 + }, + { + "epoch": 16.07, + "learning_rate": 1.3933982017489838e-06, + "loss": 0.2572, + "step": 43530 + }, + { + "epoch": 16.07, + "learning_rate": 1.3933612513856387e-06, + "loss": 0.2013, + "step": 43531 + }, + { + "epoch": 16.08, + "learning_rate": 1.3933243010222935e-06, + "loss": 0.155, + "step": 43532 + }, + { + "epoch": 16.08, + "learning_rate": 1.3932873506589481e-06, + "loss": 0.1387, + "step": 43533 + }, + { + "epoch": 16.08, + "learning_rate": 1.393250400295603e-06, + "loss": 0.1995, + "step": 43534 + }, + { + "epoch": 16.08, + "learning_rate": 1.3932134499322578e-06, + "loss": 0.2352, + "step": 43535 + }, + { + "epoch": 16.08, + "learning_rate": 1.3931764995689125e-06, + "loss": 0.326, + "step": 43536 + }, + { + "epoch": 16.08, + "learning_rate": 1.3931395492055673e-06, + "loss": 0.1892, + "step": 43537 + }, + { + "epoch": 16.08, + "learning_rate": 1.393102598842222e-06, + "loss": 0.2939, + "step": 43538 + }, + { + "epoch": 16.08, + "learning_rate": 1.3930656484788766e-06, + "loss": 0.4163, + "step": 43539 + }, + { + "epoch": 16.08, + "learning_rate": 1.3930286981155314e-06, + "loss": 0.1988, + "step": 43540 + }, + { + "epoch": 16.08, + "learning_rate": 1.3929917477521863e-06, + "loss": 0.2166, + "step": 43541 + }, + { + "epoch": 16.08, + "learning_rate": 1.392954797388841e-06, + "loss": 0.2466, + "step": 43542 + }, + { + "epoch": 16.08, + "learning_rate": 1.3929178470254958e-06, + "loss": 0.1972, + "step": 43543 + }, + { + "epoch": 16.08, + "learning_rate": 1.3928808966621506e-06, + "loss": 0.3062, + "step": 43544 + }, + { + "epoch": 16.08, + "learning_rate": 1.3928439462988053e-06, + "loss": 0.303, + "step": 43545 + }, + { + "epoch": 16.08, + "learning_rate": 1.3928069959354601e-06, + "loss": 0.2413, + "step": 43546 + }, + { + "epoch": 16.08, + "learning_rate": 1.392770045572115e-06, + "loss": 0.1549, + "step": 43547 + }, + { + "epoch": 16.08, + "learning_rate": 1.3927330952087696e-06, + "loss": 0.2518, + "step": 43548 + }, + { + "epoch": 16.08, + "learning_rate": 1.3926961448454242e-06, + "loss": 0.171, + "step": 43549 + }, + { + "epoch": 16.08, + "learning_rate": 1.392659194482079e-06, + "loss": 0.2484, + "step": 43550 + }, + { + "epoch": 16.08, + "learning_rate": 1.392622244118734e-06, + "loss": 0.1793, + "step": 43551 + }, + { + "epoch": 16.08, + "learning_rate": 1.3925852937553886e-06, + "loss": 0.1308, + "step": 43552 + }, + { + "epoch": 16.08, + "learning_rate": 1.3925483433920434e-06, + "loss": 0.2446, + "step": 43553 + }, + { + "epoch": 16.08, + "learning_rate": 1.3925113930286983e-06, + "loss": 0.2756, + "step": 43554 + }, + { + "epoch": 16.08, + "learning_rate": 1.392474442665353e-06, + "loss": 0.2056, + "step": 43555 + }, + { + "epoch": 16.08, + "learning_rate": 1.3924374923020077e-06, + "loss": 0.2209, + "step": 43556 + }, + { + "epoch": 16.08, + "learning_rate": 1.3924005419386626e-06, + "loss": 0.2112, + "step": 43557 + }, + { + "epoch": 16.08, + "learning_rate": 1.392363591575317e-06, + "loss": 0.22, + "step": 43558 + }, + { + "epoch": 16.09, + "learning_rate": 1.3923266412119719e-06, + "loss": 0.3278, + "step": 43559 + }, + { + "epoch": 16.09, + "learning_rate": 1.3922896908486267e-06, + "loss": 0.167, + "step": 43560 + }, + { + "epoch": 16.09, + "learning_rate": 1.3922527404852814e-06, + "loss": 0.3222, + "step": 43561 + }, + { + "epoch": 16.09, + "learning_rate": 1.3922157901219362e-06, + "loss": 0.4129, + "step": 43562 + }, + { + "epoch": 16.09, + "learning_rate": 1.392178839758591e-06, + "loss": 0.3188, + "step": 43563 + }, + { + "epoch": 16.09, + "learning_rate": 1.3921418893952457e-06, + "loss": 0.2313, + "step": 43564 + }, + { + "epoch": 16.09, + "learning_rate": 1.3921049390319005e-06, + "loss": 0.207, + "step": 43565 + }, + { + "epoch": 16.09, + "learning_rate": 1.3920679886685554e-06, + "loss": 0.2149, + "step": 43566 + }, + { + "epoch": 16.09, + "learning_rate": 1.39203103830521e-06, + "loss": 0.2496, + "step": 43567 + }, + { + "epoch": 16.09, + "learning_rate": 1.3919940879418647e-06, + "loss": 0.2606, + "step": 43568 + }, + { + "epoch": 16.09, + "learning_rate": 1.3919571375785195e-06, + "loss": 0.2889, + "step": 43569 + }, + { + "epoch": 16.09, + "learning_rate": 1.3919201872151743e-06, + "loss": 0.3393, + "step": 43570 + }, + { + "epoch": 16.09, + "learning_rate": 1.391883236851829e-06, + "loss": 0.1547, + "step": 43571 + }, + { + "epoch": 16.09, + "learning_rate": 1.3918462864884838e-06, + "loss": 0.2022, + "step": 43572 + }, + { + "epoch": 16.09, + "learning_rate": 1.3918093361251387e-06, + "loss": 0.3408, + "step": 43573 + }, + { + "epoch": 16.09, + "learning_rate": 1.3917723857617933e-06, + "loss": 0.1273, + "step": 43574 + }, + { + "epoch": 16.09, + "learning_rate": 1.3917354353984482e-06, + "loss": 0.2425, + "step": 43575 + }, + { + "epoch": 16.09, + "learning_rate": 1.391698485035103e-06, + "loss": 0.3528, + "step": 43576 + }, + { + "epoch": 16.09, + "learning_rate": 1.3916615346717577e-06, + "loss": 0.2593, + "step": 43577 + }, + { + "epoch": 16.09, + "learning_rate": 1.3916245843084123e-06, + "loss": 0.2484, + "step": 43578 + }, + { + "epoch": 16.09, + "learning_rate": 1.3915876339450671e-06, + "loss": 0.3063, + "step": 43579 + }, + { + "epoch": 16.09, + "learning_rate": 1.391550683581722e-06, + "loss": 0.2585, + "step": 43580 + }, + { + "epoch": 16.09, + "learning_rate": 1.3915137332183766e-06, + "loss": 0.2994, + "step": 43581 + }, + { + "epoch": 16.09, + "learning_rate": 1.3914767828550315e-06, + "loss": 0.3219, + "step": 43582 + }, + { + "epoch": 16.09, + "learning_rate": 1.3914398324916863e-06, + "loss": 0.2066, + "step": 43583 + }, + { + "epoch": 16.09, + "learning_rate": 1.391402882128341e-06, + "loss": 0.1099, + "step": 43584 + }, + { + "epoch": 16.09, + "learning_rate": 1.3913659317649958e-06, + "loss": 0.3076, + "step": 43585 + }, + { + "epoch": 16.1, + "learning_rate": 1.3913289814016507e-06, + "loss": 0.2217, + "step": 43586 + }, + { + "epoch": 16.1, + "learning_rate": 1.3912920310383053e-06, + "loss": 0.2172, + "step": 43587 + }, + { + "epoch": 16.1, + "learning_rate": 1.39125508067496e-06, + "loss": 0.206, + "step": 43588 + }, + { + "epoch": 16.1, + "learning_rate": 1.3912181303116148e-06, + "loss": 0.2626, + "step": 43589 + }, + { + "epoch": 16.1, + "learning_rate": 1.3911811799482694e-06, + "loss": 0.1138, + "step": 43590 + }, + { + "epoch": 16.1, + "learning_rate": 1.3911442295849243e-06, + "loss": 0.2934, + "step": 43591 + }, + { + "epoch": 16.1, + "learning_rate": 1.391107279221579e-06, + "loss": 0.1446, + "step": 43592 + }, + { + "epoch": 16.1, + "learning_rate": 1.3910703288582337e-06, + "loss": 0.2493, + "step": 43593 + }, + { + "epoch": 16.1, + "learning_rate": 1.3910333784948886e-06, + "loss": 0.1669, + "step": 43594 + }, + { + "epoch": 16.1, + "learning_rate": 1.3909964281315434e-06, + "loss": 0.1222, + "step": 43595 + }, + { + "epoch": 16.1, + "learning_rate": 1.390959477768198e-06, + "loss": 0.1731, + "step": 43596 + }, + { + "epoch": 16.1, + "learning_rate": 1.390922527404853e-06, + "loss": 0.1686, + "step": 43597 + }, + { + "epoch": 16.1, + "learning_rate": 1.3908855770415076e-06, + "loss": 0.2483, + "step": 43598 + }, + { + "epoch": 16.1, + "learning_rate": 1.3908486266781624e-06, + "loss": 0.2877, + "step": 43599 + }, + { + "epoch": 16.1, + "learning_rate": 1.390811676314817e-06, + "loss": 0.2339, + "step": 43600 + }, + { + "epoch": 16.1, + "learning_rate": 1.3907747259514719e-06, + "loss": 0.0987, + "step": 43601 + }, + { + "epoch": 16.1, + "learning_rate": 1.3907377755881267e-06, + "loss": 0.3257, + "step": 43602 + }, + { + "epoch": 16.1, + "learning_rate": 1.3907008252247814e-06, + "loss": 0.2128, + "step": 43603 + }, + { + "epoch": 16.1, + "learning_rate": 1.3906638748614362e-06, + "loss": 0.2105, + "step": 43604 + }, + { + "epoch": 16.1, + "learning_rate": 1.390626924498091e-06, + "loss": 0.2317, + "step": 43605 + }, + { + "epoch": 16.1, + "learning_rate": 1.3905899741347457e-06, + "loss": 0.2599, + "step": 43606 + }, + { + "epoch": 16.1, + "learning_rate": 1.3905530237714006e-06, + "loss": 0.3182, + "step": 43607 + }, + { + "epoch": 16.1, + "learning_rate": 1.3905160734080552e-06, + "loss": 0.2566, + "step": 43608 + }, + { + "epoch": 16.1, + "learning_rate": 1.3904791230447098e-06, + "loss": 0.217, + "step": 43609 + }, + { + "epoch": 16.1, + "learning_rate": 1.3904421726813647e-06, + "loss": 0.2399, + "step": 43610 + }, + { + "epoch": 16.1, + "learning_rate": 1.3904052223180195e-06, + "loss": 0.2569, + "step": 43611 + }, + { + "epoch": 16.1, + "learning_rate": 1.3903682719546742e-06, + "loss": 0.1296, + "step": 43612 + }, + { + "epoch": 16.11, + "learning_rate": 1.390331321591329e-06, + "loss": 0.2912, + "step": 43613 + }, + { + "epoch": 16.11, + "learning_rate": 1.3902943712279839e-06, + "loss": 0.2995, + "step": 43614 + }, + { + "epoch": 16.11, + "learning_rate": 1.3902574208646385e-06, + "loss": 0.1841, + "step": 43615 + }, + { + "epoch": 16.11, + "learning_rate": 1.3902204705012933e-06, + "loss": 0.1755, + "step": 43616 + }, + { + "epoch": 16.11, + "learning_rate": 1.3901835201379482e-06, + "loss": 0.2435, + "step": 43617 + }, + { + "epoch": 16.11, + "learning_rate": 1.3901465697746028e-06, + "loss": 0.1582, + "step": 43618 + }, + { + "epoch": 16.11, + "learning_rate": 1.3901096194112575e-06, + "loss": 0.3002, + "step": 43619 + }, + { + "epoch": 16.11, + "learning_rate": 1.3900726690479123e-06, + "loss": 0.2034, + "step": 43620 + }, + { + "epoch": 16.11, + "learning_rate": 1.3900357186845672e-06, + "loss": 0.2776, + "step": 43621 + }, + { + "epoch": 16.11, + "learning_rate": 1.3899987683212218e-06, + "loss": 0.1785, + "step": 43622 + }, + { + "epoch": 16.11, + "learning_rate": 1.3899618179578766e-06, + "loss": 0.2025, + "step": 43623 + }, + { + "epoch": 16.11, + "learning_rate": 1.3899248675945315e-06, + "loss": 0.2906, + "step": 43624 + }, + { + "epoch": 16.11, + "learning_rate": 1.3898879172311861e-06, + "loss": 0.2336, + "step": 43625 + }, + { + "epoch": 16.11, + "learning_rate": 1.389850966867841e-06, + "loss": 0.1962, + "step": 43626 + }, + { + "epoch": 16.11, + "learning_rate": 1.3898140165044958e-06, + "loss": 0.349, + "step": 43627 + }, + { + "epoch": 16.11, + "learning_rate": 1.3897770661411503e-06, + "loss": 0.244, + "step": 43628 + }, + { + "epoch": 16.11, + "learning_rate": 1.389740115777805e-06, + "loss": 0.2129, + "step": 43629 + }, + { + "epoch": 16.11, + "learning_rate": 1.38970316541446e-06, + "loss": 0.2887, + "step": 43630 + }, + { + "epoch": 16.11, + "learning_rate": 1.3896662150511146e-06, + "loss": 0.2944, + "step": 43631 + }, + { + "epoch": 16.11, + "learning_rate": 1.3896292646877694e-06, + "loss": 0.2172, + "step": 43632 + }, + { + "epoch": 16.11, + "learning_rate": 1.3895923143244243e-06, + "loss": 0.29, + "step": 43633 + }, + { + "epoch": 16.11, + "learning_rate": 1.389555363961079e-06, + "loss": 0.3332, + "step": 43634 + }, + { + "epoch": 16.11, + "learning_rate": 1.3895184135977338e-06, + "loss": 0.2942, + "step": 43635 + }, + { + "epoch": 16.11, + "learning_rate": 1.3894814632343886e-06, + "loss": 0.2217, + "step": 43636 + }, + { + "epoch": 16.11, + "learning_rate": 1.3894445128710432e-06, + "loss": 0.2412, + "step": 43637 + }, + { + "epoch": 16.11, + "learning_rate": 1.3894075625076979e-06, + "loss": 0.1896, + "step": 43638 + }, + { + "epoch": 16.11, + "learning_rate": 1.3893706121443527e-06, + "loss": 0.258, + "step": 43639 + }, + { + "epoch": 16.12, + "learning_rate": 1.3893336617810076e-06, + "loss": 0.3034, + "step": 43640 + }, + { + "epoch": 16.12, + "learning_rate": 1.3892967114176622e-06, + "loss": 0.2977, + "step": 43641 + }, + { + "epoch": 16.12, + "learning_rate": 1.389259761054317e-06, + "loss": 0.2434, + "step": 43642 + }, + { + "epoch": 16.12, + "learning_rate": 1.389222810690972e-06, + "loss": 0.1369, + "step": 43643 + }, + { + "epoch": 16.12, + "learning_rate": 1.3891858603276266e-06, + "loss": 0.1373, + "step": 43644 + }, + { + "epoch": 16.12, + "learning_rate": 1.3891489099642814e-06, + "loss": 0.2842, + "step": 43645 + }, + { + "epoch": 16.12, + "learning_rate": 1.3891119596009362e-06, + "loss": 0.2171, + "step": 43646 + }, + { + "epoch": 16.12, + "learning_rate": 1.3890750092375909e-06, + "loss": 0.2154, + "step": 43647 + }, + { + "epoch": 16.12, + "learning_rate": 1.3890380588742455e-06, + "loss": 0.2901, + "step": 43648 + }, + { + "epoch": 16.12, + "learning_rate": 1.3890011085109004e-06, + "loss": 0.3206, + "step": 43649 + }, + { + "epoch": 16.12, + "learning_rate": 1.3889641581475552e-06, + "loss": 0.23, + "step": 43650 + }, + { + "epoch": 16.12, + "learning_rate": 1.3889272077842099e-06, + "loss": 0.1815, + "step": 43651 + }, + { + "epoch": 16.12, + "learning_rate": 1.3888902574208647e-06, + "loss": 0.2247, + "step": 43652 + }, + { + "epoch": 16.12, + "learning_rate": 1.3888533070575196e-06, + "loss": 0.1571, + "step": 43653 + }, + { + "epoch": 16.12, + "learning_rate": 1.3888163566941742e-06, + "loss": 0.2272, + "step": 43654 + }, + { + "epoch": 16.12, + "learning_rate": 1.388779406330829e-06, + "loss": 0.1368, + "step": 43655 + }, + { + "epoch": 16.12, + "learning_rate": 1.3887424559674839e-06, + "loss": 0.2037, + "step": 43656 + }, + { + "epoch": 16.12, + "learning_rate": 1.3887055056041385e-06, + "loss": 0.2855, + "step": 43657 + }, + { + "epoch": 16.12, + "learning_rate": 1.3886685552407932e-06, + "loss": 0.3047, + "step": 43658 + }, + { + "epoch": 16.12, + "learning_rate": 1.388631604877448e-06, + "loss": 0.3688, + "step": 43659 + }, + { + "epoch": 16.12, + "learning_rate": 1.3885946545141026e-06, + "loss": 0.2194, + "step": 43660 + }, + { + "epoch": 16.12, + "learning_rate": 1.3885577041507575e-06, + "loss": 0.1562, + "step": 43661 + }, + { + "epoch": 16.12, + "learning_rate": 1.3885207537874123e-06, + "loss": 0.1921, + "step": 43662 + }, + { + "epoch": 16.12, + "learning_rate": 1.388483803424067e-06, + "loss": 0.206, + "step": 43663 + }, + { + "epoch": 16.12, + "learning_rate": 1.3884468530607218e-06, + "loss": 0.22, + "step": 43664 + }, + { + "epoch": 16.12, + "learning_rate": 1.3884099026973767e-06, + "loss": 0.3578, + "step": 43665 + }, + { + "epoch": 16.12, + "learning_rate": 1.3883729523340313e-06, + "loss": 0.1499, + "step": 43666 + }, + { + "epoch": 16.13, + "learning_rate": 1.388336001970686e-06, + "loss": 0.3874, + "step": 43667 + }, + { + "epoch": 16.13, + "learning_rate": 1.3882990516073408e-06, + "loss": 0.3878, + "step": 43668 + }, + { + "epoch": 16.13, + "learning_rate": 1.3882621012439956e-06, + "loss": 0.1404, + "step": 43669 + }, + { + "epoch": 16.13, + "learning_rate": 1.3882251508806503e-06, + "loss": 0.2749, + "step": 43670 + }, + { + "epoch": 16.13, + "learning_rate": 1.3881882005173051e-06, + "loss": 0.2101, + "step": 43671 + }, + { + "epoch": 16.13, + "learning_rate": 1.38815125015396e-06, + "loss": 0.1756, + "step": 43672 + }, + { + "epoch": 16.13, + "learning_rate": 1.3881142997906146e-06, + "loss": 0.2889, + "step": 43673 + }, + { + "epoch": 16.13, + "learning_rate": 1.3880773494272695e-06, + "loss": 0.3134, + "step": 43674 + }, + { + "epoch": 16.13, + "learning_rate": 1.3880403990639243e-06, + "loss": 0.1638, + "step": 43675 + }, + { + "epoch": 16.13, + "learning_rate": 1.388003448700579e-06, + "loss": 0.1581, + "step": 43676 + }, + { + "epoch": 16.13, + "learning_rate": 1.3879664983372336e-06, + "loss": 0.2657, + "step": 43677 + }, + { + "epoch": 16.13, + "learning_rate": 1.3879295479738884e-06, + "loss": 0.2544, + "step": 43678 + }, + { + "epoch": 16.13, + "learning_rate": 1.387892597610543e-06, + "loss": 0.4722, + "step": 43679 + }, + { + "epoch": 16.13, + "learning_rate": 1.387855647247198e-06, + "loss": 0.3264, + "step": 43680 + }, + { + "epoch": 16.13, + "learning_rate": 1.3878186968838528e-06, + "loss": 0.2389, + "step": 43681 + }, + { + "epoch": 16.13, + "learning_rate": 1.3877817465205074e-06, + "loss": 0.2851, + "step": 43682 + }, + { + "epoch": 16.13, + "learning_rate": 1.3877447961571622e-06, + "loss": 0.2742, + "step": 43683 + }, + { + "epoch": 16.13, + "learning_rate": 1.387707845793817e-06, + "loss": 0.2666, + "step": 43684 + }, + { + "epoch": 16.13, + "learning_rate": 1.3876708954304717e-06, + "loss": 0.281, + "step": 43685 + }, + { + "epoch": 16.13, + "learning_rate": 1.3876339450671266e-06, + "loss": 0.2253, + "step": 43686 + }, + { + "epoch": 16.13, + "learning_rate": 1.3875969947037812e-06, + "loss": 0.223, + "step": 43687 + }, + { + "epoch": 16.13, + "learning_rate": 1.387560044340436e-06, + "loss": 0.3317, + "step": 43688 + }, + { + "epoch": 16.13, + "learning_rate": 1.3875230939770907e-06, + "loss": 0.1958, + "step": 43689 + }, + { + "epoch": 16.13, + "learning_rate": 1.3874861436137455e-06, + "loss": 0.1213, + "step": 43690 + }, + { + "epoch": 16.13, + "learning_rate": 1.3874491932504004e-06, + "loss": 0.1767, + "step": 43691 + }, + { + "epoch": 16.13, + "learning_rate": 1.387412242887055e-06, + "loss": 0.0771, + "step": 43692 + }, + { + "epoch": 16.13, + "learning_rate": 1.3873752925237099e-06, + "loss": 0.3862, + "step": 43693 + }, + { + "epoch": 16.14, + "learning_rate": 1.3873383421603647e-06, + "loss": 0.2187, + "step": 43694 + }, + { + "epoch": 16.14, + "learning_rate": 1.3873013917970194e-06, + "loss": 0.1818, + "step": 43695 + }, + { + "epoch": 16.14, + "learning_rate": 1.3872644414336742e-06, + "loss": 0.1058, + "step": 43696 + }, + { + "epoch": 16.14, + "learning_rate": 1.3872274910703288e-06, + "loss": 0.1519, + "step": 43697 + }, + { + "epoch": 16.14, + "learning_rate": 1.3871905407069837e-06, + "loss": 0.2977, + "step": 43698 + }, + { + "epoch": 16.14, + "learning_rate": 1.3871535903436383e-06, + "loss": 0.301, + "step": 43699 + }, + { + "epoch": 16.14, + "learning_rate": 1.3871166399802932e-06, + "loss": 0.1751, + "step": 43700 + }, + { + "epoch": 16.14, + "learning_rate": 1.3870796896169478e-06, + "loss": 0.2479, + "step": 43701 + }, + { + "epoch": 16.14, + "learning_rate": 1.3870427392536027e-06, + "loss": 0.217, + "step": 43702 + }, + { + "epoch": 16.14, + "learning_rate": 1.3870057888902575e-06, + "loss": 0.2062, + "step": 43703 + }, + { + "epoch": 16.14, + "learning_rate": 1.3869688385269121e-06, + "loss": 0.4151, + "step": 43704 + }, + { + "epoch": 16.14, + "learning_rate": 1.386931888163567e-06, + "loss": 0.1724, + "step": 43705 + }, + { + "epoch": 16.14, + "learning_rate": 1.3868949378002218e-06, + "loss": 0.2558, + "step": 43706 + }, + { + "epoch": 16.14, + "learning_rate": 1.3868579874368765e-06, + "loss": 0.1612, + "step": 43707 + }, + { + "epoch": 16.14, + "learning_rate": 1.3868210370735311e-06, + "loss": 0.3134, + "step": 43708 + }, + { + "epoch": 16.14, + "learning_rate": 1.386784086710186e-06, + "loss": 0.1727, + "step": 43709 + }, + { + "epoch": 16.14, + "learning_rate": 1.3867471363468408e-06, + "loss": 0.1546, + "step": 43710 + }, + { + "epoch": 16.14, + "learning_rate": 1.3867101859834955e-06, + "loss": 0.2522, + "step": 43711 + }, + { + "epoch": 16.14, + "learning_rate": 1.3866732356201503e-06, + "loss": 0.2397, + "step": 43712 + }, + { + "epoch": 16.14, + "learning_rate": 1.3866362852568051e-06, + "loss": 0.2358, + "step": 43713 + }, + { + "epoch": 16.14, + "learning_rate": 1.3865993348934598e-06, + "loss": 0.1641, + "step": 43714 + }, + { + "epoch": 16.14, + "learning_rate": 1.3865623845301146e-06, + "loss": 0.2025, + "step": 43715 + }, + { + "epoch": 16.14, + "learning_rate": 1.3865254341667695e-06, + "loss": 0.3034, + "step": 43716 + }, + { + "epoch": 16.14, + "learning_rate": 1.3864884838034241e-06, + "loss": 0.2385, + "step": 43717 + }, + { + "epoch": 16.14, + "learning_rate": 1.3864515334400788e-06, + "loss": 0.2616, + "step": 43718 + }, + { + "epoch": 16.14, + "learning_rate": 1.3864145830767336e-06, + "loss": 0.3292, + "step": 43719 + }, + { + "epoch": 16.14, + "learning_rate": 1.3863776327133884e-06, + "loss": 0.2563, + "step": 43720 + }, + { + "epoch": 16.15, + "learning_rate": 1.386340682350043e-06, + "loss": 0.2294, + "step": 43721 + }, + { + "epoch": 16.15, + "learning_rate": 1.386303731986698e-06, + "loss": 0.2491, + "step": 43722 + }, + { + "epoch": 16.15, + "learning_rate": 1.3862667816233528e-06, + "loss": 0.2283, + "step": 43723 + }, + { + "epoch": 16.15, + "learning_rate": 1.3862298312600074e-06, + "loss": 0.1845, + "step": 43724 + }, + { + "epoch": 16.15, + "learning_rate": 1.3861928808966623e-06, + "loss": 0.3711, + "step": 43725 + }, + { + "epoch": 16.15, + "learning_rate": 1.3861559305333171e-06, + "loss": 0.2344, + "step": 43726 + }, + { + "epoch": 16.15, + "learning_rate": 1.3861189801699715e-06, + "loss": 0.1846, + "step": 43727 + }, + { + "epoch": 16.15, + "learning_rate": 1.3860820298066264e-06, + "loss": 0.2083, + "step": 43728 + }, + { + "epoch": 16.15, + "learning_rate": 1.3860450794432812e-06, + "loss": 0.2583, + "step": 43729 + }, + { + "epoch": 16.15, + "learning_rate": 1.3860081290799359e-06, + "loss": 0.2194, + "step": 43730 + }, + { + "epoch": 16.15, + "learning_rate": 1.3859711787165907e-06, + "loss": 0.3258, + "step": 43731 + }, + { + "epoch": 16.15, + "learning_rate": 1.3859342283532456e-06, + "loss": 0.2046, + "step": 43732 + }, + { + "epoch": 16.15, + "learning_rate": 1.3858972779899002e-06, + "loss": 0.1927, + "step": 43733 + }, + { + "epoch": 16.15, + "learning_rate": 1.385860327626555e-06, + "loss": 0.2228, + "step": 43734 + }, + { + "epoch": 16.15, + "learning_rate": 1.38582337726321e-06, + "loss": 0.2984, + "step": 43735 + }, + { + "epoch": 16.15, + "learning_rate": 1.3857864268998645e-06, + "loss": 0.1837, + "step": 43736 + }, + { + "epoch": 16.15, + "learning_rate": 1.3857494765365192e-06, + "loss": 0.3522, + "step": 43737 + }, + { + "epoch": 16.15, + "learning_rate": 1.385712526173174e-06, + "loss": 0.3383, + "step": 43738 + }, + { + "epoch": 16.15, + "learning_rate": 1.3856755758098289e-06, + "loss": 0.3253, + "step": 43739 + }, + { + "epoch": 16.15, + "learning_rate": 1.3856386254464835e-06, + "loss": 0.1278, + "step": 43740 + }, + { + "epoch": 16.15, + "learning_rate": 1.3856016750831384e-06, + "loss": 0.1983, + "step": 43741 + }, + { + "epoch": 16.15, + "learning_rate": 1.3855647247197932e-06, + "loss": 0.3748, + "step": 43742 + }, + { + "epoch": 16.15, + "learning_rate": 1.3855277743564478e-06, + "loss": 0.2706, + "step": 43743 + }, + { + "epoch": 16.15, + "learning_rate": 1.3854908239931027e-06, + "loss": 0.1921, + "step": 43744 + }, + { + "epoch": 16.15, + "learning_rate": 1.3854538736297575e-06, + "loss": 0.2456, + "step": 43745 + }, + { + "epoch": 16.15, + "learning_rate": 1.3854169232664122e-06, + "loss": 0.2336, + "step": 43746 + }, + { + "epoch": 16.15, + "learning_rate": 1.3853799729030668e-06, + "loss": 0.1678, + "step": 43747 + }, + { + "epoch": 16.16, + "learning_rate": 1.3853430225397217e-06, + "loss": 0.2129, + "step": 43748 + }, + { + "epoch": 16.16, + "learning_rate": 1.3853060721763763e-06, + "loss": 0.2632, + "step": 43749 + }, + { + "epoch": 16.16, + "learning_rate": 1.3852691218130311e-06, + "loss": 0.255, + "step": 43750 + }, + { + "epoch": 16.16, + "learning_rate": 1.385232171449686e-06, + "loss": 0.1986, + "step": 43751 + }, + { + "epoch": 16.16, + "learning_rate": 1.3851952210863406e-06, + "loss": 0.2704, + "step": 43752 + }, + { + "epoch": 16.16, + "learning_rate": 1.3851582707229955e-06, + "loss": 0.1514, + "step": 43753 + }, + { + "epoch": 16.16, + "learning_rate": 1.3851213203596503e-06, + "loss": 0.1797, + "step": 43754 + }, + { + "epoch": 16.16, + "learning_rate": 1.385084369996305e-06, + "loss": 0.3555, + "step": 43755 + }, + { + "epoch": 16.16, + "learning_rate": 1.3850474196329598e-06, + "loss": 0.3078, + "step": 43756 + }, + { + "epoch": 16.16, + "learning_rate": 1.3850104692696144e-06, + "loss": 0.3321, + "step": 43757 + }, + { + "epoch": 16.16, + "learning_rate": 1.3849735189062693e-06, + "loss": 0.4072, + "step": 43758 + }, + { + "epoch": 16.16, + "learning_rate": 1.384936568542924e-06, + "loss": 0.179, + "step": 43759 + }, + { + "epoch": 16.16, + "learning_rate": 1.3848996181795788e-06, + "loss": 0.335, + "step": 43760 + }, + { + "epoch": 16.16, + "learning_rate": 1.3848626678162336e-06, + "loss": 0.199, + "step": 43761 + }, + { + "epoch": 16.16, + "learning_rate": 1.3848257174528883e-06, + "loss": 0.394, + "step": 43762 + }, + { + "epoch": 16.16, + "learning_rate": 1.3847887670895431e-06, + "loss": 0.2818, + "step": 43763 + }, + { + "epoch": 16.16, + "learning_rate": 1.384751816726198e-06, + "loss": 0.1744, + "step": 43764 + }, + { + "epoch": 16.16, + "learning_rate": 1.3847148663628526e-06, + "loss": 0.1412, + "step": 43765 + }, + { + "epoch": 16.16, + "learning_rate": 1.3846779159995074e-06, + "loss": 0.2795, + "step": 43766 + }, + { + "epoch": 16.16, + "learning_rate": 1.384640965636162e-06, + "loss": 0.1826, + "step": 43767 + }, + { + "epoch": 16.16, + "learning_rate": 1.384604015272817e-06, + "loss": 0.2014, + "step": 43768 + }, + { + "epoch": 16.16, + "learning_rate": 1.3845670649094716e-06, + "loss": 0.3129, + "step": 43769 + }, + { + "epoch": 16.16, + "learning_rate": 1.3845301145461264e-06, + "loss": 0.3045, + "step": 43770 + }, + { + "epoch": 16.16, + "learning_rate": 1.3844931641827813e-06, + "loss": 0.3776, + "step": 43771 + }, + { + "epoch": 16.16, + "learning_rate": 1.384456213819436e-06, + "loss": 0.3251, + "step": 43772 + }, + { + "epoch": 16.16, + "learning_rate": 1.3844192634560907e-06, + "loss": 0.3075, + "step": 43773 + }, + { + "epoch": 16.16, + "learning_rate": 1.3843823130927456e-06, + "loss": 0.351, + "step": 43774 + }, + { + "epoch": 16.17, + "learning_rate": 1.3843453627294002e-06, + "loss": 0.2415, + "step": 43775 + }, + { + "epoch": 16.17, + "learning_rate": 1.384308412366055e-06, + "loss": 0.1507, + "step": 43776 + }, + { + "epoch": 16.17, + "learning_rate": 1.3842714620027097e-06, + "loss": 0.2659, + "step": 43777 + }, + { + "epoch": 16.17, + "learning_rate": 1.3842345116393644e-06, + "loss": 0.3102, + "step": 43778 + }, + { + "epoch": 16.17, + "learning_rate": 1.3841975612760192e-06, + "loss": 0.2059, + "step": 43779 + }, + { + "epoch": 16.17, + "learning_rate": 1.384160610912674e-06, + "loss": 0.2701, + "step": 43780 + }, + { + "epoch": 16.17, + "learning_rate": 1.3841236605493287e-06, + "loss": 0.3161, + "step": 43781 + }, + { + "epoch": 16.17, + "learning_rate": 1.3840867101859835e-06, + "loss": 0.2461, + "step": 43782 + }, + { + "epoch": 16.17, + "learning_rate": 1.3840497598226384e-06, + "loss": 0.3577, + "step": 43783 + }, + { + "epoch": 16.17, + "learning_rate": 1.384012809459293e-06, + "loss": 0.1604, + "step": 43784 + }, + { + "epoch": 16.17, + "learning_rate": 1.3839758590959479e-06, + "loss": 0.3377, + "step": 43785 + }, + { + "epoch": 16.17, + "learning_rate": 1.3839389087326027e-06, + "loss": 0.1487, + "step": 43786 + }, + { + "epoch": 16.17, + "learning_rate": 1.3839019583692573e-06, + "loss": 0.2491, + "step": 43787 + }, + { + "epoch": 16.17, + "learning_rate": 1.383865008005912e-06, + "loss": 0.2207, + "step": 43788 + }, + { + "epoch": 16.17, + "learning_rate": 1.3838280576425668e-06, + "loss": 0.3488, + "step": 43789 + }, + { + "epoch": 16.17, + "learning_rate": 1.3837911072792217e-06, + "loss": 0.2197, + "step": 43790 + }, + { + "epoch": 16.17, + "learning_rate": 1.3837541569158763e-06, + "loss": 0.3421, + "step": 43791 + }, + { + "epoch": 16.17, + "learning_rate": 1.3837172065525312e-06, + "loss": 0.2148, + "step": 43792 + }, + { + "epoch": 16.17, + "learning_rate": 1.383680256189186e-06, + "loss": 0.2452, + "step": 43793 + }, + { + "epoch": 16.17, + "learning_rate": 1.3836433058258407e-06, + "loss": 0.1367, + "step": 43794 + }, + { + "epoch": 16.17, + "learning_rate": 1.3836063554624955e-06, + "loss": 0.185, + "step": 43795 + }, + { + "epoch": 16.17, + "learning_rate": 1.3835694050991503e-06, + "loss": 0.3249, + "step": 43796 + }, + { + "epoch": 16.17, + "learning_rate": 1.3835324547358048e-06, + "loss": 0.3023, + "step": 43797 + }, + { + "epoch": 16.17, + "learning_rate": 1.3834955043724596e-06, + "loss": 0.1962, + "step": 43798 + }, + { + "epoch": 16.17, + "learning_rate": 1.3834585540091145e-06, + "loss": 0.1067, + "step": 43799 + }, + { + "epoch": 16.17, + "learning_rate": 1.383421603645769e-06, + "loss": 0.3637, + "step": 43800 + }, + { + "epoch": 16.17, + "learning_rate": 1.383384653282424e-06, + "loss": 0.2115, + "step": 43801 + }, + { + "epoch": 16.18, + "learning_rate": 1.3833477029190788e-06, + "loss": 0.2503, + "step": 43802 + }, + { + "epoch": 16.18, + "learning_rate": 1.3833107525557334e-06, + "loss": 0.1753, + "step": 43803 + }, + { + "epoch": 16.18, + "learning_rate": 1.3832738021923883e-06, + "loss": 0.3042, + "step": 43804 + }, + { + "epoch": 16.18, + "learning_rate": 1.3832368518290431e-06, + "loss": 0.1885, + "step": 43805 + }, + { + "epoch": 16.18, + "learning_rate": 1.3831999014656978e-06, + "loss": 0.2104, + "step": 43806 + }, + { + "epoch": 16.18, + "learning_rate": 1.3831629511023524e-06, + "loss": 0.2332, + "step": 43807 + }, + { + "epoch": 16.18, + "learning_rate": 1.3831260007390073e-06, + "loss": 0.2756, + "step": 43808 + }, + { + "epoch": 16.18, + "learning_rate": 1.383089050375662e-06, + "loss": 0.27, + "step": 43809 + }, + { + "epoch": 16.18, + "learning_rate": 1.3830521000123167e-06, + "loss": 0.1784, + "step": 43810 + }, + { + "epoch": 16.18, + "learning_rate": 1.3830151496489716e-06, + "loss": 0.2464, + "step": 43811 + }, + { + "epoch": 16.18, + "learning_rate": 1.3829781992856264e-06, + "loss": 0.2021, + "step": 43812 + }, + { + "epoch": 16.18, + "learning_rate": 1.382941248922281e-06, + "loss": 0.2817, + "step": 43813 + }, + { + "epoch": 16.18, + "learning_rate": 1.382904298558936e-06, + "loss": 0.2068, + "step": 43814 + }, + { + "epoch": 16.18, + "learning_rate": 1.3828673481955908e-06, + "loss": 0.1563, + "step": 43815 + }, + { + "epoch": 16.18, + "learning_rate": 1.3828303978322454e-06, + "loss": 0.1911, + "step": 43816 + }, + { + "epoch": 16.18, + "learning_rate": 1.3827934474689e-06, + "loss": 0.3521, + "step": 43817 + }, + { + "epoch": 16.18, + "learning_rate": 1.3827564971055549e-06, + "loss": 0.1576, + "step": 43818 + }, + { + "epoch": 16.18, + "learning_rate": 1.3827195467422095e-06, + "loss": 0.212, + "step": 43819 + }, + { + "epoch": 16.18, + "learning_rate": 1.3826825963788644e-06, + "loss": 0.4439, + "step": 43820 + }, + { + "epoch": 16.18, + "learning_rate": 1.3826456460155192e-06, + "loss": 0.3291, + "step": 43821 + }, + { + "epoch": 16.18, + "learning_rate": 1.3826086956521739e-06, + "loss": 0.3112, + "step": 43822 + }, + { + "epoch": 16.18, + "learning_rate": 1.3825717452888287e-06, + "loss": 0.1402, + "step": 43823 + }, + { + "epoch": 16.18, + "learning_rate": 1.3825347949254836e-06, + "loss": 0.174, + "step": 43824 + }, + { + "epoch": 16.18, + "learning_rate": 1.3824978445621382e-06, + "loss": 0.2534, + "step": 43825 + }, + { + "epoch": 16.18, + "learning_rate": 1.382460894198793e-06, + "loss": 0.2286, + "step": 43826 + }, + { + "epoch": 16.18, + "learning_rate": 1.3824239438354477e-06, + "loss": 0.2747, + "step": 43827 + }, + { + "epoch": 16.18, + "learning_rate": 1.3823869934721025e-06, + "loss": 0.2563, + "step": 43828 + }, + { + "epoch": 16.19, + "learning_rate": 1.3823500431087572e-06, + "loss": 0.145, + "step": 43829 + }, + { + "epoch": 16.19, + "learning_rate": 1.382313092745412e-06, + "loss": 0.2079, + "step": 43830 + }, + { + "epoch": 16.19, + "learning_rate": 1.3822761423820669e-06, + "loss": 0.2014, + "step": 43831 + }, + { + "epoch": 16.19, + "learning_rate": 1.3822391920187215e-06, + "loss": 0.2619, + "step": 43832 + }, + { + "epoch": 16.19, + "learning_rate": 1.3822022416553763e-06, + "loss": 0.1889, + "step": 43833 + }, + { + "epoch": 16.19, + "learning_rate": 1.3821652912920312e-06, + "loss": 0.1664, + "step": 43834 + }, + { + "epoch": 16.19, + "learning_rate": 1.3821283409286858e-06, + "loss": 0.2436, + "step": 43835 + }, + { + "epoch": 16.19, + "learning_rate": 1.3820913905653407e-06, + "loss": 0.1817, + "step": 43836 + }, + { + "epoch": 16.19, + "learning_rate": 1.3820544402019953e-06, + "loss": 0.2512, + "step": 43837 + }, + { + "epoch": 16.19, + "learning_rate": 1.3820174898386502e-06, + "loss": 0.253, + "step": 43838 + }, + { + "epoch": 16.19, + "learning_rate": 1.3819805394753048e-06, + "loss": 0.2901, + "step": 43839 + }, + { + "epoch": 16.19, + "learning_rate": 1.3819435891119596e-06, + "loss": 0.4094, + "step": 43840 + }, + { + "epoch": 16.19, + "learning_rate": 1.3819066387486145e-06, + "loss": 0.3596, + "step": 43841 + }, + { + "epoch": 16.19, + "learning_rate": 1.3818696883852691e-06, + "loss": 0.2898, + "step": 43842 + }, + { + "epoch": 16.19, + "learning_rate": 1.381832738021924e-06, + "loss": 0.2166, + "step": 43843 + }, + { + "epoch": 16.19, + "learning_rate": 1.3817957876585788e-06, + "loss": 0.1549, + "step": 43844 + }, + { + "epoch": 16.19, + "learning_rate": 1.3817588372952335e-06, + "loss": 0.1542, + "step": 43845 + }, + { + "epoch": 16.19, + "learning_rate": 1.3817218869318883e-06, + "loss": 0.2956, + "step": 43846 + }, + { + "epoch": 16.19, + "learning_rate": 1.381684936568543e-06, + "loss": 0.193, + "step": 43847 + }, + { + "epoch": 16.19, + "learning_rate": 1.3816479862051976e-06, + "loss": 0.1813, + "step": 43848 + }, + { + "epoch": 16.19, + "learning_rate": 1.3816110358418524e-06, + "loss": 0.0997, + "step": 43849 + }, + { + "epoch": 16.19, + "learning_rate": 1.3815740854785073e-06, + "loss": 0.258, + "step": 43850 + }, + { + "epoch": 16.19, + "learning_rate": 1.381537135115162e-06, + "loss": 0.3673, + "step": 43851 + }, + { + "epoch": 16.19, + "learning_rate": 1.3815001847518168e-06, + "loss": 0.2079, + "step": 43852 + }, + { + "epoch": 16.19, + "learning_rate": 1.3814632343884716e-06, + "loss": 0.3318, + "step": 43853 + }, + { + "epoch": 16.19, + "learning_rate": 1.3814262840251262e-06, + "loss": 0.1777, + "step": 43854 + }, + { + "epoch": 16.19, + "learning_rate": 1.381389333661781e-06, + "loss": 0.2195, + "step": 43855 + }, + { + "epoch": 16.19, + "learning_rate": 1.381352383298436e-06, + "loss": 0.2959, + "step": 43856 + }, + { + "epoch": 16.2, + "learning_rate": 1.3813154329350906e-06, + "loss": 0.1383, + "step": 43857 + }, + { + "epoch": 16.2, + "learning_rate": 1.3812784825717452e-06, + "loss": 0.2232, + "step": 43858 + }, + { + "epoch": 16.2, + "learning_rate": 1.3812415322084e-06, + "loss": 0.2818, + "step": 43859 + }, + { + "epoch": 16.2, + "learning_rate": 1.381204581845055e-06, + "loss": 0.3394, + "step": 43860 + }, + { + "epoch": 16.2, + "learning_rate": 1.3811676314817096e-06, + "loss": 0.2456, + "step": 43861 + }, + { + "epoch": 16.2, + "learning_rate": 1.3811306811183644e-06, + "loss": 0.2665, + "step": 43862 + }, + { + "epoch": 16.2, + "learning_rate": 1.3810937307550192e-06, + "loss": 0.2804, + "step": 43863 + }, + { + "epoch": 16.2, + "learning_rate": 1.3810567803916739e-06, + "loss": 0.1338, + "step": 43864 + }, + { + "epoch": 16.2, + "learning_rate": 1.3810198300283287e-06, + "loss": 0.184, + "step": 43865 + }, + { + "epoch": 16.2, + "learning_rate": 1.3809828796649836e-06, + "loss": 0.2888, + "step": 43866 + }, + { + "epoch": 16.2, + "learning_rate": 1.380945929301638e-06, + "loss": 0.2913, + "step": 43867 + }, + { + "epoch": 16.2, + "learning_rate": 1.3809089789382929e-06, + "loss": 0.3273, + "step": 43868 + }, + { + "epoch": 16.2, + "learning_rate": 1.3808720285749477e-06, + "loss": 0.2364, + "step": 43869 + }, + { + "epoch": 16.2, + "learning_rate": 1.3808350782116023e-06, + "loss": 0.1377, + "step": 43870 + }, + { + "epoch": 16.2, + "learning_rate": 1.3807981278482572e-06, + "loss": 0.2036, + "step": 43871 + }, + { + "epoch": 16.2, + "learning_rate": 1.380761177484912e-06, + "loss": 0.3097, + "step": 43872 + }, + { + "epoch": 16.2, + "learning_rate": 1.3807242271215667e-06, + "loss": 0.2717, + "step": 43873 + }, + { + "epoch": 16.2, + "learning_rate": 1.3806872767582215e-06, + "loss": 0.2479, + "step": 43874 + }, + { + "epoch": 16.2, + "learning_rate": 1.3806503263948764e-06, + "loss": 0.2602, + "step": 43875 + }, + { + "epoch": 16.2, + "learning_rate": 1.380613376031531e-06, + "loss": 0.2542, + "step": 43876 + }, + { + "epoch": 16.2, + "learning_rate": 1.3805764256681856e-06, + "loss": 0.1475, + "step": 43877 + }, + { + "epoch": 16.2, + "learning_rate": 1.3805394753048405e-06, + "loss": 0.1838, + "step": 43878 + }, + { + "epoch": 16.2, + "learning_rate": 1.3805025249414953e-06, + "loss": 0.2176, + "step": 43879 + }, + { + "epoch": 16.2, + "learning_rate": 1.38046557457815e-06, + "loss": 0.209, + "step": 43880 + }, + { + "epoch": 16.2, + "learning_rate": 1.3804286242148048e-06, + "loss": 0.1891, + "step": 43881 + }, + { + "epoch": 16.2, + "learning_rate": 1.3803916738514597e-06, + "loss": 0.2735, + "step": 43882 + }, + { + "epoch": 16.2, + "learning_rate": 1.3803547234881143e-06, + "loss": 0.196, + "step": 43883 + }, + { + "epoch": 16.21, + "learning_rate": 1.3803177731247692e-06, + "loss": 0.1929, + "step": 43884 + }, + { + "epoch": 16.21, + "learning_rate": 1.380280822761424e-06, + "loss": 0.1481, + "step": 43885 + }, + { + "epoch": 16.21, + "learning_rate": 1.3802438723980786e-06, + "loss": 0.3069, + "step": 43886 + }, + { + "epoch": 16.21, + "learning_rate": 1.3802069220347333e-06, + "loss": 0.3169, + "step": 43887 + }, + { + "epoch": 16.21, + "learning_rate": 1.3801699716713881e-06, + "loss": 0.3559, + "step": 43888 + }, + { + "epoch": 16.21, + "learning_rate": 1.3801330213080428e-06, + "loss": 0.184, + "step": 43889 + }, + { + "epoch": 16.21, + "learning_rate": 1.3800960709446976e-06, + "loss": 0.3284, + "step": 43890 + }, + { + "epoch": 16.21, + "learning_rate": 1.3800591205813525e-06, + "loss": 0.178, + "step": 43891 + }, + { + "epoch": 16.21, + "learning_rate": 1.380022170218007e-06, + "loss": 0.2003, + "step": 43892 + }, + { + "epoch": 16.21, + "learning_rate": 1.379985219854662e-06, + "loss": 0.214, + "step": 43893 + }, + { + "epoch": 16.21, + "learning_rate": 1.3799482694913168e-06, + "loss": 0.2722, + "step": 43894 + }, + { + "epoch": 16.21, + "learning_rate": 1.3799113191279714e-06, + "loss": 0.2024, + "step": 43895 + }, + { + "epoch": 16.21, + "learning_rate": 1.379874368764626e-06, + "loss": 0.1998, + "step": 43896 + }, + { + "epoch": 16.21, + "learning_rate": 1.379837418401281e-06, + "loss": 0.196, + "step": 43897 + }, + { + "epoch": 16.21, + "learning_rate": 1.3798004680379358e-06, + "loss": 0.1862, + "step": 43898 + }, + { + "epoch": 16.21, + "learning_rate": 1.3797635176745904e-06, + "loss": 0.1594, + "step": 43899 + }, + { + "epoch": 16.21, + "learning_rate": 1.3797265673112452e-06, + "loss": 0.1854, + "step": 43900 + }, + { + "epoch": 16.21, + "learning_rate": 1.3796896169479e-06, + "loss": 0.3674, + "step": 43901 + }, + { + "epoch": 16.21, + "learning_rate": 1.3796526665845547e-06, + "loss": 0.3957, + "step": 43902 + }, + { + "epoch": 16.21, + "learning_rate": 1.3796157162212096e-06, + "loss": 0.2941, + "step": 43903 + }, + { + "epoch": 16.21, + "learning_rate": 1.3795787658578644e-06, + "loss": 0.2035, + "step": 43904 + }, + { + "epoch": 16.21, + "learning_rate": 1.379541815494519e-06, + "loss": 0.2253, + "step": 43905 + }, + { + "epoch": 16.21, + "learning_rate": 1.3795048651311737e-06, + "loss": 0.1976, + "step": 43906 + }, + { + "epoch": 16.21, + "learning_rate": 1.3794679147678285e-06, + "loss": 0.3092, + "step": 43907 + }, + { + "epoch": 16.21, + "learning_rate": 1.3794309644044834e-06, + "loss": 0.1865, + "step": 43908 + }, + { + "epoch": 16.21, + "learning_rate": 1.379394014041138e-06, + "loss": 0.2748, + "step": 43909 + }, + { + "epoch": 16.21, + "learning_rate": 1.3793570636777929e-06, + "loss": 0.2263, + "step": 43910 + }, + { + "epoch": 16.22, + "learning_rate": 1.3793201133144477e-06, + "loss": 0.1966, + "step": 43911 + }, + { + "epoch": 16.22, + "learning_rate": 1.3792831629511024e-06, + "loss": 0.2272, + "step": 43912 + }, + { + "epoch": 16.22, + "learning_rate": 1.3792462125877572e-06, + "loss": 0.249, + "step": 43913 + }, + { + "epoch": 16.22, + "learning_rate": 1.379209262224412e-06, + "loss": 0.2948, + "step": 43914 + }, + { + "epoch": 16.22, + "learning_rate": 1.3791723118610667e-06, + "loss": 0.1897, + "step": 43915 + }, + { + "epoch": 16.22, + "learning_rate": 1.3791353614977213e-06, + "loss": 0.2745, + "step": 43916 + }, + { + "epoch": 16.22, + "learning_rate": 1.3790984111343762e-06, + "loss": 0.1567, + "step": 43917 + }, + { + "epoch": 16.22, + "learning_rate": 1.3790614607710308e-06, + "loss": 0.2324, + "step": 43918 + }, + { + "epoch": 16.22, + "learning_rate": 1.3790245104076857e-06, + "loss": 0.3303, + "step": 43919 + }, + { + "epoch": 16.22, + "learning_rate": 1.3789875600443405e-06, + "loss": 0.2464, + "step": 43920 + }, + { + "epoch": 16.22, + "learning_rate": 1.3789506096809951e-06, + "loss": 0.2607, + "step": 43921 + }, + { + "epoch": 16.22, + "learning_rate": 1.37891365931765e-06, + "loss": 0.1664, + "step": 43922 + }, + { + "epoch": 16.22, + "learning_rate": 1.3788767089543048e-06, + "loss": 0.1685, + "step": 43923 + }, + { + "epoch": 16.22, + "learning_rate": 1.3788397585909595e-06, + "loss": 0.2415, + "step": 43924 + }, + { + "epoch": 16.22, + "learning_rate": 1.3788028082276143e-06, + "loss": 0.2339, + "step": 43925 + }, + { + "epoch": 16.22, + "learning_rate": 1.378765857864269e-06, + "loss": 0.2512, + "step": 43926 + }, + { + "epoch": 16.22, + "learning_rate": 1.3787289075009238e-06, + "loss": 0.2415, + "step": 43927 + }, + { + "epoch": 16.22, + "learning_rate": 1.3786919571375785e-06, + "loss": 0.2018, + "step": 43928 + }, + { + "epoch": 16.22, + "learning_rate": 1.3786550067742333e-06, + "loss": 0.1433, + "step": 43929 + }, + { + "epoch": 16.22, + "learning_rate": 1.3786180564108881e-06, + "loss": 0.2701, + "step": 43930 + }, + { + "epoch": 16.22, + "learning_rate": 1.3785811060475428e-06, + "loss": 0.1755, + "step": 43931 + }, + { + "epoch": 16.22, + "learning_rate": 1.3785441556841976e-06, + "loss": 0.2854, + "step": 43932 + }, + { + "epoch": 16.22, + "learning_rate": 1.3785072053208525e-06, + "loss": 0.2646, + "step": 43933 + }, + { + "epoch": 16.22, + "learning_rate": 1.3784702549575071e-06, + "loss": 0.2483, + "step": 43934 + }, + { + "epoch": 16.22, + "learning_rate": 1.378433304594162e-06, + "loss": 0.1789, + "step": 43935 + }, + { + "epoch": 16.22, + "learning_rate": 1.3783963542308166e-06, + "loss": 0.2354, + "step": 43936 + }, + { + "epoch": 16.22, + "learning_rate": 1.3783594038674712e-06, + "loss": 0.1848, + "step": 43937 + }, + { + "epoch": 16.23, + "learning_rate": 1.378322453504126e-06, + "loss": 0.2096, + "step": 43938 + }, + { + "epoch": 16.23, + "learning_rate": 1.378285503140781e-06, + "loss": 0.2631, + "step": 43939 + }, + { + "epoch": 16.23, + "learning_rate": 1.3782485527774356e-06, + "loss": 0.3466, + "step": 43940 + }, + { + "epoch": 16.23, + "learning_rate": 1.3782116024140904e-06, + "loss": 0.2126, + "step": 43941 + }, + { + "epoch": 16.23, + "learning_rate": 1.3781746520507453e-06, + "loss": 0.1797, + "step": 43942 + }, + { + "epoch": 16.23, + "learning_rate": 1.3781377016874e-06, + "loss": 0.142, + "step": 43943 + }, + { + "epoch": 16.23, + "learning_rate": 1.3781007513240548e-06, + "loss": 0.1763, + "step": 43944 + }, + { + "epoch": 16.23, + "learning_rate": 1.3780638009607096e-06, + "loss": 0.2103, + "step": 43945 + }, + { + "epoch": 16.23, + "learning_rate": 1.3780268505973642e-06, + "loss": 0.2325, + "step": 43946 + }, + { + "epoch": 16.23, + "learning_rate": 1.3779899002340189e-06, + "loss": 0.3499, + "step": 43947 + }, + { + "epoch": 16.23, + "learning_rate": 1.3779529498706737e-06, + "loss": 0.2979, + "step": 43948 + }, + { + "epoch": 16.23, + "learning_rate": 1.3779159995073286e-06, + "loss": 0.2267, + "step": 43949 + }, + { + "epoch": 16.23, + "learning_rate": 1.3778790491439832e-06, + "loss": 0.2869, + "step": 43950 + }, + { + "epoch": 16.23, + "learning_rate": 1.377842098780638e-06, + "loss": 0.2941, + "step": 43951 + }, + { + "epoch": 16.23, + "learning_rate": 1.377805148417293e-06, + "loss": 0.2845, + "step": 43952 + }, + { + "epoch": 16.23, + "learning_rate": 1.3777681980539475e-06, + "loss": 0.2575, + "step": 43953 + }, + { + "epoch": 16.23, + "learning_rate": 1.3777312476906024e-06, + "loss": 0.2043, + "step": 43954 + }, + { + "epoch": 16.23, + "learning_rate": 1.3776942973272572e-06, + "loss": 0.309, + "step": 43955 + }, + { + "epoch": 16.23, + "learning_rate": 1.3776573469639119e-06, + "loss": 0.3422, + "step": 43956 + }, + { + "epoch": 16.23, + "learning_rate": 1.3776203966005665e-06, + "loss": 0.2323, + "step": 43957 + }, + { + "epoch": 16.23, + "learning_rate": 1.3775834462372214e-06, + "loss": 0.1652, + "step": 43958 + }, + { + "epoch": 16.23, + "learning_rate": 1.3775464958738762e-06, + "loss": 0.2828, + "step": 43959 + }, + { + "epoch": 16.23, + "learning_rate": 1.3775095455105308e-06, + "loss": 0.3504, + "step": 43960 + }, + { + "epoch": 16.23, + "learning_rate": 1.3774725951471857e-06, + "loss": 0.2462, + "step": 43961 + }, + { + "epoch": 16.23, + "learning_rate": 1.3774356447838405e-06, + "loss": 0.267, + "step": 43962 + }, + { + "epoch": 16.23, + "learning_rate": 1.3773986944204952e-06, + "loss": 0.252, + "step": 43963 + }, + { + "epoch": 16.23, + "learning_rate": 1.37736174405715e-06, + "loss": 0.2459, + "step": 43964 + }, + { + "epoch": 16.24, + "learning_rate": 1.3773247936938047e-06, + "loss": 0.3005, + "step": 43965 + }, + { + "epoch": 16.24, + "learning_rate": 1.3772878433304593e-06, + "loss": 0.1985, + "step": 43966 + }, + { + "epoch": 16.24, + "learning_rate": 1.3772508929671141e-06, + "loss": 0.1608, + "step": 43967 + }, + { + "epoch": 16.24, + "learning_rate": 1.377213942603769e-06, + "loss": 0.1611, + "step": 43968 + }, + { + "epoch": 16.24, + "learning_rate": 1.3771769922404236e-06, + "loss": 0.3034, + "step": 43969 + }, + { + "epoch": 16.24, + "learning_rate": 1.3771400418770785e-06, + "loss": 0.3465, + "step": 43970 + }, + { + "epoch": 16.24, + "learning_rate": 1.3771030915137333e-06, + "loss": 0.1477, + "step": 43971 + }, + { + "epoch": 16.24, + "learning_rate": 1.377066141150388e-06, + "loss": 0.2599, + "step": 43972 + }, + { + "epoch": 16.24, + "learning_rate": 1.3770291907870428e-06, + "loss": 0.2483, + "step": 43973 + }, + { + "epoch": 16.24, + "learning_rate": 1.3769922404236977e-06, + "loss": 0.3419, + "step": 43974 + }, + { + "epoch": 16.24, + "learning_rate": 1.3769552900603523e-06, + "loss": 0.3026, + "step": 43975 + }, + { + "epoch": 16.24, + "learning_rate": 1.376918339697007e-06, + "loss": 0.2953, + "step": 43976 + }, + { + "epoch": 16.24, + "learning_rate": 1.3768813893336618e-06, + "loss": 0.2316, + "step": 43977 + }, + { + "epoch": 16.24, + "learning_rate": 1.3768444389703166e-06, + "loss": 0.2942, + "step": 43978 + }, + { + "epoch": 16.24, + "learning_rate": 1.3768074886069713e-06, + "loss": 0.2274, + "step": 43979 + }, + { + "epoch": 16.24, + "learning_rate": 1.3767705382436261e-06, + "loss": 0.3517, + "step": 43980 + }, + { + "epoch": 16.24, + "learning_rate": 1.376733587880281e-06, + "loss": 0.2181, + "step": 43981 + }, + { + "epoch": 16.24, + "learning_rate": 1.3766966375169356e-06, + "loss": 0.1937, + "step": 43982 + }, + { + "epoch": 16.24, + "learning_rate": 1.3766596871535904e-06, + "loss": 0.5106, + "step": 43983 + }, + { + "epoch": 16.24, + "learning_rate": 1.3766227367902453e-06, + "loss": 0.3046, + "step": 43984 + }, + { + "epoch": 16.24, + "learning_rate": 1.3765857864269e-06, + "loss": 0.1613, + "step": 43985 + }, + { + "epoch": 16.24, + "learning_rate": 1.3765488360635546e-06, + "loss": 0.2492, + "step": 43986 + }, + { + "epoch": 16.24, + "learning_rate": 1.3765118857002094e-06, + "loss": 0.3353, + "step": 43987 + }, + { + "epoch": 16.24, + "learning_rate": 1.376474935336864e-06, + "loss": 0.2535, + "step": 43988 + }, + { + "epoch": 16.24, + "learning_rate": 1.376437984973519e-06, + "loss": 0.2553, + "step": 43989 + }, + { + "epoch": 16.24, + "learning_rate": 1.3764010346101737e-06, + "loss": 0.2286, + "step": 43990 + }, + { + "epoch": 16.24, + "learning_rate": 1.3763640842468284e-06, + "loss": 0.1641, + "step": 43991 + }, + { + "epoch": 16.25, + "learning_rate": 1.3763271338834832e-06, + "loss": 0.2663, + "step": 43992 + }, + { + "epoch": 16.25, + "learning_rate": 1.376290183520138e-06, + "loss": 0.1768, + "step": 43993 + }, + { + "epoch": 16.25, + "learning_rate": 1.3762532331567927e-06, + "loss": 0.2572, + "step": 43994 + }, + { + "epoch": 16.25, + "learning_rate": 1.3762162827934476e-06, + "loss": 0.3253, + "step": 43995 + }, + { + "epoch": 16.25, + "learning_rate": 1.3761793324301022e-06, + "loss": 0.1658, + "step": 43996 + }, + { + "epoch": 16.25, + "learning_rate": 1.376142382066757e-06, + "loss": 0.2194, + "step": 43997 + }, + { + "epoch": 16.25, + "learning_rate": 1.3761054317034117e-06, + "loss": 0.1747, + "step": 43998 + }, + { + "epoch": 16.25, + "learning_rate": 1.3760684813400665e-06, + "loss": 0.283, + "step": 43999 + }, + { + "epoch": 16.25, + "learning_rate": 1.3760315309767214e-06, + "loss": 0.2587, + "step": 44000 + }, + { + "epoch": 16.25, + "learning_rate": 1.375994580613376e-06, + "loss": 0.1771, + "step": 44001 + }, + { + "epoch": 16.25, + "learning_rate": 1.3759576302500309e-06, + "loss": 0.2386, + "step": 44002 + }, + { + "epoch": 16.25, + "learning_rate": 1.3759206798866857e-06, + "loss": 0.2619, + "step": 44003 + }, + { + "epoch": 16.25, + "learning_rate": 1.3758837295233403e-06, + "loss": 0.2524, + "step": 44004 + }, + { + "epoch": 16.25, + "learning_rate": 1.3758467791599952e-06, + "loss": 0.1708, + "step": 44005 + }, + { + "epoch": 16.25, + "learning_rate": 1.3758098287966498e-06, + "loss": 0.2814, + "step": 44006 + }, + { + "epoch": 16.25, + "learning_rate": 1.3757728784333045e-06, + "loss": 0.1698, + "step": 44007 + }, + { + "epoch": 16.25, + "learning_rate": 1.3757359280699593e-06, + "loss": 0.1631, + "step": 44008 + }, + { + "epoch": 16.25, + "learning_rate": 1.3756989777066142e-06, + "loss": 0.27, + "step": 44009 + }, + { + "epoch": 16.25, + "learning_rate": 1.3756620273432688e-06, + "loss": 0.1636, + "step": 44010 + }, + { + "epoch": 16.25, + "learning_rate": 1.3756250769799237e-06, + "loss": 0.2697, + "step": 44011 + }, + { + "epoch": 16.25, + "learning_rate": 1.3755881266165785e-06, + "loss": 0.3945, + "step": 44012 + }, + { + "epoch": 16.25, + "learning_rate": 1.3755511762532331e-06, + "loss": 0.2312, + "step": 44013 + }, + { + "epoch": 16.25, + "learning_rate": 1.375514225889888e-06, + "loss": 0.2491, + "step": 44014 + }, + { + "epoch": 16.25, + "learning_rate": 1.3754772755265428e-06, + "loss": 0.1901, + "step": 44015 + }, + { + "epoch": 16.25, + "learning_rate": 1.3754403251631975e-06, + "loss": 0.2095, + "step": 44016 + }, + { + "epoch": 16.25, + "learning_rate": 1.375403374799852e-06, + "loss": 0.2802, + "step": 44017 + }, + { + "epoch": 16.25, + "learning_rate": 1.375366424436507e-06, + "loss": 0.2503, + "step": 44018 + }, + { + "epoch": 16.26, + "learning_rate": 1.3753294740731618e-06, + "loss": 0.274, + "step": 44019 + }, + { + "epoch": 16.26, + "learning_rate": 1.3752925237098164e-06, + "loss": 0.2608, + "step": 44020 + }, + { + "epoch": 16.26, + "learning_rate": 1.3752555733464713e-06, + "loss": 0.2105, + "step": 44021 + }, + { + "epoch": 16.26, + "learning_rate": 1.3752186229831261e-06, + "loss": 0.1664, + "step": 44022 + }, + { + "epoch": 16.26, + "learning_rate": 1.3751816726197808e-06, + "loss": 0.3058, + "step": 44023 + }, + { + "epoch": 16.26, + "learning_rate": 1.3751447222564356e-06, + "loss": 0.3249, + "step": 44024 + }, + { + "epoch": 16.26, + "learning_rate": 1.3751077718930905e-06, + "loss": 0.237, + "step": 44025 + }, + { + "epoch": 16.26, + "learning_rate": 1.375070821529745e-06, + "loss": 0.223, + "step": 44026 + }, + { + "epoch": 16.26, + "learning_rate": 1.3750338711663997e-06, + "loss": 0.1294, + "step": 44027 + }, + { + "epoch": 16.26, + "learning_rate": 1.3749969208030546e-06, + "loss": 0.3025, + "step": 44028 + }, + { + "epoch": 16.26, + "learning_rate": 1.3749599704397094e-06, + "loss": 0.2655, + "step": 44029 + }, + { + "epoch": 16.26, + "learning_rate": 1.374923020076364e-06, + "loss": 0.1997, + "step": 44030 + }, + { + "epoch": 16.26, + "learning_rate": 1.374886069713019e-06, + "loss": 0.2272, + "step": 44031 + }, + { + "epoch": 16.26, + "learning_rate": 1.3748491193496738e-06, + "loss": 0.1814, + "step": 44032 + }, + { + "epoch": 16.26, + "learning_rate": 1.3748121689863284e-06, + "loss": 0.3745, + "step": 44033 + }, + { + "epoch": 16.26, + "learning_rate": 1.3747752186229833e-06, + "loss": 0.3229, + "step": 44034 + }, + { + "epoch": 16.26, + "learning_rate": 1.374738268259638e-06, + "loss": 0.2361, + "step": 44035 + }, + { + "epoch": 16.26, + "learning_rate": 1.3747013178962925e-06, + "loss": 0.1981, + "step": 44036 + }, + { + "epoch": 16.26, + "learning_rate": 1.3746643675329474e-06, + "loss": 0.2175, + "step": 44037 + }, + { + "epoch": 16.26, + "learning_rate": 1.3746274171696022e-06, + "loss": 0.2196, + "step": 44038 + }, + { + "epoch": 16.26, + "learning_rate": 1.3745904668062569e-06, + "loss": 0.2571, + "step": 44039 + }, + { + "epoch": 16.26, + "learning_rate": 1.3745535164429117e-06, + "loss": 0.177, + "step": 44040 + }, + { + "epoch": 16.26, + "learning_rate": 1.3745165660795666e-06, + "loss": 0.3126, + "step": 44041 + }, + { + "epoch": 16.26, + "learning_rate": 1.3744796157162212e-06, + "loss": 0.339, + "step": 44042 + }, + { + "epoch": 16.26, + "learning_rate": 1.374442665352876e-06, + "loss": 0.2312, + "step": 44043 + }, + { + "epoch": 16.26, + "learning_rate": 1.3744057149895309e-06, + "loss": 0.1893, + "step": 44044 + }, + { + "epoch": 16.26, + "learning_rate": 1.3743687646261855e-06, + "loss": 0.3033, + "step": 44045 + }, + { + "epoch": 16.27, + "learning_rate": 1.3743318142628402e-06, + "loss": 0.1753, + "step": 44046 + }, + { + "epoch": 16.27, + "learning_rate": 1.374294863899495e-06, + "loss": 0.3298, + "step": 44047 + }, + { + "epoch": 16.27, + "learning_rate": 1.3742579135361499e-06, + "loss": 0.2314, + "step": 44048 + }, + { + "epoch": 16.27, + "learning_rate": 1.3742209631728045e-06, + "loss": 0.1809, + "step": 44049 + }, + { + "epoch": 16.27, + "learning_rate": 1.3741840128094593e-06, + "loss": 0.1864, + "step": 44050 + }, + { + "epoch": 16.27, + "learning_rate": 1.3741470624461142e-06, + "loss": 0.2692, + "step": 44051 + }, + { + "epoch": 16.27, + "learning_rate": 1.3741101120827688e-06, + "loss": 0.1276, + "step": 44052 + }, + { + "epoch": 16.27, + "learning_rate": 1.3740731617194237e-06, + "loss": 0.2791, + "step": 44053 + }, + { + "epoch": 16.27, + "learning_rate": 1.3740362113560785e-06, + "loss": 0.2732, + "step": 44054 + }, + { + "epoch": 16.27, + "learning_rate": 1.3739992609927332e-06, + "loss": 0.2341, + "step": 44055 + }, + { + "epoch": 16.27, + "learning_rate": 1.3739623106293878e-06, + "loss": 0.3027, + "step": 44056 + }, + { + "epoch": 16.27, + "learning_rate": 1.3739253602660426e-06, + "loss": 0.2559, + "step": 44057 + }, + { + "epoch": 16.27, + "learning_rate": 1.3738884099026973e-06, + "loss": 0.2495, + "step": 44058 + }, + { + "epoch": 16.27, + "learning_rate": 1.3738514595393521e-06, + "loss": 0.2449, + "step": 44059 + }, + { + "epoch": 16.27, + "learning_rate": 1.373814509176007e-06, + "loss": 0.2771, + "step": 44060 + }, + { + "epoch": 16.27, + "learning_rate": 1.3737775588126616e-06, + "loss": 0.2399, + "step": 44061 + }, + { + "epoch": 16.27, + "learning_rate": 1.3737406084493165e-06, + "loss": 0.1615, + "step": 44062 + }, + { + "epoch": 16.27, + "learning_rate": 1.3737036580859713e-06, + "loss": 0.2174, + "step": 44063 + }, + { + "epoch": 16.27, + "learning_rate": 1.373666707722626e-06, + "loss": 0.1868, + "step": 44064 + }, + { + "epoch": 16.27, + "learning_rate": 1.3736297573592808e-06, + "loss": 0.3134, + "step": 44065 + }, + { + "epoch": 16.27, + "learning_rate": 1.3735928069959354e-06, + "loss": 0.1835, + "step": 44066 + }, + { + "epoch": 16.27, + "learning_rate": 1.3735558566325903e-06, + "loss": 0.1176, + "step": 44067 + }, + { + "epoch": 16.27, + "learning_rate": 1.373518906269245e-06, + "loss": 0.2858, + "step": 44068 + }, + { + "epoch": 16.27, + "learning_rate": 1.3734819559058998e-06, + "loss": 0.1827, + "step": 44069 + }, + { + "epoch": 16.27, + "learning_rate": 1.3734450055425546e-06, + "loss": 0.2165, + "step": 44070 + }, + { + "epoch": 16.27, + "learning_rate": 1.3734080551792092e-06, + "loss": 0.1992, + "step": 44071 + }, + { + "epoch": 16.27, + "learning_rate": 1.373371104815864e-06, + "loss": 0.2866, + "step": 44072 + }, + { + "epoch": 16.28, + "learning_rate": 1.373334154452519e-06, + "loss": 0.2205, + "step": 44073 + }, + { + "epoch": 16.28, + "learning_rate": 1.3732972040891736e-06, + "loss": 0.2787, + "step": 44074 + }, + { + "epoch": 16.28, + "learning_rate": 1.3732602537258284e-06, + "loss": 0.3213, + "step": 44075 + }, + { + "epoch": 16.28, + "learning_rate": 1.373223303362483e-06, + "loss": 0.3869, + "step": 44076 + }, + { + "epoch": 16.28, + "learning_rate": 1.3731863529991377e-06, + "loss": 0.1319, + "step": 44077 + }, + { + "epoch": 16.28, + "learning_rate": 1.3731494026357926e-06, + "loss": 0.183, + "step": 44078 + }, + { + "epoch": 16.28, + "learning_rate": 1.3731124522724474e-06, + "loss": 0.2571, + "step": 44079 + }, + { + "epoch": 16.28, + "learning_rate": 1.373075501909102e-06, + "loss": 0.2643, + "step": 44080 + }, + { + "epoch": 16.28, + "learning_rate": 1.3730385515457569e-06, + "loss": 0.267, + "step": 44081 + }, + { + "epoch": 16.28, + "learning_rate": 1.3730016011824117e-06, + "loss": 0.1947, + "step": 44082 + }, + { + "epoch": 16.28, + "learning_rate": 1.3729646508190664e-06, + "loss": 0.2419, + "step": 44083 + }, + { + "epoch": 16.28, + "learning_rate": 1.3729277004557212e-06, + "loss": 0.242, + "step": 44084 + }, + { + "epoch": 16.28, + "learning_rate": 1.372890750092376e-06, + "loss": 0.3133, + "step": 44085 + }, + { + "epoch": 16.28, + "learning_rate": 1.3728537997290307e-06, + "loss": 0.3268, + "step": 44086 + }, + { + "epoch": 16.28, + "learning_rate": 1.3728168493656853e-06, + "loss": 0.2874, + "step": 44087 + }, + { + "epoch": 16.28, + "learning_rate": 1.3727798990023402e-06, + "loss": 0.205, + "step": 44088 + }, + { + "epoch": 16.28, + "learning_rate": 1.372742948638995e-06, + "loss": 0.4226, + "step": 44089 + }, + { + "epoch": 16.28, + "learning_rate": 1.3727059982756497e-06, + "loss": 0.3601, + "step": 44090 + }, + { + "epoch": 16.28, + "learning_rate": 1.3726690479123045e-06, + "loss": 0.2874, + "step": 44091 + }, + { + "epoch": 16.28, + "learning_rate": 1.3726320975489594e-06, + "loss": 0.2405, + "step": 44092 + }, + { + "epoch": 16.28, + "learning_rate": 1.372595147185614e-06, + "loss": 0.1885, + "step": 44093 + }, + { + "epoch": 16.28, + "learning_rate": 1.3725581968222689e-06, + "loss": 0.2522, + "step": 44094 + }, + { + "epoch": 16.28, + "learning_rate": 1.3725212464589237e-06, + "loss": 0.292, + "step": 44095 + }, + { + "epoch": 16.28, + "learning_rate": 1.3724842960955783e-06, + "loss": 0.2759, + "step": 44096 + }, + { + "epoch": 16.28, + "learning_rate": 1.372447345732233e-06, + "loss": 0.1853, + "step": 44097 + }, + { + "epoch": 16.28, + "learning_rate": 1.3724103953688878e-06, + "loss": 0.2148, + "step": 44098 + }, + { + "epoch": 16.28, + "learning_rate": 1.3723734450055427e-06, + "loss": 0.1361, + "step": 44099 + }, + { + "epoch": 16.29, + "learning_rate": 1.3723364946421973e-06, + "loss": 0.2573, + "step": 44100 + }, + { + "epoch": 16.29, + "learning_rate": 1.3722995442788522e-06, + "loss": 0.1977, + "step": 44101 + }, + { + "epoch": 16.29, + "learning_rate": 1.372262593915507e-06, + "loss": 0.3608, + "step": 44102 + }, + { + "epoch": 16.29, + "learning_rate": 1.3722256435521616e-06, + "loss": 0.2054, + "step": 44103 + }, + { + "epoch": 16.29, + "learning_rate": 1.3721886931888165e-06, + "loss": 0.2285, + "step": 44104 + }, + { + "epoch": 16.29, + "learning_rate": 1.3721517428254713e-06, + "loss": 0.3211, + "step": 44105 + }, + { + "epoch": 16.29, + "learning_rate": 1.3721147924621258e-06, + "loss": 0.4205, + "step": 44106 + }, + { + "epoch": 16.29, + "learning_rate": 1.3720778420987806e-06, + "loss": 0.1851, + "step": 44107 + }, + { + "epoch": 16.29, + "learning_rate": 1.3720408917354355e-06, + "loss": 0.4448, + "step": 44108 + }, + { + "epoch": 16.29, + "learning_rate": 1.37200394137209e-06, + "loss": 0.2084, + "step": 44109 + }, + { + "epoch": 16.29, + "learning_rate": 1.371966991008745e-06, + "loss": 0.1854, + "step": 44110 + }, + { + "epoch": 16.29, + "learning_rate": 1.3719300406453998e-06, + "loss": 0.2045, + "step": 44111 + }, + { + "epoch": 16.29, + "learning_rate": 1.3718930902820544e-06, + "loss": 0.4019, + "step": 44112 + }, + { + "epoch": 16.29, + "learning_rate": 1.3718561399187093e-06, + "loss": 0.2169, + "step": 44113 + }, + { + "epoch": 16.29, + "learning_rate": 1.3718191895553641e-06, + "loss": 0.2772, + "step": 44114 + }, + { + "epoch": 16.29, + "learning_rate": 1.3717822391920188e-06, + "loss": 0.2836, + "step": 44115 + }, + { + "epoch": 16.29, + "learning_rate": 1.3717452888286734e-06, + "loss": 0.2068, + "step": 44116 + }, + { + "epoch": 16.29, + "learning_rate": 1.3717083384653282e-06, + "loss": 0.2849, + "step": 44117 + }, + { + "epoch": 16.29, + "learning_rate": 1.371671388101983e-06, + "loss": 0.3772, + "step": 44118 + }, + { + "epoch": 16.29, + "learning_rate": 1.3716344377386377e-06, + "loss": 0.2407, + "step": 44119 + }, + { + "epoch": 16.29, + "learning_rate": 1.3715974873752926e-06, + "loss": 0.2651, + "step": 44120 + }, + { + "epoch": 16.29, + "learning_rate": 1.3715605370119474e-06, + "loss": 0.2436, + "step": 44121 + }, + { + "epoch": 16.29, + "learning_rate": 1.371523586648602e-06, + "loss": 0.2317, + "step": 44122 + }, + { + "epoch": 16.29, + "learning_rate": 1.371486636285257e-06, + "loss": 0.3762, + "step": 44123 + }, + { + "epoch": 16.29, + "learning_rate": 1.3714496859219118e-06, + "loss": 0.1956, + "step": 44124 + }, + { + "epoch": 16.29, + "learning_rate": 1.3714127355585662e-06, + "loss": 0.1857, + "step": 44125 + }, + { + "epoch": 16.29, + "learning_rate": 1.371375785195221e-06, + "loss": 0.3672, + "step": 44126 + }, + { + "epoch": 16.3, + "learning_rate": 1.3713388348318759e-06, + "loss": 0.2511, + "step": 44127 + }, + { + "epoch": 16.3, + "learning_rate": 1.3713018844685305e-06, + "loss": 0.2167, + "step": 44128 + }, + { + "epoch": 16.3, + "learning_rate": 1.3712649341051854e-06, + "loss": 0.2039, + "step": 44129 + }, + { + "epoch": 16.3, + "learning_rate": 1.3712279837418402e-06, + "loss": 0.277, + "step": 44130 + }, + { + "epoch": 16.3, + "learning_rate": 1.3711910333784948e-06, + "loss": 0.2429, + "step": 44131 + }, + { + "epoch": 16.3, + "learning_rate": 1.3711540830151497e-06, + "loss": 0.1383, + "step": 44132 + }, + { + "epoch": 16.3, + "learning_rate": 1.3711171326518045e-06, + "loss": 0.1747, + "step": 44133 + }, + { + "epoch": 16.3, + "learning_rate": 1.3710801822884592e-06, + "loss": 0.2682, + "step": 44134 + }, + { + "epoch": 16.3, + "learning_rate": 1.3710432319251138e-06, + "loss": 0.2234, + "step": 44135 + }, + { + "epoch": 16.3, + "learning_rate": 1.3710062815617687e-06, + "loss": 0.3123, + "step": 44136 + }, + { + "epoch": 16.3, + "learning_rate": 1.3709693311984235e-06, + "loss": 0.2136, + "step": 44137 + }, + { + "epoch": 16.3, + "learning_rate": 1.3709323808350781e-06, + "loss": 0.1569, + "step": 44138 + }, + { + "epoch": 16.3, + "learning_rate": 1.370895430471733e-06, + "loss": 0.2597, + "step": 44139 + }, + { + "epoch": 16.3, + "learning_rate": 1.3708584801083878e-06, + "loss": 0.2842, + "step": 44140 + }, + { + "epoch": 16.3, + "learning_rate": 1.3708215297450425e-06, + "loss": 0.1657, + "step": 44141 + }, + { + "epoch": 16.3, + "learning_rate": 1.3707845793816973e-06, + "loss": 0.3759, + "step": 44142 + }, + { + "epoch": 16.3, + "learning_rate": 1.3707476290183522e-06, + "loss": 0.2455, + "step": 44143 + }, + { + "epoch": 16.3, + "learning_rate": 1.3707106786550068e-06, + "loss": 0.3274, + "step": 44144 + }, + { + "epoch": 16.3, + "learning_rate": 1.3706737282916615e-06, + "loss": 0.2623, + "step": 44145 + }, + { + "epoch": 16.3, + "learning_rate": 1.3706367779283163e-06, + "loss": 0.2362, + "step": 44146 + }, + { + "epoch": 16.3, + "learning_rate": 1.3705998275649711e-06, + "loss": 0.208, + "step": 44147 + }, + { + "epoch": 16.3, + "learning_rate": 1.3705628772016258e-06, + "loss": 0.2599, + "step": 44148 + }, + { + "epoch": 16.3, + "learning_rate": 1.3705259268382806e-06, + "loss": 0.2668, + "step": 44149 + }, + { + "epoch": 16.3, + "learning_rate": 1.3704889764749353e-06, + "loss": 0.2819, + "step": 44150 + }, + { + "epoch": 16.3, + "learning_rate": 1.3704520261115901e-06, + "loss": 0.1628, + "step": 44151 + }, + { + "epoch": 16.3, + "learning_rate": 1.370415075748245e-06, + "loss": 0.2481, + "step": 44152 + }, + { + "epoch": 16.3, + "learning_rate": 1.3703781253848996e-06, + "loss": 0.2682, + "step": 44153 + }, + { + "epoch": 16.31, + "learning_rate": 1.3703411750215544e-06, + "loss": 0.2775, + "step": 44154 + }, + { + "epoch": 16.31, + "learning_rate": 1.370304224658209e-06, + "loss": 0.233, + "step": 44155 + }, + { + "epoch": 16.31, + "learning_rate": 1.370267274294864e-06, + "loss": 0.4179, + "step": 44156 + }, + { + "epoch": 16.31, + "learning_rate": 1.3702303239315186e-06, + "loss": 0.1908, + "step": 44157 + }, + { + "epoch": 16.31, + "learning_rate": 1.3701933735681734e-06, + "loss": 0.1388, + "step": 44158 + }, + { + "epoch": 16.31, + "learning_rate": 1.3701564232048283e-06, + "loss": 0.2342, + "step": 44159 + }, + { + "epoch": 16.31, + "learning_rate": 1.370119472841483e-06, + "loss": 0.2493, + "step": 44160 + }, + { + "epoch": 16.31, + "learning_rate": 1.3700825224781378e-06, + "loss": 0.2109, + "step": 44161 + }, + { + "epoch": 16.31, + "learning_rate": 1.3700455721147926e-06, + "loss": 0.284, + "step": 44162 + }, + { + "epoch": 16.31, + "learning_rate": 1.3700086217514472e-06, + "loss": 0.2022, + "step": 44163 + }, + { + "epoch": 16.31, + "learning_rate": 1.369971671388102e-06, + "loss": 0.3888, + "step": 44164 + }, + { + "epoch": 16.31, + "learning_rate": 1.3699347210247567e-06, + "loss": 0.1375, + "step": 44165 + }, + { + "epoch": 16.31, + "learning_rate": 1.3698977706614116e-06, + "loss": 0.1233, + "step": 44166 + }, + { + "epoch": 16.31, + "learning_rate": 1.3698608202980662e-06, + "loss": 0.3063, + "step": 44167 + }, + { + "epoch": 16.31, + "learning_rate": 1.369823869934721e-06, + "loss": 0.3309, + "step": 44168 + }, + { + "epoch": 16.31, + "learning_rate": 1.369786919571376e-06, + "loss": 0.2834, + "step": 44169 + }, + { + "epoch": 16.31, + "learning_rate": 1.3697499692080305e-06, + "loss": 0.3262, + "step": 44170 + }, + { + "epoch": 16.31, + "learning_rate": 1.3697130188446854e-06, + "loss": 0.412, + "step": 44171 + }, + { + "epoch": 16.31, + "learning_rate": 1.3696760684813402e-06, + "loss": 0.2685, + "step": 44172 + }, + { + "epoch": 16.31, + "learning_rate": 1.3696391181179949e-06, + "loss": 0.1853, + "step": 44173 + }, + { + "epoch": 16.31, + "learning_rate": 1.3696021677546497e-06, + "loss": 0.3589, + "step": 44174 + }, + { + "epoch": 16.31, + "learning_rate": 1.3695652173913044e-06, + "loss": 0.1446, + "step": 44175 + }, + { + "epoch": 16.31, + "learning_rate": 1.369528267027959e-06, + "loss": 0.2573, + "step": 44176 + }, + { + "epoch": 16.31, + "learning_rate": 1.3694913166646138e-06, + "loss": 0.1304, + "step": 44177 + }, + { + "epoch": 16.31, + "learning_rate": 1.3694543663012687e-06, + "loss": 0.2476, + "step": 44178 + }, + { + "epoch": 16.31, + "learning_rate": 1.3694174159379233e-06, + "loss": 0.3925, + "step": 44179 + }, + { + "epoch": 16.31, + "learning_rate": 1.3693804655745782e-06, + "loss": 0.2537, + "step": 44180 + }, + { + "epoch": 16.31, + "learning_rate": 1.369343515211233e-06, + "loss": 0.298, + "step": 44181 + }, + { + "epoch": 16.32, + "learning_rate": 1.3693065648478877e-06, + "loss": 0.3792, + "step": 44182 + }, + { + "epoch": 16.32, + "learning_rate": 1.3692696144845425e-06, + "loss": 0.2673, + "step": 44183 + }, + { + "epoch": 16.32, + "learning_rate": 1.3692326641211974e-06, + "loss": 0.2539, + "step": 44184 + }, + { + "epoch": 16.32, + "learning_rate": 1.369195713757852e-06, + "loss": 0.2788, + "step": 44185 + }, + { + "epoch": 16.32, + "learning_rate": 1.3691587633945066e-06, + "loss": 0.1883, + "step": 44186 + }, + { + "epoch": 16.32, + "learning_rate": 1.3691218130311615e-06, + "loss": 0.2594, + "step": 44187 + }, + { + "epoch": 16.32, + "learning_rate": 1.3690848626678163e-06, + "loss": 0.198, + "step": 44188 + }, + { + "epoch": 16.32, + "learning_rate": 1.369047912304471e-06, + "loss": 0.1597, + "step": 44189 + }, + { + "epoch": 16.32, + "learning_rate": 1.3690109619411258e-06, + "loss": 0.1761, + "step": 44190 + }, + { + "epoch": 16.32, + "learning_rate": 1.3689740115777807e-06, + "loss": 0.3415, + "step": 44191 + }, + { + "epoch": 16.32, + "learning_rate": 1.3689370612144353e-06, + "loss": 0.3632, + "step": 44192 + }, + { + "epoch": 16.32, + "learning_rate": 1.3689001108510901e-06, + "loss": 0.2571, + "step": 44193 + }, + { + "epoch": 16.32, + "learning_rate": 1.368863160487745e-06, + "loss": 0.231, + "step": 44194 + }, + { + "epoch": 16.32, + "learning_rate": 1.3688262101243994e-06, + "loss": 0.2946, + "step": 44195 + }, + { + "epoch": 16.32, + "learning_rate": 1.3687892597610543e-06, + "loss": 0.237, + "step": 44196 + }, + { + "epoch": 16.32, + "learning_rate": 1.3687523093977091e-06, + "loss": 0.1998, + "step": 44197 + }, + { + "epoch": 16.32, + "learning_rate": 1.3687153590343637e-06, + "loss": 0.2337, + "step": 44198 + }, + { + "epoch": 16.32, + "learning_rate": 1.3686784086710186e-06, + "loss": 0.3225, + "step": 44199 + }, + { + "epoch": 16.32, + "learning_rate": 1.3686414583076734e-06, + "loss": 0.2591, + "step": 44200 + }, + { + "epoch": 16.32, + "learning_rate": 1.368604507944328e-06, + "loss": 0.2031, + "step": 44201 + }, + { + "epoch": 16.32, + "learning_rate": 1.368567557580983e-06, + "loss": 0.1427, + "step": 44202 + }, + { + "epoch": 16.32, + "learning_rate": 1.3685306072176378e-06, + "loss": 0.1909, + "step": 44203 + }, + { + "epoch": 16.32, + "learning_rate": 1.3684936568542924e-06, + "loss": 0.1681, + "step": 44204 + }, + { + "epoch": 16.32, + "learning_rate": 1.368456706490947e-06, + "loss": 0.1781, + "step": 44205 + }, + { + "epoch": 16.32, + "learning_rate": 1.368419756127602e-06, + "loss": 0.1837, + "step": 44206 + }, + { + "epoch": 16.32, + "learning_rate": 1.3683828057642567e-06, + "loss": 0.2487, + "step": 44207 + }, + { + "epoch": 16.32, + "learning_rate": 1.3683458554009114e-06, + "loss": 0.3075, + "step": 44208 + }, + { + "epoch": 16.33, + "learning_rate": 1.3683089050375662e-06, + "loss": 0.3564, + "step": 44209 + }, + { + "epoch": 16.33, + "learning_rate": 1.368271954674221e-06, + "loss": 0.2004, + "step": 44210 + }, + { + "epoch": 16.33, + "learning_rate": 1.3682350043108757e-06, + "loss": 0.2794, + "step": 44211 + }, + { + "epoch": 16.33, + "learning_rate": 1.3681980539475306e-06, + "loss": 0.1877, + "step": 44212 + }, + { + "epoch": 16.33, + "learning_rate": 1.3681611035841854e-06, + "loss": 0.141, + "step": 44213 + }, + { + "epoch": 16.33, + "learning_rate": 1.36812415322084e-06, + "loss": 0.2647, + "step": 44214 + }, + { + "epoch": 16.33, + "learning_rate": 1.3680872028574947e-06, + "loss": 0.2213, + "step": 44215 + }, + { + "epoch": 16.33, + "learning_rate": 1.3680502524941495e-06, + "loss": 0.2692, + "step": 44216 + }, + { + "epoch": 16.33, + "learning_rate": 1.3680133021308044e-06, + "loss": 0.2271, + "step": 44217 + }, + { + "epoch": 16.33, + "learning_rate": 1.367976351767459e-06, + "loss": 0.2064, + "step": 44218 + }, + { + "epoch": 16.33, + "learning_rate": 1.3679394014041139e-06, + "loss": 0.2124, + "step": 44219 + }, + { + "epoch": 16.33, + "learning_rate": 1.3679024510407687e-06, + "loss": 0.3713, + "step": 44220 + }, + { + "epoch": 16.33, + "learning_rate": 1.3678655006774233e-06, + "loss": 0.1893, + "step": 44221 + }, + { + "epoch": 16.33, + "learning_rate": 1.3678285503140782e-06, + "loss": 0.3161, + "step": 44222 + }, + { + "epoch": 16.33, + "learning_rate": 1.367791599950733e-06, + "loss": 0.1913, + "step": 44223 + }, + { + "epoch": 16.33, + "learning_rate": 1.3677546495873877e-06, + "loss": 0.2972, + "step": 44224 + }, + { + "epoch": 16.33, + "learning_rate": 1.3677176992240423e-06, + "loss": 0.2797, + "step": 44225 + }, + { + "epoch": 16.33, + "learning_rate": 1.3676807488606972e-06, + "loss": 0.1968, + "step": 44226 + }, + { + "epoch": 16.33, + "learning_rate": 1.3676437984973518e-06, + "loss": 0.2661, + "step": 44227 + }, + { + "epoch": 16.33, + "learning_rate": 1.3676068481340067e-06, + "loss": 0.1818, + "step": 44228 + }, + { + "epoch": 16.33, + "learning_rate": 1.3675698977706615e-06, + "loss": 0.3579, + "step": 44229 + }, + { + "epoch": 16.33, + "learning_rate": 1.3675329474073161e-06, + "loss": 0.1865, + "step": 44230 + }, + { + "epoch": 16.33, + "learning_rate": 1.367495997043971e-06, + "loss": 0.2373, + "step": 44231 + }, + { + "epoch": 16.33, + "learning_rate": 1.3674590466806258e-06, + "loss": 0.2794, + "step": 44232 + }, + { + "epoch": 16.33, + "learning_rate": 1.3674220963172805e-06, + "loss": 0.2461, + "step": 44233 + }, + { + "epoch": 16.33, + "learning_rate": 1.3673851459539353e-06, + "loss": 0.169, + "step": 44234 + }, + { + "epoch": 16.33, + "learning_rate": 1.36734819559059e-06, + "loss": 0.3152, + "step": 44235 + }, + { + "epoch": 16.34, + "learning_rate": 1.3673112452272448e-06, + "loss": 0.2711, + "step": 44236 + }, + { + "epoch": 16.34, + "learning_rate": 1.3672742948638994e-06, + "loss": 0.2689, + "step": 44237 + }, + { + "epoch": 16.34, + "learning_rate": 1.3672373445005543e-06, + "loss": 0.1835, + "step": 44238 + }, + { + "epoch": 16.34, + "learning_rate": 1.3672003941372091e-06, + "loss": 0.3201, + "step": 44239 + }, + { + "epoch": 16.34, + "learning_rate": 1.3671634437738638e-06, + "loss": 0.1522, + "step": 44240 + }, + { + "epoch": 16.34, + "learning_rate": 1.3671264934105186e-06, + "loss": 0.2194, + "step": 44241 + }, + { + "epoch": 16.34, + "learning_rate": 1.3670895430471735e-06, + "loss": 0.268, + "step": 44242 + }, + { + "epoch": 16.34, + "learning_rate": 1.367052592683828e-06, + "loss": 0.2769, + "step": 44243 + }, + { + "epoch": 16.34, + "learning_rate": 1.367015642320483e-06, + "loss": 0.1982, + "step": 44244 + }, + { + "epoch": 16.34, + "learning_rate": 1.3669786919571376e-06, + "loss": 0.1988, + "step": 44245 + }, + { + "epoch": 16.34, + "learning_rate": 1.3669417415937922e-06, + "loss": 0.2568, + "step": 44246 + }, + { + "epoch": 16.34, + "learning_rate": 1.366904791230447e-06, + "loss": 0.2129, + "step": 44247 + }, + { + "epoch": 16.34, + "learning_rate": 1.366867840867102e-06, + "loss": 0.2168, + "step": 44248 + }, + { + "epoch": 16.34, + "learning_rate": 1.3668308905037566e-06, + "loss": 0.1859, + "step": 44249 + }, + { + "epoch": 16.34, + "learning_rate": 1.3667939401404114e-06, + "loss": 0.3004, + "step": 44250 + }, + { + "epoch": 16.34, + "learning_rate": 1.3667569897770663e-06, + "loss": 0.2612, + "step": 44251 + }, + { + "epoch": 16.34, + "learning_rate": 1.3667200394137209e-06, + "loss": 0.3791, + "step": 44252 + }, + { + "epoch": 16.34, + "learning_rate": 1.3666830890503757e-06, + "loss": 0.0857, + "step": 44253 + }, + { + "epoch": 16.34, + "learning_rate": 1.3666461386870306e-06, + "loss": 0.1945, + "step": 44254 + }, + { + "epoch": 16.34, + "learning_rate": 1.3666091883236852e-06, + "loss": 0.2739, + "step": 44255 + }, + { + "epoch": 16.34, + "learning_rate": 1.3665722379603399e-06, + "loss": 0.1702, + "step": 44256 + }, + { + "epoch": 16.34, + "learning_rate": 1.3665352875969947e-06, + "loss": 0.2774, + "step": 44257 + }, + { + "epoch": 16.34, + "learning_rate": 1.3664983372336496e-06, + "loss": 0.219, + "step": 44258 + }, + { + "epoch": 16.34, + "learning_rate": 1.3664613868703042e-06, + "loss": 0.1881, + "step": 44259 + }, + { + "epoch": 16.34, + "learning_rate": 1.366424436506959e-06, + "loss": 0.2044, + "step": 44260 + }, + { + "epoch": 16.34, + "learning_rate": 1.3663874861436139e-06, + "loss": 0.3168, + "step": 44261 + }, + { + "epoch": 16.34, + "learning_rate": 1.3663505357802685e-06, + "loss": 0.3127, + "step": 44262 + }, + { + "epoch": 16.35, + "learning_rate": 1.3663135854169234e-06, + "loss": 0.3353, + "step": 44263 + }, + { + "epoch": 16.35, + "learning_rate": 1.3662766350535782e-06, + "loss": 0.2497, + "step": 44264 + }, + { + "epoch": 16.35, + "learning_rate": 1.3662396846902326e-06, + "loss": 0.323, + "step": 44265 + }, + { + "epoch": 16.35, + "learning_rate": 1.3662027343268875e-06, + "loss": 0.2853, + "step": 44266 + }, + { + "epoch": 16.35, + "learning_rate": 1.3661657839635423e-06, + "loss": 0.2274, + "step": 44267 + }, + { + "epoch": 16.35, + "learning_rate": 1.366128833600197e-06, + "loss": 0.1651, + "step": 44268 + }, + { + "epoch": 16.35, + "learning_rate": 1.3660918832368518e-06, + "loss": 0.1896, + "step": 44269 + }, + { + "epoch": 16.35, + "learning_rate": 1.3660549328735067e-06, + "loss": 0.3009, + "step": 44270 + }, + { + "epoch": 16.35, + "learning_rate": 1.3660179825101613e-06, + "loss": 0.1853, + "step": 44271 + }, + { + "epoch": 16.35, + "learning_rate": 1.3659810321468162e-06, + "loss": 0.2826, + "step": 44272 + }, + { + "epoch": 16.35, + "learning_rate": 1.365944081783471e-06, + "loss": 0.4003, + "step": 44273 + }, + { + "epoch": 16.35, + "learning_rate": 1.3659071314201256e-06, + "loss": 0.2031, + "step": 44274 + }, + { + "epoch": 16.35, + "learning_rate": 1.3658701810567803e-06, + "loss": 0.1959, + "step": 44275 + }, + { + "epoch": 16.35, + "learning_rate": 1.3658332306934351e-06, + "loss": 0.2218, + "step": 44276 + }, + { + "epoch": 16.35, + "learning_rate": 1.36579628033009e-06, + "loss": 0.2071, + "step": 44277 + }, + { + "epoch": 16.35, + "learning_rate": 1.3657593299667446e-06, + "loss": 0.1916, + "step": 44278 + }, + { + "epoch": 16.35, + "learning_rate": 1.3657223796033995e-06, + "loss": 0.292, + "step": 44279 + }, + { + "epoch": 16.35, + "learning_rate": 1.3656854292400543e-06, + "loss": 0.2596, + "step": 44280 + }, + { + "epoch": 16.35, + "learning_rate": 1.365648478876709e-06, + "loss": 0.2518, + "step": 44281 + }, + { + "epoch": 16.35, + "learning_rate": 1.3656115285133638e-06, + "loss": 0.1705, + "step": 44282 + }, + { + "epoch": 16.35, + "learning_rate": 1.3655745781500186e-06, + "loss": 0.193, + "step": 44283 + }, + { + "epoch": 16.35, + "learning_rate": 1.3655376277866733e-06, + "loss": 0.2752, + "step": 44284 + }, + { + "epoch": 16.35, + "learning_rate": 1.365500677423328e-06, + "loss": 0.2798, + "step": 44285 + }, + { + "epoch": 16.35, + "learning_rate": 1.3654637270599828e-06, + "loss": 0.2934, + "step": 44286 + }, + { + "epoch": 16.35, + "learning_rate": 1.3654267766966376e-06, + "loss": 0.1842, + "step": 44287 + }, + { + "epoch": 16.35, + "learning_rate": 1.3653898263332922e-06, + "loss": 0.3118, + "step": 44288 + }, + { + "epoch": 16.35, + "learning_rate": 1.365352875969947e-06, + "loss": 0.2885, + "step": 44289 + }, + { + "epoch": 16.36, + "learning_rate": 1.365315925606602e-06, + "loss": 0.4189, + "step": 44290 + }, + { + "epoch": 16.36, + "learning_rate": 1.3652789752432566e-06, + "loss": 0.1783, + "step": 44291 + }, + { + "epoch": 16.36, + "learning_rate": 1.3652420248799114e-06, + "loss": 0.23, + "step": 44292 + }, + { + "epoch": 16.36, + "learning_rate": 1.3652050745165663e-06, + "loss": 0.2459, + "step": 44293 + }, + { + "epoch": 16.36, + "learning_rate": 1.365168124153221e-06, + "loss": 0.2503, + "step": 44294 + }, + { + "epoch": 16.36, + "learning_rate": 1.3651311737898756e-06, + "loss": 0.1638, + "step": 44295 + }, + { + "epoch": 16.36, + "learning_rate": 1.3650942234265304e-06, + "loss": 0.1772, + "step": 44296 + }, + { + "epoch": 16.36, + "learning_rate": 1.365057273063185e-06, + "loss": 0.2641, + "step": 44297 + }, + { + "epoch": 16.36, + "learning_rate": 1.3650203226998399e-06, + "loss": 0.2446, + "step": 44298 + }, + { + "epoch": 16.36, + "learning_rate": 1.3649833723364947e-06, + "loss": 0.207, + "step": 44299 + }, + { + "epoch": 16.36, + "learning_rate": 1.3649464219731494e-06, + "loss": 0.2804, + "step": 44300 + }, + { + "epoch": 16.36, + "learning_rate": 1.3649094716098042e-06, + "loss": 0.2448, + "step": 44301 + }, + { + "epoch": 16.36, + "learning_rate": 1.364872521246459e-06, + "loss": 0.2347, + "step": 44302 + }, + { + "epoch": 16.36, + "learning_rate": 1.3648355708831137e-06, + "loss": 0.2013, + "step": 44303 + }, + { + "epoch": 16.36, + "learning_rate": 1.3647986205197685e-06, + "loss": 0.2607, + "step": 44304 + }, + { + "epoch": 16.36, + "learning_rate": 1.3647616701564232e-06, + "loss": 0.2038, + "step": 44305 + }, + { + "epoch": 16.36, + "learning_rate": 1.364724719793078e-06, + "loss": 0.3532, + "step": 44306 + }, + { + "epoch": 16.36, + "learning_rate": 1.3646877694297327e-06, + "loss": 0.1806, + "step": 44307 + }, + { + "epoch": 16.36, + "learning_rate": 1.3646508190663875e-06, + "loss": 0.3478, + "step": 44308 + }, + { + "epoch": 16.36, + "learning_rate": 1.3646138687030424e-06, + "loss": 0.2279, + "step": 44309 + }, + { + "epoch": 16.36, + "learning_rate": 1.364576918339697e-06, + "loss": 0.3375, + "step": 44310 + }, + { + "epoch": 16.36, + "learning_rate": 1.3645399679763519e-06, + "loss": 0.1603, + "step": 44311 + }, + { + "epoch": 16.36, + "learning_rate": 1.3645030176130067e-06, + "loss": 0.2667, + "step": 44312 + }, + { + "epoch": 16.36, + "learning_rate": 1.3644660672496613e-06, + "loss": 0.2363, + "step": 44313 + }, + { + "epoch": 16.36, + "learning_rate": 1.3644291168863162e-06, + "loss": 0.2491, + "step": 44314 + }, + { + "epoch": 16.36, + "learning_rate": 1.3643921665229708e-06, + "loss": 0.2086, + "step": 44315 + }, + { + "epoch": 16.36, + "learning_rate": 1.3643552161596255e-06, + "loss": 0.1692, + "step": 44316 + }, + { + "epoch": 16.37, + "learning_rate": 1.3643182657962803e-06, + "loss": 0.1364, + "step": 44317 + }, + { + "epoch": 16.37, + "learning_rate": 1.3642813154329352e-06, + "loss": 0.2005, + "step": 44318 + }, + { + "epoch": 16.37, + "learning_rate": 1.3642443650695898e-06, + "loss": 0.1491, + "step": 44319 + }, + { + "epoch": 16.37, + "learning_rate": 1.3642074147062446e-06, + "loss": 0.1608, + "step": 44320 + }, + { + "epoch": 16.37, + "learning_rate": 1.3641704643428995e-06, + "loss": 0.2227, + "step": 44321 + }, + { + "epoch": 16.37, + "learning_rate": 1.3641335139795541e-06, + "loss": 0.391, + "step": 44322 + }, + { + "epoch": 16.37, + "learning_rate": 1.364096563616209e-06, + "loss": 0.2408, + "step": 44323 + }, + { + "epoch": 16.37, + "learning_rate": 1.3640596132528638e-06, + "loss": 0.1376, + "step": 44324 + }, + { + "epoch": 16.37, + "learning_rate": 1.3640226628895185e-06, + "loss": 0.2438, + "step": 44325 + }, + { + "epoch": 16.37, + "learning_rate": 1.363985712526173e-06, + "loss": 0.2398, + "step": 44326 + }, + { + "epoch": 16.37, + "learning_rate": 1.363948762162828e-06, + "loss": 0.2252, + "step": 44327 + }, + { + "epoch": 16.37, + "learning_rate": 1.3639118117994828e-06, + "loss": 0.1914, + "step": 44328 + }, + { + "epoch": 16.37, + "learning_rate": 1.3638748614361374e-06, + "loss": 0.1991, + "step": 44329 + }, + { + "epoch": 16.37, + "learning_rate": 1.3638379110727923e-06, + "loss": 0.1623, + "step": 44330 + }, + { + "epoch": 16.37, + "learning_rate": 1.3638009607094471e-06, + "loss": 0.5133, + "step": 44331 + }, + { + "epoch": 16.37, + "learning_rate": 1.3637640103461018e-06, + "loss": 0.2522, + "step": 44332 + }, + { + "epoch": 16.37, + "learning_rate": 1.3637270599827566e-06, + "loss": 0.2631, + "step": 44333 + }, + { + "epoch": 16.37, + "learning_rate": 1.3636901096194115e-06, + "loss": 0.265, + "step": 44334 + }, + { + "epoch": 16.37, + "learning_rate": 1.363653159256066e-06, + "loss": 0.2154, + "step": 44335 + }, + { + "epoch": 16.37, + "learning_rate": 1.3636162088927207e-06, + "loss": 0.2159, + "step": 44336 + }, + { + "epoch": 16.37, + "learning_rate": 1.3635792585293756e-06, + "loss": 0.1809, + "step": 44337 + }, + { + "epoch": 16.37, + "learning_rate": 1.3635423081660302e-06, + "loss": 0.2913, + "step": 44338 + }, + { + "epoch": 16.37, + "learning_rate": 1.363505357802685e-06, + "loss": 0.3702, + "step": 44339 + }, + { + "epoch": 16.37, + "learning_rate": 1.36346840743934e-06, + "loss": 0.2876, + "step": 44340 + }, + { + "epoch": 16.37, + "learning_rate": 1.3634314570759945e-06, + "loss": 0.2311, + "step": 44341 + }, + { + "epoch": 16.37, + "learning_rate": 1.3633945067126494e-06, + "loss": 0.1791, + "step": 44342 + }, + { + "epoch": 16.37, + "learning_rate": 1.3633575563493042e-06, + "loss": 0.1973, + "step": 44343 + }, + { + "epoch": 16.38, + "learning_rate": 1.3633206059859589e-06, + "loss": 0.2083, + "step": 44344 + }, + { + "epoch": 16.38, + "learning_rate": 1.3632836556226135e-06, + "loss": 0.2166, + "step": 44345 + }, + { + "epoch": 16.38, + "learning_rate": 1.3632467052592684e-06, + "loss": 0.1792, + "step": 44346 + }, + { + "epoch": 16.38, + "learning_rate": 1.3632097548959232e-06, + "loss": 0.2272, + "step": 44347 + }, + { + "epoch": 16.38, + "learning_rate": 1.3631728045325778e-06, + "loss": 0.2944, + "step": 44348 + }, + { + "epoch": 16.38, + "learning_rate": 1.3631358541692327e-06, + "loss": 0.2733, + "step": 44349 + }, + { + "epoch": 16.38, + "learning_rate": 1.3630989038058875e-06, + "loss": 0.2673, + "step": 44350 + }, + { + "epoch": 16.38, + "learning_rate": 1.3630619534425422e-06, + "loss": 0.368, + "step": 44351 + }, + { + "epoch": 16.38, + "learning_rate": 1.363025003079197e-06, + "loss": 0.1865, + "step": 44352 + }, + { + "epoch": 16.38, + "learning_rate": 1.3629880527158519e-06, + "loss": 0.1907, + "step": 44353 + }, + { + "epoch": 16.38, + "learning_rate": 1.3629511023525065e-06, + "loss": 0.3059, + "step": 44354 + }, + { + "epoch": 16.38, + "learning_rate": 1.3629141519891611e-06, + "loss": 0.2249, + "step": 44355 + }, + { + "epoch": 16.38, + "learning_rate": 1.362877201625816e-06, + "loss": 0.1657, + "step": 44356 + }, + { + "epoch": 16.38, + "learning_rate": 1.3628402512624708e-06, + "loss": 0.2496, + "step": 44357 + }, + { + "epoch": 16.38, + "learning_rate": 1.3628033008991255e-06, + "loss": 0.3186, + "step": 44358 + }, + { + "epoch": 16.38, + "learning_rate": 1.3627663505357803e-06, + "loss": 0.1821, + "step": 44359 + }, + { + "epoch": 16.38, + "learning_rate": 1.3627294001724352e-06, + "loss": 0.1229, + "step": 44360 + }, + { + "epoch": 16.38, + "learning_rate": 1.3626924498090898e-06, + "loss": 0.1307, + "step": 44361 + }, + { + "epoch": 16.38, + "learning_rate": 1.3626554994457447e-06, + "loss": 0.177, + "step": 44362 + }, + { + "epoch": 16.38, + "learning_rate": 1.3626185490823995e-06, + "loss": 0.3365, + "step": 44363 + }, + { + "epoch": 16.38, + "learning_rate": 1.362581598719054e-06, + "loss": 0.2516, + "step": 44364 + }, + { + "epoch": 16.38, + "learning_rate": 1.3625446483557088e-06, + "loss": 0.2387, + "step": 44365 + }, + { + "epoch": 16.38, + "learning_rate": 1.3625076979923636e-06, + "loss": 0.1223, + "step": 44366 + }, + { + "epoch": 16.38, + "learning_rate": 1.3624707476290183e-06, + "loss": 0.3111, + "step": 44367 + }, + { + "epoch": 16.38, + "learning_rate": 1.3624337972656731e-06, + "loss": 0.2335, + "step": 44368 + }, + { + "epoch": 16.38, + "learning_rate": 1.362396846902328e-06, + "loss": 0.3873, + "step": 44369 + }, + { + "epoch": 16.38, + "learning_rate": 1.3623598965389826e-06, + "loss": 0.2002, + "step": 44370 + }, + { + "epoch": 16.39, + "learning_rate": 1.3623229461756374e-06, + "loss": 0.1166, + "step": 44371 + }, + { + "epoch": 16.39, + "learning_rate": 1.3622859958122923e-06, + "loss": 0.1839, + "step": 44372 + }, + { + "epoch": 16.39, + "learning_rate": 1.362249045448947e-06, + "loss": 0.3516, + "step": 44373 + }, + { + "epoch": 16.39, + "learning_rate": 1.3622120950856016e-06, + "loss": 0.3246, + "step": 44374 + }, + { + "epoch": 16.39, + "learning_rate": 1.3621751447222564e-06, + "loss": 0.304, + "step": 44375 + }, + { + "epoch": 16.39, + "learning_rate": 1.3621381943589113e-06, + "loss": 0.2383, + "step": 44376 + }, + { + "epoch": 16.39, + "learning_rate": 1.362101243995566e-06, + "loss": 0.2034, + "step": 44377 + }, + { + "epoch": 16.39, + "learning_rate": 1.3620642936322208e-06, + "loss": 0.221, + "step": 44378 + }, + { + "epoch": 16.39, + "learning_rate": 1.3620273432688756e-06, + "loss": 0.236, + "step": 44379 + }, + { + "epoch": 16.39, + "learning_rate": 1.3619903929055302e-06, + "loss": 0.3218, + "step": 44380 + }, + { + "epoch": 16.39, + "learning_rate": 1.361953442542185e-06, + "loss": 0.1866, + "step": 44381 + }, + { + "epoch": 16.39, + "learning_rate": 1.36191649217884e-06, + "loss": 0.268, + "step": 44382 + }, + { + "epoch": 16.39, + "learning_rate": 1.3618795418154946e-06, + "loss": 0.2606, + "step": 44383 + }, + { + "epoch": 16.39, + "learning_rate": 1.3618425914521492e-06, + "loss": 0.3376, + "step": 44384 + }, + { + "epoch": 16.39, + "learning_rate": 1.361805641088804e-06, + "loss": 0.2991, + "step": 44385 + }, + { + "epoch": 16.39, + "learning_rate": 1.3617686907254587e-06, + "loss": 0.4015, + "step": 44386 + }, + { + "epoch": 16.39, + "learning_rate": 1.3617317403621135e-06, + "loss": 0.3313, + "step": 44387 + }, + { + "epoch": 16.39, + "learning_rate": 1.3616947899987684e-06, + "loss": 0.303, + "step": 44388 + }, + { + "epoch": 16.39, + "learning_rate": 1.361657839635423e-06, + "loss": 0.2992, + "step": 44389 + }, + { + "epoch": 16.39, + "learning_rate": 1.3616208892720779e-06, + "loss": 0.3111, + "step": 44390 + }, + { + "epoch": 16.39, + "learning_rate": 1.3615839389087327e-06, + "loss": 0.2852, + "step": 44391 + }, + { + "epoch": 16.39, + "learning_rate": 1.3615469885453874e-06, + "loss": 0.1738, + "step": 44392 + }, + { + "epoch": 16.39, + "learning_rate": 1.3615100381820422e-06, + "loss": 0.2519, + "step": 44393 + }, + { + "epoch": 16.39, + "learning_rate": 1.3614730878186968e-06, + "loss": 0.2617, + "step": 44394 + }, + { + "epoch": 16.39, + "learning_rate": 1.3614361374553517e-06, + "loss": 0.2507, + "step": 44395 + }, + { + "epoch": 16.39, + "learning_rate": 1.3613991870920063e-06, + "loss": 0.3458, + "step": 44396 + }, + { + "epoch": 16.39, + "learning_rate": 1.3613622367286612e-06, + "loss": 0.2842, + "step": 44397 + }, + { + "epoch": 16.4, + "learning_rate": 1.361325286365316e-06, + "loss": 0.1601, + "step": 44398 + }, + { + "epoch": 16.4, + "learning_rate": 1.3612883360019707e-06, + "loss": 0.3115, + "step": 44399 + }, + { + "epoch": 16.4, + "learning_rate": 1.3612513856386255e-06, + "loss": 0.2212, + "step": 44400 + }, + { + "epoch": 16.4, + "learning_rate": 1.3612144352752804e-06, + "loss": 0.2852, + "step": 44401 + }, + { + "epoch": 16.4, + "learning_rate": 1.361177484911935e-06, + "loss": 0.2931, + "step": 44402 + }, + { + "epoch": 16.4, + "learning_rate": 1.3611405345485898e-06, + "loss": 0.1774, + "step": 44403 + }, + { + "epoch": 16.4, + "learning_rate": 1.3611035841852445e-06, + "loss": 0.3541, + "step": 44404 + }, + { + "epoch": 16.4, + "learning_rate": 1.3610666338218993e-06, + "loss": 0.1859, + "step": 44405 + }, + { + "epoch": 16.4, + "learning_rate": 1.361029683458554e-06, + "loss": 0.3268, + "step": 44406 + }, + { + "epoch": 16.4, + "learning_rate": 1.3609927330952088e-06, + "loss": 0.2705, + "step": 44407 + }, + { + "epoch": 16.4, + "learning_rate": 1.3609557827318637e-06, + "loss": 0.1884, + "step": 44408 + }, + { + "epoch": 16.4, + "learning_rate": 1.3609188323685183e-06, + "loss": 0.2407, + "step": 44409 + }, + { + "epoch": 16.4, + "learning_rate": 1.3608818820051731e-06, + "loss": 0.3274, + "step": 44410 + }, + { + "epoch": 16.4, + "learning_rate": 1.360844931641828e-06, + "loss": 0.1849, + "step": 44411 + }, + { + "epoch": 16.4, + "learning_rate": 1.3608079812784826e-06, + "loss": 0.2827, + "step": 44412 + }, + { + "epoch": 16.4, + "learning_rate": 1.3607710309151375e-06, + "loss": 0.1932, + "step": 44413 + }, + { + "epoch": 16.4, + "learning_rate": 1.3607340805517921e-06, + "loss": 0.2629, + "step": 44414 + }, + { + "epoch": 16.4, + "learning_rate": 1.3606971301884467e-06, + "loss": 0.1711, + "step": 44415 + }, + { + "epoch": 16.4, + "learning_rate": 1.3606601798251016e-06, + "loss": 0.1957, + "step": 44416 + }, + { + "epoch": 16.4, + "learning_rate": 1.3606232294617564e-06, + "loss": 0.2496, + "step": 44417 + }, + { + "epoch": 16.4, + "learning_rate": 1.360586279098411e-06, + "loss": 0.1948, + "step": 44418 + }, + { + "epoch": 16.4, + "learning_rate": 1.360549328735066e-06, + "loss": 0.2361, + "step": 44419 + }, + { + "epoch": 16.4, + "learning_rate": 1.3605123783717208e-06, + "loss": 0.2834, + "step": 44420 + }, + { + "epoch": 16.4, + "learning_rate": 1.3604754280083754e-06, + "loss": 0.3102, + "step": 44421 + }, + { + "epoch": 16.4, + "learning_rate": 1.3604384776450303e-06, + "loss": 0.3467, + "step": 44422 + }, + { + "epoch": 16.4, + "learning_rate": 1.3604015272816851e-06, + "loss": 0.2813, + "step": 44423 + }, + { + "epoch": 16.4, + "learning_rate": 1.3603645769183397e-06, + "loss": 0.3628, + "step": 44424 + }, + { + "epoch": 16.41, + "learning_rate": 1.3603276265549944e-06, + "loss": 0.1923, + "step": 44425 + }, + { + "epoch": 16.41, + "learning_rate": 1.3602906761916492e-06, + "loss": 0.2952, + "step": 44426 + }, + { + "epoch": 16.41, + "learning_rate": 1.360253725828304e-06, + "loss": 0.2328, + "step": 44427 + }, + { + "epoch": 16.41, + "learning_rate": 1.3602167754649587e-06, + "loss": 0.2917, + "step": 44428 + }, + { + "epoch": 16.41, + "learning_rate": 1.3601798251016136e-06, + "loss": 0.2589, + "step": 44429 + }, + { + "epoch": 16.41, + "learning_rate": 1.3601428747382684e-06, + "loss": 0.2491, + "step": 44430 + }, + { + "epoch": 16.41, + "learning_rate": 1.360105924374923e-06, + "loss": 0.2235, + "step": 44431 + }, + { + "epoch": 16.41, + "learning_rate": 1.3600689740115779e-06, + "loss": 0.3248, + "step": 44432 + }, + { + "epoch": 16.41, + "learning_rate": 1.3600320236482327e-06, + "loss": 0.2383, + "step": 44433 + }, + { + "epoch": 16.41, + "learning_rate": 1.3599950732848872e-06, + "loss": 0.1774, + "step": 44434 + }, + { + "epoch": 16.41, + "learning_rate": 1.359958122921542e-06, + "loss": 0.1806, + "step": 44435 + }, + { + "epoch": 16.41, + "learning_rate": 1.3599211725581969e-06, + "loss": 0.2272, + "step": 44436 + }, + { + "epoch": 16.41, + "learning_rate": 1.3598842221948515e-06, + "loss": 0.1359, + "step": 44437 + }, + { + "epoch": 16.41, + "learning_rate": 1.3598472718315063e-06, + "loss": 0.2819, + "step": 44438 + }, + { + "epoch": 16.41, + "learning_rate": 1.3598103214681612e-06, + "loss": 0.2339, + "step": 44439 + }, + { + "epoch": 16.41, + "learning_rate": 1.3597733711048158e-06, + "loss": 0.1497, + "step": 44440 + }, + { + "epoch": 16.41, + "learning_rate": 1.3597364207414707e-06, + "loss": 0.2239, + "step": 44441 + }, + { + "epoch": 16.41, + "learning_rate": 1.3596994703781255e-06, + "loss": 0.2506, + "step": 44442 + }, + { + "epoch": 16.41, + "learning_rate": 1.3596625200147802e-06, + "loss": 0.2218, + "step": 44443 + }, + { + "epoch": 16.41, + "learning_rate": 1.3596255696514348e-06, + "loss": 0.3436, + "step": 44444 + }, + { + "epoch": 16.41, + "learning_rate": 1.3595886192880896e-06, + "loss": 0.2418, + "step": 44445 + }, + { + "epoch": 16.41, + "learning_rate": 1.3595516689247445e-06, + "loss": 0.2945, + "step": 44446 + }, + { + "epoch": 16.41, + "learning_rate": 1.3595147185613991e-06, + "loss": 0.2343, + "step": 44447 + }, + { + "epoch": 16.41, + "learning_rate": 1.359477768198054e-06, + "loss": 0.3051, + "step": 44448 + }, + { + "epoch": 16.41, + "learning_rate": 1.3594408178347088e-06, + "loss": 0.2656, + "step": 44449 + }, + { + "epoch": 16.41, + "learning_rate": 1.3594038674713635e-06, + "loss": 0.1811, + "step": 44450 + }, + { + "epoch": 16.41, + "learning_rate": 1.3593669171080183e-06, + "loss": 0.267, + "step": 44451 + }, + { + "epoch": 16.42, + "learning_rate": 1.3593299667446732e-06, + "loss": 0.1695, + "step": 44452 + }, + { + "epoch": 16.42, + "learning_rate": 1.3592930163813278e-06, + "loss": 0.3691, + "step": 44453 + }, + { + "epoch": 16.42, + "learning_rate": 1.3592560660179824e-06, + "loss": 0.181, + "step": 44454 + }, + { + "epoch": 16.42, + "learning_rate": 1.3592191156546373e-06, + "loss": 0.2967, + "step": 44455 + }, + { + "epoch": 16.42, + "learning_rate": 1.359182165291292e-06, + "loss": 0.3238, + "step": 44456 + }, + { + "epoch": 16.42, + "learning_rate": 1.3591452149279468e-06, + "loss": 0.2176, + "step": 44457 + }, + { + "epoch": 16.42, + "learning_rate": 1.3591082645646016e-06, + "loss": 0.101, + "step": 44458 + }, + { + "epoch": 16.42, + "learning_rate": 1.3590713142012563e-06, + "loss": 0.3163, + "step": 44459 + }, + { + "epoch": 16.42, + "learning_rate": 1.359034363837911e-06, + "loss": 0.3328, + "step": 44460 + }, + { + "epoch": 16.42, + "learning_rate": 1.358997413474566e-06, + "loss": 0.1991, + "step": 44461 + }, + { + "epoch": 16.42, + "learning_rate": 1.3589604631112206e-06, + "loss": 0.3884, + "step": 44462 + }, + { + "epoch": 16.42, + "learning_rate": 1.3589235127478754e-06, + "loss": 0.3053, + "step": 44463 + }, + { + "epoch": 16.42, + "learning_rate": 1.35888656238453e-06, + "loss": 0.1636, + "step": 44464 + }, + { + "epoch": 16.42, + "learning_rate": 1.358849612021185e-06, + "loss": 0.2949, + "step": 44465 + }, + { + "epoch": 16.42, + "learning_rate": 1.3588126616578396e-06, + "loss": 0.2102, + "step": 44466 + }, + { + "epoch": 16.42, + "learning_rate": 1.3587757112944944e-06, + "loss": 0.1156, + "step": 44467 + }, + { + "epoch": 16.42, + "learning_rate": 1.3587387609311493e-06, + "loss": 0.1851, + "step": 44468 + }, + { + "epoch": 16.42, + "learning_rate": 1.3587018105678039e-06, + "loss": 0.2748, + "step": 44469 + }, + { + "epoch": 16.42, + "learning_rate": 1.3586648602044587e-06, + "loss": 0.2349, + "step": 44470 + }, + { + "epoch": 16.42, + "learning_rate": 1.3586279098411136e-06, + "loss": 0.2035, + "step": 44471 + }, + { + "epoch": 16.42, + "learning_rate": 1.3585909594777682e-06, + "loss": 0.2613, + "step": 44472 + }, + { + "epoch": 16.42, + "learning_rate": 1.358554009114423e-06, + "loss": 0.3546, + "step": 44473 + }, + { + "epoch": 16.42, + "learning_rate": 1.3585170587510777e-06, + "loss": 0.1993, + "step": 44474 + }, + { + "epoch": 16.42, + "learning_rate": 1.3584801083877326e-06, + "loss": 0.3704, + "step": 44475 + }, + { + "epoch": 16.42, + "learning_rate": 1.3584431580243872e-06, + "loss": 0.2004, + "step": 44476 + }, + { + "epoch": 16.42, + "learning_rate": 1.358406207661042e-06, + "loss": 0.2665, + "step": 44477 + }, + { + "epoch": 16.42, + "learning_rate": 1.3583692572976969e-06, + "loss": 0.3609, + "step": 44478 + }, + { + "epoch": 16.43, + "learning_rate": 1.3583323069343515e-06, + "loss": 0.2653, + "step": 44479 + }, + { + "epoch": 16.43, + "learning_rate": 1.3582953565710064e-06, + "loss": 0.2955, + "step": 44480 + }, + { + "epoch": 16.43, + "learning_rate": 1.3582584062076612e-06, + "loss": 0.2746, + "step": 44481 + }, + { + "epoch": 16.43, + "learning_rate": 1.3582214558443159e-06, + "loss": 0.2133, + "step": 44482 + }, + { + "epoch": 16.43, + "learning_rate": 1.3581845054809707e-06, + "loss": 0.3094, + "step": 44483 + }, + { + "epoch": 16.43, + "learning_rate": 1.3581475551176253e-06, + "loss": 0.3791, + "step": 44484 + }, + { + "epoch": 16.43, + "learning_rate": 1.35811060475428e-06, + "loss": 0.1756, + "step": 44485 + }, + { + "epoch": 16.43, + "learning_rate": 1.3580736543909348e-06, + "loss": 0.3743, + "step": 44486 + }, + { + "epoch": 16.43, + "learning_rate": 1.3580367040275897e-06, + "loss": 0.2773, + "step": 44487 + }, + { + "epoch": 16.43, + "learning_rate": 1.3579997536642443e-06, + "loss": 0.2538, + "step": 44488 + }, + { + "epoch": 16.43, + "learning_rate": 1.3579628033008992e-06, + "loss": 0.2502, + "step": 44489 + }, + { + "epoch": 16.43, + "learning_rate": 1.357925852937554e-06, + "loss": 0.2006, + "step": 44490 + }, + { + "epoch": 16.43, + "learning_rate": 1.3578889025742086e-06, + "loss": 0.1805, + "step": 44491 + }, + { + "epoch": 16.43, + "learning_rate": 1.3578519522108635e-06, + "loss": 0.3852, + "step": 44492 + }, + { + "epoch": 16.43, + "learning_rate": 1.3578150018475183e-06, + "loss": 0.1542, + "step": 44493 + }, + { + "epoch": 16.43, + "learning_rate": 1.357778051484173e-06, + "loss": 0.3354, + "step": 44494 + }, + { + "epoch": 16.43, + "learning_rate": 1.3577411011208276e-06, + "loss": 0.2535, + "step": 44495 + }, + { + "epoch": 16.43, + "learning_rate": 1.3577041507574825e-06, + "loss": 0.1832, + "step": 44496 + }, + { + "epoch": 16.43, + "learning_rate": 1.3576672003941373e-06, + "loss": 0.2498, + "step": 44497 + }, + { + "epoch": 16.43, + "learning_rate": 1.357630250030792e-06, + "loss": 0.3315, + "step": 44498 + }, + { + "epoch": 16.43, + "learning_rate": 1.3575932996674468e-06, + "loss": 0.2257, + "step": 44499 + }, + { + "epoch": 16.43, + "learning_rate": 1.3575563493041016e-06, + "loss": 0.3283, + "step": 44500 + }, + { + "epoch": 16.43, + "learning_rate": 1.3575193989407563e-06, + "loss": 0.3476, + "step": 44501 + }, + { + "epoch": 16.43, + "learning_rate": 1.3574824485774111e-06, + "loss": 0.2572, + "step": 44502 + }, + { + "epoch": 16.43, + "learning_rate": 1.357445498214066e-06, + "loss": 0.275, + "step": 44503 + }, + { + "epoch": 16.43, + "learning_rate": 1.3574085478507204e-06, + "loss": 0.2391, + "step": 44504 + }, + { + "epoch": 16.43, + "learning_rate": 1.3573715974873752e-06, + "loss": 0.2233, + "step": 44505 + }, + { + "epoch": 16.44, + "learning_rate": 1.35733464712403e-06, + "loss": 0.306, + "step": 44506 + }, + { + "epoch": 16.44, + "learning_rate": 1.3572976967606847e-06, + "loss": 0.3538, + "step": 44507 + }, + { + "epoch": 16.44, + "learning_rate": 1.3572607463973396e-06, + "loss": 0.2648, + "step": 44508 + }, + { + "epoch": 16.44, + "learning_rate": 1.3572237960339944e-06, + "loss": 0.1958, + "step": 44509 + }, + { + "epoch": 16.44, + "learning_rate": 1.357186845670649e-06, + "loss": 0.3126, + "step": 44510 + }, + { + "epoch": 16.44, + "learning_rate": 1.357149895307304e-06, + "loss": 0.2861, + "step": 44511 + }, + { + "epoch": 16.44, + "learning_rate": 1.3571129449439588e-06, + "loss": 0.2918, + "step": 44512 + }, + { + "epoch": 16.44, + "learning_rate": 1.3570759945806134e-06, + "loss": 0.2508, + "step": 44513 + }, + { + "epoch": 16.44, + "learning_rate": 1.357039044217268e-06, + "loss": 0.2897, + "step": 44514 + }, + { + "epoch": 16.44, + "learning_rate": 1.3570020938539229e-06, + "loss": 0.3101, + "step": 44515 + }, + { + "epoch": 16.44, + "learning_rate": 1.3569651434905777e-06, + "loss": 0.2358, + "step": 44516 + }, + { + "epoch": 16.44, + "learning_rate": 1.3569281931272324e-06, + "loss": 0.2697, + "step": 44517 + }, + { + "epoch": 16.44, + "learning_rate": 1.3568912427638872e-06, + "loss": 0.2907, + "step": 44518 + }, + { + "epoch": 16.44, + "learning_rate": 1.356854292400542e-06, + "loss": 0.2651, + "step": 44519 + }, + { + "epoch": 16.44, + "learning_rate": 1.3568173420371967e-06, + "loss": 0.2736, + "step": 44520 + }, + { + "epoch": 16.44, + "learning_rate": 1.3567803916738515e-06, + "loss": 0.2843, + "step": 44521 + }, + { + "epoch": 16.44, + "learning_rate": 1.3567434413105064e-06, + "loss": 0.2123, + "step": 44522 + }, + { + "epoch": 16.44, + "learning_rate": 1.356706490947161e-06, + "loss": 0.1945, + "step": 44523 + }, + { + "epoch": 16.44, + "learning_rate": 1.3566695405838157e-06, + "loss": 0.252, + "step": 44524 + }, + { + "epoch": 16.44, + "learning_rate": 1.3566325902204705e-06, + "loss": 0.2048, + "step": 44525 + }, + { + "epoch": 16.44, + "learning_rate": 1.3565956398571252e-06, + "loss": 0.2113, + "step": 44526 + }, + { + "epoch": 16.44, + "learning_rate": 1.35655868949378e-06, + "loss": 0.2904, + "step": 44527 + }, + { + "epoch": 16.44, + "learning_rate": 1.3565217391304349e-06, + "loss": 0.212, + "step": 44528 + }, + { + "epoch": 16.44, + "learning_rate": 1.3564847887670895e-06, + "loss": 0.2061, + "step": 44529 + }, + { + "epoch": 16.44, + "learning_rate": 1.3564478384037443e-06, + "loss": 0.1847, + "step": 44530 + }, + { + "epoch": 16.44, + "learning_rate": 1.3564108880403992e-06, + "loss": 0.1984, + "step": 44531 + }, + { + "epoch": 16.44, + "learning_rate": 1.3563739376770538e-06, + "loss": 0.2252, + "step": 44532 + }, + { + "epoch": 16.44, + "learning_rate": 1.3563369873137087e-06, + "loss": 0.3144, + "step": 44533 + }, + { + "epoch": 16.45, + "learning_rate": 1.3563000369503633e-06, + "loss": 0.1472, + "step": 44534 + }, + { + "epoch": 16.45, + "learning_rate": 1.3562630865870182e-06, + "loss": 0.2642, + "step": 44535 + }, + { + "epoch": 16.45, + "learning_rate": 1.3562261362236728e-06, + "loss": 0.2875, + "step": 44536 + }, + { + "epoch": 16.45, + "learning_rate": 1.3561891858603276e-06, + "loss": 0.3382, + "step": 44537 + }, + { + "epoch": 16.45, + "learning_rate": 1.3561522354969825e-06, + "loss": 0.2463, + "step": 44538 + }, + { + "epoch": 16.45, + "learning_rate": 1.3561152851336371e-06, + "loss": 0.2193, + "step": 44539 + }, + { + "epoch": 16.45, + "learning_rate": 1.356078334770292e-06, + "loss": 0.3683, + "step": 44540 + }, + { + "epoch": 16.45, + "learning_rate": 1.3560413844069468e-06, + "loss": 0.2068, + "step": 44541 + }, + { + "epoch": 16.45, + "learning_rate": 1.3560044340436015e-06, + "loss": 0.1908, + "step": 44542 + }, + { + "epoch": 16.45, + "learning_rate": 1.3559674836802563e-06, + "loss": 0.1781, + "step": 44543 + }, + { + "epoch": 16.45, + "learning_rate": 1.355930533316911e-06, + "loss": 0.2404, + "step": 44544 + }, + { + "epoch": 16.45, + "learning_rate": 1.3558935829535658e-06, + "loss": 0.219, + "step": 44545 + }, + { + "epoch": 16.45, + "learning_rate": 1.3558566325902204e-06, + "loss": 0.2235, + "step": 44546 + }, + { + "epoch": 16.45, + "learning_rate": 1.3558196822268753e-06, + "loss": 0.2642, + "step": 44547 + }, + { + "epoch": 16.45, + "learning_rate": 1.3557827318635301e-06, + "loss": 0.3977, + "step": 44548 + }, + { + "epoch": 16.45, + "learning_rate": 1.3557457815001848e-06, + "loss": 0.1883, + "step": 44549 + }, + { + "epoch": 16.45, + "learning_rate": 1.3557088311368396e-06, + "loss": 0.3892, + "step": 44550 + }, + { + "epoch": 16.45, + "learning_rate": 1.3556718807734945e-06, + "loss": 0.2026, + "step": 44551 + }, + { + "epoch": 16.45, + "learning_rate": 1.355634930410149e-06, + "loss": 0.2578, + "step": 44552 + }, + { + "epoch": 16.45, + "learning_rate": 1.355597980046804e-06, + "loss": 0.2596, + "step": 44553 + }, + { + "epoch": 16.45, + "learning_rate": 1.3555610296834586e-06, + "loss": 0.2163, + "step": 44554 + }, + { + "epoch": 16.45, + "learning_rate": 1.3555240793201132e-06, + "loss": 0.1892, + "step": 44555 + }, + { + "epoch": 16.45, + "learning_rate": 1.355487128956768e-06, + "loss": 0.4384, + "step": 44556 + }, + { + "epoch": 16.45, + "learning_rate": 1.355450178593423e-06, + "loss": 0.221, + "step": 44557 + }, + { + "epoch": 16.45, + "learning_rate": 1.3554132282300775e-06, + "loss": 0.1836, + "step": 44558 + }, + { + "epoch": 16.45, + "learning_rate": 1.3553762778667324e-06, + "loss": 0.2758, + "step": 44559 + }, + { + "epoch": 16.45, + "learning_rate": 1.3553393275033872e-06, + "loss": 0.3409, + "step": 44560 + }, + { + "epoch": 16.46, + "learning_rate": 1.3553023771400419e-06, + "loss": 0.2445, + "step": 44561 + }, + { + "epoch": 16.46, + "learning_rate": 1.3552654267766967e-06, + "loss": 0.2454, + "step": 44562 + }, + { + "epoch": 16.46, + "learning_rate": 1.3552284764133516e-06, + "loss": 0.2724, + "step": 44563 + }, + { + "epoch": 16.46, + "learning_rate": 1.3551915260500062e-06, + "loss": 0.3571, + "step": 44564 + }, + { + "epoch": 16.46, + "learning_rate": 1.3551545756866608e-06, + "loss": 0.2234, + "step": 44565 + }, + { + "epoch": 16.46, + "learning_rate": 1.3551176253233157e-06, + "loss": 0.2793, + "step": 44566 + }, + { + "epoch": 16.46, + "learning_rate": 1.3550806749599705e-06, + "loss": 0.2754, + "step": 44567 + }, + { + "epoch": 16.46, + "learning_rate": 1.3550437245966252e-06, + "loss": 0.2288, + "step": 44568 + }, + { + "epoch": 16.46, + "learning_rate": 1.35500677423328e-06, + "loss": 0.3736, + "step": 44569 + }, + { + "epoch": 16.46, + "learning_rate": 1.3549698238699349e-06, + "loss": 0.3382, + "step": 44570 + }, + { + "epoch": 16.46, + "learning_rate": 1.3549328735065895e-06, + "loss": 0.2477, + "step": 44571 + }, + { + "epoch": 16.46, + "learning_rate": 1.3548959231432444e-06, + "loss": 0.1005, + "step": 44572 + }, + { + "epoch": 16.46, + "learning_rate": 1.3548589727798992e-06, + "loss": 0.1859, + "step": 44573 + }, + { + "epoch": 16.46, + "learning_rate": 1.3548220224165536e-06, + "loss": 0.2492, + "step": 44574 + }, + { + "epoch": 16.46, + "learning_rate": 1.3547850720532085e-06, + "loss": 0.2656, + "step": 44575 + }, + { + "epoch": 16.46, + "learning_rate": 1.3547481216898633e-06, + "loss": 0.1669, + "step": 44576 + }, + { + "epoch": 16.46, + "learning_rate": 1.354711171326518e-06, + "loss": 0.2976, + "step": 44577 + }, + { + "epoch": 16.46, + "learning_rate": 1.3546742209631728e-06, + "loss": 0.3889, + "step": 44578 + }, + { + "epoch": 16.46, + "learning_rate": 1.3546372705998277e-06, + "loss": 0.2211, + "step": 44579 + }, + { + "epoch": 16.46, + "learning_rate": 1.3546003202364823e-06, + "loss": 0.2768, + "step": 44580 + }, + { + "epoch": 16.46, + "learning_rate": 1.3545633698731371e-06, + "loss": 0.2335, + "step": 44581 + }, + { + "epoch": 16.46, + "learning_rate": 1.354526419509792e-06, + "loss": 0.209, + "step": 44582 + }, + { + "epoch": 16.46, + "learning_rate": 1.3544894691464466e-06, + "loss": 0.2527, + "step": 44583 + }, + { + "epoch": 16.46, + "learning_rate": 1.3544525187831013e-06, + "loss": 0.2748, + "step": 44584 + }, + { + "epoch": 16.46, + "learning_rate": 1.3544155684197561e-06, + "loss": 0.2398, + "step": 44585 + }, + { + "epoch": 16.46, + "learning_rate": 1.354378618056411e-06, + "loss": 0.3321, + "step": 44586 + }, + { + "epoch": 16.46, + "learning_rate": 1.3543416676930656e-06, + "loss": 0.3259, + "step": 44587 + }, + { + "epoch": 16.47, + "learning_rate": 1.3543047173297204e-06, + "loss": 0.4292, + "step": 44588 + }, + { + "epoch": 16.47, + "learning_rate": 1.3542677669663753e-06, + "loss": 0.2538, + "step": 44589 + }, + { + "epoch": 16.47, + "learning_rate": 1.35423081660303e-06, + "loss": 0.2482, + "step": 44590 + }, + { + "epoch": 16.47, + "learning_rate": 1.3541938662396848e-06, + "loss": 0.2869, + "step": 44591 + }, + { + "epoch": 16.47, + "learning_rate": 1.3541569158763396e-06, + "loss": 0.2174, + "step": 44592 + }, + { + "epoch": 16.47, + "learning_rate": 1.3541199655129943e-06, + "loss": 0.3536, + "step": 44593 + }, + { + "epoch": 16.47, + "learning_rate": 1.354083015149649e-06, + "loss": 0.2491, + "step": 44594 + }, + { + "epoch": 16.47, + "learning_rate": 1.3540460647863037e-06, + "loss": 0.2092, + "step": 44595 + }, + { + "epoch": 16.47, + "learning_rate": 1.3540091144229586e-06, + "loss": 0.2952, + "step": 44596 + }, + { + "epoch": 16.47, + "learning_rate": 1.3539721640596132e-06, + "loss": 0.221, + "step": 44597 + }, + { + "epoch": 16.47, + "learning_rate": 1.353935213696268e-06, + "loss": 0.3225, + "step": 44598 + }, + { + "epoch": 16.47, + "learning_rate": 1.353898263332923e-06, + "loss": 0.233, + "step": 44599 + }, + { + "epoch": 16.47, + "learning_rate": 1.3538613129695776e-06, + "loss": 0.4566, + "step": 44600 + }, + { + "epoch": 16.47, + "learning_rate": 1.3538243626062324e-06, + "loss": 0.2009, + "step": 44601 + }, + { + "epoch": 16.47, + "learning_rate": 1.353787412242887e-06, + "loss": 0.2143, + "step": 44602 + }, + { + "epoch": 16.47, + "learning_rate": 1.3537504618795417e-06, + "loss": 0.2302, + "step": 44603 + }, + { + "epoch": 16.47, + "learning_rate": 1.3537135115161965e-06, + "loss": 0.1359, + "step": 44604 + }, + { + "epoch": 16.47, + "learning_rate": 1.3536765611528514e-06, + "loss": 0.1866, + "step": 44605 + }, + { + "epoch": 16.47, + "learning_rate": 1.353639610789506e-06, + "loss": 0.2883, + "step": 44606 + }, + { + "epoch": 16.47, + "learning_rate": 1.3536026604261609e-06, + "loss": 0.3123, + "step": 44607 + }, + { + "epoch": 16.47, + "learning_rate": 1.3535657100628157e-06, + "loss": 0.2478, + "step": 44608 + }, + { + "epoch": 16.47, + "learning_rate": 1.3535287596994704e-06, + "loss": 0.2362, + "step": 44609 + }, + { + "epoch": 16.47, + "learning_rate": 1.3534918093361252e-06, + "loss": 0.2528, + "step": 44610 + }, + { + "epoch": 16.47, + "learning_rate": 1.35345485897278e-06, + "loss": 0.2211, + "step": 44611 + }, + { + "epoch": 16.47, + "learning_rate": 1.3534179086094347e-06, + "loss": 0.2305, + "step": 44612 + }, + { + "epoch": 16.47, + "learning_rate": 1.3533809582460893e-06, + "loss": 0.3573, + "step": 44613 + }, + { + "epoch": 16.47, + "learning_rate": 1.3533440078827442e-06, + "loss": 0.2009, + "step": 44614 + }, + { + "epoch": 16.48, + "learning_rate": 1.353307057519399e-06, + "loss": 0.2106, + "step": 44615 + }, + { + "epoch": 16.48, + "learning_rate": 1.3532701071560537e-06, + "loss": 0.2329, + "step": 44616 + }, + { + "epoch": 16.48, + "learning_rate": 1.3532331567927085e-06, + "loss": 0.2048, + "step": 44617 + }, + { + "epoch": 16.48, + "learning_rate": 1.3531962064293634e-06, + "loss": 0.3796, + "step": 44618 + }, + { + "epoch": 16.48, + "learning_rate": 1.353159256066018e-06, + "loss": 0.1936, + "step": 44619 + }, + { + "epoch": 16.48, + "learning_rate": 1.3531223057026728e-06, + "loss": 0.3637, + "step": 44620 + }, + { + "epoch": 16.48, + "learning_rate": 1.3530853553393277e-06, + "loss": 0.2643, + "step": 44621 + }, + { + "epoch": 16.48, + "learning_rate": 1.3530484049759823e-06, + "loss": 0.2244, + "step": 44622 + }, + { + "epoch": 16.48, + "learning_rate": 1.353011454612637e-06, + "loss": 0.241, + "step": 44623 + }, + { + "epoch": 16.48, + "learning_rate": 1.3529745042492918e-06, + "loss": 0.1791, + "step": 44624 + }, + { + "epoch": 16.48, + "learning_rate": 1.3529375538859464e-06, + "loss": 0.2508, + "step": 44625 + }, + { + "epoch": 16.48, + "learning_rate": 1.3529006035226013e-06, + "loss": 0.3117, + "step": 44626 + }, + { + "epoch": 16.48, + "learning_rate": 1.3528636531592561e-06, + "loss": 0.283, + "step": 44627 + }, + { + "epoch": 16.48, + "learning_rate": 1.3528267027959108e-06, + "loss": 0.1456, + "step": 44628 + }, + { + "epoch": 16.48, + "learning_rate": 1.3527897524325656e-06, + "loss": 0.1389, + "step": 44629 + }, + { + "epoch": 16.48, + "learning_rate": 1.3527528020692205e-06, + "loss": 0.1203, + "step": 44630 + }, + { + "epoch": 16.48, + "learning_rate": 1.3527158517058751e-06, + "loss": 0.1634, + "step": 44631 + }, + { + "epoch": 16.48, + "learning_rate": 1.35267890134253e-06, + "loss": 0.2848, + "step": 44632 + }, + { + "epoch": 16.48, + "learning_rate": 1.3526419509791846e-06, + "loss": 0.2316, + "step": 44633 + }, + { + "epoch": 16.48, + "learning_rate": 1.3526050006158394e-06, + "loss": 0.2369, + "step": 44634 + }, + { + "epoch": 16.48, + "learning_rate": 1.352568050252494e-06, + "loss": 0.3793, + "step": 44635 + }, + { + "epoch": 16.48, + "learning_rate": 1.352531099889149e-06, + "loss": 0.3586, + "step": 44636 + }, + { + "epoch": 16.48, + "learning_rate": 1.3524941495258038e-06, + "loss": 0.1899, + "step": 44637 + }, + { + "epoch": 16.48, + "learning_rate": 1.3524571991624584e-06, + "loss": 0.2197, + "step": 44638 + }, + { + "epoch": 16.48, + "learning_rate": 1.3524202487991133e-06, + "loss": 0.2088, + "step": 44639 + }, + { + "epoch": 16.48, + "learning_rate": 1.3523832984357681e-06, + "loss": 0.2269, + "step": 44640 + }, + { + "epoch": 16.48, + "learning_rate": 1.3523463480724227e-06, + "loss": 0.3118, + "step": 44641 + }, + { + "epoch": 16.49, + "learning_rate": 1.3523093977090776e-06, + "loss": 0.247, + "step": 44642 + }, + { + "epoch": 16.49, + "learning_rate": 1.3522724473457322e-06, + "loss": 0.2951, + "step": 44643 + }, + { + "epoch": 16.49, + "learning_rate": 1.3522354969823869e-06, + "loss": 0.1609, + "step": 44644 + }, + { + "epoch": 16.49, + "learning_rate": 1.3521985466190417e-06, + "loss": 0.4081, + "step": 44645 + }, + { + "epoch": 16.49, + "learning_rate": 1.3521615962556966e-06, + "loss": 0.2672, + "step": 44646 + }, + { + "epoch": 16.49, + "learning_rate": 1.3521246458923512e-06, + "loss": 0.2367, + "step": 44647 + }, + { + "epoch": 16.49, + "learning_rate": 1.352087695529006e-06, + "loss": 0.1942, + "step": 44648 + }, + { + "epoch": 16.49, + "learning_rate": 1.3520507451656609e-06, + "loss": 0.266, + "step": 44649 + }, + { + "epoch": 16.49, + "learning_rate": 1.3520137948023155e-06, + "loss": 0.1805, + "step": 44650 + }, + { + "epoch": 16.49, + "learning_rate": 1.3519768444389704e-06, + "loss": 0.3427, + "step": 44651 + }, + { + "epoch": 16.49, + "learning_rate": 1.3519398940756252e-06, + "loss": 0.1277, + "step": 44652 + }, + { + "epoch": 16.49, + "learning_rate": 1.3519029437122799e-06, + "loss": 0.1627, + "step": 44653 + }, + { + "epoch": 16.49, + "learning_rate": 1.3518659933489345e-06, + "loss": 0.3646, + "step": 44654 + }, + { + "epoch": 16.49, + "learning_rate": 1.3518290429855893e-06, + "loss": 0.1822, + "step": 44655 + }, + { + "epoch": 16.49, + "learning_rate": 1.3517920926222442e-06, + "loss": 0.2886, + "step": 44656 + }, + { + "epoch": 16.49, + "learning_rate": 1.3517551422588988e-06, + "loss": 0.1958, + "step": 44657 + }, + { + "epoch": 16.49, + "learning_rate": 1.3517181918955537e-06, + "loss": 0.3446, + "step": 44658 + }, + { + "epoch": 16.49, + "learning_rate": 1.3516812415322085e-06, + "loss": 0.2227, + "step": 44659 + }, + { + "epoch": 16.49, + "learning_rate": 1.3516442911688632e-06, + "loss": 0.2163, + "step": 44660 + }, + { + "epoch": 16.49, + "learning_rate": 1.351607340805518e-06, + "loss": 0.1367, + "step": 44661 + }, + { + "epoch": 16.49, + "learning_rate": 1.3515703904421729e-06, + "loss": 0.2068, + "step": 44662 + }, + { + "epoch": 16.49, + "learning_rate": 1.3515334400788275e-06, + "loss": 0.305, + "step": 44663 + }, + { + "epoch": 16.49, + "learning_rate": 1.3514964897154821e-06, + "loss": 0.2594, + "step": 44664 + }, + { + "epoch": 16.49, + "learning_rate": 1.351459539352137e-06, + "loss": 0.3042, + "step": 44665 + }, + { + "epoch": 16.49, + "learning_rate": 1.3514225889887918e-06, + "loss": 0.2205, + "step": 44666 + }, + { + "epoch": 16.49, + "learning_rate": 1.3513856386254465e-06, + "loss": 0.2633, + "step": 44667 + }, + { + "epoch": 16.49, + "learning_rate": 1.3513486882621013e-06, + "loss": 0.1186, + "step": 44668 + }, + { + "epoch": 16.5, + "learning_rate": 1.3513117378987562e-06, + "loss": 0.1778, + "step": 44669 + }, + { + "epoch": 16.5, + "learning_rate": 1.3512747875354108e-06, + "loss": 0.2395, + "step": 44670 + }, + { + "epoch": 16.5, + "learning_rate": 1.3512378371720656e-06, + "loss": 0.1676, + "step": 44671 + }, + { + "epoch": 16.5, + "learning_rate": 1.3512008868087205e-06, + "loss": 0.211, + "step": 44672 + }, + { + "epoch": 16.5, + "learning_rate": 1.351163936445375e-06, + "loss": 0.4726, + "step": 44673 + }, + { + "epoch": 16.5, + "learning_rate": 1.3511269860820298e-06, + "loss": 0.1735, + "step": 44674 + }, + { + "epoch": 16.5, + "learning_rate": 1.3510900357186846e-06, + "loss": 0.3074, + "step": 44675 + }, + { + "epoch": 16.5, + "learning_rate": 1.3510530853553393e-06, + "loss": 0.2876, + "step": 44676 + }, + { + "epoch": 16.5, + "learning_rate": 1.351016134991994e-06, + "loss": 0.185, + "step": 44677 + }, + { + "epoch": 16.5, + "learning_rate": 1.350979184628649e-06, + "loss": 0.2379, + "step": 44678 + }, + { + "epoch": 16.5, + "learning_rate": 1.3509422342653036e-06, + "loss": 0.3357, + "step": 44679 + }, + { + "epoch": 16.5, + "learning_rate": 1.3509052839019584e-06, + "loss": 0.12, + "step": 44680 + }, + { + "epoch": 16.5, + "learning_rate": 1.3508683335386133e-06, + "loss": 0.2893, + "step": 44681 + }, + { + "epoch": 16.5, + "learning_rate": 1.350831383175268e-06, + "loss": 0.2175, + "step": 44682 + }, + { + "epoch": 16.5, + "learning_rate": 1.3507944328119226e-06, + "loss": 0.208, + "step": 44683 + }, + { + "epoch": 16.5, + "learning_rate": 1.3507574824485774e-06, + "loss": 0.2858, + "step": 44684 + }, + { + "epoch": 16.5, + "learning_rate": 1.3507205320852323e-06, + "loss": 0.2399, + "step": 44685 + }, + { + "epoch": 16.5, + "learning_rate": 1.3506835817218869e-06, + "loss": 0.1268, + "step": 44686 + }, + { + "epoch": 16.5, + "learning_rate": 1.3506466313585417e-06, + "loss": 0.3956, + "step": 44687 + }, + { + "epoch": 16.5, + "learning_rate": 1.3506096809951966e-06, + "loss": 0.2003, + "step": 44688 + }, + { + "epoch": 16.5, + "learning_rate": 1.3505727306318512e-06, + "loss": 0.2307, + "step": 44689 + }, + { + "epoch": 16.5, + "learning_rate": 1.350535780268506e-06, + "loss": 0.1241, + "step": 44690 + }, + { + "epoch": 16.5, + "learning_rate": 1.350498829905161e-06, + "loss": 0.32, + "step": 44691 + }, + { + "epoch": 16.5, + "learning_rate": 1.3504618795418156e-06, + "loss": 0.2239, + "step": 44692 + }, + { + "epoch": 16.5, + "learning_rate": 1.3504249291784702e-06, + "loss": 0.2783, + "step": 44693 + }, + { + "epoch": 16.5, + "learning_rate": 1.350387978815125e-06, + "loss": 0.3916, + "step": 44694 + }, + { + "epoch": 16.5, + "learning_rate": 1.3503510284517797e-06, + "loss": 0.251, + "step": 44695 + }, + { + "epoch": 16.51, + "learning_rate": 1.3503140780884345e-06, + "loss": 0.2099, + "step": 44696 + }, + { + "epoch": 16.51, + "learning_rate": 1.3502771277250894e-06, + "loss": 0.2425, + "step": 44697 + }, + { + "epoch": 16.51, + "learning_rate": 1.350240177361744e-06, + "loss": 0.1689, + "step": 44698 + }, + { + "epoch": 16.51, + "learning_rate": 1.3502032269983989e-06, + "loss": 0.33, + "step": 44699 + }, + { + "epoch": 16.51, + "learning_rate": 1.3501662766350537e-06, + "loss": 0.2219, + "step": 44700 + }, + { + "epoch": 16.51, + "learning_rate": 1.3501293262717083e-06, + "loss": 0.2408, + "step": 44701 + }, + { + "epoch": 16.51, + "learning_rate": 1.3500923759083632e-06, + "loss": 0.1526, + "step": 44702 + }, + { + "epoch": 16.51, + "learning_rate": 1.3500554255450178e-06, + "loss": 0.1664, + "step": 44703 + }, + { + "epoch": 16.51, + "learning_rate": 1.3500184751816727e-06, + "loss": 0.2661, + "step": 44704 + }, + { + "epoch": 16.51, + "learning_rate": 1.3499815248183273e-06, + "loss": 0.2777, + "step": 44705 + }, + { + "epoch": 16.51, + "learning_rate": 1.3499445744549822e-06, + "loss": 0.3311, + "step": 44706 + }, + { + "epoch": 16.51, + "learning_rate": 1.349907624091637e-06, + "loss": 0.2304, + "step": 44707 + }, + { + "epoch": 16.51, + "learning_rate": 1.3498706737282916e-06, + "loss": 0.1835, + "step": 44708 + }, + { + "epoch": 16.51, + "learning_rate": 1.3498337233649465e-06, + "loss": 0.2384, + "step": 44709 + }, + { + "epoch": 16.51, + "learning_rate": 1.3497967730016013e-06, + "loss": 0.329, + "step": 44710 + }, + { + "epoch": 16.51, + "learning_rate": 1.349759822638256e-06, + "loss": 0.2045, + "step": 44711 + }, + { + "epoch": 16.51, + "learning_rate": 1.3497228722749108e-06, + "loss": 0.2563, + "step": 44712 + }, + { + "epoch": 16.51, + "learning_rate": 1.3496859219115655e-06, + "loss": 0.1624, + "step": 44713 + }, + { + "epoch": 16.51, + "learning_rate": 1.34964897154822e-06, + "loss": 0.2242, + "step": 44714 + }, + { + "epoch": 16.51, + "learning_rate": 1.349612021184875e-06, + "loss": 0.4076, + "step": 44715 + }, + { + "epoch": 16.51, + "learning_rate": 1.3495750708215298e-06, + "loss": 0.1718, + "step": 44716 + }, + { + "epoch": 16.51, + "learning_rate": 1.3495381204581844e-06, + "loss": 0.3326, + "step": 44717 + }, + { + "epoch": 16.51, + "learning_rate": 1.3495011700948393e-06, + "loss": 0.1754, + "step": 44718 + }, + { + "epoch": 16.51, + "learning_rate": 1.3494642197314941e-06, + "loss": 0.2284, + "step": 44719 + }, + { + "epoch": 16.51, + "learning_rate": 1.3494272693681488e-06, + "loss": 0.2292, + "step": 44720 + }, + { + "epoch": 16.51, + "learning_rate": 1.3493903190048036e-06, + "loss": 0.1942, + "step": 44721 + }, + { + "epoch": 16.51, + "learning_rate": 1.3493533686414585e-06, + "loss": 0.1959, + "step": 44722 + }, + { + "epoch": 16.52, + "learning_rate": 1.349316418278113e-06, + "loss": 0.2011, + "step": 44723 + }, + { + "epoch": 16.52, + "learning_rate": 1.3492794679147677e-06, + "loss": 0.2649, + "step": 44724 + }, + { + "epoch": 16.52, + "learning_rate": 1.3492425175514226e-06, + "loss": 0.1318, + "step": 44725 + }, + { + "epoch": 16.52, + "learning_rate": 1.3492055671880774e-06, + "loss": 0.2491, + "step": 44726 + }, + { + "epoch": 16.52, + "learning_rate": 1.349168616824732e-06, + "loss": 0.2512, + "step": 44727 + }, + { + "epoch": 16.52, + "learning_rate": 1.349131666461387e-06, + "loss": 0.2131, + "step": 44728 + }, + { + "epoch": 16.52, + "learning_rate": 1.3490947160980418e-06, + "loss": 0.1798, + "step": 44729 + }, + { + "epoch": 16.52, + "learning_rate": 1.3490577657346964e-06, + "loss": 0.2154, + "step": 44730 + }, + { + "epoch": 16.52, + "learning_rate": 1.3490208153713512e-06, + "loss": 0.214, + "step": 44731 + }, + { + "epoch": 16.52, + "learning_rate": 1.348983865008006e-06, + "loss": 0.2072, + "step": 44732 + }, + { + "epoch": 16.52, + "learning_rate": 1.3489469146446607e-06, + "loss": 0.2631, + "step": 44733 + }, + { + "epoch": 16.52, + "learning_rate": 1.3489099642813154e-06, + "loss": 0.2324, + "step": 44734 + }, + { + "epoch": 16.52, + "learning_rate": 1.3488730139179702e-06, + "loss": 0.3311, + "step": 44735 + }, + { + "epoch": 16.52, + "learning_rate": 1.348836063554625e-06, + "loss": 0.2781, + "step": 44736 + }, + { + "epoch": 16.52, + "learning_rate": 1.3487991131912797e-06, + "loss": 0.2241, + "step": 44737 + }, + { + "epoch": 16.52, + "learning_rate": 1.3487621628279345e-06, + "loss": 0.1772, + "step": 44738 + }, + { + "epoch": 16.52, + "learning_rate": 1.3487252124645894e-06, + "loss": 0.202, + "step": 44739 + }, + { + "epoch": 16.52, + "learning_rate": 1.348688262101244e-06, + "loss": 0.3158, + "step": 44740 + }, + { + "epoch": 16.52, + "learning_rate": 1.3486513117378989e-06, + "loss": 0.1616, + "step": 44741 + }, + { + "epoch": 16.52, + "learning_rate": 1.3486143613745537e-06, + "loss": 0.2143, + "step": 44742 + }, + { + "epoch": 16.52, + "learning_rate": 1.3485774110112082e-06, + "loss": 0.2335, + "step": 44743 + }, + { + "epoch": 16.52, + "learning_rate": 1.348540460647863e-06, + "loss": 0.2869, + "step": 44744 + }, + { + "epoch": 16.52, + "learning_rate": 1.3485035102845178e-06, + "loss": 0.3956, + "step": 44745 + }, + { + "epoch": 16.52, + "learning_rate": 1.3484665599211725e-06, + "loss": 0.2322, + "step": 44746 + }, + { + "epoch": 16.52, + "learning_rate": 1.3484296095578273e-06, + "loss": 0.1438, + "step": 44747 + }, + { + "epoch": 16.52, + "learning_rate": 1.3483926591944822e-06, + "loss": 0.2212, + "step": 44748 + }, + { + "epoch": 16.52, + "learning_rate": 1.3483557088311368e-06, + "loss": 0.287, + "step": 44749 + }, + { + "epoch": 16.53, + "learning_rate": 1.3483187584677917e-06, + "loss": 0.1566, + "step": 44750 + }, + { + "epoch": 16.53, + "learning_rate": 1.3482818081044465e-06, + "loss": 0.3327, + "step": 44751 + }, + { + "epoch": 16.53, + "learning_rate": 1.3482448577411012e-06, + "loss": 0.2167, + "step": 44752 + }, + { + "epoch": 16.53, + "learning_rate": 1.3482079073777558e-06, + "loss": 0.2649, + "step": 44753 + }, + { + "epoch": 16.53, + "learning_rate": 1.3481709570144106e-06, + "loss": 0.2183, + "step": 44754 + }, + { + "epoch": 16.53, + "learning_rate": 1.3481340066510655e-06, + "loss": 0.4149, + "step": 44755 + }, + { + "epoch": 16.53, + "learning_rate": 1.3480970562877201e-06, + "loss": 0.3035, + "step": 44756 + }, + { + "epoch": 16.53, + "learning_rate": 1.348060105924375e-06, + "loss": 0.3706, + "step": 44757 + }, + { + "epoch": 16.53, + "learning_rate": 1.3480231555610298e-06, + "loss": 0.3413, + "step": 44758 + }, + { + "epoch": 16.53, + "learning_rate": 1.3479862051976845e-06, + "loss": 0.3408, + "step": 44759 + }, + { + "epoch": 16.53, + "learning_rate": 1.3479492548343393e-06, + "loss": 0.197, + "step": 44760 + }, + { + "epoch": 16.53, + "learning_rate": 1.3479123044709942e-06, + "loss": 0.1961, + "step": 44761 + }, + { + "epoch": 16.53, + "learning_rate": 1.3478753541076488e-06, + "loss": 0.2283, + "step": 44762 + }, + { + "epoch": 16.53, + "learning_rate": 1.3478384037443034e-06, + "loss": 0.2758, + "step": 44763 + }, + { + "epoch": 16.53, + "learning_rate": 1.3478014533809583e-06, + "loss": 0.3174, + "step": 44764 + }, + { + "epoch": 16.53, + "learning_rate": 1.347764503017613e-06, + "loss": 0.2412, + "step": 44765 + }, + { + "epoch": 16.53, + "learning_rate": 1.3477275526542678e-06, + "loss": 0.1263, + "step": 44766 + }, + { + "epoch": 16.53, + "learning_rate": 1.3476906022909226e-06, + "loss": 0.2582, + "step": 44767 + }, + { + "epoch": 16.53, + "learning_rate": 1.3476536519275772e-06, + "loss": 0.3397, + "step": 44768 + }, + { + "epoch": 16.53, + "learning_rate": 1.347616701564232e-06, + "loss": 0.2853, + "step": 44769 + }, + { + "epoch": 16.53, + "learning_rate": 1.347579751200887e-06, + "loss": 0.2231, + "step": 44770 + }, + { + "epoch": 16.53, + "learning_rate": 1.3475428008375416e-06, + "loss": 0.2802, + "step": 44771 + }, + { + "epoch": 16.53, + "learning_rate": 1.3475058504741964e-06, + "loss": 0.198, + "step": 44772 + }, + { + "epoch": 16.53, + "learning_rate": 1.347468900110851e-06, + "loss": 0.2341, + "step": 44773 + }, + { + "epoch": 16.53, + "learning_rate": 1.347431949747506e-06, + "loss": 0.193, + "step": 44774 + }, + { + "epoch": 16.53, + "learning_rate": 1.3473949993841605e-06, + "loss": 0.1107, + "step": 44775 + }, + { + "epoch": 16.53, + "learning_rate": 1.3473580490208154e-06, + "loss": 0.224, + "step": 44776 + }, + { + "epoch": 16.54, + "learning_rate": 1.3473210986574702e-06, + "loss": 0.3182, + "step": 44777 + }, + { + "epoch": 16.54, + "learning_rate": 1.3472841482941249e-06, + "loss": 0.2844, + "step": 44778 + }, + { + "epoch": 16.54, + "learning_rate": 1.3472471979307797e-06, + "loss": 0.338, + "step": 44779 + }, + { + "epoch": 16.54, + "learning_rate": 1.3472102475674346e-06, + "loss": 0.15, + "step": 44780 + }, + { + "epoch": 16.54, + "learning_rate": 1.3471732972040892e-06, + "loss": 0.3635, + "step": 44781 + }, + { + "epoch": 16.54, + "learning_rate": 1.347136346840744e-06, + "loss": 0.2833, + "step": 44782 + }, + { + "epoch": 16.54, + "learning_rate": 1.3470993964773987e-06, + "loss": 0.2463, + "step": 44783 + }, + { + "epoch": 16.54, + "learning_rate": 1.3470624461140535e-06, + "loss": 0.1954, + "step": 44784 + }, + { + "epoch": 16.54, + "learning_rate": 1.3470254957507082e-06, + "loss": 0.3637, + "step": 44785 + }, + { + "epoch": 16.54, + "learning_rate": 1.346988545387363e-06, + "loss": 0.1628, + "step": 44786 + }, + { + "epoch": 16.54, + "learning_rate": 1.3469515950240177e-06, + "loss": 0.1515, + "step": 44787 + }, + { + "epoch": 16.54, + "learning_rate": 1.3469146446606725e-06, + "loss": 0.2078, + "step": 44788 + }, + { + "epoch": 16.54, + "learning_rate": 1.3468776942973274e-06, + "loss": 0.1945, + "step": 44789 + }, + { + "epoch": 16.54, + "learning_rate": 1.346840743933982e-06, + "loss": 0.2038, + "step": 44790 + }, + { + "epoch": 16.54, + "learning_rate": 1.3468037935706368e-06, + "loss": 0.3036, + "step": 44791 + }, + { + "epoch": 16.54, + "learning_rate": 1.3467668432072917e-06, + "loss": 0.3456, + "step": 44792 + }, + { + "epoch": 16.54, + "learning_rate": 1.3467298928439463e-06, + "loss": 0.155, + "step": 44793 + }, + { + "epoch": 16.54, + "learning_rate": 1.346692942480601e-06, + "loss": 0.1488, + "step": 44794 + }, + { + "epoch": 16.54, + "learning_rate": 1.3466559921172558e-06, + "loss": 0.2097, + "step": 44795 + }, + { + "epoch": 16.54, + "learning_rate": 1.3466190417539107e-06, + "loss": 0.2706, + "step": 44796 + }, + { + "epoch": 16.54, + "learning_rate": 1.3465820913905653e-06, + "loss": 0.265, + "step": 44797 + }, + { + "epoch": 16.54, + "learning_rate": 1.3465451410272201e-06, + "loss": 0.2179, + "step": 44798 + }, + { + "epoch": 16.54, + "learning_rate": 1.346508190663875e-06, + "loss": 0.2492, + "step": 44799 + }, + { + "epoch": 16.54, + "learning_rate": 1.3464712403005296e-06, + "loss": 0.3049, + "step": 44800 + }, + { + "epoch": 16.54, + "learning_rate": 1.3464342899371845e-06, + "loss": 0.3006, + "step": 44801 + }, + { + "epoch": 16.54, + "learning_rate": 1.3463973395738393e-06, + "loss": 0.186, + "step": 44802 + }, + { + "epoch": 16.54, + "learning_rate": 1.346360389210494e-06, + "loss": 0.1763, + "step": 44803 + }, + { + "epoch": 16.55, + "learning_rate": 1.3463234388471486e-06, + "loss": 0.1877, + "step": 44804 + }, + { + "epoch": 16.55, + "learning_rate": 1.3462864884838034e-06, + "loss": 0.1367, + "step": 44805 + }, + { + "epoch": 16.55, + "learning_rate": 1.3462495381204583e-06, + "loss": 0.2249, + "step": 44806 + }, + { + "epoch": 16.55, + "learning_rate": 1.346212587757113e-06, + "loss": 0.3228, + "step": 44807 + }, + { + "epoch": 16.55, + "learning_rate": 1.3461756373937678e-06, + "loss": 0.2177, + "step": 44808 + }, + { + "epoch": 16.55, + "learning_rate": 1.3461386870304226e-06, + "loss": 0.1978, + "step": 44809 + }, + { + "epoch": 16.55, + "learning_rate": 1.3461017366670773e-06, + "loss": 0.2852, + "step": 44810 + }, + { + "epoch": 16.55, + "learning_rate": 1.3460647863037321e-06, + "loss": 0.3095, + "step": 44811 + }, + { + "epoch": 16.55, + "learning_rate": 1.346027835940387e-06, + "loss": 0.1643, + "step": 44812 + }, + { + "epoch": 16.55, + "learning_rate": 1.3459908855770414e-06, + "loss": 0.3077, + "step": 44813 + }, + { + "epoch": 16.55, + "learning_rate": 1.3459539352136962e-06, + "loss": 0.4036, + "step": 44814 + }, + { + "epoch": 16.55, + "learning_rate": 1.345916984850351e-06, + "loss": 0.3506, + "step": 44815 + }, + { + "epoch": 16.55, + "learning_rate": 1.3458800344870057e-06, + "loss": 0.2165, + "step": 44816 + }, + { + "epoch": 16.55, + "learning_rate": 1.3458430841236606e-06, + "loss": 0.2542, + "step": 44817 + }, + { + "epoch": 16.55, + "learning_rate": 1.3458061337603154e-06, + "loss": 0.1966, + "step": 44818 + }, + { + "epoch": 16.55, + "learning_rate": 1.34576918339697e-06, + "loss": 0.3417, + "step": 44819 + }, + { + "epoch": 16.55, + "learning_rate": 1.345732233033625e-06, + "loss": 0.1873, + "step": 44820 + }, + { + "epoch": 16.55, + "learning_rate": 1.3456952826702797e-06, + "loss": 0.2327, + "step": 44821 + }, + { + "epoch": 16.55, + "learning_rate": 1.3456583323069344e-06, + "loss": 0.2762, + "step": 44822 + }, + { + "epoch": 16.55, + "learning_rate": 1.345621381943589e-06, + "loss": 0.2139, + "step": 44823 + }, + { + "epoch": 16.55, + "learning_rate": 1.3455844315802439e-06, + "loss": 0.1504, + "step": 44824 + }, + { + "epoch": 16.55, + "learning_rate": 1.3455474812168987e-06, + "loss": 0.1677, + "step": 44825 + }, + { + "epoch": 16.55, + "learning_rate": 1.3455105308535534e-06, + "loss": 0.3241, + "step": 44826 + }, + { + "epoch": 16.55, + "learning_rate": 1.3454735804902082e-06, + "loss": 0.1984, + "step": 44827 + }, + { + "epoch": 16.55, + "learning_rate": 1.345436630126863e-06, + "loss": 0.297, + "step": 44828 + }, + { + "epoch": 16.55, + "learning_rate": 1.3453996797635177e-06, + "loss": 0.1816, + "step": 44829 + }, + { + "epoch": 16.55, + "learning_rate": 1.3453627294001725e-06, + "loss": 0.1824, + "step": 44830 + }, + { + "epoch": 16.56, + "learning_rate": 1.3453257790368274e-06, + "loss": 0.3098, + "step": 44831 + }, + { + "epoch": 16.56, + "learning_rate": 1.3452888286734818e-06, + "loss": 0.2939, + "step": 44832 + }, + { + "epoch": 16.56, + "learning_rate": 1.3452518783101367e-06, + "loss": 0.2059, + "step": 44833 + }, + { + "epoch": 16.56, + "learning_rate": 1.3452149279467915e-06, + "loss": 0.1278, + "step": 44834 + }, + { + "epoch": 16.56, + "learning_rate": 1.3451779775834461e-06, + "loss": 0.2027, + "step": 44835 + }, + { + "epoch": 16.56, + "learning_rate": 1.345141027220101e-06, + "loss": 0.2431, + "step": 44836 + }, + { + "epoch": 16.56, + "learning_rate": 1.3451040768567558e-06, + "loss": 0.2139, + "step": 44837 + }, + { + "epoch": 16.56, + "learning_rate": 1.3450671264934105e-06, + "loss": 0.1777, + "step": 44838 + }, + { + "epoch": 16.56, + "learning_rate": 1.3450301761300653e-06, + "loss": 0.1995, + "step": 44839 + }, + { + "epoch": 16.56, + "learning_rate": 1.3449932257667202e-06, + "loss": 0.361, + "step": 44840 + }, + { + "epoch": 16.56, + "learning_rate": 1.3449562754033748e-06, + "loss": 0.4283, + "step": 44841 + }, + { + "epoch": 16.56, + "learning_rate": 1.3449193250400294e-06, + "loss": 0.1896, + "step": 44842 + }, + { + "epoch": 16.56, + "learning_rate": 1.3448823746766843e-06, + "loss": 0.2919, + "step": 44843 + }, + { + "epoch": 16.56, + "learning_rate": 1.3448454243133391e-06, + "loss": 0.1454, + "step": 44844 + }, + { + "epoch": 16.56, + "learning_rate": 1.3448084739499938e-06, + "loss": 0.156, + "step": 44845 + }, + { + "epoch": 16.56, + "learning_rate": 1.3447715235866486e-06, + "loss": 0.2954, + "step": 44846 + }, + { + "epoch": 16.56, + "learning_rate": 1.3447345732233035e-06, + "loss": 0.3431, + "step": 44847 + }, + { + "epoch": 16.56, + "learning_rate": 1.3446976228599581e-06, + "loss": 0.2846, + "step": 44848 + }, + { + "epoch": 16.56, + "learning_rate": 1.344660672496613e-06, + "loss": 0.1319, + "step": 44849 + }, + { + "epoch": 16.56, + "learning_rate": 1.3446237221332678e-06, + "loss": 0.2512, + "step": 44850 + }, + { + "epoch": 16.56, + "learning_rate": 1.3445867717699224e-06, + "loss": 0.1543, + "step": 44851 + }, + { + "epoch": 16.56, + "learning_rate": 1.344549821406577e-06, + "loss": 0.1553, + "step": 44852 + }, + { + "epoch": 16.56, + "learning_rate": 1.344512871043232e-06, + "loss": 0.2777, + "step": 44853 + }, + { + "epoch": 16.56, + "learning_rate": 1.3444759206798868e-06, + "loss": 0.1756, + "step": 44854 + }, + { + "epoch": 16.56, + "learning_rate": 1.3444389703165414e-06, + "loss": 0.2706, + "step": 44855 + }, + { + "epoch": 16.56, + "learning_rate": 1.3444020199531963e-06, + "loss": 0.2183, + "step": 44856 + }, + { + "epoch": 16.56, + "learning_rate": 1.3443650695898511e-06, + "loss": 0.2673, + "step": 44857 + }, + { + "epoch": 16.56, + "learning_rate": 1.3443281192265057e-06, + "loss": 0.2627, + "step": 44858 + }, + { + "epoch": 16.57, + "learning_rate": 1.3442911688631606e-06, + "loss": 0.2376, + "step": 44859 + }, + { + "epoch": 16.57, + "learning_rate": 1.3442542184998154e-06, + "loss": 0.2945, + "step": 44860 + }, + { + "epoch": 16.57, + "learning_rate": 1.34421726813647e-06, + "loss": 0.2078, + "step": 44861 + }, + { + "epoch": 16.57, + "learning_rate": 1.3441803177731247e-06, + "loss": 0.1969, + "step": 44862 + }, + { + "epoch": 16.57, + "learning_rate": 1.3441433674097796e-06, + "loss": 0.4711, + "step": 44863 + }, + { + "epoch": 16.57, + "learning_rate": 1.3441064170464342e-06, + "loss": 0.3621, + "step": 44864 + }, + { + "epoch": 16.57, + "learning_rate": 1.344069466683089e-06, + "loss": 0.2007, + "step": 44865 + }, + { + "epoch": 16.57, + "learning_rate": 1.3440325163197439e-06, + "loss": 0.2553, + "step": 44866 + }, + { + "epoch": 16.57, + "learning_rate": 1.3439955659563985e-06, + "loss": 0.2509, + "step": 44867 + }, + { + "epoch": 16.57, + "learning_rate": 1.3439586155930534e-06, + "loss": 0.1502, + "step": 44868 + }, + { + "epoch": 16.57, + "learning_rate": 1.3439216652297082e-06, + "loss": 0.1582, + "step": 44869 + }, + { + "epoch": 16.57, + "learning_rate": 1.3438847148663629e-06, + "loss": 0.2006, + "step": 44870 + }, + { + "epoch": 16.57, + "learning_rate": 1.3438477645030177e-06, + "loss": 0.2453, + "step": 44871 + }, + { + "epoch": 16.57, + "learning_rate": 1.3438108141396723e-06, + "loss": 0.1178, + "step": 44872 + }, + { + "epoch": 16.57, + "learning_rate": 1.3437738637763272e-06, + "loss": 0.1973, + "step": 44873 + }, + { + "epoch": 16.57, + "learning_rate": 1.3437369134129818e-06, + "loss": 0.2031, + "step": 44874 + }, + { + "epoch": 16.57, + "learning_rate": 1.3436999630496367e-06, + "loss": 0.1711, + "step": 44875 + }, + { + "epoch": 16.57, + "learning_rate": 1.3436630126862915e-06, + "loss": 0.2058, + "step": 44876 + }, + { + "epoch": 16.57, + "learning_rate": 1.3436260623229462e-06, + "loss": 0.2509, + "step": 44877 + }, + { + "epoch": 16.57, + "learning_rate": 1.343589111959601e-06, + "loss": 0.3589, + "step": 44878 + }, + { + "epoch": 16.57, + "learning_rate": 1.3435521615962559e-06, + "loss": 0.3059, + "step": 44879 + }, + { + "epoch": 16.57, + "learning_rate": 1.3435152112329105e-06, + "loss": 0.2298, + "step": 44880 + }, + { + "epoch": 16.57, + "learning_rate": 1.3434782608695653e-06, + "loss": 0.2329, + "step": 44881 + }, + { + "epoch": 16.57, + "learning_rate": 1.34344131050622e-06, + "loss": 0.1845, + "step": 44882 + }, + { + "epoch": 16.57, + "learning_rate": 1.3434043601428746e-06, + "loss": 0.2245, + "step": 44883 + }, + { + "epoch": 16.57, + "learning_rate": 1.3433674097795295e-06, + "loss": 0.3024, + "step": 44884 + }, + { + "epoch": 16.57, + "learning_rate": 1.3433304594161843e-06, + "loss": 0.2421, + "step": 44885 + }, + { + "epoch": 16.58, + "learning_rate": 1.343293509052839e-06, + "loss": 0.2263, + "step": 44886 + }, + { + "epoch": 16.58, + "learning_rate": 1.3432565586894938e-06, + "loss": 0.2629, + "step": 44887 + }, + { + "epoch": 16.58, + "learning_rate": 1.3432196083261486e-06, + "loss": 0.1569, + "step": 44888 + }, + { + "epoch": 16.58, + "learning_rate": 1.3431826579628033e-06, + "loss": 0.3018, + "step": 44889 + }, + { + "epoch": 16.58, + "learning_rate": 1.3431457075994581e-06, + "loss": 0.2286, + "step": 44890 + }, + { + "epoch": 16.58, + "learning_rate": 1.343108757236113e-06, + "loss": 0.2274, + "step": 44891 + }, + { + "epoch": 16.58, + "learning_rate": 1.3430718068727676e-06, + "loss": 0.3547, + "step": 44892 + }, + { + "epoch": 16.58, + "learning_rate": 1.3430348565094223e-06, + "loss": 0.1699, + "step": 44893 + }, + { + "epoch": 16.58, + "learning_rate": 1.342997906146077e-06, + "loss": 0.224, + "step": 44894 + }, + { + "epoch": 16.58, + "learning_rate": 1.342960955782732e-06, + "loss": 0.1881, + "step": 44895 + }, + { + "epoch": 16.58, + "learning_rate": 1.3429240054193866e-06, + "loss": 0.1697, + "step": 44896 + }, + { + "epoch": 16.58, + "learning_rate": 1.3428870550560414e-06, + "loss": 0.3106, + "step": 44897 + }, + { + "epoch": 16.58, + "learning_rate": 1.3428501046926963e-06, + "loss": 0.4206, + "step": 44898 + }, + { + "epoch": 16.58, + "learning_rate": 1.342813154329351e-06, + "loss": 0.1863, + "step": 44899 + }, + { + "epoch": 16.58, + "learning_rate": 1.3427762039660058e-06, + "loss": 0.2534, + "step": 44900 + }, + { + "epoch": 16.58, + "learning_rate": 1.3427392536026606e-06, + "loss": 0.2199, + "step": 44901 + }, + { + "epoch": 16.58, + "learning_rate": 1.342702303239315e-06, + "loss": 0.2539, + "step": 44902 + }, + { + "epoch": 16.58, + "learning_rate": 1.3426653528759699e-06, + "loss": 0.2079, + "step": 44903 + }, + { + "epoch": 16.58, + "learning_rate": 1.3426284025126247e-06, + "loss": 0.2883, + "step": 44904 + }, + { + "epoch": 16.58, + "learning_rate": 1.3425914521492794e-06, + "loss": 0.214, + "step": 44905 + }, + { + "epoch": 16.58, + "learning_rate": 1.3425545017859342e-06, + "loss": 0.2089, + "step": 44906 + }, + { + "epoch": 16.58, + "learning_rate": 1.342517551422589e-06, + "loss": 0.1888, + "step": 44907 + }, + { + "epoch": 16.58, + "learning_rate": 1.3424806010592437e-06, + "loss": 0.1822, + "step": 44908 + }, + { + "epoch": 16.58, + "learning_rate": 1.3424436506958986e-06, + "loss": 0.0923, + "step": 44909 + }, + { + "epoch": 16.58, + "learning_rate": 1.3424067003325534e-06, + "loss": 0.3542, + "step": 44910 + }, + { + "epoch": 16.58, + "learning_rate": 1.342369749969208e-06, + "loss": 0.2538, + "step": 44911 + }, + { + "epoch": 16.58, + "learning_rate": 1.3423327996058627e-06, + "loss": 0.2462, + "step": 44912 + }, + { + "epoch": 16.59, + "learning_rate": 1.3422958492425175e-06, + "loss": 0.1873, + "step": 44913 + }, + { + "epoch": 16.59, + "learning_rate": 1.3422588988791724e-06, + "loss": 0.3098, + "step": 44914 + }, + { + "epoch": 16.59, + "learning_rate": 1.342221948515827e-06, + "loss": 0.3336, + "step": 44915 + }, + { + "epoch": 16.59, + "learning_rate": 1.3421849981524819e-06, + "loss": 0.2195, + "step": 44916 + }, + { + "epoch": 16.59, + "learning_rate": 1.3421480477891367e-06, + "loss": 0.1553, + "step": 44917 + }, + { + "epoch": 16.59, + "learning_rate": 1.3421110974257913e-06, + "loss": 0.3174, + "step": 44918 + }, + { + "epoch": 16.59, + "learning_rate": 1.3420741470624462e-06, + "loss": 0.3803, + "step": 44919 + }, + { + "epoch": 16.59, + "learning_rate": 1.342037196699101e-06, + "loss": 0.2631, + "step": 44920 + }, + { + "epoch": 16.59, + "learning_rate": 1.3420002463357557e-06, + "loss": 0.2673, + "step": 44921 + }, + { + "epoch": 16.59, + "learning_rate": 1.3419632959724103e-06, + "loss": 0.1946, + "step": 44922 + }, + { + "epoch": 16.59, + "learning_rate": 1.3419263456090652e-06, + "loss": 0.2579, + "step": 44923 + }, + { + "epoch": 16.59, + "learning_rate": 1.34188939524572e-06, + "loss": 0.3036, + "step": 44924 + }, + { + "epoch": 16.59, + "learning_rate": 1.3418524448823746e-06, + "loss": 0.3907, + "step": 44925 + }, + { + "epoch": 16.59, + "learning_rate": 1.3418154945190295e-06, + "loss": 0.273, + "step": 44926 + }, + { + "epoch": 16.59, + "learning_rate": 1.3417785441556843e-06, + "loss": 0.2885, + "step": 44927 + }, + { + "epoch": 16.59, + "learning_rate": 1.341741593792339e-06, + "loss": 0.2808, + "step": 44928 + }, + { + "epoch": 16.59, + "learning_rate": 1.3417046434289938e-06, + "loss": 0.3247, + "step": 44929 + }, + { + "epoch": 16.59, + "learning_rate": 1.3416676930656487e-06, + "loss": 0.3175, + "step": 44930 + }, + { + "epoch": 16.59, + "learning_rate": 1.3416307427023033e-06, + "loss": 0.2184, + "step": 44931 + }, + { + "epoch": 16.59, + "learning_rate": 1.341593792338958e-06, + "loss": 0.405, + "step": 44932 + }, + { + "epoch": 16.59, + "learning_rate": 1.3415568419756128e-06, + "loss": 0.2517, + "step": 44933 + }, + { + "epoch": 16.59, + "learning_rate": 1.3415198916122674e-06, + "loss": 0.3769, + "step": 44934 + }, + { + "epoch": 16.59, + "learning_rate": 1.3414829412489223e-06, + "loss": 0.2268, + "step": 44935 + }, + { + "epoch": 16.59, + "learning_rate": 1.3414459908855771e-06, + "loss": 0.3451, + "step": 44936 + }, + { + "epoch": 16.59, + "learning_rate": 1.3414090405222318e-06, + "loss": 0.2183, + "step": 44937 + }, + { + "epoch": 16.59, + "learning_rate": 1.3413720901588866e-06, + "loss": 0.1024, + "step": 44938 + }, + { + "epoch": 16.59, + "learning_rate": 1.3413351397955415e-06, + "loss": 0.1463, + "step": 44939 + }, + { + "epoch": 16.6, + "learning_rate": 1.341298189432196e-06, + "loss": 0.1632, + "step": 44940 + }, + { + "epoch": 16.6, + "learning_rate": 1.341261239068851e-06, + "loss": 0.2113, + "step": 44941 + }, + { + "epoch": 16.6, + "learning_rate": 1.3412242887055056e-06, + "loss": 0.1354, + "step": 44942 + }, + { + "epoch": 16.6, + "learning_rate": 1.3411873383421604e-06, + "loss": 0.3226, + "step": 44943 + }, + { + "epoch": 16.6, + "learning_rate": 1.341150387978815e-06, + "loss": 0.1651, + "step": 44944 + }, + { + "epoch": 16.6, + "learning_rate": 1.34111343761547e-06, + "loss": 0.2422, + "step": 44945 + }, + { + "epoch": 16.6, + "learning_rate": 1.3410764872521248e-06, + "loss": 0.2077, + "step": 44946 + }, + { + "epoch": 16.6, + "learning_rate": 1.3410395368887794e-06, + "loss": 0.2632, + "step": 44947 + }, + { + "epoch": 16.6, + "learning_rate": 1.3410025865254342e-06, + "loss": 0.2727, + "step": 44948 + }, + { + "epoch": 16.6, + "learning_rate": 1.340965636162089e-06, + "loss": 0.1779, + "step": 44949 + }, + { + "epoch": 16.6, + "learning_rate": 1.3409286857987437e-06, + "loss": 0.305, + "step": 44950 + }, + { + "epoch": 16.6, + "learning_rate": 1.3408917354353986e-06, + "loss": 0.2701, + "step": 44951 + }, + { + "epoch": 16.6, + "learning_rate": 1.3408547850720532e-06, + "loss": 0.1953, + "step": 44952 + }, + { + "epoch": 16.6, + "learning_rate": 1.3408178347087079e-06, + "loss": 0.4405, + "step": 44953 + }, + { + "epoch": 16.6, + "learning_rate": 1.3407808843453627e-06, + "loss": 0.1792, + "step": 44954 + }, + { + "epoch": 16.6, + "learning_rate": 1.3407439339820175e-06, + "loss": 0.1389, + "step": 44955 + }, + { + "epoch": 16.6, + "learning_rate": 1.3407069836186722e-06, + "loss": 0.2731, + "step": 44956 + }, + { + "epoch": 16.6, + "learning_rate": 1.340670033255327e-06, + "loss": 0.2573, + "step": 44957 + }, + { + "epoch": 16.6, + "learning_rate": 1.3406330828919819e-06, + "loss": 0.2325, + "step": 44958 + }, + { + "epoch": 16.6, + "learning_rate": 1.3405961325286365e-06, + "loss": 0.3166, + "step": 44959 + }, + { + "epoch": 16.6, + "learning_rate": 1.3405591821652914e-06, + "loss": 0.1142, + "step": 44960 + }, + { + "epoch": 16.6, + "learning_rate": 1.3405222318019462e-06, + "loss": 0.1535, + "step": 44961 + }, + { + "epoch": 16.6, + "learning_rate": 1.3404852814386008e-06, + "loss": 0.185, + "step": 44962 + }, + { + "epoch": 16.6, + "learning_rate": 1.3404483310752555e-06, + "loss": 0.1762, + "step": 44963 + }, + { + "epoch": 16.6, + "learning_rate": 1.3404113807119103e-06, + "loss": 0.2733, + "step": 44964 + }, + { + "epoch": 16.6, + "learning_rate": 1.3403744303485652e-06, + "loss": 0.3981, + "step": 44965 + }, + { + "epoch": 16.6, + "learning_rate": 1.3403374799852198e-06, + "loss": 0.2265, + "step": 44966 + }, + { + "epoch": 16.61, + "learning_rate": 1.3403005296218747e-06, + "loss": 0.2401, + "step": 44967 + }, + { + "epoch": 16.61, + "learning_rate": 1.3402635792585295e-06, + "loss": 0.1953, + "step": 44968 + }, + { + "epoch": 16.61, + "learning_rate": 1.3402266288951842e-06, + "loss": 0.4022, + "step": 44969 + }, + { + "epoch": 16.61, + "learning_rate": 1.340189678531839e-06, + "loss": 0.258, + "step": 44970 + }, + { + "epoch": 16.61, + "learning_rate": 1.3401527281684938e-06, + "loss": 0.2589, + "step": 44971 + }, + { + "epoch": 16.61, + "learning_rate": 1.3401157778051485e-06, + "loss": 0.2654, + "step": 44972 + }, + { + "epoch": 16.61, + "learning_rate": 1.3400788274418031e-06, + "loss": 0.2936, + "step": 44973 + }, + { + "epoch": 16.61, + "learning_rate": 1.340041877078458e-06, + "loss": 0.2913, + "step": 44974 + }, + { + "epoch": 16.61, + "learning_rate": 1.3400049267151126e-06, + "loss": 0.14, + "step": 44975 + }, + { + "epoch": 16.61, + "learning_rate": 1.3399679763517675e-06, + "loss": 0.1715, + "step": 44976 + }, + { + "epoch": 16.61, + "learning_rate": 1.3399310259884223e-06, + "loss": 0.2811, + "step": 44977 + }, + { + "epoch": 16.61, + "learning_rate": 1.339894075625077e-06, + "loss": 0.1602, + "step": 44978 + }, + { + "epoch": 16.61, + "learning_rate": 1.3398571252617318e-06, + "loss": 0.1393, + "step": 44979 + }, + { + "epoch": 16.61, + "learning_rate": 1.3398201748983866e-06, + "loss": 0.1454, + "step": 44980 + }, + { + "epoch": 16.61, + "learning_rate": 1.3397832245350413e-06, + "loss": 0.2877, + "step": 44981 + }, + { + "epoch": 16.61, + "learning_rate": 1.339746274171696e-06, + "loss": 0.2821, + "step": 44982 + }, + { + "epoch": 16.61, + "learning_rate": 1.3397093238083508e-06, + "loss": 0.3483, + "step": 44983 + }, + { + "epoch": 16.61, + "learning_rate": 1.3396723734450056e-06, + "loss": 0.3142, + "step": 44984 + }, + { + "epoch": 16.61, + "learning_rate": 1.3396354230816602e-06, + "loss": 0.1628, + "step": 44985 + }, + { + "epoch": 16.61, + "learning_rate": 1.339598472718315e-06, + "loss": 0.2635, + "step": 44986 + }, + { + "epoch": 16.61, + "learning_rate": 1.33956152235497e-06, + "loss": 0.2357, + "step": 44987 + }, + { + "epoch": 16.61, + "learning_rate": 1.3395245719916246e-06, + "loss": 0.1477, + "step": 44988 + }, + { + "epoch": 16.61, + "learning_rate": 1.3394876216282794e-06, + "loss": 0.1669, + "step": 44989 + }, + { + "epoch": 16.61, + "learning_rate": 1.3394506712649343e-06, + "loss": 0.1736, + "step": 44990 + }, + { + "epoch": 16.61, + "learning_rate": 1.339413720901589e-06, + "loss": 0.2452, + "step": 44991 + }, + { + "epoch": 16.61, + "learning_rate": 1.3393767705382435e-06, + "loss": 0.1945, + "step": 44992 + }, + { + "epoch": 16.61, + "learning_rate": 1.3393398201748984e-06, + "loss": 0.2832, + "step": 44993 + }, + { + "epoch": 16.62, + "learning_rate": 1.3393028698115532e-06, + "loss": 0.3694, + "step": 44994 + }, + { + "epoch": 16.62, + "learning_rate": 1.3392659194482079e-06, + "loss": 0.254, + "step": 44995 + }, + { + "epoch": 16.62, + "learning_rate": 1.3392289690848627e-06, + "loss": 0.2072, + "step": 44996 + }, + { + "epoch": 16.62, + "learning_rate": 1.3391920187215176e-06, + "loss": 0.206, + "step": 44997 + }, + { + "epoch": 16.62, + "learning_rate": 1.3391550683581722e-06, + "loss": 0.2652, + "step": 44998 + }, + { + "epoch": 16.62, + "learning_rate": 1.339118117994827e-06, + "loss": 0.2874, + "step": 44999 + }, + { + "epoch": 16.62, + "learning_rate": 1.339081167631482e-06, + "loss": 0.2525, + "step": 45000 + }, + { + "epoch": 16.62, + "learning_rate": 1.3390442172681365e-06, + "loss": 0.3039, + "step": 45001 + }, + { + "epoch": 16.62, + "learning_rate": 1.3390072669047912e-06, + "loss": 0.2083, + "step": 45002 + }, + { + "epoch": 16.62, + "learning_rate": 1.338970316541446e-06, + "loss": 0.2817, + "step": 45003 + }, + { + "epoch": 16.62, + "learning_rate": 1.3389333661781007e-06, + "loss": 0.3114, + "step": 45004 + }, + { + "epoch": 16.62, + "learning_rate": 1.3388964158147555e-06, + "loss": 0.2121, + "step": 45005 + }, + { + "epoch": 16.62, + "learning_rate": 1.3388594654514104e-06, + "loss": 0.2943, + "step": 45006 + }, + { + "epoch": 16.62, + "learning_rate": 1.338822515088065e-06, + "loss": 0.1716, + "step": 45007 + }, + { + "epoch": 16.62, + "learning_rate": 1.3387855647247198e-06, + "loss": 0.2094, + "step": 45008 + }, + { + "epoch": 16.62, + "learning_rate": 1.3387486143613747e-06, + "loss": 0.2628, + "step": 45009 + }, + { + "epoch": 16.62, + "learning_rate": 1.3387116639980293e-06, + "loss": 0.2499, + "step": 45010 + }, + { + "epoch": 16.62, + "learning_rate": 1.3386747136346842e-06, + "loss": 0.2964, + "step": 45011 + }, + { + "epoch": 16.62, + "learning_rate": 1.3386377632713388e-06, + "loss": 0.218, + "step": 45012 + }, + { + "epoch": 16.62, + "learning_rate": 1.3386008129079937e-06, + "loss": 0.2421, + "step": 45013 + }, + { + "epoch": 16.62, + "learning_rate": 1.3385638625446483e-06, + "loss": 0.2182, + "step": 45014 + }, + { + "epoch": 16.62, + "learning_rate": 1.3385269121813031e-06, + "loss": 0.2419, + "step": 45015 + }, + { + "epoch": 16.62, + "learning_rate": 1.338489961817958e-06, + "loss": 0.3618, + "step": 45016 + }, + { + "epoch": 16.62, + "learning_rate": 1.3384530114546126e-06, + "loss": 0.274, + "step": 45017 + }, + { + "epoch": 16.62, + "learning_rate": 1.3384160610912675e-06, + "loss": 0.3992, + "step": 45018 + }, + { + "epoch": 16.62, + "learning_rate": 1.3383791107279223e-06, + "loss": 0.1844, + "step": 45019 + }, + { + "epoch": 16.62, + "learning_rate": 1.338342160364577e-06, + "loss": 0.2598, + "step": 45020 + }, + { + "epoch": 16.63, + "learning_rate": 1.3383052100012318e-06, + "loss": 0.2223, + "step": 45021 + }, + { + "epoch": 16.63, + "learning_rate": 1.3382682596378864e-06, + "loss": 0.2075, + "step": 45022 + }, + { + "epoch": 16.63, + "learning_rate": 1.338231309274541e-06, + "loss": 0.2294, + "step": 45023 + }, + { + "epoch": 16.63, + "learning_rate": 1.338194358911196e-06, + "loss": 0.3289, + "step": 45024 + }, + { + "epoch": 16.63, + "learning_rate": 1.3381574085478508e-06, + "loss": 0.2241, + "step": 45025 + }, + { + "epoch": 16.63, + "learning_rate": 1.3381204581845054e-06, + "loss": 0.2837, + "step": 45026 + }, + { + "epoch": 16.63, + "learning_rate": 1.3380835078211603e-06, + "loss": 0.1741, + "step": 45027 + }, + { + "epoch": 16.63, + "learning_rate": 1.3380465574578151e-06, + "loss": 0.1591, + "step": 45028 + }, + { + "epoch": 16.63, + "learning_rate": 1.3380096070944697e-06, + "loss": 0.1855, + "step": 45029 + }, + { + "epoch": 16.63, + "learning_rate": 1.3379726567311246e-06, + "loss": 0.2732, + "step": 45030 + }, + { + "epoch": 16.63, + "learning_rate": 1.3379357063677794e-06, + "loss": 0.1156, + "step": 45031 + }, + { + "epoch": 16.63, + "learning_rate": 1.337898756004434e-06, + "loss": 0.3148, + "step": 45032 + }, + { + "epoch": 16.63, + "learning_rate": 1.3378618056410887e-06, + "loss": 0.21, + "step": 45033 + }, + { + "epoch": 16.63, + "learning_rate": 1.3378248552777436e-06, + "loss": 0.306, + "step": 45034 + }, + { + "epoch": 16.63, + "learning_rate": 1.3377879049143984e-06, + "loss": 0.2942, + "step": 45035 + }, + { + "epoch": 16.63, + "learning_rate": 1.337750954551053e-06, + "loss": 0.2096, + "step": 45036 + }, + { + "epoch": 16.63, + "learning_rate": 1.337714004187708e-06, + "loss": 0.1618, + "step": 45037 + }, + { + "epoch": 16.63, + "learning_rate": 1.3376770538243627e-06, + "loss": 0.198, + "step": 45038 + }, + { + "epoch": 16.63, + "learning_rate": 1.3376401034610174e-06, + "loss": 0.2294, + "step": 45039 + }, + { + "epoch": 16.63, + "learning_rate": 1.3376031530976722e-06, + "loss": 0.3864, + "step": 45040 + }, + { + "epoch": 16.63, + "learning_rate": 1.337566202734327e-06, + "loss": 0.1748, + "step": 45041 + }, + { + "epoch": 16.63, + "learning_rate": 1.3375292523709817e-06, + "loss": 0.2456, + "step": 45042 + }, + { + "epoch": 16.63, + "learning_rate": 1.3374923020076364e-06, + "loss": 0.1978, + "step": 45043 + }, + { + "epoch": 16.63, + "learning_rate": 1.3374553516442912e-06, + "loss": 0.2001, + "step": 45044 + }, + { + "epoch": 16.63, + "learning_rate": 1.337418401280946e-06, + "loss": 0.2694, + "step": 45045 + }, + { + "epoch": 16.63, + "learning_rate": 1.3373814509176007e-06, + "loss": 0.1959, + "step": 45046 + }, + { + "epoch": 16.63, + "learning_rate": 1.3373445005542555e-06, + "loss": 0.236, + "step": 45047 + }, + { + "epoch": 16.64, + "learning_rate": 1.3373075501909104e-06, + "loss": 0.3147, + "step": 45048 + }, + { + "epoch": 16.64, + "learning_rate": 1.337270599827565e-06, + "loss": 0.3604, + "step": 45049 + }, + { + "epoch": 16.64, + "learning_rate": 1.3372336494642199e-06, + "loss": 0.2373, + "step": 45050 + }, + { + "epoch": 16.64, + "learning_rate": 1.3371966991008745e-06, + "loss": 0.2204, + "step": 45051 + }, + { + "epoch": 16.64, + "learning_rate": 1.3371597487375291e-06, + "loss": 0.4326, + "step": 45052 + }, + { + "epoch": 16.64, + "learning_rate": 1.337122798374184e-06, + "loss": 0.2152, + "step": 45053 + }, + { + "epoch": 16.64, + "learning_rate": 1.3370858480108388e-06, + "loss": 0.5299, + "step": 45054 + }, + { + "epoch": 16.64, + "learning_rate": 1.3370488976474935e-06, + "loss": 0.2618, + "step": 45055 + }, + { + "epoch": 16.64, + "learning_rate": 1.3370119472841483e-06, + "loss": 0.1606, + "step": 45056 + }, + { + "epoch": 16.64, + "learning_rate": 1.3369749969208032e-06, + "loss": 0.1601, + "step": 45057 + }, + { + "epoch": 16.64, + "learning_rate": 1.3369380465574578e-06, + "loss": 0.2175, + "step": 45058 + }, + { + "epoch": 16.64, + "learning_rate": 1.3369010961941127e-06, + "loss": 0.2546, + "step": 45059 + }, + { + "epoch": 16.64, + "learning_rate": 1.3368641458307675e-06, + "loss": 0.3352, + "step": 45060 + }, + { + "epoch": 16.64, + "learning_rate": 1.3368271954674221e-06, + "loss": 0.2828, + "step": 45061 + }, + { + "epoch": 16.64, + "learning_rate": 1.3367902451040768e-06, + "loss": 0.3117, + "step": 45062 + }, + { + "epoch": 16.64, + "learning_rate": 1.3367532947407316e-06, + "loss": 0.2929, + "step": 45063 + }, + { + "epoch": 16.64, + "learning_rate": 1.3367163443773865e-06, + "loss": 0.2822, + "step": 45064 + }, + { + "epoch": 16.64, + "learning_rate": 1.3366793940140411e-06, + "loss": 0.2577, + "step": 45065 + }, + { + "epoch": 16.64, + "learning_rate": 1.336642443650696e-06, + "loss": 0.2991, + "step": 45066 + }, + { + "epoch": 16.64, + "learning_rate": 1.3366054932873508e-06, + "loss": 0.1835, + "step": 45067 + }, + { + "epoch": 16.64, + "learning_rate": 1.3365685429240054e-06, + "loss": 0.1973, + "step": 45068 + }, + { + "epoch": 16.64, + "learning_rate": 1.3365315925606603e-06, + "loss": 0.137, + "step": 45069 + }, + { + "epoch": 16.64, + "learning_rate": 1.3364946421973151e-06, + "loss": 0.199, + "step": 45070 + }, + { + "epoch": 16.64, + "learning_rate": 1.3364576918339696e-06, + "loss": 0.2011, + "step": 45071 + }, + { + "epoch": 16.64, + "learning_rate": 1.3364207414706244e-06, + "loss": 0.4421, + "step": 45072 + }, + { + "epoch": 16.64, + "learning_rate": 1.3363837911072793e-06, + "loss": 0.2398, + "step": 45073 + }, + { + "epoch": 16.64, + "learning_rate": 1.3363468407439339e-06, + "loss": 0.3007, + "step": 45074 + }, + { + "epoch": 16.65, + "learning_rate": 1.3363098903805887e-06, + "loss": 0.2837, + "step": 45075 + }, + { + "epoch": 16.65, + "learning_rate": 1.3362729400172436e-06, + "loss": 0.389, + "step": 45076 + }, + { + "epoch": 16.65, + "learning_rate": 1.3362359896538982e-06, + "loss": 0.373, + "step": 45077 + }, + { + "epoch": 16.65, + "learning_rate": 1.336199039290553e-06, + "loss": 0.2795, + "step": 45078 + }, + { + "epoch": 16.65, + "learning_rate": 1.336162088927208e-06, + "loss": 0.3349, + "step": 45079 + }, + { + "epoch": 16.65, + "learning_rate": 1.3361251385638626e-06, + "loss": 0.1655, + "step": 45080 + }, + { + "epoch": 16.65, + "learning_rate": 1.3360881882005172e-06, + "loss": 0.3582, + "step": 45081 + }, + { + "epoch": 16.65, + "learning_rate": 1.336051237837172e-06, + "loss": 0.2454, + "step": 45082 + }, + { + "epoch": 16.65, + "learning_rate": 1.3360142874738269e-06, + "loss": 0.3807, + "step": 45083 + }, + { + "epoch": 16.65, + "learning_rate": 1.3359773371104815e-06, + "loss": 0.1838, + "step": 45084 + }, + { + "epoch": 16.65, + "learning_rate": 1.3359403867471364e-06, + "loss": 0.27, + "step": 45085 + }, + { + "epoch": 16.65, + "learning_rate": 1.3359034363837912e-06, + "loss": 0.2376, + "step": 45086 + }, + { + "epoch": 16.65, + "learning_rate": 1.3358664860204459e-06, + "loss": 0.319, + "step": 45087 + }, + { + "epoch": 16.65, + "learning_rate": 1.3358295356571007e-06, + "loss": 0.1595, + "step": 45088 + }, + { + "epoch": 16.65, + "learning_rate": 1.3357925852937556e-06, + "loss": 0.1836, + "step": 45089 + }, + { + "epoch": 16.65, + "learning_rate": 1.3357556349304102e-06, + "loss": 0.1979, + "step": 45090 + }, + { + "epoch": 16.65, + "learning_rate": 1.3357186845670648e-06, + "loss": 0.1556, + "step": 45091 + }, + { + "epoch": 16.65, + "learning_rate": 1.3356817342037197e-06, + "loss": 0.2879, + "step": 45092 + }, + { + "epoch": 16.65, + "learning_rate": 1.3356447838403743e-06, + "loss": 0.3226, + "step": 45093 + }, + { + "epoch": 16.65, + "learning_rate": 1.3356078334770292e-06, + "loss": 0.1939, + "step": 45094 + }, + { + "epoch": 16.65, + "learning_rate": 1.335570883113684e-06, + "loss": 0.2241, + "step": 45095 + }, + { + "epoch": 16.65, + "learning_rate": 1.3355339327503386e-06, + "loss": 0.2613, + "step": 45096 + }, + { + "epoch": 16.65, + "learning_rate": 1.3354969823869935e-06, + "loss": 0.2582, + "step": 45097 + }, + { + "epoch": 16.65, + "learning_rate": 1.3354600320236483e-06, + "loss": 0.2502, + "step": 45098 + }, + { + "epoch": 16.65, + "learning_rate": 1.335423081660303e-06, + "loss": 0.2792, + "step": 45099 + }, + { + "epoch": 16.65, + "learning_rate": 1.3353861312969578e-06, + "loss": 0.2599, + "step": 45100 + }, + { + "epoch": 16.65, + "learning_rate": 1.3353491809336125e-06, + "loss": 0.3558, + "step": 45101 + }, + { + "epoch": 16.66, + "learning_rate": 1.3353122305702673e-06, + "loss": 0.2186, + "step": 45102 + }, + { + "epoch": 16.66, + "learning_rate": 1.335275280206922e-06, + "loss": 0.2024, + "step": 45103 + }, + { + "epoch": 16.66, + "learning_rate": 1.3352383298435768e-06, + "loss": 0.339, + "step": 45104 + }, + { + "epoch": 16.66, + "learning_rate": 1.3352013794802316e-06, + "loss": 0.2383, + "step": 45105 + }, + { + "epoch": 16.66, + "learning_rate": 1.3351644291168863e-06, + "loss": 0.2009, + "step": 45106 + }, + { + "epoch": 16.66, + "learning_rate": 1.3351274787535411e-06, + "loss": 0.2745, + "step": 45107 + }, + { + "epoch": 16.66, + "learning_rate": 1.335090528390196e-06, + "loss": 0.2298, + "step": 45108 + }, + { + "epoch": 16.66, + "learning_rate": 1.3350535780268506e-06, + "loss": 0.2136, + "step": 45109 + }, + { + "epoch": 16.66, + "learning_rate": 1.3350166276635055e-06, + "loss": 0.213, + "step": 45110 + }, + { + "epoch": 16.66, + "learning_rate": 1.33497967730016e-06, + "loss": 0.2802, + "step": 45111 + }, + { + "epoch": 16.66, + "learning_rate": 1.334942726936815e-06, + "loss": 0.1292, + "step": 45112 + }, + { + "epoch": 16.66, + "learning_rate": 1.3349057765734696e-06, + "loss": 0.2513, + "step": 45113 + }, + { + "epoch": 16.66, + "learning_rate": 1.3348688262101244e-06, + "loss": 0.2382, + "step": 45114 + }, + { + "epoch": 16.66, + "learning_rate": 1.3348318758467793e-06, + "loss": 0.3594, + "step": 45115 + }, + { + "epoch": 16.66, + "learning_rate": 1.334794925483434e-06, + "loss": 0.1901, + "step": 45116 + }, + { + "epoch": 16.66, + "learning_rate": 1.3347579751200888e-06, + "loss": 0.2651, + "step": 45117 + }, + { + "epoch": 16.66, + "learning_rate": 1.3347210247567436e-06, + "loss": 0.2036, + "step": 45118 + }, + { + "epoch": 16.66, + "learning_rate": 1.3346840743933983e-06, + "loss": 0.2918, + "step": 45119 + }, + { + "epoch": 16.66, + "learning_rate": 1.334647124030053e-06, + "loss": 0.2682, + "step": 45120 + }, + { + "epoch": 16.66, + "learning_rate": 1.3346101736667077e-06, + "loss": 0.3781, + "step": 45121 + }, + { + "epoch": 16.66, + "learning_rate": 1.3345732233033624e-06, + "loss": 0.2664, + "step": 45122 + }, + { + "epoch": 16.66, + "learning_rate": 1.3345362729400172e-06, + "loss": 0.2741, + "step": 45123 + }, + { + "epoch": 16.66, + "learning_rate": 1.334499322576672e-06, + "loss": 0.1933, + "step": 45124 + }, + { + "epoch": 16.66, + "learning_rate": 1.3344623722133267e-06, + "loss": 0.2192, + "step": 45125 + }, + { + "epoch": 16.66, + "learning_rate": 1.3344254218499816e-06, + "loss": 0.2981, + "step": 45126 + }, + { + "epoch": 16.66, + "learning_rate": 1.3343884714866364e-06, + "loss": 0.2261, + "step": 45127 + }, + { + "epoch": 16.66, + "learning_rate": 1.334351521123291e-06, + "loss": 0.4505, + "step": 45128 + }, + { + "epoch": 16.67, + "learning_rate": 1.3343145707599459e-06, + "loss": 0.1864, + "step": 45129 + }, + { + "epoch": 16.67, + "learning_rate": 1.3342776203966007e-06, + "loss": 0.168, + "step": 45130 + }, + { + "epoch": 16.67, + "learning_rate": 1.3342406700332554e-06, + "loss": 0.2448, + "step": 45131 + }, + { + "epoch": 16.67, + "learning_rate": 1.33420371966991e-06, + "loss": 0.1919, + "step": 45132 + }, + { + "epoch": 16.67, + "learning_rate": 1.3341667693065649e-06, + "loss": 0.3409, + "step": 45133 + }, + { + "epoch": 16.67, + "learning_rate": 1.3341298189432197e-06, + "loss": 0.6247, + "step": 45134 + }, + { + "epoch": 16.67, + "learning_rate": 1.3340928685798743e-06, + "loss": 0.2442, + "step": 45135 + }, + { + "epoch": 16.67, + "learning_rate": 1.3340559182165292e-06, + "loss": 0.2572, + "step": 45136 + }, + { + "epoch": 16.67, + "learning_rate": 1.334018967853184e-06, + "loss": 0.2562, + "step": 45137 + }, + { + "epoch": 16.67, + "learning_rate": 1.3339820174898387e-06, + "loss": 0.2216, + "step": 45138 + }, + { + "epoch": 16.67, + "learning_rate": 1.3339450671264935e-06, + "loss": 0.3622, + "step": 45139 + }, + { + "epoch": 16.67, + "learning_rate": 1.3339081167631484e-06, + "loss": 0.2224, + "step": 45140 + }, + { + "epoch": 16.67, + "learning_rate": 1.3338711663998028e-06, + "loss": 0.2009, + "step": 45141 + }, + { + "epoch": 16.67, + "learning_rate": 1.3338342160364576e-06, + "loss": 0.1534, + "step": 45142 + }, + { + "epoch": 16.67, + "learning_rate": 1.3337972656731125e-06, + "loss": 0.2514, + "step": 45143 + }, + { + "epoch": 16.67, + "learning_rate": 1.3337603153097671e-06, + "loss": 0.2095, + "step": 45144 + }, + { + "epoch": 16.67, + "learning_rate": 1.333723364946422e-06, + "loss": 0.1739, + "step": 45145 + }, + { + "epoch": 16.67, + "learning_rate": 1.3336864145830768e-06, + "loss": 0.1631, + "step": 45146 + }, + { + "epoch": 16.67, + "learning_rate": 1.3336494642197315e-06, + "loss": 0.2187, + "step": 45147 + }, + { + "epoch": 16.67, + "learning_rate": 1.3336125138563863e-06, + "loss": 0.143, + "step": 45148 + }, + { + "epoch": 16.67, + "learning_rate": 1.3335755634930412e-06, + "loss": 0.2315, + "step": 45149 + }, + { + "epoch": 16.67, + "learning_rate": 1.3335386131296958e-06, + "loss": 0.1308, + "step": 45150 + }, + { + "epoch": 16.67, + "learning_rate": 1.3335016627663504e-06, + "loss": 0.1528, + "step": 45151 + }, + { + "epoch": 16.67, + "learning_rate": 1.3334647124030053e-06, + "loss": 0.3536, + "step": 45152 + }, + { + "epoch": 16.67, + "learning_rate": 1.3334277620396601e-06, + "loss": 0.298, + "step": 45153 + }, + { + "epoch": 16.67, + "learning_rate": 1.3333908116763148e-06, + "loss": 0.3009, + "step": 45154 + }, + { + "epoch": 16.67, + "learning_rate": 1.3333538613129696e-06, + "loss": 0.369, + "step": 45155 + }, + { + "epoch": 16.68, + "learning_rate": 1.3333169109496245e-06, + "loss": 0.225, + "step": 45156 + }, + { + "epoch": 16.68, + "learning_rate": 1.333279960586279e-06, + "loss": 0.3145, + "step": 45157 + }, + { + "epoch": 16.68, + "learning_rate": 1.333243010222934e-06, + "loss": 0.121, + "step": 45158 + }, + { + "epoch": 16.68, + "learning_rate": 1.3332060598595888e-06, + "loss": 0.1293, + "step": 45159 + }, + { + "epoch": 16.68, + "learning_rate": 1.3331691094962434e-06, + "loss": 0.1881, + "step": 45160 + }, + { + "epoch": 16.68, + "learning_rate": 1.333132159132898e-06, + "loss": 0.2234, + "step": 45161 + }, + { + "epoch": 16.68, + "learning_rate": 1.333095208769553e-06, + "loss": 0.287, + "step": 45162 + }, + { + "epoch": 16.68, + "learning_rate": 1.3330582584062075e-06, + "loss": 0.2715, + "step": 45163 + }, + { + "epoch": 16.68, + "learning_rate": 1.3330213080428624e-06, + "loss": 0.1936, + "step": 45164 + }, + { + "epoch": 16.68, + "learning_rate": 1.3329843576795172e-06, + "loss": 0.3118, + "step": 45165 + }, + { + "epoch": 16.68, + "learning_rate": 1.3329474073161719e-06, + "loss": 0.1956, + "step": 45166 + }, + { + "epoch": 16.68, + "learning_rate": 1.3329104569528267e-06, + "loss": 0.2351, + "step": 45167 + }, + { + "epoch": 16.68, + "learning_rate": 1.3328735065894816e-06, + "loss": 0.2966, + "step": 45168 + }, + { + "epoch": 16.68, + "learning_rate": 1.3328365562261362e-06, + "loss": 0.1604, + "step": 45169 + }, + { + "epoch": 16.68, + "learning_rate": 1.332799605862791e-06, + "loss": 0.1979, + "step": 45170 + }, + { + "epoch": 16.68, + "learning_rate": 1.3327626554994457e-06, + "loss": 0.2579, + "step": 45171 + }, + { + "epoch": 16.68, + "learning_rate": 1.3327257051361005e-06, + "loss": 0.2993, + "step": 45172 + }, + { + "epoch": 16.68, + "learning_rate": 1.3326887547727552e-06, + "loss": 0.2603, + "step": 45173 + }, + { + "epoch": 16.68, + "learning_rate": 1.33265180440941e-06, + "loss": 0.3122, + "step": 45174 + }, + { + "epoch": 16.68, + "learning_rate": 1.3326148540460649e-06, + "loss": 0.1856, + "step": 45175 + }, + { + "epoch": 16.68, + "learning_rate": 1.3325779036827195e-06, + "loss": 0.1549, + "step": 45176 + }, + { + "epoch": 16.68, + "learning_rate": 1.3325409533193744e-06, + "loss": 0.2765, + "step": 45177 + }, + { + "epoch": 16.68, + "learning_rate": 1.3325040029560292e-06, + "loss": 0.2245, + "step": 45178 + }, + { + "epoch": 16.68, + "learning_rate": 1.3324670525926838e-06, + "loss": 0.2522, + "step": 45179 + }, + { + "epoch": 16.68, + "learning_rate": 1.3324301022293387e-06, + "loss": 0.2388, + "step": 45180 + }, + { + "epoch": 16.68, + "learning_rate": 1.3323931518659933e-06, + "loss": 0.1455, + "step": 45181 + }, + { + "epoch": 16.68, + "learning_rate": 1.3323562015026482e-06, + "loss": 0.2491, + "step": 45182 + }, + { + "epoch": 16.69, + "learning_rate": 1.3323192511393028e-06, + "loss": 0.2259, + "step": 45183 + }, + { + "epoch": 16.69, + "learning_rate": 1.3322823007759577e-06, + "loss": 0.3022, + "step": 45184 + }, + { + "epoch": 16.69, + "learning_rate": 1.3322453504126125e-06, + "loss": 0.3201, + "step": 45185 + }, + { + "epoch": 16.69, + "learning_rate": 1.3322084000492672e-06, + "loss": 0.2038, + "step": 45186 + }, + { + "epoch": 16.69, + "learning_rate": 1.332171449685922e-06, + "loss": 0.2058, + "step": 45187 + }, + { + "epoch": 16.69, + "learning_rate": 1.3321344993225768e-06, + "loss": 0.1897, + "step": 45188 + }, + { + "epoch": 16.69, + "learning_rate": 1.3320975489592315e-06, + "loss": 0.2578, + "step": 45189 + }, + { + "epoch": 16.69, + "learning_rate": 1.3320605985958863e-06, + "loss": 0.3548, + "step": 45190 + }, + { + "epoch": 16.69, + "learning_rate": 1.332023648232541e-06, + "loss": 0.1913, + "step": 45191 + }, + { + "epoch": 16.69, + "learning_rate": 1.3319866978691956e-06, + "loss": 0.2148, + "step": 45192 + }, + { + "epoch": 16.69, + "learning_rate": 1.3319497475058505e-06, + "loss": 0.2883, + "step": 45193 + }, + { + "epoch": 16.69, + "learning_rate": 1.3319127971425053e-06, + "loss": 0.3067, + "step": 45194 + }, + { + "epoch": 16.69, + "learning_rate": 1.33187584677916e-06, + "loss": 0.2801, + "step": 45195 + }, + { + "epoch": 16.69, + "learning_rate": 1.3318388964158148e-06, + "loss": 0.2305, + "step": 45196 + }, + { + "epoch": 16.69, + "learning_rate": 1.3318019460524696e-06, + "loss": 0.2305, + "step": 45197 + }, + { + "epoch": 16.69, + "learning_rate": 1.3317649956891243e-06, + "loss": 0.2031, + "step": 45198 + }, + { + "epoch": 16.69, + "learning_rate": 1.3317280453257791e-06, + "loss": 0.2547, + "step": 45199 + }, + { + "epoch": 16.69, + "learning_rate": 1.331691094962434e-06, + "loss": 0.1468, + "step": 45200 + }, + { + "epoch": 16.69, + "learning_rate": 1.3316541445990886e-06, + "loss": 0.2751, + "step": 45201 + }, + { + "epoch": 16.69, + "learning_rate": 1.3316171942357432e-06, + "loss": 0.341, + "step": 45202 + }, + { + "epoch": 16.69, + "learning_rate": 1.331580243872398e-06, + "loss": 0.1946, + "step": 45203 + }, + { + "epoch": 16.69, + "learning_rate": 1.331543293509053e-06, + "loss": 0.3082, + "step": 45204 + }, + { + "epoch": 16.69, + "learning_rate": 1.3315063431457076e-06, + "loss": 0.2074, + "step": 45205 + }, + { + "epoch": 16.69, + "learning_rate": 1.3314693927823624e-06, + "loss": 0.2504, + "step": 45206 + }, + { + "epoch": 16.69, + "learning_rate": 1.3314324424190173e-06, + "loss": 0.2452, + "step": 45207 + }, + { + "epoch": 16.69, + "learning_rate": 1.331395492055672e-06, + "loss": 0.2807, + "step": 45208 + }, + { + "epoch": 16.69, + "learning_rate": 1.3313585416923268e-06, + "loss": 0.2439, + "step": 45209 + }, + { + "epoch": 16.69, + "learning_rate": 1.3313215913289816e-06, + "loss": 0.1447, + "step": 45210 + }, + { + "epoch": 16.7, + "learning_rate": 1.331284640965636e-06, + "loss": 0.1655, + "step": 45211 + }, + { + "epoch": 16.7, + "learning_rate": 1.3312476906022909e-06, + "loss": 0.1209, + "step": 45212 + }, + { + "epoch": 16.7, + "learning_rate": 1.3312107402389457e-06, + "loss": 0.138, + "step": 45213 + }, + { + "epoch": 16.7, + "learning_rate": 1.3311737898756004e-06, + "loss": 0.1913, + "step": 45214 + }, + { + "epoch": 16.7, + "learning_rate": 1.3311368395122552e-06, + "loss": 0.1718, + "step": 45215 + }, + { + "epoch": 16.7, + "learning_rate": 1.33109988914891e-06, + "loss": 0.3331, + "step": 45216 + }, + { + "epoch": 16.7, + "learning_rate": 1.3310629387855647e-06, + "loss": 0.2772, + "step": 45217 + }, + { + "epoch": 16.7, + "learning_rate": 1.3310259884222195e-06, + "loss": 0.2387, + "step": 45218 + }, + { + "epoch": 16.7, + "learning_rate": 1.3309890380588744e-06, + "loss": 0.2342, + "step": 45219 + }, + { + "epoch": 16.7, + "learning_rate": 1.330952087695529e-06, + "loss": 0.3777, + "step": 45220 + }, + { + "epoch": 16.7, + "learning_rate": 1.3309151373321837e-06, + "loss": 0.3747, + "step": 45221 + }, + { + "epoch": 16.7, + "learning_rate": 1.3308781869688385e-06, + "loss": 0.2309, + "step": 45222 + }, + { + "epoch": 16.7, + "learning_rate": 1.3308412366054934e-06, + "loss": 0.2535, + "step": 45223 + }, + { + "epoch": 16.7, + "learning_rate": 1.330804286242148e-06, + "loss": 0.2005, + "step": 45224 + }, + { + "epoch": 16.7, + "learning_rate": 1.3307673358788028e-06, + "loss": 0.2864, + "step": 45225 + }, + { + "epoch": 16.7, + "learning_rate": 1.3307303855154577e-06, + "loss": 0.4239, + "step": 45226 + }, + { + "epoch": 16.7, + "learning_rate": 1.3306934351521123e-06, + "loss": 0.325, + "step": 45227 + }, + { + "epoch": 16.7, + "learning_rate": 1.3306564847887672e-06, + "loss": 0.2263, + "step": 45228 + }, + { + "epoch": 16.7, + "learning_rate": 1.330619534425422e-06, + "loss": 0.2773, + "step": 45229 + }, + { + "epoch": 16.7, + "learning_rate": 1.3305825840620767e-06, + "loss": 0.3509, + "step": 45230 + }, + { + "epoch": 16.7, + "learning_rate": 1.3305456336987313e-06, + "loss": 0.183, + "step": 45231 + }, + { + "epoch": 16.7, + "learning_rate": 1.3305086833353861e-06, + "loss": 0.3352, + "step": 45232 + }, + { + "epoch": 16.7, + "learning_rate": 1.330471732972041e-06, + "loss": 0.279, + "step": 45233 + }, + { + "epoch": 16.7, + "learning_rate": 1.3304347826086956e-06, + "loss": 0.2308, + "step": 45234 + }, + { + "epoch": 16.7, + "learning_rate": 1.3303978322453505e-06, + "loss": 0.1118, + "step": 45235 + }, + { + "epoch": 16.7, + "learning_rate": 1.3303608818820053e-06, + "loss": 0.2513, + "step": 45236 + }, + { + "epoch": 16.7, + "learning_rate": 1.33032393151866e-06, + "loss": 0.2339, + "step": 45237 + }, + { + "epoch": 16.71, + "learning_rate": 1.3302869811553148e-06, + "loss": 0.3261, + "step": 45238 + }, + { + "epoch": 16.71, + "learning_rate": 1.3302500307919694e-06, + "loss": 0.2859, + "step": 45239 + }, + { + "epoch": 16.71, + "learning_rate": 1.3302130804286243e-06, + "loss": 0.3183, + "step": 45240 + }, + { + "epoch": 16.71, + "learning_rate": 1.330176130065279e-06, + "loss": 0.1956, + "step": 45241 + }, + { + "epoch": 16.71, + "learning_rate": 1.3301391797019338e-06, + "loss": 0.2854, + "step": 45242 + }, + { + "epoch": 16.71, + "learning_rate": 1.3301022293385884e-06, + "loss": 0.1317, + "step": 45243 + }, + { + "epoch": 16.71, + "learning_rate": 1.3300652789752433e-06, + "loss": 0.2799, + "step": 45244 + }, + { + "epoch": 16.71, + "learning_rate": 1.3300283286118981e-06, + "loss": 0.2885, + "step": 45245 + }, + { + "epoch": 16.71, + "learning_rate": 1.3299913782485527e-06, + "loss": 0.4636, + "step": 45246 + }, + { + "epoch": 16.71, + "learning_rate": 1.3299544278852076e-06, + "loss": 0.2919, + "step": 45247 + }, + { + "epoch": 16.71, + "learning_rate": 1.3299174775218624e-06, + "loss": 0.2658, + "step": 45248 + }, + { + "epoch": 16.71, + "learning_rate": 1.329880527158517e-06, + "loss": 0.1288, + "step": 45249 + }, + { + "epoch": 16.71, + "learning_rate": 1.329843576795172e-06, + "loss": 0.3273, + "step": 45250 + }, + { + "epoch": 16.71, + "learning_rate": 1.3298066264318266e-06, + "loss": 0.2326, + "step": 45251 + }, + { + "epoch": 16.71, + "learning_rate": 1.3297696760684814e-06, + "loss": 0.1888, + "step": 45252 + }, + { + "epoch": 16.71, + "learning_rate": 1.329732725705136e-06, + "loss": 0.2719, + "step": 45253 + }, + { + "epoch": 16.71, + "learning_rate": 1.329695775341791e-06, + "loss": 0.1897, + "step": 45254 + }, + { + "epoch": 16.71, + "learning_rate": 1.3296588249784457e-06, + "loss": 0.3206, + "step": 45255 + }, + { + "epoch": 16.71, + "learning_rate": 1.3296218746151004e-06, + "loss": 0.2297, + "step": 45256 + }, + { + "epoch": 16.71, + "learning_rate": 1.3295849242517552e-06, + "loss": 0.1343, + "step": 45257 + }, + { + "epoch": 16.71, + "learning_rate": 1.32954797388841e-06, + "loss": 0.2201, + "step": 45258 + }, + { + "epoch": 16.71, + "learning_rate": 1.3295110235250647e-06, + "loss": 0.1842, + "step": 45259 + }, + { + "epoch": 16.71, + "learning_rate": 1.3294740731617196e-06, + "loss": 0.3214, + "step": 45260 + }, + { + "epoch": 16.71, + "learning_rate": 1.3294371227983742e-06, + "loss": 0.183, + "step": 45261 + }, + { + "epoch": 16.71, + "learning_rate": 1.3294001724350288e-06, + "loss": 0.3357, + "step": 45262 + }, + { + "epoch": 16.71, + "learning_rate": 1.3293632220716837e-06, + "loss": 0.2549, + "step": 45263 + }, + { + "epoch": 16.71, + "learning_rate": 1.3293262717083385e-06, + "loss": 0.4619, + "step": 45264 + }, + { + "epoch": 16.72, + "learning_rate": 1.3292893213449932e-06, + "loss": 0.196, + "step": 45265 + }, + { + "epoch": 16.72, + "learning_rate": 1.329252370981648e-06, + "loss": 0.2792, + "step": 45266 + }, + { + "epoch": 16.72, + "learning_rate": 1.3292154206183029e-06, + "loss": 0.1675, + "step": 45267 + }, + { + "epoch": 16.72, + "learning_rate": 1.3291784702549575e-06, + "loss": 0.5637, + "step": 45268 + }, + { + "epoch": 16.72, + "learning_rate": 1.3291415198916124e-06, + "loss": 0.318, + "step": 45269 + }, + { + "epoch": 16.72, + "learning_rate": 1.3291045695282672e-06, + "loss": 0.3927, + "step": 45270 + }, + { + "epoch": 16.72, + "learning_rate": 1.3290676191649218e-06, + "loss": 0.3361, + "step": 45271 + }, + { + "epoch": 16.72, + "learning_rate": 1.3290306688015765e-06, + "loss": 0.3845, + "step": 45272 + }, + { + "epoch": 16.72, + "learning_rate": 1.3289937184382313e-06, + "loss": 0.2038, + "step": 45273 + }, + { + "epoch": 16.72, + "learning_rate": 1.3289567680748862e-06, + "loss": 0.3189, + "step": 45274 + }, + { + "epoch": 16.72, + "learning_rate": 1.3289198177115408e-06, + "loss": 0.1902, + "step": 45275 + }, + { + "epoch": 16.72, + "learning_rate": 1.3288828673481957e-06, + "loss": 0.3214, + "step": 45276 + }, + { + "epoch": 16.72, + "learning_rate": 1.3288459169848505e-06, + "loss": 0.372, + "step": 45277 + }, + { + "epoch": 16.72, + "learning_rate": 1.3288089666215051e-06, + "loss": 0.1912, + "step": 45278 + }, + { + "epoch": 16.72, + "learning_rate": 1.32877201625816e-06, + "loss": 0.1366, + "step": 45279 + }, + { + "epoch": 16.72, + "learning_rate": 1.3287350658948148e-06, + "loss": 0.2861, + "step": 45280 + }, + { + "epoch": 16.72, + "learning_rate": 1.3286981155314693e-06, + "loss": 0.2099, + "step": 45281 + }, + { + "epoch": 16.72, + "learning_rate": 1.3286611651681241e-06, + "loss": 0.3228, + "step": 45282 + }, + { + "epoch": 16.72, + "learning_rate": 1.328624214804779e-06, + "loss": 0.1347, + "step": 45283 + }, + { + "epoch": 16.72, + "learning_rate": 1.3285872644414336e-06, + "loss": 0.1967, + "step": 45284 + }, + { + "epoch": 16.72, + "learning_rate": 1.3285503140780884e-06, + "loss": 0.361, + "step": 45285 + }, + { + "epoch": 16.72, + "learning_rate": 1.3285133637147433e-06, + "loss": 0.3625, + "step": 45286 + }, + { + "epoch": 16.72, + "learning_rate": 1.328476413351398e-06, + "loss": 0.2651, + "step": 45287 + }, + { + "epoch": 16.72, + "learning_rate": 1.3284394629880528e-06, + "loss": 0.274, + "step": 45288 + }, + { + "epoch": 16.72, + "learning_rate": 1.3284025126247076e-06, + "loss": 0.2162, + "step": 45289 + }, + { + "epoch": 16.72, + "learning_rate": 1.3283655622613623e-06, + "loss": 0.3206, + "step": 45290 + }, + { + "epoch": 16.72, + "learning_rate": 1.3283286118980169e-06, + "loss": 0.2025, + "step": 45291 + }, + { + "epoch": 16.73, + "learning_rate": 1.3282916615346717e-06, + "loss": 0.2662, + "step": 45292 + }, + { + "epoch": 16.73, + "learning_rate": 1.3282547111713266e-06, + "loss": 0.3747, + "step": 45293 + }, + { + "epoch": 16.73, + "learning_rate": 1.3282177608079812e-06, + "loss": 0.322, + "step": 45294 + }, + { + "epoch": 16.73, + "learning_rate": 1.328180810444636e-06, + "loss": 0.1782, + "step": 45295 + }, + { + "epoch": 16.73, + "learning_rate": 1.328143860081291e-06, + "loss": 0.1659, + "step": 45296 + }, + { + "epoch": 16.73, + "learning_rate": 1.3281069097179456e-06, + "loss": 0.3869, + "step": 45297 + }, + { + "epoch": 16.73, + "learning_rate": 1.3280699593546004e-06, + "loss": 0.2632, + "step": 45298 + }, + { + "epoch": 16.73, + "learning_rate": 1.3280330089912553e-06, + "loss": 0.2904, + "step": 45299 + }, + { + "epoch": 16.73, + "learning_rate": 1.3279960586279099e-06, + "loss": 0.4035, + "step": 45300 + }, + { + "epoch": 16.73, + "learning_rate": 1.3279591082645645e-06, + "loss": 0.1659, + "step": 45301 + }, + { + "epoch": 16.73, + "learning_rate": 1.3279221579012194e-06, + "loss": 0.2128, + "step": 45302 + }, + { + "epoch": 16.73, + "learning_rate": 1.3278852075378742e-06, + "loss": 0.2942, + "step": 45303 + }, + { + "epoch": 16.73, + "learning_rate": 1.3278482571745289e-06, + "loss": 0.3221, + "step": 45304 + }, + { + "epoch": 16.73, + "learning_rate": 1.3278113068111837e-06, + "loss": 0.2852, + "step": 45305 + }, + { + "epoch": 16.73, + "learning_rate": 1.3277743564478386e-06, + "loss": 0.1867, + "step": 45306 + }, + { + "epoch": 16.73, + "learning_rate": 1.3277374060844932e-06, + "loss": 0.2267, + "step": 45307 + }, + { + "epoch": 16.73, + "learning_rate": 1.327700455721148e-06, + "loss": 0.2156, + "step": 45308 + }, + { + "epoch": 16.73, + "learning_rate": 1.3276635053578029e-06, + "loss": 0.2867, + "step": 45309 + }, + { + "epoch": 16.73, + "learning_rate": 1.3276265549944573e-06, + "loss": 0.333, + "step": 45310 + }, + { + "epoch": 16.73, + "learning_rate": 1.3275896046311122e-06, + "loss": 0.2871, + "step": 45311 + }, + { + "epoch": 16.73, + "learning_rate": 1.327552654267767e-06, + "loss": 0.2752, + "step": 45312 + }, + { + "epoch": 16.73, + "learning_rate": 1.3275157039044216e-06, + "loss": 0.3222, + "step": 45313 + }, + { + "epoch": 16.73, + "learning_rate": 1.3274787535410765e-06, + "loss": 0.2415, + "step": 45314 + }, + { + "epoch": 16.73, + "learning_rate": 1.3274418031777313e-06, + "loss": 0.1721, + "step": 45315 + }, + { + "epoch": 16.73, + "learning_rate": 1.327404852814386e-06, + "loss": 0.2911, + "step": 45316 + }, + { + "epoch": 16.73, + "learning_rate": 1.3273679024510408e-06, + "loss": 0.1389, + "step": 45317 + }, + { + "epoch": 16.73, + "learning_rate": 1.3273309520876957e-06, + "loss": 0.228, + "step": 45318 + }, + { + "epoch": 16.74, + "learning_rate": 1.3272940017243503e-06, + "loss": 0.2343, + "step": 45319 + }, + { + "epoch": 16.74, + "learning_rate": 1.327257051361005e-06, + "loss": 0.1361, + "step": 45320 + }, + { + "epoch": 16.74, + "learning_rate": 1.3272201009976598e-06, + "loss": 0.3556, + "step": 45321 + }, + { + "epoch": 16.74, + "learning_rate": 1.3271831506343146e-06, + "loss": 0.1966, + "step": 45322 + }, + { + "epoch": 16.74, + "learning_rate": 1.3271462002709693e-06, + "loss": 0.2808, + "step": 45323 + }, + { + "epoch": 16.74, + "learning_rate": 1.3271092499076241e-06, + "loss": 0.3654, + "step": 45324 + }, + { + "epoch": 16.74, + "learning_rate": 1.327072299544279e-06, + "loss": 0.3854, + "step": 45325 + }, + { + "epoch": 16.74, + "learning_rate": 1.3270353491809336e-06, + "loss": 0.3178, + "step": 45326 + }, + { + "epoch": 16.74, + "learning_rate": 1.3269983988175885e-06, + "loss": 0.2294, + "step": 45327 + }, + { + "epoch": 16.74, + "learning_rate": 1.3269614484542433e-06, + "loss": 0.3263, + "step": 45328 + }, + { + "epoch": 16.74, + "learning_rate": 1.326924498090898e-06, + "loss": 0.1344, + "step": 45329 + }, + { + "epoch": 16.74, + "learning_rate": 1.3268875477275526e-06, + "loss": 0.3303, + "step": 45330 + }, + { + "epoch": 16.74, + "learning_rate": 1.3268505973642074e-06, + "loss": 0.2565, + "step": 45331 + }, + { + "epoch": 16.74, + "learning_rate": 1.326813647000862e-06, + "loss": 0.3212, + "step": 45332 + }, + { + "epoch": 16.74, + "learning_rate": 1.326776696637517e-06, + "loss": 0.1727, + "step": 45333 + }, + { + "epoch": 16.74, + "learning_rate": 1.3267397462741718e-06, + "loss": 0.435, + "step": 45334 + }, + { + "epoch": 16.74, + "learning_rate": 1.3267027959108264e-06, + "loss": 0.2366, + "step": 45335 + }, + { + "epoch": 16.74, + "learning_rate": 1.3266658455474813e-06, + "loss": 0.2793, + "step": 45336 + }, + { + "epoch": 16.74, + "learning_rate": 1.326628895184136e-06, + "loss": 0.1728, + "step": 45337 + }, + { + "epoch": 16.74, + "learning_rate": 1.3265919448207907e-06, + "loss": 0.2462, + "step": 45338 + }, + { + "epoch": 16.74, + "learning_rate": 1.3265549944574456e-06, + "loss": 0.2426, + "step": 45339 + }, + { + "epoch": 16.74, + "learning_rate": 1.3265180440941002e-06, + "loss": 0.3433, + "step": 45340 + }, + { + "epoch": 16.74, + "learning_rate": 1.326481093730755e-06, + "loss": 0.4097, + "step": 45341 + }, + { + "epoch": 16.74, + "learning_rate": 1.3264441433674097e-06, + "loss": 0.1641, + "step": 45342 + }, + { + "epoch": 16.74, + "learning_rate": 1.3264071930040646e-06, + "loss": 0.1712, + "step": 45343 + }, + { + "epoch": 16.74, + "learning_rate": 1.3263702426407194e-06, + "loss": 0.1371, + "step": 45344 + }, + { + "epoch": 16.74, + "learning_rate": 1.326333292277374e-06, + "loss": 0.2741, + "step": 45345 + }, + { + "epoch": 16.75, + "learning_rate": 1.3262963419140289e-06, + "loss": 0.2183, + "step": 45346 + }, + { + "epoch": 16.75, + "learning_rate": 1.3262593915506837e-06, + "loss": 0.1871, + "step": 45347 + }, + { + "epoch": 16.75, + "learning_rate": 1.3262224411873384e-06, + "loss": 0.1301, + "step": 45348 + }, + { + "epoch": 16.75, + "learning_rate": 1.3261854908239932e-06, + "loss": 0.2919, + "step": 45349 + }, + { + "epoch": 16.75, + "learning_rate": 1.3261485404606479e-06, + "loss": 0.2366, + "step": 45350 + }, + { + "epoch": 16.75, + "learning_rate": 1.3261115900973025e-06, + "loss": 0.2793, + "step": 45351 + }, + { + "epoch": 16.75, + "learning_rate": 1.3260746397339573e-06, + "loss": 0.2122, + "step": 45352 + }, + { + "epoch": 16.75, + "learning_rate": 1.3260376893706122e-06, + "loss": 0.1519, + "step": 45353 + }, + { + "epoch": 16.75, + "learning_rate": 1.3260007390072668e-06, + "loss": 0.2953, + "step": 45354 + }, + { + "epoch": 16.75, + "learning_rate": 1.3259637886439217e-06, + "loss": 0.3414, + "step": 45355 + }, + { + "epoch": 16.75, + "learning_rate": 1.3259268382805765e-06, + "loss": 0.195, + "step": 45356 + }, + { + "epoch": 16.75, + "learning_rate": 1.3258898879172312e-06, + "loss": 0.187, + "step": 45357 + }, + { + "epoch": 16.75, + "learning_rate": 1.325852937553886e-06, + "loss": 0.303, + "step": 45358 + }, + { + "epoch": 16.75, + "learning_rate": 1.3258159871905409e-06, + "loss": 0.2167, + "step": 45359 + }, + { + "epoch": 16.75, + "learning_rate": 1.3257790368271955e-06, + "loss": 0.1497, + "step": 45360 + }, + { + "epoch": 16.75, + "learning_rate": 1.3257420864638501e-06, + "loss": 0.4246, + "step": 45361 + }, + { + "epoch": 16.75, + "learning_rate": 1.325705136100505e-06, + "loss": 0.3011, + "step": 45362 + }, + { + "epoch": 16.75, + "learning_rate": 1.3256681857371598e-06, + "loss": 0.1799, + "step": 45363 + }, + { + "epoch": 16.75, + "learning_rate": 1.3256312353738145e-06, + "loss": 0.3489, + "step": 45364 + }, + { + "epoch": 16.75, + "learning_rate": 1.3255942850104693e-06, + "loss": 0.2837, + "step": 45365 + }, + { + "epoch": 16.75, + "learning_rate": 1.3255573346471242e-06, + "loss": 0.3564, + "step": 45366 + }, + { + "epoch": 16.75, + "learning_rate": 1.3255203842837788e-06, + "loss": 0.2953, + "step": 45367 + }, + { + "epoch": 16.75, + "learning_rate": 1.3254834339204336e-06, + "loss": 0.2628, + "step": 45368 + }, + { + "epoch": 16.75, + "learning_rate": 1.3254464835570885e-06, + "loss": 0.1441, + "step": 45369 + }, + { + "epoch": 16.75, + "learning_rate": 1.3254095331937431e-06, + "loss": 0.3417, + "step": 45370 + }, + { + "epoch": 16.75, + "learning_rate": 1.3253725828303978e-06, + "loss": 0.1558, + "step": 45371 + }, + { + "epoch": 16.75, + "learning_rate": 1.3253356324670526e-06, + "loss": 0.2232, + "step": 45372 + }, + { + "epoch": 16.76, + "learning_rate": 1.3252986821037075e-06, + "loss": 0.2924, + "step": 45373 + }, + { + "epoch": 16.76, + "learning_rate": 1.325261731740362e-06, + "loss": 0.175, + "step": 45374 + }, + { + "epoch": 16.76, + "learning_rate": 1.325224781377017e-06, + "loss": 0.2669, + "step": 45375 + }, + { + "epoch": 16.76, + "learning_rate": 1.3251878310136718e-06, + "loss": 0.3578, + "step": 45376 + }, + { + "epoch": 16.76, + "learning_rate": 1.3251508806503264e-06, + "loss": 0.4223, + "step": 45377 + }, + { + "epoch": 16.76, + "learning_rate": 1.3251139302869813e-06, + "loss": 0.2574, + "step": 45378 + }, + { + "epoch": 16.76, + "learning_rate": 1.3250769799236361e-06, + "loss": 0.2621, + "step": 45379 + }, + { + "epoch": 16.76, + "learning_rate": 1.3250400295602905e-06, + "loss": 0.1839, + "step": 45380 + }, + { + "epoch": 16.76, + "learning_rate": 1.3250030791969454e-06, + "loss": 0.1028, + "step": 45381 + }, + { + "epoch": 16.76, + "learning_rate": 1.3249661288336002e-06, + "loss": 0.1374, + "step": 45382 + }, + { + "epoch": 16.76, + "learning_rate": 1.3249291784702549e-06, + "loss": 0.4115, + "step": 45383 + }, + { + "epoch": 16.76, + "learning_rate": 1.3248922281069097e-06, + "loss": 0.2428, + "step": 45384 + }, + { + "epoch": 16.76, + "learning_rate": 1.3248552777435646e-06, + "loss": 0.3004, + "step": 45385 + }, + { + "epoch": 16.76, + "learning_rate": 1.3248183273802192e-06, + "loss": 0.1474, + "step": 45386 + }, + { + "epoch": 16.76, + "learning_rate": 1.324781377016874e-06, + "loss": 0.1224, + "step": 45387 + }, + { + "epoch": 16.76, + "learning_rate": 1.324744426653529e-06, + "loss": 0.2342, + "step": 45388 + }, + { + "epoch": 16.76, + "learning_rate": 1.3247074762901835e-06, + "loss": 0.3207, + "step": 45389 + }, + { + "epoch": 16.76, + "learning_rate": 1.3246705259268382e-06, + "loss": 0.3125, + "step": 45390 + }, + { + "epoch": 16.76, + "learning_rate": 1.324633575563493e-06, + "loss": 0.1631, + "step": 45391 + }, + { + "epoch": 16.76, + "learning_rate": 1.3245966252001479e-06, + "loss": 0.2274, + "step": 45392 + }, + { + "epoch": 16.76, + "learning_rate": 1.3245596748368025e-06, + "loss": 0.2827, + "step": 45393 + }, + { + "epoch": 16.76, + "learning_rate": 1.3245227244734574e-06, + "loss": 0.2568, + "step": 45394 + }, + { + "epoch": 16.76, + "learning_rate": 1.3244857741101122e-06, + "loss": 0.2365, + "step": 45395 + }, + { + "epoch": 16.76, + "learning_rate": 1.3244488237467668e-06, + "loss": 0.1506, + "step": 45396 + }, + { + "epoch": 16.76, + "learning_rate": 1.3244118733834217e-06, + "loss": 0.3183, + "step": 45397 + }, + { + "epoch": 16.76, + "learning_rate": 1.3243749230200765e-06, + "loss": 0.2163, + "step": 45398 + }, + { + "epoch": 16.76, + "learning_rate": 1.3243379726567312e-06, + "loss": 0.149, + "step": 45399 + }, + { + "epoch": 16.77, + "learning_rate": 1.3243010222933858e-06, + "loss": 0.151, + "step": 45400 + }, + { + "epoch": 16.77, + "learning_rate": 1.3242640719300407e-06, + "loss": 0.2077, + "step": 45401 + }, + { + "epoch": 16.77, + "learning_rate": 1.3242271215666953e-06, + "loss": 0.1985, + "step": 45402 + }, + { + "epoch": 16.77, + "learning_rate": 1.3241901712033502e-06, + "loss": 0.2172, + "step": 45403 + }, + { + "epoch": 16.77, + "learning_rate": 1.324153220840005e-06, + "loss": 0.148, + "step": 45404 + }, + { + "epoch": 16.77, + "learning_rate": 1.3241162704766596e-06, + "loss": 0.2953, + "step": 45405 + }, + { + "epoch": 16.77, + "learning_rate": 1.3240793201133145e-06, + "loss": 0.3867, + "step": 45406 + }, + { + "epoch": 16.77, + "learning_rate": 1.3240423697499693e-06, + "loss": 0.2009, + "step": 45407 + }, + { + "epoch": 16.77, + "learning_rate": 1.324005419386624e-06, + "loss": 0.2457, + "step": 45408 + }, + { + "epoch": 16.77, + "learning_rate": 1.3239684690232788e-06, + "loss": 0.2367, + "step": 45409 + }, + { + "epoch": 16.77, + "learning_rate": 1.3239315186599335e-06, + "loss": 0.2416, + "step": 45410 + }, + { + "epoch": 16.77, + "learning_rate": 1.3238945682965883e-06, + "loss": 0.4494, + "step": 45411 + }, + { + "epoch": 16.77, + "learning_rate": 1.323857617933243e-06, + "loss": 0.2058, + "step": 45412 + }, + { + "epoch": 16.77, + "learning_rate": 1.3238206675698978e-06, + "loss": 0.2062, + "step": 45413 + }, + { + "epoch": 16.77, + "learning_rate": 1.3237837172065526e-06, + "loss": 0.3214, + "step": 45414 + }, + { + "epoch": 16.77, + "learning_rate": 1.3237467668432073e-06, + "loss": 0.2972, + "step": 45415 + }, + { + "epoch": 16.77, + "learning_rate": 1.3237098164798621e-06, + "loss": 0.3407, + "step": 45416 + }, + { + "epoch": 16.77, + "learning_rate": 1.323672866116517e-06, + "loss": 0.2281, + "step": 45417 + }, + { + "epoch": 16.77, + "learning_rate": 1.3236359157531716e-06, + "loss": 0.1442, + "step": 45418 + }, + { + "epoch": 16.77, + "learning_rate": 1.3235989653898265e-06, + "loss": 0.1389, + "step": 45419 + }, + { + "epoch": 16.77, + "learning_rate": 1.323562015026481e-06, + "loss": 0.286, + "step": 45420 + }, + { + "epoch": 16.77, + "learning_rate": 1.323525064663136e-06, + "loss": 0.1529, + "step": 45421 + }, + { + "epoch": 16.77, + "learning_rate": 1.3234881142997906e-06, + "loss": 0.1814, + "step": 45422 + }, + { + "epoch": 16.77, + "learning_rate": 1.3234511639364454e-06, + "loss": 0.3235, + "step": 45423 + }, + { + "epoch": 16.77, + "learning_rate": 1.3234142135731003e-06, + "loss": 0.1721, + "step": 45424 + }, + { + "epoch": 16.77, + "learning_rate": 1.323377263209755e-06, + "loss": 0.4126, + "step": 45425 + }, + { + "epoch": 16.77, + "learning_rate": 1.3233403128464098e-06, + "loss": 0.2431, + "step": 45426 + }, + { + "epoch": 16.78, + "learning_rate": 1.3233033624830644e-06, + "loss": 0.363, + "step": 45427 + }, + { + "epoch": 16.78, + "learning_rate": 1.3232664121197192e-06, + "loss": 0.1321, + "step": 45428 + }, + { + "epoch": 16.78, + "learning_rate": 1.323229461756374e-06, + "loss": 0.2587, + "step": 45429 + }, + { + "epoch": 16.78, + "learning_rate": 1.3231925113930287e-06, + "loss": 0.1362, + "step": 45430 + }, + { + "epoch": 16.78, + "learning_rate": 1.3231555610296834e-06, + "loss": 0.2454, + "step": 45431 + }, + { + "epoch": 16.78, + "learning_rate": 1.3231186106663382e-06, + "loss": 0.3582, + "step": 45432 + }, + { + "epoch": 16.78, + "learning_rate": 1.323081660302993e-06, + "loss": 0.2803, + "step": 45433 + }, + { + "epoch": 16.78, + "learning_rate": 1.3230447099396477e-06, + "loss": 0.1539, + "step": 45434 + }, + { + "epoch": 16.78, + "learning_rate": 1.3230077595763025e-06, + "loss": 0.2925, + "step": 45435 + }, + { + "epoch": 16.78, + "learning_rate": 1.3229708092129574e-06, + "loss": 0.2932, + "step": 45436 + }, + { + "epoch": 16.78, + "learning_rate": 1.322933858849612e-06, + "loss": 0.272, + "step": 45437 + }, + { + "epoch": 16.78, + "learning_rate": 1.3228969084862669e-06, + "loss": 0.1624, + "step": 45438 + }, + { + "epoch": 16.78, + "learning_rate": 1.3228599581229217e-06, + "loss": 0.2631, + "step": 45439 + }, + { + "epoch": 16.78, + "learning_rate": 1.3228230077595764e-06, + "loss": 0.1666, + "step": 45440 + }, + { + "epoch": 16.78, + "learning_rate": 1.322786057396231e-06, + "loss": 0.3002, + "step": 45441 + }, + { + "epoch": 16.78, + "learning_rate": 1.3227491070328858e-06, + "loss": 0.3228, + "step": 45442 + }, + { + "epoch": 16.78, + "learning_rate": 1.3227121566695407e-06, + "loss": 0.3765, + "step": 45443 + }, + { + "epoch": 16.78, + "learning_rate": 1.3226752063061953e-06, + "loss": 0.284, + "step": 45444 + }, + { + "epoch": 16.78, + "learning_rate": 1.3226382559428502e-06, + "loss": 0.2164, + "step": 45445 + }, + { + "epoch": 16.78, + "learning_rate": 1.322601305579505e-06, + "loss": 0.2331, + "step": 45446 + }, + { + "epoch": 16.78, + "learning_rate": 1.3225643552161597e-06, + "loss": 0.2576, + "step": 45447 + }, + { + "epoch": 16.78, + "learning_rate": 1.3225274048528145e-06, + "loss": 0.2291, + "step": 45448 + }, + { + "epoch": 16.78, + "learning_rate": 1.3224904544894694e-06, + "loss": 0.1605, + "step": 45449 + }, + { + "epoch": 16.78, + "learning_rate": 1.3224535041261238e-06, + "loss": 0.2133, + "step": 45450 + }, + { + "epoch": 16.78, + "learning_rate": 1.3224165537627786e-06, + "loss": 0.2627, + "step": 45451 + }, + { + "epoch": 16.78, + "learning_rate": 1.3223796033994335e-06, + "loss": 0.3636, + "step": 45452 + }, + { + "epoch": 16.78, + "learning_rate": 1.3223426530360881e-06, + "loss": 0.3219, + "step": 45453 + }, + { + "epoch": 16.79, + "learning_rate": 1.322305702672743e-06, + "loss": 0.2598, + "step": 45454 + }, + { + "epoch": 16.79, + "learning_rate": 1.3222687523093978e-06, + "loss": 0.1325, + "step": 45455 + }, + { + "epoch": 16.79, + "learning_rate": 1.3222318019460524e-06, + "loss": 0.4068, + "step": 45456 + }, + { + "epoch": 16.79, + "learning_rate": 1.3221948515827073e-06, + "loss": 0.2109, + "step": 45457 + }, + { + "epoch": 16.79, + "learning_rate": 1.3221579012193621e-06, + "loss": 0.2358, + "step": 45458 + }, + { + "epoch": 16.79, + "learning_rate": 1.3221209508560168e-06, + "loss": 0.4009, + "step": 45459 + }, + { + "epoch": 16.79, + "learning_rate": 1.3220840004926714e-06, + "loss": 0.2613, + "step": 45460 + }, + { + "epoch": 16.79, + "learning_rate": 1.3220470501293263e-06, + "loss": 0.221, + "step": 45461 + }, + { + "epoch": 16.79, + "learning_rate": 1.3220100997659811e-06, + "loss": 0.1796, + "step": 45462 + }, + { + "epoch": 16.79, + "learning_rate": 1.3219731494026357e-06, + "loss": 0.1256, + "step": 45463 + }, + { + "epoch": 16.79, + "learning_rate": 1.3219361990392906e-06, + "loss": 0.2601, + "step": 45464 + }, + { + "epoch": 16.79, + "learning_rate": 1.3218992486759454e-06, + "loss": 0.2336, + "step": 45465 + }, + { + "epoch": 16.79, + "learning_rate": 1.3218622983126e-06, + "loss": 0.3541, + "step": 45466 + }, + { + "epoch": 16.79, + "learning_rate": 1.321825347949255e-06, + "loss": 0.2361, + "step": 45467 + }, + { + "epoch": 16.79, + "learning_rate": 1.3217883975859098e-06, + "loss": 0.244, + "step": 45468 + }, + { + "epoch": 16.79, + "learning_rate": 1.3217514472225644e-06, + "loss": 0.208, + "step": 45469 + }, + { + "epoch": 16.79, + "learning_rate": 1.321714496859219e-06, + "loss": 0.2234, + "step": 45470 + }, + { + "epoch": 16.79, + "learning_rate": 1.321677546495874e-06, + "loss": 0.3664, + "step": 45471 + }, + { + "epoch": 16.79, + "learning_rate": 1.3216405961325285e-06, + "loss": 0.177, + "step": 45472 + }, + { + "epoch": 16.79, + "learning_rate": 1.3216036457691834e-06, + "loss": 0.2473, + "step": 45473 + }, + { + "epoch": 16.79, + "learning_rate": 1.3215666954058382e-06, + "loss": 0.13, + "step": 45474 + }, + { + "epoch": 16.79, + "learning_rate": 1.3215297450424929e-06, + "loss": 0.2115, + "step": 45475 + }, + { + "epoch": 16.79, + "learning_rate": 1.3214927946791477e-06, + "loss": 0.2009, + "step": 45476 + }, + { + "epoch": 16.79, + "learning_rate": 1.3214558443158026e-06, + "loss": 0.2165, + "step": 45477 + }, + { + "epoch": 16.79, + "learning_rate": 1.3214188939524572e-06, + "loss": 0.3237, + "step": 45478 + }, + { + "epoch": 16.79, + "learning_rate": 1.321381943589112e-06, + "loss": 0.3592, + "step": 45479 + }, + { + "epoch": 16.79, + "learning_rate": 1.3213449932257667e-06, + "loss": 0.2186, + "step": 45480 + }, + { + "epoch": 16.8, + "learning_rate": 1.3213080428624215e-06, + "loss": 0.414, + "step": 45481 + }, + { + "epoch": 16.8, + "learning_rate": 1.3212710924990762e-06, + "loss": 0.3352, + "step": 45482 + }, + { + "epoch": 16.8, + "learning_rate": 1.321234142135731e-06, + "loss": 0.277, + "step": 45483 + }, + { + "epoch": 16.8, + "learning_rate": 1.3211971917723859e-06, + "loss": 0.1428, + "step": 45484 + }, + { + "epoch": 16.8, + "learning_rate": 1.3211602414090405e-06, + "loss": 0.3554, + "step": 45485 + }, + { + "epoch": 16.8, + "learning_rate": 1.3211232910456954e-06, + "loss": 0.2056, + "step": 45486 + }, + { + "epoch": 16.8, + "learning_rate": 1.3210863406823502e-06, + "loss": 0.2438, + "step": 45487 + }, + { + "epoch": 16.8, + "learning_rate": 1.3210493903190048e-06, + "loss": 0.3057, + "step": 45488 + }, + { + "epoch": 16.8, + "learning_rate": 1.3210124399556597e-06, + "loss": 0.2265, + "step": 45489 + }, + { + "epoch": 16.8, + "learning_rate": 1.3209754895923143e-06, + "loss": 0.152, + "step": 45490 + }, + { + "epoch": 16.8, + "learning_rate": 1.3209385392289692e-06, + "loss": 0.2268, + "step": 45491 + }, + { + "epoch": 16.8, + "learning_rate": 1.3209015888656238e-06, + "loss": 0.2222, + "step": 45492 + }, + { + "epoch": 16.8, + "learning_rate": 1.3208646385022787e-06, + "loss": 0.4064, + "step": 45493 + }, + { + "epoch": 16.8, + "learning_rate": 1.3208276881389335e-06, + "loss": 0.2944, + "step": 45494 + }, + { + "epoch": 16.8, + "learning_rate": 1.3207907377755881e-06, + "loss": 0.3309, + "step": 45495 + }, + { + "epoch": 16.8, + "learning_rate": 1.320753787412243e-06, + "loss": 0.2666, + "step": 45496 + }, + { + "epoch": 16.8, + "learning_rate": 1.3207168370488978e-06, + "loss": 0.2602, + "step": 45497 + }, + { + "epoch": 16.8, + "learning_rate": 1.3206798866855525e-06, + "loss": 0.3429, + "step": 45498 + }, + { + "epoch": 16.8, + "learning_rate": 1.3206429363222073e-06, + "loss": 0.1881, + "step": 45499 + }, + { + "epoch": 16.8, + "learning_rate": 1.320605985958862e-06, + "loss": 0.1596, + "step": 45500 + }, + { + "epoch": 16.8, + "learning_rate": 1.3205690355955166e-06, + "loss": 0.1788, + "step": 45501 + }, + { + "epoch": 16.8, + "learning_rate": 1.3205320852321714e-06, + "loss": 0.1954, + "step": 45502 + }, + { + "epoch": 16.8, + "learning_rate": 1.3204951348688263e-06, + "loss": 0.4499, + "step": 45503 + }, + { + "epoch": 16.8, + "learning_rate": 1.320458184505481e-06, + "loss": 0.1272, + "step": 45504 + }, + { + "epoch": 16.8, + "learning_rate": 1.3204212341421358e-06, + "loss": 0.2977, + "step": 45505 + }, + { + "epoch": 16.8, + "learning_rate": 1.3203842837787906e-06, + "loss": 0.3423, + "step": 45506 + }, + { + "epoch": 16.8, + "learning_rate": 1.3203473334154453e-06, + "loss": 0.2275, + "step": 45507 + }, + { + "epoch": 16.81, + "learning_rate": 1.3203103830521e-06, + "loss": 0.2642, + "step": 45508 + }, + { + "epoch": 16.81, + "learning_rate": 1.320273432688755e-06, + "loss": 0.1963, + "step": 45509 + }, + { + "epoch": 16.81, + "learning_rate": 1.3202364823254096e-06, + "loss": 0.3393, + "step": 45510 + }, + { + "epoch": 16.81, + "learning_rate": 1.3201995319620642e-06, + "loss": 0.2746, + "step": 45511 + }, + { + "epoch": 16.81, + "learning_rate": 1.320162581598719e-06, + "loss": 0.2278, + "step": 45512 + }, + { + "epoch": 16.81, + "learning_rate": 1.320125631235374e-06, + "loss": 0.3117, + "step": 45513 + }, + { + "epoch": 16.81, + "learning_rate": 1.3200886808720286e-06, + "loss": 0.2027, + "step": 45514 + }, + { + "epoch": 16.81, + "learning_rate": 1.3200517305086834e-06, + "loss": 0.3432, + "step": 45515 + }, + { + "epoch": 16.81, + "learning_rate": 1.3200147801453383e-06, + "loss": 0.1704, + "step": 45516 + }, + { + "epoch": 16.81, + "learning_rate": 1.3199778297819929e-06, + "loss": 0.2379, + "step": 45517 + }, + { + "epoch": 16.81, + "learning_rate": 1.3199408794186477e-06, + "loss": 0.307, + "step": 45518 + }, + { + "epoch": 16.81, + "learning_rate": 1.3199039290553024e-06, + "loss": 0.1628, + "step": 45519 + }, + { + "epoch": 16.81, + "learning_rate": 1.319866978691957e-06, + "loss": 0.2299, + "step": 45520 + }, + { + "epoch": 16.81, + "learning_rate": 1.3198300283286119e-06, + "loss": 0.1884, + "step": 45521 + }, + { + "epoch": 16.81, + "learning_rate": 1.3197930779652667e-06, + "loss": 0.2471, + "step": 45522 + }, + { + "epoch": 16.81, + "learning_rate": 1.3197561276019213e-06, + "loss": 0.2755, + "step": 45523 + }, + { + "epoch": 16.81, + "learning_rate": 1.3197191772385762e-06, + "loss": 0.2515, + "step": 45524 + }, + { + "epoch": 16.81, + "learning_rate": 1.319682226875231e-06, + "loss": 0.2035, + "step": 45525 + }, + { + "epoch": 16.81, + "learning_rate": 1.3196452765118857e-06, + "loss": 0.287, + "step": 45526 + }, + { + "epoch": 16.81, + "learning_rate": 1.3196083261485405e-06, + "loss": 0.2443, + "step": 45527 + }, + { + "epoch": 16.81, + "learning_rate": 1.3195713757851954e-06, + "loss": 0.0894, + "step": 45528 + }, + { + "epoch": 16.81, + "learning_rate": 1.31953442542185e-06, + "loss": 0.2581, + "step": 45529 + }, + { + "epoch": 16.81, + "learning_rate": 1.3194974750585046e-06, + "loss": 0.2609, + "step": 45530 + }, + { + "epoch": 16.81, + "learning_rate": 1.3194605246951595e-06, + "loss": 0.2806, + "step": 45531 + }, + { + "epoch": 16.81, + "learning_rate": 1.3194235743318143e-06, + "loss": 0.1753, + "step": 45532 + }, + { + "epoch": 16.81, + "learning_rate": 1.319386623968469e-06, + "loss": 0.2564, + "step": 45533 + }, + { + "epoch": 16.81, + "learning_rate": 1.3193496736051238e-06, + "loss": 0.248, + "step": 45534 + }, + { + "epoch": 16.81, + "learning_rate": 1.3193127232417787e-06, + "loss": 0.1969, + "step": 45535 + }, + { + "epoch": 16.82, + "learning_rate": 1.3192757728784333e-06, + "loss": 0.2569, + "step": 45536 + }, + { + "epoch": 16.82, + "learning_rate": 1.3192388225150882e-06, + "loss": 0.2954, + "step": 45537 + }, + { + "epoch": 16.82, + "learning_rate": 1.319201872151743e-06, + "loss": 0.2074, + "step": 45538 + }, + { + "epoch": 16.82, + "learning_rate": 1.3191649217883974e-06, + "loss": 0.1565, + "step": 45539 + }, + { + "epoch": 16.82, + "learning_rate": 1.3191279714250523e-06, + "loss": 0.2296, + "step": 45540 + }, + { + "epoch": 16.82, + "learning_rate": 1.3190910210617071e-06, + "loss": 0.3098, + "step": 45541 + }, + { + "epoch": 16.82, + "learning_rate": 1.3190540706983618e-06, + "loss": 0.2884, + "step": 45542 + }, + { + "epoch": 16.82, + "learning_rate": 1.3190171203350166e-06, + "loss": 0.1908, + "step": 45543 + }, + { + "epoch": 16.82, + "learning_rate": 1.3189801699716715e-06, + "loss": 0.3004, + "step": 45544 + }, + { + "epoch": 16.82, + "learning_rate": 1.318943219608326e-06, + "loss": 0.161, + "step": 45545 + }, + { + "epoch": 16.82, + "learning_rate": 1.318906269244981e-06, + "loss": 0.1981, + "step": 45546 + }, + { + "epoch": 16.82, + "learning_rate": 1.3188693188816358e-06, + "loss": 0.3452, + "step": 45547 + }, + { + "epoch": 16.82, + "learning_rate": 1.3188323685182904e-06, + "loss": 0.274, + "step": 45548 + }, + { + "epoch": 16.82, + "learning_rate": 1.318795418154945e-06, + "loss": 0.3909, + "step": 45549 + }, + { + "epoch": 16.82, + "learning_rate": 1.3187584677916e-06, + "loss": 0.2494, + "step": 45550 + }, + { + "epoch": 16.82, + "learning_rate": 1.3187215174282548e-06, + "loss": 0.1798, + "step": 45551 + }, + { + "epoch": 16.82, + "learning_rate": 1.3186845670649094e-06, + "loss": 0.2744, + "step": 45552 + }, + { + "epoch": 16.82, + "learning_rate": 1.3186476167015643e-06, + "loss": 0.2646, + "step": 45553 + }, + { + "epoch": 16.82, + "learning_rate": 1.318610666338219e-06, + "loss": 0.3172, + "step": 45554 + }, + { + "epoch": 16.82, + "learning_rate": 1.3185737159748737e-06, + "loss": 0.2334, + "step": 45555 + }, + { + "epoch": 16.82, + "learning_rate": 1.3185367656115286e-06, + "loss": 0.1642, + "step": 45556 + }, + { + "epoch": 16.82, + "learning_rate": 1.3184998152481834e-06, + "loss": 0.2848, + "step": 45557 + }, + { + "epoch": 16.82, + "learning_rate": 1.318462864884838e-06, + "loss": 0.324, + "step": 45558 + }, + { + "epoch": 16.82, + "learning_rate": 1.3184259145214927e-06, + "loss": 0.2229, + "step": 45559 + }, + { + "epoch": 16.82, + "learning_rate": 1.3183889641581476e-06, + "loss": 0.2679, + "step": 45560 + }, + { + "epoch": 16.82, + "learning_rate": 1.3183520137948024e-06, + "loss": 0.2338, + "step": 45561 + }, + { + "epoch": 16.82, + "learning_rate": 1.318315063431457e-06, + "loss": 0.2021, + "step": 45562 + }, + { + "epoch": 16.83, + "learning_rate": 1.3182781130681119e-06, + "loss": 0.279, + "step": 45563 + }, + { + "epoch": 16.83, + "learning_rate": 1.3182411627047667e-06, + "loss": 0.2389, + "step": 45564 + }, + { + "epoch": 16.83, + "learning_rate": 1.3182042123414214e-06, + "loss": 0.2413, + "step": 45565 + }, + { + "epoch": 16.83, + "learning_rate": 1.3181672619780762e-06, + "loss": 0.1723, + "step": 45566 + }, + { + "epoch": 16.83, + "learning_rate": 1.318130311614731e-06, + "loss": 0.1996, + "step": 45567 + }, + { + "epoch": 16.83, + "learning_rate": 1.3180933612513857e-06, + "loss": 0.1405, + "step": 45568 + }, + { + "epoch": 16.83, + "learning_rate": 1.3180564108880403e-06, + "loss": 0.3215, + "step": 45569 + }, + { + "epoch": 16.83, + "learning_rate": 1.3180194605246952e-06, + "loss": 0.1702, + "step": 45570 + }, + { + "epoch": 16.83, + "learning_rate": 1.3179825101613498e-06, + "loss": 0.2771, + "step": 45571 + }, + { + "epoch": 16.83, + "learning_rate": 1.3179455597980047e-06, + "loss": 0.2223, + "step": 45572 + }, + { + "epoch": 16.83, + "learning_rate": 1.3179086094346595e-06, + "loss": 0.2624, + "step": 45573 + }, + { + "epoch": 16.83, + "learning_rate": 1.3178716590713142e-06, + "loss": 0.2187, + "step": 45574 + }, + { + "epoch": 16.83, + "learning_rate": 1.317834708707969e-06, + "loss": 0.2177, + "step": 45575 + }, + { + "epoch": 16.83, + "learning_rate": 1.3177977583446239e-06, + "loss": 0.3171, + "step": 45576 + }, + { + "epoch": 16.83, + "learning_rate": 1.3177608079812785e-06, + "loss": 0.2447, + "step": 45577 + }, + { + "epoch": 16.83, + "learning_rate": 1.3177238576179333e-06, + "loss": 0.2652, + "step": 45578 + }, + { + "epoch": 16.83, + "learning_rate": 1.317686907254588e-06, + "loss": 0.175, + "step": 45579 + }, + { + "epoch": 16.83, + "learning_rate": 1.3176499568912428e-06, + "loss": 0.1386, + "step": 45580 + }, + { + "epoch": 16.83, + "learning_rate": 1.3176130065278975e-06, + "loss": 0.1564, + "step": 45581 + }, + { + "epoch": 16.83, + "learning_rate": 1.3175760561645523e-06, + "loss": 0.1213, + "step": 45582 + }, + { + "epoch": 16.83, + "learning_rate": 1.3175391058012072e-06, + "loss": 0.1935, + "step": 45583 + }, + { + "epoch": 16.83, + "learning_rate": 1.3175021554378618e-06, + "loss": 0.3306, + "step": 45584 + }, + { + "epoch": 16.83, + "learning_rate": 1.3174652050745166e-06, + "loss": 0.3749, + "step": 45585 + }, + { + "epoch": 16.83, + "learning_rate": 1.3174282547111715e-06, + "loss": 0.3783, + "step": 45586 + }, + { + "epoch": 16.83, + "learning_rate": 1.3173913043478261e-06, + "loss": 0.3063, + "step": 45587 + }, + { + "epoch": 16.83, + "learning_rate": 1.317354353984481e-06, + "loss": 0.238, + "step": 45588 + }, + { + "epoch": 16.83, + "learning_rate": 1.3173174036211356e-06, + "loss": 0.2773, + "step": 45589 + }, + { + "epoch": 16.84, + "learning_rate": 1.3172804532577902e-06, + "loss": 0.2488, + "step": 45590 + }, + { + "epoch": 16.84, + "learning_rate": 1.317243502894445e-06, + "loss": 0.3207, + "step": 45591 + }, + { + "epoch": 16.84, + "learning_rate": 1.3172065525311e-06, + "loss": 0.1, + "step": 45592 + }, + { + "epoch": 16.84, + "learning_rate": 1.3171696021677546e-06, + "loss": 0.2289, + "step": 45593 + }, + { + "epoch": 16.84, + "learning_rate": 1.3171326518044094e-06, + "loss": 0.245, + "step": 45594 + }, + { + "epoch": 16.84, + "learning_rate": 1.3170957014410643e-06, + "loss": 0.2673, + "step": 45595 + }, + { + "epoch": 16.84, + "learning_rate": 1.317058751077719e-06, + "loss": 0.1995, + "step": 45596 + }, + { + "epoch": 16.84, + "learning_rate": 1.3170218007143738e-06, + "loss": 0.1687, + "step": 45597 + }, + { + "epoch": 16.84, + "learning_rate": 1.3169848503510286e-06, + "loss": 0.2319, + "step": 45598 + }, + { + "epoch": 16.84, + "learning_rate": 1.3169478999876832e-06, + "loss": 0.1713, + "step": 45599 + }, + { + "epoch": 16.84, + "learning_rate": 1.3169109496243379e-06, + "loss": 0.1836, + "step": 45600 + }, + { + "epoch": 16.84, + "learning_rate": 1.3168739992609927e-06, + "loss": 0.2726, + "step": 45601 + }, + { + "epoch": 16.84, + "learning_rate": 1.3168370488976476e-06, + "loss": 0.2354, + "step": 45602 + }, + { + "epoch": 16.84, + "learning_rate": 1.3168000985343022e-06, + "loss": 0.2977, + "step": 45603 + }, + { + "epoch": 16.84, + "learning_rate": 1.316763148170957e-06, + "loss": 0.2177, + "step": 45604 + }, + { + "epoch": 16.84, + "learning_rate": 1.316726197807612e-06, + "loss": 0.3085, + "step": 45605 + }, + { + "epoch": 16.84, + "learning_rate": 1.3166892474442665e-06, + "loss": 0.227, + "step": 45606 + }, + { + "epoch": 16.84, + "learning_rate": 1.3166522970809214e-06, + "loss": 0.2235, + "step": 45607 + }, + { + "epoch": 16.84, + "learning_rate": 1.3166153467175762e-06, + "loss": 0.2512, + "step": 45608 + }, + { + "epoch": 16.84, + "learning_rate": 1.3165783963542309e-06, + "loss": 0.1194, + "step": 45609 + }, + { + "epoch": 16.84, + "learning_rate": 1.3165414459908855e-06, + "loss": 0.2625, + "step": 45610 + }, + { + "epoch": 16.84, + "learning_rate": 1.3165044956275404e-06, + "loss": 0.2165, + "step": 45611 + }, + { + "epoch": 16.84, + "learning_rate": 1.316467545264195e-06, + "loss": 0.1631, + "step": 45612 + }, + { + "epoch": 16.84, + "learning_rate": 1.3164305949008498e-06, + "loss": 0.2041, + "step": 45613 + }, + { + "epoch": 16.84, + "learning_rate": 1.3163936445375047e-06, + "loss": 0.3498, + "step": 45614 + }, + { + "epoch": 16.84, + "learning_rate": 1.3163566941741593e-06, + "loss": 0.2522, + "step": 45615 + }, + { + "epoch": 16.84, + "learning_rate": 1.3163197438108142e-06, + "loss": 0.2393, + "step": 45616 + }, + { + "epoch": 16.85, + "learning_rate": 1.316282793447469e-06, + "loss": 0.3158, + "step": 45617 + }, + { + "epoch": 16.85, + "learning_rate": 1.3162458430841237e-06, + "loss": 0.2378, + "step": 45618 + }, + { + "epoch": 16.85, + "learning_rate": 1.3162088927207783e-06, + "loss": 0.3056, + "step": 45619 + }, + { + "epoch": 16.85, + "learning_rate": 1.3161719423574331e-06, + "loss": 0.3121, + "step": 45620 + }, + { + "epoch": 16.85, + "learning_rate": 1.316134991994088e-06, + "loss": 0.1948, + "step": 45621 + }, + { + "epoch": 16.85, + "learning_rate": 1.3160980416307426e-06, + "loss": 0.3143, + "step": 45622 + }, + { + "epoch": 16.85, + "learning_rate": 1.3160610912673975e-06, + "loss": 0.1907, + "step": 45623 + }, + { + "epoch": 16.85, + "learning_rate": 1.3160241409040523e-06, + "loss": 0.2996, + "step": 45624 + }, + { + "epoch": 16.85, + "learning_rate": 1.315987190540707e-06, + "loss": 0.2315, + "step": 45625 + }, + { + "epoch": 16.85, + "learning_rate": 1.3159502401773618e-06, + "loss": 0.2908, + "step": 45626 + }, + { + "epoch": 16.85, + "learning_rate": 1.3159132898140167e-06, + "loss": 0.3096, + "step": 45627 + }, + { + "epoch": 16.85, + "learning_rate": 1.3158763394506713e-06, + "loss": 0.2518, + "step": 45628 + }, + { + "epoch": 16.85, + "learning_rate": 1.315839389087326e-06, + "loss": 0.2433, + "step": 45629 + }, + { + "epoch": 16.85, + "learning_rate": 1.3158024387239808e-06, + "loss": 0.1752, + "step": 45630 + }, + { + "epoch": 16.85, + "learning_rate": 1.3157654883606356e-06, + "loss": 0.2272, + "step": 45631 + }, + { + "epoch": 16.85, + "learning_rate": 1.3157285379972903e-06, + "loss": 0.1425, + "step": 45632 + }, + { + "epoch": 16.85, + "learning_rate": 1.3156915876339451e-06, + "loss": 0.1512, + "step": 45633 + }, + { + "epoch": 16.85, + "learning_rate": 1.3156546372706e-06, + "loss": 0.2454, + "step": 45634 + }, + { + "epoch": 16.85, + "learning_rate": 1.3156176869072546e-06, + "loss": 0.2472, + "step": 45635 + }, + { + "epoch": 16.85, + "learning_rate": 1.3155807365439095e-06, + "loss": 0.3056, + "step": 45636 + }, + { + "epoch": 16.85, + "learning_rate": 1.3155437861805643e-06, + "loss": 0.2561, + "step": 45637 + }, + { + "epoch": 16.85, + "learning_rate": 1.315506835817219e-06, + "loss": 0.2301, + "step": 45638 + }, + { + "epoch": 16.85, + "learning_rate": 1.3154698854538736e-06, + "loss": 0.3083, + "step": 45639 + }, + { + "epoch": 16.85, + "learning_rate": 1.3154329350905284e-06, + "loss": 0.1273, + "step": 45640 + }, + { + "epoch": 16.85, + "learning_rate": 1.315395984727183e-06, + "loss": 0.1353, + "step": 45641 + }, + { + "epoch": 16.85, + "learning_rate": 1.315359034363838e-06, + "loss": 0.2028, + "step": 45642 + }, + { + "epoch": 16.85, + "learning_rate": 1.3153220840004928e-06, + "loss": 0.137, + "step": 45643 + }, + { + "epoch": 16.86, + "learning_rate": 1.3152851336371474e-06, + "loss": 0.1814, + "step": 45644 + }, + { + "epoch": 16.86, + "learning_rate": 1.3152481832738022e-06, + "loss": 0.2221, + "step": 45645 + }, + { + "epoch": 16.86, + "learning_rate": 1.315211232910457e-06, + "loss": 0.32, + "step": 45646 + }, + { + "epoch": 16.86, + "learning_rate": 1.3151742825471117e-06, + "loss": 0.1981, + "step": 45647 + }, + { + "epoch": 16.86, + "learning_rate": 1.3151373321837666e-06, + "loss": 0.159, + "step": 45648 + }, + { + "epoch": 16.86, + "learning_rate": 1.3151003818204212e-06, + "loss": 0.1343, + "step": 45649 + }, + { + "epoch": 16.86, + "learning_rate": 1.315063431457076e-06, + "loss": 0.3181, + "step": 45650 + }, + { + "epoch": 16.86, + "learning_rate": 1.3150264810937307e-06, + "loss": 0.2463, + "step": 45651 + }, + { + "epoch": 16.86, + "learning_rate": 1.3149895307303855e-06, + "loss": 0.2253, + "step": 45652 + }, + { + "epoch": 16.86, + "learning_rate": 1.3149525803670404e-06, + "loss": 0.25, + "step": 45653 + }, + { + "epoch": 16.86, + "learning_rate": 1.314915630003695e-06, + "loss": 0.2587, + "step": 45654 + }, + { + "epoch": 16.86, + "learning_rate": 1.3148786796403499e-06, + "loss": 0.29, + "step": 45655 + }, + { + "epoch": 16.86, + "learning_rate": 1.3148417292770047e-06, + "loss": 0.2322, + "step": 45656 + }, + { + "epoch": 16.86, + "learning_rate": 1.3148047789136594e-06, + "loss": 0.267, + "step": 45657 + }, + { + "epoch": 16.86, + "learning_rate": 1.3147678285503142e-06, + "loss": 0.126, + "step": 45658 + }, + { + "epoch": 16.86, + "learning_rate": 1.3147308781869688e-06, + "loss": 0.2111, + "step": 45659 + }, + { + "epoch": 16.86, + "learning_rate": 1.3146939278236235e-06, + "loss": 0.2428, + "step": 45660 + }, + { + "epoch": 16.86, + "learning_rate": 1.3146569774602783e-06, + "loss": 0.2501, + "step": 45661 + }, + { + "epoch": 16.86, + "learning_rate": 1.3146200270969332e-06, + "loss": 0.2413, + "step": 45662 + }, + { + "epoch": 16.86, + "learning_rate": 1.3145830767335878e-06, + "loss": 0.2635, + "step": 45663 + }, + { + "epoch": 16.86, + "learning_rate": 1.3145461263702427e-06, + "loss": 0.165, + "step": 45664 + }, + { + "epoch": 16.86, + "learning_rate": 1.3145091760068975e-06, + "loss": 0.3393, + "step": 45665 + }, + { + "epoch": 16.86, + "learning_rate": 1.3144722256435521e-06, + "loss": 0.2763, + "step": 45666 + }, + { + "epoch": 16.86, + "learning_rate": 1.314435275280207e-06, + "loss": 0.1464, + "step": 45667 + }, + { + "epoch": 16.86, + "learning_rate": 1.3143983249168618e-06, + "loss": 0.2124, + "step": 45668 + }, + { + "epoch": 16.86, + "learning_rate": 1.3143613745535165e-06, + "loss": 0.3628, + "step": 45669 + }, + { + "epoch": 16.86, + "learning_rate": 1.3143244241901711e-06, + "loss": 0.2382, + "step": 45670 + }, + { + "epoch": 16.87, + "learning_rate": 1.314287473826826e-06, + "loss": 0.1972, + "step": 45671 + }, + { + "epoch": 16.87, + "learning_rate": 1.3142505234634808e-06, + "loss": 0.2387, + "step": 45672 + }, + { + "epoch": 16.87, + "learning_rate": 1.3142135731001354e-06, + "loss": 0.1867, + "step": 45673 + }, + { + "epoch": 16.87, + "learning_rate": 1.3141766227367903e-06, + "loss": 0.2908, + "step": 45674 + }, + { + "epoch": 16.87, + "learning_rate": 1.3141396723734451e-06, + "loss": 0.1822, + "step": 45675 + }, + { + "epoch": 16.87, + "learning_rate": 1.3141027220100998e-06, + "loss": 0.3432, + "step": 45676 + }, + { + "epoch": 16.87, + "learning_rate": 1.3140657716467546e-06, + "loss": 0.2353, + "step": 45677 + }, + { + "epoch": 16.87, + "learning_rate": 1.3140288212834095e-06, + "loss": 0.1662, + "step": 45678 + }, + { + "epoch": 16.87, + "learning_rate": 1.3139918709200641e-06, + "loss": 0.1986, + "step": 45679 + }, + { + "epoch": 16.87, + "learning_rate": 1.3139549205567187e-06, + "loss": 0.2007, + "step": 45680 + }, + { + "epoch": 16.87, + "learning_rate": 1.3139179701933736e-06, + "loss": 0.2904, + "step": 45681 + }, + { + "epoch": 16.87, + "learning_rate": 1.3138810198300284e-06, + "loss": 0.413, + "step": 45682 + }, + { + "epoch": 16.87, + "learning_rate": 1.313844069466683e-06, + "loss": 0.195, + "step": 45683 + }, + { + "epoch": 16.87, + "learning_rate": 1.313807119103338e-06, + "loss": 0.3076, + "step": 45684 + }, + { + "epoch": 16.87, + "learning_rate": 1.3137701687399928e-06, + "loss": 0.1511, + "step": 45685 + }, + { + "epoch": 16.87, + "learning_rate": 1.3137332183766474e-06, + "loss": 0.3929, + "step": 45686 + }, + { + "epoch": 16.87, + "learning_rate": 1.3136962680133023e-06, + "loss": 0.257, + "step": 45687 + }, + { + "epoch": 16.87, + "learning_rate": 1.313659317649957e-06, + "loss": 0.2288, + "step": 45688 + }, + { + "epoch": 16.87, + "learning_rate": 1.3136223672866115e-06, + "loss": 0.142, + "step": 45689 + }, + { + "epoch": 16.87, + "learning_rate": 1.3135854169232664e-06, + "loss": 0.1367, + "step": 45690 + }, + { + "epoch": 16.87, + "learning_rate": 1.3135484665599212e-06, + "loss": 0.3796, + "step": 45691 + }, + { + "epoch": 16.87, + "learning_rate": 1.3135115161965759e-06, + "loss": 0.2423, + "step": 45692 + }, + { + "epoch": 16.87, + "learning_rate": 1.3134745658332307e-06, + "loss": 0.2322, + "step": 45693 + }, + { + "epoch": 16.87, + "learning_rate": 1.3134376154698856e-06, + "loss": 0.3326, + "step": 45694 + }, + { + "epoch": 16.87, + "learning_rate": 1.3134006651065402e-06, + "loss": 0.2858, + "step": 45695 + }, + { + "epoch": 16.87, + "learning_rate": 1.313363714743195e-06, + "loss": 0.2656, + "step": 45696 + }, + { + "epoch": 16.87, + "learning_rate": 1.31332676437985e-06, + "loss": 0.3887, + "step": 45697 + }, + { + "epoch": 16.88, + "learning_rate": 1.3132898140165045e-06, + "loss": 0.2396, + "step": 45698 + }, + { + "epoch": 16.88, + "learning_rate": 1.3132528636531592e-06, + "loss": 0.2464, + "step": 45699 + }, + { + "epoch": 16.88, + "learning_rate": 1.313215913289814e-06, + "loss": 0.1967, + "step": 45700 + }, + { + "epoch": 16.88, + "learning_rate": 1.3131789629264689e-06, + "loss": 0.219, + "step": 45701 + }, + { + "epoch": 16.88, + "learning_rate": 1.3131420125631235e-06, + "loss": 0.163, + "step": 45702 + }, + { + "epoch": 16.88, + "learning_rate": 1.3131050621997783e-06, + "loss": 0.3131, + "step": 45703 + }, + { + "epoch": 16.88, + "learning_rate": 1.3130681118364332e-06, + "loss": 0.2071, + "step": 45704 + }, + { + "epoch": 16.88, + "learning_rate": 1.3130311614730878e-06, + "loss": 0.2685, + "step": 45705 + }, + { + "epoch": 16.88, + "learning_rate": 1.3129942111097427e-06, + "loss": 0.139, + "step": 45706 + }, + { + "epoch": 16.88, + "learning_rate": 1.3129572607463975e-06, + "loss": 0.2914, + "step": 45707 + }, + { + "epoch": 16.88, + "learning_rate": 1.3129203103830522e-06, + "loss": 0.2316, + "step": 45708 + }, + { + "epoch": 16.88, + "learning_rate": 1.3128833600197068e-06, + "loss": 0.2268, + "step": 45709 + }, + { + "epoch": 16.88, + "learning_rate": 1.3128464096563617e-06, + "loss": 0.2132, + "step": 45710 + }, + { + "epoch": 16.88, + "learning_rate": 1.3128094592930163e-06, + "loss": 0.2216, + "step": 45711 + }, + { + "epoch": 16.88, + "learning_rate": 1.3127725089296711e-06, + "loss": 0.1992, + "step": 45712 + }, + { + "epoch": 16.88, + "learning_rate": 1.312735558566326e-06, + "loss": 0.2428, + "step": 45713 + }, + { + "epoch": 16.88, + "learning_rate": 1.3126986082029806e-06, + "loss": 0.42, + "step": 45714 + }, + { + "epoch": 16.88, + "learning_rate": 1.3126616578396355e-06, + "loss": 0.2388, + "step": 45715 + }, + { + "epoch": 16.88, + "learning_rate": 1.3126247074762903e-06, + "loss": 0.1998, + "step": 45716 + }, + { + "epoch": 16.88, + "learning_rate": 1.312587757112945e-06, + "loss": 0.2751, + "step": 45717 + }, + { + "epoch": 16.88, + "learning_rate": 1.3125508067495998e-06, + "loss": 0.207, + "step": 45718 + }, + { + "epoch": 16.88, + "learning_rate": 1.3125138563862544e-06, + "loss": 0.3118, + "step": 45719 + }, + { + "epoch": 16.88, + "learning_rate": 1.3124769060229093e-06, + "loss": 0.2793, + "step": 45720 + }, + { + "epoch": 16.88, + "learning_rate": 1.312439955659564e-06, + "loss": 0.3732, + "step": 45721 + }, + { + "epoch": 16.88, + "learning_rate": 1.3124030052962188e-06, + "loss": 0.1794, + "step": 45722 + }, + { + "epoch": 16.88, + "learning_rate": 1.3123660549328736e-06, + "loss": 0.2652, + "step": 45723 + }, + { + "epoch": 16.88, + "learning_rate": 1.3123291045695283e-06, + "loss": 0.2618, + "step": 45724 + }, + { + "epoch": 16.89, + "learning_rate": 1.312292154206183e-06, + "loss": 0.1902, + "step": 45725 + }, + { + "epoch": 16.89, + "learning_rate": 1.312255203842838e-06, + "loss": 0.3438, + "step": 45726 + }, + { + "epoch": 16.89, + "learning_rate": 1.3122182534794926e-06, + "loss": 0.2579, + "step": 45727 + }, + { + "epoch": 16.89, + "learning_rate": 1.3121813031161474e-06, + "loss": 0.151, + "step": 45728 + }, + { + "epoch": 16.89, + "learning_rate": 1.312144352752802e-06, + "loss": 0.3037, + "step": 45729 + }, + { + "epoch": 16.89, + "learning_rate": 1.3121074023894567e-06, + "loss": 0.1999, + "step": 45730 + }, + { + "epoch": 16.89, + "learning_rate": 1.3120704520261116e-06, + "loss": 0.2657, + "step": 45731 + }, + { + "epoch": 16.89, + "learning_rate": 1.3120335016627664e-06, + "loss": 0.3187, + "step": 45732 + }, + { + "epoch": 16.89, + "learning_rate": 1.311996551299421e-06, + "loss": 0.2999, + "step": 45733 + }, + { + "epoch": 16.89, + "learning_rate": 1.3119596009360759e-06, + "loss": 0.1845, + "step": 45734 + }, + { + "epoch": 16.89, + "learning_rate": 1.3119226505727307e-06, + "loss": 0.1423, + "step": 45735 + }, + { + "epoch": 16.89, + "learning_rate": 1.3118857002093854e-06, + "loss": 0.2638, + "step": 45736 + }, + { + "epoch": 16.89, + "learning_rate": 1.3118487498460402e-06, + "loss": 0.1755, + "step": 45737 + }, + { + "epoch": 16.89, + "learning_rate": 1.311811799482695e-06, + "loss": 0.2458, + "step": 45738 + }, + { + "epoch": 16.89, + "learning_rate": 1.3117748491193497e-06, + "loss": 0.288, + "step": 45739 + }, + { + "epoch": 16.89, + "learning_rate": 1.3117378987560043e-06, + "loss": 0.1883, + "step": 45740 + }, + { + "epoch": 16.89, + "learning_rate": 1.3117009483926592e-06, + "loss": 0.2752, + "step": 45741 + }, + { + "epoch": 16.89, + "learning_rate": 1.311663998029314e-06, + "loss": 0.1949, + "step": 45742 + }, + { + "epoch": 16.89, + "learning_rate": 1.3116270476659687e-06, + "loss": 0.259, + "step": 45743 + }, + { + "epoch": 16.89, + "learning_rate": 1.3115900973026235e-06, + "loss": 0.2738, + "step": 45744 + }, + { + "epoch": 16.89, + "learning_rate": 1.3115531469392784e-06, + "loss": 0.3257, + "step": 45745 + }, + { + "epoch": 16.89, + "learning_rate": 1.311516196575933e-06, + "loss": 0.2847, + "step": 45746 + }, + { + "epoch": 16.89, + "learning_rate": 1.3114792462125879e-06, + "loss": 0.1957, + "step": 45747 + }, + { + "epoch": 16.89, + "learning_rate": 1.3114422958492427e-06, + "loss": 0.209, + "step": 45748 + }, + { + "epoch": 16.89, + "learning_rate": 1.3114053454858973e-06, + "loss": 0.1965, + "step": 45749 + }, + { + "epoch": 16.89, + "learning_rate": 1.311368395122552e-06, + "loss": 0.2352, + "step": 45750 + }, + { + "epoch": 16.89, + "learning_rate": 1.3113314447592068e-06, + "loss": 0.2953, + "step": 45751 + }, + { + "epoch": 16.9, + "learning_rate": 1.3112944943958617e-06, + "loss": 0.1568, + "step": 45752 + }, + { + "epoch": 16.9, + "learning_rate": 1.3112575440325163e-06, + "loss": 0.1364, + "step": 45753 + }, + { + "epoch": 16.9, + "learning_rate": 1.3112205936691712e-06, + "loss": 0.2563, + "step": 45754 + }, + { + "epoch": 16.9, + "learning_rate": 1.311183643305826e-06, + "loss": 0.2374, + "step": 45755 + }, + { + "epoch": 16.9, + "learning_rate": 1.3111466929424806e-06, + "loss": 0.3231, + "step": 45756 + }, + { + "epoch": 16.9, + "learning_rate": 1.3111097425791355e-06, + "loss": 0.3294, + "step": 45757 + }, + { + "epoch": 16.9, + "learning_rate": 1.3110727922157901e-06, + "loss": 0.3755, + "step": 45758 + }, + { + "epoch": 16.9, + "learning_rate": 1.3110358418524448e-06, + "loss": 0.16, + "step": 45759 + }, + { + "epoch": 16.9, + "learning_rate": 1.3109988914890996e-06, + "loss": 0.2501, + "step": 45760 + }, + { + "epoch": 16.9, + "learning_rate": 1.3109619411257545e-06, + "loss": 0.2811, + "step": 45761 + }, + { + "epoch": 16.9, + "learning_rate": 1.310924990762409e-06, + "loss": 0.419, + "step": 45762 + }, + { + "epoch": 16.9, + "learning_rate": 1.310888040399064e-06, + "loss": 0.2035, + "step": 45763 + }, + { + "epoch": 16.9, + "learning_rate": 1.3108510900357188e-06, + "loss": 0.1857, + "step": 45764 + }, + { + "epoch": 16.9, + "learning_rate": 1.3108141396723734e-06, + "loss": 0.288, + "step": 45765 + }, + { + "epoch": 16.9, + "learning_rate": 1.3107771893090283e-06, + "loss": 0.2932, + "step": 45766 + }, + { + "epoch": 16.9, + "learning_rate": 1.3107402389456831e-06, + "loss": 0.2031, + "step": 45767 + }, + { + "epoch": 16.9, + "learning_rate": 1.3107032885823378e-06, + "loss": 0.2339, + "step": 45768 + }, + { + "epoch": 16.9, + "learning_rate": 1.3106663382189924e-06, + "loss": 0.1698, + "step": 45769 + }, + { + "epoch": 16.9, + "learning_rate": 1.3106293878556472e-06, + "loss": 0.3242, + "step": 45770 + }, + { + "epoch": 16.9, + "learning_rate": 1.310592437492302e-06, + "loss": 0.1684, + "step": 45771 + }, + { + "epoch": 16.9, + "learning_rate": 1.3105554871289567e-06, + "loss": 0.1444, + "step": 45772 + }, + { + "epoch": 16.9, + "learning_rate": 1.3105185367656116e-06, + "loss": 0.2271, + "step": 45773 + }, + { + "epoch": 16.9, + "learning_rate": 1.3104815864022664e-06, + "loss": 0.2335, + "step": 45774 + }, + { + "epoch": 16.9, + "learning_rate": 1.310444636038921e-06, + "loss": 0.3419, + "step": 45775 + }, + { + "epoch": 16.9, + "learning_rate": 1.310407685675576e-06, + "loss": 0.2943, + "step": 45776 + }, + { + "epoch": 16.9, + "learning_rate": 1.3103707353122308e-06, + "loss": 0.1823, + "step": 45777 + }, + { + "epoch": 16.9, + "learning_rate": 1.3103337849488852e-06, + "loss": 0.1864, + "step": 45778 + }, + { + "epoch": 16.91, + "learning_rate": 1.31029683458554e-06, + "loss": 0.1766, + "step": 45779 + }, + { + "epoch": 16.91, + "learning_rate": 1.3102598842221949e-06, + "loss": 0.1845, + "step": 45780 + }, + { + "epoch": 16.91, + "learning_rate": 1.3102229338588495e-06, + "loss": 0.3818, + "step": 45781 + }, + { + "epoch": 16.91, + "learning_rate": 1.3101859834955044e-06, + "loss": 0.3048, + "step": 45782 + }, + { + "epoch": 16.91, + "learning_rate": 1.3101490331321592e-06, + "loss": 0.1536, + "step": 45783 + }, + { + "epoch": 16.91, + "learning_rate": 1.3101120827688139e-06, + "loss": 0.2645, + "step": 45784 + }, + { + "epoch": 16.91, + "learning_rate": 1.3100751324054687e-06, + "loss": 0.2411, + "step": 45785 + }, + { + "epoch": 16.91, + "learning_rate": 1.3100381820421236e-06, + "loss": 0.2983, + "step": 45786 + }, + { + "epoch": 16.91, + "learning_rate": 1.3100012316787782e-06, + "loss": 0.3283, + "step": 45787 + }, + { + "epoch": 16.91, + "learning_rate": 1.3099642813154328e-06, + "loss": 0.2105, + "step": 45788 + }, + { + "epoch": 16.91, + "learning_rate": 1.3099273309520877e-06, + "loss": 0.2097, + "step": 45789 + }, + { + "epoch": 16.91, + "learning_rate": 1.3098903805887425e-06, + "loss": 0.3521, + "step": 45790 + }, + { + "epoch": 16.91, + "learning_rate": 1.3098534302253972e-06, + "loss": 0.3267, + "step": 45791 + }, + { + "epoch": 16.91, + "learning_rate": 1.309816479862052e-06, + "loss": 0.3131, + "step": 45792 + }, + { + "epoch": 16.91, + "learning_rate": 1.3097795294987069e-06, + "loss": 0.2379, + "step": 45793 + }, + { + "epoch": 16.91, + "learning_rate": 1.3097425791353615e-06, + "loss": 0.4055, + "step": 45794 + }, + { + "epoch": 16.91, + "learning_rate": 1.3097056287720163e-06, + "loss": 0.2055, + "step": 45795 + }, + { + "epoch": 16.91, + "learning_rate": 1.3096686784086712e-06, + "loss": 0.2323, + "step": 45796 + }, + { + "epoch": 16.91, + "learning_rate": 1.3096317280453258e-06, + "loss": 0.3223, + "step": 45797 + }, + { + "epoch": 16.91, + "learning_rate": 1.3095947776819805e-06, + "loss": 0.2124, + "step": 45798 + }, + { + "epoch": 16.91, + "learning_rate": 1.3095578273186353e-06, + "loss": 0.2816, + "step": 45799 + }, + { + "epoch": 16.91, + "learning_rate": 1.30952087695529e-06, + "loss": 0.1994, + "step": 45800 + }, + { + "epoch": 16.91, + "learning_rate": 1.3094839265919448e-06, + "loss": 0.1969, + "step": 45801 + }, + { + "epoch": 16.91, + "learning_rate": 1.3094469762285996e-06, + "loss": 0.1346, + "step": 45802 + }, + { + "epoch": 16.91, + "learning_rate": 1.3094100258652543e-06, + "loss": 0.2703, + "step": 45803 + }, + { + "epoch": 16.91, + "learning_rate": 1.3093730755019091e-06, + "loss": 0.3368, + "step": 45804 + }, + { + "epoch": 16.91, + "learning_rate": 1.309336125138564e-06, + "loss": 0.1137, + "step": 45805 + }, + { + "epoch": 16.92, + "learning_rate": 1.3092991747752186e-06, + "loss": 0.258, + "step": 45806 + }, + { + "epoch": 16.92, + "learning_rate": 1.3092622244118735e-06, + "loss": 0.2818, + "step": 45807 + }, + { + "epoch": 16.92, + "learning_rate": 1.309225274048528e-06, + "loss": 0.1749, + "step": 45808 + }, + { + "epoch": 16.92, + "learning_rate": 1.309188323685183e-06, + "loss": 0.2878, + "step": 45809 + }, + { + "epoch": 16.92, + "learning_rate": 1.3091513733218376e-06, + "loss": 0.2671, + "step": 45810 + }, + { + "epoch": 16.92, + "learning_rate": 1.3091144229584924e-06, + "loss": 0.2505, + "step": 45811 + }, + { + "epoch": 16.92, + "learning_rate": 1.3090774725951473e-06, + "loss": 0.1773, + "step": 45812 + }, + { + "epoch": 16.92, + "learning_rate": 1.309040522231802e-06, + "loss": 0.205, + "step": 45813 + }, + { + "epoch": 16.92, + "learning_rate": 1.3090035718684568e-06, + "loss": 0.2738, + "step": 45814 + }, + { + "epoch": 16.92, + "learning_rate": 1.3089666215051116e-06, + "loss": 0.2199, + "step": 45815 + }, + { + "epoch": 16.92, + "learning_rate": 1.3089296711417662e-06, + "loss": 0.1337, + "step": 45816 + }, + { + "epoch": 16.92, + "learning_rate": 1.308892720778421e-06, + "loss": 0.2069, + "step": 45817 + }, + { + "epoch": 16.92, + "learning_rate": 1.3088557704150757e-06, + "loss": 0.2456, + "step": 45818 + }, + { + "epoch": 16.92, + "learning_rate": 1.3088188200517306e-06, + "loss": 0.1487, + "step": 45819 + }, + { + "epoch": 16.92, + "learning_rate": 1.3087818696883852e-06, + "loss": 0.3063, + "step": 45820 + }, + { + "epoch": 16.92, + "learning_rate": 1.30874491932504e-06, + "loss": 0.1801, + "step": 45821 + }, + { + "epoch": 16.92, + "learning_rate": 1.308707968961695e-06, + "loss": 0.3155, + "step": 45822 + }, + { + "epoch": 16.92, + "learning_rate": 1.3086710185983495e-06, + "loss": 0.2675, + "step": 45823 + }, + { + "epoch": 16.92, + "learning_rate": 1.3086340682350044e-06, + "loss": 0.2503, + "step": 45824 + }, + { + "epoch": 16.92, + "learning_rate": 1.3085971178716592e-06, + "loss": 0.3128, + "step": 45825 + }, + { + "epoch": 16.92, + "learning_rate": 1.3085601675083139e-06, + "loss": 0.2067, + "step": 45826 + }, + { + "epoch": 16.92, + "learning_rate": 1.3085232171449687e-06, + "loss": 0.2444, + "step": 45827 + }, + { + "epoch": 16.92, + "learning_rate": 1.3084862667816234e-06, + "loss": 0.3818, + "step": 45828 + }, + { + "epoch": 16.92, + "learning_rate": 1.308449316418278e-06, + "loss": 0.3099, + "step": 45829 + }, + { + "epoch": 16.92, + "learning_rate": 1.3084123660549328e-06, + "loss": 0.2373, + "step": 45830 + }, + { + "epoch": 16.92, + "learning_rate": 1.3083754156915877e-06, + "loss": 0.1653, + "step": 45831 + }, + { + "epoch": 16.92, + "learning_rate": 1.3083384653282423e-06, + "loss": 0.3253, + "step": 45832 + }, + { + "epoch": 16.93, + "learning_rate": 1.3083015149648972e-06, + "loss": 0.1541, + "step": 45833 + }, + { + "epoch": 16.93, + "learning_rate": 1.308264564601552e-06, + "loss": 0.177, + "step": 45834 + }, + { + "epoch": 16.93, + "learning_rate": 1.3082276142382067e-06, + "loss": 0.3591, + "step": 45835 + }, + { + "epoch": 16.93, + "learning_rate": 1.3081906638748615e-06, + "loss": 0.1934, + "step": 45836 + }, + { + "epoch": 16.93, + "learning_rate": 1.3081537135115164e-06, + "loss": 0.2272, + "step": 45837 + }, + { + "epoch": 16.93, + "learning_rate": 1.308116763148171e-06, + "loss": 0.2861, + "step": 45838 + }, + { + "epoch": 16.93, + "learning_rate": 1.3080798127848256e-06, + "loss": 0.2756, + "step": 45839 + }, + { + "epoch": 16.93, + "learning_rate": 1.3080428624214805e-06, + "loss": 0.4937, + "step": 45840 + }, + { + "epoch": 16.93, + "learning_rate": 1.3080059120581353e-06, + "loss": 0.1895, + "step": 45841 + }, + { + "epoch": 16.93, + "learning_rate": 1.30796896169479e-06, + "loss": 0.217, + "step": 45842 + }, + { + "epoch": 16.93, + "learning_rate": 1.3079320113314448e-06, + "loss": 0.3551, + "step": 45843 + }, + { + "epoch": 16.93, + "learning_rate": 1.3078950609680997e-06, + "loss": 0.3111, + "step": 45844 + }, + { + "epoch": 16.93, + "learning_rate": 1.3078581106047543e-06, + "loss": 0.1628, + "step": 45845 + }, + { + "epoch": 16.93, + "learning_rate": 1.3078211602414091e-06, + "loss": 0.4209, + "step": 45846 + }, + { + "epoch": 16.93, + "learning_rate": 1.307784209878064e-06, + "loss": 0.2201, + "step": 45847 + }, + { + "epoch": 16.93, + "learning_rate": 1.3077472595147184e-06, + "loss": 0.2616, + "step": 45848 + }, + { + "epoch": 16.93, + "learning_rate": 1.3077103091513733e-06, + "loss": 0.2307, + "step": 45849 + }, + { + "epoch": 16.93, + "learning_rate": 1.3076733587880281e-06, + "loss": 0.2465, + "step": 45850 + }, + { + "epoch": 16.93, + "learning_rate": 1.3076364084246828e-06, + "loss": 0.4338, + "step": 45851 + }, + { + "epoch": 16.93, + "learning_rate": 1.3075994580613376e-06, + "loss": 0.153, + "step": 45852 + }, + { + "epoch": 16.93, + "learning_rate": 1.3075625076979924e-06, + "loss": 0.1921, + "step": 45853 + }, + { + "epoch": 16.93, + "learning_rate": 1.307525557334647e-06, + "loss": 0.2019, + "step": 45854 + }, + { + "epoch": 16.93, + "learning_rate": 1.307488606971302e-06, + "loss": 0.1905, + "step": 45855 + }, + { + "epoch": 16.93, + "learning_rate": 1.3074516566079568e-06, + "loss": 0.2721, + "step": 45856 + }, + { + "epoch": 16.93, + "learning_rate": 1.3074147062446114e-06, + "loss": 0.2457, + "step": 45857 + }, + { + "epoch": 16.93, + "learning_rate": 1.307377755881266e-06, + "loss": 0.2121, + "step": 45858 + }, + { + "epoch": 16.93, + "learning_rate": 1.307340805517921e-06, + "loss": 0.1841, + "step": 45859 + }, + { + "epoch": 16.94, + "learning_rate": 1.3073038551545758e-06, + "loss": 0.1571, + "step": 45860 + }, + { + "epoch": 16.94, + "learning_rate": 1.3072669047912304e-06, + "loss": 0.2706, + "step": 45861 + }, + { + "epoch": 16.94, + "learning_rate": 1.3072299544278852e-06, + "loss": 0.1744, + "step": 45862 + }, + { + "epoch": 16.94, + "learning_rate": 1.30719300406454e-06, + "loss": 0.2471, + "step": 45863 + }, + { + "epoch": 16.94, + "learning_rate": 1.3071560537011947e-06, + "loss": 0.363, + "step": 45864 + }, + { + "epoch": 16.94, + "learning_rate": 1.3071191033378496e-06, + "loss": 0.2047, + "step": 45865 + }, + { + "epoch": 16.94, + "learning_rate": 1.3070821529745044e-06, + "loss": 0.2042, + "step": 45866 + }, + { + "epoch": 16.94, + "learning_rate": 1.307045202611159e-06, + "loss": 0.4271, + "step": 45867 + }, + { + "epoch": 16.94, + "learning_rate": 1.3070082522478137e-06, + "loss": 0.1263, + "step": 45868 + }, + { + "epoch": 16.94, + "learning_rate": 1.3069713018844685e-06, + "loss": 0.3878, + "step": 45869 + }, + { + "epoch": 16.94, + "learning_rate": 1.3069343515211234e-06, + "loss": 0.2956, + "step": 45870 + }, + { + "epoch": 16.94, + "learning_rate": 1.306897401157778e-06, + "loss": 0.2908, + "step": 45871 + }, + { + "epoch": 16.94, + "learning_rate": 1.3068604507944329e-06, + "loss": 0.2722, + "step": 45872 + }, + { + "epoch": 16.94, + "learning_rate": 1.3068235004310877e-06, + "loss": 0.283, + "step": 45873 + }, + { + "epoch": 16.94, + "learning_rate": 1.3067865500677424e-06, + "loss": 0.2595, + "step": 45874 + }, + { + "epoch": 16.94, + "learning_rate": 1.3067495997043972e-06, + "loss": 0.2016, + "step": 45875 + }, + { + "epoch": 16.94, + "learning_rate": 1.3067126493410518e-06, + "loss": 0.2425, + "step": 45876 + }, + { + "epoch": 16.94, + "learning_rate": 1.3066756989777067e-06, + "loss": 0.2285, + "step": 45877 + }, + { + "epoch": 16.94, + "learning_rate": 1.3066387486143613e-06, + "loss": 0.1747, + "step": 45878 + }, + { + "epoch": 16.94, + "learning_rate": 1.3066017982510162e-06, + "loss": 0.2841, + "step": 45879 + }, + { + "epoch": 16.94, + "learning_rate": 1.3065648478876708e-06, + "loss": 0.4813, + "step": 45880 + }, + { + "epoch": 16.94, + "learning_rate": 1.3065278975243257e-06, + "loss": 0.3186, + "step": 45881 + }, + { + "epoch": 16.94, + "learning_rate": 1.3064909471609805e-06, + "loss": 0.2893, + "step": 45882 + }, + { + "epoch": 16.94, + "learning_rate": 1.3064539967976351e-06, + "loss": 0.25, + "step": 45883 + }, + { + "epoch": 16.94, + "learning_rate": 1.30641704643429e-06, + "loss": 0.183, + "step": 45884 + }, + { + "epoch": 16.94, + "learning_rate": 1.3063800960709448e-06, + "loss": 0.291, + "step": 45885 + }, + { + "epoch": 16.94, + "learning_rate": 1.3063431457075995e-06, + "loss": 0.2931, + "step": 45886 + }, + { + "epoch": 16.94, + "learning_rate": 1.3063061953442543e-06, + "loss": 0.239, + "step": 45887 + }, + { + "epoch": 16.95, + "learning_rate": 1.306269244980909e-06, + "loss": 0.3131, + "step": 45888 + }, + { + "epoch": 16.95, + "learning_rate": 1.3062322946175638e-06, + "loss": 0.1712, + "step": 45889 + }, + { + "epoch": 16.95, + "learning_rate": 1.3061953442542184e-06, + "loss": 0.259, + "step": 45890 + }, + { + "epoch": 16.95, + "learning_rate": 1.3061583938908733e-06, + "loss": 0.3508, + "step": 45891 + }, + { + "epoch": 16.95, + "learning_rate": 1.3061214435275281e-06, + "loss": 0.1163, + "step": 45892 + }, + { + "epoch": 16.95, + "learning_rate": 1.3060844931641828e-06, + "loss": 0.1957, + "step": 45893 + }, + { + "epoch": 16.95, + "learning_rate": 1.3060475428008376e-06, + "loss": 0.1756, + "step": 45894 + }, + { + "epoch": 16.95, + "learning_rate": 1.3060105924374925e-06, + "loss": 0.3202, + "step": 45895 + }, + { + "epoch": 16.95, + "learning_rate": 1.3059736420741471e-06, + "loss": 0.2355, + "step": 45896 + }, + { + "epoch": 16.95, + "learning_rate": 1.305936691710802e-06, + "loss": 0.1565, + "step": 45897 + }, + { + "epoch": 16.95, + "learning_rate": 1.3058997413474566e-06, + "loss": 0.3614, + "step": 45898 + }, + { + "epoch": 16.95, + "learning_rate": 1.3058627909841112e-06, + "loss": 0.2469, + "step": 45899 + }, + { + "epoch": 16.95, + "learning_rate": 1.305825840620766e-06, + "loss": 0.2233, + "step": 45900 + }, + { + "epoch": 16.95, + "learning_rate": 1.305788890257421e-06, + "loss": 0.2002, + "step": 45901 + }, + { + "epoch": 16.95, + "learning_rate": 1.3057519398940756e-06, + "loss": 0.2301, + "step": 45902 + }, + { + "epoch": 16.95, + "learning_rate": 1.3057149895307304e-06, + "loss": 0.2362, + "step": 45903 + }, + { + "epoch": 16.95, + "learning_rate": 1.3056780391673853e-06, + "loss": 0.3011, + "step": 45904 + }, + { + "epoch": 16.95, + "learning_rate": 1.30564108880404e-06, + "loss": 0.2583, + "step": 45905 + }, + { + "epoch": 16.95, + "learning_rate": 1.3056041384406947e-06, + "loss": 0.3095, + "step": 45906 + }, + { + "epoch": 16.95, + "learning_rate": 1.3055671880773496e-06, + "loss": 0.2454, + "step": 45907 + }, + { + "epoch": 16.95, + "learning_rate": 1.3055302377140042e-06, + "loss": 0.2426, + "step": 45908 + }, + { + "epoch": 16.95, + "learning_rate": 1.3054932873506589e-06, + "loss": 0.2679, + "step": 45909 + }, + { + "epoch": 16.95, + "learning_rate": 1.3054563369873137e-06, + "loss": 0.2573, + "step": 45910 + }, + { + "epoch": 16.95, + "learning_rate": 1.3054193866239686e-06, + "loss": 0.1274, + "step": 45911 + }, + { + "epoch": 16.95, + "learning_rate": 1.3053824362606232e-06, + "loss": 0.2924, + "step": 45912 + }, + { + "epoch": 16.95, + "learning_rate": 1.305345485897278e-06, + "loss": 0.3263, + "step": 45913 + }, + { + "epoch": 16.95, + "learning_rate": 1.305308535533933e-06, + "loss": 0.2042, + "step": 45914 + }, + { + "epoch": 16.96, + "learning_rate": 1.3052715851705875e-06, + "loss": 0.3156, + "step": 45915 + }, + { + "epoch": 16.96, + "learning_rate": 1.3052346348072424e-06, + "loss": 0.1733, + "step": 45916 + }, + { + "epoch": 16.96, + "learning_rate": 1.3051976844438972e-06, + "loss": 0.1454, + "step": 45917 + }, + { + "epoch": 16.96, + "learning_rate": 1.3051607340805517e-06, + "loss": 0.1495, + "step": 45918 + }, + { + "epoch": 16.96, + "learning_rate": 1.3051237837172065e-06, + "loss": 0.0863, + "step": 45919 + }, + { + "epoch": 16.96, + "learning_rate": 1.3050868333538613e-06, + "loss": 0.3241, + "step": 45920 + }, + { + "epoch": 16.96, + "learning_rate": 1.305049882990516e-06, + "loss": 0.2684, + "step": 45921 + }, + { + "epoch": 16.96, + "learning_rate": 1.3050129326271708e-06, + "loss": 0.2494, + "step": 45922 + }, + { + "epoch": 16.96, + "learning_rate": 1.3049759822638257e-06, + "loss": 0.2938, + "step": 45923 + }, + { + "epoch": 16.96, + "learning_rate": 1.3049390319004803e-06, + "loss": 0.2073, + "step": 45924 + }, + { + "epoch": 16.96, + "learning_rate": 1.3049020815371352e-06, + "loss": 0.3037, + "step": 45925 + }, + { + "epoch": 16.96, + "learning_rate": 1.30486513117379e-06, + "loss": 0.1828, + "step": 45926 + }, + { + "epoch": 16.96, + "learning_rate": 1.3048281808104447e-06, + "loss": 0.4364, + "step": 45927 + }, + { + "epoch": 16.96, + "learning_rate": 1.3047912304470993e-06, + "loss": 0.1294, + "step": 45928 + }, + { + "epoch": 16.96, + "learning_rate": 1.3047542800837541e-06, + "loss": 0.2018, + "step": 45929 + }, + { + "epoch": 16.96, + "learning_rate": 1.304717329720409e-06, + "loss": 0.18, + "step": 45930 + }, + { + "epoch": 16.96, + "learning_rate": 1.3046803793570636e-06, + "loss": 0.2279, + "step": 45931 + }, + { + "epoch": 16.96, + "learning_rate": 1.3046434289937185e-06, + "loss": 0.1572, + "step": 45932 + }, + { + "epoch": 16.96, + "learning_rate": 1.3046064786303733e-06, + "loss": 0.2793, + "step": 45933 + }, + { + "epoch": 16.96, + "learning_rate": 1.304569528267028e-06, + "loss": 0.2353, + "step": 45934 + }, + { + "epoch": 16.96, + "learning_rate": 1.3045325779036828e-06, + "loss": 0.3373, + "step": 45935 + }, + { + "epoch": 16.96, + "learning_rate": 1.3044956275403376e-06, + "loss": 0.3586, + "step": 45936 + }, + { + "epoch": 16.96, + "learning_rate": 1.3044586771769923e-06, + "loss": 0.4134, + "step": 45937 + }, + { + "epoch": 16.96, + "learning_rate": 1.304421726813647e-06, + "loss": 0.3065, + "step": 45938 + }, + { + "epoch": 16.96, + "learning_rate": 1.3043847764503018e-06, + "loss": 0.2965, + "step": 45939 + }, + { + "epoch": 16.96, + "learning_rate": 1.3043478260869566e-06, + "loss": 0.4491, + "step": 45940 + }, + { + "epoch": 16.96, + "learning_rate": 1.3043108757236113e-06, + "loss": 0.4017, + "step": 45941 + }, + { + "epoch": 16.97, + "learning_rate": 1.304273925360266e-06, + "loss": 0.3245, + "step": 45942 + }, + { + "epoch": 16.97, + "learning_rate": 1.304236974996921e-06, + "loss": 0.2115, + "step": 45943 + }, + { + "epoch": 16.97, + "learning_rate": 1.3042000246335756e-06, + "loss": 0.1957, + "step": 45944 + }, + { + "epoch": 16.97, + "learning_rate": 1.3041630742702304e-06, + "loss": 0.1551, + "step": 45945 + }, + { + "epoch": 16.97, + "learning_rate": 1.3041261239068853e-06, + "loss": 0.3201, + "step": 45946 + }, + { + "epoch": 16.97, + "learning_rate": 1.30408917354354e-06, + "loss": 0.2415, + "step": 45947 + }, + { + "epoch": 16.97, + "learning_rate": 1.3040522231801946e-06, + "loss": 0.2433, + "step": 45948 + }, + { + "epoch": 16.97, + "learning_rate": 1.3040152728168494e-06, + "loss": 0.1937, + "step": 45949 + }, + { + "epoch": 16.97, + "learning_rate": 1.303978322453504e-06, + "loss": 0.221, + "step": 45950 + }, + { + "epoch": 16.97, + "learning_rate": 1.3039413720901589e-06, + "loss": 0.2757, + "step": 45951 + }, + { + "epoch": 16.97, + "learning_rate": 1.3039044217268137e-06, + "loss": 0.1871, + "step": 45952 + }, + { + "epoch": 16.97, + "learning_rate": 1.3038674713634684e-06, + "loss": 0.3428, + "step": 45953 + }, + { + "epoch": 16.97, + "learning_rate": 1.3038305210001232e-06, + "loss": 0.2478, + "step": 45954 + }, + { + "epoch": 16.97, + "learning_rate": 1.303793570636778e-06, + "loss": 0.1369, + "step": 45955 + }, + { + "epoch": 16.97, + "learning_rate": 1.3037566202734327e-06, + "loss": 0.5278, + "step": 45956 + }, + { + "epoch": 16.97, + "learning_rate": 1.3037196699100876e-06, + "loss": 0.2442, + "step": 45957 + }, + { + "epoch": 16.97, + "learning_rate": 1.3036827195467422e-06, + "loss": 0.2415, + "step": 45958 + }, + { + "epoch": 16.97, + "learning_rate": 1.303645769183397e-06, + "loss": 0.16, + "step": 45959 + }, + { + "epoch": 16.97, + "learning_rate": 1.3036088188200517e-06, + "loss": 0.1592, + "step": 45960 + }, + { + "epoch": 16.97, + "learning_rate": 1.3035718684567065e-06, + "loss": 0.2515, + "step": 45961 + }, + { + "epoch": 16.97, + "learning_rate": 1.3035349180933614e-06, + "loss": 0.1772, + "step": 45962 + }, + { + "epoch": 16.97, + "learning_rate": 1.303497967730016e-06, + "loss": 0.3236, + "step": 45963 + }, + { + "epoch": 16.97, + "learning_rate": 1.3034610173666709e-06, + "loss": 0.2519, + "step": 45964 + }, + { + "epoch": 16.97, + "learning_rate": 1.3034240670033257e-06, + "loss": 0.312, + "step": 45965 + }, + { + "epoch": 16.97, + "learning_rate": 1.3033871166399803e-06, + "loss": 0.1515, + "step": 45966 + }, + { + "epoch": 16.97, + "learning_rate": 1.3033501662766352e-06, + "loss": 0.2559, + "step": 45967 + }, + { + "epoch": 16.97, + "learning_rate": 1.3033132159132898e-06, + "loss": 0.2334, + "step": 45968 + }, + { + "epoch": 16.98, + "learning_rate": 1.3032762655499445e-06, + "loss": 0.2698, + "step": 45969 + }, + { + "epoch": 16.98, + "learning_rate": 1.3032393151865993e-06, + "loss": 0.2997, + "step": 45970 + }, + { + "epoch": 16.98, + "learning_rate": 1.3032023648232542e-06, + "loss": 0.2617, + "step": 45971 + }, + { + "epoch": 16.98, + "learning_rate": 1.3031654144599088e-06, + "loss": 0.1322, + "step": 45972 + }, + { + "epoch": 16.98, + "learning_rate": 1.3031284640965636e-06, + "loss": 0.2851, + "step": 45973 + }, + { + "epoch": 16.98, + "learning_rate": 1.3030915137332185e-06, + "loss": 0.3407, + "step": 45974 + }, + { + "epoch": 16.98, + "learning_rate": 1.3030545633698731e-06, + "loss": 0.2795, + "step": 45975 + }, + { + "epoch": 16.98, + "learning_rate": 1.303017613006528e-06, + "loss": 0.1338, + "step": 45976 + }, + { + "epoch": 16.98, + "learning_rate": 1.3029806626431828e-06, + "loss": 0.2325, + "step": 45977 + }, + { + "epoch": 16.98, + "learning_rate": 1.3029437122798375e-06, + "loss": 0.2634, + "step": 45978 + }, + { + "epoch": 16.98, + "learning_rate": 1.302906761916492e-06, + "loss": 0.3107, + "step": 45979 + }, + { + "epoch": 16.98, + "learning_rate": 1.302869811553147e-06, + "loss": 0.3141, + "step": 45980 + }, + { + "epoch": 16.98, + "learning_rate": 1.3028328611898018e-06, + "loss": 0.3222, + "step": 45981 + }, + { + "epoch": 16.98, + "learning_rate": 1.3027959108264564e-06, + "loss": 0.3711, + "step": 45982 + }, + { + "epoch": 16.98, + "learning_rate": 1.3027589604631113e-06, + "loss": 0.3358, + "step": 45983 + }, + { + "epoch": 16.98, + "learning_rate": 1.3027220100997661e-06, + "loss": 0.2494, + "step": 45984 + }, + { + "epoch": 16.98, + "learning_rate": 1.3026850597364208e-06, + "loss": 0.1494, + "step": 45985 + }, + { + "epoch": 16.98, + "learning_rate": 1.3026481093730756e-06, + "loss": 0.2799, + "step": 45986 + }, + { + "epoch": 16.98, + "learning_rate": 1.3026111590097302e-06, + "loss": 0.2356, + "step": 45987 + }, + { + "epoch": 16.98, + "learning_rate": 1.3025742086463849e-06, + "loss": 0.1742, + "step": 45988 + }, + { + "epoch": 16.98, + "learning_rate": 1.3025372582830397e-06, + "loss": 0.1323, + "step": 45989 + }, + { + "epoch": 16.98, + "learning_rate": 1.3025003079196946e-06, + "loss": 0.1708, + "step": 45990 + }, + { + "epoch": 16.98, + "learning_rate": 1.3024633575563492e-06, + "loss": 0.2505, + "step": 45991 + }, + { + "epoch": 16.98, + "learning_rate": 1.302426407193004e-06, + "loss": 0.2373, + "step": 45992 + }, + { + "epoch": 16.98, + "learning_rate": 1.302389456829659e-06, + "loss": 0.3191, + "step": 45993 + }, + { + "epoch": 16.98, + "learning_rate": 1.3023525064663136e-06, + "loss": 0.2692, + "step": 45994 + }, + { + "epoch": 16.98, + "learning_rate": 1.3023155561029684e-06, + "loss": 0.228, + "step": 45995 + }, + { + "epoch": 16.99, + "learning_rate": 1.3022786057396232e-06, + "loss": 0.2209, + "step": 45996 + }, + { + "epoch": 16.99, + "learning_rate": 1.3022416553762779e-06, + "loss": 0.2482, + "step": 45997 + }, + { + "epoch": 16.99, + "learning_rate": 1.3022047050129325e-06, + "loss": 0.3305, + "step": 45998 + }, + { + "epoch": 16.99, + "learning_rate": 1.3021677546495874e-06, + "loss": 0.2911, + "step": 45999 + }, + { + "epoch": 16.99, + "learning_rate": 1.3021308042862422e-06, + "loss": 0.1631, + "step": 46000 + }, + { + "epoch": 16.99, + "learning_rate": 1.3020938539228969e-06, + "loss": 0.2008, + "step": 46001 + }, + { + "epoch": 16.99, + "learning_rate": 1.3020569035595517e-06, + "loss": 0.3028, + "step": 46002 + }, + { + "epoch": 16.99, + "learning_rate": 1.3020199531962065e-06, + "loss": 0.2591, + "step": 46003 + }, + { + "epoch": 16.99, + "learning_rate": 1.3019830028328612e-06, + "loss": 0.222, + "step": 46004 + }, + { + "epoch": 16.99, + "learning_rate": 1.301946052469516e-06, + "loss": 0.2682, + "step": 46005 + }, + { + "epoch": 16.99, + "learning_rate": 1.3019091021061709e-06, + "loss": 0.2563, + "step": 46006 + }, + { + "epoch": 16.99, + "learning_rate": 1.3018721517428255e-06, + "loss": 0.2015, + "step": 46007 + }, + { + "epoch": 16.99, + "learning_rate": 1.3018352013794802e-06, + "loss": 0.3012, + "step": 46008 + }, + { + "epoch": 16.99, + "learning_rate": 1.301798251016135e-06, + "loss": 0.4832, + "step": 46009 + }, + { + "epoch": 16.99, + "learning_rate": 1.3017613006527899e-06, + "loss": 0.2038, + "step": 46010 + }, + { + "epoch": 16.99, + "learning_rate": 1.3017243502894445e-06, + "loss": 0.2833, + "step": 46011 + }, + { + "epoch": 16.99, + "learning_rate": 1.3016873999260993e-06, + "loss": 0.2342, + "step": 46012 + }, + { + "epoch": 16.99, + "learning_rate": 1.3016504495627542e-06, + "loss": 0.1618, + "step": 46013 + }, + { + "epoch": 16.99, + "learning_rate": 1.3016134991994088e-06, + "loss": 0.2481, + "step": 46014 + }, + { + "epoch": 16.99, + "learning_rate": 1.3015765488360637e-06, + "loss": 0.2369, + "step": 46015 + }, + { + "epoch": 16.99, + "learning_rate": 1.3015395984727185e-06, + "loss": 0.1816, + "step": 46016 + }, + { + "epoch": 16.99, + "learning_rate": 1.301502648109373e-06, + "loss": 0.2273, + "step": 46017 + }, + { + "epoch": 16.99, + "learning_rate": 1.3014656977460278e-06, + "loss": 0.1963, + "step": 46018 + }, + { + "epoch": 16.99, + "learning_rate": 1.3014287473826826e-06, + "loss": 0.1769, + "step": 46019 + }, + { + "epoch": 16.99, + "learning_rate": 1.3013917970193373e-06, + "loss": 0.1856, + "step": 46020 + }, + { + "epoch": 16.99, + "learning_rate": 1.3013548466559921e-06, + "loss": 0.1554, + "step": 46021 + }, + { + "epoch": 16.99, + "learning_rate": 1.301317896292647e-06, + "loss": 0.2524, + "step": 46022 + }, + { + "epoch": 17.0, + "learning_rate": 1.3012809459293016e-06, + "loss": 0.3699, + "step": 46023 + }, + { + "epoch": 17.0, + "learning_rate": 1.3012439955659565e-06, + "loss": 0.2878, + "step": 46024 + }, + { + "epoch": 17.0, + "learning_rate": 1.3012070452026113e-06, + "loss": 0.2285, + "step": 46025 + }, + { + "epoch": 17.0, + "learning_rate": 1.301170094839266e-06, + "loss": 0.2741, + "step": 46026 + }, + { + "epoch": 17.0, + "learning_rate": 1.3011331444759206e-06, + "loss": 0.1958, + "step": 46027 + }, + { + "epoch": 17.0, + "learning_rate": 1.3010961941125754e-06, + "loss": 0.1721, + "step": 46028 + }, + { + "epoch": 17.0, + "learning_rate": 1.3010592437492303e-06, + "loss": 0.3006, + "step": 46029 + }, + { + "epoch": 17.0, + "learning_rate": 1.301022293385885e-06, + "loss": 0.2134, + "step": 46030 + }, + { + "epoch": 17.0, + "learning_rate": 1.3009853430225398e-06, + "loss": 0.1983, + "step": 46031 + }, + { + "epoch": 17.0, + "learning_rate": 1.3009483926591946e-06, + "loss": 0.2681, + "step": 46032 + }, + { + "epoch": 17.0, + "learning_rate": 1.3009114422958492e-06, + "loss": 0.2721, + "step": 46033 + }, + { + "epoch": 17.0, + "learning_rate": 1.300874491932504e-06, + "loss": 0.274, + "step": 46034 + }, + { + "epoch": 17.0, + "learning_rate": 1.300837541569159e-06, + "loss": 0.1914, + "step": 46035 + }, + { + "epoch": 17.0, + "learning_rate": 1.3008005912058136e-06, + "loss": 0.7425, + "step": 46036 + }, + { + "epoch": 17.0, + "eval_accuracy": 0.8605300826522603, + "eval_loss": 0.3595193028450012, + "eval_runtime": 25.1607, + "eval_samples_per_second": 1721.496, + "eval_steps_per_second": 26.907, + "step": 46036 + }, + { + "epoch": 17.0, + "learning_rate": 1.3007636408424682e-06, + "loss": 0.3202, + "step": 46037 + }, + { + "epoch": 17.0, + "learning_rate": 1.300726690479123e-06, + "loss": 0.2473, + "step": 46038 + }, + { + "epoch": 17.0, + "learning_rate": 1.3006897401157777e-06, + "loss": 0.1982, + "step": 46039 + }, + { + "epoch": 17.0, + "learning_rate": 1.3006527897524325e-06, + "loss": 0.2346, + "step": 46040 + }, + { + "epoch": 17.0, + "learning_rate": 1.3006158393890874e-06, + "loss": 0.2255, + "step": 46041 + }, + { + "epoch": 17.0, + "learning_rate": 1.300578889025742e-06, + "loss": 0.1991, + "step": 46042 + }, + { + "epoch": 17.0, + "learning_rate": 1.3005419386623969e-06, + "loss": 0.1497, + "step": 46043 + }, + { + "epoch": 17.0, + "learning_rate": 1.3005049882990517e-06, + "loss": 0.2877, + "step": 46044 + }, + { + "epoch": 17.0, + "learning_rate": 1.3004680379357064e-06, + "loss": 0.1767, + "step": 46045 + }, + { + "epoch": 17.0, + "learning_rate": 1.3004310875723612e-06, + "loss": 0.2475, + "step": 46046 + }, + { + "epoch": 17.0, + "learning_rate": 1.3003941372090158e-06, + "loss": 0.1932, + "step": 46047 + }, + { + "epoch": 17.0, + "learning_rate": 1.3003571868456707e-06, + "loss": 0.273, + "step": 46048 + }, + { + "epoch": 17.0, + "learning_rate": 1.3003202364823253e-06, + "loss": 0.2659, + "step": 46049 + }, + { + "epoch": 17.01, + "learning_rate": 1.3002832861189802e-06, + "loss": 0.226, + "step": 46050 + }, + { + "epoch": 17.01, + "learning_rate": 1.300246335755635e-06, + "loss": 0.1748, + "step": 46051 + }, + { + "epoch": 17.01, + "learning_rate": 1.3002093853922897e-06, + "loss": 0.258, + "step": 46052 + }, + { + "epoch": 17.01, + "learning_rate": 1.3001724350289445e-06, + "loss": 0.2466, + "step": 46053 + }, + { + "epoch": 17.01, + "learning_rate": 1.3001354846655994e-06, + "loss": 0.2081, + "step": 46054 + }, + { + "epoch": 17.01, + "learning_rate": 1.300098534302254e-06, + "loss": 0.2777, + "step": 46055 + }, + { + "epoch": 17.01, + "learning_rate": 1.3000615839389088e-06, + "loss": 0.2593, + "step": 46056 + }, + { + "epoch": 17.01, + "learning_rate": 1.3000246335755635e-06, + "loss": 0.1275, + "step": 46057 + }, + { + "epoch": 17.01, + "learning_rate": 1.2999876832122183e-06, + "loss": 0.2491, + "step": 46058 + }, + { + "epoch": 17.01, + "learning_rate": 1.299950732848873e-06, + "loss": 0.2882, + "step": 46059 + }, + { + "epoch": 17.01, + "learning_rate": 1.2999137824855278e-06, + "loss": 0.3183, + "step": 46060 + }, + { + "epoch": 17.01, + "learning_rate": 1.2998768321221827e-06, + "loss": 0.2777, + "step": 46061 + }, + { + "epoch": 17.01, + "learning_rate": 1.2998398817588373e-06, + "loss": 0.178, + "step": 46062 + }, + { + "epoch": 17.01, + "learning_rate": 1.2998029313954921e-06, + "loss": 0.2685, + "step": 46063 + }, + { + "epoch": 17.01, + "learning_rate": 1.2997659810321468e-06, + "loss": 0.2633, + "step": 46064 + }, + { + "epoch": 17.01, + "learning_rate": 1.2997290306688016e-06, + "loss": 0.1652, + "step": 46065 + }, + { + "epoch": 17.01, + "learning_rate": 1.2996920803054565e-06, + "loss": 0.2802, + "step": 46066 + }, + { + "epoch": 17.01, + "learning_rate": 1.2996551299421111e-06, + "loss": 0.2548, + "step": 46067 + }, + { + "epoch": 17.01, + "learning_rate": 1.2996181795787658e-06, + "loss": 0.2389, + "step": 46068 + }, + { + "epoch": 17.01, + "learning_rate": 1.2995812292154206e-06, + "loss": 0.2955, + "step": 46069 + }, + { + "epoch": 17.01, + "learning_rate": 1.2995442788520754e-06, + "loss": 0.3466, + "step": 46070 + }, + { + "epoch": 17.01, + "learning_rate": 1.29950732848873e-06, + "loss": 0.2667, + "step": 46071 + }, + { + "epoch": 17.01, + "learning_rate": 1.299470378125385e-06, + "loss": 0.3544, + "step": 46072 + }, + { + "epoch": 17.01, + "learning_rate": 1.2994334277620398e-06, + "loss": 0.2849, + "step": 46073 + }, + { + "epoch": 17.01, + "learning_rate": 1.2993964773986944e-06, + "loss": 0.2249, + "step": 46074 + }, + { + "epoch": 17.01, + "learning_rate": 1.2993595270353493e-06, + "loss": 0.243, + "step": 46075 + }, + { + "epoch": 17.01, + "learning_rate": 1.2993225766720041e-06, + "loss": 0.238, + "step": 46076 + }, + { + "epoch": 17.02, + "learning_rate": 1.2992856263086588e-06, + "loss": 0.2493, + "step": 46077 + }, + { + "epoch": 17.02, + "learning_rate": 1.2992486759453134e-06, + "loss": 0.3198, + "step": 46078 + }, + { + "epoch": 17.02, + "learning_rate": 1.2992117255819682e-06, + "loss": 0.215, + "step": 46079 + }, + { + "epoch": 17.02, + "learning_rate": 1.299174775218623e-06, + "loss": 0.2383, + "step": 46080 + }, + { + "epoch": 17.02, + "learning_rate": 1.2991378248552777e-06, + "loss": 0.3011, + "step": 46081 + }, + { + "epoch": 17.02, + "learning_rate": 1.2991008744919326e-06, + "loss": 0.2206, + "step": 46082 + }, + { + "epoch": 17.02, + "learning_rate": 1.2990639241285874e-06, + "loss": 0.1335, + "step": 46083 + }, + { + "epoch": 17.02, + "learning_rate": 1.299026973765242e-06, + "loss": 0.2852, + "step": 46084 + }, + { + "epoch": 17.02, + "learning_rate": 1.298990023401897e-06, + "loss": 0.3313, + "step": 46085 + }, + { + "epoch": 17.02, + "learning_rate": 1.2989530730385517e-06, + "loss": 0.2241, + "step": 46086 + }, + { + "epoch": 17.02, + "learning_rate": 1.2989161226752062e-06, + "loss": 0.313, + "step": 46087 + }, + { + "epoch": 17.02, + "learning_rate": 1.298879172311861e-06, + "loss": 0.2441, + "step": 46088 + }, + { + "epoch": 17.02, + "learning_rate": 1.2988422219485159e-06, + "loss": 0.1749, + "step": 46089 + }, + { + "epoch": 17.02, + "learning_rate": 1.2988052715851705e-06, + "loss": 0.1879, + "step": 46090 + }, + { + "epoch": 17.02, + "learning_rate": 1.2987683212218254e-06, + "loss": 0.2061, + "step": 46091 + }, + { + "epoch": 17.02, + "learning_rate": 1.2987313708584802e-06, + "loss": 0.2592, + "step": 46092 + }, + { + "epoch": 17.02, + "learning_rate": 1.2986944204951348e-06, + "loss": 0.2903, + "step": 46093 + }, + { + "epoch": 17.02, + "learning_rate": 1.2986574701317897e-06, + "loss": 0.2523, + "step": 46094 + }, + { + "epoch": 17.02, + "learning_rate": 1.2986205197684445e-06, + "loss": 0.2412, + "step": 46095 + }, + { + "epoch": 17.02, + "learning_rate": 1.2985835694050992e-06, + "loss": 0.3219, + "step": 46096 + }, + { + "epoch": 17.02, + "learning_rate": 1.2985466190417538e-06, + "loss": 0.292, + "step": 46097 + }, + { + "epoch": 17.02, + "learning_rate": 1.2985096686784087e-06, + "loss": 0.212, + "step": 46098 + }, + { + "epoch": 17.02, + "learning_rate": 1.2984727183150635e-06, + "loss": 0.3313, + "step": 46099 + }, + { + "epoch": 17.02, + "learning_rate": 1.2984357679517181e-06, + "loss": 0.2763, + "step": 46100 + }, + { + "epoch": 17.02, + "learning_rate": 1.298398817588373e-06, + "loss": 0.2557, + "step": 46101 + }, + { + "epoch": 17.02, + "learning_rate": 1.2983618672250278e-06, + "loss": 0.1997, + "step": 46102 + }, + { + "epoch": 17.02, + "learning_rate": 1.2983249168616825e-06, + "loss": 0.2353, + "step": 46103 + }, + { + "epoch": 17.03, + "learning_rate": 1.2982879664983373e-06, + "loss": 0.1645, + "step": 46104 + }, + { + "epoch": 17.03, + "learning_rate": 1.2982510161349922e-06, + "loss": 0.1761, + "step": 46105 + }, + { + "epoch": 17.03, + "learning_rate": 1.2982140657716468e-06, + "loss": 0.2172, + "step": 46106 + }, + { + "epoch": 17.03, + "learning_rate": 1.2981771154083014e-06, + "loss": 0.1678, + "step": 46107 + }, + { + "epoch": 17.03, + "learning_rate": 1.2981401650449563e-06, + "loss": 0.1924, + "step": 46108 + }, + { + "epoch": 17.03, + "learning_rate": 1.298103214681611e-06, + "loss": 0.2686, + "step": 46109 + }, + { + "epoch": 17.03, + "learning_rate": 1.2980662643182658e-06, + "loss": 0.1964, + "step": 46110 + }, + { + "epoch": 17.03, + "learning_rate": 1.2980293139549206e-06, + "loss": 0.1383, + "step": 46111 + }, + { + "epoch": 17.03, + "learning_rate": 1.2979923635915753e-06, + "loss": 0.1268, + "step": 46112 + }, + { + "epoch": 17.03, + "learning_rate": 1.2979554132282301e-06, + "loss": 0.2494, + "step": 46113 + }, + { + "epoch": 17.03, + "learning_rate": 1.297918462864885e-06, + "loss": 0.1929, + "step": 46114 + }, + { + "epoch": 17.03, + "learning_rate": 1.2978815125015396e-06, + "loss": 0.215, + "step": 46115 + }, + { + "epoch": 17.03, + "learning_rate": 1.2978445621381944e-06, + "loss": 0.1986, + "step": 46116 + }, + { + "epoch": 17.03, + "learning_rate": 1.297807611774849e-06, + "loss": 0.2793, + "step": 46117 + }, + { + "epoch": 17.03, + "learning_rate": 1.297770661411504e-06, + "loss": 0.327, + "step": 46118 + }, + { + "epoch": 17.03, + "learning_rate": 1.2977337110481586e-06, + "loss": 0.2541, + "step": 46119 + }, + { + "epoch": 17.03, + "learning_rate": 1.2976967606848134e-06, + "loss": 0.2001, + "step": 46120 + }, + { + "epoch": 17.03, + "learning_rate": 1.2976598103214683e-06, + "loss": 0.2516, + "step": 46121 + }, + { + "epoch": 17.03, + "learning_rate": 1.297622859958123e-06, + "loss": 0.0986, + "step": 46122 + }, + { + "epoch": 17.03, + "learning_rate": 1.2975859095947777e-06, + "loss": 0.224, + "step": 46123 + }, + { + "epoch": 17.03, + "learning_rate": 1.2975489592314326e-06, + "loss": 0.279, + "step": 46124 + }, + { + "epoch": 17.03, + "learning_rate": 1.2975120088680872e-06, + "loss": 0.2781, + "step": 46125 + }, + { + "epoch": 17.03, + "learning_rate": 1.297475058504742e-06, + "loss": 0.2321, + "step": 46126 + }, + { + "epoch": 17.03, + "learning_rate": 1.2974381081413967e-06, + "loss": 0.2513, + "step": 46127 + }, + { + "epoch": 17.03, + "learning_rate": 1.2974011577780516e-06, + "loss": 0.2334, + "step": 46128 + }, + { + "epoch": 17.03, + "learning_rate": 1.2973642074147062e-06, + "loss": 0.2646, + "step": 46129 + }, + { + "epoch": 17.03, + "learning_rate": 1.297327257051361e-06, + "loss": 0.1664, + "step": 46130 + }, + { + "epoch": 17.04, + "learning_rate": 1.297290306688016e-06, + "loss": 0.2896, + "step": 46131 + }, + { + "epoch": 17.04, + "learning_rate": 1.2972533563246705e-06, + "loss": 0.1777, + "step": 46132 + }, + { + "epoch": 17.04, + "learning_rate": 1.2972164059613254e-06, + "loss": 0.238, + "step": 46133 + }, + { + "epoch": 17.04, + "learning_rate": 1.2971794555979802e-06, + "loss": 0.3361, + "step": 46134 + }, + { + "epoch": 17.04, + "learning_rate": 1.2971425052346349e-06, + "loss": 0.3856, + "step": 46135 + }, + { + "epoch": 17.04, + "learning_rate": 1.2971055548712897e-06, + "loss": 0.1875, + "step": 46136 + }, + { + "epoch": 17.04, + "learning_rate": 1.2970686045079443e-06, + "loss": 0.3201, + "step": 46137 + }, + { + "epoch": 17.04, + "learning_rate": 1.297031654144599e-06, + "loss": 0.22, + "step": 46138 + }, + { + "epoch": 17.04, + "learning_rate": 1.2969947037812538e-06, + "loss": 0.3215, + "step": 46139 + }, + { + "epoch": 17.04, + "learning_rate": 1.2969577534179087e-06, + "loss": 0.3438, + "step": 46140 + }, + { + "epoch": 17.04, + "learning_rate": 1.2969208030545633e-06, + "loss": 0.1653, + "step": 46141 + }, + { + "epoch": 17.04, + "learning_rate": 1.2968838526912182e-06, + "loss": 0.1887, + "step": 46142 + }, + { + "epoch": 17.04, + "learning_rate": 1.296846902327873e-06, + "loss": 0.1466, + "step": 46143 + }, + { + "epoch": 17.04, + "learning_rate": 1.2968099519645277e-06, + "loss": 0.1917, + "step": 46144 + }, + { + "epoch": 17.04, + "learning_rate": 1.2967730016011825e-06, + "loss": 0.11, + "step": 46145 + }, + { + "epoch": 17.04, + "learning_rate": 1.2967360512378373e-06, + "loss": 0.1478, + "step": 46146 + }, + { + "epoch": 17.04, + "learning_rate": 1.296699100874492e-06, + "loss": 0.2219, + "step": 46147 + }, + { + "epoch": 17.04, + "learning_rate": 1.2966621505111466e-06, + "loss": 0.202, + "step": 46148 + }, + { + "epoch": 17.04, + "learning_rate": 1.2966252001478015e-06, + "loss": 0.258, + "step": 46149 + }, + { + "epoch": 17.04, + "learning_rate": 1.2965882497844563e-06, + "loss": 0.2731, + "step": 46150 + }, + { + "epoch": 17.04, + "learning_rate": 1.296551299421111e-06, + "loss": 0.3393, + "step": 46151 + }, + { + "epoch": 17.04, + "learning_rate": 1.2965143490577658e-06, + "loss": 0.2424, + "step": 46152 + }, + { + "epoch": 17.04, + "learning_rate": 1.2964773986944206e-06, + "loss": 0.1713, + "step": 46153 + }, + { + "epoch": 17.04, + "learning_rate": 1.2964404483310753e-06, + "loss": 0.2813, + "step": 46154 + }, + { + "epoch": 17.04, + "learning_rate": 1.2964034979677301e-06, + "loss": 0.1908, + "step": 46155 + }, + { + "epoch": 17.04, + "learning_rate": 1.296366547604385e-06, + "loss": 0.2942, + "step": 46156 + }, + { + "epoch": 17.04, + "learning_rate": 1.2963295972410394e-06, + "loss": 0.1675, + "step": 46157 + }, + { + "epoch": 17.05, + "learning_rate": 1.2962926468776943e-06, + "loss": 0.2153, + "step": 46158 + }, + { + "epoch": 17.05, + "learning_rate": 1.296255696514349e-06, + "loss": 0.1743, + "step": 46159 + }, + { + "epoch": 17.05, + "learning_rate": 1.2962187461510037e-06, + "loss": 0.2828, + "step": 46160 + }, + { + "epoch": 17.05, + "learning_rate": 1.2961817957876586e-06, + "loss": 0.1562, + "step": 46161 + }, + { + "epoch": 17.05, + "learning_rate": 1.2961448454243134e-06, + "loss": 0.2418, + "step": 46162 + }, + { + "epoch": 17.05, + "learning_rate": 1.296107895060968e-06, + "loss": 0.1947, + "step": 46163 + }, + { + "epoch": 17.05, + "learning_rate": 1.296070944697623e-06, + "loss": 0.1647, + "step": 46164 + }, + { + "epoch": 17.05, + "learning_rate": 1.2960339943342778e-06, + "loss": 0.2493, + "step": 46165 + }, + { + "epoch": 17.05, + "learning_rate": 1.2959970439709324e-06, + "loss": 0.2449, + "step": 46166 + }, + { + "epoch": 17.05, + "learning_rate": 1.295960093607587e-06, + "loss": 0.1479, + "step": 46167 + }, + { + "epoch": 17.05, + "learning_rate": 1.2959231432442419e-06, + "loss": 0.311, + "step": 46168 + }, + { + "epoch": 17.05, + "learning_rate": 1.2958861928808967e-06, + "loss": 0.1486, + "step": 46169 + }, + { + "epoch": 17.05, + "learning_rate": 1.2958492425175514e-06, + "loss": 0.1767, + "step": 46170 + }, + { + "epoch": 17.05, + "learning_rate": 1.2958122921542062e-06, + "loss": 0.2226, + "step": 46171 + }, + { + "epoch": 17.05, + "learning_rate": 1.295775341790861e-06, + "loss": 0.1079, + "step": 46172 + }, + { + "epoch": 17.05, + "learning_rate": 1.2957383914275157e-06, + "loss": 0.1281, + "step": 46173 + }, + { + "epoch": 17.05, + "learning_rate": 1.2957014410641706e-06, + "loss": 0.2105, + "step": 46174 + }, + { + "epoch": 17.05, + "learning_rate": 1.2956644907008254e-06, + "loss": 0.129, + "step": 46175 + }, + { + "epoch": 17.05, + "learning_rate": 1.29562754033748e-06, + "loss": 0.1702, + "step": 46176 + }, + { + "epoch": 17.05, + "learning_rate": 1.2955905899741347e-06, + "loss": 0.1166, + "step": 46177 + }, + { + "epoch": 17.05, + "learning_rate": 1.2955536396107895e-06, + "loss": 0.1871, + "step": 46178 + }, + { + "epoch": 17.05, + "learning_rate": 1.2955166892474442e-06, + "loss": 0.2742, + "step": 46179 + }, + { + "epoch": 17.05, + "learning_rate": 1.295479738884099e-06, + "loss": 0.2743, + "step": 46180 + }, + { + "epoch": 17.05, + "learning_rate": 1.2954427885207539e-06, + "loss": 0.2282, + "step": 46181 + }, + { + "epoch": 17.05, + "learning_rate": 1.2954058381574085e-06, + "loss": 0.2737, + "step": 46182 + }, + { + "epoch": 17.05, + "learning_rate": 1.2953688877940633e-06, + "loss": 0.2215, + "step": 46183 + }, + { + "epoch": 17.05, + "learning_rate": 1.2953319374307182e-06, + "loss": 0.2291, + "step": 46184 + }, + { + "epoch": 17.06, + "learning_rate": 1.2952949870673728e-06, + "loss": 0.1387, + "step": 46185 + }, + { + "epoch": 17.06, + "learning_rate": 1.2952580367040277e-06, + "loss": 0.2339, + "step": 46186 + }, + { + "epoch": 17.06, + "learning_rate": 1.2952210863406823e-06, + "loss": 0.198, + "step": 46187 + }, + { + "epoch": 17.06, + "learning_rate": 1.2951841359773372e-06, + "loss": 0.2017, + "step": 46188 + }, + { + "epoch": 17.06, + "learning_rate": 1.2951471856139918e-06, + "loss": 0.1356, + "step": 46189 + }, + { + "epoch": 17.06, + "learning_rate": 1.2951102352506466e-06, + "loss": 0.1891, + "step": 46190 + }, + { + "epoch": 17.06, + "learning_rate": 1.2950732848873015e-06, + "loss": 0.1492, + "step": 46191 + }, + { + "epoch": 17.06, + "learning_rate": 1.2950363345239561e-06, + "loss": 0.2905, + "step": 46192 + }, + { + "epoch": 17.06, + "learning_rate": 1.294999384160611e-06, + "loss": 0.2424, + "step": 46193 + }, + { + "epoch": 17.06, + "learning_rate": 1.2949624337972658e-06, + "loss": 0.2267, + "step": 46194 + }, + { + "epoch": 17.06, + "learning_rate": 1.2949254834339205e-06, + "loss": 0.2776, + "step": 46195 + }, + { + "epoch": 17.06, + "learning_rate": 1.2948885330705753e-06, + "loss": 0.2125, + "step": 46196 + }, + { + "epoch": 17.06, + "learning_rate": 1.29485158270723e-06, + "loss": 0.1808, + "step": 46197 + }, + { + "epoch": 17.06, + "learning_rate": 1.2948146323438848e-06, + "loss": 0.2773, + "step": 46198 + }, + { + "epoch": 17.06, + "learning_rate": 1.2947776819805394e-06, + "loss": 0.1478, + "step": 46199 + }, + { + "epoch": 17.06, + "learning_rate": 1.2947407316171943e-06, + "loss": 0.2438, + "step": 46200 + }, + { + "epoch": 17.06, + "learning_rate": 1.2947037812538491e-06, + "loss": 0.2355, + "step": 46201 + }, + { + "epoch": 17.06, + "learning_rate": 1.2946668308905038e-06, + "loss": 0.2438, + "step": 46202 + }, + { + "epoch": 17.06, + "learning_rate": 1.2946298805271586e-06, + "loss": 0.2042, + "step": 46203 + }, + { + "epoch": 17.06, + "learning_rate": 1.2945929301638135e-06, + "loss": 0.1616, + "step": 46204 + }, + { + "epoch": 17.06, + "learning_rate": 1.294555979800468e-06, + "loss": 0.1969, + "step": 46205 + }, + { + "epoch": 17.06, + "learning_rate": 1.294519029437123e-06, + "loss": 0.3055, + "step": 46206 + }, + { + "epoch": 17.06, + "learning_rate": 1.2944820790737776e-06, + "loss": 0.1582, + "step": 46207 + }, + { + "epoch": 17.06, + "learning_rate": 1.2944451287104322e-06, + "loss": 0.2133, + "step": 46208 + }, + { + "epoch": 17.06, + "learning_rate": 1.294408178347087e-06, + "loss": 0.354, + "step": 46209 + }, + { + "epoch": 17.06, + "learning_rate": 1.294371227983742e-06, + "loss": 0.378, + "step": 46210 + }, + { + "epoch": 17.06, + "learning_rate": 1.2943342776203966e-06, + "loss": 0.2762, + "step": 46211 + }, + { + "epoch": 17.06, + "learning_rate": 1.2942973272570514e-06, + "loss": 0.1684, + "step": 46212 + }, + { + "epoch": 17.07, + "learning_rate": 1.2942603768937062e-06, + "loss": 0.231, + "step": 46213 + }, + { + "epoch": 17.07, + "learning_rate": 1.2942234265303609e-06, + "loss": 0.3147, + "step": 46214 + }, + { + "epoch": 17.07, + "learning_rate": 1.2941864761670157e-06, + "loss": 0.3167, + "step": 46215 + }, + { + "epoch": 17.07, + "learning_rate": 1.2941495258036704e-06, + "loss": 0.1723, + "step": 46216 + }, + { + "epoch": 17.07, + "learning_rate": 1.2941125754403252e-06, + "loss": 0.2674, + "step": 46217 + }, + { + "epoch": 17.07, + "learning_rate": 1.2940756250769799e-06, + "loss": 0.3403, + "step": 46218 + }, + { + "epoch": 17.07, + "learning_rate": 1.2940386747136347e-06, + "loss": 0.2426, + "step": 46219 + }, + { + "epoch": 17.07, + "learning_rate": 1.2940017243502895e-06, + "loss": 0.2189, + "step": 46220 + }, + { + "epoch": 17.07, + "learning_rate": 1.2939647739869442e-06, + "loss": 0.2448, + "step": 46221 + }, + { + "epoch": 17.07, + "learning_rate": 1.293927823623599e-06, + "loss": 0.2194, + "step": 46222 + }, + { + "epoch": 17.07, + "learning_rate": 1.2938908732602539e-06, + "loss": 0.3363, + "step": 46223 + }, + { + "epoch": 17.07, + "learning_rate": 1.2938539228969085e-06, + "loss": 0.1882, + "step": 46224 + }, + { + "epoch": 17.07, + "learning_rate": 1.2938169725335634e-06, + "loss": 0.1545, + "step": 46225 + }, + { + "epoch": 17.07, + "learning_rate": 1.293780022170218e-06, + "loss": 0.3024, + "step": 46226 + }, + { + "epoch": 17.07, + "learning_rate": 1.2937430718068726e-06, + "loss": 0.2752, + "step": 46227 + }, + { + "epoch": 17.07, + "learning_rate": 1.2937061214435275e-06, + "loss": 0.3339, + "step": 46228 + }, + { + "epoch": 17.07, + "learning_rate": 1.2936691710801823e-06, + "loss": 0.2226, + "step": 46229 + }, + { + "epoch": 17.07, + "learning_rate": 1.293632220716837e-06, + "loss": 0.2518, + "step": 46230 + }, + { + "epoch": 17.07, + "learning_rate": 1.2935952703534918e-06, + "loss": 0.2755, + "step": 46231 + }, + { + "epoch": 17.07, + "learning_rate": 1.2935583199901467e-06, + "loss": 0.1889, + "step": 46232 + }, + { + "epoch": 17.07, + "learning_rate": 1.2935213696268013e-06, + "loss": 0.1642, + "step": 46233 + }, + { + "epoch": 17.07, + "learning_rate": 1.2934844192634562e-06, + "loss": 0.2823, + "step": 46234 + }, + { + "epoch": 17.07, + "learning_rate": 1.293447468900111e-06, + "loss": 0.2846, + "step": 46235 + }, + { + "epoch": 17.07, + "learning_rate": 1.2934105185367656e-06, + "loss": 0.2306, + "step": 46236 + }, + { + "epoch": 17.07, + "learning_rate": 1.2933735681734203e-06, + "loss": 0.2665, + "step": 46237 + }, + { + "epoch": 17.07, + "learning_rate": 1.2933366178100751e-06, + "loss": 0.3573, + "step": 46238 + }, + { + "epoch": 17.07, + "learning_rate": 1.29329966744673e-06, + "loss": 0.2887, + "step": 46239 + }, + { + "epoch": 17.08, + "learning_rate": 1.2932627170833846e-06, + "loss": 0.3165, + "step": 46240 + }, + { + "epoch": 17.08, + "learning_rate": 1.2932257667200395e-06, + "loss": 0.2918, + "step": 46241 + }, + { + "epoch": 17.08, + "learning_rate": 1.2931888163566943e-06, + "loss": 0.3099, + "step": 46242 + }, + { + "epoch": 17.08, + "learning_rate": 1.293151865993349e-06, + "loss": 0.1261, + "step": 46243 + }, + { + "epoch": 17.08, + "learning_rate": 1.2931149156300038e-06, + "loss": 0.4351, + "step": 46244 + }, + { + "epoch": 17.08, + "learning_rate": 1.2930779652666586e-06, + "loss": 0.2373, + "step": 46245 + }, + { + "epoch": 17.08, + "learning_rate": 1.2930410149033133e-06, + "loss": 0.2144, + "step": 46246 + }, + { + "epoch": 17.08, + "learning_rate": 1.293004064539968e-06, + "loss": 0.1941, + "step": 46247 + }, + { + "epoch": 17.08, + "learning_rate": 1.2929671141766228e-06, + "loss": 0.2824, + "step": 46248 + }, + { + "epoch": 17.08, + "learning_rate": 1.2929301638132774e-06, + "loss": 0.2313, + "step": 46249 + }, + { + "epoch": 17.08, + "learning_rate": 1.2928932134499322e-06, + "loss": 0.1868, + "step": 46250 + }, + { + "epoch": 17.08, + "learning_rate": 1.292856263086587e-06, + "loss": 0.2937, + "step": 46251 + }, + { + "epoch": 17.08, + "learning_rate": 1.2928193127232417e-06, + "loss": 0.203, + "step": 46252 + }, + { + "epoch": 17.08, + "learning_rate": 1.2927823623598966e-06, + "loss": 0.1556, + "step": 46253 + }, + { + "epoch": 17.08, + "learning_rate": 1.2927454119965514e-06, + "loss": 0.2543, + "step": 46254 + }, + { + "epoch": 17.08, + "learning_rate": 1.292708461633206e-06, + "loss": 0.1444, + "step": 46255 + }, + { + "epoch": 17.08, + "learning_rate": 1.2926715112698607e-06, + "loss": 0.1352, + "step": 46256 + }, + { + "epoch": 17.08, + "learning_rate": 1.2926345609065155e-06, + "loss": 0.4011, + "step": 46257 + }, + { + "epoch": 17.08, + "learning_rate": 1.2925976105431704e-06, + "loss": 0.2596, + "step": 46258 + }, + { + "epoch": 17.08, + "learning_rate": 1.292560660179825e-06, + "loss": 0.2616, + "step": 46259 + }, + { + "epoch": 17.08, + "learning_rate": 1.2925237098164799e-06, + "loss": 0.1962, + "step": 46260 + }, + { + "epoch": 17.08, + "learning_rate": 1.2924867594531347e-06, + "loss": 0.1509, + "step": 46261 + }, + { + "epoch": 17.08, + "learning_rate": 1.2924498090897894e-06, + "loss": 0.2405, + "step": 46262 + }, + { + "epoch": 17.08, + "learning_rate": 1.2924128587264442e-06, + "loss": 0.1597, + "step": 46263 + }, + { + "epoch": 17.08, + "learning_rate": 1.292375908363099e-06, + "loss": 0.2149, + "step": 46264 + }, + { + "epoch": 17.08, + "learning_rate": 1.2923389579997537e-06, + "loss": 0.3676, + "step": 46265 + }, + { + "epoch": 17.08, + "learning_rate": 1.2923020076364083e-06, + "loss": 0.199, + "step": 46266 + }, + { + "epoch": 17.09, + "learning_rate": 1.2922650572730632e-06, + "loss": 0.2143, + "step": 46267 + }, + { + "epoch": 17.09, + "learning_rate": 1.292228106909718e-06, + "loss": 0.2179, + "step": 46268 + }, + { + "epoch": 17.09, + "learning_rate": 1.2921911565463727e-06, + "loss": 0.3537, + "step": 46269 + }, + { + "epoch": 17.09, + "learning_rate": 1.2921542061830275e-06, + "loss": 0.1651, + "step": 46270 + }, + { + "epoch": 17.09, + "learning_rate": 1.2921172558196824e-06, + "loss": 0.2669, + "step": 46271 + }, + { + "epoch": 17.09, + "learning_rate": 1.292080305456337e-06, + "loss": 0.3419, + "step": 46272 + }, + { + "epoch": 17.09, + "learning_rate": 1.2920433550929918e-06, + "loss": 0.1544, + "step": 46273 + }, + { + "epoch": 17.09, + "learning_rate": 1.2920064047296467e-06, + "loss": 0.1462, + "step": 46274 + }, + { + "epoch": 17.09, + "learning_rate": 1.2919694543663013e-06, + "loss": 0.2789, + "step": 46275 + }, + { + "epoch": 17.09, + "learning_rate": 1.291932504002956e-06, + "loss": 0.2365, + "step": 46276 + }, + { + "epoch": 17.09, + "learning_rate": 1.2918955536396108e-06, + "loss": 0.2522, + "step": 46277 + }, + { + "epoch": 17.09, + "learning_rate": 1.2918586032762655e-06, + "loss": 0.1742, + "step": 46278 + }, + { + "epoch": 17.09, + "learning_rate": 1.2918216529129203e-06, + "loss": 0.2699, + "step": 46279 + }, + { + "epoch": 17.09, + "learning_rate": 1.2917847025495751e-06, + "loss": 0.2172, + "step": 46280 + }, + { + "epoch": 17.09, + "learning_rate": 1.2917477521862298e-06, + "loss": 0.1193, + "step": 46281 + }, + { + "epoch": 17.09, + "learning_rate": 1.2917108018228846e-06, + "loss": 0.219, + "step": 46282 + }, + { + "epoch": 17.09, + "learning_rate": 1.2916738514595395e-06, + "loss": 0.2239, + "step": 46283 + }, + { + "epoch": 17.09, + "learning_rate": 1.2916369010961941e-06, + "loss": 0.2252, + "step": 46284 + }, + { + "epoch": 17.09, + "learning_rate": 1.291599950732849e-06, + "loss": 0.3799, + "step": 46285 + }, + { + "epoch": 17.09, + "learning_rate": 1.2915630003695036e-06, + "loss": 0.3163, + "step": 46286 + }, + { + "epoch": 17.09, + "learning_rate": 1.2915260500061584e-06, + "loss": 0.1773, + "step": 46287 + }, + { + "epoch": 17.09, + "learning_rate": 1.291489099642813e-06, + "loss": 0.2975, + "step": 46288 + }, + { + "epoch": 17.09, + "learning_rate": 1.291452149279468e-06, + "loss": 0.1994, + "step": 46289 + }, + { + "epoch": 17.09, + "learning_rate": 1.2914151989161228e-06, + "loss": 0.2584, + "step": 46290 + }, + { + "epoch": 17.09, + "learning_rate": 1.2913782485527774e-06, + "loss": 0.2071, + "step": 46291 + }, + { + "epoch": 17.09, + "learning_rate": 1.2913412981894323e-06, + "loss": 0.1371, + "step": 46292 + }, + { + "epoch": 17.09, + "learning_rate": 1.2913043478260871e-06, + "loss": 0.2524, + "step": 46293 + }, + { + "epoch": 17.1, + "learning_rate": 1.2912673974627418e-06, + "loss": 0.2395, + "step": 46294 + }, + { + "epoch": 17.1, + "learning_rate": 1.2912304470993966e-06, + "loss": 0.3527, + "step": 46295 + }, + { + "epoch": 17.1, + "learning_rate": 1.2911934967360512e-06, + "loss": 0.1748, + "step": 46296 + }, + { + "epoch": 17.1, + "learning_rate": 1.2911565463727059e-06, + "loss": 0.2258, + "step": 46297 + }, + { + "epoch": 17.1, + "learning_rate": 1.2911195960093607e-06, + "loss": 0.2085, + "step": 46298 + }, + { + "epoch": 17.1, + "learning_rate": 1.2910826456460156e-06, + "loss": 0.2025, + "step": 46299 + }, + { + "epoch": 17.1, + "learning_rate": 1.2910456952826702e-06, + "loss": 0.2559, + "step": 46300 + }, + { + "epoch": 17.1, + "learning_rate": 1.291008744919325e-06, + "loss": 0.1476, + "step": 46301 + }, + { + "epoch": 17.1, + "learning_rate": 1.29097179455598e-06, + "loss": 0.2665, + "step": 46302 + }, + { + "epoch": 17.1, + "learning_rate": 1.2909348441926345e-06, + "loss": 0.1745, + "step": 46303 + }, + { + "epoch": 17.1, + "learning_rate": 1.2908978938292894e-06, + "loss": 0.2875, + "step": 46304 + }, + { + "epoch": 17.1, + "learning_rate": 1.2908609434659442e-06, + "loss": 0.2642, + "step": 46305 + }, + { + "epoch": 17.1, + "learning_rate": 1.2908239931025989e-06, + "loss": 0.3819, + "step": 46306 + }, + { + "epoch": 17.1, + "learning_rate": 1.2907870427392535e-06, + "loss": 0.1694, + "step": 46307 + }, + { + "epoch": 17.1, + "learning_rate": 1.2907500923759084e-06, + "loss": 0.2689, + "step": 46308 + }, + { + "epoch": 17.1, + "learning_rate": 1.2907131420125632e-06, + "loss": 0.2108, + "step": 46309 + }, + { + "epoch": 17.1, + "learning_rate": 1.2906761916492178e-06, + "loss": 0.3194, + "step": 46310 + }, + { + "epoch": 17.1, + "learning_rate": 1.2906392412858727e-06, + "loss": 0.2291, + "step": 46311 + }, + { + "epoch": 17.1, + "learning_rate": 1.2906022909225275e-06, + "loss": 0.2577, + "step": 46312 + }, + { + "epoch": 17.1, + "learning_rate": 1.2905653405591822e-06, + "loss": 0.2814, + "step": 46313 + }, + { + "epoch": 17.1, + "learning_rate": 1.290528390195837e-06, + "loss": 0.2061, + "step": 46314 + }, + { + "epoch": 17.1, + "learning_rate": 1.2904914398324919e-06, + "loss": 0.2832, + "step": 46315 + }, + { + "epoch": 17.1, + "learning_rate": 1.2904544894691465e-06, + "loss": 0.1969, + "step": 46316 + }, + { + "epoch": 17.1, + "learning_rate": 1.2904175391058011e-06, + "loss": 0.1869, + "step": 46317 + }, + { + "epoch": 17.1, + "learning_rate": 1.290380588742456e-06, + "loss": 0.1129, + "step": 46318 + }, + { + "epoch": 17.1, + "learning_rate": 1.2903436383791108e-06, + "loss": 0.2699, + "step": 46319 + }, + { + "epoch": 17.1, + "learning_rate": 1.2903066880157655e-06, + "loss": 0.2068, + "step": 46320 + }, + { + "epoch": 17.11, + "learning_rate": 1.2902697376524203e-06, + "loss": 0.2382, + "step": 46321 + }, + { + "epoch": 17.11, + "learning_rate": 1.2902327872890752e-06, + "loss": 0.3221, + "step": 46322 + }, + { + "epoch": 17.11, + "learning_rate": 1.2901958369257298e-06, + "loss": 0.28, + "step": 46323 + }, + { + "epoch": 17.11, + "learning_rate": 1.2901588865623847e-06, + "loss": 0.2246, + "step": 46324 + }, + { + "epoch": 17.11, + "learning_rate": 1.2901219361990395e-06, + "loss": 0.2746, + "step": 46325 + }, + { + "epoch": 17.11, + "learning_rate": 1.290084985835694e-06, + "loss": 0.332, + "step": 46326 + }, + { + "epoch": 17.11, + "learning_rate": 1.2900480354723488e-06, + "loss": 0.1824, + "step": 46327 + }, + { + "epoch": 17.11, + "learning_rate": 1.2900110851090036e-06, + "loss": 0.321, + "step": 46328 + }, + { + "epoch": 17.11, + "learning_rate": 1.2899741347456583e-06, + "loss": 0.2417, + "step": 46329 + }, + { + "epoch": 17.11, + "learning_rate": 1.2899371843823131e-06, + "loss": 0.2589, + "step": 46330 + }, + { + "epoch": 17.11, + "learning_rate": 1.289900234018968e-06, + "loss": 0.2859, + "step": 46331 + }, + { + "epoch": 17.11, + "learning_rate": 1.2898632836556226e-06, + "loss": 0.1745, + "step": 46332 + }, + { + "epoch": 17.11, + "learning_rate": 1.2898263332922774e-06, + "loss": 0.1731, + "step": 46333 + }, + { + "epoch": 17.11, + "learning_rate": 1.2897893829289323e-06, + "loss": 0.2106, + "step": 46334 + }, + { + "epoch": 17.11, + "learning_rate": 1.289752432565587e-06, + "loss": 0.2432, + "step": 46335 + }, + { + "epoch": 17.11, + "learning_rate": 1.2897154822022416e-06, + "loss": 0.2028, + "step": 46336 + }, + { + "epoch": 17.11, + "learning_rate": 1.2896785318388964e-06, + "loss": 0.1379, + "step": 46337 + }, + { + "epoch": 17.11, + "learning_rate": 1.2896415814755513e-06, + "loss": 0.3591, + "step": 46338 + }, + { + "epoch": 17.11, + "learning_rate": 1.289604631112206e-06, + "loss": 0.2023, + "step": 46339 + }, + { + "epoch": 17.11, + "learning_rate": 1.2895676807488607e-06, + "loss": 0.3174, + "step": 46340 + }, + { + "epoch": 17.11, + "learning_rate": 1.2895307303855156e-06, + "loss": 0.3279, + "step": 46341 + }, + { + "epoch": 17.11, + "learning_rate": 1.2894937800221702e-06, + "loss": 0.175, + "step": 46342 + }, + { + "epoch": 17.11, + "learning_rate": 1.289456829658825e-06, + "loss": 0.108, + "step": 46343 + }, + { + "epoch": 17.11, + "learning_rate": 1.28941987929548e-06, + "loss": 0.1549, + "step": 46344 + }, + { + "epoch": 17.11, + "learning_rate": 1.2893829289321346e-06, + "loss": 0.354, + "step": 46345 + }, + { + "epoch": 17.11, + "learning_rate": 1.2893459785687892e-06, + "loss": 0.265, + "step": 46346 + }, + { + "epoch": 17.11, + "learning_rate": 1.289309028205444e-06, + "loss": 0.1961, + "step": 46347 + }, + { + "epoch": 17.12, + "learning_rate": 1.2892720778420987e-06, + "loss": 0.2817, + "step": 46348 + }, + { + "epoch": 17.12, + "learning_rate": 1.2892351274787535e-06, + "loss": 0.1422, + "step": 46349 + }, + { + "epoch": 17.12, + "learning_rate": 1.2891981771154084e-06, + "loss": 0.3571, + "step": 46350 + }, + { + "epoch": 17.12, + "learning_rate": 1.289161226752063e-06, + "loss": 0.2204, + "step": 46351 + }, + { + "epoch": 17.12, + "learning_rate": 1.2891242763887179e-06, + "loss": 0.2135, + "step": 46352 + }, + { + "epoch": 17.12, + "learning_rate": 1.2890873260253727e-06, + "loss": 0.296, + "step": 46353 + }, + { + "epoch": 17.12, + "learning_rate": 1.2890503756620273e-06, + "loss": 0.131, + "step": 46354 + }, + { + "epoch": 17.12, + "learning_rate": 1.2890134252986822e-06, + "loss": 0.2317, + "step": 46355 + }, + { + "epoch": 17.12, + "learning_rate": 1.2889764749353368e-06, + "loss": 0.2494, + "step": 46356 + }, + { + "epoch": 17.12, + "learning_rate": 1.2889395245719917e-06, + "loss": 0.2657, + "step": 46357 + }, + { + "epoch": 17.12, + "learning_rate": 1.2889025742086463e-06, + "loss": 0.4087, + "step": 46358 + }, + { + "epoch": 17.12, + "learning_rate": 1.2888656238453012e-06, + "loss": 0.1884, + "step": 46359 + }, + { + "epoch": 17.12, + "learning_rate": 1.288828673481956e-06, + "loss": 0.1925, + "step": 46360 + }, + { + "epoch": 17.12, + "learning_rate": 1.2887917231186107e-06, + "loss": 0.1223, + "step": 46361 + }, + { + "epoch": 17.12, + "learning_rate": 1.2887547727552655e-06, + "loss": 0.2102, + "step": 46362 + }, + { + "epoch": 17.12, + "learning_rate": 1.2887178223919203e-06, + "loss": 0.1966, + "step": 46363 + }, + { + "epoch": 17.12, + "learning_rate": 1.288680872028575e-06, + "loss": 0.2657, + "step": 46364 + }, + { + "epoch": 17.12, + "learning_rate": 1.2886439216652298e-06, + "loss": 0.3312, + "step": 46365 + }, + { + "epoch": 17.12, + "learning_rate": 1.2886069713018845e-06, + "loss": 0.2246, + "step": 46366 + }, + { + "epoch": 17.12, + "learning_rate": 1.288570020938539e-06, + "loss": 0.13, + "step": 46367 + }, + { + "epoch": 17.12, + "learning_rate": 1.288533070575194e-06, + "loss": 0.2771, + "step": 46368 + }, + { + "epoch": 17.12, + "learning_rate": 1.2884961202118488e-06, + "loss": 0.2429, + "step": 46369 + }, + { + "epoch": 17.12, + "learning_rate": 1.2884591698485034e-06, + "loss": 0.2121, + "step": 46370 + }, + { + "epoch": 17.12, + "learning_rate": 1.2884222194851583e-06, + "loss": 0.2197, + "step": 46371 + }, + { + "epoch": 17.12, + "learning_rate": 1.2883852691218131e-06, + "loss": 0.1492, + "step": 46372 + }, + { + "epoch": 17.12, + "learning_rate": 1.2883483187584678e-06, + "loss": 0.0996, + "step": 46373 + }, + { + "epoch": 17.12, + "learning_rate": 1.2883113683951226e-06, + "loss": 0.2782, + "step": 46374 + }, + { + "epoch": 17.13, + "learning_rate": 1.2882744180317775e-06, + "loss": 0.3333, + "step": 46375 + }, + { + "epoch": 17.13, + "learning_rate": 1.288237467668432e-06, + "loss": 0.1921, + "step": 46376 + }, + { + "epoch": 17.13, + "learning_rate": 1.2882005173050867e-06, + "loss": 0.2697, + "step": 46377 + }, + { + "epoch": 17.13, + "learning_rate": 1.2881635669417416e-06, + "loss": 0.3435, + "step": 46378 + }, + { + "epoch": 17.13, + "learning_rate": 1.2881266165783964e-06, + "loss": 0.1525, + "step": 46379 + }, + { + "epoch": 17.13, + "learning_rate": 1.288089666215051e-06, + "loss": 0.2033, + "step": 46380 + }, + { + "epoch": 17.13, + "learning_rate": 1.288052715851706e-06, + "loss": 0.2847, + "step": 46381 + }, + { + "epoch": 17.13, + "learning_rate": 1.2880157654883608e-06, + "loss": 0.4647, + "step": 46382 + }, + { + "epoch": 17.13, + "learning_rate": 1.2879788151250154e-06, + "loss": 0.3075, + "step": 46383 + }, + { + "epoch": 17.13, + "learning_rate": 1.2879418647616703e-06, + "loss": 0.2509, + "step": 46384 + }, + { + "epoch": 17.13, + "learning_rate": 1.287904914398325e-06, + "loss": 0.165, + "step": 46385 + }, + { + "epoch": 17.13, + "learning_rate": 1.2878679640349797e-06, + "loss": 0.1479, + "step": 46386 + }, + { + "epoch": 17.13, + "learning_rate": 1.2878310136716344e-06, + "loss": 0.1643, + "step": 46387 + }, + { + "epoch": 17.13, + "learning_rate": 1.2877940633082892e-06, + "loss": 0.2343, + "step": 46388 + }, + { + "epoch": 17.13, + "learning_rate": 1.287757112944944e-06, + "loss": 0.2886, + "step": 46389 + }, + { + "epoch": 17.13, + "learning_rate": 1.2877201625815987e-06, + "loss": 0.2487, + "step": 46390 + }, + { + "epoch": 17.13, + "learning_rate": 1.2876832122182536e-06, + "loss": 0.2706, + "step": 46391 + }, + { + "epoch": 17.13, + "learning_rate": 1.2876462618549084e-06, + "loss": 0.2044, + "step": 46392 + }, + { + "epoch": 17.13, + "learning_rate": 1.287609311491563e-06, + "loss": 0.2662, + "step": 46393 + }, + { + "epoch": 17.13, + "learning_rate": 1.2875723611282179e-06, + "loss": 0.2844, + "step": 46394 + }, + { + "epoch": 17.13, + "learning_rate": 1.2875354107648727e-06, + "loss": 0.3446, + "step": 46395 + }, + { + "epoch": 17.13, + "learning_rate": 1.2874984604015272e-06, + "loss": 0.2752, + "step": 46396 + }, + { + "epoch": 17.13, + "learning_rate": 1.287461510038182e-06, + "loss": 0.2795, + "step": 46397 + }, + { + "epoch": 17.13, + "learning_rate": 1.2874245596748369e-06, + "loss": 0.1963, + "step": 46398 + }, + { + "epoch": 17.13, + "learning_rate": 1.2873876093114915e-06, + "loss": 0.264, + "step": 46399 + }, + { + "epoch": 17.13, + "learning_rate": 1.2873506589481463e-06, + "loss": 0.3106, + "step": 46400 + }, + { + "epoch": 17.13, + "learning_rate": 1.2873137085848012e-06, + "loss": 0.3655, + "step": 46401 + }, + { + "epoch": 17.14, + "learning_rate": 1.2872767582214558e-06, + "loss": 0.2194, + "step": 46402 + }, + { + "epoch": 17.14, + "learning_rate": 1.2872398078581107e-06, + "loss": 0.2061, + "step": 46403 + }, + { + "epoch": 17.14, + "learning_rate": 1.2872028574947655e-06, + "loss": 0.2093, + "step": 46404 + }, + { + "epoch": 17.14, + "learning_rate": 1.2871659071314202e-06, + "loss": 0.3189, + "step": 46405 + }, + { + "epoch": 17.14, + "learning_rate": 1.2871289567680748e-06, + "loss": 0.2608, + "step": 46406 + }, + { + "epoch": 17.14, + "learning_rate": 1.2870920064047296e-06, + "loss": 0.2137, + "step": 46407 + }, + { + "epoch": 17.14, + "learning_rate": 1.2870550560413845e-06, + "loss": 0.2144, + "step": 46408 + }, + { + "epoch": 17.14, + "learning_rate": 1.2870181056780391e-06, + "loss": 0.3428, + "step": 46409 + }, + { + "epoch": 17.14, + "learning_rate": 1.286981155314694e-06, + "loss": 0.1915, + "step": 46410 + }, + { + "epoch": 17.14, + "learning_rate": 1.2869442049513488e-06, + "loss": 0.1628, + "step": 46411 + }, + { + "epoch": 17.14, + "learning_rate": 1.2869072545880035e-06, + "loss": 0.2761, + "step": 46412 + }, + { + "epoch": 17.14, + "learning_rate": 1.2868703042246583e-06, + "loss": 0.1241, + "step": 46413 + }, + { + "epoch": 17.14, + "learning_rate": 1.2868333538613132e-06, + "loss": 0.1939, + "step": 46414 + }, + { + "epoch": 17.14, + "learning_rate": 1.2867964034979678e-06, + "loss": 0.2481, + "step": 46415 + }, + { + "epoch": 17.14, + "learning_rate": 1.2867594531346224e-06, + "loss": 0.1357, + "step": 46416 + }, + { + "epoch": 17.14, + "learning_rate": 1.2867225027712773e-06, + "loss": 0.254, + "step": 46417 + }, + { + "epoch": 17.14, + "learning_rate": 1.286685552407932e-06, + "loss": 0.2185, + "step": 46418 + }, + { + "epoch": 17.14, + "learning_rate": 1.2866486020445868e-06, + "loss": 0.1916, + "step": 46419 + }, + { + "epoch": 17.14, + "learning_rate": 1.2866116516812416e-06, + "loss": 0.2377, + "step": 46420 + }, + { + "epoch": 17.14, + "learning_rate": 1.2865747013178962e-06, + "loss": 0.19, + "step": 46421 + }, + { + "epoch": 17.14, + "learning_rate": 1.286537750954551e-06, + "loss": 0.1951, + "step": 46422 + }, + { + "epoch": 17.14, + "learning_rate": 1.286500800591206e-06, + "loss": 0.231, + "step": 46423 + }, + { + "epoch": 17.14, + "learning_rate": 1.2864638502278606e-06, + "loss": 0.2717, + "step": 46424 + }, + { + "epoch": 17.14, + "learning_rate": 1.2864268998645154e-06, + "loss": 0.1819, + "step": 46425 + }, + { + "epoch": 17.14, + "learning_rate": 1.28638994950117e-06, + "loss": 0.318, + "step": 46426 + }, + { + "epoch": 17.14, + "learning_rate": 1.286352999137825e-06, + "loss": 0.1877, + "step": 46427 + }, + { + "epoch": 17.14, + "learning_rate": 1.2863160487744796e-06, + "loss": 0.2107, + "step": 46428 + }, + { + "epoch": 17.15, + "learning_rate": 1.2862790984111344e-06, + "loss": 0.1179, + "step": 46429 + }, + { + "epoch": 17.15, + "learning_rate": 1.2862421480477892e-06, + "loss": 0.2465, + "step": 46430 + }, + { + "epoch": 17.15, + "learning_rate": 1.2862051976844439e-06, + "loss": 0.2277, + "step": 46431 + }, + { + "epoch": 17.15, + "learning_rate": 1.2861682473210987e-06, + "loss": 0.3563, + "step": 46432 + }, + { + "epoch": 17.15, + "learning_rate": 1.2861312969577536e-06, + "loss": 0.1748, + "step": 46433 + }, + { + "epoch": 17.15, + "learning_rate": 1.2860943465944082e-06, + "loss": 0.3161, + "step": 46434 + }, + { + "epoch": 17.15, + "learning_rate": 1.286057396231063e-06, + "loss": 0.1769, + "step": 46435 + }, + { + "epoch": 17.15, + "learning_rate": 1.2860204458677177e-06, + "loss": 0.2014, + "step": 46436 + }, + { + "epoch": 17.15, + "learning_rate": 1.2859834955043723e-06, + "loss": 0.2499, + "step": 46437 + }, + { + "epoch": 17.15, + "learning_rate": 1.2859465451410272e-06, + "loss": 0.276, + "step": 46438 + }, + { + "epoch": 17.15, + "learning_rate": 1.285909594777682e-06, + "loss": 0.3035, + "step": 46439 + }, + { + "epoch": 17.15, + "learning_rate": 1.2858726444143367e-06, + "loss": 0.2292, + "step": 46440 + }, + { + "epoch": 17.15, + "learning_rate": 1.2858356940509915e-06, + "loss": 0.3369, + "step": 46441 + }, + { + "epoch": 17.15, + "learning_rate": 1.2857987436876464e-06, + "loss": 0.2517, + "step": 46442 + }, + { + "epoch": 17.15, + "learning_rate": 1.285761793324301e-06, + "loss": 0.2245, + "step": 46443 + }, + { + "epoch": 17.15, + "learning_rate": 1.2857248429609559e-06, + "loss": 0.1444, + "step": 46444 + }, + { + "epoch": 17.15, + "learning_rate": 1.2856878925976105e-06, + "loss": 0.1895, + "step": 46445 + }, + { + "epoch": 17.15, + "learning_rate": 1.2856509422342653e-06, + "loss": 0.2779, + "step": 46446 + }, + { + "epoch": 17.15, + "learning_rate": 1.28561399187092e-06, + "loss": 0.2571, + "step": 46447 + }, + { + "epoch": 17.15, + "learning_rate": 1.2855770415075748e-06, + "loss": 0.1632, + "step": 46448 + }, + { + "epoch": 17.15, + "learning_rate": 1.2855400911442297e-06, + "loss": 0.2187, + "step": 46449 + }, + { + "epoch": 17.15, + "learning_rate": 1.2855031407808843e-06, + "loss": 0.2771, + "step": 46450 + }, + { + "epoch": 17.15, + "learning_rate": 1.2854661904175392e-06, + "loss": 0.1913, + "step": 46451 + }, + { + "epoch": 17.15, + "learning_rate": 1.285429240054194e-06, + "loss": 0.2637, + "step": 46452 + }, + { + "epoch": 17.15, + "learning_rate": 1.2853922896908486e-06, + "loss": 0.244, + "step": 46453 + }, + { + "epoch": 17.15, + "learning_rate": 1.2853553393275035e-06, + "loss": 0.2951, + "step": 46454 + }, + { + "epoch": 17.15, + "learning_rate": 1.2853183889641581e-06, + "loss": 0.3883, + "step": 46455 + }, + { + "epoch": 17.16, + "learning_rate": 1.285281438600813e-06, + "loss": 0.2281, + "step": 46456 + }, + { + "epoch": 17.16, + "learning_rate": 1.2852444882374676e-06, + "loss": 0.2319, + "step": 46457 + }, + { + "epoch": 17.16, + "learning_rate": 1.2852075378741225e-06, + "loss": 0.2696, + "step": 46458 + }, + { + "epoch": 17.16, + "learning_rate": 1.2851705875107773e-06, + "loss": 0.1796, + "step": 46459 + }, + { + "epoch": 17.16, + "learning_rate": 1.285133637147432e-06, + "loss": 0.1549, + "step": 46460 + }, + { + "epoch": 17.16, + "learning_rate": 1.2850966867840868e-06, + "loss": 0.3228, + "step": 46461 + }, + { + "epoch": 17.16, + "learning_rate": 1.2850597364207416e-06, + "loss": 0.2669, + "step": 46462 + }, + { + "epoch": 17.16, + "learning_rate": 1.2850227860573963e-06, + "loss": 0.2268, + "step": 46463 + }, + { + "epoch": 17.16, + "learning_rate": 1.2849858356940511e-06, + "loss": 0.1687, + "step": 46464 + }, + { + "epoch": 17.16, + "learning_rate": 1.2849488853307058e-06, + "loss": 0.219, + "step": 46465 + }, + { + "epoch": 17.16, + "learning_rate": 1.2849119349673604e-06, + "loss": 0.1654, + "step": 46466 + }, + { + "epoch": 17.16, + "learning_rate": 1.2848749846040152e-06, + "loss": 0.1849, + "step": 46467 + }, + { + "epoch": 17.16, + "learning_rate": 1.28483803424067e-06, + "loss": 0.1933, + "step": 46468 + }, + { + "epoch": 17.16, + "learning_rate": 1.2848010838773247e-06, + "loss": 0.2999, + "step": 46469 + }, + { + "epoch": 17.16, + "learning_rate": 1.2847641335139796e-06, + "loss": 0.2405, + "step": 46470 + }, + { + "epoch": 17.16, + "learning_rate": 1.2847271831506344e-06, + "loss": 0.1979, + "step": 46471 + }, + { + "epoch": 17.16, + "learning_rate": 1.284690232787289e-06, + "loss": 0.1168, + "step": 46472 + }, + { + "epoch": 17.16, + "learning_rate": 1.284653282423944e-06, + "loss": 0.2101, + "step": 46473 + }, + { + "epoch": 17.16, + "learning_rate": 1.2846163320605988e-06, + "loss": 0.2372, + "step": 46474 + }, + { + "epoch": 17.16, + "learning_rate": 1.2845793816972534e-06, + "loss": 0.3996, + "step": 46475 + }, + { + "epoch": 17.16, + "learning_rate": 1.284542431333908e-06, + "loss": 0.2464, + "step": 46476 + }, + { + "epoch": 17.16, + "learning_rate": 1.2845054809705629e-06, + "loss": 0.165, + "step": 46477 + }, + { + "epoch": 17.16, + "learning_rate": 1.2844685306072177e-06, + "loss": 0.1655, + "step": 46478 + }, + { + "epoch": 17.16, + "learning_rate": 1.2844315802438724e-06, + "loss": 0.1201, + "step": 46479 + }, + { + "epoch": 17.16, + "learning_rate": 1.2843946298805272e-06, + "loss": 0.1115, + "step": 46480 + }, + { + "epoch": 17.16, + "learning_rate": 1.284357679517182e-06, + "loss": 0.1855, + "step": 46481 + }, + { + "epoch": 17.16, + "learning_rate": 1.2843207291538367e-06, + "loss": 0.1702, + "step": 46482 + }, + { + "epoch": 17.17, + "learning_rate": 1.2842837787904915e-06, + "loss": 0.3339, + "step": 46483 + }, + { + "epoch": 17.17, + "learning_rate": 1.2842468284271464e-06, + "loss": 0.275, + "step": 46484 + }, + { + "epoch": 17.17, + "learning_rate": 1.2842098780638008e-06, + "loss": 0.21, + "step": 46485 + }, + { + "epoch": 17.17, + "learning_rate": 1.2841729277004557e-06, + "loss": 0.2319, + "step": 46486 + }, + { + "epoch": 17.17, + "learning_rate": 1.2841359773371105e-06, + "loss": 0.2155, + "step": 46487 + }, + { + "epoch": 17.17, + "learning_rate": 1.2840990269737651e-06, + "loss": 0.2778, + "step": 46488 + }, + { + "epoch": 17.17, + "learning_rate": 1.28406207661042e-06, + "loss": 0.2047, + "step": 46489 + }, + { + "epoch": 17.17, + "learning_rate": 1.2840251262470748e-06, + "loss": 0.2336, + "step": 46490 + }, + { + "epoch": 17.17, + "learning_rate": 1.2839881758837295e-06, + "loss": 0.2952, + "step": 46491 + }, + { + "epoch": 17.17, + "learning_rate": 1.2839512255203843e-06, + "loss": 0.1366, + "step": 46492 + }, + { + "epoch": 17.17, + "learning_rate": 1.2839142751570392e-06, + "loss": 0.2766, + "step": 46493 + }, + { + "epoch": 17.17, + "learning_rate": 1.2838773247936938e-06, + "loss": 0.2133, + "step": 46494 + }, + { + "epoch": 17.17, + "learning_rate": 1.2838403744303484e-06, + "loss": 0.1072, + "step": 46495 + }, + { + "epoch": 17.17, + "learning_rate": 1.2838034240670033e-06, + "loss": 0.3231, + "step": 46496 + }, + { + "epoch": 17.17, + "learning_rate": 1.2837664737036581e-06, + "loss": 0.1761, + "step": 46497 + }, + { + "epoch": 17.17, + "learning_rate": 1.2837295233403128e-06, + "loss": 0.189, + "step": 46498 + }, + { + "epoch": 17.17, + "learning_rate": 1.2836925729769676e-06, + "loss": 0.1045, + "step": 46499 + }, + { + "epoch": 17.17, + "learning_rate": 1.2836556226136225e-06, + "loss": 0.3178, + "step": 46500 + }, + { + "epoch": 17.17, + "learning_rate": 1.2836186722502771e-06, + "loss": 0.2034, + "step": 46501 + }, + { + "epoch": 17.17, + "learning_rate": 1.283581721886932e-06, + "loss": 0.3736, + "step": 46502 + }, + { + "epoch": 17.17, + "learning_rate": 1.2835447715235868e-06, + "loss": 0.2833, + "step": 46503 + }, + { + "epoch": 17.17, + "learning_rate": 1.2835078211602414e-06, + "loss": 0.1736, + "step": 46504 + }, + { + "epoch": 17.17, + "learning_rate": 1.283470870796896e-06, + "loss": 0.2152, + "step": 46505 + }, + { + "epoch": 17.17, + "learning_rate": 1.283433920433551e-06, + "loss": 0.1208, + "step": 46506 + }, + { + "epoch": 17.17, + "learning_rate": 1.2833969700702058e-06, + "loss": 0.2574, + "step": 46507 + }, + { + "epoch": 17.17, + "learning_rate": 1.2833600197068604e-06, + "loss": 0.2502, + "step": 46508 + }, + { + "epoch": 17.17, + "learning_rate": 1.2833230693435153e-06, + "loss": 0.1323, + "step": 46509 + }, + { + "epoch": 17.18, + "learning_rate": 1.2832861189801701e-06, + "loss": 0.2247, + "step": 46510 + }, + { + "epoch": 17.18, + "learning_rate": 1.2832491686168248e-06, + "loss": 0.2737, + "step": 46511 + }, + { + "epoch": 17.18, + "learning_rate": 1.2832122182534796e-06, + "loss": 0.1848, + "step": 46512 + }, + { + "epoch": 17.18, + "learning_rate": 1.2831752678901342e-06, + "loss": 0.3124, + "step": 46513 + }, + { + "epoch": 17.18, + "learning_rate": 1.283138317526789e-06, + "loss": 0.2698, + "step": 46514 + }, + { + "epoch": 17.18, + "learning_rate": 1.2831013671634437e-06, + "loss": 0.2724, + "step": 46515 + }, + { + "epoch": 17.18, + "learning_rate": 1.2830644168000986e-06, + "loss": 0.1908, + "step": 46516 + }, + { + "epoch": 17.18, + "learning_rate": 1.2830274664367532e-06, + "loss": 0.132, + "step": 46517 + }, + { + "epoch": 17.18, + "learning_rate": 1.282990516073408e-06, + "loss": 0.1898, + "step": 46518 + }, + { + "epoch": 17.18, + "learning_rate": 1.282953565710063e-06, + "loss": 0.1081, + "step": 46519 + }, + { + "epoch": 17.18, + "learning_rate": 1.2829166153467175e-06, + "loss": 0.1741, + "step": 46520 + }, + { + "epoch": 17.18, + "learning_rate": 1.2828796649833724e-06, + "loss": 0.3535, + "step": 46521 + }, + { + "epoch": 17.18, + "learning_rate": 1.2828427146200272e-06, + "loss": 0.2809, + "step": 46522 + }, + { + "epoch": 17.18, + "learning_rate": 1.2828057642566819e-06, + "loss": 0.278, + "step": 46523 + }, + { + "epoch": 17.18, + "learning_rate": 1.2827688138933367e-06, + "loss": 0.2118, + "step": 46524 + }, + { + "epoch": 17.18, + "learning_rate": 1.2827318635299914e-06, + "loss": 0.2106, + "step": 46525 + }, + { + "epoch": 17.18, + "learning_rate": 1.2826949131666462e-06, + "loss": 0.2578, + "step": 46526 + }, + { + "epoch": 17.18, + "learning_rate": 1.2826579628033008e-06, + "loss": 0.1177, + "step": 46527 + }, + { + "epoch": 17.18, + "learning_rate": 1.2826210124399557e-06, + "loss": 0.329, + "step": 46528 + }, + { + "epoch": 17.18, + "learning_rate": 1.2825840620766105e-06, + "loss": 0.2672, + "step": 46529 + }, + { + "epoch": 17.18, + "learning_rate": 1.2825471117132652e-06, + "loss": 0.2187, + "step": 46530 + }, + { + "epoch": 17.18, + "learning_rate": 1.28251016134992e-06, + "loss": 0.2113, + "step": 46531 + }, + { + "epoch": 17.18, + "learning_rate": 1.2824732109865749e-06, + "loss": 0.2138, + "step": 46532 + }, + { + "epoch": 17.18, + "learning_rate": 1.2824362606232295e-06, + "loss": 0.1619, + "step": 46533 + }, + { + "epoch": 17.18, + "learning_rate": 1.2823993102598844e-06, + "loss": 0.1599, + "step": 46534 + }, + { + "epoch": 17.18, + "learning_rate": 1.282362359896539e-06, + "loss": 0.2897, + "step": 46535 + }, + { + "epoch": 17.18, + "learning_rate": 1.2823254095331936e-06, + "loss": 0.2123, + "step": 46536 + }, + { + "epoch": 17.19, + "learning_rate": 1.2822884591698485e-06, + "loss": 0.1972, + "step": 46537 + }, + { + "epoch": 17.19, + "learning_rate": 1.2822515088065033e-06, + "loss": 0.3156, + "step": 46538 + }, + { + "epoch": 17.19, + "learning_rate": 1.282214558443158e-06, + "loss": 0.2485, + "step": 46539 + }, + { + "epoch": 17.19, + "learning_rate": 1.2821776080798128e-06, + "loss": 0.1478, + "step": 46540 + }, + { + "epoch": 17.19, + "learning_rate": 1.2821406577164677e-06, + "loss": 0.2713, + "step": 46541 + }, + { + "epoch": 17.19, + "learning_rate": 1.2821037073531223e-06, + "loss": 0.2342, + "step": 46542 + }, + { + "epoch": 17.19, + "learning_rate": 1.2820667569897771e-06, + "loss": 0.2667, + "step": 46543 + }, + { + "epoch": 17.19, + "learning_rate": 1.282029806626432e-06, + "loss": 0.1509, + "step": 46544 + }, + { + "epoch": 17.19, + "learning_rate": 1.2819928562630866e-06, + "loss": 0.2188, + "step": 46545 + }, + { + "epoch": 17.19, + "learning_rate": 1.2819559058997413e-06, + "loss": 0.314, + "step": 46546 + }, + { + "epoch": 17.19, + "learning_rate": 1.2819189555363961e-06, + "loss": 0.2698, + "step": 46547 + }, + { + "epoch": 17.19, + "learning_rate": 1.281882005173051e-06, + "loss": 0.1906, + "step": 46548 + }, + { + "epoch": 17.19, + "learning_rate": 1.2818450548097056e-06, + "loss": 0.2566, + "step": 46549 + }, + { + "epoch": 17.19, + "learning_rate": 1.2818081044463604e-06, + "loss": 0.1613, + "step": 46550 + }, + { + "epoch": 17.19, + "learning_rate": 1.2817711540830153e-06, + "loss": 0.2997, + "step": 46551 + }, + { + "epoch": 17.19, + "learning_rate": 1.28173420371967e-06, + "loss": 0.2408, + "step": 46552 + }, + { + "epoch": 17.19, + "learning_rate": 1.2816972533563248e-06, + "loss": 0.3831, + "step": 46553 + }, + { + "epoch": 17.19, + "learning_rate": 1.2816603029929796e-06, + "loss": 0.1254, + "step": 46554 + }, + { + "epoch": 17.19, + "learning_rate": 1.281623352629634e-06, + "loss": 0.2423, + "step": 46555 + }, + { + "epoch": 17.19, + "learning_rate": 1.281586402266289e-06, + "loss": 0.2072, + "step": 46556 + }, + { + "epoch": 17.19, + "learning_rate": 1.2815494519029437e-06, + "loss": 0.269, + "step": 46557 + }, + { + "epoch": 17.19, + "learning_rate": 1.2815125015395984e-06, + "loss": 0.2428, + "step": 46558 + }, + { + "epoch": 17.19, + "learning_rate": 1.2814755511762532e-06, + "loss": 0.2933, + "step": 46559 + }, + { + "epoch": 17.19, + "learning_rate": 1.281438600812908e-06, + "loss": 0.1695, + "step": 46560 + }, + { + "epoch": 17.19, + "learning_rate": 1.2814016504495627e-06, + "loss": 0.3517, + "step": 46561 + }, + { + "epoch": 17.19, + "learning_rate": 1.2813647000862176e-06, + "loss": 0.1864, + "step": 46562 + }, + { + "epoch": 17.19, + "learning_rate": 1.2813277497228724e-06, + "loss": 0.2623, + "step": 46563 + }, + { + "epoch": 17.19, + "learning_rate": 1.281290799359527e-06, + "loss": 0.3493, + "step": 46564 + }, + { + "epoch": 17.2, + "learning_rate": 1.2812538489961817e-06, + "loss": 0.3176, + "step": 46565 + }, + { + "epoch": 17.2, + "learning_rate": 1.2812168986328365e-06, + "loss": 0.1899, + "step": 46566 + }, + { + "epoch": 17.2, + "learning_rate": 1.2811799482694914e-06, + "loss": 0.2068, + "step": 46567 + }, + { + "epoch": 17.2, + "learning_rate": 1.281142997906146e-06, + "loss": 0.3605, + "step": 46568 + }, + { + "epoch": 17.2, + "learning_rate": 1.2811060475428009e-06, + "loss": 0.3109, + "step": 46569 + }, + { + "epoch": 17.2, + "learning_rate": 1.2810690971794557e-06, + "loss": 0.1148, + "step": 46570 + }, + { + "epoch": 17.2, + "learning_rate": 1.2810321468161103e-06, + "loss": 0.2601, + "step": 46571 + }, + { + "epoch": 17.2, + "learning_rate": 1.2809951964527652e-06, + "loss": 0.2047, + "step": 46572 + }, + { + "epoch": 17.2, + "learning_rate": 1.28095824608942e-06, + "loss": 0.1873, + "step": 46573 + }, + { + "epoch": 17.2, + "learning_rate": 1.2809212957260747e-06, + "loss": 0.1781, + "step": 46574 + }, + { + "epoch": 17.2, + "learning_rate": 1.2808843453627293e-06, + "loss": 0.1941, + "step": 46575 + }, + { + "epoch": 17.2, + "learning_rate": 1.2808473949993842e-06, + "loss": 0.2692, + "step": 46576 + }, + { + "epoch": 17.2, + "learning_rate": 1.280810444636039e-06, + "loss": 0.2919, + "step": 46577 + }, + { + "epoch": 17.2, + "learning_rate": 1.2807734942726937e-06, + "loss": 0.1998, + "step": 46578 + }, + { + "epoch": 17.2, + "learning_rate": 1.2807365439093485e-06, + "loss": 0.2223, + "step": 46579 + }, + { + "epoch": 17.2, + "learning_rate": 1.2806995935460033e-06, + "loss": 0.3173, + "step": 46580 + }, + { + "epoch": 17.2, + "learning_rate": 1.280662643182658e-06, + "loss": 0.2564, + "step": 46581 + }, + { + "epoch": 17.2, + "learning_rate": 1.2806256928193128e-06, + "loss": 0.2732, + "step": 46582 + }, + { + "epoch": 17.2, + "learning_rate": 1.2805887424559677e-06, + "loss": 0.1702, + "step": 46583 + }, + { + "epoch": 17.2, + "learning_rate": 1.2805517920926223e-06, + "loss": 0.2105, + "step": 46584 + }, + { + "epoch": 17.2, + "learning_rate": 1.280514841729277e-06, + "loss": 0.2682, + "step": 46585 + }, + { + "epoch": 17.2, + "learning_rate": 1.2804778913659318e-06, + "loss": 0.1556, + "step": 46586 + }, + { + "epoch": 17.2, + "learning_rate": 1.2804409410025864e-06, + "loss": 0.3493, + "step": 46587 + }, + { + "epoch": 17.2, + "learning_rate": 1.2804039906392413e-06, + "loss": 0.2758, + "step": 46588 + }, + { + "epoch": 17.2, + "learning_rate": 1.2803670402758961e-06, + "loss": 0.1805, + "step": 46589 + }, + { + "epoch": 17.2, + "learning_rate": 1.2803300899125508e-06, + "loss": 0.2703, + "step": 46590 + }, + { + "epoch": 17.2, + "learning_rate": 1.2802931395492056e-06, + "loss": 0.2115, + "step": 46591 + }, + { + "epoch": 17.21, + "learning_rate": 1.2802561891858605e-06, + "loss": 0.1947, + "step": 46592 + }, + { + "epoch": 17.21, + "learning_rate": 1.280219238822515e-06, + "loss": 0.2062, + "step": 46593 + }, + { + "epoch": 17.21, + "learning_rate": 1.28018228845917e-06, + "loss": 0.2057, + "step": 46594 + }, + { + "epoch": 17.21, + "learning_rate": 1.2801453380958246e-06, + "loss": 0.348, + "step": 46595 + }, + { + "epoch": 17.21, + "learning_rate": 1.2801083877324794e-06, + "loss": 0.3148, + "step": 46596 + }, + { + "epoch": 17.21, + "learning_rate": 1.280071437369134e-06, + "loss": 0.2929, + "step": 46597 + }, + { + "epoch": 17.21, + "learning_rate": 1.280034487005789e-06, + "loss": 0.1738, + "step": 46598 + }, + { + "epoch": 17.21, + "learning_rate": 1.2799975366424438e-06, + "loss": 0.2109, + "step": 46599 + }, + { + "epoch": 17.21, + "learning_rate": 1.2799605862790984e-06, + "loss": 0.3831, + "step": 46600 + }, + { + "epoch": 17.21, + "learning_rate": 1.2799236359157533e-06, + "loss": 0.2224, + "step": 46601 + }, + { + "epoch": 17.21, + "learning_rate": 1.279886685552408e-06, + "loss": 0.1772, + "step": 46602 + }, + { + "epoch": 17.21, + "learning_rate": 1.2798497351890627e-06, + "loss": 0.2351, + "step": 46603 + }, + { + "epoch": 17.21, + "learning_rate": 1.2798127848257176e-06, + "loss": 0.2657, + "step": 46604 + }, + { + "epoch": 17.21, + "learning_rate": 1.2797758344623722e-06, + "loss": 0.2352, + "step": 46605 + }, + { + "epoch": 17.21, + "learning_rate": 1.2797388840990269e-06, + "loss": 0.2826, + "step": 46606 + }, + { + "epoch": 17.21, + "learning_rate": 1.2797019337356817e-06, + "loss": 0.1831, + "step": 46607 + }, + { + "epoch": 17.21, + "learning_rate": 1.2796649833723366e-06, + "loss": 0.279, + "step": 46608 + }, + { + "epoch": 17.21, + "learning_rate": 1.2796280330089912e-06, + "loss": 0.1514, + "step": 46609 + }, + { + "epoch": 17.21, + "learning_rate": 1.279591082645646e-06, + "loss": 0.349, + "step": 46610 + }, + { + "epoch": 17.21, + "learning_rate": 1.2795541322823009e-06, + "loss": 0.3038, + "step": 46611 + }, + { + "epoch": 17.21, + "learning_rate": 1.2795171819189555e-06, + "loss": 0.265, + "step": 46612 + }, + { + "epoch": 17.21, + "learning_rate": 1.2794802315556104e-06, + "loss": 0.1795, + "step": 46613 + }, + { + "epoch": 17.21, + "learning_rate": 1.2794432811922652e-06, + "loss": 0.3222, + "step": 46614 + }, + { + "epoch": 17.21, + "learning_rate": 1.2794063308289199e-06, + "loss": 0.324, + "step": 46615 + }, + { + "epoch": 17.21, + "learning_rate": 1.2793693804655745e-06, + "loss": 0.1736, + "step": 46616 + }, + { + "epoch": 17.21, + "learning_rate": 1.2793324301022293e-06, + "loss": 0.1749, + "step": 46617 + }, + { + "epoch": 17.21, + "learning_rate": 1.2792954797388842e-06, + "loss": 0.3071, + "step": 46618 + }, + { + "epoch": 17.22, + "learning_rate": 1.2792585293755388e-06, + "loss": 0.175, + "step": 46619 + }, + { + "epoch": 17.22, + "learning_rate": 1.2792215790121937e-06, + "loss": 0.1758, + "step": 46620 + }, + { + "epoch": 17.22, + "learning_rate": 1.2791846286488485e-06, + "loss": 0.2816, + "step": 46621 + }, + { + "epoch": 17.22, + "learning_rate": 1.2791476782855032e-06, + "loss": 0.1675, + "step": 46622 + }, + { + "epoch": 17.22, + "learning_rate": 1.279110727922158e-06, + "loss": 0.3216, + "step": 46623 + }, + { + "epoch": 17.22, + "learning_rate": 1.2790737775588129e-06, + "loss": 0.2146, + "step": 46624 + }, + { + "epoch": 17.22, + "learning_rate": 1.2790368271954673e-06, + "loss": 0.2236, + "step": 46625 + }, + { + "epoch": 17.22, + "learning_rate": 1.2789998768321221e-06, + "loss": 0.1486, + "step": 46626 + }, + { + "epoch": 17.22, + "learning_rate": 1.278962926468777e-06, + "loss": 0.1477, + "step": 46627 + }, + { + "epoch": 17.22, + "learning_rate": 1.2789259761054316e-06, + "loss": 0.2072, + "step": 46628 + }, + { + "epoch": 17.22, + "learning_rate": 1.2788890257420865e-06, + "loss": 0.2188, + "step": 46629 + }, + { + "epoch": 17.22, + "learning_rate": 1.2788520753787413e-06, + "loss": 0.2247, + "step": 46630 + }, + { + "epoch": 17.22, + "learning_rate": 1.278815125015396e-06, + "loss": 0.1317, + "step": 46631 + }, + { + "epoch": 17.22, + "learning_rate": 1.2787781746520508e-06, + "loss": 0.263, + "step": 46632 + }, + { + "epoch": 17.22, + "learning_rate": 1.2787412242887056e-06, + "loss": 0.1283, + "step": 46633 + }, + { + "epoch": 17.22, + "learning_rate": 1.2787042739253603e-06, + "loss": 0.1962, + "step": 46634 + }, + { + "epoch": 17.22, + "learning_rate": 1.278667323562015e-06, + "loss": 0.2317, + "step": 46635 + }, + { + "epoch": 17.22, + "learning_rate": 1.2786303731986698e-06, + "loss": 0.2176, + "step": 46636 + }, + { + "epoch": 17.22, + "learning_rate": 1.2785934228353246e-06, + "loss": 0.1709, + "step": 46637 + }, + { + "epoch": 17.22, + "learning_rate": 1.2785564724719792e-06, + "loss": 0.1799, + "step": 46638 + }, + { + "epoch": 17.22, + "learning_rate": 1.278519522108634e-06, + "loss": 0.2046, + "step": 46639 + }, + { + "epoch": 17.22, + "learning_rate": 1.278482571745289e-06, + "loss": 0.2705, + "step": 46640 + }, + { + "epoch": 17.22, + "learning_rate": 1.2784456213819436e-06, + "loss": 0.2399, + "step": 46641 + }, + { + "epoch": 17.22, + "learning_rate": 1.2784086710185984e-06, + "loss": 0.2446, + "step": 46642 + }, + { + "epoch": 17.22, + "learning_rate": 1.2783717206552533e-06, + "loss": 0.3633, + "step": 46643 + }, + { + "epoch": 17.22, + "learning_rate": 1.278334770291908e-06, + "loss": 0.1721, + "step": 46644 + }, + { + "epoch": 17.22, + "learning_rate": 1.2782978199285625e-06, + "loss": 0.2889, + "step": 46645 + }, + { + "epoch": 17.23, + "learning_rate": 1.2782608695652174e-06, + "loss": 0.2794, + "step": 46646 + }, + { + "epoch": 17.23, + "learning_rate": 1.2782239192018722e-06, + "loss": 0.329, + "step": 46647 + }, + { + "epoch": 17.23, + "learning_rate": 1.2781869688385269e-06, + "loss": 0.384, + "step": 46648 + }, + { + "epoch": 17.23, + "learning_rate": 1.2781500184751817e-06, + "loss": 0.395, + "step": 46649 + }, + { + "epoch": 17.23, + "learning_rate": 1.2781130681118366e-06, + "loss": 0.2118, + "step": 46650 + }, + { + "epoch": 17.23, + "learning_rate": 1.2780761177484912e-06, + "loss": 0.2333, + "step": 46651 + }, + { + "epoch": 17.23, + "learning_rate": 1.278039167385146e-06, + "loss": 0.2706, + "step": 46652 + }, + { + "epoch": 17.23, + "learning_rate": 1.278002217021801e-06, + "loss": 0.3228, + "step": 46653 + }, + { + "epoch": 17.23, + "learning_rate": 1.2779652666584555e-06, + "loss": 0.1934, + "step": 46654 + }, + { + "epoch": 17.23, + "learning_rate": 1.2779283162951102e-06, + "loss": 0.3088, + "step": 46655 + }, + { + "epoch": 17.23, + "learning_rate": 1.277891365931765e-06, + "loss": 0.1915, + "step": 46656 + }, + { + "epoch": 17.23, + "learning_rate": 1.2778544155684197e-06, + "loss": 0.185, + "step": 46657 + }, + { + "epoch": 17.23, + "learning_rate": 1.2778174652050745e-06, + "loss": 0.2409, + "step": 46658 + }, + { + "epoch": 17.23, + "learning_rate": 1.2777805148417294e-06, + "loss": 0.2168, + "step": 46659 + }, + { + "epoch": 17.23, + "learning_rate": 1.277743564478384e-06, + "loss": 0.3112, + "step": 46660 + }, + { + "epoch": 17.23, + "learning_rate": 1.2777066141150389e-06, + "loss": 0.234, + "step": 46661 + }, + { + "epoch": 17.23, + "learning_rate": 1.2776696637516937e-06, + "loss": 0.153, + "step": 46662 + }, + { + "epoch": 17.23, + "learning_rate": 1.2776327133883483e-06, + "loss": 0.131, + "step": 46663 + }, + { + "epoch": 17.23, + "learning_rate": 1.2775957630250032e-06, + "loss": 0.2581, + "step": 46664 + }, + { + "epoch": 17.23, + "learning_rate": 1.2775588126616578e-06, + "loss": 0.1287, + "step": 46665 + }, + { + "epoch": 17.23, + "learning_rate": 1.2775218622983127e-06, + "loss": 0.3053, + "step": 46666 + }, + { + "epoch": 17.23, + "learning_rate": 1.2774849119349673e-06, + "loss": 0.172, + "step": 46667 + }, + { + "epoch": 17.23, + "learning_rate": 1.2774479615716222e-06, + "loss": 0.3485, + "step": 46668 + }, + { + "epoch": 17.23, + "learning_rate": 1.277411011208277e-06, + "loss": 0.277, + "step": 46669 + }, + { + "epoch": 17.23, + "learning_rate": 1.2773740608449316e-06, + "loss": 0.173, + "step": 46670 + }, + { + "epoch": 17.23, + "learning_rate": 1.2773371104815865e-06, + "loss": 0.2427, + "step": 46671 + }, + { + "epoch": 17.23, + "learning_rate": 1.2773001601182413e-06, + "loss": 0.3658, + "step": 46672 + }, + { + "epoch": 17.24, + "learning_rate": 1.277263209754896e-06, + "loss": 0.327, + "step": 46673 + }, + { + "epoch": 17.24, + "learning_rate": 1.2772262593915508e-06, + "loss": 0.1697, + "step": 46674 + }, + { + "epoch": 17.24, + "learning_rate": 1.2771893090282055e-06, + "loss": 0.5663, + "step": 46675 + }, + { + "epoch": 17.24, + "learning_rate": 1.27715235866486e-06, + "loss": 0.2549, + "step": 46676 + }, + { + "epoch": 17.24, + "learning_rate": 1.277115408301515e-06, + "loss": 0.1646, + "step": 46677 + }, + { + "epoch": 17.24, + "learning_rate": 1.2770784579381698e-06, + "loss": 0.1802, + "step": 46678 + }, + { + "epoch": 17.24, + "learning_rate": 1.2770415075748244e-06, + "loss": 0.2413, + "step": 46679 + }, + { + "epoch": 17.24, + "learning_rate": 1.2770045572114793e-06, + "loss": 0.184, + "step": 46680 + }, + { + "epoch": 17.24, + "learning_rate": 1.2769676068481341e-06, + "loss": 0.184, + "step": 46681 + }, + { + "epoch": 17.24, + "learning_rate": 1.2769306564847888e-06, + "loss": 0.2248, + "step": 46682 + }, + { + "epoch": 17.24, + "learning_rate": 1.2768937061214436e-06, + "loss": 0.1875, + "step": 46683 + }, + { + "epoch": 17.24, + "learning_rate": 1.2768567557580982e-06, + "loss": 0.2053, + "step": 46684 + }, + { + "epoch": 17.24, + "learning_rate": 1.276819805394753e-06, + "loss": 0.3256, + "step": 46685 + }, + { + "epoch": 17.24, + "learning_rate": 1.2767828550314077e-06, + "loss": 0.171, + "step": 46686 + }, + { + "epoch": 17.24, + "learning_rate": 1.2767459046680626e-06, + "loss": 0.1685, + "step": 46687 + }, + { + "epoch": 17.24, + "learning_rate": 1.2767089543047174e-06, + "loss": 0.2391, + "step": 46688 + }, + { + "epoch": 17.24, + "learning_rate": 1.276672003941372e-06, + "loss": 0.2422, + "step": 46689 + }, + { + "epoch": 17.24, + "learning_rate": 1.276635053578027e-06, + "loss": 0.2319, + "step": 46690 + }, + { + "epoch": 17.24, + "learning_rate": 1.2765981032146818e-06, + "loss": 0.1699, + "step": 46691 + }, + { + "epoch": 17.24, + "learning_rate": 1.2765611528513364e-06, + "loss": 0.181, + "step": 46692 + }, + { + "epoch": 17.24, + "learning_rate": 1.2765242024879912e-06, + "loss": 0.3675, + "step": 46693 + }, + { + "epoch": 17.24, + "learning_rate": 1.2764872521246459e-06, + "loss": 0.3242, + "step": 46694 + }, + { + "epoch": 17.24, + "learning_rate": 1.2764503017613007e-06, + "loss": 0.361, + "step": 46695 + }, + { + "epoch": 17.24, + "learning_rate": 1.2764133513979554e-06, + "loss": 0.2434, + "step": 46696 + }, + { + "epoch": 17.24, + "learning_rate": 1.2763764010346102e-06, + "loss": 0.1559, + "step": 46697 + }, + { + "epoch": 17.24, + "learning_rate": 1.276339450671265e-06, + "loss": 0.1629, + "step": 46698 + }, + { + "epoch": 17.24, + "learning_rate": 1.2763025003079197e-06, + "loss": 0.0785, + "step": 46699 + }, + { + "epoch": 17.25, + "learning_rate": 1.2762655499445745e-06, + "loss": 0.2043, + "step": 46700 + }, + { + "epoch": 17.25, + "learning_rate": 1.2762285995812292e-06, + "loss": 0.2693, + "step": 46701 + }, + { + "epoch": 17.25, + "learning_rate": 1.276191649217884e-06, + "loss": 0.2442, + "step": 46702 + }, + { + "epoch": 17.25, + "learning_rate": 1.2761546988545389e-06, + "loss": 0.136, + "step": 46703 + }, + { + "epoch": 17.25, + "learning_rate": 1.2761177484911935e-06, + "loss": 0.4216, + "step": 46704 + }, + { + "epoch": 17.25, + "learning_rate": 1.2760807981278481e-06, + "loss": 0.1901, + "step": 46705 + }, + { + "epoch": 17.25, + "learning_rate": 1.276043847764503e-06, + "loss": 0.185, + "step": 46706 + }, + { + "epoch": 17.25, + "learning_rate": 1.2760068974011578e-06, + "loss": 0.2009, + "step": 46707 + }, + { + "epoch": 17.25, + "learning_rate": 1.2759699470378125e-06, + "loss": 0.2247, + "step": 46708 + }, + { + "epoch": 17.25, + "learning_rate": 1.2759329966744673e-06, + "loss": 0.2915, + "step": 46709 + }, + { + "epoch": 17.25, + "learning_rate": 1.2758960463111222e-06, + "loss": 0.208, + "step": 46710 + }, + { + "epoch": 17.25, + "learning_rate": 1.2758590959477768e-06, + "loss": 0.1998, + "step": 46711 + }, + { + "epoch": 17.25, + "learning_rate": 1.2758221455844317e-06, + "loss": 0.3496, + "step": 46712 + }, + { + "epoch": 17.25, + "learning_rate": 1.2757851952210865e-06, + "loss": 0.234, + "step": 46713 + }, + { + "epoch": 17.25, + "learning_rate": 1.2757482448577411e-06, + "loss": 0.2009, + "step": 46714 + }, + { + "epoch": 17.25, + "learning_rate": 1.2757112944943958e-06, + "loss": 0.2338, + "step": 46715 + }, + { + "epoch": 17.25, + "learning_rate": 1.2756743441310506e-06, + "loss": 0.1501, + "step": 46716 + }, + { + "epoch": 17.25, + "learning_rate": 1.2756373937677055e-06, + "loss": 0.2137, + "step": 46717 + }, + { + "epoch": 17.25, + "learning_rate": 1.2756004434043601e-06, + "loss": 0.2111, + "step": 46718 + }, + { + "epoch": 17.25, + "learning_rate": 1.275563493041015e-06, + "loss": 0.3087, + "step": 46719 + }, + { + "epoch": 17.25, + "learning_rate": 1.2755265426776698e-06, + "loss": 0.2776, + "step": 46720 + }, + { + "epoch": 17.25, + "learning_rate": 1.2754895923143244e-06, + "loss": 0.3222, + "step": 46721 + }, + { + "epoch": 17.25, + "learning_rate": 1.2754526419509793e-06, + "loss": 0.3184, + "step": 46722 + }, + { + "epoch": 17.25, + "learning_rate": 1.2754156915876341e-06, + "loss": 0.1896, + "step": 46723 + }, + { + "epoch": 17.25, + "learning_rate": 1.2753787412242886e-06, + "loss": 0.2685, + "step": 46724 + }, + { + "epoch": 17.25, + "learning_rate": 1.2753417908609434e-06, + "loss": 0.1659, + "step": 46725 + }, + { + "epoch": 17.25, + "learning_rate": 1.2753048404975983e-06, + "loss": 0.2196, + "step": 46726 + }, + { + "epoch": 17.26, + "learning_rate": 1.275267890134253e-06, + "loss": 0.2327, + "step": 46727 + }, + { + "epoch": 17.26, + "learning_rate": 1.2752309397709077e-06, + "loss": 0.3227, + "step": 46728 + }, + { + "epoch": 17.26, + "learning_rate": 1.2751939894075626e-06, + "loss": 0.2403, + "step": 46729 + }, + { + "epoch": 17.26, + "learning_rate": 1.2751570390442172e-06, + "loss": 0.1791, + "step": 46730 + }, + { + "epoch": 17.26, + "learning_rate": 1.275120088680872e-06, + "loss": 0.2032, + "step": 46731 + }, + { + "epoch": 17.26, + "learning_rate": 1.275083138317527e-06, + "loss": 0.1337, + "step": 46732 + }, + { + "epoch": 17.26, + "learning_rate": 1.2750461879541816e-06, + "loss": 0.2993, + "step": 46733 + }, + { + "epoch": 17.26, + "learning_rate": 1.2750092375908362e-06, + "loss": 0.3505, + "step": 46734 + }, + { + "epoch": 17.26, + "learning_rate": 1.274972287227491e-06, + "loss": 0.2239, + "step": 46735 + }, + { + "epoch": 17.26, + "learning_rate": 1.274935336864146e-06, + "loss": 0.203, + "step": 46736 + }, + { + "epoch": 17.26, + "learning_rate": 1.2748983865008005e-06, + "loss": 0.2537, + "step": 46737 + }, + { + "epoch": 17.26, + "learning_rate": 1.2748614361374554e-06, + "loss": 0.1447, + "step": 46738 + }, + { + "epoch": 17.26, + "learning_rate": 1.2748244857741102e-06, + "loss": 0.1913, + "step": 46739 + }, + { + "epoch": 17.26, + "learning_rate": 1.2747875354107649e-06, + "loss": 0.3133, + "step": 46740 + }, + { + "epoch": 17.26, + "learning_rate": 1.2747505850474197e-06, + "loss": 0.2105, + "step": 46741 + }, + { + "epoch": 17.26, + "learning_rate": 1.2747136346840746e-06, + "loss": 0.2585, + "step": 46742 + }, + { + "epoch": 17.26, + "learning_rate": 1.2746766843207292e-06, + "loss": 0.2751, + "step": 46743 + }, + { + "epoch": 17.26, + "learning_rate": 1.2746397339573838e-06, + "loss": 0.1293, + "step": 46744 + }, + { + "epoch": 17.26, + "learning_rate": 1.2746027835940387e-06, + "loss": 0.2952, + "step": 46745 + }, + { + "epoch": 17.26, + "learning_rate": 1.2745658332306933e-06, + "loss": 0.2731, + "step": 46746 + }, + { + "epoch": 17.26, + "learning_rate": 1.2745288828673482e-06, + "loss": 0.1696, + "step": 46747 + }, + { + "epoch": 17.26, + "learning_rate": 1.274491932504003e-06, + "loss": 0.1907, + "step": 46748 + }, + { + "epoch": 17.26, + "learning_rate": 1.2744549821406577e-06, + "loss": 0.2001, + "step": 46749 + }, + { + "epoch": 17.26, + "learning_rate": 1.2744180317773125e-06, + "loss": 0.224, + "step": 46750 + }, + { + "epoch": 17.26, + "learning_rate": 1.2743810814139674e-06, + "loss": 0.1974, + "step": 46751 + }, + { + "epoch": 17.26, + "learning_rate": 1.274344131050622e-06, + "loss": 0.2242, + "step": 46752 + }, + { + "epoch": 17.26, + "learning_rate": 1.2743071806872768e-06, + "loss": 0.1826, + "step": 46753 + }, + { + "epoch": 17.27, + "learning_rate": 1.2742702303239315e-06, + "loss": 0.2896, + "step": 46754 + }, + { + "epoch": 17.27, + "learning_rate": 1.2742332799605863e-06, + "loss": 0.1477, + "step": 46755 + }, + { + "epoch": 17.27, + "learning_rate": 1.274196329597241e-06, + "loss": 0.1833, + "step": 46756 + }, + { + "epoch": 17.27, + "learning_rate": 1.2741593792338958e-06, + "loss": 0.2745, + "step": 46757 + }, + { + "epoch": 17.27, + "learning_rate": 1.2741224288705507e-06, + "loss": 0.3324, + "step": 46758 + }, + { + "epoch": 17.27, + "learning_rate": 1.2740854785072053e-06, + "loss": 0.2591, + "step": 46759 + }, + { + "epoch": 17.27, + "learning_rate": 1.2740485281438601e-06, + "loss": 0.2143, + "step": 46760 + }, + { + "epoch": 17.27, + "learning_rate": 1.274011577780515e-06, + "loss": 0.117, + "step": 46761 + }, + { + "epoch": 17.27, + "learning_rate": 1.2739746274171696e-06, + "loss": 0.3182, + "step": 46762 + }, + { + "epoch": 17.27, + "learning_rate": 1.2739376770538245e-06, + "loss": 0.2217, + "step": 46763 + }, + { + "epoch": 17.27, + "learning_rate": 1.2739007266904791e-06, + "loss": 0.1725, + "step": 46764 + }, + { + "epoch": 17.27, + "learning_rate": 1.273863776327134e-06, + "loss": 0.4288, + "step": 46765 + }, + { + "epoch": 17.27, + "learning_rate": 1.2738268259637886e-06, + "loss": 0.1606, + "step": 46766 + }, + { + "epoch": 17.27, + "learning_rate": 1.2737898756004434e-06, + "loss": 0.2762, + "step": 46767 + }, + { + "epoch": 17.27, + "learning_rate": 1.2737529252370983e-06, + "loss": 0.2788, + "step": 46768 + }, + { + "epoch": 17.27, + "learning_rate": 1.273715974873753e-06, + "loss": 0.2051, + "step": 46769 + }, + { + "epoch": 17.27, + "learning_rate": 1.2736790245104078e-06, + "loss": 0.2682, + "step": 46770 + }, + { + "epoch": 17.27, + "learning_rate": 1.2736420741470626e-06, + "loss": 0.2597, + "step": 46771 + }, + { + "epoch": 17.27, + "learning_rate": 1.2736051237837173e-06, + "loss": 0.2038, + "step": 46772 + }, + { + "epoch": 17.27, + "learning_rate": 1.2735681734203721e-06, + "loss": 0.2269, + "step": 46773 + }, + { + "epoch": 17.27, + "learning_rate": 1.2735312230570267e-06, + "loss": 0.3376, + "step": 46774 + }, + { + "epoch": 17.27, + "learning_rate": 1.2734942726936814e-06, + "loss": 0.1873, + "step": 46775 + }, + { + "epoch": 17.27, + "learning_rate": 1.2734573223303362e-06, + "loss": 0.1501, + "step": 46776 + }, + { + "epoch": 17.27, + "learning_rate": 1.273420371966991e-06, + "loss": 0.3041, + "step": 46777 + }, + { + "epoch": 17.27, + "learning_rate": 1.2733834216036457e-06, + "loss": 0.1857, + "step": 46778 + }, + { + "epoch": 17.27, + "learning_rate": 1.2733464712403006e-06, + "loss": 0.2276, + "step": 46779 + }, + { + "epoch": 17.27, + "learning_rate": 1.2733095208769554e-06, + "loss": 0.1085, + "step": 46780 + }, + { + "epoch": 17.28, + "learning_rate": 1.27327257051361e-06, + "loss": 0.3148, + "step": 46781 + }, + { + "epoch": 17.28, + "learning_rate": 1.2732356201502649e-06, + "loss": 0.1919, + "step": 46782 + }, + { + "epoch": 17.28, + "learning_rate": 1.2731986697869197e-06, + "loss": 0.1795, + "step": 46783 + }, + { + "epoch": 17.28, + "learning_rate": 1.2731617194235744e-06, + "loss": 0.2404, + "step": 46784 + }, + { + "epoch": 17.28, + "learning_rate": 1.273124769060229e-06, + "loss": 0.2603, + "step": 46785 + }, + { + "epoch": 17.28, + "learning_rate": 1.2730878186968839e-06, + "loss": 0.1072, + "step": 46786 + }, + { + "epoch": 17.28, + "learning_rate": 1.2730508683335387e-06, + "loss": 0.2778, + "step": 46787 + }, + { + "epoch": 17.28, + "learning_rate": 1.2730139179701933e-06, + "loss": 0.1765, + "step": 46788 + }, + { + "epoch": 17.28, + "learning_rate": 1.2729769676068482e-06, + "loss": 0.2441, + "step": 46789 + }, + { + "epoch": 17.28, + "learning_rate": 1.272940017243503e-06, + "loss": 0.2227, + "step": 46790 + }, + { + "epoch": 17.28, + "learning_rate": 1.2729030668801577e-06, + "loss": 0.0856, + "step": 46791 + }, + { + "epoch": 17.28, + "learning_rate": 1.2728661165168125e-06, + "loss": 0.2691, + "step": 46792 + }, + { + "epoch": 17.28, + "learning_rate": 1.2728291661534674e-06, + "loss": 0.2591, + "step": 46793 + }, + { + "epoch": 17.28, + "learning_rate": 1.2727922157901218e-06, + "loss": 0.1819, + "step": 46794 + }, + { + "epoch": 17.28, + "learning_rate": 1.2727552654267766e-06, + "loss": 0.4077, + "step": 46795 + }, + { + "epoch": 17.28, + "learning_rate": 1.2727183150634315e-06, + "loss": 0.2209, + "step": 46796 + }, + { + "epoch": 17.28, + "learning_rate": 1.2726813647000861e-06, + "loss": 0.2745, + "step": 46797 + }, + { + "epoch": 17.28, + "learning_rate": 1.272644414336741e-06, + "loss": 0.2994, + "step": 46798 + }, + { + "epoch": 17.28, + "learning_rate": 1.2726074639733958e-06, + "loss": 0.2788, + "step": 46799 + }, + { + "epoch": 17.28, + "learning_rate": 1.2725705136100505e-06, + "loss": 0.2971, + "step": 46800 + }, + { + "epoch": 17.28, + "learning_rate": 1.2725335632467053e-06, + "loss": 0.2763, + "step": 46801 + }, + { + "epoch": 17.28, + "learning_rate": 1.2724966128833602e-06, + "loss": 0.2116, + "step": 46802 + }, + { + "epoch": 17.28, + "learning_rate": 1.2724596625200148e-06, + "loss": 0.2452, + "step": 46803 + }, + { + "epoch": 17.28, + "learning_rate": 1.2724227121566694e-06, + "loss": 0.1816, + "step": 46804 + }, + { + "epoch": 17.28, + "learning_rate": 1.2723857617933243e-06, + "loss": 0.3281, + "step": 46805 + }, + { + "epoch": 17.28, + "learning_rate": 1.2723488114299791e-06, + "loss": 0.1333, + "step": 46806 + }, + { + "epoch": 17.28, + "learning_rate": 1.2723118610666338e-06, + "loss": 0.1886, + "step": 46807 + }, + { + "epoch": 17.29, + "learning_rate": 1.2722749107032886e-06, + "loss": 0.2013, + "step": 46808 + }, + { + "epoch": 17.29, + "learning_rate": 1.2722379603399435e-06, + "loss": 0.2796, + "step": 46809 + }, + { + "epoch": 17.29, + "learning_rate": 1.272201009976598e-06, + "loss": 0.1232, + "step": 46810 + }, + { + "epoch": 17.29, + "learning_rate": 1.272164059613253e-06, + "loss": 0.2426, + "step": 46811 + }, + { + "epoch": 17.29, + "learning_rate": 1.2721271092499078e-06, + "loss": 0.2264, + "step": 46812 + }, + { + "epoch": 17.29, + "learning_rate": 1.2720901588865624e-06, + "loss": 0.2294, + "step": 46813 + }, + { + "epoch": 17.29, + "learning_rate": 1.272053208523217e-06, + "loss": 0.2469, + "step": 46814 + }, + { + "epoch": 17.29, + "learning_rate": 1.272016258159872e-06, + "loss": 0.1548, + "step": 46815 + }, + { + "epoch": 17.29, + "learning_rate": 1.2719793077965266e-06, + "loss": 0.2567, + "step": 46816 + }, + { + "epoch": 17.29, + "learning_rate": 1.2719423574331814e-06, + "loss": 0.2117, + "step": 46817 + }, + { + "epoch": 17.29, + "learning_rate": 1.2719054070698363e-06, + "loss": 0.167, + "step": 46818 + }, + { + "epoch": 17.29, + "learning_rate": 1.2718684567064909e-06, + "loss": 0.0972, + "step": 46819 + }, + { + "epoch": 17.29, + "learning_rate": 1.2718315063431457e-06, + "loss": 0.2785, + "step": 46820 + }, + { + "epoch": 17.29, + "learning_rate": 1.2717945559798006e-06, + "loss": 0.3088, + "step": 46821 + }, + { + "epoch": 17.29, + "learning_rate": 1.2717576056164552e-06, + "loss": 0.1622, + "step": 46822 + }, + { + "epoch": 17.29, + "learning_rate": 1.27172065525311e-06, + "loss": 0.2346, + "step": 46823 + }, + { + "epoch": 17.29, + "learning_rate": 1.2716837048897647e-06, + "loss": 0.1692, + "step": 46824 + }, + { + "epoch": 17.29, + "learning_rate": 1.2716467545264196e-06, + "loss": 0.1362, + "step": 46825 + }, + { + "epoch": 17.29, + "learning_rate": 1.2716098041630742e-06, + "loss": 0.3966, + "step": 46826 + }, + { + "epoch": 17.29, + "learning_rate": 1.271572853799729e-06, + "loss": 0.1962, + "step": 46827 + }, + { + "epoch": 17.29, + "learning_rate": 1.2715359034363839e-06, + "loss": 0.153, + "step": 46828 + }, + { + "epoch": 17.29, + "learning_rate": 1.2714989530730385e-06, + "loss": 0.1593, + "step": 46829 + }, + { + "epoch": 17.29, + "learning_rate": 1.2714620027096934e-06, + "loss": 0.1825, + "step": 46830 + }, + { + "epoch": 17.29, + "learning_rate": 1.2714250523463482e-06, + "loss": 0.3273, + "step": 46831 + }, + { + "epoch": 17.29, + "learning_rate": 1.2713881019830029e-06, + "loss": 0.266, + "step": 46832 + }, + { + "epoch": 17.29, + "learning_rate": 1.2713511516196577e-06, + "loss": 0.1277, + "step": 46833 + }, + { + "epoch": 17.29, + "learning_rate": 1.2713142012563123e-06, + "loss": 0.2846, + "step": 46834 + }, + { + "epoch": 17.3, + "learning_rate": 1.2712772508929672e-06, + "loss": 0.3359, + "step": 46835 + }, + { + "epoch": 17.3, + "learning_rate": 1.2712403005296218e-06, + "loss": 0.2451, + "step": 46836 + }, + { + "epoch": 17.3, + "learning_rate": 1.2712033501662767e-06, + "loss": 0.2954, + "step": 46837 + }, + { + "epoch": 17.3, + "learning_rate": 1.2711663998029315e-06, + "loss": 0.2589, + "step": 46838 + }, + { + "epoch": 17.3, + "learning_rate": 1.2711294494395862e-06, + "loss": 0.2305, + "step": 46839 + }, + { + "epoch": 17.3, + "learning_rate": 1.271092499076241e-06, + "loss": 0.1755, + "step": 46840 + }, + { + "epoch": 17.3, + "learning_rate": 1.2710555487128959e-06, + "loss": 0.1733, + "step": 46841 + }, + { + "epoch": 17.3, + "learning_rate": 1.2710185983495505e-06, + "loss": 0.3169, + "step": 46842 + }, + { + "epoch": 17.3, + "learning_rate": 1.2709816479862053e-06, + "loss": 0.3845, + "step": 46843 + }, + { + "epoch": 17.3, + "learning_rate": 1.27094469762286e-06, + "loss": 0.2783, + "step": 46844 + }, + { + "epoch": 17.3, + "learning_rate": 1.2709077472595146e-06, + "loss": 0.2839, + "step": 46845 + }, + { + "epoch": 17.3, + "learning_rate": 1.2708707968961695e-06, + "loss": 0.1548, + "step": 46846 + }, + { + "epoch": 17.3, + "learning_rate": 1.2708338465328243e-06, + "loss": 0.1719, + "step": 46847 + }, + { + "epoch": 17.3, + "learning_rate": 1.270796896169479e-06, + "loss": 0.1791, + "step": 46848 + }, + { + "epoch": 17.3, + "learning_rate": 1.2707599458061338e-06, + "loss": 0.1796, + "step": 46849 + }, + { + "epoch": 17.3, + "learning_rate": 1.2707229954427886e-06, + "loss": 0.2386, + "step": 46850 + }, + { + "epoch": 17.3, + "learning_rate": 1.2706860450794433e-06, + "loss": 0.1937, + "step": 46851 + }, + { + "epoch": 17.3, + "learning_rate": 1.2706490947160981e-06, + "loss": 0.2873, + "step": 46852 + }, + { + "epoch": 17.3, + "learning_rate": 1.270612144352753e-06, + "loss": 0.2651, + "step": 46853 + }, + { + "epoch": 17.3, + "learning_rate": 1.2705751939894076e-06, + "loss": 0.3429, + "step": 46854 + }, + { + "epoch": 17.3, + "learning_rate": 1.2705382436260622e-06, + "loss": 0.1832, + "step": 46855 + }, + { + "epoch": 17.3, + "learning_rate": 1.270501293262717e-06, + "loss": 0.1543, + "step": 46856 + }, + { + "epoch": 17.3, + "learning_rate": 1.270464342899372e-06, + "loss": 0.1696, + "step": 46857 + }, + { + "epoch": 17.3, + "learning_rate": 1.2704273925360266e-06, + "loss": 0.3498, + "step": 46858 + }, + { + "epoch": 17.3, + "learning_rate": 1.2703904421726814e-06, + "loss": 0.1629, + "step": 46859 + }, + { + "epoch": 17.3, + "learning_rate": 1.2703534918093363e-06, + "loss": 0.2387, + "step": 46860 + }, + { + "epoch": 17.3, + "learning_rate": 1.270316541445991e-06, + "loss": 0.1898, + "step": 46861 + }, + { + "epoch": 17.31, + "learning_rate": 1.2702795910826458e-06, + "loss": 0.1539, + "step": 46862 + }, + { + "epoch": 17.31, + "learning_rate": 1.2702426407193006e-06, + "loss": 0.2219, + "step": 46863 + }, + { + "epoch": 17.31, + "learning_rate": 1.270205690355955e-06, + "loss": 0.1496, + "step": 46864 + }, + { + "epoch": 17.31, + "learning_rate": 1.2701687399926099e-06, + "loss": 0.2281, + "step": 46865 + }, + { + "epoch": 17.31, + "learning_rate": 1.2701317896292647e-06, + "loss": 0.1389, + "step": 46866 + }, + { + "epoch": 17.31, + "learning_rate": 1.2700948392659194e-06, + "loss": 0.206, + "step": 46867 + }, + { + "epoch": 17.31, + "learning_rate": 1.2700578889025742e-06, + "loss": 0.2124, + "step": 46868 + }, + { + "epoch": 17.31, + "learning_rate": 1.270020938539229e-06, + "loss": 0.1471, + "step": 46869 + }, + { + "epoch": 17.31, + "learning_rate": 1.2699839881758837e-06, + "loss": 0.2516, + "step": 46870 + }, + { + "epoch": 17.31, + "learning_rate": 1.2699470378125385e-06, + "loss": 0.2147, + "step": 46871 + }, + { + "epoch": 17.31, + "learning_rate": 1.2699100874491934e-06, + "loss": 0.5075, + "step": 46872 + }, + { + "epoch": 17.31, + "learning_rate": 1.269873137085848e-06, + "loss": 0.1525, + "step": 46873 + }, + { + "epoch": 17.31, + "learning_rate": 1.2698361867225027e-06, + "loss": 0.1882, + "step": 46874 + }, + { + "epoch": 17.31, + "learning_rate": 1.2697992363591575e-06, + "loss": 0.2594, + "step": 46875 + }, + { + "epoch": 17.31, + "learning_rate": 1.2697622859958124e-06, + "loss": 0.4123, + "step": 46876 + }, + { + "epoch": 17.31, + "learning_rate": 1.269725335632467e-06, + "loss": 0.2383, + "step": 46877 + }, + { + "epoch": 17.31, + "learning_rate": 1.2696883852691218e-06, + "loss": 0.2102, + "step": 46878 + }, + { + "epoch": 17.31, + "learning_rate": 1.2696514349057767e-06, + "loss": 0.2627, + "step": 46879 + }, + { + "epoch": 17.31, + "learning_rate": 1.2696144845424313e-06, + "loss": 0.3061, + "step": 46880 + }, + { + "epoch": 17.31, + "learning_rate": 1.2695775341790862e-06, + "loss": 0.1477, + "step": 46881 + }, + { + "epoch": 17.31, + "learning_rate": 1.269540583815741e-06, + "loss": 0.1795, + "step": 46882 + }, + { + "epoch": 17.31, + "learning_rate": 1.2695036334523957e-06, + "loss": 0.1525, + "step": 46883 + }, + { + "epoch": 17.31, + "learning_rate": 1.2694666830890503e-06, + "loss": 0.3013, + "step": 46884 + }, + { + "epoch": 17.31, + "learning_rate": 1.2694297327257052e-06, + "loss": 0.2479, + "step": 46885 + }, + { + "epoch": 17.31, + "learning_rate": 1.2693927823623598e-06, + "loss": 0.3397, + "step": 46886 + }, + { + "epoch": 17.31, + "learning_rate": 1.2693558319990146e-06, + "loss": 0.2288, + "step": 46887 + }, + { + "epoch": 17.31, + "learning_rate": 1.2693188816356695e-06, + "loss": 0.2581, + "step": 46888 + }, + { + "epoch": 17.31, + "learning_rate": 1.2692819312723241e-06, + "loss": 0.1266, + "step": 46889 + }, + { + "epoch": 17.32, + "learning_rate": 1.269244980908979e-06, + "loss": 0.2651, + "step": 46890 + }, + { + "epoch": 17.32, + "learning_rate": 1.2692080305456338e-06, + "loss": 0.2844, + "step": 46891 + }, + { + "epoch": 17.32, + "learning_rate": 1.2691710801822885e-06, + "loss": 0.173, + "step": 46892 + }, + { + "epoch": 17.32, + "learning_rate": 1.2691341298189433e-06, + "loss": 0.2569, + "step": 46893 + }, + { + "epoch": 17.32, + "learning_rate": 1.269097179455598e-06, + "loss": 0.1183, + "step": 46894 + }, + { + "epoch": 17.32, + "learning_rate": 1.2690602290922528e-06, + "loss": 0.1693, + "step": 46895 + }, + { + "epoch": 17.32, + "learning_rate": 1.2690232787289074e-06, + "loss": 0.2214, + "step": 46896 + }, + { + "epoch": 17.32, + "learning_rate": 1.2689863283655623e-06, + "loss": 0.4039, + "step": 46897 + }, + { + "epoch": 17.32, + "learning_rate": 1.2689493780022171e-06, + "loss": 0.2298, + "step": 46898 + }, + { + "epoch": 17.32, + "learning_rate": 1.2689124276388718e-06, + "loss": 0.2972, + "step": 46899 + }, + { + "epoch": 17.32, + "learning_rate": 1.2688754772755266e-06, + "loss": 0.2098, + "step": 46900 + }, + { + "epoch": 17.32, + "learning_rate": 1.2688385269121815e-06, + "loss": 0.1689, + "step": 46901 + }, + { + "epoch": 17.32, + "learning_rate": 1.268801576548836e-06, + "loss": 0.1405, + "step": 46902 + }, + { + "epoch": 17.32, + "learning_rate": 1.268764626185491e-06, + "loss": 0.1686, + "step": 46903 + }, + { + "epoch": 17.32, + "learning_rate": 1.2687276758221456e-06, + "loss": 0.1465, + "step": 46904 + }, + { + "epoch": 17.32, + "learning_rate": 1.2686907254588004e-06, + "loss": 0.1996, + "step": 46905 + }, + { + "epoch": 17.32, + "learning_rate": 1.268653775095455e-06, + "loss": 0.3612, + "step": 46906 + }, + { + "epoch": 17.32, + "learning_rate": 1.26861682473211e-06, + "loss": 0.2486, + "step": 46907 + }, + { + "epoch": 17.32, + "learning_rate": 1.2685798743687648e-06, + "loss": 0.2631, + "step": 46908 + }, + { + "epoch": 17.32, + "learning_rate": 1.2685429240054194e-06, + "loss": 0.2083, + "step": 46909 + }, + { + "epoch": 17.32, + "learning_rate": 1.2685059736420742e-06, + "loss": 0.1856, + "step": 46910 + }, + { + "epoch": 17.32, + "learning_rate": 1.268469023278729e-06, + "loss": 0.2788, + "step": 46911 + }, + { + "epoch": 17.32, + "learning_rate": 1.2684320729153837e-06, + "loss": 0.1766, + "step": 46912 + }, + { + "epoch": 17.32, + "learning_rate": 1.2683951225520384e-06, + "loss": 0.2723, + "step": 46913 + }, + { + "epoch": 17.32, + "learning_rate": 1.2683581721886932e-06, + "loss": 0.3008, + "step": 46914 + }, + { + "epoch": 17.32, + "learning_rate": 1.2683212218253478e-06, + "loss": 0.2529, + "step": 46915 + }, + { + "epoch": 17.32, + "learning_rate": 1.2682842714620027e-06, + "loss": 0.165, + "step": 46916 + }, + { + "epoch": 17.33, + "learning_rate": 1.2682473210986575e-06, + "loss": 0.2248, + "step": 46917 + }, + { + "epoch": 17.33, + "learning_rate": 1.2682103707353122e-06, + "loss": 0.2036, + "step": 46918 + }, + { + "epoch": 17.33, + "learning_rate": 1.268173420371967e-06, + "loss": 0.2232, + "step": 46919 + }, + { + "epoch": 17.33, + "learning_rate": 1.2681364700086219e-06, + "loss": 0.2568, + "step": 46920 + }, + { + "epoch": 17.33, + "learning_rate": 1.2680995196452765e-06, + "loss": 0.2525, + "step": 46921 + }, + { + "epoch": 17.33, + "learning_rate": 1.2680625692819314e-06, + "loss": 0.2694, + "step": 46922 + }, + { + "epoch": 17.33, + "learning_rate": 1.268025618918586e-06, + "loss": 0.1776, + "step": 46923 + }, + { + "epoch": 17.33, + "learning_rate": 1.2679886685552408e-06, + "loss": 0.3232, + "step": 46924 + }, + { + "epoch": 17.33, + "learning_rate": 1.2679517181918955e-06, + "loss": 0.2579, + "step": 46925 + }, + { + "epoch": 17.33, + "learning_rate": 1.2679147678285503e-06, + "loss": 0.2316, + "step": 46926 + }, + { + "epoch": 17.33, + "learning_rate": 1.2678778174652052e-06, + "loss": 0.261, + "step": 46927 + }, + { + "epoch": 17.33, + "learning_rate": 1.2678408671018598e-06, + "loss": 0.1798, + "step": 46928 + }, + { + "epoch": 17.33, + "learning_rate": 1.2678039167385147e-06, + "loss": 0.2356, + "step": 46929 + }, + { + "epoch": 17.33, + "learning_rate": 1.2677669663751695e-06, + "loss": 0.2066, + "step": 46930 + }, + { + "epoch": 17.33, + "learning_rate": 1.2677300160118241e-06, + "loss": 0.3076, + "step": 46931 + }, + { + "epoch": 17.33, + "learning_rate": 1.267693065648479e-06, + "loss": 0.2969, + "step": 46932 + }, + { + "epoch": 17.33, + "learning_rate": 1.2676561152851336e-06, + "loss": 0.3416, + "step": 46933 + }, + { + "epoch": 17.33, + "learning_rate": 1.2676191649217883e-06, + "loss": 0.1769, + "step": 46934 + }, + { + "epoch": 17.33, + "learning_rate": 1.2675822145584431e-06, + "loss": 0.1935, + "step": 46935 + }, + { + "epoch": 17.33, + "learning_rate": 1.267545264195098e-06, + "loss": 0.272, + "step": 46936 + }, + { + "epoch": 17.33, + "learning_rate": 1.2675083138317526e-06, + "loss": 0.1071, + "step": 46937 + }, + { + "epoch": 17.33, + "learning_rate": 1.2674713634684074e-06, + "loss": 0.3465, + "step": 46938 + }, + { + "epoch": 17.33, + "learning_rate": 1.2674344131050623e-06, + "loss": 0.4306, + "step": 46939 + }, + { + "epoch": 17.33, + "learning_rate": 1.267397462741717e-06, + "loss": 0.2954, + "step": 46940 + }, + { + "epoch": 17.33, + "learning_rate": 1.2673605123783718e-06, + "loss": 0.1161, + "step": 46941 + }, + { + "epoch": 17.33, + "learning_rate": 1.2673235620150266e-06, + "loss": 0.3162, + "step": 46942 + }, + { + "epoch": 17.33, + "learning_rate": 1.2672866116516813e-06, + "loss": 0.1805, + "step": 46943 + }, + { + "epoch": 17.34, + "learning_rate": 1.267249661288336e-06, + "loss": 0.321, + "step": 46944 + }, + { + "epoch": 17.34, + "learning_rate": 1.2672127109249907e-06, + "loss": 0.2997, + "step": 46945 + }, + { + "epoch": 17.34, + "learning_rate": 1.2671757605616456e-06, + "loss": 0.3606, + "step": 46946 + }, + { + "epoch": 17.34, + "learning_rate": 1.2671388101983002e-06, + "loss": 0.1266, + "step": 46947 + }, + { + "epoch": 17.34, + "learning_rate": 1.267101859834955e-06, + "loss": 0.1937, + "step": 46948 + }, + { + "epoch": 17.34, + "learning_rate": 1.26706490947161e-06, + "loss": 0.1703, + "step": 46949 + }, + { + "epoch": 17.34, + "learning_rate": 1.2670279591082646e-06, + "loss": 0.2087, + "step": 46950 + }, + { + "epoch": 17.34, + "learning_rate": 1.2669910087449194e-06, + "loss": 0.3109, + "step": 46951 + }, + { + "epoch": 17.34, + "learning_rate": 1.2669540583815743e-06, + "loss": 0.2663, + "step": 46952 + }, + { + "epoch": 17.34, + "learning_rate": 1.266917108018229e-06, + "loss": 0.1736, + "step": 46953 + }, + { + "epoch": 17.34, + "learning_rate": 1.2668801576548835e-06, + "loss": 0.1378, + "step": 46954 + }, + { + "epoch": 17.34, + "learning_rate": 1.2668432072915384e-06, + "loss": 0.2026, + "step": 46955 + }, + { + "epoch": 17.34, + "learning_rate": 1.2668062569281932e-06, + "loss": 0.3654, + "step": 46956 + }, + { + "epoch": 17.34, + "learning_rate": 1.2667693065648479e-06, + "loss": 0.2523, + "step": 46957 + }, + { + "epoch": 17.34, + "learning_rate": 1.2667323562015027e-06, + "loss": 0.1683, + "step": 46958 + }, + { + "epoch": 17.34, + "learning_rate": 1.2666954058381576e-06, + "loss": 0.1312, + "step": 46959 + }, + { + "epoch": 17.34, + "learning_rate": 1.2666584554748122e-06, + "loss": 0.1978, + "step": 46960 + }, + { + "epoch": 17.34, + "learning_rate": 1.266621505111467e-06, + "loss": 0.2552, + "step": 46961 + }, + { + "epoch": 17.34, + "learning_rate": 1.266584554748122e-06, + "loss": 0.2863, + "step": 46962 + }, + { + "epoch": 17.34, + "learning_rate": 1.2665476043847763e-06, + "loss": 0.2559, + "step": 46963 + }, + { + "epoch": 17.34, + "learning_rate": 1.2665106540214312e-06, + "loss": 0.1578, + "step": 46964 + }, + { + "epoch": 17.34, + "learning_rate": 1.266473703658086e-06, + "loss": 0.1155, + "step": 46965 + }, + { + "epoch": 17.34, + "learning_rate": 1.2664367532947407e-06, + "loss": 0.2093, + "step": 46966 + }, + { + "epoch": 17.34, + "learning_rate": 1.2663998029313955e-06, + "loss": 0.156, + "step": 46967 + }, + { + "epoch": 17.34, + "learning_rate": 1.2663628525680504e-06, + "loss": 0.1724, + "step": 46968 + }, + { + "epoch": 17.34, + "learning_rate": 1.266325902204705e-06, + "loss": 0.2562, + "step": 46969 + }, + { + "epoch": 17.34, + "learning_rate": 1.2662889518413598e-06, + "loss": 0.1462, + "step": 46970 + }, + { + "epoch": 17.35, + "learning_rate": 1.2662520014780147e-06, + "loss": 0.1744, + "step": 46971 + }, + { + "epoch": 17.35, + "learning_rate": 1.2662150511146693e-06, + "loss": 0.1519, + "step": 46972 + }, + { + "epoch": 17.35, + "learning_rate": 1.266178100751324e-06, + "loss": 0.1814, + "step": 46973 + }, + { + "epoch": 17.35, + "learning_rate": 1.2661411503879788e-06, + "loss": 0.167, + "step": 46974 + }, + { + "epoch": 17.35, + "learning_rate": 1.2661042000246337e-06, + "loss": 0.1925, + "step": 46975 + }, + { + "epoch": 17.35, + "learning_rate": 1.2660672496612883e-06, + "loss": 0.283, + "step": 46976 + }, + { + "epoch": 17.35, + "learning_rate": 1.2660302992979431e-06, + "loss": 0.2093, + "step": 46977 + }, + { + "epoch": 17.35, + "learning_rate": 1.265993348934598e-06, + "loss": 0.2106, + "step": 46978 + }, + { + "epoch": 17.35, + "learning_rate": 1.2659563985712526e-06, + "loss": 0.2904, + "step": 46979 + }, + { + "epoch": 17.35, + "learning_rate": 1.2659194482079075e-06, + "loss": 0.3147, + "step": 46980 + }, + { + "epoch": 17.35, + "learning_rate": 1.2658824978445623e-06, + "loss": 0.0997, + "step": 46981 + }, + { + "epoch": 17.35, + "learning_rate": 1.265845547481217e-06, + "loss": 0.2013, + "step": 46982 + }, + { + "epoch": 17.35, + "learning_rate": 1.2658085971178716e-06, + "loss": 0.2978, + "step": 46983 + }, + { + "epoch": 17.35, + "learning_rate": 1.2657716467545264e-06, + "loss": 0.2739, + "step": 46984 + }, + { + "epoch": 17.35, + "learning_rate": 1.265734696391181e-06, + "loss": 0.2424, + "step": 46985 + }, + { + "epoch": 17.35, + "learning_rate": 1.265697746027836e-06, + "loss": 0.2612, + "step": 46986 + }, + { + "epoch": 17.35, + "learning_rate": 1.2656607956644908e-06, + "loss": 0.3161, + "step": 46987 + }, + { + "epoch": 17.35, + "learning_rate": 1.2656238453011454e-06, + "loss": 0.2528, + "step": 46988 + }, + { + "epoch": 17.35, + "learning_rate": 1.2655868949378003e-06, + "loss": 0.1651, + "step": 46989 + }, + { + "epoch": 17.35, + "learning_rate": 1.2655499445744551e-06, + "loss": 0.2007, + "step": 46990 + }, + { + "epoch": 17.35, + "learning_rate": 1.2655129942111097e-06, + "loss": 0.1971, + "step": 46991 + }, + { + "epoch": 17.35, + "learning_rate": 1.2654760438477646e-06, + "loss": 0.1976, + "step": 46992 + }, + { + "epoch": 17.35, + "learning_rate": 1.2654390934844192e-06, + "loss": 0.3024, + "step": 46993 + }, + { + "epoch": 17.35, + "learning_rate": 1.265402143121074e-06, + "loss": 0.3096, + "step": 46994 + }, + { + "epoch": 17.35, + "learning_rate": 1.2653651927577287e-06, + "loss": 0.1847, + "step": 46995 + }, + { + "epoch": 17.35, + "learning_rate": 1.2653282423943836e-06, + "loss": 0.3411, + "step": 46996 + }, + { + "epoch": 17.35, + "learning_rate": 1.2652912920310384e-06, + "loss": 0.1871, + "step": 46997 + }, + { + "epoch": 17.36, + "learning_rate": 1.265254341667693e-06, + "loss": 0.157, + "step": 46998 + }, + { + "epoch": 17.36, + "learning_rate": 1.2652173913043479e-06, + "loss": 0.216, + "step": 46999 + }, + { + "epoch": 17.36, + "learning_rate": 1.2651804409410027e-06, + "loss": 0.258, + "step": 47000 + }, + { + "epoch": 17.36, + "learning_rate": 1.2651434905776574e-06, + "loss": 0.4155, + "step": 47001 + }, + { + "epoch": 17.36, + "learning_rate": 1.2651065402143122e-06, + "loss": 0.1924, + "step": 47002 + }, + { + "epoch": 17.36, + "learning_rate": 1.2650695898509669e-06, + "loss": 0.1714, + "step": 47003 + }, + { + "epoch": 17.36, + "learning_rate": 1.2650326394876215e-06, + "loss": 0.2297, + "step": 47004 + }, + { + "epoch": 17.36, + "learning_rate": 1.2649956891242763e-06, + "loss": 0.2359, + "step": 47005 + }, + { + "epoch": 17.36, + "learning_rate": 1.2649587387609312e-06, + "loss": 0.1567, + "step": 47006 + }, + { + "epoch": 17.36, + "learning_rate": 1.2649217883975858e-06, + "loss": 0.1995, + "step": 47007 + }, + { + "epoch": 17.36, + "learning_rate": 1.2648848380342407e-06, + "loss": 0.2121, + "step": 47008 + }, + { + "epoch": 17.36, + "learning_rate": 1.2648478876708955e-06, + "loss": 0.1875, + "step": 47009 + }, + { + "epoch": 17.36, + "learning_rate": 1.2648109373075502e-06, + "loss": 0.182, + "step": 47010 + }, + { + "epoch": 17.36, + "learning_rate": 1.264773986944205e-06, + "loss": 0.1471, + "step": 47011 + }, + { + "epoch": 17.36, + "learning_rate": 1.2647370365808599e-06, + "loss": 0.1782, + "step": 47012 + }, + { + "epoch": 17.36, + "learning_rate": 1.2647000862175145e-06, + "loss": 0.2866, + "step": 47013 + }, + { + "epoch": 17.36, + "learning_rate": 1.2646631358541691e-06, + "loss": 0.1648, + "step": 47014 + }, + { + "epoch": 17.36, + "learning_rate": 1.264626185490824e-06, + "loss": 0.2148, + "step": 47015 + }, + { + "epoch": 17.36, + "learning_rate": 1.2645892351274788e-06, + "loss": 0.2044, + "step": 47016 + }, + { + "epoch": 17.36, + "learning_rate": 1.2645522847641335e-06, + "loss": 0.3403, + "step": 47017 + }, + { + "epoch": 17.36, + "learning_rate": 1.2645153344007883e-06, + "loss": 0.2673, + "step": 47018 + }, + { + "epoch": 17.36, + "learning_rate": 1.2644783840374432e-06, + "loss": 0.1994, + "step": 47019 + }, + { + "epoch": 17.36, + "learning_rate": 1.2644414336740978e-06, + "loss": 0.2907, + "step": 47020 + }, + { + "epoch": 17.36, + "learning_rate": 1.2644044833107526e-06, + "loss": 0.2923, + "step": 47021 + }, + { + "epoch": 17.36, + "learning_rate": 1.2643675329474075e-06, + "loss": 0.1618, + "step": 47022 + }, + { + "epoch": 17.36, + "learning_rate": 1.2643305825840621e-06, + "loss": 0.1658, + "step": 47023 + }, + { + "epoch": 17.36, + "learning_rate": 1.2642936322207168e-06, + "loss": 0.2407, + "step": 47024 + }, + { + "epoch": 17.37, + "learning_rate": 1.2642566818573716e-06, + "loss": 0.3013, + "step": 47025 + }, + { + "epoch": 17.37, + "learning_rate": 1.2642197314940265e-06, + "loss": 0.1722, + "step": 47026 + }, + { + "epoch": 17.37, + "learning_rate": 1.264182781130681e-06, + "loss": 0.2186, + "step": 47027 + }, + { + "epoch": 17.37, + "learning_rate": 1.264145830767336e-06, + "loss": 0.2464, + "step": 47028 + }, + { + "epoch": 17.37, + "learning_rate": 1.2641088804039908e-06, + "loss": 0.2418, + "step": 47029 + }, + { + "epoch": 17.37, + "learning_rate": 1.2640719300406454e-06, + "loss": 0.3342, + "step": 47030 + }, + { + "epoch": 17.37, + "learning_rate": 1.2640349796773003e-06, + "loss": 0.2718, + "step": 47031 + }, + { + "epoch": 17.37, + "learning_rate": 1.2639980293139551e-06, + "loss": 0.2975, + "step": 47032 + }, + { + "epoch": 17.37, + "learning_rate": 1.2639610789506096e-06, + "loss": 0.2523, + "step": 47033 + }, + { + "epoch": 17.37, + "learning_rate": 1.2639241285872644e-06, + "loss": 0.1669, + "step": 47034 + }, + { + "epoch": 17.37, + "learning_rate": 1.2638871782239193e-06, + "loss": 0.2227, + "step": 47035 + }, + { + "epoch": 17.37, + "learning_rate": 1.2638502278605739e-06, + "loss": 0.4052, + "step": 47036 + }, + { + "epoch": 17.37, + "learning_rate": 1.2638132774972287e-06, + "loss": 0.14, + "step": 47037 + }, + { + "epoch": 17.37, + "learning_rate": 1.2637763271338836e-06, + "loss": 0.1813, + "step": 47038 + }, + { + "epoch": 17.37, + "learning_rate": 1.2637393767705382e-06, + "loss": 0.1668, + "step": 47039 + }, + { + "epoch": 17.37, + "learning_rate": 1.263702426407193e-06, + "loss": 0.2656, + "step": 47040 + }, + { + "epoch": 17.37, + "learning_rate": 1.263665476043848e-06, + "loss": 0.276, + "step": 47041 + }, + { + "epoch": 17.37, + "learning_rate": 1.2636285256805026e-06, + "loss": 0.1676, + "step": 47042 + }, + { + "epoch": 17.37, + "learning_rate": 1.2635915753171572e-06, + "loss": 0.4106, + "step": 47043 + }, + { + "epoch": 17.37, + "learning_rate": 1.263554624953812e-06, + "loss": 0.181, + "step": 47044 + }, + { + "epoch": 17.37, + "learning_rate": 1.2635176745904669e-06, + "loss": 0.3568, + "step": 47045 + }, + { + "epoch": 17.37, + "learning_rate": 1.2634807242271215e-06, + "loss": 0.1617, + "step": 47046 + }, + { + "epoch": 17.37, + "learning_rate": 1.2634437738637764e-06, + "loss": 0.2023, + "step": 47047 + }, + { + "epoch": 17.37, + "learning_rate": 1.2634068235004312e-06, + "loss": 0.2868, + "step": 47048 + }, + { + "epoch": 17.37, + "learning_rate": 1.2633698731370859e-06, + "loss": 0.2844, + "step": 47049 + }, + { + "epoch": 17.37, + "learning_rate": 1.2633329227737407e-06, + "loss": 0.1695, + "step": 47050 + }, + { + "epoch": 17.37, + "learning_rate": 1.2632959724103956e-06, + "loss": 0.2317, + "step": 47051 + }, + { + "epoch": 17.38, + "learning_rate": 1.2632590220470502e-06, + "loss": 0.2463, + "step": 47052 + }, + { + "epoch": 17.38, + "learning_rate": 1.2632220716837048e-06, + "loss": 0.3846, + "step": 47053 + }, + { + "epoch": 17.38, + "learning_rate": 1.2631851213203597e-06, + "loss": 0.1263, + "step": 47054 + }, + { + "epoch": 17.38, + "learning_rate": 1.2631481709570143e-06, + "loss": 0.3402, + "step": 47055 + }, + { + "epoch": 17.38, + "learning_rate": 1.2631112205936692e-06, + "loss": 0.2021, + "step": 47056 + }, + { + "epoch": 17.38, + "learning_rate": 1.263074270230324e-06, + "loss": 0.2141, + "step": 47057 + }, + { + "epoch": 17.38, + "learning_rate": 1.2630373198669786e-06, + "loss": 0.1755, + "step": 47058 + }, + { + "epoch": 17.38, + "learning_rate": 1.2630003695036335e-06, + "loss": 0.2665, + "step": 47059 + }, + { + "epoch": 17.38, + "learning_rate": 1.2629634191402883e-06, + "loss": 0.1931, + "step": 47060 + }, + { + "epoch": 17.38, + "learning_rate": 1.262926468776943e-06, + "loss": 0.204, + "step": 47061 + }, + { + "epoch": 17.38, + "learning_rate": 1.2628895184135978e-06, + "loss": 0.1905, + "step": 47062 + }, + { + "epoch": 17.38, + "learning_rate": 1.2628525680502525e-06, + "loss": 0.3106, + "step": 47063 + }, + { + "epoch": 17.38, + "learning_rate": 1.2628156176869073e-06, + "loss": 0.2352, + "step": 47064 + }, + { + "epoch": 17.38, + "learning_rate": 1.262778667323562e-06, + "loss": 0.1737, + "step": 47065 + }, + { + "epoch": 17.38, + "learning_rate": 1.2627417169602168e-06, + "loss": 0.2558, + "step": 47066 + }, + { + "epoch": 17.38, + "learning_rate": 1.2627047665968716e-06, + "loss": 0.1317, + "step": 47067 + }, + { + "epoch": 17.38, + "learning_rate": 1.2626678162335263e-06, + "loss": 0.2067, + "step": 47068 + }, + { + "epoch": 17.38, + "learning_rate": 1.2626308658701811e-06, + "loss": 0.2803, + "step": 47069 + }, + { + "epoch": 17.38, + "learning_rate": 1.262593915506836e-06, + "loss": 0.3028, + "step": 47070 + }, + { + "epoch": 17.38, + "learning_rate": 1.2625569651434906e-06, + "loss": 0.3603, + "step": 47071 + }, + { + "epoch": 17.38, + "learning_rate": 1.2625200147801455e-06, + "loss": 0.3658, + "step": 47072 + }, + { + "epoch": 17.38, + "learning_rate": 1.2624830644168e-06, + "loss": 0.1435, + "step": 47073 + }, + { + "epoch": 17.38, + "learning_rate": 1.2624461140534547e-06, + "loss": 0.2268, + "step": 47074 + }, + { + "epoch": 17.38, + "learning_rate": 1.2624091636901096e-06, + "loss": 0.2426, + "step": 47075 + }, + { + "epoch": 17.38, + "learning_rate": 1.2623722133267644e-06, + "loss": 0.4103, + "step": 47076 + }, + { + "epoch": 17.38, + "learning_rate": 1.262335262963419e-06, + "loss": 0.1591, + "step": 47077 + }, + { + "epoch": 17.38, + "learning_rate": 1.262298312600074e-06, + "loss": 0.1738, + "step": 47078 + }, + { + "epoch": 17.39, + "learning_rate": 1.2622613622367288e-06, + "loss": 0.2344, + "step": 47079 + }, + { + "epoch": 17.39, + "learning_rate": 1.2622244118733834e-06, + "loss": 0.2333, + "step": 47080 + }, + { + "epoch": 17.39, + "learning_rate": 1.2621874615100382e-06, + "loss": 0.1565, + "step": 47081 + }, + { + "epoch": 17.39, + "learning_rate": 1.262150511146693e-06, + "loss": 0.1513, + "step": 47082 + }, + { + "epoch": 17.39, + "learning_rate": 1.2621135607833477e-06, + "loss": 0.2237, + "step": 47083 + }, + { + "epoch": 17.39, + "learning_rate": 1.2620766104200024e-06, + "loss": 0.2631, + "step": 47084 + }, + { + "epoch": 17.39, + "learning_rate": 1.2620396600566572e-06, + "loss": 0.3543, + "step": 47085 + }, + { + "epoch": 17.39, + "learning_rate": 1.262002709693312e-06, + "loss": 0.2278, + "step": 47086 + }, + { + "epoch": 17.39, + "learning_rate": 1.2619657593299667e-06, + "loss": 0.2792, + "step": 47087 + }, + { + "epoch": 17.39, + "learning_rate": 1.2619288089666215e-06, + "loss": 0.4044, + "step": 47088 + }, + { + "epoch": 17.39, + "learning_rate": 1.2618918586032764e-06, + "loss": 0.2973, + "step": 47089 + }, + { + "epoch": 17.39, + "learning_rate": 1.261854908239931e-06, + "loss": 0.2903, + "step": 47090 + }, + { + "epoch": 17.39, + "learning_rate": 1.2618179578765859e-06, + "loss": 0.2291, + "step": 47091 + }, + { + "epoch": 17.39, + "learning_rate": 1.2617810075132407e-06, + "loss": 0.3324, + "step": 47092 + }, + { + "epoch": 17.39, + "learning_rate": 1.2617440571498954e-06, + "loss": 0.3127, + "step": 47093 + }, + { + "epoch": 17.39, + "learning_rate": 1.26170710678655e-06, + "loss": 0.2835, + "step": 47094 + }, + { + "epoch": 17.39, + "learning_rate": 1.2616701564232048e-06, + "loss": 0.2997, + "step": 47095 + }, + { + "epoch": 17.39, + "learning_rate": 1.2616332060598597e-06, + "loss": 0.3237, + "step": 47096 + }, + { + "epoch": 17.39, + "learning_rate": 1.2615962556965143e-06, + "loss": 0.1048, + "step": 47097 + }, + { + "epoch": 17.39, + "learning_rate": 1.2615593053331692e-06, + "loss": 0.1409, + "step": 47098 + }, + { + "epoch": 17.39, + "learning_rate": 1.261522354969824e-06, + "loss": 0.2292, + "step": 47099 + }, + { + "epoch": 17.39, + "learning_rate": 1.2614854046064787e-06, + "loss": 0.1391, + "step": 47100 + }, + { + "epoch": 17.39, + "learning_rate": 1.2614484542431335e-06, + "loss": 0.262, + "step": 47101 + }, + { + "epoch": 17.39, + "learning_rate": 1.2614115038797884e-06, + "loss": 0.1457, + "step": 47102 + }, + { + "epoch": 17.39, + "learning_rate": 1.2613745535164428e-06, + "loss": 0.257, + "step": 47103 + }, + { + "epoch": 17.39, + "learning_rate": 1.2613376031530976e-06, + "loss": 0.2841, + "step": 47104 + }, + { + "epoch": 17.39, + "learning_rate": 1.2613006527897525e-06, + "loss": 0.3727, + "step": 47105 + }, + { + "epoch": 17.4, + "learning_rate": 1.2612637024264071e-06, + "loss": 0.1425, + "step": 47106 + }, + { + "epoch": 17.4, + "learning_rate": 1.261226752063062e-06, + "loss": 0.1745, + "step": 47107 + }, + { + "epoch": 17.4, + "learning_rate": 1.2611898016997168e-06, + "loss": 0.3171, + "step": 47108 + }, + { + "epoch": 17.4, + "learning_rate": 1.2611528513363715e-06, + "loss": 0.2239, + "step": 47109 + }, + { + "epoch": 17.4, + "learning_rate": 1.2611159009730263e-06, + "loss": 0.2485, + "step": 47110 + }, + { + "epoch": 17.4, + "learning_rate": 1.2610789506096811e-06, + "loss": 0.2658, + "step": 47111 + }, + { + "epoch": 17.4, + "learning_rate": 1.2610420002463358e-06, + "loss": 0.2593, + "step": 47112 + }, + { + "epoch": 17.4, + "learning_rate": 1.2610050498829904e-06, + "loss": 0.3463, + "step": 47113 + }, + { + "epoch": 17.4, + "learning_rate": 1.2609680995196453e-06, + "loss": 0.209, + "step": 47114 + }, + { + "epoch": 17.4, + "learning_rate": 1.2609311491563001e-06, + "loss": 0.2703, + "step": 47115 + }, + { + "epoch": 17.4, + "learning_rate": 1.2608941987929548e-06, + "loss": 0.1319, + "step": 47116 + }, + { + "epoch": 17.4, + "learning_rate": 1.2608572484296096e-06, + "loss": 0.108, + "step": 47117 + }, + { + "epoch": 17.4, + "learning_rate": 1.2608202980662645e-06, + "loss": 0.2049, + "step": 47118 + }, + { + "epoch": 17.4, + "learning_rate": 1.260783347702919e-06, + "loss": 0.2008, + "step": 47119 + }, + { + "epoch": 17.4, + "learning_rate": 1.260746397339574e-06, + "loss": 0.2186, + "step": 47120 + }, + { + "epoch": 17.4, + "learning_rate": 1.2607094469762288e-06, + "loss": 0.1696, + "step": 47121 + }, + { + "epoch": 17.4, + "learning_rate": 1.2606724966128834e-06, + "loss": 0.1958, + "step": 47122 + }, + { + "epoch": 17.4, + "learning_rate": 1.260635546249538e-06, + "loss": 0.3184, + "step": 47123 + }, + { + "epoch": 17.4, + "learning_rate": 1.260598595886193e-06, + "loss": 0.2712, + "step": 47124 + }, + { + "epoch": 17.4, + "learning_rate": 1.2605616455228475e-06, + "loss": 0.1805, + "step": 47125 + }, + { + "epoch": 17.4, + "learning_rate": 1.2605246951595024e-06, + "loss": 0.2694, + "step": 47126 + }, + { + "epoch": 17.4, + "learning_rate": 1.2604877447961572e-06, + "loss": 0.3014, + "step": 47127 + }, + { + "epoch": 17.4, + "learning_rate": 1.2604507944328119e-06, + "loss": 0.3386, + "step": 47128 + }, + { + "epoch": 17.4, + "learning_rate": 1.2604138440694667e-06, + "loss": 0.2311, + "step": 47129 + }, + { + "epoch": 17.4, + "learning_rate": 1.2603768937061216e-06, + "loss": 0.159, + "step": 47130 + }, + { + "epoch": 17.4, + "learning_rate": 1.2603399433427762e-06, + "loss": 0.2335, + "step": 47131 + }, + { + "epoch": 17.4, + "learning_rate": 1.260302992979431e-06, + "loss": 0.2281, + "step": 47132 + }, + { + "epoch": 17.41, + "learning_rate": 1.2602660426160857e-06, + "loss": 0.2694, + "step": 47133 + }, + { + "epoch": 17.41, + "learning_rate": 1.2602290922527405e-06, + "loss": 0.1267, + "step": 47134 + }, + { + "epoch": 17.41, + "learning_rate": 1.2601921418893952e-06, + "loss": 0.2682, + "step": 47135 + }, + { + "epoch": 17.41, + "learning_rate": 1.26015519152605e-06, + "loss": 0.1912, + "step": 47136 + }, + { + "epoch": 17.41, + "learning_rate": 1.2601182411627049e-06, + "loss": 0.2287, + "step": 47137 + }, + { + "epoch": 17.41, + "learning_rate": 1.2600812907993595e-06, + "loss": 0.1578, + "step": 47138 + }, + { + "epoch": 17.41, + "learning_rate": 1.2600443404360144e-06, + "loss": 0.17, + "step": 47139 + }, + { + "epoch": 17.41, + "learning_rate": 1.2600073900726692e-06, + "loss": 0.2283, + "step": 47140 + }, + { + "epoch": 17.41, + "learning_rate": 1.2599704397093238e-06, + "loss": 0.1445, + "step": 47141 + }, + { + "epoch": 17.41, + "learning_rate": 1.2599334893459785e-06, + "loss": 0.2271, + "step": 47142 + }, + { + "epoch": 17.41, + "learning_rate": 1.2598965389826333e-06, + "loss": 0.2183, + "step": 47143 + }, + { + "epoch": 17.41, + "learning_rate": 1.2598595886192882e-06, + "loss": 0.2558, + "step": 47144 + }, + { + "epoch": 17.41, + "learning_rate": 1.2598226382559428e-06, + "loss": 0.1281, + "step": 47145 + }, + { + "epoch": 17.41, + "learning_rate": 1.2597856878925977e-06, + "loss": 0.2007, + "step": 47146 + }, + { + "epoch": 17.41, + "learning_rate": 1.2597487375292525e-06, + "loss": 0.3022, + "step": 47147 + }, + { + "epoch": 17.41, + "learning_rate": 1.2597117871659071e-06, + "loss": 0.3192, + "step": 47148 + }, + { + "epoch": 17.41, + "learning_rate": 1.259674836802562e-06, + "loss": 0.2285, + "step": 47149 + }, + { + "epoch": 17.41, + "learning_rate": 1.2596378864392166e-06, + "loss": 0.2212, + "step": 47150 + }, + { + "epoch": 17.41, + "learning_rate": 1.2596009360758715e-06, + "loss": 0.301, + "step": 47151 + }, + { + "epoch": 17.41, + "learning_rate": 1.2595639857125261e-06, + "loss": 0.239, + "step": 47152 + }, + { + "epoch": 17.41, + "learning_rate": 1.259527035349181e-06, + "loss": 0.2873, + "step": 47153 + }, + { + "epoch": 17.41, + "learning_rate": 1.2594900849858356e-06, + "loss": 0.1983, + "step": 47154 + }, + { + "epoch": 17.41, + "learning_rate": 1.2594531346224904e-06, + "loss": 0.148, + "step": 47155 + }, + { + "epoch": 17.41, + "learning_rate": 1.2594161842591453e-06, + "loss": 0.1733, + "step": 47156 + }, + { + "epoch": 17.41, + "learning_rate": 1.2593792338958e-06, + "loss": 0.1563, + "step": 47157 + }, + { + "epoch": 17.41, + "learning_rate": 1.2593422835324548e-06, + "loss": 0.2741, + "step": 47158 + }, + { + "epoch": 17.41, + "learning_rate": 1.2593053331691096e-06, + "loss": 0.198, + "step": 47159 + }, + { + "epoch": 17.42, + "learning_rate": 1.2592683828057643e-06, + "loss": 0.165, + "step": 47160 + }, + { + "epoch": 17.42, + "learning_rate": 1.2592314324424191e-06, + "loss": 0.2633, + "step": 47161 + }, + { + "epoch": 17.42, + "learning_rate": 1.2591944820790737e-06, + "loss": 0.1634, + "step": 47162 + }, + { + "epoch": 17.42, + "learning_rate": 1.2591575317157286e-06, + "loss": 0.3167, + "step": 47163 + }, + { + "epoch": 17.42, + "learning_rate": 1.2591205813523832e-06, + "loss": 0.2038, + "step": 47164 + }, + { + "epoch": 17.42, + "learning_rate": 1.259083630989038e-06, + "loss": 0.2509, + "step": 47165 + }, + { + "epoch": 17.42, + "learning_rate": 1.259046680625693e-06, + "loss": 0.2042, + "step": 47166 + }, + { + "epoch": 17.42, + "learning_rate": 1.2590097302623476e-06, + "loss": 0.1862, + "step": 47167 + }, + { + "epoch": 17.42, + "learning_rate": 1.2589727798990024e-06, + "loss": 0.2159, + "step": 47168 + }, + { + "epoch": 17.42, + "learning_rate": 1.2589358295356573e-06, + "loss": 0.2882, + "step": 47169 + }, + { + "epoch": 17.42, + "learning_rate": 1.258898879172312e-06, + "loss": 0.1761, + "step": 47170 + }, + { + "epoch": 17.42, + "learning_rate": 1.2588619288089667e-06, + "loss": 0.2876, + "step": 47171 + }, + { + "epoch": 17.42, + "learning_rate": 1.2588249784456214e-06, + "loss": 0.1823, + "step": 47172 + }, + { + "epoch": 17.42, + "learning_rate": 1.258788028082276e-06, + "loss": 0.443, + "step": 47173 + }, + { + "epoch": 17.42, + "learning_rate": 1.2587510777189309e-06, + "loss": 0.1638, + "step": 47174 + }, + { + "epoch": 17.42, + "learning_rate": 1.2587141273555857e-06, + "loss": 0.2786, + "step": 47175 + }, + { + "epoch": 17.42, + "learning_rate": 1.2586771769922404e-06, + "loss": 0.2175, + "step": 47176 + }, + { + "epoch": 17.42, + "learning_rate": 1.2586402266288952e-06, + "loss": 0.2006, + "step": 47177 + }, + { + "epoch": 17.42, + "learning_rate": 1.25860327626555e-06, + "loss": 0.2201, + "step": 47178 + }, + { + "epoch": 17.42, + "learning_rate": 1.2585663259022047e-06, + "loss": 0.1515, + "step": 47179 + }, + { + "epoch": 17.42, + "learning_rate": 1.2585293755388595e-06, + "loss": 0.2901, + "step": 47180 + }, + { + "epoch": 17.42, + "learning_rate": 1.2584924251755144e-06, + "loss": 0.2144, + "step": 47181 + }, + { + "epoch": 17.42, + "learning_rate": 1.258455474812169e-06, + "loss": 0.2033, + "step": 47182 + }, + { + "epoch": 17.42, + "learning_rate": 1.2584185244488237e-06, + "loss": 0.3406, + "step": 47183 + }, + { + "epoch": 17.42, + "learning_rate": 1.2583815740854785e-06, + "loss": 0.1775, + "step": 47184 + }, + { + "epoch": 17.42, + "learning_rate": 1.2583446237221334e-06, + "loss": 0.251, + "step": 47185 + }, + { + "epoch": 17.42, + "learning_rate": 1.258307673358788e-06, + "loss": 0.2472, + "step": 47186 + }, + { + "epoch": 17.43, + "learning_rate": 1.2582707229954428e-06, + "loss": 0.1305, + "step": 47187 + }, + { + "epoch": 17.43, + "learning_rate": 1.2582337726320977e-06, + "loss": 0.275, + "step": 47188 + }, + { + "epoch": 17.43, + "learning_rate": 1.2581968222687523e-06, + "loss": 0.2102, + "step": 47189 + }, + { + "epoch": 17.43, + "learning_rate": 1.2581598719054072e-06, + "loss": 0.3053, + "step": 47190 + }, + { + "epoch": 17.43, + "learning_rate": 1.258122921542062e-06, + "loss": 0.1462, + "step": 47191 + }, + { + "epoch": 17.43, + "learning_rate": 1.2580859711787164e-06, + "loss": 0.2641, + "step": 47192 + }, + { + "epoch": 17.43, + "learning_rate": 1.2580490208153713e-06, + "loss": 0.1768, + "step": 47193 + }, + { + "epoch": 17.43, + "learning_rate": 1.2580120704520261e-06, + "loss": 0.2475, + "step": 47194 + }, + { + "epoch": 17.43, + "learning_rate": 1.2579751200886808e-06, + "loss": 0.18, + "step": 47195 + }, + { + "epoch": 17.43, + "learning_rate": 1.2579381697253356e-06, + "loss": 0.1917, + "step": 47196 + }, + { + "epoch": 17.43, + "learning_rate": 1.2579012193619905e-06, + "loss": 0.335, + "step": 47197 + }, + { + "epoch": 17.43, + "learning_rate": 1.2578642689986451e-06, + "loss": 0.285, + "step": 47198 + }, + { + "epoch": 17.43, + "learning_rate": 1.2578273186353e-06, + "loss": 0.1909, + "step": 47199 + }, + { + "epoch": 17.43, + "learning_rate": 1.2577903682719548e-06, + "loss": 0.1974, + "step": 47200 + }, + { + "epoch": 17.43, + "learning_rate": 1.2577534179086094e-06, + "loss": 0.1683, + "step": 47201 + }, + { + "epoch": 17.43, + "learning_rate": 1.257716467545264e-06, + "loss": 0.2641, + "step": 47202 + }, + { + "epoch": 17.43, + "learning_rate": 1.257679517181919e-06, + "loss": 0.3, + "step": 47203 + }, + { + "epoch": 17.43, + "learning_rate": 1.2576425668185738e-06, + "loss": 0.3113, + "step": 47204 + }, + { + "epoch": 17.43, + "learning_rate": 1.2576056164552284e-06, + "loss": 0.1877, + "step": 47205 + }, + { + "epoch": 17.43, + "learning_rate": 1.2575686660918833e-06, + "loss": 0.2143, + "step": 47206 + }, + { + "epoch": 17.43, + "learning_rate": 1.2575317157285381e-06, + "loss": 0.2136, + "step": 47207 + }, + { + "epoch": 17.43, + "learning_rate": 1.2574947653651927e-06, + "loss": 0.1724, + "step": 47208 + }, + { + "epoch": 17.43, + "learning_rate": 1.2574578150018476e-06, + "loss": 0.2836, + "step": 47209 + }, + { + "epoch": 17.43, + "learning_rate": 1.2574208646385024e-06, + "loss": 0.2192, + "step": 47210 + }, + { + "epoch": 17.43, + "learning_rate": 1.257383914275157e-06, + "loss": 0.281, + "step": 47211 + }, + { + "epoch": 17.43, + "learning_rate": 1.2573469639118117e-06, + "loss": 0.253, + "step": 47212 + }, + { + "epoch": 17.43, + "learning_rate": 1.2573100135484666e-06, + "loss": 0.1587, + "step": 47213 + }, + { + "epoch": 17.44, + "learning_rate": 1.2572730631851214e-06, + "loss": 0.2003, + "step": 47214 + }, + { + "epoch": 17.44, + "learning_rate": 1.257236112821776e-06, + "loss": 0.252, + "step": 47215 + }, + { + "epoch": 17.44, + "learning_rate": 1.2571991624584309e-06, + "loss": 0.2433, + "step": 47216 + }, + { + "epoch": 17.44, + "learning_rate": 1.2571622120950857e-06, + "loss": 0.2148, + "step": 47217 + }, + { + "epoch": 17.44, + "learning_rate": 1.2571252617317404e-06, + "loss": 0.1563, + "step": 47218 + }, + { + "epoch": 17.44, + "learning_rate": 1.2570883113683952e-06, + "loss": 0.2424, + "step": 47219 + }, + { + "epoch": 17.44, + "learning_rate": 1.25705136100505e-06, + "loss": 0.1984, + "step": 47220 + }, + { + "epoch": 17.44, + "learning_rate": 1.2570144106417047e-06, + "loss": 0.1591, + "step": 47221 + }, + { + "epoch": 17.44, + "learning_rate": 1.2569774602783593e-06, + "loss": 0.1014, + "step": 47222 + }, + { + "epoch": 17.44, + "learning_rate": 1.2569405099150142e-06, + "loss": 0.1549, + "step": 47223 + }, + { + "epoch": 17.44, + "learning_rate": 1.2569035595516688e-06, + "loss": 0.1919, + "step": 47224 + }, + { + "epoch": 17.44, + "learning_rate": 1.2568666091883237e-06, + "loss": 0.3075, + "step": 47225 + }, + { + "epoch": 17.44, + "learning_rate": 1.2568296588249785e-06, + "loss": 0.2762, + "step": 47226 + }, + { + "epoch": 17.44, + "learning_rate": 1.2567927084616332e-06, + "loss": 0.2375, + "step": 47227 + }, + { + "epoch": 17.44, + "learning_rate": 1.256755758098288e-06, + "loss": 0.2069, + "step": 47228 + }, + { + "epoch": 17.44, + "learning_rate": 1.2567188077349429e-06, + "loss": 0.1482, + "step": 47229 + }, + { + "epoch": 17.44, + "learning_rate": 1.2566818573715975e-06, + "loss": 0.2058, + "step": 47230 + }, + { + "epoch": 17.44, + "learning_rate": 1.2566449070082523e-06, + "loss": 0.1379, + "step": 47231 + }, + { + "epoch": 17.44, + "learning_rate": 1.256607956644907e-06, + "loss": 0.1518, + "step": 47232 + }, + { + "epoch": 17.44, + "learning_rate": 1.2565710062815618e-06, + "loss": 0.257, + "step": 47233 + }, + { + "epoch": 17.44, + "learning_rate": 1.2565340559182165e-06, + "loss": 0.1594, + "step": 47234 + }, + { + "epoch": 17.44, + "learning_rate": 1.2564971055548713e-06, + "loss": 0.3354, + "step": 47235 + }, + { + "epoch": 17.44, + "learning_rate": 1.2564601551915262e-06, + "loss": 0.2326, + "step": 47236 + }, + { + "epoch": 17.44, + "learning_rate": 1.2564232048281808e-06, + "loss": 0.2668, + "step": 47237 + }, + { + "epoch": 17.44, + "learning_rate": 1.2563862544648356e-06, + "loss": 0.1905, + "step": 47238 + }, + { + "epoch": 17.44, + "learning_rate": 1.2563493041014905e-06, + "loss": 0.2717, + "step": 47239 + }, + { + "epoch": 17.44, + "learning_rate": 1.2563123537381451e-06, + "loss": 0.29, + "step": 47240 + }, + { + "epoch": 17.44, + "learning_rate": 1.2562754033748e-06, + "loss": 0.1625, + "step": 47241 + }, + { + "epoch": 17.45, + "learning_rate": 1.2562384530114546e-06, + "loss": 0.2735, + "step": 47242 + }, + { + "epoch": 17.45, + "learning_rate": 1.2562015026481093e-06, + "loss": 0.2749, + "step": 47243 + }, + { + "epoch": 17.45, + "learning_rate": 1.256164552284764e-06, + "loss": 0.2497, + "step": 47244 + }, + { + "epoch": 17.45, + "learning_rate": 1.256127601921419e-06, + "loss": 0.3029, + "step": 47245 + }, + { + "epoch": 17.45, + "learning_rate": 1.2560906515580736e-06, + "loss": 0.2351, + "step": 47246 + }, + { + "epoch": 17.45, + "learning_rate": 1.2560537011947284e-06, + "loss": 0.2961, + "step": 47247 + }, + { + "epoch": 17.45, + "learning_rate": 1.2560167508313833e-06, + "loss": 0.258, + "step": 47248 + }, + { + "epoch": 17.45, + "learning_rate": 1.255979800468038e-06, + "loss": 0.3323, + "step": 47249 + }, + { + "epoch": 17.45, + "learning_rate": 1.2559428501046928e-06, + "loss": 0.2179, + "step": 47250 + }, + { + "epoch": 17.45, + "learning_rate": 1.2559058997413476e-06, + "loss": 0.1224, + "step": 47251 + }, + { + "epoch": 17.45, + "learning_rate": 1.2558689493780023e-06, + "loss": 0.3217, + "step": 47252 + }, + { + "epoch": 17.45, + "learning_rate": 1.2558319990146569e-06, + "loss": 0.2319, + "step": 47253 + }, + { + "epoch": 17.45, + "learning_rate": 1.2557950486513117e-06, + "loss": 0.2096, + "step": 47254 + }, + { + "epoch": 17.45, + "learning_rate": 1.2557580982879666e-06, + "loss": 0.3498, + "step": 47255 + }, + { + "epoch": 17.45, + "learning_rate": 1.2557211479246212e-06, + "loss": 0.1304, + "step": 47256 + }, + { + "epoch": 17.45, + "learning_rate": 1.255684197561276e-06, + "loss": 0.1166, + "step": 47257 + }, + { + "epoch": 17.45, + "learning_rate": 1.255647247197931e-06, + "loss": 0.2251, + "step": 47258 + }, + { + "epoch": 17.45, + "learning_rate": 1.2556102968345856e-06, + "loss": 0.272, + "step": 47259 + }, + { + "epoch": 17.45, + "learning_rate": 1.2555733464712404e-06, + "loss": 0.2728, + "step": 47260 + }, + { + "epoch": 17.45, + "learning_rate": 1.2555363961078952e-06, + "loss": 0.2635, + "step": 47261 + }, + { + "epoch": 17.45, + "learning_rate": 1.2554994457445497e-06, + "loss": 0.1152, + "step": 47262 + }, + { + "epoch": 17.45, + "learning_rate": 1.2554624953812045e-06, + "loss": 0.2779, + "step": 47263 + }, + { + "epoch": 17.45, + "learning_rate": 1.2554255450178594e-06, + "loss": 0.324, + "step": 47264 + }, + { + "epoch": 17.45, + "learning_rate": 1.255388594654514e-06, + "loss": 0.1809, + "step": 47265 + }, + { + "epoch": 17.45, + "learning_rate": 1.2553516442911689e-06, + "loss": 0.1312, + "step": 47266 + }, + { + "epoch": 17.45, + "learning_rate": 1.2553146939278237e-06, + "loss": 0.1891, + "step": 47267 + }, + { + "epoch": 17.45, + "learning_rate": 1.2552777435644783e-06, + "loss": 0.2951, + "step": 47268 + }, + { + "epoch": 17.46, + "learning_rate": 1.2552407932011332e-06, + "loss": 0.2455, + "step": 47269 + }, + { + "epoch": 17.46, + "learning_rate": 1.255203842837788e-06, + "loss": 0.3175, + "step": 47270 + }, + { + "epoch": 17.46, + "learning_rate": 1.2551668924744427e-06, + "loss": 0.1351, + "step": 47271 + }, + { + "epoch": 17.46, + "learning_rate": 1.2551299421110973e-06, + "loss": 0.281, + "step": 47272 + }, + { + "epoch": 17.46, + "learning_rate": 1.2550929917477522e-06, + "loss": 0.1211, + "step": 47273 + }, + { + "epoch": 17.46, + "learning_rate": 1.255056041384407e-06, + "loss": 0.2396, + "step": 47274 + }, + { + "epoch": 17.46, + "learning_rate": 1.2550190910210616e-06, + "loss": 0.1936, + "step": 47275 + }, + { + "epoch": 17.46, + "learning_rate": 1.2549821406577165e-06, + "loss": 0.3871, + "step": 47276 + }, + { + "epoch": 17.46, + "learning_rate": 1.2549451902943713e-06, + "loss": 0.2909, + "step": 47277 + }, + { + "epoch": 17.46, + "learning_rate": 1.254908239931026e-06, + "loss": 0.1204, + "step": 47278 + }, + { + "epoch": 17.46, + "learning_rate": 1.2548712895676808e-06, + "loss": 0.179, + "step": 47279 + }, + { + "epoch": 17.46, + "learning_rate": 1.2548343392043357e-06, + "loss": 0.3093, + "step": 47280 + }, + { + "epoch": 17.46, + "learning_rate": 1.2547973888409903e-06, + "loss": 0.1712, + "step": 47281 + }, + { + "epoch": 17.46, + "learning_rate": 1.254760438477645e-06, + "loss": 0.1498, + "step": 47282 + }, + { + "epoch": 17.46, + "learning_rate": 1.2547234881142998e-06, + "loss": 0.3007, + "step": 47283 + }, + { + "epoch": 17.46, + "learning_rate": 1.2546865377509546e-06, + "loss": 0.2086, + "step": 47284 + }, + { + "epoch": 17.46, + "learning_rate": 1.2546495873876093e-06, + "loss": 0.2778, + "step": 47285 + }, + { + "epoch": 17.46, + "learning_rate": 1.2546126370242641e-06, + "loss": 0.1373, + "step": 47286 + }, + { + "epoch": 17.46, + "learning_rate": 1.254575686660919e-06, + "loss": 0.1881, + "step": 47287 + }, + { + "epoch": 17.46, + "learning_rate": 1.2545387362975736e-06, + "loss": 0.2536, + "step": 47288 + }, + { + "epoch": 17.46, + "learning_rate": 1.2545017859342285e-06, + "loss": 0.1249, + "step": 47289 + }, + { + "epoch": 17.46, + "learning_rate": 1.2544648355708833e-06, + "loss": 0.2942, + "step": 47290 + }, + { + "epoch": 17.46, + "learning_rate": 1.254427885207538e-06, + "loss": 0.2759, + "step": 47291 + }, + { + "epoch": 17.46, + "learning_rate": 1.2543909348441926e-06, + "loss": 0.2154, + "step": 47292 + }, + { + "epoch": 17.46, + "learning_rate": 1.2543539844808474e-06, + "loss": 0.1737, + "step": 47293 + }, + { + "epoch": 17.46, + "learning_rate": 1.254317034117502e-06, + "loss": 0.1908, + "step": 47294 + }, + { + "epoch": 17.46, + "learning_rate": 1.254280083754157e-06, + "loss": 0.1935, + "step": 47295 + }, + { + "epoch": 17.47, + "learning_rate": 1.2542431333908118e-06, + "loss": 0.1668, + "step": 47296 + }, + { + "epoch": 17.47, + "learning_rate": 1.2542061830274664e-06, + "loss": 0.2528, + "step": 47297 + }, + { + "epoch": 17.47, + "learning_rate": 1.2541692326641212e-06, + "loss": 0.2703, + "step": 47298 + }, + { + "epoch": 17.47, + "learning_rate": 1.254132282300776e-06, + "loss": 0.3811, + "step": 47299 + }, + { + "epoch": 17.47, + "learning_rate": 1.2540953319374307e-06, + "loss": 0.2352, + "step": 47300 + }, + { + "epoch": 17.47, + "learning_rate": 1.2540583815740856e-06, + "loss": 0.1858, + "step": 47301 + }, + { + "epoch": 17.47, + "learning_rate": 1.2540214312107402e-06, + "loss": 0.2987, + "step": 47302 + }, + { + "epoch": 17.47, + "learning_rate": 1.253984480847395e-06, + "loss": 0.422, + "step": 47303 + }, + { + "epoch": 17.47, + "learning_rate": 1.2539475304840497e-06, + "loss": 0.2839, + "step": 47304 + }, + { + "epoch": 17.47, + "learning_rate": 1.2539105801207045e-06, + "loss": 0.1566, + "step": 47305 + }, + { + "epoch": 17.47, + "learning_rate": 1.2538736297573594e-06, + "loss": 0.1476, + "step": 47306 + }, + { + "epoch": 17.47, + "learning_rate": 1.253836679394014e-06, + "loss": 0.1957, + "step": 47307 + }, + { + "epoch": 17.47, + "learning_rate": 1.2537997290306689e-06, + "loss": 0.3469, + "step": 47308 + }, + { + "epoch": 17.47, + "learning_rate": 1.2537627786673237e-06, + "loss": 0.2762, + "step": 47309 + }, + { + "epoch": 17.47, + "learning_rate": 1.2537258283039784e-06, + "loss": 0.2479, + "step": 47310 + }, + { + "epoch": 17.47, + "learning_rate": 1.2536888779406332e-06, + "loss": 0.2273, + "step": 47311 + }, + { + "epoch": 17.47, + "learning_rate": 1.2536519275772878e-06, + "loss": 0.1084, + "step": 47312 + }, + { + "epoch": 17.47, + "learning_rate": 1.2536149772139425e-06, + "loss": 0.1809, + "step": 47313 + }, + { + "epoch": 17.47, + "learning_rate": 1.2535780268505973e-06, + "loss": 0.202, + "step": 47314 + }, + { + "epoch": 17.47, + "learning_rate": 1.2535410764872522e-06, + "loss": 0.3071, + "step": 47315 + }, + { + "epoch": 17.47, + "learning_rate": 1.2535041261239068e-06, + "loss": 0.247, + "step": 47316 + }, + { + "epoch": 17.47, + "learning_rate": 1.2534671757605617e-06, + "loss": 0.2178, + "step": 47317 + }, + { + "epoch": 17.47, + "learning_rate": 1.2534302253972165e-06, + "loss": 0.3716, + "step": 47318 + }, + { + "epoch": 17.47, + "learning_rate": 1.2533932750338712e-06, + "loss": 0.246, + "step": 47319 + }, + { + "epoch": 17.47, + "learning_rate": 1.253356324670526e-06, + "loss": 0.2842, + "step": 47320 + }, + { + "epoch": 17.47, + "learning_rate": 1.2533193743071808e-06, + "loss": 0.1766, + "step": 47321 + }, + { + "epoch": 17.47, + "learning_rate": 1.2532824239438355e-06, + "loss": 0.295, + "step": 47322 + }, + { + "epoch": 17.48, + "learning_rate": 1.2532454735804901e-06, + "loss": 0.2779, + "step": 47323 + }, + { + "epoch": 17.48, + "learning_rate": 1.253208523217145e-06, + "loss": 0.2784, + "step": 47324 + }, + { + "epoch": 17.48, + "learning_rate": 1.2531715728537998e-06, + "loss": 0.142, + "step": 47325 + }, + { + "epoch": 17.48, + "learning_rate": 1.2531346224904545e-06, + "loss": 0.1509, + "step": 47326 + }, + { + "epoch": 17.48, + "learning_rate": 1.2530976721271093e-06, + "loss": 0.2264, + "step": 47327 + }, + { + "epoch": 17.48, + "learning_rate": 1.2530607217637641e-06, + "loss": 0.2648, + "step": 47328 + }, + { + "epoch": 17.48, + "learning_rate": 1.2530237714004188e-06, + "loss": 0.2212, + "step": 47329 + }, + { + "epoch": 17.48, + "learning_rate": 1.2529868210370736e-06, + "loss": 0.1082, + "step": 47330 + }, + { + "epoch": 17.48, + "learning_rate": 1.2529498706737285e-06, + "loss": 0.2269, + "step": 47331 + }, + { + "epoch": 17.48, + "learning_rate": 1.2529129203103831e-06, + "loss": 0.2457, + "step": 47332 + }, + { + "epoch": 17.48, + "learning_rate": 1.2528759699470378e-06, + "loss": 0.176, + "step": 47333 + }, + { + "epoch": 17.48, + "learning_rate": 1.2528390195836926e-06, + "loss": 0.1364, + "step": 47334 + }, + { + "epoch": 17.48, + "learning_rate": 1.2528020692203475e-06, + "loss": 0.1595, + "step": 47335 + }, + { + "epoch": 17.48, + "learning_rate": 1.252765118857002e-06, + "loss": 0.1721, + "step": 47336 + }, + { + "epoch": 17.48, + "learning_rate": 1.252728168493657e-06, + "loss": 0.219, + "step": 47337 + }, + { + "epoch": 17.48, + "learning_rate": 1.2526912181303116e-06, + "loss": 0.2104, + "step": 47338 + }, + { + "epoch": 17.48, + "learning_rate": 1.2526542677669664e-06, + "loss": 0.2911, + "step": 47339 + }, + { + "epoch": 17.48, + "learning_rate": 1.2526173174036213e-06, + "loss": 0.2473, + "step": 47340 + }, + { + "epoch": 17.48, + "learning_rate": 1.252580367040276e-06, + "loss": 0.1803, + "step": 47341 + }, + { + "epoch": 17.48, + "learning_rate": 1.2525434166769305e-06, + "loss": 0.2292, + "step": 47342 + }, + { + "epoch": 17.48, + "learning_rate": 1.2525064663135854e-06, + "loss": 0.1834, + "step": 47343 + }, + { + "epoch": 17.48, + "learning_rate": 1.2524695159502402e-06, + "loss": 0.1801, + "step": 47344 + }, + { + "epoch": 17.48, + "learning_rate": 1.2524325655868949e-06, + "loss": 0.2348, + "step": 47345 + }, + { + "epoch": 17.48, + "learning_rate": 1.2523956152235497e-06, + "loss": 0.2211, + "step": 47346 + }, + { + "epoch": 17.48, + "learning_rate": 1.2523586648602046e-06, + "loss": 0.2006, + "step": 47347 + }, + { + "epoch": 17.48, + "learning_rate": 1.2523217144968592e-06, + "loss": 0.1595, + "step": 47348 + }, + { + "epoch": 17.48, + "learning_rate": 1.252284764133514e-06, + "loss": 0.1955, + "step": 47349 + }, + { + "epoch": 17.49, + "learning_rate": 1.252247813770169e-06, + "loss": 0.2174, + "step": 47350 + }, + { + "epoch": 17.49, + "learning_rate": 1.2522108634068235e-06, + "loss": 0.1756, + "step": 47351 + }, + { + "epoch": 17.49, + "learning_rate": 1.2521739130434782e-06, + "loss": 0.2281, + "step": 47352 + }, + { + "epoch": 17.49, + "learning_rate": 1.252136962680133e-06, + "loss": 0.2502, + "step": 47353 + }, + { + "epoch": 17.49, + "learning_rate": 1.2521000123167879e-06, + "loss": 0.2306, + "step": 47354 + }, + { + "epoch": 17.49, + "learning_rate": 1.2520630619534425e-06, + "loss": 0.2393, + "step": 47355 + }, + { + "epoch": 17.49, + "learning_rate": 1.2520261115900974e-06, + "loss": 0.2675, + "step": 47356 + }, + { + "epoch": 17.49, + "learning_rate": 1.2519891612267522e-06, + "loss": 0.1651, + "step": 47357 + }, + { + "epoch": 17.49, + "learning_rate": 1.2519522108634068e-06, + "loss": 0.1978, + "step": 47358 + }, + { + "epoch": 17.49, + "learning_rate": 1.2519152605000617e-06, + "loss": 0.1191, + "step": 47359 + }, + { + "epoch": 17.49, + "learning_rate": 1.2518783101367165e-06, + "loss": 0.2386, + "step": 47360 + }, + { + "epoch": 17.49, + "learning_rate": 1.2518413597733712e-06, + "loss": 0.1155, + "step": 47361 + }, + { + "epoch": 17.49, + "learning_rate": 1.2518044094100258e-06, + "loss": 0.2236, + "step": 47362 + }, + { + "epoch": 17.49, + "learning_rate": 1.2517674590466807e-06, + "loss": 0.2497, + "step": 47363 + }, + { + "epoch": 17.49, + "learning_rate": 1.2517305086833353e-06, + "loss": 0.2853, + "step": 47364 + }, + { + "epoch": 17.49, + "learning_rate": 1.2516935583199901e-06, + "loss": 0.2071, + "step": 47365 + }, + { + "epoch": 17.49, + "learning_rate": 1.251656607956645e-06, + "loss": 0.2413, + "step": 47366 + }, + { + "epoch": 17.49, + "learning_rate": 1.2516196575932996e-06, + "loss": 0.1644, + "step": 47367 + }, + { + "epoch": 17.49, + "learning_rate": 1.2515827072299545e-06, + "loss": 0.2109, + "step": 47368 + }, + { + "epoch": 17.49, + "learning_rate": 1.2515457568666093e-06, + "loss": 0.3229, + "step": 47369 + }, + { + "epoch": 17.49, + "learning_rate": 1.251508806503264e-06, + "loss": 0.2377, + "step": 47370 + }, + { + "epoch": 17.49, + "learning_rate": 1.2514718561399188e-06, + "loss": 0.3183, + "step": 47371 + }, + { + "epoch": 17.49, + "learning_rate": 1.2514349057765734e-06, + "loss": 0.1459, + "step": 47372 + }, + { + "epoch": 17.49, + "learning_rate": 1.2513979554132283e-06, + "loss": 0.2411, + "step": 47373 + }, + { + "epoch": 17.49, + "learning_rate": 1.251361005049883e-06, + "loss": 0.3888, + "step": 47374 + }, + { + "epoch": 17.49, + "learning_rate": 1.2513240546865378e-06, + "loss": 0.1863, + "step": 47375 + }, + { + "epoch": 17.49, + "learning_rate": 1.2512871043231926e-06, + "loss": 0.1556, + "step": 47376 + }, + { + "epoch": 17.5, + "learning_rate": 1.2512501539598473e-06, + "loss": 0.2555, + "step": 47377 + }, + { + "epoch": 17.5, + "learning_rate": 1.2512132035965021e-06, + "loss": 0.3041, + "step": 47378 + }, + { + "epoch": 17.5, + "learning_rate": 1.251176253233157e-06, + "loss": 0.3244, + "step": 47379 + }, + { + "epoch": 17.5, + "learning_rate": 1.2511393028698116e-06, + "loss": 0.1603, + "step": 47380 + }, + { + "epoch": 17.5, + "learning_rate": 1.2511023525064662e-06, + "loss": 0.2138, + "step": 47381 + }, + { + "epoch": 17.5, + "learning_rate": 1.251065402143121e-06, + "loss": 0.2138, + "step": 47382 + }, + { + "epoch": 17.5, + "learning_rate": 1.2510284517797757e-06, + "loss": 0.2589, + "step": 47383 + }, + { + "epoch": 17.5, + "learning_rate": 1.2509915014164306e-06, + "loss": 0.2474, + "step": 47384 + }, + { + "epoch": 17.5, + "learning_rate": 1.2509545510530854e-06, + "loss": 0.2277, + "step": 47385 + }, + { + "epoch": 17.5, + "learning_rate": 1.25091760068974e-06, + "loss": 0.1575, + "step": 47386 + }, + { + "epoch": 17.5, + "learning_rate": 1.250880650326395e-06, + "loss": 0.2385, + "step": 47387 + }, + { + "epoch": 17.5, + "learning_rate": 1.2508436999630497e-06, + "loss": 0.1481, + "step": 47388 + }, + { + "epoch": 17.5, + "learning_rate": 1.2508067495997044e-06, + "loss": 0.1772, + "step": 47389 + }, + { + "epoch": 17.5, + "learning_rate": 1.2507697992363592e-06, + "loss": 0.1981, + "step": 47390 + }, + { + "epoch": 17.5, + "learning_rate": 1.2507328488730139e-06, + "loss": 0.3318, + "step": 47391 + }, + { + "epoch": 17.5, + "learning_rate": 1.2506958985096687e-06, + "loss": 0.2612, + "step": 47392 + }, + { + "epoch": 17.5, + "learning_rate": 1.2506589481463234e-06, + "loss": 0.384, + "step": 47393 + }, + { + "epoch": 17.5, + "learning_rate": 1.2506219977829782e-06, + "loss": 0.2037, + "step": 47394 + }, + { + "epoch": 17.5, + "learning_rate": 1.250585047419633e-06, + "loss": 0.3236, + "step": 47395 + }, + { + "epoch": 17.5, + "learning_rate": 1.2505480970562877e-06, + "loss": 0.2912, + "step": 47396 + }, + { + "epoch": 17.5, + "learning_rate": 1.2505111466929425e-06, + "loss": 0.2508, + "step": 47397 + }, + { + "epoch": 17.5, + "learning_rate": 1.2504741963295974e-06, + "loss": 0.2402, + "step": 47398 + }, + { + "epoch": 17.5, + "learning_rate": 1.250437245966252e-06, + "loss": 0.1624, + "step": 47399 + }, + { + "epoch": 17.5, + "learning_rate": 1.2504002956029069e-06, + "loss": 0.239, + "step": 47400 + }, + { + "epoch": 17.5, + "learning_rate": 1.2503633452395615e-06, + "loss": 0.2283, + "step": 47401 + }, + { + "epoch": 17.5, + "learning_rate": 1.2503263948762164e-06, + "loss": 0.2157, + "step": 47402 + }, + { + "epoch": 17.5, + "learning_rate": 1.250289444512871e-06, + "loss": 0.1586, + "step": 47403 + }, + { + "epoch": 17.51, + "learning_rate": 1.2502524941495258e-06, + "loss": 0.3075, + "step": 47404 + }, + { + "epoch": 17.51, + "learning_rate": 1.2502155437861807e-06, + "loss": 0.3429, + "step": 47405 + }, + { + "epoch": 17.51, + "learning_rate": 1.2501785934228353e-06, + "loss": 0.2128, + "step": 47406 + }, + { + "epoch": 17.51, + "learning_rate": 1.2501416430594902e-06, + "loss": 0.207, + "step": 47407 + }, + { + "epoch": 17.51, + "learning_rate": 1.250104692696145e-06, + "loss": 0.0881, + "step": 47408 + }, + { + "epoch": 17.51, + "learning_rate": 1.2500677423327997e-06, + "loss": 0.2683, + "step": 47409 + }, + { + "epoch": 17.51, + "learning_rate": 1.2500307919694545e-06, + "loss": 0.2775, + "step": 47410 + }, + { + "epoch": 17.51, + "learning_rate": 1.2499938416061091e-06, + "loss": 0.1634, + "step": 47411 + }, + { + "epoch": 17.51, + "learning_rate": 1.2499568912427638e-06, + "loss": 0.2823, + "step": 47412 + }, + { + "epoch": 17.51, + "learning_rate": 1.2499199408794186e-06, + "loss": 0.3887, + "step": 47413 + }, + { + "epoch": 17.51, + "learning_rate": 1.2498829905160735e-06, + "loss": 0.2832, + "step": 47414 + }, + { + "epoch": 17.51, + "learning_rate": 1.2498460401527281e-06, + "loss": 0.161, + "step": 47415 + }, + { + "epoch": 17.51, + "learning_rate": 1.249809089789383e-06, + "loss": 0.154, + "step": 47416 + }, + { + "epoch": 17.51, + "learning_rate": 1.2497721394260378e-06, + "loss": 0.2635, + "step": 47417 + }, + { + "epoch": 17.51, + "learning_rate": 1.2497351890626924e-06, + "loss": 0.2053, + "step": 47418 + }, + { + "epoch": 17.51, + "learning_rate": 1.2496982386993473e-06, + "loss": 0.3828, + "step": 47419 + }, + { + "epoch": 17.51, + "learning_rate": 1.2496612883360021e-06, + "loss": 0.1722, + "step": 47420 + }, + { + "epoch": 17.51, + "learning_rate": 1.2496243379726568e-06, + "loss": 0.2207, + "step": 47421 + }, + { + "epoch": 17.51, + "learning_rate": 1.2495873876093114e-06, + "loss": 0.3101, + "step": 47422 + }, + { + "epoch": 17.51, + "learning_rate": 1.2495504372459663e-06, + "loss": 0.1478, + "step": 47423 + }, + { + "epoch": 17.51, + "learning_rate": 1.249513486882621e-06, + "loss": 0.2128, + "step": 47424 + }, + { + "epoch": 17.51, + "learning_rate": 1.2494765365192757e-06, + "loss": 0.2374, + "step": 47425 + }, + { + "epoch": 17.51, + "learning_rate": 1.2494395861559306e-06, + "loss": 0.1894, + "step": 47426 + }, + { + "epoch": 17.51, + "learning_rate": 1.2494026357925854e-06, + "loss": 0.2523, + "step": 47427 + }, + { + "epoch": 17.51, + "learning_rate": 1.24936568542924e-06, + "loss": 0.1664, + "step": 47428 + }, + { + "epoch": 17.51, + "learning_rate": 1.249328735065895e-06, + "loss": 0.2363, + "step": 47429 + }, + { + "epoch": 17.51, + "learning_rate": 1.2492917847025498e-06, + "loss": 0.1156, + "step": 47430 + }, + { + "epoch": 17.52, + "learning_rate": 1.2492548343392042e-06, + "loss": 0.2455, + "step": 47431 + }, + { + "epoch": 17.52, + "learning_rate": 1.249217883975859e-06, + "loss": 0.3468, + "step": 47432 + }, + { + "epoch": 17.52, + "learning_rate": 1.2491809336125139e-06, + "loss": 0.2209, + "step": 47433 + }, + { + "epoch": 17.52, + "learning_rate": 1.2491439832491685e-06, + "loss": 0.2353, + "step": 47434 + }, + { + "epoch": 17.52, + "learning_rate": 1.2491070328858234e-06, + "loss": 0.336, + "step": 47435 + }, + { + "epoch": 17.52, + "learning_rate": 1.2490700825224782e-06, + "loss": 0.2577, + "step": 47436 + }, + { + "epoch": 17.52, + "learning_rate": 1.2490331321591329e-06, + "loss": 0.346, + "step": 47437 + }, + { + "epoch": 17.52, + "learning_rate": 1.2489961817957877e-06, + "loss": 0.1886, + "step": 47438 + }, + { + "epoch": 17.52, + "learning_rate": 1.2489592314324426e-06, + "loss": 0.2311, + "step": 47439 + }, + { + "epoch": 17.52, + "learning_rate": 1.2489222810690972e-06, + "loss": 0.1948, + "step": 47440 + }, + { + "epoch": 17.52, + "learning_rate": 1.2488853307057518e-06, + "loss": 0.4037, + "step": 47441 + }, + { + "epoch": 17.52, + "learning_rate": 1.2488483803424067e-06, + "loss": 0.1618, + "step": 47442 + }, + { + "epoch": 17.52, + "learning_rate": 1.2488114299790615e-06, + "loss": 0.1057, + "step": 47443 + }, + { + "epoch": 17.52, + "learning_rate": 1.2487744796157162e-06, + "loss": 0.2685, + "step": 47444 + }, + { + "epoch": 17.52, + "learning_rate": 1.248737529252371e-06, + "loss": 0.2871, + "step": 47445 + }, + { + "epoch": 17.52, + "learning_rate": 1.2487005788890259e-06, + "loss": 0.155, + "step": 47446 + }, + { + "epoch": 17.52, + "learning_rate": 1.2486636285256805e-06, + "loss": 0.1503, + "step": 47447 + }, + { + "epoch": 17.52, + "learning_rate": 1.2486266781623353e-06, + "loss": 0.1889, + "step": 47448 + }, + { + "epoch": 17.52, + "learning_rate": 1.2485897277989902e-06, + "loss": 0.3006, + "step": 47449 + }, + { + "epoch": 17.52, + "learning_rate": 1.2485527774356448e-06, + "loss": 0.296, + "step": 47450 + }, + { + "epoch": 17.52, + "learning_rate": 1.2485158270722995e-06, + "loss": 0.307, + "step": 47451 + }, + { + "epoch": 17.52, + "learning_rate": 1.2484788767089543e-06, + "loss": 0.2063, + "step": 47452 + }, + { + "epoch": 17.52, + "learning_rate": 1.248441926345609e-06, + "loss": 0.0887, + "step": 47453 + }, + { + "epoch": 17.52, + "learning_rate": 1.2484049759822638e-06, + "loss": 0.2223, + "step": 47454 + }, + { + "epoch": 17.52, + "learning_rate": 1.2483680256189186e-06, + "loss": 0.203, + "step": 47455 + }, + { + "epoch": 17.52, + "learning_rate": 1.2483310752555733e-06, + "loss": 0.1588, + "step": 47456 + }, + { + "epoch": 17.52, + "learning_rate": 1.2482941248922281e-06, + "loss": 0.1572, + "step": 47457 + }, + { + "epoch": 17.53, + "learning_rate": 1.248257174528883e-06, + "loss": 0.1234, + "step": 47458 + }, + { + "epoch": 17.53, + "learning_rate": 1.2482202241655376e-06, + "loss": 0.2592, + "step": 47459 + }, + { + "epoch": 17.53, + "learning_rate": 1.2481832738021925e-06, + "loss": 0.1938, + "step": 47460 + }, + { + "epoch": 17.53, + "learning_rate": 1.248146323438847e-06, + "loss": 0.137, + "step": 47461 + }, + { + "epoch": 17.53, + "learning_rate": 1.248109373075502e-06, + "loss": 0.1681, + "step": 47462 + }, + { + "epoch": 17.53, + "learning_rate": 1.2480724227121566e-06, + "loss": 0.2136, + "step": 47463 + }, + { + "epoch": 17.53, + "learning_rate": 1.2480354723488114e-06, + "loss": 0.1076, + "step": 47464 + }, + { + "epoch": 17.53, + "learning_rate": 1.2479985219854663e-06, + "loss": 0.2407, + "step": 47465 + }, + { + "epoch": 17.53, + "learning_rate": 1.247961571622121e-06, + "loss": 0.3532, + "step": 47466 + }, + { + "epoch": 17.53, + "learning_rate": 1.2479246212587758e-06, + "loss": 0.2835, + "step": 47467 + }, + { + "epoch": 17.53, + "learning_rate": 1.2478876708954306e-06, + "loss": 0.1786, + "step": 47468 + }, + { + "epoch": 17.53, + "learning_rate": 1.2478507205320853e-06, + "loss": 0.2362, + "step": 47469 + }, + { + "epoch": 17.53, + "learning_rate": 1.24781377016874e-06, + "loss": 0.2297, + "step": 47470 + }, + { + "epoch": 17.53, + "learning_rate": 1.2477768198053947e-06, + "loss": 0.198, + "step": 47471 + }, + { + "epoch": 17.53, + "learning_rate": 1.2477398694420496e-06, + "loss": 0.3477, + "step": 47472 + }, + { + "epoch": 17.53, + "learning_rate": 1.2477029190787042e-06, + "loss": 0.2477, + "step": 47473 + }, + { + "epoch": 17.53, + "learning_rate": 1.247665968715359e-06, + "loss": 0.12, + "step": 47474 + }, + { + "epoch": 17.53, + "learning_rate": 1.247629018352014e-06, + "loss": 0.3724, + "step": 47475 + }, + { + "epoch": 17.53, + "learning_rate": 1.2475920679886686e-06, + "loss": 0.1584, + "step": 47476 + }, + { + "epoch": 17.53, + "learning_rate": 1.2475551176253234e-06, + "loss": 0.2361, + "step": 47477 + }, + { + "epoch": 17.53, + "learning_rate": 1.2475181672619782e-06, + "loss": 0.1521, + "step": 47478 + }, + { + "epoch": 17.53, + "learning_rate": 1.2474812168986329e-06, + "loss": 0.2629, + "step": 47479 + }, + { + "epoch": 17.53, + "learning_rate": 1.2474442665352877e-06, + "loss": 0.3059, + "step": 47480 + }, + { + "epoch": 17.53, + "learning_rate": 1.2474073161719424e-06, + "loss": 0.2142, + "step": 47481 + }, + { + "epoch": 17.53, + "learning_rate": 1.247370365808597e-06, + "loss": 0.1374, + "step": 47482 + }, + { + "epoch": 17.53, + "learning_rate": 1.2473334154452519e-06, + "loss": 0.387, + "step": 47483 + }, + { + "epoch": 17.53, + "learning_rate": 1.2472964650819067e-06, + "loss": 0.2687, + "step": 47484 + }, + { + "epoch": 17.54, + "learning_rate": 1.2472595147185613e-06, + "loss": 0.2018, + "step": 47485 + }, + { + "epoch": 17.54, + "learning_rate": 1.2472225643552162e-06, + "loss": 0.1656, + "step": 47486 + }, + { + "epoch": 17.54, + "learning_rate": 1.247185613991871e-06, + "loss": 0.3081, + "step": 47487 + }, + { + "epoch": 17.54, + "learning_rate": 1.2471486636285257e-06, + "loss": 0.4327, + "step": 47488 + }, + { + "epoch": 17.54, + "learning_rate": 1.2471117132651805e-06, + "loss": 0.3242, + "step": 47489 + }, + { + "epoch": 17.54, + "learning_rate": 1.2470747629018354e-06, + "loss": 0.1757, + "step": 47490 + }, + { + "epoch": 17.54, + "learning_rate": 1.24703781253849e-06, + "loss": 0.132, + "step": 47491 + }, + { + "epoch": 17.54, + "learning_rate": 1.2470008621751446e-06, + "loss": 0.1856, + "step": 47492 + }, + { + "epoch": 17.54, + "learning_rate": 1.2469639118117995e-06, + "loss": 0.2871, + "step": 47493 + }, + { + "epoch": 17.54, + "learning_rate": 1.2469269614484543e-06, + "loss": 0.2846, + "step": 47494 + }, + { + "epoch": 17.54, + "learning_rate": 1.246890011085109e-06, + "loss": 0.2246, + "step": 47495 + }, + { + "epoch": 17.54, + "learning_rate": 1.2468530607217638e-06, + "loss": 0.1696, + "step": 47496 + }, + { + "epoch": 17.54, + "learning_rate": 1.2468161103584187e-06, + "loss": 0.2284, + "step": 47497 + }, + { + "epoch": 17.54, + "learning_rate": 1.2467791599950733e-06, + "loss": 0.2066, + "step": 47498 + }, + { + "epoch": 17.54, + "learning_rate": 1.2467422096317282e-06, + "loss": 0.1469, + "step": 47499 + }, + { + "epoch": 17.54, + "learning_rate": 1.246705259268383e-06, + "loss": 0.2097, + "step": 47500 + }, + { + "epoch": 17.54, + "learning_rate": 1.2466683089050374e-06, + "loss": 0.2768, + "step": 47501 + }, + { + "epoch": 17.54, + "learning_rate": 1.2466313585416923e-06, + "loss": 0.1848, + "step": 47502 + }, + { + "epoch": 17.54, + "learning_rate": 1.2465944081783471e-06, + "loss": 0.1107, + "step": 47503 + }, + { + "epoch": 17.54, + "learning_rate": 1.2465574578150018e-06, + "loss": 0.1589, + "step": 47504 + }, + { + "epoch": 17.54, + "learning_rate": 1.2465205074516566e-06, + "loss": 0.1894, + "step": 47505 + }, + { + "epoch": 17.54, + "learning_rate": 1.2464835570883115e-06, + "loss": 0.3119, + "step": 47506 + }, + { + "epoch": 17.54, + "learning_rate": 1.246446606724966e-06, + "loss": 0.2312, + "step": 47507 + }, + { + "epoch": 17.54, + "learning_rate": 1.246409656361621e-06, + "loss": 0.3167, + "step": 47508 + }, + { + "epoch": 17.54, + "learning_rate": 1.2463727059982758e-06, + "loss": 0.2106, + "step": 47509 + }, + { + "epoch": 17.54, + "learning_rate": 1.2463357556349304e-06, + "loss": 0.2161, + "step": 47510 + }, + { + "epoch": 17.54, + "learning_rate": 1.246298805271585e-06, + "loss": 0.2854, + "step": 47511 + }, + { + "epoch": 17.55, + "learning_rate": 1.24626185490824e-06, + "loss": 0.2181, + "step": 47512 + }, + { + "epoch": 17.55, + "learning_rate": 1.2462249045448948e-06, + "loss": 0.1347, + "step": 47513 + }, + { + "epoch": 17.55, + "learning_rate": 1.2461879541815494e-06, + "loss": 0.238, + "step": 47514 + }, + { + "epoch": 17.55, + "learning_rate": 1.2461510038182042e-06, + "loss": 0.2575, + "step": 47515 + }, + { + "epoch": 17.55, + "learning_rate": 1.246114053454859e-06, + "loss": 0.3765, + "step": 47516 + }, + { + "epoch": 17.55, + "learning_rate": 1.2460771030915137e-06, + "loss": 0.2305, + "step": 47517 + }, + { + "epoch": 17.55, + "learning_rate": 1.2460401527281686e-06, + "loss": 0.1697, + "step": 47518 + }, + { + "epoch": 17.55, + "learning_rate": 1.2460032023648234e-06, + "loss": 0.2774, + "step": 47519 + }, + { + "epoch": 17.55, + "learning_rate": 1.245966252001478e-06, + "loss": 0.228, + "step": 47520 + }, + { + "epoch": 17.55, + "learning_rate": 1.2459293016381327e-06, + "loss": 0.2473, + "step": 47521 + }, + { + "epoch": 17.55, + "learning_rate": 1.2458923512747875e-06, + "loss": 0.2831, + "step": 47522 + }, + { + "epoch": 17.55, + "learning_rate": 1.2458554009114422e-06, + "loss": 0.2345, + "step": 47523 + }, + { + "epoch": 17.55, + "learning_rate": 1.245818450548097e-06, + "loss": 0.1929, + "step": 47524 + }, + { + "epoch": 17.55, + "learning_rate": 1.2457815001847519e-06, + "loss": 0.2603, + "step": 47525 + }, + { + "epoch": 17.55, + "learning_rate": 1.2457445498214065e-06, + "loss": 0.1845, + "step": 47526 + }, + { + "epoch": 17.55, + "learning_rate": 1.2457075994580614e-06, + "loss": 0.1141, + "step": 47527 + }, + { + "epoch": 17.55, + "learning_rate": 1.2456706490947162e-06, + "loss": 0.2316, + "step": 47528 + }, + { + "epoch": 17.55, + "learning_rate": 1.2456336987313708e-06, + "loss": 0.2615, + "step": 47529 + }, + { + "epoch": 17.55, + "learning_rate": 1.2455967483680257e-06, + "loss": 0.1492, + "step": 47530 + }, + { + "epoch": 17.55, + "learning_rate": 1.2455597980046803e-06, + "loss": 0.1903, + "step": 47531 + }, + { + "epoch": 17.55, + "learning_rate": 1.2455228476413352e-06, + "loss": 0.1596, + "step": 47532 + }, + { + "epoch": 17.55, + "learning_rate": 1.2454858972779898e-06, + "loss": 0.1912, + "step": 47533 + }, + { + "epoch": 17.55, + "learning_rate": 1.2454489469146447e-06, + "loss": 0.2375, + "step": 47534 + }, + { + "epoch": 17.55, + "learning_rate": 1.2454119965512995e-06, + "loss": 0.234, + "step": 47535 + }, + { + "epoch": 17.55, + "learning_rate": 1.2453750461879542e-06, + "loss": 0.182, + "step": 47536 + }, + { + "epoch": 17.55, + "learning_rate": 1.245338095824609e-06, + "loss": 0.1994, + "step": 47537 + }, + { + "epoch": 17.55, + "learning_rate": 1.2453011454612638e-06, + "loss": 0.2932, + "step": 47538 + }, + { + "epoch": 17.56, + "learning_rate": 1.2452641950979185e-06, + "loss": 0.2393, + "step": 47539 + }, + { + "epoch": 17.56, + "learning_rate": 1.2452272447345733e-06, + "loss": 0.1919, + "step": 47540 + }, + { + "epoch": 17.56, + "learning_rate": 1.245190294371228e-06, + "loss": 0.1698, + "step": 47541 + }, + { + "epoch": 17.56, + "learning_rate": 1.2451533440078828e-06, + "loss": 0.2875, + "step": 47542 + }, + { + "epoch": 17.56, + "learning_rate": 1.2451163936445375e-06, + "loss": 0.2383, + "step": 47543 + }, + { + "epoch": 17.56, + "learning_rate": 1.2450794432811923e-06, + "loss": 0.2703, + "step": 47544 + }, + { + "epoch": 17.56, + "learning_rate": 1.2450424929178471e-06, + "loss": 0.1586, + "step": 47545 + }, + { + "epoch": 17.56, + "learning_rate": 1.2450055425545018e-06, + "loss": 0.1404, + "step": 47546 + }, + { + "epoch": 17.56, + "learning_rate": 1.2449685921911566e-06, + "loss": 0.2629, + "step": 47547 + }, + { + "epoch": 17.56, + "learning_rate": 1.2449316418278115e-06, + "loss": 0.1518, + "step": 47548 + }, + { + "epoch": 17.56, + "learning_rate": 1.2448946914644661e-06, + "loss": 0.2276, + "step": 47549 + }, + { + "epoch": 17.56, + "learning_rate": 1.244857741101121e-06, + "loss": 0.1753, + "step": 47550 + }, + { + "epoch": 17.56, + "learning_rate": 1.2448207907377756e-06, + "loss": 0.2416, + "step": 47551 + }, + { + "epoch": 17.56, + "learning_rate": 1.2447838403744302e-06, + "loss": 0.1503, + "step": 47552 + }, + { + "epoch": 17.56, + "learning_rate": 1.244746890011085e-06, + "loss": 0.1583, + "step": 47553 + }, + { + "epoch": 17.56, + "learning_rate": 1.24470993964774e-06, + "loss": 0.098, + "step": 47554 + }, + { + "epoch": 17.56, + "learning_rate": 1.2446729892843946e-06, + "loss": 0.2666, + "step": 47555 + }, + { + "epoch": 17.56, + "learning_rate": 1.2446360389210494e-06, + "loss": 0.1312, + "step": 47556 + }, + { + "epoch": 17.56, + "learning_rate": 1.2445990885577043e-06, + "loss": 0.4229, + "step": 47557 + }, + { + "epoch": 17.56, + "learning_rate": 1.244562138194359e-06, + "loss": 0.1935, + "step": 47558 + }, + { + "epoch": 17.56, + "learning_rate": 1.2445251878310138e-06, + "loss": 0.2871, + "step": 47559 + }, + { + "epoch": 17.56, + "learning_rate": 1.2444882374676686e-06, + "loss": 0.2118, + "step": 47560 + }, + { + "epoch": 17.56, + "learning_rate": 1.2444512871043232e-06, + "loss": 0.2731, + "step": 47561 + }, + { + "epoch": 17.56, + "learning_rate": 1.2444143367409779e-06, + "loss": 0.1769, + "step": 47562 + }, + { + "epoch": 17.56, + "learning_rate": 1.2443773863776327e-06, + "loss": 0.1908, + "step": 47563 + }, + { + "epoch": 17.56, + "learning_rate": 1.2443404360142876e-06, + "loss": 0.1636, + "step": 47564 + }, + { + "epoch": 17.56, + "learning_rate": 1.2443034856509422e-06, + "loss": 0.2125, + "step": 47565 + }, + { + "epoch": 17.56, + "learning_rate": 1.244266535287597e-06, + "loss": 0.2448, + "step": 47566 + }, + { + "epoch": 17.57, + "learning_rate": 1.244229584924252e-06, + "loss": 0.3224, + "step": 47567 + }, + { + "epoch": 17.57, + "learning_rate": 1.2441926345609065e-06, + "loss": 0.3378, + "step": 47568 + }, + { + "epoch": 17.57, + "learning_rate": 1.2441556841975614e-06, + "loss": 0.1987, + "step": 47569 + }, + { + "epoch": 17.57, + "learning_rate": 1.2441187338342162e-06, + "loss": 0.2972, + "step": 47570 + }, + { + "epoch": 17.57, + "learning_rate": 1.2440817834708707e-06, + "loss": 0.3789, + "step": 47571 + }, + { + "epoch": 17.57, + "learning_rate": 1.2440448331075255e-06, + "loss": 0.2788, + "step": 47572 + }, + { + "epoch": 17.57, + "learning_rate": 1.2440078827441804e-06, + "loss": 0.3367, + "step": 47573 + }, + { + "epoch": 17.57, + "learning_rate": 1.243970932380835e-06, + "loss": 0.2502, + "step": 47574 + }, + { + "epoch": 17.57, + "learning_rate": 1.2439339820174898e-06, + "loss": 0.1826, + "step": 47575 + }, + { + "epoch": 17.57, + "learning_rate": 1.2438970316541447e-06, + "loss": 0.2089, + "step": 47576 + }, + { + "epoch": 17.57, + "learning_rate": 1.2438600812907993e-06, + "loss": 0.2574, + "step": 47577 + }, + { + "epoch": 17.57, + "learning_rate": 1.2438231309274542e-06, + "loss": 0.1479, + "step": 47578 + }, + { + "epoch": 17.57, + "learning_rate": 1.243786180564109e-06, + "loss": 0.2803, + "step": 47579 + }, + { + "epoch": 17.57, + "learning_rate": 1.2437492302007637e-06, + "loss": 0.2346, + "step": 47580 + }, + { + "epoch": 17.57, + "learning_rate": 1.2437122798374183e-06, + "loss": 0.2028, + "step": 47581 + }, + { + "epoch": 17.57, + "learning_rate": 1.2436753294740731e-06, + "loss": 0.2247, + "step": 47582 + }, + { + "epoch": 17.57, + "learning_rate": 1.243638379110728e-06, + "loss": 0.2835, + "step": 47583 + }, + { + "epoch": 17.57, + "learning_rate": 1.2436014287473826e-06, + "loss": 0.2498, + "step": 47584 + }, + { + "epoch": 17.57, + "learning_rate": 1.2435644783840375e-06, + "loss": 0.3646, + "step": 47585 + }, + { + "epoch": 17.57, + "learning_rate": 1.2435275280206923e-06, + "loss": 0.1794, + "step": 47586 + }, + { + "epoch": 17.57, + "learning_rate": 1.243490577657347e-06, + "loss": 0.2085, + "step": 47587 + }, + { + "epoch": 17.57, + "learning_rate": 1.2434536272940018e-06, + "loss": 0.2052, + "step": 47588 + }, + { + "epoch": 17.57, + "learning_rate": 1.2434166769306567e-06, + "loss": 0.4079, + "step": 47589 + }, + { + "epoch": 17.57, + "learning_rate": 1.2433797265673113e-06, + "loss": 0.2889, + "step": 47590 + }, + { + "epoch": 17.57, + "learning_rate": 1.243342776203966e-06, + "loss": 0.2146, + "step": 47591 + }, + { + "epoch": 17.57, + "learning_rate": 1.2433058258406208e-06, + "loss": 0.2329, + "step": 47592 + }, + { + "epoch": 17.57, + "learning_rate": 1.2432688754772756e-06, + "loss": 0.2417, + "step": 47593 + }, + { + "epoch": 17.58, + "learning_rate": 1.2432319251139303e-06, + "loss": 0.1845, + "step": 47594 + }, + { + "epoch": 17.58, + "learning_rate": 1.2431949747505851e-06, + "loss": 0.3745, + "step": 47595 + }, + { + "epoch": 17.58, + "learning_rate": 1.24315802438724e-06, + "loss": 0.32, + "step": 47596 + }, + { + "epoch": 17.58, + "learning_rate": 1.2431210740238946e-06, + "loss": 0.2684, + "step": 47597 + }, + { + "epoch": 17.58, + "learning_rate": 1.2430841236605494e-06, + "loss": 0.2872, + "step": 47598 + }, + { + "epoch": 17.58, + "learning_rate": 1.2430471732972043e-06, + "loss": 0.2611, + "step": 47599 + }, + { + "epoch": 17.58, + "learning_rate": 1.243010222933859e-06, + "loss": 0.2144, + "step": 47600 + }, + { + "epoch": 17.58, + "learning_rate": 1.2429732725705136e-06, + "loss": 0.2551, + "step": 47601 + }, + { + "epoch": 17.58, + "learning_rate": 1.2429363222071684e-06, + "loss": 0.1794, + "step": 47602 + }, + { + "epoch": 17.58, + "learning_rate": 1.242899371843823e-06, + "loss": 0.1813, + "step": 47603 + }, + { + "epoch": 17.58, + "learning_rate": 1.242862421480478e-06, + "loss": 0.2113, + "step": 47604 + }, + { + "epoch": 17.58, + "learning_rate": 1.2428254711171327e-06, + "loss": 0.2397, + "step": 47605 + }, + { + "epoch": 17.58, + "learning_rate": 1.2427885207537874e-06, + "loss": 0.257, + "step": 47606 + }, + { + "epoch": 17.58, + "learning_rate": 1.2427515703904422e-06, + "loss": 0.4206, + "step": 47607 + }, + { + "epoch": 17.58, + "learning_rate": 1.242714620027097e-06, + "loss": 0.2941, + "step": 47608 + }, + { + "epoch": 17.58, + "learning_rate": 1.2426776696637517e-06, + "loss": 0.2487, + "step": 47609 + }, + { + "epoch": 17.58, + "learning_rate": 1.2426407193004064e-06, + "loss": 0.3552, + "step": 47610 + }, + { + "epoch": 17.58, + "learning_rate": 1.2426037689370612e-06, + "loss": 0.1907, + "step": 47611 + }, + { + "epoch": 17.58, + "learning_rate": 1.242566818573716e-06, + "loss": 0.2616, + "step": 47612 + }, + { + "epoch": 17.58, + "learning_rate": 1.2425298682103707e-06, + "loss": 0.2154, + "step": 47613 + }, + { + "epoch": 17.58, + "learning_rate": 1.2424929178470255e-06, + "loss": 0.1532, + "step": 47614 + }, + { + "epoch": 17.58, + "learning_rate": 1.2424559674836804e-06, + "loss": 0.2951, + "step": 47615 + }, + { + "epoch": 17.58, + "learning_rate": 1.242419017120335e-06, + "loss": 0.2865, + "step": 47616 + }, + { + "epoch": 17.58, + "learning_rate": 1.2423820667569899e-06, + "loss": 0.1692, + "step": 47617 + }, + { + "epoch": 17.58, + "learning_rate": 1.2423451163936447e-06, + "loss": 0.1513, + "step": 47618 + }, + { + "epoch": 17.58, + "learning_rate": 1.2423081660302994e-06, + "loss": 0.2136, + "step": 47619 + }, + { + "epoch": 17.58, + "learning_rate": 1.242271215666954e-06, + "loss": 0.1937, + "step": 47620 + }, + { + "epoch": 17.59, + "learning_rate": 1.2422342653036088e-06, + "loss": 0.2486, + "step": 47621 + }, + { + "epoch": 17.59, + "learning_rate": 1.2421973149402635e-06, + "loss": 0.3595, + "step": 47622 + }, + { + "epoch": 17.59, + "learning_rate": 1.2421603645769183e-06, + "loss": 0.2178, + "step": 47623 + }, + { + "epoch": 17.59, + "learning_rate": 1.2421234142135732e-06, + "loss": 0.4496, + "step": 47624 + }, + { + "epoch": 17.59, + "learning_rate": 1.2420864638502278e-06, + "loss": 0.2252, + "step": 47625 + }, + { + "epoch": 17.59, + "learning_rate": 1.2420495134868827e-06, + "loss": 0.3212, + "step": 47626 + }, + { + "epoch": 17.59, + "learning_rate": 1.2420125631235375e-06, + "loss": 0.1427, + "step": 47627 + }, + { + "epoch": 17.59, + "learning_rate": 1.2419756127601921e-06, + "loss": 0.2932, + "step": 47628 + }, + { + "epoch": 17.59, + "learning_rate": 1.241938662396847e-06, + "loss": 0.2325, + "step": 47629 + }, + { + "epoch": 17.59, + "learning_rate": 1.2419017120335016e-06, + "loss": 0.2599, + "step": 47630 + }, + { + "epoch": 17.59, + "learning_rate": 1.2418647616701565e-06, + "loss": 0.286, + "step": 47631 + }, + { + "epoch": 17.59, + "learning_rate": 1.2418278113068111e-06, + "loss": 0.1456, + "step": 47632 + }, + { + "epoch": 17.59, + "learning_rate": 1.241790860943466e-06, + "loss": 0.3279, + "step": 47633 + }, + { + "epoch": 17.59, + "learning_rate": 1.2417539105801208e-06, + "loss": 0.2207, + "step": 47634 + }, + { + "epoch": 17.59, + "learning_rate": 1.2417169602167754e-06, + "loss": 0.143, + "step": 47635 + }, + { + "epoch": 17.59, + "learning_rate": 1.2416800098534303e-06, + "loss": 0.242, + "step": 47636 + }, + { + "epoch": 17.59, + "learning_rate": 1.2416430594900851e-06, + "loss": 0.3104, + "step": 47637 + }, + { + "epoch": 17.59, + "learning_rate": 1.2416061091267398e-06, + "loss": 0.2059, + "step": 47638 + }, + { + "epoch": 17.59, + "learning_rate": 1.2415691587633946e-06, + "loss": 0.2523, + "step": 47639 + }, + { + "epoch": 17.59, + "learning_rate": 1.2415322084000493e-06, + "loss": 0.2165, + "step": 47640 + }, + { + "epoch": 17.59, + "learning_rate": 1.2414952580367039e-06, + "loss": 0.2562, + "step": 47641 + }, + { + "epoch": 17.59, + "learning_rate": 1.2414583076733587e-06, + "loss": 0.1677, + "step": 47642 + }, + { + "epoch": 17.59, + "learning_rate": 1.2414213573100136e-06, + "loss": 0.2015, + "step": 47643 + }, + { + "epoch": 17.59, + "learning_rate": 1.2413844069466682e-06, + "loss": 0.3078, + "step": 47644 + }, + { + "epoch": 17.59, + "learning_rate": 1.241347456583323e-06, + "loss": 0.1856, + "step": 47645 + }, + { + "epoch": 17.59, + "learning_rate": 1.241310506219978e-06, + "loss": 0.2587, + "step": 47646 + }, + { + "epoch": 17.59, + "learning_rate": 1.2412735558566326e-06, + "loss": 0.1951, + "step": 47647 + }, + { + "epoch": 17.6, + "learning_rate": 1.2412366054932874e-06, + "loss": 0.2515, + "step": 47648 + }, + { + "epoch": 17.6, + "learning_rate": 1.2411996551299423e-06, + "loss": 0.1985, + "step": 47649 + }, + { + "epoch": 17.6, + "learning_rate": 1.2411627047665969e-06, + "loss": 0.2439, + "step": 47650 + }, + { + "epoch": 17.6, + "learning_rate": 1.2411257544032515e-06, + "loss": 0.2987, + "step": 47651 + }, + { + "epoch": 17.6, + "learning_rate": 1.2410888040399064e-06, + "loss": 0.2551, + "step": 47652 + }, + { + "epoch": 17.6, + "learning_rate": 1.2410518536765612e-06, + "loss": 0.1382, + "step": 47653 + }, + { + "epoch": 17.6, + "learning_rate": 1.2410149033132159e-06, + "loss": 0.2712, + "step": 47654 + }, + { + "epoch": 17.6, + "learning_rate": 1.2409779529498707e-06, + "loss": 0.1799, + "step": 47655 + }, + { + "epoch": 17.6, + "learning_rate": 1.2409410025865256e-06, + "loss": 0.3178, + "step": 47656 + }, + { + "epoch": 17.6, + "learning_rate": 1.2409040522231802e-06, + "loss": 0.1696, + "step": 47657 + }, + { + "epoch": 17.6, + "learning_rate": 1.240867101859835e-06, + "loss": 0.2785, + "step": 47658 + }, + { + "epoch": 17.6, + "learning_rate": 1.2408301514964899e-06, + "loss": 0.2633, + "step": 47659 + }, + { + "epoch": 17.6, + "learning_rate": 1.2407932011331445e-06, + "loss": 0.2574, + "step": 47660 + }, + { + "epoch": 17.6, + "learning_rate": 1.2407562507697992e-06, + "loss": 0.137, + "step": 47661 + }, + { + "epoch": 17.6, + "learning_rate": 1.240719300406454e-06, + "loss": 0.1288, + "step": 47662 + }, + { + "epoch": 17.6, + "learning_rate": 1.2406823500431089e-06, + "loss": 0.2115, + "step": 47663 + }, + { + "epoch": 17.6, + "learning_rate": 1.2406453996797635e-06, + "loss": 0.3907, + "step": 47664 + }, + { + "epoch": 17.6, + "learning_rate": 1.2406084493164183e-06, + "loss": 0.1523, + "step": 47665 + }, + { + "epoch": 17.6, + "learning_rate": 1.2405714989530732e-06, + "loss": 0.2424, + "step": 47666 + }, + { + "epoch": 17.6, + "learning_rate": 1.2405345485897278e-06, + "loss": 0.1684, + "step": 47667 + }, + { + "epoch": 17.6, + "learning_rate": 1.2404975982263827e-06, + "loss": 0.1401, + "step": 47668 + }, + { + "epoch": 17.6, + "learning_rate": 1.2404606478630375e-06, + "loss": 0.1304, + "step": 47669 + }, + { + "epoch": 17.6, + "learning_rate": 1.240423697499692e-06, + "loss": 0.2112, + "step": 47670 + }, + { + "epoch": 17.6, + "learning_rate": 1.2403867471363468e-06, + "loss": 0.1756, + "step": 47671 + }, + { + "epoch": 17.6, + "learning_rate": 1.2403497967730016e-06, + "loss": 0.3608, + "step": 47672 + }, + { + "epoch": 17.6, + "learning_rate": 1.2403128464096563e-06, + "loss": 0.1862, + "step": 47673 + }, + { + "epoch": 17.6, + "learning_rate": 1.2402758960463111e-06, + "loss": 0.294, + "step": 47674 + }, + { + "epoch": 17.61, + "learning_rate": 1.240238945682966e-06, + "loss": 0.222, + "step": 47675 + }, + { + "epoch": 17.61, + "learning_rate": 1.2402019953196206e-06, + "loss": 0.3196, + "step": 47676 + }, + { + "epoch": 17.61, + "learning_rate": 1.2401650449562755e-06, + "loss": 0.297, + "step": 47677 + }, + { + "epoch": 17.61, + "learning_rate": 1.2401280945929303e-06, + "loss": 0.2538, + "step": 47678 + }, + { + "epoch": 17.61, + "learning_rate": 1.240091144229585e-06, + "loss": 0.1954, + "step": 47679 + }, + { + "epoch": 17.61, + "learning_rate": 1.2400541938662396e-06, + "loss": 0.2392, + "step": 47680 + }, + { + "epoch": 17.61, + "learning_rate": 1.2400172435028944e-06, + "loss": 0.2702, + "step": 47681 + }, + { + "epoch": 17.61, + "learning_rate": 1.2399802931395493e-06, + "loss": 0.318, + "step": 47682 + }, + { + "epoch": 17.61, + "learning_rate": 1.239943342776204e-06, + "loss": 0.2973, + "step": 47683 + }, + { + "epoch": 17.61, + "learning_rate": 1.2399063924128588e-06, + "loss": 0.1662, + "step": 47684 + }, + { + "epoch": 17.61, + "learning_rate": 1.2398694420495136e-06, + "loss": 0.1935, + "step": 47685 + }, + { + "epoch": 17.61, + "learning_rate": 1.2398324916861683e-06, + "loss": 0.2144, + "step": 47686 + }, + { + "epoch": 17.61, + "learning_rate": 1.239795541322823e-06, + "loss": 0.2636, + "step": 47687 + }, + { + "epoch": 17.61, + "learning_rate": 1.239758590959478e-06, + "loss": 0.2353, + "step": 47688 + }, + { + "epoch": 17.61, + "learning_rate": 1.2397216405961326e-06, + "loss": 0.1384, + "step": 47689 + }, + { + "epoch": 17.61, + "learning_rate": 1.2396846902327872e-06, + "loss": 0.1654, + "step": 47690 + }, + { + "epoch": 17.61, + "learning_rate": 1.239647739869442e-06, + "loss": 0.3072, + "step": 47691 + }, + { + "epoch": 17.61, + "learning_rate": 1.2396107895060967e-06, + "loss": 0.2921, + "step": 47692 + }, + { + "epoch": 17.61, + "learning_rate": 1.2395738391427516e-06, + "loss": 0.2062, + "step": 47693 + }, + { + "epoch": 17.61, + "learning_rate": 1.2395368887794064e-06, + "loss": 0.2316, + "step": 47694 + }, + { + "epoch": 17.61, + "learning_rate": 1.239499938416061e-06, + "loss": 0.2214, + "step": 47695 + }, + { + "epoch": 17.61, + "learning_rate": 1.2394629880527159e-06, + "loss": 0.4204, + "step": 47696 + }, + { + "epoch": 17.61, + "learning_rate": 1.2394260376893707e-06, + "loss": 0.2012, + "step": 47697 + }, + { + "epoch": 17.61, + "learning_rate": 1.2393890873260254e-06, + "loss": 0.3102, + "step": 47698 + }, + { + "epoch": 17.61, + "learning_rate": 1.2393521369626802e-06, + "loss": 0.1882, + "step": 47699 + }, + { + "epoch": 17.61, + "learning_rate": 1.2393151865993349e-06, + "loss": 0.2512, + "step": 47700 + }, + { + "epoch": 17.61, + "learning_rate": 1.2392782362359897e-06, + "loss": 0.2158, + "step": 47701 + }, + { + "epoch": 17.62, + "learning_rate": 1.2392412858726443e-06, + "loss": 0.1554, + "step": 47702 + }, + { + "epoch": 17.62, + "learning_rate": 1.2392043355092992e-06, + "loss": 0.173, + "step": 47703 + }, + { + "epoch": 17.62, + "learning_rate": 1.239167385145954e-06, + "loss": 0.1174, + "step": 47704 + }, + { + "epoch": 17.62, + "learning_rate": 1.2391304347826087e-06, + "loss": 0.2205, + "step": 47705 + }, + { + "epoch": 17.62, + "learning_rate": 1.2390934844192635e-06, + "loss": 0.1457, + "step": 47706 + }, + { + "epoch": 17.62, + "learning_rate": 1.2390565340559184e-06, + "loss": 0.2101, + "step": 47707 + }, + { + "epoch": 17.62, + "learning_rate": 1.239019583692573e-06, + "loss": 0.1644, + "step": 47708 + }, + { + "epoch": 17.62, + "learning_rate": 1.2389826333292279e-06, + "loss": 0.2066, + "step": 47709 + }, + { + "epoch": 17.62, + "learning_rate": 1.2389456829658825e-06, + "loss": 0.2023, + "step": 47710 + }, + { + "epoch": 17.62, + "learning_rate": 1.2389087326025371e-06, + "loss": 0.2786, + "step": 47711 + }, + { + "epoch": 17.62, + "learning_rate": 1.238871782239192e-06, + "loss": 0.172, + "step": 47712 + }, + { + "epoch": 17.62, + "learning_rate": 1.2388348318758468e-06, + "loss": 0.2596, + "step": 47713 + }, + { + "epoch": 17.62, + "learning_rate": 1.2387978815125015e-06, + "loss": 0.1685, + "step": 47714 + }, + { + "epoch": 17.62, + "learning_rate": 1.2387609311491563e-06, + "loss": 0.2228, + "step": 47715 + }, + { + "epoch": 17.62, + "learning_rate": 1.2387239807858112e-06, + "loss": 0.1204, + "step": 47716 + }, + { + "epoch": 17.62, + "learning_rate": 1.2386870304224658e-06, + "loss": 0.2159, + "step": 47717 + }, + { + "epoch": 17.62, + "learning_rate": 1.2386500800591206e-06, + "loss": 0.2166, + "step": 47718 + }, + { + "epoch": 17.62, + "learning_rate": 1.2386131296957755e-06, + "loss": 0.2823, + "step": 47719 + }, + { + "epoch": 17.62, + "learning_rate": 1.2385761793324301e-06, + "loss": 0.2546, + "step": 47720 + }, + { + "epoch": 17.62, + "learning_rate": 1.2385392289690848e-06, + "loss": 0.1728, + "step": 47721 + }, + { + "epoch": 17.62, + "learning_rate": 1.2385022786057396e-06, + "loss": 0.1539, + "step": 47722 + }, + { + "epoch": 17.62, + "learning_rate": 1.2384653282423945e-06, + "loss": 0.1827, + "step": 47723 + }, + { + "epoch": 17.62, + "learning_rate": 1.238428377879049e-06, + "loss": 0.1836, + "step": 47724 + }, + { + "epoch": 17.62, + "learning_rate": 1.238391427515704e-06, + "loss": 0.2751, + "step": 47725 + }, + { + "epoch": 17.62, + "learning_rate": 1.2383544771523588e-06, + "loss": 0.3432, + "step": 47726 + }, + { + "epoch": 17.62, + "learning_rate": 1.2383175267890134e-06, + "loss": 0.163, + "step": 47727 + }, + { + "epoch": 17.62, + "learning_rate": 1.2382805764256683e-06, + "loss": 0.2318, + "step": 47728 + }, + { + "epoch": 17.63, + "learning_rate": 1.2382436260623231e-06, + "loss": 0.1717, + "step": 47729 + }, + { + "epoch": 17.63, + "learning_rate": 1.2382066756989778e-06, + "loss": 0.1519, + "step": 47730 + }, + { + "epoch": 17.63, + "learning_rate": 1.2381697253356324e-06, + "loss": 0.2564, + "step": 47731 + }, + { + "epoch": 17.63, + "learning_rate": 1.2381327749722872e-06, + "loss": 0.2704, + "step": 47732 + }, + { + "epoch": 17.63, + "learning_rate": 1.238095824608942e-06, + "loss": 0.1972, + "step": 47733 + }, + { + "epoch": 17.63, + "learning_rate": 1.2380588742455967e-06, + "loss": 0.2774, + "step": 47734 + }, + { + "epoch": 17.63, + "learning_rate": 1.2380219238822516e-06, + "loss": 0.1646, + "step": 47735 + }, + { + "epoch": 17.63, + "learning_rate": 1.2379849735189064e-06, + "loss": 0.3547, + "step": 47736 + }, + { + "epoch": 17.63, + "learning_rate": 1.237948023155561e-06, + "loss": 0.3524, + "step": 47737 + }, + { + "epoch": 17.63, + "learning_rate": 1.237911072792216e-06, + "loss": 0.2699, + "step": 47738 + }, + { + "epoch": 17.63, + "learning_rate": 1.2378741224288708e-06, + "loss": 0.1714, + "step": 47739 + }, + { + "epoch": 17.63, + "learning_rate": 1.2378371720655252e-06, + "loss": 0.2209, + "step": 47740 + }, + { + "epoch": 17.63, + "learning_rate": 1.23780022170218e-06, + "loss": 0.2957, + "step": 47741 + }, + { + "epoch": 17.63, + "learning_rate": 1.2377632713388349e-06, + "loss": 0.2941, + "step": 47742 + }, + { + "epoch": 17.63, + "learning_rate": 1.2377263209754895e-06, + "loss": 0.2733, + "step": 47743 + }, + { + "epoch": 17.63, + "learning_rate": 1.2376893706121444e-06, + "loss": 0.274, + "step": 47744 + }, + { + "epoch": 17.63, + "learning_rate": 1.2376524202487992e-06, + "loss": 0.265, + "step": 47745 + }, + { + "epoch": 17.63, + "learning_rate": 1.2376154698854538e-06, + "loss": 0.2765, + "step": 47746 + }, + { + "epoch": 17.63, + "learning_rate": 1.2375785195221087e-06, + "loss": 0.3158, + "step": 47747 + }, + { + "epoch": 17.63, + "learning_rate": 1.2375415691587635e-06, + "loss": 0.1245, + "step": 47748 + }, + { + "epoch": 17.63, + "learning_rate": 1.2375046187954182e-06, + "loss": 0.3526, + "step": 47749 + }, + { + "epoch": 17.63, + "learning_rate": 1.2374676684320728e-06, + "loss": 0.2144, + "step": 47750 + }, + { + "epoch": 17.63, + "learning_rate": 1.2374307180687277e-06, + "loss": 0.1806, + "step": 47751 + }, + { + "epoch": 17.63, + "learning_rate": 1.2373937677053825e-06, + "loss": 0.1632, + "step": 47752 + }, + { + "epoch": 17.63, + "learning_rate": 1.2373568173420371e-06, + "loss": 0.2605, + "step": 47753 + }, + { + "epoch": 17.63, + "learning_rate": 1.237319866978692e-06, + "loss": 0.2885, + "step": 47754 + }, + { + "epoch": 17.63, + "learning_rate": 1.2372829166153468e-06, + "loss": 0.3032, + "step": 47755 + }, + { + "epoch": 17.64, + "learning_rate": 1.2372459662520015e-06, + "loss": 0.1224, + "step": 47756 + }, + { + "epoch": 17.64, + "learning_rate": 1.2372090158886563e-06, + "loss": 0.2612, + "step": 47757 + }, + { + "epoch": 17.64, + "learning_rate": 1.2371720655253112e-06, + "loss": 0.1859, + "step": 47758 + }, + { + "epoch": 17.64, + "learning_rate": 1.2371351151619658e-06, + "loss": 0.3436, + "step": 47759 + }, + { + "epoch": 17.64, + "learning_rate": 1.2370981647986205e-06, + "loss": 0.3071, + "step": 47760 + }, + { + "epoch": 17.64, + "learning_rate": 1.2370612144352753e-06, + "loss": 0.101, + "step": 47761 + }, + { + "epoch": 17.64, + "learning_rate": 1.23702426407193e-06, + "loss": 0.2283, + "step": 47762 + }, + { + "epoch": 17.64, + "learning_rate": 1.2369873137085848e-06, + "loss": 0.285, + "step": 47763 + }, + { + "epoch": 17.64, + "learning_rate": 1.2369503633452396e-06, + "loss": 0.2403, + "step": 47764 + }, + { + "epoch": 17.64, + "learning_rate": 1.2369134129818943e-06, + "loss": 0.2618, + "step": 47765 + }, + { + "epoch": 17.64, + "learning_rate": 1.2368764626185491e-06, + "loss": 0.2283, + "step": 47766 + }, + { + "epoch": 17.64, + "learning_rate": 1.236839512255204e-06, + "loss": 0.3416, + "step": 47767 + }, + { + "epoch": 17.64, + "learning_rate": 1.2368025618918586e-06, + "loss": 0.1813, + "step": 47768 + }, + { + "epoch": 17.64, + "learning_rate": 1.2367656115285135e-06, + "loss": 0.1773, + "step": 47769 + }, + { + "epoch": 17.64, + "learning_rate": 1.236728661165168e-06, + "loss": 0.3809, + "step": 47770 + }, + { + "epoch": 17.64, + "learning_rate": 1.236691710801823e-06, + "loss": 0.2375, + "step": 47771 + }, + { + "epoch": 17.64, + "learning_rate": 1.2366547604384776e-06, + "loss": 0.1779, + "step": 47772 + }, + { + "epoch": 17.64, + "learning_rate": 1.2366178100751324e-06, + "loss": 0.174, + "step": 47773 + }, + { + "epoch": 17.64, + "learning_rate": 1.2365808597117873e-06, + "loss": 0.4273, + "step": 47774 + }, + { + "epoch": 17.64, + "learning_rate": 1.236543909348442e-06, + "loss": 0.2715, + "step": 47775 + }, + { + "epoch": 17.64, + "learning_rate": 1.2365069589850968e-06, + "loss": 0.1821, + "step": 47776 + }, + { + "epoch": 17.64, + "learning_rate": 1.2364700086217516e-06, + "loss": 0.126, + "step": 47777 + }, + { + "epoch": 17.64, + "learning_rate": 1.2364330582584062e-06, + "loss": 0.1839, + "step": 47778 + }, + { + "epoch": 17.64, + "learning_rate": 1.236396107895061e-06, + "loss": 0.2795, + "step": 47779 + }, + { + "epoch": 17.64, + "learning_rate": 1.2363591575317157e-06, + "loss": 0.1529, + "step": 47780 + }, + { + "epoch": 17.64, + "learning_rate": 1.2363222071683706e-06, + "loss": 0.2196, + "step": 47781 + }, + { + "epoch": 17.64, + "learning_rate": 1.2362852568050252e-06, + "loss": 0.0592, + "step": 47782 + }, + { + "epoch": 17.65, + "learning_rate": 1.23624830644168e-06, + "loss": 0.1742, + "step": 47783 + }, + { + "epoch": 17.65, + "learning_rate": 1.236211356078335e-06, + "loss": 0.1895, + "step": 47784 + }, + { + "epoch": 17.65, + "learning_rate": 1.2361744057149895e-06, + "loss": 0.2408, + "step": 47785 + }, + { + "epoch": 17.65, + "learning_rate": 1.2361374553516444e-06, + "loss": 0.1452, + "step": 47786 + }, + { + "epoch": 17.65, + "learning_rate": 1.236100504988299e-06, + "loss": 0.1621, + "step": 47787 + }, + { + "epoch": 17.65, + "learning_rate": 1.2360635546249539e-06, + "loss": 0.2797, + "step": 47788 + }, + { + "epoch": 17.65, + "learning_rate": 1.2360266042616087e-06, + "loss": 0.3883, + "step": 47789 + }, + { + "epoch": 17.65, + "learning_rate": 1.2359896538982634e-06, + "loss": 0.2248, + "step": 47790 + }, + { + "epoch": 17.65, + "learning_rate": 1.235952703534918e-06, + "loss": 0.2572, + "step": 47791 + }, + { + "epoch": 17.65, + "learning_rate": 1.2359157531715728e-06, + "loss": 0.2501, + "step": 47792 + }, + { + "epoch": 17.65, + "learning_rate": 1.2358788028082277e-06, + "loss": 0.3482, + "step": 47793 + }, + { + "epoch": 17.65, + "learning_rate": 1.2358418524448823e-06, + "loss": 0.2667, + "step": 47794 + }, + { + "epoch": 17.65, + "learning_rate": 1.2358049020815372e-06, + "loss": 0.1622, + "step": 47795 + }, + { + "epoch": 17.65, + "learning_rate": 1.235767951718192e-06, + "loss": 0.1554, + "step": 47796 + }, + { + "epoch": 17.65, + "learning_rate": 1.2357310013548467e-06, + "loss": 0.2098, + "step": 47797 + }, + { + "epoch": 17.65, + "learning_rate": 1.2356940509915015e-06, + "loss": 0.1748, + "step": 47798 + }, + { + "epoch": 17.65, + "learning_rate": 1.2356571006281564e-06, + "loss": 0.1393, + "step": 47799 + }, + { + "epoch": 17.65, + "learning_rate": 1.235620150264811e-06, + "loss": 0.1895, + "step": 47800 + }, + { + "epoch": 17.65, + "learning_rate": 1.2355831999014656e-06, + "loss": 0.2803, + "step": 47801 + }, + { + "epoch": 17.65, + "learning_rate": 1.2355462495381205e-06, + "loss": 0.1753, + "step": 47802 + }, + { + "epoch": 17.65, + "learning_rate": 1.2355092991747753e-06, + "loss": 0.2551, + "step": 47803 + }, + { + "epoch": 17.65, + "learning_rate": 1.23547234881143e-06, + "loss": 0.2781, + "step": 47804 + }, + { + "epoch": 17.65, + "learning_rate": 1.2354353984480848e-06, + "loss": 0.1539, + "step": 47805 + }, + { + "epoch": 17.65, + "learning_rate": 1.2353984480847397e-06, + "loss": 0.1758, + "step": 47806 + }, + { + "epoch": 17.65, + "learning_rate": 1.2353614977213943e-06, + "loss": 0.2517, + "step": 47807 + }, + { + "epoch": 17.65, + "learning_rate": 1.2353245473580491e-06, + "loss": 0.2351, + "step": 47808 + }, + { + "epoch": 17.65, + "learning_rate": 1.235287596994704e-06, + "loss": 0.2168, + "step": 47809 + }, + { + "epoch": 17.66, + "learning_rate": 1.2352506466313584e-06, + "loss": 0.2533, + "step": 47810 + }, + { + "epoch": 17.66, + "learning_rate": 1.2352136962680133e-06, + "loss": 0.2101, + "step": 47811 + }, + { + "epoch": 17.66, + "learning_rate": 1.2351767459046681e-06, + "loss": 0.3227, + "step": 47812 + }, + { + "epoch": 17.66, + "learning_rate": 1.2351397955413227e-06, + "loss": 0.1415, + "step": 47813 + }, + { + "epoch": 17.66, + "learning_rate": 1.2351028451779776e-06, + "loss": 0.3677, + "step": 47814 + }, + { + "epoch": 17.66, + "learning_rate": 1.2350658948146324e-06, + "loss": 0.1881, + "step": 47815 + }, + { + "epoch": 17.66, + "learning_rate": 1.235028944451287e-06, + "loss": 0.3629, + "step": 47816 + }, + { + "epoch": 17.66, + "learning_rate": 1.234991994087942e-06, + "loss": 0.3135, + "step": 47817 + }, + { + "epoch": 17.66, + "learning_rate": 1.2349550437245968e-06, + "loss": 0.1493, + "step": 47818 + }, + { + "epoch": 17.66, + "learning_rate": 1.2349180933612514e-06, + "loss": 0.2221, + "step": 47819 + }, + { + "epoch": 17.66, + "learning_rate": 1.234881142997906e-06, + "loss": 0.2503, + "step": 47820 + }, + { + "epoch": 17.66, + "learning_rate": 1.234844192634561e-06, + "loss": 0.3022, + "step": 47821 + }, + { + "epoch": 17.66, + "learning_rate": 1.2348072422712157e-06, + "loss": 0.3235, + "step": 47822 + }, + { + "epoch": 17.66, + "learning_rate": 1.2347702919078704e-06, + "loss": 0.2896, + "step": 47823 + }, + { + "epoch": 17.66, + "learning_rate": 1.2347333415445252e-06, + "loss": 0.1835, + "step": 47824 + }, + { + "epoch": 17.66, + "learning_rate": 1.23469639118118e-06, + "loss": 0.1973, + "step": 47825 + }, + { + "epoch": 17.66, + "learning_rate": 1.2346594408178347e-06, + "loss": 0.1818, + "step": 47826 + }, + { + "epoch": 17.66, + "learning_rate": 1.2346224904544896e-06, + "loss": 0.2167, + "step": 47827 + }, + { + "epoch": 17.66, + "learning_rate": 1.2345855400911444e-06, + "loss": 0.2229, + "step": 47828 + }, + { + "epoch": 17.66, + "learning_rate": 1.234548589727799e-06, + "loss": 0.1692, + "step": 47829 + }, + { + "epoch": 17.66, + "learning_rate": 1.2345116393644537e-06, + "loss": 0.1482, + "step": 47830 + }, + { + "epoch": 17.66, + "learning_rate": 1.2344746890011085e-06, + "loss": 0.3254, + "step": 47831 + }, + { + "epoch": 17.66, + "learning_rate": 1.2344377386377632e-06, + "loss": 0.2463, + "step": 47832 + }, + { + "epoch": 17.66, + "learning_rate": 1.234400788274418e-06, + "loss": 0.1771, + "step": 47833 + }, + { + "epoch": 17.66, + "learning_rate": 1.2343638379110729e-06, + "loss": 0.3125, + "step": 47834 + }, + { + "epoch": 17.66, + "learning_rate": 1.2343268875477275e-06, + "loss": 0.2334, + "step": 47835 + }, + { + "epoch": 17.66, + "learning_rate": 1.2342899371843824e-06, + "loss": 0.1742, + "step": 47836 + }, + { + "epoch": 17.67, + "learning_rate": 1.2342529868210372e-06, + "loss": 0.1853, + "step": 47837 + }, + { + "epoch": 17.67, + "learning_rate": 1.2342160364576918e-06, + "loss": 0.1809, + "step": 47838 + }, + { + "epoch": 17.67, + "learning_rate": 1.2341790860943465e-06, + "loss": 0.259, + "step": 47839 + }, + { + "epoch": 17.67, + "learning_rate": 1.2341421357310013e-06, + "loss": 0.1833, + "step": 47840 + }, + { + "epoch": 17.67, + "learning_rate": 1.2341051853676562e-06, + "loss": 0.3494, + "step": 47841 + }, + { + "epoch": 17.67, + "learning_rate": 1.2340682350043108e-06, + "loss": 0.1816, + "step": 47842 + }, + { + "epoch": 17.67, + "learning_rate": 1.2340312846409657e-06, + "loss": 0.2747, + "step": 47843 + }, + { + "epoch": 17.67, + "learning_rate": 1.2339943342776205e-06, + "loss": 0.3607, + "step": 47844 + }, + { + "epoch": 17.67, + "learning_rate": 1.2339573839142751e-06, + "loss": 0.1411, + "step": 47845 + }, + { + "epoch": 17.67, + "learning_rate": 1.23392043355093e-06, + "loss": 0.2113, + "step": 47846 + }, + { + "epoch": 17.67, + "learning_rate": 1.2338834831875848e-06, + "loss": 0.2665, + "step": 47847 + }, + { + "epoch": 17.67, + "learning_rate": 1.2338465328242395e-06, + "loss": 0.3863, + "step": 47848 + }, + { + "epoch": 17.67, + "learning_rate": 1.2338095824608941e-06, + "loss": 0.1343, + "step": 47849 + }, + { + "epoch": 17.67, + "learning_rate": 1.233772632097549e-06, + "loss": 0.2701, + "step": 47850 + }, + { + "epoch": 17.67, + "learning_rate": 1.2337356817342038e-06, + "loss": 0.1883, + "step": 47851 + }, + { + "epoch": 17.67, + "learning_rate": 1.2336987313708584e-06, + "loss": 0.2525, + "step": 47852 + }, + { + "epoch": 17.67, + "learning_rate": 1.2336617810075133e-06, + "loss": 0.1753, + "step": 47853 + }, + { + "epoch": 17.67, + "learning_rate": 1.2336248306441681e-06, + "loss": 0.1814, + "step": 47854 + }, + { + "epoch": 17.67, + "learning_rate": 1.2335878802808228e-06, + "loss": 0.2953, + "step": 47855 + }, + { + "epoch": 17.67, + "learning_rate": 1.2335509299174776e-06, + "loss": 0.1595, + "step": 47856 + }, + { + "epoch": 17.67, + "learning_rate": 1.2335139795541325e-06, + "loss": 0.2032, + "step": 47857 + }, + { + "epoch": 17.67, + "learning_rate": 1.233477029190787e-06, + "loss": 0.1867, + "step": 47858 + }, + { + "epoch": 17.67, + "learning_rate": 1.2334400788274417e-06, + "loss": 0.1479, + "step": 47859 + }, + { + "epoch": 17.67, + "learning_rate": 1.2334031284640966e-06, + "loss": 0.2867, + "step": 47860 + }, + { + "epoch": 17.67, + "learning_rate": 1.2333661781007512e-06, + "loss": 0.2368, + "step": 47861 + }, + { + "epoch": 17.67, + "learning_rate": 1.233329227737406e-06, + "loss": 0.3541, + "step": 47862 + }, + { + "epoch": 17.67, + "learning_rate": 1.233292277374061e-06, + "loss": 0.2588, + "step": 47863 + }, + { + "epoch": 17.68, + "learning_rate": 1.2332553270107156e-06, + "loss": 0.2308, + "step": 47864 + }, + { + "epoch": 17.68, + "learning_rate": 1.2332183766473704e-06, + "loss": 0.183, + "step": 47865 + }, + { + "epoch": 17.68, + "learning_rate": 1.2331814262840253e-06, + "loss": 0.1224, + "step": 47866 + }, + { + "epoch": 17.68, + "learning_rate": 1.2331444759206799e-06, + "loss": 0.3448, + "step": 47867 + }, + { + "epoch": 17.68, + "learning_rate": 1.2331075255573347e-06, + "loss": 0.1845, + "step": 47868 + }, + { + "epoch": 17.68, + "learning_rate": 1.2330705751939894e-06, + "loss": 0.4213, + "step": 47869 + }, + { + "epoch": 17.68, + "learning_rate": 1.2330336248306442e-06, + "loss": 0.3384, + "step": 47870 + }, + { + "epoch": 17.68, + "learning_rate": 1.2329966744672989e-06, + "loss": 0.1847, + "step": 47871 + }, + { + "epoch": 17.68, + "learning_rate": 1.2329597241039537e-06, + "loss": 0.1415, + "step": 47872 + }, + { + "epoch": 17.68, + "learning_rate": 1.2329227737406086e-06, + "loss": 0.1966, + "step": 47873 + }, + { + "epoch": 17.68, + "learning_rate": 1.2328858233772632e-06, + "loss": 0.1568, + "step": 47874 + }, + { + "epoch": 17.68, + "learning_rate": 1.232848873013918e-06, + "loss": 0.174, + "step": 47875 + }, + { + "epoch": 17.68, + "learning_rate": 1.2328119226505729e-06, + "loss": 0.22, + "step": 47876 + }, + { + "epoch": 17.68, + "learning_rate": 1.2327749722872275e-06, + "loss": 0.2616, + "step": 47877 + }, + { + "epoch": 17.68, + "learning_rate": 1.2327380219238824e-06, + "loss": 0.1428, + "step": 47878 + }, + { + "epoch": 17.68, + "learning_rate": 1.232701071560537e-06, + "loss": 0.2305, + "step": 47879 + }, + { + "epoch": 17.68, + "learning_rate": 1.2326641211971916e-06, + "loss": 0.117, + "step": 47880 + }, + { + "epoch": 17.68, + "learning_rate": 1.2326271708338465e-06, + "loss": 0.2333, + "step": 47881 + }, + { + "epoch": 17.68, + "learning_rate": 1.2325902204705013e-06, + "loss": 0.1783, + "step": 47882 + }, + { + "epoch": 17.68, + "learning_rate": 1.232553270107156e-06, + "loss": 0.2427, + "step": 47883 + }, + { + "epoch": 17.68, + "learning_rate": 1.2325163197438108e-06, + "loss": 0.1878, + "step": 47884 + }, + { + "epoch": 17.68, + "learning_rate": 1.2324793693804657e-06, + "loss": 0.3902, + "step": 47885 + }, + { + "epoch": 17.68, + "learning_rate": 1.2324424190171203e-06, + "loss": 0.3172, + "step": 47886 + }, + { + "epoch": 17.68, + "learning_rate": 1.2324054686537752e-06, + "loss": 0.2271, + "step": 47887 + }, + { + "epoch": 17.68, + "learning_rate": 1.23236851829043e-06, + "loss": 0.24, + "step": 47888 + }, + { + "epoch": 17.68, + "learning_rate": 1.2323315679270846e-06, + "loss": 0.2781, + "step": 47889 + }, + { + "epoch": 17.68, + "learning_rate": 1.2322946175637393e-06, + "loss": 0.1975, + "step": 47890 + }, + { + "epoch": 17.69, + "learning_rate": 1.2322576672003941e-06, + "loss": 0.3266, + "step": 47891 + }, + { + "epoch": 17.69, + "learning_rate": 1.232220716837049e-06, + "loss": 0.1932, + "step": 47892 + }, + { + "epoch": 17.69, + "learning_rate": 1.2321837664737036e-06, + "loss": 0.2311, + "step": 47893 + }, + { + "epoch": 17.69, + "learning_rate": 1.2321468161103585e-06, + "loss": 0.2556, + "step": 47894 + }, + { + "epoch": 17.69, + "learning_rate": 1.2321098657470133e-06, + "loss": 0.2139, + "step": 47895 + }, + { + "epoch": 17.69, + "learning_rate": 1.232072915383668e-06, + "loss": 0.206, + "step": 47896 + }, + { + "epoch": 17.69, + "learning_rate": 1.2320359650203228e-06, + "loss": 0.096, + "step": 47897 + }, + { + "epoch": 17.69, + "learning_rate": 1.2319990146569776e-06, + "loss": 0.2115, + "step": 47898 + }, + { + "epoch": 17.69, + "learning_rate": 1.231962064293632e-06, + "loss": 0.386, + "step": 47899 + }, + { + "epoch": 17.69, + "learning_rate": 1.231925113930287e-06, + "loss": 0.2518, + "step": 47900 + }, + { + "epoch": 17.69, + "learning_rate": 1.2318881635669418e-06, + "loss": 0.2918, + "step": 47901 + }, + { + "epoch": 17.69, + "learning_rate": 1.2318512132035964e-06, + "loss": 0.3101, + "step": 47902 + }, + { + "epoch": 17.69, + "learning_rate": 1.2318142628402512e-06, + "loss": 0.2696, + "step": 47903 + }, + { + "epoch": 17.69, + "learning_rate": 1.231777312476906e-06, + "loss": 0.224, + "step": 47904 + }, + { + "epoch": 17.69, + "learning_rate": 1.2317403621135607e-06, + "loss": 0.2487, + "step": 47905 + }, + { + "epoch": 17.69, + "learning_rate": 1.2317034117502156e-06, + "loss": 0.2164, + "step": 47906 + }, + { + "epoch": 17.69, + "learning_rate": 1.2316664613868704e-06, + "loss": 0.1664, + "step": 47907 + }, + { + "epoch": 17.69, + "learning_rate": 1.231629511023525e-06, + "loss": 0.2041, + "step": 47908 + }, + { + "epoch": 17.69, + "learning_rate": 1.2315925606601797e-06, + "loss": 0.1286, + "step": 47909 + }, + { + "epoch": 17.69, + "learning_rate": 1.2315556102968346e-06, + "loss": 0.2051, + "step": 47910 + }, + { + "epoch": 17.69, + "learning_rate": 1.2315186599334894e-06, + "loss": 0.2315, + "step": 47911 + }, + { + "epoch": 17.69, + "learning_rate": 1.231481709570144e-06, + "loss": 0.2652, + "step": 47912 + }, + { + "epoch": 17.69, + "learning_rate": 1.2314447592067989e-06, + "loss": 0.3196, + "step": 47913 + }, + { + "epoch": 17.69, + "learning_rate": 1.2314078088434537e-06, + "loss": 0.2306, + "step": 47914 + }, + { + "epoch": 17.69, + "learning_rate": 1.2313708584801084e-06, + "loss": 0.3296, + "step": 47915 + }, + { + "epoch": 17.69, + "learning_rate": 1.2313339081167632e-06, + "loss": 0.2746, + "step": 47916 + }, + { + "epoch": 17.69, + "learning_rate": 1.231296957753418e-06, + "loss": 0.3146, + "step": 47917 + }, + { + "epoch": 17.69, + "learning_rate": 1.2312600073900727e-06, + "loss": 0.3027, + "step": 47918 + }, + { + "epoch": 17.7, + "learning_rate": 1.2312230570267273e-06, + "loss": 0.2919, + "step": 47919 + }, + { + "epoch": 17.7, + "learning_rate": 1.2311861066633822e-06, + "loss": 0.2979, + "step": 47920 + }, + { + "epoch": 17.7, + "learning_rate": 1.231149156300037e-06, + "loss": 0.2336, + "step": 47921 + }, + { + "epoch": 17.7, + "learning_rate": 1.2311122059366917e-06, + "loss": 0.3329, + "step": 47922 + }, + { + "epoch": 17.7, + "learning_rate": 1.2310752555733465e-06, + "loss": 0.1731, + "step": 47923 + }, + { + "epoch": 17.7, + "learning_rate": 1.2310383052100014e-06, + "loss": 0.3066, + "step": 47924 + }, + { + "epoch": 17.7, + "learning_rate": 1.231001354846656e-06, + "loss": 0.1961, + "step": 47925 + }, + { + "epoch": 17.7, + "learning_rate": 1.2309644044833109e-06, + "loss": 0.2128, + "step": 47926 + }, + { + "epoch": 17.7, + "learning_rate": 1.2309274541199657e-06, + "loss": 0.4244, + "step": 47927 + }, + { + "epoch": 17.7, + "learning_rate": 1.2308905037566203e-06, + "loss": 0.1402, + "step": 47928 + }, + { + "epoch": 17.7, + "learning_rate": 1.230853553393275e-06, + "loss": 0.282, + "step": 47929 + }, + { + "epoch": 17.7, + "learning_rate": 1.2308166030299298e-06, + "loss": 0.191, + "step": 47930 + }, + { + "epoch": 17.7, + "learning_rate": 1.2307796526665845e-06, + "loss": 0.2634, + "step": 47931 + }, + { + "epoch": 17.7, + "learning_rate": 1.2307427023032393e-06, + "loss": 0.4388, + "step": 47932 + }, + { + "epoch": 17.7, + "learning_rate": 1.2307057519398942e-06, + "loss": 0.1884, + "step": 47933 + }, + { + "epoch": 17.7, + "learning_rate": 1.2306688015765488e-06, + "loss": 0.3059, + "step": 47934 + }, + { + "epoch": 17.7, + "learning_rate": 1.2306318512132036e-06, + "loss": 0.3345, + "step": 47935 + }, + { + "epoch": 17.7, + "learning_rate": 1.2305949008498585e-06, + "loss": 0.3369, + "step": 47936 + }, + { + "epoch": 17.7, + "learning_rate": 1.2305579504865131e-06, + "loss": 0.19, + "step": 47937 + }, + { + "epoch": 17.7, + "learning_rate": 1.230521000123168e-06, + "loss": 0.3685, + "step": 47938 + }, + { + "epoch": 17.7, + "learning_rate": 1.2304840497598226e-06, + "loss": 0.1897, + "step": 47939 + }, + { + "epoch": 17.7, + "learning_rate": 1.2304470993964775e-06, + "loss": 0.2044, + "step": 47940 + }, + { + "epoch": 17.7, + "learning_rate": 1.230410149033132e-06, + "loss": 0.2295, + "step": 47941 + }, + { + "epoch": 17.7, + "learning_rate": 1.230373198669787e-06, + "loss": 0.2541, + "step": 47942 + }, + { + "epoch": 17.7, + "learning_rate": 1.2303362483064418e-06, + "loss": 0.2155, + "step": 47943 + }, + { + "epoch": 17.7, + "learning_rate": 1.2302992979430964e-06, + "loss": 0.2143, + "step": 47944 + }, + { + "epoch": 17.7, + "learning_rate": 1.2302623475797513e-06, + "loss": 0.3612, + "step": 47945 + }, + { + "epoch": 17.71, + "learning_rate": 1.2302253972164061e-06, + "loss": 0.189, + "step": 47946 + }, + { + "epoch": 17.71, + "learning_rate": 1.2301884468530608e-06, + "loss": 0.1512, + "step": 47947 + }, + { + "epoch": 17.71, + "learning_rate": 1.2301514964897156e-06, + "loss": 0.1474, + "step": 47948 + }, + { + "epoch": 17.71, + "learning_rate": 1.2301145461263702e-06, + "loss": 0.2949, + "step": 47949 + }, + { + "epoch": 17.71, + "learning_rate": 1.2300775957630249e-06, + "loss": 0.2097, + "step": 47950 + }, + { + "epoch": 17.71, + "learning_rate": 1.2300406453996797e-06, + "loss": 0.1877, + "step": 47951 + }, + { + "epoch": 17.71, + "learning_rate": 1.2300036950363346e-06, + "loss": 0.227, + "step": 47952 + }, + { + "epoch": 17.71, + "learning_rate": 1.2299667446729892e-06, + "loss": 0.2736, + "step": 47953 + }, + { + "epoch": 17.71, + "learning_rate": 1.229929794309644e-06, + "loss": 0.1682, + "step": 47954 + }, + { + "epoch": 17.71, + "learning_rate": 1.229892843946299e-06, + "loss": 0.1867, + "step": 47955 + }, + { + "epoch": 17.71, + "learning_rate": 1.2298558935829535e-06, + "loss": 0.2266, + "step": 47956 + }, + { + "epoch": 17.71, + "learning_rate": 1.2298189432196084e-06, + "loss": 0.2224, + "step": 47957 + }, + { + "epoch": 17.71, + "learning_rate": 1.2297819928562632e-06, + "loss": 0.2367, + "step": 47958 + }, + { + "epoch": 17.71, + "learning_rate": 1.2297450424929179e-06, + "loss": 0.2485, + "step": 47959 + }, + { + "epoch": 17.71, + "learning_rate": 1.2297080921295725e-06, + "loss": 0.3223, + "step": 47960 + }, + { + "epoch": 17.71, + "learning_rate": 1.2296711417662274e-06, + "loss": 0.2438, + "step": 47961 + }, + { + "epoch": 17.71, + "learning_rate": 1.2296341914028822e-06, + "loss": 0.1788, + "step": 47962 + }, + { + "epoch": 17.71, + "learning_rate": 1.2295972410395368e-06, + "loss": 0.1705, + "step": 47963 + }, + { + "epoch": 17.71, + "learning_rate": 1.2295602906761917e-06, + "loss": 0.2913, + "step": 47964 + }, + { + "epoch": 17.71, + "learning_rate": 1.2295233403128465e-06, + "loss": 0.3026, + "step": 47965 + }, + { + "epoch": 17.71, + "learning_rate": 1.2294863899495012e-06, + "loss": 0.3361, + "step": 47966 + }, + { + "epoch": 17.71, + "learning_rate": 1.229449439586156e-06, + "loss": 0.207, + "step": 47967 + }, + { + "epoch": 17.71, + "learning_rate": 1.2294124892228109e-06, + "loss": 0.2622, + "step": 47968 + }, + { + "epoch": 17.71, + "learning_rate": 1.2293755388594655e-06, + "loss": 0.276, + "step": 47969 + }, + { + "epoch": 17.71, + "learning_rate": 1.2293385884961201e-06, + "loss": 0.2408, + "step": 47970 + }, + { + "epoch": 17.71, + "learning_rate": 1.229301638132775e-06, + "loss": 0.1372, + "step": 47971 + }, + { + "epoch": 17.71, + "learning_rate": 1.2292646877694298e-06, + "loss": 0.1608, + "step": 47972 + }, + { + "epoch": 17.72, + "learning_rate": 1.2292277374060845e-06, + "loss": 0.1581, + "step": 47973 + }, + { + "epoch": 17.72, + "learning_rate": 1.2291907870427393e-06, + "loss": 0.3202, + "step": 47974 + }, + { + "epoch": 17.72, + "learning_rate": 1.229153836679394e-06, + "loss": 0.2409, + "step": 47975 + }, + { + "epoch": 17.72, + "learning_rate": 1.2291168863160488e-06, + "loss": 0.0987, + "step": 47976 + }, + { + "epoch": 17.72, + "learning_rate": 1.2290799359527037e-06, + "loss": 0.2315, + "step": 47977 + }, + { + "epoch": 17.72, + "learning_rate": 1.2290429855893583e-06, + "loss": 0.1562, + "step": 47978 + }, + { + "epoch": 17.72, + "learning_rate": 1.229006035226013e-06, + "loss": 0.4971, + "step": 47979 + }, + { + "epoch": 17.72, + "learning_rate": 1.2289690848626678e-06, + "loss": 0.2928, + "step": 47980 + }, + { + "epoch": 17.72, + "learning_rate": 1.2289321344993226e-06, + "loss": 0.2079, + "step": 47981 + }, + { + "epoch": 17.72, + "learning_rate": 1.2288951841359773e-06, + "loss": 0.2879, + "step": 47982 + }, + { + "epoch": 17.72, + "learning_rate": 1.2288582337726321e-06, + "loss": 0.28, + "step": 47983 + }, + { + "epoch": 17.72, + "learning_rate": 1.228821283409287e-06, + "loss": 0.1931, + "step": 47984 + }, + { + "epoch": 17.72, + "learning_rate": 1.2287843330459416e-06, + "loss": 0.1969, + "step": 47985 + }, + { + "epoch": 17.72, + "learning_rate": 1.2287473826825964e-06, + "loss": 0.2249, + "step": 47986 + }, + { + "epoch": 17.72, + "learning_rate": 1.2287104323192513e-06, + "loss": 0.1391, + "step": 47987 + }, + { + "epoch": 17.72, + "learning_rate": 1.228673481955906e-06, + "loss": 0.1494, + "step": 47988 + }, + { + "epoch": 17.72, + "learning_rate": 1.2286365315925606e-06, + "loss": 0.366, + "step": 47989 + }, + { + "epoch": 17.72, + "learning_rate": 1.2285995812292154e-06, + "loss": 0.2531, + "step": 47990 + }, + { + "epoch": 17.72, + "learning_rate": 1.2285626308658703e-06, + "loss": 0.1731, + "step": 47991 + }, + { + "epoch": 17.72, + "learning_rate": 1.228525680502525e-06, + "loss": 0.2925, + "step": 47992 + }, + { + "epoch": 17.72, + "learning_rate": 1.2284887301391798e-06, + "loss": 0.1983, + "step": 47993 + }, + { + "epoch": 17.72, + "learning_rate": 1.2284517797758346e-06, + "loss": 0.189, + "step": 47994 + }, + { + "epoch": 17.72, + "learning_rate": 1.2284148294124892e-06, + "loss": 0.232, + "step": 47995 + }, + { + "epoch": 17.72, + "learning_rate": 1.228377879049144e-06, + "loss": 0.1844, + "step": 47996 + }, + { + "epoch": 17.72, + "learning_rate": 1.228340928685799e-06, + "loss": 0.1185, + "step": 47997 + }, + { + "epoch": 17.72, + "learning_rate": 1.2283039783224536e-06, + "loss": 0.2286, + "step": 47998 + }, + { + "epoch": 17.72, + "learning_rate": 1.2282670279591082e-06, + "loss": 0.2368, + "step": 47999 + }, + { + "epoch": 17.73, + "learning_rate": 1.228230077595763e-06, + "loss": 0.3886, + "step": 48000 + }, + { + "epoch": 17.73, + "learning_rate": 1.2281931272324177e-06, + "loss": 0.2352, + "step": 48001 + }, + { + "epoch": 17.73, + "learning_rate": 1.2281561768690725e-06, + "loss": 0.141, + "step": 48002 + }, + { + "epoch": 17.73, + "learning_rate": 1.2281192265057274e-06, + "loss": 0.2015, + "step": 48003 + }, + { + "epoch": 17.73, + "learning_rate": 1.228082276142382e-06, + "loss": 0.1889, + "step": 48004 + }, + { + "epoch": 17.73, + "learning_rate": 1.2280453257790369e-06, + "loss": 0.2751, + "step": 48005 + }, + { + "epoch": 17.73, + "learning_rate": 1.2280083754156917e-06, + "loss": 0.275, + "step": 48006 + }, + { + "epoch": 17.73, + "learning_rate": 1.2279714250523464e-06, + "loss": 0.1473, + "step": 48007 + }, + { + "epoch": 17.73, + "learning_rate": 1.2279344746890012e-06, + "loss": 0.3258, + "step": 48008 + }, + { + "epoch": 17.73, + "learning_rate": 1.2278975243256558e-06, + "loss": 0.2558, + "step": 48009 + }, + { + "epoch": 17.73, + "learning_rate": 1.2278605739623107e-06, + "loss": 0.2024, + "step": 48010 + }, + { + "epoch": 17.73, + "learning_rate": 1.2278236235989653e-06, + "loss": 0.3315, + "step": 48011 + }, + { + "epoch": 17.73, + "learning_rate": 1.2277866732356202e-06, + "loss": 0.4375, + "step": 48012 + }, + { + "epoch": 17.73, + "learning_rate": 1.227749722872275e-06, + "loss": 0.3294, + "step": 48013 + }, + { + "epoch": 17.73, + "learning_rate": 1.2277127725089297e-06, + "loss": 0.29, + "step": 48014 + }, + { + "epoch": 17.73, + "learning_rate": 1.2276758221455845e-06, + "loss": 0.3905, + "step": 48015 + }, + { + "epoch": 17.73, + "learning_rate": 1.2276388717822394e-06, + "loss": 0.3298, + "step": 48016 + }, + { + "epoch": 17.73, + "learning_rate": 1.227601921418894e-06, + "loss": 0.2756, + "step": 48017 + }, + { + "epoch": 17.73, + "learning_rate": 1.2275649710555488e-06, + "loss": 0.2356, + "step": 48018 + }, + { + "epoch": 17.73, + "learning_rate": 1.2275280206922035e-06, + "loss": 0.2812, + "step": 48019 + }, + { + "epoch": 17.73, + "learning_rate": 1.2274910703288581e-06, + "loss": 0.2459, + "step": 48020 + }, + { + "epoch": 17.73, + "learning_rate": 1.227454119965513e-06, + "loss": 0.2753, + "step": 48021 + }, + { + "epoch": 17.73, + "learning_rate": 1.2274171696021678e-06, + "loss": 0.2966, + "step": 48022 + }, + { + "epoch": 17.73, + "learning_rate": 1.2273802192388224e-06, + "loss": 0.2076, + "step": 48023 + }, + { + "epoch": 17.73, + "learning_rate": 1.2273432688754773e-06, + "loss": 0.2705, + "step": 48024 + }, + { + "epoch": 17.73, + "learning_rate": 1.2273063185121321e-06, + "loss": 0.3119, + "step": 48025 + }, + { + "epoch": 17.73, + "learning_rate": 1.2272693681487868e-06, + "loss": 0.2134, + "step": 48026 + }, + { + "epoch": 17.74, + "learning_rate": 1.2272324177854416e-06, + "loss": 0.1969, + "step": 48027 + }, + { + "epoch": 17.74, + "learning_rate": 1.2271954674220965e-06, + "loss": 0.1956, + "step": 48028 + }, + { + "epoch": 17.74, + "learning_rate": 1.2271585170587511e-06, + "loss": 0.1942, + "step": 48029 + }, + { + "epoch": 17.74, + "learning_rate": 1.2271215666954057e-06, + "loss": 0.2297, + "step": 48030 + }, + { + "epoch": 17.74, + "learning_rate": 1.2270846163320606e-06, + "loss": 0.22, + "step": 48031 + }, + { + "epoch": 17.74, + "learning_rate": 1.2270476659687154e-06, + "loss": 0.2461, + "step": 48032 + }, + { + "epoch": 17.74, + "learning_rate": 1.22701071560537e-06, + "loss": 0.3168, + "step": 48033 + }, + { + "epoch": 17.74, + "learning_rate": 1.226973765242025e-06, + "loss": 0.2069, + "step": 48034 + }, + { + "epoch": 17.74, + "learning_rate": 1.2269368148786798e-06, + "loss": 0.2189, + "step": 48035 + }, + { + "epoch": 17.74, + "learning_rate": 1.2268998645153344e-06, + "loss": 0.1717, + "step": 48036 + }, + { + "epoch": 17.74, + "learning_rate": 1.2268629141519893e-06, + "loss": 0.1666, + "step": 48037 + }, + { + "epoch": 17.74, + "learning_rate": 1.2268259637886441e-06, + "loss": 0.2297, + "step": 48038 + }, + { + "epoch": 17.74, + "learning_rate": 1.2267890134252987e-06, + "loss": 0.2422, + "step": 48039 + }, + { + "epoch": 17.74, + "learning_rate": 1.2267520630619534e-06, + "loss": 0.1885, + "step": 48040 + }, + { + "epoch": 17.74, + "learning_rate": 1.2267151126986082e-06, + "loss": 0.2592, + "step": 48041 + }, + { + "epoch": 17.74, + "learning_rate": 1.226678162335263e-06, + "loss": 0.1369, + "step": 48042 + }, + { + "epoch": 17.74, + "learning_rate": 1.2266412119719177e-06, + "loss": 0.3554, + "step": 48043 + }, + { + "epoch": 17.74, + "learning_rate": 1.2266042616085726e-06, + "loss": 0.2149, + "step": 48044 + }, + { + "epoch": 17.74, + "learning_rate": 1.2265673112452274e-06, + "loss": 0.2881, + "step": 48045 + }, + { + "epoch": 17.74, + "learning_rate": 1.226530360881882e-06, + "loss": 0.2333, + "step": 48046 + }, + { + "epoch": 17.74, + "learning_rate": 1.226493410518537e-06, + "loss": 0.2896, + "step": 48047 + }, + { + "epoch": 17.74, + "learning_rate": 1.2264564601551917e-06, + "loss": 0.2448, + "step": 48048 + }, + { + "epoch": 17.74, + "learning_rate": 1.2264195097918462e-06, + "loss": 0.1961, + "step": 48049 + }, + { + "epoch": 17.74, + "learning_rate": 1.226382559428501e-06, + "loss": 0.2642, + "step": 48050 + }, + { + "epoch": 17.74, + "learning_rate": 1.2263456090651559e-06, + "loss": 0.303, + "step": 48051 + }, + { + "epoch": 17.74, + "learning_rate": 1.2263086587018105e-06, + "loss": 0.3304, + "step": 48052 + }, + { + "epoch": 17.74, + "learning_rate": 1.2262717083384653e-06, + "loss": 0.1646, + "step": 48053 + }, + { + "epoch": 17.75, + "learning_rate": 1.2262347579751202e-06, + "loss": 0.1978, + "step": 48054 + }, + { + "epoch": 17.75, + "learning_rate": 1.2261978076117748e-06, + "loss": 0.1919, + "step": 48055 + }, + { + "epoch": 17.75, + "learning_rate": 1.2261608572484297e-06, + "loss": 0.2152, + "step": 48056 + }, + { + "epoch": 17.75, + "learning_rate": 1.2261239068850845e-06, + "loss": 0.2258, + "step": 48057 + }, + { + "epoch": 17.75, + "learning_rate": 1.2260869565217392e-06, + "loss": 0.423, + "step": 48058 + }, + { + "epoch": 17.75, + "learning_rate": 1.2260500061583938e-06, + "loss": 0.3125, + "step": 48059 + }, + { + "epoch": 17.75, + "learning_rate": 1.2260130557950487e-06, + "loss": 0.2118, + "step": 48060 + }, + { + "epoch": 17.75, + "learning_rate": 1.2259761054317035e-06, + "loss": 0.1068, + "step": 48061 + }, + { + "epoch": 17.75, + "learning_rate": 1.2259391550683581e-06, + "loss": 0.2184, + "step": 48062 + }, + { + "epoch": 17.75, + "learning_rate": 1.225902204705013e-06, + "loss": 0.1367, + "step": 48063 + }, + { + "epoch": 17.75, + "learning_rate": 1.2258652543416678e-06, + "loss": 0.3329, + "step": 48064 + }, + { + "epoch": 17.75, + "learning_rate": 1.2258283039783225e-06, + "loss": 0.2084, + "step": 48065 + }, + { + "epoch": 17.75, + "learning_rate": 1.2257913536149773e-06, + "loss": 0.2308, + "step": 48066 + }, + { + "epoch": 17.75, + "learning_rate": 1.2257544032516322e-06, + "loss": 0.1505, + "step": 48067 + }, + { + "epoch": 17.75, + "learning_rate": 1.2257174528882868e-06, + "loss": 0.1904, + "step": 48068 + }, + { + "epoch": 17.75, + "learning_rate": 1.2256805025249414e-06, + "loss": 0.2121, + "step": 48069 + }, + { + "epoch": 17.75, + "learning_rate": 1.2256435521615963e-06, + "loss": 0.3619, + "step": 48070 + }, + { + "epoch": 17.75, + "learning_rate": 1.225606601798251e-06, + "loss": 0.1692, + "step": 48071 + }, + { + "epoch": 17.75, + "learning_rate": 1.2255696514349058e-06, + "loss": 0.2934, + "step": 48072 + }, + { + "epoch": 17.75, + "learning_rate": 1.2255327010715606e-06, + "loss": 0.3189, + "step": 48073 + }, + { + "epoch": 17.75, + "learning_rate": 1.2254957507082153e-06, + "loss": 0.3453, + "step": 48074 + }, + { + "epoch": 17.75, + "learning_rate": 1.22545880034487e-06, + "loss": 0.2056, + "step": 48075 + }, + { + "epoch": 17.75, + "learning_rate": 1.225421849981525e-06, + "loss": 0.1864, + "step": 48076 + }, + { + "epoch": 17.75, + "learning_rate": 1.2253848996181796e-06, + "loss": 0.1739, + "step": 48077 + }, + { + "epoch": 17.75, + "learning_rate": 1.2253479492548342e-06, + "loss": 0.259, + "step": 48078 + }, + { + "epoch": 17.75, + "learning_rate": 1.225310998891489e-06, + "loss": 0.1634, + "step": 48079 + }, + { + "epoch": 17.75, + "learning_rate": 1.225274048528144e-06, + "loss": 0.2553, + "step": 48080 + }, + { + "epoch": 17.76, + "learning_rate": 1.2252370981647986e-06, + "loss": 0.3279, + "step": 48081 + }, + { + "epoch": 17.76, + "learning_rate": 1.2252001478014534e-06, + "loss": 0.3051, + "step": 48082 + }, + { + "epoch": 17.76, + "learning_rate": 1.2251631974381083e-06, + "loss": 0.261, + "step": 48083 + }, + { + "epoch": 17.76, + "learning_rate": 1.2251262470747629e-06, + "loss": 0.3423, + "step": 48084 + }, + { + "epoch": 17.76, + "learning_rate": 1.2250892967114177e-06, + "loss": 0.2136, + "step": 48085 + }, + { + "epoch": 17.76, + "learning_rate": 1.2250523463480726e-06, + "loss": 0.1644, + "step": 48086 + }, + { + "epoch": 17.76, + "learning_rate": 1.2250153959847272e-06, + "loss": 0.186, + "step": 48087 + }, + { + "epoch": 17.76, + "learning_rate": 1.2249784456213819e-06, + "loss": 0.1525, + "step": 48088 + }, + { + "epoch": 17.76, + "learning_rate": 1.2249414952580367e-06, + "loss": 0.3437, + "step": 48089 + }, + { + "epoch": 17.76, + "learning_rate": 1.2249045448946913e-06, + "loss": 0.2527, + "step": 48090 + }, + { + "epoch": 17.76, + "learning_rate": 1.2248675945313462e-06, + "loss": 0.3387, + "step": 48091 + }, + { + "epoch": 17.76, + "learning_rate": 1.224830644168001e-06, + "loss": 0.2516, + "step": 48092 + }, + { + "epoch": 17.76, + "learning_rate": 1.2247936938046557e-06, + "loss": 0.2574, + "step": 48093 + }, + { + "epoch": 17.76, + "learning_rate": 1.2247567434413105e-06, + "loss": 0.1999, + "step": 48094 + }, + { + "epoch": 17.76, + "learning_rate": 1.2247197930779654e-06, + "loss": 0.1852, + "step": 48095 + }, + { + "epoch": 17.76, + "learning_rate": 1.22468284271462e-06, + "loss": 0.1929, + "step": 48096 + }, + { + "epoch": 17.76, + "learning_rate": 1.2246458923512749e-06, + "loss": 0.1713, + "step": 48097 + }, + { + "epoch": 17.76, + "learning_rate": 1.2246089419879295e-06, + "loss": 0.1736, + "step": 48098 + }, + { + "epoch": 17.76, + "learning_rate": 1.2245719916245843e-06, + "loss": 0.1426, + "step": 48099 + }, + { + "epoch": 17.76, + "learning_rate": 1.224535041261239e-06, + "loss": 0.3481, + "step": 48100 + }, + { + "epoch": 17.76, + "learning_rate": 1.2244980908978938e-06, + "loss": 0.2461, + "step": 48101 + }, + { + "epoch": 17.76, + "learning_rate": 1.2244611405345487e-06, + "loss": 0.3246, + "step": 48102 + }, + { + "epoch": 17.76, + "learning_rate": 1.2244241901712033e-06, + "loss": 0.1968, + "step": 48103 + }, + { + "epoch": 17.76, + "learning_rate": 1.2243872398078582e-06, + "loss": 0.2509, + "step": 48104 + }, + { + "epoch": 17.76, + "learning_rate": 1.224350289444513e-06, + "loss": 0.3063, + "step": 48105 + }, + { + "epoch": 17.76, + "learning_rate": 1.2243133390811676e-06, + "loss": 0.2233, + "step": 48106 + }, + { + "epoch": 17.76, + "learning_rate": 1.2242763887178225e-06, + "loss": 0.4444, + "step": 48107 + }, + { + "epoch": 17.77, + "learning_rate": 1.2242394383544771e-06, + "loss": 0.1632, + "step": 48108 + }, + { + "epoch": 17.77, + "learning_rate": 1.224202487991132e-06, + "loss": 0.2987, + "step": 48109 + }, + { + "epoch": 17.77, + "learning_rate": 1.2241655376277866e-06, + "loss": 0.2973, + "step": 48110 + }, + { + "epoch": 17.77, + "learning_rate": 1.2241285872644415e-06, + "loss": 0.3772, + "step": 48111 + }, + { + "epoch": 17.77, + "learning_rate": 1.2240916369010963e-06, + "loss": 0.3116, + "step": 48112 + }, + { + "epoch": 17.77, + "learning_rate": 1.224054686537751e-06, + "loss": 0.3122, + "step": 48113 + }, + { + "epoch": 17.77, + "learning_rate": 1.2240177361744058e-06, + "loss": 0.1852, + "step": 48114 + }, + { + "epoch": 17.77, + "learning_rate": 1.2239807858110606e-06, + "loss": 0.2058, + "step": 48115 + }, + { + "epoch": 17.77, + "learning_rate": 1.2239438354477153e-06, + "loss": 0.3209, + "step": 48116 + }, + { + "epoch": 17.77, + "learning_rate": 1.2239068850843701e-06, + "loss": 0.1718, + "step": 48117 + }, + { + "epoch": 17.77, + "learning_rate": 1.2238699347210248e-06, + "loss": 0.2879, + "step": 48118 + }, + { + "epoch": 17.77, + "learning_rate": 1.2238329843576794e-06, + "loss": 0.266, + "step": 48119 + }, + { + "epoch": 17.77, + "learning_rate": 1.2237960339943342e-06, + "loss": 0.3386, + "step": 48120 + }, + { + "epoch": 17.77, + "learning_rate": 1.223759083630989e-06, + "loss": 0.302, + "step": 48121 + }, + { + "epoch": 17.77, + "learning_rate": 1.2237221332676437e-06, + "loss": 0.2203, + "step": 48122 + }, + { + "epoch": 17.77, + "learning_rate": 1.2236851829042986e-06, + "loss": 0.2489, + "step": 48123 + }, + { + "epoch": 17.77, + "learning_rate": 1.2236482325409534e-06, + "loss": 0.2223, + "step": 48124 + }, + { + "epoch": 17.77, + "learning_rate": 1.223611282177608e-06, + "loss": 0.3067, + "step": 48125 + }, + { + "epoch": 17.77, + "learning_rate": 1.223574331814263e-06, + "loss": 0.2199, + "step": 48126 + }, + { + "epoch": 17.77, + "learning_rate": 1.2235373814509178e-06, + "loss": 0.2412, + "step": 48127 + }, + { + "epoch": 17.77, + "learning_rate": 1.2235004310875724e-06, + "loss": 0.5262, + "step": 48128 + }, + { + "epoch": 17.77, + "learning_rate": 1.223463480724227e-06, + "loss": 0.1652, + "step": 48129 + }, + { + "epoch": 17.77, + "learning_rate": 1.2234265303608819e-06, + "loss": 0.1418, + "step": 48130 + }, + { + "epoch": 17.77, + "learning_rate": 1.2233895799975367e-06, + "loss": 0.2196, + "step": 48131 + }, + { + "epoch": 17.77, + "learning_rate": 1.2233526296341914e-06, + "loss": 0.4053, + "step": 48132 + }, + { + "epoch": 17.77, + "learning_rate": 1.2233156792708462e-06, + "loss": 0.2182, + "step": 48133 + }, + { + "epoch": 17.77, + "learning_rate": 1.223278728907501e-06, + "loss": 0.3089, + "step": 48134 + }, + { + "epoch": 17.78, + "learning_rate": 1.2232417785441557e-06, + "loss": 0.197, + "step": 48135 + }, + { + "epoch": 17.78, + "learning_rate": 1.2232048281808105e-06, + "loss": 0.2484, + "step": 48136 + }, + { + "epoch": 17.78, + "learning_rate": 1.2231678778174654e-06, + "loss": 0.2505, + "step": 48137 + }, + { + "epoch": 17.78, + "learning_rate": 1.2231309274541198e-06, + "loss": 0.2305, + "step": 48138 + }, + { + "epoch": 17.78, + "learning_rate": 1.2230939770907747e-06, + "loss": 0.1454, + "step": 48139 + }, + { + "epoch": 17.78, + "learning_rate": 1.2230570267274295e-06, + "loss": 0.2465, + "step": 48140 + }, + { + "epoch": 17.78, + "learning_rate": 1.2230200763640842e-06, + "loss": 0.2814, + "step": 48141 + }, + { + "epoch": 17.78, + "learning_rate": 1.222983126000739e-06, + "loss": 0.1219, + "step": 48142 + }, + { + "epoch": 17.78, + "learning_rate": 1.2229461756373939e-06, + "loss": 0.2766, + "step": 48143 + }, + { + "epoch": 17.78, + "learning_rate": 1.2229092252740485e-06, + "loss": 0.1795, + "step": 48144 + }, + { + "epoch": 17.78, + "learning_rate": 1.2228722749107033e-06, + "loss": 0.2635, + "step": 48145 + }, + { + "epoch": 17.78, + "learning_rate": 1.2228353245473582e-06, + "loss": 0.2213, + "step": 48146 + }, + { + "epoch": 17.78, + "learning_rate": 1.2227983741840128e-06, + "loss": 0.1467, + "step": 48147 + }, + { + "epoch": 17.78, + "learning_rate": 1.2227614238206675e-06, + "loss": 0.2262, + "step": 48148 + }, + { + "epoch": 17.78, + "learning_rate": 1.2227244734573223e-06, + "loss": 0.2576, + "step": 48149 + }, + { + "epoch": 17.78, + "learning_rate": 1.2226875230939772e-06, + "loss": 0.1564, + "step": 48150 + }, + { + "epoch": 17.78, + "learning_rate": 1.2226505727306318e-06, + "loss": 0.1663, + "step": 48151 + }, + { + "epoch": 17.78, + "learning_rate": 1.2226136223672866e-06, + "loss": 0.4287, + "step": 48152 + }, + { + "epoch": 17.78, + "learning_rate": 1.2225766720039415e-06, + "loss": 0.274, + "step": 48153 + }, + { + "epoch": 17.78, + "learning_rate": 1.2225397216405961e-06, + "loss": 0.1528, + "step": 48154 + }, + { + "epoch": 17.78, + "learning_rate": 1.222502771277251e-06, + "loss": 0.2797, + "step": 48155 + }, + { + "epoch": 17.78, + "learning_rate": 1.2224658209139058e-06, + "loss": 0.1838, + "step": 48156 + }, + { + "epoch": 17.78, + "learning_rate": 1.2224288705505605e-06, + "loss": 0.2171, + "step": 48157 + }, + { + "epoch": 17.78, + "learning_rate": 1.222391920187215e-06, + "loss": 0.129, + "step": 48158 + }, + { + "epoch": 17.78, + "learning_rate": 1.22235496982387e-06, + "loss": 0.1964, + "step": 48159 + }, + { + "epoch": 17.78, + "learning_rate": 1.2223180194605246e-06, + "loss": 0.2794, + "step": 48160 + }, + { + "epoch": 17.78, + "learning_rate": 1.2222810690971794e-06, + "loss": 0.1519, + "step": 48161 + }, + { + "epoch": 17.79, + "learning_rate": 1.2222441187338343e-06, + "loss": 0.2236, + "step": 48162 + }, + { + "epoch": 17.79, + "learning_rate": 1.222207168370489e-06, + "loss": 0.1782, + "step": 48163 + }, + { + "epoch": 17.79, + "learning_rate": 1.2221702180071438e-06, + "loss": 0.2667, + "step": 48164 + }, + { + "epoch": 17.79, + "learning_rate": 1.2221332676437986e-06, + "loss": 0.2303, + "step": 48165 + }, + { + "epoch": 17.79, + "learning_rate": 1.2220963172804532e-06, + "loss": 0.3601, + "step": 48166 + }, + { + "epoch": 17.79, + "learning_rate": 1.222059366917108e-06, + "loss": 0.2673, + "step": 48167 + }, + { + "epoch": 17.79, + "learning_rate": 1.2220224165537627e-06, + "loss": 0.2558, + "step": 48168 + }, + { + "epoch": 17.79, + "learning_rate": 1.2219854661904176e-06, + "loss": 0.1248, + "step": 48169 + }, + { + "epoch": 17.79, + "learning_rate": 1.2219485158270722e-06, + "loss": 0.1815, + "step": 48170 + }, + { + "epoch": 17.79, + "learning_rate": 1.221911565463727e-06, + "loss": 0.2623, + "step": 48171 + }, + { + "epoch": 17.79, + "learning_rate": 1.221874615100382e-06, + "loss": 0.215, + "step": 48172 + }, + { + "epoch": 17.79, + "learning_rate": 1.2218376647370365e-06, + "loss": 0.2255, + "step": 48173 + }, + { + "epoch": 17.79, + "learning_rate": 1.2218007143736914e-06, + "loss": 0.2071, + "step": 48174 + }, + { + "epoch": 17.79, + "learning_rate": 1.2217637640103462e-06, + "loss": 0.1298, + "step": 48175 + }, + { + "epoch": 17.79, + "learning_rate": 1.2217268136470009e-06, + "loss": 0.134, + "step": 48176 + }, + { + "epoch": 17.79, + "learning_rate": 1.2216898632836557e-06, + "loss": 0.2803, + "step": 48177 + }, + { + "epoch": 17.79, + "learning_rate": 1.2216529129203104e-06, + "loss": 0.2004, + "step": 48178 + }, + { + "epoch": 17.79, + "learning_rate": 1.2216159625569652e-06, + "loss": 0.32, + "step": 48179 + }, + { + "epoch": 17.79, + "learning_rate": 1.2215790121936198e-06, + "loss": 0.1937, + "step": 48180 + }, + { + "epoch": 17.79, + "learning_rate": 1.2215420618302747e-06, + "loss": 0.23, + "step": 48181 + }, + { + "epoch": 17.79, + "learning_rate": 1.2215051114669295e-06, + "loss": 0.3352, + "step": 48182 + }, + { + "epoch": 17.79, + "learning_rate": 1.2214681611035842e-06, + "loss": 0.2023, + "step": 48183 + }, + { + "epoch": 17.79, + "learning_rate": 1.221431210740239e-06, + "loss": 0.1464, + "step": 48184 + }, + { + "epoch": 17.79, + "learning_rate": 1.2213942603768939e-06, + "loss": 0.1716, + "step": 48185 + }, + { + "epoch": 17.79, + "learning_rate": 1.2213573100135485e-06, + "loss": 0.1322, + "step": 48186 + }, + { + "epoch": 17.79, + "learning_rate": 1.2213203596502034e-06, + "loss": 0.1066, + "step": 48187 + }, + { + "epoch": 17.79, + "learning_rate": 1.221283409286858e-06, + "loss": 0.324, + "step": 48188 + }, + { + "epoch": 17.8, + "learning_rate": 1.2212464589235126e-06, + "loss": 0.1986, + "step": 48189 + }, + { + "epoch": 17.8, + "learning_rate": 1.2212095085601675e-06, + "loss": 0.2815, + "step": 48190 + }, + { + "epoch": 17.8, + "learning_rate": 1.2211725581968223e-06, + "loss": 0.2064, + "step": 48191 + }, + { + "epoch": 17.8, + "learning_rate": 1.221135607833477e-06, + "loss": 0.1845, + "step": 48192 + }, + { + "epoch": 17.8, + "learning_rate": 1.2210986574701318e-06, + "loss": 0.1637, + "step": 48193 + }, + { + "epoch": 17.8, + "learning_rate": 1.2210617071067867e-06, + "loss": 0.2424, + "step": 48194 + }, + { + "epoch": 17.8, + "learning_rate": 1.2210247567434413e-06, + "loss": 0.2789, + "step": 48195 + }, + { + "epoch": 17.8, + "learning_rate": 1.2209878063800961e-06, + "loss": 0.235, + "step": 48196 + }, + { + "epoch": 17.8, + "learning_rate": 1.220950856016751e-06, + "loss": 0.2151, + "step": 48197 + }, + { + "epoch": 17.8, + "learning_rate": 1.2209139056534056e-06, + "loss": 0.2112, + "step": 48198 + }, + { + "epoch": 17.8, + "learning_rate": 1.2208769552900603e-06, + "loss": 0.1385, + "step": 48199 + }, + { + "epoch": 17.8, + "learning_rate": 1.2208400049267151e-06, + "loss": 0.148, + "step": 48200 + }, + { + "epoch": 17.8, + "learning_rate": 1.22080305456337e-06, + "loss": 0.301, + "step": 48201 + }, + { + "epoch": 17.8, + "learning_rate": 1.2207661042000246e-06, + "loss": 0.2122, + "step": 48202 + }, + { + "epoch": 17.8, + "learning_rate": 1.2207291538366794e-06, + "loss": 0.2901, + "step": 48203 + }, + { + "epoch": 17.8, + "learning_rate": 1.2206922034733343e-06, + "loss": 0.1969, + "step": 48204 + }, + { + "epoch": 17.8, + "learning_rate": 1.220655253109989e-06, + "loss": 0.2182, + "step": 48205 + }, + { + "epoch": 17.8, + "learning_rate": 1.2206183027466438e-06, + "loss": 0.246, + "step": 48206 + }, + { + "epoch": 17.8, + "learning_rate": 1.2205813523832986e-06, + "loss": 0.2192, + "step": 48207 + }, + { + "epoch": 17.8, + "learning_rate": 1.220544402019953e-06, + "loss": 0.1292, + "step": 48208 + }, + { + "epoch": 17.8, + "learning_rate": 1.220507451656608e-06, + "loss": 0.2729, + "step": 48209 + }, + { + "epoch": 17.8, + "learning_rate": 1.2204705012932628e-06, + "loss": 0.2386, + "step": 48210 + }, + { + "epoch": 17.8, + "learning_rate": 1.2204335509299174e-06, + "loss": 0.2594, + "step": 48211 + }, + { + "epoch": 17.8, + "learning_rate": 1.2203966005665722e-06, + "loss": 0.3365, + "step": 48212 + }, + { + "epoch": 17.8, + "learning_rate": 1.220359650203227e-06, + "loss": 0.1494, + "step": 48213 + }, + { + "epoch": 17.8, + "learning_rate": 1.2203226998398817e-06, + "loss": 0.2827, + "step": 48214 + }, + { + "epoch": 17.8, + "learning_rate": 1.2202857494765366e-06, + "loss": 0.14, + "step": 48215 + }, + { + "epoch": 17.81, + "learning_rate": 1.2202487991131914e-06, + "loss": 0.1986, + "step": 48216 + }, + { + "epoch": 17.81, + "learning_rate": 1.220211848749846e-06, + "loss": 0.2226, + "step": 48217 + }, + { + "epoch": 17.81, + "learning_rate": 1.2201748983865007e-06, + "loss": 0.2536, + "step": 48218 + }, + { + "epoch": 17.81, + "learning_rate": 1.2201379480231555e-06, + "loss": 0.1433, + "step": 48219 + }, + { + "epoch": 17.81, + "learning_rate": 1.2201009976598104e-06, + "loss": 0.1958, + "step": 48220 + }, + { + "epoch": 17.81, + "learning_rate": 1.220064047296465e-06, + "loss": 0.2194, + "step": 48221 + }, + { + "epoch": 17.81, + "learning_rate": 1.2200270969331199e-06, + "loss": 0.0989, + "step": 48222 + }, + { + "epoch": 17.81, + "learning_rate": 1.2199901465697747e-06, + "loss": 0.2045, + "step": 48223 + }, + { + "epoch": 17.81, + "learning_rate": 1.2199531962064294e-06, + "loss": 0.2038, + "step": 48224 + }, + { + "epoch": 17.81, + "learning_rate": 1.2199162458430842e-06, + "loss": 0.3541, + "step": 48225 + }, + { + "epoch": 17.81, + "learning_rate": 1.219879295479739e-06, + "loss": 0.1542, + "step": 48226 + }, + { + "epoch": 17.81, + "learning_rate": 1.2198423451163937e-06, + "loss": 0.2347, + "step": 48227 + }, + { + "epoch": 17.81, + "learning_rate": 1.2198053947530483e-06, + "loss": 0.1511, + "step": 48228 + }, + { + "epoch": 17.81, + "learning_rate": 1.2197684443897032e-06, + "loss": 0.1855, + "step": 48229 + }, + { + "epoch": 17.81, + "learning_rate": 1.219731494026358e-06, + "loss": 0.1466, + "step": 48230 + }, + { + "epoch": 17.81, + "learning_rate": 1.2196945436630127e-06, + "loss": 0.2325, + "step": 48231 + }, + { + "epoch": 17.81, + "learning_rate": 1.2196575932996675e-06, + "loss": 0.1985, + "step": 48232 + }, + { + "epoch": 17.81, + "learning_rate": 1.2196206429363224e-06, + "loss": 0.2706, + "step": 48233 + }, + { + "epoch": 17.81, + "learning_rate": 1.219583692572977e-06, + "loss": 0.1736, + "step": 48234 + }, + { + "epoch": 17.81, + "learning_rate": 1.2195467422096318e-06, + "loss": 0.2793, + "step": 48235 + }, + { + "epoch": 17.81, + "learning_rate": 1.2195097918462867e-06, + "loss": 0.2595, + "step": 48236 + }, + { + "epoch": 17.81, + "learning_rate": 1.2194728414829413e-06, + "loss": 0.1562, + "step": 48237 + }, + { + "epoch": 17.81, + "learning_rate": 1.219435891119596e-06, + "loss": 0.296, + "step": 48238 + }, + { + "epoch": 17.81, + "learning_rate": 1.2193989407562508e-06, + "loss": 0.1596, + "step": 48239 + }, + { + "epoch": 17.81, + "learning_rate": 1.2193619903929054e-06, + "loss": 0.1914, + "step": 48240 + }, + { + "epoch": 17.81, + "learning_rate": 1.2193250400295603e-06, + "loss": 0.2399, + "step": 48241 + }, + { + "epoch": 17.81, + "learning_rate": 1.2192880896662151e-06, + "loss": 0.3108, + "step": 48242 + }, + { + "epoch": 17.81, + "learning_rate": 1.2192511393028698e-06, + "loss": 0.2327, + "step": 48243 + }, + { + "epoch": 17.82, + "learning_rate": 1.2192141889395246e-06, + "loss": 0.2483, + "step": 48244 + }, + { + "epoch": 17.82, + "learning_rate": 1.2191772385761795e-06, + "loss": 0.2002, + "step": 48245 + }, + { + "epoch": 17.82, + "learning_rate": 1.2191402882128341e-06, + "loss": 0.1744, + "step": 48246 + }, + { + "epoch": 17.82, + "learning_rate": 1.219103337849489e-06, + "loss": 0.3298, + "step": 48247 + }, + { + "epoch": 17.82, + "learning_rate": 1.2190663874861436e-06, + "loss": 0.2374, + "step": 48248 + }, + { + "epoch": 17.82, + "learning_rate": 1.2190294371227984e-06, + "loss": 0.2962, + "step": 48249 + }, + { + "epoch": 17.82, + "learning_rate": 1.218992486759453e-06, + "loss": 0.1637, + "step": 48250 + }, + { + "epoch": 17.82, + "learning_rate": 1.218955536396108e-06, + "loss": 0.3301, + "step": 48251 + }, + { + "epoch": 17.82, + "learning_rate": 1.2189185860327628e-06, + "loss": 0.1833, + "step": 48252 + }, + { + "epoch": 17.82, + "learning_rate": 1.2188816356694174e-06, + "loss": 0.2237, + "step": 48253 + }, + { + "epoch": 17.82, + "learning_rate": 1.2188446853060723e-06, + "loss": 0.2908, + "step": 48254 + }, + { + "epoch": 17.82, + "learning_rate": 1.2188077349427271e-06, + "loss": 0.1737, + "step": 48255 + }, + { + "epoch": 17.82, + "learning_rate": 1.2187707845793817e-06, + "loss": 0.3933, + "step": 48256 + }, + { + "epoch": 17.82, + "learning_rate": 1.2187338342160366e-06, + "loss": 0.1472, + "step": 48257 + }, + { + "epoch": 17.82, + "learning_rate": 1.2186968838526912e-06, + "loss": 0.3799, + "step": 48258 + }, + { + "epoch": 17.82, + "learning_rate": 1.2186599334893459e-06, + "loss": 0.3875, + "step": 48259 + }, + { + "epoch": 17.82, + "learning_rate": 1.2186229831260007e-06, + "loss": 0.2836, + "step": 48260 + }, + { + "epoch": 17.82, + "learning_rate": 1.2185860327626556e-06, + "loss": 0.403, + "step": 48261 + }, + { + "epoch": 17.82, + "learning_rate": 1.2185490823993102e-06, + "loss": 0.379, + "step": 48262 + }, + { + "epoch": 17.82, + "learning_rate": 1.218512132035965e-06, + "loss": 0.2698, + "step": 48263 + }, + { + "epoch": 17.82, + "learning_rate": 1.21847518167262e-06, + "loss": 0.1391, + "step": 48264 + }, + { + "epoch": 17.82, + "learning_rate": 1.2184382313092745e-06, + "loss": 0.1867, + "step": 48265 + }, + { + "epoch": 17.82, + "learning_rate": 1.2184012809459294e-06, + "loss": 0.2897, + "step": 48266 + }, + { + "epoch": 17.82, + "learning_rate": 1.2183643305825842e-06, + "loss": 0.2002, + "step": 48267 + }, + { + "epoch": 17.82, + "learning_rate": 1.2183273802192389e-06, + "loss": 0.2584, + "step": 48268 + }, + { + "epoch": 17.82, + "learning_rate": 1.2182904298558935e-06, + "loss": 0.1871, + "step": 48269 + }, + { + "epoch": 17.82, + "learning_rate": 1.2182534794925483e-06, + "loss": 0.4324, + "step": 48270 + }, + { + "epoch": 17.83, + "learning_rate": 1.2182165291292032e-06, + "loss": 0.2587, + "step": 48271 + }, + { + "epoch": 17.83, + "learning_rate": 1.2181795787658578e-06, + "loss": 0.1654, + "step": 48272 + }, + { + "epoch": 17.83, + "learning_rate": 1.2181426284025127e-06, + "loss": 0.1628, + "step": 48273 + }, + { + "epoch": 17.83, + "learning_rate": 1.2181056780391675e-06, + "loss": 0.2533, + "step": 48274 + }, + { + "epoch": 17.83, + "learning_rate": 1.2180687276758222e-06, + "loss": 0.2491, + "step": 48275 + }, + { + "epoch": 17.83, + "learning_rate": 1.218031777312477e-06, + "loss": 0.215, + "step": 48276 + }, + { + "epoch": 17.83, + "learning_rate": 1.2179948269491319e-06, + "loss": 0.2645, + "step": 48277 + }, + { + "epoch": 17.83, + "learning_rate": 1.2179578765857863e-06, + "loss": 0.2576, + "step": 48278 + }, + { + "epoch": 17.83, + "learning_rate": 1.2179209262224411e-06, + "loss": 0.2682, + "step": 48279 + }, + { + "epoch": 17.83, + "learning_rate": 1.217883975859096e-06, + "loss": 0.1737, + "step": 48280 + }, + { + "epoch": 17.83, + "learning_rate": 1.2178470254957506e-06, + "loss": 0.1358, + "step": 48281 + }, + { + "epoch": 17.83, + "learning_rate": 1.2178100751324055e-06, + "loss": 0.15, + "step": 48282 + }, + { + "epoch": 17.83, + "learning_rate": 1.2177731247690603e-06, + "loss": 0.2152, + "step": 48283 + }, + { + "epoch": 17.83, + "learning_rate": 1.217736174405715e-06, + "loss": 0.2618, + "step": 48284 + }, + { + "epoch": 17.83, + "learning_rate": 1.2176992240423698e-06, + "loss": 0.2677, + "step": 48285 + }, + { + "epoch": 17.83, + "learning_rate": 1.2176622736790246e-06, + "loss": 0.2075, + "step": 48286 + }, + { + "epoch": 17.83, + "learning_rate": 1.2176253233156793e-06, + "loss": 0.4113, + "step": 48287 + }, + { + "epoch": 17.83, + "learning_rate": 1.217588372952334e-06, + "loss": 0.3625, + "step": 48288 + }, + { + "epoch": 17.83, + "learning_rate": 1.2175514225889888e-06, + "loss": 0.2215, + "step": 48289 + }, + { + "epoch": 17.83, + "learning_rate": 1.2175144722256436e-06, + "loss": 0.2079, + "step": 48290 + }, + { + "epoch": 17.83, + "learning_rate": 1.2174775218622983e-06, + "loss": 0.1866, + "step": 48291 + }, + { + "epoch": 17.83, + "learning_rate": 1.217440571498953e-06, + "loss": 0.1633, + "step": 48292 + }, + { + "epoch": 17.83, + "learning_rate": 1.217403621135608e-06, + "loss": 0.1787, + "step": 48293 + }, + { + "epoch": 17.83, + "learning_rate": 1.2173666707722626e-06, + "loss": 0.4672, + "step": 48294 + }, + { + "epoch": 17.83, + "learning_rate": 1.2173297204089174e-06, + "loss": 0.2586, + "step": 48295 + }, + { + "epoch": 17.83, + "learning_rate": 1.2172927700455723e-06, + "loss": 0.2215, + "step": 48296 + }, + { + "epoch": 17.83, + "learning_rate": 1.217255819682227e-06, + "loss": 0.1904, + "step": 48297 + }, + { + "epoch": 17.84, + "learning_rate": 1.2172188693188816e-06, + "loss": 0.2505, + "step": 48298 + }, + { + "epoch": 17.84, + "learning_rate": 1.2171819189555364e-06, + "loss": 0.2778, + "step": 48299 + }, + { + "epoch": 17.84, + "learning_rate": 1.2171449685921913e-06, + "loss": 0.2281, + "step": 48300 + }, + { + "epoch": 17.84, + "learning_rate": 1.2171080182288459e-06, + "loss": 0.1735, + "step": 48301 + }, + { + "epoch": 17.84, + "learning_rate": 1.2170710678655007e-06, + "loss": 0.1764, + "step": 48302 + }, + { + "epoch": 17.84, + "learning_rate": 1.2170341175021556e-06, + "loss": 0.2017, + "step": 48303 + }, + { + "epoch": 17.84, + "learning_rate": 1.2169971671388102e-06, + "loss": 0.1298, + "step": 48304 + }, + { + "epoch": 17.84, + "learning_rate": 1.216960216775465e-06, + "loss": 0.1987, + "step": 48305 + }, + { + "epoch": 17.84, + "learning_rate": 1.21692326641212e-06, + "loss": 0.1609, + "step": 48306 + }, + { + "epoch": 17.84, + "learning_rate": 1.2168863160487743e-06, + "loss": 0.3113, + "step": 48307 + }, + { + "epoch": 17.84, + "learning_rate": 1.2168493656854292e-06, + "loss": 0.2761, + "step": 48308 + }, + { + "epoch": 17.84, + "learning_rate": 1.216812415322084e-06, + "loss": 0.17, + "step": 48309 + }, + { + "epoch": 17.84, + "learning_rate": 1.2167754649587387e-06, + "loss": 0.2194, + "step": 48310 + }, + { + "epoch": 17.84, + "learning_rate": 1.2167385145953935e-06, + "loss": 0.2006, + "step": 48311 + }, + { + "epoch": 17.84, + "learning_rate": 1.2167015642320484e-06, + "loss": 0.2138, + "step": 48312 + }, + { + "epoch": 17.84, + "learning_rate": 1.216664613868703e-06, + "loss": 0.2643, + "step": 48313 + }, + { + "epoch": 17.84, + "learning_rate": 1.2166276635053579e-06, + "loss": 0.2259, + "step": 48314 + }, + { + "epoch": 17.84, + "learning_rate": 1.2165907131420127e-06, + "loss": 0.196, + "step": 48315 + }, + { + "epoch": 17.84, + "learning_rate": 1.2165537627786673e-06, + "loss": 0.3762, + "step": 48316 + }, + { + "epoch": 17.84, + "learning_rate": 1.216516812415322e-06, + "loss": 0.2966, + "step": 48317 + }, + { + "epoch": 17.84, + "learning_rate": 1.2164798620519768e-06, + "loss": 0.3178, + "step": 48318 + }, + { + "epoch": 17.84, + "learning_rate": 1.2164429116886317e-06, + "loss": 0.3199, + "step": 48319 + }, + { + "epoch": 17.84, + "learning_rate": 1.2164059613252863e-06, + "loss": 0.1549, + "step": 48320 + }, + { + "epoch": 17.84, + "learning_rate": 1.2163690109619412e-06, + "loss": 0.1852, + "step": 48321 + }, + { + "epoch": 17.84, + "learning_rate": 1.216332060598596e-06, + "loss": 0.1827, + "step": 48322 + }, + { + "epoch": 17.84, + "learning_rate": 1.2162951102352506e-06, + "loss": 0.2776, + "step": 48323 + }, + { + "epoch": 17.84, + "learning_rate": 1.2162581598719055e-06, + "loss": 0.2041, + "step": 48324 + }, + { + "epoch": 17.85, + "learning_rate": 1.2162212095085603e-06, + "loss": 0.1178, + "step": 48325 + }, + { + "epoch": 17.85, + "learning_rate": 1.216184259145215e-06, + "loss": 0.3377, + "step": 48326 + }, + { + "epoch": 17.85, + "learning_rate": 1.2161473087818696e-06, + "loss": 0.2058, + "step": 48327 + }, + { + "epoch": 17.85, + "learning_rate": 1.2161103584185245e-06, + "loss": 0.2072, + "step": 48328 + }, + { + "epoch": 17.85, + "learning_rate": 1.216073408055179e-06, + "loss": 0.2608, + "step": 48329 + }, + { + "epoch": 17.85, + "learning_rate": 1.216036457691834e-06, + "loss": 0.2641, + "step": 48330 + }, + { + "epoch": 17.85, + "learning_rate": 1.2159995073284888e-06, + "loss": 0.2839, + "step": 48331 + }, + { + "epoch": 17.85, + "learning_rate": 1.2159625569651434e-06, + "loss": 0.1273, + "step": 48332 + }, + { + "epoch": 17.85, + "learning_rate": 1.2159256066017983e-06, + "loss": 0.2118, + "step": 48333 + }, + { + "epoch": 17.85, + "learning_rate": 1.2158886562384531e-06, + "loss": 0.3545, + "step": 48334 + }, + { + "epoch": 17.85, + "learning_rate": 1.2158517058751078e-06, + "loss": 0.3912, + "step": 48335 + }, + { + "epoch": 17.85, + "learning_rate": 1.2158147555117626e-06, + "loss": 0.2327, + "step": 48336 + }, + { + "epoch": 17.85, + "learning_rate": 1.2157778051484172e-06, + "loss": 0.241, + "step": 48337 + }, + { + "epoch": 17.85, + "learning_rate": 1.215740854785072e-06, + "loss": 0.3059, + "step": 48338 + }, + { + "epoch": 17.85, + "learning_rate": 1.2157039044217267e-06, + "loss": 0.3605, + "step": 48339 + }, + { + "epoch": 17.85, + "learning_rate": 1.2156669540583816e-06, + "loss": 0.3132, + "step": 48340 + }, + { + "epoch": 17.85, + "learning_rate": 1.2156300036950364e-06, + "loss": 0.2565, + "step": 48341 + }, + { + "epoch": 17.85, + "learning_rate": 1.215593053331691e-06, + "loss": 0.2522, + "step": 48342 + }, + { + "epoch": 17.85, + "learning_rate": 1.215556102968346e-06, + "loss": 0.172, + "step": 48343 + }, + { + "epoch": 17.85, + "learning_rate": 1.2155191526050008e-06, + "loss": 0.2379, + "step": 48344 + }, + { + "epoch": 17.85, + "learning_rate": 1.2154822022416554e-06, + "loss": 0.2247, + "step": 48345 + }, + { + "epoch": 17.85, + "learning_rate": 1.2154452518783102e-06, + "loss": 0.1746, + "step": 48346 + }, + { + "epoch": 17.85, + "learning_rate": 1.2154083015149649e-06, + "loss": 0.2493, + "step": 48347 + }, + { + "epoch": 17.85, + "learning_rate": 1.2153713511516195e-06, + "loss": 0.2959, + "step": 48348 + }, + { + "epoch": 17.85, + "learning_rate": 1.2153344007882744e-06, + "loss": 0.156, + "step": 48349 + }, + { + "epoch": 17.85, + "learning_rate": 1.2152974504249292e-06, + "loss": 0.1546, + "step": 48350 + }, + { + "epoch": 17.85, + "learning_rate": 1.2152605000615839e-06, + "loss": 0.2531, + "step": 48351 + }, + { + "epoch": 17.86, + "learning_rate": 1.2152235496982387e-06, + "loss": 0.1367, + "step": 48352 + }, + { + "epoch": 17.86, + "learning_rate": 1.2151865993348935e-06, + "loss": 0.2769, + "step": 48353 + }, + { + "epoch": 17.86, + "learning_rate": 1.2151496489715482e-06, + "loss": 0.2627, + "step": 48354 + }, + { + "epoch": 17.86, + "learning_rate": 1.215112698608203e-06, + "loss": 0.2343, + "step": 48355 + }, + { + "epoch": 17.86, + "learning_rate": 1.2150757482448579e-06, + "loss": 0.2017, + "step": 48356 + }, + { + "epoch": 17.86, + "learning_rate": 1.2150387978815125e-06, + "loss": 0.3721, + "step": 48357 + }, + { + "epoch": 17.86, + "learning_rate": 1.2150018475181672e-06, + "loss": 0.1577, + "step": 48358 + }, + { + "epoch": 17.86, + "learning_rate": 1.214964897154822e-06, + "loss": 0.2289, + "step": 48359 + }, + { + "epoch": 17.86, + "learning_rate": 1.2149279467914769e-06, + "loss": 0.1852, + "step": 48360 + }, + { + "epoch": 17.86, + "learning_rate": 1.2148909964281315e-06, + "loss": 0.3201, + "step": 48361 + }, + { + "epoch": 17.86, + "learning_rate": 1.2148540460647863e-06, + "loss": 0.1626, + "step": 48362 + }, + { + "epoch": 17.86, + "learning_rate": 1.2148170957014412e-06, + "loss": 0.2592, + "step": 48363 + }, + { + "epoch": 17.86, + "learning_rate": 1.2147801453380958e-06, + "loss": 0.277, + "step": 48364 + }, + { + "epoch": 17.86, + "learning_rate": 1.2147431949747507e-06, + "loss": 0.282, + "step": 48365 + }, + { + "epoch": 17.86, + "learning_rate": 1.2147062446114055e-06, + "loss": 0.4317, + "step": 48366 + }, + { + "epoch": 17.86, + "learning_rate": 1.2146692942480602e-06, + "loss": 0.1466, + "step": 48367 + }, + { + "epoch": 17.86, + "learning_rate": 1.2146323438847148e-06, + "loss": 0.1952, + "step": 48368 + }, + { + "epoch": 17.86, + "learning_rate": 1.2145953935213696e-06, + "loss": 0.2138, + "step": 48369 + }, + { + "epoch": 17.86, + "learning_rate": 1.2145584431580245e-06, + "loss": 0.2155, + "step": 48370 + }, + { + "epoch": 17.86, + "learning_rate": 1.2145214927946791e-06, + "loss": 0.16, + "step": 48371 + }, + { + "epoch": 17.86, + "learning_rate": 1.214484542431334e-06, + "loss": 0.2043, + "step": 48372 + }, + { + "epoch": 17.86, + "learning_rate": 1.2144475920679888e-06, + "loss": 0.1916, + "step": 48373 + }, + { + "epoch": 17.86, + "learning_rate": 1.2144106417046435e-06, + "loss": 0.185, + "step": 48374 + }, + { + "epoch": 17.86, + "learning_rate": 1.2143736913412983e-06, + "loss": 0.2608, + "step": 48375 + }, + { + "epoch": 17.86, + "learning_rate": 1.2143367409779532e-06, + "loss": 0.2772, + "step": 48376 + }, + { + "epoch": 17.86, + "learning_rate": 1.2142997906146076e-06, + "loss": 0.295, + "step": 48377 + }, + { + "epoch": 17.86, + "learning_rate": 1.2142628402512624e-06, + "loss": 0.1054, + "step": 48378 + }, + { + "epoch": 17.87, + "learning_rate": 1.2142258898879173e-06, + "loss": 0.1937, + "step": 48379 + }, + { + "epoch": 17.87, + "learning_rate": 1.214188939524572e-06, + "loss": 0.3175, + "step": 48380 + }, + { + "epoch": 17.87, + "learning_rate": 1.2141519891612268e-06, + "loss": 0.2533, + "step": 48381 + }, + { + "epoch": 17.87, + "learning_rate": 1.2141150387978816e-06, + "loss": 0.217, + "step": 48382 + }, + { + "epoch": 17.87, + "learning_rate": 1.2140780884345362e-06, + "loss": 0.2051, + "step": 48383 + }, + { + "epoch": 17.87, + "learning_rate": 1.214041138071191e-06, + "loss": 0.2549, + "step": 48384 + }, + { + "epoch": 17.87, + "learning_rate": 1.214004187707846e-06, + "loss": 0.2063, + "step": 48385 + }, + { + "epoch": 17.87, + "learning_rate": 1.2139672373445006e-06, + "loss": 0.1688, + "step": 48386 + }, + { + "epoch": 17.87, + "learning_rate": 1.2139302869811552e-06, + "loss": 0.2431, + "step": 48387 + }, + { + "epoch": 17.87, + "learning_rate": 1.21389333661781e-06, + "loss": 0.2689, + "step": 48388 + }, + { + "epoch": 17.87, + "learning_rate": 1.213856386254465e-06, + "loss": 0.3378, + "step": 48389 + }, + { + "epoch": 17.87, + "learning_rate": 1.2138194358911195e-06, + "loss": 0.167, + "step": 48390 + }, + { + "epoch": 17.87, + "learning_rate": 1.2137824855277744e-06, + "loss": 0.3272, + "step": 48391 + }, + { + "epoch": 17.87, + "learning_rate": 1.2137455351644292e-06, + "loss": 0.257, + "step": 48392 + }, + { + "epoch": 17.87, + "learning_rate": 1.2137085848010839e-06, + "loss": 0.2263, + "step": 48393 + }, + { + "epoch": 17.87, + "learning_rate": 1.2136716344377387e-06, + "loss": 0.1845, + "step": 48394 + }, + { + "epoch": 17.87, + "learning_rate": 1.2136346840743936e-06, + "loss": 0.227, + "step": 48395 + }, + { + "epoch": 17.87, + "learning_rate": 1.2135977337110482e-06, + "loss": 0.31, + "step": 48396 + }, + { + "epoch": 17.87, + "learning_rate": 1.2135607833477028e-06, + "loss": 0.2922, + "step": 48397 + }, + { + "epoch": 17.87, + "learning_rate": 1.2135238329843577e-06, + "loss": 0.1758, + "step": 48398 + }, + { + "epoch": 17.87, + "learning_rate": 1.2134868826210123e-06, + "loss": 0.1342, + "step": 48399 + }, + { + "epoch": 17.87, + "learning_rate": 1.2134499322576672e-06, + "loss": 0.4152, + "step": 48400 + }, + { + "epoch": 17.87, + "learning_rate": 1.213412981894322e-06, + "loss": 0.2654, + "step": 48401 + }, + { + "epoch": 17.87, + "learning_rate": 1.2133760315309767e-06, + "loss": 0.2166, + "step": 48402 + }, + { + "epoch": 17.87, + "learning_rate": 1.2133390811676315e-06, + "loss": 0.3115, + "step": 48403 + }, + { + "epoch": 17.87, + "learning_rate": 1.2133021308042864e-06, + "loss": 0.2084, + "step": 48404 + }, + { + "epoch": 17.87, + "learning_rate": 1.213265180440941e-06, + "loss": 0.2783, + "step": 48405 + }, + { + "epoch": 17.88, + "learning_rate": 1.2132282300775958e-06, + "loss": 0.4119, + "step": 48406 + }, + { + "epoch": 17.88, + "learning_rate": 1.2131912797142505e-06, + "loss": 0.3148, + "step": 48407 + }, + { + "epoch": 17.88, + "learning_rate": 1.2131543293509053e-06, + "loss": 0.245, + "step": 48408 + }, + { + "epoch": 17.88, + "learning_rate": 1.21311737898756e-06, + "loss": 0.1557, + "step": 48409 + }, + { + "epoch": 17.88, + "learning_rate": 1.2130804286242148e-06, + "loss": 0.3777, + "step": 48410 + }, + { + "epoch": 17.88, + "learning_rate": 1.2130434782608697e-06, + "loss": 0.2078, + "step": 48411 + }, + { + "epoch": 17.88, + "learning_rate": 1.2130065278975243e-06, + "loss": 0.3643, + "step": 48412 + }, + { + "epoch": 17.88, + "learning_rate": 1.2129695775341791e-06, + "loss": 0.1534, + "step": 48413 + }, + { + "epoch": 17.88, + "learning_rate": 1.212932627170834e-06, + "loss": 0.2546, + "step": 48414 + }, + { + "epoch": 17.88, + "learning_rate": 1.2128956768074886e-06, + "loss": 0.2624, + "step": 48415 + }, + { + "epoch": 17.88, + "learning_rate": 1.2128587264441435e-06, + "loss": 0.1948, + "step": 48416 + }, + { + "epoch": 17.88, + "learning_rate": 1.2128217760807981e-06, + "loss": 0.3658, + "step": 48417 + }, + { + "epoch": 17.88, + "learning_rate": 1.212784825717453e-06, + "loss": 0.267, + "step": 48418 + }, + { + "epoch": 17.88, + "learning_rate": 1.2127478753541076e-06, + "loss": 0.2629, + "step": 48419 + }, + { + "epoch": 17.88, + "learning_rate": 1.2127109249907624e-06, + "loss": 0.2393, + "step": 48420 + }, + { + "epoch": 17.88, + "learning_rate": 1.2126739746274173e-06, + "loss": 0.2591, + "step": 48421 + }, + { + "epoch": 17.88, + "learning_rate": 1.212637024264072e-06, + "loss": 0.2227, + "step": 48422 + }, + { + "epoch": 17.88, + "learning_rate": 1.2126000739007268e-06, + "loss": 0.1972, + "step": 48423 + }, + { + "epoch": 17.88, + "learning_rate": 1.2125631235373814e-06, + "loss": 0.3885, + "step": 48424 + }, + { + "epoch": 17.88, + "learning_rate": 1.2125261731740363e-06, + "loss": 0.2354, + "step": 48425 + }, + { + "epoch": 17.88, + "learning_rate": 1.2124892228106911e-06, + "loss": 0.2995, + "step": 48426 + }, + { + "epoch": 17.88, + "learning_rate": 1.2124522724473458e-06, + "loss": 0.2569, + "step": 48427 + }, + { + "epoch": 17.88, + "learning_rate": 1.2124153220840004e-06, + "loss": 0.1762, + "step": 48428 + }, + { + "epoch": 17.88, + "learning_rate": 1.2123783717206552e-06, + "loss": 0.2608, + "step": 48429 + }, + { + "epoch": 17.88, + "learning_rate": 1.21234142135731e-06, + "loss": 0.2386, + "step": 48430 + }, + { + "epoch": 17.88, + "learning_rate": 1.2123044709939647e-06, + "loss": 0.2287, + "step": 48431 + }, + { + "epoch": 17.88, + "learning_rate": 1.2122675206306196e-06, + "loss": 0.1532, + "step": 48432 + }, + { + "epoch": 17.89, + "learning_rate": 1.2122305702672744e-06, + "loss": 0.1796, + "step": 48433 + }, + { + "epoch": 17.89, + "learning_rate": 1.212193619903929e-06, + "loss": 0.1672, + "step": 48434 + }, + { + "epoch": 17.89, + "learning_rate": 1.212156669540584e-06, + "loss": 0.2098, + "step": 48435 + }, + { + "epoch": 17.89, + "learning_rate": 1.2121197191772387e-06, + "loss": 0.1388, + "step": 48436 + }, + { + "epoch": 17.89, + "learning_rate": 1.2120827688138934e-06, + "loss": 0.2801, + "step": 48437 + }, + { + "epoch": 17.89, + "learning_rate": 1.212045818450548e-06, + "loss": 0.1683, + "step": 48438 + }, + { + "epoch": 17.89, + "learning_rate": 1.2120088680872029e-06, + "loss": 0.1948, + "step": 48439 + }, + { + "epoch": 17.89, + "learning_rate": 1.2119719177238577e-06, + "loss": 0.2816, + "step": 48440 + }, + { + "epoch": 17.89, + "learning_rate": 1.2119349673605124e-06, + "loss": 0.1634, + "step": 48441 + }, + { + "epoch": 17.89, + "learning_rate": 1.2118980169971672e-06, + "loss": 0.2146, + "step": 48442 + }, + { + "epoch": 17.89, + "learning_rate": 1.211861066633822e-06, + "loss": 0.1951, + "step": 48443 + }, + { + "epoch": 17.89, + "learning_rate": 1.2118241162704767e-06, + "loss": 0.3048, + "step": 48444 + }, + { + "epoch": 17.89, + "learning_rate": 1.2117871659071315e-06, + "loss": 0.3093, + "step": 48445 + }, + { + "epoch": 17.89, + "learning_rate": 1.2117502155437864e-06, + "loss": 0.243, + "step": 48446 + }, + { + "epoch": 17.89, + "learning_rate": 1.2117132651804408e-06, + "loss": 0.3099, + "step": 48447 + }, + { + "epoch": 17.89, + "learning_rate": 1.2116763148170957e-06, + "loss": 0.271, + "step": 48448 + }, + { + "epoch": 17.89, + "learning_rate": 1.2116393644537505e-06, + "loss": 0.1937, + "step": 48449 + }, + { + "epoch": 17.89, + "learning_rate": 1.2116024140904051e-06, + "loss": 0.2427, + "step": 48450 + }, + { + "epoch": 17.89, + "learning_rate": 1.21156546372706e-06, + "loss": 0.2767, + "step": 48451 + }, + { + "epoch": 17.89, + "learning_rate": 1.2115285133637148e-06, + "loss": 0.2155, + "step": 48452 + }, + { + "epoch": 17.89, + "learning_rate": 1.2114915630003695e-06, + "loss": 0.2539, + "step": 48453 + }, + { + "epoch": 17.89, + "learning_rate": 1.2114546126370243e-06, + "loss": 0.1898, + "step": 48454 + }, + { + "epoch": 17.89, + "learning_rate": 1.2114176622736792e-06, + "loss": 0.1728, + "step": 48455 + }, + { + "epoch": 17.89, + "learning_rate": 1.2113807119103338e-06, + "loss": 0.2608, + "step": 48456 + }, + { + "epoch": 17.89, + "learning_rate": 1.2113437615469884e-06, + "loss": 0.1765, + "step": 48457 + }, + { + "epoch": 17.89, + "learning_rate": 1.2113068111836433e-06, + "loss": 0.1064, + "step": 48458 + }, + { + "epoch": 17.89, + "learning_rate": 1.2112698608202981e-06, + "loss": 0.2603, + "step": 48459 + }, + { + "epoch": 17.9, + "learning_rate": 1.2112329104569528e-06, + "loss": 0.3448, + "step": 48460 + }, + { + "epoch": 17.9, + "learning_rate": 1.2111959600936076e-06, + "loss": 0.165, + "step": 48461 + }, + { + "epoch": 17.9, + "learning_rate": 1.2111590097302625e-06, + "loss": 0.2212, + "step": 48462 + }, + { + "epoch": 17.9, + "learning_rate": 1.2111220593669171e-06, + "loss": 0.2034, + "step": 48463 + }, + { + "epoch": 17.9, + "learning_rate": 1.211085109003572e-06, + "loss": 0.1296, + "step": 48464 + }, + { + "epoch": 17.9, + "learning_rate": 1.2110481586402268e-06, + "loss": 0.2636, + "step": 48465 + }, + { + "epoch": 17.9, + "learning_rate": 1.2110112082768814e-06, + "loss": 0.3228, + "step": 48466 + }, + { + "epoch": 17.9, + "learning_rate": 1.210974257913536e-06, + "loss": 0.331, + "step": 48467 + }, + { + "epoch": 17.9, + "learning_rate": 1.210937307550191e-06, + "loss": 0.282, + "step": 48468 + }, + { + "epoch": 17.9, + "learning_rate": 1.2109003571868456e-06, + "loss": 0.3299, + "step": 48469 + }, + { + "epoch": 17.9, + "learning_rate": 1.2108634068235004e-06, + "loss": 0.1797, + "step": 48470 + }, + { + "epoch": 17.9, + "learning_rate": 1.2108264564601553e-06, + "loss": 0.2037, + "step": 48471 + }, + { + "epoch": 17.9, + "learning_rate": 1.21078950609681e-06, + "loss": 0.24, + "step": 48472 + }, + { + "epoch": 17.9, + "learning_rate": 1.2107525557334647e-06, + "loss": 0.1669, + "step": 48473 + }, + { + "epoch": 17.9, + "learning_rate": 1.2107156053701196e-06, + "loss": 0.3423, + "step": 48474 + }, + { + "epoch": 17.9, + "learning_rate": 1.2106786550067742e-06, + "loss": 0.2481, + "step": 48475 + }, + { + "epoch": 17.9, + "learning_rate": 1.210641704643429e-06, + "loss": 0.3516, + "step": 48476 + }, + { + "epoch": 17.9, + "learning_rate": 1.2106047542800837e-06, + "loss": 0.3239, + "step": 48477 + }, + { + "epoch": 17.9, + "learning_rate": 1.2105678039167386e-06, + "loss": 0.3296, + "step": 48478 + }, + { + "epoch": 17.9, + "learning_rate": 1.2105308535533932e-06, + "loss": 0.2865, + "step": 48479 + }, + { + "epoch": 17.9, + "learning_rate": 1.210493903190048e-06, + "loss": 0.2374, + "step": 48480 + }, + { + "epoch": 17.9, + "learning_rate": 1.210456952826703e-06, + "loss": 0.3908, + "step": 48481 + }, + { + "epoch": 17.9, + "learning_rate": 1.2104200024633575e-06, + "loss": 0.2709, + "step": 48482 + }, + { + "epoch": 17.9, + "learning_rate": 1.2103830521000124e-06, + "loss": 0.3022, + "step": 48483 + }, + { + "epoch": 17.9, + "learning_rate": 1.2103461017366672e-06, + "loss": 0.2467, + "step": 48484 + }, + { + "epoch": 17.9, + "learning_rate": 1.2103091513733219e-06, + "loss": 0.1712, + "step": 48485 + }, + { + "epoch": 17.9, + "learning_rate": 1.2102722010099767e-06, + "loss": 0.2028, + "step": 48486 + }, + { + "epoch": 17.91, + "learning_rate": 1.2102352506466313e-06, + "loss": 0.1734, + "step": 48487 + }, + { + "epoch": 17.91, + "learning_rate": 1.2101983002832862e-06, + "loss": 0.253, + "step": 48488 + }, + { + "epoch": 17.91, + "learning_rate": 1.2101613499199408e-06, + "loss": 0.2222, + "step": 48489 + }, + { + "epoch": 17.91, + "learning_rate": 1.2101243995565957e-06, + "loss": 0.2728, + "step": 48490 + }, + { + "epoch": 17.91, + "learning_rate": 1.2100874491932505e-06, + "loss": 0.3806, + "step": 48491 + }, + { + "epoch": 17.91, + "learning_rate": 1.2100504988299052e-06, + "loss": 0.2509, + "step": 48492 + }, + { + "epoch": 17.91, + "learning_rate": 1.21001354846656e-06, + "loss": 0.28, + "step": 48493 + }, + { + "epoch": 17.91, + "learning_rate": 1.2099765981032149e-06, + "loss": 0.2196, + "step": 48494 + }, + { + "epoch": 17.91, + "learning_rate": 1.2099396477398695e-06, + "loss": 0.1488, + "step": 48495 + }, + { + "epoch": 17.91, + "learning_rate": 1.2099026973765243e-06, + "loss": 0.2271, + "step": 48496 + }, + { + "epoch": 17.91, + "learning_rate": 1.209865747013179e-06, + "loss": 0.1959, + "step": 48497 + }, + { + "epoch": 17.91, + "learning_rate": 1.2098287966498336e-06, + "loss": 0.1684, + "step": 48498 + }, + { + "epoch": 17.91, + "learning_rate": 1.2097918462864885e-06, + "loss": 0.2611, + "step": 48499 + }, + { + "epoch": 17.91, + "learning_rate": 1.2097548959231433e-06, + "loss": 0.3517, + "step": 48500 + }, + { + "epoch": 17.91, + "learning_rate": 1.209717945559798e-06, + "loss": 0.1983, + "step": 48501 + }, + { + "epoch": 17.91, + "learning_rate": 1.2096809951964528e-06, + "loss": 0.2169, + "step": 48502 + }, + { + "epoch": 17.91, + "learning_rate": 1.2096440448331076e-06, + "loss": 0.2689, + "step": 48503 + }, + { + "epoch": 17.91, + "learning_rate": 1.2096070944697623e-06, + "loss": 0.1526, + "step": 48504 + }, + { + "epoch": 17.91, + "learning_rate": 1.2095701441064171e-06, + "loss": 0.1839, + "step": 48505 + }, + { + "epoch": 17.91, + "learning_rate": 1.209533193743072e-06, + "loss": 0.2153, + "step": 48506 + }, + { + "epoch": 17.91, + "learning_rate": 1.2094962433797266e-06, + "loss": 0.3007, + "step": 48507 + }, + { + "epoch": 17.91, + "learning_rate": 1.2094592930163813e-06, + "loss": 0.3261, + "step": 48508 + }, + { + "epoch": 17.91, + "learning_rate": 1.209422342653036e-06, + "loss": 0.3456, + "step": 48509 + }, + { + "epoch": 17.91, + "learning_rate": 1.209385392289691e-06, + "loss": 0.1672, + "step": 48510 + }, + { + "epoch": 17.91, + "learning_rate": 1.2093484419263456e-06, + "loss": 0.3449, + "step": 48511 + }, + { + "epoch": 17.91, + "learning_rate": 1.2093114915630004e-06, + "loss": 0.2817, + "step": 48512 + }, + { + "epoch": 17.91, + "learning_rate": 1.2092745411996553e-06, + "loss": 0.244, + "step": 48513 + }, + { + "epoch": 17.92, + "learning_rate": 1.20923759083631e-06, + "loss": 0.1623, + "step": 48514 + }, + { + "epoch": 17.92, + "learning_rate": 1.2092006404729648e-06, + "loss": 0.237, + "step": 48515 + }, + { + "epoch": 17.92, + "learning_rate": 1.2091636901096196e-06, + "loss": 0.2077, + "step": 48516 + }, + { + "epoch": 17.92, + "learning_rate": 1.209126739746274e-06, + "loss": 0.3522, + "step": 48517 + }, + { + "epoch": 17.92, + "learning_rate": 1.2090897893829289e-06, + "loss": 0.1184, + "step": 48518 + }, + { + "epoch": 17.92, + "learning_rate": 1.2090528390195837e-06, + "loss": 0.2649, + "step": 48519 + }, + { + "epoch": 17.92, + "learning_rate": 1.2090158886562384e-06, + "loss": 0.2273, + "step": 48520 + }, + { + "epoch": 17.92, + "learning_rate": 1.2089789382928932e-06, + "loss": 0.1894, + "step": 48521 + }, + { + "epoch": 17.92, + "learning_rate": 1.208941987929548e-06, + "loss": 0.2007, + "step": 48522 + }, + { + "epoch": 17.92, + "learning_rate": 1.2089050375662027e-06, + "loss": 0.2241, + "step": 48523 + }, + { + "epoch": 17.92, + "learning_rate": 1.2088680872028576e-06, + "loss": 0.3415, + "step": 48524 + }, + { + "epoch": 17.92, + "learning_rate": 1.2088311368395124e-06, + "loss": 0.1817, + "step": 48525 + }, + { + "epoch": 17.92, + "learning_rate": 1.208794186476167e-06, + "loss": 0.1975, + "step": 48526 + }, + { + "epoch": 17.92, + "learning_rate": 1.2087572361128217e-06, + "loss": 0.238, + "step": 48527 + }, + { + "epoch": 17.92, + "learning_rate": 1.2087202857494765e-06, + "loss": 0.2737, + "step": 48528 + }, + { + "epoch": 17.92, + "learning_rate": 1.2086833353861314e-06, + "loss": 0.2044, + "step": 48529 + }, + { + "epoch": 17.92, + "learning_rate": 1.208646385022786e-06, + "loss": 0.3031, + "step": 48530 + }, + { + "epoch": 17.92, + "learning_rate": 1.2086094346594409e-06, + "loss": 0.1526, + "step": 48531 + }, + { + "epoch": 17.92, + "learning_rate": 1.2085724842960957e-06, + "loss": 0.1918, + "step": 48532 + }, + { + "epoch": 17.92, + "learning_rate": 1.2085355339327503e-06, + "loss": 0.2235, + "step": 48533 + }, + { + "epoch": 17.92, + "learning_rate": 1.2084985835694052e-06, + "loss": 0.2602, + "step": 48534 + }, + { + "epoch": 17.92, + "learning_rate": 1.20846163320606e-06, + "loss": 0.2515, + "step": 48535 + }, + { + "epoch": 17.92, + "learning_rate": 1.2084246828427145e-06, + "loss": 0.2772, + "step": 48536 + }, + { + "epoch": 17.92, + "learning_rate": 1.2083877324793693e-06, + "loss": 0.249, + "step": 48537 + }, + { + "epoch": 17.92, + "learning_rate": 1.2083507821160242e-06, + "loss": 0.3006, + "step": 48538 + }, + { + "epoch": 17.92, + "learning_rate": 1.2083138317526788e-06, + "loss": 0.2758, + "step": 48539 + }, + { + "epoch": 17.92, + "learning_rate": 1.2082768813893336e-06, + "loss": 0.2623, + "step": 48540 + }, + { + "epoch": 17.93, + "learning_rate": 1.2082399310259885e-06, + "loss": 0.1818, + "step": 48541 + }, + { + "epoch": 17.93, + "learning_rate": 1.2082029806626431e-06, + "loss": 0.2159, + "step": 48542 + }, + { + "epoch": 17.93, + "learning_rate": 1.208166030299298e-06, + "loss": 0.3169, + "step": 48543 + }, + { + "epoch": 17.93, + "learning_rate": 1.2081290799359528e-06, + "loss": 0.3487, + "step": 48544 + }, + { + "epoch": 17.93, + "learning_rate": 1.2080921295726075e-06, + "loss": 0.2364, + "step": 48545 + }, + { + "epoch": 17.93, + "learning_rate": 1.208055179209262e-06, + "loss": 0.2298, + "step": 48546 + }, + { + "epoch": 17.93, + "learning_rate": 1.208018228845917e-06, + "loss": 0.1246, + "step": 48547 + }, + { + "epoch": 17.93, + "learning_rate": 1.2079812784825718e-06, + "loss": 0.1167, + "step": 48548 + }, + { + "epoch": 17.93, + "learning_rate": 1.2079443281192264e-06, + "loss": 0.4158, + "step": 48549 + }, + { + "epoch": 17.93, + "learning_rate": 1.2079073777558813e-06, + "loss": 0.1737, + "step": 48550 + }, + { + "epoch": 17.93, + "learning_rate": 1.2078704273925361e-06, + "loss": 0.2468, + "step": 48551 + }, + { + "epoch": 17.93, + "learning_rate": 1.2078334770291908e-06, + "loss": 0.2031, + "step": 48552 + }, + { + "epoch": 17.93, + "learning_rate": 1.2077965266658456e-06, + "loss": 0.2235, + "step": 48553 + }, + { + "epoch": 17.93, + "learning_rate": 1.2077595763025005e-06, + "loss": 0.1186, + "step": 48554 + }, + { + "epoch": 17.93, + "learning_rate": 1.207722625939155e-06, + "loss": 0.3117, + "step": 48555 + }, + { + "epoch": 17.93, + "learning_rate": 1.2076856755758097e-06, + "loss": 0.255, + "step": 48556 + }, + { + "epoch": 17.93, + "learning_rate": 1.2076487252124646e-06, + "loss": 0.224, + "step": 48557 + }, + { + "epoch": 17.93, + "learning_rate": 1.2076117748491194e-06, + "loss": 0.3557, + "step": 48558 + }, + { + "epoch": 17.93, + "learning_rate": 1.207574824485774e-06, + "loss": 0.2003, + "step": 48559 + }, + { + "epoch": 17.93, + "learning_rate": 1.207537874122429e-06, + "loss": 0.1511, + "step": 48560 + }, + { + "epoch": 17.93, + "learning_rate": 1.2075009237590838e-06, + "loss": 0.1996, + "step": 48561 + }, + { + "epoch": 17.93, + "learning_rate": 1.2074639733957384e-06, + "loss": 0.2096, + "step": 48562 + }, + { + "epoch": 17.93, + "learning_rate": 1.2074270230323932e-06, + "loss": 0.2105, + "step": 48563 + }, + { + "epoch": 17.93, + "learning_rate": 1.207390072669048e-06, + "loss": 0.2427, + "step": 48564 + }, + { + "epoch": 17.93, + "learning_rate": 1.2073531223057027e-06, + "loss": 0.2253, + "step": 48565 + }, + { + "epoch": 17.93, + "learning_rate": 1.2073161719423574e-06, + "loss": 0.2015, + "step": 48566 + }, + { + "epoch": 17.93, + "learning_rate": 1.2072792215790122e-06, + "loss": 0.3258, + "step": 48567 + }, + { + "epoch": 17.94, + "learning_rate": 1.2072422712156669e-06, + "loss": 0.2044, + "step": 48568 + }, + { + "epoch": 17.94, + "learning_rate": 1.2072053208523217e-06, + "loss": 0.3374, + "step": 48569 + }, + { + "epoch": 17.94, + "learning_rate": 1.2071683704889765e-06, + "loss": 0.2112, + "step": 48570 + }, + { + "epoch": 17.94, + "learning_rate": 1.2071314201256312e-06, + "loss": 0.2574, + "step": 48571 + }, + { + "epoch": 17.94, + "learning_rate": 1.207094469762286e-06, + "loss": 0.1514, + "step": 48572 + }, + { + "epoch": 17.94, + "learning_rate": 1.2070575193989409e-06, + "loss": 0.1841, + "step": 48573 + }, + { + "epoch": 17.94, + "learning_rate": 1.2070205690355955e-06, + "loss": 0.2814, + "step": 48574 + }, + { + "epoch": 17.94, + "learning_rate": 1.2069836186722504e-06, + "loss": 0.1296, + "step": 48575 + }, + { + "epoch": 17.94, + "learning_rate": 1.206946668308905e-06, + "loss": 0.3338, + "step": 48576 + }, + { + "epoch": 17.94, + "learning_rate": 1.2069097179455599e-06, + "loss": 0.3105, + "step": 48577 + }, + { + "epoch": 17.94, + "learning_rate": 1.2068727675822145e-06, + "loss": 0.1577, + "step": 48578 + }, + { + "epoch": 17.94, + "learning_rate": 1.2068358172188693e-06, + "loss": 0.1948, + "step": 48579 + }, + { + "epoch": 17.94, + "learning_rate": 1.2067988668555242e-06, + "loss": 0.4405, + "step": 48580 + }, + { + "epoch": 17.94, + "learning_rate": 1.2067619164921788e-06, + "loss": 0.222, + "step": 48581 + }, + { + "epoch": 17.94, + "learning_rate": 1.2067249661288337e-06, + "loss": 0.3316, + "step": 48582 + }, + { + "epoch": 17.94, + "learning_rate": 1.2066880157654885e-06, + "loss": 0.2128, + "step": 48583 + }, + { + "epoch": 17.94, + "learning_rate": 1.2066510654021432e-06, + "loss": 0.2765, + "step": 48584 + }, + { + "epoch": 17.94, + "learning_rate": 1.206614115038798e-06, + "loss": 0.2813, + "step": 48585 + }, + { + "epoch": 17.94, + "learning_rate": 1.2065771646754526e-06, + "loss": 0.186, + "step": 48586 + }, + { + "epoch": 17.94, + "learning_rate": 1.2065402143121073e-06, + "loss": 0.2214, + "step": 48587 + }, + { + "epoch": 17.94, + "learning_rate": 1.2065032639487621e-06, + "loss": 0.2673, + "step": 48588 + }, + { + "epoch": 17.94, + "learning_rate": 1.206466313585417e-06, + "loss": 0.2659, + "step": 48589 + }, + { + "epoch": 17.94, + "learning_rate": 1.2064293632220716e-06, + "loss": 0.2546, + "step": 48590 + }, + { + "epoch": 17.94, + "learning_rate": 1.2063924128587265e-06, + "loss": 0.1406, + "step": 48591 + }, + { + "epoch": 17.94, + "learning_rate": 1.2063554624953813e-06, + "loss": 0.2605, + "step": 48592 + }, + { + "epoch": 17.94, + "learning_rate": 1.206318512132036e-06, + "loss": 0.3232, + "step": 48593 + }, + { + "epoch": 17.94, + "learning_rate": 1.2062815617686908e-06, + "loss": 0.4483, + "step": 48594 + }, + { + "epoch": 17.94, + "learning_rate": 1.2062446114053456e-06, + "loss": 0.3051, + "step": 48595 + }, + { + "epoch": 17.95, + "learning_rate": 1.2062076610420003e-06, + "loss": 0.2945, + "step": 48596 + }, + { + "epoch": 17.95, + "learning_rate": 1.206170710678655e-06, + "loss": 0.2343, + "step": 48597 + }, + { + "epoch": 17.95, + "learning_rate": 1.2061337603153098e-06, + "loss": 0.1783, + "step": 48598 + }, + { + "epoch": 17.95, + "learning_rate": 1.2060968099519646e-06, + "loss": 0.115, + "step": 48599 + }, + { + "epoch": 17.95, + "learning_rate": 1.2060598595886192e-06, + "loss": 0.1544, + "step": 48600 + }, + { + "epoch": 17.95, + "learning_rate": 1.206022909225274e-06, + "loss": 0.3647, + "step": 48601 + }, + { + "epoch": 17.95, + "learning_rate": 1.205985958861929e-06, + "loss": 0.1469, + "step": 48602 + }, + { + "epoch": 17.95, + "learning_rate": 1.2059490084985836e-06, + "loss": 0.2042, + "step": 48603 + }, + { + "epoch": 17.95, + "learning_rate": 1.2059120581352384e-06, + "loss": 0.21, + "step": 48604 + }, + { + "epoch": 17.95, + "learning_rate": 1.2058751077718933e-06, + "loss": 0.2562, + "step": 48605 + }, + { + "epoch": 17.95, + "learning_rate": 1.205838157408548e-06, + "loss": 0.295, + "step": 48606 + }, + { + "epoch": 17.95, + "learning_rate": 1.2058012070452025e-06, + "loss": 0.3154, + "step": 48607 + }, + { + "epoch": 17.95, + "learning_rate": 1.2057642566818574e-06, + "loss": 0.3165, + "step": 48608 + }, + { + "epoch": 17.95, + "learning_rate": 1.2057273063185122e-06, + "loss": 0.1478, + "step": 48609 + }, + { + "epoch": 17.95, + "learning_rate": 1.2056903559551669e-06, + "loss": 0.3548, + "step": 48610 + }, + { + "epoch": 17.95, + "learning_rate": 1.2056534055918217e-06, + "loss": 0.2738, + "step": 48611 + }, + { + "epoch": 17.95, + "learning_rate": 1.2056164552284764e-06, + "loss": 0.3275, + "step": 48612 + }, + { + "epoch": 17.95, + "learning_rate": 1.2055795048651312e-06, + "loss": 0.1928, + "step": 48613 + }, + { + "epoch": 17.95, + "learning_rate": 1.205542554501786e-06, + "loss": 0.311, + "step": 48614 + }, + { + "epoch": 17.95, + "learning_rate": 1.2055056041384407e-06, + "loss": 0.282, + "step": 48615 + }, + { + "epoch": 17.95, + "learning_rate": 1.2054686537750953e-06, + "loss": 0.2419, + "step": 48616 + }, + { + "epoch": 17.95, + "learning_rate": 1.2054317034117502e-06, + "loss": 0.1481, + "step": 48617 + }, + { + "epoch": 17.95, + "learning_rate": 1.205394753048405e-06, + "loss": 0.3617, + "step": 48618 + }, + { + "epoch": 17.95, + "learning_rate": 1.2053578026850597e-06, + "loss": 0.132, + "step": 48619 + }, + { + "epoch": 17.95, + "learning_rate": 1.2053208523217145e-06, + "loss": 0.3008, + "step": 48620 + }, + { + "epoch": 17.95, + "learning_rate": 1.2052839019583694e-06, + "loss": 0.3132, + "step": 48621 + }, + { + "epoch": 17.95, + "learning_rate": 1.205246951595024e-06, + "loss": 0.3, + "step": 48622 + }, + { + "epoch": 17.96, + "learning_rate": 1.2052100012316788e-06, + "loss": 0.3108, + "step": 48623 + }, + { + "epoch": 17.96, + "learning_rate": 1.2051730508683337e-06, + "loss": 0.1762, + "step": 48624 + }, + { + "epoch": 17.96, + "learning_rate": 1.2051361005049883e-06, + "loss": 0.2838, + "step": 48625 + }, + { + "epoch": 17.96, + "learning_rate": 1.205099150141643e-06, + "loss": 0.4537, + "step": 48626 + }, + { + "epoch": 17.96, + "learning_rate": 1.2050621997782978e-06, + "loss": 0.1757, + "step": 48627 + }, + { + "epoch": 17.96, + "learning_rate": 1.2050252494149527e-06, + "loss": 0.1805, + "step": 48628 + }, + { + "epoch": 17.96, + "learning_rate": 1.2049882990516073e-06, + "loss": 0.1834, + "step": 48629 + }, + { + "epoch": 17.96, + "learning_rate": 1.2049513486882621e-06, + "loss": 0.1989, + "step": 48630 + }, + { + "epoch": 17.96, + "learning_rate": 1.204914398324917e-06, + "loss": 0.2664, + "step": 48631 + }, + { + "epoch": 17.96, + "learning_rate": 1.2048774479615716e-06, + "loss": 0.1837, + "step": 48632 + }, + { + "epoch": 17.96, + "learning_rate": 1.2048404975982265e-06, + "loss": 0.2767, + "step": 48633 + }, + { + "epoch": 17.96, + "learning_rate": 1.2048035472348813e-06, + "loss": 0.2518, + "step": 48634 + }, + { + "epoch": 17.96, + "learning_rate": 1.204766596871536e-06, + "loss": 0.1942, + "step": 48635 + }, + { + "epoch": 17.96, + "learning_rate": 1.2047296465081906e-06, + "loss": 0.1738, + "step": 48636 + }, + { + "epoch": 17.96, + "learning_rate": 1.2046926961448454e-06, + "loss": 0.2741, + "step": 48637 + }, + { + "epoch": 17.96, + "learning_rate": 1.2046557457815e-06, + "loss": 0.126, + "step": 48638 + }, + { + "epoch": 17.96, + "learning_rate": 1.204618795418155e-06, + "loss": 0.197, + "step": 48639 + }, + { + "epoch": 17.96, + "learning_rate": 1.2045818450548098e-06, + "loss": 0.1762, + "step": 48640 + }, + { + "epoch": 17.96, + "learning_rate": 1.2045448946914644e-06, + "loss": 0.1512, + "step": 48641 + }, + { + "epoch": 17.96, + "learning_rate": 1.2045079443281193e-06, + "loss": 0.2617, + "step": 48642 + }, + { + "epoch": 17.96, + "learning_rate": 1.2044709939647741e-06, + "loss": 0.0959, + "step": 48643 + }, + { + "epoch": 17.96, + "learning_rate": 1.2044340436014288e-06, + "loss": 0.2436, + "step": 48644 + }, + { + "epoch": 17.96, + "learning_rate": 1.2043970932380836e-06, + "loss": 0.1503, + "step": 48645 + }, + { + "epoch": 17.96, + "learning_rate": 1.2043601428747382e-06, + "loss": 0.2272, + "step": 48646 + }, + { + "epoch": 17.96, + "learning_rate": 1.204323192511393e-06, + "loss": 0.3311, + "step": 48647 + }, + { + "epoch": 17.96, + "learning_rate": 1.2042862421480477e-06, + "loss": 0.2386, + "step": 48648 + }, + { + "epoch": 17.96, + "learning_rate": 1.2042492917847026e-06, + "loss": 0.1505, + "step": 48649 + }, + { + "epoch": 17.97, + "learning_rate": 1.2042123414213574e-06, + "loss": 0.2922, + "step": 48650 + }, + { + "epoch": 17.97, + "learning_rate": 1.204175391058012e-06, + "loss": 0.2491, + "step": 48651 + }, + { + "epoch": 17.97, + "learning_rate": 1.204138440694667e-06, + "loss": 0.1723, + "step": 48652 + }, + { + "epoch": 17.97, + "learning_rate": 1.2041014903313217e-06, + "loss": 0.1265, + "step": 48653 + }, + { + "epoch": 17.97, + "learning_rate": 1.2040645399679764e-06, + "loss": 0.2293, + "step": 48654 + }, + { + "epoch": 17.97, + "learning_rate": 1.2040275896046312e-06, + "loss": 0.1321, + "step": 48655 + }, + { + "epoch": 17.97, + "learning_rate": 1.2039906392412859e-06, + "loss": 0.11, + "step": 48656 + }, + { + "epoch": 17.97, + "learning_rate": 1.2039536888779405e-06, + "loss": 0.1921, + "step": 48657 + }, + { + "epoch": 17.97, + "learning_rate": 1.2039167385145954e-06, + "loss": 0.2253, + "step": 48658 + }, + { + "epoch": 17.97, + "learning_rate": 1.2038797881512502e-06, + "loss": 0.2806, + "step": 48659 + }, + { + "epoch": 17.97, + "learning_rate": 1.2038428377879048e-06, + "loss": 0.2633, + "step": 48660 + }, + { + "epoch": 17.97, + "learning_rate": 1.2038058874245597e-06, + "loss": 0.2894, + "step": 48661 + }, + { + "epoch": 17.97, + "learning_rate": 1.2037689370612145e-06, + "loss": 0.2846, + "step": 48662 + }, + { + "epoch": 17.97, + "learning_rate": 1.2037319866978692e-06, + "loss": 0.503, + "step": 48663 + }, + { + "epoch": 17.97, + "learning_rate": 1.203695036334524e-06, + "loss": 0.3517, + "step": 48664 + }, + { + "epoch": 17.97, + "learning_rate": 1.2036580859711789e-06, + "loss": 0.216, + "step": 48665 + }, + { + "epoch": 17.97, + "learning_rate": 1.2036211356078335e-06, + "loss": 0.2225, + "step": 48666 + }, + { + "epoch": 17.97, + "learning_rate": 1.2035841852444881e-06, + "loss": 0.3783, + "step": 48667 + }, + { + "epoch": 17.97, + "learning_rate": 1.203547234881143e-06, + "loss": 0.2351, + "step": 48668 + }, + { + "epoch": 17.97, + "learning_rate": 1.2035102845177978e-06, + "loss": 0.2474, + "step": 48669 + }, + { + "epoch": 17.97, + "learning_rate": 1.2034733341544525e-06, + "loss": 0.2757, + "step": 48670 + }, + { + "epoch": 17.97, + "learning_rate": 1.2034363837911073e-06, + "loss": 0.2579, + "step": 48671 + }, + { + "epoch": 17.97, + "learning_rate": 1.2033994334277622e-06, + "loss": 0.1652, + "step": 48672 + }, + { + "epoch": 17.97, + "learning_rate": 1.2033624830644168e-06, + "loss": 0.2206, + "step": 48673 + }, + { + "epoch": 17.97, + "learning_rate": 1.2033255327010717e-06, + "loss": 0.2564, + "step": 48674 + }, + { + "epoch": 17.97, + "learning_rate": 1.2032885823377265e-06, + "loss": 0.2696, + "step": 48675 + }, + { + "epoch": 17.97, + "learning_rate": 1.2032516319743811e-06, + "loss": 0.3895, + "step": 48676 + }, + { + "epoch": 17.98, + "learning_rate": 1.2032146816110358e-06, + "loss": 0.2199, + "step": 48677 + }, + { + "epoch": 17.98, + "learning_rate": 1.2031777312476906e-06, + "loss": 0.3182, + "step": 48678 + }, + { + "epoch": 17.98, + "learning_rate": 1.2031407808843455e-06, + "loss": 0.2042, + "step": 48679 + }, + { + "epoch": 17.98, + "learning_rate": 1.2031038305210001e-06, + "loss": 0.2062, + "step": 48680 + }, + { + "epoch": 17.98, + "learning_rate": 1.203066880157655e-06, + "loss": 0.1166, + "step": 48681 + }, + { + "epoch": 17.98, + "learning_rate": 1.2030299297943098e-06, + "loss": 0.1243, + "step": 48682 + }, + { + "epoch": 17.98, + "learning_rate": 1.2029929794309644e-06, + "loss": 0.2689, + "step": 48683 + }, + { + "epoch": 17.98, + "learning_rate": 1.2029560290676193e-06, + "loss": 0.2919, + "step": 48684 + }, + { + "epoch": 17.98, + "learning_rate": 1.2029190787042741e-06, + "loss": 0.1371, + "step": 48685 + }, + { + "epoch": 17.98, + "learning_rate": 1.2028821283409286e-06, + "loss": 0.1981, + "step": 48686 + }, + { + "epoch": 17.98, + "learning_rate": 1.2028451779775834e-06, + "loss": 0.2508, + "step": 48687 + }, + { + "epoch": 17.98, + "learning_rate": 1.2028082276142383e-06, + "loss": 0.2418, + "step": 48688 + }, + { + "epoch": 17.98, + "learning_rate": 1.202771277250893e-06, + "loss": 0.086, + "step": 48689 + }, + { + "epoch": 17.98, + "learning_rate": 1.2027343268875477e-06, + "loss": 0.2821, + "step": 48690 + }, + { + "epoch": 17.98, + "learning_rate": 1.2026973765242026e-06, + "loss": 0.2962, + "step": 48691 + }, + { + "epoch": 17.98, + "learning_rate": 1.2026604261608572e-06, + "loss": 0.2414, + "step": 48692 + }, + { + "epoch": 17.98, + "learning_rate": 1.202623475797512e-06, + "loss": 0.2298, + "step": 48693 + }, + { + "epoch": 17.98, + "learning_rate": 1.202586525434167e-06, + "loss": 0.2537, + "step": 48694 + }, + { + "epoch": 17.98, + "learning_rate": 1.2025495750708216e-06, + "loss": 0.2842, + "step": 48695 + }, + { + "epoch": 17.98, + "learning_rate": 1.2025126247074762e-06, + "loss": 0.2327, + "step": 48696 + }, + { + "epoch": 17.98, + "learning_rate": 1.202475674344131e-06, + "loss": 0.1303, + "step": 48697 + }, + { + "epoch": 17.98, + "learning_rate": 1.202438723980786e-06, + "loss": 0.2059, + "step": 48698 + }, + { + "epoch": 17.98, + "learning_rate": 1.2024017736174405e-06, + "loss": 0.2559, + "step": 48699 + }, + { + "epoch": 17.98, + "learning_rate": 1.2023648232540954e-06, + "loss": 0.1662, + "step": 48700 + }, + { + "epoch": 17.98, + "learning_rate": 1.2023278728907502e-06, + "loss": 0.2665, + "step": 48701 + }, + { + "epoch": 17.98, + "learning_rate": 1.2022909225274049e-06, + "loss": 0.1649, + "step": 48702 + }, + { + "epoch": 17.98, + "learning_rate": 1.2022539721640597e-06, + "loss": 0.1449, + "step": 48703 + }, + { + "epoch": 17.99, + "learning_rate": 1.2022170218007146e-06, + "loss": 0.2573, + "step": 48704 + }, + { + "epoch": 17.99, + "learning_rate": 1.2021800714373692e-06, + "loss": 0.1973, + "step": 48705 + }, + { + "epoch": 17.99, + "learning_rate": 1.2021431210740238e-06, + "loss": 0.211, + "step": 48706 + }, + { + "epoch": 17.99, + "learning_rate": 1.2021061707106787e-06, + "loss": 0.1966, + "step": 48707 + }, + { + "epoch": 17.99, + "learning_rate": 1.2020692203473333e-06, + "loss": 0.1469, + "step": 48708 + }, + { + "epoch": 17.99, + "learning_rate": 1.2020322699839882e-06, + "loss": 0.1195, + "step": 48709 + }, + { + "epoch": 17.99, + "learning_rate": 1.201995319620643e-06, + "loss": 0.2672, + "step": 48710 + }, + { + "epoch": 17.99, + "learning_rate": 1.2019583692572977e-06, + "loss": 0.2964, + "step": 48711 + }, + { + "epoch": 17.99, + "learning_rate": 1.2019214188939525e-06, + "loss": 0.1718, + "step": 48712 + }, + { + "epoch": 17.99, + "learning_rate": 1.2018844685306073e-06, + "loss": 0.2128, + "step": 48713 + }, + { + "epoch": 17.99, + "learning_rate": 1.201847518167262e-06, + "loss": 0.2954, + "step": 48714 + }, + { + "epoch": 17.99, + "learning_rate": 1.2018105678039168e-06, + "loss": 0.3514, + "step": 48715 + }, + { + "epoch": 17.99, + "learning_rate": 1.2017736174405715e-06, + "loss": 0.1815, + "step": 48716 + }, + { + "epoch": 17.99, + "learning_rate": 1.2017366670772263e-06, + "loss": 0.1521, + "step": 48717 + }, + { + "epoch": 17.99, + "learning_rate": 1.201699716713881e-06, + "loss": 0.1606, + "step": 48718 + }, + { + "epoch": 17.99, + "learning_rate": 1.2016627663505358e-06, + "loss": 0.1515, + "step": 48719 + }, + { + "epoch": 17.99, + "learning_rate": 1.2016258159871906e-06, + "loss": 0.3848, + "step": 48720 + }, + { + "epoch": 17.99, + "learning_rate": 1.2015888656238453e-06, + "loss": 0.1897, + "step": 48721 + }, + { + "epoch": 17.99, + "learning_rate": 1.2015519152605001e-06, + "loss": 0.2118, + "step": 48722 + }, + { + "epoch": 17.99, + "learning_rate": 1.201514964897155e-06, + "loss": 0.2277, + "step": 48723 + }, + { + "epoch": 17.99, + "learning_rate": 1.2014780145338096e-06, + "loss": 0.1894, + "step": 48724 + }, + { + "epoch": 17.99, + "learning_rate": 1.2014410641704645e-06, + "loss": 0.228, + "step": 48725 + }, + { + "epoch": 17.99, + "learning_rate": 1.201404113807119e-06, + "loss": 0.3136, + "step": 48726 + }, + { + "epoch": 17.99, + "learning_rate": 1.2013671634437737e-06, + "loss": 0.2888, + "step": 48727 + }, + { + "epoch": 17.99, + "learning_rate": 1.2013302130804286e-06, + "loss": 0.2514, + "step": 48728 + }, + { + "epoch": 17.99, + "learning_rate": 1.2012932627170834e-06, + "loss": 0.2629, + "step": 48729 + }, + { + "epoch": 17.99, + "learning_rate": 1.201256312353738e-06, + "loss": 0.2775, + "step": 48730 + }, + { + "epoch": 18.0, + "learning_rate": 1.201219361990393e-06, + "loss": 0.3292, + "step": 48731 + }, + { + "epoch": 18.0, + "learning_rate": 1.2011824116270478e-06, + "loss": 0.1284, + "step": 48732 + }, + { + "epoch": 18.0, + "learning_rate": 1.2011454612637024e-06, + "loss": 0.2142, + "step": 48733 + }, + { + "epoch": 18.0, + "learning_rate": 1.2011085109003573e-06, + "loss": 0.147, + "step": 48734 + }, + { + "epoch": 18.0, + "learning_rate": 1.201071560537012e-06, + "loss": 0.1496, + "step": 48735 + }, + { + "epoch": 18.0, + "learning_rate": 1.2010346101736667e-06, + "loss": 0.1895, + "step": 48736 + }, + { + "epoch": 18.0, + "learning_rate": 1.2009976598103214e-06, + "loss": 0.1531, + "step": 48737 + }, + { + "epoch": 18.0, + "learning_rate": 1.2009607094469762e-06, + "loss": 0.3023, + "step": 48738 + }, + { + "epoch": 18.0, + "learning_rate": 1.200923759083631e-06, + "loss": 0.2426, + "step": 48739 + }, + { + "epoch": 18.0, + "learning_rate": 1.2008868087202857e-06, + "loss": 0.1675, + "step": 48740 + }, + { + "epoch": 18.0, + "learning_rate": 1.2008498583569406e-06, + "loss": 0.1558, + "step": 48741 + }, + { + "epoch": 18.0, + "learning_rate": 1.2008129079935954e-06, + "loss": 0.3024, + "step": 48742 + }, + { + "epoch": 18.0, + "learning_rate": 1.20077595763025e-06, + "loss": 0.2764, + "step": 48743 + }, + { + "epoch": 18.0, + "learning_rate": 1.2007390072669049e-06, + "loss": 0.0026, + "step": 48744 + }, + { + "epoch": 18.0, + "eval_accuracy": 0.8536731772637023, + "eval_loss": 0.3786056339740753, + "eval_runtime": 25.2979, + "eval_samples_per_second": 1712.156, + "eval_steps_per_second": 26.761, + "step": 48744 + }, + { + "epoch": 18.0, + "learning_rate": 1.2007020569035597e-06, + "loss": 0.2417, + "step": 48745 + }, + { + "epoch": 18.0, + "learning_rate": 1.2006651065402144e-06, + "loss": 0.2153, + "step": 48746 + }, + { + "epoch": 18.0, + "learning_rate": 1.200628156176869e-06, + "loss": 0.2255, + "step": 48747 + }, + { + "epoch": 18.0, + "learning_rate": 1.2005912058135239e-06, + "loss": 0.1722, + "step": 48748 + }, + { + "epoch": 18.0, + "learning_rate": 1.2005542554501787e-06, + "loss": 0.3166, + "step": 48749 + }, + { + "epoch": 18.0, + "learning_rate": 1.2005173050868333e-06, + "loss": 0.1821, + "step": 48750 + }, + { + "epoch": 18.0, + "learning_rate": 1.2004803547234882e-06, + "loss": 0.2477, + "step": 48751 + }, + { + "epoch": 18.0, + "learning_rate": 1.200443404360143e-06, + "loss": 0.2488, + "step": 48752 + }, + { + "epoch": 18.0, + "learning_rate": 1.2004064539967977e-06, + "loss": 0.3355, + "step": 48753 + }, + { + "epoch": 18.0, + "learning_rate": 1.2003695036334525e-06, + "loss": 0.2096, + "step": 48754 + }, + { + "epoch": 18.0, + "learning_rate": 1.2003325532701074e-06, + "loss": 0.2087, + "step": 48755 + }, + { + "epoch": 18.0, + "learning_rate": 1.2002956029067618e-06, + "loss": 0.2612, + "step": 48756 + }, + { + "epoch": 18.0, + "learning_rate": 1.2002586525434166e-06, + "loss": 0.1787, + "step": 48757 + }, + { + "epoch": 18.01, + "learning_rate": 1.2002217021800715e-06, + "loss": 0.1646, + "step": 48758 + }, + { + "epoch": 18.01, + "learning_rate": 1.2001847518167261e-06, + "loss": 0.2085, + "step": 48759 + }, + { + "epoch": 18.01, + "learning_rate": 1.200147801453381e-06, + "loss": 0.4204, + "step": 48760 + }, + { + "epoch": 18.01, + "learning_rate": 1.2001108510900358e-06, + "loss": 0.2062, + "step": 48761 + }, + { + "epoch": 18.01, + "learning_rate": 1.2000739007266905e-06, + "loss": 0.0742, + "step": 48762 + }, + { + "epoch": 18.01, + "learning_rate": 1.2000369503633453e-06, + "loss": 0.2366, + "step": 48763 + }, + { + "epoch": 18.01, + "learning_rate": 1.2000000000000002e-06, + "loss": 0.2927, + "step": 48764 + }, + { + "epoch": 18.01, + "learning_rate": 1.1999630496366548e-06, + "loss": 0.2551, + "step": 48765 + }, + { + "epoch": 18.01, + "learning_rate": 1.1999260992733094e-06, + "loss": 0.2667, + "step": 48766 + }, + { + "epoch": 18.01, + "learning_rate": 1.1998891489099643e-06, + "loss": 0.2217, + "step": 48767 + }, + { + "epoch": 18.01, + "learning_rate": 1.1998521985466191e-06, + "loss": 0.193, + "step": 48768 + }, + { + "epoch": 18.01, + "learning_rate": 1.1998152481832738e-06, + "loss": 0.2139, + "step": 48769 + }, + { + "epoch": 18.01, + "learning_rate": 1.1997782978199286e-06, + "loss": 0.2014, + "step": 48770 + }, + { + "epoch": 18.01, + "learning_rate": 1.1997413474565835e-06, + "loss": 0.1882, + "step": 48771 + }, + { + "epoch": 18.01, + "learning_rate": 1.199704397093238e-06, + "loss": 0.1323, + "step": 48772 + }, + { + "epoch": 18.01, + "learning_rate": 1.199667446729893e-06, + "loss": 0.2482, + "step": 48773 + }, + { + "epoch": 18.01, + "learning_rate": 1.1996304963665478e-06, + "loss": 0.2128, + "step": 48774 + }, + { + "epoch": 18.01, + "learning_rate": 1.1995935460032022e-06, + "loss": 0.3463, + "step": 48775 + }, + { + "epoch": 18.01, + "learning_rate": 1.199556595639857e-06, + "loss": 0.1739, + "step": 48776 + }, + { + "epoch": 18.01, + "learning_rate": 1.199519645276512e-06, + "loss": 0.1398, + "step": 48777 + }, + { + "epoch": 18.01, + "learning_rate": 1.1994826949131665e-06, + "loss": 0.2485, + "step": 48778 + }, + { + "epoch": 18.01, + "learning_rate": 1.1994457445498214e-06, + "loss": 0.2447, + "step": 48779 + }, + { + "epoch": 18.01, + "learning_rate": 1.1994087941864762e-06, + "loss": 0.1282, + "step": 48780 + }, + { + "epoch": 18.01, + "learning_rate": 1.1993718438231309e-06, + "loss": 0.2881, + "step": 48781 + }, + { + "epoch": 18.01, + "learning_rate": 1.1993348934597857e-06, + "loss": 0.137, + "step": 48782 + }, + { + "epoch": 18.01, + "learning_rate": 1.1992979430964406e-06, + "loss": 0.2469, + "step": 48783 + }, + { + "epoch": 18.01, + "learning_rate": 1.1992609927330952e-06, + "loss": 0.0767, + "step": 48784 + }, + { + "epoch": 18.02, + "learning_rate": 1.1992240423697499e-06, + "loss": 0.2713, + "step": 48785 + }, + { + "epoch": 18.02, + "learning_rate": 1.1991870920064047e-06, + "loss": 0.457, + "step": 48786 + }, + { + "epoch": 18.02, + "learning_rate": 1.1991501416430595e-06, + "loss": 0.1708, + "step": 48787 + }, + { + "epoch": 18.02, + "learning_rate": 1.1991131912797142e-06, + "loss": 0.2481, + "step": 48788 + }, + { + "epoch": 18.02, + "learning_rate": 1.199076240916369e-06, + "loss": 0.2189, + "step": 48789 + }, + { + "epoch": 18.02, + "learning_rate": 1.1990392905530239e-06, + "loss": 0.3033, + "step": 48790 + }, + { + "epoch": 18.02, + "learning_rate": 1.1990023401896785e-06, + "loss": 0.1137, + "step": 48791 + }, + { + "epoch": 18.02, + "learning_rate": 1.1989653898263334e-06, + "loss": 0.2536, + "step": 48792 + }, + { + "epoch": 18.02, + "learning_rate": 1.1989284394629882e-06, + "loss": 0.1926, + "step": 48793 + }, + { + "epoch": 18.02, + "learning_rate": 1.1988914890996429e-06, + "loss": 0.4094, + "step": 48794 + }, + { + "epoch": 18.02, + "learning_rate": 1.1988545387362975e-06, + "loss": 0.2274, + "step": 48795 + }, + { + "epoch": 18.02, + "learning_rate": 1.1988175883729523e-06, + "loss": 0.2105, + "step": 48796 + }, + { + "epoch": 18.02, + "learning_rate": 1.198780638009607e-06, + "loss": 0.1661, + "step": 48797 + }, + { + "epoch": 18.02, + "learning_rate": 1.1987436876462618e-06, + "loss": 0.1613, + "step": 48798 + }, + { + "epoch": 18.02, + "learning_rate": 1.1987067372829167e-06, + "loss": 0.3403, + "step": 48799 + }, + { + "epoch": 18.02, + "learning_rate": 1.1986697869195713e-06, + "loss": 0.1542, + "step": 48800 + }, + { + "epoch": 18.02, + "learning_rate": 1.1986328365562262e-06, + "loss": 0.2577, + "step": 48801 + }, + { + "epoch": 18.02, + "learning_rate": 1.198595886192881e-06, + "loss": 0.228, + "step": 48802 + }, + { + "epoch": 18.02, + "learning_rate": 1.1985589358295356e-06, + "loss": 0.2304, + "step": 48803 + }, + { + "epoch": 18.02, + "learning_rate": 1.1985219854661905e-06, + "loss": 0.1495, + "step": 48804 + }, + { + "epoch": 18.02, + "learning_rate": 1.1984850351028451e-06, + "loss": 0.2736, + "step": 48805 + }, + { + "epoch": 18.02, + "learning_rate": 1.1984480847395e-06, + "loss": 0.224, + "step": 48806 + }, + { + "epoch": 18.02, + "learning_rate": 1.1984111343761546e-06, + "loss": 0.2039, + "step": 48807 + }, + { + "epoch": 18.02, + "learning_rate": 1.1983741840128095e-06, + "loss": 0.212, + "step": 48808 + }, + { + "epoch": 18.02, + "learning_rate": 1.1983372336494643e-06, + "loss": 0.4216, + "step": 48809 + }, + { + "epoch": 18.02, + "learning_rate": 1.198300283286119e-06, + "loss": 0.1979, + "step": 48810 + }, + { + "epoch": 18.02, + "learning_rate": 1.1982633329227738e-06, + "loss": 0.2164, + "step": 48811 + }, + { + "epoch": 18.03, + "learning_rate": 1.1982263825594286e-06, + "loss": 0.3111, + "step": 48812 + }, + { + "epoch": 18.03, + "learning_rate": 1.1981894321960833e-06, + "loss": 0.1286, + "step": 48813 + }, + { + "epoch": 18.03, + "learning_rate": 1.1981524818327381e-06, + "loss": 0.278, + "step": 48814 + }, + { + "epoch": 18.03, + "learning_rate": 1.1981155314693928e-06, + "loss": 0.1337, + "step": 48815 + }, + { + "epoch": 18.03, + "learning_rate": 1.1980785811060476e-06, + "loss": 0.2902, + "step": 48816 + }, + { + "epoch": 18.03, + "learning_rate": 1.1980416307427022e-06, + "loss": 0.1406, + "step": 48817 + }, + { + "epoch": 18.03, + "learning_rate": 1.198004680379357e-06, + "loss": 0.3241, + "step": 48818 + }, + { + "epoch": 18.03, + "learning_rate": 1.197967730016012e-06, + "loss": 0.3868, + "step": 48819 + }, + { + "epoch": 18.03, + "learning_rate": 1.1979307796526666e-06, + "loss": 0.2098, + "step": 48820 + }, + { + "epoch": 18.03, + "learning_rate": 1.1978938292893214e-06, + "loss": 0.1644, + "step": 48821 + }, + { + "epoch": 18.03, + "learning_rate": 1.1978568789259763e-06, + "loss": 0.2251, + "step": 48822 + }, + { + "epoch": 18.03, + "learning_rate": 1.197819928562631e-06, + "loss": 0.2243, + "step": 48823 + }, + { + "epoch": 18.03, + "learning_rate": 1.1977829781992858e-06, + "loss": 0.2681, + "step": 48824 + }, + { + "epoch": 18.03, + "learning_rate": 1.1977460278359404e-06, + "loss": 0.1948, + "step": 48825 + }, + { + "epoch": 18.03, + "learning_rate": 1.197709077472595e-06, + "loss": 0.2444, + "step": 48826 + }, + { + "epoch": 18.03, + "learning_rate": 1.1976721271092499e-06, + "loss": 0.2701, + "step": 48827 + }, + { + "epoch": 18.03, + "learning_rate": 1.1976351767459047e-06, + "loss": 0.2253, + "step": 48828 + }, + { + "epoch": 18.03, + "learning_rate": 1.1975982263825594e-06, + "loss": 0.218, + "step": 48829 + }, + { + "epoch": 18.03, + "learning_rate": 1.1975612760192142e-06, + "loss": 0.2128, + "step": 48830 + }, + { + "epoch": 18.03, + "learning_rate": 1.197524325655869e-06, + "loss": 0.1787, + "step": 48831 + }, + { + "epoch": 18.03, + "learning_rate": 1.1974873752925237e-06, + "loss": 0.2827, + "step": 48832 + }, + { + "epoch": 18.03, + "learning_rate": 1.1974504249291785e-06, + "loss": 0.0988, + "step": 48833 + }, + { + "epoch": 18.03, + "learning_rate": 1.1974134745658334e-06, + "loss": 0.2857, + "step": 48834 + }, + { + "epoch": 18.03, + "learning_rate": 1.197376524202488e-06, + "loss": 0.2674, + "step": 48835 + }, + { + "epoch": 18.03, + "learning_rate": 1.1973395738391427e-06, + "loss": 0.313, + "step": 48836 + }, + { + "epoch": 18.03, + "learning_rate": 1.1973026234757975e-06, + "loss": 0.2618, + "step": 48837 + }, + { + "epoch": 18.03, + "learning_rate": 1.1972656731124524e-06, + "loss": 0.1933, + "step": 48838 + }, + { + "epoch": 18.04, + "learning_rate": 1.197228722749107e-06, + "loss": 0.2846, + "step": 48839 + }, + { + "epoch": 18.04, + "learning_rate": 1.1971917723857618e-06, + "loss": 0.2939, + "step": 48840 + }, + { + "epoch": 18.04, + "learning_rate": 1.1971548220224167e-06, + "loss": 0.2745, + "step": 48841 + }, + { + "epoch": 18.04, + "learning_rate": 1.1971178716590713e-06, + "loss": 0.1931, + "step": 48842 + }, + { + "epoch": 18.04, + "learning_rate": 1.1970809212957262e-06, + "loss": 0.2141, + "step": 48843 + }, + { + "epoch": 18.04, + "learning_rate": 1.197043970932381e-06, + "loss": 0.1939, + "step": 48844 + }, + { + "epoch": 18.04, + "learning_rate": 1.1970070205690354e-06, + "loss": 0.2084, + "step": 48845 + }, + { + "epoch": 18.04, + "learning_rate": 1.1969700702056903e-06, + "loss": 0.1591, + "step": 48846 + }, + { + "epoch": 18.04, + "learning_rate": 1.1969331198423451e-06, + "loss": 0.2333, + "step": 48847 + }, + { + "epoch": 18.04, + "learning_rate": 1.1968961694789998e-06, + "loss": 0.123, + "step": 48848 + }, + { + "epoch": 18.04, + "learning_rate": 1.1968592191156546e-06, + "loss": 0.1937, + "step": 48849 + }, + { + "epoch": 18.04, + "learning_rate": 1.1968222687523095e-06, + "loss": 0.2564, + "step": 48850 + }, + { + "epoch": 18.04, + "learning_rate": 1.1967853183889641e-06, + "loss": 0.1169, + "step": 48851 + }, + { + "epoch": 18.04, + "learning_rate": 1.196748368025619e-06, + "loss": 0.1917, + "step": 48852 + }, + { + "epoch": 18.04, + "learning_rate": 1.1967114176622738e-06, + "loss": 0.1999, + "step": 48853 + }, + { + "epoch": 18.04, + "learning_rate": 1.1966744672989284e-06, + "loss": 0.3105, + "step": 48854 + }, + { + "epoch": 18.04, + "learning_rate": 1.196637516935583e-06, + "loss": 0.1917, + "step": 48855 + }, + { + "epoch": 18.04, + "learning_rate": 1.196600566572238e-06, + "loss": 0.2064, + "step": 48856 + }, + { + "epoch": 18.04, + "learning_rate": 1.1965636162088928e-06, + "loss": 0.2266, + "step": 48857 + }, + { + "epoch": 18.04, + "learning_rate": 1.1965266658455474e-06, + "loss": 0.1325, + "step": 48858 + }, + { + "epoch": 18.04, + "learning_rate": 1.1964897154822023e-06, + "loss": 0.2539, + "step": 48859 + }, + { + "epoch": 18.04, + "learning_rate": 1.1964527651188571e-06, + "loss": 0.2244, + "step": 48860 + }, + { + "epoch": 18.04, + "learning_rate": 1.1964158147555118e-06, + "loss": 0.2084, + "step": 48861 + }, + { + "epoch": 18.04, + "learning_rate": 1.1963788643921666e-06, + "loss": 0.1624, + "step": 48862 + }, + { + "epoch": 18.04, + "learning_rate": 1.1963419140288214e-06, + "loss": 0.2303, + "step": 48863 + }, + { + "epoch": 18.04, + "learning_rate": 1.196304963665476e-06, + "loss": 0.2066, + "step": 48864 + }, + { + "epoch": 18.04, + "learning_rate": 1.1962680133021307e-06, + "loss": 0.085, + "step": 48865 + }, + { + "epoch": 18.05, + "learning_rate": 1.1962310629387856e-06, + "loss": 0.1252, + "step": 48866 + }, + { + "epoch": 18.05, + "learning_rate": 1.1961941125754404e-06, + "loss": 0.1289, + "step": 48867 + }, + { + "epoch": 18.05, + "learning_rate": 1.196157162212095e-06, + "loss": 0.3296, + "step": 48868 + }, + { + "epoch": 18.05, + "learning_rate": 1.19612021184875e-06, + "loss": 0.1606, + "step": 48869 + }, + { + "epoch": 18.05, + "learning_rate": 1.1960832614854047e-06, + "loss": 0.1512, + "step": 48870 + }, + { + "epoch": 18.05, + "learning_rate": 1.1960463111220594e-06, + "loss": 0.2137, + "step": 48871 + }, + { + "epoch": 18.05, + "learning_rate": 1.1960093607587142e-06, + "loss": 0.2338, + "step": 48872 + }, + { + "epoch": 18.05, + "learning_rate": 1.195972410395369e-06, + "loss": 0.3375, + "step": 48873 + }, + { + "epoch": 18.05, + "learning_rate": 1.1959354600320237e-06, + "loss": 0.0741, + "step": 48874 + }, + { + "epoch": 18.05, + "learning_rate": 1.1958985096686784e-06, + "loss": 0.1852, + "step": 48875 + }, + { + "epoch": 18.05, + "learning_rate": 1.1958615593053332e-06, + "loss": 0.119, + "step": 48876 + }, + { + "epoch": 18.05, + "learning_rate": 1.1958246089419878e-06, + "loss": 0.3769, + "step": 48877 + }, + { + "epoch": 18.05, + "learning_rate": 1.1957876585786427e-06, + "loss": 0.1534, + "step": 48878 + }, + { + "epoch": 18.05, + "learning_rate": 1.1957507082152975e-06, + "loss": 0.1882, + "step": 48879 + }, + { + "epoch": 18.05, + "learning_rate": 1.1957137578519522e-06, + "loss": 0.1456, + "step": 48880 + }, + { + "epoch": 18.05, + "learning_rate": 1.195676807488607e-06, + "loss": 0.2752, + "step": 48881 + }, + { + "epoch": 18.05, + "learning_rate": 1.1956398571252619e-06, + "loss": 0.179, + "step": 48882 + }, + { + "epoch": 18.05, + "learning_rate": 1.1956029067619165e-06, + "loss": 0.1911, + "step": 48883 + }, + { + "epoch": 18.05, + "learning_rate": 1.1955659563985714e-06, + "loss": 0.3082, + "step": 48884 + }, + { + "epoch": 18.05, + "learning_rate": 1.195529006035226e-06, + "loss": 0.1821, + "step": 48885 + }, + { + "epoch": 18.05, + "learning_rate": 1.1954920556718808e-06, + "loss": 0.2274, + "step": 48886 + }, + { + "epoch": 18.05, + "learning_rate": 1.1954551053085355e-06, + "loss": 0.2236, + "step": 48887 + }, + { + "epoch": 18.05, + "learning_rate": 1.1954181549451903e-06, + "loss": 0.3282, + "step": 48888 + }, + { + "epoch": 18.05, + "learning_rate": 1.1953812045818452e-06, + "loss": 0.1423, + "step": 48889 + }, + { + "epoch": 18.05, + "learning_rate": 1.1953442542184998e-06, + "loss": 0.2715, + "step": 48890 + }, + { + "epoch": 18.05, + "learning_rate": 1.1953073038551547e-06, + "loss": 0.1955, + "step": 48891 + }, + { + "epoch": 18.05, + "learning_rate": 1.1952703534918095e-06, + "loss": 0.3369, + "step": 48892 + }, + { + "epoch": 18.06, + "learning_rate": 1.1952334031284641e-06, + "loss": 0.2678, + "step": 48893 + }, + { + "epoch": 18.06, + "learning_rate": 1.195196452765119e-06, + "loss": 0.2818, + "step": 48894 + }, + { + "epoch": 18.06, + "learning_rate": 1.1951595024017736e-06, + "loss": 0.2618, + "step": 48895 + }, + { + "epoch": 18.06, + "learning_rate": 1.1951225520384283e-06, + "loss": 0.2397, + "step": 48896 + }, + { + "epoch": 18.06, + "learning_rate": 1.1950856016750831e-06, + "loss": 0.1975, + "step": 48897 + }, + { + "epoch": 18.06, + "learning_rate": 1.195048651311738e-06, + "loss": 0.2821, + "step": 48898 + }, + { + "epoch": 18.06, + "learning_rate": 1.1950117009483926e-06, + "loss": 0.3084, + "step": 48899 + }, + { + "epoch": 18.06, + "learning_rate": 1.1949747505850474e-06, + "loss": 0.2429, + "step": 48900 + }, + { + "epoch": 18.06, + "learning_rate": 1.1949378002217023e-06, + "loss": 0.2869, + "step": 48901 + }, + { + "epoch": 18.06, + "learning_rate": 1.194900849858357e-06, + "loss": 0.2934, + "step": 48902 + }, + { + "epoch": 18.06, + "learning_rate": 1.1948638994950118e-06, + "loss": 0.1651, + "step": 48903 + }, + { + "epoch": 18.06, + "learning_rate": 1.1948269491316666e-06, + "loss": 0.1466, + "step": 48904 + }, + { + "epoch": 18.06, + "learning_rate": 1.1947899987683213e-06, + "loss": 0.2118, + "step": 48905 + }, + { + "epoch": 18.06, + "learning_rate": 1.194753048404976e-06, + "loss": 0.1783, + "step": 48906 + }, + { + "epoch": 18.06, + "learning_rate": 1.1947160980416307e-06, + "loss": 0.1997, + "step": 48907 + }, + { + "epoch": 18.06, + "learning_rate": 1.1946791476782856e-06, + "loss": 0.2155, + "step": 48908 + }, + { + "epoch": 18.06, + "learning_rate": 1.1946421973149402e-06, + "loss": 0.1341, + "step": 48909 + }, + { + "epoch": 18.06, + "learning_rate": 1.194605246951595e-06, + "loss": 0.3327, + "step": 48910 + }, + { + "epoch": 18.06, + "learning_rate": 1.19456829658825e-06, + "loss": 0.2526, + "step": 48911 + }, + { + "epoch": 18.06, + "learning_rate": 1.1945313462249046e-06, + "loss": 0.2669, + "step": 48912 + }, + { + "epoch": 18.06, + "learning_rate": 1.1944943958615594e-06, + "loss": 0.1115, + "step": 48913 + }, + { + "epoch": 18.06, + "learning_rate": 1.1944574454982143e-06, + "loss": 0.3138, + "step": 48914 + }, + { + "epoch": 18.06, + "learning_rate": 1.1944204951348687e-06, + "loss": 0.2298, + "step": 48915 + }, + { + "epoch": 18.06, + "learning_rate": 1.1943835447715235e-06, + "loss": 0.3038, + "step": 48916 + }, + { + "epoch": 18.06, + "learning_rate": 1.1943465944081784e-06, + "loss": 0.1378, + "step": 48917 + }, + { + "epoch": 18.06, + "learning_rate": 1.194309644044833e-06, + "loss": 0.293, + "step": 48918 + }, + { + "epoch": 18.06, + "learning_rate": 1.1942726936814879e-06, + "loss": 0.2396, + "step": 48919 + }, + { + "epoch": 18.06, + "learning_rate": 1.1942357433181427e-06, + "loss": 0.1474, + "step": 48920 + }, + { + "epoch": 18.07, + "learning_rate": 1.1941987929547973e-06, + "loss": 0.3253, + "step": 48921 + }, + { + "epoch": 18.07, + "learning_rate": 1.1941618425914522e-06, + "loss": 0.1857, + "step": 48922 + }, + { + "epoch": 18.07, + "learning_rate": 1.194124892228107e-06, + "loss": 0.2775, + "step": 48923 + }, + { + "epoch": 18.07, + "learning_rate": 1.1940879418647617e-06, + "loss": 0.0834, + "step": 48924 + }, + { + "epoch": 18.07, + "learning_rate": 1.1940509915014163e-06, + "loss": 0.2746, + "step": 48925 + }, + { + "epoch": 18.07, + "learning_rate": 1.1940140411380712e-06, + "loss": 0.1622, + "step": 48926 + }, + { + "epoch": 18.07, + "learning_rate": 1.193977090774726e-06, + "loss": 0.1166, + "step": 48927 + }, + { + "epoch": 18.07, + "learning_rate": 1.1939401404113806e-06, + "loss": 0.1775, + "step": 48928 + }, + { + "epoch": 18.07, + "learning_rate": 1.1939031900480355e-06, + "loss": 0.3757, + "step": 48929 + }, + { + "epoch": 18.07, + "learning_rate": 1.1938662396846903e-06, + "loss": 0.2333, + "step": 48930 + }, + { + "epoch": 18.07, + "learning_rate": 1.193829289321345e-06, + "loss": 0.192, + "step": 48931 + }, + { + "epoch": 18.07, + "learning_rate": 1.1937923389579998e-06, + "loss": 0.1913, + "step": 48932 + }, + { + "epoch": 18.07, + "learning_rate": 1.1937553885946547e-06, + "loss": 0.2266, + "step": 48933 + }, + { + "epoch": 18.07, + "learning_rate": 1.1937184382313093e-06, + "loss": 0.2902, + "step": 48934 + }, + { + "epoch": 18.07, + "learning_rate": 1.193681487867964e-06, + "loss": 0.3123, + "step": 48935 + }, + { + "epoch": 18.07, + "learning_rate": 1.1936445375046188e-06, + "loss": 0.2111, + "step": 48936 + }, + { + "epoch": 18.07, + "learning_rate": 1.1936075871412736e-06, + "loss": 0.317, + "step": 48937 + }, + { + "epoch": 18.07, + "learning_rate": 1.1935706367779283e-06, + "loss": 0.3066, + "step": 48938 + }, + { + "epoch": 18.07, + "learning_rate": 1.1935336864145831e-06, + "loss": 0.18, + "step": 48939 + }, + { + "epoch": 18.07, + "learning_rate": 1.193496736051238e-06, + "loss": 0.2809, + "step": 48940 + }, + { + "epoch": 18.07, + "learning_rate": 1.1934597856878926e-06, + "loss": 0.1949, + "step": 48941 + }, + { + "epoch": 18.07, + "learning_rate": 1.1934228353245475e-06, + "loss": 0.3052, + "step": 48942 + }, + { + "epoch": 18.07, + "learning_rate": 1.1933858849612023e-06, + "loss": 0.127, + "step": 48943 + }, + { + "epoch": 18.07, + "learning_rate": 1.193348934597857e-06, + "loss": 0.1901, + "step": 48944 + }, + { + "epoch": 18.07, + "learning_rate": 1.1933119842345116e-06, + "loss": 0.3755, + "step": 48945 + }, + { + "epoch": 18.07, + "learning_rate": 1.1932750338711664e-06, + "loss": 0.1809, + "step": 48946 + }, + { + "epoch": 18.07, + "learning_rate": 1.193238083507821e-06, + "loss": 0.1965, + "step": 48947 + }, + { + "epoch": 18.08, + "learning_rate": 1.193201133144476e-06, + "loss": 0.1908, + "step": 48948 + }, + { + "epoch": 18.08, + "learning_rate": 1.1931641827811308e-06, + "loss": 0.3013, + "step": 48949 + }, + { + "epoch": 18.08, + "learning_rate": 1.1931272324177854e-06, + "loss": 0.3536, + "step": 48950 + }, + { + "epoch": 18.08, + "learning_rate": 1.1930902820544403e-06, + "loss": 0.1895, + "step": 48951 + }, + { + "epoch": 18.08, + "learning_rate": 1.193053331691095e-06, + "loss": 0.1241, + "step": 48952 + }, + { + "epoch": 18.08, + "learning_rate": 1.1930163813277497e-06, + "loss": 0.2146, + "step": 48953 + }, + { + "epoch": 18.08, + "learning_rate": 1.1929794309644046e-06, + "loss": 0.2596, + "step": 48954 + }, + { + "epoch": 18.08, + "learning_rate": 1.1929424806010592e-06, + "loss": 0.2046, + "step": 48955 + }, + { + "epoch": 18.08, + "learning_rate": 1.192905530237714e-06, + "loss": 0.218, + "step": 48956 + }, + { + "epoch": 18.08, + "learning_rate": 1.1928685798743687e-06, + "loss": 0.18, + "step": 48957 + }, + { + "epoch": 18.08, + "learning_rate": 1.1928316295110236e-06, + "loss": 0.2059, + "step": 48958 + }, + { + "epoch": 18.08, + "learning_rate": 1.1927946791476784e-06, + "loss": 0.2983, + "step": 48959 + }, + { + "epoch": 18.08, + "learning_rate": 1.192757728784333e-06, + "loss": 0.2412, + "step": 48960 + }, + { + "epoch": 18.08, + "learning_rate": 1.1927207784209879e-06, + "loss": 0.1256, + "step": 48961 + }, + { + "epoch": 18.08, + "learning_rate": 1.1926838280576427e-06, + "loss": 0.2404, + "step": 48962 + }, + { + "epoch": 18.08, + "learning_rate": 1.1926468776942974e-06, + "loss": 0.208, + "step": 48963 + }, + { + "epoch": 18.08, + "learning_rate": 1.1926099273309522e-06, + "loss": 0.2181, + "step": 48964 + }, + { + "epoch": 18.08, + "learning_rate": 1.1925729769676069e-06, + "loss": 0.3245, + "step": 48965 + }, + { + "epoch": 18.08, + "learning_rate": 1.1925360266042615e-06, + "loss": 0.2563, + "step": 48966 + }, + { + "epoch": 18.08, + "learning_rate": 1.1924990762409163e-06, + "loss": 0.2833, + "step": 48967 + }, + { + "epoch": 18.08, + "learning_rate": 1.1924621258775712e-06, + "loss": 0.1806, + "step": 48968 + }, + { + "epoch": 18.08, + "learning_rate": 1.1924251755142258e-06, + "loss": 0.2165, + "step": 48969 + }, + { + "epoch": 18.08, + "learning_rate": 1.1923882251508807e-06, + "loss": 0.3236, + "step": 48970 + }, + { + "epoch": 18.08, + "learning_rate": 1.1923512747875355e-06, + "loss": 0.2767, + "step": 48971 + }, + { + "epoch": 18.08, + "learning_rate": 1.1923143244241902e-06, + "loss": 0.1567, + "step": 48972 + }, + { + "epoch": 18.08, + "learning_rate": 1.192277374060845e-06, + "loss": 0.0996, + "step": 48973 + }, + { + "epoch": 18.08, + "learning_rate": 1.1922404236974999e-06, + "loss": 0.3817, + "step": 48974 + }, + { + "epoch": 18.09, + "learning_rate": 1.1922034733341545e-06, + "loss": 0.0834, + "step": 48975 + }, + { + "epoch": 18.09, + "learning_rate": 1.1921665229708091e-06, + "loss": 0.2836, + "step": 48976 + }, + { + "epoch": 18.09, + "learning_rate": 1.192129572607464e-06, + "loss": 0.2292, + "step": 48977 + }, + { + "epoch": 18.09, + "learning_rate": 1.1920926222441188e-06, + "loss": 0.1999, + "step": 48978 + }, + { + "epoch": 18.09, + "learning_rate": 1.1920556718807735e-06, + "loss": 0.1953, + "step": 48979 + }, + { + "epoch": 18.09, + "learning_rate": 1.1920187215174283e-06, + "loss": 0.2592, + "step": 48980 + }, + { + "epoch": 18.09, + "learning_rate": 1.1919817711540832e-06, + "loss": 0.2431, + "step": 48981 + }, + { + "epoch": 18.09, + "learning_rate": 1.1919448207907378e-06, + "loss": 0.1915, + "step": 48982 + }, + { + "epoch": 18.09, + "learning_rate": 1.1919078704273926e-06, + "loss": 0.1492, + "step": 48983 + }, + { + "epoch": 18.09, + "learning_rate": 1.1918709200640475e-06, + "loss": 0.3537, + "step": 48984 + }, + { + "epoch": 18.09, + "learning_rate": 1.191833969700702e-06, + "loss": 0.1547, + "step": 48985 + }, + { + "epoch": 18.09, + "learning_rate": 1.1917970193373568e-06, + "loss": 0.3442, + "step": 48986 + }, + { + "epoch": 18.09, + "learning_rate": 1.1917600689740116e-06, + "loss": 0.1983, + "step": 48987 + }, + { + "epoch": 18.09, + "learning_rate": 1.1917231186106662e-06, + "loss": 0.2259, + "step": 48988 + }, + { + "epoch": 18.09, + "learning_rate": 1.191686168247321e-06, + "loss": 0.1758, + "step": 48989 + }, + { + "epoch": 18.09, + "learning_rate": 1.191649217883976e-06, + "loss": 0.2647, + "step": 48990 + }, + { + "epoch": 18.09, + "learning_rate": 1.1916122675206306e-06, + "loss": 0.117, + "step": 48991 + }, + { + "epoch": 18.09, + "learning_rate": 1.1915753171572854e-06, + "loss": 0.2047, + "step": 48992 + }, + { + "epoch": 18.09, + "learning_rate": 1.1915383667939403e-06, + "loss": 0.1912, + "step": 48993 + }, + { + "epoch": 18.09, + "learning_rate": 1.191501416430595e-06, + "loss": 0.3229, + "step": 48994 + }, + { + "epoch": 18.09, + "learning_rate": 1.1914644660672495e-06, + "loss": 0.1343, + "step": 48995 + }, + { + "epoch": 18.09, + "learning_rate": 1.1914275157039044e-06, + "loss": 0.1058, + "step": 48996 + }, + { + "epoch": 18.09, + "learning_rate": 1.1913905653405592e-06, + "loss": 0.1944, + "step": 48997 + }, + { + "epoch": 18.09, + "learning_rate": 1.1913536149772139e-06, + "loss": 0.165, + "step": 48998 + }, + { + "epoch": 18.09, + "learning_rate": 1.1913166646138687e-06, + "loss": 0.2685, + "step": 48999 + }, + { + "epoch": 18.09, + "learning_rate": 1.1912797142505236e-06, + "loss": 0.1256, + "step": 49000 + }, + { + "epoch": 18.09, + "learning_rate": 1.1912427638871782e-06, + "loss": 0.3377, + "step": 49001 + }, + { + "epoch": 18.1, + "learning_rate": 1.191205813523833e-06, + "loss": 0.2202, + "step": 49002 + }, + { + "epoch": 18.1, + "learning_rate": 1.191168863160488e-06, + "loss": 0.2824, + "step": 49003 + }, + { + "epoch": 18.1, + "learning_rate": 1.1911319127971425e-06, + "loss": 0.2073, + "step": 49004 + }, + { + "epoch": 18.1, + "learning_rate": 1.1910949624337972e-06, + "loss": 0.1578, + "step": 49005 + }, + { + "epoch": 18.1, + "learning_rate": 1.191058012070452e-06, + "loss": 0.144, + "step": 49006 + }, + { + "epoch": 18.1, + "learning_rate": 1.1910210617071069e-06, + "loss": 0.1758, + "step": 49007 + }, + { + "epoch": 18.1, + "learning_rate": 1.1909841113437615e-06, + "loss": 0.2088, + "step": 49008 + }, + { + "epoch": 18.1, + "learning_rate": 1.1909471609804164e-06, + "loss": 0.1266, + "step": 49009 + }, + { + "epoch": 18.1, + "learning_rate": 1.1909102106170712e-06, + "loss": 0.1742, + "step": 49010 + }, + { + "epoch": 18.1, + "learning_rate": 1.1908732602537258e-06, + "loss": 0.1251, + "step": 49011 + }, + { + "epoch": 18.1, + "learning_rate": 1.1908363098903807e-06, + "loss": 0.3945, + "step": 49012 + }, + { + "epoch": 18.1, + "learning_rate": 1.1907993595270355e-06, + "loss": 0.2606, + "step": 49013 + }, + { + "epoch": 18.1, + "learning_rate": 1.19076240916369e-06, + "loss": 0.2042, + "step": 49014 + }, + { + "epoch": 18.1, + "learning_rate": 1.1907254588003448e-06, + "loss": 0.1538, + "step": 49015 + }, + { + "epoch": 18.1, + "learning_rate": 1.1906885084369997e-06, + "loss": 0.3351, + "step": 49016 + }, + { + "epoch": 18.1, + "learning_rate": 1.1906515580736543e-06, + "loss": 0.3153, + "step": 49017 + }, + { + "epoch": 18.1, + "learning_rate": 1.1906146077103092e-06, + "loss": 0.2888, + "step": 49018 + }, + { + "epoch": 18.1, + "learning_rate": 1.190577657346964e-06, + "loss": 0.4148, + "step": 49019 + }, + { + "epoch": 18.1, + "learning_rate": 1.1905407069836186e-06, + "loss": 0.3773, + "step": 49020 + }, + { + "epoch": 18.1, + "learning_rate": 1.1905037566202735e-06, + "loss": 0.2741, + "step": 49021 + }, + { + "epoch": 18.1, + "learning_rate": 1.1904668062569283e-06, + "loss": 0.2064, + "step": 49022 + }, + { + "epoch": 18.1, + "learning_rate": 1.190429855893583e-06, + "loss": 0.1269, + "step": 49023 + }, + { + "epoch": 18.1, + "learning_rate": 1.1903929055302376e-06, + "loss": 0.2328, + "step": 49024 + }, + { + "epoch": 18.1, + "learning_rate": 1.1903559551668925e-06, + "loss": 0.1299, + "step": 49025 + }, + { + "epoch": 18.1, + "learning_rate": 1.1903190048035473e-06, + "loss": 0.2792, + "step": 49026 + }, + { + "epoch": 18.1, + "learning_rate": 1.190282054440202e-06, + "loss": 0.3563, + "step": 49027 + }, + { + "epoch": 18.1, + "learning_rate": 1.1902451040768568e-06, + "loss": 0.1896, + "step": 49028 + }, + { + "epoch": 18.11, + "learning_rate": 1.1902081537135116e-06, + "loss": 0.1627, + "step": 49029 + }, + { + "epoch": 18.11, + "learning_rate": 1.1901712033501663e-06, + "loss": 0.1728, + "step": 49030 + }, + { + "epoch": 18.11, + "learning_rate": 1.1901342529868211e-06, + "loss": 0.1888, + "step": 49031 + }, + { + "epoch": 18.11, + "learning_rate": 1.190097302623476e-06, + "loss": 0.3045, + "step": 49032 + }, + { + "epoch": 18.11, + "learning_rate": 1.1900603522601306e-06, + "loss": 0.203, + "step": 49033 + }, + { + "epoch": 18.11, + "learning_rate": 1.1900234018967852e-06, + "loss": 0.1837, + "step": 49034 + }, + { + "epoch": 18.11, + "learning_rate": 1.18998645153344e-06, + "loss": 0.2784, + "step": 49035 + }, + { + "epoch": 18.11, + "learning_rate": 1.1899495011700947e-06, + "loss": 0.1035, + "step": 49036 + }, + { + "epoch": 18.11, + "learning_rate": 1.1899125508067496e-06, + "loss": 0.1954, + "step": 49037 + }, + { + "epoch": 18.11, + "learning_rate": 1.1898756004434044e-06, + "loss": 0.2601, + "step": 49038 + }, + { + "epoch": 18.11, + "learning_rate": 1.189838650080059e-06, + "loss": 0.3267, + "step": 49039 + }, + { + "epoch": 18.11, + "learning_rate": 1.189801699716714e-06, + "loss": 0.2067, + "step": 49040 + }, + { + "epoch": 18.11, + "learning_rate": 1.1897647493533688e-06, + "loss": 0.2429, + "step": 49041 + }, + { + "epoch": 18.11, + "learning_rate": 1.1897277989900234e-06, + "loss": 0.2212, + "step": 49042 + }, + { + "epoch": 18.11, + "learning_rate": 1.1896908486266782e-06, + "loss": 0.289, + "step": 49043 + }, + { + "epoch": 18.11, + "learning_rate": 1.1896538982633329e-06, + "loss": 0.2211, + "step": 49044 + }, + { + "epoch": 18.11, + "learning_rate": 1.1896169478999877e-06, + "loss": 0.2334, + "step": 49045 + }, + { + "epoch": 18.11, + "learning_rate": 1.1895799975366424e-06, + "loss": 0.2213, + "step": 49046 + }, + { + "epoch": 18.11, + "learning_rate": 1.1895430471732972e-06, + "loss": 0.2212, + "step": 49047 + }, + { + "epoch": 18.11, + "learning_rate": 1.189506096809952e-06, + "loss": 0.3506, + "step": 49048 + }, + { + "epoch": 18.11, + "learning_rate": 1.1894691464466067e-06, + "loss": 0.1826, + "step": 49049 + }, + { + "epoch": 18.11, + "learning_rate": 1.1894321960832615e-06, + "loss": 0.1992, + "step": 49050 + }, + { + "epoch": 18.11, + "learning_rate": 1.1893952457199164e-06, + "loss": 0.261, + "step": 49051 + }, + { + "epoch": 18.11, + "learning_rate": 1.189358295356571e-06, + "loss": 0.3701, + "step": 49052 + }, + { + "epoch": 18.11, + "learning_rate": 1.1893213449932259e-06, + "loss": 0.1754, + "step": 49053 + }, + { + "epoch": 18.11, + "learning_rate": 1.1892843946298805e-06, + "loss": 0.1579, + "step": 49054 + }, + { + "epoch": 18.11, + "learning_rate": 1.1892474442665354e-06, + "loss": 0.1178, + "step": 49055 + }, + { + "epoch": 18.12, + "learning_rate": 1.18921049390319e-06, + "loss": 0.3358, + "step": 49056 + }, + { + "epoch": 18.12, + "learning_rate": 1.1891735435398448e-06, + "loss": 0.3541, + "step": 49057 + }, + { + "epoch": 18.12, + "learning_rate": 1.1891365931764997e-06, + "loss": 0.2786, + "step": 49058 + }, + { + "epoch": 18.12, + "learning_rate": 1.1890996428131543e-06, + "loss": 0.2265, + "step": 49059 + }, + { + "epoch": 18.12, + "learning_rate": 1.1890626924498092e-06, + "loss": 0.1906, + "step": 49060 + }, + { + "epoch": 18.12, + "learning_rate": 1.1890257420864638e-06, + "loss": 0.2511, + "step": 49061 + }, + { + "epoch": 18.12, + "learning_rate": 1.1889887917231187e-06, + "loss": 0.1803, + "step": 49062 + }, + { + "epoch": 18.12, + "learning_rate": 1.1889518413597735e-06, + "loss": 0.2312, + "step": 49063 + }, + { + "epoch": 18.12, + "learning_rate": 1.1889148909964281e-06, + "loss": 0.2321, + "step": 49064 + }, + { + "epoch": 18.12, + "learning_rate": 1.1888779406330828e-06, + "loss": 0.1769, + "step": 49065 + }, + { + "epoch": 18.12, + "learning_rate": 1.1888409902697376e-06, + "loss": 0.3138, + "step": 49066 + }, + { + "epoch": 18.12, + "learning_rate": 1.1888040399063925e-06, + "loss": 0.2678, + "step": 49067 + }, + { + "epoch": 18.12, + "learning_rate": 1.1887670895430471e-06, + "loss": 0.2278, + "step": 49068 + }, + { + "epoch": 18.12, + "learning_rate": 1.188730139179702e-06, + "loss": 0.1985, + "step": 49069 + }, + { + "epoch": 18.12, + "learning_rate": 1.1886931888163568e-06, + "loss": 0.1924, + "step": 49070 + }, + { + "epoch": 18.12, + "learning_rate": 1.1886562384530114e-06, + "loss": 0.2841, + "step": 49071 + }, + { + "epoch": 18.12, + "learning_rate": 1.1886192880896663e-06, + "loss": 0.2979, + "step": 49072 + }, + { + "epoch": 18.12, + "learning_rate": 1.1885823377263211e-06, + "loss": 0.1305, + "step": 49073 + }, + { + "epoch": 18.12, + "learning_rate": 1.1885453873629758e-06, + "loss": 0.116, + "step": 49074 + }, + { + "epoch": 18.12, + "learning_rate": 1.1885084369996304e-06, + "loss": 0.1465, + "step": 49075 + }, + { + "epoch": 18.12, + "learning_rate": 1.1884714866362853e-06, + "loss": 0.3258, + "step": 49076 + }, + { + "epoch": 18.12, + "learning_rate": 1.1884345362729401e-06, + "loss": 0.1989, + "step": 49077 + }, + { + "epoch": 18.12, + "learning_rate": 1.1883975859095947e-06, + "loss": 0.2225, + "step": 49078 + }, + { + "epoch": 18.12, + "learning_rate": 1.1883606355462496e-06, + "loss": 0.1793, + "step": 49079 + }, + { + "epoch": 18.12, + "learning_rate": 1.1883236851829044e-06, + "loss": 0.2477, + "step": 49080 + }, + { + "epoch": 18.12, + "learning_rate": 1.188286734819559e-06, + "loss": 0.1984, + "step": 49081 + }, + { + "epoch": 18.12, + "learning_rate": 1.188249784456214e-06, + "loss": 0.2479, + "step": 49082 + }, + { + "epoch": 18.13, + "learning_rate": 1.1882128340928688e-06, + "loss": 0.2473, + "step": 49083 + }, + { + "epoch": 18.13, + "learning_rate": 1.1881758837295232e-06, + "loss": 0.2414, + "step": 49084 + }, + { + "epoch": 18.13, + "learning_rate": 1.188138933366178e-06, + "loss": 0.2182, + "step": 49085 + }, + { + "epoch": 18.13, + "learning_rate": 1.188101983002833e-06, + "loss": 0.1432, + "step": 49086 + }, + { + "epoch": 18.13, + "learning_rate": 1.1880650326394875e-06, + "loss": 0.2176, + "step": 49087 + }, + { + "epoch": 18.13, + "learning_rate": 1.1880280822761424e-06, + "loss": 0.2209, + "step": 49088 + }, + { + "epoch": 18.13, + "learning_rate": 1.1879911319127972e-06, + "loss": 0.1265, + "step": 49089 + }, + { + "epoch": 18.13, + "learning_rate": 1.1879541815494519e-06, + "loss": 0.2909, + "step": 49090 + }, + { + "epoch": 18.13, + "learning_rate": 1.1879172311861067e-06, + "loss": 0.231, + "step": 49091 + }, + { + "epoch": 18.13, + "learning_rate": 1.1878802808227616e-06, + "loss": 0.1422, + "step": 49092 + }, + { + "epoch": 18.13, + "learning_rate": 1.1878433304594162e-06, + "loss": 0.1866, + "step": 49093 + }, + { + "epoch": 18.13, + "learning_rate": 1.1878063800960708e-06, + "loss": 0.195, + "step": 49094 + }, + { + "epoch": 18.13, + "learning_rate": 1.1877694297327257e-06, + "loss": 0.1558, + "step": 49095 + }, + { + "epoch": 18.13, + "learning_rate": 1.1877324793693805e-06, + "loss": 0.1682, + "step": 49096 + }, + { + "epoch": 18.13, + "learning_rate": 1.1876955290060352e-06, + "loss": 0.3338, + "step": 49097 + }, + { + "epoch": 18.13, + "learning_rate": 1.18765857864269e-06, + "loss": 0.362, + "step": 49098 + }, + { + "epoch": 18.13, + "learning_rate": 1.1876216282793449e-06, + "loss": 0.2433, + "step": 49099 + }, + { + "epoch": 18.13, + "learning_rate": 1.1875846779159995e-06, + "loss": 0.1764, + "step": 49100 + }, + { + "epoch": 18.13, + "learning_rate": 1.1875477275526544e-06, + "loss": 0.2155, + "step": 49101 + }, + { + "epoch": 18.13, + "learning_rate": 1.1875107771893092e-06, + "loss": 0.2766, + "step": 49102 + }, + { + "epoch": 18.13, + "learning_rate": 1.1874738268259638e-06, + "loss": 0.1487, + "step": 49103 + }, + { + "epoch": 18.13, + "learning_rate": 1.1874368764626185e-06, + "loss": 0.2445, + "step": 49104 + }, + { + "epoch": 18.13, + "learning_rate": 1.1873999260992733e-06, + "loss": 0.3092, + "step": 49105 + }, + { + "epoch": 18.13, + "learning_rate": 1.187362975735928e-06, + "loss": 0.1839, + "step": 49106 + }, + { + "epoch": 18.13, + "learning_rate": 1.1873260253725828e-06, + "loss": 0.2251, + "step": 49107 + }, + { + "epoch": 18.13, + "learning_rate": 1.1872890750092377e-06, + "loss": 0.2279, + "step": 49108 + }, + { + "epoch": 18.13, + "learning_rate": 1.1872521246458923e-06, + "loss": 0.1802, + "step": 49109 + }, + { + "epoch": 18.14, + "learning_rate": 1.1872151742825471e-06, + "loss": 0.1842, + "step": 49110 + }, + { + "epoch": 18.14, + "learning_rate": 1.187178223919202e-06, + "loss": 0.262, + "step": 49111 + }, + { + "epoch": 18.14, + "learning_rate": 1.1871412735558566e-06, + "loss": 0.2483, + "step": 49112 + }, + { + "epoch": 18.14, + "learning_rate": 1.1871043231925115e-06, + "loss": 0.2529, + "step": 49113 + }, + { + "epoch": 18.14, + "learning_rate": 1.1870673728291661e-06, + "loss": 0.1877, + "step": 49114 + }, + { + "epoch": 18.14, + "learning_rate": 1.187030422465821e-06, + "loss": 0.2126, + "step": 49115 + }, + { + "epoch": 18.14, + "learning_rate": 1.1869934721024756e-06, + "loss": 0.3422, + "step": 49116 + }, + { + "epoch": 18.14, + "learning_rate": 1.1869565217391304e-06, + "loss": 0.3211, + "step": 49117 + }, + { + "epoch": 18.14, + "learning_rate": 1.1869195713757853e-06, + "loss": 0.1253, + "step": 49118 + }, + { + "epoch": 18.14, + "learning_rate": 1.18688262101244e-06, + "loss": 0.227, + "step": 49119 + }, + { + "epoch": 18.14, + "learning_rate": 1.1868456706490948e-06, + "loss": 0.3371, + "step": 49120 + }, + { + "epoch": 18.14, + "learning_rate": 1.1868087202857496e-06, + "loss": 0.235, + "step": 49121 + }, + { + "epoch": 18.14, + "learning_rate": 1.1867717699224043e-06, + "loss": 0.4232, + "step": 49122 + }, + { + "epoch": 18.14, + "learning_rate": 1.1867348195590591e-06, + "loss": 0.2319, + "step": 49123 + }, + { + "epoch": 18.14, + "learning_rate": 1.1866978691957137e-06, + "loss": 0.1831, + "step": 49124 + }, + { + "epoch": 18.14, + "learning_rate": 1.1866609188323686e-06, + "loss": 0.2503, + "step": 49125 + }, + { + "epoch": 18.14, + "learning_rate": 1.1866239684690232e-06, + "loss": 0.2575, + "step": 49126 + }, + { + "epoch": 18.14, + "learning_rate": 1.186587018105678e-06, + "loss": 0.2612, + "step": 49127 + }, + { + "epoch": 18.14, + "learning_rate": 1.186550067742333e-06, + "loss": 0.2019, + "step": 49128 + }, + { + "epoch": 18.14, + "learning_rate": 1.1865131173789876e-06, + "loss": 0.231, + "step": 49129 + }, + { + "epoch": 18.14, + "learning_rate": 1.1864761670156424e-06, + "loss": 0.2787, + "step": 49130 + }, + { + "epoch": 18.14, + "learning_rate": 1.1864392166522973e-06, + "loss": 0.1572, + "step": 49131 + }, + { + "epoch": 18.14, + "learning_rate": 1.1864022662889519e-06, + "loss": 0.2438, + "step": 49132 + }, + { + "epoch": 18.14, + "learning_rate": 1.1863653159256067e-06, + "loss": 0.2524, + "step": 49133 + }, + { + "epoch": 18.14, + "learning_rate": 1.1863283655622614e-06, + "loss": 0.1654, + "step": 49134 + }, + { + "epoch": 18.14, + "learning_rate": 1.186291415198916e-06, + "loss": 0.1522, + "step": 49135 + }, + { + "epoch": 18.14, + "learning_rate": 1.1862544648355709e-06, + "loss": 0.3285, + "step": 49136 + }, + { + "epoch": 18.15, + "learning_rate": 1.1862175144722257e-06, + "loss": 0.164, + "step": 49137 + }, + { + "epoch": 18.15, + "learning_rate": 1.1861805641088803e-06, + "loss": 0.2693, + "step": 49138 + }, + { + "epoch": 18.15, + "learning_rate": 1.1861436137455352e-06, + "loss": 0.2254, + "step": 49139 + }, + { + "epoch": 18.15, + "learning_rate": 1.18610666338219e-06, + "loss": 0.2652, + "step": 49140 + }, + { + "epoch": 18.15, + "learning_rate": 1.1860697130188447e-06, + "loss": 0.1886, + "step": 49141 + }, + { + "epoch": 18.15, + "learning_rate": 1.1860327626554995e-06, + "loss": 0.0883, + "step": 49142 + }, + { + "epoch": 18.15, + "learning_rate": 1.1859958122921544e-06, + "loss": 0.2563, + "step": 49143 + }, + { + "epoch": 18.15, + "learning_rate": 1.185958861928809e-06, + "loss": 0.2056, + "step": 49144 + }, + { + "epoch": 18.15, + "learning_rate": 1.1859219115654636e-06, + "loss": 0.201, + "step": 49145 + }, + { + "epoch": 18.15, + "learning_rate": 1.1858849612021185e-06, + "loss": 0.3002, + "step": 49146 + }, + { + "epoch": 18.15, + "learning_rate": 1.1858480108387733e-06, + "loss": 0.2515, + "step": 49147 + }, + { + "epoch": 18.15, + "learning_rate": 1.185811060475428e-06, + "loss": 0.1596, + "step": 49148 + }, + { + "epoch": 18.15, + "learning_rate": 1.1857741101120828e-06, + "loss": 0.3216, + "step": 49149 + }, + { + "epoch": 18.15, + "learning_rate": 1.1857371597487377e-06, + "loss": 0.2173, + "step": 49150 + }, + { + "epoch": 18.15, + "learning_rate": 1.1857002093853923e-06, + "loss": 0.2202, + "step": 49151 + }, + { + "epoch": 18.15, + "learning_rate": 1.1856632590220472e-06, + "loss": 0.1567, + "step": 49152 + }, + { + "epoch": 18.15, + "learning_rate": 1.185626308658702e-06, + "loss": 0.2127, + "step": 49153 + }, + { + "epoch": 18.15, + "learning_rate": 1.1855893582953564e-06, + "loss": 0.3129, + "step": 49154 + }, + { + "epoch": 18.15, + "learning_rate": 1.1855524079320113e-06, + "loss": 0.1756, + "step": 49155 + }, + { + "epoch": 18.15, + "learning_rate": 1.1855154575686661e-06, + "loss": 0.2608, + "step": 49156 + }, + { + "epoch": 18.15, + "learning_rate": 1.1854785072053208e-06, + "loss": 0.2699, + "step": 49157 + }, + { + "epoch": 18.15, + "learning_rate": 1.1854415568419756e-06, + "loss": 0.1441, + "step": 49158 + }, + { + "epoch": 18.15, + "learning_rate": 1.1854046064786305e-06, + "loss": 0.2576, + "step": 49159 + }, + { + "epoch": 18.15, + "learning_rate": 1.185367656115285e-06, + "loss": 0.1965, + "step": 49160 + }, + { + "epoch": 18.15, + "learning_rate": 1.18533070575194e-06, + "loss": 0.2713, + "step": 49161 + }, + { + "epoch": 18.15, + "learning_rate": 1.1852937553885948e-06, + "loss": 0.3824, + "step": 49162 + }, + { + "epoch": 18.15, + "learning_rate": 1.1852568050252494e-06, + "loss": 0.1582, + "step": 49163 + }, + { + "epoch": 18.16, + "learning_rate": 1.185219854661904e-06, + "loss": 0.3268, + "step": 49164 + }, + { + "epoch": 18.16, + "learning_rate": 1.185182904298559e-06, + "loss": 0.1427, + "step": 49165 + }, + { + "epoch": 18.16, + "learning_rate": 1.1851459539352138e-06, + "loss": 0.213, + "step": 49166 + }, + { + "epoch": 18.16, + "learning_rate": 1.1851090035718684e-06, + "loss": 0.368, + "step": 49167 + }, + { + "epoch": 18.16, + "learning_rate": 1.1850720532085233e-06, + "loss": 0.1986, + "step": 49168 + }, + { + "epoch": 18.16, + "learning_rate": 1.185035102845178e-06, + "loss": 0.2126, + "step": 49169 + }, + { + "epoch": 18.16, + "learning_rate": 1.1849981524818327e-06, + "loss": 0.1181, + "step": 49170 + }, + { + "epoch": 18.16, + "learning_rate": 1.1849612021184876e-06, + "loss": 0.1093, + "step": 49171 + }, + { + "epoch": 18.16, + "learning_rate": 1.1849242517551424e-06, + "loss": 0.2031, + "step": 49172 + }, + { + "epoch": 18.16, + "learning_rate": 1.184887301391797e-06, + "loss": 0.1863, + "step": 49173 + }, + { + "epoch": 18.16, + "learning_rate": 1.1848503510284517e-06, + "loss": 0.1751, + "step": 49174 + }, + { + "epoch": 18.16, + "learning_rate": 1.1848134006651066e-06, + "loss": 0.1325, + "step": 49175 + }, + { + "epoch": 18.16, + "learning_rate": 1.1847764503017612e-06, + "loss": 0.2429, + "step": 49176 + }, + { + "epoch": 18.16, + "learning_rate": 1.184739499938416e-06, + "loss": 0.2587, + "step": 49177 + }, + { + "epoch": 18.16, + "learning_rate": 1.1847025495750709e-06, + "loss": 0.2271, + "step": 49178 + }, + { + "epoch": 18.16, + "learning_rate": 1.1846655992117255e-06, + "loss": 0.2637, + "step": 49179 + }, + { + "epoch": 18.16, + "learning_rate": 1.1846286488483804e-06, + "loss": 0.2143, + "step": 49180 + }, + { + "epoch": 18.16, + "learning_rate": 1.1845916984850352e-06, + "loss": 0.1989, + "step": 49181 + }, + { + "epoch": 18.16, + "learning_rate": 1.1845547481216899e-06, + "loss": 0.1719, + "step": 49182 + }, + { + "epoch": 18.16, + "learning_rate": 1.1845177977583447e-06, + "loss": 0.258, + "step": 49183 + }, + { + "epoch": 18.16, + "learning_rate": 1.1844808473949993e-06, + "loss": 0.1889, + "step": 49184 + }, + { + "epoch": 18.16, + "learning_rate": 1.1844438970316542e-06, + "loss": 0.1924, + "step": 49185 + }, + { + "epoch": 18.16, + "learning_rate": 1.1844069466683088e-06, + "loss": 0.3715, + "step": 49186 + }, + { + "epoch": 18.16, + "learning_rate": 1.1843699963049637e-06, + "loss": 0.3145, + "step": 49187 + }, + { + "epoch": 18.16, + "learning_rate": 1.1843330459416185e-06, + "loss": 0.1579, + "step": 49188 + }, + { + "epoch": 18.16, + "learning_rate": 1.1842960955782732e-06, + "loss": 0.2278, + "step": 49189 + }, + { + "epoch": 18.16, + "learning_rate": 1.184259145214928e-06, + "loss": 0.1499, + "step": 49190 + }, + { + "epoch": 18.17, + "learning_rate": 1.1842221948515829e-06, + "loss": 0.1671, + "step": 49191 + }, + { + "epoch": 18.17, + "learning_rate": 1.1841852444882375e-06, + "loss": 0.2732, + "step": 49192 + }, + { + "epoch": 18.17, + "learning_rate": 1.1841482941248923e-06, + "loss": 0.3499, + "step": 49193 + }, + { + "epoch": 18.17, + "learning_rate": 1.184111343761547e-06, + "loss": 0.258, + "step": 49194 + }, + { + "epoch": 18.17, + "learning_rate": 1.1840743933982018e-06, + "loss": 0.1517, + "step": 49195 + }, + { + "epoch": 18.17, + "learning_rate": 1.1840374430348565e-06, + "loss": 0.1515, + "step": 49196 + }, + { + "epoch": 18.17, + "learning_rate": 1.1840004926715113e-06, + "loss": 0.1863, + "step": 49197 + }, + { + "epoch": 18.17, + "learning_rate": 1.1839635423081662e-06, + "loss": 0.2401, + "step": 49198 + }, + { + "epoch": 18.17, + "learning_rate": 1.1839265919448208e-06, + "loss": 0.2674, + "step": 49199 + }, + { + "epoch": 18.17, + "learning_rate": 1.1838896415814756e-06, + "loss": 0.2736, + "step": 49200 + }, + { + "epoch": 18.17, + "learning_rate": 1.1838526912181305e-06, + "loss": 0.2545, + "step": 49201 + }, + { + "epoch": 18.17, + "learning_rate": 1.1838157408547851e-06, + "loss": 0.2561, + "step": 49202 + }, + { + "epoch": 18.17, + "learning_rate": 1.18377879049144e-06, + "loss": 0.1888, + "step": 49203 + }, + { + "epoch": 18.17, + "learning_rate": 1.1837418401280946e-06, + "loss": 0.3056, + "step": 49204 + }, + { + "epoch": 18.17, + "learning_rate": 1.1837048897647492e-06, + "loss": 0.3095, + "step": 49205 + }, + { + "epoch": 18.17, + "learning_rate": 1.183667939401404e-06, + "loss": 0.1304, + "step": 49206 + }, + { + "epoch": 18.17, + "learning_rate": 1.183630989038059e-06, + "loss": 0.2194, + "step": 49207 + }, + { + "epoch": 18.17, + "learning_rate": 1.1835940386747136e-06, + "loss": 0.4143, + "step": 49208 + }, + { + "epoch": 18.17, + "learning_rate": 1.1835570883113684e-06, + "loss": 0.2103, + "step": 49209 + }, + { + "epoch": 18.17, + "learning_rate": 1.1835201379480233e-06, + "loss": 0.3168, + "step": 49210 + }, + { + "epoch": 18.17, + "learning_rate": 1.183483187584678e-06, + "loss": 0.2171, + "step": 49211 + }, + { + "epoch": 18.17, + "learning_rate": 1.1834462372213328e-06, + "loss": 0.2449, + "step": 49212 + }, + { + "epoch": 18.17, + "learning_rate": 1.1834092868579876e-06, + "loss": 0.1539, + "step": 49213 + }, + { + "epoch": 18.17, + "learning_rate": 1.1833723364946422e-06, + "loss": 0.1846, + "step": 49214 + }, + { + "epoch": 18.17, + "learning_rate": 1.1833353861312969e-06, + "loss": 0.204, + "step": 49215 + }, + { + "epoch": 18.17, + "learning_rate": 1.1832984357679517e-06, + "loss": 0.3281, + "step": 49216 + }, + { + "epoch": 18.17, + "learning_rate": 1.1832614854046066e-06, + "loss": 0.1272, + "step": 49217 + }, + { + "epoch": 18.18, + "learning_rate": 1.1832245350412612e-06, + "loss": 0.131, + "step": 49218 + }, + { + "epoch": 18.18, + "learning_rate": 1.183187584677916e-06, + "loss": 0.288, + "step": 49219 + }, + { + "epoch": 18.18, + "learning_rate": 1.183150634314571e-06, + "loss": 0.3792, + "step": 49220 + }, + { + "epoch": 18.18, + "learning_rate": 1.1831136839512255e-06, + "loss": 0.2261, + "step": 49221 + }, + { + "epoch": 18.18, + "learning_rate": 1.1830767335878804e-06, + "loss": 0.3091, + "step": 49222 + }, + { + "epoch": 18.18, + "learning_rate": 1.1830397832245352e-06, + "loss": 0.1626, + "step": 49223 + }, + { + "epoch": 18.18, + "learning_rate": 1.1830028328611897e-06, + "loss": 0.2688, + "step": 49224 + }, + { + "epoch": 18.18, + "learning_rate": 1.1829658824978445e-06, + "loss": 0.254, + "step": 49225 + }, + { + "epoch": 18.18, + "learning_rate": 1.1829289321344994e-06, + "loss": 0.2178, + "step": 49226 + }, + { + "epoch": 18.18, + "learning_rate": 1.182891981771154e-06, + "loss": 0.2531, + "step": 49227 + }, + { + "epoch": 18.18, + "learning_rate": 1.1828550314078088e-06, + "loss": 0.3546, + "step": 49228 + }, + { + "epoch": 18.18, + "learning_rate": 1.1828180810444637e-06, + "loss": 0.3206, + "step": 49229 + }, + { + "epoch": 18.18, + "learning_rate": 1.1827811306811183e-06, + "loss": 0.2752, + "step": 49230 + }, + { + "epoch": 18.18, + "learning_rate": 1.1827441803177732e-06, + "loss": 0.1888, + "step": 49231 + }, + { + "epoch": 18.18, + "learning_rate": 1.182707229954428e-06, + "loss": 0.261, + "step": 49232 + }, + { + "epoch": 18.18, + "learning_rate": 1.1826702795910827e-06, + "loss": 0.263, + "step": 49233 + }, + { + "epoch": 18.18, + "learning_rate": 1.1826333292277373e-06, + "loss": 0.1254, + "step": 49234 + }, + { + "epoch": 18.18, + "learning_rate": 1.1825963788643922e-06, + "loss": 0.2841, + "step": 49235 + }, + { + "epoch": 18.18, + "learning_rate": 1.182559428501047e-06, + "loss": 0.2224, + "step": 49236 + }, + { + "epoch": 18.18, + "learning_rate": 1.1825224781377016e-06, + "loss": 0.3129, + "step": 49237 + }, + { + "epoch": 18.18, + "learning_rate": 1.1824855277743565e-06, + "loss": 0.3517, + "step": 49238 + }, + { + "epoch": 18.18, + "learning_rate": 1.1824485774110113e-06, + "loss": 0.2695, + "step": 49239 + }, + { + "epoch": 18.18, + "learning_rate": 1.182411627047666e-06, + "loss": 0.1884, + "step": 49240 + }, + { + "epoch": 18.18, + "learning_rate": 1.1823746766843208e-06, + "loss": 0.2801, + "step": 49241 + }, + { + "epoch": 18.18, + "learning_rate": 1.1823377263209757e-06, + "loss": 0.2501, + "step": 49242 + }, + { + "epoch": 18.18, + "learning_rate": 1.1823007759576303e-06, + "loss": 0.1752, + "step": 49243 + }, + { + "epoch": 18.18, + "learning_rate": 1.182263825594285e-06, + "loss": 0.3065, + "step": 49244 + }, + { + "epoch": 18.19, + "learning_rate": 1.1822268752309398e-06, + "loss": 0.1716, + "step": 49245 + }, + { + "epoch": 18.19, + "learning_rate": 1.1821899248675946e-06, + "loss": 0.2019, + "step": 49246 + }, + { + "epoch": 18.19, + "learning_rate": 1.1821529745042493e-06, + "loss": 0.2238, + "step": 49247 + }, + { + "epoch": 18.19, + "learning_rate": 1.1821160241409041e-06, + "loss": 0.1812, + "step": 49248 + }, + { + "epoch": 18.19, + "learning_rate": 1.1820790737775588e-06, + "loss": 0.3015, + "step": 49249 + }, + { + "epoch": 18.19, + "learning_rate": 1.1820421234142136e-06, + "loss": 0.146, + "step": 49250 + }, + { + "epoch": 18.19, + "learning_rate": 1.1820051730508685e-06, + "loss": 0.2451, + "step": 49251 + }, + { + "epoch": 18.19, + "learning_rate": 1.181968222687523e-06, + "loss": 0.1457, + "step": 49252 + }, + { + "epoch": 18.19, + "learning_rate": 1.1819312723241777e-06, + "loss": 0.2131, + "step": 49253 + }, + { + "epoch": 18.19, + "learning_rate": 1.1818943219608326e-06, + "loss": 0.3336, + "step": 49254 + }, + { + "epoch": 18.19, + "learning_rate": 1.1818573715974874e-06, + "loss": 0.3228, + "step": 49255 + }, + { + "epoch": 18.19, + "learning_rate": 1.181820421234142e-06, + "loss": 0.2456, + "step": 49256 + }, + { + "epoch": 18.19, + "learning_rate": 1.181783470870797e-06, + "loss": 0.1559, + "step": 49257 + }, + { + "epoch": 18.19, + "learning_rate": 1.1817465205074518e-06, + "loss": 0.1924, + "step": 49258 + }, + { + "epoch": 18.19, + "learning_rate": 1.1817095701441064e-06, + "loss": 0.2157, + "step": 49259 + }, + { + "epoch": 18.19, + "learning_rate": 1.1816726197807612e-06, + "loss": 0.1729, + "step": 49260 + }, + { + "epoch": 18.19, + "learning_rate": 1.181635669417416e-06, + "loss": 0.2709, + "step": 49261 + }, + { + "epoch": 18.19, + "learning_rate": 1.1815987190540707e-06, + "loss": 0.111, + "step": 49262 + }, + { + "epoch": 18.19, + "learning_rate": 1.1815617686907254e-06, + "loss": 0.1797, + "step": 49263 + }, + { + "epoch": 18.19, + "learning_rate": 1.1815248183273802e-06, + "loss": 0.177, + "step": 49264 + }, + { + "epoch": 18.19, + "learning_rate": 1.181487867964035e-06, + "loss": 0.2217, + "step": 49265 + }, + { + "epoch": 18.19, + "learning_rate": 1.1814509176006897e-06, + "loss": 0.1512, + "step": 49266 + }, + { + "epoch": 18.19, + "learning_rate": 1.1814139672373445e-06, + "loss": 0.2661, + "step": 49267 + }, + { + "epoch": 18.19, + "learning_rate": 1.1813770168739994e-06, + "loss": 0.1684, + "step": 49268 + }, + { + "epoch": 18.19, + "learning_rate": 1.181340066510654e-06, + "loss": 0.1062, + "step": 49269 + }, + { + "epoch": 18.19, + "learning_rate": 1.1813031161473089e-06, + "loss": 0.3001, + "step": 49270 + }, + { + "epoch": 18.19, + "learning_rate": 1.1812661657839637e-06, + "loss": 0.1266, + "step": 49271 + }, + { + "epoch": 18.19, + "learning_rate": 1.1812292154206184e-06, + "loss": 0.2193, + "step": 49272 + }, + { + "epoch": 18.2, + "learning_rate": 1.181192265057273e-06, + "loss": 0.2067, + "step": 49273 + }, + { + "epoch": 18.2, + "learning_rate": 1.1811553146939278e-06, + "loss": 0.2926, + "step": 49274 + }, + { + "epoch": 18.2, + "learning_rate": 1.1811183643305825e-06, + "loss": 0.2041, + "step": 49275 + }, + { + "epoch": 18.2, + "learning_rate": 1.1810814139672373e-06, + "loss": 0.256, + "step": 49276 + }, + { + "epoch": 18.2, + "learning_rate": 1.1810444636038922e-06, + "loss": 0.2878, + "step": 49277 + }, + { + "epoch": 18.2, + "learning_rate": 1.1810075132405468e-06, + "loss": 0.2076, + "step": 49278 + }, + { + "epoch": 18.2, + "learning_rate": 1.1809705628772017e-06, + "loss": 0.2339, + "step": 49279 + }, + { + "epoch": 18.2, + "learning_rate": 1.1809336125138565e-06, + "loss": 0.2058, + "step": 49280 + }, + { + "epoch": 18.2, + "learning_rate": 1.1808966621505111e-06, + "loss": 0.1934, + "step": 49281 + }, + { + "epoch": 18.2, + "learning_rate": 1.180859711787166e-06, + "loss": 0.2795, + "step": 49282 + }, + { + "epoch": 18.2, + "learning_rate": 1.1808227614238206e-06, + "loss": 0.1328, + "step": 49283 + }, + { + "epoch": 18.2, + "learning_rate": 1.1807858110604755e-06, + "loss": 0.2157, + "step": 49284 + }, + { + "epoch": 18.2, + "learning_rate": 1.1807488606971301e-06, + "loss": 0.2541, + "step": 49285 + }, + { + "epoch": 18.2, + "learning_rate": 1.180711910333785e-06, + "loss": 0.2956, + "step": 49286 + }, + { + "epoch": 18.2, + "learning_rate": 1.1806749599704398e-06, + "loss": 0.2203, + "step": 49287 + }, + { + "epoch": 18.2, + "learning_rate": 1.1806380096070944e-06, + "loss": 0.3874, + "step": 49288 + }, + { + "epoch": 18.2, + "learning_rate": 1.1806010592437493e-06, + "loss": 0.2771, + "step": 49289 + }, + { + "epoch": 18.2, + "learning_rate": 1.1805641088804041e-06, + "loss": 0.1817, + "step": 49290 + }, + { + "epoch": 18.2, + "learning_rate": 1.1805271585170588e-06, + "loss": 0.2219, + "step": 49291 + }, + { + "epoch": 18.2, + "learning_rate": 1.1804902081537136e-06, + "loss": 0.1542, + "step": 49292 + }, + { + "epoch": 18.2, + "learning_rate": 1.1804532577903683e-06, + "loss": 0.1651, + "step": 49293 + }, + { + "epoch": 18.2, + "learning_rate": 1.180416307427023e-06, + "loss": 0.106, + "step": 49294 + }, + { + "epoch": 18.2, + "learning_rate": 1.1803793570636777e-06, + "loss": 0.2523, + "step": 49295 + }, + { + "epoch": 18.2, + "learning_rate": 1.1803424067003326e-06, + "loss": 0.2408, + "step": 49296 + }, + { + "epoch": 18.2, + "learning_rate": 1.1803054563369872e-06, + "loss": 0.154, + "step": 49297 + }, + { + "epoch": 18.2, + "learning_rate": 1.180268505973642e-06, + "loss": 0.2913, + "step": 49298 + }, + { + "epoch": 18.2, + "learning_rate": 1.180231555610297e-06, + "loss": 0.2085, + "step": 49299 + }, + { + "epoch": 18.21, + "learning_rate": 1.1801946052469516e-06, + "loss": 0.1615, + "step": 49300 + }, + { + "epoch": 18.21, + "learning_rate": 1.1801576548836064e-06, + "loss": 0.2583, + "step": 49301 + }, + { + "epoch": 18.21, + "learning_rate": 1.1801207045202613e-06, + "loss": 0.2962, + "step": 49302 + }, + { + "epoch": 18.21, + "learning_rate": 1.180083754156916e-06, + "loss": 0.2533, + "step": 49303 + }, + { + "epoch": 18.21, + "learning_rate": 1.1800468037935705e-06, + "loss": 0.2867, + "step": 49304 + }, + { + "epoch": 18.21, + "learning_rate": 1.1800098534302254e-06, + "loss": 0.2529, + "step": 49305 + }, + { + "epoch": 18.21, + "learning_rate": 1.1799729030668802e-06, + "loss": 0.2924, + "step": 49306 + }, + { + "epoch": 18.21, + "learning_rate": 1.1799359527035349e-06, + "loss": 0.1215, + "step": 49307 + }, + { + "epoch": 18.21, + "learning_rate": 1.1798990023401897e-06, + "loss": 0.2227, + "step": 49308 + }, + { + "epoch": 18.21, + "learning_rate": 1.1798620519768446e-06, + "loss": 0.1641, + "step": 49309 + }, + { + "epoch": 18.21, + "learning_rate": 1.1798251016134992e-06, + "loss": 0.3687, + "step": 49310 + }, + { + "epoch": 18.21, + "learning_rate": 1.179788151250154e-06, + "loss": 0.2743, + "step": 49311 + }, + { + "epoch": 18.21, + "learning_rate": 1.179751200886809e-06, + "loss": 0.1542, + "step": 49312 + }, + { + "epoch": 18.21, + "learning_rate": 1.1797142505234635e-06, + "loss": 0.2684, + "step": 49313 + }, + { + "epoch": 18.21, + "learning_rate": 1.1796773001601182e-06, + "loss": 0.1713, + "step": 49314 + }, + { + "epoch": 18.21, + "learning_rate": 1.179640349796773e-06, + "loss": 0.2341, + "step": 49315 + }, + { + "epoch": 18.21, + "learning_rate": 1.1796033994334279e-06, + "loss": 0.2855, + "step": 49316 + }, + { + "epoch": 18.21, + "learning_rate": 1.1795664490700825e-06, + "loss": 0.2324, + "step": 49317 + }, + { + "epoch": 18.21, + "learning_rate": 1.1795294987067374e-06, + "loss": 0.1493, + "step": 49318 + }, + { + "epoch": 18.21, + "learning_rate": 1.1794925483433922e-06, + "loss": 0.228, + "step": 49319 + }, + { + "epoch": 18.21, + "learning_rate": 1.1794555979800468e-06, + "loss": 0.3004, + "step": 49320 + }, + { + "epoch": 18.21, + "learning_rate": 1.1794186476167017e-06, + "loss": 0.2011, + "step": 49321 + }, + { + "epoch": 18.21, + "learning_rate": 1.1793816972533565e-06, + "loss": 0.1837, + "step": 49322 + }, + { + "epoch": 18.21, + "learning_rate": 1.179344746890011e-06, + "loss": 0.2566, + "step": 49323 + }, + { + "epoch": 18.21, + "learning_rate": 1.1793077965266658e-06, + "loss": 0.2432, + "step": 49324 + }, + { + "epoch": 18.21, + "learning_rate": 1.1792708461633207e-06, + "loss": 0.2025, + "step": 49325 + }, + { + "epoch": 18.21, + "learning_rate": 1.1792338957999753e-06, + "loss": 0.1699, + "step": 49326 + }, + { + "epoch": 18.22, + "learning_rate": 1.1791969454366301e-06, + "loss": 0.2256, + "step": 49327 + }, + { + "epoch": 18.22, + "learning_rate": 1.179159995073285e-06, + "loss": 0.1944, + "step": 49328 + }, + { + "epoch": 18.22, + "learning_rate": 1.1791230447099396e-06, + "loss": 0.1401, + "step": 49329 + }, + { + "epoch": 18.22, + "learning_rate": 1.1790860943465945e-06, + "loss": 0.2697, + "step": 49330 + }, + { + "epoch": 18.22, + "learning_rate": 1.1790491439832493e-06, + "loss": 0.2465, + "step": 49331 + }, + { + "epoch": 18.22, + "learning_rate": 1.179012193619904e-06, + "loss": 0.1501, + "step": 49332 + }, + { + "epoch": 18.22, + "learning_rate": 1.1789752432565586e-06, + "loss": 0.2158, + "step": 49333 + }, + { + "epoch": 18.22, + "learning_rate": 1.1789382928932134e-06, + "loss": 0.3677, + "step": 49334 + }, + { + "epoch": 18.22, + "learning_rate": 1.1789013425298683e-06, + "loss": 0.2734, + "step": 49335 + }, + { + "epoch": 18.22, + "learning_rate": 1.178864392166523e-06, + "loss": 0.1088, + "step": 49336 + }, + { + "epoch": 18.22, + "learning_rate": 1.1788274418031778e-06, + "loss": 0.2584, + "step": 49337 + }, + { + "epoch": 18.22, + "learning_rate": 1.1787904914398326e-06, + "loss": 0.2851, + "step": 49338 + }, + { + "epoch": 18.22, + "learning_rate": 1.1787535410764873e-06, + "loss": 0.1829, + "step": 49339 + }, + { + "epoch": 18.22, + "learning_rate": 1.1787165907131421e-06, + "loss": 0.2521, + "step": 49340 + }, + { + "epoch": 18.22, + "learning_rate": 1.178679640349797e-06, + "loss": 0.1956, + "step": 49341 + }, + { + "epoch": 18.22, + "learning_rate": 1.1786426899864516e-06, + "loss": 0.1596, + "step": 49342 + }, + { + "epoch": 18.22, + "learning_rate": 1.1786057396231062e-06, + "loss": 0.3626, + "step": 49343 + }, + { + "epoch": 18.22, + "learning_rate": 1.178568789259761e-06, + "loss": 0.2324, + "step": 49344 + }, + { + "epoch": 18.22, + "learning_rate": 1.1785318388964157e-06, + "loss": 0.2663, + "step": 49345 + }, + { + "epoch": 18.22, + "learning_rate": 1.1784948885330706e-06, + "loss": 0.152, + "step": 49346 + }, + { + "epoch": 18.22, + "learning_rate": 1.1784579381697254e-06, + "loss": 0.3075, + "step": 49347 + }, + { + "epoch": 18.22, + "learning_rate": 1.17842098780638e-06, + "loss": 0.2239, + "step": 49348 + }, + { + "epoch": 18.22, + "learning_rate": 1.1783840374430349e-06, + "loss": 0.3484, + "step": 49349 + }, + { + "epoch": 18.22, + "learning_rate": 1.1783470870796897e-06, + "loss": 0.4001, + "step": 49350 + }, + { + "epoch": 18.22, + "learning_rate": 1.1783101367163444e-06, + "loss": 0.1868, + "step": 49351 + }, + { + "epoch": 18.22, + "learning_rate": 1.1782731863529992e-06, + "loss": 0.1563, + "step": 49352 + }, + { + "epoch": 18.22, + "learning_rate": 1.1782362359896539e-06, + "loss": 0.2699, + "step": 49353 + }, + { + "epoch": 18.23, + "learning_rate": 1.1781992856263087e-06, + "loss": 0.2173, + "step": 49354 + }, + { + "epoch": 18.23, + "learning_rate": 1.1781623352629633e-06, + "loss": 0.1938, + "step": 49355 + }, + { + "epoch": 18.23, + "learning_rate": 1.1781253848996182e-06, + "loss": 0.2523, + "step": 49356 + }, + { + "epoch": 18.23, + "learning_rate": 1.178088434536273e-06, + "loss": 0.1291, + "step": 49357 + }, + { + "epoch": 18.23, + "learning_rate": 1.1780514841729277e-06, + "loss": 0.3502, + "step": 49358 + }, + { + "epoch": 18.23, + "learning_rate": 1.1780145338095825e-06, + "loss": 0.2481, + "step": 49359 + }, + { + "epoch": 18.23, + "learning_rate": 1.1779775834462374e-06, + "loss": 0.1305, + "step": 49360 + }, + { + "epoch": 18.23, + "learning_rate": 1.177940633082892e-06, + "loss": 0.2743, + "step": 49361 + }, + { + "epoch": 18.23, + "learning_rate": 1.1779036827195469e-06, + "loss": 0.3347, + "step": 49362 + }, + { + "epoch": 18.23, + "learning_rate": 1.1778667323562015e-06, + "loss": 0.244, + "step": 49363 + }, + { + "epoch": 18.23, + "learning_rate": 1.1778297819928561e-06, + "loss": 0.303, + "step": 49364 + }, + { + "epoch": 18.23, + "learning_rate": 1.177792831629511e-06, + "loss": 0.2443, + "step": 49365 + }, + { + "epoch": 18.23, + "learning_rate": 1.1777558812661658e-06, + "loss": 0.356, + "step": 49366 + }, + { + "epoch": 18.23, + "learning_rate": 1.1777189309028205e-06, + "loss": 0.3338, + "step": 49367 + }, + { + "epoch": 18.23, + "learning_rate": 1.1776819805394753e-06, + "loss": 0.2328, + "step": 49368 + }, + { + "epoch": 18.23, + "learning_rate": 1.1776450301761302e-06, + "loss": 0.173, + "step": 49369 + }, + { + "epoch": 18.23, + "learning_rate": 1.1776080798127848e-06, + "loss": 0.2926, + "step": 49370 + }, + { + "epoch": 18.23, + "learning_rate": 1.1775711294494396e-06, + "loss": 0.2661, + "step": 49371 + }, + { + "epoch": 18.23, + "learning_rate": 1.1775341790860945e-06, + "loss": 0.1968, + "step": 49372 + }, + { + "epoch": 18.23, + "learning_rate": 1.1774972287227491e-06, + "loss": 0.1258, + "step": 49373 + }, + { + "epoch": 18.23, + "learning_rate": 1.1774602783594038e-06, + "loss": 0.2675, + "step": 49374 + }, + { + "epoch": 18.23, + "learning_rate": 1.1774233279960586e-06, + "loss": 0.2212, + "step": 49375 + }, + { + "epoch": 18.23, + "learning_rate": 1.1773863776327135e-06, + "loss": 0.21, + "step": 49376 + }, + { + "epoch": 18.23, + "learning_rate": 1.177349427269368e-06, + "loss": 0.36, + "step": 49377 + }, + { + "epoch": 18.23, + "learning_rate": 1.177312476906023e-06, + "loss": 0.2071, + "step": 49378 + }, + { + "epoch": 18.23, + "learning_rate": 1.1772755265426778e-06, + "loss": 0.1567, + "step": 49379 + }, + { + "epoch": 18.23, + "learning_rate": 1.1772385761793324e-06, + "loss": 0.1843, + "step": 49380 + }, + { + "epoch": 18.24, + "learning_rate": 1.1772016258159873e-06, + "loss": 0.2689, + "step": 49381 + }, + { + "epoch": 18.24, + "learning_rate": 1.1771646754526421e-06, + "loss": 0.3023, + "step": 49382 + }, + { + "epoch": 18.24, + "learning_rate": 1.1771277250892968e-06, + "loss": 0.2496, + "step": 49383 + }, + { + "epoch": 18.24, + "learning_rate": 1.1770907747259514e-06, + "loss": 0.1764, + "step": 49384 + }, + { + "epoch": 18.24, + "learning_rate": 1.1770538243626063e-06, + "loss": 0.2735, + "step": 49385 + }, + { + "epoch": 18.24, + "learning_rate": 1.177016873999261e-06, + "loss": 0.3865, + "step": 49386 + }, + { + "epoch": 18.24, + "learning_rate": 1.1769799236359157e-06, + "loss": 0.3369, + "step": 49387 + }, + { + "epoch": 18.24, + "learning_rate": 1.1769429732725706e-06, + "loss": 0.1751, + "step": 49388 + }, + { + "epoch": 18.24, + "learning_rate": 1.1769060229092254e-06, + "loss": 0.3922, + "step": 49389 + }, + { + "epoch": 18.24, + "learning_rate": 1.17686907254588e-06, + "loss": 0.2047, + "step": 49390 + }, + { + "epoch": 18.24, + "learning_rate": 1.176832122182535e-06, + "loss": 0.1839, + "step": 49391 + }, + { + "epoch": 18.24, + "learning_rate": 1.1767951718191898e-06, + "loss": 0.1598, + "step": 49392 + }, + { + "epoch": 18.24, + "learning_rate": 1.1767582214558442e-06, + "loss": 0.1392, + "step": 49393 + }, + { + "epoch": 18.24, + "learning_rate": 1.176721271092499e-06, + "loss": 0.2739, + "step": 49394 + }, + { + "epoch": 18.24, + "learning_rate": 1.1766843207291539e-06, + "loss": 0.2062, + "step": 49395 + }, + { + "epoch": 18.24, + "learning_rate": 1.1766473703658085e-06, + "loss": 0.1562, + "step": 49396 + }, + { + "epoch": 18.24, + "learning_rate": 1.1766104200024634e-06, + "loss": 0.1917, + "step": 49397 + }, + { + "epoch": 18.24, + "learning_rate": 1.1765734696391182e-06, + "loss": 0.3542, + "step": 49398 + }, + { + "epoch": 18.24, + "learning_rate": 1.1765365192757729e-06, + "loss": 0.2223, + "step": 49399 + }, + { + "epoch": 18.24, + "learning_rate": 1.1764995689124277e-06, + "loss": 0.1874, + "step": 49400 + }, + { + "epoch": 18.24, + "learning_rate": 1.1764626185490826e-06, + "loss": 0.4546, + "step": 49401 + }, + { + "epoch": 18.24, + "learning_rate": 1.1764256681857372e-06, + "loss": 0.1717, + "step": 49402 + }, + { + "epoch": 18.24, + "learning_rate": 1.1763887178223918e-06, + "loss": 0.2705, + "step": 49403 + }, + { + "epoch": 18.24, + "learning_rate": 1.1763517674590467e-06, + "loss": 0.3351, + "step": 49404 + }, + { + "epoch": 18.24, + "learning_rate": 1.1763148170957015e-06, + "loss": 0.1849, + "step": 49405 + }, + { + "epoch": 18.24, + "learning_rate": 1.1762778667323562e-06, + "loss": 0.3088, + "step": 49406 + }, + { + "epoch": 18.24, + "learning_rate": 1.176240916369011e-06, + "loss": 0.1573, + "step": 49407 + }, + { + "epoch": 18.25, + "learning_rate": 1.1762039660056659e-06, + "loss": 0.2451, + "step": 49408 + }, + { + "epoch": 18.25, + "learning_rate": 1.1761670156423205e-06, + "loss": 0.105, + "step": 49409 + }, + { + "epoch": 18.25, + "learning_rate": 1.1761300652789753e-06, + "loss": 0.247, + "step": 49410 + }, + { + "epoch": 18.25, + "learning_rate": 1.1760931149156302e-06, + "loss": 0.186, + "step": 49411 + }, + { + "epoch": 18.25, + "learning_rate": 1.1760561645522848e-06, + "loss": 0.2418, + "step": 49412 + }, + { + "epoch": 18.25, + "learning_rate": 1.1760192141889395e-06, + "loss": 0.2055, + "step": 49413 + }, + { + "epoch": 18.25, + "learning_rate": 1.1759822638255943e-06, + "loss": 0.1504, + "step": 49414 + }, + { + "epoch": 18.25, + "learning_rate": 1.175945313462249e-06, + "loss": 0.3397, + "step": 49415 + }, + { + "epoch": 18.25, + "learning_rate": 1.1759083630989038e-06, + "loss": 0.202, + "step": 49416 + }, + { + "epoch": 18.25, + "learning_rate": 1.1758714127355586e-06, + "loss": 0.2214, + "step": 49417 + }, + { + "epoch": 18.25, + "learning_rate": 1.1758344623722133e-06, + "loss": 0.2182, + "step": 49418 + }, + { + "epoch": 18.25, + "learning_rate": 1.1757975120088681e-06, + "loss": 0.1739, + "step": 49419 + }, + { + "epoch": 18.25, + "learning_rate": 1.175760561645523e-06, + "loss": 0.1983, + "step": 49420 + }, + { + "epoch": 18.25, + "learning_rate": 1.1757236112821776e-06, + "loss": 0.1618, + "step": 49421 + }, + { + "epoch": 18.25, + "learning_rate": 1.1756866609188325e-06, + "loss": 0.2167, + "step": 49422 + }, + { + "epoch": 18.25, + "learning_rate": 1.175649710555487e-06, + "loss": 0.249, + "step": 49423 + }, + { + "epoch": 18.25, + "learning_rate": 1.175612760192142e-06, + "loss": 0.4036, + "step": 49424 + }, + { + "epoch": 18.25, + "learning_rate": 1.1755758098287966e-06, + "loss": 0.1616, + "step": 49425 + }, + { + "epoch": 18.25, + "learning_rate": 1.1755388594654514e-06, + "loss": 0.2745, + "step": 49426 + }, + { + "epoch": 18.25, + "learning_rate": 1.1755019091021063e-06, + "loss": 0.1874, + "step": 49427 + }, + { + "epoch": 18.25, + "learning_rate": 1.175464958738761e-06, + "loss": 0.2312, + "step": 49428 + }, + { + "epoch": 18.25, + "learning_rate": 1.1754280083754158e-06, + "loss": 0.1661, + "step": 49429 + }, + { + "epoch": 18.25, + "learning_rate": 1.1753910580120706e-06, + "loss": 0.2123, + "step": 49430 + }, + { + "epoch": 18.25, + "learning_rate": 1.1753541076487252e-06, + "loss": 0.1696, + "step": 49431 + }, + { + "epoch": 18.25, + "learning_rate": 1.17531715728538e-06, + "loss": 0.1447, + "step": 49432 + }, + { + "epoch": 18.25, + "learning_rate": 1.1752802069220347e-06, + "loss": 0.287, + "step": 49433 + }, + { + "epoch": 18.25, + "learning_rate": 1.1752432565586896e-06, + "loss": 0.1708, + "step": 49434 + }, + { + "epoch": 18.26, + "learning_rate": 1.1752063061953442e-06, + "loss": 0.2432, + "step": 49435 + }, + { + "epoch": 18.26, + "learning_rate": 1.175169355831999e-06, + "loss": 0.1727, + "step": 49436 + }, + { + "epoch": 18.26, + "learning_rate": 1.1751324054686537e-06, + "loss": 0.2065, + "step": 49437 + }, + { + "epoch": 18.26, + "learning_rate": 1.1750954551053085e-06, + "loss": 0.2074, + "step": 49438 + }, + { + "epoch": 18.26, + "learning_rate": 1.1750585047419634e-06, + "loss": 0.31, + "step": 49439 + }, + { + "epoch": 18.26, + "learning_rate": 1.175021554378618e-06, + "loss": 0.1811, + "step": 49440 + }, + { + "epoch": 18.26, + "learning_rate": 1.1749846040152729e-06, + "loss": 0.2863, + "step": 49441 + }, + { + "epoch": 18.26, + "learning_rate": 1.1749476536519277e-06, + "loss": 0.3643, + "step": 49442 + }, + { + "epoch": 18.26, + "learning_rate": 1.1749107032885824e-06, + "loss": 0.2556, + "step": 49443 + }, + { + "epoch": 18.26, + "learning_rate": 1.174873752925237e-06, + "loss": 0.2644, + "step": 49444 + }, + { + "epoch": 18.26, + "learning_rate": 1.1748368025618918e-06, + "loss": 0.2469, + "step": 49445 + }, + { + "epoch": 18.26, + "learning_rate": 1.1747998521985467e-06, + "loss": 0.2533, + "step": 49446 + }, + { + "epoch": 18.26, + "learning_rate": 1.1747629018352013e-06, + "loss": 0.3188, + "step": 49447 + }, + { + "epoch": 18.26, + "learning_rate": 1.1747259514718562e-06, + "loss": 0.2592, + "step": 49448 + }, + { + "epoch": 18.26, + "learning_rate": 1.174689001108511e-06, + "loss": 0.1749, + "step": 49449 + }, + { + "epoch": 18.26, + "learning_rate": 1.1746520507451657e-06, + "loss": 0.2675, + "step": 49450 + }, + { + "epoch": 18.26, + "learning_rate": 1.1746151003818205e-06, + "loss": 0.1503, + "step": 49451 + }, + { + "epoch": 18.26, + "learning_rate": 1.1745781500184754e-06, + "loss": 0.1043, + "step": 49452 + }, + { + "epoch": 18.26, + "learning_rate": 1.17454119965513e-06, + "loss": 0.1308, + "step": 49453 + }, + { + "epoch": 18.26, + "learning_rate": 1.1745042492917846e-06, + "loss": 0.1956, + "step": 49454 + }, + { + "epoch": 18.26, + "learning_rate": 1.1744672989284395e-06, + "loss": 0.1987, + "step": 49455 + }, + { + "epoch": 18.26, + "learning_rate": 1.1744303485650943e-06, + "loss": 0.274, + "step": 49456 + }, + { + "epoch": 18.26, + "learning_rate": 1.174393398201749e-06, + "loss": 0.1332, + "step": 49457 + }, + { + "epoch": 18.26, + "learning_rate": 1.1743564478384038e-06, + "loss": 0.1962, + "step": 49458 + }, + { + "epoch": 18.26, + "learning_rate": 1.1743194974750587e-06, + "loss": 0.2673, + "step": 49459 + }, + { + "epoch": 18.26, + "learning_rate": 1.1742825471117133e-06, + "loss": 0.0855, + "step": 49460 + }, + { + "epoch": 18.26, + "learning_rate": 1.1742455967483681e-06, + "loss": 0.2267, + "step": 49461 + }, + { + "epoch": 18.27, + "learning_rate": 1.1742086463850228e-06, + "loss": 0.2012, + "step": 49462 + }, + { + "epoch": 18.27, + "learning_rate": 1.1741716960216774e-06, + "loss": 0.129, + "step": 49463 + }, + { + "epoch": 18.27, + "learning_rate": 1.1741347456583323e-06, + "loss": 0.2071, + "step": 49464 + }, + { + "epoch": 18.27, + "learning_rate": 1.1740977952949871e-06, + "loss": 0.2017, + "step": 49465 + }, + { + "epoch": 18.27, + "learning_rate": 1.1740608449316418e-06, + "loss": 0.1872, + "step": 49466 + }, + { + "epoch": 18.27, + "learning_rate": 1.1740238945682966e-06, + "loss": 0.1677, + "step": 49467 + }, + { + "epoch": 18.27, + "learning_rate": 1.1739869442049515e-06, + "loss": 0.2845, + "step": 49468 + }, + { + "epoch": 18.27, + "learning_rate": 1.173949993841606e-06, + "loss": 0.1795, + "step": 49469 + }, + { + "epoch": 18.27, + "learning_rate": 1.173913043478261e-06, + "loss": 0.2115, + "step": 49470 + }, + { + "epoch": 18.27, + "learning_rate": 1.1738760931149158e-06, + "loss": 0.1777, + "step": 49471 + }, + { + "epoch": 18.27, + "learning_rate": 1.1738391427515704e-06, + "loss": 0.1936, + "step": 49472 + }, + { + "epoch": 18.27, + "learning_rate": 1.173802192388225e-06, + "loss": 0.2454, + "step": 49473 + }, + { + "epoch": 18.27, + "learning_rate": 1.17376524202488e-06, + "loss": 0.219, + "step": 49474 + }, + { + "epoch": 18.27, + "learning_rate": 1.1737282916615348e-06, + "loss": 0.1953, + "step": 49475 + }, + { + "epoch": 18.27, + "learning_rate": 1.1736913412981894e-06, + "loss": 0.2129, + "step": 49476 + }, + { + "epoch": 18.27, + "learning_rate": 1.1736543909348442e-06, + "loss": 0.2706, + "step": 49477 + }, + { + "epoch": 18.27, + "learning_rate": 1.173617440571499e-06, + "loss": 0.2235, + "step": 49478 + }, + { + "epoch": 18.27, + "learning_rate": 1.1735804902081537e-06, + "loss": 0.2533, + "step": 49479 + }, + { + "epoch": 18.27, + "learning_rate": 1.1735435398448086e-06, + "loss": 0.1966, + "step": 49480 + }, + { + "epoch": 18.27, + "learning_rate": 1.1735065894814634e-06, + "loss": 0.2514, + "step": 49481 + }, + { + "epoch": 18.27, + "learning_rate": 1.1734696391181178e-06, + "loss": 0.1804, + "step": 49482 + }, + { + "epoch": 18.27, + "learning_rate": 1.1734326887547727e-06, + "loss": 0.2221, + "step": 49483 + }, + { + "epoch": 18.27, + "learning_rate": 1.1733957383914275e-06, + "loss": 0.2637, + "step": 49484 + }, + { + "epoch": 18.27, + "learning_rate": 1.1733587880280822e-06, + "loss": 0.2585, + "step": 49485 + }, + { + "epoch": 18.27, + "learning_rate": 1.173321837664737e-06, + "loss": 0.23, + "step": 49486 + }, + { + "epoch": 18.27, + "learning_rate": 1.1732848873013919e-06, + "loss": 0.1466, + "step": 49487 + }, + { + "epoch": 18.27, + "learning_rate": 1.1732479369380465e-06, + "loss": 0.2616, + "step": 49488 + }, + { + "epoch": 18.28, + "learning_rate": 1.1732109865747014e-06, + "loss": 0.2179, + "step": 49489 + }, + { + "epoch": 18.28, + "learning_rate": 1.1731740362113562e-06, + "loss": 0.2009, + "step": 49490 + }, + { + "epoch": 18.28, + "learning_rate": 1.1731370858480108e-06, + "loss": 0.3654, + "step": 49491 + }, + { + "epoch": 18.28, + "learning_rate": 1.1731001354846655e-06, + "loss": 0.2775, + "step": 49492 + }, + { + "epoch": 18.28, + "learning_rate": 1.1730631851213203e-06, + "loss": 0.1744, + "step": 49493 + }, + { + "epoch": 18.28, + "learning_rate": 1.1730262347579752e-06, + "loss": 0.2377, + "step": 49494 + }, + { + "epoch": 18.28, + "learning_rate": 1.1729892843946298e-06, + "loss": 0.2106, + "step": 49495 + }, + { + "epoch": 18.28, + "learning_rate": 1.1729523340312847e-06, + "loss": 0.1505, + "step": 49496 + }, + { + "epoch": 18.28, + "learning_rate": 1.1729153836679395e-06, + "loss": 0.1414, + "step": 49497 + }, + { + "epoch": 18.28, + "learning_rate": 1.1728784333045941e-06, + "loss": 0.1822, + "step": 49498 + }, + { + "epoch": 18.28, + "learning_rate": 1.172841482941249e-06, + "loss": 0.3938, + "step": 49499 + }, + { + "epoch": 18.28, + "learning_rate": 1.1728045325779038e-06, + "loss": 0.1771, + "step": 49500 + }, + { + "epoch": 18.28, + "learning_rate": 1.1727675822145585e-06, + "loss": 0.261, + "step": 49501 + }, + { + "epoch": 18.28, + "learning_rate": 1.1727306318512131e-06, + "loss": 0.1978, + "step": 49502 + }, + { + "epoch": 18.28, + "learning_rate": 1.172693681487868e-06, + "loss": 0.1374, + "step": 49503 + }, + { + "epoch": 18.28, + "learning_rate": 1.1726567311245228e-06, + "loss": 0.2289, + "step": 49504 + }, + { + "epoch": 18.28, + "learning_rate": 1.1726197807611774e-06, + "loss": 0.2126, + "step": 49505 + }, + { + "epoch": 18.28, + "learning_rate": 1.1725828303978323e-06, + "loss": 0.202, + "step": 49506 + }, + { + "epoch": 18.28, + "learning_rate": 1.1725458800344871e-06, + "loss": 0.1691, + "step": 49507 + }, + { + "epoch": 18.28, + "learning_rate": 1.1725089296711418e-06, + "loss": 0.2338, + "step": 49508 + }, + { + "epoch": 18.28, + "learning_rate": 1.1724719793077966e-06, + "loss": 0.246, + "step": 49509 + }, + { + "epoch": 18.28, + "learning_rate": 1.1724350289444515e-06, + "loss": 0.2485, + "step": 49510 + }, + { + "epoch": 18.28, + "learning_rate": 1.1723980785811061e-06, + "loss": 0.3159, + "step": 49511 + }, + { + "epoch": 18.28, + "learning_rate": 1.1723611282177607e-06, + "loss": 0.2816, + "step": 49512 + }, + { + "epoch": 18.28, + "learning_rate": 1.1723241778544156e-06, + "loss": 0.1838, + "step": 49513 + }, + { + "epoch": 18.28, + "learning_rate": 1.1722872274910702e-06, + "loss": 0.1606, + "step": 49514 + }, + { + "epoch": 18.28, + "learning_rate": 1.172250277127725e-06, + "loss": 0.1613, + "step": 49515 + }, + { + "epoch": 18.29, + "learning_rate": 1.17221332676438e-06, + "loss": 0.2418, + "step": 49516 + }, + { + "epoch": 18.29, + "learning_rate": 1.1721763764010346e-06, + "loss": 0.1981, + "step": 49517 + }, + { + "epoch": 18.29, + "learning_rate": 1.1721394260376894e-06, + "loss": 0.2069, + "step": 49518 + }, + { + "epoch": 18.29, + "learning_rate": 1.1721024756743443e-06, + "loss": 0.1941, + "step": 49519 + }, + { + "epoch": 18.29, + "learning_rate": 1.172065525310999e-06, + "loss": 0.131, + "step": 49520 + }, + { + "epoch": 18.29, + "learning_rate": 1.1720285749476537e-06, + "loss": 0.2542, + "step": 49521 + }, + { + "epoch": 18.29, + "learning_rate": 1.1719916245843084e-06, + "loss": 0.2503, + "step": 49522 + }, + { + "epoch": 18.29, + "learning_rate": 1.1719546742209632e-06, + "loss": 0.5103, + "step": 49523 + }, + { + "epoch": 18.29, + "learning_rate": 1.1719177238576179e-06, + "loss": 0.1976, + "step": 49524 + }, + { + "epoch": 18.29, + "learning_rate": 1.1718807734942727e-06, + "loss": 0.1748, + "step": 49525 + }, + { + "epoch": 18.29, + "learning_rate": 1.1718438231309276e-06, + "loss": 0.3228, + "step": 49526 + }, + { + "epoch": 18.29, + "learning_rate": 1.1718068727675822e-06, + "loss": 0.2129, + "step": 49527 + }, + { + "epoch": 18.29, + "learning_rate": 1.171769922404237e-06, + "loss": 0.1607, + "step": 49528 + }, + { + "epoch": 18.29, + "learning_rate": 1.171732972040892e-06, + "loss": 0.2231, + "step": 49529 + }, + { + "epoch": 18.29, + "learning_rate": 1.1716960216775465e-06, + "loss": 0.2664, + "step": 49530 + }, + { + "epoch": 18.29, + "learning_rate": 1.1716590713142014e-06, + "loss": 0.1547, + "step": 49531 + }, + { + "epoch": 18.29, + "learning_rate": 1.171622120950856e-06, + "loss": 0.1453, + "step": 49532 + }, + { + "epoch": 18.29, + "learning_rate": 1.1715851705875107e-06, + "loss": 0.225, + "step": 49533 + }, + { + "epoch": 18.29, + "learning_rate": 1.1715482202241655e-06, + "loss": 0.2849, + "step": 49534 + }, + { + "epoch": 18.29, + "learning_rate": 1.1715112698608204e-06, + "loss": 0.1444, + "step": 49535 + }, + { + "epoch": 18.29, + "learning_rate": 1.171474319497475e-06, + "loss": 0.2163, + "step": 49536 + }, + { + "epoch": 18.29, + "learning_rate": 1.1714373691341298e-06, + "loss": 0.2049, + "step": 49537 + }, + { + "epoch": 18.29, + "learning_rate": 1.1714004187707847e-06, + "loss": 0.2431, + "step": 49538 + }, + { + "epoch": 18.29, + "learning_rate": 1.1713634684074393e-06, + "loss": 0.2797, + "step": 49539 + }, + { + "epoch": 18.29, + "learning_rate": 1.1713265180440942e-06, + "loss": 0.1217, + "step": 49540 + }, + { + "epoch": 18.29, + "learning_rate": 1.171289567680749e-06, + "loss": 0.1813, + "step": 49541 + }, + { + "epoch": 18.29, + "learning_rate": 1.1712526173174037e-06, + "loss": 0.216, + "step": 49542 + }, + { + "epoch": 18.3, + "learning_rate": 1.1712156669540583e-06, + "loss": 0.153, + "step": 49543 + }, + { + "epoch": 18.3, + "learning_rate": 1.1711787165907131e-06, + "loss": 0.2959, + "step": 49544 + }, + { + "epoch": 18.3, + "learning_rate": 1.171141766227368e-06, + "loss": 0.2224, + "step": 49545 + }, + { + "epoch": 18.3, + "learning_rate": 1.1711048158640226e-06, + "loss": 0.2322, + "step": 49546 + }, + { + "epoch": 18.3, + "learning_rate": 1.1710678655006775e-06, + "loss": 0.1973, + "step": 49547 + }, + { + "epoch": 18.3, + "learning_rate": 1.1710309151373323e-06, + "loss": 0.1634, + "step": 49548 + }, + { + "epoch": 18.3, + "learning_rate": 1.170993964773987e-06, + "loss": 0.1765, + "step": 49549 + }, + { + "epoch": 18.3, + "learning_rate": 1.1709570144106418e-06, + "loss": 0.1088, + "step": 49550 + }, + { + "epoch": 18.3, + "learning_rate": 1.1709200640472967e-06, + "loss": 0.179, + "step": 49551 + }, + { + "epoch": 18.3, + "learning_rate": 1.170883113683951e-06, + "loss": 0.2171, + "step": 49552 + }, + { + "epoch": 18.3, + "learning_rate": 1.170846163320606e-06, + "loss": 0.2391, + "step": 49553 + }, + { + "epoch": 18.3, + "learning_rate": 1.1708092129572608e-06, + "loss": 0.2469, + "step": 49554 + }, + { + "epoch": 18.3, + "learning_rate": 1.1707722625939154e-06, + "loss": 0.1942, + "step": 49555 + }, + { + "epoch": 18.3, + "learning_rate": 1.1707353122305703e-06, + "loss": 0.1538, + "step": 49556 + }, + { + "epoch": 18.3, + "learning_rate": 1.170698361867225e-06, + "loss": 0.1436, + "step": 49557 + }, + { + "epoch": 18.3, + "learning_rate": 1.1706614115038797e-06, + "loss": 0.1692, + "step": 49558 + }, + { + "epoch": 18.3, + "learning_rate": 1.1706244611405346e-06, + "loss": 0.382, + "step": 49559 + }, + { + "epoch": 18.3, + "learning_rate": 1.1705875107771894e-06, + "loss": 0.192, + "step": 49560 + }, + { + "epoch": 18.3, + "learning_rate": 1.170550560413844e-06, + "loss": 0.1713, + "step": 49561 + }, + { + "epoch": 18.3, + "learning_rate": 1.1705136100504987e-06, + "loss": 0.2238, + "step": 49562 + }, + { + "epoch": 18.3, + "learning_rate": 1.1704766596871536e-06, + "loss": 0.3404, + "step": 49563 + }, + { + "epoch": 18.3, + "learning_rate": 1.1704397093238084e-06, + "loss": 0.1858, + "step": 49564 + }, + { + "epoch": 18.3, + "learning_rate": 1.170402758960463e-06, + "loss": 0.2655, + "step": 49565 + }, + { + "epoch": 18.3, + "learning_rate": 1.1703658085971179e-06, + "loss": 0.1707, + "step": 49566 + }, + { + "epoch": 18.3, + "learning_rate": 1.1703288582337727e-06, + "loss": 0.2287, + "step": 49567 + }, + { + "epoch": 18.3, + "learning_rate": 1.1702919078704274e-06, + "loss": 0.2649, + "step": 49568 + }, + { + "epoch": 18.3, + "learning_rate": 1.1702549575070822e-06, + "loss": 0.2912, + "step": 49569 + }, + { + "epoch": 18.31, + "learning_rate": 1.170218007143737e-06, + "loss": 0.2656, + "step": 49570 + }, + { + "epoch": 18.31, + "learning_rate": 1.1701810567803917e-06, + "loss": 0.3095, + "step": 49571 + }, + { + "epoch": 18.31, + "learning_rate": 1.1701441064170463e-06, + "loss": 0.2787, + "step": 49572 + }, + { + "epoch": 18.31, + "learning_rate": 1.1701071560537012e-06, + "loss": 0.2323, + "step": 49573 + }, + { + "epoch": 18.31, + "learning_rate": 1.170070205690356e-06, + "loss": 0.1766, + "step": 49574 + }, + { + "epoch": 18.31, + "learning_rate": 1.1700332553270107e-06, + "loss": 0.1943, + "step": 49575 + }, + { + "epoch": 18.31, + "learning_rate": 1.1699963049636655e-06, + "loss": 0.1748, + "step": 49576 + }, + { + "epoch": 18.31, + "learning_rate": 1.1699593546003204e-06, + "loss": 0.15, + "step": 49577 + }, + { + "epoch": 18.31, + "learning_rate": 1.169922404236975e-06, + "loss": 0.1651, + "step": 49578 + }, + { + "epoch": 18.31, + "learning_rate": 1.1698854538736299e-06, + "loss": 0.1905, + "step": 49579 + }, + { + "epoch": 18.31, + "learning_rate": 1.1698485035102847e-06, + "loss": 0.1961, + "step": 49580 + }, + { + "epoch": 18.31, + "learning_rate": 1.1698115531469393e-06, + "loss": 0.2321, + "step": 49581 + }, + { + "epoch": 18.31, + "learning_rate": 1.169774602783594e-06, + "loss": 0.2138, + "step": 49582 + }, + { + "epoch": 18.31, + "learning_rate": 1.1697376524202488e-06, + "loss": 0.261, + "step": 49583 + }, + { + "epoch": 18.31, + "learning_rate": 1.1697007020569035e-06, + "loss": 0.1953, + "step": 49584 + }, + { + "epoch": 18.31, + "learning_rate": 1.1696637516935583e-06, + "loss": 0.2445, + "step": 49585 + }, + { + "epoch": 18.31, + "learning_rate": 1.1696268013302132e-06, + "loss": 0.0929, + "step": 49586 + }, + { + "epoch": 18.31, + "learning_rate": 1.1695898509668678e-06, + "loss": 0.1976, + "step": 49587 + }, + { + "epoch": 18.31, + "learning_rate": 1.1695529006035226e-06, + "loss": 0.2079, + "step": 49588 + }, + { + "epoch": 18.31, + "learning_rate": 1.1695159502401775e-06, + "loss": 0.3083, + "step": 49589 + }, + { + "epoch": 18.31, + "learning_rate": 1.1694789998768321e-06, + "loss": 0.2463, + "step": 49590 + }, + { + "epoch": 18.31, + "learning_rate": 1.169442049513487e-06, + "loss": 0.2546, + "step": 49591 + }, + { + "epoch": 18.31, + "learning_rate": 1.1694050991501416e-06, + "loss": 0.2356, + "step": 49592 + }, + { + "epoch": 18.31, + "learning_rate": 1.1693681487867965e-06, + "loss": 0.3938, + "step": 49593 + }, + { + "epoch": 18.31, + "learning_rate": 1.169331198423451e-06, + "loss": 0.1702, + "step": 49594 + }, + { + "epoch": 18.31, + "learning_rate": 1.169294248060106e-06, + "loss": 0.111, + "step": 49595 + }, + { + "epoch": 18.31, + "learning_rate": 1.1692572976967608e-06, + "loss": 0.2758, + "step": 49596 + }, + { + "epoch": 18.31, + "learning_rate": 1.1692203473334154e-06, + "loss": 0.1326, + "step": 49597 + }, + { + "epoch": 18.32, + "learning_rate": 1.1691833969700703e-06, + "loss": 0.3139, + "step": 49598 + }, + { + "epoch": 18.32, + "learning_rate": 1.1691464466067251e-06, + "loss": 0.2083, + "step": 49599 + }, + { + "epoch": 18.32, + "learning_rate": 1.1691094962433798e-06, + "loss": 0.3718, + "step": 49600 + }, + { + "epoch": 18.32, + "learning_rate": 1.1690725458800346e-06, + "loss": 0.2029, + "step": 49601 + }, + { + "epoch": 18.32, + "learning_rate": 1.1690355955166893e-06, + "loss": 0.2448, + "step": 49602 + }, + { + "epoch": 18.32, + "learning_rate": 1.1689986451533439e-06, + "loss": 0.2704, + "step": 49603 + }, + { + "epoch": 18.32, + "learning_rate": 1.1689616947899987e-06, + "loss": 0.2489, + "step": 49604 + }, + { + "epoch": 18.32, + "learning_rate": 1.1689247444266536e-06, + "loss": 0.2701, + "step": 49605 + }, + { + "epoch": 18.32, + "learning_rate": 1.1688877940633082e-06, + "loss": 0.1501, + "step": 49606 + }, + { + "epoch": 18.32, + "learning_rate": 1.168850843699963e-06, + "loss": 0.1904, + "step": 49607 + }, + { + "epoch": 18.32, + "learning_rate": 1.168813893336618e-06, + "loss": 0.1807, + "step": 49608 + }, + { + "epoch": 18.32, + "learning_rate": 1.1687769429732726e-06, + "loss": 0.2818, + "step": 49609 + }, + { + "epoch": 18.32, + "learning_rate": 1.1687399926099274e-06, + "loss": 0.2499, + "step": 49610 + }, + { + "epoch": 18.32, + "learning_rate": 1.1687030422465822e-06, + "loss": 0.3374, + "step": 49611 + }, + { + "epoch": 18.32, + "learning_rate": 1.1686660918832369e-06, + "loss": 0.2128, + "step": 49612 + }, + { + "epoch": 18.32, + "learning_rate": 1.1686291415198915e-06, + "loss": 0.139, + "step": 49613 + }, + { + "epoch": 18.32, + "learning_rate": 1.1685921911565464e-06, + "loss": 0.1503, + "step": 49614 + }, + { + "epoch": 18.32, + "learning_rate": 1.1685552407932012e-06, + "loss": 0.2885, + "step": 49615 + }, + { + "epoch": 18.32, + "learning_rate": 1.1685182904298559e-06, + "loss": 0.1307, + "step": 49616 + }, + { + "epoch": 18.32, + "learning_rate": 1.1684813400665107e-06, + "loss": 0.3206, + "step": 49617 + }, + { + "epoch": 18.32, + "learning_rate": 1.1684443897031656e-06, + "loss": 0.1682, + "step": 49618 + }, + { + "epoch": 18.32, + "learning_rate": 1.1684074393398202e-06, + "loss": 0.303, + "step": 49619 + }, + { + "epoch": 18.32, + "learning_rate": 1.168370488976475e-06, + "loss": 0.2547, + "step": 49620 + }, + { + "epoch": 18.32, + "learning_rate": 1.1683335386131299e-06, + "loss": 0.2284, + "step": 49621 + }, + { + "epoch": 18.32, + "learning_rate": 1.1682965882497843e-06, + "loss": 0.1306, + "step": 49622 + }, + { + "epoch": 18.32, + "learning_rate": 1.1682596378864392e-06, + "loss": 0.3234, + "step": 49623 + }, + { + "epoch": 18.32, + "learning_rate": 1.168222687523094e-06, + "loss": 0.2362, + "step": 49624 + }, + { + "epoch": 18.33, + "learning_rate": 1.1681857371597486e-06, + "loss": 0.1522, + "step": 49625 + }, + { + "epoch": 18.33, + "learning_rate": 1.1681487867964035e-06, + "loss": 0.2874, + "step": 49626 + }, + { + "epoch": 18.33, + "learning_rate": 1.1681118364330583e-06, + "loss": 0.2579, + "step": 49627 + }, + { + "epoch": 18.33, + "learning_rate": 1.168074886069713e-06, + "loss": 0.1294, + "step": 49628 + }, + { + "epoch": 18.33, + "learning_rate": 1.1680379357063678e-06, + "loss": 0.2593, + "step": 49629 + }, + { + "epoch": 18.33, + "learning_rate": 1.1680009853430227e-06, + "loss": 0.1638, + "step": 49630 + }, + { + "epoch": 18.33, + "learning_rate": 1.1679640349796773e-06, + "loss": 0.2962, + "step": 49631 + }, + { + "epoch": 18.33, + "learning_rate": 1.167927084616332e-06, + "loss": 0.246, + "step": 49632 + }, + { + "epoch": 18.33, + "learning_rate": 1.1678901342529868e-06, + "loss": 0.2073, + "step": 49633 + }, + { + "epoch": 18.33, + "learning_rate": 1.1678531838896416e-06, + "loss": 0.213, + "step": 49634 + }, + { + "epoch": 18.33, + "learning_rate": 1.1678162335262963e-06, + "loss": 0.3564, + "step": 49635 + }, + { + "epoch": 18.33, + "learning_rate": 1.1677792831629511e-06, + "loss": 0.3304, + "step": 49636 + }, + { + "epoch": 18.33, + "learning_rate": 1.167742332799606e-06, + "loss": 0.3047, + "step": 49637 + }, + { + "epoch": 18.33, + "learning_rate": 1.1677053824362606e-06, + "loss": 0.1455, + "step": 49638 + }, + { + "epoch": 18.33, + "learning_rate": 1.1676684320729155e-06, + "loss": 0.4695, + "step": 49639 + }, + { + "epoch": 18.33, + "learning_rate": 1.1676314817095703e-06, + "loss": 0.2168, + "step": 49640 + }, + { + "epoch": 18.33, + "learning_rate": 1.167594531346225e-06, + "loss": 0.1307, + "step": 49641 + }, + { + "epoch": 18.33, + "learning_rate": 1.1675575809828796e-06, + "loss": 0.2512, + "step": 49642 + }, + { + "epoch": 18.33, + "learning_rate": 1.1675206306195344e-06, + "loss": 0.2993, + "step": 49643 + }, + { + "epoch": 18.33, + "learning_rate": 1.1674836802561893e-06, + "loss": 0.1485, + "step": 49644 + }, + { + "epoch": 18.33, + "learning_rate": 1.167446729892844e-06, + "loss": 0.219, + "step": 49645 + }, + { + "epoch": 18.33, + "learning_rate": 1.1674097795294988e-06, + "loss": 0.1072, + "step": 49646 + }, + { + "epoch": 18.33, + "learning_rate": 1.1673728291661536e-06, + "loss": 0.2376, + "step": 49647 + }, + { + "epoch": 18.33, + "learning_rate": 1.1673358788028082e-06, + "loss": 0.1622, + "step": 49648 + }, + { + "epoch": 18.33, + "learning_rate": 1.167298928439463e-06, + "loss": 0.3036, + "step": 49649 + }, + { + "epoch": 18.33, + "learning_rate": 1.167261978076118e-06, + "loss": 0.2173, + "step": 49650 + }, + { + "epoch": 18.33, + "learning_rate": 1.1672250277127726e-06, + "loss": 0.2871, + "step": 49651 + }, + { + "epoch": 18.34, + "learning_rate": 1.1671880773494272e-06, + "loss": 0.2623, + "step": 49652 + }, + { + "epoch": 18.34, + "learning_rate": 1.167151126986082e-06, + "loss": 0.2782, + "step": 49653 + }, + { + "epoch": 18.34, + "learning_rate": 1.1671141766227367e-06, + "loss": 0.1467, + "step": 49654 + }, + { + "epoch": 18.34, + "learning_rate": 1.1670772262593915e-06, + "loss": 0.3047, + "step": 49655 + }, + { + "epoch": 18.34, + "learning_rate": 1.1670402758960464e-06, + "loss": 0.1876, + "step": 49656 + }, + { + "epoch": 18.34, + "learning_rate": 1.167003325532701e-06, + "loss": 0.214, + "step": 49657 + }, + { + "epoch": 18.34, + "learning_rate": 1.1669663751693559e-06, + "loss": 0.3192, + "step": 49658 + }, + { + "epoch": 18.34, + "learning_rate": 1.1669294248060107e-06, + "loss": 0.169, + "step": 49659 + }, + { + "epoch": 18.34, + "learning_rate": 1.1668924744426654e-06, + "loss": 0.3213, + "step": 49660 + }, + { + "epoch": 18.34, + "learning_rate": 1.1668555240793202e-06, + "loss": 0.1621, + "step": 49661 + }, + { + "epoch": 18.34, + "learning_rate": 1.1668185737159748e-06, + "loss": 0.2076, + "step": 49662 + }, + { + "epoch": 18.34, + "learning_rate": 1.1667816233526297e-06, + "loss": 0.2638, + "step": 49663 + }, + { + "epoch": 18.34, + "learning_rate": 1.1667446729892843e-06, + "loss": 0.2356, + "step": 49664 + }, + { + "epoch": 18.34, + "learning_rate": 1.1667077226259392e-06, + "loss": 0.3227, + "step": 49665 + }, + { + "epoch": 18.34, + "learning_rate": 1.166670772262594e-06, + "loss": 0.2605, + "step": 49666 + }, + { + "epoch": 18.34, + "learning_rate": 1.1666338218992487e-06, + "loss": 0.1758, + "step": 49667 + }, + { + "epoch": 18.34, + "learning_rate": 1.1665968715359035e-06, + "loss": 0.227, + "step": 49668 + }, + { + "epoch": 18.34, + "learning_rate": 1.1665599211725584e-06, + "loss": 0.2567, + "step": 49669 + }, + { + "epoch": 18.34, + "learning_rate": 1.166522970809213e-06, + "loss": 0.3212, + "step": 49670 + }, + { + "epoch": 18.34, + "learning_rate": 1.1664860204458678e-06, + "loss": 0.293, + "step": 49671 + }, + { + "epoch": 18.34, + "learning_rate": 1.1664490700825225e-06, + "loss": 0.2314, + "step": 49672 + }, + { + "epoch": 18.34, + "learning_rate": 1.1664121197191771e-06, + "loss": 0.2612, + "step": 49673 + }, + { + "epoch": 18.34, + "learning_rate": 1.166375169355832e-06, + "loss": 0.232, + "step": 49674 + }, + { + "epoch": 18.34, + "learning_rate": 1.1663382189924868e-06, + "loss": 0.2662, + "step": 49675 + }, + { + "epoch": 18.34, + "learning_rate": 1.1663012686291415e-06, + "loss": 0.2347, + "step": 49676 + }, + { + "epoch": 18.34, + "learning_rate": 1.1662643182657963e-06, + "loss": 0.2621, + "step": 49677 + }, + { + "epoch": 18.34, + "learning_rate": 1.1662273679024511e-06, + "loss": 0.1693, + "step": 49678 + }, + { + "epoch": 18.35, + "learning_rate": 1.1661904175391058e-06, + "loss": 0.2109, + "step": 49679 + }, + { + "epoch": 18.35, + "learning_rate": 1.1661534671757606e-06, + "loss": 0.336, + "step": 49680 + }, + { + "epoch": 18.35, + "learning_rate": 1.1661165168124155e-06, + "loss": 0.1839, + "step": 49681 + }, + { + "epoch": 18.35, + "learning_rate": 1.1660795664490701e-06, + "loss": 0.1696, + "step": 49682 + }, + { + "epoch": 18.35, + "learning_rate": 1.1660426160857248e-06, + "loss": 0.2148, + "step": 49683 + }, + { + "epoch": 18.35, + "learning_rate": 1.1660056657223796e-06, + "loss": 0.2895, + "step": 49684 + }, + { + "epoch": 18.35, + "learning_rate": 1.1659687153590345e-06, + "loss": 0.1941, + "step": 49685 + }, + { + "epoch": 18.35, + "learning_rate": 1.165931764995689e-06, + "loss": 0.1811, + "step": 49686 + }, + { + "epoch": 18.35, + "learning_rate": 1.165894814632344e-06, + "loss": 0.2144, + "step": 49687 + }, + { + "epoch": 18.35, + "learning_rate": 1.1658578642689988e-06, + "loss": 0.2782, + "step": 49688 + }, + { + "epoch": 18.35, + "learning_rate": 1.1658209139056534e-06, + "loss": 0.1335, + "step": 49689 + }, + { + "epoch": 18.35, + "learning_rate": 1.1657839635423083e-06, + "loss": 0.1654, + "step": 49690 + }, + { + "epoch": 18.35, + "learning_rate": 1.1657470131789631e-06, + "loss": 0.1871, + "step": 49691 + }, + { + "epoch": 18.35, + "learning_rate": 1.1657100628156178e-06, + "loss": 0.1442, + "step": 49692 + }, + { + "epoch": 18.35, + "learning_rate": 1.1656731124522724e-06, + "loss": 0.21, + "step": 49693 + }, + { + "epoch": 18.35, + "learning_rate": 1.1656361620889272e-06, + "loss": 0.221, + "step": 49694 + }, + { + "epoch": 18.35, + "learning_rate": 1.165599211725582e-06, + "loss": 0.2374, + "step": 49695 + }, + { + "epoch": 18.35, + "learning_rate": 1.1655622613622367e-06, + "loss": 0.2294, + "step": 49696 + }, + { + "epoch": 18.35, + "learning_rate": 1.1655253109988916e-06, + "loss": 0.2566, + "step": 49697 + }, + { + "epoch": 18.35, + "learning_rate": 1.1654883606355462e-06, + "loss": 0.1832, + "step": 49698 + }, + { + "epoch": 18.35, + "learning_rate": 1.165451410272201e-06, + "loss": 0.3298, + "step": 49699 + }, + { + "epoch": 18.35, + "learning_rate": 1.165414459908856e-06, + "loss": 0.1243, + "step": 49700 + }, + { + "epoch": 18.35, + "learning_rate": 1.1653775095455105e-06, + "loss": 0.0879, + "step": 49701 + }, + { + "epoch": 18.35, + "learning_rate": 1.1653405591821652e-06, + "loss": 0.2106, + "step": 49702 + }, + { + "epoch": 18.35, + "learning_rate": 1.16530360881882e-06, + "loss": 0.2654, + "step": 49703 + }, + { + "epoch": 18.35, + "learning_rate": 1.1652666584554749e-06, + "loss": 0.2274, + "step": 49704 + }, + { + "epoch": 18.35, + "learning_rate": 1.1652297080921295e-06, + "loss": 0.138, + "step": 49705 + }, + { + "epoch": 18.36, + "learning_rate": 1.1651927577287844e-06, + "loss": 0.1871, + "step": 49706 + }, + { + "epoch": 18.36, + "learning_rate": 1.1651558073654392e-06, + "loss": 0.1357, + "step": 49707 + }, + { + "epoch": 18.36, + "learning_rate": 1.1651188570020938e-06, + "loss": 0.2001, + "step": 49708 + }, + { + "epoch": 18.36, + "learning_rate": 1.1650819066387487e-06, + "loss": 0.1087, + "step": 49709 + }, + { + "epoch": 18.36, + "learning_rate": 1.1650449562754035e-06, + "loss": 0.2229, + "step": 49710 + }, + { + "epoch": 18.36, + "learning_rate": 1.1650080059120582e-06, + "loss": 0.0841, + "step": 49711 + }, + { + "epoch": 18.36, + "learning_rate": 1.1649710555487128e-06, + "loss": 0.1834, + "step": 49712 + }, + { + "epoch": 18.36, + "learning_rate": 1.1649341051853677e-06, + "loss": 0.2999, + "step": 49713 + }, + { + "epoch": 18.36, + "learning_rate": 1.1648971548220225e-06, + "loss": 0.2111, + "step": 49714 + }, + { + "epoch": 18.36, + "learning_rate": 1.1648602044586771e-06, + "loss": 0.1815, + "step": 49715 + }, + { + "epoch": 18.36, + "learning_rate": 1.164823254095332e-06, + "loss": 0.3689, + "step": 49716 + }, + { + "epoch": 18.36, + "learning_rate": 1.1647863037319868e-06, + "loss": 0.2335, + "step": 49717 + }, + { + "epoch": 18.36, + "learning_rate": 1.1647493533686415e-06, + "loss": 0.1563, + "step": 49718 + }, + { + "epoch": 18.36, + "learning_rate": 1.1647124030052963e-06, + "loss": 0.1133, + "step": 49719 + }, + { + "epoch": 18.36, + "learning_rate": 1.1646754526419512e-06, + "loss": 0.1669, + "step": 49720 + }, + { + "epoch": 18.36, + "learning_rate": 1.1646385022786056e-06, + "loss": 0.1512, + "step": 49721 + }, + { + "epoch": 18.36, + "learning_rate": 1.1646015519152604e-06, + "loss": 0.2634, + "step": 49722 + }, + { + "epoch": 18.36, + "learning_rate": 1.1645646015519153e-06, + "loss": 0.2567, + "step": 49723 + }, + { + "epoch": 18.36, + "learning_rate": 1.16452765118857e-06, + "loss": 0.3005, + "step": 49724 + }, + { + "epoch": 18.36, + "learning_rate": 1.1644907008252248e-06, + "loss": 0.2655, + "step": 49725 + }, + { + "epoch": 18.36, + "learning_rate": 1.1644537504618796e-06, + "loss": 0.166, + "step": 49726 + }, + { + "epoch": 18.36, + "learning_rate": 1.1644168000985343e-06, + "loss": 0.269, + "step": 49727 + }, + { + "epoch": 18.36, + "learning_rate": 1.1643798497351891e-06, + "loss": 0.1591, + "step": 49728 + }, + { + "epoch": 18.36, + "learning_rate": 1.164342899371844e-06, + "loss": 0.1876, + "step": 49729 + }, + { + "epoch": 18.36, + "learning_rate": 1.1643059490084986e-06, + "loss": 0.2352, + "step": 49730 + }, + { + "epoch": 18.36, + "learning_rate": 1.1642689986451532e-06, + "loss": 0.1893, + "step": 49731 + }, + { + "epoch": 18.36, + "learning_rate": 1.164232048281808e-06, + "loss": 0.2204, + "step": 49732 + }, + { + "epoch": 18.37, + "learning_rate": 1.164195097918463e-06, + "loss": 0.1355, + "step": 49733 + }, + { + "epoch": 18.37, + "learning_rate": 1.1641581475551176e-06, + "loss": 0.1704, + "step": 49734 + }, + { + "epoch": 18.37, + "learning_rate": 1.1641211971917724e-06, + "loss": 0.2191, + "step": 49735 + }, + { + "epoch": 18.37, + "learning_rate": 1.1640842468284273e-06, + "loss": 0.2258, + "step": 49736 + }, + { + "epoch": 18.37, + "learning_rate": 1.164047296465082e-06, + "loss": 0.3376, + "step": 49737 + }, + { + "epoch": 18.37, + "learning_rate": 1.1640103461017367e-06, + "loss": 0.1976, + "step": 49738 + }, + { + "epoch": 18.37, + "learning_rate": 1.1639733957383916e-06, + "loss": 0.2645, + "step": 49739 + }, + { + "epoch": 18.37, + "learning_rate": 1.1639364453750462e-06, + "loss": 0.1625, + "step": 49740 + }, + { + "epoch": 18.37, + "learning_rate": 1.1638994950117009e-06, + "loss": 0.2875, + "step": 49741 + }, + { + "epoch": 18.37, + "learning_rate": 1.1638625446483557e-06, + "loss": 0.2635, + "step": 49742 + }, + { + "epoch": 18.37, + "learning_rate": 1.1638255942850104e-06, + "loss": 0.1897, + "step": 49743 + }, + { + "epoch": 18.37, + "learning_rate": 1.1637886439216652e-06, + "loss": 0.2152, + "step": 49744 + }, + { + "epoch": 18.37, + "learning_rate": 1.16375169355832e-06, + "loss": 0.1791, + "step": 49745 + }, + { + "epoch": 18.37, + "learning_rate": 1.1637147431949747e-06, + "loss": 0.1366, + "step": 49746 + }, + { + "epoch": 18.37, + "learning_rate": 1.1636777928316295e-06, + "loss": 0.1289, + "step": 49747 + }, + { + "epoch": 18.37, + "learning_rate": 1.1636408424682844e-06, + "loss": 0.1942, + "step": 49748 + }, + { + "epoch": 18.37, + "learning_rate": 1.163603892104939e-06, + "loss": 0.2969, + "step": 49749 + }, + { + "epoch": 18.37, + "learning_rate": 1.1635669417415939e-06, + "loss": 0.1882, + "step": 49750 + }, + { + "epoch": 18.37, + "learning_rate": 1.1635299913782485e-06, + "loss": 0.2086, + "step": 49751 + }, + { + "epoch": 18.37, + "learning_rate": 1.1634930410149034e-06, + "loss": 0.1778, + "step": 49752 + }, + { + "epoch": 18.37, + "learning_rate": 1.163456090651558e-06, + "loss": 0.2825, + "step": 49753 + }, + { + "epoch": 18.37, + "learning_rate": 1.1634191402882128e-06, + "loss": 0.2116, + "step": 49754 + }, + { + "epoch": 18.37, + "learning_rate": 1.1633821899248677e-06, + "loss": 0.3165, + "step": 49755 + }, + { + "epoch": 18.37, + "learning_rate": 1.1633452395615223e-06, + "loss": 0.2253, + "step": 49756 + }, + { + "epoch": 18.37, + "learning_rate": 1.1633082891981772e-06, + "loss": 0.3299, + "step": 49757 + }, + { + "epoch": 18.37, + "learning_rate": 1.163271338834832e-06, + "loss": 0.2418, + "step": 49758 + }, + { + "epoch": 18.37, + "learning_rate": 1.1632343884714867e-06, + "loss": 0.2042, + "step": 49759 + }, + { + "epoch": 18.38, + "learning_rate": 1.1631974381081415e-06, + "loss": 0.166, + "step": 49760 + }, + { + "epoch": 18.38, + "learning_rate": 1.1631604877447961e-06, + "loss": 0.1988, + "step": 49761 + }, + { + "epoch": 18.38, + "learning_rate": 1.163123537381451e-06, + "loss": 0.2375, + "step": 49762 + }, + { + "epoch": 18.38, + "learning_rate": 1.1630865870181056e-06, + "loss": 0.1736, + "step": 49763 + }, + { + "epoch": 18.38, + "learning_rate": 1.1630496366547605e-06, + "loss": 0.179, + "step": 49764 + }, + { + "epoch": 18.38, + "learning_rate": 1.1630126862914153e-06, + "loss": 0.2425, + "step": 49765 + }, + { + "epoch": 18.38, + "learning_rate": 1.16297573592807e-06, + "loss": 0.212, + "step": 49766 + }, + { + "epoch": 18.38, + "learning_rate": 1.1629387855647248e-06, + "loss": 0.2334, + "step": 49767 + }, + { + "epoch": 18.38, + "learning_rate": 1.1629018352013797e-06, + "loss": 0.0939, + "step": 49768 + }, + { + "epoch": 18.38, + "learning_rate": 1.1628648848380343e-06, + "loss": 0.1223, + "step": 49769 + }, + { + "epoch": 18.38, + "learning_rate": 1.1628279344746891e-06, + "loss": 0.2584, + "step": 49770 + }, + { + "epoch": 18.38, + "learning_rate": 1.1627909841113438e-06, + "loss": 0.149, + "step": 49771 + }, + { + "epoch": 18.38, + "learning_rate": 1.1627540337479984e-06, + "loss": 0.2621, + "step": 49772 + }, + { + "epoch": 18.38, + "learning_rate": 1.1627170833846533e-06, + "loss": 0.2174, + "step": 49773 + }, + { + "epoch": 18.38, + "learning_rate": 1.162680133021308e-06, + "loss": 0.1487, + "step": 49774 + }, + { + "epoch": 18.38, + "learning_rate": 1.1626431826579627e-06, + "loss": 0.1723, + "step": 49775 + }, + { + "epoch": 18.38, + "learning_rate": 1.1626062322946176e-06, + "loss": 0.1378, + "step": 49776 + }, + { + "epoch": 18.38, + "learning_rate": 1.1625692819312724e-06, + "loss": 0.2203, + "step": 49777 + }, + { + "epoch": 18.38, + "learning_rate": 1.162532331567927e-06, + "loss": 0.1825, + "step": 49778 + }, + { + "epoch": 18.38, + "learning_rate": 1.162495381204582e-06, + "loss": 0.2315, + "step": 49779 + }, + { + "epoch": 18.38, + "learning_rate": 1.1624584308412368e-06, + "loss": 0.1992, + "step": 49780 + }, + { + "epoch": 18.38, + "learning_rate": 1.1624214804778914e-06, + "loss": 0.1615, + "step": 49781 + }, + { + "epoch": 18.38, + "learning_rate": 1.162384530114546e-06, + "loss": 0.2292, + "step": 49782 + }, + { + "epoch": 18.38, + "learning_rate": 1.1623475797512009e-06, + "loss": 0.2856, + "step": 49783 + }, + { + "epoch": 18.38, + "learning_rate": 1.1623106293878557e-06, + "loss": 0.1864, + "step": 49784 + }, + { + "epoch": 18.38, + "learning_rate": 1.1622736790245104e-06, + "loss": 0.1847, + "step": 49785 + }, + { + "epoch": 18.38, + "learning_rate": 1.1622367286611652e-06, + "loss": 0.2303, + "step": 49786 + }, + { + "epoch": 18.39, + "learning_rate": 1.16219977829782e-06, + "loss": 0.1448, + "step": 49787 + }, + { + "epoch": 18.39, + "learning_rate": 1.1621628279344747e-06, + "loss": 0.1486, + "step": 49788 + }, + { + "epoch": 18.39, + "learning_rate": 1.1621258775711296e-06, + "loss": 0.2194, + "step": 49789 + }, + { + "epoch": 18.39, + "learning_rate": 1.1620889272077844e-06, + "loss": 0.3193, + "step": 49790 + }, + { + "epoch": 18.39, + "learning_rate": 1.1620519768444388e-06, + "loss": 0.2296, + "step": 49791 + }, + { + "epoch": 18.39, + "learning_rate": 1.1620150264810937e-06, + "loss": 0.3364, + "step": 49792 + }, + { + "epoch": 18.39, + "learning_rate": 1.1619780761177485e-06, + "loss": 0.2022, + "step": 49793 + }, + { + "epoch": 18.39, + "learning_rate": 1.1619411257544032e-06, + "loss": 0.2446, + "step": 49794 + }, + { + "epoch": 18.39, + "learning_rate": 1.161904175391058e-06, + "loss": 0.2082, + "step": 49795 + }, + { + "epoch": 18.39, + "learning_rate": 1.1618672250277129e-06, + "loss": 0.1155, + "step": 49796 + }, + { + "epoch": 18.39, + "learning_rate": 1.1618302746643675e-06, + "loss": 0.2754, + "step": 49797 + }, + { + "epoch": 18.39, + "learning_rate": 1.1617933243010223e-06, + "loss": 0.2606, + "step": 49798 + }, + { + "epoch": 18.39, + "learning_rate": 1.1617563739376772e-06, + "loss": 0.3605, + "step": 49799 + }, + { + "epoch": 18.39, + "learning_rate": 1.1617194235743318e-06, + "loss": 0.2614, + "step": 49800 + }, + { + "epoch": 18.39, + "learning_rate": 1.1616824732109865e-06, + "loss": 0.239, + "step": 49801 + }, + { + "epoch": 18.39, + "learning_rate": 1.1616455228476413e-06, + "loss": 0.2538, + "step": 49802 + }, + { + "epoch": 18.39, + "learning_rate": 1.1616085724842962e-06, + "loss": 0.1982, + "step": 49803 + }, + { + "epoch": 18.39, + "learning_rate": 1.1615716221209508e-06, + "loss": 0.1717, + "step": 49804 + }, + { + "epoch": 18.39, + "learning_rate": 1.1615346717576056e-06, + "loss": 0.2239, + "step": 49805 + }, + { + "epoch": 18.39, + "learning_rate": 1.1614977213942605e-06, + "loss": 0.2426, + "step": 49806 + }, + { + "epoch": 18.39, + "learning_rate": 1.1614607710309151e-06, + "loss": 0.2233, + "step": 49807 + }, + { + "epoch": 18.39, + "learning_rate": 1.16142382066757e-06, + "loss": 0.2739, + "step": 49808 + }, + { + "epoch": 18.39, + "learning_rate": 1.1613868703042248e-06, + "loss": 0.2168, + "step": 49809 + }, + { + "epoch": 18.39, + "learning_rate": 1.1613499199408795e-06, + "loss": 0.2572, + "step": 49810 + }, + { + "epoch": 18.39, + "learning_rate": 1.161312969577534e-06, + "loss": 0.2258, + "step": 49811 + }, + { + "epoch": 18.39, + "learning_rate": 1.161276019214189e-06, + "loss": 0.3567, + "step": 49812 + }, + { + "epoch": 18.39, + "learning_rate": 1.1612390688508436e-06, + "loss": 0.1441, + "step": 49813 + }, + { + "epoch": 18.4, + "learning_rate": 1.1612021184874984e-06, + "loss": 0.203, + "step": 49814 + }, + { + "epoch": 18.4, + "learning_rate": 1.1611651681241533e-06, + "loss": 0.237, + "step": 49815 + }, + { + "epoch": 18.4, + "learning_rate": 1.161128217760808e-06, + "loss": 0.2343, + "step": 49816 + }, + { + "epoch": 18.4, + "learning_rate": 1.1610912673974628e-06, + "loss": 0.2986, + "step": 49817 + }, + { + "epoch": 18.4, + "learning_rate": 1.1610543170341176e-06, + "loss": 0.1799, + "step": 49818 + }, + { + "epoch": 18.4, + "learning_rate": 1.1610173666707723e-06, + "loss": 0.377, + "step": 49819 + }, + { + "epoch": 18.4, + "learning_rate": 1.160980416307427e-06, + "loss": 0.2245, + "step": 49820 + }, + { + "epoch": 18.4, + "learning_rate": 1.1609434659440817e-06, + "loss": 0.1559, + "step": 49821 + }, + { + "epoch": 18.4, + "learning_rate": 1.1609065155807366e-06, + "loss": 0.3049, + "step": 49822 + }, + { + "epoch": 18.4, + "learning_rate": 1.1608695652173912e-06, + "loss": 0.2999, + "step": 49823 + }, + { + "epoch": 18.4, + "learning_rate": 1.160832614854046e-06, + "loss": 0.3312, + "step": 49824 + }, + { + "epoch": 18.4, + "learning_rate": 1.160795664490701e-06, + "loss": 0.1792, + "step": 49825 + }, + { + "epoch": 18.4, + "learning_rate": 1.1607587141273556e-06, + "loss": 0.2399, + "step": 49826 + }, + { + "epoch": 18.4, + "learning_rate": 1.1607217637640104e-06, + "loss": 0.1742, + "step": 49827 + }, + { + "epoch": 18.4, + "learning_rate": 1.1606848134006652e-06, + "loss": 0.2814, + "step": 49828 + }, + { + "epoch": 18.4, + "learning_rate": 1.1606478630373199e-06, + "loss": 0.127, + "step": 49829 + }, + { + "epoch": 18.4, + "learning_rate": 1.1606109126739747e-06, + "loss": 0.1682, + "step": 49830 + }, + { + "epoch": 18.4, + "learning_rate": 1.1605739623106294e-06, + "loss": 0.1598, + "step": 49831 + }, + { + "epoch": 18.4, + "learning_rate": 1.1605370119472842e-06, + "loss": 0.1811, + "step": 49832 + }, + { + "epoch": 18.4, + "learning_rate": 1.1605000615839389e-06, + "loss": 0.144, + "step": 49833 + }, + { + "epoch": 18.4, + "learning_rate": 1.1604631112205937e-06, + "loss": 0.2715, + "step": 49834 + }, + { + "epoch": 18.4, + "learning_rate": 1.1604261608572486e-06, + "loss": 0.2955, + "step": 49835 + }, + { + "epoch": 18.4, + "learning_rate": 1.1603892104939032e-06, + "loss": 0.2487, + "step": 49836 + }, + { + "epoch": 18.4, + "learning_rate": 1.160352260130558e-06, + "loss": 0.1222, + "step": 49837 + }, + { + "epoch": 18.4, + "learning_rate": 1.1603153097672129e-06, + "loss": 0.1677, + "step": 49838 + }, + { + "epoch": 18.4, + "learning_rate": 1.1602783594038675e-06, + "loss": 0.2258, + "step": 49839 + }, + { + "epoch": 18.4, + "learning_rate": 1.1602414090405224e-06, + "loss": 0.2991, + "step": 49840 + }, + { + "epoch": 18.41, + "learning_rate": 1.160204458677177e-06, + "loss": 0.2046, + "step": 49841 + }, + { + "epoch": 18.41, + "learning_rate": 1.1601675083138316e-06, + "loss": 0.1148, + "step": 49842 + }, + { + "epoch": 18.41, + "learning_rate": 1.1601305579504865e-06, + "loss": 0.1401, + "step": 49843 + }, + { + "epoch": 18.41, + "learning_rate": 1.1600936075871413e-06, + "loss": 0.2999, + "step": 49844 + }, + { + "epoch": 18.41, + "learning_rate": 1.160056657223796e-06, + "loss": 0.1464, + "step": 49845 + }, + { + "epoch": 18.41, + "learning_rate": 1.1600197068604508e-06, + "loss": 0.1675, + "step": 49846 + }, + { + "epoch": 18.41, + "learning_rate": 1.1599827564971057e-06, + "loss": 0.2511, + "step": 49847 + }, + { + "epoch": 18.41, + "learning_rate": 1.1599458061337603e-06, + "loss": 0.358, + "step": 49848 + }, + { + "epoch": 18.41, + "learning_rate": 1.1599088557704152e-06, + "loss": 0.2373, + "step": 49849 + }, + { + "epoch": 18.41, + "learning_rate": 1.15987190540707e-06, + "loss": 0.2886, + "step": 49850 + }, + { + "epoch": 18.41, + "learning_rate": 1.1598349550437246e-06, + "loss": 0.1593, + "step": 49851 + }, + { + "epoch": 18.41, + "learning_rate": 1.1597980046803793e-06, + "loss": 0.2909, + "step": 49852 + }, + { + "epoch": 18.41, + "learning_rate": 1.1597610543170341e-06, + "loss": 0.4271, + "step": 49853 + }, + { + "epoch": 18.41, + "learning_rate": 1.159724103953689e-06, + "loss": 0.2683, + "step": 49854 + }, + { + "epoch": 18.41, + "learning_rate": 1.1596871535903436e-06, + "loss": 0.1272, + "step": 49855 + }, + { + "epoch": 18.41, + "learning_rate": 1.1596502032269985e-06, + "loss": 0.4289, + "step": 49856 + }, + { + "epoch": 18.41, + "learning_rate": 1.1596132528636533e-06, + "loss": 0.2689, + "step": 49857 + }, + { + "epoch": 18.41, + "learning_rate": 1.159576302500308e-06, + "loss": 0.2538, + "step": 49858 + }, + { + "epoch": 18.41, + "learning_rate": 1.1595393521369628e-06, + "loss": 0.1636, + "step": 49859 + }, + { + "epoch": 18.41, + "learning_rate": 1.1595024017736176e-06, + "loss": 0.195, + "step": 49860 + }, + { + "epoch": 18.41, + "learning_rate": 1.159465451410272e-06, + "loss": 0.249, + "step": 49861 + }, + { + "epoch": 18.41, + "learning_rate": 1.159428501046927e-06, + "loss": 0.108, + "step": 49862 + }, + { + "epoch": 18.41, + "learning_rate": 1.1593915506835818e-06, + "loss": 0.1919, + "step": 49863 + }, + { + "epoch": 18.41, + "learning_rate": 1.1593546003202364e-06, + "loss": 0.1794, + "step": 49864 + }, + { + "epoch": 18.41, + "learning_rate": 1.1593176499568912e-06, + "loss": 0.1562, + "step": 49865 + }, + { + "epoch": 18.41, + "learning_rate": 1.159280699593546e-06, + "loss": 0.2464, + "step": 49866 + }, + { + "epoch": 18.41, + "learning_rate": 1.1592437492302007e-06, + "loss": 0.2963, + "step": 49867 + }, + { + "epoch": 18.42, + "learning_rate": 1.1592067988668556e-06, + "loss": 0.1583, + "step": 49868 + }, + { + "epoch": 18.42, + "learning_rate": 1.1591698485035104e-06, + "loss": 0.2413, + "step": 49869 + }, + { + "epoch": 18.42, + "learning_rate": 1.159132898140165e-06, + "loss": 0.3492, + "step": 49870 + }, + { + "epoch": 18.42, + "learning_rate": 1.1590959477768197e-06, + "loss": 0.2092, + "step": 49871 + }, + { + "epoch": 18.42, + "learning_rate": 1.1590589974134745e-06, + "loss": 0.1866, + "step": 49872 + }, + { + "epoch": 18.42, + "learning_rate": 1.1590220470501294e-06, + "loss": 0.2329, + "step": 49873 + }, + { + "epoch": 18.42, + "learning_rate": 1.158985096686784e-06, + "loss": 0.1171, + "step": 49874 + }, + { + "epoch": 18.42, + "learning_rate": 1.1589481463234389e-06, + "loss": 0.2276, + "step": 49875 + }, + { + "epoch": 18.42, + "learning_rate": 1.1589111959600937e-06, + "loss": 0.1816, + "step": 49876 + }, + { + "epoch": 18.42, + "learning_rate": 1.1588742455967484e-06, + "loss": 0.1481, + "step": 49877 + }, + { + "epoch": 18.42, + "learning_rate": 1.1588372952334032e-06, + "loss": 0.2789, + "step": 49878 + }, + { + "epoch": 18.42, + "learning_rate": 1.158800344870058e-06, + "loss": 0.1777, + "step": 49879 + }, + { + "epoch": 18.42, + "learning_rate": 1.1587633945067127e-06, + "loss": 0.1995, + "step": 49880 + }, + { + "epoch": 18.42, + "learning_rate": 1.1587264441433673e-06, + "loss": 0.2223, + "step": 49881 + }, + { + "epoch": 18.42, + "learning_rate": 1.1586894937800222e-06, + "loss": 0.3946, + "step": 49882 + }, + { + "epoch": 18.42, + "learning_rate": 1.158652543416677e-06, + "loss": 0.1518, + "step": 49883 + }, + { + "epoch": 18.42, + "learning_rate": 1.1586155930533317e-06, + "loss": 0.2057, + "step": 49884 + }, + { + "epoch": 18.42, + "learning_rate": 1.1585786426899865e-06, + "loss": 0.2607, + "step": 49885 + }, + { + "epoch": 18.42, + "learning_rate": 1.1585416923266411e-06, + "loss": 0.1001, + "step": 49886 + }, + { + "epoch": 18.42, + "learning_rate": 1.158504741963296e-06, + "loss": 0.1659, + "step": 49887 + }, + { + "epoch": 18.42, + "learning_rate": 1.1584677915999508e-06, + "loss": 0.2172, + "step": 49888 + }, + { + "epoch": 18.42, + "learning_rate": 1.1584308412366055e-06, + "loss": 0.2131, + "step": 49889 + }, + { + "epoch": 18.42, + "learning_rate": 1.1583938908732603e-06, + "loss": 0.1491, + "step": 49890 + }, + { + "epoch": 18.42, + "learning_rate": 1.158356940509915e-06, + "loss": 0.2412, + "step": 49891 + }, + { + "epoch": 18.42, + "learning_rate": 1.1583199901465698e-06, + "loss": 0.3617, + "step": 49892 + }, + { + "epoch": 18.42, + "learning_rate": 1.1582830397832245e-06, + "loss": 0.3193, + "step": 49893 + }, + { + "epoch": 18.42, + "learning_rate": 1.1582460894198793e-06, + "loss": 0.2484, + "step": 49894 + }, + { + "epoch": 18.43, + "learning_rate": 1.1582091390565341e-06, + "loss": 0.1186, + "step": 49895 + }, + { + "epoch": 18.43, + "learning_rate": 1.1581721886931888e-06, + "loss": 0.2445, + "step": 49896 + }, + { + "epoch": 18.43, + "learning_rate": 1.1581352383298436e-06, + "loss": 0.2711, + "step": 49897 + }, + { + "epoch": 18.43, + "learning_rate": 1.1580982879664985e-06, + "loss": 0.2406, + "step": 49898 + }, + { + "epoch": 18.43, + "learning_rate": 1.1580613376031531e-06, + "loss": 0.1233, + "step": 49899 + }, + { + "epoch": 18.43, + "learning_rate": 1.158024387239808e-06, + "loss": 0.1759, + "step": 49900 + }, + { + "epoch": 18.43, + "learning_rate": 1.1579874368764626e-06, + "loss": 0.2502, + "step": 49901 + }, + { + "epoch": 18.43, + "learning_rate": 1.1579504865131175e-06, + "loss": 0.3019, + "step": 49902 + }, + { + "epoch": 18.43, + "learning_rate": 1.157913536149772e-06, + "loss": 0.1282, + "step": 49903 + }, + { + "epoch": 18.43, + "learning_rate": 1.157876585786427e-06, + "loss": 0.281, + "step": 49904 + }, + { + "epoch": 18.43, + "learning_rate": 1.1578396354230818e-06, + "loss": 0.2168, + "step": 49905 + }, + { + "epoch": 18.43, + "learning_rate": 1.1578026850597364e-06, + "loss": 0.1388, + "step": 49906 + }, + { + "epoch": 18.43, + "learning_rate": 1.1577657346963913e-06, + "loss": 0.1738, + "step": 49907 + }, + { + "epoch": 18.43, + "learning_rate": 1.1577287843330461e-06, + "loss": 0.135, + "step": 49908 + }, + { + "epoch": 18.43, + "learning_rate": 1.1576918339697008e-06, + "loss": 0.1408, + "step": 49909 + }, + { + "epoch": 18.43, + "learning_rate": 1.1576548836063556e-06, + "loss": 0.2232, + "step": 49910 + }, + { + "epoch": 18.43, + "learning_rate": 1.1576179332430102e-06, + "loss": 0.1659, + "step": 49911 + }, + { + "epoch": 18.43, + "learning_rate": 1.1575809828796649e-06, + "loss": 0.2267, + "step": 49912 + }, + { + "epoch": 18.43, + "learning_rate": 1.1575440325163197e-06, + "loss": 0.2768, + "step": 49913 + }, + { + "epoch": 18.43, + "learning_rate": 1.1575070821529746e-06, + "loss": 0.2454, + "step": 49914 + }, + { + "epoch": 18.43, + "learning_rate": 1.1574701317896292e-06, + "loss": 0.1818, + "step": 49915 + }, + { + "epoch": 18.43, + "learning_rate": 1.157433181426284e-06, + "loss": 0.1589, + "step": 49916 + }, + { + "epoch": 18.43, + "learning_rate": 1.157396231062939e-06, + "loss": 0.2207, + "step": 49917 + }, + { + "epoch": 18.43, + "learning_rate": 1.1573592806995935e-06, + "loss": 0.2512, + "step": 49918 + }, + { + "epoch": 18.43, + "learning_rate": 1.1573223303362484e-06, + "loss": 0.2698, + "step": 49919 + }, + { + "epoch": 18.43, + "learning_rate": 1.1572853799729032e-06, + "loss": 0.2261, + "step": 49920 + }, + { + "epoch": 18.43, + "learning_rate": 1.1572484296095579e-06, + "loss": 0.335, + "step": 49921 + }, + { + "epoch": 18.44, + "learning_rate": 1.1572114792462125e-06, + "loss": 0.341, + "step": 49922 + }, + { + "epoch": 18.44, + "learning_rate": 1.1571745288828674e-06, + "loss": 0.3634, + "step": 49923 + }, + { + "epoch": 18.44, + "learning_rate": 1.1571375785195222e-06, + "loss": 0.1409, + "step": 49924 + }, + { + "epoch": 18.44, + "learning_rate": 1.1571006281561768e-06, + "loss": 0.2183, + "step": 49925 + }, + { + "epoch": 18.44, + "learning_rate": 1.1570636777928317e-06, + "loss": 0.2433, + "step": 49926 + }, + { + "epoch": 18.44, + "learning_rate": 1.1570267274294865e-06, + "loss": 0.1433, + "step": 49927 + }, + { + "epoch": 18.44, + "learning_rate": 1.1569897770661412e-06, + "loss": 0.3996, + "step": 49928 + }, + { + "epoch": 18.44, + "learning_rate": 1.156952826702796e-06, + "loss": 0.1861, + "step": 49929 + }, + { + "epoch": 18.44, + "learning_rate": 1.1569158763394507e-06, + "loss": 0.2126, + "step": 49930 + }, + { + "epoch": 18.44, + "learning_rate": 1.1568789259761053e-06, + "loss": 0.2473, + "step": 49931 + }, + { + "epoch": 18.44, + "learning_rate": 1.1568419756127601e-06, + "loss": 0.1992, + "step": 49932 + }, + { + "epoch": 18.44, + "learning_rate": 1.156805025249415e-06, + "loss": 0.1717, + "step": 49933 + }, + { + "epoch": 18.44, + "learning_rate": 1.1567680748860696e-06, + "loss": 0.159, + "step": 49934 + }, + { + "epoch": 18.44, + "learning_rate": 1.1567311245227245e-06, + "loss": 0.2368, + "step": 49935 + }, + { + "epoch": 18.44, + "learning_rate": 1.1566941741593793e-06, + "loss": 0.1789, + "step": 49936 + }, + { + "epoch": 18.44, + "learning_rate": 1.156657223796034e-06, + "loss": 0.1105, + "step": 49937 + }, + { + "epoch": 18.44, + "learning_rate": 1.1566202734326888e-06, + "loss": 0.1563, + "step": 49938 + }, + { + "epoch": 18.44, + "learning_rate": 1.1565833230693437e-06, + "loss": 0.1998, + "step": 49939 + }, + { + "epoch": 18.44, + "learning_rate": 1.1565463727059983e-06, + "loss": 0.2244, + "step": 49940 + }, + { + "epoch": 18.44, + "learning_rate": 1.156509422342653e-06, + "loss": 0.2804, + "step": 49941 + }, + { + "epoch": 18.44, + "learning_rate": 1.1564724719793078e-06, + "loss": 0.3293, + "step": 49942 + }, + { + "epoch": 18.44, + "learning_rate": 1.1564355216159626e-06, + "loss": 0.1435, + "step": 49943 + }, + { + "epoch": 18.44, + "learning_rate": 1.1563985712526173e-06, + "loss": 0.2753, + "step": 49944 + }, + { + "epoch": 18.44, + "learning_rate": 1.1563616208892721e-06, + "loss": 0.21, + "step": 49945 + }, + { + "epoch": 18.44, + "learning_rate": 1.156324670525927e-06, + "loss": 0.114, + "step": 49946 + }, + { + "epoch": 18.44, + "learning_rate": 1.1562877201625816e-06, + "loss": 0.2283, + "step": 49947 + }, + { + "epoch": 18.44, + "learning_rate": 1.1562507697992364e-06, + "loss": 0.1256, + "step": 49948 + }, + { + "epoch": 18.44, + "learning_rate": 1.1562138194358913e-06, + "loss": 0.1527, + "step": 49949 + }, + { + "epoch": 18.45, + "learning_rate": 1.156176869072546e-06, + "loss": 0.212, + "step": 49950 + }, + { + "epoch": 18.45, + "learning_rate": 1.1561399187092006e-06, + "loss": 0.2595, + "step": 49951 + }, + { + "epoch": 18.45, + "learning_rate": 1.1561029683458554e-06, + "loss": 0.1502, + "step": 49952 + }, + { + "epoch": 18.45, + "learning_rate": 1.1560660179825103e-06, + "loss": 0.1667, + "step": 49953 + }, + { + "epoch": 18.45, + "learning_rate": 1.156029067619165e-06, + "loss": 0.2324, + "step": 49954 + }, + { + "epoch": 18.45, + "learning_rate": 1.1559921172558197e-06, + "loss": 0.1961, + "step": 49955 + }, + { + "epoch": 18.45, + "learning_rate": 1.1559551668924746e-06, + "loss": 0.1863, + "step": 49956 + }, + { + "epoch": 18.45, + "learning_rate": 1.1559182165291292e-06, + "loss": 0.2629, + "step": 49957 + }, + { + "epoch": 18.45, + "learning_rate": 1.155881266165784e-06, + "loss": 0.1349, + "step": 49958 + }, + { + "epoch": 18.45, + "learning_rate": 1.155844315802439e-06, + "loss": 0.3838, + "step": 49959 + }, + { + "epoch": 18.45, + "learning_rate": 1.1558073654390934e-06, + "loss": 0.2824, + "step": 49960 + }, + { + "epoch": 18.45, + "learning_rate": 1.1557704150757482e-06, + "loss": 0.1856, + "step": 49961 + }, + { + "epoch": 18.45, + "learning_rate": 1.155733464712403e-06, + "loss": 0.2826, + "step": 49962 + }, + { + "epoch": 18.45, + "learning_rate": 1.1556965143490577e-06, + "loss": 0.2725, + "step": 49963 + }, + { + "epoch": 18.45, + "learning_rate": 1.1556595639857125e-06, + "loss": 0.2487, + "step": 49964 + }, + { + "epoch": 18.45, + "learning_rate": 1.1556226136223674e-06, + "loss": 0.2007, + "step": 49965 + }, + { + "epoch": 18.45, + "learning_rate": 1.155585663259022e-06, + "loss": 0.2965, + "step": 49966 + }, + { + "epoch": 18.45, + "learning_rate": 1.1555487128956769e-06, + "loss": 0.1212, + "step": 49967 + }, + { + "epoch": 18.45, + "learning_rate": 1.1555117625323317e-06, + "loss": 0.2952, + "step": 49968 + }, + { + "epoch": 18.45, + "learning_rate": 1.1554748121689864e-06, + "loss": 0.298, + "step": 49969 + }, + { + "epoch": 18.45, + "learning_rate": 1.155437861805641e-06, + "loss": 0.3127, + "step": 49970 + }, + { + "epoch": 18.45, + "learning_rate": 1.1554009114422958e-06, + "loss": 0.0862, + "step": 49971 + }, + { + "epoch": 18.45, + "learning_rate": 1.1553639610789507e-06, + "loss": 0.1967, + "step": 49972 + }, + { + "epoch": 18.45, + "learning_rate": 1.1553270107156053e-06, + "loss": 0.3521, + "step": 49973 + }, + { + "epoch": 18.45, + "learning_rate": 1.1552900603522602e-06, + "loss": 0.1651, + "step": 49974 + }, + { + "epoch": 18.45, + "learning_rate": 1.155253109988915e-06, + "loss": 0.2285, + "step": 49975 + }, + { + "epoch": 18.45, + "learning_rate": 1.1552161596255697e-06, + "loss": 0.3065, + "step": 49976 + }, + { + "epoch": 18.46, + "learning_rate": 1.1551792092622245e-06, + "loss": 0.2799, + "step": 49977 + }, + { + "epoch": 18.46, + "learning_rate": 1.1551422588988793e-06, + "loss": 0.2756, + "step": 49978 + }, + { + "epoch": 18.46, + "learning_rate": 1.155105308535534e-06, + "loss": 0.241, + "step": 49979 + }, + { + "epoch": 18.46, + "learning_rate": 1.1550683581721886e-06, + "loss": 0.2799, + "step": 49980 + }, + { + "epoch": 18.46, + "learning_rate": 1.1550314078088435e-06, + "loss": 0.3513, + "step": 49981 + }, + { + "epoch": 18.46, + "learning_rate": 1.1549944574454981e-06, + "loss": 0.1352, + "step": 49982 + }, + { + "epoch": 18.46, + "learning_rate": 1.154957507082153e-06, + "loss": 0.157, + "step": 49983 + }, + { + "epoch": 18.46, + "learning_rate": 1.1549205567188078e-06, + "loss": 0.1936, + "step": 49984 + }, + { + "epoch": 18.46, + "learning_rate": 1.1548836063554624e-06, + "loss": 0.1232, + "step": 49985 + }, + { + "epoch": 18.46, + "learning_rate": 1.1548466559921173e-06, + "loss": 0.2064, + "step": 49986 + }, + { + "epoch": 18.46, + "learning_rate": 1.1548097056287721e-06, + "loss": 0.2501, + "step": 49987 + }, + { + "epoch": 18.46, + "learning_rate": 1.1547727552654268e-06, + "loss": 0.2726, + "step": 49988 + }, + { + "epoch": 18.46, + "learning_rate": 1.1547358049020816e-06, + "loss": 0.2175, + "step": 49989 + }, + { + "epoch": 18.46, + "learning_rate": 1.1546988545387363e-06, + "loss": 0.2504, + "step": 49990 + }, + { + "epoch": 18.46, + "learning_rate": 1.154661904175391e-06, + "loss": 0.218, + "step": 49991 + }, + { + "epoch": 18.46, + "learning_rate": 1.1546249538120457e-06, + "loss": 0.2, + "step": 49992 + }, + { + "epoch": 18.46, + "learning_rate": 1.1545880034487006e-06, + "loss": 0.1756, + "step": 49993 + }, + { + "epoch": 18.46, + "learning_rate": 1.1545510530853554e-06, + "loss": 0.3514, + "step": 49994 + }, + { + "epoch": 18.46, + "learning_rate": 1.15451410272201e-06, + "loss": 0.2595, + "step": 49995 + }, + { + "epoch": 18.46, + "learning_rate": 1.154477152358665e-06, + "loss": 0.2061, + "step": 49996 + }, + { + "epoch": 18.46, + "learning_rate": 1.1544402019953198e-06, + "loss": 0.1521, + "step": 49997 + }, + { + "epoch": 18.46, + "learning_rate": 1.1544032516319744e-06, + "loss": 0.2484, + "step": 49998 + }, + { + "epoch": 18.46, + "learning_rate": 1.1543663012686293e-06, + "loss": 0.3053, + "step": 49999 + }, + { + "epoch": 18.46, + "learning_rate": 1.1543293509052839e-06, + "loss": 0.1718, + "step": 50000 + }, + { + "epoch": 18.46, + "learning_rate": 1.1542924005419385e-06, + "loss": 0.3511, + "step": 50001 + }, + { + "epoch": 18.46, + "learning_rate": 1.1542554501785934e-06, + "loss": 0.2062, + "step": 50002 + }, + { + "epoch": 18.46, + "learning_rate": 1.1542184998152482e-06, + "loss": 0.1584, + "step": 50003 + }, + { + "epoch": 18.47, + "learning_rate": 1.1541815494519029e-06, + "loss": 0.2558, + "step": 50004 + }, + { + "epoch": 18.47, + "learning_rate": 1.1541445990885577e-06, + "loss": 0.1593, + "step": 50005 + }, + { + "epoch": 18.47, + "learning_rate": 1.1541076487252126e-06, + "loss": 0.222, + "step": 50006 + }, + { + "epoch": 18.47, + "learning_rate": 1.1540706983618672e-06, + "loss": 0.1956, + "step": 50007 + }, + { + "epoch": 18.47, + "learning_rate": 1.154033747998522e-06, + "loss": 0.2713, + "step": 50008 + }, + { + "epoch": 18.47, + "learning_rate": 1.1539967976351769e-06, + "loss": 0.2223, + "step": 50009 + }, + { + "epoch": 18.47, + "learning_rate": 1.1539598472718315e-06, + "loss": 0.2574, + "step": 50010 + }, + { + "epoch": 18.47, + "learning_rate": 1.1539228969084862e-06, + "loss": 0.2386, + "step": 50011 + }, + { + "epoch": 18.47, + "learning_rate": 1.153885946545141e-06, + "loss": 0.1344, + "step": 50012 + }, + { + "epoch": 18.47, + "learning_rate": 1.1538489961817959e-06, + "loss": 0.2877, + "step": 50013 + }, + { + "epoch": 18.47, + "learning_rate": 1.1538120458184505e-06, + "loss": 0.2182, + "step": 50014 + }, + { + "epoch": 18.47, + "learning_rate": 1.1537750954551053e-06, + "loss": 0.2086, + "step": 50015 + }, + { + "epoch": 18.47, + "learning_rate": 1.1537381450917602e-06, + "loss": 0.2011, + "step": 50016 + }, + { + "epoch": 18.47, + "learning_rate": 1.1537011947284148e-06, + "loss": 0.1933, + "step": 50017 + }, + { + "epoch": 18.47, + "learning_rate": 1.1536642443650697e-06, + "loss": 0.139, + "step": 50018 + }, + { + "epoch": 18.47, + "learning_rate": 1.1536272940017245e-06, + "loss": 0.2456, + "step": 50019 + }, + { + "epoch": 18.47, + "learning_rate": 1.1535903436383792e-06, + "loss": 0.2193, + "step": 50020 + }, + { + "epoch": 18.47, + "learning_rate": 1.1535533932750338e-06, + "loss": 0.2189, + "step": 50021 + }, + { + "epoch": 18.47, + "learning_rate": 1.1535164429116886e-06, + "loss": 0.1174, + "step": 50022 + }, + { + "epoch": 18.47, + "learning_rate": 1.1534794925483435e-06, + "loss": 0.1879, + "step": 50023 + }, + { + "epoch": 18.47, + "learning_rate": 1.1534425421849981e-06, + "loss": 0.237, + "step": 50024 + }, + { + "epoch": 18.47, + "learning_rate": 1.153405591821653e-06, + "loss": 0.1361, + "step": 50025 + }, + { + "epoch": 18.47, + "learning_rate": 1.1533686414583078e-06, + "loss": 0.1782, + "step": 50026 + }, + { + "epoch": 18.47, + "learning_rate": 1.1533316910949625e-06, + "loss": 0.1954, + "step": 50027 + }, + { + "epoch": 18.47, + "learning_rate": 1.1532947407316173e-06, + "loss": 0.1661, + "step": 50028 + }, + { + "epoch": 18.47, + "learning_rate": 1.1532577903682722e-06, + "loss": 0.1736, + "step": 50029 + }, + { + "epoch": 18.47, + "learning_rate": 1.1532208400049266e-06, + "loss": 0.1645, + "step": 50030 + }, + { + "epoch": 18.48, + "learning_rate": 1.1531838896415814e-06, + "loss": 0.238, + "step": 50031 + }, + { + "epoch": 18.48, + "learning_rate": 1.1531469392782363e-06, + "loss": 0.2905, + "step": 50032 + }, + { + "epoch": 18.48, + "learning_rate": 1.153109988914891e-06, + "loss": 0.0943, + "step": 50033 + }, + { + "epoch": 18.48, + "learning_rate": 1.1530730385515458e-06, + "loss": 0.1656, + "step": 50034 + }, + { + "epoch": 18.48, + "learning_rate": 1.1530360881882006e-06, + "loss": 0.2744, + "step": 50035 + }, + { + "epoch": 18.48, + "learning_rate": 1.1529991378248552e-06, + "loss": 0.1585, + "step": 50036 + }, + { + "epoch": 18.48, + "learning_rate": 1.15296218746151e-06, + "loss": 0.2601, + "step": 50037 + }, + { + "epoch": 18.48, + "learning_rate": 1.152925237098165e-06, + "loss": 0.1823, + "step": 50038 + }, + { + "epoch": 18.48, + "learning_rate": 1.1528882867348196e-06, + "loss": 0.2293, + "step": 50039 + }, + { + "epoch": 18.48, + "learning_rate": 1.1528513363714742e-06, + "loss": 0.2871, + "step": 50040 + }, + { + "epoch": 18.48, + "learning_rate": 1.152814386008129e-06, + "loss": 0.1843, + "step": 50041 + }, + { + "epoch": 18.48, + "learning_rate": 1.152777435644784e-06, + "loss": 0.1895, + "step": 50042 + }, + { + "epoch": 18.48, + "learning_rate": 1.1527404852814386e-06, + "loss": 0.2695, + "step": 50043 + }, + { + "epoch": 18.48, + "learning_rate": 1.1527035349180934e-06, + "loss": 0.2111, + "step": 50044 + }, + { + "epoch": 18.48, + "learning_rate": 1.1526665845547482e-06, + "loss": 0.3411, + "step": 50045 + }, + { + "epoch": 18.48, + "learning_rate": 1.1526296341914029e-06, + "loss": 0.2073, + "step": 50046 + }, + { + "epoch": 18.48, + "learning_rate": 1.1525926838280577e-06, + "loss": 0.3154, + "step": 50047 + }, + { + "epoch": 18.48, + "learning_rate": 1.1525557334647126e-06, + "loss": 0.2042, + "step": 50048 + }, + { + "epoch": 18.48, + "learning_rate": 1.1525187831013672e-06, + "loss": 0.1942, + "step": 50049 + }, + { + "epoch": 18.48, + "learning_rate": 1.1524818327380219e-06, + "loss": 0.2903, + "step": 50050 + }, + { + "epoch": 18.48, + "learning_rate": 1.1524448823746767e-06, + "loss": 0.2534, + "step": 50051 + }, + { + "epoch": 18.48, + "learning_rate": 1.1524079320113313e-06, + "loss": 0.1715, + "step": 50052 + }, + { + "epoch": 18.48, + "learning_rate": 1.1523709816479862e-06, + "loss": 0.3354, + "step": 50053 + }, + { + "epoch": 18.48, + "learning_rate": 1.152334031284641e-06, + "loss": 0.2331, + "step": 50054 + }, + { + "epoch": 18.48, + "learning_rate": 1.1522970809212957e-06, + "loss": 0.2676, + "step": 50055 + }, + { + "epoch": 18.48, + "learning_rate": 1.1522601305579505e-06, + "loss": 0.2774, + "step": 50056 + }, + { + "epoch": 18.48, + "learning_rate": 1.1522231801946054e-06, + "loss": 0.3619, + "step": 50057 + }, + { + "epoch": 18.49, + "learning_rate": 1.15218622983126e-06, + "loss": 0.1203, + "step": 50058 + }, + { + "epoch": 18.49, + "learning_rate": 1.1521492794679149e-06, + "loss": 0.2218, + "step": 50059 + }, + { + "epoch": 18.49, + "learning_rate": 1.1521123291045695e-06, + "loss": 0.2421, + "step": 50060 + }, + { + "epoch": 18.49, + "learning_rate": 1.1520753787412243e-06, + "loss": 0.3394, + "step": 50061 + }, + { + "epoch": 18.49, + "learning_rate": 1.152038428377879e-06, + "loss": 0.1968, + "step": 50062 + }, + { + "epoch": 18.49, + "learning_rate": 1.1520014780145338e-06, + "loss": 0.1916, + "step": 50063 + }, + { + "epoch": 18.49, + "learning_rate": 1.1519645276511887e-06, + "loss": 0.1895, + "step": 50064 + }, + { + "epoch": 18.49, + "learning_rate": 1.1519275772878433e-06, + "loss": 0.1855, + "step": 50065 + }, + { + "epoch": 18.49, + "learning_rate": 1.1518906269244982e-06, + "loss": 0.2705, + "step": 50066 + }, + { + "epoch": 18.49, + "learning_rate": 1.151853676561153e-06, + "loss": 0.1648, + "step": 50067 + }, + { + "epoch": 18.49, + "learning_rate": 1.1518167261978076e-06, + "loss": 0.2149, + "step": 50068 + }, + { + "epoch": 18.49, + "learning_rate": 1.1517797758344625e-06, + "loss": 0.391, + "step": 50069 + }, + { + "epoch": 18.49, + "learning_rate": 1.1517428254711171e-06, + "loss": 0.3143, + "step": 50070 + }, + { + "epoch": 18.49, + "learning_rate": 1.151705875107772e-06, + "loss": 0.2603, + "step": 50071 + }, + { + "epoch": 18.49, + "learning_rate": 1.1516689247444266e-06, + "loss": 0.2275, + "step": 50072 + }, + { + "epoch": 18.49, + "learning_rate": 1.1516319743810815e-06, + "loss": 0.1257, + "step": 50073 + }, + { + "epoch": 18.49, + "learning_rate": 1.151595024017736e-06, + "loss": 0.131, + "step": 50074 + }, + { + "epoch": 18.49, + "learning_rate": 1.151558073654391e-06, + "loss": 0.1346, + "step": 50075 + }, + { + "epoch": 18.49, + "learning_rate": 1.1515211232910458e-06, + "loss": 0.1903, + "step": 50076 + }, + { + "epoch": 18.49, + "learning_rate": 1.1514841729277004e-06, + "loss": 0.336, + "step": 50077 + }, + { + "epoch": 18.49, + "learning_rate": 1.1514472225643553e-06, + "loss": 0.1945, + "step": 50078 + }, + { + "epoch": 18.49, + "learning_rate": 1.1514102722010101e-06, + "loss": 0.1907, + "step": 50079 + }, + { + "epoch": 18.49, + "learning_rate": 1.1513733218376648e-06, + "loss": 0.2121, + "step": 50080 + }, + { + "epoch": 18.49, + "learning_rate": 1.1513363714743194e-06, + "loss": 0.2078, + "step": 50081 + }, + { + "epoch": 18.49, + "learning_rate": 1.1512994211109742e-06, + "loss": 0.1704, + "step": 50082 + }, + { + "epoch": 18.49, + "learning_rate": 1.151262470747629e-06, + "loss": 0.2247, + "step": 50083 + }, + { + "epoch": 18.49, + "learning_rate": 1.1512255203842837e-06, + "loss": 0.1529, + "step": 50084 + }, + { + "epoch": 18.5, + "learning_rate": 1.1511885700209386e-06, + "loss": 0.1899, + "step": 50085 + }, + { + "epoch": 18.5, + "learning_rate": 1.1511516196575934e-06, + "loss": 0.2606, + "step": 50086 + }, + { + "epoch": 18.5, + "learning_rate": 1.151114669294248e-06, + "loss": 0.2629, + "step": 50087 + }, + { + "epoch": 18.5, + "learning_rate": 1.151077718930903e-06, + "loss": 0.1453, + "step": 50088 + }, + { + "epoch": 18.5, + "learning_rate": 1.1510407685675578e-06, + "loss": 0.2172, + "step": 50089 + }, + { + "epoch": 18.5, + "learning_rate": 1.1510038182042124e-06, + "loss": 0.1571, + "step": 50090 + }, + { + "epoch": 18.5, + "learning_rate": 1.150966867840867e-06, + "loss": 0.1887, + "step": 50091 + }, + { + "epoch": 18.5, + "learning_rate": 1.1509299174775219e-06, + "loss": 0.4349, + "step": 50092 + }, + { + "epoch": 18.5, + "learning_rate": 1.1508929671141767e-06, + "loss": 0.1173, + "step": 50093 + }, + { + "epoch": 18.5, + "learning_rate": 1.1508560167508314e-06, + "loss": 0.1477, + "step": 50094 + }, + { + "epoch": 18.5, + "learning_rate": 1.1508190663874862e-06, + "loss": 0.2095, + "step": 50095 + }, + { + "epoch": 18.5, + "learning_rate": 1.150782116024141e-06, + "loss": 0.2847, + "step": 50096 + }, + { + "epoch": 18.5, + "learning_rate": 1.1507451656607957e-06, + "loss": 0.2541, + "step": 50097 + }, + { + "epoch": 18.5, + "learning_rate": 1.1507082152974505e-06, + "loss": 0.2098, + "step": 50098 + }, + { + "epoch": 18.5, + "learning_rate": 1.1506712649341054e-06, + "loss": 0.2749, + "step": 50099 + }, + { + "epoch": 18.5, + "learning_rate": 1.1506343145707598e-06, + "loss": 0.2429, + "step": 50100 + }, + { + "epoch": 18.5, + "learning_rate": 1.1505973642074147e-06, + "loss": 0.306, + "step": 50101 + }, + { + "epoch": 18.5, + "learning_rate": 1.1505604138440695e-06, + "loss": 0.1914, + "step": 50102 + }, + { + "epoch": 18.5, + "learning_rate": 1.1505234634807241e-06, + "loss": 0.1532, + "step": 50103 + }, + { + "epoch": 18.5, + "learning_rate": 1.150486513117379e-06, + "loss": 0.1507, + "step": 50104 + }, + { + "epoch": 18.5, + "learning_rate": 1.1504495627540338e-06, + "loss": 0.2518, + "step": 50105 + }, + { + "epoch": 18.5, + "learning_rate": 1.1504126123906885e-06, + "loss": 0.2909, + "step": 50106 + }, + { + "epoch": 18.5, + "learning_rate": 1.1503756620273433e-06, + "loss": 0.3977, + "step": 50107 + }, + { + "epoch": 18.5, + "learning_rate": 1.1503387116639982e-06, + "loss": 0.316, + "step": 50108 + }, + { + "epoch": 18.5, + "learning_rate": 1.1503017613006528e-06, + "loss": 0.2233, + "step": 50109 + }, + { + "epoch": 18.5, + "learning_rate": 1.1502648109373075e-06, + "loss": 0.2172, + "step": 50110 + }, + { + "epoch": 18.5, + "learning_rate": 1.1502278605739623e-06, + "loss": 0.2971, + "step": 50111 + }, + { + "epoch": 18.51, + "learning_rate": 1.1501909102106171e-06, + "loss": 0.1575, + "step": 50112 + }, + { + "epoch": 18.51, + "learning_rate": 1.1501539598472718e-06, + "loss": 0.3702, + "step": 50113 + }, + { + "epoch": 18.51, + "learning_rate": 1.1501170094839266e-06, + "loss": 0.2511, + "step": 50114 + }, + { + "epoch": 18.51, + "learning_rate": 1.1500800591205815e-06, + "loss": 0.3696, + "step": 50115 + }, + { + "epoch": 18.51, + "learning_rate": 1.1500431087572361e-06, + "loss": 0.2527, + "step": 50116 + }, + { + "epoch": 18.51, + "learning_rate": 1.150006158393891e-06, + "loss": 0.2252, + "step": 50117 + }, + { + "epoch": 18.51, + "learning_rate": 1.1499692080305458e-06, + "loss": 0.3035, + "step": 50118 + }, + { + "epoch": 18.51, + "learning_rate": 1.1499322576672005e-06, + "loss": 0.1196, + "step": 50119 + }, + { + "epoch": 18.51, + "learning_rate": 1.149895307303855e-06, + "loss": 0.3483, + "step": 50120 + }, + { + "epoch": 18.51, + "learning_rate": 1.14985835694051e-06, + "loss": 0.2089, + "step": 50121 + }, + { + "epoch": 18.51, + "learning_rate": 1.1498214065771646e-06, + "loss": 0.1654, + "step": 50122 + }, + { + "epoch": 18.51, + "learning_rate": 1.1497844562138194e-06, + "loss": 0.2656, + "step": 50123 + }, + { + "epoch": 18.51, + "learning_rate": 1.1497475058504743e-06, + "loss": 0.3269, + "step": 50124 + }, + { + "epoch": 18.51, + "learning_rate": 1.149710555487129e-06, + "loss": 0.1918, + "step": 50125 + }, + { + "epoch": 18.51, + "learning_rate": 1.1496736051237838e-06, + "loss": 0.1807, + "step": 50126 + }, + { + "epoch": 18.51, + "learning_rate": 1.1496366547604386e-06, + "loss": 0.3016, + "step": 50127 + }, + { + "epoch": 18.51, + "learning_rate": 1.1495997043970932e-06, + "loss": 0.3201, + "step": 50128 + }, + { + "epoch": 18.51, + "learning_rate": 1.149562754033748e-06, + "loss": 0.1149, + "step": 50129 + }, + { + "epoch": 18.51, + "learning_rate": 1.1495258036704027e-06, + "loss": 0.1709, + "step": 50130 + }, + { + "epoch": 18.51, + "learning_rate": 1.1494888533070576e-06, + "loss": 0.2829, + "step": 50131 + }, + { + "epoch": 18.51, + "learning_rate": 1.1494519029437122e-06, + "loss": 0.2441, + "step": 50132 + }, + { + "epoch": 18.51, + "learning_rate": 1.149414952580367e-06, + "loss": 0.27, + "step": 50133 + }, + { + "epoch": 18.51, + "learning_rate": 1.149378002217022e-06, + "loss": 0.3614, + "step": 50134 + }, + { + "epoch": 18.51, + "learning_rate": 1.1493410518536765e-06, + "loss": 0.3358, + "step": 50135 + }, + { + "epoch": 18.51, + "learning_rate": 1.1493041014903314e-06, + "loss": 0.3313, + "step": 50136 + }, + { + "epoch": 18.51, + "learning_rate": 1.1492671511269862e-06, + "loss": 0.3516, + "step": 50137 + }, + { + "epoch": 18.51, + "learning_rate": 1.1492302007636409e-06, + "loss": 0.19, + "step": 50138 + }, + { + "epoch": 18.52, + "learning_rate": 1.1491932504002957e-06, + "loss": 0.1822, + "step": 50139 + }, + { + "epoch": 18.52, + "learning_rate": 1.1491563000369504e-06, + "loss": 0.1689, + "step": 50140 + }, + { + "epoch": 18.52, + "learning_rate": 1.1491193496736052e-06, + "loss": 0.2267, + "step": 50141 + }, + { + "epoch": 18.52, + "learning_rate": 1.1490823993102598e-06, + "loss": 0.1549, + "step": 50142 + }, + { + "epoch": 18.52, + "learning_rate": 1.1490454489469147e-06, + "loss": 0.3104, + "step": 50143 + }, + { + "epoch": 18.52, + "learning_rate": 1.1490084985835695e-06, + "loss": 0.2508, + "step": 50144 + }, + { + "epoch": 18.52, + "learning_rate": 1.1489715482202242e-06, + "loss": 0.2064, + "step": 50145 + }, + { + "epoch": 18.52, + "learning_rate": 1.148934597856879e-06, + "loss": 0.2843, + "step": 50146 + }, + { + "epoch": 18.52, + "learning_rate": 1.1488976474935339e-06, + "loss": 0.2104, + "step": 50147 + }, + { + "epoch": 18.52, + "learning_rate": 1.1488606971301885e-06, + "loss": 0.1092, + "step": 50148 + }, + { + "epoch": 18.52, + "learning_rate": 1.1488237467668434e-06, + "loss": 0.4366, + "step": 50149 + }, + { + "epoch": 18.52, + "learning_rate": 1.148786796403498e-06, + "loss": 0.1613, + "step": 50150 + }, + { + "epoch": 18.52, + "learning_rate": 1.1487498460401526e-06, + "loss": 0.2798, + "step": 50151 + }, + { + "epoch": 18.52, + "learning_rate": 1.1487128956768075e-06, + "loss": 0.1974, + "step": 50152 + }, + { + "epoch": 18.52, + "learning_rate": 1.1486759453134623e-06, + "loss": 0.2364, + "step": 50153 + }, + { + "epoch": 18.52, + "learning_rate": 1.148638994950117e-06, + "loss": 0.2625, + "step": 50154 + }, + { + "epoch": 18.52, + "learning_rate": 1.1486020445867718e-06, + "loss": 0.2159, + "step": 50155 + }, + { + "epoch": 18.52, + "learning_rate": 1.1485650942234267e-06, + "loss": 0.2529, + "step": 50156 + }, + { + "epoch": 18.52, + "learning_rate": 1.1485281438600813e-06, + "loss": 0.204, + "step": 50157 + }, + { + "epoch": 18.52, + "learning_rate": 1.1484911934967361e-06, + "loss": 0.2352, + "step": 50158 + }, + { + "epoch": 18.52, + "learning_rate": 1.1484542431333908e-06, + "loss": 0.2612, + "step": 50159 + }, + { + "epoch": 18.52, + "learning_rate": 1.1484172927700456e-06, + "loss": 0.261, + "step": 50160 + }, + { + "epoch": 18.52, + "learning_rate": 1.1483803424067003e-06, + "loss": 0.2132, + "step": 50161 + }, + { + "epoch": 18.52, + "learning_rate": 1.1483433920433551e-06, + "loss": 0.0758, + "step": 50162 + }, + { + "epoch": 18.52, + "learning_rate": 1.14830644168001e-06, + "loss": 0.1811, + "step": 50163 + }, + { + "epoch": 18.52, + "learning_rate": 1.1482694913166646e-06, + "loss": 0.3686, + "step": 50164 + }, + { + "epoch": 18.52, + "learning_rate": 1.1482325409533194e-06, + "loss": 0.1196, + "step": 50165 + }, + { + "epoch": 18.53, + "learning_rate": 1.1481955905899743e-06, + "loss": 0.2667, + "step": 50166 + }, + { + "epoch": 18.53, + "learning_rate": 1.148158640226629e-06, + "loss": 0.2581, + "step": 50167 + }, + { + "epoch": 18.53, + "learning_rate": 1.1481216898632838e-06, + "loss": 0.3012, + "step": 50168 + }, + { + "epoch": 18.53, + "learning_rate": 1.1480847394999384e-06, + "loss": 0.3066, + "step": 50169 + }, + { + "epoch": 18.53, + "learning_rate": 1.148047789136593e-06, + "loss": 0.1855, + "step": 50170 + }, + { + "epoch": 18.53, + "learning_rate": 1.148010838773248e-06, + "loss": 0.2581, + "step": 50171 + }, + { + "epoch": 18.53, + "learning_rate": 1.1479738884099027e-06, + "loss": 0.1295, + "step": 50172 + }, + { + "epoch": 18.53, + "learning_rate": 1.1479369380465574e-06, + "loss": 0.1704, + "step": 50173 + }, + { + "epoch": 18.53, + "learning_rate": 1.1478999876832122e-06, + "loss": 0.1599, + "step": 50174 + }, + { + "epoch": 18.53, + "learning_rate": 1.147863037319867e-06, + "loss": 0.2437, + "step": 50175 + }, + { + "epoch": 18.53, + "learning_rate": 1.1478260869565217e-06, + "loss": 0.2201, + "step": 50176 + }, + { + "epoch": 18.53, + "learning_rate": 1.1477891365931766e-06, + "loss": 0.1676, + "step": 50177 + }, + { + "epoch": 18.53, + "learning_rate": 1.1477521862298314e-06, + "loss": 0.2213, + "step": 50178 + }, + { + "epoch": 18.53, + "learning_rate": 1.147715235866486e-06, + "loss": 0.274, + "step": 50179 + }, + { + "epoch": 18.53, + "learning_rate": 1.1476782855031407e-06, + "loss": 0.3163, + "step": 50180 + }, + { + "epoch": 18.53, + "learning_rate": 1.1476413351397955e-06, + "loss": 0.183, + "step": 50181 + }, + { + "epoch": 18.53, + "learning_rate": 1.1476043847764504e-06, + "loss": 0.3098, + "step": 50182 + }, + { + "epoch": 18.53, + "learning_rate": 1.147567434413105e-06, + "loss": 0.3295, + "step": 50183 + }, + { + "epoch": 18.53, + "learning_rate": 1.1475304840497599e-06, + "loss": 0.1682, + "step": 50184 + }, + { + "epoch": 18.53, + "learning_rate": 1.1474935336864147e-06, + "loss": 0.312, + "step": 50185 + }, + { + "epoch": 18.53, + "learning_rate": 1.1474565833230693e-06, + "loss": 0.147, + "step": 50186 + }, + { + "epoch": 18.53, + "learning_rate": 1.1474196329597242e-06, + "loss": 0.2469, + "step": 50187 + }, + { + "epoch": 18.53, + "learning_rate": 1.147382682596379e-06, + "loss": 0.3307, + "step": 50188 + }, + { + "epoch": 18.53, + "learning_rate": 1.1473457322330335e-06, + "loss": 0.3387, + "step": 50189 + }, + { + "epoch": 18.53, + "learning_rate": 1.1473087818696883e-06, + "loss": 0.2195, + "step": 50190 + }, + { + "epoch": 18.53, + "learning_rate": 1.1472718315063432e-06, + "loss": 0.1552, + "step": 50191 + }, + { + "epoch": 18.53, + "learning_rate": 1.1472348811429978e-06, + "loss": 0.4138, + "step": 50192 + }, + { + "epoch": 18.54, + "learning_rate": 1.1471979307796527e-06, + "loss": 0.2612, + "step": 50193 + }, + { + "epoch": 18.54, + "learning_rate": 1.1471609804163075e-06, + "loss": 0.2784, + "step": 50194 + }, + { + "epoch": 18.54, + "learning_rate": 1.1471240300529621e-06, + "loss": 0.2341, + "step": 50195 + }, + { + "epoch": 18.54, + "learning_rate": 1.147087079689617e-06, + "loss": 0.2592, + "step": 50196 + }, + { + "epoch": 18.54, + "learning_rate": 1.1470501293262718e-06, + "loss": 0.2247, + "step": 50197 + }, + { + "epoch": 18.54, + "learning_rate": 1.1470131789629265e-06, + "loss": 0.0914, + "step": 50198 + }, + { + "epoch": 18.54, + "learning_rate": 1.146976228599581e-06, + "loss": 0.2442, + "step": 50199 + }, + { + "epoch": 18.54, + "learning_rate": 1.146939278236236e-06, + "loss": 0.2061, + "step": 50200 + }, + { + "epoch": 18.54, + "learning_rate": 1.1469023278728908e-06, + "loss": 0.2214, + "step": 50201 + }, + { + "epoch": 18.54, + "learning_rate": 1.1468653775095454e-06, + "loss": 0.1828, + "step": 50202 + }, + { + "epoch": 18.54, + "learning_rate": 1.1468284271462003e-06, + "loss": 0.1624, + "step": 50203 + }, + { + "epoch": 18.54, + "learning_rate": 1.1467914767828551e-06, + "loss": 0.2146, + "step": 50204 + }, + { + "epoch": 18.54, + "learning_rate": 1.1467545264195098e-06, + "loss": 0.1714, + "step": 50205 + }, + { + "epoch": 18.54, + "learning_rate": 1.1467175760561646e-06, + "loss": 0.1832, + "step": 50206 + }, + { + "epoch": 18.54, + "learning_rate": 1.1466806256928195e-06, + "loss": 0.2285, + "step": 50207 + }, + { + "epoch": 18.54, + "learning_rate": 1.146643675329474e-06, + "loss": 0.3082, + "step": 50208 + }, + { + "epoch": 18.54, + "learning_rate": 1.1466067249661287e-06, + "loss": 0.2359, + "step": 50209 + }, + { + "epoch": 18.54, + "learning_rate": 1.1465697746027836e-06, + "loss": 0.2853, + "step": 50210 + }, + { + "epoch": 18.54, + "learning_rate": 1.1465328242394384e-06, + "loss": 0.1979, + "step": 50211 + }, + { + "epoch": 18.54, + "learning_rate": 1.146495873876093e-06, + "loss": 0.1855, + "step": 50212 + }, + { + "epoch": 18.54, + "learning_rate": 1.146458923512748e-06, + "loss": 0.2361, + "step": 50213 + }, + { + "epoch": 18.54, + "learning_rate": 1.1464219731494028e-06, + "loss": 0.1326, + "step": 50214 + }, + { + "epoch": 18.54, + "learning_rate": 1.1463850227860574e-06, + "loss": 0.3495, + "step": 50215 + }, + { + "epoch": 18.54, + "learning_rate": 1.1463480724227123e-06, + "loss": 0.2067, + "step": 50216 + }, + { + "epoch": 18.54, + "learning_rate": 1.146311122059367e-06, + "loss": 0.1608, + "step": 50217 + }, + { + "epoch": 18.54, + "learning_rate": 1.1462741716960217e-06, + "loss": 0.2669, + "step": 50218 + }, + { + "epoch": 18.54, + "learning_rate": 1.1462372213326764e-06, + "loss": 0.2264, + "step": 50219 + }, + { + "epoch": 18.55, + "learning_rate": 1.1462002709693312e-06, + "loss": 0.2399, + "step": 50220 + }, + { + "epoch": 18.55, + "learning_rate": 1.1461633206059859e-06, + "loss": 0.108, + "step": 50221 + }, + { + "epoch": 18.55, + "learning_rate": 1.1461263702426407e-06, + "loss": 0.111, + "step": 50222 + }, + { + "epoch": 18.55, + "learning_rate": 1.1460894198792956e-06, + "loss": 0.1883, + "step": 50223 + }, + { + "epoch": 18.55, + "learning_rate": 1.1460524695159502e-06, + "loss": 0.1732, + "step": 50224 + }, + { + "epoch": 18.55, + "learning_rate": 1.146015519152605e-06, + "loss": 0.3722, + "step": 50225 + }, + { + "epoch": 18.55, + "learning_rate": 1.1459785687892599e-06, + "loss": 0.1694, + "step": 50226 + }, + { + "epoch": 18.55, + "learning_rate": 1.1459416184259145e-06, + "loss": 0.1284, + "step": 50227 + }, + { + "epoch": 18.55, + "learning_rate": 1.1459046680625694e-06, + "loss": 0.1834, + "step": 50228 + }, + { + "epoch": 18.55, + "learning_rate": 1.145867717699224e-06, + "loss": 0.1188, + "step": 50229 + }, + { + "epoch": 18.55, + "learning_rate": 1.1458307673358789e-06, + "loss": 0.164, + "step": 50230 + }, + { + "epoch": 18.55, + "learning_rate": 1.1457938169725335e-06, + "loss": 0.2003, + "step": 50231 + }, + { + "epoch": 18.55, + "learning_rate": 1.1457568666091883e-06, + "loss": 0.2325, + "step": 50232 + }, + { + "epoch": 18.55, + "learning_rate": 1.1457199162458432e-06, + "loss": 0.2306, + "step": 50233 + }, + { + "epoch": 18.55, + "learning_rate": 1.1456829658824978e-06, + "loss": 0.242, + "step": 50234 + }, + { + "epoch": 18.55, + "learning_rate": 1.1456460155191527e-06, + "loss": 0.1931, + "step": 50235 + }, + { + "epoch": 18.55, + "learning_rate": 1.1456090651558075e-06, + "loss": 0.1939, + "step": 50236 + }, + { + "epoch": 18.55, + "learning_rate": 1.1455721147924622e-06, + "loss": 0.2626, + "step": 50237 + }, + { + "epoch": 18.55, + "learning_rate": 1.145535164429117e-06, + "loss": 0.2297, + "step": 50238 + }, + { + "epoch": 18.55, + "learning_rate": 1.1454982140657716e-06, + "loss": 0.187, + "step": 50239 + }, + { + "epoch": 18.55, + "learning_rate": 1.1454612637024263e-06, + "loss": 0.1886, + "step": 50240 + }, + { + "epoch": 18.55, + "learning_rate": 1.1454243133390811e-06, + "loss": 0.1321, + "step": 50241 + }, + { + "epoch": 18.55, + "learning_rate": 1.145387362975736e-06, + "loss": 0.197, + "step": 50242 + }, + { + "epoch": 18.55, + "learning_rate": 1.1453504126123906e-06, + "loss": 0.3421, + "step": 50243 + }, + { + "epoch": 18.55, + "learning_rate": 1.1453134622490455e-06, + "loss": 0.1572, + "step": 50244 + }, + { + "epoch": 18.55, + "learning_rate": 1.1452765118857003e-06, + "loss": 0.2506, + "step": 50245 + }, + { + "epoch": 18.55, + "learning_rate": 1.145239561522355e-06, + "loss": 0.4103, + "step": 50246 + }, + { + "epoch": 18.56, + "learning_rate": 1.1452026111590098e-06, + "loss": 0.1715, + "step": 50247 + }, + { + "epoch": 18.56, + "learning_rate": 1.1451656607956646e-06, + "loss": 0.1765, + "step": 50248 + }, + { + "epoch": 18.56, + "learning_rate": 1.1451287104323193e-06, + "loss": 0.2426, + "step": 50249 + }, + { + "epoch": 18.56, + "learning_rate": 1.145091760068974e-06, + "loss": 0.28, + "step": 50250 + }, + { + "epoch": 18.56, + "learning_rate": 1.1450548097056288e-06, + "loss": 0.2891, + "step": 50251 + }, + { + "epoch": 18.56, + "learning_rate": 1.1450178593422836e-06, + "loss": 0.2039, + "step": 50252 + }, + { + "epoch": 18.56, + "learning_rate": 1.1449809089789382e-06, + "loss": 0.2086, + "step": 50253 + }, + { + "epoch": 18.56, + "learning_rate": 1.144943958615593e-06, + "loss": 0.2858, + "step": 50254 + }, + { + "epoch": 18.56, + "learning_rate": 1.144907008252248e-06, + "loss": 0.1065, + "step": 50255 + }, + { + "epoch": 18.56, + "learning_rate": 1.1448700578889026e-06, + "loss": 0.2421, + "step": 50256 + }, + { + "epoch": 18.56, + "learning_rate": 1.1448331075255574e-06, + "loss": 0.2264, + "step": 50257 + }, + { + "epoch": 18.56, + "learning_rate": 1.1447961571622123e-06, + "loss": 0.1647, + "step": 50258 + }, + { + "epoch": 18.56, + "learning_rate": 1.1447592067988667e-06, + "loss": 0.1468, + "step": 50259 + }, + { + "epoch": 18.56, + "learning_rate": 1.1447222564355216e-06, + "loss": 0.2244, + "step": 50260 + }, + { + "epoch": 18.56, + "learning_rate": 1.1446853060721764e-06, + "loss": 0.1668, + "step": 50261 + }, + { + "epoch": 18.56, + "learning_rate": 1.144648355708831e-06, + "loss": 0.1534, + "step": 50262 + }, + { + "epoch": 18.56, + "learning_rate": 1.1446114053454859e-06, + "loss": 0.2659, + "step": 50263 + }, + { + "epoch": 18.56, + "learning_rate": 1.1445744549821407e-06, + "loss": 0.2105, + "step": 50264 + }, + { + "epoch": 18.56, + "learning_rate": 1.1445375046187954e-06, + "loss": 0.1958, + "step": 50265 + }, + { + "epoch": 18.56, + "learning_rate": 1.1445005542554502e-06, + "loss": 0.2406, + "step": 50266 + }, + { + "epoch": 18.56, + "learning_rate": 1.144463603892105e-06, + "loss": 0.1744, + "step": 50267 + }, + { + "epoch": 18.56, + "learning_rate": 1.1444266535287597e-06, + "loss": 0.2094, + "step": 50268 + }, + { + "epoch": 18.56, + "learning_rate": 1.1443897031654143e-06, + "loss": 0.2565, + "step": 50269 + }, + { + "epoch": 18.56, + "learning_rate": 1.1443527528020692e-06, + "loss": 0.1709, + "step": 50270 + }, + { + "epoch": 18.56, + "learning_rate": 1.144315802438724e-06, + "loss": 0.1915, + "step": 50271 + }, + { + "epoch": 18.56, + "learning_rate": 1.1442788520753787e-06, + "loss": 0.1724, + "step": 50272 + }, + { + "epoch": 18.56, + "learning_rate": 1.1442419017120335e-06, + "loss": 0.1242, + "step": 50273 + }, + { + "epoch": 18.56, + "learning_rate": 1.1442049513486884e-06, + "loss": 0.2716, + "step": 50274 + }, + { + "epoch": 18.57, + "learning_rate": 1.144168000985343e-06, + "loss": 0.2319, + "step": 50275 + }, + { + "epoch": 18.57, + "learning_rate": 1.1441310506219979e-06, + "loss": 0.1874, + "step": 50276 + }, + { + "epoch": 18.57, + "learning_rate": 1.1440941002586527e-06, + "loss": 0.1251, + "step": 50277 + }, + { + "epoch": 18.57, + "learning_rate": 1.1440571498953073e-06, + "loss": 0.2619, + "step": 50278 + }, + { + "epoch": 18.57, + "learning_rate": 1.144020199531962e-06, + "loss": 0.2013, + "step": 50279 + }, + { + "epoch": 18.57, + "learning_rate": 1.1439832491686168e-06, + "loss": 0.2606, + "step": 50280 + }, + { + "epoch": 18.57, + "learning_rate": 1.1439462988052717e-06, + "loss": 0.1755, + "step": 50281 + }, + { + "epoch": 18.57, + "learning_rate": 1.1439093484419263e-06, + "loss": 0.1251, + "step": 50282 + }, + { + "epoch": 18.57, + "learning_rate": 1.1438723980785812e-06, + "loss": 0.2086, + "step": 50283 + }, + { + "epoch": 18.57, + "learning_rate": 1.143835447715236e-06, + "loss": 0.1447, + "step": 50284 + }, + { + "epoch": 18.57, + "learning_rate": 1.1437984973518906e-06, + "loss": 0.3007, + "step": 50285 + }, + { + "epoch": 18.57, + "learning_rate": 1.1437615469885455e-06, + "loss": 0.372, + "step": 50286 + }, + { + "epoch": 18.57, + "learning_rate": 1.1437245966252003e-06, + "loss": 0.2202, + "step": 50287 + }, + { + "epoch": 18.57, + "learning_rate": 1.143687646261855e-06, + "loss": 0.2534, + "step": 50288 + }, + { + "epoch": 18.57, + "learning_rate": 1.1436506958985096e-06, + "loss": 0.1908, + "step": 50289 + }, + { + "epoch": 18.57, + "learning_rate": 1.1436137455351645e-06, + "loss": 0.1612, + "step": 50290 + }, + { + "epoch": 18.57, + "learning_rate": 1.143576795171819e-06, + "loss": 0.3244, + "step": 50291 + }, + { + "epoch": 18.57, + "learning_rate": 1.143539844808474e-06, + "loss": 0.3199, + "step": 50292 + }, + { + "epoch": 18.57, + "learning_rate": 1.1435028944451288e-06, + "loss": 0.2312, + "step": 50293 + }, + { + "epoch": 18.57, + "learning_rate": 1.1434659440817834e-06, + "loss": 0.2483, + "step": 50294 + }, + { + "epoch": 18.57, + "learning_rate": 1.1434289937184383e-06, + "loss": 0.2584, + "step": 50295 + }, + { + "epoch": 18.57, + "learning_rate": 1.1433920433550931e-06, + "loss": 0.2668, + "step": 50296 + }, + { + "epoch": 18.57, + "learning_rate": 1.1433550929917478e-06, + "loss": 0.254, + "step": 50297 + }, + { + "epoch": 18.57, + "learning_rate": 1.1433181426284026e-06, + "loss": 0.3125, + "step": 50298 + }, + { + "epoch": 18.57, + "learning_rate": 1.1432811922650572e-06, + "loss": 0.2341, + "step": 50299 + }, + { + "epoch": 18.57, + "learning_rate": 1.143244241901712e-06, + "loss": 0.2281, + "step": 50300 + }, + { + "epoch": 18.57, + "learning_rate": 1.1432072915383667e-06, + "loss": 0.293, + "step": 50301 + }, + { + "epoch": 18.58, + "learning_rate": 1.1431703411750216e-06, + "loss": 0.1906, + "step": 50302 + }, + { + "epoch": 18.58, + "learning_rate": 1.1431333908116764e-06, + "loss": 0.268, + "step": 50303 + }, + { + "epoch": 18.58, + "learning_rate": 1.143096440448331e-06, + "loss": 0.1711, + "step": 50304 + }, + { + "epoch": 18.58, + "learning_rate": 1.143059490084986e-06, + "loss": 0.2591, + "step": 50305 + }, + { + "epoch": 18.58, + "learning_rate": 1.1430225397216408e-06, + "loss": 0.2374, + "step": 50306 + }, + { + "epoch": 18.58, + "learning_rate": 1.1429855893582954e-06, + "loss": 0.2303, + "step": 50307 + }, + { + "epoch": 18.58, + "learning_rate": 1.1429486389949502e-06, + "loss": 0.2677, + "step": 50308 + }, + { + "epoch": 18.58, + "learning_rate": 1.1429116886316049e-06, + "loss": 0.2351, + "step": 50309 + }, + { + "epoch": 18.58, + "learning_rate": 1.1428747382682595e-06, + "loss": 0.2032, + "step": 50310 + }, + { + "epoch": 18.58, + "learning_rate": 1.1428377879049144e-06, + "loss": 0.2765, + "step": 50311 + }, + { + "epoch": 18.58, + "learning_rate": 1.1428008375415692e-06, + "loss": 0.2998, + "step": 50312 + }, + { + "epoch": 18.58, + "learning_rate": 1.1427638871782238e-06, + "loss": 0.1915, + "step": 50313 + }, + { + "epoch": 18.58, + "learning_rate": 1.1427269368148787e-06, + "loss": 0.2802, + "step": 50314 + }, + { + "epoch": 18.58, + "learning_rate": 1.1426899864515335e-06, + "loss": 0.1427, + "step": 50315 + }, + { + "epoch": 18.58, + "learning_rate": 1.1426530360881882e-06, + "loss": 0.3186, + "step": 50316 + }, + { + "epoch": 18.58, + "learning_rate": 1.142616085724843e-06, + "loss": 0.1776, + "step": 50317 + }, + { + "epoch": 18.58, + "learning_rate": 1.1425791353614979e-06, + "loss": 0.2084, + "step": 50318 + }, + { + "epoch": 18.58, + "learning_rate": 1.1425421849981525e-06, + "loss": 0.1292, + "step": 50319 + }, + { + "epoch": 18.58, + "learning_rate": 1.1425052346348071e-06, + "loss": 0.2267, + "step": 50320 + }, + { + "epoch": 18.58, + "learning_rate": 1.142468284271462e-06, + "loss": 0.4192, + "step": 50321 + }, + { + "epoch": 18.58, + "learning_rate": 1.1424313339081168e-06, + "loss": 0.2111, + "step": 50322 + }, + { + "epoch": 18.58, + "learning_rate": 1.1423943835447715e-06, + "loss": 0.1239, + "step": 50323 + }, + { + "epoch": 18.58, + "learning_rate": 1.1423574331814263e-06, + "loss": 0.25, + "step": 50324 + }, + { + "epoch": 18.58, + "learning_rate": 1.1423204828180812e-06, + "loss": 0.3763, + "step": 50325 + }, + { + "epoch": 18.58, + "learning_rate": 1.1422835324547358e-06, + "loss": 0.1831, + "step": 50326 + }, + { + "epoch": 18.58, + "learning_rate": 1.1422465820913907e-06, + "loss": 0.1895, + "step": 50327 + }, + { + "epoch": 18.58, + "learning_rate": 1.1422096317280455e-06, + "loss": 0.2497, + "step": 50328 + }, + { + "epoch": 18.59, + "learning_rate": 1.1421726813647001e-06, + "loss": 0.2352, + "step": 50329 + }, + { + "epoch": 18.59, + "learning_rate": 1.1421357310013548e-06, + "loss": 0.4883, + "step": 50330 + }, + { + "epoch": 18.59, + "learning_rate": 1.1420987806380096e-06, + "loss": 0.3353, + "step": 50331 + }, + { + "epoch": 18.59, + "learning_rate": 1.1420618302746645e-06, + "loss": 0.1779, + "step": 50332 + }, + { + "epoch": 18.59, + "learning_rate": 1.1420248799113191e-06, + "loss": 0.2562, + "step": 50333 + }, + { + "epoch": 18.59, + "learning_rate": 1.141987929547974e-06, + "loss": 0.2372, + "step": 50334 + }, + { + "epoch": 18.59, + "learning_rate": 1.1419509791846288e-06, + "loss": 0.161, + "step": 50335 + }, + { + "epoch": 18.59, + "learning_rate": 1.1419140288212834e-06, + "loss": 0.1718, + "step": 50336 + }, + { + "epoch": 18.59, + "learning_rate": 1.1418770784579383e-06, + "loss": 0.1707, + "step": 50337 + }, + { + "epoch": 18.59, + "learning_rate": 1.141840128094593e-06, + "loss": 0.194, + "step": 50338 + }, + { + "epoch": 18.59, + "learning_rate": 1.1418031777312476e-06, + "loss": 0.1617, + "step": 50339 + }, + { + "epoch": 18.59, + "learning_rate": 1.1417662273679024e-06, + "loss": 0.2696, + "step": 50340 + }, + { + "epoch": 18.59, + "learning_rate": 1.1417292770045573e-06, + "loss": 0.1111, + "step": 50341 + }, + { + "epoch": 18.59, + "learning_rate": 1.141692326641212e-06, + "loss": 0.1816, + "step": 50342 + }, + { + "epoch": 18.59, + "learning_rate": 1.1416553762778668e-06, + "loss": 0.2262, + "step": 50343 + }, + { + "epoch": 18.59, + "learning_rate": 1.1416184259145216e-06, + "loss": 0.2441, + "step": 50344 + }, + { + "epoch": 18.59, + "learning_rate": 1.1415814755511762e-06, + "loss": 0.1499, + "step": 50345 + }, + { + "epoch": 18.59, + "learning_rate": 1.141544525187831e-06, + "loss": 0.2061, + "step": 50346 + }, + { + "epoch": 18.59, + "learning_rate": 1.141507574824486e-06, + "loss": 0.1875, + "step": 50347 + }, + { + "epoch": 18.59, + "learning_rate": 1.1414706244611406e-06, + "loss": 0.2576, + "step": 50348 + }, + { + "epoch": 18.59, + "learning_rate": 1.1414336740977952e-06, + "loss": 0.1853, + "step": 50349 + }, + { + "epoch": 18.59, + "learning_rate": 1.14139672373445e-06, + "loss": 0.2061, + "step": 50350 + }, + { + "epoch": 18.59, + "learning_rate": 1.141359773371105e-06, + "loss": 0.263, + "step": 50351 + }, + { + "epoch": 18.59, + "learning_rate": 1.1413228230077595e-06, + "loss": 0.2841, + "step": 50352 + }, + { + "epoch": 18.59, + "learning_rate": 1.1412858726444144e-06, + "loss": 0.2799, + "step": 50353 + }, + { + "epoch": 18.59, + "learning_rate": 1.1412489222810692e-06, + "loss": 0.1739, + "step": 50354 + }, + { + "epoch": 18.59, + "learning_rate": 1.1412119719177239e-06, + "loss": 0.0928, + "step": 50355 + }, + { + "epoch": 18.6, + "learning_rate": 1.1411750215543787e-06, + "loss": 0.1902, + "step": 50356 + }, + { + "epoch": 18.6, + "learning_rate": 1.1411380711910336e-06, + "loss": 0.2429, + "step": 50357 + }, + { + "epoch": 18.6, + "learning_rate": 1.1411011208276882e-06, + "loss": 0.1373, + "step": 50358 + }, + { + "epoch": 18.6, + "learning_rate": 1.1410641704643428e-06, + "loss": 0.3499, + "step": 50359 + }, + { + "epoch": 18.6, + "learning_rate": 1.1410272201009977e-06, + "loss": 0.4284, + "step": 50360 + }, + { + "epoch": 18.6, + "learning_rate": 1.1409902697376523e-06, + "loss": 0.1918, + "step": 50361 + }, + { + "epoch": 18.6, + "learning_rate": 1.1409533193743072e-06, + "loss": 0.2834, + "step": 50362 + }, + { + "epoch": 18.6, + "learning_rate": 1.140916369010962e-06, + "loss": 0.1998, + "step": 50363 + }, + { + "epoch": 18.6, + "learning_rate": 1.1408794186476167e-06, + "loss": 0.1868, + "step": 50364 + }, + { + "epoch": 18.6, + "learning_rate": 1.1408424682842715e-06, + "loss": 0.2742, + "step": 50365 + }, + { + "epoch": 18.6, + "learning_rate": 1.1408055179209264e-06, + "loss": 0.3156, + "step": 50366 + }, + { + "epoch": 18.6, + "learning_rate": 1.140768567557581e-06, + "loss": 0.2175, + "step": 50367 + }, + { + "epoch": 18.6, + "learning_rate": 1.1407316171942358e-06, + "loss": 0.1815, + "step": 50368 + }, + { + "epoch": 18.6, + "learning_rate": 1.1406946668308905e-06, + "loss": 0.2916, + "step": 50369 + }, + { + "epoch": 18.6, + "learning_rate": 1.1406577164675453e-06, + "loss": 0.2489, + "step": 50370 + }, + { + "epoch": 18.6, + "learning_rate": 1.1406207661042e-06, + "loss": 0.1337, + "step": 50371 + }, + { + "epoch": 18.6, + "learning_rate": 1.1405838157408548e-06, + "loss": 0.2645, + "step": 50372 + }, + { + "epoch": 18.6, + "learning_rate": 1.1405468653775097e-06, + "loss": 0.176, + "step": 50373 + }, + { + "epoch": 18.6, + "learning_rate": 1.1405099150141643e-06, + "loss": 0.2893, + "step": 50374 + }, + { + "epoch": 18.6, + "learning_rate": 1.1404729646508191e-06, + "loss": 0.1876, + "step": 50375 + }, + { + "epoch": 18.6, + "learning_rate": 1.140436014287474e-06, + "loss": 0.13, + "step": 50376 + }, + { + "epoch": 18.6, + "learning_rate": 1.1403990639241286e-06, + "loss": 0.146, + "step": 50377 + }, + { + "epoch": 18.6, + "learning_rate": 1.1403621135607835e-06, + "loss": 0.267, + "step": 50378 + }, + { + "epoch": 18.6, + "learning_rate": 1.1403251631974381e-06, + "loss": 0.2278, + "step": 50379 + }, + { + "epoch": 18.6, + "learning_rate": 1.1402882128340927e-06, + "loss": 0.2254, + "step": 50380 + }, + { + "epoch": 18.6, + "learning_rate": 1.1402512624707476e-06, + "loss": 0.3599, + "step": 50381 + }, + { + "epoch": 18.6, + "learning_rate": 1.1402143121074024e-06, + "loss": 0.1898, + "step": 50382 + }, + { + "epoch": 18.61, + "learning_rate": 1.140177361744057e-06, + "loss": 0.2624, + "step": 50383 + }, + { + "epoch": 18.61, + "learning_rate": 1.140140411380712e-06, + "loss": 0.2647, + "step": 50384 + }, + { + "epoch": 18.61, + "learning_rate": 1.1401034610173668e-06, + "loss": 0.1641, + "step": 50385 + }, + { + "epoch": 18.61, + "learning_rate": 1.1400665106540214e-06, + "loss": 0.2605, + "step": 50386 + }, + { + "epoch": 18.61, + "learning_rate": 1.1400295602906763e-06, + "loss": 0.2119, + "step": 50387 + }, + { + "epoch": 18.61, + "learning_rate": 1.139992609927331e-06, + "loss": 0.2669, + "step": 50388 + }, + { + "epoch": 18.61, + "learning_rate": 1.1399556595639857e-06, + "loss": 0.1656, + "step": 50389 + }, + { + "epoch": 18.61, + "learning_rate": 1.1399187092006404e-06, + "loss": 0.1239, + "step": 50390 + }, + { + "epoch": 18.61, + "learning_rate": 1.1398817588372952e-06, + "loss": 0.1849, + "step": 50391 + }, + { + "epoch": 18.61, + "learning_rate": 1.13984480847395e-06, + "loss": 0.3002, + "step": 50392 + }, + { + "epoch": 18.61, + "learning_rate": 1.1398078581106047e-06, + "loss": 0.3285, + "step": 50393 + }, + { + "epoch": 18.61, + "learning_rate": 1.1397709077472596e-06, + "loss": 0.3287, + "step": 50394 + }, + { + "epoch": 18.61, + "learning_rate": 1.1397339573839144e-06, + "loss": 0.1968, + "step": 50395 + }, + { + "epoch": 18.61, + "learning_rate": 1.139697007020569e-06, + "loss": 0.1708, + "step": 50396 + }, + { + "epoch": 18.61, + "learning_rate": 1.139660056657224e-06, + "loss": 0.241, + "step": 50397 + }, + { + "epoch": 18.61, + "learning_rate": 1.1396231062938785e-06, + "loss": 0.1039, + "step": 50398 + }, + { + "epoch": 18.61, + "learning_rate": 1.1395861559305334e-06, + "loss": 0.3409, + "step": 50399 + }, + { + "epoch": 18.61, + "learning_rate": 1.139549205567188e-06, + "loss": 0.1527, + "step": 50400 + }, + { + "epoch": 18.61, + "learning_rate": 1.1395122552038429e-06, + "loss": 0.1463, + "step": 50401 + }, + { + "epoch": 18.61, + "learning_rate": 1.1394753048404977e-06, + "loss": 0.2206, + "step": 50402 + }, + { + "epoch": 18.61, + "learning_rate": 1.1394383544771523e-06, + "loss": 0.1476, + "step": 50403 + }, + { + "epoch": 18.61, + "learning_rate": 1.1394014041138072e-06, + "loss": 0.1031, + "step": 50404 + }, + { + "epoch": 18.61, + "learning_rate": 1.139364453750462e-06, + "loss": 0.4054, + "step": 50405 + }, + { + "epoch": 18.61, + "learning_rate": 1.1393275033871167e-06, + "loss": 0.2423, + "step": 50406 + }, + { + "epoch": 18.61, + "learning_rate": 1.1392905530237715e-06, + "loss": 0.1407, + "step": 50407 + }, + { + "epoch": 18.61, + "learning_rate": 1.1392536026604262e-06, + "loss": 0.3055, + "step": 50408 + }, + { + "epoch": 18.61, + "learning_rate": 1.1392166522970808e-06, + "loss": 0.1854, + "step": 50409 + }, + { + "epoch": 18.62, + "learning_rate": 1.1391797019337357e-06, + "loss": 0.3137, + "step": 50410 + }, + { + "epoch": 18.62, + "learning_rate": 1.1391427515703905e-06, + "loss": 0.1965, + "step": 50411 + }, + { + "epoch": 18.62, + "learning_rate": 1.1391058012070451e-06, + "loss": 0.2168, + "step": 50412 + }, + { + "epoch": 18.62, + "learning_rate": 1.1390688508437e-06, + "loss": 0.2857, + "step": 50413 + }, + { + "epoch": 18.62, + "learning_rate": 1.1390319004803548e-06, + "loss": 0.2132, + "step": 50414 + }, + { + "epoch": 18.62, + "learning_rate": 1.1389949501170095e-06, + "loss": 0.2399, + "step": 50415 + }, + { + "epoch": 18.62, + "learning_rate": 1.1389579997536643e-06, + "loss": 0.2921, + "step": 50416 + }, + { + "epoch": 18.62, + "learning_rate": 1.1389210493903192e-06, + "loss": 0.2077, + "step": 50417 + }, + { + "epoch": 18.62, + "learning_rate": 1.1388840990269738e-06, + "loss": 0.2745, + "step": 50418 + }, + { + "epoch": 18.62, + "learning_rate": 1.1388471486636284e-06, + "loss": 0.1039, + "step": 50419 + }, + { + "epoch": 18.62, + "learning_rate": 1.1388101983002833e-06, + "loss": 0.1889, + "step": 50420 + }, + { + "epoch": 18.62, + "learning_rate": 1.1387732479369381e-06, + "loss": 0.2079, + "step": 50421 + }, + { + "epoch": 18.62, + "learning_rate": 1.1387362975735928e-06, + "loss": 0.2153, + "step": 50422 + }, + { + "epoch": 18.62, + "learning_rate": 1.1386993472102476e-06, + "loss": 0.2267, + "step": 50423 + }, + { + "epoch": 18.62, + "learning_rate": 1.1386623968469025e-06, + "loss": 0.19, + "step": 50424 + }, + { + "epoch": 18.62, + "learning_rate": 1.138625446483557e-06, + "loss": 0.0977, + "step": 50425 + }, + { + "epoch": 18.62, + "learning_rate": 1.138588496120212e-06, + "loss": 0.1307, + "step": 50426 + }, + { + "epoch": 18.62, + "learning_rate": 1.1385515457568668e-06, + "loss": 0.3103, + "step": 50427 + }, + { + "epoch": 18.62, + "learning_rate": 1.1385145953935212e-06, + "loss": 0.2154, + "step": 50428 + }, + { + "epoch": 18.62, + "learning_rate": 1.138477645030176e-06, + "loss": 0.3139, + "step": 50429 + }, + { + "epoch": 18.62, + "learning_rate": 1.138440694666831e-06, + "loss": 0.2054, + "step": 50430 + }, + { + "epoch": 18.62, + "learning_rate": 1.1384037443034856e-06, + "loss": 0.2232, + "step": 50431 + }, + { + "epoch": 18.62, + "learning_rate": 1.1383667939401404e-06, + "loss": 0.1862, + "step": 50432 + }, + { + "epoch": 18.62, + "learning_rate": 1.1383298435767953e-06, + "loss": 0.1764, + "step": 50433 + }, + { + "epoch": 18.62, + "learning_rate": 1.1382928932134499e-06, + "loss": 0.218, + "step": 50434 + }, + { + "epoch": 18.62, + "learning_rate": 1.1382559428501047e-06, + "loss": 0.1449, + "step": 50435 + }, + { + "epoch": 18.62, + "learning_rate": 1.1382189924867596e-06, + "loss": 0.1977, + "step": 50436 + }, + { + "epoch": 18.63, + "learning_rate": 1.1381820421234142e-06, + "loss": 0.2254, + "step": 50437 + }, + { + "epoch": 18.63, + "learning_rate": 1.1381450917600689e-06, + "loss": 0.3749, + "step": 50438 + }, + { + "epoch": 18.63, + "learning_rate": 1.1381081413967237e-06, + "loss": 0.2293, + "step": 50439 + }, + { + "epoch": 18.63, + "learning_rate": 1.1380711910333786e-06, + "loss": 0.2231, + "step": 50440 + }, + { + "epoch": 18.63, + "learning_rate": 1.1380342406700332e-06, + "loss": 0.1596, + "step": 50441 + }, + { + "epoch": 18.63, + "learning_rate": 1.137997290306688e-06, + "loss": 0.2457, + "step": 50442 + }, + { + "epoch": 18.63, + "learning_rate": 1.1379603399433429e-06, + "loss": 0.269, + "step": 50443 + }, + { + "epoch": 18.63, + "learning_rate": 1.1379233895799975e-06, + "loss": 0.1695, + "step": 50444 + }, + { + "epoch": 18.63, + "learning_rate": 1.1378864392166524e-06, + "loss": 0.154, + "step": 50445 + }, + { + "epoch": 18.63, + "learning_rate": 1.1378494888533072e-06, + "loss": 0.2077, + "step": 50446 + }, + { + "epoch": 18.63, + "learning_rate": 1.1378125384899619e-06, + "loss": 0.113, + "step": 50447 + }, + { + "epoch": 18.63, + "learning_rate": 1.1377755881266165e-06, + "loss": 0.4366, + "step": 50448 + }, + { + "epoch": 18.63, + "learning_rate": 1.1377386377632713e-06, + "loss": 0.1767, + "step": 50449 + }, + { + "epoch": 18.63, + "learning_rate": 1.137701687399926e-06, + "loss": 0.172, + "step": 50450 + }, + { + "epoch": 18.63, + "learning_rate": 1.1376647370365808e-06, + "loss": 0.135, + "step": 50451 + }, + { + "epoch": 18.63, + "learning_rate": 1.1376277866732357e-06, + "loss": 0.3099, + "step": 50452 + }, + { + "epoch": 18.63, + "learning_rate": 1.1375908363098903e-06, + "loss": 0.1173, + "step": 50453 + }, + { + "epoch": 18.63, + "learning_rate": 1.1375538859465452e-06, + "loss": 0.2613, + "step": 50454 + }, + { + "epoch": 18.63, + "learning_rate": 1.1375169355832e-06, + "loss": 0.2664, + "step": 50455 + }, + { + "epoch": 18.63, + "learning_rate": 1.1374799852198546e-06, + "loss": 0.1447, + "step": 50456 + }, + { + "epoch": 18.63, + "learning_rate": 1.1374430348565095e-06, + "loss": 0.4147, + "step": 50457 + }, + { + "epoch": 18.63, + "learning_rate": 1.1374060844931641e-06, + "loss": 0.1639, + "step": 50458 + }, + { + "epoch": 18.63, + "learning_rate": 1.137369134129819e-06, + "loss": 0.2167, + "step": 50459 + }, + { + "epoch": 18.63, + "learning_rate": 1.1373321837664736e-06, + "loss": 0.2084, + "step": 50460 + }, + { + "epoch": 18.63, + "learning_rate": 1.1372952334031285e-06, + "loss": 0.2631, + "step": 50461 + }, + { + "epoch": 18.63, + "learning_rate": 1.1372582830397833e-06, + "loss": 0.1731, + "step": 50462 + }, + { + "epoch": 18.63, + "learning_rate": 1.137221332676438e-06, + "loss": 0.3938, + "step": 50463 + }, + { + "epoch": 18.64, + "learning_rate": 1.1371843823130928e-06, + "loss": 0.3042, + "step": 50464 + }, + { + "epoch": 18.64, + "learning_rate": 1.1371474319497476e-06, + "loss": 0.3786, + "step": 50465 + }, + { + "epoch": 18.64, + "learning_rate": 1.1371104815864023e-06, + "loss": 0.1712, + "step": 50466 + }, + { + "epoch": 18.64, + "learning_rate": 1.1370735312230571e-06, + "loss": 0.1995, + "step": 50467 + }, + { + "epoch": 18.64, + "learning_rate": 1.1370365808597118e-06, + "loss": 0.2189, + "step": 50468 + }, + { + "epoch": 18.64, + "learning_rate": 1.1369996304963666e-06, + "loss": 0.2431, + "step": 50469 + }, + { + "epoch": 18.64, + "learning_rate": 1.1369626801330212e-06, + "loss": 0.1276, + "step": 50470 + }, + { + "epoch": 18.64, + "learning_rate": 1.136925729769676e-06, + "loss": 0.2727, + "step": 50471 + }, + { + "epoch": 18.64, + "learning_rate": 1.136888779406331e-06, + "loss": 0.2055, + "step": 50472 + }, + { + "epoch": 18.64, + "learning_rate": 1.1368518290429856e-06, + "loss": 0.2531, + "step": 50473 + }, + { + "epoch": 18.64, + "learning_rate": 1.1368148786796404e-06, + "loss": 0.2143, + "step": 50474 + }, + { + "epoch": 18.64, + "learning_rate": 1.1367779283162953e-06, + "loss": 0.1737, + "step": 50475 + }, + { + "epoch": 18.64, + "learning_rate": 1.13674097795295e-06, + "loss": 0.098, + "step": 50476 + }, + { + "epoch": 18.64, + "learning_rate": 1.1367040275896048e-06, + "loss": 0.2317, + "step": 50477 + }, + { + "epoch": 18.64, + "learning_rate": 1.1366670772262594e-06, + "loss": 0.1678, + "step": 50478 + }, + { + "epoch": 18.64, + "learning_rate": 1.136630126862914e-06, + "loss": 0.3442, + "step": 50479 + }, + { + "epoch": 18.64, + "learning_rate": 1.1365931764995689e-06, + "loss": 0.1858, + "step": 50480 + }, + { + "epoch": 18.64, + "learning_rate": 1.1365562261362237e-06, + "loss": 0.1312, + "step": 50481 + }, + { + "epoch": 18.64, + "learning_rate": 1.1365192757728784e-06, + "loss": 0.1789, + "step": 50482 + }, + { + "epoch": 18.64, + "learning_rate": 1.1364823254095332e-06, + "loss": 0.202, + "step": 50483 + }, + { + "epoch": 18.64, + "learning_rate": 1.136445375046188e-06, + "loss": 0.1685, + "step": 50484 + }, + { + "epoch": 18.64, + "learning_rate": 1.1364084246828427e-06, + "loss": 0.2194, + "step": 50485 + }, + { + "epoch": 18.64, + "learning_rate": 1.1363714743194975e-06, + "loss": 0.1569, + "step": 50486 + }, + { + "epoch": 18.64, + "learning_rate": 1.1363345239561524e-06, + "loss": 0.3756, + "step": 50487 + }, + { + "epoch": 18.64, + "learning_rate": 1.136297573592807e-06, + "loss": 0.1632, + "step": 50488 + }, + { + "epoch": 18.64, + "learning_rate": 1.1362606232294617e-06, + "loss": 0.1531, + "step": 50489 + }, + { + "epoch": 18.64, + "learning_rate": 1.1362236728661165e-06, + "loss": 0.1715, + "step": 50490 + }, + { + "epoch": 18.65, + "learning_rate": 1.1361867225027714e-06, + "loss": 0.2495, + "step": 50491 + }, + { + "epoch": 18.65, + "learning_rate": 1.136149772139426e-06, + "loss": 0.239, + "step": 50492 + }, + { + "epoch": 18.65, + "learning_rate": 1.1361128217760809e-06, + "loss": 0.2039, + "step": 50493 + }, + { + "epoch": 18.65, + "learning_rate": 1.1360758714127357e-06, + "loss": 0.1158, + "step": 50494 + }, + { + "epoch": 18.65, + "learning_rate": 1.1360389210493903e-06, + "loss": 0.2683, + "step": 50495 + }, + { + "epoch": 18.65, + "learning_rate": 1.1360019706860452e-06, + "loss": 0.2114, + "step": 50496 + }, + { + "epoch": 18.65, + "learning_rate": 1.1359650203227e-06, + "loss": 0.2603, + "step": 50497 + }, + { + "epoch": 18.65, + "learning_rate": 1.1359280699593545e-06, + "loss": 0.2228, + "step": 50498 + }, + { + "epoch": 18.65, + "learning_rate": 1.1358911195960093e-06, + "loss": 0.2401, + "step": 50499 + }, + { + "epoch": 18.65, + "learning_rate": 1.1358541692326642e-06, + "loss": 0.2287, + "step": 50500 + }, + { + "epoch": 18.65, + "learning_rate": 1.1358172188693188e-06, + "loss": 0.3609, + "step": 50501 + }, + { + "epoch": 18.65, + "learning_rate": 1.1357802685059736e-06, + "loss": 0.2235, + "step": 50502 + }, + { + "epoch": 18.65, + "learning_rate": 1.1357433181426285e-06, + "loss": 0.1514, + "step": 50503 + }, + { + "epoch": 18.65, + "learning_rate": 1.1357063677792831e-06, + "loss": 0.1381, + "step": 50504 + }, + { + "epoch": 18.65, + "learning_rate": 1.135669417415938e-06, + "loss": 0.1328, + "step": 50505 + }, + { + "epoch": 18.65, + "learning_rate": 1.1356324670525928e-06, + "loss": 0.1998, + "step": 50506 + }, + { + "epoch": 18.65, + "learning_rate": 1.1355955166892475e-06, + "loss": 0.1441, + "step": 50507 + }, + { + "epoch": 18.65, + "learning_rate": 1.135558566325902e-06, + "loss": 0.3726, + "step": 50508 + }, + { + "epoch": 18.65, + "learning_rate": 1.135521615962557e-06, + "loss": 0.2307, + "step": 50509 + }, + { + "epoch": 18.65, + "learning_rate": 1.1354846655992118e-06, + "loss": 0.2284, + "step": 50510 + }, + { + "epoch": 18.65, + "learning_rate": 1.1354477152358664e-06, + "loss": 0.2607, + "step": 50511 + }, + { + "epoch": 18.65, + "learning_rate": 1.1354107648725213e-06, + "loss": 0.2635, + "step": 50512 + }, + { + "epoch": 18.65, + "learning_rate": 1.1353738145091761e-06, + "loss": 0.2455, + "step": 50513 + }, + { + "epoch": 18.65, + "learning_rate": 1.1353368641458308e-06, + "loss": 0.1732, + "step": 50514 + }, + { + "epoch": 18.65, + "learning_rate": 1.1352999137824856e-06, + "loss": 0.2132, + "step": 50515 + }, + { + "epoch": 18.65, + "learning_rate": 1.1352629634191405e-06, + "loss": 0.2741, + "step": 50516 + }, + { + "epoch": 18.65, + "learning_rate": 1.135226013055795e-06, + "loss": 0.2897, + "step": 50517 + }, + { + "epoch": 18.66, + "learning_rate": 1.1351890626924497e-06, + "loss": 0.1624, + "step": 50518 + }, + { + "epoch": 18.66, + "learning_rate": 1.1351521123291046e-06, + "loss": 0.1695, + "step": 50519 + }, + { + "epoch": 18.66, + "learning_rate": 1.1351151619657594e-06, + "loss": 0.191, + "step": 50520 + }, + { + "epoch": 18.66, + "learning_rate": 1.135078211602414e-06, + "loss": 0.2661, + "step": 50521 + }, + { + "epoch": 18.66, + "learning_rate": 1.135041261239069e-06, + "loss": 0.2424, + "step": 50522 + }, + { + "epoch": 18.66, + "learning_rate": 1.1350043108757235e-06, + "loss": 0.2061, + "step": 50523 + }, + { + "epoch": 18.66, + "learning_rate": 1.1349673605123784e-06, + "loss": 0.213, + "step": 50524 + }, + { + "epoch": 18.66, + "learning_rate": 1.1349304101490332e-06, + "loss": 0.3157, + "step": 50525 + }, + { + "epoch": 18.66, + "learning_rate": 1.1348934597856879e-06, + "loss": 0.2079, + "step": 50526 + }, + { + "epoch": 18.66, + "learning_rate": 1.1348565094223427e-06, + "loss": 0.2564, + "step": 50527 + }, + { + "epoch": 18.66, + "learning_rate": 1.1348195590589974e-06, + "loss": 0.1835, + "step": 50528 + }, + { + "epoch": 18.66, + "learning_rate": 1.1347826086956522e-06, + "loss": 0.2879, + "step": 50529 + }, + { + "epoch": 18.66, + "learning_rate": 1.1347456583323068e-06, + "loss": 0.2805, + "step": 50530 + }, + { + "epoch": 18.66, + "learning_rate": 1.1347087079689617e-06, + "loss": 0.1715, + "step": 50531 + }, + { + "epoch": 18.66, + "learning_rate": 1.1346717576056165e-06, + "loss": 0.2175, + "step": 50532 + }, + { + "epoch": 18.66, + "learning_rate": 1.1346348072422712e-06, + "loss": 0.2478, + "step": 50533 + }, + { + "epoch": 18.66, + "learning_rate": 1.134597856878926e-06, + "loss": 0.2407, + "step": 50534 + }, + { + "epoch": 18.66, + "learning_rate": 1.1345609065155809e-06, + "loss": 0.2828, + "step": 50535 + }, + { + "epoch": 18.66, + "learning_rate": 1.1345239561522355e-06, + "loss": 0.2241, + "step": 50536 + }, + { + "epoch": 18.66, + "learning_rate": 1.1344870057888904e-06, + "loss": 0.1384, + "step": 50537 + }, + { + "epoch": 18.66, + "learning_rate": 1.134450055425545e-06, + "loss": 0.3036, + "step": 50538 + }, + { + "epoch": 18.66, + "learning_rate": 1.1344131050621998e-06, + "loss": 0.2228, + "step": 50539 + }, + { + "epoch": 18.66, + "learning_rate": 1.1343761546988545e-06, + "loss": 0.174, + "step": 50540 + }, + { + "epoch": 18.66, + "learning_rate": 1.1343392043355093e-06, + "loss": 0.1505, + "step": 50541 + }, + { + "epoch": 18.66, + "learning_rate": 1.1343022539721642e-06, + "loss": 0.2857, + "step": 50542 + }, + { + "epoch": 18.66, + "learning_rate": 1.1342653036088188e-06, + "loss": 0.2609, + "step": 50543 + }, + { + "epoch": 18.66, + "learning_rate": 1.1342283532454737e-06, + "loss": 0.3377, + "step": 50544 + }, + { + "epoch": 18.67, + "learning_rate": 1.1341914028821285e-06, + "loss": 0.3841, + "step": 50545 + }, + { + "epoch": 18.67, + "learning_rate": 1.1341544525187831e-06, + "loss": 0.2344, + "step": 50546 + }, + { + "epoch": 18.67, + "learning_rate": 1.134117502155438e-06, + "loss": 0.2879, + "step": 50547 + }, + { + "epoch": 18.67, + "learning_rate": 1.1340805517920926e-06, + "loss": 0.2095, + "step": 50548 + }, + { + "epoch": 18.67, + "learning_rate": 1.1340436014287473e-06, + "loss": 0.2024, + "step": 50549 + }, + { + "epoch": 18.67, + "learning_rate": 1.1340066510654021e-06, + "loss": 0.1376, + "step": 50550 + }, + { + "epoch": 18.67, + "learning_rate": 1.133969700702057e-06, + "loss": 0.2235, + "step": 50551 + }, + { + "epoch": 18.67, + "learning_rate": 1.1339327503387116e-06, + "loss": 0.2053, + "step": 50552 + }, + { + "epoch": 18.67, + "learning_rate": 1.1338957999753664e-06, + "loss": 0.2891, + "step": 50553 + }, + { + "epoch": 18.67, + "learning_rate": 1.1338588496120213e-06, + "loss": 0.1879, + "step": 50554 + }, + { + "epoch": 18.67, + "learning_rate": 1.133821899248676e-06, + "loss": 0.1917, + "step": 50555 + }, + { + "epoch": 18.67, + "learning_rate": 1.1337849488853308e-06, + "loss": 0.2855, + "step": 50556 + }, + { + "epoch": 18.67, + "learning_rate": 1.1337479985219856e-06, + "loss": 0.2307, + "step": 50557 + }, + { + "epoch": 18.67, + "learning_rate": 1.1337110481586403e-06, + "loss": 0.2273, + "step": 50558 + }, + { + "epoch": 18.67, + "learning_rate": 1.133674097795295e-06, + "loss": 0.2406, + "step": 50559 + }, + { + "epoch": 18.67, + "learning_rate": 1.1336371474319498e-06, + "loss": 0.2333, + "step": 50560 + }, + { + "epoch": 18.67, + "learning_rate": 1.1336001970686046e-06, + "loss": 0.2113, + "step": 50561 + }, + { + "epoch": 18.67, + "learning_rate": 1.1335632467052592e-06, + "loss": 0.3238, + "step": 50562 + }, + { + "epoch": 18.67, + "learning_rate": 1.133526296341914e-06, + "loss": 0.2924, + "step": 50563 + }, + { + "epoch": 18.67, + "learning_rate": 1.133489345978569e-06, + "loss": 0.1409, + "step": 50564 + }, + { + "epoch": 18.67, + "learning_rate": 1.1334523956152236e-06, + "loss": 0.2338, + "step": 50565 + }, + { + "epoch": 18.67, + "learning_rate": 1.1334154452518784e-06, + "loss": 0.2866, + "step": 50566 + }, + { + "epoch": 18.67, + "learning_rate": 1.1333784948885333e-06, + "loss": 0.1483, + "step": 50567 + }, + { + "epoch": 18.67, + "learning_rate": 1.1333415445251877e-06, + "loss": 0.2348, + "step": 50568 + }, + { + "epoch": 18.67, + "learning_rate": 1.1333045941618425e-06, + "loss": 0.2327, + "step": 50569 + }, + { + "epoch": 18.67, + "learning_rate": 1.1332676437984974e-06, + "loss": 0.2249, + "step": 50570 + }, + { + "epoch": 18.67, + "learning_rate": 1.133230693435152e-06, + "loss": 0.1983, + "step": 50571 + }, + { + "epoch": 18.68, + "learning_rate": 1.1331937430718069e-06, + "loss": 0.2308, + "step": 50572 + }, + { + "epoch": 18.68, + "learning_rate": 1.1331567927084617e-06, + "loss": 0.381, + "step": 50573 + }, + { + "epoch": 18.68, + "learning_rate": 1.1331198423451164e-06, + "loss": 0.2416, + "step": 50574 + }, + { + "epoch": 18.68, + "learning_rate": 1.1330828919817712e-06, + "loss": 0.2569, + "step": 50575 + }, + { + "epoch": 18.68, + "learning_rate": 1.133045941618426e-06, + "loss": 0.1209, + "step": 50576 + }, + { + "epoch": 18.68, + "learning_rate": 1.1330089912550807e-06, + "loss": 0.3512, + "step": 50577 + }, + { + "epoch": 18.68, + "learning_rate": 1.1329720408917353e-06, + "loss": 0.1782, + "step": 50578 + }, + { + "epoch": 18.68, + "learning_rate": 1.1329350905283902e-06, + "loss": 0.291, + "step": 50579 + }, + { + "epoch": 18.68, + "learning_rate": 1.132898140165045e-06, + "loss": 0.1904, + "step": 50580 + }, + { + "epoch": 18.68, + "learning_rate": 1.1328611898016997e-06, + "loss": 0.168, + "step": 50581 + }, + { + "epoch": 18.68, + "learning_rate": 1.1328242394383545e-06, + "loss": 0.1605, + "step": 50582 + }, + { + "epoch": 18.68, + "learning_rate": 1.1327872890750094e-06, + "loss": 0.1877, + "step": 50583 + }, + { + "epoch": 18.68, + "learning_rate": 1.132750338711664e-06, + "loss": 0.186, + "step": 50584 + }, + { + "epoch": 18.68, + "learning_rate": 1.1327133883483188e-06, + "loss": 0.1073, + "step": 50585 + }, + { + "epoch": 18.68, + "learning_rate": 1.1326764379849737e-06, + "loss": 0.335, + "step": 50586 + }, + { + "epoch": 18.68, + "learning_rate": 1.1326394876216283e-06, + "loss": 0.2941, + "step": 50587 + }, + { + "epoch": 18.68, + "learning_rate": 1.132602537258283e-06, + "loss": 0.2977, + "step": 50588 + }, + { + "epoch": 18.68, + "learning_rate": 1.1325655868949378e-06, + "loss": 0.1589, + "step": 50589 + }, + { + "epoch": 18.68, + "learning_rate": 1.1325286365315927e-06, + "loss": 0.2332, + "step": 50590 + }, + { + "epoch": 18.68, + "learning_rate": 1.1324916861682473e-06, + "loss": 0.2731, + "step": 50591 + }, + { + "epoch": 18.68, + "learning_rate": 1.1324547358049021e-06, + "loss": 0.2997, + "step": 50592 + }, + { + "epoch": 18.68, + "learning_rate": 1.132417785441557e-06, + "loss": 0.2286, + "step": 50593 + }, + { + "epoch": 18.68, + "learning_rate": 1.1323808350782116e-06, + "loss": 0.1501, + "step": 50594 + }, + { + "epoch": 18.68, + "learning_rate": 1.1323438847148665e-06, + "loss": 0.2644, + "step": 50595 + }, + { + "epoch": 18.68, + "learning_rate": 1.1323069343515213e-06, + "loss": 0.1447, + "step": 50596 + }, + { + "epoch": 18.68, + "learning_rate": 1.132269983988176e-06, + "loss": 0.2138, + "step": 50597 + }, + { + "epoch": 18.68, + "learning_rate": 1.1322330336248306e-06, + "loss": 0.202, + "step": 50598 + }, + { + "epoch": 18.69, + "learning_rate": 1.1321960832614854e-06, + "loss": 0.2987, + "step": 50599 + }, + { + "epoch": 18.69, + "learning_rate": 1.13215913289814e-06, + "loss": 0.2459, + "step": 50600 + }, + { + "epoch": 18.69, + "learning_rate": 1.132122182534795e-06, + "loss": 0.1665, + "step": 50601 + }, + { + "epoch": 18.69, + "learning_rate": 1.1320852321714498e-06, + "loss": 0.2821, + "step": 50602 + }, + { + "epoch": 18.69, + "learning_rate": 1.1320482818081044e-06, + "loss": 0.1796, + "step": 50603 + }, + { + "epoch": 18.69, + "learning_rate": 1.1320113314447593e-06, + "loss": 0.1646, + "step": 50604 + }, + { + "epoch": 18.69, + "learning_rate": 1.1319743810814141e-06, + "loss": 0.2206, + "step": 50605 + }, + { + "epoch": 18.69, + "learning_rate": 1.1319374307180687e-06, + "loss": 0.2791, + "step": 50606 + }, + { + "epoch": 18.69, + "learning_rate": 1.1319004803547236e-06, + "loss": 0.1768, + "step": 50607 + }, + { + "epoch": 18.69, + "learning_rate": 1.1318635299913782e-06, + "loss": 0.1186, + "step": 50608 + }, + { + "epoch": 18.69, + "learning_rate": 1.131826579628033e-06, + "loss": 0.1613, + "step": 50609 + }, + { + "epoch": 18.69, + "learning_rate": 1.1317896292646877e-06, + "loss": 0.2251, + "step": 50610 + }, + { + "epoch": 18.69, + "learning_rate": 1.1317526789013426e-06, + "loss": 0.2416, + "step": 50611 + }, + { + "epoch": 18.69, + "learning_rate": 1.1317157285379974e-06, + "loss": 0.1677, + "step": 50612 + }, + { + "epoch": 18.69, + "learning_rate": 1.131678778174652e-06, + "loss": 0.1108, + "step": 50613 + }, + { + "epoch": 18.69, + "learning_rate": 1.131641827811307e-06, + "loss": 0.2767, + "step": 50614 + }, + { + "epoch": 18.69, + "learning_rate": 1.1316048774479617e-06, + "loss": 0.1119, + "step": 50615 + }, + { + "epoch": 18.69, + "learning_rate": 1.1315679270846164e-06, + "loss": 0.192, + "step": 50616 + }, + { + "epoch": 18.69, + "learning_rate": 1.1315309767212712e-06, + "loss": 0.2421, + "step": 50617 + }, + { + "epoch": 18.69, + "learning_rate": 1.1314940263579259e-06, + "loss": 0.2487, + "step": 50618 + }, + { + "epoch": 18.69, + "learning_rate": 1.1314570759945805e-06, + "loss": 0.1815, + "step": 50619 + }, + { + "epoch": 18.69, + "learning_rate": 1.1314201256312353e-06, + "loss": 0.2838, + "step": 50620 + }, + { + "epoch": 18.69, + "learning_rate": 1.1313831752678902e-06, + "loss": 0.2192, + "step": 50621 + }, + { + "epoch": 18.69, + "learning_rate": 1.1313462249045448e-06, + "loss": 0.1584, + "step": 50622 + }, + { + "epoch": 18.69, + "learning_rate": 1.1313092745411997e-06, + "loss": 0.1323, + "step": 50623 + }, + { + "epoch": 18.69, + "learning_rate": 1.1312723241778545e-06, + "loss": 0.1981, + "step": 50624 + }, + { + "epoch": 18.69, + "learning_rate": 1.1312353738145092e-06, + "loss": 0.1963, + "step": 50625 + }, + { + "epoch": 18.69, + "learning_rate": 1.131198423451164e-06, + "loss": 0.2364, + "step": 50626 + }, + { + "epoch": 18.7, + "learning_rate": 1.1311614730878187e-06, + "loss": 0.2503, + "step": 50627 + }, + { + "epoch": 18.7, + "learning_rate": 1.1311245227244735e-06, + "loss": 0.3283, + "step": 50628 + }, + { + "epoch": 18.7, + "learning_rate": 1.1310875723611281e-06, + "loss": 0.2215, + "step": 50629 + }, + { + "epoch": 18.7, + "learning_rate": 1.131050621997783e-06, + "loss": 0.1846, + "step": 50630 + }, + { + "epoch": 18.7, + "learning_rate": 1.1310136716344378e-06, + "loss": 0.2673, + "step": 50631 + }, + { + "epoch": 18.7, + "learning_rate": 1.1309767212710925e-06, + "loss": 0.3047, + "step": 50632 + }, + { + "epoch": 18.7, + "learning_rate": 1.1309397709077473e-06, + "loss": 0.1999, + "step": 50633 + }, + { + "epoch": 18.7, + "learning_rate": 1.1309028205444022e-06, + "loss": 0.2144, + "step": 50634 + }, + { + "epoch": 18.7, + "learning_rate": 1.1308658701810568e-06, + "loss": 0.1783, + "step": 50635 + }, + { + "epoch": 18.7, + "learning_rate": 1.1308289198177116e-06, + "loss": 0.2739, + "step": 50636 + }, + { + "epoch": 18.7, + "learning_rate": 1.1307919694543663e-06, + "loss": 0.1823, + "step": 50637 + }, + { + "epoch": 18.7, + "learning_rate": 1.130755019091021e-06, + "loss": 0.1883, + "step": 50638 + }, + { + "epoch": 18.7, + "learning_rate": 1.1307180687276758e-06, + "loss": 0.2852, + "step": 50639 + }, + { + "epoch": 18.7, + "learning_rate": 1.1306811183643306e-06, + "loss": 0.3135, + "step": 50640 + }, + { + "epoch": 18.7, + "learning_rate": 1.1306441680009853e-06, + "loss": 0.2682, + "step": 50641 + }, + { + "epoch": 18.7, + "learning_rate": 1.13060721763764e-06, + "loss": 0.2349, + "step": 50642 + }, + { + "epoch": 18.7, + "learning_rate": 1.130570267274295e-06, + "loss": 0.1763, + "step": 50643 + }, + { + "epoch": 18.7, + "learning_rate": 1.1305333169109496e-06, + "loss": 0.2154, + "step": 50644 + }, + { + "epoch": 18.7, + "learning_rate": 1.1304963665476044e-06, + "loss": 0.2375, + "step": 50645 + }, + { + "epoch": 18.7, + "learning_rate": 1.1304594161842593e-06, + "loss": 0.2207, + "step": 50646 + }, + { + "epoch": 18.7, + "learning_rate": 1.130422465820914e-06, + "loss": 0.3418, + "step": 50647 + }, + { + "epoch": 18.7, + "learning_rate": 1.1303855154575686e-06, + "loss": 0.2279, + "step": 50648 + }, + { + "epoch": 18.7, + "learning_rate": 1.1303485650942234e-06, + "loss": 0.1748, + "step": 50649 + }, + { + "epoch": 18.7, + "learning_rate": 1.1303116147308783e-06, + "loss": 0.2243, + "step": 50650 + }, + { + "epoch": 18.7, + "learning_rate": 1.1302746643675329e-06, + "loss": 0.2047, + "step": 50651 + }, + { + "epoch": 18.7, + "learning_rate": 1.1302377140041877e-06, + "loss": 0.2477, + "step": 50652 + }, + { + "epoch": 18.7, + "learning_rate": 1.1302007636408426e-06, + "loss": 0.2778, + "step": 50653 + }, + { + "epoch": 18.71, + "learning_rate": 1.1301638132774972e-06, + "loss": 0.1574, + "step": 50654 + }, + { + "epoch": 18.71, + "learning_rate": 1.130126862914152e-06, + "loss": 0.1631, + "step": 50655 + }, + { + "epoch": 18.71, + "learning_rate": 1.130089912550807e-06, + "loss": 0.1797, + "step": 50656 + }, + { + "epoch": 18.71, + "learning_rate": 1.1300529621874616e-06, + "loss": 0.2605, + "step": 50657 + }, + { + "epoch": 18.71, + "learning_rate": 1.1300160118241162e-06, + "loss": 0.3223, + "step": 50658 + }, + { + "epoch": 18.71, + "learning_rate": 1.129979061460771e-06, + "loss": 0.2308, + "step": 50659 + }, + { + "epoch": 18.71, + "learning_rate": 1.1299421110974259e-06, + "loss": 0.2025, + "step": 50660 + }, + { + "epoch": 18.71, + "learning_rate": 1.1299051607340805e-06, + "loss": 0.2302, + "step": 50661 + }, + { + "epoch": 18.71, + "learning_rate": 1.1298682103707354e-06, + "loss": 0.2582, + "step": 50662 + }, + { + "epoch": 18.71, + "learning_rate": 1.1298312600073902e-06, + "loss": 0.1612, + "step": 50663 + }, + { + "epoch": 18.71, + "learning_rate": 1.1297943096440449e-06, + "loss": 0.2094, + "step": 50664 + }, + { + "epoch": 18.71, + "learning_rate": 1.1297573592806997e-06, + "loss": 0.2337, + "step": 50665 + }, + { + "epoch": 18.71, + "learning_rate": 1.1297204089173546e-06, + "loss": 0.129, + "step": 50666 + }, + { + "epoch": 18.71, + "learning_rate": 1.129683458554009e-06, + "loss": 0.1962, + "step": 50667 + }, + { + "epoch": 18.71, + "learning_rate": 1.1296465081906638e-06, + "loss": 0.2029, + "step": 50668 + }, + { + "epoch": 18.71, + "learning_rate": 1.1296095578273187e-06, + "loss": 0.1454, + "step": 50669 + }, + { + "epoch": 18.71, + "learning_rate": 1.1295726074639733e-06, + "loss": 0.1385, + "step": 50670 + }, + { + "epoch": 18.71, + "learning_rate": 1.1295356571006282e-06, + "loss": 0.1314, + "step": 50671 + }, + { + "epoch": 18.71, + "learning_rate": 1.129498706737283e-06, + "loss": 0.4145, + "step": 50672 + }, + { + "epoch": 18.71, + "learning_rate": 1.1294617563739376e-06, + "loss": 0.132, + "step": 50673 + }, + { + "epoch": 18.71, + "learning_rate": 1.1294248060105925e-06, + "loss": 0.2145, + "step": 50674 + }, + { + "epoch": 18.71, + "learning_rate": 1.1293878556472473e-06, + "loss": 0.3451, + "step": 50675 + }, + { + "epoch": 18.71, + "learning_rate": 1.129350905283902e-06, + "loss": 0.2608, + "step": 50676 + }, + { + "epoch": 18.71, + "learning_rate": 1.1293139549205566e-06, + "loss": 0.2526, + "step": 50677 + }, + { + "epoch": 18.71, + "learning_rate": 1.1292770045572115e-06, + "loss": 0.155, + "step": 50678 + }, + { + "epoch": 18.71, + "learning_rate": 1.1292400541938663e-06, + "loss": 0.234, + "step": 50679 + }, + { + "epoch": 18.71, + "learning_rate": 1.129203103830521e-06, + "loss": 0.2178, + "step": 50680 + }, + { + "epoch": 18.72, + "learning_rate": 1.1291661534671758e-06, + "loss": 0.2955, + "step": 50681 + }, + { + "epoch": 18.72, + "learning_rate": 1.1291292031038306e-06, + "loss": 0.3235, + "step": 50682 + }, + { + "epoch": 18.72, + "learning_rate": 1.1290922527404853e-06, + "loss": 0.355, + "step": 50683 + }, + { + "epoch": 18.72, + "learning_rate": 1.1290553023771401e-06, + "loss": 0.2095, + "step": 50684 + }, + { + "epoch": 18.72, + "learning_rate": 1.129018352013795e-06, + "loss": 0.2405, + "step": 50685 + }, + { + "epoch": 18.72, + "learning_rate": 1.1289814016504496e-06, + "loss": 0.1242, + "step": 50686 + }, + { + "epoch": 18.72, + "learning_rate": 1.1289444512871042e-06, + "loss": 0.1888, + "step": 50687 + }, + { + "epoch": 18.72, + "learning_rate": 1.128907500923759e-06, + "loss": 0.19, + "step": 50688 + }, + { + "epoch": 18.72, + "learning_rate": 1.1288705505604137e-06, + "loss": 0.1848, + "step": 50689 + }, + { + "epoch": 18.72, + "learning_rate": 1.1288336001970686e-06, + "loss": 0.1485, + "step": 50690 + }, + { + "epoch": 18.72, + "learning_rate": 1.1287966498337234e-06, + "loss": 0.3304, + "step": 50691 + }, + { + "epoch": 18.72, + "learning_rate": 1.128759699470378e-06, + "loss": 0.2285, + "step": 50692 + }, + { + "epoch": 18.72, + "learning_rate": 1.128722749107033e-06, + "loss": 0.1675, + "step": 50693 + }, + { + "epoch": 18.72, + "learning_rate": 1.1286857987436878e-06, + "loss": 0.2998, + "step": 50694 + }, + { + "epoch": 18.72, + "learning_rate": 1.1286488483803424e-06, + "loss": 0.19, + "step": 50695 + }, + { + "epoch": 18.72, + "learning_rate": 1.1286118980169972e-06, + "loss": 0.2323, + "step": 50696 + }, + { + "epoch": 18.72, + "learning_rate": 1.1285749476536519e-06, + "loss": 0.2812, + "step": 50697 + }, + { + "epoch": 18.72, + "learning_rate": 1.1285379972903067e-06, + "loss": 0.2464, + "step": 50698 + }, + { + "epoch": 18.72, + "learning_rate": 1.1285010469269614e-06, + "loss": 0.211, + "step": 50699 + }, + { + "epoch": 18.72, + "learning_rate": 1.1284640965636162e-06, + "loss": 0.1694, + "step": 50700 + }, + { + "epoch": 18.72, + "learning_rate": 1.128427146200271e-06, + "loss": 0.3907, + "step": 50701 + }, + { + "epoch": 18.72, + "learning_rate": 1.1283901958369257e-06, + "loss": 0.2071, + "step": 50702 + }, + { + "epoch": 18.72, + "learning_rate": 1.1283532454735805e-06, + "loss": 0.1701, + "step": 50703 + }, + { + "epoch": 18.72, + "learning_rate": 1.1283162951102354e-06, + "loss": 0.1956, + "step": 50704 + }, + { + "epoch": 18.72, + "learning_rate": 1.12827934474689e-06, + "loss": 0.1214, + "step": 50705 + }, + { + "epoch": 18.72, + "learning_rate": 1.1282423943835449e-06, + "loss": 0.122, + "step": 50706 + }, + { + "epoch": 18.72, + "learning_rate": 1.1282054440201995e-06, + "loss": 0.3335, + "step": 50707 + }, + { + "epoch": 18.73, + "learning_rate": 1.1281684936568544e-06, + "loss": 0.218, + "step": 50708 + }, + { + "epoch": 18.73, + "learning_rate": 1.128131543293509e-06, + "loss": 0.234, + "step": 50709 + }, + { + "epoch": 18.73, + "learning_rate": 1.1280945929301639e-06, + "loss": 0.3565, + "step": 50710 + }, + { + "epoch": 18.73, + "learning_rate": 1.1280576425668185e-06, + "loss": 0.1414, + "step": 50711 + }, + { + "epoch": 18.73, + "learning_rate": 1.1280206922034733e-06, + "loss": 0.1531, + "step": 50712 + }, + { + "epoch": 18.73, + "learning_rate": 1.1279837418401282e-06, + "loss": 0.1703, + "step": 50713 + }, + { + "epoch": 18.73, + "learning_rate": 1.1279467914767828e-06, + "loss": 0.2235, + "step": 50714 + }, + { + "epoch": 18.73, + "learning_rate": 1.1279098411134377e-06, + "loss": 0.2655, + "step": 50715 + }, + { + "epoch": 18.73, + "learning_rate": 1.1278728907500925e-06, + "loss": 0.1798, + "step": 50716 + }, + { + "epoch": 18.73, + "learning_rate": 1.1278359403867472e-06, + "loss": 0.2927, + "step": 50717 + }, + { + "epoch": 18.73, + "learning_rate": 1.1277989900234018e-06, + "loss": 0.2228, + "step": 50718 + }, + { + "epoch": 18.73, + "learning_rate": 1.1277620396600566e-06, + "loss": 0.3394, + "step": 50719 + }, + { + "epoch": 18.73, + "learning_rate": 1.1277250892967115e-06, + "loss": 0.2908, + "step": 50720 + }, + { + "epoch": 18.73, + "learning_rate": 1.1276881389333661e-06, + "loss": 0.1824, + "step": 50721 + }, + { + "epoch": 18.73, + "learning_rate": 1.127651188570021e-06, + "loss": 0.25, + "step": 50722 + }, + { + "epoch": 18.73, + "learning_rate": 1.1276142382066758e-06, + "loss": 0.3578, + "step": 50723 + }, + { + "epoch": 18.73, + "learning_rate": 1.1275772878433305e-06, + "loss": 0.2378, + "step": 50724 + }, + { + "epoch": 18.73, + "learning_rate": 1.1275403374799853e-06, + "loss": 0.2206, + "step": 50725 + }, + { + "epoch": 18.73, + "learning_rate": 1.1275033871166402e-06, + "loss": 0.2698, + "step": 50726 + }, + { + "epoch": 18.73, + "learning_rate": 1.1274664367532948e-06, + "loss": 0.1908, + "step": 50727 + }, + { + "epoch": 18.73, + "learning_rate": 1.1274294863899494e-06, + "loss": 0.1147, + "step": 50728 + }, + { + "epoch": 18.73, + "learning_rate": 1.1273925360266043e-06, + "loss": 0.2123, + "step": 50729 + }, + { + "epoch": 18.73, + "learning_rate": 1.1273555856632591e-06, + "loss": 0.291, + "step": 50730 + }, + { + "epoch": 18.73, + "learning_rate": 1.1273186352999138e-06, + "loss": 0.1187, + "step": 50731 + }, + { + "epoch": 18.73, + "learning_rate": 1.1272816849365686e-06, + "loss": 0.143, + "step": 50732 + }, + { + "epoch": 18.73, + "learning_rate": 1.1272447345732235e-06, + "loss": 0.1779, + "step": 50733 + }, + { + "epoch": 18.73, + "learning_rate": 1.127207784209878e-06, + "loss": 0.1659, + "step": 50734 + }, + { + "epoch": 18.74, + "learning_rate": 1.127170833846533e-06, + "loss": 0.2317, + "step": 50735 + }, + { + "epoch": 18.74, + "learning_rate": 1.1271338834831878e-06, + "loss": 0.2252, + "step": 50736 + }, + { + "epoch": 18.74, + "learning_rate": 1.1270969331198422e-06, + "loss": 0.3003, + "step": 50737 + }, + { + "epoch": 18.74, + "learning_rate": 1.127059982756497e-06, + "loss": 0.177, + "step": 50738 + }, + { + "epoch": 18.74, + "learning_rate": 1.127023032393152e-06, + "loss": 0.3285, + "step": 50739 + }, + { + "epoch": 18.74, + "learning_rate": 1.1269860820298065e-06, + "loss": 0.2033, + "step": 50740 + }, + { + "epoch": 18.74, + "learning_rate": 1.1269491316664614e-06, + "loss": 0.249, + "step": 50741 + }, + { + "epoch": 18.74, + "learning_rate": 1.1269121813031162e-06, + "loss": 0.2909, + "step": 50742 + }, + { + "epoch": 18.74, + "learning_rate": 1.1268752309397709e-06, + "loss": 0.191, + "step": 50743 + }, + { + "epoch": 18.74, + "learning_rate": 1.1268382805764257e-06, + "loss": 0.1469, + "step": 50744 + }, + { + "epoch": 18.74, + "learning_rate": 1.1268013302130806e-06, + "loss": 0.2691, + "step": 50745 + }, + { + "epoch": 18.74, + "learning_rate": 1.1267643798497352e-06, + "loss": 0.1696, + "step": 50746 + }, + { + "epoch": 18.74, + "learning_rate": 1.1267274294863898e-06, + "loss": 0.2309, + "step": 50747 + }, + { + "epoch": 18.74, + "learning_rate": 1.1266904791230447e-06, + "loss": 0.3466, + "step": 50748 + }, + { + "epoch": 18.74, + "learning_rate": 1.1266535287596995e-06, + "loss": 0.162, + "step": 50749 + }, + { + "epoch": 18.74, + "learning_rate": 1.1266165783963542e-06, + "loss": 0.2254, + "step": 50750 + }, + { + "epoch": 18.74, + "learning_rate": 1.126579628033009e-06, + "loss": 0.3355, + "step": 50751 + }, + { + "epoch": 18.74, + "learning_rate": 1.1265426776696639e-06, + "loss": 0.26, + "step": 50752 + }, + { + "epoch": 18.74, + "learning_rate": 1.1265057273063185e-06, + "loss": 0.2758, + "step": 50753 + }, + { + "epoch": 18.74, + "learning_rate": 1.1264687769429734e-06, + "loss": 0.3695, + "step": 50754 + }, + { + "epoch": 18.74, + "learning_rate": 1.1264318265796282e-06, + "loss": 0.1208, + "step": 50755 + }, + { + "epoch": 18.74, + "learning_rate": 1.1263948762162828e-06, + "loss": 0.29, + "step": 50756 + }, + { + "epoch": 18.74, + "learning_rate": 1.1263579258529375e-06, + "loss": 0.1943, + "step": 50757 + }, + { + "epoch": 18.74, + "learning_rate": 1.1263209754895923e-06, + "loss": 0.1747, + "step": 50758 + }, + { + "epoch": 18.74, + "learning_rate": 1.126284025126247e-06, + "loss": 0.2399, + "step": 50759 + }, + { + "epoch": 18.74, + "learning_rate": 1.1262470747629018e-06, + "loss": 0.2506, + "step": 50760 + }, + { + "epoch": 18.74, + "learning_rate": 1.1262101243995567e-06, + "loss": 0.318, + "step": 50761 + }, + { + "epoch": 18.75, + "learning_rate": 1.1261731740362113e-06, + "loss": 0.2889, + "step": 50762 + }, + { + "epoch": 18.75, + "learning_rate": 1.1261362236728661e-06, + "loss": 0.3122, + "step": 50763 + }, + { + "epoch": 18.75, + "learning_rate": 1.126099273309521e-06, + "loss": 0.3025, + "step": 50764 + }, + { + "epoch": 18.75, + "learning_rate": 1.1260623229461756e-06, + "loss": 0.2833, + "step": 50765 + }, + { + "epoch": 18.75, + "learning_rate": 1.1260253725828305e-06, + "loss": 0.1558, + "step": 50766 + }, + { + "epoch": 18.75, + "learning_rate": 1.1259884222194851e-06, + "loss": 0.256, + "step": 50767 + }, + { + "epoch": 18.75, + "learning_rate": 1.12595147185614e-06, + "loss": 0.2434, + "step": 50768 + }, + { + "epoch": 18.75, + "learning_rate": 1.1259145214927946e-06, + "loss": 0.1453, + "step": 50769 + }, + { + "epoch": 18.75, + "learning_rate": 1.1258775711294494e-06, + "loss": 0.2541, + "step": 50770 + }, + { + "epoch": 18.75, + "learning_rate": 1.1258406207661043e-06, + "loss": 0.3883, + "step": 50771 + }, + { + "epoch": 18.75, + "learning_rate": 1.125803670402759e-06, + "loss": 0.1703, + "step": 50772 + }, + { + "epoch": 18.75, + "learning_rate": 1.1257667200394138e-06, + "loss": 0.3709, + "step": 50773 + }, + { + "epoch": 18.75, + "learning_rate": 1.1257297696760686e-06, + "loss": 0.1532, + "step": 50774 + }, + { + "epoch": 18.75, + "learning_rate": 1.1256928193127233e-06, + "loss": 0.2271, + "step": 50775 + }, + { + "epoch": 18.75, + "learning_rate": 1.1256558689493781e-06, + "loss": 0.1378, + "step": 50776 + }, + { + "epoch": 18.75, + "learning_rate": 1.1256189185860328e-06, + "loss": 0.3022, + "step": 50777 + }, + { + "epoch": 18.75, + "learning_rate": 1.1255819682226876e-06, + "loss": 0.1238, + "step": 50778 + }, + { + "epoch": 18.75, + "learning_rate": 1.1255450178593422e-06, + "loss": 0.2959, + "step": 50779 + }, + { + "epoch": 18.75, + "learning_rate": 1.125508067495997e-06, + "loss": 0.2044, + "step": 50780 + }, + { + "epoch": 18.75, + "learning_rate": 1.125471117132652e-06, + "loss": 0.2101, + "step": 50781 + }, + { + "epoch": 18.75, + "learning_rate": 1.1254341667693066e-06, + "loss": 0.1981, + "step": 50782 + }, + { + "epoch": 18.75, + "learning_rate": 1.1253972164059614e-06, + "loss": 0.2828, + "step": 50783 + }, + { + "epoch": 18.75, + "learning_rate": 1.1253602660426163e-06, + "loss": 0.1279, + "step": 50784 + }, + { + "epoch": 18.75, + "learning_rate": 1.125323315679271e-06, + "loss": 0.2061, + "step": 50785 + }, + { + "epoch": 18.75, + "learning_rate": 1.1252863653159257e-06, + "loss": 0.2085, + "step": 50786 + }, + { + "epoch": 18.75, + "learning_rate": 1.1252494149525804e-06, + "loss": 0.239, + "step": 50787 + }, + { + "epoch": 18.75, + "learning_rate": 1.125212464589235e-06, + "loss": 0.1445, + "step": 50788 + }, + { + "epoch": 18.76, + "learning_rate": 1.1251755142258899e-06, + "loss": 0.3307, + "step": 50789 + }, + { + "epoch": 18.76, + "learning_rate": 1.1251385638625447e-06, + "loss": 0.2085, + "step": 50790 + }, + { + "epoch": 18.76, + "learning_rate": 1.1251016134991994e-06, + "loss": 0.1481, + "step": 50791 + }, + { + "epoch": 18.76, + "learning_rate": 1.1250646631358542e-06, + "loss": 0.1831, + "step": 50792 + }, + { + "epoch": 18.76, + "learning_rate": 1.125027712772509e-06, + "loss": 0.2134, + "step": 50793 + }, + { + "epoch": 18.76, + "learning_rate": 1.1249907624091637e-06, + "loss": 0.2121, + "step": 50794 + }, + { + "epoch": 18.76, + "learning_rate": 1.1249538120458185e-06, + "loss": 0.1382, + "step": 50795 + }, + { + "epoch": 18.76, + "learning_rate": 1.1249168616824734e-06, + "loss": 0.1888, + "step": 50796 + }, + { + "epoch": 18.76, + "learning_rate": 1.124879911319128e-06, + "loss": 0.2711, + "step": 50797 + }, + { + "epoch": 18.76, + "learning_rate": 1.1248429609557827e-06, + "loss": 0.1852, + "step": 50798 + }, + { + "epoch": 18.76, + "learning_rate": 1.1248060105924375e-06, + "loss": 0.2212, + "step": 50799 + }, + { + "epoch": 18.76, + "learning_rate": 1.1247690602290924e-06, + "loss": 0.2382, + "step": 50800 + }, + { + "epoch": 18.76, + "learning_rate": 1.124732109865747e-06, + "loss": 0.2969, + "step": 50801 + }, + { + "epoch": 18.76, + "learning_rate": 1.1246951595024018e-06, + "loss": 0.2195, + "step": 50802 + }, + { + "epoch": 18.76, + "learning_rate": 1.1246582091390567e-06, + "loss": 0.3063, + "step": 50803 + }, + { + "epoch": 18.76, + "learning_rate": 1.1246212587757113e-06, + "loss": 0.3432, + "step": 50804 + }, + { + "epoch": 18.76, + "learning_rate": 1.1245843084123662e-06, + "loss": 0.3883, + "step": 50805 + }, + { + "epoch": 18.76, + "learning_rate": 1.124547358049021e-06, + "loss": 0.3707, + "step": 50806 + }, + { + "epoch": 18.76, + "learning_rate": 1.1245104076856754e-06, + "loss": 0.1928, + "step": 50807 + }, + { + "epoch": 18.76, + "learning_rate": 1.1244734573223303e-06, + "loss": 0.1346, + "step": 50808 + }, + { + "epoch": 18.76, + "learning_rate": 1.1244365069589851e-06, + "loss": 0.2949, + "step": 50809 + }, + { + "epoch": 18.76, + "learning_rate": 1.1243995565956398e-06, + "loss": 0.2226, + "step": 50810 + }, + { + "epoch": 18.76, + "learning_rate": 1.1243626062322946e-06, + "loss": 0.1976, + "step": 50811 + }, + { + "epoch": 18.76, + "learning_rate": 1.1243256558689495e-06, + "loss": 0.2926, + "step": 50812 + }, + { + "epoch": 18.76, + "learning_rate": 1.1242887055056041e-06, + "loss": 0.2158, + "step": 50813 + }, + { + "epoch": 18.76, + "learning_rate": 1.124251755142259e-06, + "loss": 0.2183, + "step": 50814 + }, + { + "epoch": 18.76, + "learning_rate": 1.1242148047789138e-06, + "loss": 0.2061, + "step": 50815 + }, + { + "epoch": 18.77, + "learning_rate": 1.1241778544155684e-06, + "loss": 0.2095, + "step": 50816 + }, + { + "epoch": 18.77, + "learning_rate": 1.124140904052223e-06, + "loss": 0.2184, + "step": 50817 + }, + { + "epoch": 18.77, + "learning_rate": 1.124103953688878e-06, + "loss": 0.4168, + "step": 50818 + }, + { + "epoch": 18.77, + "learning_rate": 1.1240670033255328e-06, + "loss": 0.0822, + "step": 50819 + }, + { + "epoch": 18.77, + "learning_rate": 1.1240300529621874e-06, + "loss": 0.3429, + "step": 50820 + }, + { + "epoch": 18.77, + "learning_rate": 1.1239931025988423e-06, + "loss": 0.2342, + "step": 50821 + }, + { + "epoch": 18.77, + "learning_rate": 1.1239561522354971e-06, + "loss": 0.1604, + "step": 50822 + }, + { + "epoch": 18.77, + "learning_rate": 1.1239192018721517e-06, + "loss": 0.1206, + "step": 50823 + }, + { + "epoch": 18.77, + "learning_rate": 1.1238822515088066e-06, + "loss": 0.3366, + "step": 50824 + }, + { + "epoch": 18.77, + "learning_rate": 1.1238453011454614e-06, + "loss": 0.1639, + "step": 50825 + }, + { + "epoch": 18.77, + "learning_rate": 1.123808350782116e-06, + "loss": 0.1275, + "step": 50826 + }, + { + "epoch": 18.77, + "learning_rate": 1.1237714004187707e-06, + "loss": 0.2926, + "step": 50827 + }, + { + "epoch": 18.77, + "learning_rate": 1.1237344500554256e-06, + "loss": 0.1773, + "step": 50828 + }, + { + "epoch": 18.77, + "learning_rate": 1.1236974996920802e-06, + "loss": 0.3101, + "step": 50829 + }, + { + "epoch": 18.77, + "learning_rate": 1.123660549328735e-06, + "loss": 0.2079, + "step": 50830 + }, + { + "epoch": 18.77, + "learning_rate": 1.12362359896539e-06, + "loss": 0.2588, + "step": 50831 + }, + { + "epoch": 18.77, + "learning_rate": 1.1235866486020445e-06, + "loss": 0.1187, + "step": 50832 + }, + { + "epoch": 18.77, + "learning_rate": 1.1235496982386994e-06, + "loss": 0.2183, + "step": 50833 + }, + { + "epoch": 18.77, + "learning_rate": 1.1235127478753542e-06, + "loss": 0.2117, + "step": 50834 + }, + { + "epoch": 18.77, + "learning_rate": 1.1234757975120089e-06, + "loss": 0.2573, + "step": 50835 + }, + { + "epoch": 18.77, + "learning_rate": 1.1234388471486637e-06, + "loss": 0.2215, + "step": 50836 + }, + { + "epoch": 18.77, + "learning_rate": 1.1234018967853183e-06, + "loss": 0.1892, + "step": 50837 + }, + { + "epoch": 18.77, + "learning_rate": 1.1233649464219732e-06, + "loss": 0.3067, + "step": 50838 + }, + { + "epoch": 18.77, + "learning_rate": 1.1233279960586278e-06, + "loss": 0.2907, + "step": 50839 + }, + { + "epoch": 18.77, + "learning_rate": 1.1232910456952827e-06, + "loss": 0.2808, + "step": 50840 + }, + { + "epoch": 18.77, + "learning_rate": 1.1232540953319375e-06, + "loss": 0.1172, + "step": 50841 + }, + { + "epoch": 18.77, + "learning_rate": 1.1232171449685922e-06, + "loss": 0.1919, + "step": 50842 + }, + { + "epoch": 18.78, + "learning_rate": 1.123180194605247e-06, + "loss": 0.1972, + "step": 50843 + }, + { + "epoch": 18.78, + "learning_rate": 1.1231432442419019e-06, + "loss": 0.1781, + "step": 50844 + }, + { + "epoch": 18.78, + "learning_rate": 1.1231062938785565e-06, + "loss": 0.1956, + "step": 50845 + }, + { + "epoch": 18.78, + "learning_rate": 1.1230693435152113e-06, + "loss": 0.1794, + "step": 50846 + }, + { + "epoch": 18.78, + "learning_rate": 1.123032393151866e-06, + "loss": 0.2021, + "step": 50847 + }, + { + "epoch": 18.78, + "learning_rate": 1.1229954427885208e-06, + "loss": 0.2901, + "step": 50848 + }, + { + "epoch": 18.78, + "learning_rate": 1.1229584924251755e-06, + "loss": 0.2029, + "step": 50849 + }, + { + "epoch": 18.78, + "learning_rate": 1.1229215420618303e-06, + "loss": 0.3364, + "step": 50850 + }, + { + "epoch": 18.78, + "learning_rate": 1.1228845916984852e-06, + "loss": 0.2997, + "step": 50851 + }, + { + "epoch": 18.78, + "learning_rate": 1.1228476413351398e-06, + "loss": 0.3682, + "step": 50852 + }, + { + "epoch": 18.78, + "learning_rate": 1.1228106909717946e-06, + "loss": 0.2078, + "step": 50853 + }, + { + "epoch": 18.78, + "learning_rate": 1.1227737406084495e-06, + "loss": 0.2132, + "step": 50854 + }, + { + "epoch": 18.78, + "learning_rate": 1.1227367902451041e-06, + "loss": 0.29, + "step": 50855 + }, + { + "epoch": 18.78, + "learning_rate": 1.1226998398817588e-06, + "loss": 0.1979, + "step": 50856 + }, + { + "epoch": 18.78, + "learning_rate": 1.1226628895184136e-06, + "loss": 0.1782, + "step": 50857 + }, + { + "epoch": 18.78, + "learning_rate": 1.1226259391550683e-06, + "loss": 0.245, + "step": 50858 + }, + { + "epoch": 18.78, + "learning_rate": 1.122588988791723e-06, + "loss": 0.2062, + "step": 50859 + }, + { + "epoch": 18.78, + "learning_rate": 1.122552038428378e-06, + "loss": 0.1296, + "step": 50860 + }, + { + "epoch": 18.78, + "learning_rate": 1.1225150880650326e-06, + "loss": 0.1968, + "step": 50861 + }, + { + "epoch": 18.78, + "learning_rate": 1.1224781377016874e-06, + "loss": 0.3242, + "step": 50862 + }, + { + "epoch": 18.78, + "learning_rate": 1.1224411873383423e-06, + "loss": 0.1949, + "step": 50863 + }, + { + "epoch": 18.78, + "learning_rate": 1.122404236974997e-06, + "loss": 0.2204, + "step": 50864 + }, + { + "epoch": 18.78, + "learning_rate": 1.1223672866116518e-06, + "loss": 0.2799, + "step": 50865 + }, + { + "epoch": 18.78, + "learning_rate": 1.1223303362483064e-06, + "loss": 0.2099, + "step": 50866 + }, + { + "epoch": 18.78, + "learning_rate": 1.1222933858849613e-06, + "loss": 0.2104, + "step": 50867 + }, + { + "epoch": 18.78, + "learning_rate": 1.1222564355216159e-06, + "loss": 0.1521, + "step": 50868 + }, + { + "epoch": 18.78, + "learning_rate": 1.1222194851582707e-06, + "loss": 0.2211, + "step": 50869 + }, + { + "epoch": 18.79, + "learning_rate": 1.1221825347949256e-06, + "loss": 0.2253, + "step": 50870 + }, + { + "epoch": 18.79, + "learning_rate": 1.1221455844315802e-06, + "loss": 0.2448, + "step": 50871 + }, + { + "epoch": 18.79, + "learning_rate": 1.122108634068235e-06, + "loss": 0.1423, + "step": 50872 + }, + { + "epoch": 18.79, + "learning_rate": 1.12207168370489e-06, + "loss": 0.1869, + "step": 50873 + }, + { + "epoch": 18.79, + "learning_rate": 1.1220347333415446e-06, + "loss": 0.2017, + "step": 50874 + }, + { + "epoch": 18.79, + "learning_rate": 1.1219977829781994e-06, + "loss": 0.1741, + "step": 50875 + }, + { + "epoch": 18.79, + "learning_rate": 1.121960832614854e-06, + "loss": 0.1992, + "step": 50876 + }, + { + "epoch": 18.79, + "learning_rate": 1.1219238822515087e-06, + "loss": 0.1286, + "step": 50877 + }, + { + "epoch": 18.79, + "learning_rate": 1.1218869318881635e-06, + "loss": 0.1747, + "step": 50878 + }, + { + "epoch": 18.79, + "learning_rate": 1.1218499815248184e-06, + "loss": 0.1676, + "step": 50879 + }, + { + "epoch": 18.79, + "learning_rate": 1.121813031161473e-06, + "loss": 0.2381, + "step": 50880 + }, + { + "epoch": 18.79, + "learning_rate": 1.1217760807981279e-06, + "loss": 0.1848, + "step": 50881 + }, + { + "epoch": 18.79, + "learning_rate": 1.1217391304347827e-06, + "loss": 0.1035, + "step": 50882 + }, + { + "epoch": 18.79, + "learning_rate": 1.1217021800714373e-06, + "loss": 0.1296, + "step": 50883 + }, + { + "epoch": 18.79, + "learning_rate": 1.1216652297080922e-06, + "loss": 0.1345, + "step": 50884 + }, + { + "epoch": 18.79, + "learning_rate": 1.121628279344747e-06, + "loss": 0.2526, + "step": 50885 + }, + { + "epoch": 18.79, + "learning_rate": 1.1215913289814017e-06, + "loss": 0.19, + "step": 50886 + }, + { + "epoch": 18.79, + "learning_rate": 1.1215543786180563e-06, + "loss": 0.3447, + "step": 50887 + }, + { + "epoch": 18.79, + "learning_rate": 1.1215174282547112e-06, + "loss": 0.1666, + "step": 50888 + }, + { + "epoch": 18.79, + "learning_rate": 1.121480477891366e-06, + "loss": 0.2457, + "step": 50889 + }, + { + "epoch": 18.79, + "learning_rate": 1.1214435275280206e-06, + "loss": 0.1062, + "step": 50890 + }, + { + "epoch": 18.79, + "learning_rate": 1.1214065771646755e-06, + "loss": 0.2904, + "step": 50891 + }, + { + "epoch": 18.79, + "learning_rate": 1.1213696268013303e-06, + "loss": 0.2808, + "step": 50892 + }, + { + "epoch": 18.79, + "learning_rate": 1.121332676437985e-06, + "loss": 0.1372, + "step": 50893 + }, + { + "epoch": 18.79, + "learning_rate": 1.1212957260746398e-06, + "loss": 0.2812, + "step": 50894 + }, + { + "epoch": 18.79, + "learning_rate": 1.1212587757112947e-06, + "loss": 0.2625, + "step": 50895 + }, + { + "epoch": 18.79, + "learning_rate": 1.121221825347949e-06, + "loss": 0.1339, + "step": 50896 + }, + { + "epoch": 18.8, + "learning_rate": 1.121184874984604e-06, + "loss": 0.1916, + "step": 50897 + }, + { + "epoch": 18.8, + "learning_rate": 1.1211479246212588e-06, + "loss": 0.2661, + "step": 50898 + }, + { + "epoch": 18.8, + "learning_rate": 1.1211109742579134e-06, + "loss": 0.2222, + "step": 50899 + }, + { + "epoch": 18.8, + "learning_rate": 1.1210740238945683e-06, + "loss": 0.1163, + "step": 50900 + }, + { + "epoch": 18.8, + "learning_rate": 1.1210370735312231e-06, + "loss": 0.3054, + "step": 50901 + }, + { + "epoch": 18.8, + "learning_rate": 1.1210001231678778e-06, + "loss": 0.2258, + "step": 50902 + }, + { + "epoch": 18.8, + "learning_rate": 1.1209631728045326e-06, + "loss": 0.2725, + "step": 50903 + }, + { + "epoch": 18.8, + "learning_rate": 1.1209262224411875e-06, + "loss": 0.2633, + "step": 50904 + }, + { + "epoch": 18.8, + "learning_rate": 1.120889272077842e-06, + "loss": 0.2565, + "step": 50905 + }, + { + "epoch": 18.8, + "learning_rate": 1.1208523217144967e-06, + "loss": 0.2155, + "step": 50906 + }, + { + "epoch": 18.8, + "learning_rate": 1.1208153713511516e-06, + "loss": 0.1972, + "step": 50907 + }, + { + "epoch": 18.8, + "learning_rate": 1.1207784209878064e-06, + "loss": 0.3061, + "step": 50908 + }, + { + "epoch": 18.8, + "learning_rate": 1.120741470624461e-06, + "loss": 0.2338, + "step": 50909 + }, + { + "epoch": 18.8, + "learning_rate": 1.120704520261116e-06, + "loss": 0.176, + "step": 50910 + }, + { + "epoch": 18.8, + "learning_rate": 1.1206675698977708e-06, + "loss": 0.2556, + "step": 50911 + }, + { + "epoch": 18.8, + "learning_rate": 1.1206306195344254e-06, + "loss": 0.3213, + "step": 50912 + }, + { + "epoch": 18.8, + "learning_rate": 1.1205936691710802e-06, + "loss": 0.2949, + "step": 50913 + }, + { + "epoch": 18.8, + "learning_rate": 1.120556718807735e-06, + "loss": 0.1451, + "step": 50914 + }, + { + "epoch": 18.8, + "learning_rate": 1.1205197684443897e-06, + "loss": 0.3641, + "step": 50915 + }, + { + "epoch": 18.8, + "learning_rate": 1.1204828180810444e-06, + "loss": 0.2996, + "step": 50916 + }, + { + "epoch": 18.8, + "learning_rate": 1.1204458677176992e-06, + "loss": 0.3184, + "step": 50917 + }, + { + "epoch": 18.8, + "learning_rate": 1.120408917354354e-06, + "loss": 0.1944, + "step": 50918 + }, + { + "epoch": 18.8, + "learning_rate": 1.1203719669910087e-06, + "loss": 0.2358, + "step": 50919 + }, + { + "epoch": 18.8, + "learning_rate": 1.1203350166276635e-06, + "loss": 0.1483, + "step": 50920 + }, + { + "epoch": 18.8, + "learning_rate": 1.1202980662643184e-06, + "loss": 0.1901, + "step": 50921 + }, + { + "epoch": 18.8, + "learning_rate": 1.120261115900973e-06, + "loss": 0.1706, + "step": 50922 + }, + { + "epoch": 18.8, + "learning_rate": 1.1202241655376279e-06, + "loss": 0.2502, + "step": 50923 + }, + { + "epoch": 18.81, + "learning_rate": 1.1201872151742827e-06, + "loss": 0.2937, + "step": 50924 + }, + { + "epoch": 18.81, + "learning_rate": 1.1201502648109374e-06, + "loss": 0.1537, + "step": 50925 + }, + { + "epoch": 18.81, + "learning_rate": 1.120113314447592e-06, + "loss": 0.2504, + "step": 50926 + }, + { + "epoch": 18.81, + "learning_rate": 1.1200763640842469e-06, + "loss": 0.2431, + "step": 50927 + }, + { + "epoch": 18.81, + "learning_rate": 1.1200394137209015e-06, + "loss": 0.2185, + "step": 50928 + }, + { + "epoch": 18.81, + "learning_rate": 1.1200024633575563e-06, + "loss": 0.2576, + "step": 50929 + }, + { + "epoch": 18.81, + "learning_rate": 1.1199655129942112e-06, + "loss": 0.1719, + "step": 50930 + }, + { + "epoch": 18.81, + "learning_rate": 1.1199285626308658e-06, + "loss": 0.1389, + "step": 50931 + }, + { + "epoch": 18.81, + "learning_rate": 1.1198916122675207e-06, + "loss": 0.2866, + "step": 50932 + }, + { + "epoch": 18.81, + "learning_rate": 1.1198546619041755e-06, + "loss": 0.1945, + "step": 50933 + }, + { + "epoch": 18.81, + "learning_rate": 1.1198177115408302e-06, + "loss": 0.2207, + "step": 50934 + }, + { + "epoch": 18.81, + "learning_rate": 1.119780761177485e-06, + "loss": 0.1108, + "step": 50935 + }, + { + "epoch": 18.81, + "learning_rate": 1.1197438108141396e-06, + "loss": 0.2964, + "step": 50936 + }, + { + "epoch": 18.81, + "learning_rate": 1.1197068604507945e-06, + "loss": 0.2425, + "step": 50937 + }, + { + "epoch": 18.81, + "learning_rate": 1.1196699100874491e-06, + "loss": 0.2266, + "step": 50938 + }, + { + "epoch": 18.81, + "learning_rate": 1.119632959724104e-06, + "loss": 0.3289, + "step": 50939 + }, + { + "epoch": 18.81, + "learning_rate": 1.1195960093607588e-06, + "loss": 0.2374, + "step": 50940 + }, + { + "epoch": 18.81, + "learning_rate": 1.1195590589974135e-06, + "loss": 0.3655, + "step": 50941 + }, + { + "epoch": 18.81, + "learning_rate": 1.1195221086340683e-06, + "loss": 0.2353, + "step": 50942 + }, + { + "epoch": 18.81, + "learning_rate": 1.1194851582707232e-06, + "loss": 0.2902, + "step": 50943 + }, + { + "epoch": 18.81, + "learning_rate": 1.1194482079073778e-06, + "loss": 0.1882, + "step": 50944 + }, + { + "epoch": 18.81, + "learning_rate": 1.1194112575440326e-06, + "loss": 0.2992, + "step": 50945 + }, + { + "epoch": 18.81, + "learning_rate": 1.1193743071806873e-06, + "loss": 0.1865, + "step": 50946 + }, + { + "epoch": 18.81, + "learning_rate": 1.119337356817342e-06, + "loss": 0.3067, + "step": 50947 + }, + { + "epoch": 18.81, + "learning_rate": 1.1193004064539968e-06, + "loss": 0.2085, + "step": 50948 + }, + { + "epoch": 18.81, + "learning_rate": 1.1192634560906516e-06, + "loss": 0.182, + "step": 50949 + }, + { + "epoch": 18.81, + "learning_rate": 1.1192265057273062e-06, + "loss": 0.1899, + "step": 50950 + }, + { + "epoch": 18.81, + "learning_rate": 1.119189555363961e-06, + "loss": 0.2824, + "step": 50951 + }, + { + "epoch": 18.82, + "learning_rate": 1.119152605000616e-06, + "loss": 0.2083, + "step": 50952 + }, + { + "epoch": 18.82, + "learning_rate": 1.1191156546372706e-06, + "loss": 0.2374, + "step": 50953 + }, + { + "epoch": 18.82, + "learning_rate": 1.1190787042739254e-06, + "loss": 0.2507, + "step": 50954 + }, + { + "epoch": 18.82, + "learning_rate": 1.1190417539105803e-06, + "loss": 0.2064, + "step": 50955 + }, + { + "epoch": 18.82, + "learning_rate": 1.119004803547235e-06, + "loss": 0.1734, + "step": 50956 + }, + { + "epoch": 18.82, + "learning_rate": 1.1189678531838895e-06, + "loss": 0.2728, + "step": 50957 + }, + { + "epoch": 18.82, + "learning_rate": 1.1189309028205444e-06, + "loss": 0.2846, + "step": 50958 + }, + { + "epoch": 18.82, + "learning_rate": 1.1188939524571992e-06, + "loss": 0.3529, + "step": 50959 + }, + { + "epoch": 18.82, + "learning_rate": 1.1188570020938539e-06, + "loss": 0.1677, + "step": 50960 + }, + { + "epoch": 18.82, + "learning_rate": 1.1188200517305087e-06, + "loss": 0.1451, + "step": 50961 + }, + { + "epoch": 18.82, + "learning_rate": 1.1187831013671636e-06, + "loss": 0.2657, + "step": 50962 + }, + { + "epoch": 18.82, + "learning_rate": 1.1187461510038182e-06, + "loss": 0.1375, + "step": 50963 + }, + { + "epoch": 18.82, + "learning_rate": 1.118709200640473e-06, + "loss": 0.1431, + "step": 50964 + }, + { + "epoch": 18.82, + "learning_rate": 1.118672250277128e-06, + "loss": 0.3121, + "step": 50965 + }, + { + "epoch": 18.82, + "learning_rate": 1.1186352999137825e-06, + "loss": 0.2636, + "step": 50966 + }, + { + "epoch": 18.82, + "learning_rate": 1.1185983495504372e-06, + "loss": 0.1326, + "step": 50967 + }, + { + "epoch": 18.82, + "learning_rate": 1.118561399187092e-06, + "loss": 0.3767, + "step": 50968 + }, + { + "epoch": 18.82, + "learning_rate": 1.1185244488237469e-06, + "loss": 0.1355, + "step": 50969 + }, + { + "epoch": 18.82, + "learning_rate": 1.1184874984604015e-06, + "loss": 0.1972, + "step": 50970 + }, + { + "epoch": 18.82, + "learning_rate": 1.1184505480970564e-06, + "loss": 0.1435, + "step": 50971 + }, + { + "epoch": 18.82, + "learning_rate": 1.1184135977337112e-06, + "loss": 0.2551, + "step": 50972 + }, + { + "epoch": 18.82, + "learning_rate": 1.1183766473703658e-06, + "loss": 0.1699, + "step": 50973 + }, + { + "epoch": 18.82, + "learning_rate": 1.1183396970070207e-06, + "loss": 0.311, + "step": 50974 + }, + { + "epoch": 18.82, + "learning_rate": 1.1183027466436753e-06, + "loss": 0.3542, + "step": 50975 + }, + { + "epoch": 18.82, + "learning_rate": 1.11826579628033e-06, + "loss": 0.3559, + "step": 50976 + }, + { + "epoch": 18.82, + "learning_rate": 1.1182288459169848e-06, + "loss": 0.296, + "step": 50977 + }, + { + "epoch": 18.82, + "learning_rate": 1.1181918955536397e-06, + "loss": 0.297, + "step": 50978 + }, + { + "epoch": 18.83, + "learning_rate": 1.1181549451902943e-06, + "loss": 0.246, + "step": 50979 + }, + { + "epoch": 18.83, + "learning_rate": 1.1181179948269491e-06, + "loss": 0.3191, + "step": 50980 + }, + { + "epoch": 18.83, + "learning_rate": 1.118081044463604e-06, + "loss": 0.4005, + "step": 50981 + }, + { + "epoch": 18.83, + "learning_rate": 1.1180440941002586e-06, + "loss": 0.2258, + "step": 50982 + }, + { + "epoch": 18.83, + "learning_rate": 1.1180071437369135e-06, + "loss": 0.2383, + "step": 50983 + }, + { + "epoch": 18.83, + "learning_rate": 1.1179701933735683e-06, + "loss": 0.219, + "step": 50984 + }, + { + "epoch": 18.83, + "learning_rate": 1.117933243010223e-06, + "loss": 0.2801, + "step": 50985 + }, + { + "epoch": 18.83, + "learning_rate": 1.1178962926468776e-06, + "loss": 0.175, + "step": 50986 + }, + { + "epoch": 18.83, + "learning_rate": 1.1178593422835324e-06, + "loss": 0.1687, + "step": 50987 + }, + { + "epoch": 18.83, + "learning_rate": 1.1178223919201873e-06, + "loss": 0.2168, + "step": 50988 + }, + { + "epoch": 18.83, + "learning_rate": 1.117785441556842e-06, + "loss": 0.248, + "step": 50989 + }, + { + "epoch": 18.83, + "learning_rate": 1.1177484911934968e-06, + "loss": 0.1056, + "step": 50990 + }, + { + "epoch": 18.83, + "learning_rate": 1.1177115408301516e-06, + "loss": 0.3382, + "step": 50991 + }, + { + "epoch": 18.83, + "learning_rate": 1.1176745904668063e-06, + "loss": 0.2462, + "step": 50992 + }, + { + "epoch": 18.83, + "learning_rate": 1.1176376401034611e-06, + "loss": 0.2115, + "step": 50993 + }, + { + "epoch": 18.83, + "learning_rate": 1.117600689740116e-06, + "loss": 0.2175, + "step": 50994 + }, + { + "epoch": 18.83, + "learning_rate": 1.1175637393767706e-06, + "loss": 0.3577, + "step": 50995 + }, + { + "epoch": 18.83, + "learning_rate": 1.1175267890134252e-06, + "loss": 0.2227, + "step": 50996 + }, + { + "epoch": 18.83, + "learning_rate": 1.11748983865008e-06, + "loss": 0.1373, + "step": 50997 + }, + { + "epoch": 18.83, + "learning_rate": 1.1174528882867347e-06, + "loss": 0.2242, + "step": 50998 + }, + { + "epoch": 18.83, + "learning_rate": 1.1174159379233896e-06, + "loss": 0.135, + "step": 50999 + }, + { + "epoch": 18.83, + "learning_rate": 1.1173789875600444e-06, + "loss": 0.194, + "step": 51000 + }, + { + "epoch": 18.83, + "learning_rate": 1.117342037196699e-06, + "loss": 0.1981, + "step": 51001 + }, + { + "epoch": 18.83, + "learning_rate": 1.117305086833354e-06, + "loss": 0.2158, + "step": 51002 + }, + { + "epoch": 18.83, + "learning_rate": 1.1172681364700087e-06, + "loss": 0.157, + "step": 51003 + }, + { + "epoch": 18.83, + "learning_rate": 1.1172311861066634e-06, + "loss": 0.2524, + "step": 51004 + }, + { + "epoch": 18.83, + "learning_rate": 1.1171942357433182e-06, + "loss": 0.3231, + "step": 51005 + }, + { + "epoch": 18.84, + "learning_rate": 1.1171572853799729e-06, + "loss": 0.2125, + "step": 51006 + }, + { + "epoch": 18.84, + "learning_rate": 1.1171203350166277e-06, + "loss": 0.1068, + "step": 51007 + }, + { + "epoch": 18.84, + "learning_rate": 1.1170833846532824e-06, + "loss": 0.1258, + "step": 51008 + }, + { + "epoch": 18.84, + "learning_rate": 1.1170464342899372e-06, + "loss": 0.2204, + "step": 51009 + }, + { + "epoch": 18.84, + "learning_rate": 1.117009483926592e-06, + "loss": 0.1187, + "step": 51010 + }, + { + "epoch": 18.84, + "learning_rate": 1.1169725335632467e-06, + "loss": 0.2908, + "step": 51011 + }, + { + "epoch": 18.84, + "learning_rate": 1.1169355831999015e-06, + "loss": 0.3269, + "step": 51012 + }, + { + "epoch": 18.84, + "learning_rate": 1.1168986328365564e-06, + "loss": 0.3144, + "step": 51013 + }, + { + "epoch": 18.84, + "learning_rate": 1.116861682473211e-06, + "loss": 0.2199, + "step": 51014 + }, + { + "epoch": 18.84, + "learning_rate": 1.1168247321098659e-06, + "loss": 0.2005, + "step": 51015 + }, + { + "epoch": 18.84, + "learning_rate": 1.1167877817465205e-06, + "loss": 0.2407, + "step": 51016 + }, + { + "epoch": 18.84, + "learning_rate": 1.1167508313831751e-06, + "loss": 0.1882, + "step": 51017 + }, + { + "epoch": 18.84, + "learning_rate": 1.11671388101983e-06, + "loss": 0.2319, + "step": 51018 + }, + { + "epoch": 18.84, + "learning_rate": 1.1166769306564848e-06, + "loss": 0.2632, + "step": 51019 + }, + { + "epoch": 18.84, + "learning_rate": 1.1166399802931395e-06, + "loss": 0.155, + "step": 51020 + }, + { + "epoch": 18.84, + "learning_rate": 1.1166030299297943e-06, + "loss": 0.242, + "step": 51021 + }, + { + "epoch": 18.84, + "learning_rate": 1.1165660795664492e-06, + "loss": 0.2285, + "step": 51022 + }, + { + "epoch": 18.84, + "learning_rate": 1.1165291292031038e-06, + "loss": 0.1259, + "step": 51023 + }, + { + "epoch": 18.84, + "learning_rate": 1.1164921788397587e-06, + "loss": 0.2724, + "step": 51024 + }, + { + "epoch": 18.84, + "learning_rate": 1.1164552284764135e-06, + "loss": 0.1899, + "step": 51025 + }, + { + "epoch": 18.84, + "learning_rate": 1.1164182781130681e-06, + "loss": 0.2311, + "step": 51026 + }, + { + "epoch": 18.84, + "learning_rate": 1.1163813277497228e-06, + "loss": 0.1617, + "step": 51027 + }, + { + "epoch": 18.84, + "learning_rate": 1.1163443773863776e-06, + "loss": 0.2689, + "step": 51028 + }, + { + "epoch": 18.84, + "learning_rate": 1.1163074270230325e-06, + "loss": 0.1487, + "step": 51029 + }, + { + "epoch": 18.84, + "learning_rate": 1.1162704766596871e-06, + "loss": 0.2575, + "step": 51030 + }, + { + "epoch": 18.84, + "learning_rate": 1.116233526296342e-06, + "loss": 0.488, + "step": 51031 + }, + { + "epoch": 18.84, + "learning_rate": 1.1161965759329968e-06, + "loss": 0.1663, + "step": 51032 + }, + { + "epoch": 18.85, + "learning_rate": 1.1161596255696514e-06, + "loss": 0.2257, + "step": 51033 + }, + { + "epoch": 18.85, + "learning_rate": 1.1161226752063063e-06, + "loss": 0.2058, + "step": 51034 + }, + { + "epoch": 18.85, + "learning_rate": 1.1160857248429611e-06, + "loss": 0.2996, + "step": 51035 + }, + { + "epoch": 18.85, + "learning_rate": 1.1160487744796158e-06, + "loss": 0.2442, + "step": 51036 + }, + { + "epoch": 18.85, + "learning_rate": 1.1160118241162704e-06, + "loss": 0.2117, + "step": 51037 + }, + { + "epoch": 18.85, + "learning_rate": 1.1159748737529253e-06, + "loss": 0.1196, + "step": 51038 + }, + { + "epoch": 18.85, + "learning_rate": 1.1159379233895801e-06, + "loss": 0.3313, + "step": 51039 + }, + { + "epoch": 18.85, + "learning_rate": 1.1159009730262347e-06, + "loss": 0.1885, + "step": 51040 + }, + { + "epoch": 18.85, + "learning_rate": 1.1158640226628896e-06, + "loss": 0.1958, + "step": 51041 + }, + { + "epoch": 18.85, + "learning_rate": 1.1158270722995444e-06, + "loss": 0.3031, + "step": 51042 + }, + { + "epoch": 18.85, + "learning_rate": 1.115790121936199e-06, + "loss": 0.2174, + "step": 51043 + }, + { + "epoch": 18.85, + "learning_rate": 1.115753171572854e-06, + "loss": 0.2165, + "step": 51044 + }, + { + "epoch": 18.85, + "learning_rate": 1.1157162212095088e-06, + "loss": 0.2292, + "step": 51045 + }, + { + "epoch": 18.85, + "learning_rate": 1.1156792708461632e-06, + "loss": 0.1027, + "step": 51046 + }, + { + "epoch": 18.85, + "learning_rate": 1.115642320482818e-06, + "loss": 0.2077, + "step": 51047 + }, + { + "epoch": 18.85, + "learning_rate": 1.1156053701194729e-06, + "loss": 0.4343, + "step": 51048 + }, + { + "epoch": 18.85, + "learning_rate": 1.1155684197561275e-06, + "loss": 0.3373, + "step": 51049 + }, + { + "epoch": 18.85, + "learning_rate": 1.1155314693927824e-06, + "loss": 0.2103, + "step": 51050 + }, + { + "epoch": 18.85, + "learning_rate": 1.1154945190294372e-06, + "loss": 0.2148, + "step": 51051 + }, + { + "epoch": 18.85, + "learning_rate": 1.1154575686660919e-06, + "loss": 0.2019, + "step": 51052 + }, + { + "epoch": 18.85, + "learning_rate": 1.1154206183027467e-06, + "loss": 0.282, + "step": 51053 + }, + { + "epoch": 18.85, + "learning_rate": 1.1153836679394016e-06, + "loss": 0.1519, + "step": 51054 + }, + { + "epoch": 18.85, + "learning_rate": 1.1153467175760562e-06, + "loss": 0.2309, + "step": 51055 + }, + { + "epoch": 18.85, + "learning_rate": 1.1153097672127108e-06, + "loss": 0.2332, + "step": 51056 + }, + { + "epoch": 18.85, + "learning_rate": 1.1152728168493657e-06, + "loss": 0.1852, + "step": 51057 + }, + { + "epoch": 18.85, + "learning_rate": 1.1152358664860205e-06, + "loss": 0.1704, + "step": 51058 + }, + { + "epoch": 18.85, + "learning_rate": 1.1151989161226752e-06, + "loss": 0.2017, + "step": 51059 + }, + { + "epoch": 18.86, + "learning_rate": 1.11516196575933e-06, + "loss": 0.3155, + "step": 51060 + }, + { + "epoch": 18.86, + "learning_rate": 1.1151250153959849e-06, + "loss": 0.2244, + "step": 51061 + }, + { + "epoch": 18.86, + "learning_rate": 1.1150880650326395e-06, + "loss": 0.2798, + "step": 51062 + }, + { + "epoch": 18.86, + "learning_rate": 1.1150511146692943e-06, + "loss": 0.2122, + "step": 51063 + }, + { + "epoch": 18.86, + "learning_rate": 1.1150141643059492e-06, + "loss": 0.184, + "step": 51064 + }, + { + "epoch": 18.86, + "learning_rate": 1.1149772139426038e-06, + "loss": 0.332, + "step": 51065 + }, + { + "epoch": 18.86, + "learning_rate": 1.1149402635792585e-06, + "loss": 0.2169, + "step": 51066 + }, + { + "epoch": 18.86, + "learning_rate": 1.1149033132159133e-06, + "loss": 0.1545, + "step": 51067 + }, + { + "epoch": 18.86, + "learning_rate": 1.114866362852568e-06, + "loss": 0.283, + "step": 51068 + }, + { + "epoch": 18.86, + "learning_rate": 1.1148294124892228e-06, + "loss": 0.3074, + "step": 51069 + }, + { + "epoch": 18.86, + "learning_rate": 1.1147924621258776e-06, + "loss": 0.3241, + "step": 51070 + }, + { + "epoch": 18.86, + "learning_rate": 1.1147555117625323e-06, + "loss": 0.195, + "step": 51071 + }, + { + "epoch": 18.86, + "learning_rate": 1.1147185613991871e-06, + "loss": 0.2565, + "step": 51072 + }, + { + "epoch": 18.86, + "learning_rate": 1.114681611035842e-06, + "loss": 0.3139, + "step": 51073 + }, + { + "epoch": 18.86, + "learning_rate": 1.1146446606724966e-06, + "loss": 0.3751, + "step": 51074 + }, + { + "epoch": 18.86, + "learning_rate": 1.1146077103091515e-06, + "loss": 0.1596, + "step": 51075 + }, + { + "epoch": 18.86, + "learning_rate": 1.114570759945806e-06, + "loss": 0.4103, + "step": 51076 + }, + { + "epoch": 18.86, + "learning_rate": 1.114533809582461e-06, + "loss": 0.2609, + "step": 51077 + }, + { + "epoch": 18.86, + "learning_rate": 1.1144968592191156e-06, + "loss": 0.2503, + "step": 51078 + }, + { + "epoch": 18.86, + "learning_rate": 1.1144599088557704e-06, + "loss": 0.1613, + "step": 51079 + }, + { + "epoch": 18.86, + "learning_rate": 1.1144229584924253e-06, + "loss": 0.2615, + "step": 51080 + }, + { + "epoch": 18.86, + "learning_rate": 1.11438600812908e-06, + "loss": 0.1999, + "step": 51081 + }, + { + "epoch": 18.86, + "learning_rate": 1.1143490577657348e-06, + "loss": 0.2665, + "step": 51082 + }, + { + "epoch": 18.86, + "learning_rate": 1.1143121074023896e-06, + "loss": 0.1603, + "step": 51083 + }, + { + "epoch": 18.86, + "learning_rate": 1.1142751570390443e-06, + "loss": 0.1495, + "step": 51084 + }, + { + "epoch": 18.86, + "learning_rate": 1.1142382066756989e-06, + "loss": 0.223, + "step": 51085 + }, + { + "epoch": 18.86, + "learning_rate": 1.1142012563123537e-06, + "loss": 0.3172, + "step": 51086 + }, + { + "epoch": 18.87, + "learning_rate": 1.1141643059490084e-06, + "loss": 0.1698, + "step": 51087 + }, + { + "epoch": 18.87, + "learning_rate": 1.1141273555856632e-06, + "loss": 0.2373, + "step": 51088 + }, + { + "epoch": 18.87, + "learning_rate": 1.114090405222318e-06, + "loss": 0.2856, + "step": 51089 + }, + { + "epoch": 18.87, + "learning_rate": 1.1140534548589727e-06, + "loss": 0.2956, + "step": 51090 + }, + { + "epoch": 18.87, + "learning_rate": 1.1140165044956276e-06, + "loss": 0.2217, + "step": 51091 + }, + { + "epoch": 18.87, + "learning_rate": 1.1139795541322824e-06, + "loss": 0.2196, + "step": 51092 + }, + { + "epoch": 18.87, + "learning_rate": 1.113942603768937e-06, + "loss": 0.2, + "step": 51093 + }, + { + "epoch": 18.87, + "learning_rate": 1.1139056534055919e-06, + "loss": 0.1871, + "step": 51094 + }, + { + "epoch": 18.87, + "learning_rate": 1.1138687030422465e-06, + "loss": 0.232, + "step": 51095 + }, + { + "epoch": 18.87, + "learning_rate": 1.1138317526789014e-06, + "loss": 0.1754, + "step": 51096 + }, + { + "epoch": 18.87, + "learning_rate": 1.113794802315556e-06, + "loss": 0.1344, + "step": 51097 + }, + { + "epoch": 18.87, + "learning_rate": 1.1137578519522109e-06, + "loss": 0.1537, + "step": 51098 + }, + { + "epoch": 18.87, + "learning_rate": 1.1137209015888657e-06, + "loss": 0.243, + "step": 51099 + }, + { + "epoch": 18.87, + "learning_rate": 1.1136839512255203e-06, + "loss": 0.1503, + "step": 51100 + }, + { + "epoch": 18.87, + "learning_rate": 1.1136470008621752e-06, + "loss": 0.3369, + "step": 51101 + }, + { + "epoch": 18.87, + "learning_rate": 1.11361005049883e-06, + "loss": 0.2838, + "step": 51102 + }, + { + "epoch": 18.87, + "learning_rate": 1.1135731001354847e-06, + "loss": 0.1589, + "step": 51103 + }, + { + "epoch": 18.87, + "learning_rate": 1.1135361497721395e-06, + "loss": 0.1801, + "step": 51104 + }, + { + "epoch": 18.87, + "learning_rate": 1.1134991994087942e-06, + "loss": 0.2508, + "step": 51105 + }, + { + "epoch": 18.87, + "learning_rate": 1.113462249045449e-06, + "loss": 0.1295, + "step": 51106 + }, + { + "epoch": 18.87, + "learning_rate": 1.1134252986821036e-06, + "loss": 0.1902, + "step": 51107 + }, + { + "epoch": 18.87, + "learning_rate": 1.1133883483187585e-06, + "loss": 0.154, + "step": 51108 + }, + { + "epoch": 18.87, + "learning_rate": 1.1133513979554133e-06, + "loss": 0.1501, + "step": 51109 + }, + { + "epoch": 18.87, + "learning_rate": 1.113314447592068e-06, + "loss": 0.1998, + "step": 51110 + }, + { + "epoch": 18.87, + "learning_rate": 1.1132774972287228e-06, + "loss": 0.1799, + "step": 51111 + }, + { + "epoch": 18.87, + "learning_rate": 1.1132405468653777e-06, + "loss": 0.1847, + "step": 51112 + }, + { + "epoch": 18.87, + "learning_rate": 1.1132035965020323e-06, + "loss": 0.1296, + "step": 51113 + }, + { + "epoch": 18.88, + "learning_rate": 1.1131666461386872e-06, + "loss": 0.2336, + "step": 51114 + }, + { + "epoch": 18.88, + "learning_rate": 1.1131296957753418e-06, + "loss": 0.1823, + "step": 51115 + }, + { + "epoch": 18.88, + "learning_rate": 1.1130927454119964e-06, + "loss": 0.1112, + "step": 51116 + }, + { + "epoch": 18.88, + "learning_rate": 1.1130557950486513e-06, + "loss": 0.1988, + "step": 51117 + }, + { + "epoch": 18.88, + "learning_rate": 1.1130188446853061e-06, + "loss": 0.3503, + "step": 51118 + }, + { + "epoch": 18.88, + "learning_rate": 1.1129818943219608e-06, + "loss": 0.2457, + "step": 51119 + }, + { + "epoch": 18.88, + "learning_rate": 1.1129449439586156e-06, + "loss": 0.1337, + "step": 51120 + }, + { + "epoch": 18.88, + "learning_rate": 1.1129079935952705e-06, + "loss": 0.2411, + "step": 51121 + }, + { + "epoch": 18.88, + "learning_rate": 1.112871043231925e-06, + "loss": 0.134, + "step": 51122 + }, + { + "epoch": 18.88, + "learning_rate": 1.11283409286858e-06, + "loss": 0.1867, + "step": 51123 + }, + { + "epoch": 18.88, + "learning_rate": 1.1127971425052348e-06, + "loss": 0.2553, + "step": 51124 + }, + { + "epoch": 18.88, + "learning_rate": 1.1127601921418894e-06, + "loss": 0.1023, + "step": 51125 + }, + { + "epoch": 18.88, + "learning_rate": 1.112723241778544e-06, + "loss": 0.3114, + "step": 51126 + }, + { + "epoch": 18.88, + "learning_rate": 1.112686291415199e-06, + "loss": 0.1422, + "step": 51127 + }, + { + "epoch": 18.88, + "learning_rate": 1.1126493410518538e-06, + "loss": 0.18, + "step": 51128 + }, + { + "epoch": 18.88, + "learning_rate": 1.1126123906885084e-06, + "loss": 0.22, + "step": 51129 + }, + { + "epoch": 18.88, + "learning_rate": 1.1125754403251632e-06, + "loss": 0.2691, + "step": 51130 + }, + { + "epoch": 18.88, + "learning_rate": 1.112538489961818e-06, + "loss": 0.2044, + "step": 51131 + }, + { + "epoch": 18.88, + "learning_rate": 1.1125015395984727e-06, + "loss": 0.3837, + "step": 51132 + }, + { + "epoch": 18.88, + "learning_rate": 1.1124645892351276e-06, + "loss": 0.1914, + "step": 51133 + }, + { + "epoch": 18.88, + "learning_rate": 1.1124276388717824e-06, + "loss": 0.0937, + "step": 51134 + }, + { + "epoch": 18.88, + "learning_rate": 1.1123906885084369e-06, + "loss": 0.123, + "step": 51135 + }, + { + "epoch": 18.88, + "learning_rate": 1.1123537381450917e-06, + "loss": 0.1562, + "step": 51136 + }, + { + "epoch": 18.88, + "learning_rate": 1.1123167877817465e-06, + "loss": 0.2151, + "step": 51137 + }, + { + "epoch": 18.88, + "learning_rate": 1.1122798374184012e-06, + "loss": 0.187, + "step": 51138 + }, + { + "epoch": 18.88, + "learning_rate": 1.112242887055056e-06, + "loss": 0.3553, + "step": 51139 + }, + { + "epoch": 18.88, + "learning_rate": 1.1122059366917109e-06, + "loss": 0.3796, + "step": 51140 + }, + { + "epoch": 18.89, + "learning_rate": 1.1121689863283655e-06, + "loss": 0.1352, + "step": 51141 + }, + { + "epoch": 18.89, + "learning_rate": 1.1121320359650204e-06, + "loss": 0.2136, + "step": 51142 + }, + { + "epoch": 18.89, + "learning_rate": 1.1120950856016752e-06, + "loss": 0.206, + "step": 51143 + }, + { + "epoch": 18.89, + "learning_rate": 1.1120581352383299e-06, + "loss": 0.1616, + "step": 51144 + }, + { + "epoch": 18.89, + "learning_rate": 1.1120211848749845e-06, + "loss": 0.2406, + "step": 51145 + }, + { + "epoch": 18.89, + "learning_rate": 1.1119842345116393e-06, + "loss": 0.2307, + "step": 51146 + }, + { + "epoch": 18.89, + "learning_rate": 1.1119472841482942e-06, + "loss": 0.2533, + "step": 51147 + }, + { + "epoch": 18.89, + "learning_rate": 1.1119103337849488e-06, + "loss": 0.2417, + "step": 51148 + }, + { + "epoch": 18.89, + "learning_rate": 1.1118733834216037e-06, + "loss": 0.1524, + "step": 51149 + }, + { + "epoch": 18.89, + "learning_rate": 1.1118364330582585e-06, + "loss": 0.2112, + "step": 51150 + }, + { + "epoch": 18.89, + "learning_rate": 1.1117994826949132e-06, + "loss": 0.1453, + "step": 51151 + }, + { + "epoch": 18.89, + "learning_rate": 1.111762532331568e-06, + "loss": 0.1868, + "step": 51152 + }, + { + "epoch": 18.89, + "learning_rate": 1.1117255819682228e-06, + "loss": 0.2373, + "step": 51153 + }, + { + "epoch": 18.89, + "learning_rate": 1.1116886316048775e-06, + "loss": 0.1959, + "step": 51154 + }, + { + "epoch": 18.89, + "learning_rate": 1.1116516812415321e-06, + "loss": 0.2629, + "step": 51155 + }, + { + "epoch": 18.89, + "learning_rate": 1.111614730878187e-06, + "loss": 0.1995, + "step": 51156 + }, + { + "epoch": 18.89, + "learning_rate": 1.1115777805148418e-06, + "loss": 0.2118, + "step": 51157 + }, + { + "epoch": 18.89, + "learning_rate": 1.1115408301514965e-06, + "loss": 0.1797, + "step": 51158 + }, + { + "epoch": 18.89, + "learning_rate": 1.1115038797881513e-06, + "loss": 0.1765, + "step": 51159 + }, + { + "epoch": 18.89, + "learning_rate": 1.111466929424806e-06, + "loss": 0.3259, + "step": 51160 + }, + { + "epoch": 18.89, + "learning_rate": 1.1114299790614608e-06, + "loss": 0.1281, + "step": 51161 + }, + { + "epoch": 18.89, + "learning_rate": 1.1113930286981156e-06, + "loss": 0.2043, + "step": 51162 + }, + { + "epoch": 18.89, + "learning_rate": 1.1113560783347703e-06, + "loss": 0.1895, + "step": 51163 + }, + { + "epoch": 18.89, + "learning_rate": 1.1113191279714251e-06, + "loss": 0.2847, + "step": 51164 + }, + { + "epoch": 18.89, + "learning_rate": 1.1112821776080798e-06, + "loss": 0.2505, + "step": 51165 + }, + { + "epoch": 18.89, + "learning_rate": 1.1112452272447346e-06, + "loss": 0.1816, + "step": 51166 + }, + { + "epoch": 18.89, + "learning_rate": 1.1112082768813892e-06, + "loss": 0.3443, + "step": 51167 + }, + { + "epoch": 18.9, + "learning_rate": 1.111171326518044e-06, + "loss": 0.1923, + "step": 51168 + }, + { + "epoch": 18.9, + "learning_rate": 1.111134376154699e-06, + "loss": 0.134, + "step": 51169 + }, + { + "epoch": 18.9, + "learning_rate": 1.1110974257913536e-06, + "loss": 0.16, + "step": 51170 + }, + { + "epoch": 18.9, + "learning_rate": 1.1110604754280084e-06, + "loss": 0.322, + "step": 51171 + }, + { + "epoch": 18.9, + "learning_rate": 1.1110235250646633e-06, + "loss": 0.2505, + "step": 51172 + }, + { + "epoch": 18.9, + "learning_rate": 1.110986574701318e-06, + "loss": 0.1791, + "step": 51173 + }, + { + "epoch": 18.9, + "learning_rate": 1.1109496243379728e-06, + "loss": 0.1396, + "step": 51174 + }, + { + "epoch": 18.9, + "learning_rate": 1.1109126739746274e-06, + "loss": 0.3266, + "step": 51175 + }, + { + "epoch": 18.9, + "learning_rate": 1.1108757236112822e-06, + "loss": 0.2571, + "step": 51176 + }, + { + "epoch": 18.9, + "learning_rate": 1.1108387732479369e-06, + "loss": 0.1325, + "step": 51177 + }, + { + "epoch": 18.9, + "learning_rate": 1.1108018228845917e-06, + "loss": 0.2503, + "step": 51178 + }, + { + "epoch": 18.9, + "learning_rate": 1.1107648725212466e-06, + "loss": 0.3144, + "step": 51179 + }, + { + "epoch": 18.9, + "learning_rate": 1.1107279221579012e-06, + "loss": 0.1692, + "step": 51180 + }, + { + "epoch": 18.9, + "learning_rate": 1.110690971794556e-06, + "loss": 0.1902, + "step": 51181 + }, + { + "epoch": 18.9, + "learning_rate": 1.110654021431211e-06, + "loss": 0.0863, + "step": 51182 + }, + { + "epoch": 18.9, + "learning_rate": 1.1106170710678655e-06, + "loss": 0.309, + "step": 51183 + }, + { + "epoch": 18.9, + "learning_rate": 1.1105801207045204e-06, + "loss": 0.4021, + "step": 51184 + }, + { + "epoch": 18.9, + "learning_rate": 1.110543170341175e-06, + "loss": 0.1802, + "step": 51185 + }, + { + "epoch": 18.9, + "learning_rate": 1.1105062199778297e-06, + "loss": 0.2429, + "step": 51186 + }, + { + "epoch": 18.9, + "learning_rate": 1.1104692696144845e-06, + "loss": 0.2312, + "step": 51187 + }, + { + "epoch": 18.9, + "learning_rate": 1.1104323192511394e-06, + "loss": 0.1436, + "step": 51188 + }, + { + "epoch": 18.9, + "learning_rate": 1.110395368887794e-06, + "loss": 0.1333, + "step": 51189 + }, + { + "epoch": 18.9, + "learning_rate": 1.1103584185244488e-06, + "loss": 0.2088, + "step": 51190 + }, + { + "epoch": 18.9, + "learning_rate": 1.1103214681611037e-06, + "loss": 0.2145, + "step": 51191 + }, + { + "epoch": 18.9, + "learning_rate": 1.1102845177977583e-06, + "loss": 0.2097, + "step": 51192 + }, + { + "epoch": 18.9, + "learning_rate": 1.1102475674344132e-06, + "loss": 0.1234, + "step": 51193 + }, + { + "epoch": 18.9, + "learning_rate": 1.110210617071068e-06, + "loss": 0.1575, + "step": 51194 + }, + { + "epoch": 18.91, + "learning_rate": 1.1101736667077227e-06, + "loss": 0.1566, + "step": 51195 + }, + { + "epoch": 18.91, + "learning_rate": 1.1101367163443773e-06, + "loss": 0.163, + "step": 51196 + }, + { + "epoch": 18.91, + "learning_rate": 1.1100997659810321e-06, + "loss": 0.1427, + "step": 51197 + }, + { + "epoch": 18.91, + "learning_rate": 1.110062815617687e-06, + "loss": 0.2021, + "step": 51198 + }, + { + "epoch": 18.91, + "learning_rate": 1.1100258652543416e-06, + "loss": 0.3254, + "step": 51199 + }, + { + "epoch": 18.91, + "learning_rate": 1.1099889148909965e-06, + "loss": 0.1819, + "step": 51200 + }, + { + "epoch": 18.91, + "learning_rate": 1.1099519645276513e-06, + "loss": 0.3303, + "step": 51201 + }, + { + "epoch": 18.91, + "learning_rate": 1.109915014164306e-06, + "loss": 0.1754, + "step": 51202 + }, + { + "epoch": 18.91, + "learning_rate": 1.1098780638009608e-06, + "loss": 0.1766, + "step": 51203 + }, + { + "epoch": 18.91, + "learning_rate": 1.1098411134376157e-06, + "loss": 0.1714, + "step": 51204 + }, + { + "epoch": 18.91, + "learning_rate": 1.10980416307427e-06, + "loss": 0.3772, + "step": 51205 + }, + { + "epoch": 18.91, + "learning_rate": 1.109767212710925e-06, + "loss": 0.2445, + "step": 51206 + }, + { + "epoch": 18.91, + "learning_rate": 1.1097302623475798e-06, + "loss": 0.3112, + "step": 51207 + }, + { + "epoch": 18.91, + "learning_rate": 1.1096933119842344e-06, + "loss": 0.2846, + "step": 51208 + }, + { + "epoch": 18.91, + "learning_rate": 1.1096563616208893e-06, + "loss": 0.166, + "step": 51209 + }, + { + "epoch": 18.91, + "learning_rate": 1.1096194112575441e-06, + "loss": 0.2813, + "step": 51210 + }, + { + "epoch": 18.91, + "learning_rate": 1.1095824608941987e-06, + "loss": 0.2142, + "step": 51211 + }, + { + "epoch": 18.91, + "learning_rate": 1.1095455105308536e-06, + "loss": 0.2552, + "step": 51212 + }, + { + "epoch": 18.91, + "learning_rate": 1.1095085601675084e-06, + "loss": 0.3676, + "step": 51213 + }, + { + "epoch": 18.91, + "learning_rate": 1.109471609804163e-06, + "loss": 0.3224, + "step": 51214 + }, + { + "epoch": 18.91, + "learning_rate": 1.1094346594408177e-06, + "loss": 0.1425, + "step": 51215 + }, + { + "epoch": 18.91, + "learning_rate": 1.1093977090774726e-06, + "loss": 0.3986, + "step": 51216 + }, + { + "epoch": 18.91, + "learning_rate": 1.1093607587141274e-06, + "loss": 0.1614, + "step": 51217 + }, + { + "epoch": 18.91, + "learning_rate": 1.109323808350782e-06, + "loss": 0.1909, + "step": 51218 + }, + { + "epoch": 18.91, + "learning_rate": 1.109286857987437e-06, + "loss": 0.1624, + "step": 51219 + }, + { + "epoch": 18.91, + "learning_rate": 1.1092499076240917e-06, + "loss": 0.2018, + "step": 51220 + }, + { + "epoch": 18.91, + "learning_rate": 1.1092129572607464e-06, + "loss": 0.2496, + "step": 51221 + }, + { + "epoch": 18.92, + "learning_rate": 1.1091760068974012e-06, + "loss": 0.2387, + "step": 51222 + }, + { + "epoch": 18.92, + "learning_rate": 1.109139056534056e-06, + "loss": 0.1654, + "step": 51223 + }, + { + "epoch": 18.92, + "learning_rate": 1.1091021061707107e-06, + "loss": 0.2046, + "step": 51224 + }, + { + "epoch": 18.92, + "learning_rate": 1.1090651558073654e-06, + "loss": 0.1584, + "step": 51225 + }, + { + "epoch": 18.92, + "learning_rate": 1.1090282054440202e-06, + "loss": 0.2643, + "step": 51226 + }, + { + "epoch": 18.92, + "learning_rate": 1.108991255080675e-06, + "loss": 0.1425, + "step": 51227 + }, + { + "epoch": 18.92, + "learning_rate": 1.1089543047173297e-06, + "loss": 0.1319, + "step": 51228 + }, + { + "epoch": 18.92, + "learning_rate": 1.1089173543539845e-06, + "loss": 0.1555, + "step": 51229 + }, + { + "epoch": 18.92, + "learning_rate": 1.1088804039906394e-06, + "loss": 0.211, + "step": 51230 + }, + { + "epoch": 18.92, + "learning_rate": 1.108843453627294e-06, + "loss": 0.2652, + "step": 51231 + }, + { + "epoch": 18.92, + "learning_rate": 1.1088065032639489e-06, + "loss": 0.1663, + "step": 51232 + }, + { + "epoch": 18.92, + "learning_rate": 1.1087695529006037e-06, + "loss": 0.1323, + "step": 51233 + }, + { + "epoch": 18.92, + "learning_rate": 1.1087326025372584e-06, + "loss": 0.3673, + "step": 51234 + }, + { + "epoch": 18.92, + "learning_rate": 1.108695652173913e-06, + "loss": 0.2258, + "step": 51235 + }, + { + "epoch": 18.92, + "learning_rate": 1.1086587018105678e-06, + "loss": 0.1861, + "step": 51236 + }, + { + "epoch": 18.92, + "learning_rate": 1.1086217514472225e-06, + "loss": 0.2604, + "step": 51237 + }, + { + "epoch": 18.92, + "learning_rate": 1.1085848010838773e-06, + "loss": 0.1352, + "step": 51238 + }, + { + "epoch": 18.92, + "learning_rate": 1.1085478507205322e-06, + "loss": 0.1678, + "step": 51239 + }, + { + "epoch": 18.92, + "learning_rate": 1.1085109003571868e-06, + "loss": 0.3009, + "step": 51240 + }, + { + "epoch": 18.92, + "learning_rate": 1.1084739499938417e-06, + "loss": 0.1669, + "step": 51241 + }, + { + "epoch": 18.92, + "learning_rate": 1.1084369996304965e-06, + "loss": 0.3104, + "step": 51242 + }, + { + "epoch": 18.92, + "learning_rate": 1.1084000492671511e-06, + "loss": 0.2427, + "step": 51243 + }, + { + "epoch": 18.92, + "learning_rate": 1.108363098903806e-06, + "loss": 0.3187, + "step": 51244 + }, + { + "epoch": 18.92, + "learning_rate": 1.1083261485404606e-06, + "loss": 0.1699, + "step": 51245 + }, + { + "epoch": 18.92, + "learning_rate": 1.1082891981771155e-06, + "loss": 0.2813, + "step": 51246 + }, + { + "epoch": 18.92, + "learning_rate": 1.1082522478137701e-06, + "loss": 0.1803, + "step": 51247 + }, + { + "epoch": 18.92, + "learning_rate": 1.108215297450425e-06, + "loss": 0.2647, + "step": 51248 + }, + { + "epoch": 18.93, + "learning_rate": 1.1081783470870798e-06, + "loss": 0.1913, + "step": 51249 + }, + { + "epoch": 18.93, + "learning_rate": 1.1081413967237344e-06, + "loss": 0.1161, + "step": 51250 + }, + { + "epoch": 18.93, + "learning_rate": 1.1081044463603893e-06, + "loss": 0.189, + "step": 51251 + }, + { + "epoch": 18.93, + "learning_rate": 1.1080674959970441e-06, + "loss": 0.2706, + "step": 51252 + }, + { + "epoch": 18.93, + "learning_rate": 1.1080305456336988e-06, + "loss": 0.2966, + "step": 51253 + }, + { + "epoch": 18.93, + "learning_rate": 1.1079935952703536e-06, + "loss": 0.3178, + "step": 51254 + }, + { + "epoch": 18.93, + "learning_rate": 1.1079566449070083e-06, + "loss": 0.1683, + "step": 51255 + }, + { + "epoch": 18.93, + "learning_rate": 1.107919694543663e-06, + "loss": 0.1561, + "step": 51256 + }, + { + "epoch": 18.93, + "learning_rate": 1.1078827441803177e-06, + "loss": 0.1674, + "step": 51257 + }, + { + "epoch": 18.93, + "learning_rate": 1.1078457938169726e-06, + "loss": 0.3058, + "step": 51258 + }, + { + "epoch": 18.93, + "learning_rate": 1.1078088434536272e-06, + "loss": 0.1536, + "step": 51259 + }, + { + "epoch": 18.93, + "learning_rate": 1.107771893090282e-06, + "loss": 0.2726, + "step": 51260 + }, + { + "epoch": 18.93, + "learning_rate": 1.107734942726937e-06, + "loss": 0.3451, + "step": 51261 + }, + { + "epoch": 18.93, + "learning_rate": 1.1076979923635916e-06, + "loss": 0.3115, + "step": 51262 + }, + { + "epoch": 18.93, + "learning_rate": 1.1076610420002464e-06, + "loss": 0.1194, + "step": 51263 + }, + { + "epoch": 18.93, + "learning_rate": 1.1076240916369013e-06, + "loss": 0.2873, + "step": 51264 + }, + { + "epoch": 18.93, + "learning_rate": 1.1075871412735559e-06, + "loss": 0.2719, + "step": 51265 + }, + { + "epoch": 18.93, + "learning_rate": 1.1075501909102105e-06, + "loss": 0.2208, + "step": 51266 + }, + { + "epoch": 18.93, + "learning_rate": 1.1075132405468654e-06, + "loss": 0.1656, + "step": 51267 + }, + { + "epoch": 18.93, + "learning_rate": 1.1074762901835202e-06, + "loss": 0.2008, + "step": 51268 + }, + { + "epoch": 18.93, + "learning_rate": 1.1074393398201749e-06, + "loss": 0.2666, + "step": 51269 + }, + { + "epoch": 18.93, + "learning_rate": 1.1074023894568297e-06, + "loss": 0.2633, + "step": 51270 + }, + { + "epoch": 18.93, + "learning_rate": 1.1073654390934846e-06, + "loss": 0.2481, + "step": 51271 + }, + { + "epoch": 18.93, + "learning_rate": 1.1073284887301392e-06, + "loss": 0.2978, + "step": 51272 + }, + { + "epoch": 18.93, + "learning_rate": 1.107291538366794e-06, + "loss": 0.2141, + "step": 51273 + }, + { + "epoch": 18.93, + "learning_rate": 1.1072545880034489e-06, + "loss": 0.1715, + "step": 51274 + }, + { + "epoch": 18.93, + "learning_rate": 1.1072176376401033e-06, + "loss": 0.0929, + "step": 51275 + }, + { + "epoch": 18.94, + "learning_rate": 1.1071806872767582e-06, + "loss": 0.3448, + "step": 51276 + }, + { + "epoch": 18.94, + "learning_rate": 1.107143736913413e-06, + "loss": 0.2708, + "step": 51277 + }, + { + "epoch": 18.94, + "learning_rate": 1.1071067865500676e-06, + "loss": 0.2554, + "step": 51278 + }, + { + "epoch": 18.94, + "learning_rate": 1.1070698361867225e-06, + "loss": 0.1817, + "step": 51279 + }, + { + "epoch": 18.94, + "learning_rate": 1.1070328858233773e-06, + "loss": 0.2744, + "step": 51280 + }, + { + "epoch": 18.94, + "learning_rate": 1.106995935460032e-06, + "loss": 0.2386, + "step": 51281 + }, + { + "epoch": 18.94, + "learning_rate": 1.1069589850966868e-06, + "loss": 0.1126, + "step": 51282 + }, + { + "epoch": 18.94, + "learning_rate": 1.1069220347333417e-06, + "loss": 0.3301, + "step": 51283 + }, + { + "epoch": 18.94, + "learning_rate": 1.1068850843699963e-06, + "loss": 0.2656, + "step": 51284 + }, + { + "epoch": 18.94, + "learning_rate": 1.106848134006651e-06, + "loss": 0.1871, + "step": 51285 + }, + { + "epoch": 18.94, + "learning_rate": 1.1068111836433058e-06, + "loss": 0.2039, + "step": 51286 + }, + { + "epoch": 18.94, + "learning_rate": 1.1067742332799606e-06, + "loss": 0.2025, + "step": 51287 + }, + { + "epoch": 18.94, + "learning_rate": 1.1067372829166153e-06, + "loss": 0.2932, + "step": 51288 + }, + { + "epoch": 18.94, + "learning_rate": 1.1067003325532701e-06, + "loss": 0.1967, + "step": 51289 + }, + { + "epoch": 18.94, + "learning_rate": 1.106663382189925e-06, + "loss": 0.2692, + "step": 51290 + }, + { + "epoch": 18.94, + "learning_rate": 1.1066264318265796e-06, + "loss": 0.208, + "step": 51291 + }, + { + "epoch": 18.94, + "learning_rate": 1.1065894814632345e-06, + "loss": 0.2817, + "step": 51292 + }, + { + "epoch": 18.94, + "learning_rate": 1.1065525310998893e-06, + "loss": 0.2534, + "step": 51293 + }, + { + "epoch": 18.94, + "learning_rate": 1.106515580736544e-06, + "loss": 0.29, + "step": 51294 + }, + { + "epoch": 18.94, + "learning_rate": 1.1064786303731986e-06, + "loss": 0.2609, + "step": 51295 + }, + { + "epoch": 18.94, + "learning_rate": 1.1064416800098534e-06, + "loss": 0.405, + "step": 51296 + }, + { + "epoch": 18.94, + "learning_rate": 1.1064047296465083e-06, + "loss": 0.1954, + "step": 51297 + }, + { + "epoch": 18.94, + "learning_rate": 1.106367779283163e-06, + "loss": 0.221, + "step": 51298 + }, + { + "epoch": 18.94, + "learning_rate": 1.1063308289198178e-06, + "loss": 0.2185, + "step": 51299 + }, + { + "epoch": 18.94, + "learning_rate": 1.1062938785564726e-06, + "loss": 0.2786, + "step": 51300 + }, + { + "epoch": 18.94, + "learning_rate": 1.1062569281931273e-06, + "loss": 0.1864, + "step": 51301 + }, + { + "epoch": 18.94, + "learning_rate": 1.106219977829782e-06, + "loss": 0.2283, + "step": 51302 + }, + { + "epoch": 18.94, + "learning_rate": 1.106183027466437e-06, + "loss": 0.1497, + "step": 51303 + }, + { + "epoch": 18.95, + "learning_rate": 1.1061460771030916e-06, + "loss": 0.0981, + "step": 51304 + }, + { + "epoch": 18.95, + "learning_rate": 1.1061091267397462e-06, + "loss": 0.1873, + "step": 51305 + }, + { + "epoch": 18.95, + "learning_rate": 1.106072176376401e-06, + "loss": 0.1384, + "step": 51306 + }, + { + "epoch": 18.95, + "learning_rate": 1.1060352260130557e-06, + "loss": 0.1679, + "step": 51307 + }, + { + "epoch": 18.95, + "learning_rate": 1.1059982756497106e-06, + "loss": 0.2208, + "step": 51308 + }, + { + "epoch": 18.95, + "learning_rate": 1.1059613252863654e-06, + "loss": 0.261, + "step": 51309 + }, + { + "epoch": 18.95, + "learning_rate": 1.10592437492302e-06, + "loss": 0.2287, + "step": 51310 + }, + { + "epoch": 18.95, + "learning_rate": 1.1058874245596749e-06, + "loss": 0.3563, + "step": 51311 + }, + { + "epoch": 18.95, + "learning_rate": 1.1058504741963297e-06, + "loss": 0.4167, + "step": 51312 + }, + { + "epoch": 18.95, + "learning_rate": 1.1058135238329844e-06, + "loss": 0.1908, + "step": 51313 + }, + { + "epoch": 18.95, + "learning_rate": 1.1057765734696392e-06, + "loss": 0.1416, + "step": 51314 + }, + { + "epoch": 18.95, + "learning_rate": 1.1057396231062939e-06, + "loss": 0.1629, + "step": 51315 + }, + { + "epoch": 18.95, + "learning_rate": 1.1057026727429487e-06, + "loss": 0.2288, + "step": 51316 + }, + { + "epoch": 18.95, + "learning_rate": 1.1056657223796033e-06, + "loss": 0.1767, + "step": 51317 + }, + { + "epoch": 18.95, + "learning_rate": 1.1056287720162582e-06, + "loss": 0.2535, + "step": 51318 + }, + { + "epoch": 18.95, + "learning_rate": 1.105591821652913e-06, + "loss": 0.2461, + "step": 51319 + }, + { + "epoch": 18.95, + "learning_rate": 1.1055548712895677e-06, + "loss": 0.1362, + "step": 51320 + }, + { + "epoch": 18.95, + "learning_rate": 1.1055179209262225e-06, + "loss": 0.3242, + "step": 51321 + }, + { + "epoch": 18.95, + "learning_rate": 1.1054809705628774e-06, + "loss": 0.0838, + "step": 51322 + }, + { + "epoch": 18.95, + "learning_rate": 1.105444020199532e-06, + "loss": 0.1939, + "step": 51323 + }, + { + "epoch": 18.95, + "learning_rate": 1.1054070698361866e-06, + "loss": 0.1266, + "step": 51324 + }, + { + "epoch": 18.95, + "learning_rate": 1.1053701194728415e-06, + "loss": 0.1289, + "step": 51325 + }, + { + "epoch": 18.95, + "learning_rate": 1.1053331691094961e-06, + "loss": 0.2323, + "step": 51326 + }, + { + "epoch": 18.95, + "learning_rate": 1.105296218746151e-06, + "loss": 0.2386, + "step": 51327 + }, + { + "epoch": 18.95, + "learning_rate": 1.1052592683828058e-06, + "loss": 0.1588, + "step": 51328 + }, + { + "epoch": 18.95, + "learning_rate": 1.1052223180194605e-06, + "loss": 0.3032, + "step": 51329 + }, + { + "epoch": 18.95, + "learning_rate": 1.1051853676561153e-06, + "loss": 0.3216, + "step": 51330 + }, + { + "epoch": 18.96, + "learning_rate": 1.1051484172927702e-06, + "loss": 0.2359, + "step": 51331 + }, + { + "epoch": 18.96, + "learning_rate": 1.1051114669294248e-06, + "loss": 0.2641, + "step": 51332 + }, + { + "epoch": 18.96, + "learning_rate": 1.1050745165660796e-06, + "loss": 0.3443, + "step": 51333 + }, + { + "epoch": 18.96, + "learning_rate": 1.1050375662027343e-06, + "loss": 0.2665, + "step": 51334 + }, + { + "epoch": 18.96, + "learning_rate": 1.1050006158393891e-06, + "loss": 0.2788, + "step": 51335 + }, + { + "epoch": 18.96, + "learning_rate": 1.1049636654760438e-06, + "loss": 0.0797, + "step": 51336 + }, + { + "epoch": 18.96, + "learning_rate": 1.1049267151126986e-06, + "loss": 0.3498, + "step": 51337 + }, + { + "epoch": 18.96, + "learning_rate": 1.1048897647493535e-06, + "loss": 0.3487, + "step": 51338 + }, + { + "epoch": 18.96, + "learning_rate": 1.104852814386008e-06, + "loss": 0.2937, + "step": 51339 + }, + { + "epoch": 18.96, + "learning_rate": 1.104815864022663e-06, + "loss": 0.4496, + "step": 51340 + }, + { + "epoch": 18.96, + "learning_rate": 1.1047789136593178e-06, + "loss": 0.2162, + "step": 51341 + }, + { + "epoch": 18.96, + "learning_rate": 1.1047419632959724e-06, + "loss": 0.1314, + "step": 51342 + }, + { + "epoch": 18.96, + "learning_rate": 1.1047050129326273e-06, + "loss": 0.2414, + "step": 51343 + }, + { + "epoch": 18.96, + "learning_rate": 1.104668062569282e-06, + "loss": 0.23, + "step": 51344 + }, + { + "epoch": 18.96, + "learning_rate": 1.1046311122059368e-06, + "loss": 0.216, + "step": 51345 + }, + { + "epoch": 18.96, + "learning_rate": 1.1045941618425914e-06, + "loss": 0.1863, + "step": 51346 + }, + { + "epoch": 18.96, + "learning_rate": 1.1045572114792462e-06, + "loss": 0.234, + "step": 51347 + }, + { + "epoch": 18.96, + "learning_rate": 1.1045202611159009e-06, + "loss": 0.1164, + "step": 51348 + }, + { + "epoch": 18.96, + "learning_rate": 1.1044833107525557e-06, + "loss": 0.1599, + "step": 51349 + }, + { + "epoch": 18.96, + "learning_rate": 1.1044463603892106e-06, + "loss": 0.2935, + "step": 51350 + }, + { + "epoch": 18.96, + "learning_rate": 1.1044094100258652e-06, + "loss": 0.2885, + "step": 51351 + }, + { + "epoch": 18.96, + "learning_rate": 1.10437245966252e-06, + "loss": 0.3505, + "step": 51352 + }, + { + "epoch": 18.96, + "learning_rate": 1.104335509299175e-06, + "loss": 0.3197, + "step": 51353 + }, + { + "epoch": 18.96, + "learning_rate": 1.1042985589358295e-06, + "loss": 0.1876, + "step": 51354 + }, + { + "epoch": 18.96, + "learning_rate": 1.1042616085724842e-06, + "loss": 0.2545, + "step": 51355 + }, + { + "epoch": 18.96, + "learning_rate": 1.104224658209139e-06, + "loss": 0.1637, + "step": 51356 + }, + { + "epoch": 18.96, + "learning_rate": 1.1041877078457939e-06, + "loss": 0.1853, + "step": 51357 + }, + { + "epoch": 18.97, + "learning_rate": 1.1041507574824485e-06, + "loss": 0.1236, + "step": 51358 + }, + { + "epoch": 18.97, + "learning_rate": 1.1041138071191034e-06, + "loss": 0.2718, + "step": 51359 + }, + { + "epoch": 18.97, + "learning_rate": 1.1040768567557582e-06, + "loss": 0.2445, + "step": 51360 + }, + { + "epoch": 18.97, + "learning_rate": 1.1040399063924128e-06, + "loss": 0.3263, + "step": 51361 + }, + { + "epoch": 18.97, + "learning_rate": 1.1040029560290677e-06, + "loss": 0.2184, + "step": 51362 + }, + { + "epoch": 18.97, + "learning_rate": 1.1039660056657225e-06, + "loss": 0.2268, + "step": 51363 + }, + { + "epoch": 18.97, + "learning_rate": 1.1039290553023772e-06, + "loss": 0.1971, + "step": 51364 + }, + { + "epoch": 18.97, + "learning_rate": 1.1038921049390318e-06, + "loss": 0.2433, + "step": 51365 + }, + { + "epoch": 18.97, + "learning_rate": 1.1038551545756867e-06, + "loss": 0.2044, + "step": 51366 + }, + { + "epoch": 18.97, + "learning_rate": 1.1038182042123415e-06, + "loss": 0.192, + "step": 51367 + }, + { + "epoch": 18.97, + "learning_rate": 1.1037812538489962e-06, + "loss": 0.3011, + "step": 51368 + }, + { + "epoch": 18.97, + "learning_rate": 1.103744303485651e-06, + "loss": 0.1245, + "step": 51369 + }, + { + "epoch": 18.97, + "learning_rate": 1.1037073531223058e-06, + "loss": 0.2331, + "step": 51370 + }, + { + "epoch": 18.97, + "learning_rate": 1.1036704027589605e-06, + "loss": 0.1908, + "step": 51371 + }, + { + "epoch": 18.97, + "learning_rate": 1.1036334523956153e-06, + "loss": 0.1656, + "step": 51372 + }, + { + "epoch": 18.97, + "learning_rate": 1.1035965020322702e-06, + "loss": 0.33, + "step": 51373 + }, + { + "epoch": 18.97, + "learning_rate": 1.1035595516689246e-06, + "loss": 0.1739, + "step": 51374 + }, + { + "epoch": 18.97, + "learning_rate": 1.1035226013055795e-06, + "loss": 0.1582, + "step": 51375 + }, + { + "epoch": 18.97, + "learning_rate": 1.1034856509422343e-06, + "loss": 0.2544, + "step": 51376 + }, + { + "epoch": 18.97, + "learning_rate": 1.103448700578889e-06, + "loss": 0.112, + "step": 51377 + }, + { + "epoch": 18.97, + "learning_rate": 1.1034117502155438e-06, + "loss": 0.1527, + "step": 51378 + }, + { + "epoch": 18.97, + "learning_rate": 1.1033747998521986e-06, + "loss": 0.1174, + "step": 51379 + }, + { + "epoch": 18.97, + "learning_rate": 1.1033378494888533e-06, + "loss": 0.3061, + "step": 51380 + }, + { + "epoch": 18.97, + "learning_rate": 1.1033008991255081e-06, + "loss": 0.2445, + "step": 51381 + }, + { + "epoch": 18.97, + "learning_rate": 1.103263948762163e-06, + "loss": 0.1352, + "step": 51382 + }, + { + "epoch": 18.97, + "learning_rate": 1.1032269983988176e-06, + "loss": 0.3038, + "step": 51383 + }, + { + "epoch": 18.97, + "learning_rate": 1.1031900480354722e-06, + "loss": 0.257, + "step": 51384 + }, + { + "epoch": 18.98, + "learning_rate": 1.103153097672127e-06, + "loss": 0.1343, + "step": 51385 + }, + { + "epoch": 18.98, + "learning_rate": 1.103116147308782e-06, + "loss": 0.1602, + "step": 51386 + }, + { + "epoch": 18.98, + "learning_rate": 1.1030791969454366e-06, + "loss": 0.247, + "step": 51387 + }, + { + "epoch": 18.98, + "learning_rate": 1.1030422465820914e-06, + "loss": 0.3468, + "step": 51388 + }, + { + "epoch": 18.98, + "learning_rate": 1.1030052962187463e-06, + "loss": 0.1478, + "step": 51389 + }, + { + "epoch": 18.98, + "learning_rate": 1.102968345855401e-06, + "loss": 0.1524, + "step": 51390 + }, + { + "epoch": 18.98, + "learning_rate": 1.1029313954920558e-06, + "loss": 0.2423, + "step": 51391 + }, + { + "epoch": 18.98, + "learning_rate": 1.1028944451287106e-06, + "loss": 0.2155, + "step": 51392 + }, + { + "epoch": 18.98, + "learning_rate": 1.1028574947653652e-06, + "loss": 0.2208, + "step": 51393 + }, + { + "epoch": 18.98, + "learning_rate": 1.1028205444020199e-06, + "loss": 0.2675, + "step": 51394 + }, + { + "epoch": 18.98, + "learning_rate": 1.1027835940386747e-06, + "loss": 0.1767, + "step": 51395 + }, + { + "epoch": 18.98, + "learning_rate": 1.1027466436753294e-06, + "loss": 0.2183, + "step": 51396 + }, + { + "epoch": 18.98, + "learning_rate": 1.1027096933119842e-06, + "loss": 0.3572, + "step": 51397 + }, + { + "epoch": 18.98, + "learning_rate": 1.102672742948639e-06, + "loss": 0.3311, + "step": 51398 + }, + { + "epoch": 18.98, + "learning_rate": 1.1026357925852937e-06, + "loss": 0.2355, + "step": 51399 + }, + { + "epoch": 18.98, + "learning_rate": 1.1025988422219485e-06, + "loss": 0.2181, + "step": 51400 + }, + { + "epoch": 18.98, + "learning_rate": 1.1025618918586034e-06, + "loss": 0.1228, + "step": 51401 + }, + { + "epoch": 18.98, + "learning_rate": 1.102524941495258e-06, + "loss": 0.2144, + "step": 51402 + }, + { + "epoch": 18.98, + "learning_rate": 1.1024879911319129e-06, + "loss": 0.2688, + "step": 51403 + }, + { + "epoch": 18.98, + "learning_rate": 1.1024510407685675e-06, + "loss": 0.1795, + "step": 51404 + }, + { + "epoch": 18.98, + "learning_rate": 1.1024140904052224e-06, + "loss": 0.3209, + "step": 51405 + }, + { + "epoch": 18.98, + "learning_rate": 1.102377140041877e-06, + "loss": 0.3628, + "step": 51406 + }, + { + "epoch": 18.98, + "learning_rate": 1.1023401896785318e-06, + "loss": 0.2457, + "step": 51407 + }, + { + "epoch": 18.98, + "learning_rate": 1.1023032393151867e-06, + "loss": 0.1642, + "step": 51408 + }, + { + "epoch": 18.98, + "learning_rate": 1.1022662889518413e-06, + "loss": 0.1826, + "step": 51409 + }, + { + "epoch": 18.98, + "learning_rate": 1.1022293385884962e-06, + "loss": 0.1794, + "step": 51410 + }, + { + "epoch": 18.98, + "learning_rate": 1.102192388225151e-06, + "loss": 0.2499, + "step": 51411 + }, + { + "epoch": 18.99, + "learning_rate": 1.1021554378618057e-06, + "loss": 0.1634, + "step": 51412 + }, + { + "epoch": 18.99, + "learning_rate": 1.1021184874984605e-06, + "loss": 0.1921, + "step": 51413 + }, + { + "epoch": 18.99, + "learning_rate": 1.1020815371351151e-06, + "loss": 0.2045, + "step": 51414 + }, + { + "epoch": 18.99, + "learning_rate": 1.10204458677177e-06, + "loss": 0.2033, + "step": 51415 + }, + { + "epoch": 18.99, + "learning_rate": 1.1020076364084246e-06, + "loss": 0.1457, + "step": 51416 + }, + { + "epoch": 18.99, + "learning_rate": 1.1019706860450795e-06, + "loss": 0.3377, + "step": 51417 + }, + { + "epoch": 18.99, + "learning_rate": 1.1019337356817343e-06, + "loss": 0.2155, + "step": 51418 + }, + { + "epoch": 18.99, + "learning_rate": 1.101896785318389e-06, + "loss": 0.2957, + "step": 51419 + }, + { + "epoch": 18.99, + "learning_rate": 1.1018598349550438e-06, + "loss": 0.3556, + "step": 51420 + }, + { + "epoch": 18.99, + "learning_rate": 1.1018228845916987e-06, + "loss": 0.1289, + "step": 51421 + }, + { + "epoch": 18.99, + "learning_rate": 1.1017859342283533e-06, + "loss": 0.2045, + "step": 51422 + }, + { + "epoch": 18.99, + "learning_rate": 1.1017489838650081e-06, + "loss": 0.1741, + "step": 51423 + }, + { + "epoch": 18.99, + "learning_rate": 1.1017120335016628e-06, + "loss": 0.2507, + "step": 51424 + }, + { + "epoch": 18.99, + "learning_rate": 1.1016750831383174e-06, + "loss": 0.265, + "step": 51425 + }, + { + "epoch": 18.99, + "learning_rate": 1.1016381327749723e-06, + "loss": 0.1691, + "step": 51426 + }, + { + "epoch": 18.99, + "learning_rate": 1.1016011824116271e-06, + "loss": 0.2976, + "step": 51427 + }, + { + "epoch": 18.99, + "learning_rate": 1.1015642320482817e-06, + "loss": 0.1836, + "step": 51428 + }, + { + "epoch": 18.99, + "learning_rate": 1.1015272816849366e-06, + "loss": 0.1796, + "step": 51429 + }, + { + "epoch": 18.99, + "learning_rate": 1.1014903313215914e-06, + "loss": 0.2946, + "step": 51430 + }, + { + "epoch": 18.99, + "learning_rate": 1.101453380958246e-06, + "loss": 0.3951, + "step": 51431 + }, + { + "epoch": 18.99, + "learning_rate": 1.101416430594901e-06, + "loss": 0.302, + "step": 51432 + }, + { + "epoch": 18.99, + "learning_rate": 1.1013794802315558e-06, + "loss": 0.2216, + "step": 51433 + }, + { + "epoch": 18.99, + "learning_rate": 1.1013425298682104e-06, + "loss": 0.336, + "step": 51434 + }, + { + "epoch": 18.99, + "learning_rate": 1.101305579504865e-06, + "loss": 0.179, + "step": 51435 + }, + { + "epoch": 18.99, + "learning_rate": 1.10126862914152e-06, + "loss": 0.1772, + "step": 51436 + }, + { + "epoch": 18.99, + "learning_rate": 1.1012316787781747e-06, + "loss": 0.1705, + "step": 51437 + }, + { + "epoch": 18.99, + "learning_rate": 1.1011947284148294e-06, + "loss": 0.1153, + "step": 51438 + }, + { + "epoch": 19.0, + "learning_rate": 1.1011577780514842e-06, + "loss": 0.1674, + "step": 51439 + }, + { + "epoch": 19.0, + "learning_rate": 1.101120827688139e-06, + "loss": 0.1695, + "step": 51440 + }, + { + "epoch": 19.0, + "learning_rate": 1.1010838773247937e-06, + "loss": 0.2427, + "step": 51441 + }, + { + "epoch": 19.0, + "learning_rate": 1.1010469269614486e-06, + "loss": 0.1998, + "step": 51442 + }, + { + "epoch": 19.0, + "learning_rate": 1.1010099765981034e-06, + "loss": 0.1706, + "step": 51443 + }, + { + "epoch": 19.0, + "learning_rate": 1.1009730262347578e-06, + "loss": 0.2172, + "step": 51444 + }, + { + "epoch": 19.0, + "learning_rate": 1.1009360758714127e-06, + "loss": 0.2513, + "step": 51445 + }, + { + "epoch": 19.0, + "learning_rate": 1.1008991255080675e-06, + "loss": 0.3305, + "step": 51446 + }, + { + "epoch": 19.0, + "learning_rate": 1.1008621751447222e-06, + "loss": 0.2327, + "step": 51447 + }, + { + "epoch": 19.0, + "learning_rate": 1.100825224781377e-06, + "loss": 0.3428, + "step": 51448 + }, + { + "epoch": 19.0, + "learning_rate": 1.1007882744180319e-06, + "loss": 0.2005, + "step": 51449 + }, + { + "epoch": 19.0, + "learning_rate": 1.1007513240546865e-06, + "loss": 0.2159, + "step": 51450 + }, + { + "epoch": 19.0, + "learning_rate": 1.1007143736913414e-06, + "loss": 0.1548, + "step": 51451 + }, + { + "epoch": 19.0, + "learning_rate": 1.1006774233279962e-06, + "loss": 0.0603, + "step": 51452 + }, + { + "epoch": 19.0, + "eval_accuracy": 0.8628157177817796, + "eval_loss": 0.3613313138484955, + "eval_runtime": 25.1223, + "eval_samples_per_second": 1724.128, + "eval_steps_per_second": 26.948, + "step": 51452 + }, + { + "epoch": 19.0, + "learning_rate": 1.1006404729646508e-06, + "loss": 0.256, + "step": 51453 + }, + { + "epoch": 19.0, + "learning_rate": 1.1006035226013055e-06, + "loss": 0.1757, + "step": 51454 + }, + { + "epoch": 19.0, + "learning_rate": 1.1005665722379603e-06, + "loss": 0.1678, + "step": 51455 + }, + { + "epoch": 19.0, + "learning_rate": 1.1005296218746152e-06, + "loss": 0.1623, + "step": 51456 + }, + { + "epoch": 19.0, + "learning_rate": 1.1004926715112698e-06, + "loss": 0.3234, + "step": 51457 + }, + { + "epoch": 19.0, + "learning_rate": 1.1004557211479247e-06, + "loss": 0.1478, + "step": 51458 + }, + { + "epoch": 19.0, + "learning_rate": 1.1004187707845795e-06, + "loss": 0.2328, + "step": 51459 + }, + { + "epoch": 19.0, + "learning_rate": 1.1003818204212341e-06, + "loss": 0.1887, + "step": 51460 + }, + { + "epoch": 19.0, + "learning_rate": 1.100344870057889e-06, + "loss": 0.1592, + "step": 51461 + }, + { + "epoch": 19.0, + "learning_rate": 1.1003079196945438e-06, + "loss": 0.3243, + "step": 51462 + }, + { + "epoch": 19.0, + "learning_rate": 1.1002709693311985e-06, + "loss": 0.2083, + "step": 51463 + }, + { + "epoch": 19.0, + "learning_rate": 1.1002340189678531e-06, + "loss": 0.2618, + "step": 51464 + }, + { + "epoch": 19.0, + "learning_rate": 1.100197068604508e-06, + "loss": 0.2002, + "step": 51465 + }, + { + "epoch": 19.01, + "learning_rate": 1.1001601182411626e-06, + "loss": 0.1227, + "step": 51466 + }, + { + "epoch": 19.01, + "learning_rate": 1.1001231678778174e-06, + "loss": 0.2059, + "step": 51467 + }, + { + "epoch": 19.01, + "learning_rate": 1.1000862175144723e-06, + "loss": 0.2728, + "step": 51468 + }, + { + "epoch": 19.01, + "learning_rate": 1.100049267151127e-06, + "loss": 0.1878, + "step": 51469 + }, + { + "epoch": 19.01, + "learning_rate": 1.1000123167877818e-06, + "loss": 0.2172, + "step": 51470 + }, + { + "epoch": 19.01, + "learning_rate": 1.0999753664244366e-06, + "loss": 0.2234, + "step": 51471 + }, + { + "epoch": 19.01, + "learning_rate": 1.0999384160610913e-06, + "loss": 0.1694, + "step": 51472 + }, + { + "epoch": 19.01, + "learning_rate": 1.0999014656977461e-06, + "loss": 0.2401, + "step": 51473 + }, + { + "epoch": 19.01, + "learning_rate": 1.0998645153344007e-06, + "loss": 0.3955, + "step": 51474 + }, + { + "epoch": 19.01, + "learning_rate": 1.0998275649710556e-06, + "loss": 0.2201, + "step": 51475 + }, + { + "epoch": 19.01, + "learning_rate": 1.0997906146077102e-06, + "loss": 0.2425, + "step": 51476 + }, + { + "epoch": 19.01, + "learning_rate": 1.099753664244365e-06, + "loss": 0.2227, + "step": 51477 + }, + { + "epoch": 19.01, + "learning_rate": 1.09971671388102e-06, + "loss": 0.2178, + "step": 51478 + }, + { + "epoch": 19.01, + "learning_rate": 1.0996797635176746e-06, + "loss": 0.1609, + "step": 51479 + }, + { + "epoch": 19.01, + "learning_rate": 1.0996428131543294e-06, + "loss": 0.221, + "step": 51480 + }, + { + "epoch": 19.01, + "learning_rate": 1.0996058627909843e-06, + "loss": 0.2273, + "step": 51481 + }, + { + "epoch": 19.01, + "learning_rate": 1.0995689124276389e-06, + "loss": 0.2407, + "step": 51482 + }, + { + "epoch": 19.01, + "learning_rate": 1.0995319620642937e-06, + "loss": 0.3123, + "step": 51483 + }, + { + "epoch": 19.01, + "learning_rate": 1.0994950117009484e-06, + "loss": 0.1658, + "step": 51484 + }, + { + "epoch": 19.01, + "learning_rate": 1.0994580613376032e-06, + "loss": 0.1538, + "step": 51485 + }, + { + "epoch": 19.01, + "learning_rate": 1.0994211109742579e-06, + "loss": 0.1547, + "step": 51486 + }, + { + "epoch": 19.01, + "learning_rate": 1.0993841606109127e-06, + "loss": 0.3251, + "step": 51487 + }, + { + "epoch": 19.01, + "learning_rate": 1.0993472102475676e-06, + "loss": 0.2675, + "step": 51488 + }, + { + "epoch": 19.01, + "learning_rate": 1.0993102598842222e-06, + "loss": 0.1367, + "step": 51489 + }, + { + "epoch": 19.01, + "learning_rate": 1.099273309520877e-06, + "loss": 0.1367, + "step": 51490 + }, + { + "epoch": 19.01, + "learning_rate": 1.0992363591575319e-06, + "loss": 0.1723, + "step": 51491 + }, + { + "epoch": 19.01, + "learning_rate": 1.0991994087941865e-06, + "loss": 0.1312, + "step": 51492 + }, + { + "epoch": 19.02, + "learning_rate": 1.0991624584308414e-06, + "loss": 0.1578, + "step": 51493 + }, + { + "epoch": 19.02, + "learning_rate": 1.099125508067496e-06, + "loss": 0.1602, + "step": 51494 + }, + { + "epoch": 19.02, + "learning_rate": 1.0990885577041506e-06, + "loss": 0.1688, + "step": 51495 + }, + { + "epoch": 19.02, + "learning_rate": 1.0990516073408055e-06, + "loss": 0.1457, + "step": 51496 + }, + { + "epoch": 19.02, + "learning_rate": 1.0990146569774603e-06, + "loss": 0.1519, + "step": 51497 + }, + { + "epoch": 19.02, + "learning_rate": 1.098977706614115e-06, + "loss": 0.256, + "step": 51498 + }, + { + "epoch": 19.02, + "learning_rate": 1.0989407562507698e-06, + "loss": 0.2991, + "step": 51499 + }, + { + "epoch": 19.02, + "learning_rate": 1.0989038058874247e-06, + "loss": 0.2567, + "step": 51500 + }, + { + "epoch": 19.02, + "learning_rate": 1.0988668555240793e-06, + "loss": 0.226, + "step": 51501 + }, + { + "epoch": 19.02, + "learning_rate": 1.0988299051607342e-06, + "loss": 0.3302, + "step": 51502 + }, + { + "epoch": 19.02, + "learning_rate": 1.098792954797389e-06, + "loss": 0.2617, + "step": 51503 + }, + { + "epoch": 19.02, + "learning_rate": 1.0987560044340436e-06, + "loss": 0.2668, + "step": 51504 + }, + { + "epoch": 19.02, + "learning_rate": 1.0987190540706983e-06, + "loss": 0.2967, + "step": 51505 + }, + { + "epoch": 19.02, + "learning_rate": 1.0986821037073531e-06, + "loss": 0.1779, + "step": 51506 + }, + { + "epoch": 19.02, + "learning_rate": 1.098645153344008e-06, + "loss": 0.2284, + "step": 51507 + }, + { + "epoch": 19.02, + "learning_rate": 1.0986082029806626e-06, + "loss": 0.1592, + "step": 51508 + }, + { + "epoch": 19.02, + "learning_rate": 1.0985712526173175e-06, + "loss": 0.2625, + "step": 51509 + }, + { + "epoch": 19.02, + "learning_rate": 1.0985343022539723e-06, + "loss": 0.1578, + "step": 51510 + }, + { + "epoch": 19.02, + "learning_rate": 1.098497351890627e-06, + "loss": 0.2625, + "step": 51511 + }, + { + "epoch": 19.02, + "learning_rate": 1.0984604015272818e-06, + "loss": 0.1814, + "step": 51512 + }, + { + "epoch": 19.02, + "learning_rate": 1.0984234511639366e-06, + "loss": 0.2688, + "step": 51513 + }, + { + "epoch": 19.02, + "learning_rate": 1.098386500800591e-06, + "loss": 0.2408, + "step": 51514 + }, + { + "epoch": 19.02, + "learning_rate": 1.098349550437246e-06, + "loss": 0.1858, + "step": 51515 + }, + { + "epoch": 19.02, + "learning_rate": 1.0983126000739008e-06, + "loss": 0.1276, + "step": 51516 + }, + { + "epoch": 19.02, + "learning_rate": 1.0982756497105554e-06, + "loss": 0.1303, + "step": 51517 + }, + { + "epoch": 19.02, + "learning_rate": 1.0982386993472103e-06, + "loss": 0.2336, + "step": 51518 + }, + { + "epoch": 19.02, + "learning_rate": 1.098201748983865e-06, + "loss": 0.2059, + "step": 51519 + }, + { + "epoch": 19.03, + "learning_rate": 1.0981647986205197e-06, + "loss": 0.1131, + "step": 51520 + }, + { + "epoch": 19.03, + "learning_rate": 1.0981278482571746e-06, + "loss": 0.1672, + "step": 51521 + }, + { + "epoch": 19.03, + "learning_rate": 1.0980908978938294e-06, + "loss": 0.1929, + "step": 51522 + }, + { + "epoch": 19.03, + "learning_rate": 1.098053947530484e-06, + "loss": 0.2116, + "step": 51523 + }, + { + "epoch": 19.03, + "learning_rate": 1.0980169971671387e-06, + "loss": 0.2279, + "step": 51524 + }, + { + "epoch": 19.03, + "learning_rate": 1.0979800468037936e-06, + "loss": 0.2153, + "step": 51525 + }, + { + "epoch": 19.03, + "learning_rate": 1.0979430964404484e-06, + "loss": 0.1852, + "step": 51526 + }, + { + "epoch": 19.03, + "learning_rate": 1.097906146077103e-06, + "loss": 0.2828, + "step": 51527 + }, + { + "epoch": 19.03, + "learning_rate": 1.0978691957137579e-06, + "loss": 0.229, + "step": 51528 + }, + { + "epoch": 19.03, + "learning_rate": 1.0978322453504127e-06, + "loss": 0.2081, + "step": 51529 + }, + { + "epoch": 19.03, + "learning_rate": 1.0977952949870674e-06, + "loss": 0.3635, + "step": 51530 + }, + { + "epoch": 19.03, + "learning_rate": 1.0977583446237222e-06, + "loss": 0.1883, + "step": 51531 + }, + { + "epoch": 19.03, + "learning_rate": 1.097721394260377e-06, + "loss": 0.2016, + "step": 51532 + }, + { + "epoch": 19.03, + "learning_rate": 1.0976844438970317e-06, + "loss": 0.0959, + "step": 51533 + }, + { + "epoch": 19.03, + "learning_rate": 1.0976474935336863e-06, + "loss": 0.2299, + "step": 51534 + }, + { + "epoch": 19.03, + "learning_rate": 1.0976105431703412e-06, + "loss": 0.3226, + "step": 51535 + }, + { + "epoch": 19.03, + "learning_rate": 1.0975735928069958e-06, + "loss": 0.2715, + "step": 51536 + }, + { + "epoch": 19.03, + "learning_rate": 1.0975366424436507e-06, + "loss": 0.1858, + "step": 51537 + }, + { + "epoch": 19.03, + "learning_rate": 1.0974996920803055e-06, + "loss": 0.2025, + "step": 51538 + }, + { + "epoch": 19.03, + "learning_rate": 1.0974627417169602e-06, + "loss": 0.3178, + "step": 51539 + }, + { + "epoch": 19.03, + "learning_rate": 1.097425791353615e-06, + "loss": 0.2088, + "step": 51540 + }, + { + "epoch": 19.03, + "learning_rate": 1.0973888409902699e-06, + "loss": 0.223, + "step": 51541 + }, + { + "epoch": 19.03, + "learning_rate": 1.0973518906269245e-06, + "loss": 0.1621, + "step": 51542 + }, + { + "epoch": 19.03, + "learning_rate": 1.0973149402635793e-06, + "loss": 0.2893, + "step": 51543 + }, + { + "epoch": 19.03, + "learning_rate": 1.097277989900234e-06, + "loss": 0.1802, + "step": 51544 + }, + { + "epoch": 19.03, + "learning_rate": 1.0972410395368888e-06, + "loss": 0.2206, + "step": 51545 + }, + { + "epoch": 19.03, + "learning_rate": 1.0972040891735435e-06, + "loss": 0.1699, + "step": 51546 + }, + { + "epoch": 19.04, + "learning_rate": 1.0971671388101983e-06, + "loss": 0.1605, + "step": 51547 + }, + { + "epoch": 19.04, + "learning_rate": 1.0971301884468532e-06, + "loss": 0.2918, + "step": 51548 + }, + { + "epoch": 19.04, + "learning_rate": 1.0970932380835078e-06, + "loss": 0.2455, + "step": 51549 + }, + { + "epoch": 19.04, + "learning_rate": 1.0970562877201626e-06, + "loss": 0.1555, + "step": 51550 + }, + { + "epoch": 19.04, + "learning_rate": 1.0970193373568175e-06, + "loss": 0.1836, + "step": 51551 + }, + { + "epoch": 19.04, + "learning_rate": 1.0969823869934721e-06, + "loss": 0.1366, + "step": 51552 + }, + { + "epoch": 19.04, + "learning_rate": 1.0969454366301268e-06, + "loss": 0.2313, + "step": 51553 + }, + { + "epoch": 19.04, + "learning_rate": 1.0969084862667816e-06, + "loss": 0.2124, + "step": 51554 + }, + { + "epoch": 19.04, + "learning_rate": 1.0968715359034365e-06, + "loss": 0.1044, + "step": 51555 + }, + { + "epoch": 19.04, + "learning_rate": 1.096834585540091e-06, + "loss": 0.1791, + "step": 51556 + }, + { + "epoch": 19.04, + "learning_rate": 1.096797635176746e-06, + "loss": 0.1819, + "step": 51557 + }, + { + "epoch": 19.04, + "learning_rate": 1.0967606848134008e-06, + "loss": 0.1371, + "step": 51558 + }, + { + "epoch": 19.04, + "learning_rate": 1.0967237344500554e-06, + "loss": 0.1488, + "step": 51559 + }, + { + "epoch": 19.04, + "learning_rate": 1.0966867840867103e-06, + "loss": 0.166, + "step": 51560 + }, + { + "epoch": 19.04, + "learning_rate": 1.0966498337233651e-06, + "loss": 0.2426, + "step": 51561 + }, + { + "epoch": 19.04, + "learning_rate": 1.0966128833600198e-06, + "loss": 0.2507, + "step": 51562 + }, + { + "epoch": 19.04, + "learning_rate": 1.0965759329966744e-06, + "loss": 0.1143, + "step": 51563 + }, + { + "epoch": 19.04, + "learning_rate": 1.0965389826333292e-06, + "loss": 0.1654, + "step": 51564 + }, + { + "epoch": 19.04, + "learning_rate": 1.0965020322699839e-06, + "loss": 0.2156, + "step": 51565 + }, + { + "epoch": 19.04, + "learning_rate": 1.0964650819066387e-06, + "loss": 0.2275, + "step": 51566 + }, + { + "epoch": 19.04, + "learning_rate": 1.0964281315432936e-06, + "loss": 0.1348, + "step": 51567 + }, + { + "epoch": 19.04, + "learning_rate": 1.0963911811799482e-06, + "loss": 0.0709, + "step": 51568 + }, + { + "epoch": 19.04, + "learning_rate": 1.096354230816603e-06, + "loss": 0.2682, + "step": 51569 + }, + { + "epoch": 19.04, + "learning_rate": 1.096317280453258e-06, + "loss": 0.2838, + "step": 51570 + }, + { + "epoch": 19.04, + "learning_rate": 1.0962803300899125e-06, + "loss": 0.2333, + "step": 51571 + }, + { + "epoch": 19.04, + "learning_rate": 1.0962433797265674e-06, + "loss": 0.2828, + "step": 51572 + }, + { + "epoch": 19.04, + "learning_rate": 1.096206429363222e-06, + "loss": 0.2545, + "step": 51573 + }, + { + "epoch": 19.05, + "learning_rate": 1.0961694789998769e-06, + "loss": 0.2898, + "step": 51574 + }, + { + "epoch": 19.05, + "learning_rate": 1.0961325286365315e-06, + "loss": 0.3022, + "step": 51575 + }, + { + "epoch": 19.05, + "learning_rate": 1.0960955782731864e-06, + "loss": 0.1852, + "step": 51576 + }, + { + "epoch": 19.05, + "learning_rate": 1.0960586279098412e-06, + "loss": 0.2817, + "step": 51577 + }, + { + "epoch": 19.05, + "learning_rate": 1.0960216775464958e-06, + "loss": 0.1778, + "step": 51578 + }, + { + "epoch": 19.05, + "learning_rate": 1.0959847271831507e-06, + "loss": 0.2217, + "step": 51579 + }, + { + "epoch": 19.05, + "learning_rate": 1.0959477768198055e-06, + "loss": 0.1731, + "step": 51580 + }, + { + "epoch": 19.05, + "learning_rate": 1.0959108264564602e-06, + "loss": 0.252, + "step": 51581 + }, + { + "epoch": 19.05, + "learning_rate": 1.095873876093115e-06, + "loss": 0.1661, + "step": 51582 + }, + { + "epoch": 19.05, + "learning_rate": 1.0958369257297697e-06, + "loss": 0.2641, + "step": 51583 + }, + { + "epoch": 19.05, + "learning_rate": 1.0957999753664243e-06, + "loss": 0.2436, + "step": 51584 + }, + { + "epoch": 19.05, + "learning_rate": 1.0957630250030792e-06, + "loss": 0.1758, + "step": 51585 + }, + { + "epoch": 19.05, + "learning_rate": 1.095726074639734e-06, + "loss": 0.3647, + "step": 51586 + }, + { + "epoch": 19.05, + "learning_rate": 1.0956891242763886e-06, + "loss": 0.2963, + "step": 51587 + }, + { + "epoch": 19.05, + "learning_rate": 1.0956521739130435e-06, + "loss": 0.285, + "step": 51588 + }, + { + "epoch": 19.05, + "learning_rate": 1.0956152235496983e-06, + "loss": 0.2671, + "step": 51589 + }, + { + "epoch": 19.05, + "learning_rate": 1.095578273186353e-06, + "loss": 0.2504, + "step": 51590 + }, + { + "epoch": 19.05, + "learning_rate": 1.0955413228230078e-06, + "loss": 0.157, + "step": 51591 + }, + { + "epoch": 19.05, + "learning_rate": 1.0955043724596627e-06, + "loss": 0.2128, + "step": 51592 + }, + { + "epoch": 19.05, + "learning_rate": 1.0954674220963173e-06, + "loss": 0.1719, + "step": 51593 + }, + { + "epoch": 19.05, + "learning_rate": 1.095430471732972e-06, + "loss": 0.219, + "step": 51594 + }, + { + "epoch": 19.05, + "learning_rate": 1.0953935213696268e-06, + "loss": 0.2259, + "step": 51595 + }, + { + "epoch": 19.05, + "learning_rate": 1.0953565710062816e-06, + "loss": 0.1716, + "step": 51596 + }, + { + "epoch": 19.05, + "learning_rate": 1.0953196206429363e-06, + "loss": 0.1961, + "step": 51597 + }, + { + "epoch": 19.05, + "learning_rate": 1.0952826702795911e-06, + "loss": 0.2156, + "step": 51598 + }, + { + "epoch": 19.05, + "learning_rate": 1.095245719916246e-06, + "loss": 0.1222, + "step": 51599 + }, + { + "epoch": 19.05, + "learning_rate": 1.0952087695529006e-06, + "loss": 0.1933, + "step": 51600 + }, + { + "epoch": 19.06, + "learning_rate": 1.0951718191895555e-06, + "loss": 0.132, + "step": 51601 + }, + { + "epoch": 19.06, + "learning_rate": 1.0951348688262103e-06, + "loss": 0.32, + "step": 51602 + }, + { + "epoch": 19.06, + "learning_rate": 1.095097918462865e-06, + "loss": 0.2061, + "step": 51603 + }, + { + "epoch": 19.06, + "learning_rate": 1.0950609680995196e-06, + "loss": 0.2498, + "step": 51604 + }, + { + "epoch": 19.06, + "learning_rate": 1.0950240177361744e-06, + "loss": 0.072, + "step": 51605 + }, + { + "epoch": 19.06, + "learning_rate": 1.0949870673728293e-06, + "loss": 0.1689, + "step": 51606 + }, + { + "epoch": 19.06, + "learning_rate": 1.094950117009484e-06, + "loss": 0.1568, + "step": 51607 + }, + { + "epoch": 19.06, + "learning_rate": 1.0949131666461388e-06, + "loss": 0.1629, + "step": 51608 + }, + { + "epoch": 19.06, + "learning_rate": 1.0948762162827936e-06, + "loss": 0.1433, + "step": 51609 + }, + { + "epoch": 19.06, + "learning_rate": 1.0948392659194482e-06, + "loss": 0.1624, + "step": 51610 + }, + { + "epoch": 19.06, + "learning_rate": 1.094802315556103e-06, + "loss": 0.1838, + "step": 51611 + }, + { + "epoch": 19.06, + "learning_rate": 1.0947653651927577e-06, + "loss": 0.0984, + "step": 51612 + }, + { + "epoch": 19.06, + "learning_rate": 1.0947284148294124e-06, + "loss": 0.1531, + "step": 51613 + }, + { + "epoch": 19.06, + "learning_rate": 1.0946914644660672e-06, + "loss": 0.3454, + "step": 51614 + }, + { + "epoch": 19.06, + "learning_rate": 1.094654514102722e-06, + "loss": 0.1685, + "step": 51615 + }, + { + "epoch": 19.06, + "learning_rate": 1.0946175637393767e-06, + "loss": 0.1052, + "step": 51616 + }, + { + "epoch": 19.06, + "learning_rate": 1.0945806133760315e-06, + "loss": 0.1883, + "step": 51617 + }, + { + "epoch": 19.06, + "learning_rate": 1.0945436630126864e-06, + "loss": 0.2815, + "step": 51618 + }, + { + "epoch": 19.06, + "learning_rate": 1.094506712649341e-06, + "loss": 0.1463, + "step": 51619 + }, + { + "epoch": 19.06, + "learning_rate": 1.0944697622859959e-06, + "loss": 0.2066, + "step": 51620 + }, + { + "epoch": 19.06, + "learning_rate": 1.0944328119226507e-06, + "loss": 0.1375, + "step": 51621 + }, + { + "epoch": 19.06, + "learning_rate": 1.0943958615593054e-06, + "loss": 0.1034, + "step": 51622 + }, + { + "epoch": 19.06, + "learning_rate": 1.09435891119596e-06, + "loss": 0.217, + "step": 51623 + }, + { + "epoch": 19.06, + "learning_rate": 1.0943219608326148e-06, + "loss": 0.1648, + "step": 51624 + }, + { + "epoch": 19.06, + "learning_rate": 1.0942850104692697e-06, + "loss": 0.2531, + "step": 51625 + }, + { + "epoch": 19.06, + "learning_rate": 1.0942480601059243e-06, + "loss": 0.1474, + "step": 51626 + }, + { + "epoch": 19.06, + "learning_rate": 1.0942111097425792e-06, + "loss": 0.2773, + "step": 51627 + }, + { + "epoch": 19.06, + "learning_rate": 1.094174159379234e-06, + "loss": 0.1221, + "step": 51628 + }, + { + "epoch": 19.07, + "learning_rate": 1.0941372090158887e-06, + "loss": 0.1054, + "step": 51629 + }, + { + "epoch": 19.07, + "learning_rate": 1.0941002586525435e-06, + "loss": 0.2527, + "step": 51630 + }, + { + "epoch": 19.07, + "learning_rate": 1.0940633082891984e-06, + "loss": 0.2302, + "step": 51631 + }, + { + "epoch": 19.07, + "learning_rate": 1.094026357925853e-06, + "loss": 0.263, + "step": 51632 + }, + { + "epoch": 19.07, + "learning_rate": 1.0939894075625076e-06, + "loss": 0.2027, + "step": 51633 + }, + { + "epoch": 19.07, + "learning_rate": 1.0939524571991625e-06, + "loss": 0.1228, + "step": 51634 + }, + { + "epoch": 19.07, + "learning_rate": 1.0939155068358171e-06, + "loss": 0.1014, + "step": 51635 + }, + { + "epoch": 19.07, + "learning_rate": 1.093878556472472e-06, + "loss": 0.3088, + "step": 51636 + }, + { + "epoch": 19.07, + "learning_rate": 1.0938416061091268e-06, + "loss": 0.1693, + "step": 51637 + }, + { + "epoch": 19.07, + "learning_rate": 1.0938046557457814e-06, + "loss": 0.1411, + "step": 51638 + }, + { + "epoch": 19.07, + "learning_rate": 1.0937677053824363e-06, + "loss": 0.2526, + "step": 51639 + }, + { + "epoch": 19.07, + "learning_rate": 1.0937307550190911e-06, + "loss": 0.1436, + "step": 51640 + }, + { + "epoch": 19.07, + "learning_rate": 1.0936938046557458e-06, + "loss": 0.1977, + "step": 51641 + }, + { + "epoch": 19.07, + "learning_rate": 1.0936568542924006e-06, + "loss": 0.1815, + "step": 51642 + }, + { + "epoch": 19.07, + "learning_rate": 1.0936199039290553e-06, + "loss": 0.1997, + "step": 51643 + }, + { + "epoch": 19.07, + "learning_rate": 1.0935829535657101e-06, + "loss": 0.3896, + "step": 51644 + }, + { + "epoch": 19.07, + "learning_rate": 1.0935460032023647e-06, + "loss": 0.2262, + "step": 51645 + }, + { + "epoch": 19.07, + "learning_rate": 1.0935090528390196e-06, + "loss": 0.3592, + "step": 51646 + }, + { + "epoch": 19.07, + "learning_rate": 1.0934721024756744e-06, + "loss": 0.4277, + "step": 51647 + }, + { + "epoch": 19.07, + "learning_rate": 1.093435152112329e-06, + "loss": 0.1769, + "step": 51648 + }, + { + "epoch": 19.07, + "learning_rate": 1.093398201748984e-06, + "loss": 0.2301, + "step": 51649 + }, + { + "epoch": 19.07, + "learning_rate": 1.0933612513856388e-06, + "loss": 0.2863, + "step": 51650 + }, + { + "epoch": 19.07, + "learning_rate": 1.0933243010222934e-06, + "loss": 0.2311, + "step": 51651 + }, + { + "epoch": 19.07, + "learning_rate": 1.0932873506589483e-06, + "loss": 0.1237, + "step": 51652 + }, + { + "epoch": 19.07, + "learning_rate": 1.093250400295603e-06, + "loss": 0.1982, + "step": 51653 + }, + { + "epoch": 19.07, + "learning_rate": 1.0932134499322575e-06, + "loss": 0.2714, + "step": 51654 + }, + { + "epoch": 19.07, + "learning_rate": 1.0931764995689124e-06, + "loss": 0.1983, + "step": 51655 + }, + { + "epoch": 19.08, + "learning_rate": 1.0931395492055672e-06, + "loss": 0.1078, + "step": 51656 + }, + { + "epoch": 19.08, + "learning_rate": 1.0931025988422219e-06, + "loss": 0.1583, + "step": 51657 + }, + { + "epoch": 19.08, + "learning_rate": 1.0930656484788767e-06, + "loss": 0.2063, + "step": 51658 + }, + { + "epoch": 19.08, + "learning_rate": 1.0930286981155316e-06, + "loss": 0.1769, + "step": 51659 + }, + { + "epoch": 19.08, + "learning_rate": 1.0929917477521862e-06, + "loss": 0.3234, + "step": 51660 + }, + { + "epoch": 19.08, + "learning_rate": 1.092954797388841e-06, + "loss": 0.1989, + "step": 51661 + }, + { + "epoch": 19.08, + "learning_rate": 1.092917847025496e-06, + "loss": 0.1587, + "step": 51662 + }, + { + "epoch": 19.08, + "learning_rate": 1.0928808966621505e-06, + "loss": 0.1998, + "step": 51663 + }, + { + "epoch": 19.08, + "learning_rate": 1.0928439462988052e-06, + "loss": 0.152, + "step": 51664 + }, + { + "epoch": 19.08, + "learning_rate": 1.09280699593546e-06, + "loss": 0.2437, + "step": 51665 + }, + { + "epoch": 19.08, + "learning_rate": 1.0927700455721149e-06, + "loss": 0.3185, + "step": 51666 + }, + { + "epoch": 19.08, + "learning_rate": 1.0927330952087695e-06, + "loss": 0.1982, + "step": 51667 + }, + { + "epoch": 19.08, + "learning_rate": 1.0926961448454244e-06, + "loss": 0.2403, + "step": 51668 + }, + { + "epoch": 19.08, + "learning_rate": 1.0926591944820792e-06, + "loss": 0.235, + "step": 51669 + }, + { + "epoch": 19.08, + "learning_rate": 1.0926222441187338e-06, + "loss": 0.3113, + "step": 51670 + }, + { + "epoch": 19.08, + "learning_rate": 1.0925852937553887e-06, + "loss": 0.109, + "step": 51671 + }, + { + "epoch": 19.08, + "learning_rate": 1.0925483433920435e-06, + "loss": 0.2676, + "step": 51672 + }, + { + "epoch": 19.08, + "learning_rate": 1.0925113930286982e-06, + "loss": 0.2431, + "step": 51673 + }, + { + "epoch": 19.08, + "learning_rate": 1.0924744426653528e-06, + "loss": 0.2302, + "step": 51674 + }, + { + "epoch": 19.08, + "learning_rate": 1.0924374923020077e-06, + "loss": 0.1854, + "step": 51675 + }, + { + "epoch": 19.08, + "learning_rate": 1.0924005419386625e-06, + "loss": 0.3856, + "step": 51676 + }, + { + "epoch": 19.08, + "learning_rate": 1.0923635915753171e-06, + "loss": 0.3147, + "step": 51677 + }, + { + "epoch": 19.08, + "learning_rate": 1.092326641211972e-06, + "loss": 0.2226, + "step": 51678 + }, + { + "epoch": 19.08, + "learning_rate": 1.0922896908486268e-06, + "loss": 0.1679, + "step": 51679 + }, + { + "epoch": 19.08, + "learning_rate": 1.0922527404852815e-06, + "loss": 0.2739, + "step": 51680 + }, + { + "epoch": 19.08, + "learning_rate": 1.0922157901219363e-06, + "loss": 0.1619, + "step": 51681 + }, + { + "epoch": 19.08, + "learning_rate": 1.0921788397585912e-06, + "loss": 0.1685, + "step": 51682 + }, + { + "epoch": 19.09, + "learning_rate": 1.0921418893952456e-06, + "loss": 0.2457, + "step": 51683 + }, + { + "epoch": 19.09, + "learning_rate": 1.0921049390319004e-06, + "loss": 0.3308, + "step": 51684 + }, + { + "epoch": 19.09, + "learning_rate": 1.0920679886685553e-06, + "loss": 0.2414, + "step": 51685 + }, + { + "epoch": 19.09, + "learning_rate": 1.09203103830521e-06, + "loss": 0.1925, + "step": 51686 + }, + { + "epoch": 19.09, + "learning_rate": 1.0919940879418648e-06, + "loss": 0.2187, + "step": 51687 + }, + { + "epoch": 19.09, + "learning_rate": 1.0919571375785196e-06, + "loss": 0.2137, + "step": 51688 + }, + { + "epoch": 19.09, + "learning_rate": 1.0919201872151743e-06, + "loss": 0.1525, + "step": 51689 + }, + { + "epoch": 19.09, + "learning_rate": 1.091883236851829e-06, + "loss": 0.3783, + "step": 51690 + }, + { + "epoch": 19.09, + "learning_rate": 1.091846286488484e-06, + "loss": 0.1751, + "step": 51691 + }, + { + "epoch": 19.09, + "learning_rate": 1.0918093361251386e-06, + "loss": 0.2163, + "step": 51692 + }, + { + "epoch": 19.09, + "learning_rate": 1.0917723857617932e-06, + "loss": 0.1645, + "step": 51693 + }, + { + "epoch": 19.09, + "learning_rate": 1.091735435398448e-06, + "loss": 0.2696, + "step": 51694 + }, + { + "epoch": 19.09, + "learning_rate": 1.091698485035103e-06, + "loss": 0.1254, + "step": 51695 + }, + { + "epoch": 19.09, + "learning_rate": 1.0916615346717576e-06, + "loss": 0.2925, + "step": 51696 + }, + { + "epoch": 19.09, + "learning_rate": 1.0916245843084124e-06, + "loss": 0.3042, + "step": 51697 + }, + { + "epoch": 19.09, + "learning_rate": 1.0915876339450673e-06, + "loss": 0.2612, + "step": 51698 + }, + { + "epoch": 19.09, + "learning_rate": 1.0915506835817219e-06, + "loss": 0.3325, + "step": 51699 + }, + { + "epoch": 19.09, + "learning_rate": 1.0915137332183767e-06, + "loss": 0.3129, + "step": 51700 + }, + { + "epoch": 19.09, + "learning_rate": 1.0914767828550316e-06, + "loss": 0.3064, + "step": 51701 + }, + { + "epoch": 19.09, + "learning_rate": 1.0914398324916862e-06, + "loss": 0.1483, + "step": 51702 + }, + { + "epoch": 19.09, + "learning_rate": 1.0914028821283409e-06, + "loss": 0.2093, + "step": 51703 + }, + { + "epoch": 19.09, + "learning_rate": 1.0913659317649957e-06, + "loss": 0.2828, + "step": 51704 + }, + { + "epoch": 19.09, + "learning_rate": 1.0913289814016503e-06, + "loss": 0.1861, + "step": 51705 + }, + { + "epoch": 19.09, + "learning_rate": 1.0912920310383052e-06, + "loss": 0.1439, + "step": 51706 + }, + { + "epoch": 19.09, + "learning_rate": 1.09125508067496e-06, + "loss": 0.1712, + "step": 51707 + }, + { + "epoch": 19.09, + "learning_rate": 1.0912181303116147e-06, + "loss": 0.2126, + "step": 51708 + }, + { + "epoch": 19.09, + "learning_rate": 1.0911811799482695e-06, + "loss": 0.3148, + "step": 51709 + }, + { + "epoch": 19.1, + "learning_rate": 1.0911442295849244e-06, + "loss": 0.2266, + "step": 51710 + }, + { + "epoch": 19.1, + "learning_rate": 1.091107279221579e-06, + "loss": 0.2746, + "step": 51711 + }, + { + "epoch": 19.1, + "learning_rate": 1.0910703288582339e-06, + "loss": 0.1441, + "step": 51712 + }, + { + "epoch": 19.1, + "learning_rate": 1.0910333784948885e-06, + "loss": 0.2109, + "step": 51713 + }, + { + "epoch": 19.1, + "learning_rate": 1.0909964281315433e-06, + "loss": 0.1856, + "step": 51714 + }, + { + "epoch": 19.1, + "learning_rate": 1.090959477768198e-06, + "loss": 0.1835, + "step": 51715 + }, + { + "epoch": 19.1, + "learning_rate": 1.0909225274048528e-06, + "loss": 0.2642, + "step": 51716 + }, + { + "epoch": 19.1, + "learning_rate": 1.0908855770415077e-06, + "loss": 0.1755, + "step": 51717 + }, + { + "epoch": 19.1, + "learning_rate": 1.0908486266781623e-06, + "loss": 0.3026, + "step": 51718 + }, + { + "epoch": 19.1, + "learning_rate": 1.0908116763148172e-06, + "loss": 0.3042, + "step": 51719 + }, + { + "epoch": 19.1, + "learning_rate": 1.090774725951472e-06, + "loss": 0.162, + "step": 51720 + }, + { + "epoch": 19.1, + "learning_rate": 1.0907377755881266e-06, + "loss": 0.2828, + "step": 51721 + }, + { + "epoch": 19.1, + "learning_rate": 1.0907008252247815e-06, + "loss": 0.1575, + "step": 51722 + }, + { + "epoch": 19.1, + "learning_rate": 1.0906638748614361e-06, + "loss": 0.1929, + "step": 51723 + }, + { + "epoch": 19.1, + "learning_rate": 1.0906269244980908e-06, + "loss": 0.2184, + "step": 51724 + }, + { + "epoch": 19.1, + "learning_rate": 1.0905899741347456e-06, + "loss": 0.2257, + "step": 51725 + }, + { + "epoch": 19.1, + "learning_rate": 1.0905530237714005e-06, + "loss": 0.2118, + "step": 51726 + }, + { + "epoch": 19.1, + "learning_rate": 1.090516073408055e-06, + "loss": 0.2057, + "step": 51727 + }, + { + "epoch": 19.1, + "learning_rate": 1.09047912304471e-06, + "loss": 0.1403, + "step": 51728 + }, + { + "epoch": 19.1, + "learning_rate": 1.0904421726813648e-06, + "loss": 0.2622, + "step": 51729 + }, + { + "epoch": 19.1, + "learning_rate": 1.0904052223180194e-06, + "loss": 0.149, + "step": 51730 + }, + { + "epoch": 19.1, + "learning_rate": 1.0903682719546743e-06, + "loss": 0.1461, + "step": 51731 + }, + { + "epoch": 19.1, + "learning_rate": 1.0903313215913291e-06, + "loss": 0.3135, + "step": 51732 + }, + { + "epoch": 19.1, + "learning_rate": 1.0902943712279838e-06, + "loss": 0.1434, + "step": 51733 + }, + { + "epoch": 19.1, + "learning_rate": 1.0902574208646384e-06, + "loss": 0.161, + "step": 51734 + }, + { + "epoch": 19.1, + "learning_rate": 1.0902204705012933e-06, + "loss": 0.1779, + "step": 51735 + }, + { + "epoch": 19.1, + "learning_rate": 1.090183520137948e-06, + "loss": 0.3661, + "step": 51736 + }, + { + "epoch": 19.11, + "learning_rate": 1.0901465697746027e-06, + "loss": 0.1565, + "step": 51737 + }, + { + "epoch": 19.11, + "learning_rate": 1.0901096194112576e-06, + "loss": 0.1016, + "step": 51738 + }, + { + "epoch": 19.11, + "learning_rate": 1.0900726690479124e-06, + "loss": 0.2172, + "step": 51739 + }, + { + "epoch": 19.11, + "learning_rate": 1.090035718684567e-06, + "loss": 0.2112, + "step": 51740 + }, + { + "epoch": 19.11, + "learning_rate": 1.089998768321222e-06, + "loss": 0.3288, + "step": 51741 + }, + { + "epoch": 19.11, + "learning_rate": 1.0899618179578768e-06, + "loss": 0.1781, + "step": 51742 + }, + { + "epoch": 19.11, + "learning_rate": 1.0899248675945314e-06, + "loss": 0.1555, + "step": 51743 + }, + { + "epoch": 19.11, + "learning_rate": 1.089887917231186e-06, + "loss": 0.1157, + "step": 51744 + }, + { + "epoch": 19.11, + "learning_rate": 1.0898509668678409e-06, + "loss": 0.2822, + "step": 51745 + }, + { + "epoch": 19.11, + "learning_rate": 1.0898140165044957e-06, + "loss": 0.1775, + "step": 51746 + }, + { + "epoch": 19.11, + "learning_rate": 1.0897770661411504e-06, + "loss": 0.2327, + "step": 51747 + }, + { + "epoch": 19.11, + "learning_rate": 1.0897401157778052e-06, + "loss": 0.3011, + "step": 51748 + }, + { + "epoch": 19.11, + "learning_rate": 1.08970316541446e-06, + "loss": 0.2228, + "step": 51749 + }, + { + "epoch": 19.11, + "learning_rate": 1.0896662150511147e-06, + "loss": 0.3309, + "step": 51750 + }, + { + "epoch": 19.11, + "learning_rate": 1.0896292646877696e-06, + "loss": 0.2185, + "step": 51751 + }, + { + "epoch": 19.11, + "learning_rate": 1.0895923143244244e-06, + "loss": 0.1395, + "step": 51752 + }, + { + "epoch": 19.11, + "learning_rate": 1.0895553639610788e-06, + "loss": 0.381, + "step": 51753 + }, + { + "epoch": 19.11, + "learning_rate": 1.0895184135977337e-06, + "loss": 0.1713, + "step": 51754 + }, + { + "epoch": 19.11, + "learning_rate": 1.0894814632343885e-06, + "loss": 0.2212, + "step": 51755 + }, + { + "epoch": 19.11, + "learning_rate": 1.0894445128710432e-06, + "loss": 0.1904, + "step": 51756 + }, + { + "epoch": 19.11, + "learning_rate": 1.089407562507698e-06, + "loss": 0.2028, + "step": 51757 + }, + { + "epoch": 19.11, + "learning_rate": 1.0893706121443529e-06, + "loss": 0.2098, + "step": 51758 + }, + { + "epoch": 19.11, + "learning_rate": 1.0893336617810075e-06, + "loss": 0.1517, + "step": 51759 + }, + { + "epoch": 19.11, + "learning_rate": 1.0892967114176623e-06, + "loss": 0.3256, + "step": 51760 + }, + { + "epoch": 19.11, + "learning_rate": 1.0892597610543172e-06, + "loss": 0.1577, + "step": 51761 + }, + { + "epoch": 19.11, + "learning_rate": 1.0892228106909718e-06, + "loss": 0.1841, + "step": 51762 + }, + { + "epoch": 19.11, + "learning_rate": 1.0891858603276265e-06, + "loss": 0.1262, + "step": 51763 + }, + { + "epoch": 19.12, + "learning_rate": 1.0891489099642813e-06, + "loss": 0.2082, + "step": 51764 + }, + { + "epoch": 19.12, + "learning_rate": 1.0891119596009362e-06, + "loss": 0.2303, + "step": 51765 + }, + { + "epoch": 19.12, + "learning_rate": 1.0890750092375908e-06, + "loss": 0.2274, + "step": 51766 + }, + { + "epoch": 19.12, + "learning_rate": 1.0890380588742456e-06, + "loss": 0.1966, + "step": 51767 + }, + { + "epoch": 19.12, + "learning_rate": 1.0890011085109005e-06, + "loss": 0.2427, + "step": 51768 + }, + { + "epoch": 19.12, + "learning_rate": 1.0889641581475551e-06, + "loss": 0.2331, + "step": 51769 + }, + { + "epoch": 19.12, + "learning_rate": 1.08892720778421e-06, + "loss": 0.2386, + "step": 51770 + }, + { + "epoch": 19.12, + "learning_rate": 1.0888902574208648e-06, + "loss": 0.2139, + "step": 51771 + }, + { + "epoch": 19.12, + "learning_rate": 1.0888533070575195e-06, + "loss": 0.2373, + "step": 51772 + }, + { + "epoch": 19.12, + "learning_rate": 1.088816356694174e-06, + "loss": 0.138, + "step": 51773 + }, + { + "epoch": 19.12, + "learning_rate": 1.088779406330829e-06, + "loss": 0.1629, + "step": 51774 + }, + { + "epoch": 19.12, + "learning_rate": 1.0887424559674836e-06, + "loss": 0.2155, + "step": 51775 + }, + { + "epoch": 19.12, + "learning_rate": 1.0887055056041384e-06, + "loss": 0.1621, + "step": 51776 + }, + { + "epoch": 19.12, + "learning_rate": 1.0886685552407933e-06, + "loss": 0.142, + "step": 51777 + }, + { + "epoch": 19.12, + "learning_rate": 1.088631604877448e-06, + "loss": 0.3495, + "step": 51778 + }, + { + "epoch": 19.12, + "learning_rate": 1.0885946545141028e-06, + "loss": 0.2505, + "step": 51779 + }, + { + "epoch": 19.12, + "learning_rate": 1.0885577041507576e-06, + "loss": 0.3408, + "step": 51780 + }, + { + "epoch": 19.12, + "learning_rate": 1.0885207537874122e-06, + "loss": 0.2806, + "step": 51781 + }, + { + "epoch": 19.12, + "learning_rate": 1.0884838034240669e-06, + "loss": 0.1996, + "step": 51782 + }, + { + "epoch": 19.12, + "learning_rate": 1.0884468530607217e-06, + "loss": 0.1891, + "step": 51783 + }, + { + "epoch": 19.12, + "learning_rate": 1.0884099026973766e-06, + "loss": 0.233, + "step": 51784 + }, + { + "epoch": 19.12, + "learning_rate": 1.0883729523340312e-06, + "loss": 0.2418, + "step": 51785 + }, + { + "epoch": 19.12, + "learning_rate": 1.088336001970686e-06, + "loss": 0.2816, + "step": 51786 + }, + { + "epoch": 19.12, + "learning_rate": 1.088299051607341e-06, + "loss": 0.2083, + "step": 51787 + }, + { + "epoch": 19.12, + "learning_rate": 1.0882621012439955e-06, + "loss": 0.1408, + "step": 51788 + }, + { + "epoch": 19.12, + "learning_rate": 1.0882251508806504e-06, + "loss": 0.3386, + "step": 51789 + }, + { + "epoch": 19.12, + "learning_rate": 1.0881882005173052e-06, + "loss": 0.1848, + "step": 51790 + }, + { + "epoch": 19.13, + "learning_rate": 1.0881512501539599e-06, + "loss": 0.2534, + "step": 51791 + }, + { + "epoch": 19.13, + "learning_rate": 1.0881142997906145e-06, + "loss": 0.1852, + "step": 51792 + }, + { + "epoch": 19.13, + "learning_rate": 1.0880773494272694e-06, + "loss": 0.1939, + "step": 51793 + }, + { + "epoch": 19.13, + "learning_rate": 1.0880403990639242e-06, + "loss": 0.1501, + "step": 51794 + }, + { + "epoch": 19.13, + "learning_rate": 1.0880034487005788e-06, + "loss": 0.2524, + "step": 51795 + }, + { + "epoch": 19.13, + "learning_rate": 1.0879664983372337e-06, + "loss": 0.2659, + "step": 51796 + }, + { + "epoch": 19.13, + "learning_rate": 1.0879295479738883e-06, + "loss": 0.1909, + "step": 51797 + }, + { + "epoch": 19.13, + "learning_rate": 1.0878925976105432e-06, + "loss": 0.4495, + "step": 51798 + }, + { + "epoch": 19.13, + "learning_rate": 1.087855647247198e-06, + "loss": 0.2357, + "step": 51799 + }, + { + "epoch": 19.13, + "learning_rate": 1.0878186968838527e-06, + "loss": 0.2486, + "step": 51800 + }, + { + "epoch": 19.13, + "learning_rate": 1.0877817465205075e-06, + "loss": 0.3062, + "step": 51801 + }, + { + "epoch": 19.13, + "learning_rate": 1.0877447961571622e-06, + "loss": 0.21, + "step": 51802 + }, + { + "epoch": 19.13, + "learning_rate": 1.087707845793817e-06, + "loss": 0.3307, + "step": 51803 + }, + { + "epoch": 19.13, + "learning_rate": 1.0876708954304716e-06, + "loss": 0.2767, + "step": 51804 + }, + { + "epoch": 19.13, + "learning_rate": 1.0876339450671265e-06, + "loss": 0.1958, + "step": 51805 + }, + { + "epoch": 19.13, + "learning_rate": 1.0875969947037813e-06, + "loss": 0.1515, + "step": 51806 + }, + { + "epoch": 19.13, + "learning_rate": 1.087560044340436e-06, + "loss": 0.2093, + "step": 51807 + }, + { + "epoch": 19.13, + "learning_rate": 1.0875230939770908e-06, + "loss": 0.226, + "step": 51808 + }, + { + "epoch": 19.13, + "learning_rate": 1.0874861436137457e-06, + "loss": 0.2337, + "step": 51809 + }, + { + "epoch": 19.13, + "learning_rate": 1.0874491932504003e-06, + "loss": 0.3115, + "step": 51810 + }, + { + "epoch": 19.13, + "learning_rate": 1.0874122428870551e-06, + "loss": 0.2978, + "step": 51811 + }, + { + "epoch": 19.13, + "learning_rate": 1.0873752925237098e-06, + "loss": 0.1529, + "step": 51812 + }, + { + "epoch": 19.13, + "learning_rate": 1.0873383421603646e-06, + "loss": 0.204, + "step": 51813 + }, + { + "epoch": 19.13, + "learning_rate": 1.0873013917970193e-06, + "loss": 0.302, + "step": 51814 + }, + { + "epoch": 19.13, + "learning_rate": 1.0872644414336741e-06, + "loss": 0.1989, + "step": 51815 + }, + { + "epoch": 19.13, + "learning_rate": 1.087227491070329e-06, + "loss": 0.2452, + "step": 51816 + }, + { + "epoch": 19.13, + "learning_rate": 1.0871905407069836e-06, + "loss": 0.1408, + "step": 51817 + }, + { + "epoch": 19.14, + "learning_rate": 1.0871535903436385e-06, + "loss": 0.1425, + "step": 51818 + }, + { + "epoch": 19.14, + "learning_rate": 1.0871166399802933e-06, + "loss": 0.1578, + "step": 51819 + }, + { + "epoch": 19.14, + "learning_rate": 1.087079689616948e-06, + "loss": 0.2618, + "step": 51820 + }, + { + "epoch": 19.14, + "learning_rate": 1.0870427392536028e-06, + "loss": 0.2577, + "step": 51821 + }, + { + "epoch": 19.14, + "learning_rate": 1.0870057888902574e-06, + "loss": 0.1697, + "step": 51822 + }, + { + "epoch": 19.14, + "learning_rate": 1.086968838526912e-06, + "loss": 0.1952, + "step": 51823 + }, + { + "epoch": 19.14, + "learning_rate": 1.086931888163567e-06, + "loss": 0.1292, + "step": 51824 + }, + { + "epoch": 19.14, + "learning_rate": 1.0868949378002218e-06, + "loss": 0.0782, + "step": 51825 + }, + { + "epoch": 19.14, + "learning_rate": 1.0868579874368764e-06, + "loss": 0.1231, + "step": 51826 + }, + { + "epoch": 19.14, + "learning_rate": 1.0868210370735312e-06, + "loss": 0.2168, + "step": 51827 + }, + { + "epoch": 19.14, + "learning_rate": 1.086784086710186e-06, + "loss": 0.1469, + "step": 51828 + }, + { + "epoch": 19.14, + "learning_rate": 1.0867471363468407e-06, + "loss": 0.1604, + "step": 51829 + }, + { + "epoch": 19.14, + "learning_rate": 1.0867101859834956e-06, + "loss": 0.3548, + "step": 51830 + }, + { + "epoch": 19.14, + "learning_rate": 1.0866732356201504e-06, + "loss": 0.2369, + "step": 51831 + }, + { + "epoch": 19.14, + "learning_rate": 1.086636285256805e-06, + "loss": 0.2737, + "step": 51832 + }, + { + "epoch": 19.14, + "learning_rate": 1.0865993348934597e-06, + "loss": 0.143, + "step": 51833 + }, + { + "epoch": 19.14, + "learning_rate": 1.0865623845301145e-06, + "loss": 0.1924, + "step": 51834 + }, + { + "epoch": 19.14, + "learning_rate": 1.0865254341667694e-06, + "loss": 0.3788, + "step": 51835 + }, + { + "epoch": 19.14, + "learning_rate": 1.086488483803424e-06, + "loss": 0.223, + "step": 51836 + }, + { + "epoch": 19.14, + "learning_rate": 1.0864515334400789e-06, + "loss": 0.1796, + "step": 51837 + }, + { + "epoch": 19.14, + "learning_rate": 1.0864145830767337e-06, + "loss": 0.3012, + "step": 51838 + }, + { + "epoch": 19.14, + "learning_rate": 1.0863776327133884e-06, + "loss": 0.2613, + "step": 51839 + }, + { + "epoch": 19.14, + "learning_rate": 1.0863406823500432e-06, + "loss": 0.2624, + "step": 51840 + }, + { + "epoch": 19.14, + "learning_rate": 1.086303731986698e-06, + "loss": 0.337, + "step": 51841 + }, + { + "epoch": 19.14, + "learning_rate": 1.0862667816233525e-06, + "loss": 0.0952, + "step": 51842 + }, + { + "epoch": 19.14, + "learning_rate": 1.0862298312600073e-06, + "loss": 0.2076, + "step": 51843 + }, + { + "epoch": 19.14, + "learning_rate": 1.0861928808966622e-06, + "loss": 0.2999, + "step": 51844 + }, + { + "epoch": 19.15, + "learning_rate": 1.0861559305333168e-06, + "loss": 0.0964, + "step": 51845 + }, + { + "epoch": 19.15, + "learning_rate": 1.0861189801699717e-06, + "loss": 0.1632, + "step": 51846 + }, + { + "epoch": 19.15, + "learning_rate": 1.0860820298066265e-06, + "loss": 0.211, + "step": 51847 + }, + { + "epoch": 19.15, + "learning_rate": 1.0860450794432811e-06, + "loss": 0.1829, + "step": 51848 + }, + { + "epoch": 19.15, + "learning_rate": 1.086008129079936e-06, + "loss": 0.213, + "step": 51849 + }, + { + "epoch": 19.15, + "learning_rate": 1.0859711787165908e-06, + "loss": 0.2435, + "step": 51850 + }, + { + "epoch": 19.15, + "learning_rate": 1.0859342283532455e-06, + "loss": 0.221, + "step": 51851 + }, + { + "epoch": 19.15, + "learning_rate": 1.0858972779899001e-06, + "loss": 0.1733, + "step": 51852 + }, + { + "epoch": 19.15, + "learning_rate": 1.085860327626555e-06, + "loss": 0.3069, + "step": 51853 + }, + { + "epoch": 19.15, + "learning_rate": 1.0858233772632098e-06, + "loss": 0.1499, + "step": 51854 + }, + { + "epoch": 19.15, + "learning_rate": 1.0857864268998644e-06, + "loss": 0.2648, + "step": 51855 + }, + { + "epoch": 19.15, + "learning_rate": 1.0857494765365193e-06, + "loss": 0.1525, + "step": 51856 + }, + { + "epoch": 19.15, + "learning_rate": 1.0857125261731741e-06, + "loss": 0.2639, + "step": 51857 + }, + { + "epoch": 19.15, + "learning_rate": 1.0856755758098288e-06, + "loss": 0.2798, + "step": 51858 + }, + { + "epoch": 19.15, + "learning_rate": 1.0856386254464836e-06, + "loss": 0.1539, + "step": 51859 + }, + { + "epoch": 19.15, + "learning_rate": 1.0856016750831385e-06, + "loss": 0.294, + "step": 51860 + }, + { + "epoch": 19.15, + "learning_rate": 1.0855647247197931e-06, + "loss": 0.2046, + "step": 51861 + }, + { + "epoch": 19.15, + "learning_rate": 1.0855277743564477e-06, + "loss": 0.1805, + "step": 51862 + }, + { + "epoch": 19.15, + "learning_rate": 1.0854908239931026e-06, + "loss": 0.1838, + "step": 51863 + }, + { + "epoch": 19.15, + "learning_rate": 1.0854538736297574e-06, + "loss": 0.2553, + "step": 51864 + }, + { + "epoch": 19.15, + "learning_rate": 1.085416923266412e-06, + "loss": 0.1975, + "step": 51865 + }, + { + "epoch": 19.15, + "learning_rate": 1.085379972903067e-06, + "loss": 0.1506, + "step": 51866 + }, + { + "epoch": 19.15, + "learning_rate": 1.0853430225397218e-06, + "loss": 0.1979, + "step": 51867 + }, + { + "epoch": 19.15, + "learning_rate": 1.0853060721763764e-06, + "loss": 0.3124, + "step": 51868 + }, + { + "epoch": 19.15, + "learning_rate": 1.0852691218130313e-06, + "loss": 0.1672, + "step": 51869 + }, + { + "epoch": 19.15, + "learning_rate": 1.0852321714496861e-06, + "loss": 0.1978, + "step": 51870 + }, + { + "epoch": 19.15, + "learning_rate": 1.0851952210863407e-06, + "loss": 0.2232, + "step": 51871 + }, + { + "epoch": 19.16, + "learning_rate": 1.0851582707229954e-06, + "loss": 0.1893, + "step": 51872 + }, + { + "epoch": 19.16, + "learning_rate": 1.0851213203596502e-06, + "loss": 0.2116, + "step": 51873 + }, + { + "epoch": 19.16, + "learning_rate": 1.0850843699963049e-06, + "loss": 0.2886, + "step": 51874 + }, + { + "epoch": 19.16, + "learning_rate": 1.0850474196329597e-06, + "loss": 0.2077, + "step": 51875 + }, + { + "epoch": 19.16, + "learning_rate": 1.0850104692696146e-06, + "loss": 0.2143, + "step": 51876 + }, + { + "epoch": 19.16, + "learning_rate": 1.0849735189062692e-06, + "loss": 0.2322, + "step": 51877 + }, + { + "epoch": 19.16, + "learning_rate": 1.084936568542924e-06, + "loss": 0.193, + "step": 51878 + }, + { + "epoch": 19.16, + "learning_rate": 1.084899618179579e-06, + "loss": 0.1499, + "step": 51879 + }, + { + "epoch": 19.16, + "learning_rate": 1.0848626678162335e-06, + "loss": 0.207, + "step": 51880 + }, + { + "epoch": 19.16, + "learning_rate": 1.0848257174528884e-06, + "loss": 0.2507, + "step": 51881 + }, + { + "epoch": 19.16, + "learning_rate": 1.084788767089543e-06, + "loss": 0.1296, + "step": 51882 + }, + { + "epoch": 19.16, + "learning_rate": 1.0847518167261979e-06, + "loss": 0.2492, + "step": 51883 + }, + { + "epoch": 19.16, + "learning_rate": 1.0847148663628525e-06, + "loss": 0.2274, + "step": 51884 + }, + { + "epoch": 19.16, + "learning_rate": 1.0846779159995074e-06, + "loss": 0.1292, + "step": 51885 + }, + { + "epoch": 19.16, + "learning_rate": 1.0846409656361622e-06, + "loss": 0.1978, + "step": 51886 + }, + { + "epoch": 19.16, + "learning_rate": 1.0846040152728168e-06, + "loss": 0.1299, + "step": 51887 + }, + { + "epoch": 19.16, + "learning_rate": 1.0845670649094717e-06, + "loss": 0.185, + "step": 51888 + }, + { + "epoch": 19.16, + "learning_rate": 1.0845301145461265e-06, + "loss": 0.252, + "step": 51889 + }, + { + "epoch": 19.16, + "learning_rate": 1.0844931641827812e-06, + "loss": 0.1529, + "step": 51890 + }, + { + "epoch": 19.16, + "learning_rate": 1.084456213819436e-06, + "loss": 0.1596, + "step": 51891 + }, + { + "epoch": 19.16, + "learning_rate": 1.0844192634560907e-06, + "loss": 0.3665, + "step": 51892 + }, + { + "epoch": 19.16, + "learning_rate": 1.0843823130927453e-06, + "loss": 0.225, + "step": 51893 + }, + { + "epoch": 19.16, + "learning_rate": 1.0843453627294001e-06, + "loss": 0.2949, + "step": 51894 + }, + { + "epoch": 19.16, + "learning_rate": 1.084308412366055e-06, + "loss": 0.208, + "step": 51895 + }, + { + "epoch": 19.16, + "learning_rate": 1.0842714620027096e-06, + "loss": 0.3165, + "step": 51896 + }, + { + "epoch": 19.16, + "learning_rate": 1.0842345116393645e-06, + "loss": 0.1963, + "step": 51897 + }, + { + "epoch": 19.16, + "learning_rate": 1.0841975612760193e-06, + "loss": 0.2684, + "step": 51898 + }, + { + "epoch": 19.17, + "learning_rate": 1.084160610912674e-06, + "loss": 0.2558, + "step": 51899 + }, + { + "epoch": 19.17, + "learning_rate": 1.0841236605493288e-06, + "loss": 0.2034, + "step": 51900 + }, + { + "epoch": 19.17, + "learning_rate": 1.0840867101859837e-06, + "loss": 0.2332, + "step": 51901 + }, + { + "epoch": 19.17, + "learning_rate": 1.0840497598226383e-06, + "loss": 0.2373, + "step": 51902 + }, + { + "epoch": 19.17, + "learning_rate": 1.084012809459293e-06, + "loss": 0.3262, + "step": 51903 + }, + { + "epoch": 19.17, + "learning_rate": 1.0839758590959478e-06, + "loss": 0.3113, + "step": 51904 + }, + { + "epoch": 19.17, + "learning_rate": 1.0839389087326026e-06, + "loss": 0.2436, + "step": 51905 + }, + { + "epoch": 19.17, + "learning_rate": 1.0839019583692573e-06, + "loss": 0.285, + "step": 51906 + }, + { + "epoch": 19.17, + "learning_rate": 1.083865008005912e-06, + "loss": 0.2377, + "step": 51907 + }, + { + "epoch": 19.17, + "learning_rate": 1.083828057642567e-06, + "loss": 0.1981, + "step": 51908 + }, + { + "epoch": 19.17, + "learning_rate": 1.0837911072792216e-06, + "loss": 0.219, + "step": 51909 + }, + { + "epoch": 19.17, + "learning_rate": 1.0837541569158764e-06, + "loss": 0.3007, + "step": 51910 + }, + { + "epoch": 19.17, + "learning_rate": 1.0837172065525313e-06, + "loss": 0.2421, + "step": 51911 + }, + { + "epoch": 19.17, + "learning_rate": 1.0836802561891857e-06, + "loss": 0.2183, + "step": 51912 + }, + { + "epoch": 19.17, + "learning_rate": 1.0836433058258406e-06, + "loss": 0.1389, + "step": 51913 + }, + { + "epoch": 19.17, + "learning_rate": 1.0836063554624954e-06, + "loss": 0.1768, + "step": 51914 + }, + { + "epoch": 19.17, + "learning_rate": 1.08356940509915e-06, + "loss": 0.1852, + "step": 51915 + }, + { + "epoch": 19.17, + "learning_rate": 1.0835324547358049e-06, + "loss": 0.2098, + "step": 51916 + }, + { + "epoch": 19.17, + "learning_rate": 1.0834955043724597e-06, + "loss": 0.1583, + "step": 51917 + }, + { + "epoch": 19.17, + "learning_rate": 1.0834585540091144e-06, + "loss": 0.2439, + "step": 51918 + }, + { + "epoch": 19.17, + "learning_rate": 1.0834216036457692e-06, + "loss": 0.1532, + "step": 51919 + }, + { + "epoch": 19.17, + "learning_rate": 1.083384653282424e-06, + "loss": 0.2217, + "step": 51920 + }, + { + "epoch": 19.17, + "learning_rate": 1.0833477029190787e-06, + "loss": 0.2366, + "step": 51921 + }, + { + "epoch": 19.17, + "learning_rate": 1.0833107525557333e-06, + "loss": 0.2242, + "step": 51922 + }, + { + "epoch": 19.17, + "learning_rate": 1.0832738021923882e-06, + "loss": 0.2952, + "step": 51923 + }, + { + "epoch": 19.17, + "learning_rate": 1.083236851829043e-06, + "loss": 0.2179, + "step": 51924 + }, + { + "epoch": 19.17, + "learning_rate": 1.0831999014656977e-06, + "loss": 0.2934, + "step": 51925 + }, + { + "epoch": 19.18, + "learning_rate": 1.0831629511023525e-06, + "loss": 0.1957, + "step": 51926 + }, + { + "epoch": 19.18, + "learning_rate": 1.0831260007390074e-06, + "loss": 0.3047, + "step": 51927 + }, + { + "epoch": 19.18, + "learning_rate": 1.083089050375662e-06, + "loss": 0.2549, + "step": 51928 + }, + { + "epoch": 19.18, + "learning_rate": 1.0830521000123169e-06, + "loss": 0.3055, + "step": 51929 + }, + { + "epoch": 19.18, + "learning_rate": 1.0830151496489717e-06, + "loss": 0.3185, + "step": 51930 + }, + { + "epoch": 19.18, + "learning_rate": 1.0829781992856263e-06, + "loss": 0.2695, + "step": 51931 + }, + { + "epoch": 19.18, + "learning_rate": 1.082941248922281e-06, + "loss": 0.1451, + "step": 51932 + }, + { + "epoch": 19.18, + "learning_rate": 1.0829042985589358e-06, + "loss": 0.2747, + "step": 51933 + }, + { + "epoch": 19.18, + "learning_rate": 1.0828673481955907e-06, + "loss": 0.1915, + "step": 51934 + }, + { + "epoch": 19.18, + "learning_rate": 1.0828303978322453e-06, + "loss": 0.2417, + "step": 51935 + }, + { + "epoch": 19.18, + "learning_rate": 1.0827934474689002e-06, + "loss": 0.2735, + "step": 51936 + }, + { + "epoch": 19.18, + "learning_rate": 1.082756497105555e-06, + "loss": 0.1496, + "step": 51937 + }, + { + "epoch": 19.18, + "learning_rate": 1.0827195467422096e-06, + "loss": 0.3094, + "step": 51938 + }, + { + "epoch": 19.18, + "learning_rate": 1.0826825963788645e-06, + "loss": 0.2924, + "step": 51939 + }, + { + "epoch": 19.18, + "learning_rate": 1.0826456460155193e-06, + "loss": 0.1961, + "step": 51940 + }, + { + "epoch": 19.18, + "learning_rate": 1.082608695652174e-06, + "loss": 0.1435, + "step": 51941 + }, + { + "epoch": 19.18, + "learning_rate": 1.0825717452888286e-06, + "loss": 0.358, + "step": 51942 + }, + { + "epoch": 19.18, + "learning_rate": 1.0825347949254835e-06, + "loss": 0.1573, + "step": 51943 + }, + { + "epoch": 19.18, + "learning_rate": 1.082497844562138e-06, + "loss": 0.2078, + "step": 51944 + }, + { + "epoch": 19.18, + "learning_rate": 1.082460894198793e-06, + "loss": 0.2693, + "step": 51945 + }, + { + "epoch": 19.18, + "learning_rate": 1.0824239438354478e-06, + "loss": 0.2848, + "step": 51946 + }, + { + "epoch": 19.18, + "learning_rate": 1.0823869934721024e-06, + "loss": 0.1783, + "step": 51947 + }, + { + "epoch": 19.18, + "learning_rate": 1.0823500431087573e-06, + "loss": 0.2128, + "step": 51948 + }, + { + "epoch": 19.18, + "learning_rate": 1.0823130927454121e-06, + "loss": 0.3835, + "step": 51949 + }, + { + "epoch": 19.18, + "learning_rate": 1.0822761423820668e-06, + "loss": 0.3335, + "step": 51950 + }, + { + "epoch": 19.18, + "learning_rate": 1.0822391920187216e-06, + "loss": 0.1227, + "step": 51951 + }, + { + "epoch": 19.18, + "learning_rate": 1.0822022416553763e-06, + "loss": 0.2281, + "step": 51952 + }, + { + "epoch": 19.19, + "learning_rate": 1.082165291292031e-06, + "loss": 0.2422, + "step": 51953 + }, + { + "epoch": 19.19, + "learning_rate": 1.0821283409286857e-06, + "loss": 0.2229, + "step": 51954 + }, + { + "epoch": 19.19, + "learning_rate": 1.0820913905653406e-06, + "loss": 0.2051, + "step": 51955 + }, + { + "epoch": 19.19, + "learning_rate": 1.0820544402019954e-06, + "loss": 0.2067, + "step": 51956 + }, + { + "epoch": 19.19, + "learning_rate": 1.08201748983865e-06, + "loss": 0.1495, + "step": 51957 + }, + { + "epoch": 19.19, + "learning_rate": 1.081980539475305e-06, + "loss": 0.3431, + "step": 51958 + }, + { + "epoch": 19.19, + "learning_rate": 1.0819435891119598e-06, + "loss": 0.0937, + "step": 51959 + }, + { + "epoch": 19.19, + "learning_rate": 1.0819066387486144e-06, + "loss": 0.2143, + "step": 51960 + }, + { + "epoch": 19.19, + "learning_rate": 1.0818696883852692e-06, + "loss": 0.1547, + "step": 51961 + }, + { + "epoch": 19.19, + "learning_rate": 1.0818327380219239e-06, + "loss": 0.1695, + "step": 51962 + }, + { + "epoch": 19.19, + "learning_rate": 1.0817957876585785e-06, + "loss": 0.233, + "step": 51963 + }, + { + "epoch": 19.19, + "learning_rate": 1.0817588372952334e-06, + "loss": 0.2255, + "step": 51964 + }, + { + "epoch": 19.19, + "learning_rate": 1.0817218869318882e-06, + "loss": 0.4288, + "step": 51965 + }, + { + "epoch": 19.19, + "learning_rate": 1.0816849365685429e-06, + "loss": 0.2619, + "step": 51966 + }, + { + "epoch": 19.19, + "learning_rate": 1.0816479862051977e-06, + "loss": 0.305, + "step": 51967 + }, + { + "epoch": 19.19, + "learning_rate": 1.0816110358418526e-06, + "loss": 0.1473, + "step": 51968 + }, + { + "epoch": 19.19, + "learning_rate": 1.0815740854785072e-06, + "loss": 0.1952, + "step": 51969 + }, + { + "epoch": 19.19, + "learning_rate": 1.081537135115162e-06, + "loss": 0.2017, + "step": 51970 + }, + { + "epoch": 19.19, + "learning_rate": 1.0815001847518169e-06, + "loss": 0.1228, + "step": 51971 + }, + { + "epoch": 19.19, + "learning_rate": 1.0814632343884715e-06, + "loss": 0.1419, + "step": 51972 + }, + { + "epoch": 19.19, + "learning_rate": 1.0814262840251262e-06, + "loss": 0.214, + "step": 51973 + }, + { + "epoch": 19.19, + "learning_rate": 1.081389333661781e-06, + "loss": 0.2285, + "step": 51974 + }, + { + "epoch": 19.19, + "learning_rate": 1.0813523832984359e-06, + "loss": 0.1774, + "step": 51975 + }, + { + "epoch": 19.19, + "learning_rate": 1.0813154329350905e-06, + "loss": 0.2396, + "step": 51976 + }, + { + "epoch": 19.19, + "learning_rate": 1.0812784825717453e-06, + "loss": 0.1717, + "step": 51977 + }, + { + "epoch": 19.19, + "learning_rate": 1.0812415322084002e-06, + "loss": 0.1692, + "step": 51978 + }, + { + "epoch": 19.19, + "learning_rate": 1.0812045818450548e-06, + "loss": 0.2094, + "step": 51979 + }, + { + "epoch": 19.19, + "learning_rate": 1.0811676314817097e-06, + "loss": 0.2655, + "step": 51980 + }, + { + "epoch": 19.2, + "learning_rate": 1.0811306811183645e-06, + "loss": 0.2772, + "step": 51981 + }, + { + "epoch": 19.2, + "learning_rate": 1.0810937307550192e-06, + "loss": 0.1946, + "step": 51982 + }, + { + "epoch": 19.2, + "learning_rate": 1.0810567803916738e-06, + "loss": 0.2548, + "step": 51983 + }, + { + "epoch": 19.2, + "learning_rate": 1.0810198300283286e-06, + "loss": 0.2597, + "step": 51984 + }, + { + "epoch": 19.2, + "learning_rate": 1.0809828796649833e-06, + "loss": 0.1365, + "step": 51985 + }, + { + "epoch": 19.2, + "learning_rate": 1.0809459293016381e-06, + "loss": 0.2924, + "step": 51986 + }, + { + "epoch": 19.2, + "learning_rate": 1.080908978938293e-06, + "loss": 0.1935, + "step": 51987 + }, + { + "epoch": 19.2, + "learning_rate": 1.0808720285749476e-06, + "loss": 0.3496, + "step": 51988 + }, + { + "epoch": 19.2, + "learning_rate": 1.0808350782116025e-06, + "loss": 0.2552, + "step": 51989 + }, + { + "epoch": 19.2, + "learning_rate": 1.0807981278482573e-06, + "loss": 0.1823, + "step": 51990 + }, + { + "epoch": 19.2, + "learning_rate": 1.080761177484912e-06, + "loss": 0.181, + "step": 51991 + }, + { + "epoch": 19.2, + "learning_rate": 1.0807242271215666e-06, + "loss": 0.1719, + "step": 51992 + }, + { + "epoch": 19.2, + "learning_rate": 1.0806872767582214e-06, + "loss": 0.2518, + "step": 51993 + }, + { + "epoch": 19.2, + "learning_rate": 1.0806503263948763e-06, + "loss": 0.2876, + "step": 51994 + }, + { + "epoch": 19.2, + "learning_rate": 1.080613376031531e-06, + "loss": 0.2049, + "step": 51995 + }, + { + "epoch": 19.2, + "learning_rate": 1.0805764256681858e-06, + "loss": 0.1562, + "step": 51996 + }, + { + "epoch": 19.2, + "learning_rate": 1.0805394753048406e-06, + "loss": 0.3399, + "step": 51997 + }, + { + "epoch": 19.2, + "learning_rate": 1.0805025249414952e-06, + "loss": 0.1978, + "step": 51998 + }, + { + "epoch": 19.2, + "learning_rate": 1.08046557457815e-06, + "loss": 0.2251, + "step": 51999 + }, + { + "epoch": 19.2, + "learning_rate": 1.080428624214805e-06, + "loss": 0.1251, + "step": 52000 + }, + { + "epoch": 19.2, + "learning_rate": 1.0803916738514596e-06, + "loss": 0.3242, + "step": 52001 + }, + { + "epoch": 19.2, + "learning_rate": 1.0803547234881142e-06, + "loss": 0.1247, + "step": 52002 + }, + { + "epoch": 19.2, + "learning_rate": 1.080317773124769e-06, + "loss": 0.1364, + "step": 52003 + }, + { + "epoch": 19.2, + "learning_rate": 1.080280822761424e-06, + "loss": 0.3026, + "step": 52004 + }, + { + "epoch": 19.2, + "learning_rate": 1.0802438723980785e-06, + "loss": 0.2435, + "step": 52005 + }, + { + "epoch": 19.2, + "learning_rate": 1.0802069220347334e-06, + "loss": 0.2377, + "step": 52006 + }, + { + "epoch": 19.2, + "learning_rate": 1.0801699716713882e-06, + "loss": 0.2091, + "step": 52007 + }, + { + "epoch": 19.21, + "learning_rate": 1.0801330213080429e-06, + "loss": 0.2592, + "step": 52008 + }, + { + "epoch": 19.21, + "learning_rate": 1.0800960709446977e-06, + "loss": 0.2333, + "step": 52009 + }, + { + "epoch": 19.21, + "learning_rate": 1.0800591205813526e-06, + "loss": 0.2348, + "step": 52010 + }, + { + "epoch": 19.21, + "learning_rate": 1.0800221702180072e-06, + "loss": 0.3247, + "step": 52011 + }, + { + "epoch": 19.21, + "learning_rate": 1.0799852198546618e-06, + "loss": 0.192, + "step": 52012 + }, + { + "epoch": 19.21, + "learning_rate": 1.0799482694913167e-06, + "loss": 0.1126, + "step": 52013 + }, + { + "epoch": 19.21, + "learning_rate": 1.0799113191279713e-06, + "loss": 0.216, + "step": 52014 + }, + { + "epoch": 19.21, + "learning_rate": 1.0798743687646262e-06, + "loss": 0.2346, + "step": 52015 + }, + { + "epoch": 19.21, + "learning_rate": 1.079837418401281e-06, + "loss": 0.2167, + "step": 52016 + }, + { + "epoch": 19.21, + "learning_rate": 1.0798004680379357e-06, + "loss": 0.2298, + "step": 52017 + }, + { + "epoch": 19.21, + "learning_rate": 1.0797635176745905e-06, + "loss": 0.1953, + "step": 52018 + }, + { + "epoch": 19.21, + "learning_rate": 1.0797265673112454e-06, + "loss": 0.3155, + "step": 52019 + }, + { + "epoch": 19.21, + "learning_rate": 1.0796896169479e-06, + "loss": 0.2437, + "step": 52020 + }, + { + "epoch": 19.21, + "learning_rate": 1.0796526665845546e-06, + "loss": 0.2432, + "step": 52021 + }, + { + "epoch": 19.21, + "learning_rate": 1.0796157162212095e-06, + "loss": 0.1939, + "step": 52022 + }, + { + "epoch": 19.21, + "learning_rate": 1.0795787658578643e-06, + "loss": 0.2908, + "step": 52023 + }, + { + "epoch": 19.21, + "learning_rate": 1.079541815494519e-06, + "loss": 0.2772, + "step": 52024 + }, + { + "epoch": 19.21, + "learning_rate": 1.0795048651311738e-06, + "loss": 0.1356, + "step": 52025 + }, + { + "epoch": 19.21, + "learning_rate": 1.0794679147678287e-06, + "loss": 0.2327, + "step": 52026 + }, + { + "epoch": 19.21, + "learning_rate": 1.0794309644044833e-06, + "loss": 0.2102, + "step": 52027 + }, + { + "epoch": 19.21, + "learning_rate": 1.0793940140411381e-06, + "loss": 0.2152, + "step": 52028 + }, + { + "epoch": 19.21, + "learning_rate": 1.079357063677793e-06, + "loss": 0.1598, + "step": 52029 + }, + { + "epoch": 19.21, + "learning_rate": 1.0793201133144476e-06, + "loss": 0.2053, + "step": 52030 + }, + { + "epoch": 19.21, + "learning_rate": 1.0792831629511023e-06, + "loss": 0.0945, + "step": 52031 + }, + { + "epoch": 19.21, + "learning_rate": 1.0792462125877571e-06, + "loss": 0.2181, + "step": 52032 + }, + { + "epoch": 19.21, + "learning_rate": 1.0792092622244118e-06, + "loss": 0.2784, + "step": 52033 + }, + { + "epoch": 19.21, + "learning_rate": 1.0791723118610666e-06, + "loss": 0.2316, + "step": 52034 + }, + { + "epoch": 19.22, + "learning_rate": 1.0791353614977215e-06, + "loss": 0.1023, + "step": 52035 + }, + { + "epoch": 19.22, + "learning_rate": 1.079098411134376e-06, + "loss": 0.1339, + "step": 52036 + }, + { + "epoch": 19.22, + "learning_rate": 1.079061460771031e-06, + "loss": 0.0702, + "step": 52037 + }, + { + "epoch": 19.22, + "learning_rate": 1.0790245104076858e-06, + "loss": 0.3076, + "step": 52038 + }, + { + "epoch": 19.22, + "learning_rate": 1.0789875600443404e-06, + "loss": 0.2866, + "step": 52039 + }, + { + "epoch": 19.22, + "learning_rate": 1.0789506096809953e-06, + "loss": 0.1634, + "step": 52040 + }, + { + "epoch": 19.22, + "learning_rate": 1.07891365931765e-06, + "loss": 0.2523, + "step": 52041 + }, + { + "epoch": 19.22, + "learning_rate": 1.0788767089543048e-06, + "loss": 0.1681, + "step": 52042 + }, + { + "epoch": 19.22, + "learning_rate": 1.0788397585909594e-06, + "loss": 0.1328, + "step": 52043 + }, + { + "epoch": 19.22, + "learning_rate": 1.0788028082276142e-06, + "loss": 0.115, + "step": 52044 + }, + { + "epoch": 19.22, + "learning_rate": 1.078765857864269e-06, + "loss": 0.1725, + "step": 52045 + }, + { + "epoch": 19.22, + "learning_rate": 1.0787289075009237e-06, + "loss": 0.1434, + "step": 52046 + }, + { + "epoch": 19.22, + "learning_rate": 1.0786919571375786e-06, + "loss": 0.358, + "step": 52047 + }, + { + "epoch": 19.22, + "learning_rate": 1.0786550067742334e-06, + "loss": 0.3086, + "step": 52048 + }, + { + "epoch": 19.22, + "learning_rate": 1.078618056410888e-06, + "loss": 0.1754, + "step": 52049 + }, + { + "epoch": 19.22, + "learning_rate": 1.078581106047543e-06, + "loss": 0.1697, + "step": 52050 + }, + { + "epoch": 19.22, + "learning_rate": 1.0785441556841975e-06, + "loss": 0.2119, + "step": 52051 + }, + { + "epoch": 19.22, + "learning_rate": 1.0785072053208524e-06, + "loss": 0.1429, + "step": 52052 + }, + { + "epoch": 19.22, + "learning_rate": 1.078470254957507e-06, + "loss": 0.2132, + "step": 52053 + }, + { + "epoch": 19.22, + "learning_rate": 1.0784333045941619e-06, + "loss": 0.2669, + "step": 52054 + }, + { + "epoch": 19.22, + "learning_rate": 1.0783963542308167e-06, + "loss": 0.1985, + "step": 52055 + }, + { + "epoch": 19.22, + "learning_rate": 1.0783594038674714e-06, + "loss": 0.1506, + "step": 52056 + }, + { + "epoch": 19.22, + "learning_rate": 1.0783224535041262e-06, + "loss": 0.1712, + "step": 52057 + }, + { + "epoch": 19.22, + "learning_rate": 1.078285503140781e-06, + "loss": 0.1891, + "step": 52058 + }, + { + "epoch": 19.22, + "learning_rate": 1.0782485527774357e-06, + "loss": 0.1688, + "step": 52059 + }, + { + "epoch": 19.22, + "learning_rate": 1.0782116024140905e-06, + "loss": 0.3037, + "step": 52060 + }, + { + "epoch": 19.22, + "learning_rate": 1.0781746520507452e-06, + "loss": 0.1778, + "step": 52061 + }, + { + "epoch": 19.23, + "learning_rate": 1.0781377016873998e-06, + "loss": 0.2054, + "step": 52062 + }, + { + "epoch": 19.23, + "learning_rate": 1.0781007513240547e-06, + "loss": 0.1782, + "step": 52063 + }, + { + "epoch": 19.23, + "learning_rate": 1.0780638009607095e-06, + "loss": 0.1251, + "step": 52064 + }, + { + "epoch": 19.23, + "learning_rate": 1.0780268505973641e-06, + "loss": 0.2428, + "step": 52065 + }, + { + "epoch": 19.23, + "learning_rate": 1.077989900234019e-06, + "loss": 0.2235, + "step": 52066 + }, + { + "epoch": 19.23, + "learning_rate": 1.0779529498706738e-06, + "loss": 0.1444, + "step": 52067 + }, + { + "epoch": 19.23, + "learning_rate": 1.0779159995073285e-06, + "loss": 0.3137, + "step": 52068 + }, + { + "epoch": 19.23, + "learning_rate": 1.0778790491439833e-06, + "loss": 0.181, + "step": 52069 + }, + { + "epoch": 19.23, + "learning_rate": 1.0778420987806382e-06, + "loss": 0.4267, + "step": 52070 + }, + { + "epoch": 19.23, + "learning_rate": 1.0778051484172928e-06, + "loss": 0.149, + "step": 52071 + }, + { + "epoch": 19.23, + "learning_rate": 1.0777681980539474e-06, + "loss": 0.192, + "step": 52072 + }, + { + "epoch": 19.23, + "learning_rate": 1.0777312476906023e-06, + "loss": 0.2081, + "step": 52073 + }, + { + "epoch": 19.23, + "learning_rate": 1.0776942973272571e-06, + "loss": 0.1666, + "step": 52074 + }, + { + "epoch": 19.23, + "learning_rate": 1.0776573469639118e-06, + "loss": 0.2313, + "step": 52075 + }, + { + "epoch": 19.23, + "learning_rate": 1.0776203966005666e-06, + "loss": 0.3002, + "step": 52076 + }, + { + "epoch": 19.23, + "learning_rate": 1.0775834462372215e-06, + "loss": 0.3081, + "step": 52077 + }, + { + "epoch": 19.23, + "learning_rate": 1.0775464958738761e-06, + "loss": 0.1985, + "step": 52078 + }, + { + "epoch": 19.23, + "learning_rate": 1.077509545510531e-06, + "loss": 0.2823, + "step": 52079 + }, + { + "epoch": 19.23, + "learning_rate": 1.0774725951471858e-06, + "loss": 0.2936, + "step": 52080 + }, + { + "epoch": 19.23, + "learning_rate": 1.0774356447838402e-06, + "loss": 0.2659, + "step": 52081 + }, + { + "epoch": 19.23, + "learning_rate": 1.077398694420495e-06, + "loss": 0.2009, + "step": 52082 + }, + { + "epoch": 19.23, + "learning_rate": 1.07736174405715e-06, + "loss": 0.2158, + "step": 52083 + }, + { + "epoch": 19.23, + "learning_rate": 1.0773247936938046e-06, + "loss": 0.1868, + "step": 52084 + }, + { + "epoch": 19.23, + "learning_rate": 1.0772878433304594e-06, + "loss": 0.2063, + "step": 52085 + }, + { + "epoch": 19.23, + "learning_rate": 1.0772508929671143e-06, + "loss": 0.1341, + "step": 52086 + }, + { + "epoch": 19.23, + "learning_rate": 1.077213942603769e-06, + "loss": 0.2586, + "step": 52087 + }, + { + "epoch": 19.23, + "learning_rate": 1.0771769922404237e-06, + "loss": 0.2547, + "step": 52088 + }, + { + "epoch": 19.24, + "learning_rate": 1.0771400418770786e-06, + "loss": 0.3032, + "step": 52089 + }, + { + "epoch": 19.24, + "learning_rate": 1.0771030915137332e-06, + "loss": 0.1042, + "step": 52090 + }, + { + "epoch": 19.24, + "learning_rate": 1.0770661411503879e-06, + "loss": 0.2175, + "step": 52091 + }, + { + "epoch": 19.24, + "learning_rate": 1.0770291907870427e-06, + "loss": 0.2355, + "step": 52092 + }, + { + "epoch": 19.24, + "learning_rate": 1.0769922404236976e-06, + "loss": 0.1963, + "step": 52093 + }, + { + "epoch": 19.24, + "learning_rate": 1.0769552900603522e-06, + "loss": 0.1635, + "step": 52094 + }, + { + "epoch": 19.24, + "learning_rate": 1.076918339697007e-06, + "loss": 0.1305, + "step": 52095 + }, + { + "epoch": 19.24, + "learning_rate": 1.076881389333662e-06, + "loss": 0.2414, + "step": 52096 + }, + { + "epoch": 19.24, + "learning_rate": 1.0768444389703165e-06, + "loss": 0.4428, + "step": 52097 + }, + { + "epoch": 19.24, + "learning_rate": 1.0768074886069714e-06, + "loss": 0.1171, + "step": 52098 + }, + { + "epoch": 19.24, + "learning_rate": 1.0767705382436262e-06, + "loss": 0.1556, + "step": 52099 + }, + { + "epoch": 19.24, + "learning_rate": 1.0767335878802809e-06, + "loss": 0.2735, + "step": 52100 + }, + { + "epoch": 19.24, + "learning_rate": 1.0766966375169355e-06, + "loss": 0.1645, + "step": 52101 + }, + { + "epoch": 19.24, + "learning_rate": 1.0766596871535904e-06, + "loss": 0.1604, + "step": 52102 + }, + { + "epoch": 19.24, + "learning_rate": 1.076622736790245e-06, + "loss": 0.3955, + "step": 52103 + }, + { + "epoch": 19.24, + "learning_rate": 1.0765857864268998e-06, + "loss": 0.2161, + "step": 52104 + }, + { + "epoch": 19.24, + "learning_rate": 1.0765488360635547e-06, + "loss": 0.2821, + "step": 52105 + }, + { + "epoch": 19.24, + "learning_rate": 1.0765118857002093e-06, + "loss": 0.2654, + "step": 52106 + }, + { + "epoch": 19.24, + "learning_rate": 1.0764749353368642e-06, + "loss": 0.2824, + "step": 52107 + }, + { + "epoch": 19.24, + "learning_rate": 1.076437984973519e-06, + "loss": 0.127, + "step": 52108 + }, + { + "epoch": 19.24, + "learning_rate": 1.0764010346101737e-06, + "loss": 0.271, + "step": 52109 + }, + { + "epoch": 19.24, + "learning_rate": 1.0763640842468285e-06, + "loss": 0.1772, + "step": 52110 + }, + { + "epoch": 19.24, + "learning_rate": 1.0763271338834831e-06, + "loss": 0.2071, + "step": 52111 + }, + { + "epoch": 19.24, + "learning_rate": 1.076290183520138e-06, + "loss": 0.1663, + "step": 52112 + }, + { + "epoch": 19.24, + "learning_rate": 1.0762532331567926e-06, + "loss": 0.2086, + "step": 52113 + }, + { + "epoch": 19.24, + "learning_rate": 1.0762162827934475e-06, + "loss": 0.2372, + "step": 52114 + }, + { + "epoch": 19.24, + "learning_rate": 1.0761793324301023e-06, + "loss": 0.2415, + "step": 52115 + }, + { + "epoch": 19.25, + "learning_rate": 1.076142382066757e-06, + "loss": 0.2788, + "step": 52116 + }, + { + "epoch": 19.25, + "learning_rate": 1.0761054317034118e-06, + "loss": 0.2082, + "step": 52117 + }, + { + "epoch": 19.25, + "learning_rate": 1.0760684813400667e-06, + "loss": 0.2275, + "step": 52118 + }, + { + "epoch": 19.25, + "learning_rate": 1.0760315309767213e-06, + "loss": 0.1647, + "step": 52119 + }, + { + "epoch": 19.25, + "learning_rate": 1.0759945806133761e-06, + "loss": 0.2694, + "step": 52120 + }, + { + "epoch": 19.25, + "learning_rate": 1.0759576302500308e-06, + "loss": 0.1552, + "step": 52121 + }, + { + "epoch": 19.25, + "learning_rate": 1.0759206798866856e-06, + "loss": 0.2923, + "step": 52122 + }, + { + "epoch": 19.25, + "learning_rate": 1.0758837295233403e-06, + "loss": 0.1979, + "step": 52123 + }, + { + "epoch": 19.25, + "learning_rate": 1.075846779159995e-06, + "loss": 0.1731, + "step": 52124 + }, + { + "epoch": 19.25, + "learning_rate": 1.07580982879665e-06, + "loss": 0.1918, + "step": 52125 + }, + { + "epoch": 19.25, + "learning_rate": 1.0757728784333046e-06, + "loss": 0.2382, + "step": 52126 + }, + { + "epoch": 19.25, + "learning_rate": 1.0757359280699594e-06, + "loss": 0.1971, + "step": 52127 + }, + { + "epoch": 19.25, + "learning_rate": 1.0756989777066143e-06, + "loss": 0.1417, + "step": 52128 + }, + { + "epoch": 19.25, + "learning_rate": 1.075662027343269e-06, + "loss": 0.1732, + "step": 52129 + }, + { + "epoch": 19.25, + "learning_rate": 1.0756250769799238e-06, + "loss": 0.1618, + "step": 52130 + }, + { + "epoch": 19.25, + "learning_rate": 1.0755881266165784e-06, + "loss": 0.2767, + "step": 52131 + }, + { + "epoch": 19.25, + "learning_rate": 1.075551176253233e-06, + "loss": 0.157, + "step": 52132 + }, + { + "epoch": 19.25, + "learning_rate": 1.0755142258898879e-06, + "loss": 0.2731, + "step": 52133 + }, + { + "epoch": 19.25, + "learning_rate": 1.0754772755265427e-06, + "loss": 0.3585, + "step": 52134 + }, + { + "epoch": 19.25, + "learning_rate": 1.0754403251631974e-06, + "loss": 0.2041, + "step": 52135 + }, + { + "epoch": 19.25, + "learning_rate": 1.0754033747998522e-06, + "loss": 0.1151, + "step": 52136 + }, + { + "epoch": 19.25, + "learning_rate": 1.075366424436507e-06, + "loss": 0.1803, + "step": 52137 + }, + { + "epoch": 19.25, + "learning_rate": 1.0753294740731617e-06, + "loss": 0.2326, + "step": 52138 + }, + { + "epoch": 19.25, + "learning_rate": 1.0752925237098166e-06, + "loss": 0.1348, + "step": 52139 + }, + { + "epoch": 19.25, + "learning_rate": 1.0752555733464714e-06, + "loss": 0.1039, + "step": 52140 + }, + { + "epoch": 19.25, + "learning_rate": 1.075218622983126e-06, + "loss": 0.171, + "step": 52141 + }, + { + "epoch": 19.25, + "learning_rate": 1.0751816726197807e-06, + "loss": 0.2101, + "step": 52142 + }, + { + "epoch": 19.26, + "learning_rate": 1.0751447222564355e-06, + "loss": 0.2332, + "step": 52143 + }, + { + "epoch": 19.26, + "learning_rate": 1.0751077718930904e-06, + "loss": 0.2074, + "step": 52144 + }, + { + "epoch": 19.26, + "learning_rate": 1.075070821529745e-06, + "loss": 0.2206, + "step": 52145 + }, + { + "epoch": 19.26, + "learning_rate": 1.0750338711663999e-06, + "loss": 0.0849, + "step": 52146 + }, + { + "epoch": 19.26, + "learning_rate": 1.0749969208030547e-06, + "loss": 0.2577, + "step": 52147 + }, + { + "epoch": 19.26, + "learning_rate": 1.0749599704397093e-06, + "loss": 0.2015, + "step": 52148 + }, + { + "epoch": 19.26, + "learning_rate": 1.0749230200763642e-06, + "loss": 0.1716, + "step": 52149 + }, + { + "epoch": 19.26, + "learning_rate": 1.074886069713019e-06, + "loss": 0.3041, + "step": 52150 + }, + { + "epoch": 19.26, + "learning_rate": 1.0748491193496735e-06, + "loss": 0.327, + "step": 52151 + }, + { + "epoch": 19.26, + "learning_rate": 1.0748121689863283e-06, + "loss": 0.2351, + "step": 52152 + }, + { + "epoch": 19.26, + "learning_rate": 1.0747752186229832e-06, + "loss": 0.3828, + "step": 52153 + }, + { + "epoch": 19.26, + "learning_rate": 1.0747382682596378e-06, + "loss": 0.2807, + "step": 52154 + }, + { + "epoch": 19.26, + "learning_rate": 1.0747013178962926e-06, + "loss": 0.3844, + "step": 52155 + }, + { + "epoch": 19.26, + "learning_rate": 1.0746643675329475e-06, + "loss": 0.2205, + "step": 52156 + }, + { + "epoch": 19.26, + "learning_rate": 1.0746274171696021e-06, + "loss": 0.2462, + "step": 52157 + }, + { + "epoch": 19.26, + "learning_rate": 1.074590466806257e-06, + "loss": 0.2946, + "step": 52158 + }, + { + "epoch": 19.26, + "learning_rate": 1.0745535164429118e-06, + "loss": 0.1548, + "step": 52159 + }, + { + "epoch": 19.26, + "learning_rate": 1.0745165660795665e-06, + "loss": 0.2234, + "step": 52160 + }, + { + "epoch": 19.26, + "learning_rate": 1.074479615716221e-06, + "loss": 0.141, + "step": 52161 + }, + { + "epoch": 19.26, + "learning_rate": 1.074442665352876e-06, + "loss": 0.2614, + "step": 52162 + }, + { + "epoch": 19.26, + "learning_rate": 1.0744057149895308e-06, + "loss": 0.2551, + "step": 52163 + }, + { + "epoch": 19.26, + "learning_rate": 1.0743687646261854e-06, + "loss": 0.2723, + "step": 52164 + }, + { + "epoch": 19.26, + "learning_rate": 1.0743318142628403e-06, + "loss": 0.158, + "step": 52165 + }, + { + "epoch": 19.26, + "learning_rate": 1.0742948638994951e-06, + "loss": 0.1817, + "step": 52166 + }, + { + "epoch": 19.26, + "learning_rate": 1.0742579135361498e-06, + "loss": 0.2547, + "step": 52167 + }, + { + "epoch": 19.26, + "learning_rate": 1.0742209631728046e-06, + "loss": 0.275, + "step": 52168 + }, + { + "epoch": 19.26, + "learning_rate": 1.0741840128094595e-06, + "loss": 0.219, + "step": 52169 + }, + { + "epoch": 19.27, + "learning_rate": 1.074147062446114e-06, + "loss": 0.1974, + "step": 52170 + }, + { + "epoch": 19.27, + "learning_rate": 1.0741101120827687e-06, + "loss": 0.2277, + "step": 52171 + }, + { + "epoch": 19.27, + "learning_rate": 1.0740731617194236e-06, + "loss": 0.2041, + "step": 52172 + }, + { + "epoch": 19.27, + "learning_rate": 1.0740362113560782e-06, + "loss": 0.26, + "step": 52173 + }, + { + "epoch": 19.27, + "learning_rate": 1.073999260992733e-06, + "loss": 0.2491, + "step": 52174 + }, + { + "epoch": 19.27, + "learning_rate": 1.073962310629388e-06, + "loss": 0.3432, + "step": 52175 + }, + { + "epoch": 19.27, + "learning_rate": 1.0739253602660426e-06, + "loss": 0.2895, + "step": 52176 + }, + { + "epoch": 19.27, + "learning_rate": 1.0738884099026974e-06, + "loss": 0.1724, + "step": 52177 + }, + { + "epoch": 19.27, + "learning_rate": 1.0738514595393522e-06, + "loss": 0.1772, + "step": 52178 + }, + { + "epoch": 19.27, + "learning_rate": 1.0738145091760069e-06, + "loss": 0.3598, + "step": 52179 + }, + { + "epoch": 19.27, + "learning_rate": 1.0737775588126617e-06, + "loss": 0.1758, + "step": 52180 + }, + { + "epoch": 19.27, + "learning_rate": 1.0737406084493164e-06, + "loss": 0.0947, + "step": 52181 + }, + { + "epoch": 19.27, + "learning_rate": 1.0737036580859712e-06, + "loss": 0.163, + "step": 52182 + }, + { + "epoch": 19.27, + "learning_rate": 1.0736667077226259e-06, + "loss": 0.1303, + "step": 52183 + }, + { + "epoch": 19.27, + "learning_rate": 1.0736297573592807e-06, + "loss": 0.1742, + "step": 52184 + }, + { + "epoch": 19.27, + "learning_rate": 1.0735928069959356e-06, + "loss": 0.2744, + "step": 52185 + }, + { + "epoch": 19.27, + "learning_rate": 1.0735558566325902e-06, + "loss": 0.1792, + "step": 52186 + }, + { + "epoch": 19.27, + "learning_rate": 1.073518906269245e-06, + "loss": 0.1751, + "step": 52187 + }, + { + "epoch": 19.27, + "learning_rate": 1.0734819559058999e-06, + "loss": 0.3695, + "step": 52188 + }, + { + "epoch": 19.27, + "learning_rate": 1.0734450055425545e-06, + "loss": 0.1754, + "step": 52189 + }, + { + "epoch": 19.27, + "learning_rate": 1.0734080551792094e-06, + "loss": 0.2325, + "step": 52190 + }, + { + "epoch": 19.27, + "learning_rate": 1.073371104815864e-06, + "loss": 0.2359, + "step": 52191 + }, + { + "epoch": 19.27, + "learning_rate": 1.0733341544525189e-06, + "loss": 0.2324, + "step": 52192 + }, + { + "epoch": 19.27, + "learning_rate": 1.0732972040891735e-06, + "loss": 0.2675, + "step": 52193 + }, + { + "epoch": 19.27, + "learning_rate": 1.0732602537258283e-06, + "loss": 0.1817, + "step": 52194 + }, + { + "epoch": 19.27, + "learning_rate": 1.0732233033624832e-06, + "loss": 0.2617, + "step": 52195 + }, + { + "epoch": 19.27, + "learning_rate": 1.0731863529991378e-06, + "loss": 0.2075, + "step": 52196 + }, + { + "epoch": 19.28, + "learning_rate": 1.0731494026357927e-06, + "loss": 0.1802, + "step": 52197 + }, + { + "epoch": 19.28, + "learning_rate": 1.0731124522724475e-06, + "loss": 0.2636, + "step": 52198 + }, + { + "epoch": 19.28, + "learning_rate": 1.0730755019091022e-06, + "loss": 0.2969, + "step": 52199 + }, + { + "epoch": 19.28, + "learning_rate": 1.073038551545757e-06, + "loss": 0.1655, + "step": 52200 + }, + { + "epoch": 19.28, + "learning_rate": 1.0730016011824116e-06, + "loss": 0.199, + "step": 52201 + }, + { + "epoch": 19.28, + "learning_rate": 1.0729646508190663e-06, + "loss": 0.1905, + "step": 52202 + }, + { + "epoch": 19.28, + "learning_rate": 1.0729277004557211e-06, + "loss": 0.3412, + "step": 52203 + }, + { + "epoch": 19.28, + "learning_rate": 1.072890750092376e-06, + "loss": 0.1981, + "step": 52204 + }, + { + "epoch": 19.28, + "learning_rate": 1.0728537997290306e-06, + "loss": 0.3, + "step": 52205 + }, + { + "epoch": 19.28, + "learning_rate": 1.0728168493656855e-06, + "loss": 0.1573, + "step": 52206 + }, + { + "epoch": 19.28, + "learning_rate": 1.0727798990023403e-06, + "loss": 0.1852, + "step": 52207 + }, + { + "epoch": 19.28, + "learning_rate": 1.072742948638995e-06, + "loss": 0.2849, + "step": 52208 + }, + { + "epoch": 19.28, + "learning_rate": 1.0727059982756498e-06, + "loss": 0.2814, + "step": 52209 + }, + { + "epoch": 19.28, + "learning_rate": 1.0726690479123046e-06, + "loss": 0.2281, + "step": 52210 + }, + { + "epoch": 19.28, + "learning_rate": 1.0726320975489593e-06, + "loss": 0.1916, + "step": 52211 + }, + { + "epoch": 19.28, + "learning_rate": 1.072595147185614e-06, + "loss": 0.3648, + "step": 52212 + }, + { + "epoch": 19.28, + "learning_rate": 1.0725581968222688e-06, + "loss": 0.0833, + "step": 52213 + }, + { + "epoch": 19.28, + "learning_rate": 1.0725212464589236e-06, + "loss": 0.2576, + "step": 52214 + }, + { + "epoch": 19.28, + "learning_rate": 1.0724842960955782e-06, + "loss": 0.1522, + "step": 52215 + }, + { + "epoch": 19.28, + "learning_rate": 1.072447345732233e-06, + "loss": 0.1371, + "step": 52216 + }, + { + "epoch": 19.28, + "learning_rate": 1.072410395368888e-06, + "loss": 0.1653, + "step": 52217 + }, + { + "epoch": 19.28, + "learning_rate": 1.0723734450055426e-06, + "loss": 0.2022, + "step": 52218 + }, + { + "epoch": 19.28, + "learning_rate": 1.0723364946421974e-06, + "loss": 0.2579, + "step": 52219 + }, + { + "epoch": 19.28, + "learning_rate": 1.0722995442788523e-06, + "loss": 0.211, + "step": 52220 + }, + { + "epoch": 19.28, + "learning_rate": 1.0722625939155067e-06, + "loss": 0.2367, + "step": 52221 + }, + { + "epoch": 19.28, + "learning_rate": 1.0722256435521615e-06, + "loss": 0.2938, + "step": 52222 + }, + { + "epoch": 19.28, + "learning_rate": 1.0721886931888164e-06, + "loss": 0.1694, + "step": 52223 + }, + { + "epoch": 19.29, + "learning_rate": 1.072151742825471e-06, + "loss": 0.1517, + "step": 52224 + }, + { + "epoch": 19.29, + "learning_rate": 1.0721147924621259e-06, + "loss": 0.2858, + "step": 52225 + }, + { + "epoch": 19.29, + "learning_rate": 1.0720778420987807e-06, + "loss": 0.1659, + "step": 52226 + }, + { + "epoch": 19.29, + "learning_rate": 1.0720408917354354e-06, + "loss": 0.3334, + "step": 52227 + }, + { + "epoch": 19.29, + "learning_rate": 1.0720039413720902e-06, + "loss": 0.2692, + "step": 52228 + }, + { + "epoch": 19.29, + "learning_rate": 1.071966991008745e-06, + "loss": 0.2087, + "step": 52229 + }, + { + "epoch": 19.29, + "learning_rate": 1.0719300406453997e-06, + "loss": 0.2736, + "step": 52230 + }, + { + "epoch": 19.29, + "learning_rate": 1.0718930902820543e-06, + "loss": 0.151, + "step": 52231 + }, + { + "epoch": 19.29, + "learning_rate": 1.0718561399187092e-06, + "loss": 0.1825, + "step": 52232 + }, + { + "epoch": 19.29, + "learning_rate": 1.071819189555364e-06, + "loss": 0.2834, + "step": 52233 + }, + { + "epoch": 19.29, + "learning_rate": 1.0717822391920187e-06, + "loss": 0.1557, + "step": 52234 + }, + { + "epoch": 19.29, + "learning_rate": 1.0717452888286735e-06, + "loss": 0.1897, + "step": 52235 + }, + { + "epoch": 19.29, + "learning_rate": 1.0717083384653284e-06, + "loss": 0.1921, + "step": 52236 + }, + { + "epoch": 19.29, + "learning_rate": 1.071671388101983e-06, + "loss": 0.1672, + "step": 52237 + }, + { + "epoch": 19.29, + "learning_rate": 1.0716344377386378e-06, + "loss": 0.3309, + "step": 52238 + }, + { + "epoch": 19.29, + "learning_rate": 1.0715974873752927e-06, + "loss": 0.2782, + "step": 52239 + }, + { + "epoch": 19.29, + "learning_rate": 1.0715605370119473e-06, + "loss": 0.2069, + "step": 52240 + }, + { + "epoch": 19.29, + "learning_rate": 1.071523586648602e-06, + "loss": 0.2723, + "step": 52241 + }, + { + "epoch": 19.29, + "learning_rate": 1.0714866362852568e-06, + "loss": 0.3552, + "step": 52242 + }, + { + "epoch": 19.29, + "learning_rate": 1.0714496859219117e-06, + "loss": 0.1941, + "step": 52243 + }, + { + "epoch": 19.29, + "learning_rate": 1.0714127355585663e-06, + "loss": 0.28, + "step": 52244 + }, + { + "epoch": 19.29, + "learning_rate": 1.0713757851952211e-06, + "loss": 0.2697, + "step": 52245 + }, + { + "epoch": 19.29, + "learning_rate": 1.071338834831876e-06, + "loss": 0.1888, + "step": 52246 + }, + { + "epoch": 19.29, + "learning_rate": 1.0713018844685306e-06, + "loss": 0.2946, + "step": 52247 + }, + { + "epoch": 19.29, + "learning_rate": 1.0712649341051855e-06, + "loss": 0.1352, + "step": 52248 + }, + { + "epoch": 19.29, + "learning_rate": 1.0712279837418401e-06, + "loss": 0.1477, + "step": 52249 + }, + { + "epoch": 19.29, + "learning_rate": 1.0711910333784948e-06, + "loss": 0.1997, + "step": 52250 + }, + { + "epoch": 19.3, + "learning_rate": 1.0711540830151496e-06, + "loss": 0.1917, + "step": 52251 + }, + { + "epoch": 19.3, + "learning_rate": 1.0711171326518045e-06, + "loss": 0.2107, + "step": 52252 + }, + { + "epoch": 19.3, + "learning_rate": 1.071080182288459e-06, + "loss": 0.2366, + "step": 52253 + }, + { + "epoch": 19.3, + "learning_rate": 1.071043231925114e-06, + "loss": 0.2376, + "step": 52254 + }, + { + "epoch": 19.3, + "learning_rate": 1.0710062815617688e-06, + "loss": 0.1653, + "step": 52255 + }, + { + "epoch": 19.3, + "learning_rate": 1.0709693311984234e-06, + "loss": 0.2392, + "step": 52256 + }, + { + "epoch": 19.3, + "learning_rate": 1.0709323808350783e-06, + "loss": 0.1945, + "step": 52257 + }, + { + "epoch": 19.3, + "learning_rate": 1.0708954304717331e-06, + "loss": 0.1944, + "step": 52258 + }, + { + "epoch": 19.3, + "learning_rate": 1.0708584801083878e-06, + "loss": 0.2437, + "step": 52259 + }, + { + "epoch": 19.3, + "learning_rate": 1.0708215297450424e-06, + "loss": 0.1897, + "step": 52260 + }, + { + "epoch": 19.3, + "learning_rate": 1.0707845793816972e-06, + "loss": 0.1534, + "step": 52261 + }, + { + "epoch": 19.3, + "learning_rate": 1.070747629018352e-06, + "loss": 0.2218, + "step": 52262 + }, + { + "epoch": 19.3, + "learning_rate": 1.0707106786550067e-06, + "loss": 0.2539, + "step": 52263 + }, + { + "epoch": 19.3, + "learning_rate": 1.0706737282916616e-06, + "loss": 0.3096, + "step": 52264 + }, + { + "epoch": 19.3, + "learning_rate": 1.0706367779283164e-06, + "loss": 0.1597, + "step": 52265 + }, + { + "epoch": 19.3, + "learning_rate": 1.070599827564971e-06, + "loss": 0.2222, + "step": 52266 + }, + { + "epoch": 19.3, + "learning_rate": 1.070562877201626e-06, + "loss": 0.1877, + "step": 52267 + }, + { + "epoch": 19.3, + "learning_rate": 1.0705259268382808e-06, + "loss": 0.3069, + "step": 52268 + }, + { + "epoch": 19.3, + "learning_rate": 1.0704889764749354e-06, + "loss": 0.1687, + "step": 52269 + }, + { + "epoch": 19.3, + "learning_rate": 1.07045202611159e-06, + "loss": 0.2029, + "step": 52270 + }, + { + "epoch": 19.3, + "learning_rate": 1.0704150757482449e-06, + "loss": 0.151, + "step": 52271 + }, + { + "epoch": 19.3, + "learning_rate": 1.0703781253848995e-06, + "loss": 0.2967, + "step": 52272 + }, + { + "epoch": 19.3, + "learning_rate": 1.0703411750215544e-06, + "loss": 0.2277, + "step": 52273 + }, + { + "epoch": 19.3, + "learning_rate": 1.0703042246582092e-06, + "loss": 0.1676, + "step": 52274 + }, + { + "epoch": 19.3, + "learning_rate": 1.0702672742948638e-06, + "loss": 0.1503, + "step": 52275 + }, + { + "epoch": 19.3, + "learning_rate": 1.0702303239315187e-06, + "loss": 0.2129, + "step": 52276 + }, + { + "epoch": 19.3, + "learning_rate": 1.0701933735681735e-06, + "loss": 0.1503, + "step": 52277 + }, + { + "epoch": 19.31, + "learning_rate": 1.0701564232048282e-06, + "loss": 0.1219, + "step": 52278 + }, + { + "epoch": 19.31, + "learning_rate": 1.070119472841483e-06, + "loss": 0.2859, + "step": 52279 + }, + { + "epoch": 19.31, + "learning_rate": 1.0700825224781377e-06, + "loss": 0.2669, + "step": 52280 + }, + { + "epoch": 19.31, + "learning_rate": 1.0700455721147925e-06, + "loss": 0.2459, + "step": 52281 + }, + { + "epoch": 19.31, + "learning_rate": 1.0700086217514471e-06, + "loss": 0.1893, + "step": 52282 + }, + { + "epoch": 19.31, + "learning_rate": 1.069971671388102e-06, + "loss": 0.0639, + "step": 52283 + }, + { + "epoch": 19.31, + "learning_rate": 1.0699347210247568e-06, + "loss": 0.174, + "step": 52284 + }, + { + "epoch": 19.31, + "learning_rate": 1.0698977706614115e-06, + "loss": 0.2858, + "step": 52285 + }, + { + "epoch": 19.31, + "learning_rate": 1.0698608202980663e-06, + "loss": 0.242, + "step": 52286 + }, + { + "epoch": 19.31, + "learning_rate": 1.0698238699347212e-06, + "loss": 0.2382, + "step": 52287 + }, + { + "epoch": 19.31, + "learning_rate": 1.0697869195713758e-06, + "loss": 0.2868, + "step": 52288 + }, + { + "epoch": 19.31, + "learning_rate": 1.0697499692080307e-06, + "loss": 0.1334, + "step": 52289 + }, + { + "epoch": 19.31, + "learning_rate": 1.0697130188446853e-06, + "loss": 0.2281, + "step": 52290 + }, + { + "epoch": 19.31, + "learning_rate": 1.06967606848134e-06, + "loss": 0.2264, + "step": 52291 + }, + { + "epoch": 19.31, + "learning_rate": 1.0696391181179948e-06, + "loss": 0.244, + "step": 52292 + }, + { + "epoch": 19.31, + "learning_rate": 1.0696021677546496e-06, + "loss": 0.2858, + "step": 52293 + }, + { + "epoch": 19.31, + "learning_rate": 1.0695652173913043e-06, + "loss": 0.2393, + "step": 52294 + }, + { + "epoch": 19.31, + "learning_rate": 1.0695282670279591e-06, + "loss": 0.2465, + "step": 52295 + }, + { + "epoch": 19.31, + "learning_rate": 1.069491316664614e-06, + "loss": 0.204, + "step": 52296 + }, + { + "epoch": 19.31, + "learning_rate": 1.0694543663012686e-06, + "loss": 0.0987, + "step": 52297 + }, + { + "epoch": 19.31, + "learning_rate": 1.0694174159379234e-06, + "loss": 0.2402, + "step": 52298 + }, + { + "epoch": 19.31, + "learning_rate": 1.0693804655745783e-06, + "loss": 0.1572, + "step": 52299 + }, + { + "epoch": 19.31, + "learning_rate": 1.069343515211233e-06, + "loss": 0.2325, + "step": 52300 + }, + { + "epoch": 19.31, + "learning_rate": 1.0693065648478876e-06, + "loss": 0.1752, + "step": 52301 + }, + { + "epoch": 19.31, + "learning_rate": 1.0692696144845424e-06, + "loss": 0.144, + "step": 52302 + }, + { + "epoch": 19.31, + "learning_rate": 1.0692326641211973e-06, + "loss": 0.2904, + "step": 52303 + }, + { + "epoch": 19.31, + "learning_rate": 1.069195713757852e-06, + "loss": 0.1585, + "step": 52304 + }, + { + "epoch": 19.31, + "learning_rate": 1.0691587633945067e-06, + "loss": 0.2043, + "step": 52305 + }, + { + "epoch": 19.32, + "learning_rate": 1.0691218130311616e-06, + "loss": 0.2785, + "step": 52306 + }, + { + "epoch": 19.32, + "learning_rate": 1.0690848626678162e-06, + "loss": 0.2428, + "step": 52307 + }, + { + "epoch": 19.32, + "learning_rate": 1.069047912304471e-06, + "loss": 0.0796, + "step": 52308 + }, + { + "epoch": 19.32, + "learning_rate": 1.069010961941126e-06, + "loss": 0.2633, + "step": 52309 + }, + { + "epoch": 19.32, + "learning_rate": 1.0689740115777806e-06, + "loss": 0.2689, + "step": 52310 + }, + { + "epoch": 19.32, + "learning_rate": 1.0689370612144352e-06, + "loss": 0.1175, + "step": 52311 + }, + { + "epoch": 19.32, + "learning_rate": 1.06890011085109e-06, + "loss": 0.2863, + "step": 52312 + }, + { + "epoch": 19.32, + "learning_rate": 1.068863160487745e-06, + "loss": 0.2685, + "step": 52313 + }, + { + "epoch": 19.32, + "learning_rate": 1.0688262101243995e-06, + "loss": 0.179, + "step": 52314 + }, + { + "epoch": 19.32, + "learning_rate": 1.0687892597610544e-06, + "loss": 0.3121, + "step": 52315 + }, + { + "epoch": 19.32, + "learning_rate": 1.0687523093977092e-06, + "loss": 0.2291, + "step": 52316 + }, + { + "epoch": 19.32, + "learning_rate": 1.0687153590343639e-06, + "loss": 0.2198, + "step": 52317 + }, + { + "epoch": 19.32, + "learning_rate": 1.0686784086710187e-06, + "loss": 0.2245, + "step": 52318 + }, + { + "epoch": 19.32, + "learning_rate": 1.0686414583076736e-06, + "loss": 0.2006, + "step": 52319 + }, + { + "epoch": 19.32, + "learning_rate": 1.068604507944328e-06, + "loss": 0.2585, + "step": 52320 + }, + { + "epoch": 19.32, + "learning_rate": 1.0685675575809828e-06, + "loss": 0.2486, + "step": 52321 + }, + { + "epoch": 19.32, + "learning_rate": 1.0685306072176377e-06, + "loss": 0.2029, + "step": 52322 + }, + { + "epoch": 19.32, + "learning_rate": 1.0684936568542923e-06, + "loss": 0.2395, + "step": 52323 + }, + { + "epoch": 19.32, + "learning_rate": 1.0684567064909472e-06, + "loss": 0.1236, + "step": 52324 + }, + { + "epoch": 19.32, + "learning_rate": 1.068419756127602e-06, + "loss": 0.2857, + "step": 52325 + }, + { + "epoch": 19.32, + "learning_rate": 1.0683828057642567e-06, + "loss": 0.2492, + "step": 52326 + }, + { + "epoch": 19.32, + "learning_rate": 1.0683458554009115e-06, + "loss": 0.2212, + "step": 52327 + }, + { + "epoch": 19.32, + "learning_rate": 1.0683089050375663e-06, + "loss": 0.2198, + "step": 52328 + }, + { + "epoch": 19.32, + "learning_rate": 1.068271954674221e-06, + "loss": 0.2431, + "step": 52329 + }, + { + "epoch": 19.32, + "learning_rate": 1.0682350043108756e-06, + "loss": 0.1915, + "step": 52330 + }, + { + "epoch": 19.32, + "learning_rate": 1.0681980539475305e-06, + "loss": 0.3252, + "step": 52331 + }, + { + "epoch": 19.32, + "learning_rate": 1.0681611035841853e-06, + "loss": 0.3564, + "step": 52332 + }, + { + "epoch": 19.33, + "learning_rate": 1.06812415322084e-06, + "loss": 0.4113, + "step": 52333 + }, + { + "epoch": 19.33, + "learning_rate": 1.0680872028574948e-06, + "loss": 0.1524, + "step": 52334 + }, + { + "epoch": 19.33, + "learning_rate": 1.0680502524941497e-06, + "loss": 0.1003, + "step": 52335 + }, + { + "epoch": 19.33, + "learning_rate": 1.0680133021308043e-06, + "loss": 0.1815, + "step": 52336 + }, + { + "epoch": 19.33, + "learning_rate": 1.0679763517674591e-06, + "loss": 0.2235, + "step": 52337 + }, + { + "epoch": 19.33, + "learning_rate": 1.067939401404114e-06, + "loss": 0.2022, + "step": 52338 + }, + { + "epoch": 19.33, + "learning_rate": 1.0679024510407686e-06, + "loss": 0.1923, + "step": 52339 + }, + { + "epoch": 19.33, + "learning_rate": 1.0678655006774233e-06, + "loss": 0.1986, + "step": 52340 + }, + { + "epoch": 19.33, + "learning_rate": 1.067828550314078e-06, + "loss": 0.2631, + "step": 52341 + }, + { + "epoch": 19.33, + "learning_rate": 1.0677915999507327e-06, + "loss": 0.2039, + "step": 52342 + }, + { + "epoch": 19.33, + "learning_rate": 1.0677546495873876e-06, + "loss": 0.1094, + "step": 52343 + }, + { + "epoch": 19.33, + "learning_rate": 1.0677176992240424e-06, + "loss": 0.3184, + "step": 52344 + }, + { + "epoch": 19.33, + "learning_rate": 1.067680748860697e-06, + "loss": 0.1852, + "step": 52345 + }, + { + "epoch": 19.33, + "learning_rate": 1.067643798497352e-06, + "loss": 0.3064, + "step": 52346 + }, + { + "epoch": 19.33, + "learning_rate": 1.0676068481340068e-06, + "loss": 0.2019, + "step": 52347 + }, + { + "epoch": 19.33, + "learning_rate": 1.0675698977706614e-06, + "loss": 0.191, + "step": 52348 + }, + { + "epoch": 19.33, + "learning_rate": 1.0675329474073163e-06, + "loss": 0.2266, + "step": 52349 + }, + { + "epoch": 19.33, + "learning_rate": 1.0674959970439709e-06, + "loss": 0.1781, + "step": 52350 + }, + { + "epoch": 19.33, + "learning_rate": 1.0674590466806257e-06, + "loss": 0.1955, + "step": 52351 + }, + { + "epoch": 19.33, + "learning_rate": 1.0674220963172804e-06, + "loss": 0.2555, + "step": 52352 + }, + { + "epoch": 19.33, + "learning_rate": 1.0673851459539352e-06, + "loss": 0.2838, + "step": 52353 + }, + { + "epoch": 19.33, + "learning_rate": 1.06734819559059e-06, + "loss": 0.1178, + "step": 52354 + }, + { + "epoch": 19.33, + "learning_rate": 1.0673112452272447e-06, + "loss": 0.2627, + "step": 52355 + }, + { + "epoch": 19.33, + "learning_rate": 1.0672742948638996e-06, + "loss": 0.1929, + "step": 52356 + }, + { + "epoch": 19.33, + "learning_rate": 1.0672373445005544e-06, + "loss": 0.1933, + "step": 52357 + }, + { + "epoch": 19.33, + "learning_rate": 1.067200394137209e-06, + "loss": 0.2968, + "step": 52358 + }, + { + "epoch": 19.33, + "learning_rate": 1.0671634437738639e-06, + "loss": 0.255, + "step": 52359 + }, + { + "epoch": 19.34, + "learning_rate": 1.0671264934105185e-06, + "loss": 0.2447, + "step": 52360 + }, + { + "epoch": 19.34, + "learning_rate": 1.0670895430471732e-06, + "loss": 0.0778, + "step": 52361 + }, + { + "epoch": 19.34, + "learning_rate": 1.067052592683828e-06, + "loss": 0.2315, + "step": 52362 + }, + { + "epoch": 19.34, + "learning_rate": 1.0670156423204829e-06, + "loss": 0.2815, + "step": 52363 + }, + { + "epoch": 19.34, + "learning_rate": 1.0669786919571375e-06, + "loss": 0.2116, + "step": 52364 + }, + { + "epoch": 19.34, + "learning_rate": 1.0669417415937923e-06, + "loss": 0.1122, + "step": 52365 + }, + { + "epoch": 19.34, + "learning_rate": 1.0669047912304472e-06, + "loss": 0.1508, + "step": 52366 + }, + { + "epoch": 19.34, + "learning_rate": 1.0668678408671018e-06, + "loss": 0.2215, + "step": 52367 + }, + { + "epoch": 19.34, + "learning_rate": 1.0668308905037567e-06, + "loss": 0.2851, + "step": 52368 + }, + { + "epoch": 19.34, + "learning_rate": 1.0667939401404115e-06, + "loss": 0.2632, + "step": 52369 + }, + { + "epoch": 19.34, + "learning_rate": 1.0667569897770662e-06, + "loss": 0.3024, + "step": 52370 + }, + { + "epoch": 19.34, + "learning_rate": 1.0667200394137208e-06, + "loss": 0.2554, + "step": 52371 + }, + { + "epoch": 19.34, + "learning_rate": 1.0666830890503756e-06, + "loss": 0.2116, + "step": 52372 + }, + { + "epoch": 19.34, + "learning_rate": 1.0666461386870305e-06, + "loss": 0.216, + "step": 52373 + }, + { + "epoch": 19.34, + "learning_rate": 1.0666091883236851e-06, + "loss": 0.3447, + "step": 52374 + }, + { + "epoch": 19.34, + "learning_rate": 1.06657223796034e-06, + "loss": 0.3008, + "step": 52375 + }, + { + "epoch": 19.34, + "learning_rate": 1.0665352875969948e-06, + "loss": 0.2339, + "step": 52376 + }, + { + "epoch": 19.34, + "learning_rate": 1.0664983372336495e-06, + "loss": 0.1382, + "step": 52377 + }, + { + "epoch": 19.34, + "learning_rate": 1.0664613868703043e-06, + "loss": 0.11, + "step": 52378 + }, + { + "epoch": 19.34, + "learning_rate": 1.0664244365069592e-06, + "loss": 0.1876, + "step": 52379 + }, + { + "epoch": 19.34, + "learning_rate": 1.0663874861436138e-06, + "loss": 0.2078, + "step": 52380 + }, + { + "epoch": 19.34, + "learning_rate": 1.0663505357802684e-06, + "loss": 0.167, + "step": 52381 + }, + { + "epoch": 19.34, + "learning_rate": 1.0663135854169233e-06, + "loss": 0.1392, + "step": 52382 + }, + { + "epoch": 19.34, + "learning_rate": 1.0662766350535781e-06, + "loss": 0.1812, + "step": 52383 + }, + { + "epoch": 19.34, + "learning_rate": 1.0662396846902328e-06, + "loss": 0.133, + "step": 52384 + }, + { + "epoch": 19.34, + "learning_rate": 1.0662027343268876e-06, + "loss": 0.1922, + "step": 52385 + }, + { + "epoch": 19.34, + "learning_rate": 1.0661657839635425e-06, + "loss": 0.1052, + "step": 52386 + }, + { + "epoch": 19.35, + "learning_rate": 1.066128833600197e-06, + "loss": 0.2508, + "step": 52387 + }, + { + "epoch": 19.35, + "learning_rate": 1.066091883236852e-06, + "loss": 0.1533, + "step": 52388 + }, + { + "epoch": 19.35, + "learning_rate": 1.0660549328735068e-06, + "loss": 0.1622, + "step": 52389 + }, + { + "epoch": 19.35, + "learning_rate": 1.0660179825101612e-06, + "loss": 0.3994, + "step": 52390 + }, + { + "epoch": 19.35, + "learning_rate": 1.065981032146816e-06, + "loss": 0.1515, + "step": 52391 + }, + { + "epoch": 19.35, + "learning_rate": 1.065944081783471e-06, + "loss": 0.2407, + "step": 52392 + }, + { + "epoch": 19.35, + "learning_rate": 1.0659071314201256e-06, + "loss": 0.1952, + "step": 52393 + }, + { + "epoch": 19.35, + "learning_rate": 1.0658701810567804e-06, + "loss": 0.1823, + "step": 52394 + }, + { + "epoch": 19.35, + "learning_rate": 1.0658332306934352e-06, + "loss": 0.4104, + "step": 52395 + }, + { + "epoch": 19.35, + "learning_rate": 1.0657962803300899e-06, + "loss": 0.2366, + "step": 52396 + }, + { + "epoch": 19.35, + "learning_rate": 1.0657593299667447e-06, + "loss": 0.3294, + "step": 52397 + }, + { + "epoch": 19.35, + "learning_rate": 1.0657223796033996e-06, + "loss": 0.2957, + "step": 52398 + }, + { + "epoch": 19.35, + "learning_rate": 1.0656854292400542e-06, + "loss": 0.2161, + "step": 52399 + }, + { + "epoch": 19.35, + "learning_rate": 1.0656484788767089e-06, + "loss": 0.1217, + "step": 52400 + }, + { + "epoch": 19.35, + "learning_rate": 1.0656115285133637e-06, + "loss": 0.168, + "step": 52401 + }, + { + "epoch": 19.35, + "learning_rate": 1.0655745781500186e-06, + "loss": 0.3223, + "step": 52402 + }, + { + "epoch": 19.35, + "learning_rate": 1.0655376277866732e-06, + "loss": 0.1703, + "step": 52403 + }, + { + "epoch": 19.35, + "learning_rate": 1.065500677423328e-06, + "loss": 0.1171, + "step": 52404 + }, + { + "epoch": 19.35, + "learning_rate": 1.0654637270599829e-06, + "loss": 0.2779, + "step": 52405 + }, + { + "epoch": 19.35, + "learning_rate": 1.0654267766966375e-06, + "loss": 0.2309, + "step": 52406 + }, + { + "epoch": 19.35, + "learning_rate": 1.0653898263332924e-06, + "loss": 0.209, + "step": 52407 + }, + { + "epoch": 19.35, + "learning_rate": 1.0653528759699472e-06, + "loss": 0.1974, + "step": 52408 + }, + { + "epoch": 19.35, + "learning_rate": 1.0653159256066019e-06, + "loss": 0.3085, + "step": 52409 + }, + { + "epoch": 19.35, + "learning_rate": 1.0652789752432565e-06, + "loss": 0.1228, + "step": 52410 + }, + { + "epoch": 19.35, + "learning_rate": 1.0652420248799113e-06, + "loss": 0.1776, + "step": 52411 + }, + { + "epoch": 19.35, + "learning_rate": 1.065205074516566e-06, + "loss": 0.3536, + "step": 52412 + }, + { + "epoch": 19.35, + "learning_rate": 1.0651681241532208e-06, + "loss": 0.2936, + "step": 52413 + }, + { + "epoch": 19.36, + "learning_rate": 1.0651311737898757e-06, + "loss": 0.3178, + "step": 52414 + }, + { + "epoch": 19.36, + "learning_rate": 1.0650942234265303e-06, + "loss": 0.2582, + "step": 52415 + }, + { + "epoch": 19.36, + "learning_rate": 1.0650572730631852e-06, + "loss": 0.3481, + "step": 52416 + }, + { + "epoch": 19.36, + "learning_rate": 1.06502032269984e-06, + "loss": 0.2214, + "step": 52417 + }, + { + "epoch": 19.36, + "learning_rate": 1.0649833723364946e-06, + "loss": 0.2107, + "step": 52418 + }, + { + "epoch": 19.36, + "learning_rate": 1.0649464219731495e-06, + "loss": 0.1899, + "step": 52419 + }, + { + "epoch": 19.36, + "learning_rate": 1.0649094716098041e-06, + "loss": 0.37, + "step": 52420 + }, + { + "epoch": 19.36, + "learning_rate": 1.064872521246459e-06, + "loss": 0.1449, + "step": 52421 + }, + { + "epoch": 19.36, + "learning_rate": 1.0648355708831136e-06, + "loss": 0.1699, + "step": 52422 + }, + { + "epoch": 19.36, + "learning_rate": 1.0647986205197685e-06, + "loss": 0.2455, + "step": 52423 + }, + { + "epoch": 19.36, + "learning_rate": 1.0647616701564233e-06, + "loss": 0.3769, + "step": 52424 + }, + { + "epoch": 19.36, + "learning_rate": 1.064724719793078e-06, + "loss": 0.2267, + "step": 52425 + }, + { + "epoch": 19.36, + "learning_rate": 1.0646877694297328e-06, + "loss": 0.2758, + "step": 52426 + }, + { + "epoch": 19.36, + "learning_rate": 1.0646508190663876e-06, + "loss": 0.1568, + "step": 52427 + }, + { + "epoch": 19.36, + "learning_rate": 1.0646138687030423e-06, + "loss": 0.2713, + "step": 52428 + }, + { + "epoch": 19.36, + "learning_rate": 1.0645769183396971e-06, + "loss": 0.2269, + "step": 52429 + }, + { + "epoch": 19.36, + "learning_rate": 1.0645399679763518e-06, + "loss": 0.2093, + "step": 52430 + }, + { + "epoch": 19.36, + "learning_rate": 1.0645030176130066e-06, + "loss": 0.2898, + "step": 52431 + }, + { + "epoch": 19.36, + "learning_rate": 1.0644660672496612e-06, + "loss": 0.1575, + "step": 52432 + }, + { + "epoch": 19.36, + "learning_rate": 1.064429116886316e-06, + "loss": 0.2172, + "step": 52433 + }, + { + "epoch": 19.36, + "learning_rate": 1.0643921665229707e-06, + "loss": 0.1838, + "step": 52434 + }, + { + "epoch": 19.36, + "learning_rate": 1.0643552161596256e-06, + "loss": 0.2546, + "step": 52435 + }, + { + "epoch": 19.36, + "learning_rate": 1.0643182657962804e-06, + "loss": 0.2386, + "step": 52436 + }, + { + "epoch": 19.36, + "learning_rate": 1.064281315432935e-06, + "loss": 0.1778, + "step": 52437 + }, + { + "epoch": 19.36, + "learning_rate": 1.06424436506959e-06, + "loss": 0.2178, + "step": 52438 + }, + { + "epoch": 19.36, + "learning_rate": 1.0642074147062448e-06, + "loss": 0.171, + "step": 52439 + }, + { + "epoch": 19.36, + "learning_rate": 1.0641704643428994e-06, + "loss": 0.1926, + "step": 52440 + }, + { + "epoch": 19.37, + "learning_rate": 1.064133513979554e-06, + "loss": 0.184, + "step": 52441 + }, + { + "epoch": 19.37, + "learning_rate": 1.0640965636162089e-06, + "loss": 0.1459, + "step": 52442 + }, + { + "epoch": 19.37, + "learning_rate": 1.0640596132528637e-06, + "loss": 0.248, + "step": 52443 + }, + { + "epoch": 19.37, + "learning_rate": 1.0640226628895184e-06, + "loss": 0.177, + "step": 52444 + }, + { + "epoch": 19.37, + "learning_rate": 1.0639857125261732e-06, + "loss": 0.3532, + "step": 52445 + }, + { + "epoch": 19.37, + "learning_rate": 1.063948762162828e-06, + "loss": 0.2727, + "step": 52446 + }, + { + "epoch": 19.37, + "learning_rate": 1.0639118117994827e-06, + "loss": 0.2246, + "step": 52447 + }, + { + "epoch": 19.37, + "learning_rate": 1.0638748614361375e-06, + "loss": 0.1584, + "step": 52448 + }, + { + "epoch": 19.37, + "learning_rate": 1.0638379110727924e-06, + "loss": 0.1946, + "step": 52449 + }, + { + "epoch": 19.37, + "learning_rate": 1.063800960709447e-06, + "loss": 0.1979, + "step": 52450 + }, + { + "epoch": 19.37, + "learning_rate": 1.0637640103461017e-06, + "loss": 0.309, + "step": 52451 + }, + { + "epoch": 19.37, + "learning_rate": 1.0637270599827565e-06, + "loss": 0.1406, + "step": 52452 + }, + { + "epoch": 19.37, + "learning_rate": 1.0636901096194114e-06, + "loss": 0.1813, + "step": 52453 + }, + { + "epoch": 19.37, + "learning_rate": 1.063653159256066e-06, + "loss": 0.2159, + "step": 52454 + }, + { + "epoch": 19.37, + "learning_rate": 1.0636162088927208e-06, + "loss": 0.2896, + "step": 52455 + }, + { + "epoch": 19.37, + "learning_rate": 1.0635792585293757e-06, + "loss": 0.309, + "step": 52456 + }, + { + "epoch": 19.37, + "learning_rate": 1.0635423081660303e-06, + "loss": 0.1934, + "step": 52457 + }, + { + "epoch": 19.37, + "learning_rate": 1.0635053578026852e-06, + "loss": 0.3138, + "step": 52458 + }, + { + "epoch": 19.37, + "learning_rate": 1.06346840743934e-06, + "loss": 0.2134, + "step": 52459 + }, + { + "epoch": 19.37, + "learning_rate": 1.0634314570759945e-06, + "loss": 0.2809, + "step": 52460 + }, + { + "epoch": 19.37, + "learning_rate": 1.0633945067126493e-06, + "loss": 0.2178, + "step": 52461 + }, + { + "epoch": 19.37, + "learning_rate": 1.0633575563493041e-06, + "loss": 0.2608, + "step": 52462 + }, + { + "epoch": 19.37, + "learning_rate": 1.0633206059859588e-06, + "loss": 0.1828, + "step": 52463 + }, + { + "epoch": 19.37, + "learning_rate": 1.0632836556226136e-06, + "loss": 0.2586, + "step": 52464 + }, + { + "epoch": 19.37, + "learning_rate": 1.0632467052592685e-06, + "loss": 0.1799, + "step": 52465 + }, + { + "epoch": 19.37, + "learning_rate": 1.0632097548959231e-06, + "loss": 0.2188, + "step": 52466 + }, + { + "epoch": 19.37, + "learning_rate": 1.063172804532578e-06, + "loss": 0.3192, + "step": 52467 + }, + { + "epoch": 19.38, + "learning_rate": 1.0631358541692328e-06, + "loss": 0.2356, + "step": 52468 + }, + { + "epoch": 19.38, + "learning_rate": 1.0630989038058874e-06, + "loss": 0.2879, + "step": 52469 + }, + { + "epoch": 19.38, + "learning_rate": 1.063061953442542e-06, + "loss": 0.1448, + "step": 52470 + }, + { + "epoch": 19.38, + "learning_rate": 1.063025003079197e-06, + "loss": 0.1455, + "step": 52471 + }, + { + "epoch": 19.38, + "learning_rate": 1.0629880527158518e-06, + "loss": 0.2877, + "step": 52472 + }, + { + "epoch": 19.38, + "learning_rate": 1.0629511023525064e-06, + "loss": 0.1349, + "step": 52473 + }, + { + "epoch": 19.38, + "learning_rate": 1.0629141519891613e-06, + "loss": 0.2133, + "step": 52474 + }, + { + "epoch": 19.38, + "learning_rate": 1.0628772016258161e-06, + "loss": 0.2046, + "step": 52475 + }, + { + "epoch": 19.38, + "learning_rate": 1.0628402512624708e-06, + "loss": 0.3628, + "step": 52476 + }, + { + "epoch": 19.38, + "learning_rate": 1.0628033008991256e-06, + "loss": 0.1629, + "step": 52477 + }, + { + "epoch": 19.38, + "learning_rate": 1.0627663505357804e-06, + "loss": 0.2725, + "step": 52478 + }, + { + "epoch": 19.38, + "learning_rate": 1.0627294001724349e-06, + "loss": 0.1394, + "step": 52479 + }, + { + "epoch": 19.38, + "learning_rate": 1.0626924498090897e-06, + "loss": 0.4381, + "step": 52480 + }, + { + "epoch": 19.38, + "learning_rate": 1.0626554994457446e-06, + "loss": 0.2159, + "step": 52481 + }, + { + "epoch": 19.38, + "learning_rate": 1.0626185490823992e-06, + "loss": 0.1235, + "step": 52482 + }, + { + "epoch": 19.38, + "learning_rate": 1.062581598719054e-06, + "loss": 0.2407, + "step": 52483 + }, + { + "epoch": 19.38, + "learning_rate": 1.062544648355709e-06, + "loss": 0.1829, + "step": 52484 + }, + { + "epoch": 19.38, + "learning_rate": 1.0625076979923635e-06, + "loss": 0.2202, + "step": 52485 + }, + { + "epoch": 19.38, + "learning_rate": 1.0624707476290184e-06, + "loss": 0.1691, + "step": 52486 + }, + { + "epoch": 19.38, + "learning_rate": 1.0624337972656732e-06, + "loss": 0.1716, + "step": 52487 + }, + { + "epoch": 19.38, + "learning_rate": 1.0623968469023279e-06, + "loss": 0.2573, + "step": 52488 + }, + { + "epoch": 19.38, + "learning_rate": 1.0623598965389825e-06, + "loss": 0.1374, + "step": 52489 + }, + { + "epoch": 19.38, + "learning_rate": 1.0623229461756374e-06, + "loss": 0.3437, + "step": 52490 + }, + { + "epoch": 19.38, + "learning_rate": 1.0622859958122922e-06, + "loss": 0.1684, + "step": 52491 + }, + { + "epoch": 19.38, + "learning_rate": 1.0622490454489468e-06, + "loss": 0.2112, + "step": 52492 + }, + { + "epoch": 19.38, + "learning_rate": 1.0622120950856017e-06, + "loss": 0.3323, + "step": 52493 + }, + { + "epoch": 19.38, + "learning_rate": 1.0621751447222565e-06, + "loss": 0.275, + "step": 52494 + }, + { + "epoch": 19.39, + "learning_rate": 1.0621381943589112e-06, + "loss": 0.2352, + "step": 52495 + }, + { + "epoch": 19.39, + "learning_rate": 1.062101243995566e-06, + "loss": 0.2813, + "step": 52496 + }, + { + "epoch": 19.39, + "learning_rate": 1.0620642936322209e-06, + "loss": 0.1819, + "step": 52497 + }, + { + "epoch": 19.39, + "learning_rate": 1.0620273432688755e-06, + "loss": 0.2599, + "step": 52498 + }, + { + "epoch": 19.39, + "learning_rate": 1.0619903929055301e-06, + "loss": 0.2261, + "step": 52499 + }, + { + "epoch": 19.39, + "learning_rate": 1.061953442542185e-06, + "loss": 0.2095, + "step": 52500 + }, + { + "epoch": 19.39, + "learning_rate": 1.0619164921788398e-06, + "loss": 0.1969, + "step": 52501 + }, + { + "epoch": 19.39, + "learning_rate": 1.0618795418154945e-06, + "loss": 0.1324, + "step": 52502 + }, + { + "epoch": 19.39, + "learning_rate": 1.0618425914521493e-06, + "loss": 0.238, + "step": 52503 + }, + { + "epoch": 19.39, + "learning_rate": 1.0618056410888042e-06, + "loss": 0.125, + "step": 52504 + }, + { + "epoch": 19.39, + "learning_rate": 1.0617686907254588e-06, + "loss": 0.3211, + "step": 52505 + }, + { + "epoch": 19.39, + "learning_rate": 1.0617317403621137e-06, + "loss": 0.3152, + "step": 52506 + }, + { + "epoch": 19.39, + "learning_rate": 1.0616947899987685e-06, + "loss": 0.1547, + "step": 52507 + }, + { + "epoch": 19.39, + "learning_rate": 1.0616578396354231e-06, + "loss": 0.2784, + "step": 52508 + }, + { + "epoch": 19.39, + "learning_rate": 1.0616208892720778e-06, + "loss": 0.3195, + "step": 52509 + }, + { + "epoch": 19.39, + "learning_rate": 1.0615839389087326e-06, + "loss": 0.1808, + "step": 52510 + }, + { + "epoch": 19.39, + "learning_rate": 1.0615469885453873e-06, + "loss": 0.1522, + "step": 52511 + }, + { + "epoch": 19.39, + "learning_rate": 1.0615100381820421e-06, + "loss": 0.3416, + "step": 52512 + }, + { + "epoch": 19.39, + "learning_rate": 1.061473087818697e-06, + "loss": 0.1751, + "step": 52513 + }, + { + "epoch": 19.39, + "learning_rate": 1.0614361374553516e-06, + "loss": 0.1314, + "step": 52514 + }, + { + "epoch": 19.39, + "learning_rate": 1.0613991870920064e-06, + "loss": 0.167, + "step": 52515 + }, + { + "epoch": 19.39, + "learning_rate": 1.0613622367286613e-06, + "loss": 0.1883, + "step": 52516 + }, + { + "epoch": 19.39, + "learning_rate": 1.061325286365316e-06, + "loss": 0.2473, + "step": 52517 + }, + { + "epoch": 19.39, + "learning_rate": 1.0612883360019708e-06, + "loss": 0.0888, + "step": 52518 + }, + { + "epoch": 19.39, + "learning_rate": 1.0612513856386254e-06, + "loss": 0.2807, + "step": 52519 + }, + { + "epoch": 19.39, + "learning_rate": 1.0612144352752803e-06, + "loss": 0.1733, + "step": 52520 + }, + { + "epoch": 19.39, + "learning_rate": 1.061177484911935e-06, + "loss": 0.2657, + "step": 52521 + }, + { + "epoch": 19.4, + "learning_rate": 1.0611405345485897e-06, + "loss": 0.2441, + "step": 52522 + }, + { + "epoch": 19.4, + "learning_rate": 1.0611035841852446e-06, + "loss": 0.3338, + "step": 52523 + }, + { + "epoch": 19.4, + "learning_rate": 1.0610666338218992e-06, + "loss": 0.1898, + "step": 52524 + }, + { + "epoch": 19.4, + "learning_rate": 1.061029683458554e-06, + "loss": 0.1948, + "step": 52525 + }, + { + "epoch": 19.4, + "learning_rate": 1.060992733095209e-06, + "loss": 0.3456, + "step": 52526 + }, + { + "epoch": 19.4, + "learning_rate": 1.0609557827318636e-06, + "loss": 0.1296, + "step": 52527 + }, + { + "epoch": 19.4, + "learning_rate": 1.0609188323685184e-06, + "loss": 0.1946, + "step": 52528 + }, + { + "epoch": 19.4, + "learning_rate": 1.060881882005173e-06, + "loss": 0.1364, + "step": 52529 + }, + { + "epoch": 19.4, + "learning_rate": 1.0608449316418277e-06, + "loss": 0.3071, + "step": 52530 + }, + { + "epoch": 19.4, + "learning_rate": 1.0608079812784825e-06, + "loss": 0.2329, + "step": 52531 + }, + { + "epoch": 19.4, + "learning_rate": 1.0607710309151374e-06, + "loss": 0.1944, + "step": 52532 + }, + { + "epoch": 19.4, + "learning_rate": 1.060734080551792e-06, + "loss": 0.1685, + "step": 52533 + }, + { + "epoch": 19.4, + "learning_rate": 1.0606971301884469e-06, + "loss": 0.2413, + "step": 52534 + }, + { + "epoch": 19.4, + "learning_rate": 1.0606601798251017e-06, + "loss": 0.1959, + "step": 52535 + }, + { + "epoch": 19.4, + "learning_rate": 1.0606232294617563e-06, + "loss": 0.2361, + "step": 52536 + }, + { + "epoch": 19.4, + "learning_rate": 1.0605862790984112e-06, + "loss": 0.0995, + "step": 52537 + }, + { + "epoch": 19.4, + "learning_rate": 1.060549328735066e-06, + "loss": 0.2034, + "step": 52538 + }, + { + "epoch": 19.4, + "learning_rate": 1.0605123783717207e-06, + "loss": 0.2494, + "step": 52539 + }, + { + "epoch": 19.4, + "learning_rate": 1.0604754280083753e-06, + "loss": 0.2922, + "step": 52540 + }, + { + "epoch": 19.4, + "learning_rate": 1.0604384776450302e-06, + "loss": 0.3113, + "step": 52541 + }, + { + "epoch": 19.4, + "learning_rate": 1.060401527281685e-06, + "loss": 0.1731, + "step": 52542 + }, + { + "epoch": 19.4, + "learning_rate": 1.0603645769183397e-06, + "loss": 0.3159, + "step": 52543 + }, + { + "epoch": 19.4, + "learning_rate": 1.0603276265549945e-06, + "loss": 0.2661, + "step": 52544 + }, + { + "epoch": 19.4, + "learning_rate": 1.0602906761916493e-06, + "loss": 0.2687, + "step": 52545 + }, + { + "epoch": 19.4, + "learning_rate": 1.060253725828304e-06, + "loss": 0.1758, + "step": 52546 + }, + { + "epoch": 19.4, + "learning_rate": 1.0602167754649588e-06, + "loss": 0.1764, + "step": 52547 + }, + { + "epoch": 19.4, + "learning_rate": 1.0601798251016137e-06, + "loss": 0.2522, + "step": 52548 + }, + { + "epoch": 19.41, + "learning_rate": 1.060142874738268e-06, + "loss": 0.1677, + "step": 52549 + }, + { + "epoch": 19.41, + "learning_rate": 1.060105924374923e-06, + "loss": 0.2297, + "step": 52550 + }, + { + "epoch": 19.41, + "learning_rate": 1.0600689740115778e-06, + "loss": 0.3276, + "step": 52551 + }, + { + "epoch": 19.41, + "learning_rate": 1.0600320236482324e-06, + "loss": 0.2502, + "step": 52552 + }, + { + "epoch": 19.41, + "learning_rate": 1.0599950732848873e-06, + "loss": 0.3151, + "step": 52553 + }, + { + "epoch": 19.41, + "learning_rate": 1.0599581229215421e-06, + "loss": 0.1488, + "step": 52554 + }, + { + "epoch": 19.41, + "learning_rate": 1.0599211725581968e-06, + "loss": 0.2704, + "step": 52555 + }, + { + "epoch": 19.41, + "learning_rate": 1.0598842221948516e-06, + "loss": 0.2802, + "step": 52556 + }, + { + "epoch": 19.41, + "learning_rate": 1.0598472718315065e-06, + "loss": 0.2072, + "step": 52557 + }, + { + "epoch": 19.41, + "learning_rate": 1.059810321468161e-06, + "loss": 0.2955, + "step": 52558 + }, + { + "epoch": 19.41, + "learning_rate": 1.0597733711048157e-06, + "loss": 0.248, + "step": 52559 + }, + { + "epoch": 19.41, + "learning_rate": 1.0597364207414706e-06, + "loss": 0.2761, + "step": 52560 + }, + { + "epoch": 19.41, + "learning_rate": 1.0596994703781254e-06, + "loss": 0.109, + "step": 52561 + }, + { + "epoch": 19.41, + "learning_rate": 1.05966252001478e-06, + "loss": 0.1447, + "step": 52562 + }, + { + "epoch": 19.41, + "learning_rate": 1.059625569651435e-06, + "loss": 0.1522, + "step": 52563 + }, + { + "epoch": 19.41, + "learning_rate": 1.0595886192880898e-06, + "loss": 0.1349, + "step": 52564 + }, + { + "epoch": 19.41, + "learning_rate": 1.0595516689247444e-06, + "loss": 0.1463, + "step": 52565 + }, + { + "epoch": 19.41, + "learning_rate": 1.0595147185613993e-06, + "loss": 0.1164, + "step": 52566 + }, + { + "epoch": 19.41, + "learning_rate": 1.059477768198054e-06, + "loss": 0.1455, + "step": 52567 + }, + { + "epoch": 19.41, + "learning_rate": 1.0594408178347087e-06, + "loss": 0.1325, + "step": 52568 + }, + { + "epoch": 19.41, + "learning_rate": 1.0594038674713634e-06, + "loss": 0.1275, + "step": 52569 + }, + { + "epoch": 19.41, + "learning_rate": 1.0593669171080182e-06, + "loss": 0.2085, + "step": 52570 + }, + { + "epoch": 19.41, + "learning_rate": 1.059329966744673e-06, + "loss": 0.2453, + "step": 52571 + }, + { + "epoch": 19.41, + "learning_rate": 1.0592930163813277e-06, + "loss": 0.2152, + "step": 52572 + }, + { + "epoch": 19.41, + "learning_rate": 1.0592560660179826e-06, + "loss": 0.3546, + "step": 52573 + }, + { + "epoch": 19.41, + "learning_rate": 1.0592191156546374e-06, + "loss": 0.3349, + "step": 52574 + }, + { + "epoch": 19.41, + "learning_rate": 1.059182165291292e-06, + "loss": 0.2697, + "step": 52575 + }, + { + "epoch": 19.42, + "learning_rate": 1.0591452149279469e-06, + "loss": 0.3577, + "step": 52576 + }, + { + "epoch": 19.42, + "learning_rate": 1.0591082645646017e-06, + "loss": 0.1622, + "step": 52577 + }, + { + "epoch": 19.42, + "learning_rate": 1.0590713142012564e-06, + "loss": 0.2792, + "step": 52578 + }, + { + "epoch": 19.42, + "learning_rate": 1.059034363837911e-06, + "loss": 0.267, + "step": 52579 + }, + { + "epoch": 19.42, + "learning_rate": 1.0589974134745659e-06, + "loss": 0.1859, + "step": 52580 + }, + { + "epoch": 19.42, + "learning_rate": 1.0589604631112205e-06, + "loss": 0.1922, + "step": 52581 + }, + { + "epoch": 19.42, + "learning_rate": 1.0589235127478753e-06, + "loss": 0.3382, + "step": 52582 + }, + { + "epoch": 19.42, + "learning_rate": 1.0588865623845302e-06, + "loss": 0.2701, + "step": 52583 + }, + { + "epoch": 19.42, + "learning_rate": 1.0588496120211848e-06, + "loss": 0.3004, + "step": 52584 + }, + { + "epoch": 19.42, + "learning_rate": 1.0588126616578397e-06, + "loss": 0.1274, + "step": 52585 + }, + { + "epoch": 19.42, + "learning_rate": 1.0587757112944945e-06, + "loss": 0.1995, + "step": 52586 + }, + { + "epoch": 19.42, + "learning_rate": 1.0587387609311492e-06, + "loss": 0.1945, + "step": 52587 + }, + { + "epoch": 19.42, + "learning_rate": 1.058701810567804e-06, + "loss": 0.266, + "step": 52588 + }, + { + "epoch": 19.42, + "learning_rate": 1.0586648602044586e-06, + "loss": 0.2309, + "step": 52589 + }, + { + "epoch": 19.42, + "learning_rate": 1.0586279098411135e-06, + "loss": 0.1294, + "step": 52590 + }, + { + "epoch": 19.42, + "learning_rate": 1.0585909594777681e-06, + "loss": 0.1501, + "step": 52591 + }, + { + "epoch": 19.42, + "learning_rate": 1.058554009114423e-06, + "loss": 0.2898, + "step": 52592 + }, + { + "epoch": 19.42, + "learning_rate": 1.0585170587510778e-06, + "loss": 0.2269, + "step": 52593 + }, + { + "epoch": 19.42, + "learning_rate": 1.0584801083877325e-06, + "loss": 0.241, + "step": 52594 + }, + { + "epoch": 19.42, + "learning_rate": 1.0584431580243873e-06, + "loss": 0.2174, + "step": 52595 + }, + { + "epoch": 19.42, + "learning_rate": 1.0584062076610422e-06, + "loss": 0.3479, + "step": 52596 + }, + { + "epoch": 19.42, + "learning_rate": 1.0583692572976968e-06, + "loss": 0.2572, + "step": 52597 + }, + { + "epoch": 19.42, + "learning_rate": 1.0583323069343516e-06, + "loss": 0.13, + "step": 52598 + }, + { + "epoch": 19.42, + "learning_rate": 1.0582953565710063e-06, + "loss": 0.2136, + "step": 52599 + }, + { + "epoch": 19.42, + "learning_rate": 1.058258406207661e-06, + "loss": 0.3639, + "step": 52600 + }, + { + "epoch": 19.42, + "learning_rate": 1.0582214558443158e-06, + "loss": 0.291, + "step": 52601 + }, + { + "epoch": 19.42, + "learning_rate": 1.0581845054809706e-06, + "loss": 0.1942, + "step": 52602 + }, + { + "epoch": 19.43, + "learning_rate": 1.0581475551176252e-06, + "loss": 0.2556, + "step": 52603 + }, + { + "epoch": 19.43, + "learning_rate": 1.05811060475428e-06, + "loss": 0.2479, + "step": 52604 + }, + { + "epoch": 19.43, + "learning_rate": 1.058073654390935e-06, + "loss": 0.1585, + "step": 52605 + }, + { + "epoch": 19.43, + "learning_rate": 1.0580367040275896e-06, + "loss": 0.1914, + "step": 52606 + }, + { + "epoch": 19.43, + "learning_rate": 1.0579997536642444e-06, + "loss": 0.1404, + "step": 52607 + }, + { + "epoch": 19.43, + "learning_rate": 1.0579628033008993e-06, + "loss": 0.1202, + "step": 52608 + }, + { + "epoch": 19.43, + "learning_rate": 1.057925852937554e-06, + "loss": 0.1936, + "step": 52609 + }, + { + "epoch": 19.43, + "learning_rate": 1.0578889025742086e-06, + "loss": 0.2904, + "step": 52610 + }, + { + "epoch": 19.43, + "learning_rate": 1.0578519522108634e-06, + "loss": 0.1275, + "step": 52611 + }, + { + "epoch": 19.43, + "learning_rate": 1.0578150018475182e-06, + "loss": 0.3067, + "step": 52612 + }, + { + "epoch": 19.43, + "learning_rate": 1.0577780514841729e-06, + "loss": 0.1714, + "step": 52613 + }, + { + "epoch": 19.43, + "learning_rate": 1.0577411011208277e-06, + "loss": 0.1639, + "step": 52614 + }, + { + "epoch": 19.43, + "learning_rate": 1.0577041507574826e-06, + "loss": 0.2127, + "step": 52615 + }, + { + "epoch": 19.43, + "learning_rate": 1.0576672003941372e-06, + "loss": 0.357, + "step": 52616 + }, + { + "epoch": 19.43, + "learning_rate": 1.057630250030792e-06, + "loss": 0.2421, + "step": 52617 + }, + { + "epoch": 19.43, + "learning_rate": 1.057593299667447e-06, + "loss": 0.2348, + "step": 52618 + }, + { + "epoch": 19.43, + "learning_rate": 1.0575563493041015e-06, + "loss": 0.2465, + "step": 52619 + }, + { + "epoch": 19.43, + "learning_rate": 1.0575193989407562e-06, + "loss": 0.232, + "step": 52620 + }, + { + "epoch": 19.43, + "learning_rate": 1.057482448577411e-06, + "loss": 0.191, + "step": 52621 + }, + { + "epoch": 19.43, + "learning_rate": 1.0574454982140657e-06, + "loss": 0.1385, + "step": 52622 + }, + { + "epoch": 19.43, + "learning_rate": 1.0574085478507205e-06, + "loss": 0.1872, + "step": 52623 + }, + { + "epoch": 19.43, + "learning_rate": 1.0573715974873754e-06, + "loss": 0.2036, + "step": 52624 + }, + { + "epoch": 19.43, + "learning_rate": 1.05733464712403e-06, + "loss": 0.1923, + "step": 52625 + }, + { + "epoch": 19.43, + "learning_rate": 1.0572976967606849e-06, + "loss": 0.0914, + "step": 52626 + }, + { + "epoch": 19.43, + "learning_rate": 1.0572607463973397e-06, + "loss": 0.2452, + "step": 52627 + }, + { + "epoch": 19.43, + "learning_rate": 1.0572237960339943e-06, + "loss": 0.1107, + "step": 52628 + }, + { + "epoch": 19.43, + "learning_rate": 1.057186845670649e-06, + "loss": 0.1208, + "step": 52629 + }, + { + "epoch": 19.44, + "learning_rate": 1.0571498953073038e-06, + "loss": 0.2501, + "step": 52630 + }, + { + "epoch": 19.44, + "learning_rate": 1.0571129449439587e-06, + "loss": 0.0923, + "step": 52631 + }, + { + "epoch": 19.44, + "learning_rate": 1.0570759945806133e-06, + "loss": 0.1511, + "step": 52632 + }, + { + "epoch": 19.44, + "learning_rate": 1.0570390442172682e-06, + "loss": 0.3503, + "step": 52633 + }, + { + "epoch": 19.44, + "learning_rate": 1.057002093853923e-06, + "loss": 0.26, + "step": 52634 + }, + { + "epoch": 19.44, + "learning_rate": 1.0569651434905776e-06, + "loss": 0.1674, + "step": 52635 + }, + { + "epoch": 19.44, + "learning_rate": 1.0569281931272325e-06, + "loss": 0.1265, + "step": 52636 + }, + { + "epoch": 19.44, + "learning_rate": 1.0568912427638873e-06, + "loss": 0.17, + "step": 52637 + }, + { + "epoch": 19.44, + "learning_rate": 1.056854292400542e-06, + "loss": 0.3004, + "step": 52638 + }, + { + "epoch": 19.44, + "learning_rate": 1.0568173420371966e-06, + "loss": 0.2526, + "step": 52639 + }, + { + "epoch": 19.44, + "learning_rate": 1.0567803916738515e-06, + "loss": 0.1504, + "step": 52640 + }, + { + "epoch": 19.44, + "learning_rate": 1.0567434413105063e-06, + "loss": 0.1179, + "step": 52641 + }, + { + "epoch": 19.44, + "learning_rate": 1.056706490947161e-06, + "loss": 0.177, + "step": 52642 + }, + { + "epoch": 19.44, + "learning_rate": 1.0566695405838158e-06, + "loss": 0.2123, + "step": 52643 + }, + { + "epoch": 19.44, + "learning_rate": 1.0566325902204706e-06, + "loss": 0.2061, + "step": 52644 + }, + { + "epoch": 19.44, + "learning_rate": 1.0565956398571253e-06, + "loss": 0.2707, + "step": 52645 + }, + { + "epoch": 19.44, + "learning_rate": 1.0565586894937801e-06, + "loss": 0.3875, + "step": 52646 + }, + { + "epoch": 19.44, + "learning_rate": 1.056521739130435e-06, + "loss": 0.1576, + "step": 52647 + }, + { + "epoch": 19.44, + "learning_rate": 1.0564847887670896e-06, + "loss": 0.2874, + "step": 52648 + }, + { + "epoch": 19.44, + "learning_rate": 1.0564478384037442e-06, + "loss": 0.3016, + "step": 52649 + }, + { + "epoch": 19.44, + "learning_rate": 1.056410888040399e-06, + "loss": 0.3306, + "step": 52650 + }, + { + "epoch": 19.44, + "learning_rate": 1.0563739376770537e-06, + "loss": 0.1449, + "step": 52651 + }, + { + "epoch": 19.44, + "learning_rate": 1.0563369873137086e-06, + "loss": 0.2401, + "step": 52652 + }, + { + "epoch": 19.44, + "learning_rate": 1.0563000369503634e-06, + "loss": 0.2166, + "step": 52653 + }, + { + "epoch": 19.44, + "learning_rate": 1.056263086587018e-06, + "loss": 0.2034, + "step": 52654 + }, + { + "epoch": 19.44, + "learning_rate": 1.056226136223673e-06, + "loss": 0.2553, + "step": 52655 + }, + { + "epoch": 19.44, + "learning_rate": 1.0561891858603278e-06, + "loss": 0.1917, + "step": 52656 + }, + { + "epoch": 19.44, + "learning_rate": 1.0561522354969824e-06, + "loss": 0.1558, + "step": 52657 + }, + { + "epoch": 19.45, + "learning_rate": 1.0561152851336372e-06, + "loss": 0.1743, + "step": 52658 + }, + { + "epoch": 19.45, + "learning_rate": 1.0560783347702919e-06, + "loss": 0.3932, + "step": 52659 + }, + { + "epoch": 19.45, + "learning_rate": 1.0560413844069467e-06, + "loss": 0.2182, + "step": 52660 + }, + { + "epoch": 19.45, + "learning_rate": 1.0560044340436014e-06, + "loss": 0.2372, + "step": 52661 + }, + { + "epoch": 19.45, + "learning_rate": 1.0559674836802562e-06, + "loss": 0.1326, + "step": 52662 + }, + { + "epoch": 19.45, + "learning_rate": 1.055930533316911e-06, + "loss": 0.4286, + "step": 52663 + }, + { + "epoch": 19.45, + "learning_rate": 1.0558935829535657e-06, + "loss": 0.1427, + "step": 52664 + }, + { + "epoch": 19.45, + "learning_rate": 1.0558566325902205e-06, + "loss": 0.2328, + "step": 52665 + }, + { + "epoch": 19.45, + "learning_rate": 1.0558196822268754e-06, + "loss": 0.1808, + "step": 52666 + }, + { + "epoch": 19.45, + "learning_rate": 1.05578273186353e-06, + "loss": 0.115, + "step": 52667 + }, + { + "epoch": 19.45, + "learning_rate": 1.0557457815001849e-06, + "loss": 0.1568, + "step": 52668 + }, + { + "epoch": 19.45, + "learning_rate": 1.0557088311368395e-06, + "loss": 0.1443, + "step": 52669 + }, + { + "epoch": 19.45, + "learning_rate": 1.0556718807734941e-06, + "loss": 0.0647, + "step": 52670 + }, + { + "epoch": 19.45, + "learning_rate": 1.055634930410149e-06, + "loss": 0.1943, + "step": 52671 + }, + { + "epoch": 19.45, + "learning_rate": 1.0555979800468038e-06, + "loss": 0.1764, + "step": 52672 + }, + { + "epoch": 19.45, + "learning_rate": 1.0555610296834585e-06, + "loss": 0.2862, + "step": 52673 + }, + { + "epoch": 19.45, + "learning_rate": 1.0555240793201133e-06, + "loss": 0.3511, + "step": 52674 + }, + { + "epoch": 19.45, + "learning_rate": 1.0554871289567682e-06, + "loss": 0.2363, + "step": 52675 + }, + { + "epoch": 19.45, + "learning_rate": 1.0554501785934228e-06, + "loss": 0.4045, + "step": 52676 + }, + { + "epoch": 19.45, + "learning_rate": 1.0554132282300777e-06, + "loss": 0.1519, + "step": 52677 + }, + { + "epoch": 19.45, + "learning_rate": 1.0553762778667325e-06, + "loss": 0.1493, + "step": 52678 + }, + { + "epoch": 19.45, + "learning_rate": 1.0553393275033871e-06, + "loss": 0.1552, + "step": 52679 + }, + { + "epoch": 19.45, + "learning_rate": 1.0553023771400418e-06, + "loss": 0.1752, + "step": 52680 + }, + { + "epoch": 19.45, + "learning_rate": 1.0552654267766966e-06, + "loss": 0.3207, + "step": 52681 + }, + { + "epoch": 19.45, + "learning_rate": 1.0552284764133515e-06, + "loss": 0.2681, + "step": 52682 + }, + { + "epoch": 19.45, + "learning_rate": 1.0551915260500061e-06, + "loss": 0.356, + "step": 52683 + }, + { + "epoch": 19.45, + "learning_rate": 1.055154575686661e-06, + "loss": 0.1456, + "step": 52684 + }, + { + "epoch": 19.46, + "learning_rate": 1.0551176253233158e-06, + "loss": 0.1329, + "step": 52685 + }, + { + "epoch": 19.46, + "learning_rate": 1.0550806749599704e-06, + "loss": 0.1675, + "step": 52686 + }, + { + "epoch": 19.46, + "learning_rate": 1.0550437245966253e-06, + "loss": 0.2078, + "step": 52687 + }, + { + "epoch": 19.46, + "learning_rate": 1.0550067742332801e-06, + "loss": 0.2268, + "step": 52688 + }, + { + "epoch": 19.46, + "learning_rate": 1.0549698238699348e-06, + "loss": 0.3579, + "step": 52689 + }, + { + "epoch": 19.46, + "learning_rate": 1.0549328735065894e-06, + "loss": 0.2184, + "step": 52690 + }, + { + "epoch": 19.46, + "learning_rate": 1.0548959231432443e-06, + "loss": 0.3289, + "step": 52691 + }, + { + "epoch": 19.46, + "learning_rate": 1.0548589727798991e-06, + "loss": 0.3476, + "step": 52692 + }, + { + "epoch": 19.46, + "learning_rate": 1.0548220224165538e-06, + "loss": 0.2024, + "step": 52693 + }, + { + "epoch": 19.46, + "learning_rate": 1.0547850720532086e-06, + "loss": 0.3263, + "step": 52694 + }, + { + "epoch": 19.46, + "learning_rate": 1.0547481216898634e-06, + "loss": 0.2261, + "step": 52695 + }, + { + "epoch": 19.46, + "learning_rate": 1.054711171326518e-06, + "loss": 0.1734, + "step": 52696 + }, + { + "epoch": 19.46, + "learning_rate": 1.054674220963173e-06, + "loss": 0.3412, + "step": 52697 + }, + { + "epoch": 19.46, + "learning_rate": 1.0546372705998276e-06, + "loss": 0.1784, + "step": 52698 + }, + { + "epoch": 19.46, + "learning_rate": 1.0546003202364822e-06, + "loss": 0.1785, + "step": 52699 + }, + { + "epoch": 19.46, + "learning_rate": 1.054563369873137e-06, + "loss": 0.2037, + "step": 52700 + }, + { + "epoch": 19.46, + "learning_rate": 1.054526419509792e-06, + "loss": 0.1874, + "step": 52701 + }, + { + "epoch": 19.46, + "learning_rate": 1.0544894691464465e-06, + "loss": 0.2387, + "step": 52702 + }, + { + "epoch": 19.46, + "learning_rate": 1.0544525187831014e-06, + "loss": 0.1862, + "step": 52703 + }, + { + "epoch": 19.46, + "learning_rate": 1.0544155684197562e-06, + "loss": 0.1842, + "step": 52704 + }, + { + "epoch": 19.46, + "learning_rate": 1.0543786180564109e-06, + "loss": 0.2558, + "step": 52705 + }, + { + "epoch": 19.46, + "learning_rate": 1.0543416676930657e-06, + "loss": 0.1061, + "step": 52706 + }, + { + "epoch": 19.46, + "learning_rate": 1.0543047173297206e-06, + "loss": 0.1819, + "step": 52707 + }, + { + "epoch": 19.46, + "learning_rate": 1.0542677669663752e-06, + "loss": 0.1668, + "step": 52708 + }, + { + "epoch": 19.46, + "learning_rate": 1.0542308166030298e-06, + "loss": 0.1928, + "step": 52709 + }, + { + "epoch": 19.46, + "learning_rate": 1.0541938662396847e-06, + "loss": 0.3044, + "step": 52710 + }, + { + "epoch": 19.46, + "learning_rate": 1.0541569158763395e-06, + "loss": 0.2567, + "step": 52711 + }, + { + "epoch": 19.47, + "learning_rate": 1.0541199655129942e-06, + "loss": 0.1478, + "step": 52712 + }, + { + "epoch": 19.47, + "learning_rate": 1.054083015149649e-06, + "loss": 0.0827, + "step": 52713 + }, + { + "epoch": 19.47, + "learning_rate": 1.0540460647863039e-06, + "loss": 0.291, + "step": 52714 + }, + { + "epoch": 19.47, + "learning_rate": 1.0540091144229585e-06, + "loss": 0.2279, + "step": 52715 + }, + { + "epoch": 19.47, + "learning_rate": 1.0539721640596134e-06, + "loss": 0.3369, + "step": 52716 + }, + { + "epoch": 19.47, + "learning_rate": 1.0539352136962682e-06, + "loss": 0.3651, + "step": 52717 + }, + { + "epoch": 19.47, + "learning_rate": 1.0538982633329226e-06, + "loss": 0.2295, + "step": 52718 + }, + { + "epoch": 19.47, + "learning_rate": 1.0538613129695775e-06, + "loss": 0.1641, + "step": 52719 + }, + { + "epoch": 19.47, + "learning_rate": 1.0538243626062323e-06, + "loss": 0.1893, + "step": 52720 + }, + { + "epoch": 19.47, + "learning_rate": 1.053787412242887e-06, + "loss": 0.2581, + "step": 52721 + }, + { + "epoch": 19.47, + "learning_rate": 1.0537504618795418e-06, + "loss": 0.1853, + "step": 52722 + }, + { + "epoch": 19.47, + "learning_rate": 1.0537135115161967e-06, + "loss": 0.1723, + "step": 52723 + }, + { + "epoch": 19.47, + "learning_rate": 1.0536765611528513e-06, + "loss": 0.344, + "step": 52724 + }, + { + "epoch": 19.47, + "learning_rate": 1.0536396107895061e-06, + "loss": 0.2274, + "step": 52725 + }, + { + "epoch": 19.47, + "learning_rate": 1.053602660426161e-06, + "loss": 0.2281, + "step": 52726 + }, + { + "epoch": 19.47, + "learning_rate": 1.0535657100628156e-06, + "loss": 0.1482, + "step": 52727 + }, + { + "epoch": 19.47, + "learning_rate": 1.0535287596994703e-06, + "loss": 0.1478, + "step": 52728 + }, + { + "epoch": 19.47, + "learning_rate": 1.0534918093361251e-06, + "loss": 0.2604, + "step": 52729 + }, + { + "epoch": 19.47, + "learning_rate": 1.05345485897278e-06, + "loss": 0.1511, + "step": 52730 + }, + { + "epoch": 19.47, + "learning_rate": 1.0534179086094346e-06, + "loss": 0.1616, + "step": 52731 + }, + { + "epoch": 19.47, + "learning_rate": 1.0533809582460894e-06, + "loss": 0.1557, + "step": 52732 + }, + { + "epoch": 19.47, + "learning_rate": 1.0533440078827443e-06, + "loss": 0.2025, + "step": 52733 + }, + { + "epoch": 19.47, + "learning_rate": 1.053307057519399e-06, + "loss": 0.1717, + "step": 52734 + }, + { + "epoch": 19.47, + "learning_rate": 1.0532701071560538e-06, + "loss": 0.2933, + "step": 52735 + }, + { + "epoch": 19.47, + "learning_rate": 1.0532331567927086e-06, + "loss": 0.2287, + "step": 52736 + }, + { + "epoch": 19.47, + "learning_rate": 1.0531962064293633e-06, + "loss": 0.2144, + "step": 52737 + }, + { + "epoch": 19.47, + "learning_rate": 1.053159256066018e-06, + "loss": 0.2857, + "step": 52738 + }, + { + "epoch": 19.48, + "learning_rate": 1.0531223057026727e-06, + "loss": 0.2213, + "step": 52739 + }, + { + "epoch": 19.48, + "learning_rate": 1.0530853553393274e-06, + "loss": 0.25, + "step": 52740 + }, + { + "epoch": 19.48, + "learning_rate": 1.0530484049759822e-06, + "loss": 0.24, + "step": 52741 + }, + { + "epoch": 19.48, + "learning_rate": 1.053011454612637e-06, + "loss": 0.2847, + "step": 52742 + }, + { + "epoch": 19.48, + "learning_rate": 1.0529745042492917e-06, + "loss": 0.2637, + "step": 52743 + }, + { + "epoch": 19.48, + "learning_rate": 1.0529375538859466e-06, + "loss": 0.316, + "step": 52744 + }, + { + "epoch": 19.48, + "learning_rate": 1.0529006035226014e-06, + "loss": 0.1371, + "step": 52745 + }, + { + "epoch": 19.48, + "learning_rate": 1.052863653159256e-06, + "loss": 0.2259, + "step": 52746 + }, + { + "epoch": 19.48, + "learning_rate": 1.052826702795911e-06, + "loss": 0.1155, + "step": 52747 + }, + { + "epoch": 19.48, + "learning_rate": 1.0527897524325655e-06, + "loss": 0.1897, + "step": 52748 + }, + { + "epoch": 19.48, + "learning_rate": 1.0527528020692204e-06, + "loss": 0.4386, + "step": 52749 + }, + { + "epoch": 19.48, + "learning_rate": 1.052715851705875e-06, + "loss": 0.1464, + "step": 52750 + }, + { + "epoch": 19.48, + "learning_rate": 1.0526789013425299e-06, + "loss": 0.267, + "step": 52751 + }, + { + "epoch": 19.48, + "learning_rate": 1.0526419509791847e-06, + "loss": 0.1612, + "step": 52752 + }, + { + "epoch": 19.48, + "learning_rate": 1.0526050006158393e-06, + "loss": 0.1803, + "step": 52753 + }, + { + "epoch": 19.48, + "learning_rate": 1.0525680502524942e-06, + "loss": 0.2304, + "step": 52754 + }, + { + "epoch": 19.48, + "learning_rate": 1.052531099889149e-06, + "loss": 0.2801, + "step": 52755 + }, + { + "epoch": 19.48, + "learning_rate": 1.0524941495258037e-06, + "loss": 0.2805, + "step": 52756 + }, + { + "epoch": 19.48, + "learning_rate": 1.0524571991624585e-06, + "loss": 0.2701, + "step": 52757 + }, + { + "epoch": 19.48, + "learning_rate": 1.0524202487991132e-06, + "loss": 0.2002, + "step": 52758 + }, + { + "epoch": 19.48, + "learning_rate": 1.052383298435768e-06, + "loss": 0.1056, + "step": 52759 + }, + { + "epoch": 19.48, + "learning_rate": 1.0523463480724227e-06, + "loss": 0.229, + "step": 52760 + }, + { + "epoch": 19.48, + "learning_rate": 1.0523093977090775e-06, + "loss": 0.1535, + "step": 52761 + }, + { + "epoch": 19.48, + "learning_rate": 1.0522724473457323e-06, + "loss": 0.2172, + "step": 52762 + }, + { + "epoch": 19.48, + "learning_rate": 1.052235496982387e-06, + "loss": 0.139, + "step": 52763 + }, + { + "epoch": 19.48, + "learning_rate": 1.0521985466190418e-06, + "loss": 0.238, + "step": 52764 + }, + { + "epoch": 19.48, + "learning_rate": 1.0521615962556967e-06, + "loss": 0.4047, + "step": 52765 + }, + { + "epoch": 19.49, + "learning_rate": 1.0521246458923513e-06, + "loss": 0.1848, + "step": 52766 + }, + { + "epoch": 19.49, + "learning_rate": 1.0520876955290062e-06, + "loss": 0.1747, + "step": 52767 + }, + { + "epoch": 19.49, + "learning_rate": 1.0520507451656608e-06, + "loss": 0.1447, + "step": 52768 + }, + { + "epoch": 19.49, + "learning_rate": 1.0520137948023154e-06, + "loss": 0.2275, + "step": 52769 + }, + { + "epoch": 19.49, + "learning_rate": 1.0519768444389703e-06, + "loss": 0.2225, + "step": 52770 + }, + { + "epoch": 19.49, + "learning_rate": 1.0519398940756251e-06, + "loss": 0.1084, + "step": 52771 + }, + { + "epoch": 19.49, + "learning_rate": 1.0519029437122798e-06, + "loss": 0.167, + "step": 52772 + }, + { + "epoch": 19.49, + "learning_rate": 1.0518659933489346e-06, + "loss": 0.2132, + "step": 52773 + }, + { + "epoch": 19.49, + "learning_rate": 1.0518290429855895e-06, + "loss": 0.2573, + "step": 52774 + }, + { + "epoch": 19.49, + "learning_rate": 1.051792092622244e-06, + "loss": 0.1041, + "step": 52775 + }, + { + "epoch": 19.49, + "learning_rate": 1.051755142258899e-06, + "loss": 0.2619, + "step": 52776 + }, + { + "epoch": 19.49, + "learning_rate": 1.0517181918955538e-06, + "loss": 0.2186, + "step": 52777 + }, + { + "epoch": 19.49, + "learning_rate": 1.0516812415322084e-06, + "loss": 0.1752, + "step": 52778 + }, + { + "epoch": 19.49, + "learning_rate": 1.051644291168863e-06, + "loss": 0.2507, + "step": 52779 + }, + { + "epoch": 19.49, + "learning_rate": 1.051607340805518e-06, + "loss": 0.212, + "step": 52780 + }, + { + "epoch": 19.49, + "learning_rate": 1.0515703904421728e-06, + "loss": 0.2412, + "step": 52781 + }, + { + "epoch": 19.49, + "learning_rate": 1.0515334400788274e-06, + "loss": 0.183, + "step": 52782 + }, + { + "epoch": 19.49, + "learning_rate": 1.0514964897154823e-06, + "loss": 0.2577, + "step": 52783 + }, + { + "epoch": 19.49, + "learning_rate": 1.051459539352137e-06, + "loss": 0.234, + "step": 52784 + }, + { + "epoch": 19.49, + "learning_rate": 1.0514225889887917e-06, + "loss": 0.2229, + "step": 52785 + }, + { + "epoch": 19.49, + "learning_rate": 1.0513856386254466e-06, + "loss": 0.2202, + "step": 52786 + }, + { + "epoch": 19.49, + "learning_rate": 1.0513486882621014e-06, + "loss": 0.2723, + "step": 52787 + }, + { + "epoch": 19.49, + "learning_rate": 1.0513117378987559e-06, + "loss": 0.1488, + "step": 52788 + }, + { + "epoch": 19.49, + "learning_rate": 1.0512747875354107e-06, + "loss": 0.2543, + "step": 52789 + }, + { + "epoch": 19.49, + "learning_rate": 1.0512378371720656e-06, + "loss": 0.2209, + "step": 52790 + }, + { + "epoch": 19.49, + "learning_rate": 1.0512008868087202e-06, + "loss": 0.1688, + "step": 52791 + }, + { + "epoch": 19.49, + "learning_rate": 1.051163936445375e-06, + "loss": 0.2529, + "step": 52792 + }, + { + "epoch": 19.5, + "learning_rate": 1.0511269860820299e-06, + "loss": 0.1555, + "step": 52793 + }, + { + "epoch": 19.5, + "learning_rate": 1.0510900357186845e-06, + "loss": 0.182, + "step": 52794 + }, + { + "epoch": 19.5, + "learning_rate": 1.0510530853553394e-06, + "loss": 0.121, + "step": 52795 + }, + { + "epoch": 19.5, + "learning_rate": 1.0510161349919942e-06, + "loss": 0.1768, + "step": 52796 + }, + { + "epoch": 19.5, + "learning_rate": 1.0509791846286489e-06, + "loss": 0.1329, + "step": 52797 + }, + { + "epoch": 19.5, + "learning_rate": 1.0509422342653035e-06, + "loss": 0.2207, + "step": 52798 + }, + { + "epoch": 19.5, + "learning_rate": 1.0509052839019583e-06, + "loss": 0.2307, + "step": 52799 + }, + { + "epoch": 19.5, + "learning_rate": 1.0508683335386132e-06, + "loss": 0.3389, + "step": 52800 + }, + { + "epoch": 19.5, + "learning_rate": 1.0508313831752678e-06, + "loss": 0.2613, + "step": 52801 + }, + { + "epoch": 19.5, + "learning_rate": 1.0507944328119227e-06, + "loss": 0.1884, + "step": 52802 + }, + { + "epoch": 19.5, + "learning_rate": 1.0507574824485775e-06, + "loss": 0.1683, + "step": 52803 + }, + { + "epoch": 19.5, + "learning_rate": 1.0507205320852322e-06, + "loss": 0.1696, + "step": 52804 + }, + { + "epoch": 19.5, + "learning_rate": 1.050683581721887e-06, + "loss": 0.2223, + "step": 52805 + }, + { + "epoch": 19.5, + "learning_rate": 1.0506466313585419e-06, + "loss": 0.1749, + "step": 52806 + }, + { + "epoch": 19.5, + "learning_rate": 1.0506096809951965e-06, + "loss": 0.145, + "step": 52807 + }, + { + "epoch": 19.5, + "learning_rate": 1.0505727306318511e-06, + "loss": 0.1374, + "step": 52808 + }, + { + "epoch": 19.5, + "learning_rate": 1.050535780268506e-06, + "loss": 0.342, + "step": 52809 + }, + { + "epoch": 19.5, + "learning_rate": 1.0504988299051606e-06, + "loss": 0.2461, + "step": 52810 + }, + { + "epoch": 19.5, + "learning_rate": 1.0504618795418155e-06, + "loss": 0.3494, + "step": 52811 + }, + { + "epoch": 19.5, + "learning_rate": 1.0504249291784703e-06, + "loss": 0.1135, + "step": 52812 + }, + { + "epoch": 19.5, + "learning_rate": 1.050387978815125e-06, + "loss": 0.308, + "step": 52813 + }, + { + "epoch": 19.5, + "learning_rate": 1.0503510284517798e-06, + "loss": 0.3502, + "step": 52814 + }, + { + "epoch": 19.5, + "learning_rate": 1.0503140780884346e-06, + "loss": 0.3214, + "step": 52815 + }, + { + "epoch": 19.5, + "learning_rate": 1.0502771277250893e-06, + "loss": 0.1572, + "step": 52816 + }, + { + "epoch": 19.5, + "learning_rate": 1.0502401773617441e-06, + "loss": 0.1618, + "step": 52817 + }, + { + "epoch": 19.5, + "learning_rate": 1.0502032269983988e-06, + "loss": 0.3234, + "step": 52818 + }, + { + "epoch": 19.5, + "learning_rate": 1.0501662766350536e-06, + "loss": 0.1449, + "step": 52819 + }, + { + "epoch": 19.51, + "learning_rate": 1.0501293262717082e-06, + "loss": 0.192, + "step": 52820 + }, + { + "epoch": 19.51, + "learning_rate": 1.050092375908363e-06, + "loss": 0.1613, + "step": 52821 + }, + { + "epoch": 19.51, + "learning_rate": 1.050055425545018e-06, + "loss": 0.2936, + "step": 52822 + }, + { + "epoch": 19.51, + "learning_rate": 1.0500184751816726e-06, + "loss": 0.1735, + "step": 52823 + }, + { + "epoch": 19.51, + "learning_rate": 1.0499815248183274e-06, + "loss": 0.234, + "step": 52824 + }, + { + "epoch": 19.51, + "learning_rate": 1.0499445744549823e-06, + "loss": 0.209, + "step": 52825 + }, + { + "epoch": 19.51, + "learning_rate": 1.049907624091637e-06, + "loss": 0.309, + "step": 52826 + }, + { + "epoch": 19.51, + "learning_rate": 1.0498706737282918e-06, + "loss": 0.2085, + "step": 52827 + }, + { + "epoch": 19.51, + "learning_rate": 1.0498337233649464e-06, + "loss": 0.1705, + "step": 52828 + }, + { + "epoch": 19.51, + "learning_rate": 1.0497967730016012e-06, + "loss": 0.1556, + "step": 52829 + }, + { + "epoch": 19.51, + "learning_rate": 1.0497598226382559e-06, + "loss": 0.1631, + "step": 52830 + }, + { + "epoch": 19.51, + "learning_rate": 1.0497228722749107e-06, + "loss": 0.2665, + "step": 52831 + }, + { + "epoch": 19.51, + "learning_rate": 1.0496859219115656e-06, + "loss": 0.1955, + "step": 52832 + }, + { + "epoch": 19.51, + "learning_rate": 1.0496489715482202e-06, + "loss": 0.3212, + "step": 52833 + }, + { + "epoch": 19.51, + "learning_rate": 1.049612021184875e-06, + "loss": 0.1885, + "step": 52834 + }, + { + "epoch": 19.51, + "learning_rate": 1.04957507082153e-06, + "loss": 0.2537, + "step": 52835 + }, + { + "epoch": 19.51, + "learning_rate": 1.0495381204581845e-06, + "loss": 0.1736, + "step": 52836 + }, + { + "epoch": 19.51, + "learning_rate": 1.0495011700948394e-06, + "loss": 0.0929, + "step": 52837 + }, + { + "epoch": 19.51, + "learning_rate": 1.049464219731494e-06, + "loss": 0.2741, + "step": 52838 + }, + { + "epoch": 19.51, + "learning_rate": 1.0494272693681487e-06, + "loss": 0.0978, + "step": 52839 + }, + { + "epoch": 19.51, + "learning_rate": 1.0493903190048035e-06, + "loss": 0.1661, + "step": 52840 + }, + { + "epoch": 19.51, + "learning_rate": 1.0493533686414584e-06, + "loss": 0.2856, + "step": 52841 + }, + { + "epoch": 19.51, + "learning_rate": 1.049316418278113e-06, + "loss": 0.2282, + "step": 52842 + }, + { + "epoch": 19.51, + "learning_rate": 1.0492794679147679e-06, + "loss": 0.2508, + "step": 52843 + }, + { + "epoch": 19.51, + "learning_rate": 1.0492425175514227e-06, + "loss": 0.1812, + "step": 52844 + }, + { + "epoch": 19.51, + "learning_rate": 1.0492055671880773e-06, + "loss": 0.1993, + "step": 52845 + }, + { + "epoch": 19.51, + "learning_rate": 1.0491686168247322e-06, + "loss": 0.2302, + "step": 52846 + }, + { + "epoch": 19.52, + "learning_rate": 1.049131666461387e-06, + "loss": 0.1446, + "step": 52847 + }, + { + "epoch": 19.52, + "learning_rate": 1.0490947160980417e-06, + "loss": 0.1405, + "step": 52848 + }, + { + "epoch": 19.52, + "learning_rate": 1.0490577657346963e-06, + "loss": 0.2341, + "step": 52849 + }, + { + "epoch": 19.52, + "learning_rate": 1.0490208153713512e-06, + "loss": 0.2252, + "step": 52850 + }, + { + "epoch": 19.52, + "learning_rate": 1.048983865008006e-06, + "loss": 0.3557, + "step": 52851 + }, + { + "epoch": 19.52, + "learning_rate": 1.0489469146446606e-06, + "loss": 0.2469, + "step": 52852 + }, + { + "epoch": 19.52, + "learning_rate": 1.0489099642813155e-06, + "loss": 0.1774, + "step": 52853 + }, + { + "epoch": 19.52, + "learning_rate": 1.0488730139179703e-06, + "loss": 0.2016, + "step": 52854 + }, + { + "epoch": 19.52, + "learning_rate": 1.048836063554625e-06, + "loss": 0.189, + "step": 52855 + }, + { + "epoch": 19.52, + "learning_rate": 1.0487991131912798e-06, + "loss": 0.2375, + "step": 52856 + }, + { + "epoch": 19.52, + "learning_rate": 1.0487621628279347e-06, + "loss": 0.1799, + "step": 52857 + }, + { + "epoch": 19.52, + "learning_rate": 1.048725212464589e-06, + "loss": 0.2242, + "step": 52858 + }, + { + "epoch": 19.52, + "learning_rate": 1.048688262101244e-06, + "loss": 0.2778, + "step": 52859 + }, + { + "epoch": 19.52, + "learning_rate": 1.0486513117378988e-06, + "loss": 0.2801, + "step": 52860 + }, + { + "epoch": 19.52, + "learning_rate": 1.0486143613745534e-06, + "loss": 0.2762, + "step": 52861 + }, + { + "epoch": 19.52, + "learning_rate": 1.0485774110112083e-06, + "loss": 0.2053, + "step": 52862 + }, + { + "epoch": 19.52, + "learning_rate": 1.0485404606478631e-06, + "loss": 0.2617, + "step": 52863 + }, + { + "epoch": 19.52, + "learning_rate": 1.0485035102845178e-06, + "loss": 0.1439, + "step": 52864 + }, + { + "epoch": 19.52, + "learning_rate": 1.0484665599211726e-06, + "loss": 0.1066, + "step": 52865 + }, + { + "epoch": 19.52, + "learning_rate": 1.0484296095578275e-06, + "loss": 0.214, + "step": 52866 + }, + { + "epoch": 19.52, + "learning_rate": 1.048392659194482e-06, + "loss": 0.2846, + "step": 52867 + }, + { + "epoch": 19.52, + "learning_rate": 1.0483557088311367e-06, + "loss": 0.366, + "step": 52868 + }, + { + "epoch": 19.52, + "learning_rate": 1.0483187584677916e-06, + "loss": 0.1726, + "step": 52869 + }, + { + "epoch": 19.52, + "learning_rate": 1.0482818081044464e-06, + "loss": 0.3076, + "step": 52870 + }, + { + "epoch": 19.52, + "learning_rate": 1.048244857741101e-06, + "loss": 0.1526, + "step": 52871 + }, + { + "epoch": 19.52, + "learning_rate": 1.048207907377756e-06, + "loss": 0.1838, + "step": 52872 + }, + { + "epoch": 19.52, + "learning_rate": 1.0481709570144108e-06, + "loss": 0.1855, + "step": 52873 + }, + { + "epoch": 19.53, + "learning_rate": 1.0481340066510654e-06, + "loss": 0.1878, + "step": 52874 + }, + { + "epoch": 19.53, + "learning_rate": 1.0480970562877202e-06, + "loss": 0.113, + "step": 52875 + }, + { + "epoch": 19.53, + "learning_rate": 1.048060105924375e-06, + "loss": 0.166, + "step": 52876 + }, + { + "epoch": 19.53, + "learning_rate": 1.0480231555610297e-06, + "loss": 0.2311, + "step": 52877 + }, + { + "epoch": 19.53, + "learning_rate": 1.0479862051976844e-06, + "loss": 0.2712, + "step": 52878 + }, + { + "epoch": 19.53, + "learning_rate": 1.0479492548343392e-06, + "loss": 0.1904, + "step": 52879 + }, + { + "epoch": 19.53, + "learning_rate": 1.047912304470994e-06, + "loss": 0.1986, + "step": 52880 + }, + { + "epoch": 19.53, + "learning_rate": 1.0478753541076487e-06, + "loss": 0.1951, + "step": 52881 + }, + { + "epoch": 19.53, + "learning_rate": 1.0478384037443035e-06, + "loss": 0.2264, + "step": 52882 + }, + { + "epoch": 19.53, + "learning_rate": 1.0478014533809584e-06, + "loss": 0.1098, + "step": 52883 + }, + { + "epoch": 19.53, + "learning_rate": 1.047764503017613e-06, + "loss": 0.2691, + "step": 52884 + }, + { + "epoch": 19.53, + "learning_rate": 1.0477275526542679e-06, + "loss": 0.1502, + "step": 52885 + }, + { + "epoch": 19.53, + "learning_rate": 1.0476906022909225e-06, + "loss": 0.1839, + "step": 52886 + }, + { + "epoch": 19.53, + "learning_rate": 1.0476536519275774e-06, + "loss": 0.172, + "step": 52887 + }, + { + "epoch": 19.53, + "learning_rate": 1.047616701564232e-06, + "loss": 0.2421, + "step": 52888 + }, + { + "epoch": 19.53, + "learning_rate": 1.0475797512008868e-06, + "loss": 0.2026, + "step": 52889 + }, + { + "epoch": 19.53, + "learning_rate": 1.0475428008375415e-06, + "loss": 0.1743, + "step": 52890 + }, + { + "epoch": 19.53, + "learning_rate": 1.0475058504741963e-06, + "loss": 0.235, + "step": 52891 + }, + { + "epoch": 19.53, + "learning_rate": 1.0474689001108512e-06, + "loss": 0.1388, + "step": 52892 + }, + { + "epoch": 19.53, + "learning_rate": 1.0474319497475058e-06, + "loss": 0.312, + "step": 52893 + }, + { + "epoch": 19.53, + "learning_rate": 1.0473949993841607e-06, + "loss": 0.0948, + "step": 52894 + }, + { + "epoch": 19.53, + "learning_rate": 1.0473580490208155e-06, + "loss": 0.0803, + "step": 52895 + }, + { + "epoch": 19.53, + "learning_rate": 1.0473210986574701e-06, + "loss": 0.1198, + "step": 52896 + }, + { + "epoch": 19.53, + "learning_rate": 1.047284148294125e-06, + "loss": 0.1818, + "step": 52897 + }, + { + "epoch": 19.53, + "learning_rate": 1.0472471979307796e-06, + "loss": 0.2542, + "step": 52898 + }, + { + "epoch": 19.53, + "learning_rate": 1.0472102475674345e-06, + "loss": 0.2462, + "step": 52899 + }, + { + "epoch": 19.53, + "learning_rate": 1.0471732972040891e-06, + "loss": 0.2104, + "step": 52900 + }, + { + "epoch": 19.54, + "learning_rate": 1.047136346840744e-06, + "loss": 0.0915, + "step": 52901 + }, + { + "epoch": 19.54, + "learning_rate": 1.0470993964773988e-06, + "loss": 0.2232, + "step": 52902 + }, + { + "epoch": 19.54, + "learning_rate": 1.0470624461140534e-06, + "loss": 0.2257, + "step": 52903 + }, + { + "epoch": 19.54, + "learning_rate": 1.0470254957507083e-06, + "loss": 0.1917, + "step": 52904 + }, + { + "epoch": 19.54, + "learning_rate": 1.0469885453873631e-06, + "loss": 0.1424, + "step": 52905 + }, + { + "epoch": 19.54, + "learning_rate": 1.0469515950240178e-06, + "loss": 0.2739, + "step": 52906 + }, + { + "epoch": 19.54, + "learning_rate": 1.0469146446606726e-06, + "loss": 0.2903, + "step": 52907 + }, + { + "epoch": 19.54, + "learning_rate": 1.0468776942973273e-06, + "loss": 0.2373, + "step": 52908 + }, + { + "epoch": 19.54, + "learning_rate": 1.046840743933982e-06, + "loss": 0.1092, + "step": 52909 + }, + { + "epoch": 19.54, + "learning_rate": 1.0468037935706368e-06, + "loss": 0.1825, + "step": 52910 + }, + { + "epoch": 19.54, + "learning_rate": 1.0467668432072916e-06, + "loss": 0.1617, + "step": 52911 + }, + { + "epoch": 19.54, + "learning_rate": 1.0467298928439462e-06, + "loss": 0.2061, + "step": 52912 + }, + { + "epoch": 19.54, + "learning_rate": 1.046692942480601e-06, + "loss": 0.2222, + "step": 52913 + }, + { + "epoch": 19.54, + "learning_rate": 1.046655992117256e-06, + "loss": 0.2125, + "step": 52914 + }, + { + "epoch": 19.54, + "learning_rate": 1.0466190417539106e-06, + "loss": 0.225, + "step": 52915 + }, + { + "epoch": 19.54, + "learning_rate": 1.0465820913905654e-06, + "loss": 0.1826, + "step": 52916 + }, + { + "epoch": 19.54, + "learning_rate": 1.0465451410272203e-06, + "loss": 0.2238, + "step": 52917 + }, + { + "epoch": 19.54, + "learning_rate": 1.046508190663875e-06, + "loss": 0.3357, + "step": 52918 + }, + { + "epoch": 19.54, + "learning_rate": 1.0464712403005295e-06, + "loss": 0.246, + "step": 52919 + }, + { + "epoch": 19.54, + "learning_rate": 1.0464342899371844e-06, + "loss": 0.1547, + "step": 52920 + }, + { + "epoch": 19.54, + "learning_rate": 1.0463973395738392e-06, + "loss": 0.1695, + "step": 52921 + }, + { + "epoch": 19.54, + "learning_rate": 1.0463603892104939e-06, + "loss": 0.1585, + "step": 52922 + }, + { + "epoch": 19.54, + "learning_rate": 1.0463234388471487e-06, + "loss": 0.2575, + "step": 52923 + }, + { + "epoch": 19.54, + "learning_rate": 1.0462864884838036e-06, + "loss": 0.1408, + "step": 52924 + }, + { + "epoch": 19.54, + "learning_rate": 1.0462495381204582e-06, + "loss": 0.1528, + "step": 52925 + }, + { + "epoch": 19.54, + "learning_rate": 1.046212587757113e-06, + "loss": 0.1789, + "step": 52926 + }, + { + "epoch": 19.54, + "learning_rate": 1.046175637393768e-06, + "loss": 0.3046, + "step": 52927 + }, + { + "epoch": 19.55, + "learning_rate": 1.0461386870304223e-06, + "loss": 0.1442, + "step": 52928 + }, + { + "epoch": 19.55, + "learning_rate": 1.0461017366670772e-06, + "loss": 0.1781, + "step": 52929 + }, + { + "epoch": 19.55, + "learning_rate": 1.046064786303732e-06, + "loss": 0.133, + "step": 52930 + }, + { + "epoch": 19.55, + "learning_rate": 1.0460278359403867e-06, + "loss": 0.1968, + "step": 52931 + }, + { + "epoch": 19.55, + "learning_rate": 1.0459908855770415e-06, + "loss": 0.3052, + "step": 52932 + }, + { + "epoch": 19.55, + "learning_rate": 1.0459539352136964e-06, + "loss": 0.2748, + "step": 52933 + }, + { + "epoch": 19.55, + "learning_rate": 1.045916984850351e-06, + "loss": 0.1454, + "step": 52934 + }, + { + "epoch": 19.55, + "learning_rate": 1.0458800344870058e-06, + "loss": 0.1815, + "step": 52935 + }, + { + "epoch": 19.55, + "learning_rate": 1.0458430841236607e-06, + "loss": 0.1548, + "step": 52936 + }, + { + "epoch": 19.55, + "learning_rate": 1.0458061337603153e-06, + "loss": 0.1857, + "step": 52937 + }, + { + "epoch": 19.55, + "learning_rate": 1.04576918339697e-06, + "loss": 0.2109, + "step": 52938 + }, + { + "epoch": 19.55, + "learning_rate": 1.0457322330336248e-06, + "loss": 0.2241, + "step": 52939 + }, + { + "epoch": 19.55, + "learning_rate": 1.0456952826702797e-06, + "loss": 0.3159, + "step": 52940 + }, + { + "epoch": 19.55, + "learning_rate": 1.0456583323069343e-06, + "loss": 0.1757, + "step": 52941 + }, + { + "epoch": 19.55, + "learning_rate": 1.0456213819435891e-06, + "loss": 0.2181, + "step": 52942 + }, + { + "epoch": 19.55, + "learning_rate": 1.045584431580244e-06, + "loss": 0.1638, + "step": 52943 + }, + { + "epoch": 19.55, + "learning_rate": 1.0455474812168986e-06, + "loss": 0.1491, + "step": 52944 + }, + { + "epoch": 19.55, + "learning_rate": 1.0455105308535535e-06, + "loss": 0.1677, + "step": 52945 + }, + { + "epoch": 19.55, + "learning_rate": 1.0454735804902083e-06, + "loss": 0.2879, + "step": 52946 + }, + { + "epoch": 19.55, + "learning_rate": 1.045436630126863e-06, + "loss": 0.1718, + "step": 52947 + }, + { + "epoch": 19.55, + "learning_rate": 1.0453996797635176e-06, + "loss": 0.2568, + "step": 52948 + }, + { + "epoch": 19.55, + "learning_rate": 1.0453627294001724e-06, + "loss": 0.1542, + "step": 52949 + }, + { + "epoch": 19.55, + "learning_rate": 1.0453257790368273e-06, + "loss": 0.3402, + "step": 52950 + }, + { + "epoch": 19.55, + "learning_rate": 1.045288828673482e-06, + "loss": 0.2036, + "step": 52951 + }, + { + "epoch": 19.55, + "learning_rate": 1.0452518783101368e-06, + "loss": 0.2584, + "step": 52952 + }, + { + "epoch": 19.55, + "learning_rate": 1.0452149279467916e-06, + "loss": 0.1002, + "step": 52953 + }, + { + "epoch": 19.55, + "learning_rate": 1.0451779775834463e-06, + "loss": 0.1764, + "step": 52954 + }, + { + "epoch": 19.56, + "learning_rate": 1.0451410272201011e-06, + "loss": 0.3694, + "step": 52955 + }, + { + "epoch": 19.56, + "learning_rate": 1.045104076856756e-06, + "loss": 0.135, + "step": 52956 + }, + { + "epoch": 19.56, + "learning_rate": 1.0450671264934104e-06, + "loss": 0.3014, + "step": 52957 + }, + { + "epoch": 19.56, + "learning_rate": 1.0450301761300652e-06, + "loss": 0.1952, + "step": 52958 + }, + { + "epoch": 19.56, + "learning_rate": 1.04499322576672e-06, + "loss": 0.1974, + "step": 52959 + }, + { + "epoch": 19.56, + "learning_rate": 1.0449562754033747e-06, + "loss": 0.2041, + "step": 52960 + }, + { + "epoch": 19.56, + "learning_rate": 1.0449193250400296e-06, + "loss": 0.3136, + "step": 52961 + }, + { + "epoch": 19.56, + "learning_rate": 1.0448823746766844e-06, + "loss": 0.132, + "step": 52962 + }, + { + "epoch": 19.56, + "learning_rate": 1.044845424313339e-06, + "loss": 0.1632, + "step": 52963 + }, + { + "epoch": 19.56, + "learning_rate": 1.044808473949994e-06, + "loss": 0.1668, + "step": 52964 + }, + { + "epoch": 19.56, + "learning_rate": 1.0447715235866487e-06, + "loss": 0.2103, + "step": 52965 + }, + { + "epoch": 19.56, + "learning_rate": 1.0447345732233034e-06, + "loss": 0.1961, + "step": 52966 + }, + { + "epoch": 19.56, + "learning_rate": 1.044697622859958e-06, + "loss": 0.2599, + "step": 52967 + }, + { + "epoch": 19.56, + "learning_rate": 1.0446606724966129e-06, + "loss": 0.1993, + "step": 52968 + }, + { + "epoch": 19.56, + "learning_rate": 1.0446237221332677e-06, + "loss": 0.3547, + "step": 52969 + }, + { + "epoch": 19.56, + "learning_rate": 1.0445867717699223e-06, + "loss": 0.3181, + "step": 52970 + }, + { + "epoch": 19.56, + "learning_rate": 1.0445498214065772e-06, + "loss": 0.1554, + "step": 52971 + }, + { + "epoch": 19.56, + "learning_rate": 1.044512871043232e-06, + "loss": 0.3786, + "step": 52972 + }, + { + "epoch": 19.56, + "learning_rate": 1.0444759206798867e-06, + "loss": 0.2982, + "step": 52973 + }, + { + "epoch": 19.56, + "learning_rate": 1.0444389703165415e-06, + "loss": 0.2261, + "step": 52974 + }, + { + "epoch": 19.56, + "learning_rate": 1.0444020199531964e-06, + "loss": 0.2575, + "step": 52975 + }, + { + "epoch": 19.56, + "learning_rate": 1.044365069589851e-06, + "loss": 0.122, + "step": 52976 + }, + { + "epoch": 19.56, + "learning_rate": 1.0443281192265057e-06, + "loss": 0.2123, + "step": 52977 + }, + { + "epoch": 19.56, + "learning_rate": 1.0442911688631605e-06, + "loss": 0.2169, + "step": 52978 + }, + { + "epoch": 19.56, + "learning_rate": 1.0442542184998151e-06, + "loss": 0.2586, + "step": 52979 + }, + { + "epoch": 19.56, + "learning_rate": 1.04421726813647e-06, + "loss": 0.1279, + "step": 52980 + }, + { + "epoch": 19.56, + "learning_rate": 1.0441803177731248e-06, + "loss": 0.1644, + "step": 52981 + }, + { + "epoch": 19.56, + "learning_rate": 1.0441433674097795e-06, + "loss": 0.2573, + "step": 52982 + }, + { + "epoch": 19.57, + "learning_rate": 1.0441064170464343e-06, + "loss": 0.1708, + "step": 52983 + }, + { + "epoch": 19.57, + "learning_rate": 1.0440694666830892e-06, + "loss": 0.2001, + "step": 52984 + }, + { + "epoch": 19.57, + "learning_rate": 1.0440325163197438e-06, + "loss": 0.2165, + "step": 52985 + }, + { + "epoch": 19.57, + "learning_rate": 1.0439955659563986e-06, + "loss": 0.1903, + "step": 52986 + }, + { + "epoch": 19.57, + "learning_rate": 1.0439586155930533e-06, + "loss": 0.1648, + "step": 52987 + }, + { + "epoch": 19.57, + "learning_rate": 1.0439216652297081e-06, + "loss": 0.2724, + "step": 52988 + }, + { + "epoch": 19.57, + "learning_rate": 1.0438847148663628e-06, + "loss": 0.135, + "step": 52989 + }, + { + "epoch": 19.57, + "learning_rate": 1.0438477645030176e-06, + "loss": 0.4157, + "step": 52990 + }, + { + "epoch": 19.57, + "learning_rate": 1.0438108141396725e-06, + "loss": 0.2366, + "step": 52991 + }, + { + "epoch": 19.57, + "learning_rate": 1.043773863776327e-06, + "loss": 0.254, + "step": 52992 + }, + { + "epoch": 19.57, + "learning_rate": 1.043736913412982e-06, + "loss": 0.1594, + "step": 52993 + }, + { + "epoch": 19.57, + "learning_rate": 1.0436999630496368e-06, + "loss": 0.2517, + "step": 52994 + }, + { + "epoch": 19.57, + "learning_rate": 1.0436630126862914e-06, + "loss": 0.2849, + "step": 52995 + }, + { + "epoch": 19.57, + "learning_rate": 1.0436260623229463e-06, + "loss": 0.1972, + "step": 52996 + }, + { + "epoch": 19.57, + "learning_rate": 1.043589111959601e-06, + "loss": 0.2959, + "step": 52997 + }, + { + "epoch": 19.57, + "learning_rate": 1.0435521615962556e-06, + "loss": 0.1224, + "step": 52998 + }, + { + "epoch": 19.57, + "learning_rate": 1.0435152112329104e-06, + "loss": 0.2366, + "step": 52999 + }, + { + "epoch": 19.57, + "learning_rate": 1.0434782608695653e-06, + "loss": 0.3374, + "step": 53000 + }, + { + "epoch": 19.57, + "learning_rate": 1.0434413105062199e-06, + "loss": 0.1442, + "step": 53001 + }, + { + "epoch": 19.57, + "learning_rate": 1.0434043601428747e-06, + "loss": 0.2265, + "step": 53002 + }, + { + "epoch": 19.57, + "learning_rate": 1.0433674097795296e-06, + "loss": 0.2121, + "step": 53003 + }, + { + "epoch": 19.57, + "learning_rate": 1.0433304594161842e-06, + "loss": 0.2068, + "step": 53004 + }, + { + "epoch": 19.57, + "learning_rate": 1.043293509052839e-06, + "loss": 0.2102, + "step": 53005 + }, + { + "epoch": 19.57, + "learning_rate": 1.043256558689494e-06, + "loss": 0.2012, + "step": 53006 + }, + { + "epoch": 19.57, + "learning_rate": 1.0432196083261486e-06, + "loss": 0.1706, + "step": 53007 + }, + { + "epoch": 19.57, + "learning_rate": 1.0431826579628032e-06, + "loss": 0.2564, + "step": 53008 + }, + { + "epoch": 19.57, + "learning_rate": 1.043145707599458e-06, + "loss": 0.2798, + "step": 53009 + }, + { + "epoch": 19.58, + "learning_rate": 1.0431087572361129e-06, + "loss": 0.1286, + "step": 53010 + }, + { + "epoch": 19.58, + "learning_rate": 1.0430718068727675e-06, + "loss": 0.1742, + "step": 53011 + }, + { + "epoch": 19.58, + "learning_rate": 1.0430348565094224e-06, + "loss": 0.2142, + "step": 53012 + }, + { + "epoch": 19.58, + "learning_rate": 1.0429979061460772e-06, + "loss": 0.2897, + "step": 53013 + }, + { + "epoch": 19.58, + "learning_rate": 1.0429609557827319e-06, + "loss": 0.1164, + "step": 53014 + }, + { + "epoch": 19.58, + "learning_rate": 1.0429240054193867e-06, + "loss": 0.1597, + "step": 53015 + }, + { + "epoch": 19.58, + "learning_rate": 1.0428870550560416e-06, + "loss": 0.2638, + "step": 53016 + }, + { + "epoch": 19.58, + "learning_rate": 1.0428501046926962e-06, + "loss": 0.2052, + "step": 53017 + }, + { + "epoch": 19.58, + "learning_rate": 1.0428131543293508e-06, + "loss": 0.2591, + "step": 53018 + }, + { + "epoch": 19.58, + "learning_rate": 1.0427762039660057e-06, + "loss": 0.2567, + "step": 53019 + }, + { + "epoch": 19.58, + "learning_rate": 1.0427392536026605e-06, + "loss": 0.2805, + "step": 53020 + }, + { + "epoch": 19.58, + "learning_rate": 1.0427023032393152e-06, + "loss": 0.3164, + "step": 53021 + }, + { + "epoch": 19.58, + "learning_rate": 1.04266535287597e-06, + "loss": 0.3491, + "step": 53022 + }, + { + "epoch": 19.58, + "learning_rate": 1.0426284025126249e-06, + "loss": 0.0773, + "step": 53023 + }, + { + "epoch": 19.58, + "learning_rate": 1.0425914521492795e-06, + "loss": 0.3287, + "step": 53024 + }, + { + "epoch": 19.58, + "learning_rate": 1.0425545017859343e-06, + "loss": 0.2543, + "step": 53025 + }, + { + "epoch": 19.58, + "learning_rate": 1.0425175514225892e-06, + "loss": 0.2056, + "step": 53026 + }, + { + "epoch": 19.58, + "learning_rate": 1.0424806010592436e-06, + "loss": 0.357, + "step": 53027 + }, + { + "epoch": 19.58, + "learning_rate": 1.0424436506958985e-06, + "loss": 0.1935, + "step": 53028 + }, + { + "epoch": 19.58, + "learning_rate": 1.0424067003325533e-06, + "loss": 0.3719, + "step": 53029 + }, + { + "epoch": 19.58, + "learning_rate": 1.042369749969208e-06, + "loss": 0.3327, + "step": 53030 + }, + { + "epoch": 19.58, + "learning_rate": 1.0423327996058628e-06, + "loss": 0.2136, + "step": 53031 + }, + { + "epoch": 19.58, + "learning_rate": 1.0422958492425176e-06, + "loss": 0.3881, + "step": 53032 + }, + { + "epoch": 19.58, + "learning_rate": 1.0422588988791723e-06, + "loss": 0.2131, + "step": 53033 + }, + { + "epoch": 19.58, + "learning_rate": 1.0422219485158271e-06, + "loss": 0.1625, + "step": 53034 + }, + { + "epoch": 19.58, + "learning_rate": 1.042184998152482e-06, + "loss": 0.2314, + "step": 53035 + }, + { + "epoch": 19.58, + "learning_rate": 1.0421480477891366e-06, + "loss": 0.0938, + "step": 53036 + }, + { + "epoch": 19.59, + "learning_rate": 1.0421110974257912e-06, + "loss": 0.2744, + "step": 53037 + }, + { + "epoch": 19.59, + "learning_rate": 1.042074147062446e-06, + "loss": 0.2045, + "step": 53038 + }, + { + "epoch": 19.59, + "learning_rate": 1.042037196699101e-06, + "loss": 0.2232, + "step": 53039 + }, + { + "epoch": 19.59, + "learning_rate": 1.0420002463357556e-06, + "loss": 0.1458, + "step": 53040 + }, + { + "epoch": 19.59, + "learning_rate": 1.0419632959724104e-06, + "loss": 0.1594, + "step": 53041 + }, + { + "epoch": 19.59, + "learning_rate": 1.0419263456090653e-06, + "loss": 0.2529, + "step": 53042 + }, + { + "epoch": 19.59, + "learning_rate": 1.04188939524572e-06, + "loss": 0.2262, + "step": 53043 + }, + { + "epoch": 19.59, + "learning_rate": 1.0418524448823748e-06, + "loss": 0.1663, + "step": 53044 + }, + { + "epoch": 19.59, + "learning_rate": 1.0418154945190296e-06, + "loss": 0.205, + "step": 53045 + }, + { + "epoch": 19.59, + "learning_rate": 1.0417785441556842e-06, + "loss": 0.1693, + "step": 53046 + }, + { + "epoch": 19.59, + "learning_rate": 1.0417415937923389e-06, + "loss": 0.2567, + "step": 53047 + }, + { + "epoch": 19.59, + "learning_rate": 1.0417046434289937e-06, + "loss": 0.2791, + "step": 53048 + }, + { + "epoch": 19.59, + "learning_rate": 1.0416676930656484e-06, + "loss": 0.1855, + "step": 53049 + }, + { + "epoch": 19.59, + "learning_rate": 1.0416307427023032e-06, + "loss": 0.3464, + "step": 53050 + }, + { + "epoch": 19.59, + "learning_rate": 1.041593792338958e-06, + "loss": 0.157, + "step": 53051 + }, + { + "epoch": 19.59, + "learning_rate": 1.0415568419756127e-06, + "loss": 0.2464, + "step": 53052 + }, + { + "epoch": 19.59, + "learning_rate": 1.0415198916122675e-06, + "loss": 0.271, + "step": 53053 + }, + { + "epoch": 19.59, + "learning_rate": 1.0414829412489224e-06, + "loss": 0.2442, + "step": 53054 + }, + { + "epoch": 19.59, + "learning_rate": 1.041445990885577e-06, + "loss": 0.2525, + "step": 53055 + }, + { + "epoch": 19.59, + "learning_rate": 1.0414090405222319e-06, + "loss": 0.1192, + "step": 53056 + }, + { + "epoch": 19.59, + "learning_rate": 1.0413720901588865e-06, + "loss": 0.3239, + "step": 53057 + }, + { + "epoch": 19.59, + "learning_rate": 1.0413351397955414e-06, + "loss": 0.3181, + "step": 53058 + }, + { + "epoch": 19.59, + "learning_rate": 1.041298189432196e-06, + "loss": 0.1762, + "step": 53059 + }, + { + "epoch": 19.59, + "learning_rate": 1.0412612390688509e-06, + "loss": 0.2308, + "step": 53060 + }, + { + "epoch": 19.59, + "learning_rate": 1.0412242887055057e-06, + "loss": 0.1814, + "step": 53061 + }, + { + "epoch": 19.59, + "learning_rate": 1.0411873383421603e-06, + "loss": 0.3318, + "step": 53062 + }, + { + "epoch": 19.59, + "learning_rate": 1.0411503879788152e-06, + "loss": 0.2469, + "step": 53063 + }, + { + "epoch": 19.6, + "learning_rate": 1.04111343761547e-06, + "loss": 0.2305, + "step": 53064 + }, + { + "epoch": 19.6, + "learning_rate": 1.0410764872521247e-06, + "loss": 0.2224, + "step": 53065 + }, + { + "epoch": 19.6, + "learning_rate": 1.0410395368887795e-06, + "loss": 0.2654, + "step": 53066 + }, + { + "epoch": 19.6, + "learning_rate": 1.0410025865254342e-06, + "loss": 0.2404, + "step": 53067 + }, + { + "epoch": 19.6, + "learning_rate": 1.040965636162089e-06, + "loss": 0.161, + "step": 53068 + }, + { + "epoch": 19.6, + "learning_rate": 1.0409286857987436e-06, + "loss": 0.2588, + "step": 53069 + }, + { + "epoch": 19.6, + "learning_rate": 1.0408917354353985e-06, + "loss": 0.1827, + "step": 53070 + }, + { + "epoch": 19.6, + "learning_rate": 1.0408547850720531e-06, + "loss": 0.192, + "step": 53071 + }, + { + "epoch": 19.6, + "learning_rate": 1.040817834708708e-06, + "loss": 0.1413, + "step": 53072 + }, + { + "epoch": 19.6, + "learning_rate": 1.0407808843453628e-06, + "loss": 0.1755, + "step": 53073 + }, + { + "epoch": 19.6, + "learning_rate": 1.0407439339820175e-06, + "loss": 0.2341, + "step": 53074 + }, + { + "epoch": 19.6, + "learning_rate": 1.0407069836186723e-06, + "loss": 0.2655, + "step": 53075 + }, + { + "epoch": 19.6, + "learning_rate": 1.0406700332553272e-06, + "loss": 0.2242, + "step": 53076 + }, + { + "epoch": 19.6, + "learning_rate": 1.0406330828919818e-06, + "loss": 0.2297, + "step": 53077 + }, + { + "epoch": 19.6, + "learning_rate": 1.0405961325286364e-06, + "loss": 0.2173, + "step": 53078 + }, + { + "epoch": 19.6, + "learning_rate": 1.0405591821652913e-06, + "loss": 0.2249, + "step": 53079 + }, + { + "epoch": 19.6, + "learning_rate": 1.0405222318019461e-06, + "loss": 0.2015, + "step": 53080 + }, + { + "epoch": 19.6, + "learning_rate": 1.0404852814386008e-06, + "loss": 0.1428, + "step": 53081 + }, + { + "epoch": 19.6, + "learning_rate": 1.0404483310752556e-06, + "loss": 0.1889, + "step": 53082 + }, + { + "epoch": 19.6, + "learning_rate": 1.0404113807119105e-06, + "loss": 0.2593, + "step": 53083 + }, + { + "epoch": 19.6, + "learning_rate": 1.040374430348565e-06, + "loss": 0.2116, + "step": 53084 + }, + { + "epoch": 19.6, + "learning_rate": 1.04033747998522e-06, + "loss": 0.1411, + "step": 53085 + }, + { + "epoch": 19.6, + "learning_rate": 1.0403005296218748e-06, + "loss": 0.1713, + "step": 53086 + }, + { + "epoch": 19.6, + "learning_rate": 1.0402635792585294e-06, + "loss": 0.229, + "step": 53087 + }, + { + "epoch": 19.6, + "learning_rate": 1.040226628895184e-06, + "loss": 0.1864, + "step": 53088 + }, + { + "epoch": 19.6, + "learning_rate": 1.040189678531839e-06, + "loss": 0.1589, + "step": 53089 + }, + { + "epoch": 19.6, + "learning_rate": 1.0401527281684938e-06, + "loss": 0.2112, + "step": 53090 + }, + { + "epoch": 19.61, + "learning_rate": 1.0401157778051484e-06, + "loss": 0.2473, + "step": 53091 + }, + { + "epoch": 19.61, + "learning_rate": 1.0400788274418032e-06, + "loss": 0.1022, + "step": 53092 + }, + { + "epoch": 19.61, + "learning_rate": 1.040041877078458e-06, + "loss": 0.2148, + "step": 53093 + }, + { + "epoch": 19.61, + "learning_rate": 1.0400049267151127e-06, + "loss": 0.1903, + "step": 53094 + }, + { + "epoch": 19.61, + "learning_rate": 1.0399679763517676e-06, + "loss": 0.2356, + "step": 53095 + }, + { + "epoch": 19.61, + "learning_rate": 1.0399310259884224e-06, + "loss": 0.1833, + "step": 53096 + }, + { + "epoch": 19.61, + "learning_rate": 1.0398940756250768e-06, + "loss": 0.1789, + "step": 53097 + }, + { + "epoch": 19.61, + "learning_rate": 1.0398571252617317e-06, + "loss": 0.1432, + "step": 53098 + }, + { + "epoch": 19.61, + "learning_rate": 1.0398201748983865e-06, + "loss": 0.1491, + "step": 53099 + }, + { + "epoch": 19.61, + "learning_rate": 1.0397832245350412e-06, + "loss": 0.2448, + "step": 53100 + }, + { + "epoch": 19.61, + "learning_rate": 1.039746274171696e-06, + "loss": 0.3204, + "step": 53101 + }, + { + "epoch": 19.61, + "learning_rate": 1.0397093238083509e-06, + "loss": 0.2252, + "step": 53102 + }, + { + "epoch": 19.61, + "learning_rate": 1.0396723734450055e-06, + "loss": 0.3014, + "step": 53103 + }, + { + "epoch": 19.61, + "learning_rate": 1.0396354230816604e-06, + "loss": 0.1892, + "step": 53104 + }, + { + "epoch": 19.61, + "learning_rate": 1.0395984727183152e-06, + "loss": 0.1859, + "step": 53105 + }, + { + "epoch": 19.61, + "learning_rate": 1.0395615223549698e-06, + "loss": 0.346, + "step": 53106 + }, + { + "epoch": 19.61, + "learning_rate": 1.0395245719916245e-06, + "loss": 0.1897, + "step": 53107 + }, + { + "epoch": 19.61, + "learning_rate": 1.0394876216282793e-06, + "loss": 0.228, + "step": 53108 + }, + { + "epoch": 19.61, + "learning_rate": 1.0394506712649342e-06, + "loss": 0.3266, + "step": 53109 + }, + { + "epoch": 19.61, + "learning_rate": 1.0394137209015888e-06, + "loss": 0.174, + "step": 53110 + }, + { + "epoch": 19.61, + "learning_rate": 1.0393767705382437e-06, + "loss": 0.1312, + "step": 53111 + }, + { + "epoch": 19.61, + "learning_rate": 1.0393398201748985e-06, + "loss": 0.0724, + "step": 53112 + }, + { + "epoch": 19.61, + "learning_rate": 1.0393028698115531e-06, + "loss": 0.1246, + "step": 53113 + }, + { + "epoch": 19.61, + "learning_rate": 1.039265919448208e-06, + "loss": 0.2376, + "step": 53114 + }, + { + "epoch": 19.61, + "learning_rate": 1.0392289690848628e-06, + "loss": 0.2824, + "step": 53115 + }, + { + "epoch": 19.61, + "learning_rate": 1.0391920187215175e-06, + "loss": 0.1691, + "step": 53116 + }, + { + "epoch": 19.61, + "learning_rate": 1.0391550683581721e-06, + "loss": 0.2168, + "step": 53117 + }, + { + "epoch": 19.62, + "learning_rate": 1.039118117994827e-06, + "loss": 0.1865, + "step": 53118 + }, + { + "epoch": 19.62, + "learning_rate": 1.0390811676314816e-06, + "loss": 0.2191, + "step": 53119 + }, + { + "epoch": 19.62, + "learning_rate": 1.0390442172681364e-06, + "loss": 0.1362, + "step": 53120 + }, + { + "epoch": 19.62, + "learning_rate": 1.0390072669047913e-06, + "loss": 0.1796, + "step": 53121 + }, + { + "epoch": 19.62, + "learning_rate": 1.038970316541446e-06, + "loss": 0.1725, + "step": 53122 + }, + { + "epoch": 19.62, + "learning_rate": 1.0389333661781008e-06, + "loss": 0.2584, + "step": 53123 + }, + { + "epoch": 19.62, + "learning_rate": 1.0388964158147556e-06, + "loss": 0.2979, + "step": 53124 + }, + { + "epoch": 19.62, + "learning_rate": 1.0388594654514103e-06, + "loss": 0.3294, + "step": 53125 + }, + { + "epoch": 19.62, + "learning_rate": 1.0388225150880651e-06, + "loss": 0.2181, + "step": 53126 + }, + { + "epoch": 19.62, + "learning_rate": 1.0387855647247198e-06, + "loss": 0.246, + "step": 53127 + }, + { + "epoch": 19.62, + "learning_rate": 1.0387486143613746e-06, + "loss": 0.1124, + "step": 53128 + }, + { + "epoch": 19.62, + "learning_rate": 1.0387116639980292e-06, + "loss": 0.1943, + "step": 53129 + }, + { + "epoch": 19.62, + "learning_rate": 1.038674713634684e-06, + "loss": 0.2847, + "step": 53130 + }, + { + "epoch": 19.62, + "learning_rate": 1.038637763271339e-06, + "loss": 0.0864, + "step": 53131 + }, + { + "epoch": 19.62, + "learning_rate": 1.0386008129079936e-06, + "loss": 0.1023, + "step": 53132 + }, + { + "epoch": 19.62, + "learning_rate": 1.0385638625446484e-06, + "loss": 0.2327, + "step": 53133 + }, + { + "epoch": 19.62, + "learning_rate": 1.0385269121813033e-06, + "loss": 0.2956, + "step": 53134 + }, + { + "epoch": 19.62, + "learning_rate": 1.038489961817958e-06, + "loss": 0.1567, + "step": 53135 + }, + { + "epoch": 19.62, + "learning_rate": 1.0384530114546127e-06, + "loss": 0.2388, + "step": 53136 + }, + { + "epoch": 19.62, + "learning_rate": 1.0384160610912674e-06, + "loss": 0.1088, + "step": 53137 + }, + { + "epoch": 19.62, + "learning_rate": 1.0383791107279222e-06, + "loss": 0.2648, + "step": 53138 + }, + { + "epoch": 19.62, + "learning_rate": 1.0383421603645769e-06, + "loss": 0.1493, + "step": 53139 + }, + { + "epoch": 19.62, + "learning_rate": 1.0383052100012317e-06, + "loss": 0.209, + "step": 53140 + }, + { + "epoch": 19.62, + "learning_rate": 1.0382682596378866e-06, + "loss": 0.215, + "step": 53141 + }, + { + "epoch": 19.62, + "learning_rate": 1.0382313092745412e-06, + "loss": 0.2332, + "step": 53142 + }, + { + "epoch": 19.62, + "learning_rate": 1.038194358911196e-06, + "loss": 0.1312, + "step": 53143 + }, + { + "epoch": 19.62, + "learning_rate": 1.038157408547851e-06, + "loss": 0.1728, + "step": 53144 + }, + { + "epoch": 19.63, + "learning_rate": 1.0381204581845055e-06, + "loss": 0.1145, + "step": 53145 + }, + { + "epoch": 19.63, + "learning_rate": 1.0380835078211604e-06, + "loss": 0.2082, + "step": 53146 + }, + { + "epoch": 19.63, + "learning_rate": 1.038046557457815e-06, + "loss": 0.2086, + "step": 53147 + }, + { + "epoch": 19.63, + "learning_rate": 1.0380096070944697e-06, + "loss": 0.2775, + "step": 53148 + }, + { + "epoch": 19.63, + "learning_rate": 1.0379726567311245e-06, + "loss": 0.2936, + "step": 53149 + }, + { + "epoch": 19.63, + "learning_rate": 1.0379357063677794e-06, + "loss": 0.2576, + "step": 53150 + }, + { + "epoch": 19.63, + "learning_rate": 1.037898756004434e-06, + "loss": 0.2953, + "step": 53151 + }, + { + "epoch": 19.63, + "learning_rate": 1.0378618056410888e-06, + "loss": 0.1901, + "step": 53152 + }, + { + "epoch": 19.63, + "learning_rate": 1.0378248552777437e-06, + "loss": 0.1863, + "step": 53153 + }, + { + "epoch": 19.63, + "learning_rate": 1.0377879049143983e-06, + "loss": 0.3035, + "step": 53154 + }, + { + "epoch": 19.63, + "learning_rate": 1.0377509545510532e-06, + "loss": 0.2129, + "step": 53155 + }, + { + "epoch": 19.63, + "learning_rate": 1.037714004187708e-06, + "loss": 0.1583, + "step": 53156 + }, + { + "epoch": 19.63, + "learning_rate": 1.0376770538243627e-06, + "loss": 0.135, + "step": 53157 + }, + { + "epoch": 19.63, + "learning_rate": 1.0376401034610173e-06, + "loss": 0.2475, + "step": 53158 + }, + { + "epoch": 19.63, + "learning_rate": 1.0376031530976721e-06, + "loss": 0.2333, + "step": 53159 + }, + { + "epoch": 19.63, + "learning_rate": 1.037566202734327e-06, + "loss": 0.1, + "step": 53160 + }, + { + "epoch": 19.63, + "learning_rate": 1.0375292523709816e-06, + "loss": 0.2552, + "step": 53161 + }, + { + "epoch": 19.63, + "learning_rate": 1.0374923020076365e-06, + "loss": 0.2792, + "step": 53162 + }, + { + "epoch": 19.63, + "learning_rate": 1.0374553516442913e-06, + "loss": 0.2408, + "step": 53163 + }, + { + "epoch": 19.63, + "learning_rate": 1.037418401280946e-06, + "loss": 0.2561, + "step": 53164 + }, + { + "epoch": 19.63, + "learning_rate": 1.0373814509176008e-06, + "loss": 0.1157, + "step": 53165 + }, + { + "epoch": 19.63, + "learning_rate": 1.0373445005542557e-06, + "loss": 0.1591, + "step": 53166 + }, + { + "epoch": 19.63, + "learning_rate": 1.03730755019091e-06, + "loss": 0.1934, + "step": 53167 + }, + { + "epoch": 19.63, + "learning_rate": 1.037270599827565e-06, + "loss": 0.2921, + "step": 53168 + }, + { + "epoch": 19.63, + "learning_rate": 1.0372336494642198e-06, + "loss": 0.2523, + "step": 53169 + }, + { + "epoch": 19.63, + "learning_rate": 1.0371966991008744e-06, + "loss": 0.0812, + "step": 53170 + }, + { + "epoch": 19.63, + "learning_rate": 1.0371597487375293e-06, + "loss": 0.1643, + "step": 53171 + }, + { + "epoch": 19.64, + "learning_rate": 1.0371227983741841e-06, + "loss": 0.1413, + "step": 53172 + }, + { + "epoch": 19.64, + "learning_rate": 1.0370858480108387e-06, + "loss": 0.1913, + "step": 53173 + }, + { + "epoch": 19.64, + "learning_rate": 1.0370488976474936e-06, + "loss": 0.1873, + "step": 53174 + }, + { + "epoch": 19.64, + "learning_rate": 1.0370119472841484e-06, + "loss": 0.1632, + "step": 53175 + }, + { + "epoch": 19.64, + "learning_rate": 1.036974996920803e-06, + "loss": 0.2707, + "step": 53176 + }, + { + "epoch": 19.64, + "learning_rate": 1.0369380465574577e-06, + "loss": 0.2865, + "step": 53177 + }, + { + "epoch": 19.64, + "learning_rate": 1.0369010961941126e-06, + "loss": 0.2102, + "step": 53178 + }, + { + "epoch": 19.64, + "learning_rate": 1.0368641458307674e-06, + "loss": 0.2127, + "step": 53179 + }, + { + "epoch": 19.64, + "learning_rate": 1.036827195467422e-06, + "loss": 0.2305, + "step": 53180 + }, + { + "epoch": 19.64, + "learning_rate": 1.0367902451040769e-06, + "loss": 0.3758, + "step": 53181 + }, + { + "epoch": 19.64, + "learning_rate": 1.0367532947407317e-06, + "loss": 0.3051, + "step": 53182 + }, + { + "epoch": 19.64, + "learning_rate": 1.0367163443773864e-06, + "loss": 0.1739, + "step": 53183 + }, + { + "epoch": 19.64, + "learning_rate": 1.0366793940140412e-06, + "loss": 0.125, + "step": 53184 + }, + { + "epoch": 19.64, + "learning_rate": 1.036642443650696e-06, + "loss": 0.2084, + "step": 53185 + }, + { + "epoch": 19.64, + "learning_rate": 1.0366054932873505e-06, + "loss": 0.1729, + "step": 53186 + }, + { + "epoch": 19.64, + "learning_rate": 1.0365685429240053e-06, + "loss": 0.1323, + "step": 53187 + }, + { + "epoch": 19.64, + "learning_rate": 1.0365315925606602e-06, + "loss": 0.1719, + "step": 53188 + }, + { + "epoch": 19.64, + "learning_rate": 1.0364946421973148e-06, + "loss": 0.1773, + "step": 53189 + }, + { + "epoch": 19.64, + "learning_rate": 1.0364576918339697e-06, + "loss": 0.3146, + "step": 53190 + }, + { + "epoch": 19.64, + "learning_rate": 1.0364207414706245e-06, + "loss": 0.2313, + "step": 53191 + }, + { + "epoch": 19.64, + "learning_rate": 1.0363837911072792e-06, + "loss": 0.3436, + "step": 53192 + }, + { + "epoch": 19.64, + "learning_rate": 1.036346840743934e-06, + "loss": 0.1813, + "step": 53193 + }, + { + "epoch": 19.64, + "learning_rate": 1.0363098903805889e-06, + "loss": 0.1751, + "step": 53194 + }, + { + "epoch": 19.64, + "learning_rate": 1.0362729400172435e-06, + "loss": 0.1809, + "step": 53195 + }, + { + "epoch": 19.64, + "learning_rate": 1.0362359896538981e-06, + "loss": 0.1537, + "step": 53196 + }, + { + "epoch": 19.64, + "learning_rate": 1.036199039290553e-06, + "loss": 0.214, + "step": 53197 + }, + { + "epoch": 19.64, + "learning_rate": 1.0361620889272078e-06, + "loss": 0.1354, + "step": 53198 + }, + { + "epoch": 19.65, + "learning_rate": 1.0361251385638625e-06, + "loss": 0.2255, + "step": 53199 + }, + { + "epoch": 19.65, + "learning_rate": 1.0360881882005173e-06, + "loss": 0.1901, + "step": 53200 + }, + { + "epoch": 19.65, + "learning_rate": 1.0360512378371722e-06, + "loss": 0.0623, + "step": 53201 + }, + { + "epoch": 19.65, + "learning_rate": 1.0360142874738268e-06, + "loss": 0.1845, + "step": 53202 + }, + { + "epoch": 19.65, + "learning_rate": 1.0359773371104816e-06, + "loss": 0.165, + "step": 53203 + }, + { + "epoch": 19.65, + "learning_rate": 1.0359403867471365e-06, + "loss": 0.1833, + "step": 53204 + }, + { + "epoch": 19.65, + "learning_rate": 1.0359034363837911e-06, + "loss": 0.1852, + "step": 53205 + }, + { + "epoch": 19.65, + "learning_rate": 1.0358664860204458e-06, + "loss": 0.1612, + "step": 53206 + }, + { + "epoch": 19.65, + "learning_rate": 1.0358295356571006e-06, + "loss": 0.2508, + "step": 53207 + }, + { + "epoch": 19.65, + "learning_rate": 1.0357925852937555e-06, + "loss": 0.0972, + "step": 53208 + }, + { + "epoch": 19.65, + "learning_rate": 1.03575563493041e-06, + "loss": 0.2602, + "step": 53209 + }, + { + "epoch": 19.65, + "learning_rate": 1.035718684567065e-06, + "loss": 0.2833, + "step": 53210 + }, + { + "epoch": 19.65, + "learning_rate": 1.0356817342037198e-06, + "loss": 0.208, + "step": 53211 + }, + { + "epoch": 19.65, + "learning_rate": 1.0356447838403744e-06, + "loss": 0.2249, + "step": 53212 + }, + { + "epoch": 19.65, + "learning_rate": 1.0356078334770293e-06, + "loss": 0.1171, + "step": 53213 + }, + { + "epoch": 19.65, + "learning_rate": 1.0355708831136841e-06, + "loss": 0.1473, + "step": 53214 + }, + { + "epoch": 19.65, + "learning_rate": 1.0355339327503388e-06, + "loss": 0.1901, + "step": 53215 + }, + { + "epoch": 19.65, + "learning_rate": 1.0354969823869934e-06, + "loss": 0.3092, + "step": 53216 + }, + { + "epoch": 19.65, + "learning_rate": 1.0354600320236483e-06, + "loss": 0.1269, + "step": 53217 + }, + { + "epoch": 19.65, + "learning_rate": 1.0354230816603029e-06, + "loss": 0.3094, + "step": 53218 + }, + { + "epoch": 19.65, + "learning_rate": 1.0353861312969577e-06, + "loss": 0.3528, + "step": 53219 + }, + { + "epoch": 19.65, + "learning_rate": 1.0353491809336126e-06, + "loss": 0.2428, + "step": 53220 + }, + { + "epoch": 19.65, + "learning_rate": 1.0353122305702672e-06, + "loss": 0.2151, + "step": 53221 + }, + { + "epoch": 19.65, + "learning_rate": 1.035275280206922e-06, + "loss": 0.1549, + "step": 53222 + }, + { + "epoch": 19.65, + "learning_rate": 1.035238329843577e-06, + "loss": 0.213, + "step": 53223 + }, + { + "epoch": 19.65, + "learning_rate": 1.0352013794802316e-06, + "loss": 0.1606, + "step": 53224 + }, + { + "epoch": 19.65, + "learning_rate": 1.0351644291168864e-06, + "loss": 0.2764, + "step": 53225 + }, + { + "epoch": 19.66, + "learning_rate": 1.035127478753541e-06, + "loss": 0.1834, + "step": 53226 + }, + { + "epoch": 19.66, + "learning_rate": 1.0350905283901959e-06, + "loss": 0.1578, + "step": 53227 + }, + { + "epoch": 19.66, + "learning_rate": 1.0350535780268505e-06, + "loss": 0.259, + "step": 53228 + }, + { + "epoch": 19.66, + "learning_rate": 1.0350166276635054e-06, + "loss": 0.2821, + "step": 53229 + }, + { + "epoch": 19.66, + "learning_rate": 1.0349796773001602e-06, + "loss": 0.217, + "step": 53230 + }, + { + "epoch": 19.66, + "learning_rate": 1.0349427269368149e-06, + "loss": 0.2355, + "step": 53231 + }, + { + "epoch": 19.66, + "learning_rate": 1.0349057765734697e-06, + "loss": 0.2756, + "step": 53232 + }, + { + "epoch": 19.66, + "learning_rate": 1.0348688262101246e-06, + "loss": 0.2731, + "step": 53233 + }, + { + "epoch": 19.66, + "learning_rate": 1.0348318758467792e-06, + "loss": 0.2522, + "step": 53234 + }, + { + "epoch": 19.66, + "learning_rate": 1.034794925483434e-06, + "loss": 0.1783, + "step": 53235 + }, + { + "epoch": 19.66, + "learning_rate": 1.0347579751200887e-06, + "loss": 0.2828, + "step": 53236 + }, + { + "epoch": 19.66, + "learning_rate": 1.0347210247567433e-06, + "loss": 0.3213, + "step": 53237 + }, + { + "epoch": 19.66, + "learning_rate": 1.0346840743933982e-06, + "loss": 0.201, + "step": 53238 + }, + { + "epoch": 19.66, + "learning_rate": 1.034647124030053e-06, + "loss": 0.2357, + "step": 53239 + }, + { + "epoch": 19.66, + "learning_rate": 1.0346101736667076e-06, + "loss": 0.2828, + "step": 53240 + }, + { + "epoch": 19.66, + "learning_rate": 1.0345732233033625e-06, + "loss": 0.2552, + "step": 53241 + }, + { + "epoch": 19.66, + "learning_rate": 1.0345362729400173e-06, + "loss": 0.137, + "step": 53242 + }, + { + "epoch": 19.66, + "learning_rate": 1.034499322576672e-06, + "loss": 0.1724, + "step": 53243 + }, + { + "epoch": 19.66, + "learning_rate": 1.0344623722133268e-06, + "loss": 0.1928, + "step": 53244 + }, + { + "epoch": 19.66, + "learning_rate": 1.0344254218499817e-06, + "loss": 0.2404, + "step": 53245 + }, + { + "epoch": 19.66, + "learning_rate": 1.0343884714866363e-06, + "loss": 0.3221, + "step": 53246 + }, + { + "epoch": 19.66, + "learning_rate": 1.034351521123291e-06, + "loss": 0.1678, + "step": 53247 + }, + { + "epoch": 19.66, + "learning_rate": 1.0343145707599458e-06, + "loss": 0.2865, + "step": 53248 + }, + { + "epoch": 19.66, + "learning_rate": 1.0342776203966006e-06, + "loss": 0.2416, + "step": 53249 + }, + { + "epoch": 19.66, + "learning_rate": 1.0342406700332553e-06, + "loss": 0.2435, + "step": 53250 + }, + { + "epoch": 19.66, + "learning_rate": 1.0342037196699101e-06, + "loss": 0.124, + "step": 53251 + }, + { + "epoch": 19.66, + "learning_rate": 1.034166769306565e-06, + "loss": 0.2018, + "step": 53252 + }, + { + "epoch": 19.67, + "learning_rate": 1.0341298189432196e-06, + "loss": 0.2523, + "step": 53253 + }, + { + "epoch": 19.67, + "learning_rate": 1.0340928685798745e-06, + "loss": 0.2794, + "step": 53254 + }, + { + "epoch": 19.67, + "learning_rate": 1.0340559182165293e-06, + "loss": 0.1914, + "step": 53255 + }, + { + "epoch": 19.67, + "learning_rate": 1.034018967853184e-06, + "loss": 0.1847, + "step": 53256 + }, + { + "epoch": 19.67, + "learning_rate": 1.0339820174898386e-06, + "loss": 0.2018, + "step": 53257 + }, + { + "epoch": 19.67, + "learning_rate": 1.0339450671264934e-06, + "loss": 0.1654, + "step": 53258 + }, + { + "epoch": 19.67, + "learning_rate": 1.033908116763148e-06, + "loss": 0.245, + "step": 53259 + }, + { + "epoch": 19.67, + "learning_rate": 1.033871166399803e-06, + "loss": 0.2389, + "step": 53260 + }, + { + "epoch": 19.67, + "learning_rate": 1.0338342160364578e-06, + "loss": 0.2658, + "step": 53261 + }, + { + "epoch": 19.67, + "learning_rate": 1.0337972656731124e-06, + "loss": 0.2501, + "step": 53262 + }, + { + "epoch": 19.67, + "learning_rate": 1.0337603153097672e-06, + "loss": 0.2473, + "step": 53263 + }, + { + "epoch": 19.67, + "learning_rate": 1.033723364946422e-06, + "loss": 0.2402, + "step": 53264 + }, + { + "epoch": 19.67, + "learning_rate": 1.0336864145830767e-06, + "loss": 0.2552, + "step": 53265 + }, + { + "epoch": 19.67, + "learning_rate": 1.0336494642197314e-06, + "loss": 0.3227, + "step": 53266 + }, + { + "epoch": 19.67, + "learning_rate": 1.0336125138563862e-06, + "loss": 0.1801, + "step": 53267 + }, + { + "epoch": 19.67, + "learning_rate": 1.033575563493041e-06, + "loss": 0.2895, + "step": 53268 + }, + { + "epoch": 19.67, + "learning_rate": 1.0335386131296957e-06, + "loss": 0.1245, + "step": 53269 + }, + { + "epoch": 19.67, + "learning_rate": 1.0335016627663505e-06, + "loss": 0.1592, + "step": 53270 + }, + { + "epoch": 19.67, + "learning_rate": 1.0334647124030054e-06, + "loss": 0.4119, + "step": 53271 + }, + { + "epoch": 19.67, + "learning_rate": 1.03342776203966e-06, + "loss": 0.2096, + "step": 53272 + }, + { + "epoch": 19.67, + "learning_rate": 1.0333908116763149e-06, + "loss": 0.1833, + "step": 53273 + }, + { + "epoch": 19.67, + "learning_rate": 1.0333538613129697e-06, + "loss": 0.1927, + "step": 53274 + }, + { + "epoch": 19.67, + "learning_rate": 1.0333169109496244e-06, + "loss": 0.1375, + "step": 53275 + }, + { + "epoch": 19.67, + "learning_rate": 1.033279960586279e-06, + "loss": 0.1925, + "step": 53276 + }, + { + "epoch": 19.67, + "learning_rate": 1.0332430102229339e-06, + "loss": 0.2038, + "step": 53277 + }, + { + "epoch": 19.67, + "learning_rate": 1.0332060598595887e-06, + "loss": 0.3109, + "step": 53278 + }, + { + "epoch": 19.67, + "learning_rate": 1.0331691094962433e-06, + "loss": 0.1847, + "step": 53279 + }, + { + "epoch": 19.68, + "learning_rate": 1.0331321591328982e-06, + "loss": 0.1872, + "step": 53280 + }, + { + "epoch": 19.68, + "learning_rate": 1.033095208769553e-06, + "loss": 0.2241, + "step": 53281 + }, + { + "epoch": 19.68, + "learning_rate": 1.0330582584062077e-06, + "loss": 0.2228, + "step": 53282 + }, + { + "epoch": 19.68, + "learning_rate": 1.0330213080428625e-06, + "loss": 0.1974, + "step": 53283 + }, + { + "epoch": 19.68, + "learning_rate": 1.0329843576795174e-06, + "loss": 0.1778, + "step": 53284 + }, + { + "epoch": 19.68, + "learning_rate": 1.032947407316172e-06, + "loss": 0.2299, + "step": 53285 + }, + { + "epoch": 19.68, + "learning_rate": 1.0329104569528266e-06, + "loss": 0.2405, + "step": 53286 + }, + { + "epoch": 19.68, + "learning_rate": 1.0328735065894815e-06, + "loss": 0.312, + "step": 53287 + }, + { + "epoch": 19.68, + "learning_rate": 1.0328365562261361e-06, + "loss": 0.1615, + "step": 53288 + }, + { + "epoch": 19.68, + "learning_rate": 1.032799605862791e-06, + "loss": 0.1183, + "step": 53289 + }, + { + "epoch": 19.68, + "learning_rate": 1.0327626554994458e-06, + "loss": 0.1941, + "step": 53290 + }, + { + "epoch": 19.68, + "learning_rate": 1.0327257051361005e-06, + "loss": 0.1356, + "step": 53291 + }, + { + "epoch": 19.68, + "learning_rate": 1.0326887547727553e-06, + "loss": 0.2113, + "step": 53292 + }, + { + "epoch": 19.68, + "learning_rate": 1.0326518044094102e-06, + "loss": 0.2577, + "step": 53293 + }, + { + "epoch": 19.68, + "learning_rate": 1.0326148540460648e-06, + "loss": 0.2116, + "step": 53294 + }, + { + "epoch": 19.68, + "learning_rate": 1.0325779036827196e-06, + "loss": 0.2896, + "step": 53295 + }, + { + "epoch": 19.68, + "learning_rate": 1.0325409533193743e-06, + "loss": 0.1453, + "step": 53296 + }, + { + "epoch": 19.68, + "learning_rate": 1.0325040029560291e-06, + "loss": 0.2407, + "step": 53297 + }, + { + "epoch": 19.68, + "learning_rate": 1.0324670525926838e-06, + "loss": 0.1971, + "step": 53298 + }, + { + "epoch": 19.68, + "learning_rate": 1.0324301022293386e-06, + "loss": 0.2784, + "step": 53299 + }, + { + "epoch": 19.68, + "learning_rate": 1.0323931518659935e-06, + "loss": 0.3476, + "step": 53300 + }, + { + "epoch": 19.68, + "learning_rate": 1.032356201502648e-06, + "loss": 0.2373, + "step": 53301 + }, + { + "epoch": 19.68, + "learning_rate": 1.032319251139303e-06, + "loss": 0.0745, + "step": 53302 + }, + { + "epoch": 19.68, + "learning_rate": 1.0322823007759578e-06, + "loss": 0.1037, + "step": 53303 + }, + { + "epoch": 19.68, + "learning_rate": 1.0322453504126124e-06, + "loss": 0.2458, + "step": 53304 + }, + { + "epoch": 19.68, + "learning_rate": 1.0322084000492673e-06, + "loss": 0.1488, + "step": 53305 + }, + { + "epoch": 19.68, + "learning_rate": 1.032171449685922e-06, + "loss": 0.2468, + "step": 53306 + }, + { + "epoch": 19.69, + "learning_rate": 1.0321344993225765e-06, + "loss": 0.109, + "step": 53307 + }, + { + "epoch": 19.69, + "learning_rate": 1.0320975489592314e-06, + "loss": 0.292, + "step": 53308 + }, + { + "epoch": 19.69, + "learning_rate": 1.0320605985958862e-06, + "loss": 0.268, + "step": 53309 + }, + { + "epoch": 19.69, + "learning_rate": 1.0320236482325409e-06, + "loss": 0.3149, + "step": 53310 + }, + { + "epoch": 19.69, + "learning_rate": 1.0319866978691957e-06, + "loss": 0.2261, + "step": 53311 + }, + { + "epoch": 19.69, + "learning_rate": 1.0319497475058506e-06, + "loss": 0.1916, + "step": 53312 + }, + { + "epoch": 19.69, + "learning_rate": 1.0319127971425052e-06, + "loss": 0.3417, + "step": 53313 + }, + { + "epoch": 19.69, + "learning_rate": 1.03187584677916e-06, + "loss": 0.2928, + "step": 53314 + }, + { + "epoch": 19.69, + "learning_rate": 1.031838896415815e-06, + "loss": 0.1281, + "step": 53315 + }, + { + "epoch": 19.69, + "learning_rate": 1.0318019460524695e-06, + "loss": 0.2923, + "step": 53316 + }, + { + "epoch": 19.69, + "learning_rate": 1.0317649956891242e-06, + "loss": 0.1372, + "step": 53317 + }, + { + "epoch": 19.69, + "learning_rate": 1.031728045325779e-06, + "loss": 0.2534, + "step": 53318 + }, + { + "epoch": 19.69, + "learning_rate": 1.0316910949624339e-06, + "loss": 0.2469, + "step": 53319 + }, + { + "epoch": 19.69, + "learning_rate": 1.0316541445990885e-06, + "loss": 0.2786, + "step": 53320 + }, + { + "epoch": 19.69, + "learning_rate": 1.0316171942357434e-06, + "loss": 0.1453, + "step": 53321 + }, + { + "epoch": 19.69, + "learning_rate": 1.0315802438723982e-06, + "loss": 0.2243, + "step": 53322 + }, + { + "epoch": 19.69, + "learning_rate": 1.0315432935090528e-06, + "loss": 0.1971, + "step": 53323 + }, + { + "epoch": 19.69, + "learning_rate": 1.0315063431457077e-06, + "loss": 0.2495, + "step": 53324 + }, + { + "epoch": 19.69, + "learning_rate": 1.0314693927823625e-06, + "loss": 0.1831, + "step": 53325 + }, + { + "epoch": 19.69, + "learning_rate": 1.0314324424190172e-06, + "loss": 0.2137, + "step": 53326 + }, + { + "epoch": 19.69, + "learning_rate": 1.0313954920556718e-06, + "loss": 0.2556, + "step": 53327 + }, + { + "epoch": 19.69, + "learning_rate": 1.0313585416923267e-06, + "loss": 0.1194, + "step": 53328 + }, + { + "epoch": 19.69, + "learning_rate": 1.0313215913289815e-06, + "loss": 0.1773, + "step": 53329 + }, + { + "epoch": 19.69, + "learning_rate": 1.0312846409656361e-06, + "loss": 0.1477, + "step": 53330 + }, + { + "epoch": 19.69, + "learning_rate": 1.031247690602291e-06, + "loss": 0.1315, + "step": 53331 + }, + { + "epoch": 19.69, + "learning_rate": 1.0312107402389458e-06, + "loss": 0.3044, + "step": 53332 + }, + { + "epoch": 19.69, + "learning_rate": 1.0311737898756005e-06, + "loss": 0.1574, + "step": 53333 + }, + { + "epoch": 19.69, + "learning_rate": 1.0311368395122553e-06, + "loss": 0.1515, + "step": 53334 + }, + { + "epoch": 19.7, + "learning_rate": 1.03109988914891e-06, + "loss": 0.1829, + "step": 53335 + }, + { + "epoch": 19.7, + "learning_rate": 1.0310629387855646e-06, + "loss": 0.2984, + "step": 53336 + }, + { + "epoch": 19.7, + "learning_rate": 1.0310259884222194e-06, + "loss": 0.1692, + "step": 53337 + }, + { + "epoch": 19.7, + "learning_rate": 1.0309890380588743e-06, + "loss": 0.1502, + "step": 53338 + }, + { + "epoch": 19.7, + "learning_rate": 1.030952087695529e-06, + "loss": 0.1957, + "step": 53339 + }, + { + "epoch": 19.7, + "learning_rate": 1.0309151373321838e-06, + "loss": 0.2346, + "step": 53340 + }, + { + "epoch": 19.7, + "learning_rate": 1.0308781869688386e-06, + "loss": 0.3329, + "step": 53341 + }, + { + "epoch": 19.7, + "learning_rate": 1.0308412366054933e-06, + "loss": 0.3073, + "step": 53342 + }, + { + "epoch": 19.7, + "learning_rate": 1.0308042862421481e-06, + "loss": 0.2046, + "step": 53343 + }, + { + "epoch": 19.7, + "learning_rate": 1.030767335878803e-06, + "loss": 0.2071, + "step": 53344 + }, + { + "epoch": 19.7, + "learning_rate": 1.0307303855154576e-06, + "loss": 0.2109, + "step": 53345 + }, + { + "epoch": 19.7, + "learning_rate": 1.0306934351521122e-06, + "loss": 0.2129, + "step": 53346 + }, + { + "epoch": 19.7, + "learning_rate": 1.030656484788767e-06, + "loss": 0.2881, + "step": 53347 + }, + { + "epoch": 19.7, + "learning_rate": 1.030619534425422e-06, + "loss": 0.1458, + "step": 53348 + }, + { + "epoch": 19.7, + "learning_rate": 1.0305825840620766e-06, + "loss": 0.1519, + "step": 53349 + }, + { + "epoch": 19.7, + "learning_rate": 1.0305456336987314e-06, + "loss": 0.212, + "step": 53350 + }, + { + "epoch": 19.7, + "learning_rate": 1.0305086833353863e-06, + "loss": 0.3081, + "step": 53351 + }, + { + "epoch": 19.7, + "learning_rate": 1.030471732972041e-06, + "loss": 0.2487, + "step": 53352 + }, + { + "epoch": 19.7, + "learning_rate": 1.0304347826086957e-06, + "loss": 0.242, + "step": 53353 + }, + { + "epoch": 19.7, + "learning_rate": 1.0303978322453506e-06, + "loss": 0.2835, + "step": 53354 + }, + { + "epoch": 19.7, + "learning_rate": 1.0303608818820052e-06, + "loss": 0.2149, + "step": 53355 + }, + { + "epoch": 19.7, + "learning_rate": 1.0303239315186599e-06, + "loss": 0.2164, + "step": 53356 + }, + { + "epoch": 19.7, + "learning_rate": 1.0302869811553147e-06, + "loss": 0.2027, + "step": 53357 + }, + { + "epoch": 19.7, + "learning_rate": 1.0302500307919694e-06, + "loss": 0.1374, + "step": 53358 + }, + { + "epoch": 19.7, + "learning_rate": 1.0302130804286242e-06, + "loss": 0.2678, + "step": 53359 + }, + { + "epoch": 19.7, + "learning_rate": 1.030176130065279e-06, + "loss": 0.2459, + "step": 53360 + }, + { + "epoch": 19.7, + "learning_rate": 1.0301391797019337e-06, + "loss": 0.2347, + "step": 53361 + }, + { + "epoch": 19.71, + "learning_rate": 1.0301022293385885e-06, + "loss": 0.2232, + "step": 53362 + }, + { + "epoch": 19.71, + "learning_rate": 1.0300652789752434e-06, + "loss": 0.2428, + "step": 53363 + }, + { + "epoch": 19.71, + "learning_rate": 1.030028328611898e-06, + "loss": 0.2839, + "step": 53364 + }, + { + "epoch": 19.71, + "learning_rate": 1.0299913782485529e-06, + "loss": 0.2482, + "step": 53365 + }, + { + "epoch": 19.71, + "learning_rate": 1.0299544278852075e-06, + "loss": 0.1884, + "step": 53366 + }, + { + "epoch": 19.71, + "learning_rate": 1.0299174775218624e-06, + "loss": 0.1835, + "step": 53367 + }, + { + "epoch": 19.71, + "learning_rate": 1.029880527158517e-06, + "loss": 0.1935, + "step": 53368 + }, + { + "epoch": 19.71, + "learning_rate": 1.0298435767951718e-06, + "loss": 0.1702, + "step": 53369 + }, + { + "epoch": 19.71, + "learning_rate": 1.0298066264318267e-06, + "loss": 0.2032, + "step": 53370 + }, + { + "epoch": 19.71, + "learning_rate": 1.0297696760684813e-06, + "loss": 0.2163, + "step": 53371 + }, + { + "epoch": 19.71, + "learning_rate": 1.0297327257051362e-06, + "loss": 0.2282, + "step": 53372 + }, + { + "epoch": 19.71, + "learning_rate": 1.029695775341791e-06, + "loss": 0.2374, + "step": 53373 + }, + { + "epoch": 19.71, + "learning_rate": 1.0296588249784457e-06, + "loss": 0.3429, + "step": 53374 + }, + { + "epoch": 19.71, + "learning_rate": 1.0296218746151005e-06, + "loss": 0.2353, + "step": 53375 + }, + { + "epoch": 19.71, + "learning_rate": 1.0295849242517551e-06, + "loss": 0.2016, + "step": 53376 + }, + { + "epoch": 19.71, + "learning_rate": 1.0295479738884098e-06, + "loss": 0.2638, + "step": 53377 + }, + { + "epoch": 19.71, + "learning_rate": 1.0295110235250646e-06, + "loss": 0.204, + "step": 53378 + }, + { + "epoch": 19.71, + "learning_rate": 1.0294740731617195e-06, + "loss": 0.2868, + "step": 53379 + }, + { + "epoch": 19.71, + "learning_rate": 1.0294371227983741e-06, + "loss": 0.1815, + "step": 53380 + }, + { + "epoch": 19.71, + "learning_rate": 1.029400172435029e-06, + "loss": 0.2103, + "step": 53381 + }, + { + "epoch": 19.71, + "learning_rate": 1.0293632220716838e-06, + "loss": 0.159, + "step": 53382 + }, + { + "epoch": 19.71, + "learning_rate": 1.0293262717083384e-06, + "loss": 0.1727, + "step": 53383 + }, + { + "epoch": 19.71, + "learning_rate": 1.0292893213449933e-06, + "loss": 0.129, + "step": 53384 + }, + { + "epoch": 19.71, + "learning_rate": 1.0292523709816481e-06, + "loss": 0.1846, + "step": 53385 + }, + { + "epoch": 19.71, + "learning_rate": 1.0292154206183028e-06, + "loss": 0.2392, + "step": 53386 + }, + { + "epoch": 19.71, + "learning_rate": 1.0291784702549574e-06, + "loss": 0.2841, + "step": 53387 + }, + { + "epoch": 19.71, + "learning_rate": 1.0291415198916123e-06, + "loss": 0.2318, + "step": 53388 + }, + { + "epoch": 19.72, + "learning_rate": 1.0291045695282671e-06, + "loss": 0.211, + "step": 53389 + }, + { + "epoch": 19.72, + "learning_rate": 1.0290676191649217e-06, + "loss": 0.3847, + "step": 53390 + }, + { + "epoch": 19.72, + "learning_rate": 1.0290306688015766e-06, + "loss": 0.1453, + "step": 53391 + }, + { + "epoch": 19.72, + "learning_rate": 1.0289937184382314e-06, + "loss": 0.2329, + "step": 53392 + }, + { + "epoch": 19.72, + "learning_rate": 1.028956768074886e-06, + "loss": 0.23, + "step": 53393 + }, + { + "epoch": 19.72, + "learning_rate": 1.028919817711541e-06, + "loss": 0.162, + "step": 53394 + }, + { + "epoch": 19.72, + "learning_rate": 1.0288828673481958e-06, + "loss": 0.2438, + "step": 53395 + }, + { + "epoch": 19.72, + "learning_rate": 1.0288459169848504e-06, + "loss": 0.2154, + "step": 53396 + }, + { + "epoch": 19.72, + "learning_rate": 1.028808966621505e-06, + "loss": 0.1786, + "step": 53397 + }, + { + "epoch": 19.72, + "learning_rate": 1.0287720162581599e-06, + "loss": 0.0983, + "step": 53398 + }, + { + "epoch": 19.72, + "learning_rate": 1.0287350658948147e-06, + "loss": 0.2075, + "step": 53399 + }, + { + "epoch": 19.72, + "learning_rate": 1.0286981155314694e-06, + "loss": 0.1235, + "step": 53400 + }, + { + "epoch": 19.72, + "learning_rate": 1.0286611651681242e-06, + "loss": 0.1462, + "step": 53401 + }, + { + "epoch": 19.72, + "learning_rate": 1.028624214804779e-06, + "loss": 0.1509, + "step": 53402 + }, + { + "epoch": 19.72, + "learning_rate": 1.0285872644414337e-06, + "loss": 0.3088, + "step": 53403 + }, + { + "epoch": 19.72, + "learning_rate": 1.0285503140780886e-06, + "loss": 0.285, + "step": 53404 + }, + { + "epoch": 19.72, + "learning_rate": 1.0285133637147432e-06, + "loss": 0.1829, + "step": 53405 + }, + { + "epoch": 19.72, + "learning_rate": 1.0284764133513978e-06, + "loss": 0.1304, + "step": 53406 + }, + { + "epoch": 19.72, + "learning_rate": 1.0284394629880527e-06, + "loss": 0.2491, + "step": 53407 + }, + { + "epoch": 19.72, + "learning_rate": 1.0284025126247075e-06, + "loss": 0.1472, + "step": 53408 + }, + { + "epoch": 19.72, + "learning_rate": 1.0283655622613622e-06, + "loss": 0.342, + "step": 53409 + }, + { + "epoch": 19.72, + "learning_rate": 1.028328611898017e-06, + "loss": 0.0813, + "step": 53410 + }, + { + "epoch": 19.72, + "learning_rate": 1.0282916615346719e-06, + "loss": 0.257, + "step": 53411 + }, + { + "epoch": 19.72, + "learning_rate": 1.0282547111713265e-06, + "loss": 0.1422, + "step": 53412 + }, + { + "epoch": 19.72, + "learning_rate": 1.0282177608079813e-06, + "loss": 0.1362, + "step": 53413 + }, + { + "epoch": 19.72, + "learning_rate": 1.0281808104446362e-06, + "loss": 0.1427, + "step": 53414 + }, + { + "epoch": 19.72, + "learning_rate": 1.0281438600812908e-06, + "loss": 0.248, + "step": 53415 + }, + { + "epoch": 19.73, + "learning_rate": 1.0281069097179455e-06, + "loss": 0.1308, + "step": 53416 + }, + { + "epoch": 19.73, + "learning_rate": 1.0280699593546003e-06, + "loss": 0.2069, + "step": 53417 + }, + { + "epoch": 19.73, + "learning_rate": 1.0280330089912552e-06, + "loss": 0.2432, + "step": 53418 + }, + { + "epoch": 19.73, + "learning_rate": 1.0279960586279098e-06, + "loss": 0.2488, + "step": 53419 + }, + { + "epoch": 19.73, + "learning_rate": 1.0279591082645646e-06, + "loss": 0.1452, + "step": 53420 + }, + { + "epoch": 19.73, + "learning_rate": 1.0279221579012195e-06, + "loss": 0.175, + "step": 53421 + }, + { + "epoch": 19.73, + "learning_rate": 1.0278852075378741e-06, + "loss": 0.3391, + "step": 53422 + }, + { + "epoch": 19.73, + "learning_rate": 1.027848257174529e-06, + "loss": 0.266, + "step": 53423 + }, + { + "epoch": 19.73, + "learning_rate": 1.0278113068111838e-06, + "loss": 0.1785, + "step": 53424 + }, + { + "epoch": 19.73, + "learning_rate": 1.0277743564478383e-06, + "loss": 0.2531, + "step": 53425 + }, + { + "epoch": 19.73, + "learning_rate": 1.027737406084493e-06, + "loss": 0.1981, + "step": 53426 + }, + { + "epoch": 19.73, + "learning_rate": 1.027700455721148e-06, + "loss": 0.2452, + "step": 53427 + }, + { + "epoch": 19.73, + "learning_rate": 1.0276635053578026e-06, + "loss": 0.2024, + "step": 53428 + }, + { + "epoch": 19.73, + "learning_rate": 1.0276265549944574e-06, + "loss": 0.1239, + "step": 53429 + }, + { + "epoch": 19.73, + "learning_rate": 1.0275896046311123e-06, + "loss": 0.4053, + "step": 53430 + }, + { + "epoch": 19.73, + "learning_rate": 1.027552654267767e-06, + "loss": 0.2354, + "step": 53431 + }, + { + "epoch": 19.73, + "learning_rate": 1.0275157039044218e-06, + "loss": 0.1721, + "step": 53432 + }, + { + "epoch": 19.73, + "learning_rate": 1.0274787535410766e-06, + "loss": 0.2296, + "step": 53433 + }, + { + "epoch": 19.73, + "learning_rate": 1.0274418031777313e-06, + "loss": 0.2005, + "step": 53434 + }, + { + "epoch": 19.73, + "learning_rate": 1.0274048528143859e-06, + "loss": 0.2801, + "step": 53435 + }, + { + "epoch": 19.73, + "learning_rate": 1.0273679024510407e-06, + "loss": 0.1448, + "step": 53436 + }, + { + "epoch": 19.73, + "learning_rate": 1.0273309520876956e-06, + "loss": 0.191, + "step": 53437 + }, + { + "epoch": 19.73, + "learning_rate": 1.0272940017243502e-06, + "loss": 0.2199, + "step": 53438 + }, + { + "epoch": 19.73, + "learning_rate": 1.027257051361005e-06, + "loss": 0.2601, + "step": 53439 + }, + { + "epoch": 19.73, + "learning_rate": 1.02722010099766e-06, + "loss": 0.2977, + "step": 53440 + }, + { + "epoch": 19.73, + "learning_rate": 1.0271831506343146e-06, + "loss": 0.1921, + "step": 53441 + }, + { + "epoch": 19.73, + "learning_rate": 1.0271462002709694e-06, + "loss": 0.2476, + "step": 53442 + }, + { + "epoch": 19.74, + "learning_rate": 1.0271092499076243e-06, + "loss": 0.2037, + "step": 53443 + }, + { + "epoch": 19.74, + "learning_rate": 1.0270722995442789e-06, + "loss": 0.1409, + "step": 53444 + }, + { + "epoch": 19.74, + "learning_rate": 1.0270353491809335e-06, + "loss": 0.0904, + "step": 53445 + }, + { + "epoch": 19.74, + "learning_rate": 1.0269983988175884e-06, + "loss": 0.1822, + "step": 53446 + }, + { + "epoch": 19.74, + "learning_rate": 1.026961448454243e-06, + "loss": 0.2234, + "step": 53447 + }, + { + "epoch": 19.74, + "learning_rate": 1.0269244980908979e-06, + "loss": 0.1551, + "step": 53448 + }, + { + "epoch": 19.74, + "learning_rate": 1.0268875477275527e-06, + "loss": 0.2375, + "step": 53449 + }, + { + "epoch": 19.74, + "learning_rate": 1.0268505973642073e-06, + "loss": 0.2104, + "step": 53450 + }, + { + "epoch": 19.74, + "learning_rate": 1.0268136470008622e-06, + "loss": 0.163, + "step": 53451 + }, + { + "epoch": 19.74, + "learning_rate": 1.026776696637517e-06, + "loss": 0.1702, + "step": 53452 + }, + { + "epoch": 19.74, + "learning_rate": 1.0267397462741717e-06, + "loss": 0.1331, + "step": 53453 + }, + { + "epoch": 19.74, + "learning_rate": 1.0267027959108265e-06, + "loss": 0.1315, + "step": 53454 + }, + { + "epoch": 19.74, + "learning_rate": 1.0266658455474812e-06, + "loss": 0.2505, + "step": 53455 + }, + { + "epoch": 19.74, + "learning_rate": 1.026628895184136e-06, + "loss": 0.2259, + "step": 53456 + }, + { + "epoch": 19.74, + "learning_rate": 1.0265919448207906e-06, + "loss": 0.2524, + "step": 53457 + }, + { + "epoch": 19.74, + "learning_rate": 1.0265549944574455e-06, + "loss": 0.0944, + "step": 53458 + }, + { + "epoch": 19.74, + "learning_rate": 1.0265180440941003e-06, + "loss": 0.211, + "step": 53459 + }, + { + "epoch": 19.74, + "learning_rate": 1.026481093730755e-06, + "loss": 0.1331, + "step": 53460 + }, + { + "epoch": 19.74, + "learning_rate": 1.0264441433674098e-06, + "loss": 0.2991, + "step": 53461 + }, + { + "epoch": 19.74, + "learning_rate": 1.0264071930040647e-06, + "loss": 0.18, + "step": 53462 + }, + { + "epoch": 19.74, + "learning_rate": 1.0263702426407193e-06, + "loss": 0.3278, + "step": 53463 + }, + { + "epoch": 19.74, + "learning_rate": 1.0263332922773742e-06, + "loss": 0.1836, + "step": 53464 + }, + { + "epoch": 19.74, + "learning_rate": 1.0262963419140288e-06, + "loss": 0.265, + "step": 53465 + }, + { + "epoch": 19.74, + "learning_rate": 1.0262593915506836e-06, + "loss": 0.2825, + "step": 53466 + }, + { + "epoch": 19.74, + "learning_rate": 1.0262224411873383e-06, + "loss": 0.139, + "step": 53467 + }, + { + "epoch": 19.74, + "learning_rate": 1.0261854908239931e-06, + "loss": 0.2915, + "step": 53468 + }, + { + "epoch": 19.74, + "learning_rate": 1.026148540460648e-06, + "loss": 0.2981, + "step": 53469 + }, + { + "epoch": 19.75, + "learning_rate": 1.0261115900973026e-06, + "loss": 0.2147, + "step": 53470 + }, + { + "epoch": 19.75, + "learning_rate": 1.0260746397339575e-06, + "loss": 0.1188, + "step": 53471 + }, + { + "epoch": 19.75, + "learning_rate": 1.0260376893706123e-06, + "loss": 0.1926, + "step": 53472 + }, + { + "epoch": 19.75, + "learning_rate": 1.026000739007267e-06, + "loss": 0.2236, + "step": 53473 + }, + { + "epoch": 19.75, + "learning_rate": 1.0259637886439218e-06, + "loss": 0.4009, + "step": 53474 + }, + { + "epoch": 19.75, + "learning_rate": 1.0259268382805764e-06, + "loss": 0.2386, + "step": 53475 + }, + { + "epoch": 19.75, + "learning_rate": 1.025889887917231e-06, + "loss": 0.2989, + "step": 53476 + }, + { + "epoch": 19.75, + "learning_rate": 1.025852937553886e-06, + "loss": 0.2935, + "step": 53477 + }, + { + "epoch": 19.75, + "learning_rate": 1.0258159871905408e-06, + "loss": 0.1499, + "step": 53478 + }, + { + "epoch": 19.75, + "learning_rate": 1.0257790368271954e-06, + "loss": 0.1413, + "step": 53479 + }, + { + "epoch": 19.75, + "learning_rate": 1.0257420864638502e-06, + "loss": 0.1822, + "step": 53480 + }, + { + "epoch": 19.75, + "learning_rate": 1.025705136100505e-06, + "loss": 0.2096, + "step": 53481 + }, + { + "epoch": 19.75, + "learning_rate": 1.0256681857371597e-06, + "loss": 0.2676, + "step": 53482 + }, + { + "epoch": 19.75, + "learning_rate": 1.0256312353738146e-06, + "loss": 0.1994, + "step": 53483 + }, + { + "epoch": 19.75, + "learning_rate": 1.0255942850104694e-06, + "loss": 0.2023, + "step": 53484 + }, + { + "epoch": 19.75, + "learning_rate": 1.025557334647124e-06, + "loss": 0.2343, + "step": 53485 + }, + { + "epoch": 19.75, + "learning_rate": 1.0255203842837787e-06, + "loss": 0.2362, + "step": 53486 + }, + { + "epoch": 19.75, + "learning_rate": 1.0254834339204335e-06, + "loss": 0.194, + "step": 53487 + }, + { + "epoch": 19.75, + "learning_rate": 1.0254464835570884e-06, + "loss": 0.2921, + "step": 53488 + }, + { + "epoch": 19.75, + "learning_rate": 1.025409533193743e-06, + "loss": 0.1876, + "step": 53489 + }, + { + "epoch": 19.75, + "learning_rate": 1.0253725828303979e-06, + "loss": 0.1189, + "step": 53490 + }, + { + "epoch": 19.75, + "learning_rate": 1.0253356324670527e-06, + "loss": 0.2186, + "step": 53491 + }, + { + "epoch": 19.75, + "learning_rate": 1.0252986821037074e-06, + "loss": 0.2786, + "step": 53492 + }, + { + "epoch": 19.75, + "learning_rate": 1.0252617317403622e-06, + "loss": 0.2877, + "step": 53493 + }, + { + "epoch": 19.75, + "learning_rate": 1.025224781377017e-06, + "loss": 0.1751, + "step": 53494 + }, + { + "epoch": 19.75, + "learning_rate": 1.0251878310136715e-06, + "loss": 0.2884, + "step": 53495 + }, + { + "epoch": 19.75, + "learning_rate": 1.0251508806503263e-06, + "loss": 0.3259, + "step": 53496 + }, + { + "epoch": 19.76, + "learning_rate": 1.0251139302869812e-06, + "loss": 0.2193, + "step": 53497 + }, + { + "epoch": 19.76, + "learning_rate": 1.0250769799236358e-06, + "loss": 0.3235, + "step": 53498 + }, + { + "epoch": 19.76, + "learning_rate": 1.0250400295602907e-06, + "loss": 0.108, + "step": 53499 + }, + { + "epoch": 19.76, + "learning_rate": 1.0250030791969455e-06, + "loss": 0.2107, + "step": 53500 + }, + { + "epoch": 19.76, + "learning_rate": 1.0249661288336002e-06, + "loss": 0.2495, + "step": 53501 + }, + { + "epoch": 19.76, + "learning_rate": 1.024929178470255e-06, + "loss": 0.183, + "step": 53502 + }, + { + "epoch": 19.76, + "learning_rate": 1.0248922281069098e-06, + "loss": 0.1933, + "step": 53503 + }, + { + "epoch": 19.76, + "learning_rate": 1.0248552777435645e-06, + "loss": 0.1962, + "step": 53504 + }, + { + "epoch": 19.76, + "learning_rate": 1.0248183273802191e-06, + "loss": 0.207, + "step": 53505 + }, + { + "epoch": 19.76, + "learning_rate": 1.024781377016874e-06, + "loss": 0.1953, + "step": 53506 + }, + { + "epoch": 19.76, + "learning_rate": 1.0247444266535288e-06, + "loss": 0.1439, + "step": 53507 + }, + { + "epoch": 19.76, + "learning_rate": 1.0247074762901835e-06, + "loss": 0.1454, + "step": 53508 + }, + { + "epoch": 19.76, + "learning_rate": 1.0246705259268383e-06, + "loss": 0.0963, + "step": 53509 + }, + { + "epoch": 19.76, + "learning_rate": 1.0246335755634932e-06, + "loss": 0.2871, + "step": 53510 + }, + { + "epoch": 19.76, + "learning_rate": 1.0245966252001478e-06, + "loss": 0.1785, + "step": 53511 + }, + { + "epoch": 19.76, + "learning_rate": 1.0245596748368026e-06, + "loss": 0.3182, + "step": 53512 + }, + { + "epoch": 19.76, + "learning_rate": 1.0245227244734575e-06, + "loss": 0.2198, + "step": 53513 + }, + { + "epoch": 19.76, + "learning_rate": 1.0244857741101121e-06, + "loss": 0.2176, + "step": 53514 + }, + { + "epoch": 19.76, + "learning_rate": 1.0244488237467668e-06, + "loss": 0.1566, + "step": 53515 + }, + { + "epoch": 19.76, + "learning_rate": 1.0244118733834216e-06, + "loss": 0.1928, + "step": 53516 + }, + { + "epoch": 19.76, + "learning_rate": 1.0243749230200765e-06, + "loss": 0.2158, + "step": 53517 + }, + { + "epoch": 19.76, + "learning_rate": 1.024337972656731e-06, + "loss": 0.1184, + "step": 53518 + }, + { + "epoch": 19.76, + "learning_rate": 1.024301022293386e-06, + "loss": 0.2954, + "step": 53519 + }, + { + "epoch": 19.76, + "learning_rate": 1.0242640719300408e-06, + "loss": 0.2229, + "step": 53520 + }, + { + "epoch": 19.76, + "learning_rate": 1.0242271215666954e-06, + "loss": 0.3443, + "step": 53521 + }, + { + "epoch": 19.76, + "learning_rate": 1.0241901712033503e-06, + "loss": 0.1958, + "step": 53522 + }, + { + "epoch": 19.76, + "learning_rate": 1.024153220840005e-06, + "loss": 0.1934, + "step": 53523 + }, + { + "epoch": 19.77, + "learning_rate": 1.0241162704766598e-06, + "loss": 0.1565, + "step": 53524 + }, + { + "epoch": 19.77, + "learning_rate": 1.0240793201133144e-06, + "loss": 0.2738, + "step": 53525 + }, + { + "epoch": 19.77, + "learning_rate": 1.0240423697499692e-06, + "loss": 0.1862, + "step": 53526 + }, + { + "epoch": 19.77, + "learning_rate": 1.0240054193866239e-06, + "loss": 0.2538, + "step": 53527 + }, + { + "epoch": 19.77, + "learning_rate": 1.0239684690232787e-06, + "loss": 0.3342, + "step": 53528 + }, + { + "epoch": 19.77, + "learning_rate": 1.0239315186599336e-06, + "loss": 0.191, + "step": 53529 + }, + { + "epoch": 19.77, + "learning_rate": 1.0238945682965882e-06, + "loss": 0.1224, + "step": 53530 + }, + { + "epoch": 19.77, + "learning_rate": 1.023857617933243e-06, + "loss": 0.1993, + "step": 53531 + }, + { + "epoch": 19.77, + "learning_rate": 1.023820667569898e-06, + "loss": 0.1398, + "step": 53532 + }, + { + "epoch": 19.77, + "learning_rate": 1.0237837172065525e-06, + "loss": 0.1931, + "step": 53533 + }, + { + "epoch": 19.77, + "learning_rate": 1.0237467668432074e-06, + "loss": 0.179, + "step": 53534 + }, + { + "epoch": 19.77, + "learning_rate": 1.023709816479862e-06, + "loss": 0.1819, + "step": 53535 + }, + { + "epoch": 19.77, + "learning_rate": 1.0236728661165169e-06, + "loss": 0.2432, + "step": 53536 + }, + { + "epoch": 19.77, + "learning_rate": 1.0236359157531715e-06, + "loss": 0.2791, + "step": 53537 + }, + { + "epoch": 19.77, + "learning_rate": 1.0235989653898264e-06, + "loss": 0.1471, + "step": 53538 + }, + { + "epoch": 19.77, + "learning_rate": 1.0235620150264812e-06, + "loss": 0.189, + "step": 53539 + }, + { + "epoch": 19.77, + "learning_rate": 1.0235250646631358e-06, + "loss": 0.2657, + "step": 53540 + }, + { + "epoch": 19.77, + "learning_rate": 1.0234881142997907e-06, + "loss": 0.281, + "step": 53541 + }, + { + "epoch": 19.77, + "learning_rate": 1.0234511639364455e-06, + "loss": 0.1566, + "step": 53542 + }, + { + "epoch": 19.77, + "learning_rate": 1.0234142135731002e-06, + "loss": 0.2822, + "step": 53543 + }, + { + "epoch": 19.77, + "learning_rate": 1.023377263209755e-06, + "loss": 0.1318, + "step": 53544 + }, + { + "epoch": 19.77, + "learning_rate": 1.0233403128464097e-06, + "loss": 0.2106, + "step": 53545 + }, + { + "epoch": 19.77, + "learning_rate": 1.0233033624830643e-06, + "loss": 0.2643, + "step": 53546 + }, + { + "epoch": 19.77, + "learning_rate": 1.0232664121197191e-06, + "loss": 0.3283, + "step": 53547 + }, + { + "epoch": 19.77, + "learning_rate": 1.023229461756374e-06, + "loss": 0.2725, + "step": 53548 + }, + { + "epoch": 19.77, + "learning_rate": 1.0231925113930286e-06, + "loss": 0.1478, + "step": 53549 + }, + { + "epoch": 19.77, + "learning_rate": 1.0231555610296835e-06, + "loss": 0.1382, + "step": 53550 + }, + { + "epoch": 19.78, + "learning_rate": 1.0231186106663383e-06, + "loss": 0.1703, + "step": 53551 + }, + { + "epoch": 19.78, + "learning_rate": 1.023081660302993e-06, + "loss": 0.295, + "step": 53552 + }, + { + "epoch": 19.78, + "learning_rate": 1.0230447099396478e-06, + "loss": 0.1777, + "step": 53553 + }, + { + "epoch": 19.78, + "learning_rate": 1.0230077595763027e-06, + "loss": 0.2696, + "step": 53554 + }, + { + "epoch": 19.78, + "learning_rate": 1.0229708092129573e-06, + "loss": 0.2484, + "step": 53555 + }, + { + "epoch": 19.78, + "learning_rate": 1.022933858849612e-06, + "loss": 0.2509, + "step": 53556 + }, + { + "epoch": 19.78, + "learning_rate": 1.0228969084862668e-06, + "loss": 0.283, + "step": 53557 + }, + { + "epoch": 19.78, + "learning_rate": 1.0228599581229216e-06, + "loss": 0.2954, + "step": 53558 + }, + { + "epoch": 19.78, + "learning_rate": 1.0228230077595763e-06, + "loss": 0.1456, + "step": 53559 + }, + { + "epoch": 19.78, + "learning_rate": 1.0227860573962311e-06, + "loss": 0.3504, + "step": 53560 + }, + { + "epoch": 19.78, + "learning_rate": 1.022749107032886e-06, + "loss": 0.1917, + "step": 53561 + }, + { + "epoch": 19.78, + "learning_rate": 1.0227121566695406e-06, + "loss": 0.2134, + "step": 53562 + }, + { + "epoch": 19.78, + "learning_rate": 1.0226752063061954e-06, + "loss": 0.1928, + "step": 53563 + }, + { + "epoch": 19.78, + "learning_rate": 1.0226382559428503e-06, + "loss": 0.2816, + "step": 53564 + }, + { + "epoch": 19.78, + "learning_rate": 1.0226013055795047e-06, + "loss": 0.4975, + "step": 53565 + }, + { + "epoch": 19.78, + "learning_rate": 1.0225643552161596e-06, + "loss": 0.2175, + "step": 53566 + }, + { + "epoch": 19.78, + "learning_rate": 1.0225274048528144e-06, + "loss": 0.2372, + "step": 53567 + }, + { + "epoch": 19.78, + "learning_rate": 1.022490454489469e-06, + "loss": 0.3526, + "step": 53568 + }, + { + "epoch": 19.78, + "learning_rate": 1.022453504126124e-06, + "loss": 0.3573, + "step": 53569 + }, + { + "epoch": 19.78, + "learning_rate": 1.0224165537627787e-06, + "loss": 0.2454, + "step": 53570 + }, + { + "epoch": 19.78, + "learning_rate": 1.0223796033994334e-06, + "loss": 0.3198, + "step": 53571 + }, + { + "epoch": 19.78, + "learning_rate": 1.0223426530360882e-06, + "loss": 0.3582, + "step": 53572 + }, + { + "epoch": 19.78, + "learning_rate": 1.022305702672743e-06, + "loss": 0.151, + "step": 53573 + }, + { + "epoch": 19.78, + "learning_rate": 1.0222687523093977e-06, + "loss": 0.1928, + "step": 53574 + }, + { + "epoch": 19.78, + "learning_rate": 1.0222318019460524e-06, + "loss": 0.2014, + "step": 53575 + }, + { + "epoch": 19.78, + "learning_rate": 1.0221948515827072e-06, + "loss": 0.1373, + "step": 53576 + }, + { + "epoch": 19.78, + "learning_rate": 1.022157901219362e-06, + "loss": 0.1173, + "step": 53577 + }, + { + "epoch": 19.79, + "learning_rate": 1.0221209508560167e-06, + "loss": 0.2071, + "step": 53578 + }, + { + "epoch": 19.79, + "learning_rate": 1.0220840004926715e-06, + "loss": 0.2868, + "step": 53579 + }, + { + "epoch": 19.79, + "learning_rate": 1.0220470501293264e-06, + "loss": 0.1434, + "step": 53580 + }, + { + "epoch": 19.79, + "learning_rate": 1.022010099765981e-06, + "loss": 0.1615, + "step": 53581 + }, + { + "epoch": 19.79, + "learning_rate": 1.0219731494026359e-06, + "loss": 0.1922, + "step": 53582 + }, + { + "epoch": 19.79, + "learning_rate": 1.0219361990392907e-06, + "loss": 0.1903, + "step": 53583 + }, + { + "epoch": 19.79, + "learning_rate": 1.0218992486759454e-06, + "loss": 0.163, + "step": 53584 + }, + { + "epoch": 19.79, + "learning_rate": 1.0218622983126e-06, + "loss": 0.1866, + "step": 53585 + }, + { + "epoch": 19.79, + "learning_rate": 1.0218253479492548e-06, + "loss": 0.2269, + "step": 53586 + }, + { + "epoch": 19.79, + "learning_rate": 1.0217883975859097e-06, + "loss": 0.1965, + "step": 53587 + }, + { + "epoch": 19.79, + "learning_rate": 1.0217514472225643e-06, + "loss": 0.155, + "step": 53588 + }, + { + "epoch": 19.79, + "learning_rate": 1.0217144968592192e-06, + "loss": 0.1758, + "step": 53589 + }, + { + "epoch": 19.79, + "learning_rate": 1.021677546495874e-06, + "loss": 0.3066, + "step": 53590 + }, + { + "epoch": 19.79, + "learning_rate": 1.0216405961325287e-06, + "loss": 0.2224, + "step": 53591 + }, + { + "epoch": 19.79, + "learning_rate": 1.0216036457691835e-06, + "loss": 0.208, + "step": 53592 + }, + { + "epoch": 19.79, + "learning_rate": 1.0215666954058384e-06, + "loss": 0.1694, + "step": 53593 + }, + { + "epoch": 19.79, + "learning_rate": 1.021529745042493e-06, + "loss": 0.1538, + "step": 53594 + }, + { + "epoch": 19.79, + "learning_rate": 1.0214927946791476e-06, + "loss": 0.2282, + "step": 53595 + }, + { + "epoch": 19.79, + "learning_rate": 1.0214558443158025e-06, + "loss": 0.1944, + "step": 53596 + }, + { + "epoch": 19.79, + "learning_rate": 1.0214188939524571e-06, + "loss": 0.2145, + "step": 53597 + }, + { + "epoch": 19.79, + "learning_rate": 1.021381943589112e-06, + "loss": 0.1644, + "step": 53598 + }, + { + "epoch": 19.79, + "learning_rate": 1.0213449932257668e-06, + "loss": 0.2232, + "step": 53599 + }, + { + "epoch": 19.79, + "learning_rate": 1.0213080428624214e-06, + "loss": 0.4012, + "step": 53600 + }, + { + "epoch": 19.79, + "learning_rate": 1.0212710924990763e-06, + "loss": 0.3269, + "step": 53601 + }, + { + "epoch": 19.79, + "learning_rate": 1.0212341421357311e-06, + "loss": 0.2551, + "step": 53602 + }, + { + "epoch": 19.79, + "learning_rate": 1.0211971917723858e-06, + "loss": 0.2753, + "step": 53603 + }, + { + "epoch": 19.79, + "learning_rate": 1.0211602414090406e-06, + "loss": 0.2023, + "step": 53604 + }, + { + "epoch": 19.8, + "learning_rate": 1.0211232910456953e-06, + "loss": 0.1915, + "step": 53605 + }, + { + "epoch": 19.8, + "learning_rate": 1.0210863406823501e-06, + "loss": 0.1549, + "step": 53606 + }, + { + "epoch": 19.8, + "learning_rate": 1.0210493903190047e-06, + "loss": 0.2297, + "step": 53607 + }, + { + "epoch": 19.8, + "learning_rate": 1.0210124399556596e-06, + "loss": 0.3186, + "step": 53608 + }, + { + "epoch": 19.8, + "learning_rate": 1.0209754895923144e-06, + "loss": 0.1692, + "step": 53609 + }, + { + "epoch": 19.8, + "learning_rate": 1.020938539228969e-06, + "loss": 0.1957, + "step": 53610 + }, + { + "epoch": 19.8, + "learning_rate": 1.020901588865624e-06, + "loss": 0.3131, + "step": 53611 + }, + { + "epoch": 19.8, + "learning_rate": 1.0208646385022788e-06, + "loss": 0.2033, + "step": 53612 + }, + { + "epoch": 19.8, + "learning_rate": 1.0208276881389334e-06, + "loss": 0.1416, + "step": 53613 + }, + { + "epoch": 19.8, + "learning_rate": 1.0207907377755883e-06, + "loss": 0.3157, + "step": 53614 + }, + { + "epoch": 19.8, + "learning_rate": 1.0207537874122429e-06, + "loss": 0.146, + "step": 53615 + }, + { + "epoch": 19.8, + "learning_rate": 1.0207168370488975e-06, + "loss": 0.2471, + "step": 53616 + }, + { + "epoch": 19.8, + "learning_rate": 1.0206798866855524e-06, + "loss": 0.1334, + "step": 53617 + }, + { + "epoch": 19.8, + "learning_rate": 1.0206429363222072e-06, + "loss": 0.1792, + "step": 53618 + }, + { + "epoch": 19.8, + "learning_rate": 1.0206059859588619e-06, + "loss": 0.2301, + "step": 53619 + }, + { + "epoch": 19.8, + "learning_rate": 1.0205690355955167e-06, + "loss": 0.1657, + "step": 53620 + }, + { + "epoch": 19.8, + "learning_rate": 1.0205320852321716e-06, + "loss": 0.1485, + "step": 53621 + }, + { + "epoch": 19.8, + "learning_rate": 1.0204951348688262e-06, + "loss": 0.1711, + "step": 53622 + }, + { + "epoch": 19.8, + "learning_rate": 1.020458184505481e-06, + "loss": 0.2791, + "step": 53623 + }, + { + "epoch": 19.8, + "learning_rate": 1.0204212341421359e-06, + "loss": 0.183, + "step": 53624 + }, + { + "epoch": 19.8, + "learning_rate": 1.0203842837787905e-06, + "loss": 0.1988, + "step": 53625 + }, + { + "epoch": 19.8, + "learning_rate": 1.0203473334154452e-06, + "loss": 0.2899, + "step": 53626 + }, + { + "epoch": 19.8, + "learning_rate": 1.0203103830521e-06, + "loss": 0.2777, + "step": 53627 + }, + { + "epoch": 19.8, + "learning_rate": 1.0202734326887549e-06, + "loss": 0.1605, + "step": 53628 + }, + { + "epoch": 19.8, + "learning_rate": 1.0202364823254095e-06, + "loss": 0.3425, + "step": 53629 + }, + { + "epoch": 19.8, + "learning_rate": 1.0201995319620643e-06, + "loss": 0.2272, + "step": 53630 + }, + { + "epoch": 19.8, + "learning_rate": 1.0201625815987192e-06, + "loss": 0.1567, + "step": 53631 + }, + { + "epoch": 19.81, + "learning_rate": 1.0201256312353738e-06, + "loss": 0.2567, + "step": 53632 + }, + { + "epoch": 19.81, + "learning_rate": 1.0200886808720287e-06, + "loss": 0.2067, + "step": 53633 + }, + { + "epoch": 19.81, + "learning_rate": 1.0200517305086835e-06, + "loss": 0.2405, + "step": 53634 + }, + { + "epoch": 19.81, + "learning_rate": 1.020014780145338e-06, + "loss": 0.221, + "step": 53635 + }, + { + "epoch": 19.81, + "learning_rate": 1.0199778297819928e-06, + "loss": 0.3179, + "step": 53636 + }, + { + "epoch": 19.81, + "learning_rate": 1.0199408794186476e-06, + "loss": 0.2671, + "step": 53637 + }, + { + "epoch": 19.81, + "learning_rate": 1.0199039290553023e-06, + "loss": 0.2252, + "step": 53638 + }, + { + "epoch": 19.81, + "learning_rate": 1.0198669786919571e-06, + "loss": 0.2293, + "step": 53639 + }, + { + "epoch": 19.81, + "learning_rate": 1.019830028328612e-06, + "loss": 0.1834, + "step": 53640 + }, + { + "epoch": 19.81, + "learning_rate": 1.0197930779652666e-06, + "loss": 0.1276, + "step": 53641 + }, + { + "epoch": 19.81, + "learning_rate": 1.0197561276019215e-06, + "loss": 0.346, + "step": 53642 + }, + { + "epoch": 19.81, + "learning_rate": 1.0197191772385763e-06, + "loss": 0.2711, + "step": 53643 + }, + { + "epoch": 19.81, + "learning_rate": 1.019682226875231e-06, + "loss": 0.2068, + "step": 53644 + }, + { + "epoch": 19.81, + "learning_rate": 1.0196452765118856e-06, + "loss": 0.2742, + "step": 53645 + }, + { + "epoch": 19.81, + "learning_rate": 1.0196083261485404e-06, + "loss": 0.1985, + "step": 53646 + }, + { + "epoch": 19.81, + "learning_rate": 1.0195713757851953e-06, + "loss": 0.3184, + "step": 53647 + }, + { + "epoch": 19.81, + "learning_rate": 1.01953442542185e-06, + "loss": 0.2251, + "step": 53648 + }, + { + "epoch": 19.81, + "learning_rate": 1.0194974750585048e-06, + "loss": 0.2222, + "step": 53649 + }, + { + "epoch": 19.81, + "learning_rate": 1.0194605246951596e-06, + "loss": 0.2576, + "step": 53650 + }, + { + "epoch": 19.81, + "learning_rate": 1.0194235743318143e-06, + "loss": 0.153, + "step": 53651 + }, + { + "epoch": 19.81, + "learning_rate": 1.019386623968469e-06, + "loss": 0.3006, + "step": 53652 + }, + { + "epoch": 19.81, + "learning_rate": 1.019349673605124e-06, + "loss": 0.315, + "step": 53653 + }, + { + "epoch": 19.81, + "learning_rate": 1.0193127232417786e-06, + "loss": 0.1671, + "step": 53654 + }, + { + "epoch": 19.81, + "learning_rate": 1.0192757728784332e-06, + "loss": 0.4002, + "step": 53655 + }, + { + "epoch": 19.81, + "learning_rate": 1.019238822515088e-06, + "loss": 0.1291, + "step": 53656 + }, + { + "epoch": 19.81, + "learning_rate": 1.019201872151743e-06, + "loss": 0.1611, + "step": 53657 + }, + { + "epoch": 19.81, + "learning_rate": 1.0191649217883976e-06, + "loss": 0.2439, + "step": 53658 + }, + { + "epoch": 19.81, + "learning_rate": 1.0191279714250524e-06, + "loss": 0.1281, + "step": 53659 + }, + { + "epoch": 19.82, + "learning_rate": 1.0190910210617073e-06, + "loss": 0.1809, + "step": 53660 + }, + { + "epoch": 19.82, + "learning_rate": 1.0190540706983619e-06, + "loss": 0.1505, + "step": 53661 + }, + { + "epoch": 19.82, + "learning_rate": 1.0190171203350167e-06, + "loss": 0.1904, + "step": 53662 + }, + { + "epoch": 19.82, + "learning_rate": 1.0189801699716716e-06, + "loss": 0.2006, + "step": 53663 + }, + { + "epoch": 19.82, + "learning_rate": 1.018943219608326e-06, + "loss": 0.1831, + "step": 53664 + }, + { + "epoch": 19.82, + "learning_rate": 1.0189062692449809e-06, + "loss": 0.2822, + "step": 53665 + }, + { + "epoch": 19.82, + "learning_rate": 1.0188693188816357e-06, + "loss": 0.2767, + "step": 53666 + }, + { + "epoch": 19.82, + "learning_rate": 1.0188323685182903e-06, + "loss": 0.1498, + "step": 53667 + }, + { + "epoch": 19.82, + "learning_rate": 1.0187954181549452e-06, + "loss": 0.1946, + "step": 53668 + }, + { + "epoch": 19.82, + "learning_rate": 1.0187584677916e-06, + "loss": 0.1398, + "step": 53669 + }, + { + "epoch": 19.82, + "learning_rate": 1.0187215174282547e-06, + "loss": 0.2747, + "step": 53670 + }, + { + "epoch": 19.82, + "learning_rate": 1.0186845670649095e-06, + "loss": 0.2143, + "step": 53671 + }, + { + "epoch": 19.82, + "learning_rate": 1.0186476167015644e-06, + "loss": 0.1805, + "step": 53672 + }, + { + "epoch": 19.82, + "learning_rate": 1.018610666338219e-06, + "loss": 0.1335, + "step": 53673 + }, + { + "epoch": 19.82, + "learning_rate": 1.0185737159748736e-06, + "loss": 0.3065, + "step": 53674 + }, + { + "epoch": 19.82, + "learning_rate": 1.0185367656115285e-06, + "loss": 0.1665, + "step": 53675 + }, + { + "epoch": 19.82, + "learning_rate": 1.0184998152481833e-06, + "loss": 0.3111, + "step": 53676 + }, + { + "epoch": 19.82, + "learning_rate": 1.018462864884838e-06, + "loss": 0.1502, + "step": 53677 + }, + { + "epoch": 19.82, + "learning_rate": 1.0184259145214928e-06, + "loss": 0.2916, + "step": 53678 + }, + { + "epoch": 19.82, + "learning_rate": 1.0183889641581477e-06, + "loss": 0.1624, + "step": 53679 + }, + { + "epoch": 19.82, + "learning_rate": 1.0183520137948023e-06, + "loss": 0.2421, + "step": 53680 + }, + { + "epoch": 19.82, + "learning_rate": 1.0183150634314572e-06, + "loss": 0.2628, + "step": 53681 + }, + { + "epoch": 19.82, + "learning_rate": 1.018278113068112e-06, + "loss": 0.2204, + "step": 53682 + }, + { + "epoch": 19.82, + "learning_rate": 1.0182411627047666e-06, + "loss": 0.2522, + "step": 53683 + }, + { + "epoch": 19.82, + "learning_rate": 1.0182042123414213e-06, + "loss": 0.2747, + "step": 53684 + }, + { + "epoch": 19.82, + "learning_rate": 1.0181672619780761e-06, + "loss": 0.3461, + "step": 53685 + }, + { + "epoch": 19.82, + "learning_rate": 1.0181303116147308e-06, + "loss": 0.1768, + "step": 53686 + }, + { + "epoch": 19.83, + "learning_rate": 1.0180933612513856e-06, + "loss": 0.1611, + "step": 53687 + }, + { + "epoch": 19.83, + "learning_rate": 1.0180564108880405e-06, + "loss": 0.1822, + "step": 53688 + }, + { + "epoch": 19.83, + "learning_rate": 1.018019460524695e-06, + "loss": 0.2088, + "step": 53689 + }, + { + "epoch": 19.83, + "learning_rate": 1.01798251016135e-06, + "loss": 0.2501, + "step": 53690 + }, + { + "epoch": 19.83, + "learning_rate": 1.0179455597980048e-06, + "loss": 0.22, + "step": 53691 + }, + { + "epoch": 19.83, + "learning_rate": 1.0179086094346594e-06, + "loss": 0.0975, + "step": 53692 + }, + { + "epoch": 19.83, + "learning_rate": 1.0178716590713143e-06, + "loss": 0.2218, + "step": 53693 + }, + { + "epoch": 19.83, + "learning_rate": 1.017834708707969e-06, + "loss": 0.175, + "step": 53694 + }, + { + "epoch": 19.83, + "learning_rate": 1.0177977583446238e-06, + "loss": 0.2427, + "step": 53695 + }, + { + "epoch": 19.83, + "learning_rate": 1.0177608079812784e-06, + "loss": 0.2508, + "step": 53696 + }, + { + "epoch": 19.83, + "learning_rate": 1.0177238576179332e-06, + "loss": 0.1991, + "step": 53697 + }, + { + "epoch": 19.83, + "learning_rate": 1.017686907254588e-06, + "loss": 0.2766, + "step": 53698 + }, + { + "epoch": 19.83, + "learning_rate": 1.0176499568912427e-06, + "loss": 0.2127, + "step": 53699 + }, + { + "epoch": 19.83, + "learning_rate": 1.0176130065278976e-06, + "loss": 0.2032, + "step": 53700 + }, + { + "epoch": 19.83, + "learning_rate": 1.0175760561645524e-06, + "loss": 0.2622, + "step": 53701 + }, + { + "epoch": 19.83, + "learning_rate": 1.017539105801207e-06, + "loss": 0.2308, + "step": 53702 + }, + { + "epoch": 19.83, + "learning_rate": 1.017502155437862e-06, + "loss": 0.2037, + "step": 53703 + }, + { + "epoch": 19.83, + "learning_rate": 1.0174652050745165e-06, + "loss": 0.186, + "step": 53704 + }, + { + "epoch": 19.83, + "learning_rate": 1.0174282547111714e-06, + "loss": 0.3695, + "step": 53705 + }, + { + "epoch": 19.83, + "learning_rate": 1.017391304347826e-06, + "loss": 0.293, + "step": 53706 + }, + { + "epoch": 19.83, + "learning_rate": 1.0173543539844809e-06, + "loss": 0.3161, + "step": 53707 + }, + { + "epoch": 19.83, + "learning_rate": 1.0173174036211355e-06, + "loss": 0.3157, + "step": 53708 + }, + { + "epoch": 19.83, + "learning_rate": 1.0172804532577904e-06, + "loss": 0.2237, + "step": 53709 + }, + { + "epoch": 19.83, + "learning_rate": 1.0172435028944452e-06, + "loss": 0.1308, + "step": 53710 + }, + { + "epoch": 19.83, + "learning_rate": 1.0172065525310998e-06, + "loss": 0.2262, + "step": 53711 + }, + { + "epoch": 19.83, + "learning_rate": 1.0171696021677547e-06, + "loss": 0.2935, + "step": 53712 + }, + { + "epoch": 19.83, + "learning_rate": 1.0171326518044095e-06, + "loss": 0.1333, + "step": 53713 + }, + { + "epoch": 19.84, + "learning_rate": 1.0170957014410642e-06, + "loss": 0.2006, + "step": 53714 + }, + { + "epoch": 19.84, + "learning_rate": 1.0170587510777188e-06, + "loss": 0.1991, + "step": 53715 + }, + { + "epoch": 19.84, + "learning_rate": 1.0170218007143737e-06, + "loss": 0.2128, + "step": 53716 + }, + { + "epoch": 19.84, + "learning_rate": 1.0169848503510285e-06, + "loss": 0.1746, + "step": 53717 + }, + { + "epoch": 19.84, + "learning_rate": 1.0169478999876832e-06, + "loss": 0.1759, + "step": 53718 + }, + { + "epoch": 19.84, + "learning_rate": 1.016910949624338e-06, + "loss": 0.1977, + "step": 53719 + }, + { + "epoch": 19.84, + "learning_rate": 1.0168739992609928e-06, + "loss": 0.338, + "step": 53720 + }, + { + "epoch": 19.84, + "learning_rate": 1.0168370488976475e-06, + "loss": 0.2417, + "step": 53721 + }, + { + "epoch": 19.84, + "learning_rate": 1.0168000985343023e-06, + "loss": 0.2105, + "step": 53722 + }, + { + "epoch": 19.84, + "learning_rate": 1.0167631481709572e-06, + "loss": 0.2295, + "step": 53723 + }, + { + "epoch": 19.84, + "learning_rate": 1.0167261978076118e-06, + "loss": 0.2386, + "step": 53724 + }, + { + "epoch": 19.84, + "learning_rate": 1.0166892474442665e-06, + "loss": 0.1396, + "step": 53725 + }, + { + "epoch": 19.84, + "learning_rate": 1.0166522970809213e-06, + "loss": 0.2831, + "step": 53726 + }, + { + "epoch": 19.84, + "learning_rate": 1.0166153467175761e-06, + "loss": 0.1505, + "step": 53727 + }, + { + "epoch": 19.84, + "learning_rate": 1.0165783963542308e-06, + "loss": 0.2244, + "step": 53728 + }, + { + "epoch": 19.84, + "learning_rate": 1.0165414459908856e-06, + "loss": 0.2033, + "step": 53729 + }, + { + "epoch": 19.84, + "learning_rate": 1.0165044956275405e-06, + "loss": 0.1994, + "step": 53730 + }, + { + "epoch": 19.84, + "learning_rate": 1.0164675452641951e-06, + "loss": 0.1338, + "step": 53731 + }, + { + "epoch": 19.84, + "learning_rate": 1.01643059490085e-06, + "loss": 0.2557, + "step": 53732 + }, + { + "epoch": 19.84, + "learning_rate": 1.0163936445375048e-06, + "loss": 0.1511, + "step": 53733 + }, + { + "epoch": 19.84, + "learning_rate": 1.0163566941741592e-06, + "loss": 0.2747, + "step": 53734 + }, + { + "epoch": 19.84, + "learning_rate": 1.016319743810814e-06, + "loss": 0.2228, + "step": 53735 + }, + { + "epoch": 19.84, + "learning_rate": 1.016282793447469e-06, + "loss": 0.2208, + "step": 53736 + }, + { + "epoch": 19.84, + "learning_rate": 1.0162458430841236e-06, + "loss": 0.1896, + "step": 53737 + }, + { + "epoch": 19.84, + "learning_rate": 1.0162088927207784e-06, + "loss": 0.2316, + "step": 53738 + }, + { + "epoch": 19.84, + "learning_rate": 1.0161719423574333e-06, + "loss": 0.2374, + "step": 53739 + }, + { + "epoch": 19.84, + "learning_rate": 1.016134991994088e-06, + "loss": 0.1799, + "step": 53740 + }, + { + "epoch": 19.85, + "learning_rate": 1.0160980416307428e-06, + "loss": 0.2266, + "step": 53741 + }, + { + "epoch": 19.85, + "learning_rate": 1.0160610912673976e-06, + "loss": 0.1588, + "step": 53742 + }, + { + "epoch": 19.85, + "learning_rate": 1.0160241409040522e-06, + "loss": 0.1653, + "step": 53743 + }, + { + "epoch": 19.85, + "learning_rate": 1.0159871905407069e-06, + "loss": 0.1552, + "step": 53744 + }, + { + "epoch": 19.85, + "learning_rate": 1.0159502401773617e-06, + "loss": 0.2304, + "step": 53745 + }, + { + "epoch": 19.85, + "learning_rate": 1.0159132898140166e-06, + "loss": 0.2638, + "step": 53746 + }, + { + "epoch": 19.85, + "learning_rate": 1.0158763394506712e-06, + "loss": 0.2525, + "step": 53747 + }, + { + "epoch": 19.85, + "learning_rate": 1.015839389087326e-06, + "loss": 0.1346, + "step": 53748 + }, + { + "epoch": 19.85, + "learning_rate": 1.015802438723981e-06, + "loss": 0.2328, + "step": 53749 + }, + { + "epoch": 19.85, + "learning_rate": 1.0157654883606355e-06, + "loss": 0.2431, + "step": 53750 + }, + { + "epoch": 19.85, + "learning_rate": 1.0157285379972904e-06, + "loss": 0.3579, + "step": 53751 + }, + { + "epoch": 19.85, + "learning_rate": 1.0156915876339452e-06, + "loss": 0.2593, + "step": 53752 + }, + { + "epoch": 19.85, + "learning_rate": 1.0156546372705999e-06, + "loss": 0.2433, + "step": 53753 + }, + { + "epoch": 19.85, + "learning_rate": 1.0156176869072545e-06, + "loss": 0.3743, + "step": 53754 + }, + { + "epoch": 19.85, + "learning_rate": 1.0155807365439094e-06, + "loss": 0.1673, + "step": 53755 + }, + { + "epoch": 19.85, + "learning_rate": 1.015543786180564e-06, + "loss": 0.0813, + "step": 53756 + }, + { + "epoch": 19.85, + "learning_rate": 1.0155068358172188e-06, + "loss": 0.1982, + "step": 53757 + }, + { + "epoch": 19.85, + "learning_rate": 1.0154698854538737e-06, + "loss": 0.2381, + "step": 53758 + }, + { + "epoch": 19.85, + "learning_rate": 1.0154329350905283e-06, + "loss": 0.2917, + "step": 53759 + }, + { + "epoch": 19.85, + "learning_rate": 1.0153959847271832e-06, + "loss": 0.3149, + "step": 53760 + }, + { + "epoch": 19.85, + "learning_rate": 1.015359034363838e-06, + "loss": 0.3453, + "step": 53761 + }, + { + "epoch": 19.85, + "learning_rate": 1.0153220840004927e-06, + "loss": 0.3388, + "step": 53762 + }, + { + "epoch": 19.85, + "learning_rate": 1.0152851336371475e-06, + "loss": 0.1977, + "step": 53763 + }, + { + "epoch": 19.85, + "learning_rate": 1.0152481832738021e-06, + "loss": 0.1793, + "step": 53764 + }, + { + "epoch": 19.85, + "learning_rate": 1.015211232910457e-06, + "loss": 0.2195, + "step": 53765 + }, + { + "epoch": 19.85, + "learning_rate": 1.0151742825471116e-06, + "loss": 0.1464, + "step": 53766 + }, + { + "epoch": 19.85, + "learning_rate": 1.0151373321837665e-06, + "loss": 0.0987, + "step": 53767 + }, + { + "epoch": 19.86, + "learning_rate": 1.0151003818204213e-06, + "loss": 0.1525, + "step": 53768 + }, + { + "epoch": 19.86, + "learning_rate": 1.015063431457076e-06, + "loss": 0.177, + "step": 53769 + }, + { + "epoch": 19.86, + "learning_rate": 1.0150264810937308e-06, + "loss": 0.3094, + "step": 53770 + }, + { + "epoch": 19.86, + "learning_rate": 1.0149895307303857e-06, + "loss": 0.2318, + "step": 53771 + }, + { + "epoch": 19.86, + "learning_rate": 1.0149525803670403e-06, + "loss": 0.2149, + "step": 53772 + }, + { + "epoch": 19.86, + "learning_rate": 1.0149156300036951e-06, + "loss": 0.2043, + "step": 53773 + }, + { + "epoch": 19.86, + "learning_rate": 1.0148786796403498e-06, + "loss": 0.1857, + "step": 53774 + }, + { + "epoch": 19.86, + "learning_rate": 1.0148417292770046e-06, + "loss": 0.1089, + "step": 53775 + }, + { + "epoch": 19.86, + "learning_rate": 1.0148047789136593e-06, + "loss": 0.2541, + "step": 53776 + }, + { + "epoch": 19.86, + "learning_rate": 1.0147678285503141e-06, + "loss": 0.2456, + "step": 53777 + }, + { + "epoch": 19.86, + "learning_rate": 1.014730878186969e-06, + "loss": 0.2713, + "step": 53778 + }, + { + "epoch": 19.86, + "learning_rate": 1.0146939278236236e-06, + "loss": 0.1762, + "step": 53779 + }, + { + "epoch": 19.86, + "learning_rate": 1.0146569774602784e-06, + "loss": 0.2891, + "step": 53780 + }, + { + "epoch": 19.86, + "learning_rate": 1.0146200270969333e-06, + "loss": 0.2154, + "step": 53781 + }, + { + "epoch": 19.86, + "learning_rate": 1.014583076733588e-06, + "loss": 0.2128, + "step": 53782 + }, + { + "epoch": 19.86, + "learning_rate": 1.0145461263702428e-06, + "loss": 0.2781, + "step": 53783 + }, + { + "epoch": 19.86, + "learning_rate": 1.0145091760068974e-06, + "loss": 0.2354, + "step": 53784 + }, + { + "epoch": 19.86, + "learning_rate": 1.014472225643552e-06, + "loss": 0.1844, + "step": 53785 + }, + { + "epoch": 19.86, + "learning_rate": 1.014435275280207e-06, + "loss": 0.1909, + "step": 53786 + }, + { + "epoch": 19.86, + "learning_rate": 1.0143983249168617e-06, + "loss": 0.178, + "step": 53787 + }, + { + "epoch": 19.86, + "learning_rate": 1.0143613745535164e-06, + "loss": 0.2159, + "step": 53788 + }, + { + "epoch": 19.86, + "learning_rate": 1.0143244241901712e-06, + "loss": 0.3941, + "step": 53789 + }, + { + "epoch": 19.86, + "learning_rate": 1.014287473826826e-06, + "loss": 0.1194, + "step": 53790 + }, + { + "epoch": 19.86, + "learning_rate": 1.0142505234634807e-06, + "loss": 0.1533, + "step": 53791 + }, + { + "epoch": 19.86, + "learning_rate": 1.0142135731001356e-06, + "loss": 0.241, + "step": 53792 + }, + { + "epoch": 19.86, + "learning_rate": 1.0141766227367904e-06, + "loss": 0.213, + "step": 53793 + }, + { + "epoch": 19.86, + "learning_rate": 1.014139672373445e-06, + "loss": 0.1302, + "step": 53794 + }, + { + "epoch": 19.87, + "learning_rate": 1.0141027220100997e-06, + "loss": 0.292, + "step": 53795 + }, + { + "epoch": 19.87, + "learning_rate": 1.0140657716467545e-06, + "loss": 0.2307, + "step": 53796 + }, + { + "epoch": 19.87, + "learning_rate": 1.0140288212834094e-06, + "loss": 0.1953, + "step": 53797 + }, + { + "epoch": 19.87, + "learning_rate": 1.013991870920064e-06, + "loss": 0.2833, + "step": 53798 + }, + { + "epoch": 19.87, + "learning_rate": 1.0139549205567189e-06, + "loss": 0.2598, + "step": 53799 + }, + { + "epoch": 19.87, + "learning_rate": 1.0139179701933737e-06, + "loss": 0.2233, + "step": 53800 + }, + { + "epoch": 19.87, + "learning_rate": 1.0138810198300284e-06, + "loss": 0.2512, + "step": 53801 + }, + { + "epoch": 19.87, + "learning_rate": 1.0138440694666832e-06, + "loss": 0.1429, + "step": 53802 + }, + { + "epoch": 19.87, + "learning_rate": 1.013807119103338e-06, + "loss": 0.2076, + "step": 53803 + }, + { + "epoch": 19.87, + "learning_rate": 1.0137701687399925e-06, + "loss": 0.2308, + "step": 53804 + }, + { + "epoch": 19.87, + "learning_rate": 1.0137332183766473e-06, + "loss": 0.1575, + "step": 53805 + }, + { + "epoch": 19.87, + "learning_rate": 1.0136962680133022e-06, + "loss": 0.1737, + "step": 53806 + }, + { + "epoch": 19.87, + "learning_rate": 1.0136593176499568e-06, + "loss": 0.2383, + "step": 53807 + }, + { + "epoch": 19.87, + "learning_rate": 1.0136223672866117e-06, + "loss": 0.1777, + "step": 53808 + }, + { + "epoch": 19.87, + "learning_rate": 1.0135854169232665e-06, + "loss": 0.1287, + "step": 53809 + }, + { + "epoch": 19.87, + "learning_rate": 1.0135484665599211e-06, + "loss": 0.2388, + "step": 53810 + }, + { + "epoch": 19.87, + "learning_rate": 1.013511516196576e-06, + "loss": 0.2543, + "step": 53811 + }, + { + "epoch": 19.87, + "learning_rate": 1.0134745658332308e-06, + "loss": 0.3246, + "step": 53812 + }, + { + "epoch": 19.87, + "learning_rate": 1.0134376154698855e-06, + "loss": 0.1543, + "step": 53813 + }, + { + "epoch": 19.87, + "learning_rate": 1.0134006651065401e-06, + "loss": 0.2632, + "step": 53814 + }, + { + "epoch": 19.87, + "learning_rate": 1.013363714743195e-06, + "loss": 0.2179, + "step": 53815 + }, + { + "epoch": 19.87, + "learning_rate": 1.0133267643798498e-06, + "loss": 0.1419, + "step": 53816 + }, + { + "epoch": 19.87, + "learning_rate": 1.0132898140165044e-06, + "loss": 0.2439, + "step": 53817 + }, + { + "epoch": 19.87, + "learning_rate": 1.0132528636531593e-06, + "loss": 0.2075, + "step": 53818 + }, + { + "epoch": 19.87, + "learning_rate": 1.0132159132898141e-06, + "loss": 0.2315, + "step": 53819 + }, + { + "epoch": 19.87, + "learning_rate": 1.0131789629264688e-06, + "loss": 0.2205, + "step": 53820 + }, + { + "epoch": 19.87, + "learning_rate": 1.0131420125631236e-06, + "loss": 0.1373, + "step": 53821 + }, + { + "epoch": 19.88, + "learning_rate": 1.0131050621997785e-06, + "loss": 0.2857, + "step": 53822 + }, + { + "epoch": 19.88, + "learning_rate": 1.013068111836433e-06, + "loss": 0.2885, + "step": 53823 + }, + { + "epoch": 19.88, + "learning_rate": 1.0130311614730877e-06, + "loss": 0.204, + "step": 53824 + }, + { + "epoch": 19.88, + "learning_rate": 1.0129942111097426e-06, + "loss": 0.1813, + "step": 53825 + }, + { + "epoch": 19.88, + "learning_rate": 1.0129572607463972e-06, + "loss": 0.2148, + "step": 53826 + }, + { + "epoch": 19.88, + "learning_rate": 1.012920310383052e-06, + "loss": 0.2014, + "step": 53827 + }, + { + "epoch": 19.88, + "learning_rate": 1.012883360019707e-06, + "loss": 0.1285, + "step": 53828 + }, + { + "epoch": 19.88, + "learning_rate": 1.0128464096563616e-06, + "loss": 0.1199, + "step": 53829 + }, + { + "epoch": 19.88, + "learning_rate": 1.0128094592930164e-06, + "loss": 0.2076, + "step": 53830 + }, + { + "epoch": 19.88, + "learning_rate": 1.0127725089296713e-06, + "loss": 0.1802, + "step": 53831 + }, + { + "epoch": 19.88, + "learning_rate": 1.0127355585663259e-06, + "loss": 0.201, + "step": 53832 + }, + { + "epoch": 19.88, + "learning_rate": 1.0126986082029807e-06, + "loss": 0.1882, + "step": 53833 + }, + { + "epoch": 19.88, + "learning_rate": 1.0126616578396354e-06, + "loss": 0.1351, + "step": 53834 + }, + { + "epoch": 19.88, + "learning_rate": 1.0126247074762902e-06, + "loss": 0.178, + "step": 53835 + }, + { + "epoch": 19.88, + "learning_rate": 1.0125877571129449e-06, + "loss": 0.2125, + "step": 53836 + }, + { + "epoch": 19.88, + "learning_rate": 1.0125508067495997e-06, + "loss": 0.3077, + "step": 53837 + }, + { + "epoch": 19.88, + "learning_rate": 1.0125138563862546e-06, + "loss": 0.1739, + "step": 53838 + }, + { + "epoch": 19.88, + "learning_rate": 1.0124769060229092e-06, + "loss": 0.1192, + "step": 53839 + }, + { + "epoch": 19.88, + "learning_rate": 1.012439955659564e-06, + "loss": 0.1312, + "step": 53840 + }, + { + "epoch": 19.88, + "learning_rate": 1.0124030052962189e-06, + "loss": 0.2818, + "step": 53841 + }, + { + "epoch": 19.88, + "learning_rate": 1.0123660549328735e-06, + "loss": 0.206, + "step": 53842 + }, + { + "epoch": 19.88, + "learning_rate": 1.0123291045695284e-06, + "loss": 0.1435, + "step": 53843 + }, + { + "epoch": 19.88, + "learning_rate": 1.012292154206183e-06, + "loss": 0.1971, + "step": 53844 + }, + { + "epoch": 19.88, + "learning_rate": 1.0122552038428379e-06, + "loss": 0.1542, + "step": 53845 + }, + { + "epoch": 19.88, + "learning_rate": 1.0122182534794925e-06, + "loss": 0.2789, + "step": 53846 + }, + { + "epoch": 19.88, + "learning_rate": 1.0121813031161473e-06, + "loss": 0.3615, + "step": 53847 + }, + { + "epoch": 19.88, + "learning_rate": 1.0121443527528022e-06, + "loss": 0.1643, + "step": 53848 + }, + { + "epoch": 19.89, + "learning_rate": 1.0121074023894568e-06, + "loss": 0.2598, + "step": 53849 + }, + { + "epoch": 19.89, + "learning_rate": 1.0120704520261117e-06, + "loss": 0.184, + "step": 53850 + }, + { + "epoch": 19.89, + "learning_rate": 1.0120335016627665e-06, + "loss": 0.0817, + "step": 53851 + }, + { + "epoch": 19.89, + "learning_rate": 1.0119965512994212e-06, + "loss": 0.2648, + "step": 53852 + }, + { + "epoch": 19.89, + "learning_rate": 1.011959600936076e-06, + "loss": 0.2353, + "step": 53853 + }, + { + "epoch": 19.89, + "learning_rate": 1.0119226505727306e-06, + "loss": 0.171, + "step": 53854 + }, + { + "epoch": 19.89, + "learning_rate": 1.0118857002093853e-06, + "loss": 0.2003, + "step": 53855 + }, + { + "epoch": 19.89, + "learning_rate": 1.0118487498460401e-06, + "loss": 0.3028, + "step": 53856 + }, + { + "epoch": 19.89, + "learning_rate": 1.011811799482695e-06, + "loss": 0.2191, + "step": 53857 + }, + { + "epoch": 19.89, + "learning_rate": 1.0117748491193496e-06, + "loss": 0.2885, + "step": 53858 + }, + { + "epoch": 19.89, + "learning_rate": 1.0117378987560045e-06, + "loss": 0.2552, + "step": 53859 + }, + { + "epoch": 19.89, + "learning_rate": 1.0117009483926593e-06, + "loss": 0.1697, + "step": 53860 + }, + { + "epoch": 19.89, + "learning_rate": 1.011663998029314e-06, + "loss": 0.2505, + "step": 53861 + }, + { + "epoch": 19.89, + "learning_rate": 1.0116270476659688e-06, + "loss": 0.2508, + "step": 53862 + }, + { + "epoch": 19.89, + "learning_rate": 1.0115900973026236e-06, + "loss": 0.1911, + "step": 53863 + }, + { + "epoch": 19.89, + "learning_rate": 1.0115531469392783e-06, + "loss": 0.2088, + "step": 53864 + }, + { + "epoch": 19.89, + "learning_rate": 1.011516196575933e-06, + "loss": 0.168, + "step": 53865 + }, + { + "epoch": 19.89, + "learning_rate": 1.0114792462125878e-06, + "loss": 0.3152, + "step": 53866 + }, + { + "epoch": 19.89, + "learning_rate": 1.0114422958492426e-06, + "loss": 0.1804, + "step": 53867 + }, + { + "epoch": 19.89, + "learning_rate": 1.0114053454858973e-06, + "loss": 0.1813, + "step": 53868 + }, + { + "epoch": 19.89, + "learning_rate": 1.011368395122552e-06, + "loss": 0.2305, + "step": 53869 + }, + { + "epoch": 19.89, + "learning_rate": 1.011331444759207e-06, + "loss": 0.2488, + "step": 53870 + }, + { + "epoch": 19.89, + "learning_rate": 1.0112944943958616e-06, + "loss": 0.1371, + "step": 53871 + }, + { + "epoch": 19.89, + "learning_rate": 1.0112575440325164e-06, + "loss": 0.2576, + "step": 53872 + }, + { + "epoch": 19.89, + "learning_rate": 1.011220593669171e-06, + "loss": 0.1001, + "step": 53873 + }, + { + "epoch": 19.89, + "learning_rate": 1.0111836433058257e-06, + "loss": 0.2182, + "step": 53874 + }, + { + "epoch": 19.89, + "learning_rate": 1.0111466929424806e-06, + "loss": 0.1119, + "step": 53875 + }, + { + "epoch": 19.9, + "learning_rate": 1.0111097425791354e-06, + "loss": 0.2878, + "step": 53876 + }, + { + "epoch": 19.9, + "learning_rate": 1.01107279221579e-06, + "loss": 0.3085, + "step": 53877 + }, + { + "epoch": 19.9, + "learning_rate": 1.0110358418524449e-06, + "loss": 0.2279, + "step": 53878 + }, + { + "epoch": 19.9, + "learning_rate": 1.0109988914890997e-06, + "loss": 0.142, + "step": 53879 + }, + { + "epoch": 19.9, + "learning_rate": 1.0109619411257544e-06, + "loss": 0.1038, + "step": 53880 + }, + { + "epoch": 19.9, + "learning_rate": 1.0109249907624092e-06, + "loss": 0.1513, + "step": 53881 + }, + { + "epoch": 19.9, + "learning_rate": 1.010888040399064e-06, + "loss": 0.1868, + "step": 53882 + }, + { + "epoch": 19.9, + "learning_rate": 1.0108510900357187e-06, + "loss": 0.2288, + "step": 53883 + }, + { + "epoch": 19.9, + "learning_rate": 1.0108141396723733e-06, + "loss": 0.3229, + "step": 53884 + }, + { + "epoch": 19.9, + "learning_rate": 1.0107771893090282e-06, + "loss": 0.2701, + "step": 53885 + }, + { + "epoch": 19.9, + "learning_rate": 1.010740238945683e-06, + "loss": 0.2243, + "step": 53886 + }, + { + "epoch": 19.9, + "learning_rate": 1.0107032885823377e-06, + "loss": 0.2487, + "step": 53887 + }, + { + "epoch": 19.9, + "learning_rate": 1.0106663382189925e-06, + "loss": 0.1993, + "step": 53888 + }, + { + "epoch": 19.9, + "learning_rate": 1.0106293878556474e-06, + "loss": 0.234, + "step": 53889 + }, + { + "epoch": 19.9, + "learning_rate": 1.010592437492302e-06, + "loss": 0.1637, + "step": 53890 + }, + { + "epoch": 19.9, + "learning_rate": 1.0105554871289569e-06, + "loss": 0.3116, + "step": 53891 + }, + { + "epoch": 19.9, + "learning_rate": 1.0105185367656117e-06, + "loss": 0.2041, + "step": 53892 + }, + { + "epoch": 19.9, + "learning_rate": 1.0104815864022663e-06, + "loss": 0.1999, + "step": 53893 + }, + { + "epoch": 19.9, + "learning_rate": 1.010444636038921e-06, + "loss": 0.2943, + "step": 53894 + }, + { + "epoch": 19.9, + "learning_rate": 1.0104076856755758e-06, + "loss": 0.2339, + "step": 53895 + }, + { + "epoch": 19.9, + "learning_rate": 1.0103707353122305e-06, + "loss": 0.1276, + "step": 53896 + }, + { + "epoch": 19.9, + "learning_rate": 1.0103337849488853e-06, + "loss": 0.1431, + "step": 53897 + }, + { + "epoch": 19.9, + "learning_rate": 1.0102968345855402e-06, + "loss": 0.1023, + "step": 53898 + }, + { + "epoch": 19.9, + "learning_rate": 1.0102598842221948e-06, + "loss": 0.1863, + "step": 53899 + }, + { + "epoch": 19.9, + "learning_rate": 1.0102229338588496e-06, + "loss": 0.1705, + "step": 53900 + }, + { + "epoch": 19.9, + "learning_rate": 1.0101859834955045e-06, + "loss": 0.1171, + "step": 53901 + }, + { + "epoch": 19.9, + "learning_rate": 1.0101490331321591e-06, + "loss": 0.2409, + "step": 53902 + }, + { + "epoch": 19.91, + "learning_rate": 1.0101120827688138e-06, + "loss": 0.2722, + "step": 53903 + }, + { + "epoch": 19.91, + "learning_rate": 1.0100751324054686e-06, + "loss": 0.228, + "step": 53904 + }, + { + "epoch": 19.91, + "learning_rate": 1.0100381820421235e-06, + "loss": 0.2056, + "step": 53905 + }, + { + "epoch": 19.91, + "learning_rate": 1.010001231678778e-06, + "loss": 0.1404, + "step": 53906 + }, + { + "epoch": 19.91, + "learning_rate": 1.009964281315433e-06, + "loss": 0.2802, + "step": 53907 + }, + { + "epoch": 19.91, + "learning_rate": 1.0099273309520878e-06, + "loss": 0.1589, + "step": 53908 + }, + { + "epoch": 19.91, + "learning_rate": 1.0098903805887424e-06, + "loss": 0.283, + "step": 53909 + }, + { + "epoch": 19.91, + "learning_rate": 1.0098534302253973e-06, + "loss": 0.3001, + "step": 53910 + }, + { + "epoch": 19.91, + "learning_rate": 1.0098164798620521e-06, + "loss": 0.2538, + "step": 53911 + }, + { + "epoch": 19.91, + "learning_rate": 1.0097795294987068e-06, + "loss": 0.2882, + "step": 53912 + }, + { + "epoch": 19.91, + "learning_rate": 1.0097425791353614e-06, + "loss": 0.309, + "step": 53913 + }, + { + "epoch": 19.91, + "learning_rate": 1.0097056287720162e-06, + "loss": 0.1819, + "step": 53914 + }, + { + "epoch": 19.91, + "learning_rate": 1.009668678408671e-06, + "loss": 0.1397, + "step": 53915 + }, + { + "epoch": 19.91, + "learning_rate": 1.0096317280453257e-06, + "loss": 0.2763, + "step": 53916 + }, + { + "epoch": 19.91, + "learning_rate": 1.0095947776819806e-06, + "loss": 0.2095, + "step": 53917 + }, + { + "epoch": 19.91, + "learning_rate": 1.0095578273186354e-06, + "loss": 0.3577, + "step": 53918 + }, + { + "epoch": 19.91, + "learning_rate": 1.00952087695529e-06, + "loss": 0.352, + "step": 53919 + }, + { + "epoch": 19.91, + "learning_rate": 1.009483926591945e-06, + "loss": 0.2245, + "step": 53920 + }, + { + "epoch": 19.91, + "learning_rate": 1.0094469762285998e-06, + "loss": 0.265, + "step": 53921 + }, + { + "epoch": 19.91, + "learning_rate": 1.0094100258652544e-06, + "loss": 0.2652, + "step": 53922 + }, + { + "epoch": 19.91, + "learning_rate": 1.009373075501909e-06, + "loss": 0.1824, + "step": 53923 + }, + { + "epoch": 19.91, + "learning_rate": 1.0093361251385639e-06, + "loss": 0.2051, + "step": 53924 + }, + { + "epoch": 19.91, + "learning_rate": 1.0092991747752185e-06, + "loss": 0.2034, + "step": 53925 + }, + { + "epoch": 19.91, + "learning_rate": 1.0092622244118734e-06, + "loss": 0.1169, + "step": 53926 + }, + { + "epoch": 19.91, + "learning_rate": 1.0092252740485282e-06, + "loss": 0.1894, + "step": 53927 + }, + { + "epoch": 19.91, + "learning_rate": 1.0091883236851828e-06, + "loss": 0.1092, + "step": 53928 + }, + { + "epoch": 19.91, + "learning_rate": 1.0091513733218377e-06, + "loss": 0.2864, + "step": 53929 + }, + { + "epoch": 19.92, + "learning_rate": 1.0091144229584925e-06, + "loss": 0.0827, + "step": 53930 + }, + { + "epoch": 19.92, + "learning_rate": 1.0090774725951472e-06, + "loss": 0.2942, + "step": 53931 + }, + { + "epoch": 19.92, + "learning_rate": 1.009040522231802e-06, + "loss": 0.2036, + "step": 53932 + }, + { + "epoch": 19.92, + "learning_rate": 1.0090035718684567e-06, + "loss": 0.239, + "step": 53933 + }, + { + "epoch": 19.92, + "learning_rate": 1.0089666215051115e-06, + "loss": 0.2356, + "step": 53934 + }, + { + "epoch": 19.92, + "learning_rate": 1.0089296711417662e-06, + "loss": 0.1564, + "step": 53935 + }, + { + "epoch": 19.92, + "learning_rate": 1.008892720778421e-06, + "loss": 0.25, + "step": 53936 + }, + { + "epoch": 19.92, + "learning_rate": 1.0088557704150758e-06, + "loss": 0.106, + "step": 53937 + }, + { + "epoch": 19.92, + "learning_rate": 1.0088188200517305e-06, + "loss": 0.2475, + "step": 53938 + }, + { + "epoch": 19.92, + "learning_rate": 1.0087818696883853e-06, + "loss": 0.2724, + "step": 53939 + }, + { + "epoch": 19.92, + "learning_rate": 1.0087449193250402e-06, + "loss": 0.1567, + "step": 53940 + }, + { + "epoch": 19.92, + "learning_rate": 1.0087079689616948e-06, + "loss": 0.2077, + "step": 53941 + }, + { + "epoch": 19.92, + "learning_rate": 1.0086710185983497e-06, + "loss": 0.2239, + "step": 53942 + }, + { + "epoch": 19.92, + "learning_rate": 1.0086340682350043e-06, + "loss": 0.1467, + "step": 53943 + }, + { + "epoch": 19.92, + "learning_rate": 1.008597117871659e-06, + "loss": 0.1375, + "step": 53944 + }, + { + "epoch": 19.92, + "learning_rate": 1.0085601675083138e-06, + "loss": 0.282, + "step": 53945 + }, + { + "epoch": 19.92, + "learning_rate": 1.0085232171449686e-06, + "loss": 0.3142, + "step": 53946 + }, + { + "epoch": 19.92, + "learning_rate": 1.0084862667816233e-06, + "loss": 0.1364, + "step": 53947 + }, + { + "epoch": 19.92, + "learning_rate": 1.0084493164182781e-06, + "loss": 0.2556, + "step": 53948 + }, + { + "epoch": 19.92, + "learning_rate": 1.008412366054933e-06, + "loss": 0.2679, + "step": 53949 + }, + { + "epoch": 19.92, + "learning_rate": 1.0083754156915876e-06, + "loss": 0.2514, + "step": 53950 + }, + { + "epoch": 19.92, + "learning_rate": 1.0083384653282425e-06, + "loss": 0.2807, + "step": 53951 + }, + { + "epoch": 19.92, + "learning_rate": 1.0083015149648973e-06, + "loss": 0.1879, + "step": 53952 + }, + { + "epoch": 19.92, + "learning_rate": 1.008264564601552e-06, + "loss": 0.2529, + "step": 53953 + }, + { + "epoch": 19.92, + "learning_rate": 1.0082276142382066e-06, + "loss": 0.2523, + "step": 53954 + }, + { + "epoch": 19.92, + "learning_rate": 1.0081906638748614e-06, + "loss": 0.2318, + "step": 53955 + }, + { + "epoch": 19.92, + "learning_rate": 1.0081537135115163e-06, + "loss": 0.1831, + "step": 53956 + }, + { + "epoch": 19.93, + "learning_rate": 1.008116763148171e-06, + "loss": 0.2109, + "step": 53957 + }, + { + "epoch": 19.93, + "learning_rate": 1.0080798127848258e-06, + "loss": 0.2631, + "step": 53958 + }, + { + "epoch": 19.93, + "learning_rate": 1.0080428624214806e-06, + "loss": 0.1832, + "step": 53959 + }, + { + "epoch": 19.93, + "learning_rate": 1.0080059120581352e-06, + "loss": 0.3286, + "step": 53960 + }, + { + "epoch": 19.93, + "learning_rate": 1.00796896169479e-06, + "loss": 0.2146, + "step": 53961 + }, + { + "epoch": 19.93, + "learning_rate": 1.007932011331445e-06, + "loss": 0.3792, + "step": 53962 + }, + { + "epoch": 19.93, + "learning_rate": 1.0078950609680996e-06, + "loss": 0.2629, + "step": 53963 + }, + { + "epoch": 19.93, + "learning_rate": 1.0078581106047542e-06, + "loss": 0.1872, + "step": 53964 + }, + { + "epoch": 19.93, + "learning_rate": 1.007821160241409e-06, + "loss": 0.1162, + "step": 53965 + }, + { + "epoch": 19.93, + "learning_rate": 1.007784209878064e-06, + "loss": 0.2864, + "step": 53966 + }, + { + "epoch": 19.93, + "learning_rate": 1.0077472595147185e-06, + "loss": 0.261, + "step": 53967 + }, + { + "epoch": 19.93, + "learning_rate": 1.0077103091513734e-06, + "loss": 0.1794, + "step": 53968 + }, + { + "epoch": 19.93, + "learning_rate": 1.0076733587880282e-06, + "loss": 0.1428, + "step": 53969 + }, + { + "epoch": 19.93, + "learning_rate": 1.0076364084246829e-06, + "loss": 0.1458, + "step": 53970 + }, + { + "epoch": 19.93, + "learning_rate": 1.0075994580613377e-06, + "loss": 0.105, + "step": 53971 + }, + { + "epoch": 19.93, + "learning_rate": 1.0075625076979924e-06, + "loss": 0.137, + "step": 53972 + }, + { + "epoch": 19.93, + "learning_rate": 1.007525557334647e-06, + "loss": 0.1934, + "step": 53973 + }, + { + "epoch": 19.93, + "learning_rate": 1.0074886069713018e-06, + "loss": 0.1551, + "step": 53974 + }, + { + "epoch": 19.93, + "learning_rate": 1.0074516566079567e-06, + "loss": 0.2312, + "step": 53975 + }, + { + "epoch": 19.93, + "learning_rate": 1.0074147062446113e-06, + "loss": 0.2268, + "step": 53976 + }, + { + "epoch": 19.93, + "learning_rate": 1.0073777558812662e-06, + "loss": 0.2333, + "step": 53977 + }, + { + "epoch": 19.93, + "learning_rate": 1.007340805517921e-06, + "loss": 0.2472, + "step": 53978 + }, + { + "epoch": 19.93, + "learning_rate": 1.0073038551545757e-06, + "loss": 0.256, + "step": 53979 + }, + { + "epoch": 19.93, + "learning_rate": 1.0072669047912305e-06, + "loss": 0.2151, + "step": 53980 + }, + { + "epoch": 19.93, + "learning_rate": 1.0072299544278854e-06, + "loss": 0.3576, + "step": 53981 + }, + { + "epoch": 19.93, + "learning_rate": 1.00719300406454e-06, + "loss": 0.1722, + "step": 53982 + }, + { + "epoch": 19.93, + "learning_rate": 1.0071560537011946e-06, + "loss": 0.1541, + "step": 53983 + }, + { + "epoch": 19.94, + "learning_rate": 1.0071191033378495e-06, + "loss": 0.379, + "step": 53984 + }, + { + "epoch": 19.94, + "learning_rate": 1.0070821529745043e-06, + "loss": 0.2398, + "step": 53985 + }, + { + "epoch": 19.94, + "learning_rate": 1.007045202611159e-06, + "loss": 0.136, + "step": 53986 + }, + { + "epoch": 19.94, + "learning_rate": 1.0070082522478138e-06, + "loss": 0.1902, + "step": 53987 + }, + { + "epoch": 19.94, + "learning_rate": 1.0069713018844687e-06, + "loss": 0.1551, + "step": 53988 + }, + { + "epoch": 19.94, + "learning_rate": 1.0069343515211233e-06, + "loss": 0.3015, + "step": 53989 + }, + { + "epoch": 19.94, + "learning_rate": 1.0068974011577781e-06, + "loss": 0.2806, + "step": 53990 + }, + { + "epoch": 19.94, + "learning_rate": 1.006860450794433e-06, + "loss": 0.171, + "step": 53991 + }, + { + "epoch": 19.94, + "learning_rate": 1.0068235004310876e-06, + "loss": 0.2845, + "step": 53992 + }, + { + "epoch": 19.94, + "learning_rate": 1.0067865500677423e-06, + "loss": 0.2412, + "step": 53993 + }, + { + "epoch": 19.94, + "learning_rate": 1.0067495997043971e-06, + "loss": 0.1732, + "step": 53994 + }, + { + "epoch": 19.94, + "learning_rate": 1.0067126493410517e-06, + "loss": 0.1678, + "step": 53995 + }, + { + "epoch": 19.94, + "learning_rate": 1.0066756989777066e-06, + "loss": 0.2228, + "step": 53996 + }, + { + "epoch": 19.94, + "learning_rate": 1.0066387486143614e-06, + "loss": 0.2952, + "step": 53997 + }, + { + "epoch": 19.94, + "learning_rate": 1.006601798251016e-06, + "loss": 0.2905, + "step": 53998 + }, + { + "epoch": 19.94, + "learning_rate": 1.006564847887671e-06, + "loss": 0.1093, + "step": 53999 + }, + { + "epoch": 19.94, + "learning_rate": 1.0065278975243258e-06, + "loss": 0.1437, + "step": 54000 + }, + { + "epoch": 19.94, + "learning_rate": 1.0064909471609804e-06, + "loss": 0.3121, + "step": 54001 + }, + { + "epoch": 19.94, + "learning_rate": 1.0064539967976353e-06, + "loss": 0.1475, + "step": 54002 + }, + { + "epoch": 19.94, + "learning_rate": 1.00641704643429e-06, + "loss": 0.2443, + "step": 54003 + }, + { + "epoch": 19.94, + "learning_rate": 1.0063800960709447e-06, + "loss": 0.2176, + "step": 54004 + }, + { + "epoch": 19.94, + "learning_rate": 1.0063431457075994e-06, + "loss": 0.1673, + "step": 54005 + }, + { + "epoch": 19.94, + "learning_rate": 1.0063061953442542e-06, + "loss": 0.1373, + "step": 54006 + }, + { + "epoch": 19.94, + "learning_rate": 1.006269244980909e-06, + "loss": 0.1735, + "step": 54007 + }, + { + "epoch": 19.94, + "learning_rate": 1.0062322946175637e-06, + "loss": 0.3295, + "step": 54008 + }, + { + "epoch": 19.94, + "learning_rate": 1.0061953442542186e-06, + "loss": 0.1458, + "step": 54009 + }, + { + "epoch": 19.94, + "learning_rate": 1.0061583938908734e-06, + "loss": 0.1761, + "step": 54010 + }, + { + "epoch": 19.94, + "learning_rate": 1.006121443527528e-06, + "loss": 0.0976, + "step": 54011 + }, + { + "epoch": 19.95, + "learning_rate": 1.006084493164183e-06, + "loss": 0.1471, + "step": 54012 + }, + { + "epoch": 19.95, + "learning_rate": 1.0060475428008375e-06, + "loss": 0.2821, + "step": 54013 + }, + { + "epoch": 19.95, + "learning_rate": 1.0060105924374922e-06, + "loss": 0.1342, + "step": 54014 + }, + { + "epoch": 19.95, + "learning_rate": 1.005973642074147e-06, + "loss": 0.3044, + "step": 54015 + }, + { + "epoch": 19.95, + "learning_rate": 1.0059366917108019e-06, + "loss": 0.1874, + "step": 54016 + }, + { + "epoch": 19.95, + "learning_rate": 1.0058997413474565e-06, + "loss": 0.3756, + "step": 54017 + }, + { + "epoch": 19.95, + "learning_rate": 1.0058627909841114e-06, + "loss": 0.2859, + "step": 54018 + }, + { + "epoch": 19.95, + "learning_rate": 1.0058258406207662e-06, + "loss": 0.1678, + "step": 54019 + }, + { + "epoch": 19.95, + "learning_rate": 1.0057888902574208e-06, + "loss": 0.1406, + "step": 54020 + }, + { + "epoch": 19.95, + "learning_rate": 1.0057519398940757e-06, + "loss": 0.1971, + "step": 54021 + }, + { + "epoch": 19.95, + "learning_rate": 1.0057149895307305e-06, + "loss": 0.2518, + "step": 54022 + }, + { + "epoch": 19.95, + "learning_rate": 1.0056780391673852e-06, + "loss": 0.2165, + "step": 54023 + }, + { + "epoch": 19.95, + "learning_rate": 1.0056410888040398e-06, + "loss": 0.1556, + "step": 54024 + }, + { + "epoch": 19.95, + "learning_rate": 1.0056041384406947e-06, + "loss": 0.1912, + "step": 54025 + }, + { + "epoch": 19.95, + "learning_rate": 1.0055671880773495e-06, + "loss": 0.3339, + "step": 54026 + }, + { + "epoch": 19.95, + "learning_rate": 1.0055302377140041e-06, + "loss": 0.2128, + "step": 54027 + }, + { + "epoch": 19.95, + "learning_rate": 1.005493287350659e-06, + "loss": 0.1618, + "step": 54028 + }, + { + "epoch": 19.95, + "learning_rate": 1.0054563369873138e-06, + "loss": 0.1996, + "step": 54029 + }, + { + "epoch": 19.95, + "learning_rate": 1.0054193866239685e-06, + "loss": 0.1886, + "step": 54030 + }, + { + "epoch": 19.95, + "learning_rate": 1.0053824362606233e-06, + "loss": 0.1556, + "step": 54031 + }, + { + "epoch": 19.95, + "learning_rate": 1.0053454858972782e-06, + "loss": 0.2222, + "step": 54032 + }, + { + "epoch": 19.95, + "learning_rate": 1.0053085355339328e-06, + "loss": 0.22, + "step": 54033 + }, + { + "epoch": 19.95, + "learning_rate": 1.0052715851705874e-06, + "loss": 0.1989, + "step": 54034 + }, + { + "epoch": 19.95, + "learning_rate": 1.0052346348072423e-06, + "loss": 0.2711, + "step": 54035 + }, + { + "epoch": 19.95, + "learning_rate": 1.0051976844438971e-06, + "loss": 0.2886, + "step": 54036 + }, + { + "epoch": 19.95, + "learning_rate": 1.0051607340805518e-06, + "loss": 0.1668, + "step": 54037 + }, + { + "epoch": 19.95, + "learning_rate": 1.0051237837172066e-06, + "loss": 0.155, + "step": 54038 + }, + { + "epoch": 19.96, + "learning_rate": 1.0050868333538615e-06, + "loss": 0.0947, + "step": 54039 + }, + { + "epoch": 19.96, + "learning_rate": 1.005049882990516e-06, + "loss": 0.1323, + "step": 54040 + }, + { + "epoch": 19.96, + "learning_rate": 1.005012932627171e-06, + "loss": 0.2892, + "step": 54041 + }, + { + "epoch": 19.96, + "learning_rate": 1.0049759822638258e-06, + "loss": 0.2595, + "step": 54042 + }, + { + "epoch": 19.96, + "learning_rate": 1.0049390319004802e-06, + "loss": 0.1903, + "step": 54043 + }, + { + "epoch": 19.96, + "learning_rate": 1.004902081537135e-06, + "loss": 0.1686, + "step": 54044 + }, + { + "epoch": 19.96, + "learning_rate": 1.00486513117379e-06, + "loss": 0.1765, + "step": 54045 + }, + { + "epoch": 19.96, + "learning_rate": 1.0048281808104446e-06, + "loss": 0.2474, + "step": 54046 + }, + { + "epoch": 19.96, + "learning_rate": 1.0047912304470994e-06, + "loss": 0.2079, + "step": 54047 + }, + { + "epoch": 19.96, + "learning_rate": 1.0047542800837543e-06, + "loss": 0.2747, + "step": 54048 + }, + { + "epoch": 19.96, + "learning_rate": 1.0047173297204089e-06, + "loss": 0.2469, + "step": 54049 + }, + { + "epoch": 19.96, + "learning_rate": 1.0046803793570637e-06, + "loss": 0.1963, + "step": 54050 + }, + { + "epoch": 19.96, + "learning_rate": 1.0046434289937186e-06, + "loss": 0.1949, + "step": 54051 + }, + { + "epoch": 19.96, + "learning_rate": 1.0046064786303732e-06, + "loss": 0.2507, + "step": 54052 + }, + { + "epoch": 19.96, + "learning_rate": 1.0045695282670279e-06, + "loss": 0.0879, + "step": 54053 + }, + { + "epoch": 19.96, + "learning_rate": 1.0045325779036827e-06, + "loss": 0.2134, + "step": 54054 + }, + { + "epoch": 19.96, + "learning_rate": 1.0044956275403376e-06, + "loss": 0.1734, + "step": 54055 + }, + { + "epoch": 19.96, + "learning_rate": 1.0044586771769922e-06, + "loss": 0.1421, + "step": 54056 + }, + { + "epoch": 19.96, + "learning_rate": 1.004421726813647e-06, + "loss": 0.2242, + "step": 54057 + }, + { + "epoch": 19.96, + "learning_rate": 1.0043847764503019e-06, + "loss": 0.2653, + "step": 54058 + }, + { + "epoch": 19.96, + "learning_rate": 1.0043478260869565e-06, + "loss": 0.1953, + "step": 54059 + }, + { + "epoch": 19.96, + "learning_rate": 1.0043108757236114e-06, + "loss": 0.4066, + "step": 54060 + }, + { + "epoch": 19.96, + "learning_rate": 1.0042739253602662e-06, + "loss": 0.2536, + "step": 54061 + }, + { + "epoch": 19.96, + "learning_rate": 1.0042369749969209e-06, + "loss": 0.1945, + "step": 54062 + }, + { + "epoch": 19.96, + "learning_rate": 1.0042000246335755e-06, + "loss": 0.1212, + "step": 54063 + }, + { + "epoch": 19.96, + "learning_rate": 1.0041630742702303e-06, + "loss": 0.2637, + "step": 54064 + }, + { + "epoch": 19.96, + "learning_rate": 1.004126123906885e-06, + "loss": 0.1991, + "step": 54065 + }, + { + "epoch": 19.97, + "learning_rate": 1.0040891735435398e-06, + "loss": 0.1628, + "step": 54066 + }, + { + "epoch": 19.97, + "learning_rate": 1.0040522231801947e-06, + "loss": 0.1778, + "step": 54067 + }, + { + "epoch": 19.97, + "learning_rate": 1.0040152728168493e-06, + "loss": 0.167, + "step": 54068 + }, + { + "epoch": 19.97, + "learning_rate": 1.0039783224535042e-06, + "loss": 0.2447, + "step": 54069 + }, + { + "epoch": 19.97, + "learning_rate": 1.003941372090159e-06, + "loss": 0.1509, + "step": 54070 + }, + { + "epoch": 19.97, + "learning_rate": 1.0039044217268136e-06, + "loss": 0.1274, + "step": 54071 + }, + { + "epoch": 19.97, + "learning_rate": 1.0038674713634685e-06, + "loss": 0.0755, + "step": 54072 + }, + { + "epoch": 19.97, + "learning_rate": 1.0038305210001231e-06, + "loss": 0.1942, + "step": 54073 + }, + { + "epoch": 19.97, + "learning_rate": 1.003793570636778e-06, + "loss": 0.1883, + "step": 54074 + }, + { + "epoch": 19.97, + "learning_rate": 1.0037566202734326e-06, + "loss": 0.1178, + "step": 54075 + }, + { + "epoch": 19.97, + "learning_rate": 1.0037196699100875e-06, + "loss": 0.1814, + "step": 54076 + }, + { + "epoch": 19.97, + "learning_rate": 1.0036827195467423e-06, + "loss": 0.1649, + "step": 54077 + }, + { + "epoch": 19.97, + "learning_rate": 1.003645769183397e-06, + "loss": 0.185, + "step": 54078 + }, + { + "epoch": 19.97, + "learning_rate": 1.0036088188200518e-06, + "loss": 0.3823, + "step": 54079 + }, + { + "epoch": 19.97, + "learning_rate": 1.0035718684567066e-06, + "loss": 0.2248, + "step": 54080 + }, + { + "epoch": 19.97, + "learning_rate": 1.0035349180933613e-06, + "loss": 0.3241, + "step": 54081 + }, + { + "epoch": 19.97, + "learning_rate": 1.0034979677300161e-06, + "loss": 0.1936, + "step": 54082 + }, + { + "epoch": 19.97, + "learning_rate": 1.0034610173666708e-06, + "loss": 0.1586, + "step": 54083 + }, + { + "epoch": 19.97, + "learning_rate": 1.0034240670033254e-06, + "loss": 0.1613, + "step": 54084 + }, + { + "epoch": 19.97, + "learning_rate": 1.0033871166399803e-06, + "loss": 0.2648, + "step": 54085 + }, + { + "epoch": 19.97, + "learning_rate": 1.003350166276635e-06, + "loss": 0.3139, + "step": 54086 + }, + { + "epoch": 19.97, + "learning_rate": 1.0033132159132897e-06, + "loss": 0.2618, + "step": 54087 + }, + { + "epoch": 19.97, + "learning_rate": 1.0032762655499446e-06, + "loss": 0.2503, + "step": 54088 + }, + { + "epoch": 19.97, + "learning_rate": 1.0032393151865994e-06, + "loss": 0.2502, + "step": 54089 + }, + { + "epoch": 19.97, + "learning_rate": 1.003202364823254e-06, + "loss": 0.2611, + "step": 54090 + }, + { + "epoch": 19.97, + "learning_rate": 1.003165414459909e-06, + "loss": 0.2488, + "step": 54091 + }, + { + "epoch": 19.97, + "learning_rate": 1.0031284640965638e-06, + "loss": 0.1698, + "step": 54092 + }, + { + "epoch": 19.98, + "learning_rate": 1.0030915137332184e-06, + "loss": 0.1687, + "step": 54093 + }, + { + "epoch": 19.98, + "learning_rate": 1.003054563369873e-06, + "loss": 0.1502, + "step": 54094 + }, + { + "epoch": 19.98, + "learning_rate": 1.0030176130065279e-06, + "loss": 0.155, + "step": 54095 + }, + { + "epoch": 19.98, + "learning_rate": 1.0029806626431827e-06, + "loss": 0.1603, + "step": 54096 + }, + { + "epoch": 19.98, + "learning_rate": 1.0029437122798374e-06, + "loss": 0.2463, + "step": 54097 + }, + { + "epoch": 19.98, + "learning_rate": 1.0029067619164922e-06, + "loss": 0.2942, + "step": 54098 + }, + { + "epoch": 19.98, + "learning_rate": 1.002869811553147e-06, + "loss": 0.2359, + "step": 54099 + }, + { + "epoch": 19.98, + "learning_rate": 1.0028328611898017e-06, + "loss": 0.2945, + "step": 54100 + }, + { + "epoch": 19.98, + "learning_rate": 1.0027959108264566e-06, + "loss": 0.22, + "step": 54101 + }, + { + "epoch": 19.98, + "learning_rate": 1.0027589604631112e-06, + "loss": 0.1465, + "step": 54102 + }, + { + "epoch": 19.98, + "learning_rate": 1.002722010099766e-06, + "loss": 0.153, + "step": 54103 + }, + { + "epoch": 19.98, + "learning_rate": 1.0026850597364207e-06, + "loss": 0.1996, + "step": 54104 + }, + { + "epoch": 19.98, + "learning_rate": 1.0026481093730755e-06, + "loss": 0.2681, + "step": 54105 + }, + { + "epoch": 19.98, + "learning_rate": 1.0026111590097304e-06, + "loss": 0.1917, + "step": 54106 + }, + { + "epoch": 19.98, + "learning_rate": 1.002574208646385e-06, + "loss": 0.1418, + "step": 54107 + }, + { + "epoch": 19.98, + "learning_rate": 1.0025372582830399e-06, + "loss": 0.3351, + "step": 54108 + }, + { + "epoch": 19.98, + "learning_rate": 1.0025003079196947e-06, + "loss": 0.1752, + "step": 54109 + }, + { + "epoch": 19.98, + "learning_rate": 1.0024633575563493e-06, + "loss": 0.2233, + "step": 54110 + }, + { + "epoch": 19.98, + "learning_rate": 1.0024264071930042e-06, + "loss": 0.1274, + "step": 54111 + }, + { + "epoch": 19.98, + "learning_rate": 1.0023894568296588e-06, + "loss": 0.2877, + "step": 54112 + }, + { + "epoch": 19.98, + "learning_rate": 1.0023525064663135e-06, + "loss": 0.2636, + "step": 54113 + }, + { + "epoch": 19.98, + "learning_rate": 1.0023155561029683e-06, + "loss": 0.2615, + "step": 54114 + }, + { + "epoch": 19.98, + "learning_rate": 1.0022786057396232e-06, + "loss": 0.1362, + "step": 54115 + }, + { + "epoch": 19.98, + "learning_rate": 1.0022416553762778e-06, + "loss": 0.1953, + "step": 54116 + }, + { + "epoch": 19.98, + "learning_rate": 1.0022047050129326e-06, + "loss": 0.1996, + "step": 54117 + }, + { + "epoch": 19.98, + "learning_rate": 1.0021677546495875e-06, + "loss": 0.345, + "step": 54118 + }, + { + "epoch": 19.98, + "learning_rate": 1.0021308042862421e-06, + "loss": 0.2301, + "step": 54119 + }, + { + "epoch": 19.99, + "learning_rate": 1.002093853922897e-06, + "loss": 0.1638, + "step": 54120 + }, + { + "epoch": 19.99, + "learning_rate": 1.0020569035595518e-06, + "loss": 0.1632, + "step": 54121 + }, + { + "epoch": 19.99, + "learning_rate": 1.0020199531962065e-06, + "loss": 0.2108, + "step": 54122 + }, + { + "epoch": 19.99, + "learning_rate": 1.001983002832861e-06, + "loss": 0.3092, + "step": 54123 + }, + { + "epoch": 19.99, + "learning_rate": 1.001946052469516e-06, + "loss": 0.1832, + "step": 54124 + }, + { + "epoch": 19.99, + "learning_rate": 1.0019091021061708e-06, + "loss": 0.1562, + "step": 54125 + }, + { + "epoch": 19.99, + "learning_rate": 1.0018721517428254e-06, + "loss": 0.2733, + "step": 54126 + }, + { + "epoch": 19.99, + "learning_rate": 1.0018352013794803e-06, + "loss": 0.3267, + "step": 54127 + }, + { + "epoch": 19.99, + "learning_rate": 1.0017982510161351e-06, + "loss": 0.3198, + "step": 54128 + }, + { + "epoch": 19.99, + "learning_rate": 1.0017613006527898e-06, + "loss": 0.2574, + "step": 54129 + }, + { + "epoch": 19.99, + "learning_rate": 1.0017243502894446e-06, + "loss": 0.2812, + "step": 54130 + }, + { + "epoch": 19.99, + "learning_rate": 1.0016873999260995e-06, + "loss": 0.1942, + "step": 54131 + }, + { + "epoch": 19.99, + "learning_rate": 1.0016504495627539e-06, + "loss": 0.1173, + "step": 54132 + }, + { + "epoch": 19.99, + "learning_rate": 1.0016134991994087e-06, + "loss": 0.1497, + "step": 54133 + }, + { + "epoch": 19.99, + "learning_rate": 1.0015765488360636e-06, + "loss": 0.1601, + "step": 54134 + }, + { + "epoch": 19.99, + "learning_rate": 1.0015395984727182e-06, + "loss": 0.1396, + "step": 54135 + }, + { + "epoch": 19.99, + "learning_rate": 1.001502648109373e-06, + "loss": 0.1951, + "step": 54136 + }, + { + "epoch": 19.99, + "learning_rate": 1.001465697746028e-06, + "loss": 0.1765, + "step": 54137 + }, + { + "epoch": 19.99, + "learning_rate": 1.0014287473826825e-06, + "loss": 0.1982, + "step": 54138 + }, + { + "epoch": 19.99, + "learning_rate": 1.0013917970193374e-06, + "loss": 0.1497, + "step": 54139 + }, + { + "epoch": 19.99, + "learning_rate": 1.0013548466559922e-06, + "loss": 0.21, + "step": 54140 + }, + { + "epoch": 19.99, + "learning_rate": 1.0013178962926469e-06, + "loss": 0.2874, + "step": 54141 + }, + { + "epoch": 19.99, + "learning_rate": 1.0012809459293015e-06, + "loss": 0.1425, + "step": 54142 + }, + { + "epoch": 19.99, + "learning_rate": 1.0012439955659564e-06, + "loss": 0.1128, + "step": 54143 + }, + { + "epoch": 19.99, + "learning_rate": 1.0012070452026112e-06, + "loss": 0.271, + "step": 54144 + }, + { + "epoch": 19.99, + "learning_rate": 1.0011700948392658e-06, + "loss": 0.2034, + "step": 54145 + }, + { + "epoch": 19.99, + "learning_rate": 1.0011331444759207e-06, + "loss": 0.2253, + "step": 54146 + }, + { + "epoch": 20.0, + "learning_rate": 1.0010961941125755e-06, + "loss": 0.1961, + "step": 54147 + }, + { + "epoch": 20.0, + "learning_rate": 1.0010592437492302e-06, + "loss": 0.2805, + "step": 54148 + }, + { + "epoch": 20.0, + "learning_rate": 1.001022293385885e-06, + "loss": 0.2022, + "step": 54149 + }, + { + "epoch": 20.0, + "learning_rate": 1.0009853430225399e-06, + "loss": 0.2429, + "step": 54150 + }, + { + "epoch": 20.0, + "learning_rate": 1.0009483926591945e-06, + "loss": 0.2536, + "step": 54151 + }, + { + "epoch": 20.0, + "learning_rate": 1.0009114422958492e-06, + "loss": 0.4501, + "step": 54152 + }, + { + "epoch": 20.0, + "learning_rate": 1.000874491932504e-06, + "loss": 0.2013, + "step": 54153 + }, + { + "epoch": 20.0, + "learning_rate": 1.0008375415691588e-06, + "loss": 0.2337, + "step": 54154 + }, + { + "epoch": 20.0, + "learning_rate": 1.0008005912058135e-06, + "loss": 0.1443, + "step": 54155 + }, + { + "epoch": 20.0, + "learning_rate": 1.0007636408424683e-06, + "loss": 0.204, + "step": 54156 + }, + { + "epoch": 20.0, + "learning_rate": 1.0007266904791232e-06, + "loss": 0.2013, + "step": 54157 + }, + { + "epoch": 20.0, + "learning_rate": 1.0006897401157778e-06, + "loss": 0.2424, + "step": 54158 + }, + { + "epoch": 20.0, + "learning_rate": 1.0006527897524327e-06, + "loss": 0.2057, + "step": 54159 + }, + { + "epoch": 20.0, + "learning_rate": 1.0006158393890873e-06, + "loss": 0.1891, + "step": 54160 + }, + { + "epoch": 20.0, + "eval_accuracy": 0.8713349032645334, + "eval_loss": 0.34530189633369446, + "eval_runtime": 25.2727, + "eval_samples_per_second": 1713.867, + "eval_steps_per_second": 26.788, + "step": 54160 + } + ], + "logging_steps": 1, + "max_steps": 81240, + "num_train_epochs": 30, + "save_steps": 500, + "total_flos": 2.8749199794312144e+16, + "trial_name": null, + "trial_params": null +}